Professional Documents
Culture Documents
How well have classic ideas on whole-word phonology stood the test of time?
Waterson claimed that each child has a system of their own; Ferguson and
Farwell emphasized the relative accuracy of rst words; Menn noted the
occurrence of regression and the emergence of phonological systematicity.
This volume brings together classic texts such as these with current data-rich
studies of British and American English, Arabic, Brazilian Portuguese,
Finnish, French, Japanese, Polish, and Spanish. This combination of classic
and contemporary work from the last thirty years presents the reader with
cutting-edge perspectives on child language by linking historical approaches
with current ideas such as exemplar theory and usage-based phonology and
contrasting state-of-the-art perspectives from developmental psychology and
linguistics. This is a valuable resource for cognitive scientists, developmen-
talists, linguists, psychologists, speech scientists, and therapists interested in
understanding how children begin to use language without the benet of
language-specic innate knowledge.
marilyn m. vihman is Professor of Language and Linguistic Science at the
University of York.
tamar keren-portnoy is Lecturer in Language and Linguistic Science at
the University of York.
The Emergence of Phonology:
Whole-word Approaches and
Cross-linguistic Evidence
Edited by
Marilyn M. Vihman and Tamar Keren-Portnoy
University Printing House, Cambridge CB2 8BS, United Kingdom
Published in the United States of America by Cambridge University Press, NewYork
Cambridge University Press is part of the University of Cambridge.
It furthers the Universitys mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9780521762342
Cambridge University Press 2013
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2013
Printed in the United Kingdom by CPI Group Ltd, Croydon CR0 4YY
A catalog record for this publication is available from the British Library.
Library of Congress Cataloging in Publication Data
The emergence of phonology : whole-word approaches and cross-linguistic
evidence / Edited by Marilyn M. Vihman and Tamar Keren-Portnoy
pages cm
Includes bibliographical references and index.
ISBN 978-0-521-76234-2
1. Lexical phonology. 2. Grammar, Comparative and general
Phonology. 3. Grammar, Comparative and general Morphology.
4. Reading Language experience approach. 5. Language and languages
Study and teaching. 6. Visual learning. 7. Complexity
(Linguistics) I. Vihman, Marilyn May, editor of compilation.
P217.6.E43 2013
414dc23 2013013107
ISBN 978-0-521-76234-2 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
To the memory of Fergie
Charles A. Ferguson (19211998)
who mentored a generation of child phonologists,
including many whose work is included here
Contents
List of contributors page ix
Preface xi
1 Introduction: the emergence of phonology: whole-word
approaches, cross-linguistic evidence 1
mari lyn m. vi hman and tamar keren- portnoy
Part I The current framework 15
2 Phonological development: toward a radical templatic
phonology 17
mari lyn m. vi hman and wi lli am croft
Part II Setting papers 59
3 Child phonology: a prosodic view 61
natali e waterson
4 Words and sounds in early language acquisition 93
charles a. ferguson and carol b. farwell
5 Developmental reorganization of phonology: a hierarchy of basic
units of acquisition 133
marlys a. macken
6 Development of articulatory, phonetic, and phonological
capabilities 168
li se menn
Part III Cross-linguistic studies 215
7 One idiosyncratic strategy in the acquisition of phonology 217
t. m. s. pri estly
8 Phonological reorganization: a case study 238
mari lyn m. vi hman and shelley l. velleman
vii
9 How abstract is child phonology? Towards an integration of
linguistic and psychological approaches 259
mari lyn m. vi hman, shelley l. velleman,
and lorrai ne mccune
10 Beyond early words: word template development in Brazilian
Portuguese 291
dani ela oli vei ra- gui mara es
11 Templates in French 317
sophi e wauqui er and naomi yamaguchi
12 The acquisition of consonant clusters in Polish: a case study 343
marta szreder
13 Geminate template: a model for rst Finnish words 362
tuula savi nai nen- makkonen
14 Inuence of geminate structure on early Arabic templatic patterns 374
ghada khattab and j alal al- tami mi
15 Lexical frequency effects on phonological development: the case
of word production in Japanese 415
mi tsuhi ko ota
Part IV Perspectives and challenges 439
16 Aview from developmental psychology 441
lorrai ne mccune
17 Challenges to theories, charges to a model: the Linked-Attractor
model of phonological development 460
li se menn, ellen schmi dt, and brent ni cholas
References for reprinted papers 503
Index 504
viii Contents
Contributors
j alal al- tami mi
wi lli am croft
carol b. farwell
charles a. ferguson
tamar keren- portnoy
ghada khattab
marlys a. macken
lorrai ne mccune
li se menn
brent ni cholas
dani ela oli vei ra- gui mara es
mi tsuhi ko ota
t. m. s. pri estly
tuula savi nai nen- makkonen
ellen schmi dt
marta szreder
shelley l. velleman
mari lyn m. vi hman
natali e waterson
sophi e wauqui er
naomi yamaguchi
ix
Preface
The idea of creating this volume of readings was rst conceived ve years ago.
We would like to thank our editor, Helen Barton, for her willingness to take
the project on and her patience in seeing it through what turned out to be a far
longer incubation period than any of us anticipated! We also thank the living
authors of reprinted papers for their permission to include their inspiring work
in this volume, and also the University of Wisconsin for providing nancial
support for our reprint of Macken (1979). Finally, we thank Dr. Nina Gram
Garmann, University of Oslo and Akershus University College of Applied
Sciences, for providing insightful critiques of two of the papers published
here for the rst time.
Every effort has been made to secure necessary permissions to reproduce
copyright material in this work, though in some cases it has proved impossible
to trace or contact copyright holders. If any omissions are brought to our notice,
we will be happy to include appropriate acknowledgements on reprinting, and/
or in any subsequent edition.
xi
1 Introduction: the emergence of phonology:
whole-word approaches, cross-linguistic evidence
Marilyn M. Vihman and Tamar Keren-Portnoy
Whole-word phonology is a particular approach to early phonological develop-
ment. This volume is designed to bring together the classic papers which gave
rise to it in the 1970s and current studies that build on and extend the model,
which in essence took an emergentist and usage-based stance before its time;
the book will make no attempt to cover other approaches to phonological
development in any systematic way. Many of the papers, including Vihman
and Croft (2007, this volume, Chapter 2),
1
with which we begin, use the term
template to refer to child-specic word patterns identiable within the rst
year of word use. Templates, referred to sporadically in the earlier developmen-
tal literature (e.g., Menn 1983, this volume, Chapter 6) and given formal status
for adult linguistic analyses in Prosodic Morphology (McCarthy and Prince
1995), are a more focused expression of the ideas formulated by Waterson
(1971, this volume, Chapter 3), Ferguson and Farwell (1975, this volume,
Chapter 4), and Macken (1979, this volume, Chapter 5), which provided the
core of the whole-word phonology idea (see Vihman and Croft 2007, this
volume, Chapter 2, for a summary of the basic arguments).
This volume is restricted to the study of early word production and the phono-
logical patterning that can be seen in that domain. The year in which the rst of our
setting papers was written Waterson (1971, this volume, Chapter 3) also
marks the year of publication of the rst study of infant speech perception (Eimas,
Siqueland, Jusczyk, and Vigorito 1971). Since then, perception studies have solidly
documented infants remarkable early discriminatory capacities and the rapid
advances in knowledge of the ambient language that follow over the rst year of
life (see Jusczyk 1997; Kuhl 2004; and Vihman forthcoming 2014 for reviews),
while numerous studies demonstrating infant statistical learning (in language and
other areas) from an early age have expanded our understanding of the learning
mechanism that may underlie those advances (see Thiessen and Saffran 2007,
and Johnson and Tyler 2010 for alternative positions on the role of statistical
learning; Vihman forthcoming 2014: ch. 5 provides an overview). In addition,
several distinct methodological procedures have been used to trace and explore
the nature of early word-form learning over the rst two years of life.
2
The
resultant studies are of evident relevance to phonological development but
none are included here, as the addition of even a few would result in a far
1
longer and less focused volume (and the studies are readily available elsewhere).
Nevertheless, what we have learned about perception and early learning capacities
is critical to our understanding of the course of phonological development and
clearly complements the whole-word approach presented here.
Whole-word phonology and templates in phonological
development
As understood in the chapters that follow, templates involve (idiosyncratic)
prosodic structures that appear to be generalized, in different ways by different
children, from the forms of a childs earlier babble vocalizations and rst words.
Templates typically lead to increased similarity in the forms of the childs words at
the expense of accuracy (i.e., of match to the adult target form). This corresponds
to a sequence of, rst, item (or exemplar) learning, then distributional learning,
implicitly and automatically applied to repeatedly used child output forms the
presumed source of the generalizing of patterns to new targets. Taking an
exemplar model perspective, this generalization can be thought of as the self-
organization of the exemplar space, due to connections being formed between
similarly shaped child forms; an alternative (but not necessarily incompatible),
strictly sensorimotor perspective sees the generalization as no more than
the automatization of one or more well-practiced procedures, namely, the childs
emergent neuromotor word-production routines (McCune, this volume,
Chapter 16). The resultant patterns appear to constitute (unconscious or implicit)
child responses to the phonological challenges posed by target word forms. In
other words, the childs existing resources (familiar production routines) are
deployed to deal with what is novel and thus difcult to bring to mind, plan,
and produce as needed. Although this understanding of the function of templates
and of the mechanism underlying generalization and analogy (Macken 1979,
this volume, Chapter 5, p. 144) is relatively recent, the core papers depict
essentially the same learning sequences and the same conclusion as to the role
of templatic patterns as a way of dealing with challenges by bringing familiar
routines to bear on them.
Vihman and Velleman (2000) introduced the terminological distinction
between selected words, or child word forms that (roughly) match the form
of their adult targets while conforming to a childs preferred prosodic structure,
and adapted words, or word forms based on adult targets that are less similar
to the childs pattern, which the child thus modies more radically to arrive at
an output that ts the template. Examples of both selected and adapted words
can be found in many of the chapters of this book (see also Keren-Portnoy,
Majorano, and Vihman, 2009). The earliest papers make no mention of the
term template, let alone of selecting and adapting, yet the detailed data
presented by Priestly (1977, this volume, Chapter 7), for example, make it easy
to see that some words, such as lion, produced as [lajn], and whale [wjl]
(bisyllabic ordinary forms in Priestlys terms), are selected in our sense,
2 Marilyn M. Vihman and Tamar Keren-Portnoy
while others the bisyllabic experimental forms, which Priestly found to be
not only amusing but systematic (p. 217) are adapted: e.g., berries [bjas],
chocolate [kajak], peanut [pijat], and tiger [tajak].
From the childs point of view, there is presumably no essential difference,
except perhaps of degree, between the two kinds of words: things that are
similar are treated similarly. The targets for selected words are similar to
other selected word targets as well as to the childs own forms of those words.
The targets for adapted words are not as obviously similar to one another, yet
they must sufciently resemble other words rendered within the framework of
that particular template to attract the child into associating themwith the same
type of own (child) form. It is typical of the forms used under the inuence
of a childs dominant template that no attempt (by researchers) to separately
trace or relate each segment to its presumed model in the target word will yield
a satisfactory analysis (this is well exemplied by the data in both Waterson
1971 and Macken 1979, this volume Chapters 3 and 5 respectively, as both
investigators emphasize). Instead, we see the child matching the overall shape
of the adult word (CVC[C]V[C] in Priestlys examples), often including the
target syllable count, as here, and at least one of the consonants, while simplify-
ing the overall structure through repetition of segments or syllables or through
reordering to achieve a xed output structure for multiple lexical items
(here, CVjVC). In short, the term template is used to formalize the notion
of whole-word learning as the basis of a childs phonology.
It is important to note that templates are not a lasting element in a childs
phonological system, even for children learning the classic templatic adult
language, Arabic (see Khattab and Al-Tamimi, this volume, Chapter 14).
Instead, templates typically gain increasing dominance over a period of days,
weeks or months often beginning toward the middle or end of the single-word
period but then fade thereafter, as the child comes to master (in terms of
articulation, speech planning, and memory or representation) the more complex
sequences of the adult language: see Priestlys and Oliveira-Guimares
accounts of the rise and fall of templates in the phonological development of
one English and two Brazilian children respectively (this volume, Chapters 7
and 10), as well as Macken (1979, this volume, Chapter 5), for the emergence
of templates and the subsequent advance to accurate segmental sequences in
the speech of a Spanish-learning child, and Vihman and Vihman (2011), a
longitudinal account of the emergence, use, and fading of two templatic patterns
in a diary study of an Estonian- and English-learning childs rst 500 words.
Finally, note that the templatic shape itself is dynamic, changing in more or
less subtle ways over the period of time in which it holds sway as the childs
phonological knowledge increases and stabilizes, often with a period of
competition between variant solutions to the phonological challenge (see
Priestly 1977, this volume, Chapter 7; Macken 1979, this volume, Chapter 5;
Vihman and Velleman 1989, this volume, Chapter 8; Vihman, Velleman, and
McCune 1994, this volume, Chapter 9; and Oliveira-Guimares, this volume,
Introduction: the emergence of phonology 3
Chapter 10, as well as Menn and Matthei 1992, who discuss competition in
child rules or patterns).
Universals vs. typological and individual differences: the role
of rhythm
How does the child get started learning the phonetics and the phonology of the
ambient language? What resources are available for kick-starting the process?
It is worth considering the role of rhythm, in both perception and production,
as a theoretical and developmental starting point for the child, and one which
may go some way toward accounting for three separate aspects of child vocal
production: its initial universality, its typological variability by language of
exposure, and the individual differences found even within a single language
group, all of which are amply illustrated in the chapters of this book.
The earliest theoretical statements about the course of phonological
development those of Jakobson (1941/1968) were based on diary studies,
with their inevitable focus on the individual child and his or her early word
production. Nevertheless, the conclusions of that highly inuential rst attempt
at systematization, heavily shaped by the structuralist theoretical principles of
the Prague School of linguistics of which Jakobson was a key member, were
meant to serve as putative universals. Somewhat later, Brown (1958) provoca-
tively hypothesized that babbling thought by Jakobson to be unrelated to later
phonological development involved a phonetic drift in the direction of
the ambient language. It was only later still, when the wide availability of
rst audio and then video recording devices made possible far more reliable and
detailed phonetic observations of childrens speech and especially of their
prelinguistic vocalizations, that the wide range of individual differences in
pathways to language (even for children acquiring the same language) began
to become evident from production studies (see, e.g., Vihman, Macken, Miller,
Simmons, and Miller 1985; Vihman, Ferguson, and Elbert 1986; Menn and
Vihman 2011). All three of these characteristics of phonological development
must be encompassed in our understanding of this complex process: universals,
or the commonalities to be found in the babble and rst word production of
children learning any language; ambient language effects and their implications
for the mapping of what is perceived onto vocal production; and the variability
due to the contribution of the individual child, within the constraints of percep-
tion, the neurophysiology of vocal production, and cognitive development.
Perceptual experience of the dominant rhythms of the ambient language can
be taken to provide a phonological frame suitable for supporting rst word
forms (see Wauquier and Yamaguchi, this volume, Chapter 11, for evidence
of the impact of rhythm on template formation in French). In other words,
perceptual experience of the specic rhythms of the language will yield a typical
one- or two-syllable unit, based on stress and syllable type and weight, which
a childs immature and inexperienced phonological memory will retain and
4 Marilyn M. Vihman and Tamar Keren-Portnoy
use, rst in implicit segmentation (Nazzi, Iakimova, Bertoncini, Frdonie, and
Alcantara 2006; Hhle, Bijeljac-Babic, Herold, Weissenborn, and Nazzi 2009;
Pons and Bosch 2010), then in early attempts at production; this in turn will
tend to strengthen the patterns that the child has tuned into, resulting in more
ambitious targeting of adult words (i.e., of word targets beyond the childs
production abilities), which are thus adapted to a well-practiced pattern or
template. (For evidence that phonological memory is constructed through
use, see Keren-Portnoy et al. 2010.) The cross-linguistic data provided in this
volume are largely consistent with this proposal, as child templates are shaped
by target language affordances whose scope is typically a lexical unit (a word or
a short phrase) in interaction with the childs own babbling practice and rst
word production experience (through selecting and adapting).
The evidence from templates suggests that rhythm is critical here, providing a
perceptual envelope into which the childs individual production patterns can
be tted. As Brown anticipated, individual childrens vocal practice (babble)
gradually drifts toward (or is shaped by) the rhythms of input speech
(Boysson-Bardies, Hall, Sagart, and Durand 1989; Boysson-Bardies and
Vihman 1991); this implicit sensorimotor experience of babbling is a critical
mechanism for transforming heard speech patterns into the production base for
word learning a different base for different children, despite broadly similar
input and neuromotor constraints. It is this prosodic framing of speech sequences
that eventually leads to the individual but ambient-language-inuenced phono-
logical templates.
In contrast to the implicit shaping of babble by perceived input speech, the
integration of what is heard with what can be produced as learned word forms
is neither automatic nor effortless. Furthermore, because this integration will
depend on such individually variable factors as the particular characteristics of a
childs babble, emergent representational ability, and volubility or sociability,
among other things, we should not be surprised at the wide variability identied
in production even among children learning the same language. In general, the
patterns that we nd described in this volume, for one or more children per
language, broadly reect the prosody of the individual language and support
the notion that rhythm is an important starting point for phonology (for further
discussion, see Wauquier and Yamaguchi, this volume, Chapter 11).
Whole-word learning from the perspective
of an exemplar model
In what sense is the phenomenon that we have been describing whole-word
learning? The childs rendition of the word shows sensitivity to some of the
segments or phonetic features that occur in it. However, it does not necessarily
maintain the order in which the segments or features appear, but may instead
redistribute, merge or spread some of those features. This is seen as whole-word
learning because it is within the lexical unit (a word or a short but often repeated
Introduction: the emergence of phonology 5
phrase: all gone, in there, whats this?) that this lack of conformity to the
identity or ordering of parts is observed. Within the lexical unit there may be
no clear evidence that the child has registered information about the identity
and number of all of the segments (as perceived by adult speakers) or their
relative order. Based on evidence fromproduction, then, children seemto have a
representation or memory trace of the adult form, but that representation is not
constructed out of an ordered sequence of segments.
This claim, that the childs representation lacks a clear structure made up of
neatly ordered parts, has often been misunderstood. The failure to appreciate
what is meant here has led some researchers to ascribe to the proponents of
whole-word phonology, or holistic representations, the claim that such rep-
resentations are vague or underspecied (Gerken, Murphy, and Aslin 1995;
Swingley and Aslin 2002, 2007; Storkel and Maekawa 2005); holistic repre-
sentations are contrasted here with segmentally detailed representations
(Storkel and Maekawa 2005) that are characterized by phonetic specication
(Swingley and Aslin 2002). Gerken et al. (1995) aptly present this viewpoint:
Children represent early words in terms of holistic properties, such as prosodic
structure and acoustic shape, or in terms of phonetic features that are not
bundled into individual segments (p. 476). In fact, as we understand them,
these child representations include abundant detail much more than is appa-
rent in phonemic or even broad phonetic description.
Taking an exemplar model perspective on whole-word learning, let us con-
sider what whole-word learning might be like. As suggested by Pierrehumbert
(2003), the perceptual input for speech is an auditory coding of the speech
signal. Acovering map provides an analog representation of the phonetic space,
with the dimensions being the many phonetic parameters which are relevant to
speech perception (p. 132; see also Edwards, Munson, and Beckman 2011).
Thus, for infants the representation is highly detailed, perhaps hyper-detailed,
or even overly detailed in some aspects but less so in others. In addition,
since infants need not at rst know which acoustic parameters are relevant for
speech perception, they may assign weights to parameters differently than
adults would.
Something Pierrehumbert does not mention, but which may also affect infor-
mation processing in the young child, is salience: parts of the acoustic signal
which are less readily perceived (shorter, lower pitch, quieter typically,
unstressed) may be processed less successfully, with more error or more loss
of information (as shown in Vihman, Nakai, DePaolis, and Hall 2004). Since
unstressed parts of words also tend to be produced with increased motoric
variability (Goffman, Gerken, and Lucchesi 2007), the unstressed parts of differ-
ent exemplars of the same word would differ more, leading those parts to be less
coherently represented; that is, their representations would contain more varia-
bility or noise. In a noisy or variable exemplar space the treatment of a newly
encountered exemplar as belonging, or not, to the particular category will be less
consistent.
6 Marilyn M. Vihman and Tamar Keren-Portnoy
Note that our claim is that young childrens exemplar space is sparser and
more variable than adults, with less clearly dened clumps or categories, and
that it therefore functions with less clearly dened boundaries for what does or
does not fall within each category. The more frequently a child encounters or
produces the exemplars of a given lexical type or structure, the sharper will be
the organization of the corresponding portion of exemplar space (see Ota, this
volume, Chapter 15).
There is no vagueness or lack of detail in this scenario. What is lacking
is segmental organization, or a tidy organization into sequentially ordered
time-bound units, each built of a unique co-occurring set of features. It is this
abstract level of categorization that is missing, not ne detail. In this sense
the childs representations, based on the evidence of phonological templates,
is both richer and poorer than what is implied by standard phonetic tran-
scription: It is rich in featural texture but poor in sequential organization. Nor
does the interpretation of child phonological representations as lacking seg-
mental units constitute a problem for continuity between child and adult
phonological knowledge, to the extent that some theoretical models similarly
deny any such organization for adult representations (Browman and Goldstein
1989, 1991, 1992; Pierrehumbert 2003; Edwards, Beckman, and Munson
2004; Edwards, Munson, and Beckman 2011; Munson, Edwards, and
Beckman 2012).
The orientation of this volume
In the 1970s three papers appeared that have since become classics: Waterson
(1971, this volume, Chapter 3) took a Firthian approach to one childs phonol-
ogy and introduced the notion of schemas, or child-specic word patterns;
Ferguson and Farwell (1975, this volume, Chapter 4) argued for whole-word
or lexical patterns as the core of adult as well as child phonological knowledge;
and Macken (1979, this volume, Chapter 5) demonstrated the unusual adult-to-
child-form mappings that can be found in early phonology to meet the childs
constraints on output forms. These papers all stood outside of phonological
theory as it was understood at the time, shortly after publication of the denitive
statement of generative phonology, Chomsky and Halle (1968). As it happened,
that formalization was about to be superseded by the range of new perspectives
that emerged in response to the perceived limitations of Chomsky and Halles
approach (see Van der Hulst and Smith 1982; Anderson 1985; Goldsmith 1995;
and Scheer 2013).
3
This period in the study of phonological development
culminated in the widely cited paper by Menn (1983, this volume, Chapter 6),
who adopted a psycholinguistic perspective and formulated the two-lexicon
model (for a rethinking of this model, see Menn and Matthei 1992; Menn,
Schmidt, and Nicholas 2009, as well as this volume, Chapter 17).
In the period that followed, phonological theory blossomed and expanded,
diversifying into a range of distinct theories, including CV phonology (Clements
Introduction: the emergence of phonology 7
and Keyser 1983), Lexical Phonology (Mohanan 1986), Autosegmental and
Metrical Phonology (Goldsmith 1990), Dependency Phonology (Durand
1990), Government Phonology (Kaye, Lowenstamm, and Vergnaud 1990),
Declarative Phonology (Coleman 1998), and most recently CVCV phonology
(Scheer 2004).
4
However, during the 1990s one of the new models, Optimality
Theory (OT: Prince and Smolensky 1992/2004), began to dominate the eld, to
the point that it came often to be the only theoretical perspective presented to
linguistics students. A number of attempts have been made to cast phonological
development in terms of OT(see Boersma and Levelt 2003); the studies collected
in Kager, Pater, and Zonneveld (2004) are dedicated to the presentation of
acquisition data from an OT perspective. Yet no extensive OT treatment of data
fromone or more children has appeared to date. The present volume returns to the
whole-word phonology approach, which has much in common with the early
work of McCarthy and Prince (1986, 1993, 1995) but which, on the basis of
extensive cross-linguistic studies of child data, diverges sharply from OT, with
its reliance on Universal Grammar and markedness theory and the tendency of
its advocates to expect linear advances along with set stages of development and
across-the-board changes in child forms.
As Menn and her colleagues point out in their recent efforts to model what we
know about how children learn phonology (Menn et al. 2009; see also Menn
et al. this volume, Chapter 17), any adequate theory of phonological develop-
ment must be able to account for three key ndings, all solidly grounded in forty
years of empirical research:
1. individual differences across children,
2. lexical variation within a given child,
3. the phenomenon of regression (nonlinear advance, or the U-shaped curve),
in which early accuracy is succeeded by less accurate, more child-specic
word forms only to be followed, much later, by a return to adultlike forms,
or relative accuracy.
No theoretical approach that sees phonological development as the automatic
suppression of innate processes (Stampe 1969), across-the-board changes in rule
application (Smith 1973), triggering of parameters (Fikkert 1994) or reordering
of constraints (various chapters in Kager et al. 2004) can account for these core
characteristics in any straightforward way.
Vihman and Croft (2007, this volume, Chapter 2) propose a new way of
thinking about phonology, based on the ideas of Charles Ferguson and the
evidence from child data, with specic reference to the three characteristics
listed above. Menn and her colleagues (this volume, Chapter 17), who focused
on OT in their 2009 critique of what has been missing in theories of
phonological acquisition to date, now propose to extend Vihman and Crofts
exemplar model by including some key missing elements namely, (i) a role
for representation of the adult target form; (ii) mappings from input to output
(corresponding to the inuential rules or processes of earlier generative
models such as Stampe 1969 and Smith 1973); and (iii) mappings from output
8 Marilyn M. Vihman and Tamar Keren-Portnoy
to input, this latter an important characteristic not yet incorporated in any purely
phonological theory.
The concluding section of this volume includes both Menn et al.s future-
oriented proposals for extension and revision of Vihman and Crofts model
and McCunes developmentally oriented thoughts on the importance of the
study of phonological development as a whole, and of early word templates
in particular, for child language development. These discussion chapters
clash on a number of specic points, which can only be a healthy symptom
of the liveliness of creative thinking in our eld, even within the scope of
a broadly similar theoretical (here, functionalist) inclination. Beneath the
evident differences in dening the notion of representation, in particular
the editors of this volume nd some deeply rooted similarities. Specically,
the interrelatedness of representations in a network of potential associations,
both formal and meaning-related, within a broadly neurological framework
seems to us to emerge from both of these chapters. Based on this idea we can
conceptualize the potentiality or transitory nature of representations, such
that they only come into existence in moments of what, in adults, could be
termed consciousness (a better expression, for developmental purposes,
might be moments of use, for either speaker or listener). To return to the
exemplar metaphor, individual instances (or their subparts) are activated to
differing degrees in different situations of use, when activation lights up
differing elements in the network of associations. This would mean that
the representation of a given item has no essential stability over time,
especially in the early period of phonological development. It is our hope
that the contrasting perspectives provided will lead to discussion, debate, and
further empirical research.
The contents of this volume
Three basic considerations guided the choice of papers for this volume. First,
in Part II (Setting papers) we sought to provide a fair representation of the
core papers that gave rise to the whole-word approach. Secondly, in Part III
(Cross-linguistic studies) we included empirical papers that work through the
implications of this approach. We limited ourselves to data-oriented papers in
which the childs word forms are presented in sufcient numbers to give the
reader a clear understanding of the shape of his or her emerging phonology,
excluding papers with only anecdotal mention of specic forms to support a
rule or constraint. Most of the chapters in this volume also take an overtly whole-
word approach, but that is not the case with all of them(neither Ota, Chapter 15,
nor Priestly, Chapter 7, have any such explicit orientation, for example).
Thirdly, this volume is specically designed to provide data and analysis
exemplifying the ways in which these fundamental properties of phonological
development are manifested in a range of different languages and child learners.
The diversity of languages included makes it possible to document, for
Introduction: the emergence of phonology 9
example, the exceptional salience of geminates for children acquiring languages
that have them here, Finnish (Savinainen-Makkonen, Chapter 13) and
Arabic (Khattab and Al-Tamimi, Chapter 14) or the unavoidability of
learning clusters early in a language in which they are particularly common
(and one childs solution to that problem: see Szreder, Chapter 12). (Of the
chapters in Part II, only Mackens, Chapter 5, concerns a language other than
English.) This will enable a serious student of child phonology to see just what
kinds of data need to be accommodated, not in one child or one language alone
but in a broad sample and also to see the lines of similarity, patterns, and
limitations or constraints that recur in one child and one language after another,
although not always in the form predicted by adult-theory-based notions of
markedness (see also Vihman and Kunnari 2008). However, the typological
study of child phonological development is certainly still in its infancy. We hope
that this volume will stimulate empirical studies of children learning a far wider
range of languages.
We believe that whole-word phonology (in Ferguson and Farwells
terms), or Templatic Phonology (Vihman and Croft 2007), provides a
model that, while still limited in many ways, is at a minimum faithful to the
evidence afforded by large quantities of child data. This is one of the key
legacies of Charles Fergusons approach: the empirical data are allowed to
speak although the interpretation will necessarily be inuenced by the
investigators training and habits of mind. We hope that this edited volume,
with its mix of classics and both old and new data-based papers as well as
contemporary re-evaluations from the points of view of both linguistics and
developmental psychology, will bring the whole-word phonology approach to
the attention of a new generation of linguists, psychologists, psycholinguists,
and speech scientists.
notes
1. Papers included in this volume are indicated in bold face; those published here for
the rst time are cited without mention of year of publication.
2. Some of these latter studies are specically designed to challenge the notion of
whole-word phonology; we discuss below some of the ways in which this idea
has been interpreted (or misinterpreted). Although we cannot here enter into a
discussion of the differences between experimental responses to a limited number
of stimuli and spontaneous speech production, Vihman, DePaolis, and Keren-Portnoy
(2009) discuss the issue briey, while Vihman (forthcoming 2014: ch. 7) is devoted to
Experimental studies of word form learning.
3. Smith (1973), still the most extensive analysis of any one childs phonology, came out
in the same period. Smiths study held closely to the mainstream formalization
represented by Chomsky and Halle and purported to demonstrate in direct contra-
diction to the studies included in this volume that there is no basis for assuming that
the childs word forms reect an independent system.
4. Some of these models overlap with or subsume others.
10 Marilyn M. Vihman and Tamar Keren-Portnoy
References
Anderson, S. R. (1985). Phonology in the twentieth century: theories of rules and
theories of representations. Chicago University Press.
Boersma, P. and Levelt, C. (2003). Optimality theory and phonological acquisition.
Annual Review of Language Acquisition, 3, 150.
Boysson-Bardies, B. de, Hall, P., Sagart, L., and Durand, C. (1989). A crosslinguistic
investigation of vowel formants in babbling. Journal of Child Language, 16, 117.
Boysson-Bardies, B. de and Vihman, M. M. (1991). Adaptation to language: evidence
from babbling and rst words in four languages. Language, 67, 297319.
Browman, C. P. and Goldstein, L. (1989). Articulatory gestures as phonological units.
Phonology, 6, 20151.
(1991). Gestural structures: distinctiveness, phonological processes and historical
change. In I. G. Mattingly and M. Studdert-Kennedy (eds.), Modularity and the
motor theory of speech perception, pp. 31338. Hillsdale, NJ: Lawrence Erlbaum.
(1992). Articulatory phonology: an overview. Phonetica, 49, 15580.
Brown, R. (1958). Words and things. Glencoe, IL: Free Press.
Chomsky, N. and Halle, M. (1968). The sound pattern of English. New York: Harper &
Row.
Clements, G. N. and Keyser, S. J. (1983). CV Phonology: a generative theory of the
syllable. Cambridge, MA: MIT Press.
Coleman, J. S. 1998. Phonological representations: their names, forms and powers.
Cambridge University Press.
Durand, J. (1990). Generative and nonlinear phonology. London: Longman.
Edwards, J., Beckman, M. E., and Munson, B. (2004). The interaction between vocabulary
size and phonotactic probability effects on childrens production accuracy and uency
in nonword repetition. Journal of Speech, Language, and Hearing Research, 47,
42136.
Edwards, J., Munson, B., and Beckman, M. E. (2011). Lexiconphonology relationships
and dynamics of early language development a commentary on Stoel-Gammons
Relationships between lexical and phonological development in young children.
Journal of Child Language, 38, 3540.
Eimas, P. D., Siqueland, E. R., Jusczyk, P. W., and Vigorito, J. (1971). Speech perception
in infants. Science, 171, 3036.
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language acquis-
ition. Language, 51, 41939. Reprinted in W. S-Y. Wang, The lexicon in phono-
logical change. The Hague: Mouton (1977). Reprinted in this volume as Chapter 4.
Fikkert, P. (1994). On the acquisition of prosodic structure. PhD dissertation, University
of Leiden (HIL Dissertations 6). The Hague: Holland Academic Graphics.
Gerken, L. A., Murphy, W. D., and Aslin, R. N. (1995). Three- and four-year-olds
perceptual confusions for spoken words. Perception and Psychophysics, 57,
47586.
Goffman, L., Gerken, L. A., and Lucchesi, J. (2007). Relations between segmental and
motor variability in prosodically complex nonword sequences. Journal of Speech,
Language, and Hearing Research, 50, 44458.
Goldsmith, J. A. (1990). Autosegmental and metrical phonology. Oxford: Blackwell.
(ed.) (1995). The handbook of phonological theory. Oxford: Blackwell.
Hhle, B., Bijeljac-Babic, R., Herold, B., Weissenborn, J., and Nazzi, T. (2009).
Language-specic prosodic preferences during the rst half year of life: evidence
from German and French infants. Infant Behavior and Development, 32, 26274.
Introduction: the emergence of phonology 11
Jakobson, R. (1941/1968). Child language, aphasia, and phonological universals, trans.
A. R. Keiler. The Hague: Mouton. (Originally published as Kindersprache, Aphasie
und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell, 1941.)
Johnson, E. K. and Tyler, M. (2010). Testing the limits of statistical learning for word
segmentation. Developmental Science, 13, 33945.
Jusczyk, P. W. (1997). The discovery of spoken language. Cambridge, MA: MIT Press.
Kager, R., Pater, J., and Zonneveld, W. (eds.) (2004). Constraints in phonological
acquisition. Cambridge University Press.
Kaye, J., Lowenstamm, J., and Vergnaud, J. (1990). Constituent structure and govern-
ment in phonology. Phonology, 7, 193231.
Keren-Portnoy, T., Majorano, M., and Vihman, M. M. (2008). From phonetics to phonol-
ogy: the emergence of rst words in Italian. Journal of Child Language, 36, 23567.
Keren-Portnoy, T., Vihman, M. M., DePaolis, R., Whitaker, C., and Williams, N. M.
(2010). The role of vocal practice in constructing phonological working memory.
Journal of Speech, Language, and Hearing Research, 53, 128093.
Kuhl, P. K. (2004). Early language acquisition: cracking the speech code. Nature
Reviews Neuroscience, 5, 83143.
Macken, M. A. (1979). Developmental reorganization of phonology: a hierarchy of basic
units of acquisition. Lingua, 49, 1149. Reprinted in this volume as Chapter 5.
McCarthy, J. J. and Prince, A. S. (1986). Prosodic morphology. MS, University of
Massachusetts, Amherst, and Brandeis University.
(1993). Prosodic morphology, I: Constraint interaction and satisfaction. MS, University
of Massachusetts, Amherst, and Rutgers University, New Brunswick.
(1995). Prosodic morphology 1. In J. A. Goldsmith (ed.), The handbook of phono-
logical theory, pp. 31866. Oxford: Blackwell.
Menn, L. (1983). Development of articulatory, phonetic and phonological capabilities. In
B. Butterworth (ed.), Language production, vol. 2, pp. 350. London: Academic
Press. Reprinted in this volume as Chapter 6.
Menn, L. and Vihman, M. M. (2011). Features in child phonology: inherent, emergent, or
artefacts of analysis? In N. Clements and R. Ridouane (eds.), Where do phonolog-
ical features come from? The nature and sources of phonological primitives,
pp. 261301. Amsterdam: John Benjamins.
Menn, L. and Matthei, E. (1992). The two-lexicon account of child phonology. In
C. A. Ferguson, L. Menn, and C. Stoel-Gammon (eds.), Phonological development:
models, research, implications, pp. 21147. Timonium, MD: York Press.
Menn, L., Schmidt, E., and Nicholas, B. (2009). Conspiracy and sabotage in the
acquisition of phonology: dense data undermine existing theories, provide scaffold-
ing for a new one. Language Sciences, 31, 285304.
Mohanan, K. P. (1986). The theory of lexical phonology. Dordrecht: D. Reidel.
Munson, B., Edwards, J., and Beckman, M. E. (2012). Phonological representations in
language acquisition: Climbing the ladder of abstraction. In A. Cohn, C. Fougeron,
and M. Huffman (eds.), Oxford handbook of laboratory phonology, pp. 288309.
Oxford University Press.
Nazzi, T., Iakimova, G., Bertoncini, J., Frdonie, S., and Alcantara, C. (2006). Early
segmentation of uent speech by infants acquiring French. Journal of Memory and
Language, 54, 28399.
Pierrehumbert, J. (2003). Phonetic diversity, statistical learning and acquisition of
phonology. Language and Speech, 46, 11554.
Pons, F. and Bosch, L. (2010). Stress pattern preference in Spanish-learning infants: the
role of syllable weight. Infancy, 15, 22345.
12 Marilyn M. Vihman and Tamar Keren-Portnoy
Priestly, T. M. S. (1977). One idiosyncratic strategy in the acquisition of phonology.
Journal of Child Language, 4, 4566. Reprinted in this volume as Chapter 7.
Prince, A. and Smolensky, P. (2004). Constraint interaction in generative grammar. In
J. J. McCarthy (ed.), Optimality Theory in phonology: a reader, pp. 371. Oxford:
Blackwell. (Originally circulated as Optimality: constraint interaction in genera-
tive grammar, MS, 1992.)
Savinainen-Makkonen, T. (2007). Geminate template: a model for rst Finnish words.
First Language, 27, 34759. Reprinted in this volume as Chapter 13.
Scheer, T. (2004). A lateral theory of phonology: what is CVCVand why should it be?
Berlin: Mouton de Gruyter.
(2013). Aspects of the development of generative phonology. In N. C. Kula, B. Botma,
and K. Nasukawa (eds.), The Bloomsbury companion to phonology, pp. 397446.
London: Bloomsbury.
Smith, N. V. (1973). The acquisition of phonology: a case study. Cambridge University
Press.
Stampe, D. (1969). The acquisition of phonetic representation. Papers from the Fifth
Regional Meeting of the Chicago Linguistic Society, Chicago, IL. Reprinted in
D. Stampe, A dissertation on natural phonology. New York: Garland, 1979.
Storkel, H. L. and Maekawa, J. (2005). A comparison of homonym and novel word
learning: the role of phonotactic probability and word frequency. Journal of Child
Language, 32, 82753.
Swingley, D. and Aslin, R. N. (2002). Lexical neighborhoods and the word-form repre-
sentations of 14-month-olds. Psychological Science, 13, 4804.
(2007). Lexical competition in young childrens word learning. Cognitive Psychology,
54, 99132.
Thiessen, E. D. and Saffran, J. (2007). Learning to learn: infants acquisition of stress-based
strategies for word segmentation. Language Learning & Development, 3, 73100.
van der Hulst, H. and Smith, N. (1982). The structure of phonological representations.
Dordrecht: Foris.
Vihman, M. M. (Forthcoming 2014). Phonological development: the rst two years. 2nd
edn. Oxford: Blackwell.
Vihman, M. M. and Croft, W. (2007). Phonological development: toward a radical
templatic phonology. Linguistics, 45, 683725. Reprinted in this volume as
Chapter 2.
Vihman, M. M., DePaolis, R. A., and Keren-Portnoy, T. (2009). A Dynamic Systems
approach to babbling and words. In E. Bavin (ed.), Handbook of child language,
pp. 16382. Cambridge University Press.
Vihman, M. M., Ferguson, C. A., and Elbert, M. (1986). Phonological development from
babbling to speech: common tendencies and individual differences. Applied
Psycholinguistics, 7, 340.
Vihman, M. M. and Kunnari, S. (2008). The sources of phonological knowledge: a cross-
linguistic perspective. Recherches Linguistiques de Vincennes, 35, 13364.
Vihman, M. M., Macken, M. A., Miller, R., Simmons, H., and Miller, J. (1985). From
babbling to speech: a reassessment of the continuity issue. Language, 61,
395443.
Vihman, M. M., Nakai, S., DePaolis, R. A., and Hall, P. (2004). The role of accentual
pattern in early lexical representation. Journal of Memory and Language, 50,
33653.
Vihman, M. M. and Velleman, S. (1989). Phonological reorganization: a case study.
Language and Speech, 32, 14970. Reprinted in this volume as Chapter 8.
Introduction: the emergence of phonology 13
(2000). Phonetics and the origins of phonology. In N. Burton-Roberts, P. Carr, and
G. Docherty (eds.), Phonological knowledge: its nature and status, pp. 30539.
Oxford University Press.
Vihman, M. M., Velleman, S. L., and McCune, L. (1994). How abstract is child phonol-
ogy? Towards an integration of linguistic and psychological approaches. In M. Yavas
(ed.), First and second language phonology, pp. 944. San Diego: Singular
Publishing. Reprinted in this volume as Chapter 9.
Vihman, M. M. and Vihman, V-A. (2011). From rst words to segments: a case study in
phonological development. In E. V. Clark and I. Arnon (eds.), Experience, variation
and generalization: learning a rst language, pp. 10933. Amsterdam: John
Benjamins.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
14 Marilyn M. Vihman and Tamar Keren-Portnoy
Part I
The current framework
2 Phonological development: toward a radical
templatic phonology
Marilyn M. Vihman and William Croft
1. Introduction
In this chapter we argue for a template-based approach to segmental phonological
representation. Our central theoretical hypothesis is that the segmental phono-
logical structure of words is represented as language-specic phonotactic
templates (the latter including syllable structure and other higher-order
structures such as metrical structure).
1
We present cross-linguistic evidence
from phonological development that supports a template-based approach to
phonological representation. We argue, however, that the template-based
approach is equally suited to the analysis of adult phonology. Research in
more phonetically oriented approaches towards phonological categories, and in
usage-based or exemplar models of the representation of phonological knowl-
edge, also supports a template-based approach to the representation of the
phonological structure of words. We take this research (and ours) to its logical
conclusion, arguing that it applies to more abstract phonological categories and
adult phonologies as well. Before turning to this evidence, we briey discuss three
general issues that have led us to this approach to phonological representation.
The rst issue is the relationship between language structure and language
function, namely, communication for the purposes of social interaction
(see Clark 1996; Keller 1994). The hypothesis that we propose, following
many others, is that the starting point for the analysis of linguistic structure
should be the soundmeaning link that denes linguistic signs or symbols. This
hypothesis does not rule out the possibility that generalizations about linguistic
structure, including phonological structure, may be separated from general-
izations about their function. Indeed, there is much arbitrariness in language,
most notably the arbitrariness of the association of a phonological form with a
particular meaning in a particular language. Also, as is well known, the phono-
logical organization of a word into syllables often fails to match the morpho-
logical composition of a word. But we will argue below that the basic
phonological unit is a word template, specically dened on a phonological
unit that is also a fundamental symbolic unit.
2
We will argue that starting from
words can solve certain theoretical and empirical problems that arise for reasons
not directly connected to language function and, furthermore, that this reects
the developmental learning sequence.
17
The second issue is the empirical range of a linguistic theory. A central
fact about linguistic data is the pervasiveness of variation: variation across
languages, across dialects, across speakers, across utterances by an individual
speaker, and also variation in the behavior of linguistic units across linguistic
contexts. We do not believe it is appropriate to abstract away from empirical
variation, or to attempt to explain it away (e.g., by positing separate invariable
grammars; see, e.g., Croft 2000: 513). Instead, we seek a model of grammatical
representation that will accommodate this variation. The need to accommodate
the full range of variation observed within and across languages will play a central
role in our arguments for a template-based approach to segmental phonology.
The third issue is the relationship between a linguistic theory and psychological
plausibility. In many linguistic theories, it is common to separate grammatical
competence from performance, and to evaluate competence theories on the basis
of principles of simplicity and generality, leaving aside performance or even the
precise psychological implementation of the competence module. But simplicity
and generality are a priori formal criteria, not psychological ones. Moreover,
separating competence from performance makes it impossible to subject compe-
tence models to empirical psycholinguistic evaluation.
We consider it to be preferable (other things being equal) to posit a unied
model of grammatical representation that does not separate a competence
module from its psychological implementation, or from actual language pro-
cessing (compare Bybee 2001: 8). In particular, psycholinguistic evidence
should be relevant to the evaluation of theories of grammatical representation.
In this chapter, we focus on the representation of the phonological structure of
linguistic units. We draw on another type of psycholinguistic evidence, namely,
that afforded by language development, to support a template-based approach
to phonological representation.
The developmental data that we bring to bear on the question of word
templates in phonology raises a nal general issue: the relationship between
child language data and data derived from adult linguistic behavior. Only the
latter are normally used as a basis for theories of linguistic representation. Such
theories are then applied to rst language acquisition data. Often there are
substantial discrepancies between the hypothesized adult system and the devel-
oping child system. In this situation, two opposing proposals are typically made.
The discontinuity hypothesis maintains that the process by which
language is learned and the representations developed by the child are different
from those that are found in the adult system and must therefore somehow be
replaced by the adult system at a later stage of development. The discontinuity
hypothesis is unattractive because it seems to make little or no connection
between what the child knows and does and what the adult knows. It also
appears to insulate the theory of the adult system from any potentially discon-
rming data from child language development.
The continuity hypothesis maintains that the child already knows the
adult system (because many aspects of it are innately specied). The inability of
18 Marilyn M. Vihman and William Croft
the child to exhibit adult linguistic behavior is taken to be due to performance
and other limitations (or in one variant, to the need for innate capacities to
mature over time). The continuity hypothesis is also unattractive in that it too
appears to insulate the competence model of the adult from any potentially
disconrming developmental data.
We suggest that it is preferable to develop a theory of linguistic representation
that draws on developmental as well as adult data from the outset. Such a theory
will view the development of knowledge of linguistic structure as a gradual
process, assuming neither full adult competence from the beginning nor a
discontinuity between developmental stages and adult outcome. The template-
based approach to segmental phonology constitutes such a theory. It proposes
that a limited number of specic, actual word shapes are the rst steps in
phonological learning. The child gradually develops rst one or a small number
of phonological templates, then a wider variety of them, while at the same time
inducing a range of other phonological categories and structures from the
known word shapes. The result of differentiating and generalizing knowledge
of the phonological structure of words in the course of language acquisition is
an adult template-based model of phonological representation, with neither
discontinuity nor an assumption of pre-specied adult competence.
2. Word templates in early phonological development
2.1. A brief history
For over thirty years child phonologists have been claiming that the earliest
phonological structure is whole-word based. Perhaps the simplest expression of
the idea is that of Francescato (1968: 148) (who makes reference to Reichling
1935): Children never learn sounds: They only learn words, and the sounds are
learned through words. At the time that the idea was rst seriously put forward,
infant speech perception had not yet begun to be investigated and there were
few, if any, acoustic studies of childrens word production. Nevertheless, the
pioneering studies in child phonology made some fundamental observations,
while later, more detailed studies have provided further support for the basic
idea of whole-word phonological development.
In 1971 two diary studies, one American (Menn), one British (Waterson,
whose work is rooted in the Firthian tradition; see also Menn 1983; Waterson
1987), provided empirical data that seemed to point to the idea that the
whole word was at the core of a childs early phonology. Concluding a close
analysis of her son Daniels rst words, Menn (1971) suggested that the facts
that simplifying is principally by assimilation embracing the whole monosyl-
lable, all simplifying is done within word boundaries, [and] . . . there is no
conditioning across word boundaries indicate that the word is an entity, stored
and accessed as a block (Menn 1971: 247, emphasis ours). Daniels
assimilation embracing the whole monosyllable generally involved velar
Phonological development: toward a radical templatic phonology 19
harmony (e.g., at 22 months, when systematic forms began to appear: [gk]
cracker, [gg] bug, [gk] truck).
It has since become clear, partly through Menns own later work, that a
number of qualications have to be made to this summary of the facts.
We now know that conditioning can also occur across word boundaries, for
example (see Donahue 1986; Stemberger 1988; Matthei 1989; Menn and
Matthei 1992). Furthermore, there is no reason to equate the word with the
monosyllable, outside of an English language context. Disyllables dominate the
early lexicon of children acquiring most of the other languages in which early
word phonology has been extensively investigated, through either diary or
observational studies (Estonian, Finnish, French, Greek, Hebrew, Hindi,
Italian, Japanese, Spanish, Swedish, Welsh). The Germanic languages generally
may constitute exceptions, as monosyllables appear to be the most common
early word form in Dutch (e.g., Elbers and Ton 1985) and German (Leopold
1939; Elsen 1996) as well as English; for Swedish our data showthat mono- and
disyllabic early word forms are in close balance. Table 2.1 indicates proportions
of word targets of differing lengths in a cross-linguistic sample of early word
data, with 325 children represented in each language group. However, it
remains the case that the fundamental intuition that whole words are at
the core of early phonology was convincingly illustrated in Menn 1971 and
Waterson 1971 for the rst time.
Table 2.2 illustrates the type of phenomenon with which Waterson 1971 was
concerned, drawing on data from her son P.
This childs forms are less closely related to their adult targets than were those
that Menn reported for Daniel. Perhaps for this reason Waterson draws more
radical conclusions in attempting to account for her ndings:
Table 2.1. Mean length in syllables for early word targets in seven
languages
a
(ordered by proportion of monosyllables)
Language (N children) 1-syl. 2-syls. 3+-syls. Mean words per child
English (5) .59 .35 .06 120
Swedish (5) .44 .52 .04 106
Welsh (5) .36 .54 .10 53
Estonian (3) .33 .58 .09 48
French (5) .28 .68 .04 114
Finnish (10) .18 .79 .03 133
Italian (25) .17 .58 .26 22
Mean .34 .58 .09
a
For additional detail regarding the data summarized here, see Vihman (1996) (English
and French), Vihman, Boysson-Bardies, Durand, and Sundberg (1994a) (Swedish),
Vihman and DePaolis (2000), Vihman, Nakai, and DePaolis (2006) (Welsh), Krgvee
(2001), Salo (1993) and Vihman (1976) (Estonian), Kunnari (2000) (Finnish), and
DOdorico, Carubbi, Salerni, and Calvo (2001) (Italian).
20 Marilyn M. Vihman and William Croft
It . . . seems reasonable to consider that a child perceives some sort of schema in words or
utterances through the recognition of a particular selection of phonetic features . . . which
go into the composition of the forms of the words or groups of words, and this
recognition of a schema results in his producing words of the same type of structure
for such adult forms. (Waterson 1971: 206)
Unfortunately Watersons insistence on perception as the source of her sons
early word schemas was never convincingly supported by direct evidence
(see Waterson 1987 for some attempts to provide such evidence, however),
and the idea that the childs patterns derive from what is salient in the target
words, although plausible, remains only an idea, since the evidence so far
inheres primarily in the production data themselves a problematic circularity.
Ferguson, Peizer, and Weeks (1973) were sufciently impressed by their
data, drawn from a case study of Weeks granddaughter (see also Weeks 1974),
to assert that for the adult we may assume that the predominant [phonological]
unit is the phoneme . . . [whereas] for many children the earliest domain seems
to be the entire lexical unit . . . (p. 57). Two years later, basing themselves
primarily on their analysis of longitudinal rst word data from three children
(including those of the EnglishGerman bilingual child Hildegard, as docu-
mented by her father, Leopold 1939), Ferguson and Farwell (1975) published
the classic statement of the whole word position, which they extended to adult
phonology as well:
The data and analysis of this study suggest a model of phonological development and
hence of phonology which is very different from those in vogue among linguists.
The model would de-emphasize the separation of phonetic and phonemic development
[i.e., contra Jakobson 1941/1968], but would maintain in some way the notion of
contrast . . . It would emphasize individual variation . . . but would incorporate the
notion of universal phonetic tendencies . . . It would emphasize the primacy of lexical
items . . . but provide for a complex array of phonological elements and relations . . .
(Ferguson and Farwell 1975: 437)
This position has been cited repeatedly but has only recently begun to receive
empirical investigation. Studies with adults over the last ve years or so have
shown that phonotactic familiarity effects, based on relative frequency of
occurrence of segments and segmental sequences, facilitate (speed up) the
Table 2.2. Ps early word templates: nasal structure
(age 1;6) VV (adapted from Waterson 1971)
Child form Adult target
[aa] another
[ee], [ii] nger
[a] Randall
[ee] window
Phonological development: toward a radical templatic phonology 21
processing of nonwords, although competitive effects deriving from known
lexical items (similarity neighborhoods) tend to slow processing of real words
in dense neighborhoods (see Vitevich, Luce, Charles-Luce, and Kemmerer
1997; Vitevich and Luce 1998, 1999). Similarly, Beckman and Edwards
(2000a) found that familiarity with particular phonemic sequences resulted in
more accurate repetition of nonwords by three- to four-year-olds (see also
Edwards, Beckman, and Munson 2004).
The idea of whole-word phonology was further extended and more tightly
dened by Macken (1979), who summed up her analysis of the early phonology
of a Spanish-speaking child by noting that [a number of] unusual substitutions
can be accounted for by the overgeneralization of . . . preferred word patterns . . .
Prosodic similarity between certain adult words provides a plausible explan-
ation for the similar treatment of some words (p. 29). Macken alludes to word
templates here (preferred word patterns) and appears to be agreeing with
Waterson in nding a probable source for the childs patterns in the prosodic
similarity of words in the adult language. Based on her detailed longitudinal
case study, she goes on to adumbrate her ndings for the early word learning
period: all words had a consistent word pattern form; . . . new patterns resulted
from the expansion of previously acquired word patterns; some words changed
patterns over time as new word patterns were learned (Macken 1979: 34).
We will see that this description ts the data for any number of other children
for whom detailed phonetic lists of early words have been provided in the
intervening years. Macken (1996) indicates further that she sees word templates
as being identiable through the typical overgeneralization and conspiratorial
effects of the several rules that operate to produce [a particular] output e.g.,
metathesis (plus harmony) . . . , consonant epenthesis . . . , unusual deletion of
the input medial stressed V . . . (p. 169).
How solid, and how cross-linguistically valid, is the empirical basis for the
whole-word phonology idea in language development? The three arguments
that have been primarily used to support the concept are as follows:
1. Variability of segment production: A child may produce the same sounds
differently in different words, and some words may be more variable than
others. This suggests that the child has knowledge of particular words but
has not yet developed abstract categories of sounds for production (Ferguson
and Farwell 1975).
2. Relationship of child word to adult target: The relation of early child words
to their adult models is often found to be difcult to account for on a
segment-by-segment basis. Instead, the child seems to be targeting a whole
gestalt (Waterson 1971). The resulting patterns have been described as
whole-word processes, sometimes characterized as either harmony
(assimilation of noncontiguous vowels or consonants) or melody (patterning
in the sequencing of noncontiguous vowels or consonants) (Grunwell 1982;
Macken 1992, 1995; Vihman 1996).
22 Marilyn M. Vihman and William Croft
3. Relationship between child words: The interrelation between the childs own
words may be more evident than the relation to the adult models (Macken
1979). This is due to the childs eventual reliance on one or more word
templates, specic phonological patterns which t many of the words that
the child attempts (these words are said to be selected), but which are also
extended to words that are less close to the template (these words are then
adapted to t the template [Vihman and Velleman 2000]).
An additional argument can be proposed, with reference to the apparent basis
for developmental patterning that is distinct from the phonology of the adult
language:
4. Source of child patterns: The dominant child patterns of the early word
production period are responses to challenges posed by adult target words,
primarily, the challenge of producing distinct consonants or distinct vow-
els, or both, in different syllables or different word positions (i.e., initial
and nal consonants in a monosyllable, as in Daniel Menns forms, cited
above).
We will provide no specic developmental evidence here in relation to (1), the
variability in production of the same segment in different words, but such
evidence can be obtained from the more detailed of the various single-case or
small-group studies cited (see also Section 3.1 below). The evidence to be
provided in Section 2.2 (as well as in Table 2.2 above), based on data from
individual children, will serve to illustrate the remaining arguments, which are
complementary. Finally, we will indicate some of the differential effects of
ambient language rhythmic patterning on the shapes of early child templates in
Section 2.3, where we provide cross-linguistic data based on three to ten
children per language group.
The nature of the challenge that early word production poses to children has
yet to be satisfactorily established. Some have argued that the challenge is
primarily representational (memory difculties: see Vihman 1978; Macken
1979, among others) or articulatory (production difculties: Labov and Labov
1978; Studdert-Kennedy and Goodell 1995, among others); both speech
planning (Chiat 1979) and speech processing (Berg and Schade 2000) have
also been identied as plausible bases for childrens problems. Although
infants are known to have remarkable capacities for perceptual processing
(specically, for segmental discrimination) from the earliest months, so that
perceptual problems per se might seem an unlikely source of difculty,
3
it
has become increasingly clear that the deployment of these capacities in
relation to the discrimination of minimally distinct word forms requires addi-
tional attentional resources, at the very least, and constitutes a novel task for
one-year-olds (Stager and Werker 1997; Werker, Fennell, Corcoran, and
Stager 2002). Thus some combination of attentional or representational fac-
tors may be involved, although differences in motor control and practice must
also affect differences in production (McCune and Vihman 2001).
Phonological development: toward a radical templatic phonology 23
2.2. Evidence for word templates in early phonological development
In the earliest period of acquisition the idea of structure emerging from known
holistic phonological units can be demonstrated in its simplest, most direct
form. Menn (1971) observed that early phonological patterning is partly
determined by the shapes of the rst handful of words attempted (p. 246).
Later studies have made it clear that, contrary to Jakobsons (1941/1968) well-
known discontinuity view, the source of the shapes of the rst words is often
to be found in prelinguistic vocal practice, or babbling (Stoel-Gammon and
Cooper 1984; Vihman, Macken, Simmons, and Miller 1985; Vihman and Miller
1988; Elbers and Wijnen 1992; Vihman 1992; McCune and Vihman 2001), with
some effects of the ambient language on vocal production being identiable
even before rst word production (Boysson-Bardies, Hall, Sagart, and Durand
1989; Boysson-Bardies and Vihman 1991; for comparable effects in the seman-
tic domain, see Bowerman and Choi 2001).
The earliest word forms are thus typically closely related to the individual
childs babbling patterns (Vihman et al. 1985) as well as being relatively
accurate (Ferguson and Farwell 1975), and they may show strong selection
constraints (Ferguson, Peizer, and Weeks 1973; Schwartz 1988). That is, it is
often apparent that only a small range of the many possible adult word patterns
are attempted, with certain phonetically accessible forms characterizing most of
the rst words produced. Such forms include particular phonotactic shapes
or prosodies (CVCV, VCV, or in some cases CVC); forms with a limited
range of onset consonant types (stops, nasals, glottals, and glides); forms with
only a single consonant type; forms including only low or front vowels,
especially in the rst syllable; and forms involving associated CV sequences,
such as labial + /a/ or schwa, alveolar + front vowel, velar + back vowel (Davis
and MacNeilage 1990, 1995, 2000, 2002).
Although direct experimental evidence remains limited (but see Vihman and
Nakai 2003; DePaolis 2006), there is reason to believe that the earliest word
forms are the product of implicit infant matching of own vocal patterns to input
patterning (Vihman 1993, 2002b). This would account for the ndings of relative
accuracy and of phonologically constrained selection. Arst lexicon of some ve
to ten identiable, spontaneously produced adult-based words would be the result
of that match. As a result, the earliest word forms of children acquiring different
languages are broadly similar (with limited phonotactic shapes and consonant
and vowel patterns, as indicated above), being rooted in the physiological con-
straints that govern vocal production in the babbling and rst word period (Locke
1983; Locke and Pearson 1992; Davis and MacNeilage 1990, 1995, 2000; Kent
and Bauer 1985; Kent 1992; see Vihman 1996, Appendix B, which presents the
rst few words of 27 children acquiring seven different languages; as well as
Tables 2.6a, 2.7a, 2.8a, and 2.9a below, which also sample the rst word forms of
children acquiring different ambient languages).
24 Marilyn M. Vihman and William Croft
Within these biologically given limits, however, the ambient language shapes
the rst phonological patterns or templates, which emerge out of the rst words
as the child begins to target new word forms beyond his or her existing range,
sometimes selecting minimally new adult patterns to attempt, sometimes adapt-
ing more distant adult patterns by imposing an existing pattern on them
(Vihman and Velleman 2000). Whereas the rst words are individual by child
but broadly similar cross-linguistically, the templates that are then induced from
them, signaling the rst phonological organization, reect language-particular
differences to a limited extent, as we will illustrate below.
Individual synchronic patterns from children learning a wide range of
languages have provided evidence of word templates, with or without making
reference to whole-word phonology (for examples, see Berman 1977 [Hebrew/
English]; Macken 1978, 1979 [Spanish]; Vihman 1993 [French]; Vihman and
Velleman 1989, Vihman, Velleman, and McCune 1994b [English]; Vihman and
Velleman 2000 [Finnish]; in addition to the children whose data are presented
here). Tables 2.32.5 add to the sample in Table 2.2 with examples from
Vihmans son Raivo, acquiring both English and Estonian, Watersons son P,
and another Estonian-learning child, Madli; note the similarity of the Estonian
data in Tables 2.3 and 2.5 to Watersons data (Tables 2.2 and 2.4).
Table 2.3. Raivos early word templates: nasal structure
(Estonian; age 1;3.181;3.24) nN (N = any nasal) (adapted
from Vihman 1981)
Child form Adult target
[in(+)], [n(+)] (im.); [n] lind bird
[nnn], [nn] rind breast (nursing)
[nni], [n], [n], [nn] king shoe
[ni], [ninin], [ni] kinni closed
+ indicates several repetitions of the syllable in production; im. = imitation
Table 2.4. Ps early word templates: sibilant
structure (age 1;6) (stop)V (adapted from
Waterson 1971)
Child form Adult target
[by] brush
[di] dish
[i] fetch
[i], [] sh
[] vest
Phonological development: toward a radical templatic phonology 25
No segmental substitution account could do justice to these data or capture
the systematicity apparent here. This was the point that Waterson was making in
1971; the little word groups or schemas that she identied when her son P had
roughly 150 words turn out to roughly characterize Madlis and Raivos
Estonian early word patterns as well.
Three types of clues are generally used to identify a childs word template(s):
(a) Consistency of patterning in a substantial number of the child forms for
words produced in one or more recording sessions or over a period of some
weeks or months;
(b) The occurrence of unusual phonological correspondences between adult and
child forms (i.e., rules or processes or repairs to target word violations of
child constraints), under the inuence of a dominating pattern or template;
(c) Frequently, a sharp increase in words attempted that either t or can be tted
into the pattern.
Given these criteria, it is clear that such patterns are most reliably identied on the
basis of longitudinal data from the same child, as Macken (1996) emphasized.
The systematicity in a childs early word production tends to be evident only after
the child has produced some critical number of word forms. The number of forms
will vary from one child to the next, since the emergence of a systematic word
production plan or template depends on the child inducing this structure from the
words s/he is able to say. For example, Menn 1971 observed:
using hindsight, only 3 of [Daniels rst] 30 words fail to satisfy the constraints reected
by the rst set of phonotactic rules, those which govern stage 2 . . . One is led to the
opinion that, while phonotactic rules have not yet crystallized in stage 1, something
vaguely systematic, from which the rules will develop, is at work. (Menn 1971: 231f.)
A developmental progression can thus characteristically be tracked in longitu-
dinal studies of individual infants, from relatively accurate (but highly
constrained) earliest word forms to systematically adapted (and thus sometimes
less accurate but wider ranging) later forms. To illustrate this progression
Table 2.6 presents data from a case study of a child acquiring German in a
monolingual context (Elsen 1996). Here and in what follows we will distinguish
Table 2.5. Madlis early word templates (Estonian;
age 1;8) (p, t)Vs (adapted from Krgvee 2001)
Child form Adult target
[is] isa, issi daddy
[as] kass cat
[pis] piss pee
[us] suss slipper
[tis] tiss teat
[us] uss snake
26 Marilyn M. Vihman and William Croft
Table 2.6. Developmental progression in rst words (Annalena: German).
CV(C1V1); Vi; labialalveolar as phonological patterns, rst fty words
(data from Elsen 1996)
a. Select only (810 mos.)
Child form Adult target Characteristic pattern (based on later template)
[da] da there CV
[ba] Buch book CV
[ai] ei! (fondling expression) Vi
[ai] Ei egg Vi
[nain] nein no Vi
[mama] Mama mama CVCV: CH + VH
[baba] Papa papa CVCV: CH + VH
[pipi] pieppiep mouse CVCV: CH + VH
[dd] Teddy CVCV: CH + VH
[data] das da that one there CVCV: CH + VH
[bita] bitte please lab C . . . alv C
CH = consonant harmony; VH = vowel harmony; MET = metathesis, RED = reduplication;
TRUNC = truncation
b. Select + adapt (1012 months)
Select Adapt
Child form Adult target Template Child form Adult target Template
[ja] ja yes CV [ba] Wasser water CV
[bi] Bild CV
picture
[de] Tee tea CV
[d] Zeh toe CV [bai] Wasser water CV + Vi
[hai] heiss hot Vi [oi] oh! Vi
[ba] Baum tree V [ail] l oil Vi + Vi
[ail] Eule owl Vi, Vl
[pp] tt toot CVC: CH [mom] Baum tree CVC: CH
(blow nose) [note regression]
[mom] bong! CVC: CH
[kiki] kikeriki cock-
a-doodle-do
CVCV: CH + VH [nana] Zahn(brste)
tooth(brush)
CVCV: CH
MET + RED
[pipi] Pipi peepee CVCV: CH + VH [nana] Annalena CVCV: CH
TRUNC + MET
[nan] Banane
banana
CVCV: CH + VH [dada] Tag (good)day CVCV: RED
[bebi] Baby CVCV: CH [vava] wauwau
bowwow
CVCV: CH
Phonological development: toward a radical templatic phonology 27
the rst words, which we term selected (these are the early words in which
something vaguely systematic . . . is at work), and the later words, which may
be either adapted (e.g., the velar harmony words produced by Daniel as his
phonotactic rules began to operate) or selected, in cases in which the adult
word targeted already ts the childs existing phonotactic constraints or word
template.
We have organized the words according to their patterning, primarily their
phonotactic patterns. In the rst months of word production we nd simple
monosyllabic Ca patterns (with initial stop: da, Buch), VV and CVVC
(with the rising diphthong [ai]: ei!, Ei, nein), CVCV (with both consonants
and vowels agreeing across the two syllables: Mama, Papa, pieppiep, Teddy,
das da), and a single C
1
V
1
C
2
V
2
pattern, with a labial alveolar sequence
(bitte). The childs forms are closely related to their adult targets; in Ferguson
and Farwells terms, they are fairly accurate, although we nd some omission
of syllable-nal consonants and two instances of vowel change ([ba] for buch,
[dd] for Teddy).
4
In the following two months, as the pace of word learning quickens consid-
erably (some forty new words are added), we nd (under select) all of the
same patterns represented, with some loosening of the constraints apparent in
the earlier words. The CV patterns include newvowels and an initial glide; the
diphthong [a] occurs as well as [ai]; new syllables occur in harmonizing
disyllabic words. In addition, there are two new phonotactic shapes for
words VCV and CVC. It is notable that the CVC syllables, the only
word forms with differing C
1
vs. C
2
, either show consonant harmony or retain
the previously represented sequence labialalveolar. Under adapt, moreover,
we nd essentially the same word shapes and sequential constraints but with
more radical departures from the adult model.
Table 2.6. (cont.)
Select Adapt
Child form Adult target Template Child form Adult target Template
[babi
d
] Papier paper CVCV: CH [baba] Bauch belly CVCV: RED
[i] trinken CVCV: CH
to drink
[ata], [ada] ada bye VCV [aa] essen to eat VCV
[man] Mann! CVC: [bal] Lampe lamp CVC: lab . . . alv
oh boy! lab . . . alv MET
[man] Mann man CVC: [bl] Brille glasses CVC: lab . . . alv
lab . . . alv
[bal] Ball ball CVC:
lab . . . alv
CH = consonant harmony; VH = vowel harmony; MET = metathesis, RED = reduplication;
TRUNC = truncation
28 Marilyn M. Vihman and William Croft
One way of conceptualizing the childs adapted forms is to see them as
the result of the child (implicitly) imposing one or more preexisting tem-
plates, or familiar phonological patterns, on an adult form that is sufciently
similar to those patterns to serve as a hook. From this perspective, we can
see the effects of the childs practice or motoric familiarity with redupli-
cated patterns (resulting in [nana] for Zahnbrste and [baba] for Bauch, for
example) and with the diphthong [ai], which now appears unexpectedly in
adult words that lack it (e.g., Wasser, oh!, l). Note that the child has
consistently produced only C1C2 sequences involving labials followed
by alveolars (see bitte among her rst words, Mann, Ball, Brille among her
later words), this also being the presumed motoric-plan basis for the meta-
thesis of Lampe to [bal]. Thus, from a usage-based perspective, the childs
adoption of the pattern [bal] (identical to her production of Ball) for Lampe
is not surprising, despite the fact that it involved both (1) omission of the
nal vowel and medial nasal and (2) rearrangement of the syllable-onset
consonants.
In these data, then, we can see evidence of a shift fromthe exclusive production
of words that deviate very little from the adult model to words that may deviate
quite markedly, and in different ways for different words, with the result that
certain patterns are heavily overrepresented in the childs surface forms. In
general, the childs changes affect whole-word forms, not individual segments,
and a number of word templates or well-practiced patterns can be identied,
some of them acting jointly in certain cases (CVC + labialalveolar, for
example).
In Table 2.7 we see the rst words of a child (Virve) acquiring Estonian but
with some exposure to English as well (Vihman 1976).
This child began talking early, although not as precociously as Annalena.
Her early word production suggests tightly constrained phonological selection,
in that words attempted as well as word forms produced were restricted to
(1) a limited segmental inventory (labial and alveolar stops, [s], glides and
glottals), (2) constrained word shapes such that only a single consonant type
could occur anywhere in the word ([tete] for tere), and (3) constrained vowel
sequencing as well (lower vowel rst, higher vowel second). Note that three of
Virves rst six recorded words include the diphthong [ai], the same diphthong
favored by Annalena.
In the following two months of rapid lexical advance Virve loosened con-
straints on possible word forms step by step, as illustrated in Table 2.7b. First
manner ([tin] for kinni),
5
then place (Manni) were allowed to vary, but not both.
Within the vowel sequences, similarly, we see a consistent tendency to produce
either harmonizing forms or V(. . .)i/u patterns, these word forms being
supported by the adult models listed under select but imposed on the models
listed under adapt.
Although the nal /i/ pattern is also commonly found in English (e.g., Molly,
in Vihman and Velleman 1989; Alice, in Vihman et al. 1994a; and the subject
Phonological development: toward a radical templatic phonology 29
Table 2.7. Developmental progression in rst words (Virve: Estonian [and English]) (Vihman 1976).
a . . . i or V
1
. . . V
2
= low non-low
a. Select only (1012 months)
Child form Adult target
Characteristic pattern
(as identied in later template)
[hai] hi CVV: Vi
[pai] pai nice CVV: Vi
[aita], [aida] aith /aith/ thanks VV(CV): Vi
[ao] allo hello (into telephone) VV: Vo
[se] see this CV
[te], [tee], [tete] tere hello CV(CV)
Adult Estonian words have initial stress unless otherwise noted. CH= consonant harmony; MET= metathesis; VH= vowel harmony
b. Select + adapt (1415 months)
Select Adapt
Child form Adult target Template Child form Adult target Template
[titi] kikerikii
cock-a-doodle-do
CVCV: CH, VH [asi] isa father VCV: V
1
. . . V
2
(i) MET
[ap] habe beard VCV [ami] [ani] ema mother VCV: V
1
. . . V
2
(i) MET
[k k ] cookie, cracker CVCV: CH, VH [ati] liha meat VCV: V
1
. . . V
2
(i) MET
[tin] kinni closed C
1
VC
2
[ta | ti] lahti open CVCV: V
1
. . . V
2
(i) CH
[tata], [tai] tdi /tti/ auntie CV(CV): CH, Vi [tati] kallikalli hug CVCV: V
1
. . . V
2
(i) CH
[pebi] beebi baby CVCV: CH, V
1
. . . V
2
(i) [papu] bravo CVCV: V
1
. . . V
2
(high V) CH
[api] appidu uppy-do (jump) VCV: V
1
. . . V
2
(i)
[pai] bye CV: Vi
[ta | si] tantsi dance CVCV: V
1
. . . V
2
(i)
[atsi(h)] tsih achoo VCV: V
1
. . . V
2
(i)
[mani] Manni (name) CVCV: V
1
. . . V
2
(i)
[pawawei] papagoi parrot CVCVCV: Vi
Adult Estonian words have initial stress unless otherwise noted. CH = consonant harmony; MET = metathesis; VH = vowel harmony
of Davis and MacNeilage 1990) and can plausibly be related to the high input
frequency of diminutives such as baby, doggie, kitty, nappy, etc., it is not
necessary to invoke English inuence as a source of Virves patterns.
Table 2.8 presents all the disyllabic words attempted among the rst fty
words of a monolingual Estonian-learning child, Eeriku (Salo 1993).
Like Virve, Eeriku generally avoided the vowel sequence non-low low
(that is, he observed a sequential constraint on vowel height, which we term
SEQ) as well as nonharmonizing frontback vowel sequences (F/B), adapting
words which fail to meet those constraints by the use of truncation (TRUNC)
and metathesis (MET) as well as vowel harmony (VH). As can be seen in
Table 2.8a, the rst few longer words that Eeriku attempted had low vowels
only or were truncated to eliminate the second vowel. Word (12), isa daddy, is
the only word that violates SEQ until the very last few words produced in this
period, which covered a full year in Eerikus case. Eeriku showed a highly
unusual afnity for the difcult Estonian consonant (trilled) /r/. Of his rst 50
words 13 include an /r/; in several cases he appears to truncate specically in
order to produce a syllabic or coda /r/. Otherwise, the adaptations of adult
targets included in Table 2.8b all seem to conspire to achieve a vowel sequence
that violates neither SEQnor F/B(for each word we have indicated the violation
avoided in italics).
Finally, in Table 2.9 we see the same developmental progression that was
illustrated in Tables 2.62.8, this time based on data from a child acquiring
English, though with some exposure to Spanish (Alice: Jaeger 1997), and
starting on her rst word production at 18 months, several months later than
the two children discussed in some detail so far.
Alice again shows only minor changes from the adult model in most of her
rst words (select only). The child forms for food, bottle, and doggie con-
stitute an exception: Jaeger notes that these unusual phonetic forms, which were
produced with a strongly nasal release of the medial obstruent, correspond to
one of this childs frequent prelinguistic babbling patterns.
However, by ve months later, when Alice had acquired a lexicon of some
100 words, she had developed a striking word-form constraint or template,
restricting unlike consonants to a front-before-back sequence. This led to
extensive changes to some adult words (adapted), while other words showed
only minor consonant or vowel substitutions (selected). The constraint was
pregured by 6 (out of a total of 22) earlier words, bottle, mine, doggie, this,
and, at 2021 months, block, stocking). At 23 months the only exceptions to the
constraint were the words dummy, jump, and tum one of only two exceptions
to the constraint among Alices rst words. It seems likely that the exceptional
status of all three words at the later stage stems from entrenchment due to the
frequent use Alice made of this form in a period of great lexical expansion.
While living temporarily with her grandparents, from 1;9.15 on, she called both
of them [tm] for a few days.
32 Marilyn M. Vihman and William Croft
b. Vowel sequences admitted (but low non-low preferred)
Select (target vowels t pattern) Adapt (target vowels violate pattern)
Child form Adult target Relation of target to template Child form Adult target
Adaptation Relation of target
to template
[isa] isa daddy (12) Violates SEQ and F/B [tr:u] toru, torud pipe, pipes
(14, 15)
[produce r]
[a:u] halloo! (24) [mum:] muna egg (16) TRUNC Violates SEQ
[pa:p:a] papagoi parrot (30) VH [ame] ema mother (17) MET
Violates SEQ and F/B
Table 2.8. Developmental progression in rst words (Eeriku: Estonian) (Salo 1993) From vowel harmony (VH) constraint to
sequential constraint V
1
. . . V
2
= low non-low (SEQ) or front/back harmony (F/B)
First fty words: 1;52;5. All (non-onomatopoeic) multisyllabic target words are listed below, along with the childs word form. Numbers
in parentheses refer to the order of rst production of these forms.
a. No vowel sequences allowed
Select (target vowels t pattern) Adapt (target vowels violate pattern)
Child form Adult target Template Child form Adult target Adaptation
[ppa] pkapikk elf (3) CVCV: RED [tit] tita child (4) TRUNC
[paba] paber paper (5) CVCV: RED [en:] onu uncle (6) TRUNC
[ana] vanaema grandmother (9) VCV: VH [:] vike little (8) TRUNC
MET = metathesis, RED = reduplication; TRUNC = truncation
Table 2.8. (cont.)
