You are on page 1of 30

Language, Interaction and Acquisition 4:2 (2013), 161189. doi 10.1075/lia.4.2.

03wau
issn 18797865 / e-issn 18797873 John Benjamins Publishing Company
Convergence and divergence
in the acquisition of French liaison
by native and non-native speakers
A review of existing data and avenues
for future research
Sophie Wauquier and Ellenor Shoemaker
Universit de Paris 8 & Laboratoire Structures Formelles du Langage (UMR
7023/CNRS) / Universit Sorbonne Nouvelle Paris 3 & Laboratoire de
Phontique et Phonologie (UMR 7018/CNRS)
Tis article presents an overview of empirical fndings to date concerning the ac-
quisition of liaison in French as a frst and second language (L1 and L2, respec-
tively). We present data culled from production studies as well as from psycho-
linguistic experimentation involving various paradigms. Our aim is to highlight
both the similarities and diferences in the learning strategies and developmental
paths followed by these two groups of learners, including particular examina-
tion of how representations of liaison in a learners phonological grammar may
develop and change throughout the course of development. We conclude with
a discussion of areas where existing data are lacking and potential avenues for
future research.
Keywords: French liaison, phonology, second language phonology, frst language
acquisition, second language acquisition
1. Introduction
Liaison, a phenomenon of external sandhi in spoken French, is without doubt
one of the phonological concepts that has most inspired contemporary phonol-
ogy. From the SPE
1
model presented by Chomsky and Halle (1968) to Optimality
1. Tis theoretical framework is named afer the book in which it was frst published: Te Sound
Pattern of English.
162 Sophie Wauquier and Ellenor Shoemaker
Teory (e.g. Tranel 2000), all contemporary theoretical frameworks have attempt-
ed to account for this phenomenon, and numerous analyses have been proposed
(e.g. Schane 1968; Agren 1973; Morin & Kaye 1982; Clements & Keyser 1983;
Encrev 1988; Tranel 1995, 2000; Bybee 2001; Durand & Lyche 2008; Durand,
Laks, Calderone & Tchobanov 2011; for a review see Ct 2011). To these analyses
can be added extensive psycholinguistic research investigating lexical processing
and access in adult native speakers (Wauquier-Gravelines 1996; Gaskell, Spinelli
& Meunier 2002; Spinelli, McQueen & Cutler 2003; Wauquier-Gravelines 2005;
Nguyen, Wauquier-Gravelines, Lancia & Tuller 2007; Tremblay & Spinelli 2013;
Shoemaker, in press; among others).
Despite the vast amount of work undertaken on adult phonology in French,
the developmental aspects of liaison have only begun to be explored in the last
15 years. More recent research has examined the acquisition of liaison in French
as a frst language (Chevrot & Fayol 2000, Wauquier-Gravelines & Braud 2005;
Wauquier-Gravelines 2005; Wauquier 2009; Chevrot, Dugua & Fayol 2005, 2009;
Chevrot, Dugua, Harnois-Delpiano, Siccardi & Spinelli 2013) as well as the ac-
quisition of liaison in French as a second language (Matter 1986; Dejean de la
Btie 1993; Dejean de la Btie & Bradley 1995; Mastromonaco 1999; Tomas
2004; Stridfeldt 2005; Hannahs 2007; Shoemaker & Birdsong 2008; Shoemaker
2010; Tremblay 2011; Tremblay & Spinelli, in press). Tis body of research ofers
a rich and varied empirical foundation based on both perception and production,
incorporating longitudinal, cross-sectional, and psycholinguistic data. However,
though an extensive body of work has focused on the acquisition of this phono-
logical phenomenon, existing data are at times contradictory or lacking in sev-
eral key areas. Subsequently, fundamental questions remain concerning both the
chronology and development of this phenomenon in learners of frst and second
language (L1 and L2, respectively), and the way in which liaison is represented at
diferent stages in a learners phonological grammar.
Afer a brief explanation of the principles of liaison and laying out the theoret-
ical and empirical issues that liaison raises in the feld of psycholinguistics, this pa-
per will review data from existing production and behavioral studies regarding the
acquisition of liaison in L1 and L2 French, giving special focus to the processing
of liaison by both native and non-native speakers. Exploring the issue of process-
ing contributes to our knowledge of the mechanisms involved in the acquisition
of sandhi phenomena in general, which lie at the interface of phonology, prosody
and morphosyntax. In addition, the investigation of processing is of particular
interest for acquisition research in that it adds to our understanding of similarities
and diferences between acquisitional processes in the L1 and L2.
In the study of the acquisition of liaison, two main theoretical frameworks
have emerged to account for existing data: a lexicalist model (e.g. Bybee 2001;
Te acquisition of French liaison by native and non-native speakers 163
Ct 2005), which proposes that learning essentially emerges from usage and is
conditioned by the frequency of lexical information in the input (e.g. Chevrot et
al. 2005; Chevrot et al. 2009), and a phonological model, which proposes both
that abstract representations are established in the phonological grammar dur-
ing development and that liaison can be formalized in terms of foating segments
(Encrev 1988). For the purposes of this review, we remain neutral as to this theo-
retical debate, though reference to both models is made throughout (for in-depth
discussion of these two opposing viewpoints see Wauquier 2009, 2010, Chevrot
et al. 2013). Our aim is rather to identify areas of convergence and divergence in
L1 and L2 acquisition of liaison with an eye to comparing how liaison develops in
both the native and non-native grammar. We conclude with an overview of linger-
ing questions and unresolved debates, including a (non-exhaustive) list of topics
which could be avenues for future research such as the role of orthography, the
role of input, the role of linguistic typology and the relationship between the target
and source languages, as well as the plasticity of the adult phonological system.
2. A brief description of French liaison
Liaison is a phenomenon of external sandhi that occurs at the lef edge of major
lexical categories in spoken French. A word-fnal consonant that is normally la-
tent surfaces when two vowels come into contact at a word boundary (e.g. un ami
[nami] a friend versus un livre [liv] a book, where /n/ is not pronounced).
Te fnal consonant in Word 1 is subsequently resyllabifed (enchane) as the on-
set of Word 2,
2
therefore creating misalignment between word and syllable bound-
aries in liaison environments, e.g. un ami [.n#a.mi] versus un livre, e.g. [#li.v].
3
Te realization of consonants in liaison environments is syntactically as well
as prosodically and phonologically conditioned. Consequently, liaison is not re-
alized at every word boundary where the phonological environment would ren-
der it possible. Traditionally, there are thought to be three categories of liaison,
as proposed by Delattre (1947): obligatoire (obligatory), interdite (prohibited)
2. Tere are instances of liaison without linking (enchanement), in which liaison consonants
are realized but not resyllabifed. Instead, liaisons are realized and followed by a glottal stop,
a pause or a schwa at the word juncture (e.g. javais un rve [avz v] I had a dream as
produced by V. Giscard dEstaing, the former President of France, quoted in Encrev 1988).
However, such cases are rare and mostly limited to public discourse and carefully produced
speech in native speakers.
3. Here and throughout, the symbol # represents a word boundary, while a period . represents
a syllable boundary.
164 Sophie Wauquier and Ellenor Shoemaker
and facultative (optional). Liaison is considered to be obligatory, for example, be-
tween determiner and noun (e.g. un enfant the child [.n.f]/*[..f]) or be-
tween personal (clitic) pronoun and fnite verb (e.g. vous avez you
PLUR
have [vu.
za.ve]/*[vu.a.ve]). Environments where liaison is prohibited or blocked occur be-
tween a singular lexical noun and a verb (e.g. lenfant aime the child loves [l.
f.m]/*[l.f.tm]), or between a singular lexical noun and a following adjective
(lenfant anglais the English child [l.f..gl]/*[l.f.t.gl]). Finally, liaison can
be produced optionally, for example afer certain adverbs such as trop too in trop
aim too loved, [to.pe.me] or [to.eme], or afer the negative particle pas not
as in pas encore not yet, [pa.z.k] or [pa..k]. In contexts where liaison is
optional, the choice to realize a liaison consonant (LC) is usually a stylistic one on
the part of the speaker. A higher register of speech would call for more frequent
realizations of liaison consonants in optional contexts.
Tese three categories of liaison do not, however, comprehensively account
for contemporary usage by native speakers, who exhibit a great deal of variation in
both the realization and non-realization of liaison across all three categories. Boula
de Mareil, Adda-Decker and Gendner (2003) refned the categorization of liaison
following an analysis of a corpus of 66,500 sentences taken from the newspaper Le
Monde and read aloud by 120 native speakers (a total of 100 hours of speech). Te
corpus shows evidence of a continuum of liaison realizations rather than the strict
three-tiered division proposed by Delattre (1947). Te authors further break down
the three categories as follows: seven varieties of obligatory liaison, with realiza-
tion rates from 81 to 99.2%; seven varieties of prohibited liaison with realization
rates of less than 2% (with the exception of the context non-auxiliary verb + en,
e.g. tu pars en, you go to, which has a realization rate of 10.5%); and four vari-
eties of optional liaison, which exhibit the greatest variation with realization rates
ranging from 14 to 44 %.
Widespread variability in the production of liaison by native speakers has also
been exhibited in data from the large-scale corpus Phonologie du franais contem-
porain (Contemporary French Phonology; Durand, Laks & Lyche 2002, 2009),
which examined both read speech and guided conversation from 600 French
speakers from various regions across the globe (Durand & Lyche 2008; Durand
et al. 2011; Mallet 2008). Data show that, aside from obligatory contexts in which
Word 1 is a clitic pronoun or determiner and where liaison is realized in close
to 100% of cases (e.g. pronoun + verb: vous avez you
PLUR
have / determiner +
noun: un ami a friend
MASC
), usage is conditioned by a complex interplay of word
length, lexical frequency, and the nature of the liaison consonant itself, even in
contexts that have been heretofore considered obligatory (e.g. adjective + noun).
Tus, while liaison is realized in grand ami great friend more than 70% of the
time, it is almost never realized in grand moi great emotion, a phrase which is far
Te acquisition of French liaison by native and non-native speakers 165
less common. Similarly, liaison is realized afer the prepositions en and dans, both
in, 100% of the time, while chez at/to the home of triggers liaison in only 86%
of productions. Given such extensive variability in native speaker usage, the com-
plete acquisition of liaison in French requires the learner not only to internalize
the phonological and syntactic rules governing its realization, but also to develop
pragmatic competence based on usage.
As we will see below, the empirical variability of the input to which both L1
and L2 learners are exposed is not necessarily refected in psycholinguistic experi-
mentation, in that the majority of data is based on a single context in which liaison
is both obligatory and variable in its realization, namely, the lef edge of a noun
(un enfant, des enfants, un petit enfant, un grand enfant, un gros enfant). In both L1
and L2, the number of contexts in which most production errors are made (and
therefore the contexts which are investigated in psycholinguistic experimentation)
is very limited and consistently gives rise to the same types of errors in L1 and L2
learners. Consequently, existing experimental data are largely homogenous, con-
trary to native usage. Furthermore, research has produced disparate sets of data
that do not necessarily form a coherent picture of developmental strategies. With
this in mind, we present in Sections 3 (production data) and 4 (perceptual data) a
critical review of the data gathered over the past 15 years in an attempt to clarify
which data accurately refect learner behavior.
3. Te production of liaison
3.1 Production data from L1 French
Over the past decade and a half, diverse data have been collected from mono-
lingual children acquiring French in longitudinal studies (e.g. the Sophie corpus,
cited in Chevrot & Fayol 2000, 2001; Chevrot et al. 2005, 2009), cross-sectional
studies (Wauquier-Gravelines 2005; Chevrot & Fayol 2000), as well as behavioral
studies and psycholinguistic tests (Dugua 2006; Chevrot et al. 2009; Barbu, Nardy,
Chevrot & Juhel 2013).
In the chronology of the acquisition of L1 French, production data show de-
velopmental patterns that are relatively homogenous from one child to the other,
leading us to conclude that the acquisition of liaison is a systematic phenome-
non that occurs in all children learning French as their frst language. Analysis of
production errors suggests that children acquire liaison primarily through gram-
matical generalizations based on obligatory contexts. Te frst errors in liaison are
observed only afer the acquisition of syntax and morphology has begun and a
considerable lexicon is already in place, a period which begins at 2.53 years and
166 Sophie Wauquier and Ellenor Shoemaker
ends in monolingual children by about 4 years of age (or up to 6 years for children
growing up in a bilingual environment).
Most children produce similar and systematic errors. At the frst stage of de-
velopment, errors are observed almost exclusively in the obligatory liaison context
determiner + noun (e.g. un lphant an elephant) and less frequently in the con-
text adjective + noun (e.g. un grand lphant a big elephant). Errors in obliga-
tory liaison uniformly disappear from childrens speech afer this developmental
period and are not observed in adult speech. In contrast, errors in optional liaison
contexts, where variation is widespread in adult speech, appear later in develop-
ment, toward 78 years of age (Wauquier-Gravelines 2005), which coincides with
the development of pragmatic competence and when children start to master vari-
ation as a function of discourse (Bernicot 2000).
Both Wauquier (2009) and Chevrot et al. (2009) propose a three-stage devel-
opmental scenario in the acquisition of L1 liaison; they diverge, however, on the
development of generalization in the third stage.
1. First, children identify referential units, or chunks, in the input, which include
multiple lexical units, e.g. the NP includes the determiner and a noun, un ours
[.nus] a bear.
2. Second, children incorrectly segment these units by interpreting the LC as
a lexical consonant at the beginning of Word 2. Tey detach the determiner
from the noun without taking into account the special status of the LC, e.g.
*le nours [l.nus] for lours [lus] the bear. (See Table 1 for more details on
errors.)
3. Finally, according to Wauquier (2009, 2010), children then formulate an ab-
stract generalization of prosodic slots allowing them to understand that the
LC is prosodically the onset of Word 2, but that its segmental content is pro-
vided by Word 1, e.g. lours, [lus] the bear, un ours, [.nus] a bear, un petit
ours, [.p.ti.tus] a little bear. Chevrot et al. (2009), on the other hand, argue
in favor of a lexical surface-based generalization in which the formation of
item-based constructions emerges from usage. Te usage-based model does
not directly address the underlying phonological representation of the LC.
Based on data gathered from varying sources and experimental paradigms,
Wauquier (2009, 2010) has classifed child production errors observed in the sec-
ond stage of development as belonging to four main types (see Table 1). Te frst
and most prevalent error type involves the use of an incorrect consonant in an
obligatory liaison environment: *les nanes [lenan] for les nes [lezan] the donkeys
or *un zlphant [zelef] for un lphant [nelef] an elephant. Further examples
of this type come from Dugua (2006), who elicited productions of vowel-initial
words in isolation by asking children to call out to a series of animals (e.g. Ours !
Te acquisition of French liaison by native and non-native speakers 167
Viens ici ! Bear! Come here!). She found that 15% of child productions were of
the type Nours ! (from un ours [.nus] a bear) or Nne ! (from un ne [.nan] a
donkey), suggesting that mental representations of these items retained the resyl-
labifed liaison consonant as the lexical onset.
A second type of error involves epenthesis of a liaison consonant in order to
resolve hiatus: [papatus] or [papanus] for Papa Ours [papaus] Papa Bear. A
third type of error is the non-realization of a liaison consonant in an obligatory
environment: [ elef] for un lphant an elephant (Wauquier 2009). Finally, a
fourth category of error involves the reinterpretation of a lexical onset consonant
as a liaison consonant: Blanche-Neige et les sept ains [blnelest] for Blanche-
Neige et les sept nains [blnelestn] Snow White and the seven dwarfs.
Errors in obligatory contexts are not realized with the same frequency, nor are
they realized at exactly the same time from one child to the next. Te frst and sec-
ond error types are by far the most frequent, while the others are rare. Furthermore,
not all liaison consonants /n, t, z, , p, k/ give rise to the same number of errors,
and the choice of a fller consonant appears to be linked to the frequency of its
usage in that same context (Chevrot et al. 2009). Te consonant appearing most
ofen as a fller is /n/, followed by /z/, both of which appear in obligatory contexts
and are highly frequent (Braud 1998; Chevrot et al. 2009).
3.2 Production data from L2 French
Very little data on the production of liaison by learners of L2 French are currently
available, and no consistent developmental chronology has been documented,
although researchers have identifed liaison as a key issue in the acquisition of
French phonology in adult learners (Hannahs 2007). Nonetheless, what data do
exist show similar developmental patterns and learning strategies across learners.
However, unlike L1 acquisition, in which competence uniformly stabilizes at a
certain point in development, production errors in L2 learners have been shown
to persist at very advanced stages of learning, even when learners are otherwise
fuent speakers.
Concerning the types of liaison errors produced by L2 learners (see Table 1),
Tomas (2004) conducted a systematic study of both spontaneous and read speech
produced by English-speaking university students in their third year of French
study. He notes that, in obligatory contexts, the rate of realization in L2 learners
(91%) is close to that of native speakers. More detailed analysis, however, shows
that liaison errors accounted for 20% of all pronunciation errors in the corpus,
and included the non-realization of liaison, the use of the wrong LC, or, most
commonly, a lack of resyllabifcation of the LC (e.g. un grand ami a great friend
produced as [gt.ami] instead of [g.tami]), an error which accounted for
168 Sophie Wauquier and Ellenor Shoemaker
8.5% of all liaison productions.
4
Furthermore, Tomas (2004) observes errors of
obligatory liaison that are produced according to the graphic representation of
the LC and not according to the underlying LC (e.g. un grand ami a great friend
produced as [gdami] instead of [gtami]). A similar pattern of errors is also
revealed in a study by Mastromonaco (1999).