Select (target vowels t pattern) Adapt (target vowels violate pattern)
Child form Adult target Relation of target to template Child form Adult target
Adaptation Relation of target
to template
[aith] aith thanks (33) Violates F/B [pop:] potsataja fairy tale animal (18) TRUNC
Violates SEQ
[istu] istu sit! (37) Violates F/B [amo] homme tomorrow (19) MET
Violates SEQ and F/B
[arstd] arsti(-)tdi doctor-auntie (38) Violates F/B [aut] auto car (20) TRUNC
Violates SEQ
[priv] prillid glasses (40) TRUNC (despite VH in target) [trar] traktor tractor (21) TRUNC [produce r]
[bi] kbi pinecone (41) [o:ro] koori peel (23) VH
Violates F/B
[sin:a] sinna to there (45) Violates SEQ and F/B [trr] terita- sharpen (pencils) TRUNC [produce r]
[sis:e] sisse to inside (46) Violates SEQ [o:t] oota wait (32) TRUNC Violates SEQ
[pe] pike sun (47) [or:] orav squirrel (36) TRUNC Violates SEQ [produce r]
MET (rst two syllables)
Violates SEQ
[ara] hari brush (42) VH
Violates F/B
[pe] pea head (43) TRUNC
Violates SEQ
[avr] Aivar (44) TRUNC
[produce r]
[todo] Tota-tdi Auntie VH
Tota (49) Violates SEQ
MET = metathesis, RED = reduplication; TRUNC = truncation
2.3. Prosodic/segmental interactions and ambient language inuence
So far we have looked at longitudinal data from three children, each acquiring a
different language, as well as at sample word patterns from a few additional
children acquiring English and Estonian. We have seen that some patterns occur
cross-linguistically and that the early segmental types children produce tend to
be similar regardless of the language to which the child is exposed. Some
patterns do differ by ambient language, however. In this section we illustrate
the effect of the ambient language on early child word patterns by considering
Table 2.9. Developmental progression in rst words (Alice: English) (data
from Jaeger 1997). C1 C1 or fronting constraint: labial alveopalatal,
labial velar, alveopalatal velar
a. Select only (1819 months)
Child form Adult target Child form Adult target
[mama] mommy [hai], [ai] hi
[tata] daddy [aw] out
[nana] Anna [(p)pai] byebye
[peipi] baby [tm] music: tum(te-tum)?
[kta] look at that [main] mine
[kak] food: cracker/cookie? [ti] this
[papm] bottle [mm] no: mm-mm
[tak] doggie [o] uh-oh
b. Select + adapt (23 months)
Child form Adult target Child form Adult target
[ptu] butter [pita] MET David
lab alv alv lab lab alv
[tik
h
] cheek [taik] MET kite
alv vel vel alv alv vel
[pak
h
] frog [pi] MET sheep
lab vel pal lab lab pal
[ppi] puppy [pu] MET soup
lab lab alv lab lab alv
[ti] teeth [piti] MET TV
alv pal alv lab lab alv
Exceptions (based on entrenchment of [tm]?)
[tm] dummy
[tmp] jump
[tmi] tum music
MET = metathesis
Phonological development: toward a radical templatic phonology 35
no onset, or child omission of word-initial consonants. This pattern is dis-
favored by markedness constraints: CVis the most widely occurring syllable
pattern, universally, and is also the rst adultlike syllable infants produce (at
about 68 months [Oller 1980, 2000]). However, as we shall see, the accentual
pattern of the adult language renders some segmental positions more salient
than others, so that although the omission of initial consonants occurs only
rarely in English child words, it is far more common in other languages. We will
summarize some evidence to this effect and will then consider how differences
in adult language accentual patterning might result in this difference in early
child word patterns.
In a study of Finnish children acquiring geminate consonants Vihman and
Velleman (2000) were surprised to nd that the second most common child
phonological pattern (after consonant harmony) was no onset (31 percent,
both selected and adapted) a pattern considered to be a mark of deviant
phonology in English (see also Savinainen-Makkonen 2000). Subsequent
analyses of data from children learning other languages suggest that it is the
absence of any such pattern in data from English-speaking children that is
unusual. Table 2.10 shows the proportion of initial consonant omission in
selected and adapted word forms for each of ve languages.
The column labeled % select shows the mean proportion of the childrens
forms that are based on adult words (or phrases) that fall into the no onset
pattern. Although Finnish has the highest proportion, the languages are roughly
evenly distributed across the range, from 12 to 24 percent. The column labeled
%adapted shows the incidence of child forms in which an initial consonant of
the adult form has been omitted (a pattern seen in some earlier tables as well).
6
Here we see that four of the ve languages cluster closely together, with
incidence of initial target consonant omission ranging from 14 percent to 16
percent. Only English, in accordance with what has generally been taken to be
the universal norm, shows a very low incidence of initial consonant omission
(4 percent); see Figure 2.1.
Table 2.10. Initial consonant omission in ve languages
a
Language (N children) % select Language (N children) % adapt
Finnish (11) 23.9 French 16.4
Estonian (3) 22 Welsh 16
French (5) 15.4 Finnish 14.9
Welsh (5) 13 Estonian 14
English (6) 11.8 English 4.3
Mean 17.04 13.12
a
Data from the case study of Sini, a child acquiring Finnish (Savinainen-Makkonen
2001), and from Andrew, a child acquiring British English (French 1989), have been
added to the data cited in the footnote to Table 2.1.
36 Marilyn M. Vihman and William Croft
Thus, a similar proportion of target words and phrases lack an onset conso-
nant in all ve languages (based on words selected), but the children are less
likely to adapt target words by omitting an onset consonant in English than in
any of the other languages. We must look beyond the basic segmental structure
of the language to account for this.
The languages differ in their accentual patterns, especially their rhythmic
patterns. In English the dominant trochaic pattern is manifested, phonetically, in
a longer and louder rst syllable (which may also be higher in pitch) and a
reduced second syllable (Vihman, DePaolis, and Davis 1998; Vihman, Nakai,
and DePaolis 2006). In none of the other languages do these factors jointly
No onset (selected)
40
%
o
f
a
l
l
w
o
r
d
t
y
p
e
s
35
30
25
20
15
10
5
0
Finnish (11) French (5) Welsh (5) English (6) Estonian (3)
No onset (adapted)
40
%
o
f
a
l
l
w
o
r
d
t
y
p
e
s
35
30
25
20
15
10
5
5
0
Finnish (11) French (5) Welsh (5) English (6) Estonian (3)
Figure 2.1. No onset (selected vs. adapted) in ve languages
Phonological development: toward a radical templatic phonology 37
affect the rst syllable, despite the fact that in our sample all but one of the
languages is primarily or exclusively trochaic. In French the dominant pattern is
iambic, with lengthening of the nal syllable as the primary accentual marker. In
Welsh, although the rst syllable of a disyllable is normally stressed, this is
manifested by a short rst-syllable vowel followed by a lengthened medial
consonant and a long second vowel (see Vihman et al. 2006 for documentation
of both adult and child production). Finnish, although strictly and exclusively
trochaic, has another highly salient rhythmic characteristic frequently occur-
ring medial geminates which can deect infant attention away from the initial
consonant. Indeed, the presence of medial geminates appears to be a powerful
attractor for infant attention, since children target a disproportionate number
(49 percent, compared to an incidence in mothers content words of 37 percent)
(Vihman and Velleman 2000). In the childrens own productions, 55 percent
have long medial consonants, again suggesting attention to and overextension
of this rhythmic property.
Here then we see group results analyzed in the same way as the longitudinal
data presented in Tables 2.62.9 above. A similar proportion of VCV patterns
occurs in the input in all ve languages (mean of 17 percent), based on child
selection of words to attempt that lack an initial consonant (e.g., English uh-oh,
Table 2.9). In the case of all of the languages except English the children extend
the pattern to assimilate word targets falling outside it in the adult language.
In some cases the omitted consonant itself poses a problem for the child
(see Table 2.4, in which P, learning English, systematically omits initial frica-
tives). In most cases, however, omission of the initial consonant appears to be a
way to arrive at a pronounceable form despite the difculty posed by a word-
internal noncontiguous consonant sequence. This is a striking demonstration of
the effect of the whole-word (disyllabic) pattern on learning, since it is the
lengthening of a medial consonant or nal vowel, or both, which appears to
draw the childs attention away from the initial segment, typically considered
most critical to word learning in English.
As further evidence for the hypothesized role of geminates in supporting a
no onset template, Table 2.11 summarizes the phonological patterning in the
complete lexicon of a child V, aged 1;7, who is bilingual in Hindi and English
(with a few words from other Indian languages).
One example of each occurring pattern is provided; numbers in each cell
indicate the total child word form types conforming to the pattern (T = 198
words).
This child primarily produces monosyllables in English (83 percent far
exceeding the mean seen in other children acquiring English as well; see
Table 2.1) but disyllables in the Indic language words he knows (78 percent).
Indeed, the author/diarist sees the childs differential attention to English
monosyllables vs. Hindi disyllables as Vs way of keeping the languages
apart in a setting in which several languages are current and code mixing
is the rule. Vs English words also tend to show consonant harmony (15/41,
38 Marilyn M. Vihman and William Croft
Table 2.11. Consonant harmony and no onset in a bilingual child, V (1;7) (based on Bhaya Nair 1991)
English Hindi (+ a few Bengali and Malayalam words)
Phonological pattern Select Adapt Select Adapt Total word types
CV(V) 7 no 1 ball [b:] 4 /ta/ tea 1 /phu:l/ ower [pu:] 13
V(V)(C) 1 eye 0 4 /a:g/ re 0 5
C
1
VC
1
(or place agreement only) 3 cake 12 dog [kg] 0 1 /na:k/ nose [ka:k] 16
C
1
VC
2
10 bus 0 2 /ka:n/ ears 1 /gram/ hot [gm] 13
C
1
VC
1
V 2 dirty 0 6 /ba:ba/ grandpa 0 8
C
1
VC
2
V 1 bowwow 0 3 /k
h
ta/ thorn 0 4
VCV 0 3 cover 5 /a:pa/ aunt 7 /pa:ni/ water [a:ni] 15
VCCV 6 /nda/ egg 13 /khi/ comb [hi] 19
C
1
VC
1
C
1
V 1/i/ excrement 0 1
C
1
VC
1
C
1
VC
1
2 /ti:tti:t/ sweet 0 2
C
1
VC
2
C
1
VC
2
1 ticktick 2 /ptpt/ beating 0 3
Total 25 16 35 23 99
or 37 percent) while his Hindi words tend to show no onset instead (35/58, or
60 percent). Interestingly, three of his English words also showinitial consonant
omission: [b] cover, [ki] monkey, [t] water a probable sign of inter-
action with the Hindi pattern, since such a pattern seems highly unusual for
English words whose initial consonants are a stop, a nasal, and a glide.
Of the initial consonants omitted in non-English words, 6/20 are affricates or
/ / or /r/, segments the child does not yet produce or produces only rarely.
(Four English, three Hindi and one Bengali word are produced with initial
affricates; none have initial / / or /r/.) Yet segmental difculties are not the sole
or primary basis for no onset since in three cases the omitted consonant is a
stop or nasal that agrees in full or in place only with the medial consonant. Of
the child words that differ from their targets by virtue of initial consonant
omission, 13 out of 20 (65 percent) have a medial consonant cluster; 8 of
these (40 percent) are geminates. Thus, the medial long consonants are as
plausible a rhythmic source of the no onset: adapt pattern here as in Finnish.
2.4. Universals of early phonological development or inductive
generalizations from the lexicon?
We have considered the emergence of word templates in the course of rst word
production as recorded in several diary studies. The templates cannot be innate,
since they are not always present from the rst words, nor can they be universal,
since they differ from one child to the next and also differ to some extent by
ambient language.
Rather, we take them to be the emergent product of three sources of phono-
logical knowledge for the child: (1) familiarity with the segmental patterns
typical of the adult language, which advances steadily over the last few months
of the rst year (see Jusczyk 1992, 1997); (2) developing motoric control and
familiarity with a subset of adultlike phonological patterns due to production
practice (babbling); and (3) increasing familiarity with the structure implicit in
the childrens own rst lexicon. The childs early word forms can be taken to
reect sensitivity to matches between his or her emergent production patterns
and frequently used adult words. The wide interchild variability in early pho-
nological patterning that we see even within the limits of a single ambient
language does not derive from the adult input, however, but from the individual
lter that each child brings to the word learning process. This is evident from
the fact that while the phonological patterns found by sampling input from ve
mothers are strikingly similar, those of their ve children are widely different
(see Vihman et al. 1994a, which replicates the nding in three languages,
English, French, and Swedish).
We take the fact that cross-linguistic differences shape word templates to be a
natural consequence of the induction process, since the target lexicon necessa-
rily shapes the patterns implicit in the childs rst 50 words or so. We note that
English, Estonian, and German data often show a concentration of CVC shapes
40 Marilyn M. Vihman and William Croft
(see also Vihman and Velleman 1989). In contrast, French data do not normally
show CVC forms as early as the rst 50100 words (Vihman 1993, 1996),
although the EnglishFrench bilingual early words reported by Brulard and
Carr (2003) do include such forms, and they dominated the English lexicon of
the child V, as indicated in Table 2.11. These diary studies provide some insight
into the construction of templates under conditions of bilingual input (Vihman
2002a).
In short, we see the earliest phonological organization as constituting an
inductive generalization based on the childs rst repertoire of phonetic patterns
and their interaction with the phonological structure implicit in the words of the
ambient language that the child is attempting to reproduce. The phonological
organization itself inheres in whole-word patterns or word templates, as can be
seen from the adapted patterns illustrated above. Phonological categories will
gradually emerge later, in different ways for different children. The develop-
mental pattern is like that found in recent studies of early syntax, in which
verb islands are found in lieu of abstract grammar, with productive use of
subcategories emerging only slowly, in different ways for different children
(e.g., Tomasello 1992; Lieven, Theakston, Pine, and Rowland 2000).
3. From child to adult: toward a radical templatic phonology
In Section 2, we argued for a templatic approach to phonological development
in the child. In this section, we argue that a templatic approach is equally suited
to the analysis of adult phonology. This argument derives much from phoneti-
cally oriented, exemplar and usage-based approaches to phonology and from a
related approach to syntax, Radical Construction Grammar (Croft 2001).
3.1. Variation and phonological categories
One of the initial arguments for a templatic approach to child phonological
development is the variability of segment production. Such variability is
pervasive in adult phonological categories as well. Ohala writes, One of
the major discoveries of phonetics for the past century is the tremendous
variability that exists in what we regard as the same event in speech, whether
this sameness be phones, syllables, or words (Ohala 1993: 239). Ladefoged
and Maddiesons (1996) survey of segments across languages documents this
variability on virtually every page. Pierrehumbert, in a paper advocating an
approach to phonology that is quite similar to ours, also begins by demon-
strating the high degree of variation found not just in segments but also in
prosodic structures (Pierrehumbert 2003a: 1207; see also Pierrehumbert,
Beckmann, and Roberts 2000).
This variability occurs at all levels, fromindividual usage events to languages
(that is, cross-linguistic variation). For example, vowel productions are stand-
ardly mapped onto a two-dimensional F1F2 space, and scatter plots illustrate
Phonological development: toward a radical templatic phonology 41
variation in production in usage events within and across individuals (e.g.,
Pierrehumbert 2003b), leading to sociolinguistic variation (e.g., Labov 1994).
Ladefoged and Maddieson (1996) document this variation as it eventually
manifests itself as divergence across dialects and across languages. For exam-
ple, at the dialect level, Californian English speakers use true interdentals in a
word such as [ ik] whereas British English speakers use a dental fricative
[ik] (p. 20). Cross-linguistically, many languages distinguish dental and
alveolar stops, particularly in India, Australia, and the Americas. Most such
languages contrast a laminal dental [t] vs. an apical alveolar [] as in Toda [pot]
ten vs. [pt] cock-roach, but Temne contrasts an apical dental vs. a laminal
alveolar (Ladefoged and Maddieson 1996). Most such languages also have
greater affrication of apical alveolars than laminal dentals, as in Isako, but
Dahalo has greater affrication of the laminal dentals (p. 25).
Variation is so pervasive that an adequate theory of phonology cannot ignore
it or properly abstract away from it (see Section 1). Pierrehumbert (2003a)
argues for an approach to phonological categories based on mathematical
psychology that accommodates variation:
A category is a mental construct which relates two levels of representation, a discrete
level and a parametric level. Specically, a category denes a density distribution over
the parametric level, and a category system denes a set of such distributions.
Using the density distributions for categories in a category system, incoming signals
may be recognized, identied, and discriminated through statistical choice rules. This
understanding of categories has been generally adopted in experimental phonetics and
sociolinguistics. (Pierrehumbert 2003a: 119)
We believe that this approach to categories can and should be adopted in
phonology as well.
One result of this approach to categorization is that the segment categories
that can be formed from the actual input are not phonemes but positional
variants of phonemes (Pierrehumbert 2003a: 12930). For example, tokens
of initial and nal /s/ in English differ from each other signicantly. Within
each position, /s/ and /z/ are reasonably well differentiated, but across posi-
tions, there is substantial overlap between /s/ and /z/ tokens. Pierrehumbert
(2003a: 140) concludes that the engine of adult speech perception appears to
be positional segmental variants. Pierrehumberts conclusion is exactly that
of our templatic approach: segmental phonological categories are dened
in terms of their position in a larger structure (the word template; see
Section 3.2). The evidence that Pierrehumbert amasses supports this view
for adult phonology as well.
Pierrehumbert restricts her attention to the identication of individual
segments, that is, positionally dened allophones. She notes that phonemes,
as categories of allophones in different positions, play little if any role in adult
speech perception (Pierrehumbert 2003a: 129). But contemporary generative
phonological theory does not refer much to phonemes either; for example,
42 Marilyn M. Vihman and William Croft
phonemes are hardly mentioned in a recent survey of theories of phonological
representation (Ewen and van der Hulst 2002). Instead, a more abstract
or general category is used for phonological representation, namely features.
Afeature is a more general category that subsumes multiple segments namely,
all the segments that possess that feature.
Yet features as a more general category are problematic. For example, Ewen
and van der Hulst (2002) argue that the same vowels are categorized in different
ways depending on the relevant phonological process/phonotactic pattern
(pp. 1521, 1025). The vowels in (1), for example, are grouped according to
the category/feature of tenseness:
(1) [+tense] i e a u o
[tense] i
Ewen and van der Hulst argue that this categorization of vowels is needed to
describe a constraint on nal stressed vowels in English (e.g., [+tense] /bi:/ vs.
[tense] */bi/).
A different categorization of the same vowels, given in (2), is necessary
for representing the constraint on possible vowels in a single word (vowel
harmony) in some languages. Vowel harmony in languages such as the
Asante dialect of Akan is governed by the feature of advanced/retracted tongue
root (ATR; Ewen and van der Hulst 2002: 1920):
(2) [+ATR] i e u o
[ATR] i a
Finally, Ewen and van der Hulst (2002) argue that the categorization of
vowels in terms of the traditional feature of height is also necessary in order
to describe, for example, the stepwise shifts in vowel height of the English
Vowel Shift and also a diphthongization process in Skane Swedish (pp. 201;
we have used a multivalued height feature here but most feature theories use
various devices to avoid multivalued features):
(3) [high] i u y
[high-mid] e o
[low-mid]
[low]
Ewen and van der Hulst (2002) introduce three different features for grouping
the same sounds in the three different ways in (1)(3) (they use the single-
valued features ATR [advanced tongue root] and @ [for laxness] and some
combination of features for height: pp. 1025). That is, they have proposed a
distinct vowel feature for each of the three phonological phenomena they
describe. They write:
The range of processes surveyed in this section suggest that vowel systems can be
organized along different phonetic and phonological parameters, and hence that our
Phonological development: toward a radical templatic phonology 43
feature system must be rich enough to be able to describe all of the parameters found to
play a role in the organization of vowel systems. (Ewen and van der Hulst 2002: 21)
We agree with this statement but we raise the question, where does it stop?
For example, Ewen and van der Hulst (2002) observe in a footnote that with
respect to another English phonotactic phenomenon, occurrence before //,
the category of [tense] vowels must exclude //, and in other respects // acts
as a separate class (p. 18, fn 16). In other words, occurrence before // denes
a different natural class from that in (3), namely {i }. In principle a new
feature should be posited for that class. Otherwise one is in effect choosing
the distribution pattern dened by nal stressed vowels over that dened by
occurrence before // but there is no a priori reason to do so.
The logical conclusion to this process would be the positing of a different
feature for each category dened by each phonotactic constraint. This is in fact
what we are basically arguing for: even the more abstract categories familiar to
us from phonological theory are dened in terms of their position in phonotactic
templates. That is, phonological categories are dened in terms of their distri-
bution in templatic patterns. In other words, the phonotactic templates are basic,
and phonological categories are derivative (we return to this point in
Section 3.2).
A templatic approach to adult phonology is supported by the widespread
and well-known fact that the most general and abstract categories of sounds
(those usually described by features) actually differ in different word or
syllable positions. For example, Bybee (2001) suggests that consonants in
initial and nal position are quite different in phonetic realization (compare
Pierrehumbert 2003a above), that consonant as a category may not be
valid: onsets and codas may not be unied into a single set of consonants
(Bybee 2001: 88). She adds, This proposal would predict that a language
could have a completely mutually exclusive set of syllable onsets and syllable
codas (p. 88).
Although we are not familiar with such a language, some languages have
quite distinct sets of initial and nal consonants with only partial overlap.
Sedang exhibits this pattern for stressed syllables and in addition has a third
series of consonants for initial consonants in an unstressed syllable preceding
the stressed syllable, called a presyllable (Smith 1979: 22, 26, 37), as shown
in Table 2.12.
In addition, there are consonant clusters with stops followed by /l/ or /r/.
The total count of initial vs. nal consonants in Sedang is as given below
(clusters and the presyllabic consonants are excluded from this comparison):
7
(4) Initial: 41 consonants, 30 unique to initial position
Final: 14 consonants, 3 unique to nal position
Overlap: 11 consonants
44 Marilyn M. Vihman and William Croft
Smith writes: The dissimilarity of the nal consonant inventory from the initial
single consonant inventory . . . recommends the establishment of a separate
consonantal system for each consonantal position of the phonological word
(Smith 1979: 37). Moreover, the relationship between the syllable nucleus and
the nal consonant is also complex: nal zero and glides allow for register and
oralnasal distinctions in the nucleus, nal nasals allow for register distinctions
only, and other nals allow only oralnasal distinctions (Smith 1979: 424).
This example demonstrates not only that one must distinguish between syllable-
initial and syllable-nal consonants as distinct phonological categories, but
presyllable consonants are a distinct category as well. All three categories of
consonants are dened by their position in the Sedang word template, as
Smith recommends.
The closest example to mutually exclusive positional categories of a highly
general feature that we are aware of is found with the vowels of the
nineteenth-century Tremjugan dialect of Khanty (Abondolo 1998: 362). The
set of word-initial (stressed) vowels of Khanty (called V
1
below) is not the same
as the set of noninitial vowels (V
2
; // and // are back unrounded vowels, // is a
front low unrounded vowel and // a back low rounded vowel; // and / / are
front and back central vowels, respectively):
(5) Initial vowels: ii ee uu oo
e o a
Noninitial vowels: ii ee aa
Table 2.12. Sedang consonant inventories by position
Initial stops p t c k
m
b
n
d
j
g
m n
p
h
t
h
c
h
k
h
b d
m n
m n
Final stops p t k
m n
Presyllabic stops p t k
b
m
Initial continuants s
l r j h
l r
1 r j
nd u]
1
another [aa] [n]
Randall [a
] [r nd ]
Randall is the name of a friend and neighbor who helped to looked after P
and who was always addressed and referred to as Mrs. Randall until P started
to call her [a
]
(see p. 73); and at 1;8 honey [ah u:]. Also the following at 1;6: biscuit
[be:be:]; bucket [bbu:]; pudding [pp]; Bobby [bbu:]; Kitty [tt]; and dirty
[d :t]. A detailed phonetic account is given of these forms but only a brief
64 Natalie Waterson
account of the phonology. A detailed phonological analysis of these forms is the
subject of a separate paper in which regular correspondences are shown between
child and adult structures and these correspondences are used to predict some
child forms and structures from the adults. (See Waterson 1970.)
It was found that correlations at the phonetic level could be stated between the
childs and adults forms by reference to the following:
1. Various features of articulation such as nasality, sibilance, glottality, stop
(complete closure), continuance, frontness, backness, voicing, voicelessness,
labiality, rounding, nonrounding. (A distinction is made between labiality
and rounding. Labiality is used to refer to the action of the lips as being
concerned in the articulation of a consonant such as lip protrusion which is part
of the articulation of initial [r], or the lip contact of the labiodental stricture of
[f]. Rounding, which includes labiality, refers to lip action which extends over
the syllable, e.g., in the second syllable of barrow, [bru].)
2. Grade of vowel opening.
3. The syllabic structure of the words.
4. The prominence of syllables (cf. Jones 1962: 55, who used the term prom-
inence in relation to sounds: the prominence of sounds may be due to
inherent sonority . . ., to length or to stress or to special intonation, or to
combinations of these).
In order to make a comparison between the childs and adults forms, the childs
forms are rst examined and analyzed into features and are then grouped into
different types of structure according to the selection of features which goes into
their composition, viz.
I. Labial Structure
II. Continuant Structure
III. Sibilant Structure
IV. Stop Structure, and
V. Nasal Structure.
It will be found that some features are common to all ve types of structure, others
to only three or two; but from the detailed descriptions given under each of the
ve headings, it will be seen that no one type of structure has the same selection of
basic features as any other type. For what is meant by basic features see below.
Type I. Labial Structure
The childs forms of y, barrow, and ower at 1;5 and 1;6 belong to this type.
y barrow ower
1;5 [w/b] [ww]
2
1;6 [/v/b] [bw] v/vw
They have the following features in common: labiality at the onset of each
syllable, [w, b, b, v, ]; continuance, [w, , b, v]; voiced onset of every syllable,
Child phonology: a prosodic view 65
voiced ending of every syllable, broad degree of openness of vowel (as opposed
to closeness); prominence of one syllable, and the syllabic structure CV. These
features account for the similarity of these forms and such features will be called
the basic features. The structures may be symbolized as KVand KVKVat 1;5
and KV, KVKVand PVKVat 1;6 (K = continuant system, P = stop system).
Features which are not shared by all the forms but may be shared by some
and which account for the differences between them are as follows: friction,
[b, , v]; nonfriction [w]; affrication [b]; bilabiality [w, , b, b]; labioden-
tality [v]; stop [b]; centrality of the syllable [b]; rounding of the syllable [w];
backness of the syllable [w]; the ner distinction of frontness of syllable as
opposed to centrality, i.e., [w, b, , v] as opposed to [b] ([] in the
speech of the child and his mother is fully front, [] is advanced fromcentral but
is not fully front); word structure CVCVand CV. Such features as account for
differences of form will be called the differential features.
It is seen that the form for barrow at 1;6 has developed to a form closer to the
adult form and has the greatest number of differential features of all the child
forms belonging to this type of structure. The childs form [ww] at 1;5 had
more features in common with the other forms belonging to the Labial Structure
than his form [bw] at 1;6.
Type II. Continuant Structure
The childs forms for Rooney, honey, and hymn/angel belong to this type of
structure.
Rooney honey hymn/angel
1;6 [/h] 1;5 [ah /h/a
]
1;8 [ah u:]
The basic features of these forms are as follows: glottality, [h, ]; continuance
[h, ] ([h] and [] are analyzed as glottal continuants not as fricatives because
the stricture is at the vocal cords and there is no stricture in the supraglottal area
in common with other sounds classed as continuants sounds classed as
fricatives have supraglottal stricture); prominence of the rst syllable; voiced
onset of syllable i ([h] was a rare form and is grouped together with the other
forms under this type of structure because of its obvious similarity it is the
only one with voiceless onset); voiced ending of syllables 1 and 2; the disyllabic
structure of the word. The structure may be symbolized as VHV (H = glottal
continuant prosody, see p. 72).
The differential features of these forms are: voiceless onset in syllable 1
[h] and in syllable 2 [ah u: ah , h]; voiced onset of syllable 2 [, h,
a
]; nasality in [,
h] and in syllable 2 of [a
]
They have the following basic features: nasality []; stop []; voiced onset of
the syllable; voiced ending of the syllable; prominence of the rst syllable;
syllabic structure of word CVCV.
68 Natalie Waterson
Differential features are: frontness of syllable [e:, e, i:,
, ]; centrality
of syllable [a]; length of syllable [e:, i:]; rounding of syllable [
];
nonrounding of syllable [e:, e, i:, , a]; same grade of vowel in both
syllables [e:e:, :, aa]; more open grade vowel in syllable 1 and more
close in syllable 2 [a
] of
[e n
], and glottality in [h] of [hn] and [hm]); nasality in the stops [n] and
[m] and a certain amount in the vowels. Where the word has voiced onset there
is fairly strong nasality over the word, e.g., [r :n ] and [e n
] for angel, pointing at the angels on the cover one at a time and
naming them. On the same day he used [b a
], i.e., where the adult has voiced onset and heavy nasality, the
child has nasality in the word but where the adult has voiceless onset and weak
or no nasality over the word, the child has no nasality, e.g., adult [hn], childs
[ah u:]. It has already been shown that the words hymn and angel are not clearly
differentiated semantically for the child and it seems that they are therefore
not phonetically differentiated. It is possible to link the two forms without
nasality, [ah ] and [h], more closely with the adult form [him], which has
voiceless onset and little nasality over the word, and the form with nasality,
[a ], more closely with the adult form [e n
g] nger, [w
g] and [w
nd u] and [r nd ].
It can be seen that the basic features nasality and stop are common to all the
childs and adults forms. Prominence of the penultimate syllable is also basic to
child and adult as are the following: nonrounded syllable, voiced ending of all
syllables, voiced onset of syllable 2. The adults basic feature continuance is not
reproduced by the child.
The nasals of the adult forms, apart from [n], are homorganic with the
following oral stops and are thus complex articulations and strongly articulated.
In [n] the nasal stop is at the onset of a stressed syllable and is also strongly
articulated. These strongly articulated nasal stops are reproduced and redu-
plicated by the child; cf. the weakly articulated nasal stops of Continuant
Structure words which are not reproduced by the child. In Nasal Structure
words the nasal stops are more forcefully articulated than the continuants and
it may be that they are therefore more clearly perceived by the child and hence
are reproduced by him.
As the differential features are many, a more detailed comparison is needed
to show the close relationship of the childs and adults forms. Prominence in
the rst syllable of the adult forms for nger and window, which have strong
nasality in addition to the other qualities which go to make a syllable prominent
(see p. 65, reference to Jones 1962), is matched in the childs forms by length of
syllable, i.e., the rst syllables of [e :e /i:] nger and [e:e:] window. The
second syllable of window in the childs and adults forms, although less
prominent than the rst syllable, has more prominence than the nal unstressed
syllable of their forms for nger, i.e., [e] and [] in the childs forms for nger
Child phonology: a prosodic view 77
are less prominent than [e:] in syllable 2 of his form for window, and [g] in the
adults form for nger is less prominent than [d u] of window.
Acorrelation of vowel grade can be shown. Four grades of openness of vowel
are needed to describe vowels of the childs and adult forms being discussed
here: close [i:], close-mid [1], open-mid [, e, o], and open [, a, ], [u] is a
labial glide.
The childs forms belonging to this structure may be described as redupli-
cated structures. Some are fully reduplicated, i.e., [aa] and [e:e:], and
others are partially reduplicated, i.e., [e :e ], [i: ], and [a
g]. Where the initial consonant of the adult form functions as an initial in
the childs system belonging to a different structure, the childs form is also a
reverse reduplication (either full or partial), e.g., in the childs system initial [w]
and [r ] are used only in Labial Structure words (one of his forms for rabbit,
adult [r bt], was [r w]), therefore the consonantal elements of the childs
forms for window and Randall, which are Nasal Structure words, are a reverse
reduplication of the ending of the rst syllable of the adult form, i.e., [e:e] and
[a
], with the vowel grades and syllable features of the adult forms partially
maintained.
Quite a large number of substitutions in the examples of reduplications in
childrens speech given in the Appendices of Lewiss Infant Speech (1968) can
be explained in terms of reverse reduplication as described above. This sort of
reduplication has been observed before, but was interpreted in terms of assim-
ilation, e.g., by Leopold (1947).
By way of illustration of howthe approach used in this chapter can be applied
to another childs speech in order to explain phenomena that cannot be
accounted for by substitution, assimilation, etc., a few examples are taken
from Leopolds material (1947: 25774) which his theory could not explain,
to show how easily the irregularities can be explained by the type of analysis
suggested here. For instance, Leopold found it difcult to account for his
daughters [de] for steht (stands), [d] for stone and [l] for story, all
these being used by his child at 1;11. His daughter Hildegard was bilingual
English and German. He suggests [de] is steht with metathesis of initial [] or
steh- plus English -z. He compared [d] with [de] and at rst considered that it
had an incorrectly placed plural [z] (the word was used to refer to one stone), but
then decided it might be due to metathesis of initial [s]. However, fromhis excellent
phonetic records it is plain that at 1;10 and 1;11 his child had a type of structure
84 Natalie Waterson
with sibilant nal which had three different kinds of initial: (1) stop, (2) continuant,
and (3) nasal; cf. Ps Sibilant Structure at 1;6. The following is a representative
selection of the forms the child had, taken from Leopold (1939: 53137):
1. Stop initial and sibilant nal [bi] piece; [be] bathe; [ba]
beiss(en) (bite); [da] crash, dress, Katz (cat), kratzen (scratch), Glas
(glass); [d] kiss; [du] juice, Kuss (kiss). Cf. the adult forms which in
both English and German have the basic features onset with oral stop, and
fricative (generally sibilant) in nal position. Child and adult forms share
the basic features of voiceless and fricative ending; voiced onset is basic
for the child but not for the adult; sibilant ending is basic for the child but not
for the adult although most of the adult forms do have sibilance.
2. Continuant initial and sibilant nal [ha] heiss (hot); [hau]
Hause, house; [wa], [wa] waschen, wash; [w] abwischen (to wipe up); [ju]
lutsch(t) (sucks). Cf. the adult forms which have the basic features continuant
onset and fricative ending. Sibilant ending is common to most adult forms but is
nonbasic; it is basic for the child. The onset is voiced for child and adult except
where there is glottal continuance, e.g., [ha], heiss; voiceless ending is basic for
both in the rst or only syllable.
3. Nasal initial and sibilant nal [mau] mouse; [na] nice, knife;
[na] nass (wet); [ma] much; [mau] mouth; [n] nose. Cf. the adult forms which
all have nasal stop initial (basic) and fricative (basic), generally sibilant ending.
Voiced onset is basic for both the childs and adult forms. Voiceless ending is basic
for the child but not for the adult although it is common to most of the adult forms.
One may note that Hildegard maintains the distinction of labial and nonlabial
onset in all the examples quoted under (1), (2), and (3). Cf. Ps Stop Structure
forms where he kept this distinction.
Hildegards system appears to be based on adult forms having sibilant
fricative nal and initials with oral stop, continuant, and nasal stop. The
following irregular forms given by Leopold are now examined in relation to
the three types of structure set up for the childs phonological system: [de]
steht; [d] stone, and [l] story.
The adult form steht has the features (checked) sibilance and friction [t],
mid vowel [e], stop [t] and there is frontness over the whole word. The childs
form [de] has the features sibilance and friction [], mid vowel [e], stop [d], and
frontness over the whole word, i.e., it has features almost identical with those
of the adult form. In structures of the childs system with such a selection of
features, the stop feature comes rst and the sibilant feature last; cf. (1) above,
and the onset is always voiced whether the onset in the adult form is voiced or
voiceless, i.e., voiced onset is basic for the child. The sequence of sounds as in
the childs form [de] is therefore the only possible one to t her system and is
Child phonology: a prosodic view 85
thus perfectly regular. If one examines the features composing the childs and
adult forms for stone, they are also found to be similar to each other. Adult
[stun] has the features (checked) sibilance and friction, [st], stop (nasal), [n],
more open vowel followed by more close, [u], and rounding of the whole
word. There is voiceless onset and voiced ending. The childs form [d] has
the features sibilance and friction [], stop (oral) [d] (nasality is not a basic
feature of the childs stop initial and sibilant nal structures, so is not relevant
here; it is only relevant when onset with nasality plus stop is basic in the adult
form), more open vowel followed by more close, [], and rounding and voicing
at the onset of the word and nonrounding and voicelessness in the ending. As
noted above, the childs systemrequires the stop to be initial and the sibilance to
be nal, with voiced onset and voiceless ending. The vowel grades are the same
as in the adult form, viz. more open followed by more close, but the rounding
feature does not extend over the whole of the childs form. The form [d] is
thus just as regular as the form [de] and the other forms listed under (1) above.
The form [l] for story puzzled Leopold so much that he doubted the
interpretation of the word as story, although the context in which it was used
seemed to suggest it: it was given as the answer to the question was hat Mama dir
erzhlt? (what did mummy tell you?), the answer being [ l], which Leopold
interpreted as a story but said that if it was story, it was quite irregular as none
of his patterns of assimilation, etc., could explain it. However it is possible to
showthat it shares many features with [st:r] and as far as the childs own system
is concerned, is quite regular. The adult form [st:r] has the features (checked)
sibilance and friction, [st], liquid and continuance, [r], more open vowel followed
by more close, [] and [], and rounding and backness in the rst syllable and
nonrounding and frontness in the second, with voiceless onset and voiced ending
of the word, i.e., it has the same features as the adult forms grouped under
(2) continuant initial and sibilant nal, but in a different sequence. The childs
form [l] has sibilance and friction, [], liquid and continuance, [l], more open
vowel followed by more close, [] and [], rounding and backness in the rst
syllable and nonrounding and frontness in the second; all these features are
shared with the adult form, but the childs form has voiced continuant onset
and voiceless sibilant ending, which is a different sequence from the adults but
is required by the childs system as seen in (2) above. The childs form
thus shares most of the features of the adults while conforming to her own
system referred to above, viz. continuant initial and sibilant nal structure. This
form is thus considered to be completely regular. In none of these cases is it
necessary to bring in the concepts of substitution or metathesis to explain
the differences between the childs and adult forms. The childs forms conform
to the patterns of her own phonological system and, as noted in the case of P, the
sequence and combinations of features in the childs and adult forms are not
always the same.
It is possible that the perception of phonetic features that have been
described in articulatory terms is, in fact, some kind of perception of acoustic
86 Natalie Waterson
cues similar to what Fry suggests for the development of the phonemic system
(1966: 197):
It is clear that a very important part of this development of the phonemic system is bound
up with the use of acoustic cues, both for monitoring of the childs own speech and for
the reception of other peoples. We now have a considerable body of information about
the operation of these cues in adult speech, although we are still far from understanding
fully how they function, but have no knowledge of the ways in which the use of the cues
develops as speech is acquired.
It is possible, also, that there may be some parallel in the perception of phonetic
features (whatever their nature) with what Piaget calls verbal syncretism
(1967: 1312), he writes:
Recent research on the nature of perception particularly in connexion with tachistoscopic
reading, and with the perception of forms, has led to the view that objects are recognised
and perceived by us, not because we have analysed them and seen them in detail,
but because of general forms which are as much constructed by ourselves as given
by the elements of the perceived object, and which may be called the schema or the
gestaltqualitt of these objects. For example, a word passes through the tachistoscope
far too rapidly for the letters to be distinguished separately. But one or two of these letters
and the general dimensions of the word are perceived, and that is sufcient to ensure a
correct reading. Each word, therefore, has its own schema.
Piaget considers that such schemata are far more important for the child than
for the adult, as they develop long before the perception of detail, the natural
course of development being from syncretism to a combination of analysis
and synthesis, and not from analysis to syncretism. It thus seems reasonable to
consider that a child perceives some sort of schema in words or utterances
through the recognition of a particular selection of phonetic features (the basic
features) which go into the composition of the forms of the words or groups of
words, and this recognition of a schema results in his producing words of the
same type of structure for such adult forms, e.g., words with consonantal
features continuance and strongly articulated nasal followed by stop have in
the forms of the child P a reduplicated nasal stop pattern, i.e., the Nasal
Structure. A child also recognizes differences in form within the particular
type of structure (the differential features) and this results in his having
different forms within the Structure, and as his skill in perception and articu-
lation increases, so he perceives and reproduces more and more of the features
of the adult forms. Such a hypothesis seems to link up with what appears to be
currently a widely accepted view of the cognitive development of the child,
i.e., starting with a comparative lack of differentiation and progressing by way
of increasing differentiation. This view may be briey illustrated by the words
of Brown (1958, reprint 1968: 89):
the primitive stage in cognition is one of a comparative lack of differentiation. Probably
certain distinctions are inescapable; the difference between a loud noise and near silence,
between a bright contour and a dark ground, etc. These inevitable discriminations divide
Child phonology: a prosodic view 87
the perceived world into a small number of very large (abstract) categories. Cognitive
development is increasing differentiation. The more distinctions we make, the more
categories we have and the smaller (more concrete) these are. I think the latter view is
favored in psychology today,
and (1958, reprint 1968: 91):
Psychologists who believe that mental development is from the abstract to the concrete,
from a lack of differentiation to increased differentiation, have been embarrassed by
the fact that vocabulary often builds in the opposite direction. This fact need not trouble
them, since the sequence in which words are acquired is not determined by the cognitive
preferences of children so much as by the naming practices of adults.
In the analysis presented in this chapter, the adult forms, like the childs,
were grouped into ve types of structure on the basis of the particular selection
of features which they have in common and these were suggested as the
schemata of the words which the child perceives. Perhaps the rst reaction
to such a classication will be to ask whether it is not always possible to nd
enough common features among words to group them into any type one may
wish. This may be so when one is dealing with the adults whole lexicon, but
the child is building up his phonological system from nothing, i.e., one may
consider his competence to be nil at the start (that is to say at the time when he
rst begins to understand what is said to him, not when he rst begins to
talk), and it seems that the basis on which he builds is the input he receives,
i.e., utterances which are meaningful to him by their function in context, and
these at the start are few in number. He therefore has little or no expectancy
and no conditioning to inuence his perception of sounds until he gets some
system built in, i.e., gets some competence. Thus it seems that it is the
selection of features composing the utterances which are the input for the
child that determine the patterns he will acquire, and the input is decided more
by the adults than by the child (see Brown, above). This means that the
sequence in which he registers various utterances will determine which
features he will learn to perceive and reproduce rst and will thus determine
the different types of structures he will have in his phonological system. One
may take as an example the words Randall, window, and nger which were all
used frequently to P. He sees them operating in context. They thus become
meaningful for him and therefore claim his attention. He appears to perceive
certain features common to them, i.e., nasal stop which is forcefully articu-
lated, broad grades of vowel openness in syllables, certain syllable features,
and reproduces these features in a particular way, e.g., reduplicated nasal
stops, etc., and thus he has a new type of structure.
Every child has a different input as different children have different
environments and different things are said to them. This means that it is
possible for every child to register a different set of words and perceive
some similarity in the selection of features of different groups of words,
thus perceiving and reproducing different sets of features. This will result in
88 Natalie Waterson
different kinds of structures in their phonological systems so that children
learning the same language will have different forms. This does not of
course mean that there cannot be similarity; cf. Ps and Hildegards Sibilant
Structures. Although it has been noticed that there is a tendency for children
to acquire certain sounds earlier than others (see references to Jakobson and
others on p. 62), i.e., those of which the articulation does not require great
skill in timing and coordination, e.g., stops and nasals and an open
unrounded vowel, this does not mean that their phonological patterns will
be the same; in fact they are usually different, and that is why when
children rst begin to speak, they are often not understood by speakers of
the same language outside the family.
It is not possible on the evidence of the analysis given in this paper to suggest
whether in the very early stages a child rst observes similarity in the feature
selection of several words before he attempts to reproduce them in speech or
if he perceives particular features in each word independently and reproduces
them. Whatever way it happens, the child produces similar forms, i.e., with the
same basic features, and thus a type of structure with a particular selection of
features in a particular sequence becomes established in the childs phonolog-
ical system. Once such a structure is established, he has a framework within
which he perceives other utterances which have the same selection of features,
i.e., he has some competence which gives hima certain expectancy. This may be
illustrated from the structures of Leopolds child with nal sibilant fricative
quoted on p. 85 which were obviously based on adult forms with nal sibilant
friction. Other words with the same basic features but in a different sequence,
i.e., with sibilant onset instead of ending, e.g., story and stone, were reproduced
by her with features in the same sequence as the rest of her words belonging
to that particular structure, i.e., with sibilant ending, and not in the sequence
found in the adult forms, thus her competence conditioned her performance.
Presumably when a childs perception sharpens and the input includes more
adult forms with the same selection of features but in a sequence different from
the one on which the childs structure is based, the childs structure expands to
include the new sequence.
The writer has not yet made a thorough study of Ps acquisition of grammar
but has reason to believe that it is possible to show that the grammar of the
language was acquired in a similar way, i.e., he observes an utterance as a whole
and perceives certain basic features of grammatical structure in the utterance
which are linked with stress and prominence. It is mostly the stressed and
prominent words of a sentence that a child reproduces so that many unstressed
words are left out, hence the telegraphic effect, to borrow a term from Fraser,
Bellugi and Brown (1963, reprint 1968: 50). These then are the basic features
or units on which his sentence structures are built. Such basic units of grammar
are established on the basis of regularly recurring structures which can easily be
related to the context by the child, i.e., such as are functional for him, e.g., in the
case of P such sentences as Bobs a good boy, said to the dog when he does as he
Child phonology: a prosodic view 89
is told, and Annes a good girl, said to the goat at milking time. Ungrammatical
or anomalous sentences are unlikely to play a part because they do not recur
often enough for a child to register them. It is possible, therefore, that a child
perceives certain basic patterns of regularly used sentence types, i.e., the
schemata of the sentences. These are reproduced by him and he uses such
patterning as a model for his own sentences; cf. Ps reciting of sentence patterns
which are obviously his own creations, e.g., on 11.7.60 he said [n g g:, dada
gg:, ba: gg:] (Annes a good girl, Daddys a good girl, Bobs a good girl).
These are apparently modelled on Annes a good girl. As a child gains more
experience and as his phonological system develops and he is able to perceive
more, he appears to grow more aware of ner grammatical distinctions,
many of which occur mainly in unstressed positions in the utterance, such as
prepositions, conjunctions, weak forms, gender concords, etc., which in the
early stages are not reproduced and probably are not so clearly perceived; they
get gradually incorporated into his sentence structures and so his grammatical
system grows. The basic units of sentences can be expected to vary to some
extent from one child to another, as what would be basic for each child would
depend on the type of sentence structures which were the input for him; but
it seems that more similarity can be expected in the basic units of grammatical
structures than in phonological structures of English children because
English-speaking adults seem to use the same sort of sentence structures to
children in the main so that they have mostly the same sort of structures as
input, e.g., simplied grammatical structures such as Mummy do it, wheres
pencil?, Baby want Teddy?.
The study of the pattern of the acquisition of the grammar requires a separate
paper and the brief comment on the subject is only put forward here as the
obvious corollary of the pattern of the acquisition of the phonological system
of the child P, thus showing that the pattern of the acquisition of grammar and
phonology seems to be a coherent whole. It is somewhat rash to put forward
speculations about the acquisition of grammar before the grammatical study is
complete but they are made in the hope that those concerned with problems of
language acquisition will be provoked either to support the views expressed
here or to offer reasoned arguments to disprove them.
At the present time there is much speculation about what constitutes a childs
capacity for language acquisition, e.g., Chomsky (1965: 362, 1966: 11113,
1967: 397442), Katz (1966: 24082), McNeill (1966: 6585), Lenneberg
(1965, 1966a, 1966b, 1967). The evidence given in this chapter suggests that
P perceived some sort of schema through the recognition of a particular set of
features out of the selection of features of which groups of adult forms were
composed, and this resulted in his producing his own related forms with one
structural pattern. If this proves to be the general pattern of how a child acquires
the phonological system of his mother tongue, it may be that part of a childs
capacity for the acquisition of the phonological system is the ability to perceive
schemata in the sound patterns of utterances.
90 Natalie Waterson
notes
1. [d ] = labialized [d]; [ ] = labialized [].
2. The sign stands for no recorded form. The rst recorded form for ower was at
1;6.
References
Bellugi, U. and Brown, R. (1964). The acquisition of language. (Monograph of the
society for research in child development, serial no. 92. 29. 1.) Lafayette, IN: Child
Development Publications.
Brown, R. (1938). How shall a thing be called? Psychological Review, 65, 1421.
Reprinted in Oldeld and Marshall (eds.) (1968).
Carroll, J. B. (1961). Language development in children. In Saporta (ed.) (1961).
Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.
(1966). Current issues in linguistic theory. The Hague: Mouton.
(1967). The formal nature of language. In Lenneberg (1967), pp. 397442.
Cohen, M. (1969). Sur ltude du langage enfantin. Enfance, 34. Paris.
Fodor, J. A. and Katz, J. J. (eds.) (1965). The structure of language. Englewood Cliffs,
NJ: Prentice-Hall.
Fraser, C., Bellugi, U., and Brown, R. (1963). Control of grammar in imitation, com-
prehension and production. Journal of Verbal Learning and Behavior, 2, 12135.
Reprinted in Oldeld and Marshall (eds.) (1968).
Fry, C. B. (1966). The development of the phonological system in the normal and the
deaf child. In Smith and Miller (eds.) (1966), pp. 187206.
Gimson, A. C. (1964). An introduction to the pronunciation of English. London: Edward
Arnold.
Grgoire, A. (1933). Lapprentissage de la parole pendant les deux premires annes de
lenfance. Psychologie du langage, ch. 5. Paris.
(1947). Lapprentissage du langage. vol. 2, Facult de philosophie et lettres. Lige.
Fascicule CVI. Librairie E. Droz, Paris 25, rue de Tournon.
Ingram, T. T. S. (1966). Syntactic regularities. General discussion. In Lyons and Wales
(eds.) (1966), pp. 21419.
Jakobson, R. (1941/1968). Child language, aphasia and phonological universals, trans.
A. R. Keiler. The Hague: Mouton. (Originally Published as Kindersprache, Aphasie
und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell, 1941.)
Jakobson, R. and Halle, M. (1961). Phonemic patterning. In Saporta (ed.) (1961).
Jones, D. (1962). An outline of English phonetics, 9th edn. Cambridge: Heffer.
Katz, J. J. (1966). The philosophy of language. New York and London: Harper & Row.
Ladefoged, P. (1967). Three areas of experimental phonetics. London: Oxford University
Press.
Lenneberg, E. H. (1965). The capacity for language acquisition. In Fodor and Katz (eds.)
(1965), pp. 579603.
(1966a). Abiological perspective of language. In E. H. Lenneberg (ed.), New directions
in the study of language, pp. 6588. Cambridge, MA: MIT Press.
(1966b). The natural history of language. In Smith and Miller (eds.) (1966), pp. 21952.
(1967). Biological foundations of language. New York, London, and Sydney:
J. Wiley & Sons.
Leopold, W. F. (1939). Speech development of a bilingual child, vol. 1 (1954 reprint).
Evanston: North Western University Press.
Child phonology: a prosodic view 91
(1947). Speech development of a bilingual child, vol. 2. Evanston: North Western
University Press.
(1961). Patterning in childrens language learning. In Saporta (ed.), (1961), pp. 3508.
Lewis, M. M. (1968). Infant speech: a study in the beginnings of language. London:
Routledge & Kegan Paul.
Lyons, J. and Wales, R. J. (eds.) (1966). Psycholinguistics papers (Proceedings of the
Edinburgh Conference 1966). Edinburgh University Press.
McNeill, D. (1966). The creation of language by children. In Lyons and Wales (eds.)
(1966), pp. 99132.
Ohnesorg, K. (1959). Druh fonetick studie o dtsk ei. Bmo: spisy university v Brn
losock fakulta. 57.
Oldeld, R. C. and Marshall, J. C. (eds.) (1968). Language. Harmondsworth: Penguin.
Piaget, J. (1967). The language and thought of the child, trans. M. and R. Gabain.
London: Routledge & Kegan Paul.
Saporta, S. (ed.). (1961). Psycholinguistics: a book of readings. New York: Holt,
Rinehart & Winston.
Smith, F. and Miller, G. A. (eds.) (1966). The genesis of language: a psycholinguistic
approach. Cambridge, MA, and London: MIT Press.
Velten, H. V. (1943). The growth of phonemic and lexical patterns in infant language.
Language, 19, 28192.
Waterson, N. (1970). Some speech forms of an English child: a phonological study.
Transactions of the Philological Society, 3450.
Weir, R. H. (1962). Language in the crib (Janua Linguarum. Series maior, XIV). The
Hague: Mouton.
92 Natalie Waterson
4 Words and sounds in early language acquisition
Charles A. Ferguson and Carol B. Farwell
In acquiring full control over the language of his speech community, the child
must learn to deal with an enormous array of lexical and phonological
elements, as well as with the complex relations among these elements which
constitute the grammar of a particular language, different from all other
possible languages. In addition to the machinery of the language itself, he
must learn when and how to use the language in accordance with his own
needs and the norms of the community. And all this confronts the child not in
neat, separate units, but in conglomerate batches which he must largely sort
out for himself. Even if the speech input to which he is exposed is restricted in
scope and simplied in structure, as the talk addressed to young children tends
to be, the analytic problem is severe, and it must not be expected that the
childs early attempts will match with any great precision the adults language
behavior and its underlying principles of organization.
Thus the linguist who wishes to identify analytic units in the childs
speech encounters even greater pitfalls than he does in abstracting from
the adults speech those components at various levels which merit analytic
autonomy. Looking for distinctive features, inectional categories, syntactic
rules, and all the dozens of other possible basic units in a childs linguistic
system is a hazardous pastime; yet if we are to understand the processes of
language development indeed of language behavior in general we must
make the effort to do so, since it is manifestly impossible to deal with the
childs language in one large undifferentiated mass.
1
In the present study, we
examine the language development of the child in terms of two putative
units: words and word-initial consonants. In the description and analysis
which follow, no assertion is made that these units are independent of all
other possible units, or that recognition of these two units precludes recog-
nizing certain other possible units (e.g., morphemes/formatives, syllables,
sentences, prosodies, schemata, idioms, distinctive features, rules,
agreement . . .). What is assumed is twofold: (1) words and word-initial
consonants are valid units of analysis from the earliest productions of
meaningful speech by the child, and (2) it is instructive to study these two
units in relation to each other.
93
1. Data
The data used here are a small part of those collected in a longitudinal study of
seven children, conducted as a part of research on the development of conso-
nants in rst-language learning.
2
The children, four girls and three boys of
monolingual English background, were selected for the study when they were
reported by their parents to use several words. Ages at the beginning of the
study ranged from 0;11 to 1;2.
1.1. Procedure
Each child was visited at home at approximately weekly intervals for seven to
ten months, with occasional larger gaps because of illness and family vacations.
Three observers participated in the project, two attending each session when
possible. For about half the sessions, only one observer was present, but each
child was seen consistently by the same observer.
During each half-hour visit, attempts were made by parents and observers to
elicit as many of the childs words as possible, by the use of picture books and
things familiar to him (food, toys, etc.). The sessions were tape-recorded, and
notes were made by the observer(s) of the probable adult equivalent of each
utterance. Utterances were considered meaningful if there was sufcient con-
sistency to allow recognition of the form, and if there was some consistency in
reference or accompanying action not necessarily exactly that expected from
the meaning of the adult word. Similarly, it was not required that a specic adult
English equivalent should be identied. Occasionally, it was found that a child
would consistently use a form for which no probable adult equivalent could
be imagined. In fact, however, such uninterpretable words occurred much
less frequently than expected. They were included in the data, as well as
forms which seemed to correspond to whole adult phrases rather than words,
e.g., I see you.
Identication of words was aided by parents recognition, although observers
often obtained evidence of the use of a particular word before parents noticed
it. We assume that our judgment of the identity of meaningful forms is valid.
McCurry and Irwin (1953) demonstrated 91 percent inter-observer agreement in
the determination of meaningful utterances and their referents in naturalistic
settings, and our agreement in sessions attended by more than one observer was
similarly high.
Child utterances were transcribed using the techniques established by the
Phonetics Workshop of the Child Phonology Project, Fall 1971, and problems
were referred to that workshop. An expanded IPA symbol grid was used
(Johnson and Bush 1972). Transcription is to a level comparable to that in
Leopold (193949), with narrower transcription of initial consonants and less
attention to vowels.
94 Charles A. Ferguson and Carol B. Farwell
1.2. Subjects
This chapter reports the early stages of development of two girls, Tand K, from
the larger study. Utterances occurring from the beginning of the study to the
week in which the ftieth word type was recorded are included.
3
In order to
provide a reference point for our analysis, Hildegard Leopold (H) has been
included as a third subject, since information about her development is widely
known and generally available.
T was a rst child and spent almost all her time with her two parents. Her
mother kept a detailed list of words produced by T during each week, and
conscientiously elicited new words for us at each session. T had just begun to
walk when we started our study. She did not engage in much babbling, and she
imitated only infrequently usually words she had already produced herself.
Pivot-type syntax, especially with the words hi and where, was evident from the
rst sessions; and two-word utterances became more common soon after the
session with the ftieth word.
K had an older brother and, possibly in self-defense, was physically aggres-
sive and active. She spent time with both her parents and a housekeeper, and
was often left alone with investigators during a taping session. Our tapes of K
contain a lot of babbling or at least unintelligible speech, and she showed
willingness to imitate almost any word beginning with a sound at least close to
one she could say. Even during the rst sessions where our data are scanty, she
would occasionally imitate or even spontaneously say three-word sentences;
and our general impression was that she was more adventuresome and less
concerned with details than T and H.
H, a rst child, was deliberately raised as a bilingual: her father spoke to her
only in German, her mother in English. She spent two months (age 1;0 and 1;l)
in Germany where even her mother spoke only German to her, and for some
time on her return she did not understand English. To make the Leopold data
comparable, only the words which H still said at age 1;0 were included in the
study; but those words are followed from their beginnings, back to 0;10. She
imitated very rarely always words which she understood. Until 1;5, the last
month considered here, many of Hs words occurred only in whispered form,
although some had full voice from the beginning, and a whisper/voice distinc-
tion sometimes separated homonyms. H learned to walk in the second half of
1;1, a month and a half after the beginning of the period studied here. On the
whole, H was cautious: It was characteristic of her that she generally avoided
altogether any words the meaning and form of which she could not successfully
cope with (Leopold, I.172).
4
The children and the number of sessions reported here are shown in Table 4.1.
Because no natural criteria present themselves for grouping weekly sessions
together, each session has been analyzed separately. For H, grouping is done
month by month, since Leopold tells us only the month in which each form
occurred. The main effect of the use of larger time divisions with H is that
Words and sounds in early language acquisition 95
uctuations from day to day are likely to be lost in the general trend of
development. This tendency coupled with the fact that Leopold often reports
only a fewphonetic variants of a word during a month, while one of our children
might produce as many as eight variants of a word in one session tends to
make Hs progress look much smoother than that of the other two children. Far
from making the two sources of data incompatible, such a difference can be put
to good use: Hs development can help us recognize overall trends within the
variant forms in our data, while our data can make clear the degree of simpli-
cation in the H data.
1.3. Imitations and other problems
In a study of child phonology, as in any other phonological work, it is common
to exclude certain problematic forms of data from analysis. For example,
utterances in which a child imitates or echoes an immediately prior adult
utterance are often separated from other, spontaneous utterances. Researchers
have sometimes found that such imitations may be more accurate phonetically
than the same forms said spontaneously; and they have excluded imitations in
order to maximize the number of utterances processed by the childs phono-
logical system, rather than by a separate imitative ability.
There are several reasons why we have not excluded imitations from analysis
in this study. For one thing, a very high percentage of what a one-year-old says
is imitated, so that there is very little purely spontaneous data. Furthermore, a
study of the forms collected shows that a separation of imitated from sponta-
neous forms, where the two can be compared, does not correspond in any
straightforward way to a separation of different forms of the same word.
Finally, even children this young can repeat or imitate things said by adults at
some distance of time ve minutes or more despite considerable intervening
speech, so that no simple denition of imitation is feasible. Hence a separation
of imitated utterances has not been carried out here, since it would lead to a great
reduction of available data without any demonstrable gains of accuracy or
homogeneity although such a separation might be methodologically sound
when dealing with older children, where data are not so limited. (For discus-
sions of this whole question from different points of view, see Templin 1947,
Olmsted 1971: 945, and Edwards and Garnica 1973.)
Table 4.1. Periods of elicitation for the three subjects T, K, and H
Child Age at beginning Number of sessions Time span Total no. of words
T 0;11 9 13 weeks 51
K 1;2 13 13 weeks 72
H 1;0 6 months 54
96 Charles A. Ferguson and Carol B. Farwell
Several kinds of data have been excluded, however. In order to make the three
children comparable, forms which Leopold himself questions or which H
seemed to repeat once have been excluded, as well as exclamations which
probably would not have been collected from our children. Some of Hs words
have been included several months later than Leopold rst lists them. Similarly,
marginal forms such as mmm, hm-m, tsk-tsk, etc., as well as onomatopoeic
words in which imitative qualities obscure the segmental phonology, have been
left out in all three children. However, Hs sch-sch has been included because of
its conventional referential meaning although it is extremely marginal phono-
logically, the [] being syllabic and not occurring before a vowel, like other
consonants.
5
Finally, certain forms have been included even though they present problems
for the analysis of word-initial consonants. A short listing of three cases in
which this occurs may help explain some of the variation observed:
(a) Backgrounding: the word-initial consonant is deleted or drastically reduced
when the child is working on another part of the word (for full discussion
of trade-off phenomena in phonological development, see Edwards and
Garnica). One example from our data shows two forms of a word: (T IX)
milk [b, k
].
6
(b) Assimilation and syllable deletion: here a word-initial consonant is
affected by a phonological rule. Such cases are familiar from the literature.
Examples of each are: (K IX) sh [i, k
h
i], (K IV) thank you [
m
kj].
7
(c) Prosodic phenomena: here the child treats the whole word, rather than its
segments, as a phonological unit. Two examples are: (T III) shoe [guti,
gutidi], (T IX) feet [
t
].
8
1.4. Phone classes and phone trees
One way to proceed in analyzing the initial consonants in the data would be to
group together all recurrences of the same phonetic symbols used in tran-
scription. Such a structureless listing is unilluminating for several reasons.
First, it simply does not show which different symbols might be regarded as
variants of one another, i.e., which sounds are in some structural sense related
and which are not. How similar must two sounds be for the analyst to decide
they belong together? Second, it does not allow the very likely possibility of
overlap in the phonetic value of different structural units or features. The phone
represented by a given phonetic symbol may be a production sometimes of one
phonological unit, sometimes of another. Finally, this procedure offers no
satisfactory way to relate the phones of one session with those of another
session. If one speech sound has changed sufciently between one session
and the next to be reported with a different symbol, how does the analyst
recognize this fact? Or if a child has nine phones (i.e., different phonetic
symbols) at one session, and twelve at the next, how is one to relate the two
systems?
Words and sounds in early language acquisition 97
What is needed is a way to determine which phones belong together or
correspond to one another, and the most obvious way is to use the word as the
framework for phone identication and classication. This is hardly a new idea,
since it is implicit in much of the phonological analysis of child language, but it
seems never to be made explicit (thus Francescato 1968 criticizes Jakobson and
others for not making explicit use of the word, although he himself does not
offer analysis of this kind).
By using the word as the basis of comparison, it is possible to establish the
notion of correspondence or corresponding phones, similar to the notion of
sound correspondence in comparative linguistics. For the purposes of our study,
in which we are dealing only with initial consonants, we may dene corre-
sponding phones essentially as any two consonants which begin different
utterances of the same word, whether at a single session or different sessions.
This denition must be modied to exclude instances of omission or assimila-
tion which may put non-corresponding phones in initial position.
The procedures employed in our analysis were as follows. For each session,
all the renditions of a given word were grouped together, and all variants of the
initial consonants in those renditions were noted. Then all words beginning with
the same phone or set of variant phones were put together. The set of initial-
consonant variants of each of these groups of words constitutes a phone class,
and is represented by the appropriate phonetic symbols in a box, or between
vertical lines.
9
Thus a phone class |d ~ t
h
| consists of the initial consonants of all
of those words whose initial-consonant sound varied between [d] and [t
h
]. All
the phone classes of one child at one session were represented by boxes in a
horizontal row, arranged roughly in order of place of articulation. Thus a child
might show three phone classes of initial consonants at a particular session:
jp~bj jmj jt~dj
After this, phone classes in different sessions were constructed according to
the occurrences of the same word. With each session making up a horizontal
level, solid vertical lines were drawn between successive phone classes if they
contained the same word. If successive phone classes did not contain the same
word but were related to phone classes which did, dotted lines were drawn
connecting them. For example, in Ts |m| class:
jmj mama
.
.
.
jmj milk
j
jmj milk; mama
In addition, and especially in the case of K, dotted lines were used to connect
phone classes which were each well-motivated and were phonetically close or
identical, but shared no words in common (especially Ks |b~p| in IX to XII).
10
98 Charles A. Ferguson and Carol B. Farwell
Diagrams of this kind which connect corresponding phone classes of suc-
cessive stages constitute phone trees. The phone trees constructed for T, K,
and H appear as Figures 4.1, 4.2, and 4.3. In each gure, the number in
parentheses to the right of each phone class indicates the number of words
belonging to that class.
Sometimes the phone classes are not as simple as described above. Thus the
phone class |b ~ ~ bw ~ p
h
~ ~ | in T VI contains the following words and
initial-consonant variations: baby [b ~ ], ball [b], blanket [b], book [b ~ ],
bounce [b], bye-bye [b ~ p
h
], paper [b ~ ]. One might reasonably make
several phone classes out of these words, perhaps separating those in which [b]
does not vary or varies only with from those in which variation is with a
fricative or voiceless stop. For our purposes, they have been grouped together in
opposition to the phone class |p
h
| in which the following words occur: pat, please,
pretty, purse all beginning only with aspirated [p]. The claim of this grouping is
that it is only accidental that some words in the |b| class were found with variation
of one sort, and some with another; but that it is not accidental that the words in the
|b| class are separate from those in the |p| class.
In fact, if we look at the corresponding classes in the next session, we nd the
following: baby [b ~ w ~ p], ball [b], bang [b], blanket [b], book [b], bounce [b],
box [b], bye-bye [b ~ ]; but paper [p
h
], pat [p
h
], purse [p]. From the data
listings, it can be seen that baby occurs seven times with an initial [b], once with
a [p
h
], and once with a [w]. Bye-bye occurs three times with a [b] and once with a
[]. Hence it seems justiable to group them with other [b] words, and again it
seems that the important split is between the [b] words and the [p] words.
The notion phone class here is similar to the notion phoneme of
American structuralism, in that it refers to a class of phonetically similar speech
sounds believed to contrast with other such classes, as shown by lexical
identications. The determination of the phone classes of a particular childs
speech is made by methods similar to linguists procedures of elicitation and
phonemic analysis, but largely without the benet of minimal pairs and speak-
ers judgments. The purpose of the exercise (as ultimately for phonemic analy-
sis as well?) is to locate valid behavioral units.
In general, an attempt was made to distinguish as few phone classes as
possible, so that any error would be in the direction of underdifferentiation.
Consider the word dog in T IVIII. It is included in phone classes with some
variation even though the word dog itself is consistently produced with an initial
d. By Session VIII, however, dog seems to belong to a phone class by itself, and
perhaps it should have been separated all along.
Even with the policy of minimal differentiation, it may happen that phone
classes are separated unjustly. Consider Ts two classes |
t
s ~ s ~ | and | ~ ~ d|
in Session V. Although the regular criteria require their separation during that
one session, the fact that they are joined in the sessions before and after
suggests that the criteria are misleading in this case. A similar example is
the separation of |d ~ t
h
| and |t
h
| classes in Sessions VIII and IX. However,
Words and sounds in early language acquisition 99
I
IV
V
VI
VII
VIII
IX
II
III
b~d
b
p
h
p
bp
h
p
bbwp
h
bp
h
w bw
p
h
p
h b
p
h
p
b
b
(1)
(1)
(2)
(2)
(8)
(8)
(4)
(4)
(4)
(3)
(3)
(7)
mb
mmn
mb
(1)
(1)
(2)
(1)
w
w
w
w
w
w~b
(2)
(2)
(2)
(1)
(1)
(1)
f~p
h
f
(1)
(1)
d~t
h
~d
dd
ddj
dt
h
ddt
d
d
d
g
(1)
(1)
(2)
(2)
(3)
(4)
(2)
(2)
(1)
dt
h
g
d~d
t
h
t
h
t
h
tt jdZ
(2)
(2)
(2)
(1)
t
h
t
t
h
(3)
(3)
n
n
n
n j
(2)
(1)
(1)
(1)
(1)
s jsC
C
(3)
(3)
t Zj
t
h
sh
C
t
C
t
d jdZ
t dZ
(2)
(3)
(1)
t
ss Cd
ct
h
(2) (1)
(2)
(3)
Ct
h
II jh
(1)
dZd j
k
h
k
x
k
h
t
h
k
h
k
k
w
g
k
x
x
g
g
(1)
(1)
(1)
(1)
(1)
(2)
k
h
h
h?
h?
h?
h?
?
?
?
? h
h
(1)
(1)
(1)
(2)
(2) (1)
(1)
(4)
(4)
(2)
(3)
(2)
t
h
(2)
(4)
Figure 4.1. Phone trees of T
m
m
m
f
m
b
b
g
g
s
1
n
g
g
g
g
h
n
d
d
h
h
?
?
h?
?
h
n
d
dt
h
szZ
b
b
b
bp
d
hd
bp
b
dg
f
mn
m
b
m
ddt
bp
h
bp
h
t
h
t
t
h
t
h
t
h
k
h
k
h
k
h
pp
h
kh
k
h
p
h
b
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
(1)
(1) (1)
(1)
(4)
(1)
(2)
(2)
(2)
(2)
(2)
(2)
(2)
(2)
(5)
(1)
(4)
(4)
(2)
(2)
(1)
(1)
(1)
(1)
(1)
(1)
(1)
(1)
(1)
(1) (1)
(1)
(1)
(1)
(1)
(1)
(1)
(1) (1)
(1)
(1)
(3)
(3)
(1)
(1)
(1)
(1)
(1)
(1)
(2)
(1)
(1)
(1)
(3)
(3)
(3)
(3)
(2)
(2)
(2)
(2)
(1)
b
?
d
d
Figure 4.2. Phone trees of K
p
p
p
p
p
b m
m
mb
td
n l
pb
pb
b
b
b
b
b
t
t
t w
w
w
w
w m
t
t
t
t
d
d
j
j h
?
?
?
h
d
g
d
d
d
d
(1)
(1)
(1)
(1)
(1)
(1)
(1)
(1)
(1) (1)
(1)
(1)
(1)
(1)
0.10
0.11
1.00
1.1
1.2
1.3
1.4
1.5
(1)
(3)
(1)
(3)
(1) (3)
(2)
(3)
(5)
(1)
(2)
(2)
(5)
(2)
(2)
(2)
(7)
(8)
(8)
(2)
(2)
(4)
(3)
(3)
(2)
(2)
(6) (2)
(2)
(1)
(1)
(1) (1)
(1) (1)
(1)
(1)
(1)
(1)
(1)
(1) (1) (2)
(2)
(6)
d
p
p
p
j
] (imitation), (2) [
] (imitation), (3)
[d
dn
], (4) [hin], (5) [
m
b], (6) [p
h
in], (7) [t
hn
t
h
n t
h
n ], (8) [ba
h
], (9) [d
h
au
N
], (10)
[bu]. K seems here to be trying to sort out the features of nasality, bilabial closure,
alveolar closure, and voicelessness.
9. The box was adopted as a convenient symbol, different from the brackets and slant
lines used in phonological transcriptions, and suggestive of the unity of the class; the
same symbol was used with a somewhat similar value in Jakobson 1949. For
typographic reasons, the box is replaced by vertical lines in the present text, although
boxes are used in the gures.
10. There is a danger that phone classes containing the same words may not actually
correspond because of an intervening reanalysis of a certain word at the input level
by the child. There is evidence that such reanalysis does take place: see Smiths
example of some and its compounds (1973:1456). Probably, however, such rean-
alysis is relatively infrequent, and in any case not directly related to the development
of the sound system.
11. Two important recent exceptions should be mentioned: Chen (1972), for a fuller
discussion of different approaches to sound change and an explication of Wangs
model as applied to the lexical parameter; and Labov (1972), for a typology of sound
change along social and lexical parameters. In the absence of any linguistically
motivated ordering principle, we assume that phonological change affects earliest
child language according to Labovs Model E, Random decomposition.
12. Editors note: the full reference is missing in the original paper.
References
Berko Gleason, J. (1973). Code switching in childrens language. In T. E. Moore (ed.),
Cognitive development and the acquisition of language, pp. 15967. New York:
Academic Press.
Bloomeld, L. (1933). Language. New York: Holt.
Chen, M. (1972). The time dimension: contribution toward a theory of sound change.
Foundations of Language, 8, 45798.
114 Charles A. Ferguson and Carol B. Farwell
Chomsky, N. (1964). Formal discussion. In U. Bellugi and R. Brown (eds.), The
acquisition of language (Monographs of the Society of Research on Child
Development 29:1), 359. Lafayette, IN: Purdue University.
Cruttenden, A. (1970). A phonetic study of babbling. British Journal of Disorders of
Communication, 5, 11017.
Edwards, M. L. and Garnica, O. K. (1973). Patterns of variation in the repetition of
utterances by young children. MS.
Ferguson, C. A. (1971). Short a in Philadelphia English. Stanford Occasional Papers
in Linguistics, 1, 227. (Also in M. Estelle Smith (ed.), Studies in linguistics in
honor of George L. Trager, pp. 25974. The Hague: Mouton, 1973.)
Ferguson, C. A. and Garnica, O. K. (1973). Theories of phonological development. In
E. H. and E. Lenneberg (eds.), Foundations of language development, pp. 15380.
New York: Academic Press.
Ferguson, C. A., Peizer, D. B., and Weeks, T. E. (1973). Model-and-replica phonological
grammar of a childs rst words. Lingua, 31, 3565.
Ferguson, C. A. and Slobin, D. I. (eds.) (1973). Studies of child language development.
New York: Holt, Rinehart & Winston.
Francescato, G. (1968). On the role of the word in rst language acquisition. Lingua 21,
14453.
Hsieh, H-I. (1972). Lexical diffusion: evidence from child language acquisition. Glossa,
6, 89104.
Ingram, D. (1972). Phonological analysis of a developmentally aphasic child. Mimeo,
Institute for Childhood Aphasia, Stanford University.
(1973). Phonological rules in young children. Journal of Child Language, 1, 4964.
Jakobson, R. (1949). Principes de phonologie historique. Appendix to Principes de
phonologie, by N. S. Troubetzkoy. Paris: Klincksieck. (Originally presented at the
Runion Phonologique Internationale, Prague, 1930 and published in German,
Travaux du cercle linguistique de Prague, 4 (1931). Also in R. Jakobson,
Selected writings I, 20220, 1962.)
(1941/1968). Child language, aphasia and phonological universals, trans. A. R. Keiler.
The Hague: Mouton. (Originally published as Kindersprache, Aphasie and allge-
meine Lautgesetze. Uppsala: Almqvist & Wiksell, 1941.)
Johnson, C. E. and Bush, C. N. (1972). A note on transcribing the speech of young
children. Papers and reports on child language development, Stanford University,
3, 95100.
Kavanagh, J. F. and Mattingly, I. G. (eds.) (1972). Language by ear and eye. Cambridge,
MA: MIT Press.
Labov, W. (1972). The internal evolution of linguistic rules. In R. P. Stockwell and
R. K. S. Macaulay (eds.), Linguistic change and generative theory, pp. 10171.
Bloomington: Indiana University Press.
Leopold, W. F. (193949). Speech development of a bilingual child, 4 vols. Evanston, IL:
Northwestern University Press.
McCurry, W. H. and Irwin, O. C. (1953). A study of word approximations in the sponta-
neous speech of infants. Journal of Speech and Hearing Disorders, 18, 1339.
Malkiel, Y. (1967). Each word has a history of its own. Glossa, 1, 13749.
Menn, L. (1971). Phonotactic rules in beginning speech. Lingua, 26, 22551.
Moskowitz, A. I. (1971). Acquisition of phonology. Dissertation, University of
California, Berkeley.
(1972). Idiomatic phonology and phonological change. Mimeo, University of
California, Los Angeles.
Words and sounds in early language acquisition 115
Olmsted, D. L. (1971). Out of the mouth of babes. The Hague: Mouton.
Smith, N. V. (1973). The acquisition of phonology. Cambridge University Press.
Templin, M. C. (1947). Spontaneous versus imitated verbalization in testing articulation
in pre-school children. Journal of Speech Disorders, 12, 293300.
Wang, W. S-Y. (1969). Competing changes as a cause of residue. Language, 45, 925.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
Weeks, T. (1971). Speech registers in young children. Child Development, 42, 111931.
Appendix 1: Primary data from T, K, and H
Data from T
Session I: 34 utterances, 4 words, 5 imitations
daddy (7) d
j, d
(2), d d
(2), dt, d
1
dog (5) d (5)
hi (20) h (2), ha (12),
h
a (3), a (3)
see (2) h
i
,
i
Session II: 30 utterances, 5 words, 8 imitations
baby (2) be
(imit.), dd
i
(imit.)
dog (4) d (2), d (2)
hi (14) a (5), ha (4), h (2), a, ha
i
, h
thank you (6) k
h
ju, k
x
ju
t
, dkk (imit.), td (imit.),
dd (imit.), dnk (imit.)
asking word (4) hlji, ljije, lij, le
Session III: 27 utterances, 9 words, 7 imitations
baby (1) de
i
bye-bye (1) bb (imit.)
daddy (1)
(3)
dog (2)
ddi
d
imit:; d
imit:
g
dd; d
; d
d
hi (4) ha (3), a
mama (8) mm (5), mmb, mnm, mm
shoe (2) gut, gutdi
thank you (3) djuk
ha
, dt
h
, d
d
tea (2) t
h
i (2)
Session IV: 26 utterances, 10 words, 6 imitations
baby (4) beb (2), bebe, p/be
ball (2) p
h
(imit), b
bye-bye (1) d b
daddy (2) dt
h
i, dd
116 Charles A. Ferguson and Carol B. Farwell
duck (1) gk
h
(imit.)
hi (4) ha (2), h, a (imit.)
see (2) :, i:
shoe(s) (5)
, t
h
u,
i
, ,
thank you (3) dt
h
, t
h
dju, n da-d
where (2) w,
Session V: 30 utterances, 12 words, 3 imitations
baby (2) be, beb
w
i
ball (3) be,
m
bi, b
h
cracker (1) t
h
di (imit.)
daddy (1) ddi
eye (3)
(1)
a (2), a
)
ha
hi (2) a, ha
rock (5) wkwk (2), wkwe, wx, bk
see (4)
t
si, i, si, s
i
shoes (3) i, d,
u
sit (1)
t
st
tea (2) t
h
i
thank you (2) d
h
di,
n
xju (imit.)