More recently, in a longitudinal study of 16 Korean speakers learning French,
Harnois-Delpiano, Cavalla and Chevrot (2012) tested learners on their perfor-
mance in a picture-naming task in two liaison contexts, DET (determiner + noun)
and ADJ (adjective + noun) at four intervals (every six months over approximately
two years of weekly French classes). Two determiners (un a/an and trois three)
and two adjectives (grand big/great and gros fat) were used as Word 1 and six
nouns were used as Word 2 (homme man, arbre tree, ami friend, enfant child,
tudiant student and appartement apartment) Korean learners produced more
errors in DET contexts than in ADJ contexts. However, in both DET and ADJ
contexts, learners primarily produced errors that respect the pronunciation of
words as they would be produced in isolation (e.g. [#ami] instead of [.nami] for
un ami a friend; [twa#ami] instead of [twa.zami] for trois amis three friends)
and spelling-like errors that respect the orthographic form of the target words
instead of the underlying form of the LC (e.g. [gradami] instead of [gratami] for
grand ami great friend; [grosami] instead of [grozami] for gros ami fat friend) as
was observed by Tomas (2004). Furthermore, learners avoid substitution errors
observed in children such as [zami], [twanami]. Across the four time intervals,
omissions of liaison decreased signifcantly, while the rate of substitutions and
spelling-related errors remained constant. Tese authors hypothesize that the L2
errors were due mainly to the fact that L2 learners are able to activate an ortho-
graphic representation of a word while L1 learners cannot. Indeed, the patterns
of errors corroborate this hypothesis and largely support Mastromonacos and
Tomas fndings. Te sole diference from the errors observed by Tomas (2004)
is that a lack of resyllabifcation of the LC was not observed in the Korean learners,
possibly due to L1 infuence, which we discuss further below.
Finally, there is even less research on the production of optional liaisons in
L2 French (for an exception see Howard, this issue). Data suggest that in optional
liaison contexts, many L2 speakers avoid liaison altogether. Tomas (2004) inter-
prets this result as a low-risk strategy employed by learners faced with contexts
where the realization of liaison is variable and stylistically determined, and for
which there is not a clear point of reference in native speech as to what constitutes
the correct form. In a study of Irish university students studying French, Howard
4. Tomas considers a lack of resyllabifcation of an LC as a production error, even though this
production pattern is observed in formal discourse situations in native speakers, as noted above.
Te acquisition of French liaison by native and non-native speakers 169
(2004) similarly noted that learners produced optional liaison in semi-guided
conversation in 8.2% of possible contexts, which is signifcantly lower than native
speaker productions of optional liaison with realization rates ranging from 14% to
44% in the Boula de Mareil et al. (2003) corpus discussed above.
3.3 Comparison of L1 and L2 production data
Two major points of divergence emerge between L1 and L2 learners in the produc-
tion of liaison. First, the behavior of the two groups of learners does not exhibit the
same homogeneity. Current L1 data suggest that the observed stages of acquisition
are homogenous and common to all children, whereas L2 acquisition shows a con-
siderable amount of inter-subject variability, which does not necessarily correlate
with the subjects fuency, morphosyntactic competence or years of study (Matter
1986; Dejean de la Btie 1993; Tomas 2004).
Based on data from Korean learners of L2 French presented in Harnois-
Delpiano et al. (2012), Chevrot et al. (2013) suggest that learners liaison produc-
tions are actually more stable than those of children learning L1 French. Te au-
thors base this conclusion on the fact that over 25% of liaison errors in children in
this study were produced afer correct liaison productions, while the proportion
of errors produced by L2 learners afer initial correct productions was only 12%.
Tey take this as evidence that once the L2 learners produced correct instances
of liaison, the representations were already stable, whereas the childrens produc-
tions remained in a state of fux throughout development. While this conclusion is
Table 1. Liaison errors observed in L1 and L2 speech.
L1 Production Errors
(Wauquier 2009, 2010)
L2 Production Errors
(Tomas 1994; Harnois-Delpiano et al. 2012)
Substitution of
wrong LC
[le.nan] for [le.zan]
les nes the donkeys
[zwazo] for [nwazo]
un oiseau, a bird
Use of the graphic C
instead of underly-
ing LC
[gdami] for [gtami]
un grand ami a great
friend
[osami] for [ozami]
gros ami fat friend
Epenthesis of LC
to resolve hiatus
[pa.pa.tus] for [pa.
pa.us]
Papa Ours Papa bear
Lack of resyllabif-
cation of the LC
[n.m] for [.nm]
un homme a man
LC not realized [.e.le.f] for [.ne.le.f]
un lphant an elephant
LC not realized [.wa.zo]for [.nwa.zo]
un oiseau a bird
Reinterpretation
of onset C as LC
[le.s.t] for [le.st.n]
les sept nains the seven
dwarfs
170 Sophie Wauquier and Ellenor Shoemaker
feasible, it is difcult to reconcile this hypothesis with data from previous studies
showing that liaison errors persist even in advanced speakers, while L1 learners
master liaison without exception.
One other possible interpretation is that the greater stability observed in
Korean learners as compared to L1 learners in Harnois-Delpiano et al. (2012) is
linked to both adult cognitive maturity and the difculty of the task for children
relative to adult learners. Given that data cited were based on a picture-naming
task including relatively few items that was repeated four times at 6-month inter-
vals, adult L2 learners may be capable of stabilizing their responses through mem-
orization or other strategies. Tis does not necessarily entail that representations
would be stable or regular in natural speech or with unknown words. Indeed, pre-
vious research, which has concentrated primarily on more conversational (spon-
taneous) French, has suggested that learners remain insecure in their productions
of liaison even afer years of study (Mastromonaco 1999; Tomas 2004).
Harnois-Delpiano et al. (2012) also compare increases in the correct pro-
ductions of obligatory liaisons in their L2 learners with native speaker data from
Dugua (2006). In Dugua (2006), children learning L1 French produce 36% correct
obligatory liaisons at 23 years of age, which increases to 83% at the age of 56.
Korean speakers, on the other hand, increase from 53 to 70% when tested every
6 months. However, with an increase of only 17% correct productions on an easy
and not extremely challenging task, adult L2 productions may not be as stable as
those of 5-year-olds who improve to 83% correct productions of obligatory liai-
sons.
Concerning the diferent types of production errors, L1 learners and L2 learn-
ers do not make the same types of errors, suggesting that the problems they face in
the resolution of liaison are also not the same. We observe only one type of error
that is common to both L1 and L2 learners, namely the complete omission of the
LC (e.g. [.e.le.f] for [.ne.le.f] un lphant an elephant). In the L2 data, there
are two particular patterns of errors that are not observed in children learning L1
French and that may be linked to literacy: the realization of the LC as the coda of
Word 1, i.e. realization without resyllabifcation, and the realization of the LC as
it is represented orthographically (e.g. *[gdelef] for un grand lphant a big
elephant instead of [gtelef]). Indeed, these two errors in L2 speech suggest
that the internalized form of the target word in L2 learners corresponds to the
form which was explicitly learned via the orthographic representation of the word.
Furthermore, as proposed by Wauquier (2009), children initially seem to learn
chunks of multiple words, while L2 learners appear to acquire individual lexical
items, which are then strung together in continuous speech without taking into
account the particular prosody of French. In other words, L2 learners know that
words such as un, petit, and grand contain a fnal consonant which is only realized
Te acquisition of French liaison by native and non-native speakers 171
before a vowel, but the segmental content and the prosodic position of the fnal
consonant are based on the orthographic form of the individual word, rather than
on formalized abstractions or constructions that are independent of the lexical
content, implying that the LC is not encoded with a particular phonological status
and/or prosodic slot in L2 speech. To put it simply, existing data suggest that the
production challenge faced by L2 learners is initially the inverse of that faced by L1
learners: L1 learners must learn to detach individual words from one another once
referential chunks are learned, while L2 learners must learn to attach individual
lexical units to one another in the continuous speech stream.
However, it should be noted that currently available L2 production data ofer
but a partial picture of L2 development. Firstly, in all of the above-cited studies,
the learners are highly literate university students learning French primarily in a
classroom environment and who therefore have a systematic orthographic knowl-
edge of the target language. One can assume that this could have an efect on the
processing of oral language and could exacerbate the learners confusion when
faced with oral input in which the canonical forms of written French are not easily
recognized, rendering the resolution of surface opacity caused by liaison all the
more difcult.
Furthermore, apart from the data of Harnois-Delpiano et al. (2012), the source
languages in the above studies are Germanic (English, Dutch). In this family of
languages the correspondence between word and syllable boundaries is more ex-
plicit than in French and word boundaries are ofen marked by relatively salient
acoustic cues such as glottal stops or aspiration. In addition, the prevalence of
tonic accent on the frst syllable of the majority of content words in English and
Dutch (Cutler & Carter 1987; Vroomen & de Gelder 1995) renders the lef edge of
words particularly prominent for speakers of these L1s, which can also render the
particularity of resyllabifcation in French more challenging.
Finally, while production data highlight the difculties faced by learners in
acquiring and mastering French prosody and resyllabifcation processes, it should
also be noted that some production studies have revealed adult learners whose
pronunciation is deemed indistinguishable from native controls (Bongaerts, van
Summeren, Planken & Schils 1997; Birdsong 2003, 2007). Tese studies were de-
signed to test global pronunciation in L2 French and therefore did not specifcally
target liaison, however the complete absence of a foreign accent in advanced learn-
ers, while rare, leads us to conclude that L2 speakers can acquire phonological
representations that are identical, or at least equivalent, to those of native speakers.
172 Sophie Wauquier and Ellenor Shoemaker
4. Perceptual data: Finding words in the speech stream
According to Durand and Lyche (2008), 99.5% of all liaisons attested in the PFC
corpus were realized as resyllabifed word onsets. From a developmental point of
view, resyllabifcation and the misalignment of syllable and word boundaries pose
signifcant challenges for spoken word recognition processes and the recovery of
vowel-initial lexical forms in the acoustic stream. Both L1 and L2 learners must
create a repertoire of lexical reference forms based on linguistic experience and
knowledge of the extra-linguistic world. Nevertheless, if we take for example the
case of nominal forms, the learner is rarely exposed to input in spoken French in
the form of bare nouns, such as lphant [elef], even though it is this bare form
that must be recovered from the signal and retained as the form of reference. Te
input that learners receive is far more frequently in context, where learners may
hear for example un petit elephant [.p.ti.te.le.f] a small elephant, les lphants
[le.ze.le.f] the elephants, un lphant [.ne.le.f] an elephant, in efect hearing
three diferent allomorphs for the bare noun [e.le.f].
When locating words in the acoustic stream, L1 learners have recourse neither
to stored lexical information nor to a previously established phonological system,
yet they must somehow sort and segment the acoustic input based only on limited
sensory experience. Tis challenge is, of course, the same for children in all the
worlds languages, but the segmentation problem is particularly acute in the case
of French, which is a cursus language (Pulgram 1970), and where word boundar-
ies are masked not only by resyllabifcation, but also by elision (le + ami becomes
lami the friend). Tis can create both local ambiguity (un petit lphant [.p.ti.
te.le.f] a small elephant / un petit tlphone [.p.ti.te.le.fn] a small telephone,
which are homophonous until the fnal syllable) and global ambiguity in phrases
(un petit ami a boyfriend / un petit tamis a little sieve, both [.p.ti.ta.mi]) and
complete sentences (il est tout vert it is completely green / il est ouvert it is open,
both [i.le.tu.v]).
L2 learners must also manage to segment the signal and encode canonical
lexical forms from potentially ambiguous input; however, the situation is indis-
putably diferent in that their L1 lexicon is already in place and learners possess
metalinguistic knowledge, allowing for the explicit learning of the L2 lexicon from
the frst moments of exposure to the target language. Complicating the issue fur-
ther for L2 learners is the fact that many leaners of French acquire the language
in a controlled classroom environment where much of the input is written, and
therefore the obstacles of resyllabifcation and the comprehension of oral input
are intensifed.
Te acquisition of French liaison by native and non-native speakers 173
4.1 Perception in L1 French
We are not aware of any existing studies that directly address the perception of li-
aison contexts and segmentation by children acquiring L1 French. For this reason,
we focus in this section on perceptual data gathered from adult native speakers in
order to examine how liaison may be represented in the native speaker phonology.
Spoken word recognition models suggest that the processing of liaison should hin-
der access, in that multiple forms must be concurrently considered by the lexicon,
thus slowing access and recognition (see for example Luce & Pisoni 1998; Norris,
McQueen & Cutler 1995; Vroomen & de Gelder 1995, among others, for discus-
sion of competition-based word recognition). However, this supposition is not
consistent with the facility with which speaker-listeners both produce and process
liaison in spoken French. Resyllabifcation and liaison have not been shown to
incur processing costs. Wauquier-Gravelines (1996) employed a word recognition
task using locally ambiguous pairs such as un petit tlphone / un petit lphant
[.p.ti.te.le.fn / .p.ti.te.le.f]: a small telephone / elephant and un navire / un
avion [.na.vi / .na.vj]: a ship / a plane and showed that lphant, which has a
resyllabifed onset when preceded by petit, was recognized as easily as tlphone,
where syllable and word boundaries are congruent.
Spinelli et al. (2003) also probed lexical access processes using a cross-modal
priming paradigm. In globally ambiguous sentence pairs such as cest le dernier
rognon, its the last kidney, and cest le dernier oignon, its the last onion, both
[se.l.d.nje..], signifcant priming efects were found for both vowel-initial
(oignon) and consonant-initial (rognon) candidates, indicating that the ambiguity
caused by liaison and subsequent resyllabifcation did not impair the lexical activa-
tion of the vowel-initial candidate. Furthermore, responses followed the intention
of the speaker, i.e. priming efects were stronger for oignon than for rognon when
the speaker intended oignon, and vice versa. Te authors take this as evidence of
the presence of acoustic cues which signal to the listener the intended segmenta-
tion, which is further discussed below.
Although liaison does not slow lexical processing in native speakers, un-
derlying consonants that surface in liaison have been shown to be processed
diferently than fxed consonants. Using a phoneme-detection task, Wauquier-
Gravelines (1996) found that native speakers are slower to detect LCs than the
same consonant in word-initial position, suggesting that resyllabifed consonants
may have a specifc status in the phonological grammar. Nguyen et al. (2007) ex-
tended Wauquier-Gravelines fndings by including comparisons with consonants
in word-fnal and word-medial positions. Tis study also employed a phoneme-
detection task and examined the detection of /n/ and /z/ in four positions: Word
2-initial (e.g. des zros [de.ze.o] some zeros), Word 1-fnal enchane (e.g. seize
174 Sophie Wauquier and Ellenor Shoemaker
lves [s.ze.lv] sixteen students), word-medial (e.g. du raisin [dy..z] some
grape), and liaison (e.g. des crous [de.ze.ku] some nuts). Nguyen et al. (2007)
replicated the results of Wauquier-Gravelines (1996). Detection rates were lower
and responses were signifcantly slower for LCs as compared to the three other
conditions. Noteworthy as well is the fact that detection rates for Word 1-fnal
and word-medial targets were intermediate to those of Word 2-initial and LC tar-
gets. For the segment /n/, the detection rates of LCs (56%) were lower than those
for both Word 1-fnal enchane consonants (76%) and medial consonants (75%),
which were in turn lower than those for Word 2-initial consonants (87%). Te
pattern of detection rates for /z/ was in the same direction, although diferences
were less robust. Tese results suggest that though LCs undergo enchanement as
do fnal fxed consonants, they retain a status in the phonological grammar that is
distinct from that of both fxed fnal and medial consonants.
Several researchers have attributed behavioral results in the processing of liai-
son to the presence of acoustic diferentiation between consonants that surface in
liaison environments and the same consonants in fxed, lexical word-initial posi-
tion. Determining whether LCs are consistently produced with specifc acoustic
properties relative to initial consonants (ICs) has been the focus of a large body of
work. Dejean de la Btie (1993) was the frst to directly measure acoustic difer-
ences between LCs and ICs. For the segment /t/, she found that closure duration
and VOT were signifcantly shorter for LCs. Wauquier-Gravelines (1996) mea-
sured the acoustic properties of both /n/ and /t/. For the plosive /t/, Wauquier-
Gravelines results confrm those of Dejean de la Btie; however, she failed to fnd
acoustic diferences for /n/. Gaskell et al. (2002) also found durational diferences;
the segments /t/, // and /z/ were signifcantly shorter when realized in liaison
(mean 73 ms) than in word-initial position (mean 88 ms). Spinelli et al. (2003)
found signifcant durational diferences for fve consonants that surface in liaison
/n, t, , , p/. LCs were on average 17% shorter than ICs. In a sample taken from
six native speakers, Shoemaker (in press) measured durational diferences of three
segments that appear in liaison environments, /n/, /t/, and /z/, and also found that
LCs were consistently shorter than ICs for all three segments. Diferences were
most robust for /n/, followed by /t/ and then /z/.