Session VI: 56 utterances, 25 words, 7 imitations
allgone (4) ag
h
o, awo, ok
h
, ok
h
u
baby (3) ei, bi, bii
ball (1) ba
please (2) p
h
e (2)
pretty (1) pr hi
purse (2) p
h
e, p
h
h
i
up (1) a
yeah (1) ij
h
Words and sounds in early language acquisition 117
Session VII: 132 utterances, 29 words, 25 imitations
allgone (4)
X
g, g (2) (imit.), axg (imit.)
baby (9) bebi (5), p
h
ebi (imit.), wepi (imit.), be, bibi
ball (1) b (imit.)
bang (1) b
blanket (2) ba, b
book (4) bk (2), bk, bx
bounce (6) b (3) (series), b b b
w
(series)
box (4)
b (2), b
h
,bk (all imit.)
bye-bye (4) bab (2), bp, w (all imit.)
chair (2)
h
(2)
cheese (6) , , sti, h, ti,
t
i
daddy (10) ddi (3), ddi (2), di, d
h
i (imit.), t
h
(imit), t
h
i (imit.), t
h
sit (5)
t
(3), (2)
tea (9) i (3), t
h
i
(3), tji, di, ti
thank you (1) t
h
t
h
tiger (3) t
h
t
h
u, tc , t
h
k
h
i
up (1) a
h
where (9) u (2), u
(2), w
(5)
Session VIII: 106 utterances, 30 words, 9 imitations
allgone (2) ak
h
ox, at
h
o (imit.)
baby (2) be bi (2)
ball (2) (2)
book (2) b
x
(limit.), b
bye-bye (1) bp
h
(imit.)
cat (2) k
h
h
, t
h
dog (4) d
, do, do, do
, , dju, tut
, ti
key (3) k
h
i (3)
mama (6) mm (2), mam (2),
, mm
118 Charles A. Ferguson and Carol B. Farwell
milk (6) mx (2), bx (2), m
e
!
x, bx
no (8) n (7),
n
out (8) ax (3), a
(2), ax, a
x, x
paper (1) p
h
ep
h
e (whisper, imit.)
pat (1) p
h
b (pat bunny)
purse (1) p
h
y (imit.)
rock (6)
u
wkwk (2), wk,
u
wx,
hu
wx,
u
wk
h
see (2) ci (2)
shoe(s) (2)
i
u,
u::
sit (5) (2), , , t
tea (4) t
h
i (3), t
h
i
thank you (4) t
h
t
h
, dt
h
a, dt
h
, t
h
ht
h
i
, t
hi (1) ha (imit.)
juice (5) dus , duc
o
u
(2), djuc (2)
milk (5) b, k
no (3) no (2), na
okay (1) k
h
e (imit.)
one (1) w
paper (3) p
h
e p
h
i (imit.)
pot-pot (1) p
h
p
h
a
pretty (3) p
h
i
, p
h
i, p
h
t
h
i (imit.)
purse (2) py, p
h
rock (3) b
h
b
h
, b, b (imit.)
shoe (5) d u, sju
(imit.), i
u
,
u (imit.)
u
sit (3) (2), .
i
, t
h
u
where (4) w
(3), w
Words and sounds in early language acquisition 119
Data from K
Session I: 33 utterances, 12 words, 6 imitations
allgone (3) k
h
o,
g , ga (imit.)
bear (1) bi
u
j (imit.)
book (1) b
w
ux
daddy (2) de
hI
,
dd (imit)
duck (1) d (imit.)
hi (1) ha
mine (1)
b
ma
monkey (1) m
k
bu (imit.)
pop (1) p
h
ap
h
(imit.)
that/there (10) d (2), d (2), id
, d, d
, d, d
, t
Session II: 6 utterances, 3 words, 1 imitation
block (1) bok
h
(imit.)
on (1)
(imit.)
off/up (2)
,
b
u
shoe (2) g (2)
Session IV: 27 utterances, 8 words, 3imitations
boom (1) bu
daddy (3) ddi
(3)
dog (4) dad
1
, d, d
w
, abd
Kimberly (1) s imu
me/mime (2) m
N
, m
no (2) , n.o
h
pen (12) m
(imit.), ~(imit.), d
dn
, hn,
m
b, p
h
n,
t
h
n (), b
h
, d
h
a
n
, bu
thank you (2) h
(imit.),
m
kj
Session V: 19 utterances, 11 words, 11 imitations
car (1) gk
h
dog (2) d (imit.), do
(imit.)
duck (2) dk
h
, d
gum (1) g
hot (4) ht
h
, h
h
, ht
h
, h (all imits.)
hot tea (1) hdi (imit.)
Max (1) m.~ (imit.)
Nona (3) mm
h
(imit.), nun
, mun
h
120 Charles A. Ferguson and Carol B. Farwell
Satchiko (1) un (imit.)
there (2) d, d
turkey (1) t
hI
g (imit.)
Session VI: 6 utterances, 4 words, 3 imitations
balloon (2) bp
(imit.), b
b
box (1) b
t
(imit.)
hot (1)
h
d (imit.)
that (2)
d,
t
h
Session VII: 11 utterances, 6 words, 9 imitations
boy (1) b (imit.)
cow (2) g
h
(imit.), gj
lady (2) lad, ld (imit.)
moo (2) b (imit.), b
(imit.)
telephone (3) t
h
d, t
h
, tddd (all imit.)
thank you (1) x
h
(imit.)
Session VIII: 5 utterances, 4 words, 3 imitations
allgone (2) ul
, a
(both imits.)
night-night (1) m
(imit.)
off (1)
watch (1) bt:
h
Session IX: 15 utterances, 12 words, 7 imitations
allgone (1) aw
kh
bear (1) b
t
(imit.)
bee (1) bi:
bird (2) b
n
d, p
u
x (both imits.)
coat (1) k
h
k
hI
cookie (1) k
h
xju
M (2) , k
h
(both imits.)
girl (1) k
h
(imit.)
no (1) n
pumpkin (1) bg
n
(imit.)
tea (1) t
h
N
turn (off) (2) t
h
r, t
h
na
Session X: 15 utterances, 13 words, 3 imitations
all done (1)
.
ba
baby (1) p
h
e (imit.)
boot (1) b
h
bottle (1) badf
Its a toothbrush (1) is i (imit.)
Max (1) m
k
h
(imit.)
me (1) m
moo (1) mu:
on (2) al
, a:dit
h
put (1) p
h
u
see (1) i
Words and sounds in early language acquisition 121
that (1) d t
h
up/off (2) ,
Analyzed sentence
me see that m i et
h
Session XI: 11 utterances, 10 words, 7 imitations
balloon (2) bo
u
wu, bow
book (1) bk
h
bow (1) pvo
(imit.)
dog (1) gak
h
i (imit.)
driver (1) dd (imit.)
girl (1) g
x (imit.)
kitten (1) k
v
h
i
d
n
pencil (1) p
hI
s
n
(imit.)
recorder (1) fod
(imit.)
sock (1) gk
h
(imit.)
Session XII: 26 utterances, 19 words, 7 imitations
baby (1) p
h
eb
h
i
break (1) p
h
k
h
(imit.)
bye-bye (1) b:
t
ears (1) irw
x
eye (2) ai
h
, ha
i
(both imits.)
go (1) go
its a (1) its
key (1) k
h
e
h
mama (1) b
w
h
moo (1)
mu.
h
(imit.)
mouth (1) m
f
no (4) n
u
, no (2), n
nose (1) n
(imit.)
puppy (1)
bbi
i
shoe (2) i
u
(imit.), hu
l
that a (2)
d
h
h
(imit.), di
there (1)
h
il
tie (1) t
h
ai
l
woofwoof (2) f,
Analyzed sentences
that a woofwoof di
f
Its a baby itsa p
h
eb
h
i
Data from H
0;10 4 utterances, 2 words
pretty (1) prti (wh/vd)
there (3) di, dii, de:
122 Charles A. Ferguson and Carol B. Farwell
0;11 3 utterances, 3 words
pretty (1) priti (wh/vd)
there (2) d:
ticktock (1) tak (wh)
1;0 11 utterances, 9 words
ball (1) ba (wh)
Blumen (1) bu (wh)
da (1) da:
Opa (a) pa (wh)
Papa (1) pa-pa (wh)
piep! (1) pi, pipi
pretty (1) pti (wh)
sch-sch (1) -
ticktock (2) ti-ta, ta, t-t (wh)
1;1 21 utterances, 15 words
ball (1) ba (wh) pieks! (1) by
bimbam (1) bi: piep! (3) pi:, pi:p, pi pi
da (1) da pretty (2) priti, prti
Gertrude (2) d:da, d:di sch-sch (1) -
kiek! (1) ti Tante (2) da-da, di-d
kritze (1) tits (wh) ticktock (1) t-ta
Opa (1) pa-o (wh) Wauwau (1) wa wa (falsetto)
Papa (2) pa-pa, ba-ba (wh)
1;2 14 utterances, 13 words
baby (1) bebi Papa (2) papa, baba
ball (2) b, p (wh) piep (1) pi pi
bimbam bi-ba pretty (0) (rare)
Carolyn (1) g-ga sch-sch (1) -
da (1) da ticktock t-ta
kritze (1) tits (wh) Wauwau (1) wa wa (falsetto)
moo (1) mu:
1;3 23 utterances, 20 words
A-a (1) a-a kitty (2) di di, ti ti (affricated)
baby (1) be bi: kritze (1) tits
ball (2) ba: i, ba (wh) Mama (3) mama, maba, ma
bath (1) ba: Papa (1) baba
Bild (1) bi piep (1) bi bi
bye-bye (1) ba ba (wh) pretty (1) pti
Carolyn (1) dada sch-sch -
da (1) da thank you (1) da da (wh)
ja (1) ia ticktock t
t
i-t
t
a
kiek! (1) ti: Wauwau wu wu wu
Words and sounds in early language acquisition 123
1;4 33 utterances, 26 words
A-a (1)
t
a
t
a kritze (1) tits
baby (1) be bi Mama (3) bama, maba, mama
ball (2) ba: i, ba:i (wh) Marion (1) mm
bath (1) ba: Papa (1) ba ba
bed (1) b peekaboo (1) bi
Bild (1) bi: piep! (1) pi pi (falsetto)
bye-bye (1) babai (wh/vd) pretty (3) pwiti, pti, pyiti (wh)
Carolyn (1) dada sch-sch -
da (1) da thank you (1) dada, dadai (wh)
da ist es (2) da:i, da:i ticktock (2) ti-ta, tik-tak (wh/vd)
down (1) da: up (1) ap
hot (1) ha (wh) Wauwau (1) wu wu
ja (1) ja (wh) yes (1) j
1;5 46 utterances, 40 words
A-a (1) a a I see you (1) ai i
all (1) a: ja (1) ja ja ja ja
apple (2) apa, aba klingelingling (1) li li li
auto (3) ata, ada, aoda mama (1) mama
baby (1) bebi man (2) m, ma
ball (1) ba Marion (1) meme
bath (1) ba: mehr (1) me:
bed (1) b mitten (1) mi:
bitte (2) bit, biti naughty (1) nana (-like)
bye-bye (1) bai bai Papa (1) baba
brush (1) b night (1) a a
Carolyn (1) dada piep! (1) pi pi
da (1) da pretty (0) (rare)
da ist es (1) da: i: Rita (1) wi wi
down (2) da:, da: o sch-sch -
heiss (1) hai (wh) thank you (2) da dai, dada
hello (1) l there (1) d
highchair (1) aita ticktock (1) ti ta
hot (1) ha up (1) ap
I (1) ai Wauwau (1) wu wu
Appendix 2: Phone classes for T, K, and H
Phone classes for T
Session I
h ~ hi, see
d ~ d daddy, dog
Session II
h ~ ~ hi
d ~ d ~ t daddy, dog, thank you
b ~ d baby
124 Charles A. Ferguson and Carol B. Farwell
k
h
~ k
x
thank you
l ~ lj ~ h asking word
Session III
h ~ hi
d ~ dj daddy, dog, thank you, baby
m ~ mn mama
g shoe
t
h
tea
b bye bye
Session IV
b ~ p
h
~ p baby, ball
d ~ t
h
thank you, daddy, bye-bye
h ~ hi
w where
~ ~ t
h
shoe, see
g duck
Session V
b ball, baby
t
s ~ s ~ see, sit
~ ~ d shoe
w ~ b rock
t
h
tea, cracker
d daddy, thank you
h ~ ~ hi, eye
Session VI
w rockrock
b ~ ~ p
h
~ bw ~ ~ baby, bounce, bye-bye, paper, blanket, ball, book
p
h
please, purse, pretty, pat
d dog
t
h
tea, thank you
j yeah
h hi
n no, night-night
~ ~ ~ t
h
cereal, cheese, shoe
~ up, allgone, ice
Session VII
w rock, where
b ~ p
h
~ w ~ ~ bu baby, ball, box, bye-bye, book, bang,
blanket, bounce
p
h
~ p pursey, pat, paper
k
h
~ k ~ k
w
~ kx ~ x ~ g key
d ~ t
h
~ d dog, daddy
t
h
~ t ~ d tea, thank you, tiger, chair
~ t ~ ~ j ~ t
h
~ s ~ h ~ ~ t ~ dj shoe, sit, cheese
m ~ b milk
Words and sounds in early language acquisition 125
n no
~ eyes, ice, allgone, up
Session VIII
t
h
~ d ~ d thank you, daddy
t
h
~ t two, tea, tiger
d dog
k
h
~ t
h
key, cat
w rock, walk
b ~ ball, book, bye-bye, baby
p
h
pat, purse, paper
n no
m ~ b mama, milk
~ see, shoes, sit
~ dj ~ d ~ t ~ d ~ cheese, juice
f ower
~ eyes, ice, out
Session IX
p
h
~ p paper, purse, pot-pot, pretty
f ~ p
h
feet
b
x
bo, bye-bye, baby, book, rock, ball, blanket
k
h
Carol, okay
g dog
h hi
n no
b ~ milk
t
h
thank you, two, cup
d ~ d tiger, daddy
w where, one
t cheese
d ~ dj juice
~ allgone, eyes
~ ~ ~ dj s shoe, cereal, sit
Phone classes for K
Session I
p
h
pop
b book, bear
m ~
b
m monkey, my
d ~ d ~ t thank you, duck, daddy, pointing word
d ~ t
h
~ s ~ z ~ see
h hi
all gone
Session II
b block
d dog
h ~ d hello
126 Charles A. Ferguson and Carol B. Farwell
Session III
d ~ (g) ~
d duck, down, daddy
g shoe
~ on, up/off, allgone
Session IV
b boom
m me/my
d ~ dog, daddy
n ~ no
s Kimberly
k thank you
Session V
m ~ n Nona, Max
d dog, duck, there
t
h
turkey
Satchiko
g car, gum
h hot, hot tea
Session VI
b balloon, box
d ~ that
h hot
Session VII
b boy, moo
t
h
~ t telephone
l lady
g cow
thank you
Session VIII
b watch
m night-night
off, allgone
Session IX
b ~ p bird, bear, bee, pumpkin
t
h
tea, turn off
n no
~ k sh
k
h
cookie, coat, girl
allgone
Session X
b ~ p put, baby, boot, bottle
m moo, Max, me
Words and sounds in early language acquisition 127
d that
see
~ all done, on, up, off
Session XI
b book, balloon
p ~ p
h
bow, pencil
f recorder
d driver
g dog, girl, sock
k
h
kitten
Session XII
b ~ p baby, puppy, bye-bye, break, mama
m moo, mouth
d ~ d there, that
t
h
tie
n no, nose
h ~ shoe
g go
k
h
key
h ~ ~ ears, eyes, itsa
f ~ woof-woof
Phone classes for H
0;10 p pretty
d there
0;11 p pretty
t ticktock
d there
1;0 p Opa, Papa, piep, pretty
b ball, Blumen
t ticktock
d da
sch-sch
1;1 p Opa, piep, pretty
p ~ b Papa
b ball, bimbam, pieks!
w wauwau
t ticktock, kiek!, kritze
d da, Gertrude, Tante
sch-sch
1;2 p piep, pretty
p ~ b ball, Papa
b baby, bimbam
m moo
w wauwau
t ticktock, kritze
d da
sch-sch
128 Charles A. Ferguson and Carol B. Farwell
g Carolyn
1;3 p pretty
b baby, ball, bath, Bild, bye-bye, Papa, piep
m mama
w wauwau
t ticktock, kritze, kiek
t ~ d kitty
d Carolyn, da, thank you
j ja
sch-sch
A-a
1;4 p pretty, piep
b baby, ball, bath, bed, Bild, bye-bye, Papa, peekaboo
m ~ b mama, Marion
w wauwau
t kritze, ticktock
d Carolyn, da, da ist es, down, thank you
j yes, ja
sch-sch
h hot
A-a, up
1;5 p piep, pretty
b baby, ball, bath, bed, bitte, brush, bye-bye, papa
m mam, man, Marion, mehr, mitten
w wau wau, Rita
t ticktock
d Carolyn, da, da ist es, down, thank you, there
n ~ naughty, night-night
l klingelingeling
j ja
sch-sch
highchair
h hot, heiss
A-a, all, apple, auto, hello, I, I see you, up
Appendix 3: Word Index for T, K, and H
A-a H-1; 3, 1;4, 1;5
all H-1;5
all gone T-VI, VII, VIII, IX; K-I, III, VIII, IX, X
apple H-1;5
auto H-1;5
baby T-II, III, IV, V, VI, VII, VIII, IX; K-X, XII; H-1;2, 1;3, 1;4, 1;5
ball T-IV, V, VI, VII, VIII, IX; H-1;0, 1;1, 1;2, 1;3, 1;4, 1;5
balloon K-VI, XI
bang T-VII
bath H-1;3, 1;4, 1;5
bear K-I, IX
bed H-1; 4, 1;5
bee K-IX
Bild H-1;3, 1;4
bimbam H-1;1, 1;2
Words and sounds in early language acquisition 129
bird K-IX
bitte H-1;5
blanket T-VI, VII, IX
block K-II
Blumen H-1;0
book T-VI, VII, VIII, IX, K-I, XI
boom K-IV
boot K-X
bottle K-X
bounce T-VI, VII
bow K-XI
box T-VII, IX; K-VI
boy K-VII
break K-XII
brush H-1;5
bye-bye T-IV, VI, VII, VIII, IX; K-XII; H-1;3, 1;4, 1;5
car K-V
Carol(yn) T-IX; H-1;2, 1;3, 1;4, 1;5
cat T-VIII
cereal T-VI, IX
chair T-VII
cheese T-VI, VII, VIII, IX
cookie K-IX
coat K-IX
cow K-VII
cracker T-V
cup T-IX
da H-1;0, 1;1, 1;2, 1;3, 1;4, 1;5
da ist es H-1;4, 1;5
daddy T-I, III, IV, V, VII, VIII, IX; K-I, III, IV
dog T-I, II, III, VI, VII, VIII, IX; K-II, IV, V, XI
down K-III; H-1;4, 1;5
driver K-XI
duck T-IC; K-I, III, V
ears K-XII
eyes T-V, VII, VIII, IX; K-XII
feet T-IX
sh K-IX
ower T-VIII
Gertrude H-1;1
girl K-IX, XI
go K-XII
gum K-V
heiss H-1;5
hello K-II; H-1;5
hi T-I, II, III, IV, V, VI, IX; K-I
high chair H-I;5
hot K-V, VI; H-I;4, 1;5
hot tea K-V
I H-1;5
ice T-VI, VII, VIII
I see you H-I;5
130 Charles A. Ferguson and Carol B. Farwell
ja H-1;3, 1;4, 1;5
juice T-VIII, IX
key T-VII, VIII; K-XII
kiek! H-1;1, 1;3
Kimberley K-IV
kitten, kitty K-XI; H-1;3
klingelingeling H-1;5
kritze H-1;1, 1;2, 1;3, 1;4
lady K-VII
mama T-III, VIII; K-XII; H-1;3, 1;4, 1;5
Mann H-1;5
Marion H-1;4, 1;5
Max K-V, X
mehr H-1;5
milk T-VII, VIII, IX
mine (me) K-I, IV
mitten H-1;5
monkey K-I
moo K-VII, X, XII; H-1;2
mouth K-XII
naughty H-1;5
night-night T-VI; K-VIII; H-1;5
no T-VI, VII, VIII, IX; K-IV, IX, XII
Nona K-V
nose K-XII
off K-VIII
okay T-IX
on K-III, X
one T-IX
Opa H-1;0, 1;1
out T-VIII
Papa H-1;0, 1;1, 1;2, 1;3, 1;4, 1;5
paper T-VI, VII, VIII, IX
pat T-VI, VII, VIII
peekaboo H-l;4
pen K-IV
pencil K-XI
pieks! H-l;l
piep H-1;0, 1;1, 1;2, 1;3, 1;4, 1;5
please T-VI
pop K-I
pot-pot T-IX
pretty T-VI, IX; H-0;10, 0;11, 1;0, 1;1, 1;2, 1;3, 1;4, l;5
pumpkin K-IX
puppy K-XII
purse T-VI, VII, VIII, IX
put K-X
recorder K-XI
Rita H-l;5
rock (rock) T-V, VI, VII, VIII, IX
Satchiko K-V
Words and sounds in early language acquisition 131
see T-I, IV, V, VIII; K-I
sch-sch H-1;0, 1;1, 1;2, 1;3, 1;4, 1;5
shoe T-III, IV, V, VI, VII, VIII, IX; K-III, XII
sit T-V, VII, VIII, IX
sock K-Xl
Tante H-l;l
tea T-III, V, VI, VII, VIII; K-IX
telephone K-VII
thank you (danke) T-II, III, IV, V, VI, VII, VIII, IX; K-I, IV,
VII; H-1;3, 1;4, 1;5
that K-V, VI, XII
there K-V, VI, XII; H-0;10, 0;11, 1;5
ticktock H-0;11, 1;0, 1;1, 1;2, 1;3,1;4, 1;5
tie K-XII
tiger T-VII,VIII, IX
turkey K-V
turn (off) K-IX
two T-VIII, IX
up T-VI, VII; K-III, X (?off); H-l ;4, 1 ;5
walk T-VIII
watch K-VIII
wauwau H-1;1, 1;2, 1;3, 1;4, 1;5
where T-IV, VII, IX
woof-woof K-XII
yeah T-VI
yes H-l;4
132 Charles A. Ferguson and Carol B. Farwell
5 Developmental reorganization of phonology:
a hierarchy of basic units of acquisition
Marlys A. Macken
1. Introduction
This chapter describes the acquisition of the consonant system by one child
acquiring Mexican Spanish as her native language. During the earliest stages
(from 1;7 to 2;1 years of age), the data from this child referred to here as Si
showed several phenomena that could best be accounted for by assuming a
central role for the word as the basic unit being acquired. Words were, for Si,
prosodic units, each being selected for a particular output form on the basis of
the component consonants and each processed in exible ways to achieve
preferred output patterns. The evidence for the centrality of words and word
pattern in Sis early development will be the major focus of this chapter.
During the later stages (from 2;2 to 2;5), most of the evidence for words and
word patterns has disappeared, and Sis phonological system during this period
can be described adequately in terms of phonemic contrasts and the more
traditional phonological rules.
Thus, the picture of phonology acquisition that emerges from these data is
one in which there are at least two and possibly three basic units the word,
the phoneme, and possibly the feature that gure signicantly in the
developmental process. It appears, moreover, that the word is more important in
the earliest stages, and that in the later stages, the phoneme replaces the word as
the basic structural unit of the phonological system. In the nal section of the
chapter, the relevance of these data for some aspects of a general model of
phonology acquisition is discussed.
2. Data collection and analysis
Si participated in a longitudinal study designed to investigate the acquisition of
consonants in monolingual Mexican-Spanish-speaking children. Approximately
This research is part of the activities of the Stanford Child Phonology Project and was supported by
a National Science Foundation Grant (#SOC 740316 AOl). I would like to thank: Charles
A. Ferguson for support during all phases of the research; Debby Ohsiek, Maria Rodriguez, and
Ana Ortiz for assistance with the data collection; and Carol Stoel-Gammon and Lise Menn for
helpful comments on a draft of the paper. A preliminary version of this paper appears in the
Stanford Papers and Reports on Child Language Development (December 1977), 14, l36.
133
once a week for a ten-month period, she was recorded for fteen to thirty minutes,
while interacting with the experimenter (E), a native speaker of Mexican Spanish.
Recording was done on a Uher 4000 tape recorder with a Sony Electret micro-
phone (attached to a soft cloth vest which the child wore). Stimulus materials
included picture books, index cards picturing common objects, and small toy
objects. When not being recorded, she remained in a playroom. Notes were taken
in the playroomby a research assistant. These notes, which included words which
the child used, were used subsequently by the Eduring recording sessions in order
to elicit as many of the childs vocabulary items as possible. Word lists were
collected from the parents every two weeks and were also used by the E during
recording sessions.
During the week following each session, transcriptions were made of all
tapes by two transcribers working independently and using Revox A77 tape
recorders with Super St-Pro B-V headphones. Two procedures were used to
combine the individual transcriptions of each utterance into a nal one (Macken
1978). The transcription system used is that of the International Phonetic
Association, with a supplemental symbology developed by the Stanford Child
Phonetics Workshop (Bush et al. 1973). Consonants were transcribed narrowly
and vowels somewhat more broadly. In this chapter, Sis phonetic segments and
phonetic sequences will be given in square brackets, [. . .], the typical notation
for such segments in descriptions of adult speech. Her phonemes and phonemic
sequences will be enclosed however by vertical straight lines, | . . . | a nota-
tional device used by Smith (1973). The more usual slant lines, /. . ./, will be
reserved for phonemes or phonemic sequences of adult Spanish only. Although
Sis data were transcribed quite narrowly, examples of phonetic sequences when
used in the following text will be given in no narrower a transcription than
necessary.
3. The subject
Si was 1;7 at the beginning of the study and 2;5 at the end. She is the youngest of
seven children; she has three brothers (ages 17, 15, and 6 years at the beginning
of the study) and three sisters (13, 11, and 8). She was born in Redwood City,
California, to parents who had moved there from Michoacn, and the family
speaks only Spanish in the home. However, both parents work and speak
English at their places of employment, and all of her siblings attend English-
speaking schools; thus, Si may have been exposed to some English. During the
rst several months of the study, Si was cared for by a monolingual Spanish-
speaking neighbor. She subsequently was enrolled in a day-care program for
Spanish-speaking children which was staffed by native speakers of Spanish;
nearly all the children enrolled were monolingual Spanish-speaking children.
During the period (prior to and) from1;7 to 2;2.15, Sis environment was almost
134 Marlys A. Macken
exclusively Spanish-speaking, and her exposure to English is presumed to have
been minimal.
When Si was 2;2.15, she was transferred to another day-care program, in
which there were many monolingual English-speaking children. The programs
staff reported to us that Si began learning English very quickly. During the
period from 2;2.15 to 2;5, she used several English words during our sessions:
apple, phone, shoe(s), car, puppy, and mommy home were used frequently; hi
and monkey were used once; and donkey, sun, and berry are possible English
glosses for utterances produced in ambiguous contexts.
During her thirty-ve sessions over the ten months, Si spontaneously pro-
duced nearly 200 recognizable words. The corpus of spontaneously produced
speech contains 2,536 tokens and accounts for 51 percent of her total recogniz-
able speech. In imitation, she produced 2,463 tokens of both the same words and
many other words that she never produced spontaneously. From the very
beginning, Si imitated more than any other subject in the study and more than
any of the children reported in Bloom, Hood, and Lightbown (1975). During a
preliminary analysis, it became clear that the relationship between spontaneous
and imitated forms was a complicated issue: imitations were neither always in
advance of spontaneous forms nor necessarily predictive of spontaneous devel-
opment. These two types of productions were analyzed separately. In this
chapter, only the analysis of spontaneously produced speech will be reported.
There were several aspects of Sis speech that were particularly striking and
are relevant to her phonological development: her use of a pre-utterance vowel;
the diminished phonetic accuracy that occurred when she produced words in
phrases; her use of consolidated unit phrases and routines; and her mis-
perceptions. The rst three characteristics of her speech, together with several
other factors, contributed to the enormous amount of phonetic variation which
was the hallmark of her productions, which set her distinctly apart from the
other subjects, and which presented the rst obstacle to the phonological
analysis. All four aspects of her production are relevant to general features of
her phonology and, in particular, to the role that the word as a phonological
unit played in her development.
From1;7 to 2;5, 19 percent of all spontaneous utterances and 20 percent of all
imitated utterances were produced with a prexed ller that was almost
always a neutralized vowel, but that sometimes was a syllabic consonant or a
syllable of the form CV; in rare cases, she added a ller (either a vowel or a
syllable) to the ends of utterances. This pre-utterance vowel occurred freely,
and no restriction of its occurrence to particular words, word classes or phonetic
environments could be discerned. Although there were several changes in the
frequency with which this vocalic segment occurred, the uniformity of its
appearance over the ten-month period indicates that it was a general character-
istic of her speech, perhaps an initiation-of-speech phenomenon.
Obviously, the pre-utterance vowels presented problems for analysis.
Particular occurrences of it could variously be interpreted as an article, a verb
Developmental reorganization of phonology 135
form, the unstressed initial syllable of a particular word, etc. In any particular
case, any of these interpretations would have consequences for the phonological
analysis. Since the interpretation of these vocalic segments was problematic, it
was decided to treat all occurrences as a single phenomenon. Thus, they were
not counted as independent words in the session tabulations for words and
tokens, and they were ignored in the phonological analysis. A possible conse-
quence of this decision is the underestimation of Sis abilities. For example, she
was not credited with the acquisition of some articles and the verb forms es and
est until late in the study when clear evidence obtained. Similarly, she was not
credited with the ability to produce three-syllable words until three-syllable
words of the form [CVCVCV] were produced; words with an initial unstressed
vowel syllable (like araa spider) remained difcult to interpret. If in fact the
pre-utterance vowel represents an attempt on Sis part to designate semantic
properties such as denite/indenite reference, then it is the case that this
attempt preceded any other evidence of such semantic knowledge by many
months. Alternatively, its use could merely represent an attempt to replicate the
length and form (without content) of adult utterances. In any event, the use of
these vocalic segments is consistent with her general tendency to embed a word
in a longer carrier. In fact during the earliest sessions, Si frequently produced
long, basically unanalyzable utterances in which only one recognizable word
occurred.
When Si began to combine two recognizable words, the phonetic accuracy of
both words decreased. In many of these two-word sentences, she consolidated
the two (polysyllabic) source words into one, smaller (e.g., two-syllable) form.
In the resulting two-syllable sentence, only one syllable of each word was
produced; in contrast, when these words were produced in isolation, they were
rarely, if ever, produced as a single syllable only. Possibly similar to this was the
way in which she consolidated learned routines: for example, qu es? what
is? |kes|.
1
It seems that Sis willingness to use large units (e.g., long words,
sentences, and routines) while at the same time having only limited production
abilities (and/or limited semantic knowledge) led to the phenomenon of such
coalesced or consolidated units and sentences. This phenomenon is directly
related to the most striking feature of her phonology her use of coalesced
word patterns. For the phonological analysis of such routines, the surface form
of the unit phrase was treated as being the phonological form also (since the
component words did not occur separately). The phonological structure of
individually occurring words was determined on the basis of their most fre-
quently produced form in isolation.
One way of looking at Sis consolidation of routines and phrases would be
that these routines were in fact units for her, stored and used as wholes. This
interpretation would also be relevant for an analysis of her misperceptions,
imitations in which she misperceived the adult model and produced a different,
but phonetically similar word. In this study, the frequency of misperceptions
was characteristic of only Si among the six subjects who participated. For
136 Marlys A. Macken
example, the Es Armando was repeated by Si as Fernando, len lion as avin
airplane, gallo rooster as caballo horse, limn lemon as jamn ham, and
taza cup as casa house. In the misperceptions, the two words involved are
phonetically similar but rarely visually or semantically similar; they would be
called slips of the ear. Since the phonetic similarity between the words
involved tended not to consist of single segment or feature changes but rather
was based on some holistic similarity between the words, the misperceptions are
in this way similar to the unit routines. For the adult, words have a specic
segmental composition, but for Si words seemed to have in addition (and in
some cases, only) a general prosodic shape. As will be discussed, the phe-
nomena of unit phrases and misperceptions are similar in important ways to
aspects of her phonological development in that all three are suggestive of a
system in which the word is a unitary whole which has a general prosodic shape
(perhaps as yet unanalyzed segmentally) and which can be prosodically similar
in a loose way to other words.
4. Sis phonology at the beginning of the study (1;7)
Si is learning the phonological system of the variety of Spanish spoken by her
family and neighbors, almost all of whom come from the state of Michoacn in
Mexico. Since the Project did not analyze the phonology of the subjects
parents, and there are no phonological analyses of Michoacn Spanish, the
adult phonology to which Si is progressing is by necessity presumed to be that
of general Mexican Spanish, a variety that has received considerable attention
from linguists. Given the basic sound classes stops, fricatives, and sonorants,
the following divisions can be made. The feature of voicing is distinctive in the
stops /ptk/ versus /bdg/ and in the fricatives as /fsx/ versus []. However, the
voiced fricatives are allophones of the voiced stops /bdg/: [bdg] occur in
utterance-initial position and after certain sonorants; [] occur in all other
positions. Depending on the reference being cited, /t/ is classied with the
voiceless stops (Alarcos Llorach 1950), with the voiceless fricatives (Stockwell
and Bowen 1965) or as the single member of a voiceless affricative class
(Dalbour 1969). In this chapter, /t/ will be referred to as a member of the
voiceless fricative class. The greatest disagreement concerns the classication
of /w/ and /j/: (1) as semivowels (Stockwell and Bowen 1965); (2) as fricatives
(Dalbour 1969); or (3) in separate classes, the /j/ with /bdg/ as voiced or lax
sounds, and [w] as a phonetic nonsyllabic variant of /u/ or /gu/ (Alarcos
Llorach1950). The two sounds /w/ and /j/ will be referred to as glide consonants
in the present chapter. The class of sonorants is divided into the nasals /mn/ and
the liquids /l/. The symbol // represents the apico-alveolar single ap r, and //
stands for the apical trill r phoneme. All eighteen consonant phonemes may
occur in intervocalic position, and all but two (/ / and //) may appear in initial
position; in nal position, however, only /nsrld/ may occur. The symbol /r/ for
Developmental reorganization of phonology 137
nal position represents a neutralization of the two r phonemes, and nal /n/
represents a neutralization of all nasal phonemes.
Si was 1;7 at the beginning of the study. During the rst month, her sponta-
neous production was limited to a very small set of words (N = 12). The set of
consonantal phonemes in her speech was correspondingly small: the voiceless
stops [p, t], the nasals [m, n], and the glides [w, j]. Sis phonemes corresponded
to the appropriate adult phonemes, with the following additional relationships:
/b/ was realized as /p/ or /w/; /k/ as |t|; /g/ as |w|; /, n/ as |j|; and // as |n| (in one
word, as a result of nasal assimilation). Phonetically, Sis nasals occasionally
were de-nasalized and the voiced bilabial stop was sometimes weakened to a
glide; a lenis articulation was common to much of Sis production throughout
the period studied. By far the greatest phonetic variation was seen in produc-
tions of mira look (e.g., [ja], [hi ja], [i ja], [mi ja], [bi j], [mi a] or [ j]).
A limited segmental system is typical of the phonology of a young child, and
the correspondence between the childs system and that of the adult is typically
captured by a set of substitution rules similar to those presented above. In
addition, childrens productions must be further described by the set of con-
straints that determine the ways in which consonants can be combined in words
of two or more syllables and the constraints that determine the number of
syllables that can be produced in any given word. The restriction of words to
one or two syllables is probably universal during early acquisition. Consonant
harmony a constraint that stipulates that if two consonants appear in a word,
they must be the same or highly similar is widespread (Vihman 1978) and is
frequently identied as a universal (Smith 1973). Although consonant harmony
may be either complete (i.e., involving both place and manner modications) or
partial (i.e., either place or manner), harmony involving all the features of a
segment as opposed to only one or two . . . is . . . characteristic of very early
speech (Smith 1973).
In Sis productions, all words were either one or two syllables long, as
expected. With regard to consonant harmony, the situation was more compli-
cated. Several words exhibited harmony but were productions of adult Spanish
words in which both consonants already agreed either partially or completely: in
beb baby, pap father, and guau guau bow wow, both syllable initial
consonants were of the same place of articulation and a highly similar manner.
One word rana frog was a clear case of complete harmonization ([na na]).
Dame give me was also (partially) harmonized ([pa me]): however, in all
subsequent productions of this word, the initial consonant was deleted.
In contrast, manzana apple was not produced with both consonants agree-
ing in place feature: manzana |ma na|. The completely harmonized produc-
tion which could have been expected is [na na], which was the production used
by all the other children. In the output formof [nana], a weak-syllable deletion
rule could be posited to explain the reduction of this word from three to two
syllables. Such a rule is quite common in early child phonology (Ingram
1974b), and a similar rule (initial syllable(s) deletion) was characteristic of
138 Marlys A. Macken
another subject in the study. In fact, Si used the initial syllable(s) deletion rule
several months later. In the beginning stages, however, she deleted the medial
syllable of manzana, and in general deleted syllables in a exible manner
consistent with the goal of producing a favored output form. The exible
syllable deletion rule, the absence of complete harmonization, and the use
of a favorite two-syllable canonical form (in which a word-initial labial con-
sonant combined with a medial dental consonant) as seen in the production of
manzana were to become typical features of Sis early phonological system.
Nio boy and nia girl were produced with an initial dental nasal and a
medial glide (i.e., // |j|). Here and during all subsequent stages, glides freely
combined with other consonants in two-syllable words. Evidently the non-
consonantal nature of glides exempted them from the constraints which limited
the co-occurrence of consonants. As will be seen, the liquids also combined
freely with other consonants during the stages in which they were realized as
either glides or liquids.
Thus, Sis phonological system during the month she was 1;7 was limited to
labial and dental stops and nasals and the glides. During the rst session, she
also occasionally babbled long sequences of syllables in which the consonant
segment was either [b], [p] or [w]; such babbling did not occur in later sessions.
However, during all four sessions of this period and during many sessions in the
rst several months, Si produced long utterances containing only one recogniz-
able word. In such sentences, the extra syllables contained the same con-
sonant as occurred in the recognizable word; thus, these sentences were
primarily labial or dental sequences. Si continued with her preference for labial
and dental stops and nasals in selecting favorite words, in her nonsense rhym-
ing, and in her sound play, all of which drew upon this set of consonants. This
set remained at the core of her phonology for several months.
5. Sis early phonological development (1;8 to 2;1)
In the preceding section, syllable deletion and consonant harmony were men-
tioned as major means by which many very young children simplify the
phonological form of adult words. Smith (1973) includes these in his list of
the four functions of child rules: (a) consonant and/or vowel harmony;
(b) consonant cluster reduction; (c) systemic simplication (e.g., the reduction
of adult contrasts); and (d) grammatical simplication (e.g., the absence of nal
s and hence of the singularplural contrast in English). Clearly, constraints on
the length of words, on the complexity of the childs phonological system, and
on the complexity of combinations of sounds in words operate universally to
affect simplication. However, there is ample evidence in recently published
papers that demonstrates that individual children may differ in the strategies
they adopt to achieve such simplication a fact recognized explicitly by Smith
in a later paper (Smith 1975): whereas the tendencies or strategies themselves
are universal, the rules which implement them . . . are child specic.
Developmental reorganization of phonology 139
It is also true that in spite of the early and strong necessity to simplify the
adult phonology, children must ultimately learn the entire set of phonological
units which are contrastive in the language being learned. The phoneme is one
such unit that is traditionally recognized in phonological theory and is used
frequently as the basic unit in studies of child phonology (Smith 1973).
However, recent studies (Ferguson and Farwell 1975; Menn 1971, 1977; see
also Ferguson 1977a) have demonstrated that a more appropriate unit of
analysis for the corpora from very young children is the word.
2
In these
studies, phonological rules are not realization rules deriving a childs surface
formfroman underlying adult phoneme (as in the Smith framework) but, rather,
are formalizations of the strategies that a particular child has adopted to
represent words and classes of phonetically similar words (see Menn 1976).
Sis early development can best be accounted for within a framework that
recognizes the signicance of early words and word shapes in the development
of the young childs phonology and the variability with which individual
children implement the simplication processes (Ferguson and Farwell 1975;
Farwell 1976).
It will be seen that the use of the word (and word patterns) rather than the
adult phoneme (and phonemic contrasts) as the basic organizing unit of Sis
early phonology better explains the variation in words over time, the develop-
ment of canonical forms, the variable correspondence between adult phonemes
and Sis phones, and several additional phenomena that would be largely
inexplicable within a framework like Smiths which maps adultlike underlying
representations onto the childs surface forms. By the end of this period,
however, much of the evidence for a word-based phonology has disappeared,
and Sis productions during the period 2;2 to 2;5 can more easily be described in
terms of phonemic contrasts and related phonological rules. During the period
up to 2;2, Si was also learning contrasts between individual sounds and the
equivalences between similar sounds in different environments (i.e., phonemic
contrasts and allophonic relationships) and in fact during the period 2;2 to 2;5,
her phonetic realizations of adult Spanish phonemes in different positions,
environments, and words were much more regular Although the framework
of Section 6 which covers the period 2;2 to 2;5 will accurately reect Sis
transition from words to phonemes, this change is most obvious in the compar-
ison of the end state (2;5) with the beginning one (1;71;8); how the transition
precisely came about is not nearly so clear, primarily because the two develop-
ments overlapped considerably.
5.1. The learning of word patterns
Table 5.1 presents the development of word patterns in two- and three-syllable
words of the form #(C)(V)CVCV#. These word patterns capture the ways in
which constraints on the co-occurrence of consonants in words were gradually
relaxed; for this reason, word patterns of the form #(V)CV# (i.e., words with
140 Marlys A. Macken
Table 5.1. Development of word patterns in two- and three-syllable words
C
1
C
2
C
1
C
2
[+ front] [+ back] [+ back] [+ front]
C
1
C
2
C
1
C
2
C
1
C
2
Stage [ place] [ place] [+ labial] [+ dental] Other [+ velar] [+ dental] Other Age
I p/b__p/b__
n__n__ p__m__ m__n__ 1;7
II m__m__ p/b__t/d__ b__k__ (t__p__t__) 1;8.7
III k__k__ p/b__n__ 1;9
t__t__ p/b__nt__
IV t__t__ t__n__ m__s__ n____ 1;10.7
t__t__ f__n__
V t__l__ 1;11.7
t__nt__ p__l__ b__ __ k__t/d__
VI t__t__t__ p__n__ __n/t____ 2;0.7
p__(n)t__(n)
f__nt__
f__t__n__
p__s__ k__s__
VII (A) n__l__ p__n k__l__ n__f__ 2:1
n__t__ f__ t__ k__m__
s__n__ m__l__ s__p__t__
t__n P__t__n__ d__k__
Table 5.1. (cont.)
C
1
C
2
C
1
C
2
[+ front] [+ back] [+ back] [+ front]
C
1
C
2
C
1
C
2
C
1
C
2
Stage [ place] [ place] [+ labial] [+ dental] Other [+ velar] [+ dental] Other Age
(B) l__l__ b__mb__ b b s
f n s
j b s
m s
g
s#
b j n
s__p
l__p__s
2:2
Notes:
1. All word patterns in the table correspond to two syllable productions except the following: t__p__t and s__p__t__ for zapato (stages II and VIIA); t__t__t__ for
chachita (stage VI); f__t__n__ for telfono (stage VI), and p__t__n__ for manzana (stage VIIA). Only one other three-syllable production occurred during this period:
k__w__j__ for caballo (stage VIIA).
2. = only one occurrence of one word.
3. ( ) = an optionally occurring consonant
4. / = either consonant may occur.
5. L = cover symbol for a liquid which had various phonetic forms.
only one consonant) are not included. The set of consonants included in this
table includes stops, nasals, fricatives, and liquids; from the beginning stage
(1;7), the glides could freely combine with other consonants and hence have
been omitted from the table.
In the rst column are the word shapes in which both consonants agreed in
place. Although in adult Spanish /t/ is a palatal affricate and /tdns/ are referred
to as dental consonants, they are here considered to agree in place because
Sis productions of these sounds did not conform to the adult contrast: she
usually produced all ve consonants as [+ alveolar], but her phonetic range
covered the entire dimension of dental to palatal. These sounds will be referred
to as the class of dental consonants, following the typical nomenclature for the
adult phonemes /tdns/.
The most interesting developments in word patterns are seen in columns II
and III. Here, it is clear that Si preferred the order of consonants in a word to be
[+ front] in initial position and [+ back] in medial position: all nal consonants
were deleted until stage VII. Moreover, the preferred initial consonant was a
labial one and the preferred medial was a dental. This preferred front + back
ordering accounts for the output form of all words containing a place contrast
from 1;7 to 1;11.15, with the exception of the pattern [t__p__t__] used only for
zapato shoe (stage II). This pattern appears in parentheses in the table, because
it was not a productive word pattern (i.e., it was used for no other words nor for
the generation of additional word patterns). The early syllable structure accu-
racy (stage II) of zapato is unusual for two reasons: (1) it was lost during stage
III, and (2) no other three-syllable productions were regularly produced until
stage VI. Two other words, elefante elephant and manzana apple, were also
produced with a three-syllable form for a brief period before being regularized
to a two-syllable form (see Moskowitz 1971 and 1973 on phonological idioms).
Up to stage V, all initial consonants in column II class words were labial.
During stages V and VI, a pattern emerged which violated the front + back
ordering: the new word pattern contained the other member of the [+ grave]
class (a velar stop) in initial position, with a dental consonant still preferred in
the medial position. That the new pattern was not of the form dental + velar is
signicant and demonstrates that Sis preference was not simply a fronting
strategy (see Ingram 1974a on fronting in child phonology).
The data in columns II and III also show that the process by which Si
expanded her repertoire of word patterns was one in which new patterns were
created out of existing ones. The patterns of [p/b__n__] and [p/b__nt__] of
stage III represent a combination of the patterns [m__n__] and [p/b__t/d__] of
the earlier two stages. In stages IV through VII, the set of possible initial
consonants was expanded to include the remaining labial consonants, while
the set of possible medial consonants was expanded to include several addi-
tional dental consonants. The creation of new word patterns on analogy with
existing ones is seen most clearly in the great expansion during stage VI, when
many new words and new word patterns were acquired. In this stage, Si
Developmental reorganization of phonology 143
expanded her general labial + dental pattern to include nearly all the possi-
bilities of appropriate consonant co-occurrence. She also overgeneralized
[+ velar] in initial position to include a velar nasal, a sound that has no phonemic
status in adult Spanish and occurs in syllable-nal position only (as an allo-
phone of /n/ before velar consonants).
The use of [] only occurred during a two-week period and was restricted to
productions of three words: rana frog [wa a, ga , a na]; bola ball
[a w]; and gato cat [ak to, a ko]. The productions for gato occurred
only in imitations; Si rarely used this word spontaneously. Gato was produced
with either an initial [n] or an initial [k] for a three-week period during the
following stage and subsequently stabilized with an initial [k]. Prior to stage VI,
gato had been produced as [ka ko].
The ways in which Si expanded her repertoire of word patterns by combining
and/or expanding existing patterns point to the signicant roles that overgener-
alization and analogy play in the acquisition of phonology (as they do in the
acquisition of syntax and semantics). Sis phonemicization of the velar nasal
strikingly demonstrates these processes and shows that it is not always the case
that the childs phonemes correspond directly to the representations and feature
assignments of adult phonology; more importantly, it points to the creativity
exhibited by the child in his/her role as the active organizer of phonology. This
latter fact has only recently begun to be recognized (see Moskowitz 1971); most
previous discussions of phonological acquisition assumed the role of the child
to be passive. Recent work by Kiparsky and Menn (1977) assigns a signicant
role to the childs creative role and characterizes phonological acquisition as
inherently a cognitive (i.e., problem-solving) task. Clearly this approach ts Sis
development well.
Words during these stages fell into several types of patterns and Table 5.1
demonstrates the regular nature of word pattern development over time. These
facts in themselves suggest the importance of words and word patterns to Sis
early phonological development. More convincing evidence comes from the
exibility with which several processes operated on individual words to pro-
duce the preferred word patterns, several unusual substitutions and the change
in form of several words which occurred as new word patterns were learned.
Table 5.2 presents the rst two-word patterns learned by Si and shows the
ways in which words were processed to t the output goal. The rst column
gives the words that were selected to t particular patterns. The next four
columns correspond to four processes syllable reduction, metathesis, substi-
tution, and consonant cluster (CC) reduction that Si used to simplify adult
words. The last column gives a phonetic transcription of the most typical output
form; when two forms occurred, both are given.
The word pattern [p/l__c/d__] was much more productive than [m__n__];
however, the same statements can be made for both pattern types, and the
general point of interest here that of the variable nature of the processes
applies to both equally. The process of syllable reduction operated to delete
144 Marlys A. Macken
Table 5.2. Analysis of the rst two word patterns acquired by Si: [m__n__] and [p/b__t/d__]
Processes
Word pattern Stage Age Words Syllable reduction* Metathesis Substitution CC-reduction Sample producton
(A) [m__n__] I 1;7 manzana manzana m w mnna
mano wainno
II 1;8.7 Fernando Fernando
Fernando
f m
m w
nd n
"
mann
wanno
nanno
#
Ramn Ramn m b ~ w ~ m mn
VII 2;l comiendo comiendo ndn minnu
(B) [p/b__t/d__] II 1;8.7 pelota
pata
pato
zapato
pelota
zapato
)
o a
(assimilation)
patda
pa ta
ptda
"
pwatto
bwaddo
#
sopa
Vicki
sopa
kt
pwta
"
wt ^ t
bjk ^ ke
#
III 1;9 elefante elefante f b ~ p nt t batte
[
librito
]
librito br b
}
pitd
libro libro l d br b
IV 1;10.7 vestido vestido i, o i bit-ti
perro i d b d
plato pl p pw t
Table 5.2. (cont.)
Processes
Word pattern Stage Age Words Syllable reduction* Metathesis Substitution CC-reduction Sample producton
VI 2;0.7 bota
pastel
l n ~ st t pattj
pttn
VII 2;1 reloj b
}
buddo
l d
p ~ l
l l }
[
hlllo
]
pllo
Notes:
1. The variability in the voicing of stops is not listed under substitutions; see Section 5.2 for discussion of the acquisition of the voicing contrast.
2. Words are listed under the rst stage in which they appeared.
3.* = Deleted portion italicized.
either the initial syllable(s) or the medial one. The choice of which was deleted
depended crucially on the consonants in the syllables. Sis goal was to achieve
an output form of labial + dental. If the labial consonant occurred in the
correct initial position, then the medial portion of the word was deleted:
manzana apple, Fernando brothers name, pelota ball, and vestido
dress. If, however, the labial consonant was in medial position in the adult
word, the preceding syllables were deleted: Ramn brothers name, comiendo
eating, zapato shoe, elefante elephant, and librito little book.
3
In the cases
in which medial segments were deleted, the situation was more complicated
than just the deletion of syllables. The general phonetic quality of the vowels
that appeared in the output forms suggests that Si tended to retain the vowel of
the stressed (penultimate) syllable; this vowel was then combined with the
word-initial consonant to form the rst syllable of the output form (e.g.,
Fernando and vestido, Table 5.2; see Menn 1974 on the essentially universal
preservation of the stressed vowel in child forms).
In both syllable reduction types, the adult word functioned as a single
prosodic unit, the features of which were changed to t the output goal. All
the adult words that were selected by Si to t her patterns of [m__n__] and
[p/b__t/d__] had a labial and a dental consonant somewhere in the word; this
requirement is crucial for distinguishing between words that underwent meta-
thesis as opposed to harmony. Metathesis an uncommon process in child
phonology occurred in Sis speech but not in the speech of another subject a
subject who used (complete) harmony almost exclusively to simplify adult
words (Macken 1978). Si also used (complete) harmony but considerably less
than would be expected on the basis of its documentation in the literature. The
only words which underwent metathesis were those words that contained labial
and dental consonants in the wrong order (e.g., sopa soup, Table 5.2B; and
telfono telephone, Table 5.3B). In contrast, words that exhibited complete
harmony were words that had an incorrect ordering of consonants, but lacked
one of the pattern-criterial consonants (e.g., gato cat [ka ko] stage III). The two
words in the earliest stage that also exhibited harmony (rana [nana] and
dame [pame]) do not t the rule just stated. However, from stage III on, rana
was produced just as often with an initial bilabial phone as with an initial [n].
During stage I, Sis production for dame changed to [ame]. It seems that these
two words were acquired prior to the point (stage II) at which Si settled on her
[labial + dental] pattern preference and thus were exempt to some extent from
her later rules. Cases of metathesis and (complete) harmony were not common
in Sis data.
Substitution processes also operated in different ways, depending on the
consonant structure of the adult word and the requirements of Sis patterns.
For example, in Fernando (which contains a word-initial labial with a nasal as
the rst consonant of the stressed syllable), /f/ |m|, while in elefante, /f/ |b/
p| (Table 5.2); in perro dog, // |d|, while in reloj watch, //|b|
(Table 5.2). Further examples of substitutions being determined by word
Developmental reorganization of phonology 147
patterns will be seen in Table 5.3. In contrast to the goal-directed and, hence,
variable nature of the processes of syllable reduction, metathesis and substitu-
tion, the fourth process needed to explain surface forms the process of
consonant cluster reduction was very systematic: nasals were deleted when
followed by a voiceless stop; voiced stops were deleted when followed by a
nasal; liquids were deleted in all clusters; and fricatives were deleted when
followed by a stop (Table 5.2).
Table 5.3 contains eleven words and charts their development through stages
I to VII. Most of these words demonstrate unusual correspondences between
adult phonemes and the phones in Sis productions. Several of these unusual
substitutions have already been mentioned. Throughout this period of 1;7 to 2;1,
/t/ |p/b| only in tenedor fork (Table 5.3B), /m/ |p/b| systematically only in
manzana apple (Table 3A) and Ramn brothers name, and // |p/b| in
initial position but |l| in intervocalic position (reloj versus perro, Table 5.3B).
The rst three fricative words that Si acquired were also unusual: /f/ [t] in
Fernando and elefante; and /s/ |f| in manzana. Although the phonetic
realization of /f/ in Fernando and elefante was [t] (phonetically very similar
to the adult phoneme /t/), it is possible that the source of the substitution error
was a confusion of /f/ and /s/. In contrast to these three words, all subsequently
acquired words were realized with correct /f, s, t/ contrasts.
With the exception of the words involved in the /f/ and /s/ reversal, all the
unusual substitutions can be accounted for by the overgeneralization of Sis
preferred word patterns, although this pattern force in itself cannot explain
why these particular words were susceptible and not others (see Labov, Yaeger,
and Steiner 1972 on the riddle of actualization). As will be seen, prosodic
similarity between certain adult words provides a plausible explanation for
the similar treatment of some words.
Table 5.3A shows the developmental changes in six words. At stage III,
manzana changed from |mana| to |p/bana|, due to the overgeneralization of the
[p/b__n__] word pattern; at stage IV, it changed to |fana|. This latter change
exemplies Sis tendency to combine features from different segments of the
adult word: the labiality of the initial /m/ and the frication and voicelessness of
the medial (stressed syllable) /s/. The odd development occurs in elefante,
which had been |pante| during stage IV but changed to |tante| during stage V;
the expected development would have been [fante]. Two possible explanations
can be offered: words as prosodic units; and confusion between adult fricatives.
In the words as prosodic units explanation, two factors may be relevant: (1)
within-word combination of features from different segments; and (2) cross-
word prosodic similarity. (1) First in manzana, the combination of features
[+labial] and [+fricative] resulted in |f|; in elefante, it may be that the change of
/f/ to |t| was a result of the inuence of the [+dental] feature of /l/, although
such an interaction would be anomalous in Sis treatment of liquid words. (2)
Note that in the preceding stage (IV), cuchara spoon was usually produced as
|ta na|; cuchara also contains a dental liquid in the adult form. To Si, perhaps,
148 Marlys A. Macken
Table 5.3. Acquisition of word patterns as shown in developmental stages of selected words
I II III IV V VI VII
[m__n__]
[
b__d__
]
[p/b__n__] [t__t__] [p__l__] [p__n__] [p__n]
p__t__ [p/b__nt__] [t__n._] [b__(n)t__(n)] [f__t__]
[f__n._] [f__nt__]
Word [m__s__] [f__t__n__1]
(A)
manzana m na
[
bwnn
]
[fann]
[
panna
]
[
panna
]
[
tanna
]
pwnn fanna tanna ptanna
manna anna
elefante
[
ba te
]
pwanti
[
bwante
]
[
bante
]
[
bante
]
pantl tante tante fante
fante fadti
cuchara
[
fanna
]
!
[
fann
]
fadd a la
Fernando
[
manno
]
[
ma no
] [
mn tu
(1)
]
tinal to
(1)
wanno janto meando
nanno
telfono
[
fn tonno
]
fwe fnno
fwa tin, nu
(B) rana na na
[
wanna
]
[
nann
]
[
manna
]
nna manna znn
wann
wanna
[
gan k
]
anna
na k
Table 5.3. (cont.)
I II III IV V VI VII
[m__n__]
[
b__d__
]
[p/b__n__] [t__t__] [p__l__] [p__n__] [p__n]
p__t__ [p/b__nt__] [t__n._] [b__(n)t__(n)] [f__t__]
[f__n._] [f__nt__]
Word [m__s__] [f__t__n__1]
reloj wi jo
(1)
nen no
(1)
ha do
[
bddo
]
winno
hlllo
?jjo
ne o
pllo
(I)
ratn (p ln) tn
(I)
pntn pdn
tenedor b
dd
(I)
pnnn pnt:
pastel
[
pt t
] [
pi t
e
]
pt tn pi t
perro (p lo) b d p o pe do
pe lo p d do
Notes:
1. Variable forms are listed in order of frequency.
2. [ ] = variable forms, given when a stage was characterized by signicant variation.
3. ( ) = questionable referent.
4. = word form remains unchanged throughout period indicated by arrow.
5. Subscript (I) following a form indicates that it occurred in imitation only; imitated forms are given only if they differed signicantly from the spontanteous form.
6. Two sets of variable forms are given for rana; those contained in the second group occurred only during the last two weeks of stage VI.
the prosodic similarity of cuchara and elefante was greater than that between
manzana and elefante, which would account for the same pattern having been
adopted for elefante. Similarly, Fernando was either [ma no] or [na no] during
stage II. Here two processes were in competition. In [ma no], the preference for
a labial consonant in initial position resulted in the feature combination of
labiality of initial /f/ and the nasality of the (stressed syllable) /n/ to produce
m. In [na no], the initial unstressed syllable was deleted. In both the cluster was
reduced. During stage VI, Fernando was added to the prosodic class to which
elefante and cuchara belonged, a change which resulted in the newform[ti nal
to]. Throughout this period, Si regularly produced an initial t in imitations of
this word. It was common throughout this period for old words to change as new
word patterns were acquired a phenomenon which presumably was due to
changes in Sis hypotheses about which words and sounds were similar and
should be said in similar ways.
The alternative explanation would be that Si simply confused /f, s. t/.
However, the production of |f| for /s/ in manzana is consistent with Sis general
tendency to combine features from segments of the adult word, and the pro-
duction of |t| for /f/ in elefante and Fernando could be a further example of Sis
general tendency to overgeneralize word patterns to new instances of words that
were prosodically similar in some way. If Si confused these fricatives, she
apparently did so only in these three words: all subsequently acquired words
containing /f, s, t/ were produced correctly, and some of these words were
acquired during the same time period that the three words showed the reversal.
Neither of these explanations words as prosodic units or fricative con-
fusion is necessarily correct, or they both might be. The word as prosodic
unit though does seem to be a factor in Sis treatment of tenedor fork. In
Table 5.3B, rana frog, reloj watch, and ratn mouse show how Si selected
initial // words (which also had a medial dental consonant) to be members of the
[p/b t/d ] word-pattern class. A possible explanation for why tenedor was also
selected to be a member of this class is that the nal /r/ (of the stressed syllable)
made this word prosodically similar to the class of initial // words, in which case
the labiality of the /r/ (phonemically |b| in initial position for Si) combined with the
voicelessness of the initial /t/ to result in the output [p].
The force exerted by preferred word patterns provides a plausible explanation
for the otherwise inexplicable treatment of many adult words and phonemes.
However, the same pattern force which frequently caused words to change
phonological form also caused variation which was not so easily interpreted.
The variation seen in Table 5.3B for rana was typical of many words and also
points to the problem of knowing whether an output change was due to the
phonological reorganization of a word (as in the overgeneralization of a newly
acquired pattern seen in manzana, stage III, Table 5.3A) or whether it was due to
a phonetic slip-of-the-tongue. In Fromkins theory (1973), slips-of-the-
tongue are not random errors made in speech production but are rule-governed
errors that systematically reveal the nature of the rules of the grammar.
Developmental reorganization of phonology 151
Although Fromkins theory of speech errors has not been applied to child
phonology, it may provide an explanation for some of the variation seen in
Sis productions of words. In rana, for example, the earliest and most com-
monly produced form was [na na]. The variation seen in other productions of
this word could be explained as slips, the nature of which was determined by
the rules of Sis grammar (i.e., the force of word patterns). During stage IVwhen
the pattern of [t__n__] was established, rana was imitated a few times as [ta
na]. Its variant form [wa na] during stage VI could either be a phonological
reorganization (patterned after ratn) or a phonetic slip in the direction of a
close and strong pattern. In either interpretation, the production results in a
form that is consistent with other aspects of her phonology. A similar argument
can be made concerning the variation seen in reloj.
The variation seen in the words presented in Table 5.3 is in many ways typical
of Sis productions of particular words, and this variation had many causes. As
previously mentioned, some variation was due to the phonological reorganiza-
tion of old words as newword patterns were learned; similarly, the variation of a
word during a single time period was often due to the overlap of stages of
development. It has also been suggested that the infrequently occurring forms of
some words variable forms which nonetheless exhibited some regular pat-
tern can be attributed to the force of the word patterns (i.e., rule-governed
slips-of-the-tongue). All the causes of variation discussed in this section
resulted in phonologically revealing productions; in Section 4, randomly occur-
ring phonetic variation was mentioned and related to Sis typically lenis pro-
duction. One other type of variation was seen in a few words during the earliest
months of the study; this was variation that apparently stemmed from Sis
experimentation as she searched for an acceptable way to pronounce a
particular word. The most striking example of this experimentation was in
Sis productions of elefante prior to the point at which it stabilized as |ba te|
(Table 5.3B, stage III): [
l
hwan tu ti], [pfan tin di], [pan ti], [
1
ban tin di] (early
stage III). Its subsequent variation i.e., between [ba te] and [p/ban te] was
due to the overlap of two stages, with only the earlier one characterized by
obligatory cluster reduction. It may be that the occurrence of three different
rules discussed previously (i.e., harmony (rana and dame), initial-consonant
deletion (dame) and labial + dental word pattern (manzana)) was also due to Sis
experimentation in this case a search for a rule that would simplify word
structures (cf. Menn 1971 on Dannys discovery of harmony rules).
In summary, the evidence for the primacy of word patterns as the organizing
principle of Sis early phonological development has been the following: (1) all
words had a consistent word pattern form; (2) the gradual development of
classes of word patterns can best be described as a process by which new
patterns resulted from the expansion of previously acquired word patterns;
(3) some words changed pattern over time as new word patterns were learned;
(4) three of the four simplication processes operated to produce favored word
patterns as output; and (5) several unusual phonological substitutions and some
152 Marlys A. Macken
phonetic slips can only be explained by the notion of pattern force. It will be
seen (Section 6) that errors during the late stage of acquisition were usually
frozen forms of earlier word patterns that proved to be particularly resistant to
change (cf. regressive idioms, Moskowitz 1971 and 1973).
5.2. The learning of phonemic contrasts
As signicant as Sis word patterns are, there is evidence throughout this period
that Si was also learning the phonemic contrasts of Spanish. This evidence is of
two types: (1) the close correspondence between Sis word patterns and the
consonant structure of words that were selected as members of each pattern
class (which shows both an early ability to segment adult words and an early
recognition of place of articulation differences); and (2) the close correspond-
ence between some of Sis phonemes (as determined by different sets of phones)
and the phonemes of adult Spanish.
Fromthe very beginning, the close correspondence between the labialdental
place contrasts of Sis patterns and the place contrasts of the adult words that
comprised each pattern class demonstrates Sis recognition of the differences
between these places of articulation a necessary precursor to phonemic
learning. The velardental contrast in stops which was merged during stage
I (/k/ |t|) was produced by Si during stage II and was established by stage III
in all words except gato cat. Gato, which was subject to the early ordering
constraints, persisted as [ka ko] until stage VII, at which time it was produced
either as [ka to] or as [d/ta ko]. The latter form contradicts the metathesis rule of
the early stages and was the only example of metathesis involving a velar +
dental sequence in the entire corpus. By stage VII, the early metathesis rule had
dropped out. It may be that a new metathesis rule had replaced the earlier one;
since only the one word (gato) was affected, no rule is set up.
In addition to the separation of places of articulation, Sis stage I productions
showed a phonemically relevant distinction between manners of articulation
nasal/oral. The resulting four-way contrast of labial/dental and nasal/oral pro-
duced the four consonantal phonemes |ptmn|. These are listed with the glides as
the phonemic inventory for stage I in Table 5.4, which lists the order in which
consonants achieved phonemic status during stages I through VII.
During stage I, voicing was not distinctive, nor was the contrast between the
dental and palatal nasals (// |n|); Section 4 discussed all aspects of the
neutralization of adult contrasts seen in the substitutions of stage I. During
stages II to VII, Si began to distinguish the voiced/voiceless contrast in at least
the labial stops, acquired the palatal nasal, the stop/fricative contrast, and a
three-way contrast within the voiceless fricative class, and achieved a rudimen-
tary two-way contrast within the liquid class (Table 5.4).
Within the stop class, a three-way place contrast was accomplished by stages
II/III; the acquisition of the voicing distinction was more complicated.
Throughout stages I to VII, /b/ was usually produced as [w]; in the infrequent
Developmental reorganization of phonology 153
Table 5.4. Acquisition of consonant phonemes
Nonnal position
Stage Stops Nasals Glides Fricatives Liquids Medial consonant clusters only Final position Age
I pt mn wj 1;7
II k (n) 1;8.7
III b t 1 nt 1;9
IV s 1;10.7
(f)
V ()* (s) 1;11.7
VI f n 2;0.7
VII (mb) (1) 2;1
s
Notes:
1. ( ) = rst appearance.
2. = drops out after two weeks.
3. * = phonemically distinct but phonetically not adultlike.
cases where it was produced as a stop, it typically was voiced. The phoneme /p/
was usually voiced during stage I (although the number of tokens was small)
and was either voiced (28 tokens) or voiceless (20 tokens) during stage II; from
stage III to stage VII, it was rarely voiced. Further evidence of Sis contrast
between /b/ and /p/ can be seen in substitution patterns: /b/ [b, , w), rarely
[p] and [] in bola ball; /p/ [p, b], once [v], and never [w]. The [] which
occurred for /b/ in bola was analyzed as having separate phonemic status. The
adult contrast between /t:d/ and /k:g/ was absent. Of the dental pair, words
containing /d/ were very rare, and the /d/ was deleted in four out of the total ve
productions of /d/ words. The phonetic variability of /t/ was very similar to that
for /p/: usually voiced in stage I; either voiced (29 tokens) or voiceless (18
tokens) during stage II; and almost always voiceless during stages IIIVII ([t],
45 tokens versus [d], 4 tokens). The allophones of |t| were [t, d], and [k] (in gato,
as a result of velar assimilation). Tenedor was consistently produced with an
initial [p] that was analyzed as being phonemically unrelated to [t| (Section 5.1).
The adult phoneme /g/ was as rare as /d/; it occurred only in gato (where Si
treated it as |k|) and in guau guau (where the initial /g/ was phonemically |b| in
Sis system). On the basis of the differences between Sis productions of /b/ and
/p/, it can be argued that Si had a phonemic contrast of voicing in labial stops by
stage III; however, she had not fully mastered the phonetic control of voicing.
Sis dental and velar stops were phonemically and phonetically voiceless. The
Spanish phonemes /bdg/ also have voiced fricative allophones in medial posi-
tion; since Si had no voiced velar and dental stops, it would be reasonable to
expect an absence of [] and [], which in fact was the case. In the two words
guau guau bow wow and agua water where [] could be expected, Si
regularly produced |w| (although the glide was often produced with some velar
friction). For the initial segment of guau guau, Si usually produced [b], which is
the adult initial phone in this word in Spanish baby talk. The fricative allophone
of /b/ began to be produced in stage IV, but it occurred as often in initial position
as in medial position; Si gave no evidence of having the complementary
distribution relationship between /b/ and [].
The nasals /m/ and /n/ were phonemic in Sis system from the beginning, and
at least |n| was phonetically stable throughout all stages. The labial nasal,
however, was occasionally denasalized: |m| [m] occasionally [w] (3 tokens),
regularly [p] in manzana, and [m, w, or w] in Ramn. These latter substitutions
(in manzana and Ramn) were discussed in Section 5.1 as the outcome of
pattern regularization: at stage II, when manzana changed from [ma na] to [pa
na], the word is assumed to have been phonologically reorganized and the initial
segment is analyzed as |p|; in Ramn, two phonological forms are in competi-
tion |mon| (due to initial syllable deletion) and |bon| (due to medial segment
reduction). The palatal nasal was merged with |n| during stages I to III; during
stage IV, it was distinguished from |n| (|| [, j, n], a pattern of phonetic
variation typical of stages IVto VII). During stage VI, Si used ||, a sound that is
not phonemic in adult Spanish (Section 5.1).
Developmental reorganization of phonology 155
When fricatives were rst being produced (stages IVV), /f/ and /s/ were
phonetically interchanged in the words manzana and elefante. As with the [] in
bola, the [p] in tenedor, the [p] in manzana (stage III), and the [w] in Ramn, the
phonemic interpretation of [f] in manzana and [t] in elefante could be handled in
various ways, as for example by positing underlying adult phonemes. However, as
in the other cases, the less abstract solution was adopted: the surface form was
assumed to be the phonemic one (|f| in manzana; |t| in elefante).
All subsequently acquired /f/ and /s/ words were correctly produced, and the
different patterns of variation for the phonemes /f, s, t/ argue for their phonemic
status in Sis system at least by stage VI. Phonetic control over voicing was not a
problem from stage III for |t| and from stage IV for |s| and |f|, stages at which the
adult phonemes were rst realized as fricatives in Sis words: |t| was voiceless 29
out of 32 times; |f| was voiceless 20 times (of which 6 occurrences were in
manzana) and was produced as [w] 3 times; |s| was always voiceless ([t] 21
tokens, [] 3 tokens, [s] 21 tokens, [, ] 6 tokens, [ts] once and [h] 4 tokens).
However, during the stages when /f/ and /s/ were realized as stops (primarily
during stages II and III), voicing was variable: /f/ was a voiced labiodental stop 12
out of 18 times, and /s/ was a voiced alveolar stop 23 out of 32 times. Although
|s| which showed the greatest phonetic variation of the three fricatives was
frequently [t], the adult phonemes /t/ and /s/ were distinct for Si: /t/ was never
produced as [s]. In spite of the fact that most (14 out of 21) of the instances in
which |s| was produced as [t] occurred in productions of manzana, it is likely that
this was due to the high frequency of manzana among Sis /s/ words, rather than to
any confusion of /s/ and / t/. Words containing the velar fricative /x/ were not
common: /x/ was deleted in both conejo rabbit and reloj watch.
The class of liquids showed the greatest variation during all stages. The
lateral was the rst liquid acquired both phonetically and phonemically. It rst
appeared during stage III in bola ball, but was regularly deleted in three other
words. From stages III to VI, it was produced only twice. During stage VII the
rst stage at which it was produced with any regularity , it was realized as [l, j]
in intervocalic position and as [n, l] in initial and nal position. It was produced
as [d] only in production of reloj where the initial // was produced as a [b] (see
the discussion of pattern force and slips-of-the-tongue in Section 5.1). The //
was completely merged with |l|: it occurred in two words where it was produced
as [n, l]. In mira look, however, // was regularly produced as [j] (see Section 4
on the exceptional nature of this word). The // phoneme was similarly merged
with |l| in stages I through IV: // [l, n] (and one time each as [w d, ],
substitutions which may argue for a preliminary contrast between /l/ and //).
During stage V, Sis productions for // changed: // [l, d, d] in intervocalic
position (occasionally as [n]). However, in initial position, // [p] in ratn and
to [b, , l] in reloj. In rana frog, the // was either [n] or a labial consonant [m,
w, w]; for two weeks during stage VI, it was a velar nasal (Section 5.1). Clearly
the phonetic variation seen in initial position // does not warrant the assignment
of phonemic status, and it was analyzed as |b|. In this analysis, the [p] in ratn
156 Marlys A. Macken
was due to voicing agreement with the intervocalic |t|. In reloj, the initial [l] phone
occurred only when the medial |l| was produced as [l]; both consonants then
agreed due to lateral assimilation. In rana (a word acquired before the labial +
dental word pattern was established), the predominant initial phone was [n]; the
phonemic status could be either |n| or |b| (if |b|, then instances of the nasal phone
would be due to nasal assimilation). In fact, the earlier form [na na] occurred
during the stages in which // was merged with |l|; that rana may have been
subsequently phonologically reorganized with an initial |b| would explain why
both forms occurred. In medial position, the differences between the phones that
were realizations of // (at least by stage V) and those that were realizations of /l/
demonstrate that Si had at least made a rudimentary two-way phonemic distinc-
tion (in medial position only) within liquids: |l| (/l, r/) versus //.
Medial consonant clusters and nal consonants are included in Table 5.4 only
for convenience; they are not considered to be separate phonemes. The cluster /nt/
was the rst cluster to be acquired. In nal position, /n/ was acquired during stage
II and /s/ during stage V. However, both these nal consonants were found in only
one word each. They were not produced regularly until stage VI for /n#/ and stage
VII for /s#/. A third consonant /l/ appeared in nal position during stage VII, but
was restricted to only one word and was not consistently produced.
In this section, the acquisition of Sis phonemic contrasts has been described.
The correspondences between adult Spanish phonemes and Sis phonemes can for
the most part be explained by the following general processes: (1) deletion;
(2) phonemic merger; (3) voicing instability; (4) weakening; (5) strengthening;
(6) nasalization; and (7) place instability. In addition, the productions of some
words were lexical exceptions to general rules. However, several of Sis produc-
tions presented formidable problems to a strict phonemic analysis. The unusual
nature of the correspondences between the phones in these productions and the
phonemes in the adult words is apparently related to the ways in which Sis output
was determined by existing word patterns and Sis tendency to combine features
fromdifferent segments in the adult words. The explanation for why certain words
were given the same output patterns seems to lie in some general prosodic
similarity of the words involved (a phenomenon perhaps similar to the prosodic
schemas reported in Waterson 1971). The changes over time in output pattern for
some words appeared to be due to changes in Sis hypotheses about which words
were similar and should be said in similar ways. The unusual correspondences
often the result of feature combinations and the changes in words incurred as a
result of pattern force made a traditional phonemic analysis of words such as
manzana, elefante, rana, ratn, and tenedor quite difcult and decisions regard-
ing phonemic structure somewhat arbitrary.
6. Sis development from 2;2 to 2;5
During the period from2;2 to 2;5, Sis phonological systemimproved greatly. Her
set of phonemes expanded to include nearly all the phonemes of adult Spanish.
Developmental reorganization of phonology 157
Although each of her phonemes had several variants, her phonetic control was
much improved and the unusual substitution patterns characteristic of the earlier
period largely disappeared. The constraints on the co-occurrence of consonants
that had previously been a function of the word pattern goals also disappeared; the
only exceptions were frozen forms from the earlier period. In terms of syllable
structure, most two-syllable and many three- and four-syllable words were pro-
duced accurately. Those longer words that did not have the correct syllable
structure were reduced by means of a single set of rules: initial syllable deletion;
consonant cluster reduction; and an optional initial and/or nal consonant deletion
rule. These rules applied to all words, irrespective of the overall consonant
structure of the word; this systematicity is in sharp contrast to the variability
seen in earlier syllable simplication processes. As in the case of rules governing
segments and the co-occurrence of consonants, the only exceptions to the new
syllable structure rules were words which were frozen forms from the earlier
period.
6.1. Segmental system
By the end of the earlier period, Sis segmental system included phonemes
equivalent to all but four of the adult Spanish consonantal phonemes. In initial
and postconsonantal position, /d/ was merged with |t| and /g/ with |k|. In medial
position, the voiced fricative allophone of /d/ was treated either as |t| or as a
member of Sis liquid class, while the voiced fricative allophone of /g/ was
phonemically |w|. Two other phonemes were absent in Sis system as of 2;1: /x/,
which was deleted, and //, which was merged with |l|. By 2;5, Sis productions
showed some evidence for the contrastive status of each of these four adult
phonemes, although, as with most of her other phonemes, the phonetic realiza-
tion was variable. The allophonic distribution of the voiced stops and voiced
fricatives was not established.
Phonetically, voicing was still not completely mastered. The stop phonemes
|pbtd| occasionally showed voicing errors; |g| was often devoiced, while |k| was
never voiced. Among the fricatives, |f| and |s| were occasionally voiced, and
both exhibited a large amount of manner variation: |f| [f, fw, w, w, pw, b, v,
pf, w]; and |s| [s, , , , h, , t, t, d, d, ). In contrast, both |t| and |x| were
phonetically relatively stable: |t| [t], rarely [t, d]; and |x| [x, h, ] and []
only in jugo juice. In the English words shoes and home, Si used [t] and [h]
respectively. The nasals were phonetically very accurate, although the palatal
nasal continued to be produced as a nonnasal glide on occasion. The liquids
were by far the least stable phonetically, and in the case of the two r phonemes
bore little resemblance to the adult articulations: |l| [l, n, ], rarely [d]; ||
[l, j, d, , ]; and || [, l, n, d]. The fricative [], the intervocalic allophone of
/d/, was usually produced as [d], but occasionally as [l, n]; thus, its set of phones
was identical to the set of phones for |l|. Since initial-position |d| was never
produced as [l] and the relative proportions of [d] and [l] phones for [] and |l|
158 Marlys A. Macken
differed signicantly, Si at least contrasted /d/ and /l/ and, in most cases,
correctly treated [] as phonemically related to /d/. The infrequent cases of
[] [l] appear to be the last traces of her earlier indecision regarding the
phonemic status of adult [] (1;7 to 2;1).
The most signicant aspect of Sis phonemic systemduring 2;2 to 2;5 is that all
the unusual substitutions that characterized the earlier time period disappeared: no
new words were produced in ways consistent with the earlier pattern-dominated
substitution rules, and most old words were produced in a manner consistent with
the above description of phonemes and allophones. Only ve words persisted in
the earlier word pattern form: vestido, |#b| |p|; la nia/o, /#l/ |n|;
4
guau guau,
/#g/ |b|; Fernando, /#f/ |t|: and tenedor [b d].
6.2. Co-occurrence of consonants
During the period 1;7 to 2;1, the co-occurrence and ordering of consonants in a
word were determined by word patterns, primarily the patterns labial + dental
and velar + dental. In a small number of words, Si used metathesis as a way to
achieve the preferred ordering of consonants. During 2;2 to 2;5, all sequences of
consonants occurred, although there were still a few words in which the order
was dental + labial (in the adult model). In only one word did Si metathesize the
consonants: this word was gato cat, which already had appeared as [d/ta ko] in
the last stage of the earlier period and which remained in this form throughout
the study. All other words preserved the adult ordering of consonants.
In initial and medial positions, singleton consonants were produced accord-
ing to the description of Sis phonemes and allophones in the preceding section.
In nal position, /n/ and /s/ were well established (although /s/ was occasionally
deleted), /l/ was usually deleted but occasionally produced correctly, and some
allophone of nal /r/ ([, d], rarely [n]) was usually produced; Si used no words
containing a nal /d/. In addition, Si usually correctly pronounced /m#/ in the
English word home; in a few tokens the /m/ was deleted.
In clusters, /s/ and all liquids were deleted, with the exception of one production
of ste this [es ti] (2;5), and two productions of manzana apple [m sa na]
(manzana was usually produced as [sa na]). The nasal + stop clusters /mb/, /nt/,
/nd/, and /nk/ ([k]) were well established; if one member of the cluster was
deleted, it typically was the nasal in nasal + voiceless stop clusters and the stop in
nasal + voiced stop clusters (see also Ferguson 1977b). The cluster /ng/ ([g]) was
nearly always produced as [k], rarely as [g], once as [], and once as [k]. Si used
no words with the cluster /mp/. The rules governing consonant occurrence in
clusters and nal position were similar in the earlier and later periods.
6.3. Syllable structure
As was seen in Section 5, the variable nature of syllable deletion rules provided
some of the clearest evidence for the existence of favored word patterns during
Developmental reorganization of phonology 159
the early period of 1;7 to 2;1 (Table 5.2). In the later period, however, Si had two
regular rules for reducing three- and four-syllable words to a shorter form:
delete the initial syllable(s), or delete the initial or nal consonants. Of the
approximately seventy words that she acquired between 2;2 and 2;5, only one
was reduced in a manner consistent with the earlier word pattern stage: pantaln
trousers [pa lon] or [bum po lon]. In addition, many words were produced
with the correct number of syllables. Only one word persisted in the form from
the earlier word pattern stage: tenedor fork [b
l
d]; this word was
produced frequently and always in the same labial + dental form.
To summarize, Sis productions from 2;2 to 2;5 were largely accurate realiza-
tions of the complex consonant and syllable structure of the words she knew.
Little evidence remained of the earlier stage where the realization of consonants
and syllables was constrained by a small number of word patterns. It is worth
noting that had Si only been seen beginning when she was 2;2 (or as early as
1;11), the phonological forms of words such as tenedor, restido, reloj, or
pantaln would have appeared anomalous; it is only within the context of Sis
earliest development that such words can be seen as an integral part of her
developing phonological system.
7. Discussion
In the past ten or so years, interest in child phonology has greatly increased
among linguists, mainly as a result of the inclusion of child phonology data in
the class of behavior for which a model of phonology must provide explan-
ations. As a result, many linguistic studies have recently appeared that docu-
ment the acquisition process for individual children and thus have contributed to
the description of phonology acquisition in general. Since phonological acquis-
ition has not been studied sufciently to separate developmental phenomena
reliably from those phenomena that reveal universal properties of phonological
structure, it is premature to make strong claims concerning the contributions of
phonology acquisition to general phonological theory. It is, however, appropri-
ate to ask what the study of an individual child in this case, Si may contribute
to what we know or hypothesize to be true of the acquisition of phonology.
Toward this goal, aspects of Sis data will be reviewed in this section as they
pertain to a general model of phonology acquisition (in particular the relevant
units of acquisition), universals of acquisition, and individual differences. It
is, of course, obvious that the validity of the following interpretation can only be
substantiated (or invalidated) on the basis of additional studies of many more
children acquiring a variety of languages.
Several recent papers have argued that acquiring phonology is basically a
cognitive process i.e., a problem-solving task (see in particular Kiparsky and
Menn 1977). At several points in the preceding sections, data were interpreted
as being explained best by such a model. Probably the most dramatic evidence
in these data of the child as the active organizer of phonology is Sis
160 Marlys A. Macken
phonemicization of the velar nasal, a phonetic segment that occurs in highly
restricted environments in adult Spanish but has no phonemic status. The theme
of the childs active role in acquisition will also gure prominently in the
following discussion of the units of acquisition. However, it will be in the
evaluation of individual differences that the problem-solving model will be
most useful.
Of central importance for any model of child phonology is the character-
ization of the units of acquisition. During the earliest stages (from 1;7 to 2;1),
Sis data showed several phenomena which could best be accounted for by
assuming a central role for the word as a basic unit being acquired. More
specically, the data argue for several levels of representation in her phonolog-
ical system. These levels correspond to two and possibly three basic units: the
word, the phoneme (i.e., some segmental unit larger than the feature), and
possibly the feature.
By word is meant a grammatical unit: in these data the word is the
morpheme and the morpheme is the word. A grammatical unit was needed to
describe the constraints on sequences of smaller units (which are best charac-
terized as phonemes) and to specify the domain within which phonological
processes operated (e.g., the simplication process of consonant harmony). In
addition, word boundaries were needed to block processes (with the exception
of some early sentences which were characterized by harmony across word
boundaries, Section 4) and to condition processes (e.g., word-initial versus
word-nal (but not necessarily syllable-nal) phenomena). Some such gram-
matical unit is frequently used in just these ways in descriptions of adult
phonologies.
However, the data support an even stronger claim regarding the word,
namely that the word and associated word-structure constraints are psycholog-
ically real. The evidence for this claim is that Si formulated hypotheses about
the nature of Spanish phonology on the basis of the similarity between words,
that she abstracted what were referred to here as word patterns, that she
expanded and generalized these word patterns to handle new words and that
she changed the output form of some old words as she learned new word
patterns (i.e., words that were similar in some ways underwent rules together).
The claim is then that the word was a basic organizational unit of her (early)
phonological development and that without considering the word as a unit,
some phonological phenomena would have appeared quite arbitrary and the
frequency and consistency of other phonological phenomena could not have
been easily explained.
Although the learning of words and word patterns, phonemes and features was
occurring simultaneously throughout this period of 1;7 to 2;5, the evidence for the
centrality of words and word patterns had largely disappeared by 2;1. In fact, the
data from 2;2 to 2;5 could be adequately described in terms of phonemic
contrasts, allophonic relationships, and phonotactic constraints. This change
suggests that the levels of representation constitute a hierarchy at least
Developmental reorganization of phonology 161
developmentally in which words and word patterns dominate phonemes, and
phonemes in turn are more central than features. That the phoneme is a basic
building block of the acquisition process can be seen in Section 6; the evidence for
the feature is considerably weaker. Although, it is clear that Si could analyze
phonemes into component features (as seen for example in the way in which she
recombined features from different segments of the adult word), there is little
evidence that once having mastered particular features, she could generalize this
knowledge to the acquisition of a new phoneme. For example, the acquisition
of the features voicelessness and frication (|f, s|) and velarity (|k|) were not
sufcient for the acquisition of /x/ (cf. also Ferguson 1977b). The strongest
evidence for the status of the feature as a unit is the overgeneralization of the +
velar feature in initial position (word pattern) which resulted in the phonemiciza-
tion of an initial velar nasal a process that (if accurately stated) clearly involved
the analysis and generalization of features.
Such a developmental hierarchy would be consistent with the even stronger
claims regarding the word made by researchers who have studied children
somewhat younger than Si (Cruttenden 1971; Ferguson and Farwell 1975;
Menn 1971, 1977; see also Ferguson 1977a). The change in the hierarchy
seen at roughly 2;2 (but probably occurring earlier) would be consistent with
the Piagetian claim that at each successive period of development in the
acquisition of complex systems, previous skills are reorganized into a new
system of knowledge (Piaget 1952; see also Bower 1974 for a similar view of
motor skills). This knowledge of word-structure constraints is not lost, and in
fact must be substantially elaborated, for it is part of an adult speakers knowl-
edge of his/her language (see Greenberg and Jenkins 1964).
Clearly, a model of phonological acquisition should also describe the uni-
versal aspects of acquisition. In Sections 3 through 6, aspects of Sis develop-
ment that correlated with putative universals as claimed by various researchers
in child phonology were noted. Here, some of those aspects will be reviewed, in
order to place them within the context of phonology acquisition in general.
One of the most signicant aspects of child phonology is the simplication of
adult words to one or two syllables, typically composed of alternating singleton
consonants and vowels. Constraints on the length of words, and on the com-
plexity of co-occurrence of sounds in words, are probably universal. Smiths
four functions of child rules (1973) describe the major types of simplication
(see also Ingram 1974b), and these are seen in Sis data as well (Section 5).
Of the theories that deal with the universal order in which consonants are
acquired, Jakobson 1941/1968 is the most explicit. Although there are problems
with this theory (see Ferguson and Garnica 1975), the general claims regarding
the order of acquisition of classes of consonants t the general pattern seen in
Sis data and in several other studies of Spanish acquisition (Macken 1975,
1978). Si acquired the classes of stops and nasals before the class of fricatives,
and liquids were acquired last. Front consonants were acquired before back
ones, and voiceless before voiced. /l/ was the rst liquid acquired, with the
162 Marlys A. Macken
contrast between the r-phonemes acquired considerably later. However, /t/ was
acquired before the rst true fricative, contrary to Jakobsons prediction (see
also Macken 1975, 1978). Before being produced correctly, the fricatives were
replaced by the homorganic voiceless stops, back consonants by front ones, and
the r-phonemes usually by |l| a set of substitutions predicted by Jakobson.
When compared to data from other children acquiring Spanish, Sis data are
identical in at least the additional, following ways: /n/ was acquired in nal
position before /s/, with both before nal /l/ or /r/; learning the allophonic
relationship of stopspirant for the voiced stop phonemes proved difcult and
the earliest stage was one in which the voiced stops were usually produced as
glides and [] in some ways patterned with the liquids (see also Stoel 1974);
many aspects of the co-occurrence of consonants in words were determined by a
strength hierarchy in which voiceless stops (and nasals in the earliest stages) are
stronger than voiceless fricatives (Macken 1975, 1978). In addition, a wide
range of phonetic variation (Section 3) and the exceptional status of some
lexical items (e.g., Sis mira and guau guau, Section 4, and the frozen
forms in Section 6) were seen in all the Spanish subjects and are probably
universal. (From even so brief a summary, it is clear that it is at the level of
segment inventories that Si is most similar to other children acquiring Spanish
and other languages.)
Some aspects of Sis development appear to be characteristic only of Si
among the six subjects studied; other aspects are unique only in the frequency
with which they occur in her corpus. For example, all subjects imitated to some
degree at some point: Si imitated more than any other subject; J, another subject,
imitated rarely until the age of 2;1, when he began to imitate a great deal. Si
coalesced several learned routines into a single unit (e.g., qu es? what is
|kes| and unit phrases like mommy home); most subjects did this only with
donde est? where is (which was produced, as in Spanish baby-talk, as |n ta|)
and one subject never used such forms. Si was the only subject who produced
many misperceptions (Section 3).
As was pointed out before (Section 4), there are certain parallels between Sis
general language behavior and aspects of her phonological development, as for
example between the coalesced routines and the coalesced word patterns. The
misperceptions suggest a global, only partly differentiated auditory processing
which is paralleled by her loose, prosodic treatment of words. The pattern which
emerges is consistent in several ways if we view Si as a child whose preferred
processing mode is a global one rather than a detail or analytic one.
5
Within this
context of an information-processing model for interpreting individual differ-
ences (see in particular Zelnicker and Jeffrey 1976), the difference between Si
and a child like J as a global- versus a detail-processing child relates several
aspects of their language behavior with specic differences between their
phonological development (see Macken 1976). This view would predict that
the tendencies to scan an entire three- or four-syllable word for pattern-criterial
consonants, to use several different syllable reduction rules, and to have
Developmental reorganization of phonology 163
substitution and metathesis rules sensitive to word position and pattern goals
would occur together and be of signicantly greater frequency in the phonology
of global children such as Si than in the phonology of nonglobal children like
J. This analysis of differences in terms of cognitive styles is not an unreasonable
extension of a cognitive model of phonology, one emphasizing the problem-
solving nature of acquisition, and is promising in that it may if successful
restrict the range of individual phonological differences to several sets or
syndromes typical of different styles.
In addition to the possible set of restrictions on individual differences
deduced from a set of basic preferences in processing styles, it appears that
the particular structure of the language being learned may also restrict the
number and type of individual differences among children learning that lan-
guage. Although Si is the only subject in the study to have such a strong
preference for word-initial labial
6
(and subsequently word-initial labial or
velar) word patterns, it may be that in part this particular preference reects a
particular property of Spanish: Hooper (1976) suggests that in Spanish labial
and velar consonants are considerably stronger than dental consonants in
syllable initial position. It seems reasonable to expect that the different ways
in which children organize their phonology will reect differential selection and
emphasis of particular aspects from the complete possible set of complex
relationships that obtain in the particular language being acquired (Macken
1976).
notes
1. Although we do not know how Sis parents pronounce this phrase, it is likely that
adults (in fast speech) also delete one of the two vowels; however, the other children
used [ke] when requesting information in comparable situations. La nia is another
example of a unit phrase (see footnote 4).
2. Although Si is chronologically older than the young children reported on by
Ferguson and Farwell and Menn, she is similar to these children in size of vocabulary.
Ferguson and Farwell use the rst fty words as the time domain for the phenomenon
they describe. Si was using a vocabulary of approximately fty-two words up to the
age of 1;10.15 and had only ninety-seven words by the age of 2;2 (n.b. as determined
by her productions during our experimental sessions). With respect to age and
vocabulary growth, she is similar to the other children in our study.
3. The gloss for this production could be libro book, in which case Si metathesized the
consonants to achieve the preferred output (see also sopa soup). This was one of
several cases in which an examination of Sis imitated corpus was of no help in
solving a problem occurring in the spontaneous corpus: Sis imitations of libro were
always closer to the adult form than her spontaneous production. It was clear from the
context that when she produced the form given in the table, she was referring to a
book, but whether the origin of her form was libro or librito could not be determined.
4. At about 1;9, Si began producing nia girl as na ni na (later, the palatal nasal was
correctly produced); the adult form of la nia the girl was evidently lexicalized as a
unit. Nio boy was also produced with an initial extra syllable na. Since Si often
164 Marlys A. Macken
neutralized nal vowels, it was frequently not clear what gender was intended; she
would, however, produce a nal schwa in imitation of either nia or nio. This word
|na ni na/o| was very common and was regularly produced with three syllables. No
other word was lexicalized with the article into one unit in this way.
5. Peters (1977) reports on the language of a somewhat unintelligible child who used
two types of speech, analytic and gestalt. Peters suggests that these two types of
speech may reect two language-learning strategies that may be used to different
degrees by different children. This contrast of analytic versus gestalt is similar in
some respects to the detail versus global cognitive style dichotomy of Zelnicker
and Jeffrey (1976).
6. Montez Giraldo (1970:488) reports that Cuando en la palabra hay una labial, es
frecuente que se produzca una mettesis que tiene coma resultado iniciar la palabra
con la labial. However, six out of the eight examples that the author gives are from
only one child (Emilia 2027 months) of the four that he studied. This child (see also
Montez Giraldo 1971) metathesized both dental + labial and velar + labial sequences
but apparently no others (e.g., zapato patato; camisa manika, 1971: 339).
References
Alarcos Llorach, E. (1950). Fonologa espaola (3rd edn. 1961). Madrid: Editoral
Gredos.
Bloom, L., Hood, L., and Lightbown, P. (1975). Imitation in language development: if,
when and why. Cognitive Psychology 6, 380420.
Bower, T. C. R. (1974). Repetition in human development. Merrill-Palmer Quarterly,
20, 3038.
Bush, C. N., Edwards, M. L., Luckau, J. M., Stoel, C. M., Macken, M. A., and
Petersen, J. D. (1973). On specifying a system for transcribing consonants in
child language: a working paper with examples from American English and
Mexican Spanish. Stanford: Stanford Univ., Dept. of Linguistics.
Cruttenden, A. (1970). A phonetic study of babbling. Brititish Journal of Disorders of
Communication, 5, 11017.
Dalbour, J. B. (1969). Spanish pronunciation: theory and practice. New York: Holt,
Rhinehart & Winston.
Farwell, C. B. (1976). Some strategies in the early production of fricatives. Papers and
Reports on Child Language Development, 12, 97104.
Ferguson, C. A. (1977a). Learning to pronounce: the earliest stages of phonological
development in the child. In F. D. Minie and L. I. Lloyd (eds.), Communicative
competence and cognitive abilities early behavioral assessment. Baltimore:
University Park Press.
(1977b). New directions in phonological theory: language acquisition and universals
research. In R. W. Cole (ed.), Current issues in linguistic theory. Bloomington:
Indiana University Press.
Ferguson, C. A. and Farwell, C.B. (1975). Words and sounds in early language acquis-
ition. Language, 51(2), 41939. Reprinted in this volume as Chapter 4.
Ferguson, C. A. and O. K. Garnica, (1975). Theories of phonological development.
In E. H. and E. Lenneberg (eds.), Foundations of language development, vol. 2.
New York: Academic Press.
Ferguson, C. A., Peizer, D. B., and Weeks, T. E. (1973). Model-and-replica phonological
grammar of a childs rst words. Lingua 31, 3565.
Developmental reorganization of phonology 165
Fromkin, V. A. (ed.), 1973. Speech errors as linguistic evidence. The Hague: Mouton.
Greenberg, J. H. and Jenkins, J. J. (1964). Studies in the psychological correlates of the
sound system of American English. Word, 20(2), 15777.
Hooper, J. (1976). An introduction to natural generative phonology. New York:
Academic Press.
Ingram, D. (1974a). Fronting in child phonology. Journal of Child Language, 1, 23341.
(1974b). Phonological rules in young children. Journal of Child Language, 1, 4964.
Jakobson, R. (1941/1968). Child language, aphasia and phonological universals, trans.
A. R. Keiler. The Hague: Mouton. (Originally published as Kindersprache, Aphasie
und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell.)
Kiparsky, P. and Menn, L. (1977). On the acquisition of phonology. In J. MacNamara
(ed.), Language learning and thought. New York: Academic Press.
Labov, W., Yaeger, M., and Steiner, R. (1972). A quantitative study of sound change in
progress, vol. I. Philadelphia: US Regional Survey.
Macken, M. A. (1975). The acquisition of intervocalic consonants in Mexican Spanish: a
cross-sectional study based on imitation data. Papers and Reports on Child
Language Development, 9, 2942.
(1976). Individual differences in phonological acquisition: strategy versus cognitive
style. Paper presented to the Child Language Seminar Series, Stanford University,
May 1976.
(1978). Permitted complexity in phonological development: one childs acquisition of
Spanish consonants. Lingua, 44, 21953.
Menn, L. (1971). Phonotactic rules in beginning speech. Lingua, 26, 22551.
(1974). A theoretical framework for child phonology. Paper given at the Summer
Meeting of the Linguistic Society of America, Amherst, MA.
(1976). Evidence for an interactionist-discovery theory of child phonology. Papers
and Reports on Child Language Development, 12, 16977.
(1977). Phonological units in beginning speech. In A. Bell and J. B. Hooper (eds.),
Syllable and segments, pp. 15771. Amsterdam: North-Holland.
Montez Giraldo, J. J. (1970). Dominancia de las labiales en el sistema fonolgico del
habla infantil. Thesaurus, 25(3), 4878.
(1971). Acerca de la apropriacin por el nio del sistema fonolgico espaol.
Thesaurus, 26, 32246.
Moskowitz, A. I. (1971). The two-year-old stage in the acquisition of English phonology.
Language, 46(2), 42641.
(1973). The acquisition of phonology and syntax: a preliminary study. In
K. J. J. Hintakka, J. M. E. Moravcsik, and P. Suppes (eds.), Approaches to natural
language. Dordrecht: Reidel.
Peters, A. (1977). Language learning strategies: does the whole equal the sum of the
parts? Language, 53, 56073.
Piaget, J. (1952). The origins of intelligence in children. New York: International
University Press.
Smith, N. V. (1973). The acquisition of phonology, a case study. London: Cambridge
University Press.
(1975). Universal tendencies in the childs acquisition of phonology. In N. OConnor
(ed.), Language, cognitive deciencies and retardation. London: Butterworths.
Stockwell, R. P. and Bowen, J. D. (1965). The sounds of English and Spanish. University
of Chicago Press.
Stoel, C. M. (1974). The acquisition of liquids in Spanish. Unpublished PhDdissertation,
Stanford University.
166 Marlys A. Macken
Vihman, M. M. (1978). Consonant harmony: its scope and function in child language. In
J. H. Greenberg, C. A Ferguson, and E. A. Moravcsik (eds.), Universals of human
language. Stanford University Press.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics. 7,
179211. Reprinted in this volume as Chapter 3.
Zelnicker, T. and Jeffrey, W. E. (1976). Reective and impulsive children: strategies of
information processing underlying differences in problem solving. Society for
Research on Child Language Development Monograph 168.
Developmental reorganization of phonology 167
6 Development of articulatory, phonetic,
and phonological capabilities
Lise Menn
I. Introduction
A. Background
A rather large body of information about the early stages of the acquisition of
the phonology of English and some other languages (Spanish, Mandarin,
Thai) has become available over the last decade or so, and the theory of
acquisition of phonology has not only grown, but has changed its nature
considerably. Since about 1974, we have moved away from a model in
which phonological development was considered to resemble the differentia-
tion of an embryo. In its place we have evolved a notion of the young child as a
creature of some intelligence who is trying to solve a problem: the problem of
sounding like her companions when communicating with them. This shift of
model took place as more diary and small-group studies were published, and
in the context of Slobins similar approach to the acquisition of morphology
and syntax (Slobin 1966, 1973).
In recent years, the study of child phonology has also become distinctly more
psychological in the explanatory concepts that it employs. This is largely
because the richer data base has made it possible to see a considerable range
of individual differences among children. Faced with such diversity, we have
had to look below the surface for an underlying unity; and in doing so, we have
begun to invoke notions of processing and storage of information in addition to
the linguistic notions of articulatory control and phonemic contrast.
In this chapter, I will review the strategies that children are presently believed
to use in acquiring phonology, and I will give an account of the psycholinguistic
model of early phonology which I think is presently the most adequate. For
more extensive discussion, in addition to the references which will be cited, the
reader should see the many important papers collected in Yeni-Komshian,
Kavanagh, and Ferguson (1980).
My grateful thanks to Sarah Hawkins, Paula Menyuk, and Ronnie Wilbur, who spent considerable
time and effort working over the rst draft of this chapter. At the urging of Charles A. Ferguson, and
with the help of Prof. Hawkins, I have attempted to overcome old habits and use IPA consistently
throughout it.
168
B. Plan of exposition
We will begin this summary of the acquisition of phonology by looking at the
transition from babble to speech in Section II. This is necessary so that we can
understand the problems of dening which early vocalizations a theory of child
phonology attempts to account for.
Then we shall undertake the construction of a model of child phonology that
will allow us to deal separately with three different kinds of information that the
child is acquiring: (1) knowledge of howwords sound, (2) knowledge of howto
pronounce them, and (3) knowledge of allomorphy or abstract phonology,
manifested as the relationships among words or morphs that sound somewhat
different but are the same in meaning.
In Section III.A, we shall discuss the childs perceptual knowledge of the
sounds of language; in Section III.B, we will pause to discuss problems with the
notion of phoneme in early stages of phonological development. In Section
III.C, we will turn to the traditional subject matter of child phonology: the ways
in which children pronounce the words of the adult language (Section III.C.2).
We will see, however, that modifying the pronunciation of a word is only one
type of reaction to the complexities of adult phonotactics; the other type of
reaction is the avoidance of particular sound patterns (Section III.C.1). These
behaviors can be unied, at least in the early stages of acquisition, by the formal
descriptive device of saying that children obey phonological output constraints
(Section III.C.3), and data from a variety of children are presented which
support this description.
In Section IV, we consider howa child may go about inventing rules to derive
her output forms fromthe forms spoken by adults; Section IV.Aextends the idea
of ease of articulation by including skill already acquired at a point in time
as a factor affecting the ease of a new sound. In Section IV.B, the notion of
naturalness in child phonology is discussed, and we consider how it can be
related to our account of rule-creation. In Section IV.C, we note the non-natural
rules that children can also create, including those which appear to stem from a
dim awareness of the fact that allomorphy exists in the adult language. Section
IV.D describes data on rule origin and growth, and Section IV.E concerns rule
overgeneralization.
Section V presents the notion that early phonological development should be
viewed as the development of skill in the ability to program and execute
complex motor sequences. It begins by noting the theoretical importance of
the two irregular phenomena that are most difcult for conventional approaches
to deal with: overgeneralizations of rules of child phonology (Section V.A) and
phonological idioms (Section V.B). Then the regular pattern of the arrangement
of childrens early words into families of canonical forms is recalled in Section
V.C. Section V.D attempts to account for these three fundamental types of data
within a unied two-stage model of articulatory motor programming.
Development of articulatory, phonetic, and phonological capabilities 169
In Section V.E, we see how this model can be tted into the overall picture of
child phonology that was set up in Section II.A; we ex its explanatory muscles
in dealing with rules and rule changes, and we consider some of its conceptual
limitations in subsection V.E.3 entitled Caution: The limitations of the pro-
gramming metaphor.
In Section VI, we deal with some difcult logical and methodological topics:
the relation of imitated to spontaneous productions and the nature of chil-
drens metalinguistic ability to focus on pronunciation as a task (Section VI.A).
These are related to the perennial problem of why some sounds may appear in
babble but not in speech (Section VI.B). (Section VI may be considered
logically prior to most of the rest of the chapter.)
Section VII, The acquisition of allophones and allomorphs, turns to the
other major branch of developmental phonology, and gives a brief outline of this
topic, especially as it relates to questions of psychological reality. The reader is
referred to MacWhinneys (1978) monograph on this topic, since including a
full account of it would double the length of the chapter.
Finally, Section VIII lists the major ndings of the past decade of research in
developmental phonology, recalls the motor programming model for the begin-
ning stages of acquisition of phonology which was proposed in Section V, and
briey contrasts the working assumptions of the current approach with those of
the preceding Jakobsonian era.
Note: Some longitudinal studies will be cited repeatedly in this chapter. For
convenience, unless otherwise noted, references to Hildegard are from
Moskowitz 1970b (originally, of course, from Leopold 1939), to Daniel
from Menn (1971), to Si from Macken (1979); to Jacob from Menn
(1976a, b); and to Amahl from Smith (1973).
II. The transition from babbling to speech
We can usually assume that phonology deals with sound patterns of words, but
even in adult languages we must decide whether the phonology that we write
should attempt to include certain marginal items, for example, onomatopoeic
representations of animal cries and noises. In studying early child phonology,
this problem, marginal in dealing with most adult languages, becomes central.
There is no ready-made solution to it; in this section, I will just attempt to show
the nature of the difculty.
There seem to be three denable types of utterances found during the
transition period that we call the onset of speech: sound-play, protowords,
and modulated babble. Modulated babble refers to the use of strings of sounds
which appear to carry meaning only by their intonation contour.
This is also called jargon and it can be very eloquent and effective vocal
communication. Since our concern is with the development of articulation, we
will not discuss modulated babble further here; the reader is referred to von
Rafer-Engel (1973) and Menn (1976a, b).
170 Lise Menn
Sound-play, which may include word-practice, is not communicative behavior;
in other words, when we classify an utterance as sound-play, we do so because
there is no indication of any association between recurrent context and recurrent
sound-play patterns. One can of course say that sound-play is expressive of a
cheerful mood, but in that weak sense, any evidence of mood is communicative.
Joint sound-play is another matter; it is certainly communicative action, but it
seems to be absent or rare in adultchild pairs in our culture when the child appears
mature enough to be on the threshold of speech, although it is certainly found with
young infants (see Sterne, Jaffe, Beebe, and Bennett 1975; Snow 1977).
Proto-words are articulated meaningful utterances; some of themare directed to
others (one can tell because the child gets annoyed if no one responds), and some
are solo performances. These are our objects of study, for only here can we be
certain that the child is trying to say a word that is, trying to match a desired
perceived target. And again, we judge that they are meaningful because of a
recurrent association between sound and situation (although obviously if what
appears to be a clear token of an adult word is uttered just once in a context for
which it is strikingly appropriate, it is usually included as a meaningful utterance).
A child may have all of these utterance types for a period of several months.
Some utterances, furthermore, may contain elements that belong to more than
one class: for example, a child may start playing sound-games with a real
word (Weir 1962; Menn 1976a), or he may address one with an utterance that
has a real word or two embedded in modulated babble (Jones 1967). And of
course, in practice, some utterances are hard to classify, since classication
depends in part on surmising the childs intent.
The important point here is that clear cases can easily be found, and that a
child may have one, two, or all three of these utterance types for a period of
many months. The silent period, despite the emphasis given to it in the older
literature, is a rare phenomenon.
There is a fourth type of utterance that we should mention. Some childrens
early attack on language proceeds by global approximation to long phrases
rather than by attempts at single words or short phrases. Their early efforts at
speech are characterized by variable and often loose articulation which is
extremely hard to transcribe; Ann Peters (1977) dubbed these children mush-
mouth kids. In this chapter, we shall consider only children who take the more
segmental word-by-word approach to phonology; the reader who is interested in
the global approach should see the Peters article and also Branigan (1979).
Proto-words now need to be dened more carefully. They are vocables
(articulated utterances) which recur in denable contexts. One might fear that
this notion of recurrent denable contexts would be very difcult to use, but it
generally is not, because a one-year-olds activities tend to fall into identiable
behavioral routines, some solo and some partnered. These include favourite
manipulations on objects (putting things into things), games (peekaboo), direct-
ing an adults attention (pointing), obtaining things (requesting/demanding),
offering things, greetings, farewells, and so on. Halliday (1975) describes such
Development of articulatory, phonetic, and phonological capabilities 171
pairs of vocalization and behavioral routine in elegant detail; see also Menn
(1976a) and Clumeck (1977). The meaning of a proto-word is originally very
limited, and is best characterized as what you say when you do X. Proto-
words may thus usefully be considered as one type of vocal signal; they are not
yet symbols, because each of them is bound to the performance of its routine; it
cannot be used freely in new contexts. At some point, however, rst singly and
then more rapidly, some of the proto-words start to be used in more situations,
and thus they begin to acquire the symbolic autonomy of the true word. For
example, a woof-woof vocable may be initially used only when a child is
pointing to a picture of a dog; then it may be generalized rapidly to pointing to
real and toy dogs, and yet it may take months to become usable in requesting a
toy dog. Incidentally, proto-words do not have to have adult words as models
(Halliday 1975), and some without adult models may even make the transition
to becoming true symbols (Menn 1976a, Menn and Haselkorn 1977).
Proto-words are, by denition, the rst units for which a child is trying to
produce a particular articulated sound pattern for communication (always
excepting the whole-phrase efforts of the mush-mouth kids). If we wish to
make generalizations about the childs rst phones, or to evaluate the applic-
ability of terms such as phoneme to the onset of speech, we must look into the
period when proto-words are rst being produced. Sometimes what we see is a
handful of nicely dened CV(CV) shapes, as tradition would have it: [papa],
[mama], [dada]. A good example is given in Ferguson, Weeks, and Peizer
(1973). But more often, apparently, the early picture shows quite a mixture of
forms: some vowelless items, perhaps, such as [m:: ], or Hildegards [::]; some
traditional CV(CV) shapes and/or some (C)VC and VCV shapes; perhaps an
isolated word with a consonant cluster (Hildegard, again); and some wildly
uctuating forms that seem to originate from rather complex adult target words
(e.g., Jacobs renditions of thank you, which showed an endless variation
including [deig], [geigu], [gigo] [g:do], [dejo], [dido], [dt], [it]).
Summarizing this section: the transition from babbling to speech is typically
gradual, and may involve any combination of four types of utterances: sound-
play, modulated babble (using meaningful or possibly meaningful intonation
contour), whole-phrase efforts, and proto-words. Proto-words are meaningful
utterances with phonetically denable targets; however, the phonetic denition
may be quite loose by the standards of adult phonetic target-matching and the
meaning may be very limited and situation-bound. We will take child phonol-
ogy as beginning with proto-words, and in Section III.B we will examine the
problem of applying adult-based phonological concepts to these rst words.
III. Constructing a model of early phonological knowledge
In this section, we will undertake the description of some aspects of early
phonological knowledge. This includes what children, in the rst months of
speaking, seem to know about the sounds of words in adult language
172 Lise Menn
(perceptual knowledge), about the relations among those sounds (phonological
knowledge, including knowledge of segmentation and phonemic contrasts),
and about how to pronounce words.
The most striking fact about early child words has always been how sim-
plied most of them are compared to their adult models. What has made child
phonology an object for study has been three realizations about these simpli-
ed forms: that there are generally systematic relations among a given childs
words, that there are generally systematic relations between the childs word
and the adult model word, and that it is possible, by comparing children who
have very different ways of dealing with adult words, to come up with a general
theory of why and how these generally systematic relations exist. These three
realizations will be developed in this section and in the two which follow.
Note: Beginning in this section, I will occasionally draw small ow-chart
diagrams in order to keep track of the various capacities for processing and
storage that we postulate in order to account for the childs language behavior. It
is important to keep in mind that the entities and processes represented by these
boxes and arrows are only hypothetical constructs, and that even the best
guesses among them must be grossly oversimplied compared to whatever it
is that we have in our heads.
III.A. The input lexicon: representation of the adult word
Lexicon is a word whose precise meaning varies from user to user, but it at
least denotes a collection of stored, accessible, memorized bits of information
about the sounds and meanings of words and/or their component meaningful
parts. We must grant that something which should be called a lexicon exists in
the human individual; that is, there must be some form of long-term storage
containing at least a sketchy encoding of the sound pattern and meaning which
is accessible when we recognize and understand a word.
In order to say a word spontaneously and meaningfully, one must also have
access to stored information about how it sounds and what it means; a standing
controversy is whether this knowledge is best represented by postulating a
separate output lexicon or whether both recognition and production informa-
tion are best conceived of as being in a single lexicon (Butterworth 1983: chs. 6
and 7).
To advocate a single lexicon in a psycholinguistic model of child phonol-
ogy is to hypothesize that the rules which create the childs output form from
her input form operate in real time; to advocate a two-lexicon model is to
claim that a form closer to the output form is also stored and that this second
form is used as a basis for production. Much of the data that we will consider
can be handled more gracefully in a two-lexicon model than in a one-lexicon
model; I think the two-lexicon model is likely to be a better approximation to
what we really utilize in speaking, and so I will use it in this chapter. It is by no
means universally accepted as the superior model (cf. N. V. Smith 1978),
Development of articulatory, phonetic, and phonological capabilities 173
however, and formally all the data that it handles can be managed in a one-
lexicon system, by the use of markings on each lexical entry specifying
which rules apply to it in the event of competing rules applying to the same
domain.
We shall say, then, that two forms may be stored for each word: a recognition
form and a production form. The collection of words (formmeaning pairs) that
a speaker can recognize and understand is called the input lexicon; it could
equally well be called the recognition lexicon or the passive lexicon. The
collection of words that a speaker can use (that is, the information necessary to
use them meaningfully and to pronounce them) is referred to as the output
lexicon, but could also be thought of as the active lexicon. (This active/
passive dichotomy is usually thought of as a matter of knowledge of word
meaning rather than pronunciation, but the extension of it to include knowledge
of pronunciation seems to capture the right distinction.) So far, then, we have the
rudimentary diagram shown in Figure 6.1.
Let us explore the properties that can be ascribed to the input lexicon. We
knowthat speech perception is an active process: the hearer lters and structures
the incoming sound. Several researchers, including Waterson (1970, 1972),
Ingram (1974), Hawkins (1973), Macken (1979), and Wilbur (1981), have
called attention to the possibility that a child may not succeed at rst in getting
a complete picture of a word he has begun to learn. Therefore, we may be more
accurate in particular cases if we represent the childs knowledge of some part of
the words sound pattern by noise (Ingram 1974) or by underspecied
phonemes (archiphonemes, macrophonemes). These are useful notational
devices whenever we have reason to believe that, for example, a child has not
gured out what sounds are present in the unstressed syllable of a word or has
been unable to tell which of several fricatives a word ends with. To be more
explicit, these devices are useful notations whenever the child apparently cannot
distinguish perceptually among particular sets of similar words.
Note that we cannot rely on the childs pronunciation to let us know what
perceptual distinctions she is making, for children can in fact frequently tell the
difference between two words while they are still unable to pronounce either
one of them. ([Ronnie] Wilbur points out in personal communication that in
adults, cross-dialect phenomena continue to give examples of perception out-
stripping production: American Midwesterners who do not distinguish among
[Collection of percepts/understandings]
[Input lexicon]
[Output lexicon]
Figure 6.1.
174 Lise Menn
/,,e/ before /r/ in their speech nevertheless can reliably distinguish merry,
marry, and Mary in the speech of those who do make the distinction.
1
To give two simple examples of the use of these notations for incomplete
phonetic input information: suppose that a certain child appears unable to
distinguish between two words which differ only in the shape of a pretonic
syllable, such as along and belong, but that she can distinguish them from long.
Then noise marker would be appropriate to represent the rst syllable of
iambic words in the input lexicon.
Now suppose that we have a child who cannot distinguish /bs, b, bf/
from one another at an above-chance level in an appropriate test situation, but
who can tell them from /bt, bv/. Here, the child has some knowledge of the
nal sound of, say, bath, so we would not use a noise marker. Instead, we would
say that bath is entered as /b(unvoiced fricative)/ in the childs input lexicon.
So, what we have been saying is that the childs ability to use acoustic features
to discriminate meaningful words is typically well ahead of her ability to control
those features for making contrasts in production, but may well be inferior to the
linguistic discrimination ability of the adult. Some discrimination which the child
appears to make may in fact be carried out partly on the basis of extra-linguistic
information and linguistic context. For, like all of us, a childs ability to hear is
conditioned by her expectations of what she is about to hear. This factor is
important to emphasize for two reasons. One will be discussed in Section VI,
where we will explore some implications of Bartons (1976) work which shows
that unfamiliar words in minimal-pair tests of discrimination ability tend to be
misheard as familiar ones. This biases the tests and increases the difculty of
ascertaining what the childs input lexical representation of a word really is.
The other reason for bringing up the notion of the childs expectations is the
following phenomenon: Macken (1979) and Platt and MacWhinney (1983)
have argued that we sometimes have good evidence for the following sequence
of events. First, a child learns to recognize the sounds of a word adequately but
cannot produce it very well: we say that the input representation is good, but not
the output representation. Usually, the child will then slowly bring the produc-
tion into line with the target, but in certain cases, expected improvements fail to
occur in particular words or sets of words. The child maintains his old pronun-
ciation in such a way that it seems that he is no longer even trying to match the
adult model. Instead, it seems that he has replaced his original input represen-
tation with a new one which is based on his own output. For example, Macken
(1979) gives this analysis for certain events reported by Smith (1973). Amahl,
his subject, produced the word take as [geik] at an early stage, using a general
velar assimilation rule (a type of rule which we will shortly be discussing in
some detail). The rule stopped operating for all other words by Smiths stage
14, but Amahl retained a velar-harmonized form for take until stage 22, and
even created a participle [kukn] for taken at stage 18.
Now, if a child maintains his own form when he is capable of improving it, it
must mean that he has temporarily stopped monitoring, stopped really listening
Development of articulatory, phonetic, and phonological capabilities 175
to himself and/or to the adult model. He expects that he is correct, and does not
bother to check up. Indeed, many of us have adult acquaintances who have an
idiosyncratic pronunciation of some word, and who seem quite unaware that
they are not speaking as other people do. Many irregularities in childrens
phonological behavior thus seem to be explainable in terms of the biasing of
perception by expectation.
III.B. Segments, phones, and phonemic contrasts
Now we will consider the early stages of the production of proto-words and
words. Early child speech is often called pre-phonemic (Nakazima 1972;
Menyuk 1977). There are very good reasons for this. One is that phonemic
contrast and phonetic control do not develop in synchrony. One example of this
sort of uneven development can occur when a child honors a contrast without
being able to handle the relevant phonetics at all. So we may nd a child who
renders the voicing contrast in word-nal position by deleting voiced nal stops
and producing the unvoiced stops as a glottal stop. In such a case, for example,
the pair bead, beat would be rendered as the pair [bi, bi]. This hypothetical
child has preserved a phonemic contrast without being able to produce either
adult phone involved.
The converse case can occur as well: phonetic control can develop ahead of
phonemic contrast. It is very common for all initial stops, regardless of target, to
be produced by a child learning English as voiced (more precisely, to have
voicing onset time between 0 and 20 msec; see Macken and Barton 1980). In
such a case, the phonetics of voiced (short-lag VOT) initial stops could be under
control, but not the phonetics of unvoiced (long-lag) initial stops. One could
correctly say that the child at this stage had acquired the phones [b, d, g], but it
would be quite wrong to say that she had acquired the phonemes /b, d, g/ since
she does not have the contrast between them and /p, t, k/. (For further discussion
with examples, see Moskowitz 1975.)
The second reason why the concept of phoneme is difcult to apply in the
early stages of language development is that for many children, minimal pairs
(pairs of words differing only by the contrast in question) are so rare as to make
statements about the presence or absence of contrast impossible (see Itkonen
1977).
And the third good reason for calling early speech pre-phonemic is even more
linguistically unsettling. At least we can speak of phones in the rst case above,
and nothing prevents us from doing so in the second case. That is, we appear to
have phonetic targets which are comparable to one another, independent of the
lexical items the particular words in which they are located. In adult
language, we expect that any difference between, say, an /a/ in one word and
an /a/ in another will be completely due to the sounds surrounding them, the
stress pattern, and possibly to some kinds of morphosyntactic factors (e.g.,
being used as a clitic) or more social factors (formality, rate). We are not
176 Lise Menn
prepared to see arbitrary variation in phonetic targeting between one lexical
item and the next. Yet it does happen; it even occurs in adult language in special
marginal cases.
Let us rst consider a special case in adult English where a segment fails to
satisfy the criteria for being a phone. The o of no is subject to a huge amount
of variation in realization because of the expressive roles it plays; it can occupy
almost all positions in the English vowel space below a diagonal from [] to
[o], including for example [, a,
]
nice [naj]; house [w]; dog [da]
please [pi]; blue [bu]; clock [ka]
drum [d]; up []; down [dw]
This child would appear to substitute glottal stop for nal /t/ and to delete other
nal consonants. Initial /h/ is also deleted. Liquids are dropped from consonant
clusters. These statements may be translated into formal terms like this:
t [] (/t/ becomes glottal stop and then all other consonants
are deleted word-nally)
C |_#
[+cons] (liquids are deleted from initial clusters)
[+ voc ] |#C_V
/h/ |_# (h is deleted word-nally)
The reader may have noted that these four rules are not the only ones that can be
devised to describe the observed behavior. It is important to understand that in
most cases we do not get enough different words from a young child to
determine her set of rules fully. Rules are always to be regarded as the analysts
tentative hypotheses about the childs mental operations. And it is also impor-
tant to remember that a rule is no more than a description of a hypothesized
regularity of behavior. It is not an explanation of anything to say that a child
has a deletion rule or a substitution rule, just as it is no explanation to say that
an apple falls because of gravity.
Development of articulatory, phonetic, and phonological capabilities 179
Now let us examine in more detail two of the best-known rule types of child
phonology: assimilation and voicing/devoicing. Towards the end of this section
we shall also see that there are other strategies that children use which produce
the same effects that these rules do.
(a) Assimilation and consonant harmony: assimilation rules. We often notice
that young children have rules which change the consonants in a word to make
them more similar to one another. As in general phonology, these are called
consonant assimilation rules. For example, a child who can say daddy with a
good initial [d] and egg with a good nal [g] may yet say [gg] for dog. Such a
child usually also says [gk] for duck and truck, etc. These rules may be so strict
for a time that all the consonants in any given output word must be homorganic
that is, made with the same position of the articulators. Boat, for example, would
have to be produced as either [bowp] or [dowt].
Assimilation involving the feature [nasal] is common, too, in child phonol-
ogy: dance may become [nns], with the [d] assimilating in nasality to match
the following nasal; or meat may become [dit], with the [m] losing both its
nasality and its labial position as it assimilates to the nal [t]. (Both of these
forms are from Daniel; Menn 1971.)
Sometimes a child may produce some non-harmonic sequences and yet
apparently require harmony in other words: he may say gate correctly, but
produce [gg] for big and [gejk] for take. In this case, the assimilation of labials
or dentals to velars occurred only if the velar was word-nal; if it was word-
initial, both stops were produced correctly. Relative position of the consonants
in a word is often a factor when some sort of asymmetry of consonant harmony
is found (Ingram 1974). Vihmans (1978) survey suggests that sounds at the
beginning of a word are somewhat more likely to be the ones which are changed
when there is an assimilation rule, but this is merely a tendency.
Assimilation rules can be found in great numbers in adult language as well,
but there is an important difference. In adult language, the usual type of
consonant assimilation is contact assimilation: a segment changes and becomes
more like one that is next to it. Although many adult languages have vowel
harmony, which occurs even when consonants lie between the vowels, very few
adult languages have consonant assimilation at a distance; Vihman nds it in
only 3 of the 88 languages in the Stanford Phonology Archive (not including
some cases in which the intervening vowel is colored by nasalization in nasal
harmony, or by pharyngealization in pharyngeal harmony; these cases are
called prosodies by Vihman). Something special is taking place in child
phonology.
When we nd deletion rules, as in our initial examples for this section, or
contact assimilations like the change of ask to [st], we usually feel that
mechanical ease of articulation should account for them. But when we
contemplate distance assimilation, we nd our intuitive notion of simplicity
challenged. Why should a child who can say dad and egg nd [gg] easier than
[dg]? Is this to be explained in natural terms? In a sense, yes, but in terms of a
180 Lise Menn
different kind than we have previously considered, terms which are very
important to the construction of a theory of child phonology.
In trying to understand distance assimilation, we can get some help from
considering general motor behavior. Under what circumstances is an ABA
pattern of behavior easier than an ABC pattern (assuming that A and C are
equally easy to carry out in themselves and as sequels to B)? The only way that
doing A again can be easier is if the sequence is to some extent preassembled
or preprogrammed, for in a memoryless series of events it would not matter
whether an element is one that has recently been used. In other words, doing
A the second time is easier than doing C only if we know that we are going
to do A again and can make use of that information. So the argument goes as
follows: young children often use distance assimilation. We take as a working
assumption that this must make words easier for them. It cannot make words
easier for them unless there is a stage of production at which a word is
programmed or assembled before it is spoken. Therefore, I think that a model
of howwords are produced by young children must have such a stage in it. Later
on, we will come back to this point and try to deduce more about the properties
of this stage from the data that we have available.
(b) Other strategies: consonant harmony as a goal. Assimilation rules are not
the only way that children deal with disharmonic sequences. Some children
omit one of the offending consonants: Daniel, who used assimilation on dog,
boat, and a good many other words, said [gej] for gate, rather than [gejk] or
[dejt]. Other children use a glottal stop in place of one of the adult sounds. Such
patterns of rule use linked by similar input and similar output strongly suggest
that we should take a functional approach to child phonology rules; that is, they
make more sense if we think of themas means to some end. And in fact, we have
been doing just that: we have been assuming that these rules are somehow
designed to eliminate disharmonic sequences.
III.C.3. Output constraints and conspiracies: rst mention At this point
it will help to develop some terms for dealing with sets of rules which appear to
serve some common function. Suppose none of the forms produced by a child
contain consonant clusters, for example, or that none have nal stops, or that
none have disharmonic sequences. A statement that a particular sound pattern
does not appear in a corpus and is not expected to appear if we get a larger
sample is a statement of an output constraint. Adult languages have output
constraints as well; consonant clusters are absent from many languages, and
every language has restrictions on how many and what kind of consonants form
a pronounceable cluster (Bell 1971). Vowel harmony, present in quite a number
of languages, is also describable as an output constraint.
Following Kisseberth (1970), when we have a set of rules that all contribute
to eliminating sound patterns which would violate a particular output constraint,
we say that those rules form a conspiracy. In the example from Daniel,
Development of articulatory, phonetic, and phonological capabilities 181
assimilation rules and a (limited) deletion rule were part of the conspiracy to
eliminate disharmonic sequences.
Conspiracies of rules are not the only devices that children use to maintain
output constraints, however. Selection strategies may also contribute: children
may avoid adult words which violate a constraint. Sometimes, this may be a
very minor strategy for a particular child (Daniel probably avoided the word
cup), but sometimes it is a major contributor to the maintenance of an output
constraint.
Let us now look at some cases involving another very common output
constraint in young children. This one actually involves a pair of phenomena
collectively referred to by Ingram (1976) as voicing: the constraint that initial
stops be voiced and nal stops be unvoiced. A child may have only one of these
or neither, but the pair is very common for English-learning children.
At the acoustic-phonetic level, the statement is slightly different: initial stops
tend to be voiceless-unaspirated (short-lag VOT) and nal stops to be partially
devoiced (see again Macken and Barton 1980; N. V. Smith 1973; B. Smith
1979). This difference in statement is not important within English phonology,
but it becomes very important cross-linguistically, since voiceless unaspirated
stops count as voiced in English phonology, but as unvoiced in Spanish,
French, and many other languages. An explanation for this pair of phenomena
should be in terms of the regulation of glottal airow for discussion see Flege
and Massey (1980) and Westbury and Keating (1980). If there is any rule which
deserves to be called a natural process, surely it is the rule of nal devoicing: it
is not only found in child language, but is one of the most frequent rules in adult
language, appearing in many forms from a low-level tendency (as in American
English) to the familiar German and Russian nal devoicing rule and Turkish
syllable-nal devoicing.
So, many children use the natural-process rule of devoicing nal stops, and
many also use the natural-process rule of voicing initial stops; Joan Velten is
undoubtedly the best-known example. She said [bat] for pocket, [ba] for pie,
[bat] for bad, [ap] for up, and [zas] for sauce, to choose from a long list (Velten
1941: 867 ). There are no examples involving velar stops in output, for at this
age (23 months) Joan changed all adult velars to coronals (except for [bup],
book). Other children who have the same voicing constraint may use a selection
strategy: words beginning with /p, t, k/ or words ending with /b, d, g/ may be
avoided, and words which begin and end with the preferred sounds may be
selected.
Now let us look at a more complicated case, one in which all the three
principal stop positions of English were being produced by the child. Here the
voicing constraint is in full force in nal position: nal [p, t, k] have been
mastered, while the nal voiced stops /b, d/ are avoided, and nal /g/ is modied
by being devoiced or deleted.
The constraint has been overcome in initial position: the contrast between
initial /d/ and /t/ has been mastered and initial [k] has been acquired. Initial /p/ is
182 Lise Menn
avoided, but so is initial /g/. (Ferguson 1975 has commented on similar asym-
metries of consonant distributions in child phonology and across adult lan-
guages.) These statements are summarized in tabular form (Table 6.1). Another
important point is exemplied by these data; notice that the voicing contrast has
been mastered for initial dentals, but not for initial velars or labials, and that in
this case we cannot even say that one value of the feature is present for all three
initial stops. A feature that has been mastered (in either the control sense or the
contrast sense) in one phoneme may or may not spread to other phonemes in the
same word position. We presently do not know whether it is possible to explain
the difference between the cases in which a feature generalizes and the cases in
which it remains bound to a particular phone.
Other rule strategies besides the use of voicing or devoicing rules can be
found in children obeying the voicing constraint. We have just mentioned
Jacobs occasional deletion of nal /g/, but there are much more interesting
cases to be found. These are the children who add extra segments in order to
render a voicing contrast. It has been claimed that some children add a vowel to
the end of a word with a nal voiced stop; this brings the sound into the interior
of the word where it could be managed. Bag might be produced as [bg] or
[bg].
Also, two cases are now reported in which children added nasals rather than
vowels in their apparent efforts to preserve the voicing contrast in nal
position. Fey and Gandour (1979) presented a study of a child who found
that he could preserve the voicing of adult nal stops by adding a nal
homorganic nasal: bag became [bg]. (Phonetically this is rather less exotic
than it looks written out; the effect is just produced by releasing the velar
closure before releasing the stop articulation. However, this cannot well be
considered a natural process; there is no evidence that there is a general
tendency for speakers attempting to maintain voicing through a nal closure
to fail with this result.) Clark and Bowerman (1986) report a different use of
added nasal segments: one of her daughters added a homorganic nasal before
nal voiced stops, so that for example Bob became [bamp]. The stops them-
selves were still devoiced, but contrast was maintained (and the insertion of
the nasal should have helped to maintain the vowel-lengthening which pre-
cedes nal voiced stops in English and which in fact serves to carry the nal
voicing contrast in some dialects).
Table 6.1. Jacobs consonants
Initial Final
p absent b mastered p mastered b absent
t mastered d mastered t mastered d absent
k mastered g absent k mastered g devoiced or deleted
Development of articulatory, phonetic, and phonological capabilities 183
Now that we have seen how the notion of output constraint can serve to bring
together several rules and/or strategies under the observation that they all serve
to maintain the same output constraint, it is time to take a critical look at the
notion itself. So far, all we have is description, not explanation. To say that a rule
serves an output constraint or is part of a conspiracy is only organization of
data. But once we organize the data in this way, a plausible explanation jumps
out at us: the child is modifying unfamiliar sound patterns to make them like the
ones he has already mastered. And that means that the child has to learn sound
patterns, not just sounds. Again, output constraints are only descriptive devices;
what they describe are those sound patterns which a child has mastered vs. those
that he has not. That is why words which do not t the constraints are almost all
avoided or modied. This is the central thesis of this chapter; we shall explore
its empirical support and its implications in many of the remaining sections.
III.C.4. Another modication strategy: template matching Now let us
consider another type of modication strategy, one evidenced primarily in work
done by Vihman (1976, 1981), Macken (1979), and Priestly (1977). These cases
involve fairly violent rearrangements of sounds of adult words to match tem-
plates of preferred sound patterns. The simpler cases can just as well be consid-
ered cases of rule use, and usually are described in terms of metathesis (place-
exchanging) rules. The more complex cases, however, cannot be described by
rules without a lot of articial special-case magic, for what makes them so
complex is the fact that the childs attack on the adult word is not fully systematic.
A good simple case to begin with is Vihman (1976). A child learning Estonian
as her rst language seemed to have learned to say words containing two different
vowel sounds only if the rst vowel was lower than the second. The Estonian
words for mother, /ema/, and for father, /isa/, do not happen to followthis pattern.
For a little while, the child said just [sa] for father; then for four months she
failed to attempt either word, although both father and mother made earnest
attempts to elicit the words /ema/ and /isa/. At 15.5 months, the child began to
rearrange those words to conform to her output constraint: /ema/ emerged as
[ami] or [ani] . . . at which time /isa/ also reappeared, now pronounced [asi], and
the word /liha/, meat, was reproduced, following the same rule, as [ati].
An example of a case where the child was less systematic about the map from
the adult word to output is given in Priestly (1977) (also discussed in Ingram,
1979). Priestlys son Christopher treated virtually all stop-nal adult two-
syllable words and a fair number of vowel/sonorant-nal two-syllable words
according to the following patterns: Consonant selection:
C
1
C
2
C
1
j C
2
examples: pillow [pijal]; Brenda [bajan]; tiger [tajak]
or
C
1
C
x
C
2
C
1
j C
2
examples: rabbit [rajat]; melon [majan]
184 Lise Menn
with a few cases of idiosyncratic rearrangements, such as streamer being
produced as [mijat]. There was also a choice of vowel treatments; sometimes
Christopher was able to match two vowels of the target, but at other times he
replaced one or both by [a]. In addition to the cases already listed, consider the
apparent metathesis of vowel features involved in his rendition of woman as
[wajum]!
Other two-syllable words which ended in a vowel or sonorants were treated
without these special medial-[j] rearrangements: examples are bacon, produced
almost correctly as [bejkan], kitchen, where the medial affricate apparently
caused the only problem, rendered [kkn, ktn], and scissors, [szz].
While it is possible to discern some tendencies in Christophers assignments
of particular adult forms to particular outputs, Priestly makes it clear that there is
considerable arbitrary variation fromword to word. This fact of lexical variation
is further emphasized by Christophers variation across tokens of the same
word: monster was recorded as [majs] in weeks 4 and 6 of the study, but as
[mjan] in week 5; dragon was given as both [dajan] (week 3) and as [dajak]
(week 4).
In Priestlys case, then, the child had a favorite output shape to ll, but only a
few constraints on which consonants and vowels he picked to ll it with.
Mackens 1979 subject Si, acquiring Spanish, shows us a much more con-
strained output template that is, one which allowed a very limited set of
consonants and a much greater abandon in her treatment of the model word.
(The latter fact probably also reects the much greater proportion of polysyl-
labic words among her targets.)
Si could produce disharmonic sequences in a word only if one target con-
sonant was labial and another was dental. Adult words which met this criterion
were produced so that the labial preceded the dental; much deletion and occa-
sional metathesis occurred.
examples: manzana [mana] pelota [patda]
zapato [patda] elefante [batte]
Fernando [wanno] sopa [pwta]
In Sis case, the details of what is deleted and what is selected defy organized
statement in terms of rewrite rules. As Macken says, this is goal-directed
behavior: the child is looking for consonants that she can t into her output
template and ignoring the rest.
IV. Rule creation
IV.A. Extending the notion of ease of articulation: one key to a newtheory
When a childs production of a word fails to match the adult model, we cannot
help assuming that there must be some sense in which what he does produce is
easier than what he has failed to produce. But what sense is this? How can
Development of articulatory, phonetic, and phonological capabilities 185
[bada] be easier for Mackens Si than [daba]? Why will some children use [l] for
/j/ and others use [j] for /l/? Why do some children exploit fricatives while others
delete them, avoid them, or replace them with stops? Clearly, if we stick to our
commonsense starting assumption, then it must be the case that what is easier
for one child can be harder for another. Perhaps a little of the variation is due to
anatomical differences, but we simply do not have the means to investigate that
hypothesis. A much more fruitful approach is to assume that a great deal of
ease and difculty is not a matter of physiology at all or, to put it another
way, that physiological causes are only one factor in determining ease of
articulation for the individual child. The other factor, and I propose that it is
the major factor, is the state of a childs knowledge at a given time.
Let me give an example. A child may, as we have said, discover how to say
[l] before how to say [j], or the reverse may be true. Suppose a particular
child has discovered [l] rst, by chance. We notate this discovery as the
invention of a rule taking /l/ into [l]. Now this child may slip into her [l] while
trying to say [j], either accidentally or on purpose. If she nds the approximation
good enough, she will continue to use it: she will have thus discovered or
invented a modication rule. Again, in this case, [l] is easier than [j] only
because this child happens to have found out how to make an [l] rst.
I suggest, in short, that a two-stage discovery process is probably involved in
a childs establishment of a new articulatory gesture as her way-of-saying a
particular target sound. The rst stage is a matter of trial-and-error attempts to
match the sound sequence; the second stage is one of deliberate or accidental
overgeneralization of the success of that articulatory gesture, that is, the use of it
to render similar adult targets.
Let us consider the hypothesized scenario here in more detail, for it is the
heart of this chapters proposal for dealing with one of the fundamental prob-
lems of child phonology, namely, howcan there be so much individual variation
and yet such strong general tendencies? We suppose, then, that variability
across children originates with each child making trial-and-error starts at match-
ing adult sound patterns. For each given sound or pattern, some children will
succeed and some will fail. External factors, such as the frequency and
salience of the sound in the speech of others, may contribute to the likelihood
of success; so will internal factors: the probability of accidentally hitting on an
acceptable way to produce it and the salience of the sound in ones own speech.
We frankly do not know why some sounds are more probable than others;
Stevens (1972) notion that favored phones are those which are acoustically
stable (i.e., permit a certain sloppiness in articulation without showing appreci-
able acoustic change) is certainly an attractive idea, but we cannot yet simulate
the childs vocal tract accurately enough to test this idea with acoustic modeling.
(However, progress has recently been made in this area see Goldstein 1980.)
The accidental aspect of learning to produce target sounds is a principal source
of individual variation, but it is also a principal source of the probabilistic
universals of order of acquisition; roughly and with all due caveats, stops
186 Lise Menn
usually are acquired before fricatives, labials usually before velars, nasals
usually early, liquids usually late. (See Sander 1972, both for data on English
and for methodological considerations.) If the reader will permit me some
licence in the statement of probabilities, we might say that a [b] is a low pair,
[k] is jacks or better, [l] is a ush, [] is a straight ush, and the fricative [r ],
which Jakobson dwelt on as the latest acquired Czech phoneme, is a royal ush
in spades: some kid somewhere in Czechoslovakia is going to get it phonetically
right in her rst ten words, but dont bet on her being in your data sample.
We should stress one more thing about this proposed initial trial-and-error
stage of discovery: a child may accept her rendition of a sound even when it is
quite inaccurate. Some rules that give inaccurate renditions of adult targets
therefore arise at this rst stage. But many more may arise in the second stage, as
the child makes use of her initial accomplishment.
IV.B. Natural processes
It is quite reasonable to say that both /l/>[j] and /j/~ [l] are natural phonetic
processes, in that articulatory factors make it quite likely that a clumsy attempt
at either of themwill produce the other, rather than, say, a [t] or a [b]. Put another
way, a child with a certain amount of experience at making speech sounds with
his mouth is likely to get some of the properties of, say, [l], correct (in a word
that does not present a host of other problems): perhaps the voicing, the
continuancy, the central tongue placement, or the lack of rounding. [l] and [j]
share all of these properties, so a child who is doing well at approximating one
of these two phones is quite likely to end up with the other as his approximation
to it.
Informal observation suggests that [l] and [j] are roughly equally likely to be
found substituting for one another ([w] or a similar sound is also found
frequently for dark L [], of course). In other cases, there is a heavy bias in
favor of one of a pair of phones. For example, in word-initial position, stops are
much more likely to be discovered before fricatives and then to be used to
substitute for them. Similarly, voiced stops are likely to be used for unvoiced
stops in initial position, as we have already seen. We certainly have enough
reason to say that stopping (use of stop for fricative) and voicing are natural
in initial position; that is, we have reason to believe that there is a high,
physiologically governed probability that the child making a rst attempt at
an initial fricative will produce an initial stop, and that the child rst attempting
an initial unvoiced stop will produce a voiced stop instead. This, I think, is the
only coherent interpretation of the notion natural process, although other
views certainly appear to be held (see Stampe 1969; Ingram 1976, but also
Ingram 1979).
In summary, I propose that natural processes are really descriptions of
those pitfalls of learning to articulate which are commoner and more heavily
determined by physiology. To build a rigorous theory of the acquisition of
Development of articulatory, phonetic, and phonological capabilities 187
phonology, one must also be able to explain why children fall into those
particular pits.
And that step would still be only a beginning, for physiology only dictates what
articulatory goals are likely to be surrounded with what traps. To explain how
children succeed in avoiding or climbing out of them, we need a problem-solving
theory, a cognitive theory. The essence of such a theory for the acquisition of
phonology, again, is the trial-and-error discovery followed by application of the
discovered skill to new cases a model which will be very unsurprising to any
developmental psychologist.
IV.C. Non-natural rules
There remain some kinds of rules that are at a considerable remove from the
solution of particular articulatory problems.
Avery important kind of non-natural rule arises as the child begins to attend
to the fact that what appears to be the same morpheme is not always produced in
the same way by adults. Sometimes that child is correct in interpreting her
observations this way that is, sometimes she has indeed run into a case of
allomorphy or of stylistic variation. However, sometimes she is incorrect; what
appears to be variation in the shape of a single morpheme is in fact a case in
which the adult is sometimes using one morpheme and sometimes using two
which the child has failed to segment. For example, if a child notices the
Z-morpheme of the English possessive and plural appearing on certain
nouns but does not yet understand that the nal sibilant has one or both of
those meanings, he may develop his own phonological hypothesis about
where those nal sibilants are supposed to appear. Daniel (Menn 1971) created
a rule adding [s] to the end of all English words ending in /r/, apparently because
there was an accidental abundance of plurals and possessives on names and
objects in /r/ in his environment. He may have gured that the sibilant-nal
forms which he heard were the full and correct forms of the words which he also
heard with nal /r/ that is, he took pears as the full form of pear, Peters as the
full form of Peter, etc.
It is also the case that rules which once had an articulatory base, after they
have been invented, seemto acquire considerable autonomy and may generalize
without any further articulatory motivation. A child may apply a rule for one
segment or (sequence of segments) to a similar one even though he could have
produced the latter correctly. This seems to be the case for several rules used by
Amahl (see Smith 1978). Rules are much more than articulatory habits, then;
they are transduction habits, habits of rendering perceived targets in particular
ways. Illustrations and further discussion will be presented in Section IV.E,
Overgeneralization.
It is too early to make strong generalizations about the ages at which trans-
duction rules of different kinds can be found, but roughly, it seems that the very
youngest childrens rules are mostly those which lend themselves to
188 Lise Menn
explanations in terms of seeking solutions to articulatory problems; as these
problems are overcome, we begin to see more instances of rules that arise from
overgeneralizations of other rules, and more rules which reect the childs
guesses about the reasons for variation in words of the adult language.
IV.D. Rule origin and growth
We have already found ourselves considering the topic of rule origin; let us now
do so in more generality and depth. We have characterized transduction rules as
systematic correspondences between adult and child sound patterns, ranging
from correct renditions (/d/[d]), omissions, and natural substitutions (//
[d]) to the idiosyncratic rule inserting [s] after word-nal /r/ that we have just
discussed. There is also a range in how systematic a rule is. Some are excep-
tionless; most have a few lexical exceptions which typically consist of forms
that were learned before the child invented the rule in question, or of forms
which are the forerunners of a new rule. And some rules have so many
exceptions that they reach the point where we are better off abandoning the
attempt to write them; the Priestly case was one example of such a state of
affairs.
The evidence for the nature of rule change is somewhat sketchy, because rule
changes can take place in a short time, sometimes within a few hours. Fine-
grained longitudinal study is needed to give a picture of Before, During, and
After in such cases. This is emphatically not to say that all rule change is rapid.
Replacement of one well-established rule by another may take place over a
period of weeks (and fossil forms created by the old rule may survive
indenitely).
IV.D.1. Rule origin We have already discussed trial-and-error experimen-
tation as a source for correct transduction rules (/d/ [d]) and for natural
transduction rules. But it should be noted that a childs trial-and-error sessions
do not always lead to the formation of a rule. Even if the child manages a perfect
rendition of some sound pattern, she may be unable to capture the trick of doing
it at will. For example, Daniel (Menn 1971) made dozens of attempts at the
word peach during the period when his consonants were subject to assimilation.
If he had been able to make the beginning of the word affricate to match the
end, he presumably would have had no problem. But he had not learned to
produce any initial affricates, and his versions of the word included [dits, cit,
nits, its, pip] and [pit] itself at various times. He settled on none of them.
Yet sometimes a rule actually emerged within hours: Daniel tried [as] and
[dts] for box at 10;16, and later the same day his assimilation rule made its rst
true appearance, with dog as [gVg], a formit kept stably for months (as far as the
consonants were concerned).
The other case of rule origin in the literature has been called consolidation
(Menn 1976a). This term is used to describe the situation in which two similar
Development of articulatory, phonetic, and phonological capabilities 189
adult target sound patterns are involved in very similar trial-and error sequen-
ces, and end up being handled in the same way. Correct versions of both of the
patterns may be produced in the course of the trials. Jacob varied between [ei]
and [i] for the vowel of both tea and table for some weeks before settling on
[i] for both. The mutual inuence of similar sound patterns is clearly demon-
strated in such cases. Template matching can also originate in this fashion see
Vihman (1981).
IV.D.2. Rule generalization Rule origin can occur through rule general-
ization, for of course dividing a rule from its predecessor is often difcult or
arbitrary there is often no sense to the question is this a new rule or a
generalized version of an old one? Rule generalization basically means the
extension of a rule to new cases, and this covers two different kinds of events.
To discuss them, we need the concept of the domain of a rule. The domain of a
rule is simply the set of cases to which it is actually applied. For example, the
domain of a rule that applies to all English voiced obstruent is just the set of all
instances of /bdgvz/.
Formally, if we have an exceptionless rule, its domain is specied in its
structural description. In the example given, the structural description could be
written [ + obstruent, + voice ].
If a rule has lexical exceptions, sounds in the excepted words are not in its
domain even if they meet its structural description. Thus, if the word bad were
simply listed as a lexical exception to a rule otherwise applying to all voiced
obstruents, the /b/ and /d/ in it would be outside the domain of the rule. If, at a
later time, bad ceased to be an exception, it would by denition have been
brought into the domain of the rule and, thus, the rule would have become more
general without any change in its structural description at all. We might termthis
type of rule generalization lexical smoothing. Lexical smoothing is important
in child phonology because lexical exceptions to rules are so frequent. Yet it is
not really a change in the rule; it is only a change in the set of exceptions to it.
The other type of rule generalization is formally expressible as a relaxation of
the structural description, allowing additional phonologically dened sets of
words to be operated on by the rule. For example, a rule which at some point
applies only to nal /b/ might at a later time apply to all nal labials, or to all
nal obstruents, or to all instances of /b/. Any of those changes would bring new
sets of sounds into the domain of the rule, thus generalizing it. A relatively
technical note: in child phonology, we often have trouble determining the
domain of a rule for various reasons. Here is one interesting problem: consider
the data fromJoan Velten given above (Section III.C.3). She had no velars in her
output; she had initial voicing and nal devoicing of other stop consonants.
Should velars be considered to be in the domain of the voicing and devoicing
rules? It is easy to write the rules either way (with voicing and devoicing rules
applying directly to all stops before the conversion of velars to dentals, or with
fronting preceding voicing and devoicing). Only in the latter order can the
190 Lise Menn
voicing rules be written excluding velars and still give us the observed distri-
bution of forms. Now the fact is that when velars show up, they may not be
subject to either of the rules obeyed by the other stops, so it is preferable to write
the rules the second way, and thus to make no vacuous claims about the velars.
If the velars do show up obeying the voicing and/or devoicing rules, that would
then count as a generalization of the two rules.
IV.E. Overgeneralization
Just as in the acquisition of morphology or syntax, rule generalization can create
incorrect forms, and thus, from the adult point of view, be overgeneralization.
The term is used loosely; typically it is used when a rule produces some good
results and some bad ones. If a rule always produces modied forms (bad
results), we do not bother to call extensions of it overgeneralizations except
when they make a childs approximations worse than they were before the rule
affected them. Let us consider some examples.
Daniel (Menn 1971) had the two words down and stone rendered as [dn]
and [don] fromthe time of his rst attempts at them. Then he developed a rule of
nasal harmony he made all of the stops in a word nasal if the nal stop was
nasal. Down and stone remained lexical exceptions to this rule; that is, after he
had been saying [nns] for dance and [ein] for train for two weeks, he still
maintained the two older words in their unassimilated form. Eventually, how-
ever, there was a period of time in which he varied between [nn] and [dn]
for down, and between [non] and [don] for stone. Finally, the assimilated forms
for these two words took over completely and they were no longer lexical
exceptions to the rule. From the adult point of view, these two words were
poorer approximations to the adult model after the rule had been applied to them
than before (indeed, down had been perfect). Therefore, the generalization
involved in extending the domain of the assimilation rule to include down and
stone (a case of lexical smoothing, to use the term introduced above) is an
overgeneralization of the assimilation rule.
A change in the structural description of a rule can also produce overgener-
alization (recidivism in N. V. Smiths terminology). Here is his example from
Amahl (1973, 1523): At stage 1, /s/ and /l/ were normally neutralised as [d],
together with all the other coronal consonants . . . (I omit his description of
exceptions to this rule, which generally made coronals into [d].) Then /l/ began
to appear in As speech before any coronal consonant for example, lady was
rendered either [d
, n
, l
skt {b + V
low
+ k + } bajak
(38) blkt {b + V
low
+ k + } bajak
(13a) tjg {t + V
low
+ g + } tajak
One idiosyncratic strategy in the acquisition of phonology 227
(15) t
kij {t + V
low
+ k + } tajak
(25) k
bd {k + V
low
+ d + } kajat
(26) k
vd {k + V
low
+ d + } kajat
(31) fwntn {f + V
low
+ n + } fajan
(42a) nl {f + V
low
+ n + } fajan
Note that the C
n
for (13a), and (256) is set up on the understanding of normal
obstruent-devoicing in nal position; and that these UFs (enclosed in braces) are
minimal estimates the UFs for (38) and (42a) may well include the [1] in the
initial cluster, for example.
Coincidences and reversions
In discussing strategies, above, we postponed analysis of relationships between
two output forms for the same input form. We now suggest that only under
certain chronological conditions may a direct relationship between two such
forms (and identity of their UFs) be postulated. We impose the following
limitation on the time-factor: if an experimental and a non-experimental form
were observed during the same week, they are treated as contemporaneous
(coincidences); and if C used an experimental form subsequently to using a
non-experimental form, it may be said that he reverted to his experiment
(reversions). In both cases, a direct relationship, and UF identity, are postulated.
In each case, the pairs of forms are inspected for minimal shared content, as in
the preceding section, and the minimal UF for the two forms is estimated as
follows:
Coincidences
(32) BEF [fajam], ORF [f
m] {f + V
low
+ m + } (W6)
(57) BEF [majs], ORF [m
?)
Here, u (the velarization of the lateral being incorporated in the V, cf. the
ORF [wisu]) giving w + + u + s; after glide-insertion, [wijus].
(2a) (Equation AII): [p
low] {p+ + 1+ }
Here, the neutral vowel a, giving p + + a + 1; after glide-insertion [pijal].
(27) (Equation AIII): [kwst] {k + ow + s + }
For all Equation III items, (which is realized as the neutral vowel in this
instance) is inserted before the V
1
of the UF: k + a + ow + s; after glide-
insertion, [kajows].
(56) (Equation BI): [m
sk
ks] {m + V
lowround
+ ks+ } ( = )
Here, the UF may well be represented as including a fully specied []; we give
only the minimal content of the UF. is inserted after V
l
, giving m + V
low round
+ + ks; after glide-insertion (and, presumably, adjustment of the V
1
) [majks].
In this case, other routes suggest themselves, e.g., the [ks] of the output may
perhaps correspond to the medial [sk] of the input, and metathesis may be
involved.
(60) (Equation BII): [rkdz] {r + + z + }
Here, the neutral a, inserted after V
1
: r + + a + z; after glide-insertion and
nal devoicing, [rjas]. The UF may in addition include the [d] of the input,
which is elided in the ORF for this item.
230 T. M. S. Priestly
(70) (Equation BIII): [
rnd] { + + d+ }
The UF consonant is suggested arbitrarily. Any form from the complete cluster
[nd] to the simple [d] would surely have the same outcome: + a + + d [ajt]
(22) (Equation X): [krt] {k + + t + } and
(61) (Equation Y): [rbt] [r + + b + }
For the MEFs, a component of length, which can be represented as a
reduplication of the V
1
: k + + + t [kt] and, with nal devoicing, r + +
+ b [rp],
(11) ORF (Equation Z): [b
skt] {b + V
lowround
, + sk + }
The a is inserted, here and in (68), between the two consonants: b + V
lowround
,
+ s + a + k [bsak].