Several studies have sought to test listeners sensitivity to and exploitation
of acoustic variation in the identifcation of word boundaries, with varying out-
comes. Gaskell et al. (2002) and Spinelli et al. (2003) explicitly suggest that native
listeners are sensitive to acoustic variation and use this variation for lexical ac-
cess. However, neither of these studies directly tested this hypothesis. Wauquier-
Gravelines (1996) found no signifcant correlation between acoustic diferences
and performance in lexical access in native speakers. Using the same recorded
stimuli as Spinelli et al. (2003) in a forced-choice identifcation task, Shoemaker
Te acquisition of French liaison by native and non-native speakers 175
and Birdsong (2008) found that listeners (native and non-native) are not able to
distinguish ambiguous pairs of phrases (e.g. both un air or un nerf [.n]) when
presented without context. However, the work of Shoemaker (in press) has fur-
ther refned the study of acoustic detail in liaison environments by showing that
when durational diferences between LCs and ICs are instrumentally exaggerated,
listeners are able to make use of the acoustic information in segmentation. In this
study, pivotal consonants in ambiguous sequences containing real and non-words
were both lengthened and shortened and presented to listeners in an AX discrimi-
nation task and a forced-choice identifcation task. Listeners were sensitive to the
instrumental manipulation of duration and identifed shortened consonants as
LCs and lengthened consonants as ICs, providing evidence that acoustic informa-
tion is indeed encoded in the phonological grammar.
While a large body of work has explored the presence and exploitation of
acoustic variation in liaison environments, only recently has research tackled the
use of multiple cues in the processing of liaison. Specifcally, the work of Tremblay
and Spinelli (2013) suggests that top-down (distributional) information may out-
weigh acoustic information in the signal. Tese authors used a visual-world eye-
tracking paradigm that compared the processing of three LCs, /n/, /t/ and /z/. Te
authors predicted that the diferential distribution of the three segments in liaison
environments would bias processing in locally ambiguous phrases (e.g. fameux
lu famous elected-one and fameux zle famous zealous-one). Te segment /t/ is
more frequent as an IC than as an LC (expected bias for consonant-initial words),
while /z/ is more frequent as an LC than as an IC (expected bias for liaison-initial
words). Te segment /n/, on the other hand, is equally as frequent as an LC and as
an IC, and therefore no bias was predicted. Te authors predictions were borne
out for /t/ and /z/, suggesting that listeners are indeed sensitive to the distribu-
tion of LCs in French. However, in the case of /n/ listeners in fact showed a bias
toward consonant-initial words. Te authors hypothesize that when distributional
information does not bias processing, listeners may prefer to segment the signal
in such a way that syllable and word boundaries are aligned, thus resulting in a
consonant-initial bias.
4.2 Perception in L2 French
Data on the perception of liaison in L2 French reveal that French is a language
that poses particular challenges for the learner in the comprehension of running
speech. Stridfeldt (2005: 14) summarizes the issue for Swedish-speaking learners
of French as follows:
5
5. Our translation from the original French.
176 Sophie Wauquier and Ellenor Shoemaker
In spoken French, the word has a relatively weak phonological status in rela-
tion to the syllable. It ofen occurs that [] syllable boundaries do not coincide
with word boundaries. Phonological processes such as schwa deletion (e muet
silent e), liaison and resyllabifcation all contribute to the misalignment of (lexi-
cal) word and syllable boundaries, which in turn creates a mismatch between the
acoustic and graphic forms of words [] For learners of French as a second lan-
guage, whose instruction has ofen been based on written forms, this difculty in
recognizing words [orally] ofen results in a traumatic experience upon their frst
visit to a French-speaking country, when they realize that they do not understand
the language they studied for years at school, or even at university.
As with L1 research, work on L2 perception has primarily focused on word recog-
nition capacities in environments in which word boundaries can be ambiguous.
Experiments have employed various experimental paradigms (e.g. forced-choice
identifcation, phoneme detection, eye-tracking) and have tested a variety of stim-
uli including both non-words (grand irbois / grand tirbois [gtibwa]) and real
words (grand ami great friend / grand tamis big sieve [gtami]) in homopho-
nous pairs, as well as real lexical items in which there is temporary (local) lexi-
cal ambiguity (fameux lan famous elan / fameux zl famous zeal [famzel] /
[famzele]).
Saunders (1988) conducted a study investigating the systematicity of compre-
hension errors in spoken French in which 45 third- and fourth-year American
university students were tested on their perception of a three-minute recording of
radio French. She found a great deal of consistency among the students regarding
which sequences were misperceived, a fact she took to be indicative of the spe-
cifc areas of perception that may pose problems for (English-speaking) learners.
Errors in the perception of resyllabifed liaison consonants were prevalent. In the
sequence il vous devient [t] ce point ncessaire it becomes so necessary for
you, where an optional liaison is realized with the fnal /t/ of devient, only six of
the 45 students heard the phoneme /t/ at all, three of whom correctly identifed it
as a liaison /t/. Seven students perceived it as word-initial /d/ and four as word-ini-
tial /l/. Te phrase on est lu [ .ne.te.ly] one is elected, also a case of optional liai-
son with /t/, generated the same misperception for over a third of the students; the
phrase was perceived as on tait lu one was read, where resyllabifed /t/ is analyzed
as word-medial. Saunders frames perceptual difculties in terms of depth of pro-
cessing. She proposes, in efect, that non-native speakers are unable to penetrate
the surface variability in French speech to hear the underlying entities (1988: 94).
Dejean de la Btie and Bradley (1995) tested both native and non-native
speakers in a word-monitoring task on the detection of /t/-initial nouns in four
stimulus conditions: a /t/-initial word preceded by a word that could potentially
trigger liaison with /t/ (e.g. grand thtre big theater); a /t/-initial word preceded
Te acquisition of French liaison by native and non-native speakers 177
by a word where liaison would not be possible (e.g. vrai thtre real theater); a
vowel-initial word preceded by liaison with /t/ (e.g. grand lphant big elephant);
and fnally a vowel-initial word preceded by a word where liaison would not be
possible (e.g. vrai lphant real elephant). Tis study showed that native listeners
are able to process the word boundary without difculty across the four condi-
tions, as indicated by error rates close to zero (1 error out of 360 responses), while
the L2 learners error rates in the two conditions that included a liaison-triggering
word (e.g. grand thtre big theater and grand lphant big elephant) were 13.8%
(L2 error rates for conditions with no possible liaison were 1.9%). Te non-native
speakers showed not only higher error rates, but far more individual variation
than was observed in the native speakers. Tis discrepancy suggests that native
speakers had stable strategies in place for the processing of ambiguous word junc-
tures, while processing strategies of L2 learners were inconsistent.
Stridfeldt (2005) tested Swedish speakers learning L2 French on the identifca-
tion of potentially ambiguous non-words (e.g. [navas] intended by the speaker
as either un navas or un avas) which incorporated four segments that surface in
liaison, /n, , t, z/. Both native and non-native participants were asked to write
what they heard and performed roughly at chance levels in identifying the word
intended by the speaker. Furthermore, non-native participants showed an over-
all tendency to identify the sequences as instances of liaison (un avas). A native
speaker control group also showed a slight bias toward perceiving the stimuli as
vowel-initial, but the bias was not as robust as for the learners. Te learners bias
toward vowel-initial words suggests that they had developed processing routines
and mental representations of liaison consonants, but that these representations
were overgeneralized to all possible instances of liaison. Shoemaker and Birdsong
(2008) also tested highly advanced English-speaking learners of French in a forced-
choice identifcation task using ambiguous minimal pairs (e.g. grand ami / grand
tamis, both [gtami]). Te results show that neither learners nor native speakers
could reliably distinguish the ambiguous pairs.
Concerning the exploitation of diferences in segmental duration in liaison
environments, Shoemaker (2010) directly tested the use of acoustic information
in the processing of liaison and showed that advanced learners can use durational
diferences in the localization of word boundaries in manipulated stimuli when the
diferences between LCs and ICs are sufciently large. Tis study employed am-
biguous pairs of phrases in which the duration of the pivotal consonant had been
instrumentally shortened or lengthened. Learners were able to use this enhanced
acoustic information in locating word boundaries in a forced-choice identifca-
tion task. In fact, several of the advanced learners in this study showed perceptual
sensitivity equal to that of native speakers. It should be noted though that both
native and non-native speakers showed a great deal of individual variation among
178 Sophie Wauquier and Ellenor Shoemaker
participants, suggesting that durational diferences in liaison may not represent a
consistently robust processing cue in natural speech.
More recent research has employed online measures of processing to examine
recognition strategies used by learners of L2 French. Using eye-tracking, Tremblay
(2011) demonstrated that in real-word pairs (e.g. fameux zl / fameux lan, )
and non-word pairs (e.g. fameux anage / fameux zanon) where there is temporary
lexical ambiguity, L2 learners recognized vowel-initial words (lan) in liaison situ-
ations more quickly than /z/-initial words (zl), demonstrating that learners had
developed processing strategies for recovering vowel-initial words from resyllabi-
fed environments. Tremblay suggests that learners did not use low-level acoustic
information to distinguish LCs and ICs, but rather that results were conditioned
by distributional frequency; words beginning with /z/ in French are highly infre-
quent, which could feasibly bias learners toward vowel-initial words in this par-
ticular liaison context. Furthermore, as Tremblay points out, the fact that learners
in this study exhibited evidence of parsing procedures for nonce words as well as
for real words ofers evidence of the acquisition of phonological abstraction as op-
posed to a strictly lexical strategy. However, it is also worth noting that the native
speaker control group in this study did not show the same processing strategy,
recognizing /z/-initial items more readily than vowel-initial items, suggesting that
even though the L2 learners have developed efcient processing strategies, they
are behaving diferently from native speakers.