A great deal of the above is suggested very tentatively, and some of the
formulations are arbitrary. In particular, the glide-insertion might well be replaced
with another formulation. Note also that some reference to stress may also be called
for; see the next section. Other examples could be adduced: thus (44), with
consonantal metathesis, and (58b), with vocalic metathesis, provide interesting
exercises in formalization; as does (55), where the nal [] of the output seems to
represent a fusion of the [g] plus the [n] of the input. (17) is especially puzzling: the
expected BEF, given the available ruses, would surely be *[dajowt] or *[dowjat].
Conclusion
Stress
It was pointed out, in discussing the consonant equations, that while the analysis
was concerned with syllable position, the stress of the vowels might in fact be
the deciding factor. Since we are dealing with English, most of the forms have
the stress on the rst vowel; the only exceptions are the rather uninformative (8),
(12), (16) and (62). It must therefore be emphasized that the canonical UF
for the experimental forms may well with equal validity be formalized in terms
of vowel stress rather than vowel position, thus: C
1
+ + C
n
+ . For arguments
concerning the importance of stressed syllables as salient portions in child
language, see, e.g., Kiterman (1913), vakin (1948: 103), Ervin-Tripp (1966:
71), Blasdell and Jensen (1970).
Noticeable consonants
The notation C
n
was used above to refer, rather vaguely, to what is called the
most noticeable consonant. The implicit assumption is that whenever C made
a choice between medial and nal consonants of the input for his canonical UF,
One idiosyncratic strategy in the acquisition of phonology 231
some kind of non-random choice was involved. It could be suggested that C
would, for example, choose the acoustically most noticeable consonant; or that
he would choose the one which was phonologically most marked; or that he
would choose the one he had learned best. Other phonological factors may also
be involved; the following remarks are therefore tentative. If uninformative data
are ignored (words with no nal consonant, and hence no choice; words with
similar or identical medial and nal consonants; and words with medial liquids,
where substitution may be involved), 21 items remain. Of these, 4 show
vacillation between Equations A and B. Of the remainder, 9 follow Equation
A, and 8 followEquation B. The facts may be summarized as follows (here, > is
used to mean is chosen in preference to): Equation A: s > n; sk > t; z > nt; k >
t; s > l; m > n; ts > n; z > k; ks > nt . Equation B: t > n (3 times); 1 > nd; d > b; d >
v; k > d; dz > k. This shows that, in general, consonant clusters are more
noticeable than single consonants; and that the strident fricatives are more
noticeable than other single consonants. The evidence with regard to other
parameters is negligible or contradictory, although nasals appear to be generally
less noticeable than other consonants.
The syllable as a basic unit
The data presented and analysed here give strong support to the position that the
syllable should be regarded as the basic unit in phonological acquisition, if not
also in the phonology of those for whom acquisition is complete. Here the
ambiguous position of the ubiquitous [j] must be discussed (see data section,
above): is this sound, as it occurs in all the BEFs, to be regarded as an offglide to
the rst vowel, or as the onset to the second syllable? The evidence appears
contradictory: on the one hand, the glide clearly seems to have developed
(in its use as the automatic medial in the BEFs) from a substitute for liquids
(see the discussion of substitutions, above), and thus to have a clear consonantal
origin. On the other hand, comparison of the MEFs with the BEFs suggests that
the glide is an automatic insertion to break the hiatus between the two vowels
and hence that it was not a component of the UFs at all (see the discussion of
strategies and ruses, above). Whatever its role, however, it is clear that the most
fundamental part of Cs strategy was the recognition of the polysyllabic nature of
the input and the manifestation of this fact as a bisyllabic form or, infrequently,
as a monosyllabic form with extraordinary length. We suggest that the rst
opposition learnt by the child, i.e., that between C + V and zero (implicit in
Jakobson 1941), is the foundation for a syllabic learning of phonology: when the
child perceives the importance of the contrast between monosyllabic and bisyl-
labic forms, he learns to produce the extra syllable in some fashion or other by
reduplication in some cases, by a fair approximation in others, and/or by resorting
to an unusual strategy, as in Cs case. Arguments for this syllabic viewpoint date
back at least to Otuszewski (1897: 201), and have been renewed recently by,
e.g., Mike and Vlahovi (1967), Ladefoged (1967: 149), Moskowitz (1970:
232 T. M. S. Priestly
43941; 1972: 515, 602), Menyuk (1971: 5971; 1972: 1163), Waterson
(1971: 2067), Drachman (1973: 146; 1975), Ferguson and Farwell (1975), and
Ingram(1974a, b). It may be pointed out that Cdealt with the syllables of the input
forms in a twofold manner. The rst (or, the stressed) syllable was highlighted,
and reproduced in a manner quite appropriate to the stage of development then
reached; the other syllable was dealt with quite cursorily and in a very general
way by recourse to either lengthening of the UF vowel, or by insertion of a
combination of the simple glide and (in many cases) the maximally open vowel.
Finally, the few MEFs which were observed, and which sounded apart fromthe
extra degree of length extremely similar to a reading of the postulated UFs, were
manifestations of an alternative ruse and were apparently employed as a kind of
last ing before the strategy was abandoned altogether.
note
1. We discard the homonyms (3) and (8) because (3) is dubious; (53), (57a), (58a) for
similar reasons, cf. the discussion of idiomatic forms; (5) and (6), (23) and (24),
because one member of the pair ts BI; (22a), because it also ts BI; and (30a) and
(34) because the vocalic correspondence in (30a) is not regular.
References
Berko, J. (1958). The childs learning of English morphology. Word, 14, 15077.
Blasdell, R. and Jensen, P. (1970). Stress and word-position as determinants of imitation
in rst-language learners. Journal of speech and Hearing Research, 13, 193202.
Drachman, G. (1973). Some strategies in the acquisition of phonology. In
M. J. Kenstowicz and C. W. Kisseberth (eds.), Issues in phonological theory, pp.
14559. The Hague: Mouton.
(1975) Generative phonology and child language acquisition. In W. U. Dressler and
F. V. Mare (eds.), Phonologica 1972. Akten der zweiten Internationalen
Phonologie-Tagung. Munich: Fink.
Edwards, M. L. (1971). One childs acquisition of English liquids. Papers and Reports
on Child Language development, 3, 1019.
Edwards, M. L. and Garnica, O. (1973). Phonological variation in imitated and sponta-
neous utterances. Papers and Reports on Child Language development, 5, 78.
Ervin-Tripp, S. M. (1966). Language development. In L. W. Hoffmann and
M. L. Hoffmann (eds.), Review of child development research, pp. 55105.
New York: Russell Sage Foundation.
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language acquis-
ition. Language, 51. 41939. Reprinted in this volume as Chapter 4.
Ferguson, C. A., Peizer, D. B., and Weeks, T. E. (1973). Model-and-replica grammar of a
childs rst words. Lingua, 31, 3565.
Grammont, M. (1902). Observations sur le langage des enfants. In Mlanges linguisti-
ques offerts A. Meillet. Paris: Klincksieck.
Hsieh, H-I. (1972). Lexical diffusion: evidence from child language acquisition. Glossa,
6, 89104.
Ingram, D. (1970). Some suggestions on the role of systematic phonemics in child
phonology. Papers and Reports on Child Language Development, 1, 4355.
One idiosyncratic strategy in the acquisition of phonology 233
(1974a). Phonological rules in young children. Journal of Child Language, 1, 4964.
(1974b). Fronting in child phonology. Journal of Child Language, 1, 23341.
Jakobson, R. (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala:
Almqvist & Wiksell.
Kiterman, B. (1913). Opyt izuenija slogovoj lizii v detskom jazyke. Russkij
lologieskij vestnik, 69.
Kresheck, J., Fisher, H., and Rutherford, D. (1972). A study of r-phones in the speech of
three-year-old children. Folia Phoniatrica, 24, 30112.
Ladefoged, P. (1967). Three areas of experimental phonetics. London: Oxford
University Press.
Ljamina, G. M. (1958). K voprosu o mexanizme ovladenija proiznoeniem slov u detej
vtorogo i tretego goda izni. Voprosy psixologii, 6, 11930.
McNeil, J. and Stone, J. (1965). Note on teaching children to hear separate sounds in
spoken words. Journal of Educational Psychology, 56, 1315.
Menyuk, P. (1971). The acquisition and development of language. Englewood Cliffs, NJ:
Prentice-Hall.
(1972). Clusters as single underlying consonants; evidence from childrens produc-
tion. In Proceedings of the Seventh International Congress of Phonetic Sciences,
pp. 11615. The Hague: Mouton.
Mike, M. and Vlahovi, P. (1967). Glasovna stuktura KVKV u razvojnom procesu
asociativnog sistema glasova. Prilozi prouavanju jezika, 3, 189203.
Moskowitz, B. A. (1970). The two-year-old stage in the acquisition of English phonol-
ogy. Language, 46, 42641.
(1972). The acquisition of phonology and syntax: a preliminary study. In K. Hintikka
et al. (eds.), Approaches to natural languages, pp. 4884. Dordrecht: Reidel.
Ohnesorg, K. (1959). Druh fonetick studie o dtsk ei. Bratislava: Brno University.
Otuszewski, W. (1897). Die geistige und sprachliche Entwicklung des Kindes. Berlin:
Fischer.
Paesov, J. (1968). The development of the vocabulary of the child. Bratislava: Brno
University.
Priestly, T. M. S. (1980). On homonymy in child phonology. Journal of Child Language,
7, 41327.
Snow, K. (1963). A detailed analysis of articulation analyses of normal rst grade
children. Journal of Speech and Hearing Research, 6, 27790.
Suxanova, N. V. (1968). Otnoenie razvitija vtoroj signalnoj sistemy v spektralnoj
kartine detskoj rei. urnal vysej nervnoj dejatelnosti imeni Paloa, 18, 9015.
vakin, N. X. (1948). Razvitie fonematieskogo vosprijatija rei v rannem vozraste.
Trudy instituta psixologii, 13, 10132.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
Appendix I Bisyllabic ordinary forms noted in week 4
A. By medial consonant
Medial [j]: ljn, dj, wjl, sjel, j, fj, wj, gj, hj, fjl, sjl, wjl,
njl, jn, bj, nj, tj
lion, deer, whale, seal, ear, re, wire, gear, here/hear, le, stile, wheel, nail,
iron, beer, near, tire
234 T. M. S. Priestly
Medial stop: ddi, g
gi, fki, k
tn, k
kn, ppi, h
g, p
ki, p
p, fp, rdi,
ttut, pk, bt, ljdi
Daddy, doggy, Suki, kitten/kitchen, kitchen/chicken, puppy, hippo, Deedee,
cracker, bacon, water, apple, potty, butter, Teddy, Buddy, monkey, sweater,
digger, picky, popper, supper, ready, toot-toot, paca, better, lady
Medial nasal: mmi, nn,
ni, n
rid
{
(a) pajas 5
}
(b) pajs 12
p
rs 12 porridge
(8) pljsk
pija 4 pljsk
13 police-car
(9) brijz bjas 6 brijs 13 berries
(10) bjsn bajas 4 bjsn 12 bison
(11) b
kij tajak 4 t
kij 12 turkey
(16) tm
t djawt 3, 7 dn
t 8 doughnut
(18)
t
kt
ka
k
( )
a kajak 3; 6; 9
bkjak7
8
<
:
c k k 7
)
ka
kat 6
ka
kt 9
8
>
<
>
:
9
>
=
>
;
chocolate
(19) djzs dijas 4 djzs 13 Jesus
(20) d
n 12 Jennifer
(22) krt
a kajat 5; 9
(
b k t 12
)
krt 12 carrot
(23) krl kajal 3 krl 12 carol
(24) kndl kajal 3 kndl 12 candle
(25) k
bd kajat 5, 9 p
bd 12 cupboard
(26) k
vd kajat 5 k
vd 13 covered (wagon)
(27) kwst kajows 5 kws 8 coaster
(28) gr gajas 4 gras 10 garage
(29) g
m fajam 3, 6 f
m 6 farmer
(33) svn sjan 7 svn 7 seven
(34) s
k fajak 3, 6
f
k 4
s
k 12
( )
sucker
(35) swld sjat soldier
(36) wld sjat shoulder
(37) prznt pjas 4 przt 13 present
(38) blkt bajak 3, 6 bkt 11 blanket
(39) brnd bjan 4, 7 bnd 10 Brenda
(40) drgn
a dajan 3
b dajak 4
( )
dgn 10 dragon
(41) krsms kijas 4 kss 12 Christmas
(42) nl
a fajan 6
b fajal 6
( )
fnl 10 annel
(43) spjd bajat 4 bjd 8 spider
(44) strjm mijat 4 djm 12 streamer
(45) skwrl gijal 12 grjl 10, 13 squirrel
(46) wsk wijak 4 ws 12 whisker
(47) wsl
a wijas 4
b wijus 6
( )
wsu 12 whistle
(48) wmn wajum 6 wbn 12 woman
(49) hdjk hajak headache
(50) h haja 4 h 12 hanger
236 T. M. S. Priestly
Input BEF Output MEF ORF Lexical item
(51) hjdrnt hajat 7 hjdt 12 hydrant
(52) mnt mijat minute
(53) mtsn mjas 3 mtsn 12 medicine
(54) mln
amjan7
(
bme n12
)
mln 12 melon
(55) mw
sk
ks majks 7 mks
ks 10 musk-ox
(57) m
nst
{
(a) mjan 5
}
(b) majs 4
(c) majs 6
m
st 6 monster
(58) mjuwzk
a mejas 3
b mijus 6
( )
mzk 12 music
(59) lzd zijan 6 lzt 12 lizard
(60) rkdz rjas 3 rkas 10 records
(61) rbt
a rajat 4
b rajap5; 7
(
cr p10; 11
)
rbt 9 rabbit
(62) rajn
srs
{
(a) rajas 4
}
(b) rajs 7
rjn
s 12 rhinoceros
(63) r
n rajan 7 rn 10 runner
(64) jmjuw
a ijumum6
b ijum6
( )
jmju 12 emu
(65) ndn jan engine
(66) lfnt a jat 4; 6 b t10; 12 f g lt 8 elephant
(67) rjl jal 5, 12 rijal 13 aerial
(68) ksdnt ajak 5 kas 11 accident
(69)
sk jas 5
s 8 Oscar
(70)
rnd
a ajat 6
b aj
t 7
( )
rn 11
rdz 12
( )
orange
One idiosyncratic strategy in the acquisition of phonology 237
8 Phonological reorganization: a case study
Marilyn M. Vihman and Shelley L. Velleman
Introduction
In 1975 Ferguson and Farwell analyzed the initial consonant use in the rst fty
words of three English-speaking children in an effort to identify the primary
characteristics of childrens early sound production in words. They began with
the assumption that both initial consonants and words are valid units of early
child phonology, but ended by concluding that much of early phonology is
word-based. Since that time, that conclusion has come to be widely accepted
(see e.g., Grunwell 1981; Menyuk, Menn, and Silber 1986; Studdert-Kennedy
1987; MacKain 1988).
If we see phonological development as beginning with pre-systematic, whole-
word-based productions, we must then ask howthe child proceeds fromthat point
to the orderly, segment-substitution-based phonology described for older chil-
dren, the hallmark of which is said to be its systematicity (Oller 1975; see also
Smith 1973; Ingram 1986). The rst ve to ten words may show little phono-
logical interrelationship. Jusczyk (1986) hypothesizes that the earliest word
recognition network may be restricted to words stored as separate entities with
no particular organization (p. 14). In production, similarly, there is little evidence
of even incipient phonological organization in the formof the rst fewwords (see
Waterson 1978 and also Vihman 1987, which lists the rst ve to six words of
each of twenty children learning six languages). As more words are added to the
lexicon, however, we begin to see a decrease in the number of individual phones
used in limited lexical contexts as fewer, broader phone categories are used in
larger classes of lexical contexts (Leonard, Newhoff, and Mesalam 1980; Stoel-
Gammon and Cooper 1984). We also see the emergence of the rst idiosyncratic
rules or word recipes which restrict the childs rst sound congurations
(Ingram 1974; Menn 1979), and the rst signs of phonological behavior which
imply some awareness of segment (Bleile 1986).
The relationship between these earliest detectable signs of systematization
and later segmental phonological systems is not yet well understood. We do not
yet know whether children acquire segmental phonology suddenly and/or
This study was supported in part by funding from the National Science Foundation (BNS 7924167
and 8520048).
238
across the board (i.e., all phonemes in all word positions), or whether segmen-
tation gradually becomes the predominant characteristic of their phonological
systems. The nature and process of emerging phonological systems must
be studied in greater detail if we are to understand how the child gets from
here to there; from whole words to segments. (Macken 1979, provided a rare
instance of a longitudinal study documenting this transition.) Furthermore,
there has been little acoustic verication of the systematization of childrens
phonologies, despite the well-known limitations on auditory transcription,
which are due primarily to listeners tendencies to hear through the lter of
the established categories of the adult language (see e.g., Stockman, Woods, and
Tishman 1971; Zlatin and Koenigsknecht 1975, 1976; Macken and Barton
1980; Maxwell 1981; Maxwell and Weismer 1982).
The purpose of the present study is to illustrate the process of reorganization
and the beginning of phonological systematization, as validated acoustically,
in the development of one particularly voluble child recorded weekly from
the age of 9 to 16 months. This childs lexical production was analyzed in
detail from the onset of word use, at 10 months, to 16 months, when she had a
cumulative lexicon, according to maternal report, of over 70 words. Both
perceptual (transcription-based) and acoustic analyses were carried out, in
order to identify and conrm the emergence of a phonological system, includ-
ing evidence of the onset of productive word-conguration patterns and of the
earliest behavior indicative of the treatment of segments as entities distin-
guishable within words.
We will document three related characteristics of the process of system-
atization in this childs production:
1. Experimentation, or the phonological variation resulting from the childs
apparent exploration of alternate solutions to production problems posed by
particular target words, given the childs articulatory constraints (Macken
and Ferguson 1981; Bleile 1986).
2. Word recipes, or the use of idiosyncratic, whole-word-sized production
patterns, sometimes involving a prosodic match to the target. The restruc-
turing of adult targets to t child output patterns provides the best
evidence of the workings of a word recipe or articulatory routine. These
recipes allow the child to expand his or her lexicon within the constraints
of a small number of possible output shapes. (See Waterson 1971; Menn
1976; Vihman 1976 1981; Macken 1978 1979 for examples of this
phenomenon.)
3. Regression, the nonlinear progression more familiar from studies of the
acquisition of morphology and syntax (e.g., Bowerman 1982), in which
early forms accurately reecting an adult model are replaced by less
advanced forms in closer conformity with the childs system (Leopold
1947).
Finally, we will consider the childs emergent system from the point of view of
the word vs. the segment as the primary unit of organization.
Phonological reorganization: a case study 239
Method
Subject
The subject of this study, Molly, was one of ten children whose language
development was followed as part of the Stanford Child Phonology Project.
Each child was audio- and video-recorded weekly at home in free play with
the mother for 30 minutes, from 9 to 16 months of age. A maternal interview
focusing on the childs progress in language comprehension, gestural and
vocal communication, and play was administered monthly (Bates, Benigni,
Bretherton, Camaioni, and Volterra 1979) and the mother was asked to maintain
a daily log in which she recorded advances relevant to the childs communica-
tive and symbolic development. The mother was also asked to so structure the
play session as to allow the child to produce any words that seemed to be in use
in the preceding week, so that the childs current lexicon would be represented
in each recording to the greatest possible extent. The number of cumulative
words reported by each mother in the larger study for the lexicon of her child
was found to be approximately twice the number produced during a given
session (Vihman and Miller 1988). Further details of subject selection and
recording procedures from the larger study are available elsewhere (Vihman,
Macken, Miller, Simmons, and Miller 1985; Vihman, Ferguson, and Elbert
1986; Vihman and Greenlee 1987; and Vihman and Miller 1988).
Mollys word production and phonology were unusual in several respects.
First, she was an exceptionally voluble child. Molly ranked rst in mean
vocalizations per session, sampled over seven sessions from 9 to 16 months
(Vihman et al. 1985). Secondly, Mollys production patterns for individual
words were remarkably consistent or stable within a given session; there was
relatively little variability from token to token of a certain word type. At
16 months Molly used the smallest number of different phonetic shapes per
word (fewer than two different phonetic shapes per word type: Vihman and
Greenlee 1987). Thus, many tokens were available for each word type, and
those tokens tended to be phonetically similar.
Furthermore, Mollys early phonology was relatively systematic as well
as stable. She attempted only a restricted set of consonants (primarily stops
and nasals: Vihman et al. 1986). Last but not least, Molly was unusual in her
early preference for nal consonant production. Twenty-three percent of her
vocalizations both words and babble included nal consonants at a time
when she had a 3050-word lexicon, as compared with a range of 4 to 19
percent nal consonants for four other children (Vihman and Greenlee 1987;
for a discussion of the rooting of this nal consonant preference in the
prelinguistic period, see McCune and Vihman 1987). Of 77 word types
recorded for Molly over seven months, 35 percent sometimes included a
nal consonant, compared to a mean, for seven subjects, of 25 percent
(Vihman and Hochberg 1986). All of these aspects of Mollys phonological
240 Marilyn M. Vihman and Shelley L. Velleman
style will be relevant to our discussion of her treatment of nal consonants
over time.
Data preparation
Each audio tape was transcribed by one of four transcribers using a ne phonetic
transcription system based upon the IPA and supplemented by a symbology
especially developed for use with children (Bush, Edwards, Luckau, Stoel,
Macken, and Peterson 1973). Reliability was tested using brief samples from
tapes of seven infants, including Molly. Agreement with respect to place and
manner in consonants, syllable shape, and vocalization length in syllables
reached 86 percent. If differences involving initial and nal glottal stops and
[h] are disregarded, reliability across the four transcribers reaches 91 percent for
the parameters tested.
The video tapes were reviewed repeatedly, once auditory transcription had
been completed, in order to determine the word status of each vocalization. Both
form and meaning were taken into consideration in evaluating word status
(see Vihman et al. 1985; Vihman et al. 1986; Vihman and Greenlee 1987; and
Vihman and Miller 1988 for discussion of the problems involved in word
identication at this age). Word tokens tentatively identied early on were
sometimes discarded from the nal analysis. This generally occurred when
one or more of the following sources of doubt about word status obtained: the
context was insufciently clear, the occurring phonetic shape was relatively
distant from the suspected adult target or was not easily distinguishable from
other vocalizations used in different contexts, or the word type in question was
used only once or in only one episode on the tape. In Mollys case, frequent,
relatively stable repetition of tokens of the same word type and, in later sessions,
increasing use of nal consonants, made word identication relatively less
problematic than for some of the other infants.
Data analysis
Approximately two sessions per month were selected for acoustic analysis,
beginning with the session in which Molly was 1;0.26 (one year and twenty-six
days old) and produced 11 different words spontaneously (11 word types) in the
course of the half-hour recording, and ending with the session in which she was
1;2.20 and produced 19 words spontaneously. Perceptual (transcription-based)
analysis covered a larger number of sessions, from 0;10.15 to 1;3.24. Mollys
mother had reported a cumulative vocabulary of 69 words by the week preced-
ing this last session analyzed. Mollys age at each session and the number of
spontaneous adult-based word types identied during that session are given in
Table 8.1.
Perceptual analysis of Mollys speech revealed that words of CVC shape
were of particular interest with respect to Mollys phonological reorganization.
Phonological reorganization: a case study 241
Spectrograms were therefore made of all word tokens of CVCshape which were
not interrupted by environmental noise or overlapping speech. An oscillogram
was also made of each utterance for verication of patterns detected within the
spectrogram.
Results
Perceptual analysis
Words with nal nasals. Longitudinal perceptual analysis of Mollys
speech reveals the development of a nasal-nal word recipe or articulatory
routine. This routine developed through the stages described below (see
Table 8.2 and Appendix):
a. Presystematic; no pattern. In the rst two months of word production no
nasal-nal words are attempted disregarding the onomatopoeic yum-yum
and vroom-vroom, which are highly variable in the adult presentation (yum
ranges from [jm] to [], and vroom ranged from [vrm] to []). In fact,
the only nasal words recorded are mama, moo, and night-night (produced by
the mother as nigh-nigh).
At 1;0.26 Molly attempts her rst conventional nasal-nal words, actually
producing a nal nasal in at least one token of bang [ba
]. (Table 8.2
therefore begins at this point.) However, each nasal-nal word attempted is
phonologically distinct from the other in the childs production; no pattern is
evident. The child appears to have a set of unrelated forms which happen to
end similarly in the adult language.
b. Apparent increase in awareness of phonetic targets; experimentation.
At 1;1.8, Molly imitates down [t
~ t
:] down [t
]
c 1;1. 15 down [d:n] button [pa
n
n]
round [han]
d 1;2.20 around [wa:n] Ernie [hn] Brian [pan]
down [tn] Granma [nm] button [pa
n
n]
hand [han] Jennie [tni] building [pa:n]
e 1;3.24 bang [pa
n
n] Nicky [nni] camera [ka
m
m] Graham [k
n
ni] open [hp]
down [ta
n
n] Nonny [na
n
ni] piano [pa]
green [kyni]
in [
n
ni]
name [ne
m
mi]
NB: One token is cited here for each word; see Appendix for other tokens.
adult pattern. At the same time, a wide range of different endings are
experimented with for one of these words (bang), only some of which
actually include a nal nasal (the free variation between initial [p] and [b]
is disregarded here, both being noted as [b]):
[b:]
[b
]
[bk]
[b:]
[b:in]
[b
:ini]
[b
]
Note that one of these variants, [b:], is a fairly exact rendering of the adult
form. As described below, this experimentation ends in the gradual domi-
nance of one particular output pattern but not, as it happens, the most
accurate variant.
c. Production pattern dominance. At 1;1.15, one of the patterns used for
bang emerges as the preferred pattern for target nasal-nal words; this
pattern dominates within word variants and is used across nasal-nal word
types as well, extending to one disyllabic model. This preferred pattern is (C)
VN:V, in which the medial consonant varies between [n] and [], and the
nal vowel between (voiced or voiceless) [i] and [] (see the last form cited
above). This pattern is the most common production type for all three nasal-
nal words attempted: two monosyllables and one disyllable formerly pro-
duced without a nal nasal.
d. Restructuring of targets. One month later (1:2.20) we see the same pattern
in use, but the competing alternate forms no longer occur. Furthermore, other
word types are restructured in accordance with the newly established word
recipe. In addition to nasal-nal target words, target disyllabic words with
medial nasals and eventually (at 1;3.24) even one word with a target initial
nasal are adapted to the childs preferred nasal pattern. Nine words are
involved at 1:2.20: three monosyllables, three disyllables with target nal
nasals, and three disyllables with target medial nasals.
1
The change in Mollys production of words involving nasals was
striking enough to provoke a spontaneous comment by her mother at this
time. She reported that button, balloon, banana, and bunny, each of which
had had its own unique phonetic shape previously, were now all produced
as [bn:] or [ban:] a regression reecting the force of Mollys new
output pattern.
e. Other patterns emerge. At nearly 16 months (1;3.24) Molly continues to
make use of the nasal production pattern, adding camera (/kmr/, produced
as [k
m
m]) and even reforming the nasal-initial Nicky: [
n
ni] to t the
pattern. On the other hand, some nasal words are now produced without a
nasal: open ([hp]) and piano ([pa]).
244 Marilyn M. Vihman and Shelley L. Velleman
Words with nal obstruents. An obstruent-nal output pattern devel-
oped in a manner which paralleled the emergence of the nasal output pattern
quite closely. The chronology of this related pattern is described below and
displayed in Table 8.3. Comparing the chronology of changes in nal obstruent
use with the stages of development of the nasal-nal production pattern, we can
characterize Mollys progress as follows:
a. Pre-systematic. Molly rst produces a nal stop in a word token at 0:10.15,
when cracker occurs in a range of variants, including the monosyllables
[kk], [w
h
k], and [kk], as well as disyllabic [pak], [tk], and
[pkwa]. In the following month block, cat, and dog are all produced as
monosyllables with no supraglottal nal consonant ([pa], [ka:], [d]). At
1;0.10 four words are produced with nal stops: baby [pe:p], good girl
[kVkVk], peek(-a-boo) [pk], and (w)oops [Vp] (V = a range of different
vowels). Thus, during this period consonant-nal words are attempted, and
monosyllabic variants with nal consonants occur in attempts to produce
disyllabic words, but no pattern is evident. (Compare 0;10.15 through 1;0.26
for nasals.)
b1. Apparent increase in awareness of phonetic targets. Two weeks later, at
1;0.26 the session in which the rst conventional nasal-nal words are
attempted Molly produces ten new words with nal obstruents (see
Table 8.3). Furthermore, a new level of attention to word endings seems
to be reected in the fact that virtually all of these words occur with heavily
aspirated nal consonants: e.g., cup [kk
h
], hot [hat
h
], teeth [tit
h
], and up
[hp
h
]. In addition, these consonant-nal words are characterized by rela-
tively little variation. In sum, there is an increase in the targeting of words of
a given type, with phonetic output patterns at 1;0.26 for obstruents similar
to those which occurred for nasals at 1;1.8, the next session analyzed.
b2. Experimentation. At 1;1.8 the session in which bang is explored through
a range of different variants Molly again produces a large number of
consonant-nal words, but the degree of variability rises sharply. Two
words in particular, hot and teeth, resemble bang in the wide range of
variants used:
hot [hat] (10) ~ [ht] ~ [at
h
] ~ [ha] (2) ~ [ha] (2)
teeth [ti
t
t
h
i] (2) ~ [titsti] ~ [tit
h
t
h
t
h
] ~ [tit] (5) ~ [tit] (2)
(Note: initial [t] = [t] or [d])
In addition to the glottal-stop-nal variants of hot, Molly produces boat
[be], cat [k
h
], and dog [ta
h
] as open monosyllables in this session. On the
other hand, the extra-heavy aspiration of teeth occurs also in an imitation of
clock: [ka
k
k
h
]. This increase in variability, especially in the production of
a few often-used words, parallels the variable production of bang, also at
1;1.8. A dialogue which occurs between Molly and her mother during the
session at 1;1.8 suggests that the child is aware in some sense that released
nal stops enhance intelligibility:
Phonological reorganization: a case study 245
Table 8.3. Obstruent-nal production pattern
Final consonant used No nal consonant used
Stage Age Target nal C No obstruent in _# in model Target nal C
Stop Fricative Stop Fricative
labial dental velar
0;10.15 cracker [kk]
0;11.9 cat [k
h
]
0;11.20 block [pa]
dog [d]
a 1;0.10 oops [p] peek [pk] baby [pe:p]
good girl [kk
h
k]
b1 1;0.26 cup [kk
h
] bird [pt] book [pk
h
] teeth [tit
h
] pumpkin [kk] close [ko
]
up [hp
h
] hot [hat
h
] box [bk
h
]
toot [tt]
cold [kok
h
]
b2 1;1.8 hot [hat] clock [ka
k
ki ] teeth [tit t
h
i] good girl [kkkkk] boat [be]
peek [pe||k
x
e] horse [ht] cat [k
h
]
squeak [k
h
k
h
] dog [ta
h
]
tick [tt
h
]
c 1;1.15 up [p] hot [ht] good girl [ggk] hat [h] box [ba]
burp [pap
h
]
d 1;2.20 up [p] foot [pt] book [pk] bus [pt] baby [pib] nose [n:]
eat [t
] stuck [tt
h
] cheese [at
] Hooper [pt]
coat [kk] block [pt] watch [wat] apple [ap]
toot [tut] tock [t
h
aki]
clock [kak]
1;3.24 Brett [pat] block [pak] glasses [ka
k
k
h
i:] nose [no:]
red [wa::t] book [pk] house [hat] beads [pi]
that [tat] click [kik] Ruth [ht]
oink [ho:|k
h
]
peek [pik]
pig [pk]
rug [wak]
. stuck [kak]
walk [wak
h
]
work [hk]
NB: One token is cited here for each word; see Appendix for other tokens.
mother: Good stuff. Good stuff to eat.
molly: [ha]. [ha]. (hot X 2)
mother: To eat.
molly: [ha
h
]. (hot)
mother: No, I didnt say teeth; I said eat. (Laughs)
molly: [ht]! (hot)
Molly repaired her misunderstood productions of hot by over-articulating the
nal consonant, which was more in keeping with her previous (and later)
pronunciations of this word.
c. Production pattern dominance. At 1:1.15, Mollys variability decreases
some what, but less than it does for nasals in the same session (1:1.15). The
production of heavily aspirated nal consonants (CVC
h
) begins to emerge as
the preferred output pattern.
d. Restructuring of targets. By 1:2.20 production of nal consonants has
again stabilized. Aspirated release characterizes most of Mollys nal con-
sonants at this point. In addition, certain targets are restructured to t this
output pattern by 1;3.24. Substitution of a dental stop or affricate for a nal
voiceless dental fricative has become a systematic phonological process for
Molly (bus [pt], house [hat], Ruth [ht], in addition to the earlier teeth
and horse [ht]). Voiced fricatives are omitted word-nally (nose [no:],
beads [pi]). But cheese is exceptional. Like the reformation of Nicky noted
earlier, the cheese variants [at ] and [it ] appear to reect whole-word
processing. The word-initial [t ], which Molly had not yet produced else-
where, is moved to nal position, where [t ] has already occurred in some
variants of words like hot and teeth.
In summary, for obstruents, just as for nasals, a gestalt-like whole-word pattern
is now imposed on a certain number of adult words with nal obstruents. The
change in nal-obstruent patterning is less dramatic than the development of
the nal-nasal production pattern, but is nonetheless unmistakable. Before the
reorganization, nal-consonant production has an unplanned, accidental look to
it: there is little relationship between the adult shape of the different words
produced with nal consonants (e.g., baby, good girl, oops). After the stages of
pattern-discovery and experimentation, a stable production pattern emerges and
is widely applied, extending also to new types of adult models.
The development of these two production patterns is interesting not only for
the chronological parallels (summarized in Table 8.4), but also for the fact that
both output patterns involve a strategy for producing nal consonants. In fact,
both patterns involve continued airow following nal consonant constriction,
and both serve to increase the salience of nal consonants. Extension of the
obstruent-nal production pattern to cheese is fully parallel to the extension of
the nasal-nal pattern to Nicky. There is thus clear phonological evidence of a
reorganization affecting consonant-nal words during this period. The similar-
ities in the shape and function of the two patterns lead us to conclude that a
unitary phonological (re)organization has occurred.
248 Marilyn M. Vihman and Shelley L. Velleman
Acoustic analysis
Acoustic analysis was used to verify the perceptions of the transcribers and the
authors. Spectrograms were made of all productions relevant to the develop-
ments described here which were relatively free of background noise, compet-
ing maternal speech, etc. (approximately 5075 percent of productions in each
category in each session; see Appendix). Wide-band spectrograms were made
on the Kay Elemetrics Corp. Digital Sona-Graph 7800, using a 500 Hz band-
width lter, and representing the frequencies from 0 to 5000 Hz. Spectrograms
of stop-nal words were measured for duration of consonant closure and for
duration of release (aspiration) following closure. Spectrograms of nasal-nal
words were measured for duration of vowel + nasal and for duration of nasal
release following the nal nasal. Because nasal-nal words were often produced
with a nasalized vowel only (e.g., [b
and (C)VC
h
, restructuring
of new target words to t the established production patterns, and apparent
regressions loss of distinction among nasal-nal words, as noted by Mollys
mother; loss of salience in nal position for obstruents during the experimenta-
tion period.
600
450
300
150
C
o
n
s
.
c
l
o
s
u
r
e
d
u
r
a
t
i
o
n
(
m
s
e
c
)
Child age
1;0.26 1;1.8 1;1.15 1;2.20
0
Figure 8.3. Consonant closure duration by age
C
o
n
s
.
r
e
l
e
a
s
e
d
u
r
a
t
i
o
n
(
m
s
e
c
)
1200
800
400
0
Child age
1;0.26 1;1.8 1;1.15 1;2.20
Figure 8.4. Consonant release duration by age
Phonological reorganization: a case study 251
Mollys unusual phonetic consistency makes the contrast between experi-
mentation sessions and preceding and following sessions particularly striking.
Mollys productions at 1;1.81;1.15 are much more variable than at any other
time, and the patterns of unreleased nal stops and of nal nasals with nasal
releases which occur during this session are both new and of high frequency.
In addition, the restructuring of targets to t output patterns following the
experimentation sessions is quite noticeable, especially with respect to nal
nasals; the nal nasal pattern is extended to a wide variety of words (Tables 8.2
and 8.3). It is clear that some sort of experimentation and reorganization occur
in Mollys phonology at 13 months. The changes which follow this session
cannot be taken to be random or accidental events.
An interpretation reducing the changes described here to motoric maturation
can be ruled out. The new development with respect to nal obstruents at 1;1.8
is the production of a variety of unreleased or glottal-stop offsets in words of
the shape CVC. This pattern is abandoned in the next session. Locke (1983)
presents data which suggest that American adults fail to release nal voiceless
stops about 36 percent of the time in conversational speech, despite a resultant
loss in intelligibility, and that young children omit or replace nal voiceless
stops by glottal stops up to 50 percent of the time (pp. 228ff). These data imply
that Mollys new pattern at 1;1.8 is articulatorily easier than is her previously
and ultimately preferred pattern of released nals. Therefore, the glottal-stop
and vowel-nal productions of forms with obstruent-nal targets cannot be held
to reect a phonetic advance over the productions with an aspirated stop release.
In the case of the nasal-nal pattern, the phonological reorganization is clear.
A pattern never used before the experimentation session one involving nasal
releases is tried out along with a variety of other patterns during that session,
and is clearly dominant in the following sessions.
2
In the case of obstruents, Molly consistently produces aspirated nals both
before and after the experimentation at 1;1.8. The change is that the pattern
becomes more productive following the childs experimentation with other
(phonetically simpler) means of approximating nal consonants. The fact that
Mollys CVCs are phonetically the same before and after 1;1.8 does not contradict
the claim that a phonological change occurred; it merely obscures the process.
Both the experimentation at 1;1.8 and the increased productivity thereafter are
evidence of phonological systematization in the case of obstruents as well as
nasals.
Furthermore, it is probably not a coincidence that nal obstruents and nal
nasals undergo reorganization at the same time. Acoustic analysis suggests that
the two patterns are related. In both cases, nal consonants are given a strong
vowel-like release in the form of heavy aspiration for obstruents and of often
voiceless nasal releases for nasals. The articulatory sequence for both involves
continued airow following the release of the consonant constriction. If voicing
is continued into this post-release airow, a vowel is likely to be perceived. If
voicing is not continued but the airow is, a heavily aspirated release is a likely
252 Marilyn M. Vihman and Shelley L. Velleman
percept. The fact that vowels were perceived by the transcribers more often
following Mollys (voiced) nasals than following her (voiceless) oral conso-
nants is not surprising. Furthermore, when vowels were perceived following
nal obstruents in Mollys speech, they were often transcribed as voiceless. This
is merely further evidence that her nal stops were indeed voiceless.
We embarked upon this study with the question of the nature and process of the
transition from whole-word to segmental phonology. It may be that production
patterns such as Mollys, which highlight segments in one word position only, are
a step in the direction of segmental phonology. Molly seems to experiment only
with variants of nal segments; there is no evidence of a change by 16 months
in Mollys organization or awareness of other parts of words. Such changes
may occur so subtly and gradually as to escape notice, or they may occur later in
her phonological development. If the latter is true, then the onset of segmental
phonology may occur gradually in some children, with segmentation emerging
rst in the shape of an articulatory routine which affects one particular word
position. A series of such specic articulatory routines would lead the child
eventually to a segment-based phonological system.
Although Molly is an unusual child phonologically in some ways, the process
which we have documented here provides strong conrmation of previous
claims about childrens development of articulatory routines (e.g., Menn
1979). Mollys volubility and unusual consistency, coupled with our use of
acoustic as well as perceptual analysis, have permitted more thorough quanti-
tative documentation than has previously been provided in the literature.
Although such changes in phonetic variability may be more difcult to identify
in other children, the search for such data should be motivated by the results
reported here. Further verication of such changes is important if we wish to
discover the process by which a child progresses from pre-systematic whole-
word based phonology to the orderly, segment-substitution-based phonology
described for older children.
notes
1. Notice that round and later hand were treated as identical in pattern to the nasal-nal
words (name, bang, down, and green). There is no evidence that Molly was distin-
guishing perceptually between nal nasals and a nal nasal + stop cluster (cf. Braines
[1976] remarks regarding the relative imperceptibility of the voiced stop in such
clusters).
2. It is interesting to note that addition of a nasal offset has been reported for three
children as a phonological strategy facilitating production of a nal voiced stop. Fey
and Gandours (1982) Lasan used a pattern involving voiced stop plus nal
nasal, while Damon Clark and Eva Bowerman both used a nasal with or without a
following voiceless stop as a way of expressing nal voiced stops (Clark and
Bowerman 1986).
Phonological reorganization: a case study 253
References
Bates, E., Benigni, L., Bretherton, I., Camaioni, L., and Volterra, V. (1979). Cognition
and communication from 9 to 13 months: correlational ndings. In E. Bates (ed.),
The emergence of symbols: cognition and communication in infancy, pp. 69140.
New York: Academic Press.
Bleile, K. (1986). Regressions in the phonological development of two children. PhD
dissertation, University of Iowa.
Bowerman, M. (1982). Reorganizational processes in lexical and syntactic development.
In E. Wanner and L. R. Gleitman (eds.), Language acquisition: the state of the art,
pp. 31946. Cambridge University Press.
Braine, M. D. S. (1976). Review of N.V. Smith, The acquisition of phonology: A case
study. Language, 52, 48998.
Bush, C. N., Edwards, M. L., Luckau, J., Stoel, C., Macken, M. A., and Peterson, J.
(1973). On specifying a system for transcribing consonants in child language: a
working paper with examples from American English and Mexican Spanish.
Stanford University: Department of Linguistics.
Clark, E.V. and Bowerman, M. (1986). On the acquisition of nal voiced stops. In
J. A. Fishman, A. Tabouret-Keller, M. Clyne, Bh. Krishnamurti, and M. Abdulaziz
(eds.), The Fergusonian impact: in honor of Charles A. Ferguson on the occasion of his
65th birthday, vol. I: Fromphonology to society, pp. 5168. Berlin: Mouton de Gruyter.
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language acquis-
ition. Language, 15, 41939. Reprinted in this volume as Chapter 4.
Fey, M. E. and Gandour, J. (1982). The pig dialogue: phonological systems in transition.
Journal of Child Language, 9, 5179.
Grunwell, P. (1981). The development of phonology. First Language, 2, 16191.
Ingram, D. (1974). Phonological rules in young children. Journal of Child Language, 1,
4964.
(1986). Phonological development: Production. In P. Fletcher and M. Garman (eds.),
Language acquisition: studies in rst language development, 2nd edn., pp. 22339.
Cambridge University Press.
Jusczyk, P. W. (1986). Toward a model of the development of speech perception. In
J. S. Perkell and D. H. Klatt (eds.), Invariance and variability of speech processes,
pp. 119. Hillsdale, NJ: Lawrence Erlbaum Associates.
Leonard, L. B., Newhoff, M., and Mesalam, L. (1980). Individual differences in early
child phonology. Applied Psycholinguistics, 1, 730.
Leopold, W. F. (1947). Speech development of a bilingual child, vol. 2: Sound learning in
the rst two years. Evanston, IL: Northwestern University Press.
Locke, J. L. (1983). Phonological acquisition and change. New York: Academic Press.
MacKain, K. S. (1988). Filling the gap between speech and language. In M. D. Smith and
J. L. Locke (eds.), The emergent lexicon: The childs development of a linguistic
vocabulary, pp. 5174. New York: Academic Press.
Macken, M. A. (1978). Permitted complexity in phonological development: one childs
acquisition of Spanish consonants. Lingua, 44, 21953.
(1979). Developmental reorganization of phonology: A hierarchy of basic units of
acquisition. Lingua, 49, 119. Reprinted in this volume as Chapter 5.
(1980). The acquisition of the voicing contrast in English: a study of voice onset time
in word-initial stop consonants. Journal of Child Language, 7, 4174.
254 Marilyn M. Vihman and Shelley L. Velleman
Macken, M. A. and Ferguson, C. A. (1981). Phonological universals of language acquis-
ition. In H. B. Winitz (ed.), Native and foreign language acquisition, pp. 11029.
New York: New York Academy of Sciences.
Maxwell, E. M. (1981). A study of misarticulation from a linguistic perspective. PhD
dissertation, Indiana University. (Reprinted by the Indiana University Linguistics
Club, 1982.)
Maxwell, E. M. and Weismer, G. (1982). The contribution of phonological, acoustic and
perceptual techniques to the characterization of a misarticulating childs voice
contrast for stops. Applied Psycholinguistics, 3, 2943.
McCune, L. and Vihman, M. M. (1987). Vocal motor schemes. Papers and Reports on
Child Language Development, 26, 229.
Menn, L. (1976). Pattern, control and contrast in beginning speech: a case study in the
development of word form and word function. PhD dissertation, University of
Illinois. (Reprinted by the Indiana University Linguistics Club, 1978.)
(1979). Towards a psychology of phonology: Child phonology as a rst step. Paper
Presented at the Conference on Applications of Linguistic Theory in the Human
Sciences, Michigan State University.
Menyuk, P., Menn, L., and Silber, R. (1986). Early strategies for the perception and
production of words and sounds. In P. Fletcher and M. Garman (eds.), Language
acquisition: studies in rst language development, 2nd edn., pp. 198222.
Cambridge University Press.
Oller, D. K. (1975). Simplication as the goal of phonological processes in child speech.
Language and Learning, 24, 299303.
Smith, N. V. (1973). The acquisition of phonology: a case study. Cambridge: Cambridge
University Press.
Stockman, I. J., Woods, D. R., and Tishman, A. (1971). Listener agreement on phonetic
segments in early infant vocalizations. Journal of Psycholinguistic Research, 10,
593617.
Stoel-Gammon, C. and Cooper, J. (1984). Patterns of early lexical and phonological
development. Journal of Child Language, 11, 24771.
Studdert-Kennedy, M. (1987). The phoneme as a perceptuomotor structure. In
A. Allport, D. MacKay, W. Prinz, and E. Scheerer (eds.), Language perception
and production, pp. 6784. London: Academic Press.
Velleman, P. F. and Hoaglin, D. C. (1981). Applications, basics and computing of
exploratory data analysis. Boston: Duxbury Press.
Vihman, M. M. (1976). From pre-speech to speech: on early phonology. Papers and
Reports on Child Language Development, 3, 5194.
(1981). Phonology and the development of the lexicon: evidence from childrens
errors. Journal of Child Language, 8, 239264.
(1987). The interaction of production and perception in the transition to speech.
Presented at the Twelfth Annual Boston University Conference on Language
Development.
Vihman, M. M., Ferguson, C. A., and Elbert, M. (1986). Phonological development from
babbling to speech: common tendencies and individual differences. Applied
Psycholinguistics, 7, 340.
Vihman, M. M. and Greenlee, M. (1987). Individual differences in phonological devel-
opment: age one and age three. Journal of Speech and Hearing Research, 30,
50321.
Vihman, M. M. and Hochberg, J. (1986). Velars and nal consonants in early words. In
J. A. Fishman, A. Tabouret-Keller, M. Clyne, Bh. Krishnamurti, and M. Abdulaziz
Phonological reorganization: a case study 255
(eds.), The Fergusonian impact: in honor of Charles A. Ferguson on the occasion of
his 65th birthday, vol. 1: From phonology to society, pp. 3749. Berlin: Mouton de
Gruyter.
Vihman, M. M., Macken, M. A., Miller R., Simmons, H., and Miller, J. (1985). From
babbling to speech: a reassessment of the continuity issue. Language, 61, 395443.
Vihman, M. M. and Miller, R. (1988). Words and babble at the threshold of language
acquisition. In M. D. Smith and J. L. Locke (eds.), The emergent lexicon: the childs
development of a linguistic vocabulary, pp. 15183. New York: Academic Press.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
(1978). Growth of complexity in phonological development. In N. Waterson and
C. E. Snow(eds.), The development of communication, pp. 41542. NewYork: Wiley.
Zlatin, M. and Koenigsknecht, R. (1975). Development of the voicing contrast: percep-
tion of stop consonants. Journal of Speech and Hearing Research, 18, 54153.
(1976). Development of the voicing contrast: a comparison of voice onset time in stop
perception and production. Journal of Speech and Hearing Research, 19, 7892.
Appendix
Acomplete list of transcribed child word types which either target a consonant-nal adult
word or are sometimes produced as a consonant-nal form, or both. (Onomatopoeic
words were generally excluded.) Types used for acoustic analysis are starred. (Note
that acoustic analyses cover ages 1;0.25 through 1;2.20.) C = consonant, V = vowel,
G = glide, I = imitated only.
Age Adult target Child token
0;10.15 cracker w
h
k, kk, kk
[also CV (CV) forms]
protoword: b, ,
hug sound [also CV forms]
0;11.9 cat (I) ka:, k
h
, k
h
, k
h
j,
kh (2)
0;11.20 block (I) a, b,
dog (I) d
1;0.10 baby pe:t, pe: p (3), pep (3)
[also CV, CVC forms]
ball pp, pep
good girl (I) kuk
h
k (2), kko:k
oops jp, p (2),
:p, p, o p, p(1)
peek pk
1;0.26 balloon (I) b.
h
*bang ba , ba
I
(2), da
I
bird b, pt, bt
h
*book bt, pk
h
*button (I) b (2)
choo-choo/toot-toot (I) tt
[also CVCV forms]
*cup (I) kk
h
, kk
256 Marilyn M. Vihman and Shelley L. Velleman
Age Adult target Child token
*hot hat
h
(3)
*up hp
h
(3), hap
h
*box (I) bk
h
, pk
close (I) k, ko
i
ni (2),
b
: (I), p
(2),
pe, pk, be, b
n
boat (I) be
cat k
h
*clock(I) ka
k
k
h
dog (I) ta
h
*down (I) t
:., t
(2), ht, at
h
*peek (I) p
h
i, p, pe||k
x
e
squeak k
h
k
h
, kk (I)
kkkiki (i)
kke
k
kk
kkkik
h
ik
h
ik
kkkk
*teeth tit t
h
i
(2),
tit t
h
t
h
t
h
, dit (3),
ti
t
(2), dits,
ti
t
s.t i , tit (2)
tick-tock (I) ti t
h
n
ni (i),
n
n
*up p, :p
1;2.20 apple (I) ap, hap (4), hap
,
Phonological reorganization: a case study 257
Age Adult target Child token
ap (6,) hp,
p, hp (2),
hap (2), hapa,
ap, ha:p,
ap (3), hap,
hp (2), hp (2)
hap (7), hap
hp (14), hap (22)
hp, u
*around (1) wa:n
baby pipi (5), pebi (10)
pib, paba, pipi (5)
*block (I) pat (5)
*book (I) pk, p
h
ik, pk (2)
*Brian pan, pani, pa:ni (2),
pa
*building (I) pa:ni
*bus (I) pt, pat(4), pa:t
*button (I) pa:a, pa
n
n
cheese ait
, it
(3), hit
*clock (I) kak (2), kk(2)
*coat kk (24), kk (4),
gk, kk (5), ko:k
kk (3)
[also CV forms]
*down tn (2), tn: (3),
tm:, t
n
t (2), ta
*eat i:t
(i), it(i),
it
(3), hit
*foot pt, ht, to, tot,
tot, tt, tt, at,
tat, tot (4)
*hand (I) han (3)
Hooper h
p
pe (3), h
p
p
b, b, p
pt (2), hpo
hp (2), p
hp (3)
nose (I) no:, n: (2)
*stuck tk, tt
h
(19)
tt, tat
h
*tock (I) k
h
t, kaki , t
h
iki,
t
h
a ki,
h
ki
toot toot t
t
t
, tt tt, tt
h
tt
h
*up p
,
:p (2), ::b
watch (I) wa
t
, wat
258 Marilyn M. Vihman and Shelley L. Velleman
9 How abstract is child phonology?
Towards an integration of linguistic and psychological
approaches
Marilyn M. Vihman, Shelley L. Velleman, and
Lorraine McCune
Our goal in this chapter is to explore the emergence of phonological systema-
ticity within a psychological framework. We begin by reviewing earlier work
which traces the initial phonological system back to its origins in babble and
proposes a model of the interaction of perception and production in emergent
vocal organization. Our account of the origins of system attempts to suggest
answers to the question: how can an initial system be constructed? That is,
how does the child move from the production of unrelated vocal forms
(sometimes known as item-based phonology; see Menn 1983; Waterson
1971) to an idiosyncratic holistic system (word-based phonology; see
Ferguson and Farwell 1975)? We will consider a number of issues con-
cerning representation, focusing on these questions: can the categorical
change in language use, from contextually embedded word production to
symbolic reference, be related to some underlying qualitative change in
mental representation? How does the linguistic notion of internal represen-
tation relate to the psychological notion of mental representation? Finally, we
explore the issues that arise in attempting to model the onset of phonological
systematicity, such as: when are we justied in imputing a phonological
internal representation to the child? That is, what is the evidence from the
childs observable behavior (word production shapes) that a formal system of
interrelated representations has begun to cohere? How much structure should
be specied in such internal representations? Or what counts as sufcient
evidence for positing contrasting levels or units in the childs emerging
system? And what is the status of extra-systemic elements, either in the
early period, when a small repertoire of vocal production patterns are used
in response to specic familiar eliciting situations, or in the later period, when
a phonological system appears to underlie the childs productions?
To develop a sufciently general basis for examining these issues we present
microanalyses of the early phonological development of two children who
differ in overall strategy as well as in units of organization and in the articulatory
basis for their rst word productions. This allows us to illustrate some of the
ways in which individual children follow distinct paths in phonological and
259
lexical development, and also to place linguistic advances within a larger
psychological framework.
Origins of system
The rst adultlike syllable production (canonical or reduplicative babbling)
emerges in normal infants within a narrow temporal frame (610 months) and
evinces strong neuromotor constraints. A small consonantal repertoire usually
is reported, reecting simple ballistic movements (stops and nasals account for
most true-consonant-like productions); the syllable nucleus is restricted largely
to low- to mid-central or front vowels, resulting from relatively wide jaw
opening with neutral tongue placement (Davis and MacNeilage 1990; Kent
1992; MacNeilage and Davis 1990). By 10 months individual differences in
production are apparent as infants explore their vocal resources, developing
vocal motor schemes (McCune and Vihman 1987), or preferred production
patterns, which reect both sensitivity to adult language phonetic tendencies
and emergent vocal control.
As the child develops articulatory control and familiarity, through self-
monitoring, with the sound as well as the feel of well-practiced phonetic
gestures, some routinely used sound patterns of the adult language become
perceptually salient through their resemblance to the childs own often repeated
vocal motor schemes. When the child reproduces such vocal patterns
in situationally appropriate contexts, caretakers may identify them as rst
words. We interpret such early words as the products of a tight developmental
interaction. They reect an interindividual construction process based on the
childs evolving vocal capacities and parental attunement to child vocal pro-
duction and to the focus of child attention. The childs vocal capacities them-
selves globally echo dominant patterns of the ambient language and, in turn,
serve as a lter for the childs more detailed (production-driven) auditory
processing of that language.
Figure 9.1 (adapted from Vihman 1993a) displays the model of the inter-
action between perception and production which we assume gives rise to the
rst phonological system. A certain number of adult words are made salient by
virtue of prosodic heightening (the combination of pitch change, increased
amplitude, and increased duration which enters into word or phrasal accenting
in most languages and is usually emphasized further in caretaker talk; see
Ferguson 1964; Fernald 1984 1991; Garnica 1977), frequent occurrence in
isolation or in sentence-nal position (Aslin 1993; Goldeld 1993), and the
inherent interest of the situation of use to a particular child (Lewis 1936;
Ferguson 1978). These words are taken to provide the child, over the rst
several months of life, with an aural impression of ambient speech patterns.
Attention to the sound patterns made prominent by these three factors can be
assumed to play an important role in channeling the childs prelinguistic vocal-
izations toward the phonetic characteristics of the ambient language; these
260 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
salient words and phrases must make up the global auditory impression which
is reected in the babbling of infants on the threshold of speech (Boysson-
Bardies, Hall, Sagart, and Durand 1989; Boysson-Bardies and Vihman 1991).
Visual effects also play a role in shaping the childs prelinguistic vocalizations.
For example, the visual image of jaw opening and closing is a likely component
in the sudden emergence of the rst canonical syllable production, which is
sometimes observed to occur silently before it is accompanied by vocalization
(Roug, Landberg, and Lundberg 1989). Similarly, the characteristic facial set
of adult caretakers could explain early ambient language effects on the use of
vowel space (Boysson-Bardies et al. 1989).
1
Finally, the predominance of
Visual effects
[a]
[jIji] [jeIdji],[jIji] [taIdi]
(lady, daddy)
(14 months)
[babi]
(bottle)
(16 + months)
[i] [ ] [ ] [ ]
C V C V C C V
Word production pattern
('canonical form')
Word
[labial]
Word
[palatal]
(9 months)
(16 + months)
(9 10 months)
(8 months)
[jajaja]
[dji]
[da:ejan]
[kijita:ji]
[a:ji]
[B]
[?pa
h
]
: syllable
() ()
: mora
[ ]: specified features
Timmy
ball Ais /peIz/
Alice
baby
baby
bracelet
balloon
good boy
Bonny
dolly
Ernie
hi
block
bell
(prosodic effects, frequency,
inherent interest of situation of use)
Auditory effects: salient adult words
Salient adult words
(as above + match to VMS)
PERCEPTION PRODUCTION
PERCEPTUO-MOTOR LINK
(ARTICULATORY FILTER)
(Exploration through babbling)
Vocal Motor Scheme
[p
h
a]
[?
a
fa]
[h u
v
v]
Figure 9.1. Model of the interaction of perception and production
How abstract is child phonology? 261
labials in the early words of sighted children (Locke 1983; Vihman Macken,
Miller, Simmons, and Miller 1985), especially the hearing-impaired (Stoel-
Gammon and Otomo 1986) but apparently not the blind (Mulford 1988),
may be ascribed to the facilitative effect of the visual cue afforded by lip
closure.
2
These global ambient language inuences on babbling are expressed in
Figure 9.1 by the dotted line linking auditory and visual effects and vocal
exploration; the interaction of the two may be taken to guide the construction
of individual vocal motor schemes, consistent motor acts performed inten-
tionally and . . . capable of variation and combination to form larger units
which evolve in the course of babbling (McCune and Vihman 1987: 72). The
vocal motor schemes are different for each child, regardless of ambient lan-
guage, but nevertheless are shaped by that language.
The placement of a perceptuo-motor link at the center of Figure 9.1 expresses
the view that many of the characteristics of the childs earliest words which have
been established over the past two decades their relative accuracy along with
their apparent selectivity with regard to adult models (Ferguson and Farwell
1975), their lack of interrelationship or piecemeal quality (Macken and
Ferguson 1983) are most readily understood if we assume that once a child
has begun to repeat a fewvocal patterns with some regularity or apparently at will,
that is, once some vocal motor schemes have developed, these patterns add to the
salience of certain adult words that are, besides, prosodically highlighted,
frequent, and inherently interesting to the child.
More specically, adult words that (more or less) match some pattern which
the child has come to produce with facility eventually will be attempted by the
child, in appropriate (remembered) context. This can be taken to be the charac-
teristic route by which the rst words are uttered by children and identied by
caretakers, often before the child has progressed cognitively to the point of
making adultlike general or symbolic reference to classes of objects and events
(Bates, Benigni, Bretherton, Camaioni and Volterra 1979; Vihman and
McCune 1994). That is, when a familiar situation arises in which a particular
word or phrase allgone, byebye, duckie, no tends to be repeatedly expressed
by adults, if one of those words also happens to be close enough to a vocal
pattern the child has come to know through self-monitoring and can now make
at will, the child is likely to be reminded of that matching pattern (Rovee-
Collier, Sullivan, Enright, Lucas and Fagen 1980), resulting in what adults
identify as rst (context-limited) word production.
The arrow at the bottom of Figure 9.1 represents the route from phonetically
salient adult words, in combination with the development of one or more vocal
motor schemes, to word production patterns, or production formulae that allow
the child to make rapid lexical progress by simplifying the number of options
available when a word is uttered (Kiparsky and Menn 1977; Menn 1983). It is
this last step with which we will primarily be concerned here, a step that
constitutes the bridge from phonetics into phonology.
262 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
Representation
The term representation is widely and somewhat ambiguously used in scien-
tic elds, including linguistics and psychology. To clarify the issues raised in
this chapter, in which the term representation is applied in several senses, we
rst will address controversy regarding mental representation in the eld of
psychology and introduce evidence for relations between mental representation
and language. We then consider the systematically ambiguous meaning of
internal representation, as used in the eld of linguistics, to prevent misun-
derstanding of the comparable usage here.
Psychological views of mental representation
Two basic positions regarding mental representation in the infant are current in
psychology today: (1) it is present from birth and demonstrable early in life,
with subsequent changes reecting maturation of innate capacities (Leslie
1987); and (2) it develops over the rst two years of life as a consequence of
the organization of motor and behavioral actions in relation to the developing
central nervous system, with onset following the rst birthday and roughly
corresponding to signicant developments in language (McCune-Nicolich
1981b). Investigations of the behavioral expression of representation from
these disparate viewpoints show little overlap in the tasks used or the ages of
the subjects tested. For example, those who espouse the innatist view study
infants in the early months of life, using differential looking time or the infants
tendency to continue the trajectory of visual following when an object disap-
pears from view (Baillargeon 1987). Studies deriving from the developmental
position begin after 6 months of age and utilize motor responses to phenomena
of absence, such as object search (Ramsay and Campos 1978; Uzgiris and Hunt
1975) or representational play (McCune 1995; Nicolich 1977).
Sartres (1966) contrast of perceptual versus imaginal or representational
experience suggested that the early phenomena are best understood as per-
ceptual processing of present reality, which may include memory
of the immediate past and expectancy regarding the immediate future.
According to Sartre, perceptual processing draws continually from the sensory
present, in which the contents of consciousness can be specied in relation to
phenomena observable in the environment. In contrast, imaginal (or mental)
representation may use a perceived event as a starting point (e.g., a portrait of a
friend), but the resulting experience is an instance of pure consciousness, an
internal contentful state that is not directed at perceived reality.
If we take perceptual processing to be infants original tendency, the devel-
opmental course of the ability to relate the present to the absent and the past to
the present provides an index to the emergence of a capacity for mental
representation. Memory research spanning the age range of 8 weeks to 6 months
by Rovee-Collier and her colleagues (e.g., Rovee-Collier et al. 1980) has
How abstract is child phonology? 263
provided the strongest indicator of both the strengths and the limitations of
infants capacity to retrieve past experience. In these studies, infants learn to set
a mobile in motion by kicking the leg attached to the mobile by a ribbon. After
several training sessions it is possible to test both immediate and long-term
retention (up to weeks and months). The results clearly indicate that even
8-week-olds are capable of retention. However, the degree of dependence
of this effect on exact replication of the training context, the limitations on the
length of the retention period, and the relative power of cuing to reinstate the
memories each show strong developmental trends. Furthermore, the behavioral
expression of memory in this task is itself a motor response (foot kick) that
occurs in the context of perceptual recognition. These studies demonstrate the
strengths of perceptual motor processes for learning and memory. The fact that
at 6 months of age infants succeed at this paradigm but fail to search for and
retrieve small hidden objects suggests that a qualitatively different type of
processing may characterize memory that depends upon contextual reinstate-
ment (here termed perceptually dependent memory) as opposed to
memory that is able to function in the presence of confusing contextual cues
(here termed mental representation). We use the term mental representation to
refer to a contentful mental state distinguished fromperception by its capacity to
reference absent and past realities.
Mental representation and language
Piaget (1962) suggested that the infants ability to retrieve an object placed in
the experimenters hand or a container and then released beneath a cloth outside
the infants view indicates an initial capacity for mental representation (Stage 6
of Object Permanence). Given the nature of language as symbolic or repre-
sentational, it was at rst assumed that development in understanding of object
permanence would correlate with language development. In fact, children in
the early stages of language acquisition are capable of solving the Stage 6 task.
However, the number of object permanence test items passed in Stage 6 shows
no continued correlation with advances in language (Bates et al. 1979;
McCune-Nicolich 1981). The lack of correlation can be attributed to the fact
that entry into Stage 6 constitutes a culminating milestone in the concept of
object permanence but marks only the onset of language use.
Representational play, which begins only toward the end of the sensorimotor
period, does show reliable relationships with later language milestones (Bates
et al. 1979; McCune 1995; McCune-Nicolich and Bruskin 1981). The transition
to representational play, in which the child rst demonstrates knowledge of the
function of small replicas (toy cups and saucers, tiny cars and trucks) and later
indicates awareness of the pretend nature of such acts by vocal elaborations
and coy smiles, corresponds to the production of early nonreferential (context-
limited) words in precocious talkers, whereas referential words are likely to be
noted at the transition to play in which two or more acts are combined. We
264 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
interpret the temporal correspondence between this combinatorial play and
referential language use as following from the more differentiated character of
mental representation, which allows an event to be portrayed with a variety of
gestures and objects and vocal forms to be produced outside of their original
context in relation to a variety of new situations. For example, doggie, learned
with reference to the family pet, is now produced in relation to the neighbors
dog and to pictures of dogs as well. It should be noted that close play-language
correspondences characterize early talkers, whereas studies have indicated
that later talkers may show the play milestone well before the corresponding
language is observed. McCune (1992) demonstrated that lack of available vocal
motor schemes accounted for the time lags between representational and lan-
guage milestones for some subjects.
Internal representation in phonology
In child phonology the term internal representation is intended to characterize
underlying aspects of the childs understanding and production of speech.
Diagrammatic descriptive models typically are used to characterize the structure
imputed to the childs system. The childs internal representation and the
linguists description thereof are sometimes assumed to be isomorphic. When
the linguist claims psychological reality for internal representation, that reality
can best be considered roughly equivalent to the psychologists term mental
representation, which is a form of mental processing, or a contentful state of
the organism. The linguists model attempts to describe the organization or
complexity of the relations among elements imputed to the childs system as
evidenced by systematic relations among the utterances produced. The model
is, therefore, a characterization of information about the childs system as we
know it, whereas the childs internal representation is a system of unknown
parameters capable of generating the utterances appropriately described by the
model (see Van Gulick 1982).
Internal representation generally is taken to refer to a form of mental storage
(e.g., Locke 1988; Menn and Matthei 1992). For example, Locke (1988) argued
that we cannot know whether a phonology is needed until it is determined that a
discrepancy exists between stored and produced patterns. Nor can we propose
explicit phonological rules until we have inferred the phonetic structure of internal
representations (p. 4). Developmental approaches to internal representation are
rare (but see Velleman 1992; Waterson 1981); to our knowledge, there has been
no previous effort to explicitly relate the linguists internal representation to the
psychologists mental representation.
In our view, the production of context-limited words, which have been
observed to occur prior to the onset of combinatorial representational play
(McCune 1992), is in some ways comparable to the perceptually dependent
memories of the Rovee-Collier experiments. Aspecic familiar event evokes an
intentional state and associated vocalizations (Bloom 1991). There is minimal
How abstract is child phonology? 265
differentiation among such components as speaker, hearer, physical context,
and vocal motor action (Werner and Kaplan 1963). Such production requires the
availability of one or more vocal motor schemes, but not a word production
pattern generalized across a variety of word types.
Somewhat later, a newcapacity for mental representation allows the elements
of the speech situation to be differentiated, yet integrated. An unfamiliar event
might now evoke the word associated with a related event. A given event might
be referenced by one of several words. The child experiences an increasing
range of potential representational meanings, whereas vocal motor skill may
remain limited. Vocal expression thus comes to rely on the word production
patterns which now evolve, made possible by the increased capacity for
separating word form from situation of use and for the internal experience of
relationships between linguistic elements, which facilitates the juxtaposition of
one or more vocal motor schemes and a range of phonetically related adult
words.
Storage need not be postulated at either of these developmental time points.
Whereas in the earlier period a familiar context may provide sufcient percep-
tual support to elicit instantiation of the one vocal motor scheme associated with
the situation, in the later period instantiation of a variety of situationally
appropriate adult-based forms is possible, given adequate articulatory capacity.
Furthermore, once the childs vocal forms are no longer embedded in a partic-
ular situation of use, they can be compared or superposed (as in connectionist
models, such as Stemberger 1992) as a basis for the development of a general-
ized word production pattern. We believe that this is the basis for the beginnings
of phonological systematization, which we nd to emerge at about the same
time as the rst referential or generalized use of words. At this point, something
more abstract than a vocal motor scheme operating in combination with per-
ceptual attention to particular auditory patterns has materialized. This is a
phonological mental representation which may become instantiated when refer-
ence to the corresponding object or event is contemplated.
Modeling the childs system
There is a wealth of persuasive evidence regarding the importance of the word
and syllable levels in child phonology, particularly in the early period (Chiat
1979; Ferguson and Farwell 1975; Kent and Bauer 1985; Macken 1979;
Vihman 1992). A hierarchical model of phonological structure is needed to
capture this important aspect of childrens systems. It appears to us that the most
viable option currently available is nonlinear phonology.
Several phonologists have suggested nonlinear models of early child words,
especially for the frequently observed patterns of harmony and reduplication.
The task has been approached in a variety of theoretical frameworks, including
prosodic (Waterson 1971), parametrical (Fikkert 1991; Lle 1992), connectionist
(Berg 1992; Menn and Matthei 1992), and cognitivist (Menn 1978; Velleman
266 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
1992). All of these approaches seek to account for the fact that the childs
phonology is simpler than the adults, at least at the output level. However,
depending on the child and/or the model, simple may have many different
meanings; these will be reviewed briey.
A child system may be simpler in its hierarchical structure, lacking whole
levels of representation (e.g., the skeletal or segmental tier). Menn and Matthei
(1992), for example, suggested within a connectionist framework that priming-
type interactions among similar articulatory patterns (words) may induce the
beginnings of autosegmental structure potentially . . . without segmentation
below the word level (p. 243). Velleman (1992) suggested that the phonological
representations of children with highly restrictive, often babble-based word
recipes may have lexical representations with almost no structure at all (e.g.,
word-level representation only), relying on their existing articulatory patterns to
provide whatever redundant phonetic detail is required to esh out productions.
3
The childs representation also may be simpler within a given level. For
example, early syllable constituents may be nonbranching (Fee 1991; Fikkert
1991; Lle 1992; Ohala 1991; Velleman 1992). That is, only CV syllables may
be possible at rst. If no branches are available at the word level (i.e., if all words
are monosyllabic), then word and syllable are synonymous and no lexical
distinction need be made.
The childs system may be less integrated, with consonant and vowel effects
occurring independently due to planar segregation (Fikkert 1991; Lle 1992;
McDonough and Myers 1991; Macken 1993; Velleman 1992), in which con-
sonants and vowels occur on separate phonological tiers. Such segregation was
originally proposed for Semitic languages in which morphological templates
require either consonants or vowels, for other types of templatic morphology in
which linear order of consonants and vowels is redundant, and for languages
with very simple CV phonotactic structures (Lle 1992; McCarthy 1989).
Planar segregation in child phonology allows vowels to be transparent to
consonant harmony, and vice versa, and accounts for the increased frequency
of such harmony processes in early phonologies. It also provides a model of
C/V metathesis that is consistent with principles of adult phonology: if conso-
nants and vowels are on separate tiers, then they may appear to switch places
without violating constraints against crossing association lines.
Childrens lexical feature specications may be minimal as well. Such
underspecication is identied in child phonologies when elements of surface
form are completely predictable. This may stem from pervasive harmony
patterns, in which the degree of feature spreading is so great that the positions
receiving harmony are thought to be vulnerable to spreading due to the lack of
any feature specication of their own. For example, if a child demonstrates
regressive and progressive harmony affecting target coronals (or dento-
alveolars) whenever a labial or dorsal (velar) consonant occurs anywhere in
the word, then we assume that [coronal] is not underlyingly specied.
Redundancy, or predictability in surface features, also may stem from phonetic
How abstract is child phonology? 267
or phonotactic restrictions. For example, if all consonants in a childs systemare
stops, then [continuant] is predictable and need not be lexically marked.
Features also may be specied but lexically unordered where their order is
predictable. This lack of lexical ordering is manifested in apparent C/C or V/V
metathesis in children (e.g., Virves productions of [asi] for /isa/ father and
[am-i] for /ma/ mother, described in Vihman 1976) and in redundantly
ordered complex clusters in some adult phonologies. (See Velleman 1992 for
further discussion.)
Sometimes an unspecied feature also will serve as a default feature value, to
be lled in at the surface level whenever the corresponding C or V slot remains
otherwise unspecied (Fikkert 1991: Lle 1992; Velleman 1992). Although
coronal has been proposed by some as both a default and a lexically unspecied
feature for adult languages (Stemberger and Stoel-Gammon 1991), a childs
unspecied features need not necessarily be defaults (Lle 1992). Either spread-
ing of a harmonic feature or the lack of any surface realization (omission) may
be the fate of such unspecied elements. Although some features may not need
specication, feature specications that are necessary may encompass a broader
domain than in adult phonology, applying to an entire mora, syllable, word, or
even phrase (Iverson and Wheeler 1987; Velleman 1992). The eventual trickle-
down of such features to the segmental level has been referred to as deauto-
segmentalization (Goldsmith 1979; Spencer 1986). Similarly, rules or processes
may showa greater breadth of application. For example, spreading of a particular
feature may affect all possible recipient segments in either direction (right or
left) over a large domain, such as an entire phrase (Lle 1992).
Whether these options are available to all children, specied by innate param-
eters, determined by some characteristics of the language to which the child is
exposed, chosen by the child based on idiosyncratic perceptual, physiological,
or cognitive biases, or some combination of the above is an open and widely
debated question. In any case, the course of phonological development includes
the addition of complexity to any or all of these aspects of the representation.
We prefer to attribute the minimal possible structure and the fewest possible
rules to the child at any given point in development and will attempt to
demonstrate that a nonlinear model can be constructed to account for devel-
opmental increments of phonological complexity, attributing complexity to
representations rather than rules and adding rather than changing structure
over time. Given our assumptions about the origins of early words in production
and perception and about the relation of emergent phonological systems to the
childs evolving representational capacity, we see no need to posit specically
linguistic innate structures (contra, e.g., Macken 1992).
Identifying the onset of system
Although the emergence of phonological systemis clear, even dramatic, in most
of the children we have observed, bits of the system typically are already
268 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
apparent (at least in retrospect) before they come together sufciently to lead to
a sudden ourishing of diverse lexical items. The system itself coheres gradu-
ally, over time, but when a critical point is reached (either cognitively or
phonologically; it may be impossible to decide which is determinative), the
system seems suddenly to have power enough to strongly affect lexical choice
and production and to assimilate adult words that do not provide an obvious t
with the childs template. Close analysis of the phonological progress of two
children, reported below, will demonstrate that the onset of system can be
recognized not only in the interesting cases of distortion of adult models
(regression in accuracy) in child word production, but also in the spurt in
acquisition of words that t the template.
Extra-systemic elements
Adult phonological systems include marginal elements, especially in words
which are salient because they are exotic (ZsaZsa), humorous (schmaltzy), chic
(au jus, karaoke), or newsworthy (dtente, Sri Lanka, Schwartzkopf). Similarly,
the childs production may include a small set of extra-systemic words, recog-
nizable by their inconsistency with the majority of the childs forms. Children
sometimes produce surprisingly accurate renditions of difcult words before a
system has coalesced (e.g., Hildegard Leopolds famous production of pretty).
Such progressive idioms may be regularized when the childs phonological
system has become established, or they may persist as extra-systemic elements.
Later, extra-systemic words may reect aspects of the adult shape as perceived
by the child as well as aspects of the childs existing template, and so include
both systemic and extra-systemic elements.
Words that are partially or wholly extra-systemic can be expected to be
shorter-lived than other forms. The childs system, by denition, is more
consistent and persistent than other aspects of production and tends to dominate
lexical production and to dictate selection, once it is in place. However, extra-
systemic items may serve as precursors or even triggers for change in the childs
system; the system may accommodate to them some time after they rst appear
as marginal elements.
Two phonological proles
In a paper that focused on syllable production, Vihman (1992) presented
sketchy proles of the initial steps in lexical development of two children as
well as the syllables they practiced at 911 months. One of these children
seemed to base his early phonology on the syllable. The other child seemed
instead to operate with a phonetic gesture involving tongue fronting and raising,
or palatal articulation; the syllable did not play an important role for her. In this
chapter we consider the phonetic and phonological development of the same
children in ner detail, attempting to trace the interaction of perceptual biases,
How abstract is child phonology? 269
vocal motor schemes, and representational capacity in the formation of phono-
logical systems.
Timmy: the syllable pattern
Timmy provides an example of a child who progressed phonetically rather
slowly and apparently effortfully. His words and babble forms were unusually
difcult to distinguish for several months (Vihman et al. 1985). His rst six
months of word production were based largely on phonetic variants of a single
syllable shape, <Ca> with a gradual increase in the consonantal choices avail-
able. Nevertheless, it is possible to distinguish an early, presystematic period
(913 months) and a later, system-based period (from 14 or 15 months on).
At 9 months Timmy already responds with <ba> to adult monosyllabic /b/
words (ball, block); by 10 months he produces <ba> spontaneously in situations
associated with those words (at 10 and 11 months <ba> is also produced in
imitation of basket, bell, boat, book, button and spontaneously for box; by 15
months bird, brush, bunny, baa(-baa) are produced as <ba>).
4
From 11 months
on Timmy responds to /k/ words (kitty, quack-quack, car, duck, key) with <ka
(ka)>.
5
The word-length distinction (between monosyllabic /b/ words and
disyllabic /k/ words) derives from the models, but is maintained somewhat
inconsistently, particularly after the rst month of use for each word.