Tremblay and Spinelli (in press) furthered this line of analysis by testing
English-speaking learners of French on the recognition of words in liaison en-
vironments using both /t/ and /z/. Te authors again hypothesized that learners
would be sensitive to the relative distribution of the two liaison consonants in
spoken French; /t/ is more commonly found in word-initial position than in liai-
son environments, while /z/ is more common in liaison environments than as a
lexical word-initial consonant. Given this diferential distribution, Tremblay and
Spinelli predicted that learners would fxate more readily and more ofen on a /t/-
initial sequence in pairs such as parfait abri/parfait tableau, perfect shelter/perfect
painting, but that conversely learners would fxate more readily on a vowel-initial
sequence with pairs incorporating /z/, such as curieux rable/ curieux zro, strange
maple/strange zero. Te authors predictions were borne out for stimuli incor-
porating /t/, but not /z/, which is somewhat surprising given that the diferential
distribution is much more pronounced for /z/ than for /t/. To account for this
discrepancy, the authors suggest that learners may prefer to parse the signal in
such a way that word and syllable boundaries are aligned, as was hypothesized
for native speaker data concerning the LC /n/ in Tremblay and Spinelli (2013).
Learners in this study were also more sensitive to the acoustic information in the
signal than native speakers, which the authors attribute to L1 infuence. Acoustic
Te acquisition of French liaison by native and non-native speakers 179
diferentiation at word boundaries tends to be more salient in English than in
French, and therefore the authors propose that learners were more sensitive to
phonetic detail than native speakers.
4.3 Convergence and divergence in L1 and L2 perception data
Te bulk of research on the perception of liaison has found that non-native
learners of French perform diferently than native speakers on the same behav-
ioral tasks (Dejean de la Btie & Bradley 1995; Stridfeldt 2005; Tremblay 2011;
Tremblay & Spinelli, in press) suggesting that processing strategies difer between
the two groups, and/or that L2 processing strategies are not as stable as those of
native speakers. However, the data also suggest that even though L2 strategies may
not be identical to those of native speakers, and even though L2 learners may give
preference to diferent sources of information in the signal than native speakers,
learner strategies are nonetheless efcient and allow successful processing of the
signal. Furthermore, some research has found that advanced learners can reach
native-like levels of performance in processing (Shoemaker 2010).
Concerning the exploitation of acoustic cues to liaison, the ensemble of results
suggests that neither native nor non-native listeners rely principally on acoustic
information in the resolution of word boundaries. Te data seem rather to support
a dynamic model in which listeners simultaneously exploit acoustic, phonologi-
cal, lexical, and post-lexical information, using the various sources of information
alternately and relative to respective availability (Nguyen et al. 2007). Tis hypoth-
esis is in line with the work of Shoemaker (2010) who showed that both native
and non-native speakers can directly exploit acoustic information, but only when
it is exaggerated in the signal, implying that other information is more robust.
Similarly, Dejean de la Batie and Bradley (1995) found that native speakers were
faster to detect /t/-initial words when presented in context than with no context,
suggesting that top-down information played a greater role than acoustic informa-
tion alone. Te work of Tremblay and Spinelli (2013, in press), who showed that
the distribution of LCs and ICs in spoken French can bias listeners segmentations
of potentially ambiguous speech, and that this information can outweigh acous-
tic information, also supports a model of speech processing in which multiple
sources of information are available to the listener.
All of these studies are in line with the work of Mattys, White and Melhorn
(2005) and of Mattys, Melhorn and White (2007) who explored the weighting
of multiple segmentation cues in an attempt to formulate a hierarchy of cues
based upon the saliency of each individual cue in speech processing. Based on
data concerning the processing of English, these authors showed that when all
cues are optimally available, knowledge-driven cues (lexical, syntactic, semantic,
180 Sophie Wauquier and Ellenor Shoemaker
and contextual cues) take precedence, while signal-driven cues (phonotactic and
acoustic-phonetic cues) gain prominence when top-down information is un-
available, impoverished, or ambiguous (2005: 487).
One particular question that emerges from this line of inquiry is whether the
weighting and hierarchy of cues is diferent for frst and second language learners.
Tremblay and Spinelli (in press) showed that both native and non-native listeners
use acoustic information, but their data also suggest that L2 learners rely more
heavily on acoustic information than native speakers in the processing of liaison, a
result which the authors attribute to the learners native language. Tis work sug-
gests that learners may weigh cues diferently than native speakers; however, fur-
ther research including learners from diferent source languages would be needed
to examine the extent to which the hierarchy of processing cues varies, both across
learners/languages and relative to native speakers.
Again, it should be noted that the L1 perceptual data reviewed here come from
adult speakers. Gathering perceptual data from children learning French therefore
constitutes a further challenge in this domain in that children do not necessarily
have the cognitive capacity required to understand and perform the tasks involved.
Tis fact signifcantly limits the conclusions that can be drawn in comparing the
development of native and non-native processing strategies.
5. L1 and L2 developmental strategies: Unanswered questions and further
research
In this section, we discuss several questions that remain unanswered in the ac-
quisition of liaison, as well as possible further avenues for research. First, data
are particularly lacking in the investigation of what L2 learners do in a learning
situation in which the orthographic system is not studied simultaneously. All of
the L2 learners included in this review were literate and in most cases highly edu-
cated, and therefore had access to orthographic support throughout their L2 de-
velopment. Only in studying adult learners whose development is based solely
(or primarily) on oral input can we begin to form a comprehensive picture of the
diferences between L1 and L2 acquisition of liaison. An L2 learning environment
that more closely mirrors that of L1 learners would be highly informative in de-
termining how learners who do not have recourse to orthographic representations
learn to segment and classify LCs. Tis type of learning situation can give valuable
insight as to whether the errors in L2 speech observed in the existing literature
are indeed infuenced by orthographic knowledge of French, as researchers have
proposed.
Te acquisition of French liaison by native and non-native speakers 181
Lyche (2013) directly addresses the role of literacy in the production of liaison
in examining data of Louisiana French gathered from two sources: 19th-century
letters written by semi-literate writers and contemporary oral data from Cajun
French speakers who are literate in English but not in French (the contemporary
data was originally published in Dubois 1997). Both the written and oral data
show various segmentation errors such as the use of an LC as the onset of Word 2
(e.g. une taction instead of une action, an action; une zoreille instead of une oreille,
an ear) and the reinterpretation of a lexical-word initial consonant as an LC (e.g.
ombril instead of nombril navel), production errors that have been observed in
25 year olds learning L1 French. Certain forms are even codifed in the modern
Dictionary of Louisiana French (Valdman & Rottet 2010), in which there are 29
entries under z with a resyllabifed liaison /z/ (e.g. une zafaire and une zoreille),
and two entries under n (e.g. le nanglais). Tese data underscore the difculties
faced by both children learning L1 French and semi-literate speakers and dem-
onstrate that the strategies of the two groups of learners to resolve difculties are
analogous. Tese data also imply that the learning of writing and orthography
constitutes a crucial stage in both the stabilization of canonical French forms and
in the resolution of segmentation problems in French, as has also been shown by
Soum (1987). Other French-based Creole languages also shed light on how spoken
French is processed in a second language environment based primarily on oral
input. Many vowel-initial words from the lexifer language were retained with the
resyllabifed liaison consonant as the lexical word onset (e.g. Haitian Creole, nonm
man from un homme or zwazo bird from les/des oiseaux), suggesting processing
strategies similar to those observed in L1 acquisition.
One further fundamental question that merits more thorough exploration is
L1 transfer and the exact role of the L1 phonological flter. Tere are no existing
data that specifcally examine the extent to which a learners L1 determines the
L2 developmental path in the acquisition of French liaison. With the exception
of Harnois-Delpiano et al. (2012), research on the acquisition of liaison has been
undertaken with native speakers of Germanic languages: Dutch (Matter 1986),
English (Dejean de la Btie 1993; Dejean de la Btie & Bradley 1995; Mastromonaco
1999; Tomas 2004; Shoemaker 2010; Tremblay 2011), and Swedish (Stridfeldt
2005), languages that share prosodic features concerning word boundaries, as well
as relatively robust acoustic cues to word boundaries. It would therefore be worth-
while to study learners of L2 French from source languages that systematically em-
ploy resyllabifcation or have less consistent acoustic marking at word boundaries
(for example, other Romance languages or Slavic languages).
In particular, the status of coda consonants and phonotactic constraints on
consonants in coda position in the L1 could serve as a phonological flter con-
straining the types of possible errors in L2 speech. In this respect, the Korean data
182 Sophie Wauquier and Ellenor Shoemaker
cited above merit closer examination. As mentioned in 3.2, Harnois-Delpiano et
al. (2012) did not observe a lack of resyllabifcation of the LC in their Korean
participants, an error which has frequently been observed in English-speaking
learners of French (Tomas 2004). English and Korean difer as to which VC se-
quences are allowed in word-fnal position (Kang 2003; Carvalho, to appear). It
would therefore be worthwhile to investigate whether these constraints constitute
a phonological flter of sorts which infuences not only the perception and pro-
cessing of LCs, but also the strategies used in production by the speakers of the
two languages (Peperkamp 2005).
Te question also remains as to whether liaison is represented in the L2 pho-
nological grammar as it is in native speakers. Wauquier (2009) proposes that L1
learners employ a strategy that is based on the establishment of abstract represen-
tations of linking consonants, while L2 learners acquire liaison on a case-by-case
basis, employing a lexical strategy based on surface (orthographic) representations
of words in liaison environments, and ultimately may fail to internalize an abstract
representation of liaison constructions. Put diferently, Wauquier proposes that
the two groups do not proceed in the same manner to arrive at the same gener-
alization: L1 learners form abstractions based on auditory input in which the LC
surfaces as the onset of Word 2 and then create a specifc prosodic position for
the LC (e.g. *nlphant [nelef]), while L2 learners are faced with the challenge of
creating the same generalization based on the orthographic lexical form, in which
the LC belongs lexically to Word 1 but phonologically and prosodically to Word
2. She further proposes that the persistence of inconsistent and sporadic errors in
L2 learners, even at advanced profciency, suggests that L2 speakers proceed by
adjusting processing on a case-by-case basis conditioned by the learning of new
words in new contexts.
While it is certainly feasible that many L2 learners fail to internalize a formal-
ized abstraction of LCs that is comparable to that of native speakers, as Wauquier
(2009) suggests, much L2 data bring into question such a cut-and-dried claim.
Tremblay (2011) presents data from learners who show evidence of generalizing
processing strategies to non-words as well as real lexical items, suggesting the pres-
ence of abstract representations that are separate from their lexical content, and
therefore providing evidence against a learning strategy in which liaison is learned
on a case-by-case basis. Furthermore, the work of Tremblay and Spinelli (in press)
suggests that learners are sensitive to the distribution of LCs in spoken French from
very early on in their development, suggesting that learners are not dependent on
the surface forms of words alone, but rather that they track multiple sources of
information in the acoustic signal. Native-like performance on the part of L2 learn-
ers in the processing of liaison (e.g. Shoemaker 2010) can also be interpreted as
evidence for the acquisition of stable (and complete) representations of liaison in
Te acquisition of French liaison by native and non-native speakers 183
highly advanced learners. Consequently, a model in which L1 and L2 learner be-
havior can be placed along a developmental continuum, rather than one which
proposes a strict division between L1 and L2 processing, appears to be more com-
patible with existing data. Such a continuum may privilege phonological strategies
at the L1 end of the spectrum and lexical strategies at the L2 end of the spectrum.
One further domain of research is underexplored. As we have seen, the vast
majority of psycholinguistic investigation deals with contexts of obligatory liaison.
Subsequently, the sociolinguistic aspects involved in the acquisition of optional
liaisons have been given little attention in native speakers (Nardy 2008; Nardy,
Chevrot & Barbu 2013; Barbu et al. 2013) and have been virtually ignored in L2
learners (see however Howard, this issue). In the study of L1 learners of French,
Nardy and colleagues examined the productions and grammaticality judgments
of 185 children in four age groups (23, 34, 45, 56 years of age) and found
that children show sensitivity to errors in obligatory liaison contexts earlier than
to errors in optional contexts. Te authors interpret these data as showing that
sensitivity to linguistic norms is based on frequency of input in that obligatory
liaisons are far more frequent than optional liaisons. Te acquisition of writing
coupled with the reinforcement of metalinguistic awareness and linguistic norms
in formalized education infuence sociolinguistic judgments in school-aged chil-
dren. Te authors further dispute the chronology ofered by Wauquier-Gravelines
(2005), who considers that the acquisition of liaison in obligatory contexts pre-
cedes the acquisition of optional liaisons and that sensitivity to variation is linked
to the development of pragmatic competence (see 3.1 above). However, further L1
data are necessary to evaluate more closely childrens sensitivity to sociolinguistic
norms, given that only two items were tested, gros fat and petit small, in a single
optional context (pre-posed adjective). Furthermore, one of the items, petit, has
shown realization rates of above 70% in adults (72% in Boula de Mareil et al.
2003; 75% in Mallet 2008) and has been classifed as invariable by Durand and
Lyche (2008) and Durand et al. (2011). Concerning learners of L2 French, there is
very little work in this particular domain, but as Tomas (2004) notes, L2 learners
appear to adopt a strategy of avoidance when faced with uncertainty in optional
liaison contexts.
Further research is also warranted in the investigation of the extent to which
the experimental task infuences participant performance. Given that current re-
search has employed a variety of both online and ofine tasks (picture-naming,
acceptability judgments, discrimination, identifcation, priming, eye-tracking,
among others), a further avenue for research could be to investigate the relation-
ship between the nature of the task and the structure of the source language, as
well as the efect of this relationship on the implementation of strategies giving
precedence to diferent sources of information (e.g. top-down or bottom-up).
184 Sophie Wauquier and Ellenor Shoemaker
Data are also lacking in the establishment of a stable and consistent chronol-
ogy in L2 learners. L1 data suggest that children follow a similar developmental
path and produce comparable errors at roughly the same point in development.
However, given the inherently idiosyncratic nature of L2 development compared
to L1 development, such a consistent developmental path is lacking in existing L2
data. Further research which, in particular, includes additional longitudinal study
of L2 learners from diferent source languages, could give some insight as to the
chronology of the L2 acquisition of liaison.
We are also not aware of any studies which examine the development of liai-
son in children growing up bilingual, nor are we aware of any work which inves-
tigates the role of the acquisition of writing in the stabilization of phonological
knowledge and in the development of segmentation strategies. Both of these topics
could also be productive in ofering comparative data. Data on the interface of
production and perception from the same participants would also be extremely
informative in both L1 and L2 acquisition.
6. Conclusion
Although liaison has been an object of study for linguists and philologists for cen-
turies (Palsgrave 1530), psycholinguistic research in this domain has been under-
taken only in the past 15 years. In this article, our frst objective was to give a state
of the art review of data gathered from production and perception studies on the
acquisition of liaison. Based on points of convergence and divergence in the ex-
isting data, we evaluated the extent to which these data allow us to explicate (1)
the linguistic performance of native speakers in the production and perception
of liaison and (2) developmental strategies employed by L1 and L2 learners. Our
second objective was to identify areas of research where data are currently lacking
and which could be fruitful in future research.
In sum, we observe that L1 and L2 production data converge in the contexts
in which production errors occur. Errors are observed primarily in nominal con-
texts that are considered to be obligatory (e.g. determiner + noun). In perception,
L1 and L2 data further converge in showing that neither L1 nor L2 learners base
their processing of the acoustic signal on low-level information alone. Behavioral
data for both frst and second language learners indicate that processing is based
on multiple sources of information from both bottom-up and top-down sources,
including acoustic-phonetic variation, lexical frequency, and the distribution of
LCs, among others.
Data diverge in three main areas. In production, although the contexts where
errors are committed are largely the same, L1 and L2 learners do not make the
Te acquisition of French liaison by native and non-native speakers 185
same types of errors. Te types of production errors observed appear to be rel-
atively homogenous within groups (i.e. all L1 learners make similar errors and
most L2 learners make similar errors), but divergent across groups (i.e. L1 learners
make diferent errors than L2 learners). Tis leads us to believe that phonologi-
cal representations also difer between L1 and L2 learners. In the development of
processing strategies, data suggest that the difering cognitive capacities of the two
groups of learners may account for a great deal of discrepancy in the data, limiting
direct comparisons between L1 and L2 data sets. Adult learners come to the table
with the L1 lexicon and metalinguistic knowledge already in place, while infants
must learn not only that words exist but also how to extract individual lexical
items from continuous speech.
Afer reviewing the existing data, we observe that, despite the vast amount
of work in this domain, liaison serves to prove Labovs (1972: 98) paradox, [t]he
more that is known about a language, the more we can fnd out. Tus, more re-
search is needed on diferent learner populations and diferent learning situations
in diferent types of interactions that have not been previously explored. Moreover,
learners from varying source languages and ages, in learning environments other
than a controlled classroom environment, should be studied in order to address
issues that have been underexplored: the role of typology, the role of literacy, and
the role of sociolinguistics in the acquisition and processing of liaison.
References
Agren, J. (1973). Enqute sur quelques liaisons facultatives dans le franais de conversation ra-
diophonique. Acta Universitatis Upsaliensis. Studia Romanica Upsaliensia 10.
Barbu, S., Nardy, A., Chevrot, J.-P. & Juhel, J. (2013). Language evaluation and use during early
childhood: Adhesion to social norms or environmental regularities? Linguistics 51, 381
412.