6
There is
little evidence of a phonological system operating here. Instead, Timmy draws
on one of the articulatorily simplest syllables, [ba] (Davis and MacNeilage
1990; Vihman 1992), when he is reminded to produce his matching vocal
motor scheme by situations in which a familiar auditory pattern is commonly
produced by adults (in relation to some of his favorite toys, balls, blocks, bells).
Similarly, he produces his second vocal motor scheme, <ka(ka)>, in situations
associated with stop-initial word forms other than /b/ (car, kitty, Teddy, later also
a deictic form which begins as a response to Great Gable, referring to a
frequently identied drawing of a mountain; Vihman and Miller 1988).
It is only at 14 months that we see the extension of this pattern, rst to a single
word, eye, assimilated to Timmys pattern as [ja], then (at 15 months) to words
that elicit a range of different consonants: [a] for words characterized by
labiality and continuant friction, rst Ruth, with its rounded initial approximant
and nal fricative, later re, ies, owers, and plum [cf. Waterson 1971];
7
[j
-
a]
for light, where the palatal place appears to derive from the nuclear diphthong
while the stop articulation derives from the nal consonant; [na] for nose and
later Nana; and [ja] for ear, hair as well as eye these latter perhaps best
glossed, together with the probable phonological model eye, as response to
questions about my body.
At 15 months, furthermore, Timmy for the rst time produces two forms
outside his vocal motor scheme, both involving special sounds or sound effects
in the adult models: hiss is reproduced as [s
()
()
(C)
V V
([V place])
([V place])
([V place])
([V place])
{(C)}
{V} {V}
16 months (b)
( ): optional elements
{ }: emerging elements
: syllable
14 months
<ba, ka, ja>
<ba, ka, ja, na, a, ja>
Word
[place, manner]
Word
[C place, manner]
Word
[C place, manner]
Word
place of
articulation
Figure 9.2. Development of a lexical representation
272 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
Table 9.1 Development of a childs lexical representation (Timmy)
Portions provided by
Motor control Perception
Lexicon (phonological representation)
Age (in months) Word forms Vocal motor scheme Adult Model Feature geometry Autosegments Skeleton
10 <ba> Ca syllable
1113 <ba>, <ka> [a], +/ labial
open syllables
iteration
14 <ba>, <ka>, <ja> [a], labial default
open syllables
iteration [3-way place contrast] {W}
15a <ba>, <ka>, <ja>
<na>, <a>, <
j a>
[a], labial default
open syllables
iteration [u], [s
m], [ni
m n m ni
j a na ja
15b ba ka a
j a ma na ja
16a ba ta ka a sa
j a ma na ja
ti ki
16b ba ta ka a sa
j a ma na ja (wa)
bi ti ki mi ni
ba tu ku (dzu) mu
Note: a = earlier in month, b = later in month.
276 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
Alice: the palatal pattern
Alices phonological development illustrates the emergence of a far more
complex initial structure. Alice appears to organize her phonology on two
independent planes at once. At the autosegmental level, she gradually works
her mastery of the motoric control needed to produce a palatal glide, [j], into a
word-based palatal melody. The melody may be seen to evolve gradually out of
the words which Alice selects or attempts to produce, words that naturally
accommodate her preferred phonetic gesture, [j] (e.g., hi, baby at 10 months). At
rst the melody is applied inconsistently in production, to whole words (no
[nj]: 9 mos., bottle [bj]: 11 mos.), then to both words and syllables (dolly
[dali:], elephant [ni a]: 13 mos.). At each of these levels Alice explores a
variety of options. We identify the beginnings of a phonological system at 14
months, when a single relatively consistent word production pattern begins to
be applied to a range of different words. Some examples:
baby [be:bi] blanket [bi]
bottle [bad i] mommy [ma:i]
Bonnie [ban i]
The pattern found in these productions is related to the form of the adult models,
but cannot derive from them alone; it ts closely with baby and Bonnie, but
distorts blanket and mommy.
Figure 9.3 tracks over time the emergence and decline of the various elements
which participate in the formation of the systemin evidence at 14 months. Three
patterns are isolated and identied in the order from most to least complex or
inclusive (beginning with the bottom-most panel): <CoVCi> , or polysyllabic
word shapes including a nal [i] (e.g., baby, Bonnie, mommy); <Vi>, or mono-
syllabic word shapes including a front rising diphthong (e.g., hi, Ais the childs
nickname, which rhymes with haze); and <jV>, or any other word shapes that
include the glide yod ([j]: e.g., yumyum).
At 8 months only babble was produced; there were no identiable words.
Babble vocalizations tended to include yod to an uncommonly great extent
(24 percent vs. a mean of 6 percent for nine other American infants; McCune
and Vihman 1987). At 9 months three words were identied, realized in seven
tokens. The <jV> pattern is incorporated into two of these tokens: hello/hi[ya]
[hije], no [n:j]. No other palatal pattern is used in words in this month.
At 10 months we see the rst and strikingly high use of the <Vi> pattern in
words, accounting for 50 percent of all word tokens. Two likely adult sources of
this pattern for Alice are hi and baby, words she produces accurately, with
syllable count and nuclear-syllable consonant and vowel matching the adult
form. Both words constitute plausible models for the shaping of a palatal
articulatory gesture in the direction of adult speech.
Figure 9.4 displays the use of all palatal patterns combined in babble as compared
with words to facilitate tracing the emergence of a word schema out of the babble
How abstract is child phonology? 277
repertoire. Here we see a sharp increase in word production at 10 months (to twenty
wordtokens), withproportionate increase inpalatal patternuse, while babblingitself
shows little change. Babbling shapes foreshadowword shapes, as we see in the rst
wave of palatal patterning in babbling at 812 months perhaps reecting the
100
90
80
70
60
50
40
30
20
10
P
e
r
c
e
n
t
u
s
e
o
f
p
a
l
a
t
a
l
p
a
t
t
e
r
n
0
<jV>
8 9 10 11 12 13 14 15 16
100
90
80
70
60
50
40
30
20
10
P
e
r
c
e
n
t
u
s
e
o
f
p
a
l
a
t
a
l
p
a
t
t
e
r
n
0
8 9 10 11 12 13 14 15 16
<VI>
8 9 10 11 12 13 14 15 16
100
90
80
70
60
50
40
30
20
10
P
e
r
c
e
n
t
u
s
e
o
f
p
a
l
a
t
a
l
p
a
t
t
e
r
n
0
Months of age
<CoVCi>
Figure 9.3. Palatal pattern use in words
278 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
childs global auditory representation of words like mommy, daddy, baby, hi, and her
own nickname, Ais. Babbling also reects newly emergent patterns rst attempted
in word production (as illustrated in Elbers and Ton 1985), as we see in the second
wave of palatal patterning in babbling, at 1316 months, covering the period in
which word production shows a dramatic palatal-pattern-based increase. Only the
emerging lexicon shows sharp or apparently categorical changes from month to
month, however, reecting the ongoing phonological work of construction and
reorganization or systematization.
Looking over the changing patterns in Figure 9.3 in the remaining months, we
see that all three patterns are used in at least 10 percent of Alices word tokens at
11 months. From 12 months on (when word production drops temporarily), the
<jV> pattern is replaced by the other more differentiated patterns. Two patterns
compete at 13 months, the <CoVCi> pattern dominating from 14 months on.
100
90
80
70
60
50
40
30
20
10
0
100
90
80
70
60
50
40
30
20
10
0
8 9 10 11 12 13 14 15 16
Months of age
Palatalization in word tokens
Word tokens
Babbling
Palatalization
T
o
t
a
l
b
a
b
b
l
e
v
o
c
a
l
i
z
a
t
i
o
n
s
T
o
t
a
l
w
o
r
d
t
o
k
e
n
s
Figure 9.4. Raw frequency of vocalizations and palatal pattern use
How abstract is child phonology? 279
Whereas one or at most two different palatal patterns had been used in earlier
months, at 14 months a full range of possibilities is explored, with some words
varying across subpatterns. For example, tokens of daddy vary between the fairly
accurate [tdi], a <Vi> form[tardi], and a more fully palatal [jrji]. Similarly, the
word hi, a staple of Alices lexicon for ve months, now receives experimental
shaping into [ha:ji]. It is worth noting that at 14 months, when the majority of her
productions (42 tokens) are disyllabic [i]-nal word shapes, almost all of these are
in relatively good conformity with the adult model; similarly, the diphthongal
production of words such as bye and eye (also kay and nigh-nigh at 15 months)
owes as much to the model and the childs evident experience of a match as to
assimilatory or creative reconstruction by the child.
Until nowAlices words seemto have been selected, at least in part, on the basis
of the increased salience of palatals. However, her palatalization pattern has
appeared to exist in some sense separate from its manifestation in any particular
word, because it is imposed inconsistently on various portions of different words,
even within the same recording session, and its effects vary from one token to the
next, even of the same word. Palatalization appears to have the status of an
autosegmental melody for Alice, independent of the segments on which it operates.
Furthermore, this palatal melody can be seen as a direct outgrowth of the phonetic
gesture [j] which marked Alices vocal production at 810 months.
We propose that the vocal motor scheme which is rst manifest as [j] is
gradually shaped into the more extensive and exible palatal melody expressed
in both mono- and disyllabic words from 14 months on. Until a range of
different words are produced in a phonologically consistent way, we have
vocal production under the dual inuence of the infants own previous vocal-
izations and prior experience of adult vocalizations, each embedded in a
familiar situation of use. Once a stable word production pattern is established,
we can infer the existence of a phonological system. This system is emergent at
14 months, as Alice experiments with different palatal patterns for old,
previously palatalized word shapes.
At 15 months Alices palatal pattern is no longer independent of the under-
lying word shapes; it has begun to shape them. Words now begin to change in a
way that cannot be accounted for by the shape of the adult model. A number of
the relatively accurate earlier shapes have been replaced by disyllabic [i]-nal
renditions that also incorporate yod, which had been submerged earlier in the
more abstract realization of palatal articulation affecting different parts of the
word vowel nucleus ([Vi]), palatalized stop or nasal ([d ], [n ]), and nal [i].
Now we see a resurgence of intervocalic yod in forms such as blanket [baji] and
dolly and daddy [daji]. The manifestation of palatalization has become system-
atic, reminiscent of the spread of tones in tone languages or of nasalization in a
language like Guarani. This is the sign of an active phonological system
exerting an inuence on production patterns, where earlier those patterns
merely reected various possible interactions between Alices well-developed
motoric capacity and her auditory experience of the adult language.
280 Marilyn M. Vihman, Shelley L. Velleman, and Lorraine McCune
A clear picture of Alices phonological system now emerges. Palatalization is
redundant everywhere except in word onsets. Medial consonant features never-
theless will be specied in most words, as they may emerge in any one of three
ways: intact, palatalized, or replaced by Alices default, [j]. First-syllable vowels
must be specied, because these are not predictable. However, each initial syllable
must include two morae, the rst of which will be specied whereas the second
will be either lled by the default palatalization or left unrealized. Production
factors may determine which of the consonantal and vocalic options occurs as the
output form. Alice seems to allot motoric attention to the onset of a word pattern
and then, in the remainder, allowher default palatal to ll in wherever attention or
articulatory agility fails. In some cases (e.g., lady at 16 months) even the initial
consonant is lexically unrealized at times, and is therefore supplied with the
default [j]. Vowels in second syllables remain unspecied as they are redundantly
palatal, with one or two extra-systemic exceptions (e.g., one production of blanket
as [bo], hammer as [hv:a]).
These aspects of Alices system provide an interesting contrast to
Timmys. Her motor skills are more developed than his, as reected in far
greater phonotactic and phonetic variety. She is willing to experiment (at 14
months) with various ways to integrate her palatalization pattern into her
phonological system. Like Timmy, she is able to underspecify some elements
in her lexical representation because she has a preexisting well-practiced
motor pattern which will ll in for them. However, she shows greater
variability in use of her pattern; production variables as well as lexical
organization inuence the extent to which it is used for any given word
token. Whereas the concept of planar segregation can serve to simplify our
model of Timmys phonology, it cannot account for Alices system. The
primary reason it cannot is that our autosegmental representation of Alices
pattern must include branching within the syllable, with two morae available
for her frequent diphthongs and occasional CVC forms (e.g., clean [kin]). In
the absence of a simple CVCV phonotactic pattern, the relative order of
consonants and vowels is not predictable, and planar segregation is ruled
out. It is in any case unnecessary here, because the palatal melody affects
consonants and vowels alike.
Let us nowconsider the last month for which phonological data are available,
16 months. Alices polysyllables have begun to return to the balance reected at
14 months; there are no new examples of regression affecting formerly
correct forms, although experimentation continues in words which are dif-
cult for the child, such as mommy, now sometimes produced with medial [m],
sometimes with [n. ], and lady, in which the initial lateral and the medial stop are
both subject to replacement by yod in variant tokens. The word that gives Alice
the most trouble is elephant. The toy set which engaged Alice at each recording
session included a Jack-in-the-box elephant. In a classic illustration of a child
valiantly attempting a situationally salient word with an alien phonological
shape, Alice progressed from [e:], [a] or [ni] at 13 months, to [anj],
How abstract is child phonology? 281
[aij] or [ j
a
i u
e o
[] ()
a
[ ] = segments produced only as substitutions for adult segment, never as match-to-target
( ) = phones produced in only one word
302 Daniela Oliveira-Guimares
Note that Lucass rst words mainly have a weakstrong stress pattern,
despite the fact that the adult targets tend to have penultimate stress (see
Table 10.8, which quanties word length and stress in the rst three sessions).
Case study 2: Paulos phonological development
Table 10.9 provides an illustration of the word templates found in Paulos
sessions 1 (1;11.13) and 2. Note that although some words are clearly selected
and others adapted, others may be close to the model and thus accurate or
selected in some respects but modied to t the childs template, or adapted, in
other respects.
The strongest template in these two sessions involves full reduplication, with
an optional offglide that occurs when stress falls in nal position. Like Lucas,
Paulo usually changes the stress from the penultimate to the nal syllable
(Table 10.10, based on the rst session). There is also a CV(V) template.
The reduplication template occurs in the rst two sessions, in both selected
and adapted words. The CV template occurs in the rst session in selected
words only and in the second session in adapted as well as selected words.
Table 10.9. Examples of Paulos word templates
Template
Session 1
(1;11)
Adult
form
Child
form
Session 2
(2;00)
Child
form
C
1
V
1
C
1
V
1
(V)
(Selected)
vov
grandmother
vv [vv] mame mother mamj [mmj]
papai father papaj [papaj] vov grandfather vovo [vovo]
CVCV(V)
(Adapted)
Letcia (name) leisa [tata] tartaruga turtle tartag [tata]
Luciana (name) lusina [uu] Roseli (name) hozeli [ii]
CV(V)
(Selected)
po bread pw [p] p foot p [p]
banho bath bj [bj] por to put por [po]
CV(V)
(Adapted)
Silene (name) silen [e]
Edmar (name) emar [ma]
Table 10.8. Word length and stress for Lucas
Adult target Child production
Monosyllables 18/72 (25%) 45/83 (37%)
Iambic 30/ 72 (42%) 42/83 (51%)
Trochaic 24/72 (33%) 4/83 (5%)
Word template development in Brazilian Portuguese 303
Table 10.11 gives an overview of the rst two sessions, those in which these
templates are strongest.
In both sessions the majority of words are selected. Harmony (a phonological
process that provides consonant assimilation in place or manner) and redupli-
cation are the main processes that give rise to the C
1
V
1
C
1
V
1
(V) template in
Paulos data. Here reduplication (or harmony) is generally, but not always, a
consequence of regressive assimilation, in which the child copies or antici-
pates the following segment. Examples, from sessions 1 and 2, are given in (1):
(1) Orthography Target form Child form
sapo sap papu
copo kp pp
Another notable characteristic of Paulos rst words is the very extensive use of
labial consonants. Table 10.12 and Figure 10.4 show the proportion of use of
labials and dentals over ve sessions, over all words produced. We count as
labials words that have labial consonants exclusively, including [m], [b], [p],
[v], [f], and as dentals words that have alveolar/dental consonants exclusively,
including [d], [t], [s], and [z]. Labials + dentals refer to words which include
both, while other refers to all words which include a labial or dental with
another consonant type.
As we can see, Paulos rst words consist mainly of labials. The proportion of
labials gradually decreases while alveolar production increases. In fact, in
sessions 4 (2;2.20) and 5 (2;3.22) we can see a kind of alveolarization of
Paulos consonants, as both alveopalatal fricatives and velar stops tend to be
substituted by alveolar stops (e.g., girafa [diafa], aqui [ati]). This is related to
the move from templates to segments, as the alveolarization seems to be related
to liberation from consonant harmony.
Table 10.11. Overview of Paulos sessions 1 (1;11) and 2 (2;00)
Session
Total number
of words
Selected
items
Adapted
items
Reduplicated
CVCV CV Others
1 24 15 (62%) 9 (38%) 13 (54%) 7 (30%) 4 (16%)
2 31 20 (64%) 11 (31%) 9 (29%) 10 (32%) 12 (39%)
Table 10.10. Word length and stress for Paulo
Adult target Child production
Monosyllables 7/64 (11%) 20/86 (23%)
Iambic 27/64 (42%) 42/86 (49%)
Trochaic 30/64 (47%) 24/86 (28%)
304 Daniela Oliveira-Guimares
Paulos reduplicated and CVpatterns are mainly seen in the rst two sessions.
In the following sessions we see some word forms adapted to t the CVCV
template generally fossilized forms (i.e., inaccurate child forms that remain as
such for some time, even when new words with the same sounds are produced
accurately), such as proper nouns. Table 10.13 shows all of the adapted forms
found in session 3 (2;1.28).
Of the 69 word types that occur in session 3, 17 child forms (29 percent)
exhibit some kind of harmony or reduplication or, in other words, some degree
of adaptation. The other words exhibit no harmony and most are produced with
variegated consonants. In session 4 (2;2.20) some adapted reduplicated forms
can still be found. These forms are illustrated in Table 10.14.
There are fewer adapted words in session 4 than in session 3 and new words
adapted to a reduplicated template are rare in the following sessions. Some
Table 10.12. Paulos labial and dental production over ve sessions
1
Session 1 Session 2 Session 3 Session 4 Session 5
N (%) N (%) N (%) N (%) N (%)
Labials only 17 (58) 26 (64) 40 (48) 27 (34) 16 (11)
Dentals only 4 (14) 5 (12) 25 (30) 29 (37) 52 (37)
Labials + dentals 4 (14) 5 (12) 1 (1) 7 (9) 25 (18)
Other 4 (14) 5 (12) 18 (21) 15 (20) 47 (33)
5 4 3 2 1
10
20
30
40
50
60
70
0
Labials
Dentals
Session
%
O
v
e
r
a
l
l
w
o
r
d
s
Figure 10.4 Labials and dentals over ve sessions
Word template development in Brazilian Portuguese 305
words are crystallized, frozen or entrenched forms, which retain the tem-
plate shape for a long time. These forms may have been adopted and reinforced
by the adults, in their child-directed speech. In the last sessions templates apply
only to specic words that appear to have become entrenched, such as proper
nouns. For example, at 1;11 tartaruga turtle /tahtauga/ is produced as [tata]
and Roseli (name) /hozeli/ as [ii]; at 2;10 tartaruga has become [tahtaga]
but Roseli remains unchanged. We can also see the child using adapted
CVCV forms in some phrases, such as, for example, comeu bolo eat cake
Table 10.13. Paulos template words produced in session 3 (2;1.28)
Orthography Gloss Adult form Childs form
1 chapu hat apw ppw
2 chapu it is the hat apw ppw
3 chapu pega chape hat take the hat apwp apw pwpppw
4 Dani tartaruga Danis turtle danitartauga danitata
5 dormi na gua sleep in water duminaaga mimiaaka
6 dormindo sleeping dumin mimiunu
7 dormir sleep dumi amimi, mimi
8 Duda (name) duda dutu
9 Felipe (name) lip pipi
10 o Felipe it is Felipe ulipi upipi
11 hipoptamo hippopotamus hipoptm ppp, papap
12 Letcia (name) leisa tata
13 peixe sh pe pepi
14 prncipe prince psip ppi
15 sapato shoe sapat papapu
16 tartaruga turtle tartauga tata
17 umbigo navel bigu bibi
Table 10.14. Paulos template words produced in session 4 (2;2.20)
Orthography Gloss Adult form Childs form
1 acabou nish akabo bobo, abo
2 atender answer atde dede
3 dormer sleep dumi mimi
4 dormir na gua sleep in the water duminaaga mimiaga
5 duro hard duu dutu
6 entendeu understood tde dedew
7 peixe sh pe pepe
8 Roseli deu Roseli gave hozelidew iideu
9 sapato shoe sapatu papapu
10 tartaruga turtle tartauga tata
11 por tartaruga put turtle portartauga potata
306 Daniela Oliveira-Guimares
/kumewbolu/ [memewbolu] (session 7, 2;5.20). In this case only the word
comeu is adapted to a reduplicated form as part of a combination. This provides
some evidence for the template being deployed as a way of dealing with
combinations, but the current data are insufcient for testing the extent of this
possible template function.
As the use of templates decreases in the later sessions new forms are emerg-
ing. These new forms do not t into a simple CVCV pattern or any other set
schema or template. At this point we can more usefully analyze the relation
between child forms and adult targets by making reference to phonological
processes or rules, which apply segment by segment, with a straightforward
alignment between target word and child form. Thus we see the template
gradually fading out and the segment emerging as an important unit of phono-
logical organisation. To illustrate, we show the evolution of the phonetic forms
of the word tartaruga turtle, in Table 10.15.
The word tartaruga is rst adapted to a reduplicated template as [tata]
(sessions 1 to 4). In session 4 the adapted form is tted into a combination:
por tartaruga /pohtahtauga/ [potata] put turtle. In session 5 we see the
emergence of a new representation. In this session tartaruga takes three differ-
ent forms, which suggests an unstable representation, characteristic of a period
of transition in this case, from template to segment-based phonology. From
session 5 on, the relationship between Paulos productions of tartaruga and the
adult target can be analyzed in terms of substitution and deletion processes;
there is no longer any reason to refer to the application of a holistic template. We
can see these changes in other words, such as Luciana (name) /lusina/, which
is rst pronounced as [uu], tted into a reduplicated template, but which in
session 7 is pronounced variably as [sina] and [na]. Here again we can see
the emergence of a form that is closer to the adult target.
Table 10.15. Changes in the production of one word over twelve
sessions, both in isolation and as part of combinations
Session
tartaruga turtle
tartauga Adult form
Gloss (if not
simply turtle
1, 2, and 3 tata
4 potata por tartaruga put turtle
tata [pohtahtauga]
5 tataluga,
tatautugu, tatau
6 tauga
7 tatauga
tatauganadanaag
w
a tartaruga nada na gua
[tahtauganadanagwa]
turtle sleeps
in the water
8 and 9 tatauga
11 tatauga
tataugapkena tartaruga pequena small turtle
Word template development in Brazilian Portuguese 307
Another important piece of evidence for the emergence of segmental
phonology is variation in the production of segments in different tokens of
the same word. From session 3 (2;1.28) onward variation in word forms
becomes more frequent. The same target segment is pronounced differently
in different words and even in different tokens of the same word. For
example, the word bruxa witch is pronounced as [bua], [buta], and
[bua] in the same session. It is as if Paulo were playing with sounds, trying
out different forms. We can see many examples of variation in the production
of specic segments, as in the case of the word bruxa. In Table 10.16 we
provide one example from each of the later sessions to show this variation,
which we take to reect the process of reorganization and the gradual
emergence of the segment as a unit in the childs phonological grammar.
Note that in those words which contain more than one supraglottal consonant,
the variability involves one of those consonants only; in almost no case does
it reect the inuence of the place of articulation of the other consonants in
the word, in contrast to the effect of harmony and reduplication seen in the
earlier sessions. Thus, this variability shows that at this point the child is
dealing with segments, not with holistic templates.
Table 10.16. Word form variability in nine sessions
Orthography Gloss Session Adult target
a
Child forms Target segment
1. tirar take it 4 ia ia
ia
dia
affricate
2. foi he/she went 5 foj toj
oj
labiodental fricative
3. cad Where is it? 6 kade kade
ade
de
velar stop
4. Letcia (name) 7 leisa isja
sa
affricate
5. aqui here 8 aki akia
ati
velar stop
6. nmero number 9 nume numi
numelu
nume
ap
7. desse aqui of this here 10 desiaki desiki
deiki
alveolar fricative
8. palhao clown 11 palasu pajasu
palasu
lateral
9. trilho rail 12 til ilu
tilu
cluster
a
Target segment is underlined.
308 Daniela Oliveira-Guimares
Overview of Paulos segmental inventory
To analyze Paulos move to segmental representation it is important to have an
overview of his phonological inventory. Table 10.17 shows Paulos segmental
inventory at three developmental points: session 1 (1;11.13), session 6 (2;4.21),
and session 12 (2;10.20). This table shows the small inventory of consonants in
the beginning, when most of his words occurred in a templatic form, and the
increase in his inventory around the time when the segment became a functional
unit of representation.
In session 1 only labials occur in word onset and medial positions, except for
[t] and [h] (which occur once each). The only fricative is the labiodental [v],
which occurs in both positions. In session 6, we see other fricatives emerge,
especially in word-medial position. Palatal sibilants occur only in substitution
for other consonants. The voiced velar [g] does not occur. We nd target-like
affricates in initial position and lateral liquids in initial and medial position. In
stressed position the vowel system is complete in session 6. In unstressed
syllables the open medial vowels [] and [] occur only once each. By session
12 the consonant inventory has become quite large; only the tap [] and [h]
(corresponding to orthographic and historical r) are missing. The [h] found in
Table 10.17. Paulos segmental inventory at three points
Session 1
(1;11.13)
Session 6
(2;4.21)
Session 12
(2;10.20)
Consonant inventory Word onset p b (t)
v
m
b p t d k
(s) [] v
m n
(l)
p b t d k g
f v s z
m n
(l)
Word medial p b
v
m
(h)
p b t d k g
(f)(v)( s) (r) [] []
m n
l l
p b t d k g
f v s z
m n
l l
Word nal j w s () j w s j w
Vowel inventory Stressed (u)
(e) o
a
i u
e o
a
i u
e o
a
Unstressed syllable i u
(e) o
()
a
i u
e o
() ()
a
i u
e o
()
a
[ ] segments produced only as substitutions for adult segment, never as match-to-target
( ) segments produced in only one word
Word template development in Brazilian Portuguese 309
the rst session does not occur in sessions 6 or 12. In session 12 there are again
target-like affricates in initial and medial position. The bilabial, alveolar, and
velar stops occur accurately in onset position in the last session analyzed, as do
the labiodental, alveolar, and palatal fricatives.
A summary of template evolution
This study has explored the use of word templates in two children acquiring
Brazilian Portuguese. We have described the emergence and evolution of
templates over one year.
Both children started out using a reduplicative CVCV template. This pattern
is quite common in babbling and in rst words in most languages (MacNeilage,
Davis, Kinney, and Matyear 2000). CVis generally taken to be the least marked
syllable, occurring in all languages and emerging in children as the rst adult-
like or canonical syllable (Oller 2000). The use of a form which reduplicates a
CV syllable can be explained as being especially easy, both because CVis an
early-learned syllable, motorically accessible to the child, and because CVis the
most frequent syllable type in Brazilian Portuguese (Almeida 2005). It may be
assumed that repeating such a syllable twice is also relatively simple in terms of
speech planning, easier than coordinating two different syllables of any kind.
Stress may also play a role in characterizing a childs templates: Both Lucas
and Paulo tend to move the stress from the penultimate to the nal syllable in
their word templates. However, although nal syllable stress is not the most
common pattern in Brazilian Portuguese, it is very frequent in child-directed
speech. Input frequency may therefore explain this preference in both cases.
We have seen that both of the children followed in this study converged on
reduplication as their solution for meeting the challenge of producing adult
words. However, following the period in which the reduplicated template is
most active, the developmental paths taken by the two children diverged.
For Lucas, the rst child discussed in this chapter, a new template emerges in
session 3, represented by a bilabial nasal /m/ in word-nal position. Later on,
Lucass phonological systemundergoes reorganization, with the emergence of a
new template with a coda glide [w], which competes with the template form
with coda [m] until the form with [w] gradually comes to dominate. This
competition can be explained in part by the articulatory similarity between the
two bilabial phones, [m] and [w]. In the transition from [m] to [w] use,
competition between [m] and [w] was observed even in the same word and in
the same session. Some words changed pattern over time as the new word
template was adopted. Thus in Lucass case progress is expressed as a move
from one template to another.
In the case of the second child, Paulo, the decline of the reduplicated template
can be related to the emergence of the segment as a unit of phonological
representation. During this shift to reliance on segments we see unstable
behavior, involving play with words and sounds, which resulted in a lot of
310 Daniela Oliveira-Guimares
variation in the production of some words (see Table 10.16). Paulos redupli-
cation template gradually fades in the last sessions, remaining active only as
regards specic entrenched words, such as proper nouns. This is similar to the
developmental pattern in the case study reported by Macken (1979), in which a
childs template gradually faded as the child learned the contrast between
individual sounds.
General discussion
The two case studies presented here raise some issues regarding the develop-
ment of phonological knowledge which we would like to briey outline here:
the signicance of the occurrence of non-adultlike structures, variability in
word or segment production, the role of frequency in template formation, and
the relative role of the word vs. the segment in representations.
Lucass nasal-coda template is noteworthy because nasal consonants do not
occur in coda position in adult Brazilian Portuguese. This nding thus presents a
challenge for acquisition theories because the template, which Lucas uses
consistently over a period of several months, cannot derive directly from his
ambient adult language, nor are codas considered to be unmarked and thus
expected to occur in early words on the basis of such principles as the emer-
gence of the unmarked (McCarthy and Prince 1994; Gnanadesikan 2004). We
are thus faced with a case in which neither universal properties nor input
frequency can explain why a child produces a nontarget form. Thus a template
analysis can be a useful tool for discovering what sounds or structures represent
an articulatory challenge for the child, or conversely, which sounds or structures
a given child may nd easy.
There is some variation in both Lucass and Paulos data, but the variation
seems to apply to different-sized units. In Paulos case the variation may well
reect segmental learning (Ferguson and Farwell 1975). Such segmental learn-
ing means that a new, more detailed level of analysis has begun to develop, such
that the segment emerges as a functional unit alongside the word. For Paulo a
focus on mastering specic segments means that a single segment in the word,
the one being targeted perhaps, is produced with variability. This variability is
not affected by the other consonants in the word, and the variably produced
segment occupies a different place in different words, which shows that the
problem is not with any one word structure but rather with a particular segment.
Lucass productions, on the other hand, are relatively more stable throughout
the period of the study, perhaps because he has not yet moved to segmental
representations over this period. In Lucass case there is also competition
between two segments, [m] and [w], in the monosyllabic templates, which
could be taken as a case of variability in segment production. However, the
competition occurs between the holistic CV[m] or CV[w] forms, and not
between [m]s and [w]s in different positions, or even in coda position in
different word structures. The competition, in Lucass case, is not between
Word template development in Brazilian Portuguese 311
different renditions of a segment that appears in the target form, but rather
between two child forms, neither of which may be an accurate rendition of the
adult form. We see, then, that different types of variability or competition
between forms can be informative as to the size or identity of the unit that a
child seems to be operating with.
We have seen in Paulos data that the reduplicative template, once it has
mostly faded out of use, continues to apply only to specic words entrenched
as proper nouns. That raises the issue of the inuence of frequency on the
consolidation of a template. The frequent use of a word in a templatic form can
strengthen that form of the word, especially if it begins to be used by adults as
well as by the child. Reuse of this particular form, and even more so reuse by
more than just the child speaker, leads to this form being more readily acces-
sible, perhaps more strongly activated, more distinctly represented in memory,
than competing forms. It is likely that having the same form produced by other
speakers not only leads to this form being judged as accurate, due to its being
very close to the adult forms, but also to its creating a better dened and richer
representation, one which contains exemplars originating in more than one
voice. As long as this strengthening affects only a single word it will not
necessarily lead to the strengthening of the template, but instead will create a
phonological idiom. If more such words are reused and their representations
are strengthened in this way, such reuse could lead to the strengthening of the
template as a pattern that affects other words as well. The effect of frequency
here is parallel to that seen in language change, such that token frequency affects
changes that pertain to individual words (or entrenchment), but type frequency
affects generalization (Bybee 2001).
And nally, what status does the word have in a childs phonological system
once segments have begun to play a role as units of representation? As sug-
gested by Menn (1983), evidence that the word continues to serve as an
important unit of organization is the fact that the phonetic form mastered in
one word fails to occur in other words with similar targets. Phonological idioms,
or idiosyncratic forms produced by the child (Ferguson and Farwell 1975;
Menn 1983), would be a case in point. Paulo provides one such example. He
produced the word esse these /es/ accurately from the third session (2;1.28).
However, at the same time other words with [s] were not pronounced accurately.
For example, the same target /s/ is pronounced as /t/ in Sirlei (name) [tilei], and
as /d/ in senta sit down [dita]. There are also other cases in Paulos corpus of
variability in production of the same sound in different words, such as the
alveopalatal sibilant //, which Paulo pronounces as an affricate in feijo bean
[teu] and as an alveolar sibilant in laranja orange [neza] (session 6). This
phenomenon supports the claim that phonological acquisition is lexically grad-
ual. Although our examples largely concern early stages of phonological
development, it should be noted that studies of lexical diffusion provide evi-
dence that the word also has status in adult phonological representation (Wang
1969). Thus a phonological model is needed which recognizes segments as
312 Daniela Oliveira-Guimares
functional units for both adults and children but that allows for the word in
lexical representation as well.
Both of the case studies reported here give support to the notion of the word
as an important unit of representation in child language acquisition. The
templates changed over the course of the year. For Paulo, with his more rapid
lexical advance, we see segments gradually becoming functional units of
representation and organization alongside the word. In contrast, Lucas persists
in relying on a template representation for all twelve sessions. However, we saw
that Lucas changed his templatic consonant [m] to [w], showing an adjustment
of his production toward the structure of the ambient language.
note
1. We include only the rst ve sessions because thereafter Paulo began to produce
predominantly long phrases and words with only a single place of articulation became
increasingly rare.
References
Almeida, L. S. (2005). Um estudo sobre sntese de fala para o portugus brasileiro. [A
speech synthesis study of Brazilian Portuguese]. MA thesis.
Bisol, L. (2005). Introduo a estudos de fonologia do portugus brasileiro [Introduction
to Brazilian Portuguese phonology studies]. Porto Alegre: EDIPUCRS.
Bybee, J. (2001). Phonology and language use. Cambridge University Press
Medeiros, B. R. and Demolin, D. (2006). Vogais nasais do portugus brasileiro: um
estudo de IRM [Nasal vowels in Brazilian Portuguese]. Revista da ABRALIN, 5,
13142.
Cristfaro-Silva, T. (2001). Fontica e fonologia do portugus: Roteiro de estudos e
guia de exerccios [Portuguese phonetics and phonology], 4th edn. So Paulo:
Contexto.
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language acquis-
ition. Language, 51, 41939. Reprinted in this volume as Chapter 4.
Ferguson, C. A, Peizer, D. B., and Weeks, T. (1973). Model-and-replica phonological
grammar of a childs rst words. Lingua, 3, 3565.
Gnanadesikan, A. E. (2004). Markedness and faithfulness constraints in child phonol-
ogy. In R. Kager, J. Pater, and W. Zonneveld (eds.), Constraints in phonological
acquisition, pp. 73108. Cambridge University Press.
Keren-Portnoy, T., Majorano, M., and Vihman, M. M. (2008). From phonetics to
phonology: the emergence of rst words in Italian. Journal of Child Language,
36, 23567.
Lamprecht, R. R., Bonilha, G. F. G., Freitas, G. C. M., Matzenauer, C. L. B., Mezzomo, C. L,
Oliveira, C. C., and Ribas, L. P. (eds.). (2004). Aquisio fonolgica do Portugus:
perl de desenvolvimento e subsdios para terapia [Phonological acquisition of
Portuguese]. So Paulo: Artmed Editora.
Lle, C. (1990). Homonymy and reduplication: on the extended availability of two
strategies in phonological acquisition. Journal of Child Language, 17, 26778.
Word template development in Brazilian Portuguese 313
Macken, M. A. (1979). Developmental reorganization of phonology: a hierarchy of basic
units of acquisition. Lingua, 49, 1149. Reprinted in this volume as Chapter 5.
MacNeilage, P. F., Davis, B. L., Kinney, A., and Matyear, C. L. (2000). The motor core of
speech: a comparison of serial organization patterns in infants and languages. Child
Development, 71, 15363.
Mattoso-Camara, J. (1970). Estrutura da lngua portuguesa [Structure of the Portuguese
language]. Petrpolis: Editora Vozes.
McCarthy, J. and Prince, A. (1994). The emergence of the unmarked: optimality in
prosodic morphology. In M. Gonzalez (ed.), Proceedings of the North East
Linguistic Society 24, pp. 33379. Amherst, MA: Graduate Linguistic Student
Association, University of Massachusetts.
McCune, L. and Vihman, M. M. (2001). Early phonetic and lexical development: a
productivity approach. Journal of Speech, Language and Hearing Research, 44,
67084.
Menn, L. (1983). Development of articulatory, phonetic and phonological capabilities.
In B. Butterworth (ed.), Language production, vol. 2, pp. 350. London: Academic
Press. Reprinted in this volume as Chapter 6.
Miranda, I. C. (2007). Aquisio e variao estrutura de encontros consonantais
tautossilbicos [Acquisition and structured variation in tautosyllabic clusters].
Unpublished PhD dissertation, Federal University of Minas Gerais.
Oliveira, M. A. (1995). O lxico como controlador de mudanas sonoras [The lexicon as
a controller of phonological change]. Revista de estudos da linguagem, 4, 7591
Oliveira-Guimares, D. M. (2008). Percurso de construo da fonologia pela criana:
uma abordagem dinmica [Childrens construction of phonology: a dynamic
approach]. Unpublished PhD dissertation, Federal University of Minas Gerais.
Oller, D. K. (2000). The emergence of the speech capacity. Mahwah, NJ: Lawrence Erlbaum.
Priestly, T. M. S. (1977). One idiosyncratic strategy in the acquisition of phonology.
Journal of Child Language, 4, 4565. Reprinted in this volume as Chapter 7.
Rangel, G. A. (1998). Uma anlise de auto-segmental da fonologia normal: estudo
longitudinal de 3 crianas de 1:6 a 3:0 [An auto-segmental analysis of normal
phonology: A longitudinal study of 3 children from 1:6 to 3:0]. MA thesis.
(2002). Aquisio do sistema voclico do Portugus Brasileiro [Vocalic system
acquistion of Brazilian Portuguese]. Unpublished PhD dissertation, Ponticia
Universidade Catlica de Rio Grande de Sub, Porto Allegre.
Savinainen-Makkonen, T. (2007). Geminate template: a model for rst Finnish words.
First Language, 17, 34759. Reprinted in this volume as Chapter 13.
Stoel-Gammon, C. and Cooper, J. A. (1984). Patterns of early lexical and phonological
development. Journal of Child Language, 11, 24771.
Teixeira, E. R. and Davis, B. L. (2002). Early sound patterns in the speech of two
Brazilian Portuguese Speakers. Language and Speech, 45, 179204.
Thelen, E. and Smith, L. B. (1994). A dynamic systems approach to the development of
cognition and action. Cambridge, MA: MIT Press.
Viegas, M. C. (2001). O alamento de vogais e itens lexicais [Pretonic vowel raising and
lexical items]. Unpublished PhD dissertation, Federal University of Minas Gerais.
Vihman, M. M. (1993). Variable paths to early word production. Journal of Phonetics,
21, 6182.
(1996). Phonological development: the origins of language in the child. Oxford:
Blackwell.
(2009). Word learning and the origins of phonological system. In S. Foster-Cohen
(ed.), Language acquisition, pp. 1539. Basingstoke: Palgrave Macmillan.
314 Daniela Oliveira-Guimares
Vihman, M. M. and Croft, W. (2007). Phonological development: toward a radical
templatic phonology. Linguistics, 45, 683725. Reprinted in this volume as
Chapter 2.
Vihman, M. M. and Kunnari, S. (2006). The sources of phonological knowledge: a
cross-linguistic perspective. Recherches linguistiques de Vincennes, 35, 13364.
Vihman, M. M. and Velleman, S. L. (1989). Phonological reorganization: A case study.
Language and Speech, 32, 14970. Reprinted in this volume as Chapter 8.
(2000). Phonetics and the origins of phonology. In N. Burton-Roberts, P. Carr, and
G. Docherty (eds.), Phonological knowledge: its nature and status, pp. 30539.
Oxford University Press.
Vihman, M. M., Velleman, S. L., and McCune, L. (1994). How abstract is child phonol-
ogy? Towards an integration of linguistic and psychological approaches. In
M. Yavas (ed.), First and second language phonology, pp. 944. San Diego:
Singular Publishing. Reprinted in this volume as Chapter 9.
Vihman, M. M. and Vihman, V-A. (2011). From rst word to segments: a case study in
phonological development. In E. V. Clark and I. Arnon (eds.), How children make
linguistic generalizations: experience and variation in learning a rst language,
pp. 10933. Amsterdam: Benjamins.
Wang, W. S-Y. (1969). Competing changes as a cause of residue. Language, 45, 925.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
Appendix: Lucass session 3
Orthography Gloss Target form Child form
1. Abre open it ab(r)(i) ab
2. gua water ag
w
a a
3. boi bull bo bo
4. Cac (name) kaka tata
5. chapu hat ap pep
6. desce go down ds d
7. dois two do s dos
8. no it isnt n n
9. embora away bra b
10. Fernanda (name) fernd vv
11. Izabel (name) izabw pp
12. Lucas (name) lukas us
13. Mame mom mm mm
14. no no n nm
15. n conrmation n n
16. nenm baby nen nen
17. po bread p p
18. ovo egg ov of
19. papai dad papa papa, papa
20. papel paper papw pp
21. parabns congratulations paab s pala
22. pato duck pat pap, pa
23. p foot p upa, pa
24. Pedro (name) pedr dudu
Word template development in Brazilian Portuguese 315
The word total here does not correspond to Table 10.2, because imitated forms are not
included here. Forms in bold reect the reduplication template.
Orthography Gloss Target form Child form
25. peixe sh pe pes
26. praia beach praa pa
27. t he/she/it is ta ta
28. tartaruga turtle tahtaruga ta
29. tchau by a ta
30. tira take off ra dili
31. uva grape uva uf
32. Viviane (name) vivin vivi
33. vov grandmother vv ff
34. vov grandfather vovo vovo
35. xixi pee i zizi
316 Daniela Oliveira-Guimares
11 Templates in French
Sophie Wauquier and Naomi Yamaguchi
1. Introduction
As must be clear from the variety of analyses and approaches proposed in the
literature on phonological acquisition, there is no straightforward way to establish
the format of childrens rst phonological units or the conditions that shape them,
independent of target language. This chapter presents a proposal to account for
the acquisition of French within a template model. At the outset (Section 1), three
issues must be considered, to clarify the basis for the proposed template and the
analyses to be provided here: the lexical status of the template (1.1), phonetic
continuity vs. typological constraints on the template (1.2), and the function of the
template (1.3). In Section 2 we address what the template should be, considering
the typological characteristics of French, and in Section 3 we present three
longitudinal data sets that illustrate what the early template might be in French
and how it evolves and changes with lexical growth.
1.1 Phonological or lexical starting point: why a lexical template?
The rst problem that arises in attempting to determine the format of the rst
phonological units is whether to analyze them as essentially phonological or
essentially lexical. Can children categorize the phonological sequences that
they hear directly from the input, to construct a representation that will enable
them to recover such elements as the syllables or the phonemes that make them
up? Or must they necessarily rst resort to the lexicon (and thus acquire their
phonological knowledge through semantic bootstrapping based on the referen-
tial dimensions of the target language)?
The proposal that phonological acquisition is established through a lexical
template was originally proposed by Menn (1978), taken up by Macken (1992,
1995), and later further developed by Vihman and her collaborators (Vihman
and Velleman 1989, 2000; Vihman, Velleman, and McCune 1994; Vihman and
Croft 2007), within the older framework of the whole-word hypothesis
(Ferguson and Farwell 1975; Macken 1979; Menn 1971, 1983; Waterson
1971, 1987). The underlying assumption of this approach, as formulated by
Francescato (1968), is that children never learn sounds, they only learn words
and the sounds are learnt through words (p. 148).
317
Previous analyses of French lead to the conclusion that children construct
their phonology on the basis of a small number of templates, shaping a mini-
lexicon that allows them to progressively develop the relevant phonological
generalizations (Wauquier-Gravelines 2005). Templates can be taken to reect
the formal side of early words.
1.2 Articulatory continuity or typological constraints?
The adoption of the whole-word hypothesis and of a lexical template for the
acquisition of French raises the problem of how to model this template for
French and how to determine the constraints that apply to the production of the
rst observable word forms in French data.
Despite being one of the pioneers in the collection and analysis of early word
production data in various languages (Vihman and Velleman 1989; Vihman
et al. 1994 for English; Vihman 1976, Vihman and Vihman 2011 for Estonian;
Vihman 1993 for French; Vihman and Velleman 2000 for Finnish; Keren-
Portnoy, Majorano, and Vihman 2009 for Italian), Vihman has not emphasized
typological constraints as a determining factor for the templates she describes.
Her initial focus was on establishing articulatory continuity frombabbling to the
rst words, which suggests for every child an individual developmental sce-
nario that is less likely to be inuenced by the target language (Vihman,
Macken, Miller, Simmons, and Miller 1985).
More recently, Vihman has undertaken more systematic cross-linguistic com-
parisons of her data (Vihman and Kunnari 2006) and opened a typological line of
inquiry by showing that the templates are at least partly constrained by the
regularities of the target language (Vihman 2010). After examining some ten or
twelve languages, she identies the following major tendencies (Vihman 2010).
The templates reect a limited number of syllabic structures that never
exceed two vocalic nuclei: CV, VC, CVC, CVCV, CVCVC.
Consonant clusters and structures are generally absent.
The templates are built on the basis of a limited segmental inventory, gen-
erally a subset of the inventory of the target language. This limited inventory
seems to vary from child to child and relies, in part, on articulatory continuity
from babbling to the rst words.
Consonantal variation across the lexical unit is restricted to manner or place
only, not both, with full harmony the most common outcome.
Melodic patterning (or a xed segmental sequence) is also found within
templates, though more rarely: in this case the consonantal sequences may
be specied for place but not for manner.
In the case of melodic patterning, either medial or nal position may be
specied, but not initial position. Recorded segmental specications include
medial glides [j] or [w], medial glottal or uvular fricatives or [l], and nal
coronal, velar, fricative or nasal.
318 Sophie Wauquier and Naomi Yamaguchi
Vowel melodies include <lowhigh> (but not the reverse), diphthongal
specication (<Vi>, <Vu> or both) and nal vowel specication (often [i]).
These tendencies are also reected in the French data presented here, but
systematic ambient-language-based contrasts with English, for example
are also evident. In particular, very few CVC structures are found in French
templates compared to data from English, Dutch, and Estonian children (Elbers
and Ton 1985; Fikkert 1994; Vihman 1976; Vihman and Velleman 1989;
Vihman and Vihman 2011) or from bilingual English/French children
(Brulard and Carr 2001). While CVC sequences are common enough in high-
frequency words typically addressed to children in French (e.g., poule hen,
vache cow, robe dress, jambe leg), the number of words of this type that
children attempt to say may be reduced due to resyllabication in continuous
oral speech (Adda-Decker, Boula de Mareil, Adda, and Lemel 2005), as
shown in (1).
(1)
la vache [la /va] => la vache est au pr [la/va/e/to/pe]
the cow the cow is in the pasture
This suggests that the rhythm of French and its strong tendency for a CV-CV-
CV syllabication a tendency which leads, in particular, to fairly systematic
resyllabication of the nal coda of a word and its attachment to the next word
shapes the word forms produced by French-speaking children, who quite
consistently avoid producing codas. As will be shown below (Section 2),
French offers a typologically unique accentual, metrical, and prosodic structure,
although its segmental phonology is not particularly complex, despite its
marked vowel inventory (Carvalho, Nguyen, and Wauquier 2010). If typolog-
ical constraints guide the formation of initial word forms in French children, this
could be expected to be more evident on the prosodic and rhythmic than on the
segmental level. Accordingly, we hypothesize that CV-CV-CV syllabication
of the input will constrain the templates produced by French children.
This brings us to another aspect of French that requires attention in the
typological denition of templates. This is the fact that in French, unlike
English, for example, common nouns seldom occur without a determiner
(Veneziano and Sinclair 2000; Bassano, Maillochon, and Mottet 2008).
Consequently, children are exposed to an input in which bare nouns are rarely
heard, so we can expect childrens prosodic templates to incorporate unanalyzed
pro-clitic determiners or to show some trace of those determiners.
1.3 Template functionality
The third point that requires examination concerns the purpose and function-
ality of templates in the word production of French-speaking children. The
answer to this question lies in our approach to the role of the template at the
interface between perception and production.
Templates in French 319
Experiments with infants show that at around 911 months, an age generally
corresponding to the transition from canonical to variegated babbling, children
can recognize the major prosodic boundaries (Hirsh-Pasek et al. 1987; Jusczyk
1992; Gerken 1994) and the accentual patterns of their language (Jusczyk, Cutler,
and Redanz 1993), and have a holistic representation of lexical labels, which are
undoubtedly underspecied phonologically as well as morphosyntactically
(Hall and Boysson-Bardies 1994). Many perception studies have shown, for
example, that children identify function words at an early stage and essentially
use them, in association with other salient information, to identify word bounda-
ries and segment the speech signal into blocks (see Echols and Marti 2004; Hall,
Durand, and Boysson-Bardies 2008). Moreover, Hall and Boysson-Bardies
(1994, 1996) conducted experiments to identify the age of familiar word form
recognition. Children aged 11 months were presented with phonotactically
matched lists of common and rare words in a headturn preference procedure.
The results suggest that at this age there is no analysis or phonological decom-
position of lexical units, which are stored either globally or underspecied
particularly as regards the unaccented syllable (Vihman, Nakai, DePaolis, and
Hall 2004). Finally, other studies have demonstrated childrens difculty, even a
few months later, in word recognition and lexical processing tasks (for a detailed
review, see Fisher, Church, and Chambers 2004), particularly as concerns dis-
tinguishing between newly learned minimal pairs, which requires attention to
phonological detail (Barton 1978; Stager and Werker 1997).
We also nowknowthat infant speech segmentation is strongly constrained by
the rhythm of the target language (Ramus, Nespor, and Mehler 1999), and that
speech rhythm is one of the rst linguistic properties that infants employ to
distinguish languages (Nazzi, Jusczyk, and Johnson 2000). On this basis, then,
the templates seem to be global unanalyzed forms that provide a formal shape
for constructions in the sense of units with a formmeaning link. They can be
seen as a way for children to deal with the temporal organization of speech in
production: they constitute, at the phonological level of development, a tempo-
rary structural response to the metrical structure, the syllabic organization, the
rhythm and the stress/accent patterns of input speech. In that sense, the tem-
plates reect in production the units perceived in the input at a very young age
(before 8 months). Thus, we can consider the templates as functionally emer-
gent units whose format is typologically constrained by the input of the ambient
language.
1.4 Outline
We will draw on the three issues that we considered above (format, constraints
and functionality) to support the template conception that we propose for
French. We begin by presenting the main aspects of French prosody, partic-
ularly with a view to countering a common alternative conception that we
consider to be misguided, namely, that French is an iambic language, a mirror
320 Sophie Wauquier and Naomi Yamaguchi
image of English, in which central status is accorded to the binary foot
(Section 2.1). Contrary to this view, we will provide evidence that the accentual
and rhythmic structure of French predicts that the rst unit of acquisition is a
at prosodic template (Section 2.2). Having proposed a model for such a
prosodic template for French (Section 2.3), we will show how data for the
early period of word production, to which these premises apply, partially
conrm the proposed theoretical model (Section 3), while data obtained at
later stages clearly demonstrate the use of this prosodic template as well as
other templatic phenomena (spreading, planar segregation) which can be
seen as effective acquisition strategies for a syllabic CV-CV language like
French.
The data that we provide to illustrate this analysis come primarily from three
sources: (a) a corpus whose collection and analysis was nanced by the ESRC
project Psychological signicance of production templates in phonological
and lexical advance: A cross-linguistic study (the PSPT Project), made up of
longitudinal data from six children aged 1729 months, (b) longitudinal data
from the Claire corpus (Wauquier-Gravelines 2005), and (c) a corpus of elicited
and semi-elicited production from 38 3- to 5-year-old French children (Braud
1998, 2003).
2. Templates in French: the promise of the input
2.1 What French is not and what it cannot do
A good deal of acquisition research has been carried out within the framework
of prosodic phonology (Selkirk 1984; Nespor and Vogel 1986; Fikkert 1994), in
which a universal hierarchy is assumed to govern the organization of prosodic
constituents, regardless of language, such that all levels are obligatorily repre-
sented and nested according to binary logic.
[2]
Phonological Phrase (PPh)
Prosodic Word (PWd)
Foot (Ft)
Syllable ()
Mora ()
This perspective leads to an empirically inadequate analysis of French,
however. Indeed, the initial assumption within this approach is that acquisition
Templates in French 321
is achieved in any language through the production of progressively more
complex units based on binary metrical feet (trochaic in English and other
Germanic languages). Thus, a strictly linear order of acquisition is assumed,
which should vary little from language to language. Moreover, this analysis
assumes that all languages have binary feet and that children necessarily go
through a stage that includes the calculation of a lexical stress (Fikkert 1994;
Hayes 1995; Demuth 2001; Demuth and Fee 1995; Rose 2000; Goad and
Buckley 2006; Goad and Prvost 2008).
Until Rose (2000), this hypothesis was maintained for Germanic languages
almost exclusively, leading to the formulation of the trochaic bias hypothesis
(Allen and Hawkins 1978, 1980). Gerken (1994), on English, and Fikkert
(1994), on Dutch, both assume that children focus on strongweak metrical
structure in constructing their rst lexical units and perform truncation oper-
ations accordingly, while maintaining that the binary foot may be parametrized
across languages. Since by no means all languages follow the pattern of lexical
stress languages using a trochaic meter, this assumption was subsequently
treated within the Principles and Parameters framework, which allows for
(binary) alternative paths into language. Hayes (1995) proposes that all children
use a binary foot, the head of which is parametrized according to contact input
(that is, initial patterns in any target language will be bisyllabic, either trochaic
strongweak or iambic weakstrong), and that the default setting for this
parameter is a trochaic foot. Finally, Rose (2000), Dos Santos (2007), Goad and
Buckley (2006), and Demuth and Tremblay (2008) have proposed that the
acquisition of French is a mirror image of the process in English, and that
children go through an iambic foot stage, centering on the last two syllables of
the units they produce according to the weakstrong pattern.
These proposals appear to implicitly consider French as a language with
word-nal lexical stress. Very little consideration is given to the prosodic
and metrical structure of French, despite the availability of good
descriptive accounts (Fonagy 1980; Verluyten 1982; Dell 1985; Di Cristo
1999). These accounts clearly show that French is not an iambic language
with lexical stress, for one good reason: unlike English, French uses the
phrase rather than the word as its accentual unit (Dell 1985; Fnagy 1980;
Di Cristo 1999).
2.1.1 Stress in French Functional categories in French do not carry stress,
while stress in lexical categories (nouns, verbs, adjectives, adverbs) always falls on
the last full syllable, but only when they are produced in isolation:
(3) Monosyllabic words: lait milk, cur heart, vache cow
(4) Bisyllabic words: cheval horse, maison house, voiture car
(5) Tri/quadrisylabic words: crocodile crocodile, lphant elephant, balanoire
swing, hippopotame hippopotamus
322 Sophie Wauquier and Naomi Yamaguchi
In continuous speech, stress placement varies according to the position of the
carrying word within a larger constituent, a syntactic, semantic or phonological
phrase (usually considered as a breath group):
(6) Marie aime son cheval Mary loves her horse
(7) Marie aime son cheval fou Mary loves her crazy horse
(8) Marie aime son cheval fou et orgueilleux Mary loves her crazy and arrogant
horse
(9) Marie et son cheval / traversent la fort au galop Mary and her horse gallop across
the forest
These examples demonstrate that stress placement in French is determined not
on the level of the accented word itself, but higher up, on the level of the larger
phrase or utterance that the word is part of. The term iambic language is
therefore inappropriate for French, given that a stressed constituent-nal
syllable does not necessarily correspond to the heavy syllable of an iamb,
which is a metrical unit. While cheval (6) can be considered an iamb, val fou
(7) cannot.
2.1.2 The foot and the syllable in French metrics The hypothesis that
French is an iambic language is clearly invalidated by metrical structure. In
classical meter, the iamb is dened as a metrical grouping of two syllables, the
rst of which is weak and the second strong. This termhas been extended to all
forms of metrical structure that contain two syllables in the wS pattern (weak
strong). Thus, iambic verse is composed of a number of iambs, that is, binary
feet whose second syllable is stressed or lengthened. A case in point is the
famous iambic pentameter, the meter frequently associated with the poetry of
English and German, both lexical-stress languages. In French, which is
considered a syllable-timed language, the verse is divided not into feet,
grouping long and short or stressed and unstressed syllables, but into syllables
of the same metrical value (Verluyten 1989). Verlaines poem Green (10),
for example, is written in the classical form of alexandrine verses of twelve
syllables, all having the same metrical value. A caesura at the hemistich
boundary after the sixth syllable (the half point of the verse, producing a 6/6
structure) creates a division into two sets of six syllables with identical
metrical structure. In this case, the nucleus of the sixth syllable (i.e., eur
[], cur [k], pas [pa], beaux [bo]) provides a phrasal accent that
announces the caesura. However, the 6/6 caesura is not essential to the under-
standing of verses that can be segmented into another structure. One could
recite these verses without a break, producing a single twelve-syllable con-
stituent, or a single phonological phrase. This is evident in the third line,
which allows for liaison or linking of [z] across the caesura.
(10) Voici des fruits, des eurs, des feuilles et des branches
Et puis voici mon cur qui ne bat que pour vous
Templates in French 323
Ne le dchirez pas avec vos deux mains blanches
Et qu vos yeux si beaux lhumble prsent soit doux
1
(Verlaine, Green, Romances sans paroles, 1874)
[vwa/si/de/fi/de/ // de/f/j/ze/de/b
e/pi/vwa/si/m
/k // ki/ n/ba/k/pu/vu
n/l/de/i/e/pa // [z]a/vk/vo/d/m
/bl/
e/ka/vo/zj/si/bo // l/bl/pe/z/swa/du]
It has been argued that there is no organizational hierarchy or intermediary
structure between the syllables and the large prosodic constituents (the six-
syllable hemistich or even the twelve-syllable verse) and that the hemistich or
the alexandrine are at structures, not hierarchical constructions based on a
binary-branching prosodic structure (Verluyten 1989). If the foot is taken to be a
necessary unit, the alexandrine could be analyzed as having twelve feet, but
these would be single feet, that is, twelve syllables.
In short, the fact that French stress placement is phrase-nal rather than word-
nal, as shown above, supports the claim that there is no intermediary structure
between the syllable level and the phrase. In most cases French employs unitary
feet
2
(i.e., syllables of equal value in a non-hierarchical constituent on the foot
level: Verluyten 1982, 1989; Dell 1985), suggesting that these are attached to
larger constituents in a at structure.
But however questionable the postulation of an iambic foot as a metrical unit in
French may be, French does have, as we sawin (6)(9), a nal prominence that has
been well documented (Fnagy 1980; Dell 1985; Di Cristo 1999; Jun and
Fougeron 2000). Should this nal prominence be interpreted as the instantiation
of a (w)S nal foot (an iambic foot)? Some French linguists have indeed interpreted
it as reecting a stressed syllable that would be the strong position of a (w)S foot
(Charette 1991), but a good deal of evidence supports an interpretation of the
French nal prominence as being a domain edge marker belonging to the intona-
tion system in a language without feet (Verluyten 1982; Jun and Fougeron 2000).
Despite this debate, the position that French has an iambic foot is generally
taken for granted in acquisition studies, often without further discussion (but see
Goad and Buckley 2006, Goad and Prvost 2008, Goad 2011). The insistence
on (i) analyzing French as having iambic structures and (ii) considering that
French children will consequently systematically produce iambic feet at the
early stages (e.g., Demuth and Tremblay 2008) appears to be based on a
systematic theoretical bias in favor of the universality of the prosodic hierarchy,
although the empirical facts of French must be taken to present important
challenges to the theory.
2.2 What French is and what it can do
It has also been proposed that French metrical structure is based not only on
phrase-nal stress but also on a phrase-initial counter-stress, symmetrical to the
nal stress. The hypothesized phrase-initial stress was originally proposed by
324 Sophie Wauquier and Naomi Yamaguchi
Fnagy (1980), and later adopted and developed primarily by Di Cristo (1999),
who provides the following description: [French exhibits] a tendency to
accentuate the rst syllable of words, which gives rise to the formation of
barytone patterns and accentual arcs in which only the initial and nal syllables
of a phrase are stressed (my translation).
3
According to Di Cristo, the existence of this initial stress in contemporary
French is accepted by most prosodists, however they conceptualize the phe-
nomenon. Differences relate to the exact interpretation of the counter-stress
(variously regarded as emphatic, an echo or secondary stress). For Di Cristo,
both initial and nal syllables are, therefore, prosodically strong positions,
forming the two pillars of an accentual arc within which the metrically equiv-
alent internal syllables are inserted and eventually reduced relative to the edges
of the constituent. Examples (11)(14) are extensions of (6)(9) with the
addition of Di Cristos proposed counter-stress.
(11) Marie aime son cheval Mary loves her horse
(12) Marie aime son cheval fou Mary loves her crazy horse
(13) Marie aime son cheval fou et orgueilleux Mary loves her crazy and arrogant horse
(14) Marie et son cheval / traversent la fort au galop Mary and her horse gallop across
the forest
2.3 Templates in French: what is the appropriate model for acquisition?
Based on the accentual arc model we may conclude, following Macken (1995),
that the units available to children at the production/ perception interface can be
schematized as in (15):
(15) [ ()
n
]
This formal structure is initially dened by the constraints that produce the
accentual arc structure and that are heard in the French input. It is bounded by
two demarcating stresses that correspond to the stress and counter-stress
described above: the last syllable () carries the demarcating phrasal stress
that delimits the right edge of the unit, while the rst syllable () bears the
counter-stress. The initial and nal syllables dene the boundaries of the
accentual arc, which thus serve as prosodically strong positions. This structure
can be expected to provide the rst lexical pattern for phonological
development in French. It derives from a prosodic unit that is perceptually
available, bounded by stress and counter-stress, and therefore segmentable in
the input. Therefore, we expect that it is this abstract phonological structure
that is targeted by the production templates that we observe and see as
temporary structural responses to the prosodic characteristics of the input.
From this we can develop the following predictions: (i) children will rst
construct the strong syllables, (ii) these will never be truncated, and (iii) will
undergo little deformation. Between these prosodically strong boundaries there
are an open number (n) of intermediary syllables, where n can theoretically
Templates in French 325
contain any number from 0 to innity. We need to consider that n may be 0: the
internal syllable is generally optional at the early stages (see Section 1.2). Braud
(2003) shows that n is consistently 2 in early production, up to the age of 2.
Unlike the rst and second syllables, the n-site should be less stable and more
variable. We also postulate that this prosodic structure will constrain the
templates produced at an early stage and will later be the domain of morpho-
phonological generalizations.
3. Data and observed template formats
We turn nowto a comparison of this schematization of the rhythmic template for
French with the evidence provided by three sets of French child data. We focus
on the question as to whether or not French childrens early word forms exhibit
the kind of systematic patterning that would reect the typological constraints
given by the rhythmic structure of French as we have described it.
3.1 The PSPT project data: six children at the early stage
Case studies have been prominent in the literature on child phonology, but they
do not provide the best way to test claims about the typological systematicity of
data; it is preferable to compare data from several children of the same age. We
rely here on data obtained and analyzed as part of an ESRC project,
Psychological Signicance of Production Templates in Phonological and
Lexical Advance: A cross-linguistic study (the PSPT project), which includes
longitudinal data from 6 French children aged 1729 months. These children (2
girls and 4 boys) were recorded once a month in 30-minute sessions of natural-
istic, non-elicited interaction with a parent. Recordings began at the 25-word
point
4
(based on a parental questionnaire and researcher verication in an initial
control recording). The data were transcribed and analyzed using PHON (Rose
and MacWhinney, 2013). Here we present our ndings based on the rst ve
recording sessions, which will enable us to observe the prosodic structure of the
rst words.
3.1.1 The idiosyncratic character of the rst templates We begin by noting
that the data conrm the tendencies identied by Vihman (2010) (see
Section 1.2), at least in part. A size constraint on these word forms is clearly
evident (they never exceed two vocalic nuclei). These data also conrm the
great variability of word forms from one child to the next and the existence of
idiosyncratic strategies that suggest individual articulatory continuity with
babbling and the personal preferences of each child.
To illustrate, Table 11.1 presents the 28 occurrences of the word micro
microphone produced by Bryl during a single recording session (19 mos.).
The data reveal Bryls strong preference for the form [ao] (13 out of 28
tokens). But beyond that, all tokens realize the VCV pattern, with a vocalic
326 Sophie Wauquier and Naomi Yamaguchi
a-o melody and a xed medial consonant consisting of a uvular fricative (14 out
of 28 tokens). The remaining patterns are roughly modeled on the main pattern,
with two systematic changes: either to the vocalic melody (o-o or a -o alternate
with a-o), or to the medial consonant (the uvular fricative alternates or (in one
case) combines with the velar stop [k]). Table 11.2 shows that this pattern was
also extended to other target words that were selected (in gray, here and else-
where) or adapted
5
to t this pattern (same session).
At rst glance, this result seems to follow an idiosyncratic articulatory logic in
the construction of a template, to the extent that the chosen structure is clearly
specic to this child whereas uvular fricatives in medial position are not particularly
characteristic of French. Thus, the medial /kr/ cluster of micro, with its uvular [r],
must be supposed to have inspired Bryls rough phonetic approximation.
3.1.2 Typological characteristics of the rst templates Yet Bryls tem-
plate is perhaps less idiosyncratic than it appears, if analyzed on the
syllabic level and compared to Vincents data (at 17 mos.), for example
(Table 11.3).
In fact, Vincents data also reveal a VCV structure, where the rst vowel is
frequently central (schwa or a). This suggests that Bryls apparently idiosyncratic
Table 11.1. Bryls 28 tokens of micro microphone (at 19
months)
Main patterns N tokens
Other patterns (one
token each)
[ao] 13 [akpo]
[oo] 2 [pako]
[o] 2 [ako]
[oo] 2 [koko]
[tao] 1 [ahko]
[oko] 1 [ak
h
o]
[o]
Table 11.2. Bryls other word forms reecting the
< aCo > template
Target words Bryls productions
agneau [ao] lamb [alo]
bateau [bato] ship [ato]
poisson [pwas
] sh
[ao]
lphant [elef] elephant [afo]
crapaud [kapo] toad [ako]
Templates in French 327
VCV template corresponds to a prosodic structure available to other French
children. This hypothesis is conrmed by the data summarized in Figure 11.1.
Figure 11.1 shows the percentages of word forms produced by the six
children over the course of their rst ve recording sessions, sorted according
to the output syllabic structure. The rst structures produced are mainly CV for
all six children (the most frequent production form: 43 percent on average). The
CV syllable may derive fully or partially from the nal syllable of the target
word (e.g., [k] < [k], encore again), less frequently from the initial
syllable ([k
] < [kana], canard duck), or from segmental reorganization
based on both the initial and the nal syllable of the target word ([bu]< [bizu],
bisou kiss). This CV syllable is not always the one receiving nal stress in the
adult target (and can therefore still less be characterized as the strong syllable
of an iambic foot).
However, we also note that CVCV and VCV structures are systematically
produced at an early stage alongside the CV pattern (unlike the remaining
structures, none of which accounts for more than 5 percent of the tokens
produced). The CVCV structure is primarily produced in the case of targets
that are reduplicated in the adult language (doudou security blanket, papa
Table 11.3. Vincents word forms, the <VCV> template
Target words Vincents production
allo [alo] hello (on telephone) [alo]
OK [oke] [oke]
attends [at] wait [at]
ici [isi] here [ii]
bravo [bavo] bravo [avo]
voil [vwala] there it is, there you are [ala]
avion [avj
] airplane
[aj
]
ferm [fme] closed [ame]
cach [kae] hidden [ae]
encore [k] again [at]
50%
40%
30%
20%
10%
0%
CV CVCV VCV CCV V CVC (C) VCCV VC
Figure 11.1. Percentages of early syllable structures of word-forms (averaged
over ve sessions for six children)
328 Sophie Wauquier and Naomi Yamaguchi
daddy, maman mommy), but for non-reduplicated targets (lapin rabbit,
chapeau hat) the structure of the child form is more likely to be VCV. The
VCV structure is obtained either through selection (allo hello, attends wait)
or adaptation (CVCV > VCV: lapin, chapeau, ferm closed). The tendency
to favor some words over others can also be observed longitudinally
(Figure 11.2). The CV structure arises early as the preferred word form and
remains the most used structure throughout the ve sessions for all the
children. Despite individual variation from one child to the next (Figure 11.3),
the preferred structures include CV, CVCV, and VCV for all but one child.
The children can be ranked from Bastien, who primarily used the CV structure
and secondarily the CVCV and VCV structures, to Bryl, who made equal
use of the CV, CVCV, and VCV structures. In all the cases the CV structure is
used early and remains the most used by all the children from the rst to the fth
session.
The generalization that emerges is that the rst word structure to stabilize
and be frequently produced by French-speaking children is built around a
CV syllable and not a binary iambic foot. This is followed by two options that
are frequently produced, in parallel, to augment and vary the rst CV template:
CV > VCV, where V
1
is mostly a central or front vowel ([a] [] [e] [])
CV > CVCV, where CV is frequently reduplicated. There are also
some cases where the two consonants differ but are harmonized
(mainly for place).
Both of these patterns express a systematic avoidance of consonant change across
the word and a preference for open syllables, word-internally as well as nally.
This means that Bryls [ao] word form, which at rst looks so idiosyncratic, can
be analyzed as the realization of a more general <aCo> template which is itself an
instantiation of a <VCV> pattern reecting one of the main typological
characteristics of French (i.e., open syllabication). And, as we saw,
1000
800
600
400
200
0
N
u
m
b
e
r
o
f
o
c
c
u
r
r
e
n
c
e
s
(
6
c
h
i
l
d
r
e
n
)
S1 S2 S3 S4 S5
1200
V
VCV
CVCV
CV
Sessions
Figure 11.2. Number of lexical items belonging to each word form: CV,
CVCV, VCV, V (averaged over ve sessions for six children)
Templates in French 329
the<VCV> pattern can itself be analyzed as a variant of the CV pattern (V + CV),
which is by far the most common structure produced by French children.
3.2 Later longitudinal data: truncation and reduplication
Will the typological constraints of the target language have the same effect on
the word forms of older children? We present below data from the longitudinal
Claire corpus, from sessions taken at the age of 2225 months, when she had a
vocabulary level higher than what is reected at the 25-word point (cumulative
lexicon of some 200 words): see Tables 11.4 and 11.5, organized by length of
the adult target. We focus on Claires truncations and reduplications and the way
they may t a prosodic template partly (or fully) expressing the typological
constraints described above (Section 1).
3.2.1 Truncation and templates As illustrated in Table 11.4, one- and two-
syllable words are produced without truncation, mostly with a schwa or [l],
[la], [], which can be interpreted as proto-determiners (Veneziano and Sinclair
2000, see Section 1.2). Note that the proto-determiner does not appear with
proper nouns in the input and is not reected in Claires forms of these either. In
contrast, three- and four-syllable words exhibit partial deletion of segmental
material between the proto-determiner and the last syllable (or even the last
vowel). Thus, Claire preserves the two edges of the target words according to
the proposed template. This is true for long words as well as for monosyllabic
words, for the former at the expense of the internal syllables and for the latter as
well as for monosyllabic words produced with an initial vowel as a proto-
determiner (Veneziano and Sinclair 2000).
Bastien Julien Romuald Vincent
Children
Marie Bryl
CV
CVC
CCV
CVCV
VCV
V
70
%
o
f
p
r
o
d
u
c
t
i
o
n
60
50
40
30
20
10
0
Figure 11.3. Individual variation in the syllable structures of word forms
(averaged over ve sessions)
330 Sophie Wauquier and Naomi Yamaguchi
3.2.2 Truncation, reduplication, and spreading In the same period Claire
developed another strategy for handling three- and four-syllable words, which
provides additional conrmation for this analysis (Table 11.5).
Here Claire lengthens her word forms by reduplicating a syllable of the
truncated word. She appears to proceed in two steps:
1. Truncate left edge of word but preserve determiner;
2. Lengthen the word by reduplicating the left-most syllable.
For example, for le chocolat [lokola] Claire begins by producing [ekola],
followed by the string [ekokola], produced by reduplication of the syllable [ko];
for un crocodile [
kokodil]).
In fact, Claires production aims at the template proposed in (15) and
elaborated on the basis of Di Cristos concept of the accentual arc. This indicates
Table 11.4. Claires word forms (2223 mos.)
Monosyllabic target words Child forms
la vache [lava] the cow [ja] / [laja]
lne [lan] the donkey [ltan]
6
le pot [lpo] the pot [lpo]
le chien [lj
]
Claire [kl] [l]
Bisyllabic target words Child forms
le bb [lbebe] the baby [lbebe]
un ballon [
/dbal
] one balloon [abal
]
deux ballons [dbal
]
Didou [didu] [didu] [tidu]
Maman [mam] Mummy [mam]
Trisyllabic target words Child forms
lphant [elef] elephant [e]
un lphant [
]
Aurlien [orelj
] [j
]
Olivier [olivje] [oje]
Quadrisyllabic target words Child forms
la brosse dents [labsad] the toothbrush [anad]
un mdicament [
kokodil] a crocodile [
koti]
un crocodile [
nunus]
un os [
ns] a bone [
nons]
Trisyllabic target words Child forms
un arrosoir [
naozwa] a watering-pot [
ozwa] / [
ezwa]
une coccinelle [ynkksinl] a ladybird [ynkokosinl]
Quadrisyllabic Target words Child forms
un accordon [
nakde
] an accordion [
aaj
]
un pouvantail [
nepuvtaj] a scarecrow [
pupuvtaj]
un hlicoptre [
nelikpt] an helicopter [
nenikt] / [
ninikpt]
un aspirateur [
naspiat] a vacuum
cleaner
[
pasat] /[
aat] /[
astat] [
piat] /
[
piat] / [
pisat]
Templates in French 333
do not tend to arrive at early CVC templates patterns while Dutch, English,
German, and Estonian children often do. Brulard and Carr (2001) also demon-
strate a CVC pattern in their English/French bilingual child. Thus, omission of
codas is not dictated by an age-related or wholly maturational constraint. Nor
does it originate from isolated words in the French input, which provide many
CVC words of high frequency like robe dress, soupe soup, dame lady, coq
rooster, vache cow, which commonly occur in child-directed speech. Why
then do monolingual French children seem to lter the input to avoid CVC
patterns? We assumed that their templatic patterns are inuenced by the rhythm
and the CV-CV syllabication of French (Section 1.2). More generally, one can
assume that the parameters of variation in the childrens surface forms in
production must be at least partially typologically constrained and limited
to the underlying structures supported by the rhythmic and syllabic structure of
the target language.
The strong generalizations that emerge from our data clearly conrm this
assumption. As we have seen, French-speaking children share a preference for
open-syllable structures, early vowel stability in the nuclei and an avoidance of
consonant clusters. The rst word forms to be produced systematically by
French-speaking children are built around the CV word structure.
The childrens word forms evolve in two ways, in parallel, then, to augment
and vary the rst CV template:
CV > VCV, where V
1
is mostly a central or front vowel ([a] [] [e] [])
CV > CVCV, where CV is frequently reduplicated. There are also
some cases where the two consonants differ but are harmonized
(mainly for place).
These regularities correspond to our predictions regarding the typological
constraints that French input imposes on the rst templates. The rhythmic
features of French, a syllabic language that favors CV-CVsyllabication (some-
times even at the expense of word boundaries, as in the case of liaison
sequences: see Section 2.1.2), lead children to construct their initial templates
on the basis of the CV syllable and to prefer open structures.
4.2 Status of the initial ller: does it support the binary foot?
Another aspect of the data is also interesting, although harder to interpret: the
presence of low or mid vowels as V
1
in VCV structures is tricky. For certain
words like [ap
astat]: [
pisat]: [
naspiat] is obtained: [
] for /
bal
n]
for /ana/ I) in the realizations. Long C: and complex realisations by the
children (54 and 30 percent respectively) are around 1.5 times as frequent
as geminate and complex targets, suggesting that the medial consonants of
many words with singleton targets were lengthened or produced with
complex articulation.
51%
21%
28%
30%
54%
16%
Disyllable medial C(C)
realization
N = 1456
Disyllable medial C(C) target
N = 1691
Singleton
Complex
Geminate
Figure 14.4. Distribution of medial consonant type (single, geminate,
complex) in disyllabic words targeted by the children (left) and their
realization (right)
Inuence of geminate structure on early Arabic templatic patterns 387
5.3 Developmental patterns
On the whole, the children target similar word structures in the early (4wp)
and later (25wp) stages of production (Figures 14.56), with a wider range of
word shapes at the more advanced stage and an emergence of more complex
shapes (not all listed in the gures below due to their very low frequency).
One notable difference is a 14 percent drop in disyllabic CVC:V targets at the
25wp (Figure 14.6), but not in realizations; in fact, lengthening of singleton
consonants is still prominent and actually increases at the more advanced
stage (Table 14.4, Figure 14.9). The structure of the realisations for the most
frequent target word shapes does not change very much as the children
progress to the 25wp (Figures 14.78); this is due to the fact that the children
45%
25%
15%
10%
40%
30%
20%
5%
0%
CVCV CVCCV CVCVC CCV:CV CCVCV
Target disyllables
CV:CV:C CVCC
V(:)C
CV:CV(:) CVC(:)
V:C
CV:C(:)
VC
CVC:V CV:CV
35%
4 wp
25 wp
N = 1052
Figure 14.6. Most frequent types of word structures targeted in disyllabic
word shapes at the 4wp and the 25wp. Shapes constituting less than 1 percent
of the data are not included
45%
4 wp
25 wp
25%
15%
10%
40%
30%
20%
5%
0%
CVC: CVC
Target monosyllables
CCV CVV
CCV:C
CCV: CCVC CVCC CVVC CCVV CCVVC CV:C CV: CV
35%
N = 497
Figure 14.5. Most frequent types of word structures targeted in monosyllabic
word shapes at the 4wp and the 25wp. Shapes constituting less than 1 percent
of the data are not included
388 Ghada Khattab and Jalal Al-Tamimi
produce target-like structures from an early age, if phonological length is set
aside. What they seem to take some time to acquire is phonological length,
and their patterns of acquisition seem to involve experimenting with adding
phonetic length to all elements of the target syllable structures rather than just
to the phonologically long ones, or strengthening consonants (denoted as
Cs in the gures above). For instance, a target CVC: can be produced not
16%
Target CVC: 4-word point
14%
12%
10%
8%
6%
4%
2%
0%
CV:Cs CVCV:C CV:C: CsV:C CVCs
N = 20
16%
Target CVC: 25-word point
14%
12%
10%
8%
6%
4%
2%
0%
C:V:C: C:VC CVC C:V:C CV:C
N = 94
Figure 14.7. Range of realizations for the most frequently targeted
monosyllabic word shape, CVC:, at the 4wp (left) and 25wp (right). Here
and elsewhere, Cs refers to a consonant that is articulated with extra strength/
tenseness
12%
Target CVCV 4 -word point
10%
8%
6%
4%
2%
0%
CV:C:V:C CVC:V:C CVC:VC C:V:C CV:CV: CV:CV:C
N=100
12%
Target CVCV 25-word point
10%
8%
6%
4%
2%
0%
CVC:V:C CVCV:C CV:C:V:C CVCVC CV:CV:C CVCV
N= 210
Figure 14.8. Range of realizations for the most frequently targeted disyllabic
word shape, CVCV, at the 4wp (left) and 25wp (right)
Table 14.4. Proportions of CVCV shapes being realized with a
singleton or a geminate consonant at each developmental stage
CVCV realization
Singleton Geminate/strong
4wp 46% 54%
25wp 35% 65%
Inuence of geminate structure on early Arabic templatic patterns 389
just with a long coda, but also with a long onset and/or a long vowel, e.g.,
/ba/ all gone realized as [ba], [bba], [ba], etc. Similarly, a target
CVCV can be realized with varying lengths for all segments, e.g., /baba/
daddy realized as [babbah], [babba], [bbabam], and [baba], etc.; the
realizations of disyllables with open nal syllables in the target frequently
contained a nal coda, often a guttural sound (glottal stop, glottal fricative or
pharyngeal fricative) but occasionally also other consonants with supraglot-
tal places of articulation. While variable phonetic length may apply to
all childrens early productions regardless of their native language, the
fact that Arabic has phonological vowel and consonant length may increase
the salience of contrastive duration for the children, leading to their extensive
experimentation with segment length and the production of syllables with
heavy rhymes and/or codas. Acoustic analysis is currently under way in
order to obtain a clearer picture of the relationship between phonetic and
phonological length in the childrens productions.
The prominence of monosyllabic CV(:)C(:) and disyllabic CV(:)C(:)V
shapes in the targets that the children are aiming for throughout the single-
word period can also be seen at the individual level, though with interesting
differences connected to each childs starting point (the structure of their earliest
words), the relative frequency of each language that they hear, and their
individual journey towards the 25wp. The next section looks at longitudinal
data from three of the children whose data are presented here in order to explore
the interaction between language-specic and individual differences in the
development of early phonological structure. In the data presented below only
one token per lexical item is presented, chosen from the most frequent and/or
most adultlike realizations.
100%
Disyllabic targets vs. realizations
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Singleton
4-Words 25-Words
Realization
Target
Singleton Complex Complex Geminate Geminate
Figure 14.9. Target medial consonant type and realizations in disyllabic
productions at the 4- and the 25wps. N = 1142
390 Ghada Khattab and Jalal Al-Tamimi
Table 14.5. Martins selected and adapted form over the one-word stage.
Words were considered selected if the adult target matched the pattern of
interest and adapted if they were modied to t the childs pattern(s). Shaded
grey is used for imitations. Here and elsewhere, the half-length symbol
following a consonant was used both for half-long and/or noticeably strong/
tense articulation. Italics = French or English target
Inuence of geminate structure on early Arabic templatic patterns 391
Table 14.5. (cont.)
ill
392 Ghada Khattab and Jalal Al-Tamimi
Table 14.5. (cont.)
ill
Inuence of geminate structure on early Arabic templatic patterns 393
5.4 Individual paths and templatic behavior
5.4.1 Martin Martin was exposed mostly to Arabic, often mixed with
French, and his production in the seven sessions that were analyzed reects
that exposure (77 percent of his utterances are Arabic, followed by French at 13
percent and English at 6 percent). His 4wp was identied at age 1;3, which is close
to the mean age at the 4wp for the children studied here. He is the most systematic
of the children in that his earliest productions fell mostly in the CVC:V(C) pattern,
and this remained his favorite structure throughout. Below is a more detailed
account of Martins phonological patterns across the one-word stage.
At the 4wp all but one of Martins word types have the CVC:V(C) shape, and
the majority of these (88 percent of tokens) show consonant harmony either in
the target or the realization or in both (Table 14.5). Martin adapts both mono-
and multisyllables to the disyllabic shape with a long medial C:, e.g., French
train /t
nn
h], [ enna
h], [jnn
h], [h
nna
h],
[n
nn
h] for
/nanna/ above. As expected, Martins consonant inventory at the 4-word point is
relatively small, mainly consisting of bilabial and alveolar nasals and stops,
along with glottal stops and fricatives (Appendix).
Over the next two sessions, Martin maintains this pattern but also produces
relatively target-like words with disyllabic CV(:)CV(C) and monosyllabic
CV(:)C(C) patterns, e.g., /mam
m]
Table 14.5. (cont.)
394 Ghada Khattab and Jalal Al-Tamimi
peek-a-boo; /nanna/ [n
nn
] Jesus).
Medial consonants in Martins imitated utterances are more target-like than
initial ones, which are more variable. Words with target fricatives and liquids
are targeted in imitations for the rst time, with frequent stopping and other
adaptations, adding to the prominence of consonant harmony.
Martins last two sessions show a marked change in the frequency of words
produced as well as a growing consonant inventory (Appendix), but Martins
preferred CVC:V(C) pattern is still prominent, with adaptations that are twice as
frequent as the selected words with this pattern in session 6. These stand out
compared with the other minor word shapes that Martin produces, which tend to
be more accurate, e.g., CV(V)C(C): English thats /ats/ [dat
wn ] where. The majority of words that Martin targets are still disyllables,
and despite his increased phonetic and phonological inventory his productions
still exhibit frequent consonant harmony. In the last session the CVC:V(C)
shape rises to 71 percent of all of Martins productions, the highest since his rst
session, which suggests that the medial long consonant template is at its most
productive for Martin as he approaches the 25wp. Consonant harmony is not as
prominent in this session, as newconsonants are attempted and coda consonants
are more frequent. Session 7 also sees the geminate/long pattern being applied
to longer words as Martin starts producing multisyllabic words; multisyllabic
words with medial geminates like /tattuna/ [tttnh] nickname for Martin
and /battajjet/ [
cc
tt
h] food.
Although Martins consonant inventory is expanding, variation in the realiza-
tion of some consonants is higher than in earlier sessions, especially in initial
position, e.g., for /b/, /k/, and /m/ (see Appendix). Medial codas are targeted but
are often assimilated to the next onset, adding to the geminate pattern, e.g.,
/mala/ [
mm
h] spoon, /min hon/ [x
nn
nh] whos there?, and
/mfte/ [c a t
/ [bm
h] mummy,
/koko/ [kl
k h] nickname.
5.4.2 Rama Rama was exposed to more English than Martin, and 30 percent
of her utterances were English. Arabic still constituted the majority of her utter-
ances at 60 percent. Ramas rst two sessions are combined, due to the small
number of spontaneous utterances in her rst session (Table 14.6). Her prole at
this early stage of production is strikingly different fromMartins, mainly due to the
high frequency of monosyllabic words that she produces (46 percent). The
Inuence of geminate structure on early Arabic templatic patterns 395
disyllabic geminate pattern is prominent as well (40 percent of utterances), with
adaptations such as /baba/ [mbb h] daddy; /kle/ [ke] eat!; /hajda nz/
[
dnnz] thats (a) nose. Perhaps due to Ramas jargon practice and older age at
this stage, her consonant inventory is more varied than Martins at the 4wp
Table 14.6. Ramas selected and adapted forms over the one-word stage;
shaded grey is used for imitations
miaow
396 Ghada Khattab and Jalal Al-Tamimi
(Appendix), with a small number of fricatives and laterals alongside stops and
nasals as well as nal consonants and occasional two-word utterances (e.g., /jalla
kle/ [jake] come on eat; /ptn mni/ [pd md
] put on Minnie).
In the next two sessions (age 1;7 and 1;8) the two patterns identied at the
4wp still make up the majority of utterances, though the prominence of the
disyllabic geminate pattern is due more to frequency of use (48 percent of
utterances) than to type (28 percent of the total of different words).
Monosyllabic CV(:)(C) is the most varied and productive shape, showing a
nal glide pattern (e.g., /hajj/ [ j] this; /ba/ [bjj] bye; /waw/ [b ww]
wow) and a front mid-high to mid-low vowel pattern (e.g., French danse /d
s/
[teh
s]; /mijaw/ [n
m ] miaow).
Table 14.6. (cont.)
Inuence of geminate structure on early Arabic templatic patterns 397
In the nal session the monosyllabic CV(:)(C) shape becomes the most prom-
inent, accounting for 40 percent of all utterances. Within this shape a subset of
productions still have the nal glide pattern, as in previous sessions (Table 14.6),
but others include other consonants as well and a rich variety of vowels (e.g., /ba/
[b] all gone; /mbu/ [mbu:u] water; English Po (name of TV character) /po/
[po]). The inuence of words of English origin is obvious in the frequency of
monosyllabic words in Ramas sessions, with words like Po, bye, wow, ball, eyes,
book, and nose making up a large proportion of her productions, especially in the last
session. The second most frequent pattern in this session is a disyllabic C(:)V(:)CV
shape (29 percent of utterances), which takes over from the medial long C(:) as the
second most frequent shape (e.g., /teta/ [teta] grandma; /mama/ [mma:ma h]
mother; English baby /bebi/ [bebi]). These and all but one of the monosyllabic
words are selected and, apart from expected developmental features, they are
fairly accurate. In fact, most of Ramas productions in the nal session are
essentially accurate; in comparison with Martin, she produces fewer utterances
and fewer repetitions of words (98 types and 203 tokens over ve sessions for
Rama, compared with 179 types and 604 tokens over seven sessions for Martin)
but the words tend to be more accurate and her production exhibits no large-scale
adaptations to any preferred shape. The only pattern that still shows more adapta-
tion than selection is the disyllabic long/geminate pattern (e.g., /tiktak/ [ti:ttih]
sweet; /ba/ [ba] all gone; English oven gloves /ovn lvz/ [au]),
though the frequency of occurrence of this pattern is now down to 20 percent.
5.4.3 Lina Lina was exposed to both French and English on a regular basis,
and her production reects that, with Arabic productions constituting only about
half of her overall utterances at 48 percent, followed by French (28 percent) and
English (21 percent). Her rst three sessions, between 1;3 and 1;5, had similar
patterns and no increase in the number of spontaneous words produced, so they are
combined here for analysis. In these sessions, Linas utterances can be grouped
into the three word shapes identied for the other two children, with the disyllabic
pattern with a long medial C: being the most frequent (e.g., French Oui Oui (PN)
/wiwi/ [wwih]; maman /mam
t h] biscuit), multisyl-
labic realizations of disyllabic targets (e.g., /doa/ [wwl l
d ] Dora; /lala/
[ ell lh] Lala), or across word boundaries (e.g., /la ma badde/
[alla
ddi
s/ juice []
*
, []
*
Kenta (2;0)
/aita/ opened [a:da], [a:da], [a:te]
/o/ blue [a:]
/b:/ boo! [ba:]
/bibai/ bye-bye [ba:ba]
/bs / bus [ba
:], [ba]
*
/b
= ingressive airow.
Appendix B: Tais word production at 1;6.04
Target and gloss Structure Tais production
/ka/ red (noun) LL1 [aka]
/aki/ red (adjective) LH2 [akai]
/akete/ open LLL0 [akete]
/a
:/ kiss H1 [:]
/da/ (copula) L0 [da]
Lexical frequency effects on phonological development 437
/d:zo/ please HL1 [do:zo]
/hi/ yes H1 [hai], [ai]
/hana/ nose LL0 [hana]
/hantai/ opposite HH0 [hantai]
/i k:ki / plane LHL2 [ik:ki], [k:]
*
, [ko]
*
/ho:o:/ kitchen knife HH0 [ho:o:], [ho:o:], [ho:to], [o:to], [o:do], [oto]
/i/ good H1 [i]
/ini/ not there LH2 [ini]
/io/ together HL0 [itto], [t:o]
*
, [d:o]
*
, [o]
*
/iti/ ouch LH2 [itai]
/ij/ no LL2 [ija]
/:a/ grandpa HH1 [i:a], [i:i]
/kkka/ mom HL1 [kakka]
/kakkoi/ cool HLH3 [kakkoi]
/ki kna/ train LHL2 [ka:ta]
*
/koko/ here LL0 [koko]
/konniiwa/ hello HLLL0 [koniwa] (acceptable informal form)
/ko:ka/ exchange HH0 [ko:ka:], [ko:ka]
/kot/ this way HL2 [koti]
/koe/ this LL0 [koe]
/mdoi/ green LLL1 [mi]
*
, [m:i]
*
/mo/ too L0 [mo]
/mimoi/ hello LLLL1 [mi]
*
/ni/ there isnt H1 [nai], [na:], [ne:]
/nna/ seven LL1 [nana]
/natta/ xed LHL2 [naotta]
/n:a/ older sister HH1 [ne:a], [ne:ta]
/oi:/ yummy HH2 [oii:], [oiji:], [o::]
/okaei/ welcome back LHL0 [ka:i:]
*
/p/ bread H1 [pa]
/pi/ toss H1 [poi]
/sakana/ sh LLL0 [nnaka], [naka]
*
/s k/ like LL2 [ski]
*
/teta/ came off LLL1 [doteta]
/toke:/ clock LH0 [ke:]
*
/ttto/ daddy HL1 [totto]
/yo/ (emphatic marker) L0 [yo]
/z:/ elephant H1 [zo:]
*
= truncated form.
= truncated form not included in statistical analysis because the target contains a
devoiced vowel or // whose omission can lead to ambiguous syllable counts.
= location of the pitch accent. Words without an accent mark are unaccented.
438 Mitsuhiko Ota
Part IV
Perspectives and challenges
16 Aview from developmental psychology
Lorraine McCune
In developing spoken language children integrate their ability to vocalize with
the experience of meaning. Language typically begins with a period of at least
several months where children express meanings with one word at a time. It is
therefore not surprising that detailed analysis of childrens word production
strategies has yielded the information that their production process, level of
phonetic skill, and phonological development are organized at the word level.
In this chapter I will rst address the critical importance of template research.
I then consider the relationship between research in early child phonology and
more general studies of the rst phase of language acquisition, and the approach
to the integration of these diverse ideas that Marilyn Vihman and I have
developed. Next is a section on use of the term representation and the manner
in which mental representation and entry into language interact in development.
I then briey review a physiologically based theory regarding linguistic repre-
sentation. Finally, I propose a dynamic systems view of the transition into
language.
The importance of the template research
The early case studies exposing individual childrens idiosyncratic patterns of
single word production (e.g., Priestly 1977; Macken 1978) did not immediately
lead to the hypothesis that such personal shaping of word productions is a phase
of typical development. That hypothesis must now be seriously entertained.
From a developmental perspective the recognition that adopting one or more
word production templates may be a typical step in the acquisition of language
for most children (Vihman and Croft 2007) is a critical discovery because it
potentially adds to the known developmental sequence of vocal behaviors
characterizing the transition to language. There was a time when babbling
was considered unrelated to speech (Jakobson, 1941/1968). Resolution of that
issue has allowed researchers to recognize prelinguistic vocalizations as inu-
ential, and to use them as a resource in plotting the childs path to language.
Recognizing individual complex and consistent motor production patterns that
are closely related across words for many children, yet differ by child, portends
an even more radical turn in our approach to childrens transition to language
than earlier recognition of the importance of babbling.
441
Vihman and Croft (2007), citing Vihman and Velleman (2000), describe
templates as follows: specic phonological patterns which t many of the
words that the child attempts (these words are said to be selected), but which are
also extended to words that are less close to the template (these words are then
adapted to t the template) (p. 690). They further describe the process of
recognizing the inuence of a template on a childs productions:
Three types of clues are generally used to identify a childs word template(s):
(a) Consistency of patterning in a substantial number of the child forms for words
produced in one or more recording sessions or over a period of some weeks or
months;
(b) The occurrence of unusual phonological correspondences between adult and child
forms (i.e., rules or processes or repairs to target word violations of child
constraints), under the inuence of a dominating pattern or template;
(c) Frequently, a sharp increase in words attempted that either t or can be tted into the
pattern. (p. 693)
The schematic pattern that is used to describe a childs template in this volume
and elsewhere describes the shapes that can be encompassed by the childs
phonetically interrelated word productions. In some sense this can be seen as a
basic phonetic score for expressing meaning. In a given case of communicative
production I envision a child as experiencing simultaneously the intention
to communicate, an internal sense of meaning, and the activation of motor
procedures. For a certain period of development the motor processes seem to be
dominated by a consistent motor framework that nevertheless is modied by
characteristics of individual words. In my view the development of one or more
templates provides that consistent motor framework. Template patterns ordina-
rily emerge from earlier processes such as the Vocal Motor Schemes (VMS)
initially described by McCune and Vihman (1987, 2001). The crosslinguistic
examples in this volume show template development and use to be a dynamic
process integrating the childs ongoing phonetic development with co-
occurring development of word meanings.
Phonology and child language study: relationships over time
As Vihman and Croft (2007) mention, Ferguson and Farwell (1975) reported
some surprises in their ndings that pointed ahead to the approach to phono-
logical development explored in this volume: One inconsistency is the
existence of a high level of variation of word forms. The range of variability
plus certain regular forms of variation together make it difcult to make state-
ments about either phonological contrasts or unique underlying forms and
systematic rules, so that traditional forms of phonological analysis are not
strictly applicable (p. 22). Additional surprising aspects included recogni-
tion of greater accuracy in earlier productions that later showed reduction in
accuracy, and the surprising selectivity in the words individual children chose
to produce. With detailed examination of individual children over the ensuing
442 Lorraine McCune
years it became clear that phonological development was nonlinear and not
easily explained by existing models. Meanwhile many researchers in the
broader eld of child language remained unaware of this news for some time.
Ferguson and Farwell were writing in the midst of a strong resurgence of
interest in the single-word period. The work of early diarists had formed the
basis for major theoretical proposals regarding the joint emergence of sound and
meaning in children acquiring spoken language (e.g., Werner and Kaplan 1963;
Piaget 1962). Diary studies had necessarily been limited in their use of phonetic
analysis, although some tried to capture childrens phonetic forms on the y,
with results that were useful to these theoretical proposals. Surprisingly, in the
1970s, when audio and video recording allowed detailed transcription of infant
utterances, the major longitudinal studies of the single-word period undertaken
by psychologists, and even some linguists, had very little to say regarding the
phonetics or phonology of single-word speech. It was only researchers devoted
to phonological development who delved deeply into the specic phonetic
forms children produced in relation to intended language targets. At the same
time, the major theoretical proposals of Piaget and Werner and Kaplan were
mentioned respectfully but not fully integrated into research paradigms address-
ing early transitions in language.
Bloom and Lahey (1978), who summarized and interpreted the develop-
mental work of this period, identied the interaction of form, content and
function as critical to both language acquisition and our understanding of
the process of acquisition. Yet early phonetic development (surely a critical
aspect of form) could not be integrated into their work, because of the
isolation of studies of phonological development from other work on acquis-
ition. Bloom and Lahey remark: Virtually all of the studies of phonological
development before the 1970s took isolated speech sounds as the unit for
analysis, and tested children in the age range of three to eight years (1978:
100). Reviewing studies of infant vocal production, they report the growing
amount of evidence about the parameters of the sounds that infants make
along with a consensus that there is little overlap between the sounds made in
infancy and the sounds of early speech (p. 88; a conclusion that was already
being disputed in the eld of child phonology). In a prescient section devoted
to phonological development, they reviewed the work of several authors
included in the current volume (e.g., Ferguson and Farwell 1975; Waterson
1971), recognizing the rising interest of the word as a unit of analysis for early
phonological development. For example, they cite Watersons (1971) view
that children are able to capture perceptual features from the words they hear
and derive structural information or schemata to guide their own productions.
Similarly, they reported Ingrams (1974) recognition of child-specic produc-
tion tendencies leading to word productions that differ from typical adult
productions. But this information is isolated in the volume, not integrated
with the developmental themes explored.
Aview from developmental psychology 443
In her 1993 monograph Bloom provided an updated review of child phonol-
ogy research, including the recognition of continuity between babbling and
speech. However, even this excellent volume maintained the division between
phonological studies of early language and the study reported in the mono-
graph. The phonological data have no recognizable effects on the questions
asked or the analyses performed in the 1993 study. Childrens early words were
assumed to have both relevance and a consistent form that make them recog-
nizable by parents and interested others as conventional words in the language
(p. 82, emphasis added). The fact that childrens forms differed from those
typical of adults was recognized, but the differences were expected not to be too
great and to be consistent for a given child. There has rarely been recognition
outside of the phonological literature that childrens several productions of a
given lexical item might vary in phonetic form, or that such variation is of
interest. Phonetic and phonological development was largely omitted from
these studies. Current emphasis on parent report for language assessment has
only exacerbated this problem. There is a great need for the integration of
phonological analysis within the broader acquisition literature. If motor organ-
ization for production is a critical acquisition, the omission of phonetic and
phonological analysis seriously limits the validity of theorizing.
In language research by developmental psychologists the kinds of word
meanings learned, their consistency across children, and rates of development,
were among the primary issues of interest (e.g., Bates, Benigni, Camaioni, and
Volterra 1979; Bloom 1973; McCune-Nicolich 1981; Nelson 1973). The devel-
opment of mental representation was considered a shift in cognitive ability
from earlier perceptual or sensory motor processes that was important for
language acquisition, and some studies targeted relationships between non-
linguistic measures, such as play, with language. But little attention was given
to a possible internal represented form of specic adult words in relation to a
childs productions, which might both differ from the adults form and vary
among themselves. For researchers in child phonology this was and remains a
critical issue.
For example, Menn (1983) presented what she considered a minimal de-
nition of a lexicon that includes attention to form: it at least denotes a
collection of stored, accessible, memorized bits of information about the sounds
and meanings of words and/or their component meaningful parts (p. 8). In
contrast, child development studies equate the lexicon with a vocabulary list!
To oversimplify, studies of phonological development aim (among other goals)
to determine what a child comes to know that allows (a) comprehension of
specic sound sequences as language and (b) production of sufciently closely
matching sequences to be recognizable as such by adults who speak the
language. The question of how the child comes to this knowledge is also of
critical interest. Menns denition of a lexicon (above) reects information
processing theory, but child phonology has utilized a variety of theoretical
models over the decades, as summarized by Vihman (forthcoming 2014).
444 Lorraine McCune
Integrating child phonology into general studies of language
acquisition
My initial studies were part of the urry of developmental data collection with
language acquisition goals that began in the 1970s, taking no account of
phonology (e.g., McCune-Nicolich 1981). A major early goal of my work
was to explore the relationship between the development of representational
play and the development of language, under the assumption that the capacity
for mental representation was an underlying development affecting various
domains of behavior. I found that children showed hypothesized language
levels only when they also exhibited the appropriate hypothesized play levels.
However, a number of children showed delay between the play and language
achievements. I suspected limitations in their ability to produce speech, but had
no idea about how to address this issue.
Collaboration with Marilyn Vihman made it possible to integrate phonolog-
ical considerations into the study of the transition into language in our joint
work (Vihman and McCune 1994; McCune and Vihman 2001), in contrast with
my earlier, more semantically based studies (and her earlier phonologically
based studies). The joint work allowed us to bridge the gap between earlier
psychological and phonological approaches. The dynamic systems approach to
language recognizes that a number of underlying developmental variables,
each with its own trajectory, contribute to a childs shift from prelinguistic
status to becoming a language user (McCune 2008; Thelen 1991). Along with
phonetic skill, the ability to represent internal meaning with external symbols
(e.g., words, signs, or play acts) is a primary requirement for the transition to
language. The childs capacity for mental representation (symbolic ability)
develops over time through interaction with objects and people in the environ-
ment, in the context of cognitive and maturational processes.
Werner and Kaplans (1963) Symbol Formation was at once ahead of and
behind its time. The theory boldly addressed the need to derive something from
nothing: to start with a child who knows nothing of language, not even that
language exists, and show how development might proceed from the earliest
sounds and bodily movements, in a human social environment, to the produc-
tion of sentences. Theirs is essentially a cognitive and embodied model written
with the grace of a philosophical treatise. It entered an intellectual world
dominated by the perspective of behaviorist thinking, soon to be overtaken by
the Chomsky revolution. This book was revered by developmental students
of child language, but its message was never integrated into mainstream
research. It remains a rich theoretical source for understanding how children
come to language. Research developments over the past half-century have only
increased its relevance. In what follows I will use their model to sketch an
interpretation of how the developments in child phonology included in this
volume allow a more complete theoretical view of how children come to
language.
Aview from developmental psychology 445
What sort of representation?
Mental representation, in the sense of Symbol Formation, refers to the relation-
ship between one element, dened as a symbol, and another element, the
symbolized, which the symbol is said to represent. Simply put, for example, a
word can be considered to symbolize an underlying meaning. Mental represen-
tation from this perspective is a contentful state of consciousness (Searle 1992)
rather than a neural code. This sort of relationship was the primary psychological
meaning of the term representation prior to broad adoption of the computer
metaphor of mind. A neurological basis was assumed, but the recognized lack
of knowledge regarding brain physiology and processes prevented theorists from
proposing models of these relationships. More recently, computer simulations
and brain-imaging techniques have emboldened theorists to model possible
neurological relationships underlying behavior. The language of neurophysio-
logical studies has been adopted in simulations, leading to a confusing supercial
similarity with physiologically based neurological research.
Underlying linguistic representation of a more abstract sort was proposed at
least as early as Chomsky (1965). Gradually the idea that brain and behavioral
processes were guided by the interactions, and perhaps computations, of under-
lying, in some sense physiologically based representations became the entrenched
assumption of many cognitive and linguistic approaches. Emphasis on the role of
such representation in the acquisition process was a logical next step, given this
view of adult language. In contrast, psychological studies of development tended
to begin with descriptions of child behavior and attempted to infer from them
interpretations of meaning and/or communicative goal. The current volume
strongly emphasizes approaches to understanding the relationship between the
underlying or represented formof an adult word that children hear and the forms
that they produce. The path between the two is puzzling. In what way might the
adult input form be represented by the child? In what way might this represented
formprovide a basis for the childs very different production? In my view, the lack
of detailed knowledge regarding the neurological mechanisms involved in com-
prehending and producing language limits our capacity to answer these questions.
Instead I suggest that the more detailed exploration of the motor aspects of speech
production implicated by the template ndings, along with close attention to the
emerging evidence of embodied aspects of meaning, may bring us closer to a
physiologically based understanding of language acquisition. In the following
I address these issues more fully.
Menn, Schmidt, and Nicholas (this volume), proposing an essentially cogni-
tive model of phonological development, address the ambiguity in the use of the
term when discussing the representation of a word: An under-acknowledged
problem in linguistics, and even in psycholinguistics, is that we use the term
representation as in mental representation, underlying representation,
surface representation, semantic representation, etc. without discussing the
concept of representation itself (pp. 4689). Menn et al. proceed to clarify
446 Lorraine McCune
usage within their model and discuss the broad range of information that must
be included in the representation of a word. The focus is entirely on assumed
neurological representation rather than mental representation as a conscious
state. The Menn et al. Linked-Attractor model aims to integrate the best of
current phonological representation theories into a workable whole, in line with
the available data. At the same time Menn and her co-authors acknowledge that
the linked attractor model is not appropriate for predictions at least until it can
be modeled on a computer (p. 485), suggesting that the basis of such analyses
is as much in computational modeling as in analyzing child data. There is some
controversy within philosophy of science between model-builders and other
philosophers, who suggest that proposed cognitive/neurological processing
theories must go beyond models, and be physiologically demonstrable in
order to prove their value. For example, Stich (1992) states that to be exploited
in a respectable scientic theory a concept must be naturalizable (p. 258), that
is, able to be described at a physiological level.
Menn et al. (this volume) recognize that current models are schematic
compared to the real level of explanation the level of patterns of neural
activity . . . but were not even close to being able to get data at that level, or
interpreting them if we had them (p. 495). The daunting distance between
research on the behavioral phenomena discussed in this volume and the neuro-
logical bases of development is partly attributable to the wide division between
behavioral and neurological investigation. Rapid development in neurological
study over the past decade or so could provide the basis for more empirically
rooted neurological theorizing about language acquisition. However, neuro-
logical investigations fail to take note of developmental ndings in phonology, a
barrier parallel to developmental linguists and psychologists lack of familiar-
ity with progress in neurological work. An additional disconnect is between
models of brain development based in cognitive psychology (most emphasized
in the phonological literature) and more physiologically based models where
research relevant to but lacking a connection with phonological development is
advancing.
The development of more biologically oriented models from the study of
embodied cognition (e.g., Johnson 1987, 2008) and those based on neuro-
logical research (e.g., Hickok and Poeppel 2004) provide additional relevant
perspectives on what physiological processes may contribute to language
comprehension and production. In neurological studies the term representa-
tion refers to physical brain locations at various levels of detail (from individ-
ual neuron to functional area) that have demonstrable physiological relations
with bodily elements or processes. Kent (2007) addressed the anatomic, motor,
and sensory foundations of speech development in children, suggesting that a
fruitful direction for theory and research may be found in the study of mirror
neurons (Gallese, Fadiga, Fogassi, and Rizzolatti 1996). Considering physio-
logical and neurological information in relation to language acquisition in
general, and childrens word productions specically, should enhance our
Aview from developmental psychology 447
approaches to understanding developmental trajectories and individual differ-
ences across children. Since the initial description of the mirror neuron system
in monkeys (Rizzolatti et al. 1988; di Pellegrino et al. 1992) there has been
sufcient progress to demonstrate the high likelihood of such a system in
humans (Rizzolatti and Craighero 2004). This has allowed the beginning of
theorizing regarding mirror neurons in infancy (Del Giudice, Manera, and
Keysers 2009) and even research demonstrating the development of such a
system in human infants by 6 months of age (e.g., Lepage and Thoret 2006;
Marshall, Young, and Meltzoff 2010; Nystrm 2008). This line of study may be
of critical importance in addressing a wide range of issues in child language
research, but it is beyond the scope of the current chapter.
In my view, both physiologically based neurological representation and
mental representation as a conscious experience of meaning are essential
aspects of language.
Development of mental representation and early language
The capacity for mental representation dened as a conscious contentful state
develops over the second year of life and is assumed to underlie different forms
of language and play (McCune 1995). Context-limited words (termed proto-
words by Menn 1983) should be distinguished from referential words
(McCune 2008). The former occur embedded within situations at the earliest
levels of representational play (contemporaneous with the transition to mental
representation, as assessed in object permanence tasks: Piaget 1962, Ramsay
and Campos 1978). Play at this period involves simple schemes limited in
application to the childs own body (e.g., putting a cup or toy bottle to the
lips, or brush to the hair; McCune 1995). A child with phonetic skill, for
example having VMS [p/b] may, with parental assistance, learn a word such
as bye-bye, always accompanied by a hand wave and often occurring when
people are departing. Context and word are inextricably linked and production
is often based on a Vocal Motor Scheme (VMS). Some initially context-limited
words may be extended to referential use when the child develops referential
capacity; others will drop out of the repertoire, or continue to be used in limited
but appropriate circumstances.
Werner and Kaplan, from detailed diary studies, traced the emergence of
clearly referential words along a pathway beginning with sounds produced
automatically in certain circumstances, such as the sounds of eating. They
report that both Hildegard Leopold and a child studied by Lewis (1936) derived
their initial vocables, as Werner and Kaplan term the childs initial words, from
eating sounds (Hildegarde: [m]; Aments niece, studied by Lewis: [mammam]).
These referred to a large sphere of events related to food-getting and food
eating (Werner and Kaplan 1963: 111). In this example, rather than being
highly restricted, the words application is diffuse yet context-embedded.
In theory, the child rst experiences the sounds generated while ingesting
448 Lorraine McCune
food, then begins to be reminded of these sounds in contexts related to eating.
This leads to production somewhat separate from the actual eating that origi-
nally generated the sounds. For Amants niece the vocable mammam began as a
context-limited word related to eating and food but was gradually shaped in
interaction with the ambient language.
Werner and Kaplan proposed that children form their initial symbols in the
process of gradually developing the ability to represent (i.e., consciously
experience) situations internally, in the absence of perceptual support.
Grounding in bodily experience is seen as critical. An external symbol, such
as a word form, is constructed in relation to the underlying meaning that it
comes to express. They used the diary examples to demonstrate that the earliest
vocables may begin in relation to a sound that is indistinguishable froma natural
activity. In both of the cases cited by Werner and Kaplan, as additional words
were learned, the reference of the initial vocable became restricted. Aments
niece initially (at 354 days of age) used mammamin reference to her mother and
sister as well as to bread, cakes, and cooked dishes. By 597 days of age clearly
referential words were produced: mammam was delimited to cooked dishes,
while bread and cakes were brodi and her mother was mama, her sister desi.
Werner and Kaplan attribute these developments to an underlying process of
differentiation. As the child differentiates the sound produced while eating,
producing it in the absence of actual ingestion, ongoing experience of the
language accompanying various events of importance to the child facilitates
development of additional meanings in relation to other adult words. As new
meanings come to be used, the original vocalization becomes restricted in its
meaning and use. Then, having established the potential for meaningful vocal-
ization outside the natural source context, the child begins to differentiate the
wider variety of meanings expressed with various vocal forms by speakers in
her environment. A parallel process of integration characterizes the relationship
between internal meanings and the words that come to represent them.
As a result of the transition to referential language the child begins extending
word meanings beyond the original context where they were learned, as well as
differentiating meanings that were overly broad (e.g., the changes in application
of mammam, above). Werner and Kaplan do not specify the referential tran-
sition by name, but it appears in the developmental trajectory of their examples.
These theorists did not consider the formmeaning link to be arbitrary. Rather,
word form and meaning are co-constructed through a process termed dynamic
schematizing. The word form (symbolic vehicle) and meaning (symbolized)
remain related at a neurological level through this developmental process.
Dynamic schematizing is dened as the process that allows the differentiation
of varied aspects of meaning and form, for example, supporting the transition
from use of the original form mammam in varied contexts to the use of separate
forms in relation to each of these contexts (e.g., brodi for cooked dishes and
mama for the childs mother). As forms become differentiated, each becomes
more fully integrated with its internal meaning. Children experience the words
Aview from developmental psychology 449
of the language from adults, but then construct both form and meaning through
their own internal processes. (The discovery of template-based production
patterns provides a window into this interactive process.) Because of the joint
developmental history of sound and meaning, hearing a word instantiates an
internal contentful state of meaning. For a child who has achieved the capacity
for referential language, various external circumstances related to that same
state for the child (but perhaps not for adults) may call the word to mind,
presumably its phonetic potential as well as its meaning, leading to production.
The fact that children generalize their productions beyond those expected by
adults testies to an internal constructive process. While such processes must
have a biological basis, this was not discussed in early works (e.g., Piaget 1962;
Sartre 1948, or Werner and Kaplan 1963). Children also demonstrate a capacity
for mental representation outside language by the time these developments
occur (e.g., McCune 1995).
Citing Werners (1957) theme of the syncretic nature of the young childs
experience, Tucker (2002) supports Werner and Kaplans claim that linguistic
representations are non-arbitrary (p. 67). Although we may not be able to
apprehend the fact in experience, the neural architecture teaches that the mean-
ing of language is multileveled[,] from the gut level that is inherently subjective
to the surface articulation that is communicable within the articulatory conven-
tions of the culture (p. 68).
A current neurological model
Hickok and Poeppel (2004) developed a large-scale model of language inte-
grating data from neuropsychology, neuroimaging, and psycholinguistics,
drawing on relatively recent analyses of the cortical organization of vision
to guide this new framework. They propose a bilateral ventral stream of neuro-
logical activity that integrates the acoustic and semantic aspects of language
and a simultaneously active bilateral dorsal stream integrating acoustic and
motor aspects of language. They see these systems as differentiated, but prob-
ably interacting. This model would support the Werner and Kaplan view.
The vehicle/meaning relationship is considered non-arbitrary, but not in the
sense that similar vocal forms should share meaning across languages. Rather,
meaning and form are co-constructed, within the individual, with (in contem-
porary terms) mutual neurological activation in relevant brain structures. Within
the eld of child phonology there is clear recognition that a childs production of
the same word varies both within a given session and across time (e.g., Ferguson
and Farwell 1975; Macken 1978; Waterson 1971), yet eventually child
word productions match those of the ambient language. Recognition of this
developmental course supports a construction process, as does the existence
of child-specic templates.
Differential neurological response to motor-specic language lends credence
to the non-arbitrary aspect of meaning. Recent research with adults has found
450 Lorraine McCune
differentiated neurological activation while participants listened to action sen-
tences and verbs. Pulvermller (2002) reported differential EEG activation at
dorsal sites closer to the cortical leg area for listening to verbs such as walking
versus stronger activation at inferior sites next to motor representation of the
hand and mouth for verbs such as talking. Tettamanti et al. (2005) found that
cortical areas that were active during action observation (mirror neuron areas)
also showed differential activation during listening to sentences describing
actions by mouth, hand/arm, or leg. Buccino et al. (2005) found motor-
evoked-potential (MEP) changes specic to the hand or foot neuromotor area
in response to action sentences describing limb actions. Results were specic
for the effector involved in the action sentence heard, and listening to abstract
control sentences had no effect in either study.
These adult effects might be learned associations going from language to
motor activation, or they may support the notion of deep bodily construction
processes in the development of meaning/vocalization relationships that have
left a lasting mark. It is also possible that such peripheral neurological reactions
occur in the moment as we comprehend or produce language and constitute an
aspect of the language meanings we experience.
Language as a dynamic system: the inuence of vocal variables
Language development is a highly complex process best understood as the
emergence of a dynamic system. As conceived by Thelen and colleagues
(e.g., Thelen 1989, 1991; Thelen and Smith 1994, 2007), a dynamic systems
approach to development entails pervasive interaction between the organism
and the environment and the possibly asymmetrical development of subsidiary
systems within the organism. (The earlier development of representational play
in comparison with language milestones exemplies this (e.g., McCune 1995).)
The development of vocal production ability is both a process in the path toward
language and the product of additional underlying developments. McCune
(1992, 2008), relying on joint work with Marilyn Vihman (e.g., McCune and
Vihman 1987, 2001; Vihman and McCune 1994), demonstrated that the assess-
ment of a number of underlying variables, including both vocal development
and the development of mental representation, interacting in a dynamic system,
could predict the timing of childrens shift to referential language.
The McCune model for a dynamic systems view of the transition into
language begins with a basis in the development of mental representation as
an underlying variable that affects the observable behaviors included in the
model. The model assumes a positive social/emotional relationship with one or
more adults in the language community. The two nonvocal skills included are
sensorimotor cognition as dened by Piaget (e.g., Piaget and Inhelder 1969)
and assessed by measures of object permanence and representational play, as
described above. The vocal components of the model were operationalized
following some initial exploratory investigation of data, and then veried on
Aview from developmental psychology 451
additional data. Two vocal variables were found to be predictive of the refer-
ential shift in studies of McCunes and Vihmans combined sample of 20
children between 9 and 16 months of age. The rst, Vocal Motor Scheme
(VMS) production, is a measure of vocal production skill that McCune and
Vihman (1987) related to the transition to reference. It is assessed by frequent
and consistent use of one or more specic supraglottal consonants.
The term Vocal Motor Scheme has its origin in Piagetian sensorimotor
cognition, where skill with a particular movement is termed a scheme. For
example, Thelen, Corbetta, and Spencer (1996) demonstrated that 6-month-old
childrens successive reaches toward an object showed random variation in
trajectory, while by 8 months each child showed a relatively consistent trajec-
tory in repeated reaches, achieving a reaching scheme which could vary with
reference to distance and target characteristics. Analogously, repeated accurate
production of the motor action yielding a given supraglottal consonant, or some
other vocal target, is considered a Vocal Motor Scheme. Children who made the
transition to referential word production by 16 months, the nal month of the
study, all showed VMS-level competence with at least two supraglottal con-
sonants by the time of that transition. Of the words used at both 15 and 16
months, on average 90 percent incorporated each childs specic VMS reper-
toire (McCune and Vihman 2001). Vihmans continuing studies have more fully
established the value of this variable (e.g., Keren-Portnoy, Vihman, DePaolis,
Whitaker, and Williams 2010; DePaolis, Vihman, and Keren-Portnoy 2011;
DePaolis, Vihman, and Nakai in press).
Vocalization, like all behavior, has its basis in the neurobiology of the organ-
ism. Tucker (2002) reported that the earlier myelination of primary sensory and
motor cortices in comparison to other brain areas provides the basis for the vocal
control of articulation needed for babble. Thus basic production processes
stabilize early. This would account for a childs frequent production of what
has sometimes been called a favorite sound (Ferguson 1978; in our terms,
VMS). Neurological and motor developments contribute to the consistency of
production that provides coherent feedback to the child herself and leads to adult
recognition of the sounds recurrence. More cognitively directed brain regions are
slower in myelination, providing ongoing opportunity for developing meanings
in relation to more complex sequences of articulatory gesture. Tucker suggests
that the retention of juvenile plasticity in limbic cortices may be integral to the
exibility of human adult cognition, allowing adults to learn the meanings of the
words in a new language, even though to native speakers, they remain only
marginally competent with the sensorimotor articulation of those words (p. 74).
The VMS measure relies on the motoric stability described by Tucker. The
development of a word production template is a more complex matter and may
await additional developments, as a template by denition involves word learning
and so the interaction of sound and meaning.
In our initial report (McCune and Vihman 1987) we noted complex patterns
evolving from initial VMS skill that affected the word shapes the children
452 Lorraine McCune
produced. Recognition that this more complex development incorporated
phonetic aspects of words in the ambient language as well as the childs own
phonetic tendencies suggested that word recipes or word production pat-
terns (Vihman and Velleman 1989), now termed templates, reect distinct
processes. However, there is typically continuity in motor development
between the two (Vihman, Velleman, and McCune 1994). This volume clearly
demonstrates the theoretical importance of the template form, but understanding
the developmental role of this process in childrens language acquisition will
require additional study.
The second vocal variable implicated in the shift to reference is the commu-
nicative grunt (McCune, Vihman, Roug-Hellichius, Delery, and Gogate 1996).
Two important ndings linked communicative grunts with the transition to
reference (McCune et al. 1996). First, we found that referential word produc-
tion, for the early talkers studied, and referential word comprehension for the
later talkers, were rst observed either in the same monthly session as the onset
of communicative grunt use or in the following session. Second, the early
talkers all showed sharp increases in word production following only limited
use of context-dependent words in earlier sessions. Both earlier and later talkers
more than doubled communicative events (including gesture) at the time they
began communicative grunts (McCune 2008).
Grunts occur autonomically following reexive laryngeal closure under
conditions of effort or physiological stress. Such reex closure, across mamma-
lian species, tends to increase oxygenation to the blood and restore homeostasis,
or facilitate ongoing effortful activity. McCune et al. (1996) found that grunts
rst co-occurred with physical effort, then with focused attention, before shift-
ing to communicative use. To account for the temporal linkage between com-
municative grunt use and the shift to reference we reasoned that childrens
experience of their own grunt under conditions of effort or attention (internal
meaningful states) might prompt an initial recognition of sound/meaning link-
age, leading to increased attention to the meanings available in their linguistic
environment.
In summary, this dynamic systems model predicts that children will make the
transition to referential language production only when (1) mental representa-
tion reaches the level shown in play by combining pretend acts, (2) phonetic
development reaches a critical point (dened in early talkers by identication of
at least two VMS), (3) communicative intent comes to be realized by production
of the natural vocalization dened as a communicative grunt. The children
studied all showed communicative gestures earlier than communicative grunts,
but we could not determine whether this is an essential variable in the model.
Children lacking the phonetic skill indexed by two VMS but exhibiting the
requisite communicative and representational skills showed referential compre-
hension by gesture, in the absence of word production. The variables identied
above are all indices of underlying abilities and so might be assessed in
other ways. The three vocal variables I have emphasized in this chapter,
Aview from developmental psychology 453
VMS, communicative grunts, and word templates, share the property that each
facilitates some aspect of language development over time. In addition they all
contribute in the moment to facilitating communicative production. This dual
behavioral and developmental role is typical of variables within a dynamic
system.
Automaticity and language production
Although we discuss motor planning for speech, such planning must occur
simultaneously with production, or disuency results. Vocal communication
reects automaticity between intention and expression. Vocal Motor Schemes,
communicative grunts and word templates all contribute to the automaticity of
production during the process of language development. VMS consonants
dominate childrens early words, no doubt due to ease of production (partly as
a result of practice and familiarity). Word templates provide formats that can
be shaped in word production to integrate child phonetic capacity with ambient
language word shape. Communicative grunts in infancy may be a by-product of
the automaticity required in vocal production.
In adults increased motor activity in the laryngeal muscles is observed
immediately before vocalization, suggesting that laryngeal activation (the
basis of communicative grunts) may be an automatic response to the intention
to vocalize (Buchtal and Faaborg-Anderson 1964; Kirchner 1987; see also
Esling 2012, on infant laryngeal initiation). Communicative grunts, which
continue with some frequency throughout the single-word period, may be
stand-in vocalizations for missing or slow-to-be-recalled words.
Consider the adult experience of searching for a word. This search may not be
silent, as the speaker lls in with sounds such as eh or um (Goffman 1978),
suggesting continuity with infant grunts, which take a similar form. These
vocalizations also serve as pragmatic or phonetic devices to maintain the
conversational rhythm. Ward (2004, 2006) terms these vocal expressions con-
versational grunts. He reports their broad use in English and other languages,
with some correspondence between phonetic aspects of such expressions and
their meaning in context. These expressions seem to be continuous with infant
communicative grunts.
The critical advantages of automaticity in speech production may be the basis
for early reliance on VMS consonants and the development of production
templates in young children. Awell-practiced phonetic repertoire should impact
directly on the transition to referential language use because as the child
experiences the intention to communicate, the internal meaningful experience
may nd expression only through fairly routinized motor activity. The childs
internal idea (or meaning) is essentially clothed with sounds and words in the
process of its formation. Having basic motoric potential at the ready when the
communicative intention is experienced must be an essential feature of
454 Lorraine McCune
communicative speech. A communicative intent, absent motor potential, may
result in a communicative grunt.
There may not be sufcient understanding of the speech-motor develop-
ment underlying template-based word production to suggest a coherent theory
of how the capacity for reference might be integrated with this important
development. Progress in understanding the source, function, and develop-
mental trajectory of word templates depends, at least in part, on determining
their phonetic bases in individual children. Can production commonalities be
identied across many childrens template formats that might suggest whether
and how the templates increase production ease and thus automaticity? While
based in the individual childs phonology, during the instantaneous process of
producing a word it is also possible that echoic inuences from adult words
heard in given contexts might affect the formation of the specic template-
based word in its production. Throughout the single-word period word pro-
duction would be affected by these various memory- and context-based
inuences.
The next phase in language acquisition is the transition to word combina-
tions. I found sharp acceleration in multiword production following a shift to
more advanced symbolic play (McCune 1995), but have not developed a
dynamic systems model of this next phase. There is minimal research evidence
on the relationship between template-based word production and the transition
to combinations. Yet the templates begin to occur just as word production is
rapidly increasing and word combinations are imminent. It would seem that the
attraction to automatic template-based production would need to be either
eliminated or incorporated into uent multiword production. The fact that
there is a two-word stage in childrens development toward grammar opens
the possibility that template effects might be seen in presyntactic combinations,
forming part of a dynamic system for this next transition. (See Donahue 1986
and Matthei 1989 for case studies demonstrating possible template effects in
this transition.)
Despite great progress in understanding neurological development and
functioning, and despite useful models of these processes, we do not actually
know, beyond our metaphors, how a word is produced or how it is recognized in
perception, although progress is being made from various directions. Except in
cases of recording individual brain cell activity, and brain-imaging techniques
identifying general areas of activation under specic conditions, the term repre-
sentation exists only in a metaphorical context. Phonologists understanding of
representation uses the metaphor of the brain as an information-processing
device. Psychologists ideas about mental representation depend upon unveri-
able states of consciousness that must also have some basis in neurophysiology.
The very limitations of our biological knowledge allow us freedom to conceptu-
alize different brain/behavior relationships. Perhaps words are based on relatively
stable internal representations, as some authors in the current volume assume,
or on more dynamic processes where such stability is lacking. It may be the case
Aview from developmental psychology 455
that the underlying neurological representation of a word exists only at the
moment of perception or production, emerging as a result of the particular task.
Physiological evidence that might specify the nature of our ongoing linguistic
knowledge is lacking. The studies of differential motor activation in response to
word meanings (mentioned above) suggest that the body itself participates in such
representation, as predicted by Varela, Thompson, and Rosch (1991) for neuro-
logical representation in general. Research is needed that combines relevant
aspects of phonological development with neurological study. Beyond this empir-
ical goal, theoretical integration is needed which will allow researchers from
various elds of endeavor to engage in the same task.
References
Bates, E., Benigni, L., Bretherton, I., Camaioni, L., and Volterra, V. (eds.) (1979). The
emergence of symbols: cognition and communication in infancy. New York: Wiley.
Bloom, L. (1993). The transition from infancy to language. New York: Cambridge
University Press.
Bloom, L. and Lahey, M. (1978). Language development and language disorders. New
York: Wiley.
Buccino, G., Riggio, T., Melli, G., Binkofski, F., Gallese, V., and Rizzolatti, G. (2005).
Listening to action-related sentences modulates the activity of the motor system: a
combined TMS and behavioral study. Cognitive Brain Research, 24, 35563.
Buchtal, F. and Faaborg-Anderson, K. L. (1964). Electromyography of laryngeal and
respiratory muscles. Annals of Otology, Rhinology and Laryngology, 73, 18121.
Chomsky, N. (1965) Aspects of a theory of syntax. Cambridge, MA: MIT Press.
Del Giudice, M., Manera, V., and Keysers, C. (2009). Programmed to learn: the ontogeny
of mirror neurons. Developmental Science, 12, 35063.
DePaolis, R., Vihman, M. M., and Keren-Portnoy, T. (2011). Do production patterns
inuence the processing of speech in prelinguistic infants? Infant Behavior and
Development, 34, 590601.
DePaolis, R., Vihman, M. M., and Nakai, S. (In press). The inuence of babbling patterns
on the processing of speech. Infant Behavior and Development.
di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., and Rizzolatti, G. (1992).
Understanding motor events. Experimental Brain Research, 91, 17680.
Donahue, M. L. (1986). Phonological constraints on the emergence of two-word utter-
ances. Journal of Child Language, 13, 20918.
Esling, J. (2012). Articulatory function of the larynx and the origins of speech. Plenary
paper presented at the 38th Annual Meeting of the Berkeley Linguistics Society.
Ferguson, C. A. (1978). Learning to pronounce: the earliest stages of phonological
development in the child. In F. D. Minie and L. L. Lloyd (eds.), Communicative
and cognitive abilities early behavioral assessment, pp. 27397. Baltimore:
University Park Press.
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language
acquisition. Language, 51, 41939. Reprinted in this volume as Chapter 4.
Gallese, V., Fadiga, L., Fogassi, L., and Rizzolatti, G. (1996). Action recognition in the
premotor cortex. Brain, 119, 593609.
Goffman, E. (1978). Response cries. Language, 54, 787815.
456 Lorraine McCune
Hickok, G. and Poeppel, D. (2004). Dorsal and ventral streams: a framework for
understanding aspects of the functional anatomy of language. Cognition, 92,
6799.
Ingram, D. (1974). Phonological rules in young children. Journal of Child Language,
1, 4964.
Jakobson, R. (1941/1968). Child language, aphasia and language universals, trans.
A. R. Keiler. The Hague: Mouton. (Orginally published as Kindersprache, Aphasic
und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell, 1941.)
Johnson, M. (1987). The body in the mind. University of Chicago Press.
(2007). The meaning of the body: aesthetics and human understanding. University of
Chicago Press.
Kent, R. (2007). In the mouths of babes: anatomic, motor and sensory foundations of
speech development in children. In R. Paul (ed.), Language disorders from a
developmental perspective: essays in honor of Robin S. Chapman, pp. 5581.
Mahwah, NJ: Lawrence Erlbaum.
Keren-Portnoy, T., Vihman, M. M., DePaolis, R., Whitaker, C., and Williams, N. M.
(2010). The role of vocal practice in constructing phonological working memory.
Journal of Speech, Language and Hearing Research, 53, 128093.
Kirchner, J. A. (1987). Laryngeal reex systems. In T. Baer, C. Sasaki, and K. Harris
(eds.), Laryngeal function in phonation and respiration, pp. 6570. Boston: Little,
Brown, and Company.
Leopold, W. F. (19301949). Speech development of a bilingual child. Evanston, IL:
Northwestern University Press.
Lepage, J. F. and Thoret, H. (2006). EEG evidence for the presence of an action
observation-execution matching system in children. European Journal of
Neuroscience, 23, 250510.
Lewis, M. M. (1936/1975). Infant speech: a study of the beginnings of language. New
York: Arno Press.
Macken, M. A. (1978). Permitted complexity in phonological development: one childs
acquisition of Spanish consonants. Lingua, 44, 21953.
Marshall, P. J., Young, T., and Meltzoff, A. N. (2010). Neural correlates of action
observation and execution in 14-month-old infants: an event-related EEG desyn-
chronization study. Developmental Science, 14, 47480.
Matthei, E. (1989). Crossing boundaries: more evidence for phonological constraints
on early multi-word utterances. Journal of Child Language, 16, 4154.
McCune, L. (1992). First words: a dynamic systems view. In C. A. Ferguson, L. Menn,
and C. Stoel-Gammon (eds.), Phonological development: models, research, impli-
cations, 31336. Parkton, MD: York Press.
(1995). A normative study of representational play at the transition to language.
Developmental Psychology, 31, 198206.
(2008) How children learn to learn language. New York: Oxford.
McCune, L. and Vihman, M. M. (1987). Vocal motor schemes. Papers and Reports in
Child Language Development, 26, 729.
(2001). Early phonetic and lexical development. Journal of Speech, Language and
Hearing Research, 44, 67084.
McCune, L., Vihman, M. M., Roug-Hellichius, L., Delery, D. B., and Gogate, L. (1996).
Grunt communication in human infants. Journal of Comparative Psychology, 110,
2737.
McCune-Nicolich, L. (1981). The cognitive basis of relational words. Journal of Child
Language, 8, 1536.
Aview from developmental psychology 457
Menn, L. (1983). Development of articulatory, phonetic and phonological capabilities. In
B. Butterworth (ed.), Language production, vol. 2, pp. 350. London: Academic
Press. Reprinted in this volume as Chapter 6.
Menn, L., Schmidt, E., and Nicholas, B. (This volume). Challenges to theories, charges
to a model: the Linked-Attractor model of phonological development.
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society
for Research in Child Development, 38, 12.
Nystrm, P. (2008). The infant mirror neuron system studied with high density EEG.
Social Neuroscience, 3, 33447.
Piaget, J. (1962). Play, dreams and imitation. New York: Norton.
Piaget, J. and Inhelder, B. (1969). The psychology of the child. New York: Basic Books.
Priestly, T. M. S. (1977). One idiosyncratic strategy in the aquisition of phonology.
Journal of Child Language, 4, 4566. Reprinted in this volume as Chapter 7.
Pulvermller, F. (2002). The neuroscience of language. Cambridge University Press.
Ramsay, D. and Campos, J. (1978). The onset of representation and entry into stage 6 of
object permanence development. Developmental Psychology, 52, 78597.
Rizzolatti, G., Camarda, R., Fogassi, L., Gentilucci, M., Luppino, G., and Matelli, M.
(1988). Functional organization of inferior area 6 in the macaque monkey. II. Area F5
and the control of distal movements. Experimental Brain Research, 71, 491507.
Rizzolatti, G. and Craighero, L. (2004). The mirror-neuron system. Annual Review
Neurosciences, 27, 16992.
Rizzolatti, G., Fadiga, L., Fogassi, L., and Gallese, V. (2002). From mirror neurons to
imitation: facts and speculations. In A. N. Meltzoff and W. Prinz (eds.), The
imitative mind: development, evolution and brain bases, pp. 24766. Cambridge
University Press.
Rosch, E. (2000). Reclaiming concepts. In R. Nunez and W. J. Freeman (eds.),
Reclaiming cognition: the primacy of action, intention and emotion, pp. 6177.
Cambridge, MA: MIT Press.
Sartre, J.-P. (1948/1962). The psychology of imagination. New York: Philosophical
Library.
Searle, J. (1992). The rediscovery of the mind. Cambridge, MA: MIT Press.
Stich, S. (1992). What is a theory of mental representation? Mind, 101, 24361.
Tettamanti, M., Buccino, G., Saccuman, M. C., Gallese, V., Danna, M., Scifo, P.,
Fazio, F., Rizzolatti, G., Cappa, S. F., and Perani, D. (2005). Listening to action-
related sentences activates frontal-parietal motor circuits. Journal of Cognitive
Neuroscience, 17, 27381.
Thelen, E. (1989). Self-organization in developmental processes: can systems
approaches work? In M. R. Gunnar and E. Thelen (eds.), The Minnesota symposia
on child psychology: systems and development, vol. 22, pp. 77117. Hillsdale, NJ:
Lawrence Erlbaum.
(1991). Motor aspects of emergent speech. In N. A. Krasnegor, D. M. Rumbaugh,
R. L. Schiefelbusch, and M. Studdert-Kennedy (eds.), Biological and behavioral
determinants of language development, pp. 33962. Hillsdale, NJ: Lawrence
Erlbaum.
Thelen, E., Corbetta, D., and Spencer, P. (1996). The development of reaching during the
rst year: the role of movement speed. Journal of Experimental Psychology:
Human Perception and Performance, 22, 105976.
Thelen, E. and Smith, L. (1994). A dynamic systems approach to the development of
cognition and action. Cambridge, MA: MIT Press.
458 Lorraine McCune
Thelen, E. and Smith, L. (2007). Dynamic systems theories. In W. Damon (ed.),
Handbook of child psychology, pp. 258311. New York: Wiley.
Tucker, D. M. (2002). Embodied meaning. In T. Givon and B. M. Malle (eds.), The
evolution of language out of pre-language, pp. 518. Amsterdam: John Benjamins.
Varela, F. J., Thompson, E., and Rosch, E. (1991). The embodied mind. Cambridge: MIT
Press.
Vihman, M. M. (Forthcoming 2014). Phonological development: the rst two years, 2nd
edn. Oxford: Blackwell.
Vihman, M. and Croft, W. (2007). Phonological development: toward a radical
templatic phonology. Linguistics, 45, 683725. Reprinted in this volume as
Chapter 2.
Vihman, M. M. and McCune, L. (1994). When is a word a word? Journal of Child
Language, 21, 51742.
Vihman, M. M. and Miller, R. (1988). Words and babble at the threshold of language
acquisition. In M. D. Smith and J. Locke (eds.), The emergent lexicon: the childs
development of a linguistic vocabulary, pp. 15184. New York: Academic Press.
Vihman, M. M. and Velleman, S. (1989). Phonological reorganization: a case study.
Language and Speech, 32, 14970. Reprinted in this volume as Chapter 8.
(2000). The construction of a rst phonology. Phonetica, 57, 25566.
Vihman, M. M., Velleman, S., and McCune, L. (1994). Howabstract is child phonology?
Towards an integration of linguistic and psychological approaches. In M. Yavas
(ed.), First and second language phonology, pp. 944. San Diego, CA: Singular
Publishing. Reprinted in this volume as Chapter 9.
Ward, N. (2004). Pragmatic functions of prosodic features in non-lexical utterances.
Speech Prosody, 3258.
(2006). Non-lexical conversational sounds in American English. Pragmatics and
Cognition, 14, 11384.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
Werner, H. (1957). The concept of development from a comparative and organismic
view. In D. B. Harris (ed.), The concept of development. Minneapolis: University of
Minnesota Press.
Werner, H. and Kaplan, B. (1963/1984). Symbol formation. New York: Wiley.
Aview from developmental psychology 459
17 Challenges to theories, charges to a model:
the Linked-Attractor model of phonological
development
Lise Menn, Ellen Schmidt, and Brent Nicholas
I. Introduction: why its time for a new model of phonological
development
A. Contemporary setting: the Linked-Attractor model
as a usage-based model
Usage-based (bottom-up, emergentist) models of phonology (e.g., Boyland,
2009; Bybee 2006, 2010; Johnson 2006; McMurray, Cole, and Munson 2011;
Peperkamp 2003, Pierrehumbert 2002, 2003), based strongly in laboratory pho-
nology and computational simulation, have become increasingly elaborated and
convincing, and they are of tremendous importance for child phonology. Probably
their most important contribution, and the one that is the focus of this chapter, is
that they permit us to consider the representation of a words form as something
that develops continuously over time in strength, precision, and accessibility a
reconceptualization of representation which has long been psycholinguistically
necessary for really understanding language development. And these models
allow us to bring frequency data to bear on development without ignoring the
equally important contributions of linguistic structure.
This volume as a whole belongs to the growing literature supplying evidence
for usage-based phonological development; that is, the data and arguments
supporting the claim that childrens phonology emerges principally from what
they hear and try to say, rather than from an innately guided grammar. Of course
there are universals, because the ambient languages are subject to general
constraints on their structures; additional constraints are imposed by the limi-
tations of the human infants articulation, memory, and perception. But the
chapters in this book, as well as other cross-linguistic work, show that differ-
ences across languages and individuals cannot be regarded as minor disturban-
ces to some basically deterministic pattern.
Usage-based models also offer ways to deal with pervasive phenomena that
older phonological theories treat as marginal e.g., effects of specic speaker
Thanks for many hours of discussion to members of the psycholinguistics and phonetics/phonology
community at the University of Colorado, notably Al Kim, Rebecca Scarborough, Les Sikos, Jill
Dufeld, and Bhuvana Narasimhan; also to Ben Munson, Janet Pierrehumbert, Ronnie Silber, and
Carol Stoel-Gammon; and especially to our editors Tamar Keren-Portnoy and Marilyn M. Vihman.
460
and of language-particular phonetics. They also raise and reframe fundamental
questions about the distinction between phonetic and phonological levels of
description, something that has long needed attention (really, how do children
learn that phones which are in complementary distribution can all be allophones
of a single phoneme? Is it possible that they dont learn any such thing?)
although consideration of these topics lies well beyond the current chapter.
Vihman and Crofts (2007) paper on Radical Templatic Phonology, reprinted
as the rst chapter in this volume, is the rst major attempt to bring usage-based
modeling into contact with substantial amounts of production data from
children developing their rst phonology. However, the Radical Templatic
model needs elaboration so that it can deal explicitly with how and why the
huge gap between a childs production and its adult target gradually closes as
she becomes a uent adult speaker of her language.
The present chapter concludes the volume by sketching our Linked-Attractor
model of phonological development (Menn, Schmidt, and Nicholas 2009). Our
model extends the Radical Templatic approach beyond its focus on the childs
developing sensory-motor output representations, augmenting it so that it can
also handle three other basic aspects of phonological representation: the childs
developing input representation of the adult model word, the web of relation-
ships among the input representations, and the childs auditory representation
of her own productions. Such an extension is necessary to account for an
individuals development of an adultlike phonology from prelinguistic begin-
nings; an important paper already moving in this direction is Munson, Edwards,
and Beckman (2012).
In a sense, the Linked-Attractor model brings together the two-lexicon model
(Menn 1983, also reprinted in this volume) and the Radical Templatic model
as ingredients in a new model of child phonology that is compatible with
current usage-based models of adult phonology, and that can grow into an
adult phonology through its experiences of speaking, storing, and understand-
ing words.
We do want to make it clear from the outset, however, that there is no reason
to stop describing child phonology in whatever terms make a particular phe-
nomenon easiest to think about. Rules have an intuitively transparent precision;
constraints describe fundamental regularities, which exist in tension with
the lumps induced by individual experience. The Linked-Attractor model
complements and enriches the insights created by generative and harmonic
approaches to phonology.
B. Historical setting: rules, constraints, abstraction
Child phonology historically concerns what happens to the childs representa-
tion of the adult surface form of a word when the child tries to say it (Kiparsky
and Menn 1977; Smith 1973; Menn 1983). For that reason, child phonology
rules were initially written from an adult-centered point of view: as formal
Challenges to theories, charges to a model 461
descriptions of the differences between the adult word and the child word.
This practice was partially justied by the observation that children appeared
to honor more distinctions in perception than in production; therefore, they
might have a complete representation of the form of the adult words they were
attempting (though Waterson [e.g., 1971] and Ingram 1974 thought otherwise).
If a child said [dk] for stick or [don] for stone, we called it [s]-deletion and
wrote [s] > |#_C.
Lets reviewa fewkey points about differences between child phonology and
general phonology in familiar theoretical frameworks. Generative Phonology
(Chomsky and Halle 1968) and its direct descendant, Autosegmental Phonology,
as well as the newer constraint-satisfaction approaches to phonology, Optimality
Theory and Harmonic Grammar (see Kager 1999; Kager, Pater, and Zonneveld
2004), were developed for the idealized adult speakerhearer. Both of these
types of theory operate in terms of what happens to an abstract underlying
segment in a particular phonological environment. How these abstract under-
lying forms get into the adults mind is not addressed that job has been left to us
developmentalists.
Because child phonology rules relate the adult surface form (or the childs
representation of it) to the childs surface form, a words underlying form in
child phonology is not nearly as abstract as its underlying form in Generative
Phonology or Optimality Theory. Achilds underlying formhas been abstracted
from what the child takes to be different tokens of the same word, so its
essentially the same as the words surface phonemic representation. In contrast,
the underlying forms of classical adult phonology are intended to account for
morphophonemic alternations (captive, captivity; critic, criticize), so they are
fairly remote from the phonemic surface.
In spite of this essential difference in the abstractness of childrens and adults
underlying forms in generative and constraint-based phonologies, the relation-
ships between underlying and output phonological patterns in the two cases
are similar enough for standard tools of phonology to have been able to bring a
reasonable amount of order to both kinds of data. After the reel-to-reel tape
recorder and the formalism of generative phonology converged to allow child
phonology to become more than a theoreticians toy, rules brought a substantial
amount of order into a messy little world, at least for children who had relatively
regular mappings from adult word to child word (air-brushing out the phonetic
details). The enormous power of generative rules handled physiologically
mysterious patterns like metathesis almost as routinely as it dealt with physio-
logically plausible ones like assimilation and deletion. (In retrospect, that was
not really a good thing, but thats a different story.)
However, the present volume documents a substantial range of phenomena
where familiar types of rules and constraints do not work smoothly enough to
bring order or insight. The problem of unruly child word forms is not new;
they were discussed very early (famously, by T. M. S. Priestly 1977, reprinted
in this volume), but there was no tool for dealing with them as mappings from
462 Lise Menn, Ellen Schmidt, and Brent Nicholas
adult form to child form. So unruly mappings were set aside, and researchers
only dealt with these childrens output forms; the patterns in these outputs
were variously called prosodies (Waterson 1971, this volume), canonical
forms (Ingram 1974, Macken 1978, Menn 1983, this volume) or templates
(Vihman 1996).
Rule-ordering was designed to handle some kinds of subregularities among
exceptions in adult language, and it could do the same for child phonology.
Interestingly, for most children, examples of rule ordering involving more than
one or two words are hard to nd. Smith (1973) see especially pp. 1322 and
158 ff. provides the major published set of ordered rules, arguing for their
ordering with the standard tools of phonological theory. One of the most
important types of rule ordering is when Rule A has to apply before Rule
B because it removes a segment that would otherwise be input to Rule B (this
is called bleeding order it bleeds off part of the potential input to a rule). To
write an example informally, Amahls cluster simplication rule /sw/ > [w]
precedes the labial postponement rule /CwVC/ > [CVC
+labial
], because sweetie
becomes [widi] while quit becomes [kip]. (If the labial postponement rule had
applied rst, the output for sweetie would be [sipi].) Such opaque patterns are
still easier to handle with ordered rules than with output constraints or any other
device.
In some cases, constraint theories like Optimality Theory can describe the
inputoutput relations quite well and this sometimes requires constraints that
appear to be unattested in adult language. For example, a number of children
acquiring English, including Hildegard (Leopold 193949), Patrick (P;
Waterson 1971), and Daniel (Menn 1971), had a top-ranked (i.e., unbreakable)
markedness constraint that sibilants may appear only in word-nal position. For
other examples, presented in the Optimality Theory framework, see Bernhardt
and Stemberger (1998).
The question of howrules work in real time and whether/in what sense they
are real has been a vexing issue for all approaches to child phonology (and
general phonology). While a sequence of ordered phonological rules or a grid of
ranked constraints has never been claimed to describe a set of events that take
place in real time, the input-to-output formalismtempts users to think of themin
that way. But children often delay in applying new rules to existing words. This
fact is strong evidence that they store (partial) articulatory specications of at
least the output forms that are not obeying the new rules in an output lexicon,
rather than completely creating them online by rules acting on stored input
forms. (If they arent doing that, they must be storing the information about
which rules apply to each word along with the word. We wont pursue that
descriptive option here.) U-shaped developmental curves and phonological
idioms also tell us that at least some output forms are established, and therefore
that they are stored in some sense (though not necessarily in every detail). So
the mappings from input to output dont need to act in real time or even to be in
the childs head, any more than rules relating Latin forms to their French
Challenges to theories, charges to a model 463
descendants do; they may simply be the observers account of correspondences.
But that doesnt seemto be the whole story, because some child phonology rules
do seem to work in real time. How can a rule be both online and ofine at the
same time?
Menn (1983) broke a childs internal rules into two types: selection rules and
articulatory rules. The articulatory rules (which were never fully described)
were invoked to specify automatic details of articulation that would apply to all
the childs words, such as nal devoicing, giving them their nal output forms;
they were considered to be online processes. The selection rules, on the other
hand, were used to describe howa child selects which items of information about
the adult word to preserve in her output form, and which information to abandon.
For example, if the target sh has output /ps/, the information that the word is a
monosyllable with onset obstruent labial, stressed lax high front vowel, and nal
obstruent sibilant appears to be preserved, while the information that the initial
consonant is a fricative and that the sibilant is post-alveolar has been abandoned.
When selection rules like these operate between two stored levels that is, when
the child is familiar with the target word and is well practiced at producing the
output form they are ofine descriptions of relationships.
But selection rules also sometimes act as real-time maps from the way a child
hears something to how she says it; even some rules that have exceptions seem
to work in real time. LMs two-and-a-half-year-old son Danny, hearing jeep for
the rst time, happily and reliably repeated it as /bip/; his rules evidently applied
online for dozens of new words over several months (LM could and did say to
her friends: Watch, hes going to say your name as . . .). When we observe a
child immediately picking up a new word and saying it condently according to
her established mapping patterns, it seems reasonable to suggest that those
patterns comprise a set of procedures operating in real time, procedures that
link how a word sounds to how it will be pronounced.
So whatever does the job of the old selection rules in a newmodel must explain
howsome mappings function both as an ofine representation of the information
about a word that a child preserves in output and as a real-time mapping that
connects what I hear to how Im going to move my articulators.
This dual role of rules (online/ofine) is not just a problem for child phonol-
ogy; its been around in generative phonological theory since Halles Sound
Pattern of Russian (1971) tried to minimize descriptive redundancy by collaps-
ing morphophonemic rules with allophonic ones the programcarried out more
fully for English, of course, by Chomsky and Halle (1968), and also envisioned
by constraint theories. But we (and I think usage-based theorists in general)
would argue that dual roles, and maybe dual representation for some kinds of
mapping patterns regardless of the formalism used to express them is not the
serious problem that it was thought to be fty years ago.
The classical goal of generative and constraint-based phonology is to describe
phonologies while minimizing the redundancy of the description. But by nowwe
know that the brain is full of redundant systems not just because evolution is
464 Lise Menn, Ellen Schmidt, and Brent Nicholas
inelegant, but because only redundancy can ensure reasonably reliable process-
ing under noisy conditions. Elegant, minimal systems are not neurologically
realistic. The Linked-Attractor model is unapologetically redundant in its attempt
to be psycholinguistically realistic, so it does not collapse online and ofine
processes that do the same thing just because its more elegant to state general-
izations only once. Being able to capture subtle patterns in the data is a vital
criterion for being a good theory; parsimony is nice, but it cant be allowed to
trump accuracy.
C. What does it mean to take a whole-word approach
to phonology?
As the termis used in this book, it has two components: 1. the exemplar claim:
learning a system starts from learning many individual examples, and 2. the
lexical claim: by about 8 months of age (and maybe earlier), the examples that
an infant learner takes as units are words (or word-like sequences), not sounds
or subword sound sequences.
Part of the evidence for word-based example-driven learning is statistical: the
sound patterns which children learn to recognize and produce are strongly
affected by the learning opportunities afforded by the words in the ambient
language not just what sounds occur and in what arrangements, but the
frequency with which a sound occurs, the probability that a sound will occur
in a particular phonological environment, and the number of other words with
highly similar sequences of sounds (neighbor words or simply neighbors).
For example, Zamuner, Gerken, and Hammond (2004) show that young chil-
dren can repeat the nal C of a CVC monosyllable more accurately if the initial
CVof the syllable is a common CV sequence rather than a relatively rare one
(controlling for the frequency of occurrence of the Cand the V) (see also Storkel
2001). This transition-probability effect is plausible if we regard learning
language as having a great deal in common with learning other things; for
example, it is parallel to the difference between repeating a phone number with
a familiar area code vs. a novel one if the area code is familiar, repeating the
rest of the digits becomes much easier. The reader will be able to think of many
parallel examples. (Thanks to Ronnie Silber for the analogy. For reviews,
arguments, and lists of citations see Edwards, Beckman, and Munson 2004,
Munson et al. 2012, and Stoel-Gammon 2011.)
Experimental studies of adult language support a whole-word view as well:
for example, Coleman and Pierrehumbert (1997) show whole-word effects in
adult judgments of how acceptable a nonword is (p. 8): When statistically
valid data on acceptability [of a pseudo-word] is gathered . . . it is found that
deviations [from phonotactic well-formedness] are partially redeemed by good
parts, and that forms which are locally well-formed, in the sense that each piece
is reasonably well-attested [for example, /sll/], can nonetheless be viewed as
Challenges to theories, charges to a model 465
improbable overall. They argue that neither classical generative phonology nor
Optimality Theory predict these results.
Generative and constraint-based grammars, as developed so far, are purely
structural: although phonotactic frequencies can be used in Optimality Theory
and its close relatives as evidence for constraint rankings, neither generative
nor optimality approaches allow linguistic phenomena to be sensitive to word
token or type frequency, which are whole-word phenomena in the second sense
above. However, token-frequency-dependent effects on how words are pro-
nounced are documented for adults. They are, in fact, the cornerstone of the
usage-based phonology literature (e.g., Bybee 2001). To take just two exam-
ples, the auxiliary verb can may be reduced to [kn
s]
vs. bunch362 > [bnts], lunch469 > [nts]
Challenges to theories, charges to a model 497
References
Au, T. K., Knightly, L. M., Jun, S.-A., and Oh, J. S. (2002). Overhearing a language
during childhood. Psychological Science, 13(3), 23843.
Beckman, M. E. and Edwards, J. (2000). The ontogeny of phonological categories and the
primacy of lexical learning in linguistic development. Child Development, 71, 2409.
Beckman, M. E., Munson, B., and Edwards, J. (2007). Vocabulary growth and devel-
opmental expansion of types of phonological knowledge. In J. Cole and J. I. Hualde
(eds.), Laboratory phonology 9, pp. 24164, Berlin: Mouton de Gruyter.
Bernhard, B. H. and Stemberger J. P. (1998). Handbook of phonological development.
San Diego: Academic Press.
Boyland, J. T. (2009). Usage-based models of language. In D. Eddington (ed.),
Experimental and quantitative linguistics, pp. 351419. Munich: Lincom.
Browman, C. P. (1978). Tip of the tongue and slip of the ear: a comparative study.
Journal of the Acoustical Society of America, 64(S1), S93.
Browman, C. P. and Goldstein, L. (1992). Articulatory phonology: an overview,
Phonetica, 49, 15580.
Brown, R. W. and Hildum, D. C. (1956). Expectancy and the perception of syllables.
Language, 32, 41119.
Bybee, J. (2001). Phonology and language use. Cambridge University Press.
(2006). Fromusage to grammar: the minds response to repetition. Language, 82, 71133.
(2010). Language, usage and cognition. Cambridge University Press.
Chomsky, N. and Halle, M. (1968). The sound pattern of English. New York: Harper &
Row.
Coleman, J. and Pierrehumbert, J. (1997). Stochastic phonological grammars and accept-
ability. 3rd Meeting of the ACL Special Interest Group in Computational
Phonology: Proceedings of the Workshop, 12 July 1997. Association for
Computational Linguistics, Somerset, NJ, pp. 4956.
Cutler, A. (2012). Native listening: language experience and the recognition of spoken
words. Cambridge, MA: MIT Press.
Davidson, L., Jusczyk, P., and Smolensky, P. (2006). Optimality in language acquisition I: the
initial and nal states of the phonological grammar. In P. Smolensky and G. Legendre
(eds.), The harmonic mind, vol 2, pp. 23178. Cambridge MA: MIT Press.
Edwards, J., Beckman, M. E., and Munson, B. (2004). The interaction between vocabu-
lary size and phonotactic probability effects on childrens production accuracy
and uency in nonword repetition. Journal of Speech, Language, and Hearing
Research, 47, 42136.
Ferguson, C. A. and Farwell C. B. (1975). Words and sounds in early language acquis-
ition. Language, 51, 419439. Reprinted in WilliamS-Y. Wang (ed.), The lexicon in
phonological change. The Hague: Mouton, 1977. Reprinted in this volume as
Chapter 4.
Fikkert, P. and Levelt, C. (2006). Howdoes Place fall into place? The lexicon and emergent
constraints in childrens developing phonological grammar. Unpublished MS.
Gerken, L. A. (2009). Acquiring linguistic structure. In E. Hoff and M. Shatz (eds.),
Handbook of language development, pp. 17390. New York: Blackwell.
Grunwell, P. (1981). The nature of phonological disability in children. London:
Academic Press.
Halle, M. (1971). The sound pattern of Russian: a linguistic and acoustical investigation.
Berlin: Walter de Gruyter.
Hebb, D. O. (1949). The organization of behavior. New York: John Wiley & Sons.
498 Lise Menn, Ellen Schmidt, and Brent Nicholas
Hume, E. (2008). Markedness and the language user. Phonological Studies, 11, 8398.
Ingram, D. (1974). Phonological rules in young children. Journal of Child Language, 1,
4964.
Johnson, K. (1997). Speech perception without speaker normalization. In K. Johnson
and J. W. Mullenix (eds.), Talker variability in speech processing, pp. 14565. San
Diego: Academic Press.
(2006). Resonance in an exemplar-based lexicon: the emergence of social identity and
phonology. Journal of Phonetics, 34, 48599.
Juszcyk, P. (1993). Fromgeneral to language-specic capacities: the WRAPSAmodel of
how speech perception develops. Journal of Phonetics, 21, 328.
1997. The discovery of spoken language. Cambridge, MA: MIT Press.
Kager, R. 1999. Optimality Theory. Cambridge University Press.
Kager, R., Pater, J., and Zonneveld W. (2004). Constraints in phonological acquisition.
Cambridge University Press.
Kaisse, E. and Shaw, P. A. (1985). On the theory of lexical phonology. Phonology
Yearbook I, 130.
Kiparsky, P. and Menn, L. (1977). On the acquisition of phonology. In John Macnamara (ed.),
Language learning and thought, pp. 4778. New York: Academic Press. Reprinted in
G. Ioup and S. H. Weinberger (eds.), Interlanguage phonology: the acquisition of a
second language sound system, pp. 2352. Cambridge, MA: Newbury House, 1987.
Knightly, L. M., Jun, S.-A., Oh, J. S., and Au, T. K.-f. (2003). Production benets of
childhood overhearing. Journal of the Acoustical Society of America, 114(1),
46574.
Kuhl, P. (1991). Human adults and human infants show a perceptual magnet effect for
the prototypes of speech categories, monkeys do not. Perception and Psychophysics,
20(2), 93107.
(2007). Is speech learning gated by the social brain? Developmental Science, 10,
11020.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., and Lindblom, B. (1992).
Linguistic experience alters phonetic perception in infants by 6 months of age.
Science, 255, 6068.
Labov, W. (1981). Resolving the neogrammarian controversy. Language, 57(2), 267308.
Labov, W. and Labov, T. (1978). The phonetics of cat and mama. Language, 54(4),
81652.
Leopold, W. F. (193949). Speech development of a bilingual child (4 vols.). Evanston,
IL: Northwestern University Press.
Legendre, G., Hagstrom P., Vainikka, A., and Todorovna, M. (2006). Optimality in
language acquisition II: inection in early French syntax. In P. Smolensky and
G. Legendre (eds.), The harmonic mind, vol. 2, pp. 276306. Cambridge, MA: MIT
Press.
Lindblom, B., Diehl R., Park, S.-H., and Salvi, G. (2011). Sound systems are shaped by
their users. In G. N. Clements and R. Ridouane (eds.), Where do phonological
features come from?. Amsterdam: John Benjamins. pp. 6797.
Macken, M. A. (1978). Permitted complexity in phonological development: one childs
acquisition of Spanish consonants. Lingua, 44, 21953.
(1979). Developmental reorganization of phonology: a hierarchy of basic units of
acquisition. Lingua, 49, 1149. Reprinted in this volume as Chapter 5.
Maye, J., Werker, J., and Gerken, L. (2002). Infant sensitivity to distributional informa-
tion can affect phonetic discrimination. Cognition, 82, 10111.
McCune, L. (This volume). Aview from developmental psychology.
Challenges to theories, charges to a model 499
McMurray, B., Cole, J., and Munson, C. (2011). Features as an emergent product of
perceptual parsing: evidence from V-to-V coarticulation. In G. N. Clements and
R. Ridouane (eds.), Where do phonological features come from? Cognitive,
physical and developmental bases of distinctive speech categories, pp. 197236.
Amsterdam: John Benjamins.
Mnard, L., Schwartz, J.-L., Bo, J.-L., Aubin, J. (2007). Articulatoryacoustic relation-
ships during vocal tract growth for French vowels: analysis of real data and
simulations with an articulatory model. Journal of Phonetics, 35, 119.
Menn, L. (1971). Phonotactic rules in beginning speech. Lingua, 26, 22541.
(1976). Pattern, control, and contrast in beginning speech: a case study in the develop-
ment of word form and word function. PhD dissertation, University of Illinois,
Urbana; published, Bloomington: Indiana University Linguistic Club, 1979.
(1983). Development of articulatory, phonetic and phonological capabilities. In
B. Butterworth (ed.), Language production, vol. 2, pp. 350. London: Academic
Press. Reprinted in this volume as Chapter 6.
Menn, L. and Matthei, E. (1992). The two-lexicon model of child phonology: Looking
back, looking ahead. In C. A. Ferguson, L. Menn, and C. Stoel-Gammon (eds.),
Phonological development: models, research, implications, pp. 21147. Timonium,
MD: York Press.
Menn, L., Schmidt, E., and Nicholas, B. (2009). Conspiracy and sabotage in the
acquisition of phonology: dense data undermine existing theories, provide scaffold-
ing for a new one. Language Sciences, 31 (26), 285304.
Menn, L. and Vihman, M. M. (2011). Features in child phonology: inherent, emergent,
or artefacts of analysis? In G. N. Clements and R. Ridouane (eds.), Where do
phonological features come from? pp. 261301. Amsterdam: John Benjamins.
Messum, P. (2005). Learning to talk: a non-imitative account of the replication of
phonetics by child learners. CamLing 2005, 99106.
Moskowitz, B. A. (1970). The two-year-old stage in the acquisition of phonology.
Language, 46, 42641.
Munson, B., Edwards, J., and Beckman, M. (2012). Phonological representations in
language acquisition: climbing the ladder of abstraction. In A. C. Cohn,
C. Fougeron, and M. K. Huffman (eds.), The Oxford handbook of laboratory phonol-
ogy, pp. 288309. Oxford University Press.
Oh, J. S., Jun, S.-A., Knightly, L. M., and Au, T. K. (2003). Holding on to childhood
language memory. Cognition, 86(3), B53B64.
Oliveira-Guimares, D. (This volume). Beyond early words: word template development
in Brazilian Portuguese.
Ota, M. (2006). Input frequency and word truncation in child Japanese: structural and
lexical effects. Language and Speech, 49, 26194.
(This volume). Lexical frequency effects on phonological development: the case of
word production in Japanese.
Ota, M. and Green, S. J. (2013). Input frequency and lexical variability in phonological
development: a survival analysis of word-initial cluster production. Journal of
Child Language, 40, 53966.
Peperkamp, S. (2003). Phonological acquisition: recent attainments and new challenges.
Language and Speech, 46, 87113.
Peters, A. M. and Menn, L. (1993). False starts and ller syllables: ways to learn
grammatical morphemes. Language, 69(4), 74277.
Pierrehumbert, J. (2002). Word-specic phonetics. In C. Gussenhoven and N. Warner
(eds.), Laboratory phonology 7, pp. 10139. Berlin: Mouton de Gruyter.
500 Lise Menn, Ellen Schmidt, and Brent Nicholas
(2003). Phonetic diversity, statistical learning and acquisition of phonology. Language
and Speech, 46, 11554.
Priestly, T. M. S. (1977). One idiosyncratic stategy in the acquisition of phonology.
Journal of Child Language, 4, 4566. Reprinted in this volume as Chapter 7.
Pye, C., Ingram, D., and List, H. (1987). Acomparison of initial consonant acquisition in
English and Quich. In K. E. Nelson and A. Van Kleeck (eds.), Childrens language,
vol. 6, pp. 17590. Hillsdale, NJ: Lawrence Erlbaum.
Rice, K. (2007). Markedness in phonology. In Paul de Lacy (ed.), Cambridge handbook
of phonology, pp. 7998. Cambridge University Press.
Roy, D. (2011). www.ted.com/talks/deb_roy_the_birth_of_a_word.html.
Saffran, J. R., Aslin, R. N., and Newport, E. L. (1996). Statistical learning by 8-month-
old infants. Science, 274, 268.
Saffran, J. R. and Thiessen, E. D. (2003). Pattern induction by infant language learners.
Developmental Psychology, 39, 48494.
Smith, N. V. (1973). The acquisition of phonology: a case study. Cambridge University
Press.
Smolensky, P. and Legendre G. (2006a). Harmony optimization and the computational
architecture of the mind/brain. In P. Smolensky and G. Legendre (eds.), The
harmonic mind, vol. 1, pp. 362. Cambridge, MA: MIT Press.
(2006b). Principles of the integrated connectionist-symbolic cognitive architecture. In
P. Smolensky and G. Legendre (eds.), The harmonic mind, vol.1, pp. 63344.
Cambridge, MA: MIT Press.
Szreder, M. (This volume). The acquisition of consonant clusters in Polish: a case study.
Stoel-Gammon, C. (1989). Prespeech and early speech development of two late talkers.
First Language, 9, 20724.
(2011). Relationships between lexical and phonological development in young chil-
dren. Journal of Child Language, 38, 134.
Stoel-Gammon, C. and Dunn C. (1985). Normal and disordered phonology in children.
Austin, TX: Pro-Ed.
Storkel, H. (2001). Learning new words: phonotactic probability in language develop-
ment. Journal of Speech, Language and Hearing Research, 44(6), 132137.
Swinney, D. (1979). Lexical access during sentence comprehension: (re)consideration of
context effects. Journal of Verbal Learning and Verbal Behavior, 18, 64559.
Tokowicz, N. and MacWhinney, B. (2005). Implicit and explicit measures of sensitivity
to violations in second language grammar: an event-related potential investigation.
Studies in Second Language Acquisition, 27, 173204.
Velleman, S. L. and Vihman, M. M. (2006). Phonological development in infancy and
early childhood: implications for theories of language learning. In M. C. Pennington
(ed.), Phonology in context, pp. 2550. Luton: Macmillan.
Vihman, M. M. (1981). Phonology and the development of the lexicon: evidence from
childrens errors. Journal of Child Language, 8, 23964.
(1982). A note on childrens lexical representations. Journal of Child Language, 9,
24953.
(1996). Phonological development. Cambridge, MA: Blackwell.
Vihman, M. M. and Croft, W. (2007). Phonological development: toward a radical
templatic phonology. Linguistics, 45, 683725. Reprinted in this volume as
Chapter 2.
Vihman, M. M., DePaolis, R. A., and Keren-Portnoy, T. (2009). A Dynamic Systems
approach to babbling and words. In E. Bavin (ed.), Handbook of child language,
pp. 16382. Cambridge University Press.
Challenges to theories, charges to a model 501
Vihman, M. M., Kay, E., Boysson-Bardies, B., Durand, C., and Sundberg, U. (1994).
External sources of individual differences? A cross-linguistic analysis of the
phonetics of mothers speech to one-year-old children. Developmental
Psychology, 30(5), 65263.
Vihman, M. M., Velleman, S. L., and McCune, L. (1994). How abstract is child phonol-
ogy? Towards an integration of linguistic and psychological approaches, In
M. Yavas (ed.), First and second language phonology, pp. 944. San Diego:
Singular Publishing. Reprinted in this volume as Chapter 9.
Vihman, M. M. and Velleman, S. L. (1989). Phonological reorganization: a case study.
Language and Speech, 32, 14970. Reprinted in this volume as Chapter 8.
(2000). The construction of a rst phonology. Phonetica, 57, 25566.
Wauquier, S. and Yamaguchi, N. (This volume). Templates in French.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211. Reprinted in this volume as Chapter 3.
(1976). Perception and production in the acquisition of phonology. In W. von Rafer
Engel and Y. Lebrun (eds.), Baby talk and infant speech, pp. 294322. Amsterdam:
Swets & Zeitlinger. Reprinted in N. Waterson. Prosodic phonology: the theory and
its application to language acquisition and speech processing, pp. 678. Newcastle
upon Tyne: Grevatt & Grevatt, 1987.
Zamuner, T. S., Gerken, L., and Hammond, M. (2004). Phonotactic probabilities in
young childrens speech production. Journal of Child Language, 31, 51536.
502 Lise Menn, Ellen Schmidt, and Brent Nicholas
References for reprinted papers
Ferguson, C. A. and Farwell, C. B. (1975). Words and sounds in early language acquis-
ition. Language, 51, 41939. Reprinted in W. S-Y. Wang, The lexicon in phono-
logical change. The Hague: Mouton, 1977.
Macken, M. A. (1979). Developmental reorganization of phonology: a hierarchy of basic
units of acquisition. Lingua, 49, 1149.
Menn, L. (1983). Development of articulatory, phonetic, and phonological capabilities.
In B. Butterworth (ed.), Language production, vol. 2, pp. 350. London: Academic
Press.
Priestly, T. M. S. (1977). One idiosyncratic strategy in the acquisition of phonology.
Journal of Child Language, 4, 4566.
Savinainen-Makkonen, T. (2007). Geminate template: a model for rst Finnish words.
First Language, 27, 34759.
Vihman, M. M. and Croft, W. (2007). Phonological development: toward a radical
templatic phonology. Linguistics, 45, 683725.
Vihman, M. M. and Velleman, S. (1989). Phonological reorganization: a case study.
Language and Speech, 32, 14970.
Vihman, M. M., Velleman, S. L., and McCune, L. (1994). How abstract is child phonol-
ogy? Towards an integration of linguistic and psychological approaches. In
M. Yavas (ed.), First and second language phonology, pp. 944. San Diego:
Singular Publishing.
Waterson, N. (1971). Child phonology: a prosodic view. Journal of Linguistics, 7,
179211.
503
Index
Abdoh, 380, 404
Abondolo, 45, 51
abstract, 7, 17, 22, 414, 468, 88, 156, 169,
266, 280, 294, 325, 343, 446, 375, 446, 451,
462, 46970, 475, 496
abstraction, 4789, 495
accent, 320, 332
accentual arc, 325, 331, 333, 338
accentual pattern, 36, 338
accuracy, 2, 8, 24, 96, 1356, 143, 2001,
2089, 262, 269, 271, 2834, 291, 297, 299,
344, 3524, 357, 366, 380, 4035, 407, 417,
421, 424, 426, 4301, 442, 465
accurate, 3, 8, 22, 24, 26, 28, 96, 103, 1068,
158, 160, 174, 195, 197, 203, 244, 269, 280,
291, 296, 303, 312, 374, 376, 380, 384, 395,
398, 399, 4057, 416, 426, 428, 452, 4878
acoustic analysis, 238, 241, 256, 294, 296
acoustic signal, 6
across-the-board, 8, 239, 415, 431, 482, 496
activation, 484, 4912
adapt, 289, 405, 488
adaptation/s, 305, 329, 334, 3745, 3956,
3989, 403, 406
adapted, 23, 5, 23, 26, 29, 32, 36, 41, 244,
2967, 303, 305, 307, 327, 339, 351, 365,
376, 391, 3956, 3989, 404, 406, 442
adapting, 25
Adda, 319, 339
Adda-Decker, 319, 339
adult phonological representation, 312
adult phonology, 17, 21, 412, 44, 106, 110,
137, 140, 144, 207, 2678, 378, 381, 405,
4612, 482
affricate/s, 353, 356
affrication, 6578, 812, 352, 355
Akan, 43
Alarcos Llorach, 137, 165
Alcantara, 12
Ali, 380
Allen, 322, 339, 362, 36970, 418, 426, 432
allomorphy, 2056
allophonic rules, 205, 207
Almeida, 310, 313
Al-Tamimi, viii, ix, 3, 10, 344, 376, 378, 4089
alternation, 480
alveolar closure, 114
alveolarity, 6578
Amayreh, 380, 405, 408
ambient language/s, 1, 4, 235, 35, 401,
2602, 313, 31920, 375, 41617, 44950,
4534, 4656, 473, 4823, 487, 494
Ammar, 380, 381, 404, 408
analogy, 2, 1434
anchor syllable, 394
Anderson, 7, 11, 290, 456
Aoyama, 364, 370
Arabic, 3, 10, 3445, 37685, 390, 3945, 398,
403, 4057, 483
articulation, 3, 65, 72, 7783, 87, 89, 112, 138,
153, 1701, 178, 183, 186, 1956, 2089,
227, 252, 26970, 2801, 300, 313, 3456,
3534, 35660, 387, 460, 467, 476, 496
articulators, 195
articulatory challenge, 301, 311
articulatory control, 168, 260, 405, 407,
468, 473
articulatory difculty, 487
articulatory effort, 224
articulatory factors, 227
articulatory gestures, 46, 49, 186, 277, 343,
452, 470
articulatory habits, 204
articulatory motor programming, 169
articulatory patterns, 267, 360
Articulatory Phonology, 49, 3435, 479
articulatory program, 1978
articulatory representation, 4923
articulatory routine, 242
articulatory score, 49
articulatory skills, 346
Aslin, 6, 11, 13, 260, 286, 475, 501
assimilation/s, 19, 22, 80, 84, 86, 978, 112,
138, 145, 155, 157, 175, 1802, 189, 1912,
197, 207, 304, 352, 3546, 362, 366, 368,
412, 415, 462, 481, 491
504
attention, 23, 38, 196, 203, 245, 260, 271, 281,
284, 345, 357, 360, 363, 3745, 406
attraction, 494
attractor/s, 38, 472, 4747, 481, 4846, 489,
491, 4945
Au, 471, 498500
Aubin, 500
auditory coding, 6
auditory input representation, 496
auditory memory, 201, 476
auditory representation, 461, 488, 4924, 496
autosegmental, 8, 277, 285, 337
Autosegmental and Metrical Phonology, 8,
337, 462
avoidance, 108, 11011, 169, 178, 209, 224,
329, 334, 488, 493, 497
awareness, 169, 206, 238, 242, 245
babble/babbling, 2, 45, 24, 32, 40, 47, 95, 139,
16972, 204, 240, 259, 2612, 267, 270,
2778, 2835, 291, 310, 318, 320, 326, 374,
382, 384, 417, 441, 444, 452, 4767, 482,
494
baby talk, 163, 199, 381
backness, 6578, 86
Bailey, 48, 51
Baillargeon, 263, 286
Barton, 1756, 182, 2012, 210, 212, 239, 320,
339
basic unit, 47, 49, 133, 140, 161, 232,
338, 343
Bassano, 319, 335, 339
Bates, 240, 254, 262, 264, 286, 364, 370,
444, 456
Bauer, 24, 53, 266, 287
Beckman, 67, 1112, 22, 47, 512, 55, 417,
4323, 461, 465, 470, 498, 500
Beebe, 213
BEFs, see bisyllabic experimental forms
Bell, 181, 210
Bellugi, 61, 89, 91
Bengali, 3940
Benigni, 240, 254, 262, 286, 444, 456
Bennett, 213
Berg, 23, 51, 266, 284, 286, 415, 433
Berko, 114, 199, 206, 210, 212, 218, 233
Berko Gleason, 114, 199, 206
Berman, 25, 51
Bernhardt, 362, 370, 463, 483
Bertoncini, 12
Bhaya Nair, 39, 51, 376, 405, 408
bias, 176, 187, 202, 2689, 282
Bijeljac-Babic, 11
bilabial closure, 114
bilabiality, 6578
bilingual, 319, 334
binarity, 48
binary, 48, 3212, 324, 335, 338
binary foot, 3213, 329, 334, 338, 380
Binkofski, 456
Bisol, 293, 313
bisyllabic experimental forms (BEFs), 3,
21732, 237
bisyllabic ordinary forms (BOFs), 2, 217,
2245, 234
Blasdell, 231, 233
Bleile, 2389, 254
Bloom, 135, 165, 265, 286, 4434, 456
Bloomeld, 103, 114
Bo, 500
Boersma, 8, 11, 382, 408
BOFs, see bisyllabic ordinary forms
boldness, 284
Bond, 201, 211
Bonilha, 313
Bosch, 5, 12
Boudelaa, 377, 408
Boula de Mareil, 339
Bowen, 137, 166
Bower, 162, 165
Bowerman, 24, 51, 183, 239, 2534
Bowman, 55
Boyland, 460, 498
Boysson-Bardies, 5, 11, 24, 51, 56, 261, 2856,
320, 340, 417, 4325, 502
Braine, 2534
Branigan, 171, 210, 283, 286
Braud, 321, 326, 332, 3379
Brazilian Portuguese, 3, 2915, 298,
31011; see also Portuguese
Bretherton, 240, 254, 262, 286, 456
Broselow, 378, 403, 408
Browman, 7, 11, 49, 51, 343, 360, 472,
479, 498
Brown, 45, 11, 61, 879, 91, 472, 498
Brulard, 41, 52, 319, 334, 339
Bruskin, 264, 288
Buccino, 451, 456, 458
Buchtal, 454, 456
Buckley, 322, 324, 340
Bush, 94, 115, 134, 165, 241, 254
Butterworth, 210
Bybee, 18, 44, 479, 52, 31213, 460,
466, 498
Calvo, 52
Camaioni, 240, 254, 262, 286, 444, 456
Camarda, 458
Campos, 263, 289, 448, 458
canonical form/s, 105, 13940, 169, 192,
194200, 208, 463, 476, 481, 488
Cappa, 458
Carr, 41, 52, 319, 334, 339
Carroll, 61, 91
Carubbi, 52
Carvalho, 319, 339
Index 505
categorization, 7
category, 67, 426, 4950, 88, 354, 376
centrality, 6578
challenge, 23, 301, 310, 333, 346, 367, 408
Chambers, 340
Champion, 416, 433
Charette, 324, 339
Charles-Luce, 22, 56
Chen, C.-C., 416, 435
Chen, M., 114
Chiat, 23, 52, 266, 286
Childers, 363, 370
Chin, 362, 370
Choi, 24, 51
Chomsky, 7, 1011, 901, 113, 115, 343, 360,
4456, 456, 462, 464, 474, 498
Chomsky, 10
Church, 340
Clark, E. V., 2534
Clark, H. H., 17, 52
Clements, 7, 11
closed syllable, 297
closeness, 67
Clumeck, 172, 210
cluster insertion, 3512
clusters, 10, 44, 108, 148, 154, 157, 159, 179,
181, 193, 196, 199200, 207, 2312, 268,
294, 318, 334, 34452, 354, 3567, 35960,
362, 365, 369, 376, 381, 387, 408, 412,
41617, 431, 463, 467, 473, 479, 487,
490; see also consonant clusters
coda/s, 2934, 298301, 310, 319, 334, 3778,
3801, 390, 3945, 4045, 407, 417
code-switching, 384
cognitive development, 4, 87, 194, 204
cognitive load, 202
cognitive style, 1645
Cohen, 61, 91
coincidences, 219, 228
Cole, 460
Coleman, 8, 11, 465, 472, 498
combination, 307, 455
communicative grunts, 4535
competence, 1819, 80, 889, 113
competition, 34, 74, 812, 151, 155, 299,
31012, 480, 4912
complex realizations, 387
complex targets, 387
confusion, 148, 151, 156
connectionist model, 284
conscious, 4478
consciousness, 9, 202, 263, 446, 455
consolidation, 136, 189, 192, 312
consonant cluster, 40, 139, 144, 148, 158, 172,
199, 354, 357, 431, 481; see also clusters
consonant sequences, 367
consonant-nal words, 245, 248, 256
conspiracy, 1812, 184
constraint/s, 45, 78, 10, 24, 26, 289, 32,
356, 434, 48, 112, 13840, 153, 158,
1612, 1815, 193, 195, 239, 260, 267, 275,
31720, 3256, 330, 3339, 3435, 3567,
367, 406, 426, 4607, 4705, 4803, 4935
context-dependent words, 453
context-limited words, 2645, 285, 448
continuance, 6578, 813, 857
continuity, 7, 18, 31718, 326, 444, 4534
contrast, 21, 412, 99, 103, 1056, 108,
11013, 133, 13940, 143, 146, 148, 153,
1559, 163, 168, 176, 1823, 192, 208, 232,
259, 2716, 283, 2856, 311, 364, 431
co-occurrence, 13940, 144, 1589, 1623,
194, 367
Cooper, 24, 55, 238, 255, 291, 314, 362, 372
Corbetta, 452, 458
Corcoran, 57
core syllable, 365, 376, 380
coronal, 3526
correspondence, 98, 138, 140, 153, 208, 2202,
227, 229, 265, 426, 495
corresponding phones, 98
counter-stress, 3245, 332, 336, 375
Craighero, 448, 458
creativity, 144
Cristfaro-Silva, 292, 313
Croft, vii, ix, 1, 810, 13, 1718, 41, 4950, 52,
291, 315, 317, 342, 345, 361, 3745, 410,
4412, 459, 461, 467, 4702, 475, 481, 489,
501, 503
cross-language variation, 490
cross-linguistic, 5, 8, 23, 318, 374, 377, 442, 460
cross-modal, 477, 495
cross-modal mappings, 485
Crowhurst, 363, 370
Cruttenden, 105, 115, 162, 165
Cutler, 341, 363, 370, 478
CV phonology, 7
CV syllable, 310, 3289, 334, 338, 364, 404
CVCV phonology, 8
DOdorico, 20, 52
Dahalo, 42
Dalbour, 137, 165
Dale, 364, 370, 4323
Danna, 458
Davidson, 466, 472, 498
Davis, 24, 32, 52, 56, 260, 270, 286, 288,
294, 314
Dean, 362, 372
deautosegmentalization, 268
Declarative Phonology, 8
Del Giudice, 448, 456
Delery, 453, 457
deletion, 22, 97, 114, 1389, 147, 1512, 155,
1579, 17985, 307, 330, 347, 362, 376, 381,
4045, 462, 497
506 Index
Dell, 322, 324, 339
Demolin, 294, 313
Demuth, 322, 324, 335, 33940, 365, 370, 376,
409, 418, 433
den Os, 363, 372, 435
dental, 353
DePaolis, 6, 10, 1213, 20, 24, 52, 56, 342, 344,
361, 416, 433, 452, 4567
Dependency Phonology, 8
devoicing, 230
Di Cristo, 322, 3245, 331, 333, 338, 340
di Pellegrino, 448, 456
diary, 34, 1920, 401, 168, 294, 364, 4489,
480, 493
Diehl, 499
differentiation, 878, 168, 449
dimensions, 4856, 494
diminutive/s, 32, 332, 349
Dinnsen, 362, 370
discontinuity, 1819, 24
dissimilation, 80
distinctive features, 93, 111
distributional learning, 2
domain edge, 324
dominance, 3, 244, 248, 279
Donahue, 20, 52, 4556
dorsal, 354, 356
Dos Santos, 322, 340
Drachman, 207, 210, 224, 233
Druss, 340
Dunn, 501
Durand, C., 5, 11, 51, 56, 286, 340, 432, 435,
502
Durand, J., 8, 11
Dutch, 20, 319, 322, 334, 365, 417, 426, 483
dynamic systems, 441, 445, 451, 453, 455
Dyson, 380, 405, 408
ease of articulation, 169, 180, 185, 224
Echols, 320, 340, 363, 370, 418, 426, 433
Edwards, J., 67, 1112, 22, 478, 512, 417,
4323, 461, 465, 470, 497, 498, 5001
Edwards, M. L., 967, 115, 165, 218, 226, 233,
241, 254, 362, 370, 416, 435
effort, 80, 224
Egyptian, 380, 381
Eimas, 1, 11
Elbers, 20, 24, 52, 279, 286, 319, 340
Elbert, 13, 240, 255, 369, 372, 384, 410
elicitation, 218
elicited forms, 217
elision, 80
Elsen, 20, 267, 52
embodied cognition, 447
emergence of the unmarked, 311
emergent, 2, 5, 40, 47, 239, 25960, 268, 271,
273, 276, 27980, 282, 285, 320
emergentist, 1, 460
English, 3, 10, 201, 25, 2930, 32, 3544,
845, 92, 945, 1345, 139, 1589, 168,
1767, 1823, 1878, 190, 199, 2067, 218,
224, 226, 231, 238, 31823, 334, 355, 3626,
369, 3756, 3815, 3945, 3989, 4034,
406, 41719, 426, 454, 464, 467, 478, 4823,
486, 488, 490
Enright, 262, 289
entrenched, 312, 4735, 484
entrenched forms, 306
entrenchment, 32, 35, 312, 475, 484
epenthesis, 22
Ervin-Tripp, 231, 233
Esling, 454, 456
Estonian, 3, 20, 256, 2933, 356, 40, 55, 184,
31819, 334, 375, 377, 379, 483
Ewen, 434, 489, 52
exceptions, 32, 1578, 18994, 231, 281,
4234, 4634, 467, 474, 494
exemplar/s, 2, 69, 17, 41, 47, 50, 465, 46970,
487, 4912, 495
exemplar models, 490
experimental forms, 223, 228, 231; see also BEFs
experimentation, 152, 189, 2389, 242, 244,
24851, 271, 281, 390
extra-systemic, 106, 259, 269, 271, 274, 281, 284
Faaborg-Anderson, 454, 456
Fadiga, 447, 456, 458
Fagen, 262, 289
faithfulness, 494
faithfulness constraints, 473, 494
familiarity, 21, 29, 40, 202, 224, 454
Farwell, ix, 1, 7, 1011, 212, 24, 28, 47, 52,
93, 140, 162, 1645, 1778, 21011,
21718, 2245, 233, 238, 254, 259, 262, 266,
286, 291, 31113, 317, 340, 343, 360, 362,
372, 375, 409, 41516, 433, 4423, 450, 456,
468, 4845, 498, 503
favorite sounds, 178
Fazio, 458
featural, 7
feature geometry, 271, 276
Fee, 267, 286, 322, 339, 365, 370, 376, 409
feedback, 202, 4923
feet, 48, 97, 119, 126, 130, 3224, 338,
381; see also foot
Fennell, 57
Fenson, 364, 370
Ferguson, ix, 1, 78, 1011, 13, 212, 24, 28,
47, 52, 93, 104, 10810, 113, 115, 140, 159,
162, 1645, 172, 1778, 183, 211, 214,
21718, 2245, 233, 23840, 2545,
25960, 262, 266, 286, 288, 291, 31113,
317, 340, 343, 356, 360, 362, 369, 372, 375,
384, 40910, 41516, 433, 4423, 450, 452,
456, 468, 4845, 498, 500, 503
Fernald, 260, 287
Index 507
Fey, 183, 211, 2534
Fikkert, 8, 11, 2668, 287, 319, 3212, 333,
340, 343, 360, 363, 372, 376, 409, 418, 426,
433, 476, 498
ller, 135, 333
ller syllable/s, 404
lter, 40, 239, 260, 334
nal consonant/s, 445, 143, 157, 179, 2401,
245, 248, 250, 252
nal segments, 253
Finnish, 10, 20, 25, 36, 38, 40, 318, 3445, 357,
3639, 376, 377, 405, 4667, 483
rst word, 45
Firth, 49, 52
Firthian, 7
Fisher, C., 320, 340
Fisher, H., 226, 234
Flege, 182, 211
Fodor, 91
Fogassi, 447, 456, 458
Folger, 178, 211, 284, 287
Fonagy, 322, 3245, 340
foot, 246, 258, 303, 315, 3214, 328, 333, 335,
338, 381, 424, 437, 451, 485; see also feet
forcefully articulated, 88
fossilized forms, 305
Fougeron, 324, 341
Francescato, 19, 53, 98, 115, 317, 340
Fraser, 91
Frdonie, 12
Freitas, G. C. M., 313
Freitas, M. J., 340
French, 4, 20, 25, 36, 38, 401, 182, 283, 293,
31739, 364, 375, 378, 3816, 3945, 397,
399, 4034, 4067, 417, 463, 466, 483
French, A., 36, 53
frequency, 47, 31012, 376, 378, 390, 407,
41617, 421, 430, 460, 478, 4867
frequency effects, 466
frequency of occurrence, 21, 377, 398, 465, 487
frication, 148
fricative/s, 85, 353, 356
friction, 6578, 812, 856
Frisch, 48, 53
Fromkin, 152, 166
fronting, 190
frontness, 6578, 856
frozen forms, 153, 158, 163
Fry, 61, 63, 87, 91
functionalist, 9
functionality, 31920
Gallese, 447, 456, 458
Gamkrelidze, 287
Gandour, 183, 211, 2534
Garlock, 4323
Garnica, 967, 109, 115, 162, 165, 218, 233,
260, 287
geminate/s, 10, 36, 38, 40, 344, 3634, 3669,
376, 378, 381, 385, 387, 389, 3959, 4047,
419, 483, 487
geminate template, 405
gemination, 375
generalization, 312
generative, 8
Generative Phonology, 7, 50, 62, 462, 464,
466, 474
generative rules, 462
Gentilucci, 458
Gerken, 6, 11, 320, 322, 340, 363, 372, 417,
435, 465, 475, 483, 498, 499, 502
German, 20, 21, 26, 27, 40, 84, 85, 95, 182, 206,
323, 334, 483
Gierut, 416, 4324
Gimson, 91
Gleason see Berko Gleason
glide/s, 353, 354, 356
gliding, 355
glottality, 6578
Gnanadesikan, 311, 313, 343, 360
Goad, 285, 287, 322, 324, 340
Goffman, E., 454, 456
Goffman, L., 6, 11
Gogate, 453, 457
Goldeld, 260, 287
Goldinger, 47, 53
Goldsmith, 78, 11, 49, 53, 268, 287
Goldstein, L., 7, 11, 49, 51, 211, 343, 360, 479,
498
Goldstein, U., 186
Goodell, 23, 55, 343, 357, 361
Goodman, 4323
Government Phonology, 8
Grammont, 217, 233
Greek, 20, 207
Green, 324, 417, 431, 434, 468, 500
Greenberg, 110, 162, 166
Greenlee, 2401, 255
Grgoire, 61, 91
Grijzenhout, 340
Grunwell, 22, 53, 238, 254, 355, 360, 362, 365,
372, 483, 498
Guarani, 280
Gussman, 380, 409
Hagstrom, 499
Hahn, 48, 51
Halle, 7, 1011, 612, 91, 343, 360, 462, 464,
474, 498
Hall, 56, 11, 13, 51, 261, 286, 320,
340, 342
Halliday, 171, 172, 211
Hammond, 417, 435, 465, 502
Hamp, 199, 211
Harmonic Grammar, 462, 473, 494
harmonized, 334
508 Index
harmonizing, 29
harmony, 20, 22, 278, 303, 36, 389, 43,
1389, 147, 152, 161, 1801, 191, 193, 195,
197, 266, 267, 271, 274, 276, 283, 3045,
308, 318, 329, 348, 353, 359, 376, 394, 395,
4067, 412, 497
Harris, J., 380, 409
Harris, Z. S., 195, 214
Haselkorn, 172, 212
Haspelmath, 50, 53
Hawkins, 174, 207, 211, 322, 339, 362, 36970,
418, 426, 432
Hayes, 322, 340, 378, 380, 409
Hebb, 470, 498
Hebrew, 20, 25
Hegazi, 380, 410
Herold, 11
hiatus, 230, 232
Hickok, 447, 450, 457
high frequency, 47, 156, 252, 274, 294, 319,
334, 378, 387, 395, 41617
Hijazi, 380
Hildum, 472, 498
Hindi, 20, 3840, 345, 3767
Hirsh-Pasek, 320, 3401
Hoaglin, 249, 255
Hochberg, 240, 255
Hodson, 417, 433
Hogan, 363, 370, 415, 4323
Hhle, 5, 11
holistic, 6, 24, 137, 207, 259, 292, 3078, 311,
320
Holmes, 418, 433
homonyms, 61, 678, 75, 95, 224, 227, 467
Hood, 165
Hooper, 164, 166
Hooper-Bybee, see Bybee
household word, 219
Howell, 362, 372
Hsieh, 104, 115, 224, 233, 416, 433
Hulst, 523
Hume, 466, 485, 499
Hunt, 263, 289
Iakimova, 12
iambic, 38, 175, 3204, 333, 3378, 378, 3801
iambic foot, 322, 324, 3289, 333, 335,
338; see also foot/feet
idiomatic, 219, 225, 228, 473, 484
idiosyncratic, 2, 111, 113, 1767, 185, 189, 217,
220, 2246, 2389, 259, 268, 312, 3267,
329, 337, 441
imitation/s, 25, 63, 78, 96, 11422, 1356, 144,
1501, 1645, 2012, 204, 218, 245, 270,
283, 285, 363, 391, 3956, 399, 471, 476,
4923
implicit, 2, 5, 24, 401, 47, 232
independent system, 10, 612
individual, 477, 479
individual differences, 4, 8, 71, 10910, 133,
1601, 1634, 168, 260, 336, 374, 390, 403,
4067, 448, 482
induction, 40
information-processing model, 163
Ingram, D., 105, 108, 115, 138, 143, 162, 166,
174, 180, 182, 184, 187, 194, 211, 217, 227,
233, 238, 254, 362, 365, 370, 372, 443, 457,
4623, 483, 499, 501
Ingram, T. T. S., 61, 91
Inhelder, 451, 458
inhibition, 492
initiation-of-speech phenomenon, 135
innate, 8, 19, 40, 263, 268, 485
innatist, 263
input attractor/s, 491, 495
input frequency, 32, 31011, 364, 41618, 421,
426, 4302, 467, 478, 4889
input lexicon, 1979
input representation/s, 461, 471, 479, 4889,
492
input templates, 472, 4745
inputoutput mapping, 475, 477, 492
inputoutput relations, 463, 473
instability, 480
inter-modal representations, 484
internal negative feedback, 4923
internal representation, 263, 265, 271, 2845
Irwin, 94, 115
Isako, 42
Italian, 20, 318
item, 177, 219, 407
item-based phonology, 259
item-learning, 405
Itkonen, 211
Iverson, 268, 287
Jaeger, 32, 35, 53, 2067, 211
Jaffe, 213
Jakobson, 4, 12, 21, 24, 53, 612, 79, 89, 91, 98,
104, 106, 10810, 11415, 162, 166, 170,
178, 187, 193, 211, 224, 232, 234, 362, 370,
441, 457
Japanese, 20, 364, 41720, 424, 428, 432, 487,
490
jargon, 170, 384, 396
Jeffrey, 163, 165, 167
Jenkins, 162
Jensen, 231, 233
Johnson, C. E., 94, 115
Johnson, E. K., 1, 12, 341
Johnson, J. S., 370, 415, 418, 4324
Johnson, K., 460, 470, 472, 499
Johnson M., 447, 457
Jones, D., 65, 77, 91
Jones, L. G., 171, 211
Jordanian Arabic, 380
Index 509
Jun, 324, 341, 498, 499, 500
Jusczyk, 1, 11, 12, 40, 53, 238, 254, 320, 3401,
363, 370, 466, 472, 475, 483, 4989
Kager, 8, 12, 462, 499
Kaisse, 474, 499
Kaplan, 266, 290, 443, 445, 44850, 459
Karlsson, 363, 365, 369, 370
Katz, 901
Kavanagh, 113, 115, 214
Kay, 56, 432, 435, 502
Kay Elemetrics Corp., 249
Kaye, 8, 12
Keating, 182, 214
Kehoe, 363, 370, 418, 433
Keller, 17, 53
Kemler-Nelson, 3401
Kemmerer, 22, 56
Kennedy, 3401
Kenstowicz, 50, 53
Kent, 24, 53, 201, 211, 260, 266, 287, 447,
457, 480
Keren-Portnoy, ix, 2, 5, 10, 1213, 291, 313,
318, 341, 416, 430, 433, 452, 4567
Kessler, 55
Keyser, 8, 11
Keysers, 448, 456
Khanty, 456
Khattab, ix, 3, 10, 344, 376, 378, 4089
Kinney, 314
Kiparsky, 144, 160, 166, 262, 287, 461, 471,
496, 499
Kirchner, 454, 457
Kisseberth, 181, 211
Kiterman, 231, 234
Knewasser, 55
Knightly, 471, 498500
Koenigsknecht, 239, 256
Krgvee, 20, 26, 53
Korte, 201, 211
Kresheck, 226, 234
Krikhaar, 363, 372, 426, 435
Kubozono, 424, 433
Kuhl, 1, 12, 472, 475, 486, 4989
Kunnari, 10, 13, 20, 53, 291, 315, 318, 342,
36371
labial/s, 68, 3526
labialalveolar, 279
labiality, 6578, 813, 148, 151, 270, 367
labiodentality, 6578
Labov, T., 23, 54, 480, 499
Labov, W., 23, 42, 50, 534, 11415, 148, 166,
480, 499
laboratory phonology, 460
Lacerda, 499
Ladd, 55
Ladefoged, 412, 54, 61, 7980, 91, 2334
Lahey, 443, 456
Lamel, 339
Lamprecht, 294, 313
Landberg, 261, 289
Large, 53
lateral release, 701
laxness, 83
Lebanese, 374, 3789, 381, 399, 405, 407
Lebanese Arabic, 379
Legendre, 466, 495, 498, 501
lemma, 471
length, 3746, 378, 38990, 394, 404, 406
Lenneberg, 61, 901
Leonard, 178, 211, 213, 238, 254, 284, 287,
289, 416, 433
Leopold, 201, 54, 612, 79, 80, 846,
8997, 103, 11011, 11415, 170, 193, 202,
211, 239, 254, 269, 448, 457, 463,
481, 499
Lepage, 448, 457
Leslie, 263, 287
Levelt, C. C., 8, 11, 340, 343, 360, 417, 433,
476, 498
Levelt, W. J., 417
Lewis, C., 284, 288
Lewis, L. B., 97, 363, 370, 415, 4323
Lewis, M. M., 612, 80, 84, 92, 260, 284, 287,
448, 457
lexeme, 471
lexical accent, 322
lexical contrast, 109, 111
lexical diffusion, 312, 416, 418, 4312, 482
lexical entries, 476
lexical frequency, viii, 48, 41518, 421,
42632, 468, 487; see also frequency
lexical identity effects, 467, 468
lexical item/s, 294, 390, 416, 427, 487
lexical parameter, 1034, 108, 114, 224
Lexical Phonology, 8, 474
lexical smoothing, 1903
lexical stress, 3223
lexical unit, 56, 21, 318, 377
lexical variation, 8, 185, 415, 418, 482, 490
lexicon, 7, 20, 24, 32, 38, 40, 46, 47, 80, 88, 93,
104, 111, 1737, 198, 199, 23840, 279, 284,
285, 295, 31718, 330, 3645, 369, 3745,
377, 407, 416, 429, 4312, 466, 470, 476,
480, 482
Li, 432, 433
liaison, 323, 334, 339
Lieven, 54
Lightbown, 165
Lindblom, 276, 287, 499
Linell, 206, 212
Linked-Attractor model, 447, 4601, 465, 472,
474, 479, 4834, 48896
liquid/s, 86, 356
liquid feature, 6578
510 Index
List, 483, 501
Ljamina, 218, 234
Lle, 2668, 287, 313, 418, 433
Locke, 24, 54, 252, 254, 262, 265, 287
long consonants, 376
long-term memory, 201, 202; see also memory
Lowenstamm, 8, 12
low frequency, 416
Lucas, 262, 289; see also frequency
Lucchesi, 11
Luce, 22, 48, 56
Luckau, 165, 241, 254
ukaszewicz, 344, 347, 350, 360
Lundberg, 261, 289
Luppino, 458
Lyons, 92
Lyytinen, 364, 370
MacKain, 238, 254
Macken, vii, ix, xi, 1, 2, 3, 7, 10, 12, 13, 223,
256, 51, 54, 56, 1334, 147, 1626, 170,
1748, 182, 1846, 196, 200, 211, 212,
23941, 2546, 262, 2668, 28791, 311,
314, 317, 325, 337, 341, 342, 360, 362, 371,
375, 404, 406, 409, 415, 418, 431, 434, 441,
450, 457, 463, 467, 481, 499, 503
MacNeilage, 24, 32, 52, 260, 270, 286, 288,
310, 314
MacWhinney, 170, 175, 206, 21213, 326, 341,
418, 434, 471, 501
Maddieson, 42
Maekawa, 6, 13
Maillochon, 339
Majorano, 2, 12, 313, 341
malapropisms, 203
Malikouti-Drachman, 207, 210
Malkiel, 103, 115
Mandarin, 168, 372
Manera, 448, 456
manner, 29, 138, 158, 271, 273, 304, 318, 346,
352, 356
manner of articulation, 153
mapping/s, 4, 7, 8, 49, 193, 199, 208,
426, 4624, 4724, 477,
482, 4845, 48893, 496
marked, 232, 310
marked forms, 494
markedness, 8, 10, 36, 344, 463, 466, 473
markedness constraints, 494
Markey, 284, 288
Marshall, J. C., 92
Marshall, P. J., 448, 457
Marslen-Wilson, 377, 408
Marti, 320, 340
Massey, 182, 211
Matelli, 458
Matthei, 4, 7, 12, 20, 54, 2657, 284, 288, 415,
432, 434, 455, 457, 471, 474, 477, 482, 500
Mattingly, 113, 115
Mattoso-Camara, 2934, 314
maturation, 252, 263, 334, 405
Matyear, 314
Matzenauer, 313
Maxwell, 239, 255
Maye, 475, 499
McCarthy, 1, 8, 12, 50, 54, 267, 288, 311, 314,
3778, 380, 403, 409
McCawley, 206
McCune, ix, 2, 3, 9, 14, 23, 24, 54, 56, 240, 255,
2605, 277, 284, 288, 2901, 31415, 342,
382, 384, 410, 442, 445, 448, 4509, 469,
480, 5023
McCune-Nicolich, 288, 4445, 457
McCurry, 94, 115
McDonough, 267, 288
McMurray, 460
McNeil, 218, 234
McNeill, 90, 92
meaning, 4423, 44951
meansends, 2045
Medeiros, 294, 313
MEFs, see monosyllabic experimental forms
Mehler, 341
Melli, 456
melodic patterning, 318
melody, 22, 277, 280, 281, 327
Meltzoff, 285, 288, 448, 457
memories, 3, 4, 6, 23, 47, 62, 2012, 2056,
209, 2635, 312, 335, 358, 360, 455, 460,
469, 476
Menn, 1, 4, 79, 12, 1920, 234, 26,
54, 110, 115, 140, 144, 147, 152, 160, 162,
164, 166, 1702, 178, 180, 188, 189,
1912, 199, 203, 206, 210, 212, 214, 2389,
253, 255, 259, 262, 2657, 284, 2878,
291, 312, 314, 317, 337, 341, 362, 367, 371,
375, 385, 409, 415, 432, 434, 4448, 458,
4604, 46785, 488, 491, 493, 496,
499500, 503
mental representation, 259, 2636, 416, 441,
44455, 468
Menyuk, 176, 204, 212, 224, 2334, 238, 255
Mesalam, 238, 254
Messum, 473, 500
metalinguistic awareness, 178
metathesis, 22, 2735, 80, 84, 86, 144, 147,
148, 153, 159, 1645, 179, 1845, 219, 221,
230, 231, 267, 268, 348, 462
metrical structure, 17, 3204
Metsala, 4323
Mezzomo, 313
Mike, 232, 234
Milewski, 3467, 360
Miller, G. A., 92
Miller, J., 4, 13, 56, 240, 256, 262, 290, 318,
342
Index 511
Miller, R., 4, 13, 24, 56, 213, 2401,
256, 262, 270, 290, 295, 318, 342, 369,
372, 459
minimal pairs, 99, 105, 176, 320
minimal word, 380
Miranda, 294, 314
mirror neuron/s, 4478, 451
misperceptions, 1357, 163
Miyata, 418, 421, 434
Mohanan, 8, 12
monosyllabic experimental forms (MEFs),
21819, 2234, 228333, 237
Montez Giraldo, 165
Moore, 285, 288
mora, 268, 378
morphophonemic rule, 2056
morphophonological rules, 205
morphophonologies, 206
Morrisette, 416, 433, 434
Morsi, 380, 408
Moskowitz, 104, 106, 115, 1434, 153, 166,
170, 176, 193, 201, 212, 2334, 415, 432,
434, 467, 481, 500
motor, 263
motor control, 23, 209, 277, 480, 490
motor procedures, 442
motor sequences, 169
motor skill, 266
motoric maturation, 252
motorically accessible, 310
Mottet, 339
Mozer, 284, 288
Mulford, 262, 289
multidimensional hyperspace, 486
multi-modal, 489
Munson, B., 67, 1112, 52, 255, 417, 433,
434, 4601, 465, 46970, 472, 485,
498, 500
Munson, C., 460, 500
Murphy, 11
Myers, 267, 288
Myerson, 206, 212
Naeser, 212
Nakai, 6, 13, 24, 56, 320, 342, 344, 361, 364,
370, 452, 456
Nakazima, 176, 212
nasal, 353, 354, 356
nasalization, 355
nasality, 6578, 823, 86, 114, 151, 180, 195,
300, 356
Nasr, 378, 409
natural process, 182, 187, 193
Nazzi, 5, 1112, 320, 341
neighborhood density, 4312
neighborhoods, 22, 432
Nelson, 341, 444, 458, 501
Nespor, 321, 341
Nettelbladt, 365, 371
network, 9, 469
network models, 478
neural activation, 484
neurological bases of development, 447
neurological model, 450
neurological research, 447
neuromotor, 2, 5, 260, 451
neurophysiological, 446
Newhoff, 238, 254
Newport, 418, 426, 433, 475, 501
Nguyen, 339
Nicholas, ix, 7, 12, 2634, 284, 286, 288, 458,
461, 481, 484, 496, 500
Ninio, 430, 434
no onset, 36, 3840, 51
nonlinear, 8, 380, 443
nonlinear model, 49, 268
nonlinear phonology, 49, 266
nonlinear representations, 49
non-natural rules, 169
nonsegmental, 613
Nystrm, 448
object permanence, 451
Obrecht, 378, 409
ofine, 465
ofine representation, 464
Oh, 471, 498500
Ohala, D., 267, 289
Ohala, J., 41, 54
Ohnesorg, 612, 92, 225, 234
Okada, 432, 434
Oldeld, 92
Oliveira, C. C., 313
Oliveira, M. A., 293, 314
Oliveira-Guimares, ix, 3, 291, 295, 467, 480
Oller, 36, 55, 238, 255, 310, 314
Olmsted, 96, 110, 116
Otuszewski, 232, 234
omission/s, 289, 36, 38, 40, 98, 179, 268, 283,
3334, 345, 354, 3623, 366, 368, 375, 418,
419, 4201, 426, 429, 4312,
438, 487
omitted, 348
online, 4635
onomatopoeic words, 97
opening, 6578
openness, 6672, 88
opposition, 106, 232
Optimality Theory (OT), 8, 4623, 466,
473, 494
ordered rules, 230, 463
ordinary replacement forms (ORFs), 21819,
22330, 237
ordinary forms, 226
ORFs, see ordinary replacement forms
OT, see Optimality Theory
512 Index
Ota, ix, 7, 8, 9, 409, 41719, 424, 426, 431, 434,
466, 468, 473, 483, 4878, 494, 500
Otomo, 262, 289
output attractor/s, 491, 495
output constraints, 169, 193, 195
output forms, 227
output frequency, 4889
output lexical entries, 476
output lexicon, 1979, 203, 208,
463, 471
output pattern/s, 133, 157, 2445, 248,
252, 472
output representation/s, 471, 488, 489, 492
output template, 475
output variation, 490
output vocabulary, 197
overgeneralization, 169, 186, 189, 1913, 208
Paesov, 219, 234
Paden, 417, 433
Pakinam, 410
palatal, 3534
palatal pattern, 2778, 280, 282, 286
palatalization, 3523, 355
palatals, 352
Palestinian, 380
parameters, 6, 8, 43, 50, 93, 112, 195200, 232,
241, 268, 334, 421, 443
Park, 499
paronymic attraction, 219
Pater, 8, 12, 462, 499
pattern attractors, 491
pattern force, 148, 151, 153, 156, 157
pausing, 365
Pearson, 24, 54
Peizer, 21, 52, 115, 165, 172, 211, 224, 233,
291, 313
Peperkamp, 460, 500
Perani, 458
perception, 1, 4, 6, 19, 21, 42, 48, 51,
7689, 174, 176, 198, 201, 204, 209, 259,
260, 264, 268, 319, 320, 325, 363, 460, 462,
483, 495
perceptual magnet, 472, 486
perceptual processing, 23, 263
perceptuomotor link, 262
performance, 18, 19, 89, 113, 172, 210
Peters, 165, 171, 208, 213, 404, 409, 483
Peterson, 241, 254
Pethick, 364, 372
Phillips, 416, 434
phone class/es, 93, 989, 103, 105, 108,
111, 114
phone tree, 93, 97, 99, 103, 105, 109
phoneme, 21
phonemicization, 161
phonetic control, 1556, 158, 176, 208
phonetic length, 38990
phonetic parameter, 50, 104
phonetic score, 442
phonetic specication, 6
phonic core, 47, 112
phonological awareness, 11213
phonological challenge, 3, 408
phonological conditioning, 418
phonological contrasts, 103, 105
phonological grammar, 48, 292, 308
phonological idiom/s, 93, 106, 113, 143, 169,
1936, 198, 208, 217, 312, 463, 467, 481,
488
phonological knowledge, 3, 7, 17, 40, 47, 172,
311, 317, 374, 474
phonological length, 374, 38990, 403, 405
phonological memory, 5
phonological organization, 17, 25, 41, 48, 50,
112, 113, 238, 282, 291, 345, 486
phonological priming, 478
phonological rule, 217
phonological structure, 421, 426, 427
phonotactic constraints, 494
phonotactic regularities, 47
Piaget, 87, 92, 162, 166, 264, 283, 289, 443,
448, 450, 451, 458
Pierrehumbert, 6, 7, 12, 41, 42, 449, 55, 460,
465, 470, 472, 475, 490, 498, 500
Pine, 54
Pisoni, 53
pitch accent, 419, 422, 4256, 437, 438
pitch patterns, 424
Piwoz, 341
place, 29, 39, 40, 71, 98, 138, 1423, 153, 157,
193, 195, 196, 202, 2704, 294, 304, 308,
313, 318, 329, 334, 352; see also place of
articulation
place harmony, 491
place of articulation, 98, 138, 153, 195, 196,
271, 308, 313, 346, 349, 357, 358, 466,
496; see also place
planar segregation, 267, 271, 274, 275, 281,
321, 337, 338
planning, 3, 23, 195, 310, 345, 357, 359, 360,
454, 493
Platt, 175, 213
play, 263, 284, 4445, 448, 455
Plnat, 332, 341
plosive, 71
Poeppel, 447, 450, 457
Polish, 3449, 353,
467, 487
Pons, 5, 12
Poole, 50, 52
Portuguese, 301, 355357, 473; see also
Brazilian Portuguese
position in the word, 3556, 357, 360, 362,
3689, 380
positional variants, 42
Index 513
practice, 5, 23, 24, 29, 40, 171, 202, 205, 374,
376, 396, 405, 454
preception bias, 176
preferred neural pathways, 284
preferred pattern, 244, 251, 252
pre-phonemic, 176
pre-utterance vowel, 1356
Prvost, 322, 324, 340
Priestly, ix, 2, 3, 9, 13, 1845, 189, 200, 213,
217, 227, 2345, 291, 3001, 314, 343, 360,
375, 406, 409, 441, 458, 462, 4667, 481,
491, 501, 503
Prince, 1, 8, 12, 13, 50, 54, 311, 314, 3778,
380, 403, 409
Principles and Parameters framework, 322
problem solving, 144, 160, 164, 188, 202,
209, 210
process, 19, 40, 43, 112, 144, 147, 148, 161,
1823, 1867, 193, 2246, 248, 304, 333,
348, 355, 3589, 362, 368, 4067
processes, 8, 22, 26, 43, 4850, 84, 112, 140,
144, 147, 1512, 157, 158, 161, 187, 193,
268, 304, 307, 3435, 3529, 362, 365, 380,
415, 4645, 491
processing, 168, 284
production, 16, 10, 1929, 402, 48, 51, 111,
1356, 138, 147, 1512, 156, 17381, 185,
192, 195204, 208, 209, 23853, 25960,
26884, 294304, 308, 31113, 31822,
325, 328, 331, 3349, 344, 348, 349, 35460,
364, 371, 3745, 395, 423, 431, 437, 455,
4612, 472, 474, 4767, 487, 490
production patterns, 40, 239, 240, 244, 245,
248, 249, 251, 253, 260, 262, 266, 280, 283,
399, 441
production routines, 282
program, 169, 170, 195200
programming, 169, 170, 195, 197, 200
progressive idiom/s, 106, 107, 269
progressive phonological idioms, 193, 208, 481
prominence, 6578, 89, 324, 368, 3746, 390,
395, 397
pronunciation, 169, 178, 204
prosodic, 2, 5, 6, 22, 41, 46, 49, 50, 61, 114,
133, 137, 148, 151, 157, 163, 209, 239, 260,
271, 31922, 3246, 330, 332, 3357, 375,
422, 427, 428,
431, 477
prosodic heightening, 260
prosodic phonology, 321
prosodic structure, 2, 6, 46, 319, 324, 326, 328,
333, 337, 378, 466
prosodic template, 330, 332
prosodic unit/s, 151, 404
prosodic word, 271
prosodically highlighted, 262
prosody, 5, 66, 320, 337, 375, 466
proto-determiner/s, 330, 3323, 335
proto-words, 1702, 176, 448
psychological reality, 161, 170, 265, 284
Pulvermller, 451, 458
Pye, 418, 434, 483, 501
quantitative, 407
quantity, 3689
quantity-sensitive, 378
Radical Construction Grammar, 41, 49
Ramsay, 263, 289, 448, 458
Ramus, 320, 341
Rangel, 294, 314
Ravid, 377, 409
real time, 4634
real-time mapping, 464
recidivism, 191, 193
recognition, 21, 47, 87, 93, 153, 1734, 179,
205, 232, 238, 242, 264, 320, 4712,
475, 480
Redanz, 341, 363, 370
reduction, 96, 107, 1389, 144, 1468, 1512,
155, 158, 163, 199, 332, 3479, 351, 353,
365, 367, 381, 398, 412, 431
redundancy, 465, 474, 495
reduplicated, 29, 61, 689, 71, 758, 82, 878,
106, 111, 2967, 305, 307, 310, 3289, 332,
334, 367
reduplicating, 331
reduplication, 278, 334, 6578, 834, 107,
110, 194, 2312, 266, 294, 3035, 308,
31011, 3302, 3367, 3645
reduplicative, 75, 301, 310, 312
referential language, 265, 4501, 4534
referential language use, 265, 454
referential words, 264, 285, 4489, 453
regression, 8, 27, 239, 244, 269, 281, 291, 374,
405
regressive assimilation, 304
regressive idiom/s, 106, 153, 225
Reichling, 19
relational words, 284
reorganization, 133, 1512, 239, 241, 248, 252,
279, 308, 310, 328, 470
representation, 3, 69, 1719, 23, 423, 4650,
112, 144, 161, 175, 205, 213, 225, 229, 259,
2638, 2713, 2756, 279, 281, 2845,
2912, 307, 30910, 31213, 317, 320,
3367, 416, 441, 4448, 4501, 455, 4602,
464, 4689, 4712, 4756, 4789, 480,
4849, 4926
representational capacity, 270
representational play, 2635, 445,
448, 451
restructuring, 244, 248, 2512
resyllabication, 319
reversions, 219, 228
Reznick, 364, 370
514 Index
rhythm, 45, 31920, 334, 338, 454
rhythmic patterning, 23
rhythmic shape, 80, 408
Ribas, 313
Rice, 501
Richardson, 363, 371
Riggio, 456
Ritterman, 416, 433
Rizzolatti, 447, 456, 458
Rosch, 456, 4589
Rose, 322, 326, 341, 382, 385, 410
Roug, 261, 289, 457
Roug-Hellichius, 453, 457
rounding, 6578, 86, 187, 496
routine/s, 2, 63, 1356, 163, 1712, 199, 239,
242, 253, 284, 405, 4767
Rovee-Collier, 2623, 265, 289
Rowland, 54
Roy, 468, 480, 501
rule generalization, 1901
rules, 4, 8, 22, 26, 28, 42, 48, 80, 83, 93, 103,
1058, 11213, 133, 13840, 147, 1512,
1579, 1613, 16970, 173, 175, 179209,
217, 230, 238, 265, 268, 271, 274, 307, 343,
345, 35860, 4617, 4714, 477, 4804,
48890, 4934
ruse/s, 22933
Russian, 104, 182, 206, 464, 490
Rutherford, 226, 234
Saaristo-Helin, 365, 371
Saccuman, 458
Saffran, 1, 13, 475, 501
Sagart, 5, 11, 51, 261, 286
Sagey, 274, 289
Saleh, 380, 410
Salem, 380, 410
Salerni, 52
Salidis, 418, 434
salience, 6, 10, 186, 248, 251, 262, 280, 282,
285, 3445, 3634, 368, 377, 387, 390
salient, 36, 283, 357, 363, 368, 376, 406
Salo, 20, 323, 55
Salvi, 499
Sander, 187, 213
Saporta, 92
Sartre, 263, 289, 450, 458
Savinainen-Makkonen, viii, ix, 10, 13, 36, 55,
314, 341, 344, 360, 36271, 376, 405, 410,
418, 434, 4667, 503
Schade, 23, 51
Scheer, 7, 8, 13, 338, 341
schema/s, 7, 21, 26, 87, 90, 111, 157, 277, 307
schemata, 359
Schiller, 417, 433
Schmidt, ix, 7, 12, 458, 461, 481,
484, 496, 500
Schwartz, J.-L., 500
Schwartz, R. G., 24, 55, 178, 211, 213, 284,
287, 289, 416, 434
Scifo, 458
Searle, 446, 458
Sedang, 44, 45
segmental, 7
segmental organization, 7
segmentation, 5, 173, 239, 253, 267, 320
segments, 56
select, 289, 32, 36, 405, 488
selected, 23, 23, 28, 32, 367, 94, 133, 144,
147, 149, 151, 153, 182, 185, 241, 280, 284,
2967, 3034, 327, 339, 351, 376, 391,
3956, 3989, 406, 425, 442, 496
selecting, 25
selection, 24, 29, 38, 79, 88, 164, 178, 182, 184,
194, 198, 209, 269, 282, 285, 329, 374, 398,
406, 415
selection rule/s, 199200, 464, 496
selectivity, 103, 108, 262, 368, 442
self-monitoring, 2023, 260, 262, 284
self-organization, 2
Selkirk, 321, 338, 341
semantic bootstrapping, 317
sensorimotor, 2, 5, 4512
sequencing, 209
Shahin, 380, 405, 410
Shaw, 474, 499
Shoeib, 380, 410
short-term memory, 2012; see also memory
sibilance, 6578, 81, 85, 86
Silber, 238, 255, 465
Simmons, 13, 56, 240, 256, 262, 290, 342
simplication, 48, 96, 13940, 152, 158, 1612,
196, 463
Sinclair, 319, 330, 3345, 342
single lexicon, 173
single-valued features, 43, 48
single-word period, 3
Siqueland, 11
skeletal tier, 275
slips of the ear, 137, 472
slips of the tongue, 1513
Slobin, 113, 168, 213
Smith, B. L., 182, 213
Smith, F., 92
Smith K. D., 445, 55
Smith, L. B., 291, 314, 451, 458
Smith, N. V., 78, 10, 13, 49, 51, 53, 114, 116,
134, 13840, 162, 166, 170, 173, 175, 182,
188, 1912, 197, 213, 238, 255, 343, 361,
461, 463, 482, 501
Smolensky, 8, 13, 466, 495, 498, 501
Snow, C. E., 171, 213
Snow, K., 226, 234
social parameter, 104
sonority, 48, 6578, 347
Sosa, 431, 434
Index 515
sound effects, 270
sound-play, 139, 1702, 204
Spanish, 3, 20, 22, 25, 32, 13344, 153, 1558,
1614, 168, 182, 185, 207, 404, 467, 471,
473
Spencer, A., 268, 289
Spencer, P., 452, 458
spreading, 321, 331
spurt, 269, 384
Stager, 23, 55, 57, 320, 341
stages of development, 8, 95, 106, 152, 245,
368, 408
Stampe, 8, 13, 187, 193, 213, 343, 3612, 371
statistical learning, 1
Steiner, 166
Stemberger, 20, 55, 266, 268, 284, 289, 362,
370, 463, 483, 498
Sterne, 171, 213
Stevens, 213, 499
Stich, 447, 458
Stockman, 239, 255
Stockwell, 137, 166
Stoel, 24, 55, 163, 1656, 238, 241, 2545, 262,
268, 314, 372, 431, 434, 483, 501
Stoel-Gammon, 24, 55, 238, 255, 262, 268,
2878, 291, 3623, 367, 370, 372, 376, 385,
410, 431, 434, 465, 476, 483, 501
Stone, 218, 234
stop/s, 25, 28, 40, 44, 51, 6578, 818, 99,
11011, 138, 143, 148, 153, 156, 1589, 163,
176, 179, 1824, 207, 235, 243, 245, 248,
2523, 274, 281, 285, 292, 308, 327, 3526,
390, 404, 47781
storage, 168, 2656
Storkel, 6, 13, 465, 501
strata, 474, 480, 482
strategies, 93, 110, 113, 13940, 165, 168,
17784, 194, 196, 2089, 217, 2249, 232,
321, 326, 355, 365, 441
stress, 4, 301, 6578, 89, 176, 200, 220, 231,
286, 293, 3034, 30910, 3205, 328, 336,
363, 368, 376, 378, 3801, 41819
stress placement, 323
stressed syllable, 44, 77, 1478, 151, 297, 324,
375
strongly articulated, 72, 77, 79, 813, 87
Studdert-Kennedy, 23, 55, 238, 255, 343,
357, 361
sublexical patterns, 48
substitution/s, 26, 80, 84, 86, 138, 144, 1478,
1529, 164, 179, 2246, 232, 238, 248, 253,
307, 309, 339, 345, 349, 3545
Sullivan, 262, 289
Sundberg, 56, 432, 435, 502
suprasegmental feature, 403
Suxanova, 218, 234
vakin, 231, 234
Swedish, 20, 40, 43, 365, 417
Swingley, 6, 13, 416, 435
Swinney, 491, 501
syllabic language, 334
syllable positions, 44, 46, 48, 344, 473
syllable structure, 48, 15860, 334, 37781,
3845, 404
syllable weight, 378, 419, 422, 426, 431
syllable-timed language, 323
symbol/s, 446, 449
symbolic, 17, 46, 50, 172, 259, 262, 264, 445,
449, 455
symbolic play, 455; see also play,
representational play
systematic, 3, 20, 26, 28, 103, 133, 148, 173,
179, 184, 189, 200, 21718, 238, 240, 245,
248, 249, 253, 265, 271, 280, 31819, 324,
3267, 329, 335, 345, 3489, 394, 403,
4056, 415, 4269, 428, 473, 487, 496
systematicity, 26, 158, 219, 238, 259, 271, 284,
291, 297, 326, 345, 351, 4067
systematization, 2389, 2512, 266,
279, 297
Szreder, ix, 10, 405, 4668, 487
target language, 317, 318, 334, 338
task, 203, 204
Teixeira, 294, 314
Temne, 42
template/s, 19, 1734, 3850, 1845, 194,
200, 267, 269, 2917, 30013, 31721,
32539, 343, 350, 35660, 362, 364, 3689,
3747, 395, 403, 406, 431, 4412, 446, 450,
4525, 463, 467, 4727, 48195
template matching, 367
templatic, 3, 394
Templin, 96, 116
Terrell, 416, 434
Tettamanti,, 451, 458
Thal, 168, 364, 372
Theakston, 54
Thelen, 284, 289, 291, 314, 445, 4512,
4579
Theoret, 448, 457
theory, 151
Thiessen, 1, 13, 475, 501
Thompson, 456, 459
Tincoff, 55
Tishman, 239, 255
Toda, 42
Todd, 51
Todorovna, 499
Toivainen, J., 366, 372
Toivainen, K., 368, 372
token frequency, 294, 312, 417, 466,
488; see also frequency
Tokowicz, 471, 501
Tomasello, 41, 55, 341
Ton, 20, 52, 279, 286, 319, 340
516 Index
trade-off, 97
Tranel, 338, 342
transduction, 1889, 198, 2003
Treiman, 48, 55
Tremblay, 322, 324, 335, 340
trial-and-error, 1869, 196, 208
trochaic, 378, 322, 338, 362, 37581
trochaic bias hypothesis, 322
trochaic foot, 322, 338; see also foot/feet
trochaic pattern, 376
truncation, 278, 324, 322, 325, 3305, 363,
365, 381, 418, 42032, 4378, 468
Tucker, 450, 452, 459
Tuomi, 363, 372
Turunen, 365, 372
two-lexicon model, 7, 173, 1978, 461, 471,
496
Tyler, A. A., 416, 435
Tyler, M., 1, 12
type frequency, 48, 312, 417, 466, 488; see also
frequency
typological, 3267
typological constraints, 31719, 326, 330,
3334, 337
typologically constrained, 320, 338
underlying form/s (UF/s), 103, 217, 22733,
462, 479
underlying representation/s, 140, 337, 469,
479, 496
underspecication, 48, 267
underspecied, 6, 320
underspecied phonemes, 174
unit phrases, 135, 137, 163
unitary feet, 324; see also foot/feet
universal, 21, 36, 40, 62, 93, 109, 112, 1389,
147, 1603, 204, 311, 321, 338, 380, 407,
466, 487
Universal Grammar, 8
universality, 4, 324, 478
universals, 4, 40, 133, 160, 162, 186, 285,
460, 466
unmarked, 311
unruly contextual effects, 467
unspecied feature, 268
usage-based, 1, 17, 29, 41, 47, 4601, 464, 466,
469, 471, 478, 4856, 4901
U-shaped curve, 8, 374, 405, 463, 481
Uzgiris, 263, 289
vague, 7
vague representation, 6
Vainikka, 499
Vainio, 368, 372
van der Hulst, 7, 13, 43, 44, 48, 49
Van Gulick, 265, 289
Vance, 432
Varela, 456, 459
variability, 46, 223, 401, 103, 140, 146, 155,
158, 186, 200, 240, 245, 24850, 253, 281,
284, 308, 31112, 326, 335, 344, 3528, 415,
466, 471, 47880, 490, 496
variable, 57, 22, 1056, 140, 144, 148, 150,
152, 156, 1589, 171, 177, 195200, 208,
242, 245, 252, 276, 283, 2923, 326, 332,
352, 354, 357, 375, 390, 395, 404
variation, 8, 18, 21, 412, 489, 51, 97, 99,
1037, 112, 133, 135, 138, 140, 1502,
1558, 163, 172, 177, 18592, 199, 2012,
209, 218, 224, 239, 2445, 262, 2746, 291,
2934, 308, 311, 318, 329, 334, 3527, 380,
395, 406, 424, 426, 429, 431, 444, 452, 468,
471, 47983, 4901, 496
velar/s, 352, 353
Velleman, P. F., 249, 255
Velleman, S., ix, 23, 1314, 23, 25, 29, 36,
38, 41, 56, 2658, 2834, 28991, 2967,
31519, 3423, 351, 361, 364, 368, 372, 376,
405, 410, 415, 435, 442, 453, 459, 469, 475,
480, 5023
Velten, 612, 68, 92, 182, 190, 213
Veneziano, 319, 330, 3345, 342
Vergnaud, 8, 12
Verlaine, 323
Verluyten, 3224, 342
Viegas, 293, 314
Vigorito, 11
Vihman, M. M., ix, 114, 17, 20, 225, 2930,
3641, 51, 545, 138, 167, 180, 184, 190,
210, 213, 23841, 2556, 2602, 26670,
277, 28390, 291, 2957, 31315, 31720,
326, 333, 335, 338, 3415, 351, 3612, 364,
36876, 382, 384, 398, 4056, 410, 41518,
4323, 435, 4412, 444, 445, 4513, 4567,
459, 461, 463, 46772, 4756, 4803, 489,
496, 5003
Vihman, V-A., 14, 291, 31819, 375, 405
Vitevich, 22, 47, 56
Vlahovi, 232, 234
vocabulary, 48, 624, 88, 107, 11011, 134,
164, 197, 2012, 206, 241, 284, 2912, 295,
330, 345, 382, 384, 399, 405, 407, 41617,
444, 467, 470, 4801, 488, 490
vocabulary size, 48, 470
vocal capacities, 260
vocal control, 260, 452
vocal exploration, 284
vocal motor action, 266
vocal motor scheme/s, 260, 262, 2656, 270,
271, 273, 280, 2825, 442, 448, 4524
vocal motor skill, 266
vocal patterns, 260, 262
Vogel, 207, 213, 321, 341
voice, 71, 83
voiced, 69
voiceless, 85
Index 517
voicelessness, 6578, 86, 114, 148, 151, 162
voicing, 51, 6578, 86, 1056, 108, 111, 137,
146, 153, 1558, 176, 180, 1823, 187,
1901, 195, 197, 207, 252, 276, 3523, 355
Volterra, 240, 254, 262, 286, 444, 456
von Rafer-Engel, 170, 213
Vorperian, 480
vowel sequences, 29, 323
Wales, 92
Walley, 4323
Wang, 1045, 114, 116, 312, 315, 416,
435, 503
Ward, 454, 459
Waterson, ix, 1, 3, 7, 14, 19, 202, 256, 57, 61,
63, 65, 92, 111, 116, 157, 167, 174, 1946,
199, 202, 210, 214, 2334, 2389, 256, 259,
2656, 270, 290, 315, 317, 3423, 361, 376,
410, 415, 431, 435, 443, 450, 459, 4623,
4667, 476, 481, 5023
Watson, 3768, 4034, 410
Wauquier, ix, 45, 33940, 342, 375, 410, 466
Wauquier-Gravelines, 318, 321, 337, 339, 342,
385, 410
weakly articulated, 72, 77, 80, 82
Weeks, 21, 52, 57, 104, 111, 11516, 165, 172,
211, 224, 233, 291, 313
Weenink, 382, 408
Weir, 61, 92, 171, 202, 214
Weismer, 239
Weissenborn, 11
Wellmers, 195, 214
Welsh, 20, 36, 38, 3445, 357
Werker, 23, 55, 57, 320, 341, 475, 499
Werner, 266, 290, 443, 445, 44850, 459
Westbury, 182, 214
Wheeler, 268, 287
whisper, 95, 111, 119, 202
Whitaker, 12, 416, 433, 452, 457
whole-word, 3
Wijnen, 24, 52, 363, 372, 426, 435
Wilbur, 174, 206, 214
Wilcox, 178, 211, 284, 287
wild variation, 467, 468
Williams, K. A., 499
Williams, N. M., 12, 416, 433, 452, 457
Woods, 239, 255
Woodward, 341
word combinations, 455
word endings, 245
word identication, 241, 382
word meanings, 444
word pattern/s, 1, 7, 22, 24, 26, 35, 41, 133,
136, 1405, 14853, 15764, 248, 281,
375, 407
word production pattern/s, 280, 283, 285
word recipe/s, 238, 239, 242, 244, 267
word structure, 49, 6578, 152, 1612,
311, 329, 334, 366, 380, 384, 388, 42930,
487
word-nal consonant, 297
working memory, 335; see also memory
Wright-Cassidy, 340
Yaeger, 166
Yamaguchi, ix, 4, 5, 375, 410
Yeni-Komshian, 168, 214
Young, 340, 448, 457
Zamuner, 417, 435, 465, 502
Zawaydeh, 48, 53
Zelnicker, 163, 165, 167
Zlatin, 239, 256
Zonneveld, 8, 12, 462, 499
Zwicky, 203, 214
Zydorowicz, 344, 349, 350, 361
518 Index