Bernicot, J. (2000). La pragmatique des noncs chez lenfant. In M. Kail & M. Fayol (eds.),
Lacquisition du langage, vol.2 : le langage en dveloppement, au-del de trois ans (4582).
Paris: PUF.
Birdsong, D. (2003). Authenticit de prononciation en franais L2 chez des apprenants tar-
difs anglophones: Analyses segmentales et globales. Acquisition et Interaction en Langue
Etrangre 18, 1736.
Birdsong, D. (2007). Nativelike pronunciation among late learners of French as a second lan-
guage. In O.-S. Bohn & M. Munro (eds.), Second language speech learning: Te role of lan-
guage experience in speech perception and production. Amsterdam: John Benjamins.
Bongaerts, T., van Summeren, C., Planken, B. & Schils, E. (1997). Age and ultimate attainment
in the pronunciation of a foreign language. Studies in Second Language Acquisition 19,
447465.
186 Sophie Wauquier and Ellenor Shoemaker
Boula de Mareil, P., Adda-Decker, M. & Gendner, V. (2003). Liaisons in French: A corpus based
study using morpho-syntactic information. Proceedings of ICPhS, Barcelona, 13291332.
Braud, V. (1998). Acquisition de l information phonologique : exemple de la liaison, Unpublished
Master thesis, University of Nantes.
Bybee, J. (2001). Phonology and language use. Cambridge: Cambridge University Press.
de Carvalho, J. Brandao. (To appear). C/V interactions in strict CV. In M. Lahrouchi, N.
Lampitelli, N. Faust & S. Bendjaballah (eds.), Te form of structure, the structure of forms:
Essays on the realization of linguistic structures. Language faculty and beyond. John
Benjamins.
Chevrot, J.-P. & Fayol, M. (2000). Lacquisition de la liaison : enjeux thoriques, premiers
rsultats, perspectives. Lidil 22, 1130.
Chevrot, J.-P. & Fayol, M. (2001). Acquisition of French liaison and related child errors. In M.
Almgren, A. Barrena, M.J. Ezeizabarrena, I. Idiazabal & B. MacWhinney (eds.), Research on
child language acquisition, vol. 2 (760774). Somerville MA: Cascadilla Press
Chevrot, J.-P., Dugua, C. & Fayol, M. (2005). Liaison et formation des mots en franais : un
scnario developpemental. Langages 158, 3852.
Chevrot, J.-P., Dugua, C. & Fayol, M. (2009). Liaison acquisition, word segmentation and con-
struction in French: a usage-based account. Journal of Child Language 36, 557596.
Chevrot, J.-P., Dugua, C., Harnois-Delpiano, M., Siccardi, A. & Spinelli, E. (2013). Liaison acqui-
sition: Debates, critical issues, future research. Language Sciences 39, 8394.
Chomsky, N. & Halle, M. (1968). Te sound pattern of English. New York: Harper and Row.
Clements, G. & Keyser S. (1983). CV phonology: A generative theory of the syllable. Cambridge,
Mass: MIT Press.
Cot, M.-H. (2005). Le statut lexical des consonnes de liaison. Langages 158, 6679.
Cot, M.-H. (2011). French Liaison. In M. van Oostendorp, C. Ewen, E. Hume & K. Rice (eds.),
Companion to phonology (26852710). Malden, MA: Wiley-Blackwell.
Cutler, A. & Carter, D. (1987). Te predominance of strong initial syllables in the English vo-
cabulary. Computer Speech and Language 2, 133142.
Dejean de La Btie, B. (1993). Word boundary ambiguity in spoken French. Unpublished Doctoral
Dissertation, Monash University, Victoria, Australia.
Dejean de la Btie, B. & Bradley, D. (1995). Resolving word boundaries in spoken French: Native
and non-native strategies. Applied Psycholinguistics 16, 5981.
Delattre, P. (1947). La liaison en franais, tendances et classifcation. Te French Review XXI,
148147.
Dubois, S. (1997). Field Method in Four Cajun Communities in Louisiana. In Albert Valdman
(ed.), French and Creole in Louisiana (4769). New York: Plenum Press.
Dugua, C. (2006). Liaison et segmentation lexicale et schmas syntaxiques entre 2 et 6 ans : un
modle dveloppemental bas sur lusage. Unpublished doctoral dissertation, University of
Grenoble III.
Durand, J., Laks B. & Lyche C. (2002). La phonologie du franais contemporain : Usages, varits
et structure. In C. Pusch & W. Raible (eds.), Romanistische Korpuslinguistik- Korpora und
gesprochene Sprache/Romance corpus linguistics corpora and spoken language (93106).
Tbingen: Gunter Narr Verlag.
Durand, J., Laks B. & Lyche C. (2009). Le projet PFC : une source de donnes primaires struc-
tures. In J. Durand, B. Laks & C. Lyche (eds.), Phonologie, variation et accents du franais
(1961). Paris: Herms.
Te acquisition of French liaison by native and non-native speakers 187
Durand, J. & Lyche, C. (2008). French liaison in the light of corpus data. Journal of French
Language Studies 18, 3366.
Durand, J., Laks, B., Calderone, B. & Tchobanov, A. (2011). Que savons-nous de la liaison au-
jourdhui ? Langue Francaise 169, 103126.
Encrev, P. (1988). La liaison avec et sans enchanement, phonologie tridimensionnelle et usages
du franais. Paris: Le Seuil.
Gaskell, G., Spinelli, E., & Meunier, F. (2002). Perception of resyllabifcation in French. Memory
and Cognition 30, 798810.
Hannahs, S. (2007). French phonology and L2 acquisition. In D. Ayoun (ed.), French Applied
Linguistics (5074). Amsterdam: John Benjamins.
Harnois-Delpiano, M., Cavalla, C. & Chevrot, J.-P. (2012). Lacquisition de la liaison en L2 :
tude longitudinale chez des apprenants corens de FLE et comparaison avec enfants fran-
cophones natifs. In F. Neveu, V. Muni Toke, P. Blumenthal, T. Klingler, P. Ligas, S. Prvost
& S. Teston-Bonnard (eds.), SHS Web of Conferences Actes du 3me Congrs Mondial de
Linguistique Franaise, Lyon, France (15751589).
Howard, M. (2004). Sociolinguistic variation and second language acquisition: A preliminary
study of advanced learners of French. Sky Journal of Linguistics 17, 143165.
Howard, M. (Tis issue). La liaison en franais langue seconde : une tude longitudinale prlimi-
naire. Language, Interaction, and Acquisition 4(2), 189230.
Kang, Y. (2003). Perceptual similarity in loanword adaptation: English post-vocalic word-fnal
stops in Korean. Phonology 20, 219273.
Labov, W. (1972). Some principles in linguistic methodology. Language in Society 1, 97120.
Luce, P. & Pisoni, D. (1998). Recognizing spoken words: Te neighborhood activation model.
Ear and Hearing 19, 136.
Lyche, C. (2013). Liaison, word formation and illiteracy: Te case of Louisiana French. Workshop
on Liaison, University of Tromso, June 28th 2013.
Mallet, G. (2008). La Liaison en franais : descriptions et analyses dans le corpus PFC. Unpublished
doctoral dissertation, University of Paris-Ouest, Nanterre, La Dfense.
Mastromonaco, S. (1999). Liaison in French as a second language. Unpublished doctoral disser-
tation, University of Toronto.
Matter, J.-F. (1986). A la recherche des frontires perdues : Etude sur la perception de la parole en
franais. Doctoral dissertation, University of Utrecht, Te Netherlands. Amsterdam: De
Werelt.
Mattys, S., White, L. & Melhorn, J. (2005). Integration of multiple speech segmentation cues: A
hierarchical framework. Journal of Experimental Psychology: General 134, 477500.
Mattys, S., Melhorn, J. & White, L. (2007). Efects of syntactic expectations on speech segmenta-
tion. Journal of Experimental Psychology: Human Perception and Performance 33, 960977.
Morin, S. & Kaye, J. (1982). Te syntactic bases for French liaison. Journal of Linguistics 18,
291330.
Nardy, A. (2008). Acquisition des variables sociolinguistiques entre 2 et 6 ans : facteurs sociolo-
giques et infuences des interactions au sein du rseau social. Unpublished doctoral disserta-
tion, University of Grenoble III
Nardy, A., Chevrot, J.-P. & Barbu, S. (2013). Te acquisition of sociolinguistic variation: Looking
back and thinking ahead. Linguistics 51, 255284.
Nguyen, N., Wauquier-Gravelines, S., Lancia, L. & Tuller, B. (2007). Detection of liaison conso-
nants in speech processing in French, experimental data and theoretical implications. In P.
188 Sophie Wauquier and Ellenor Shoemaker
Prieto, J. Mascaro & M.-J. Sol (eds.), Segmental and prosodic issues in Romance phonology
(325). Current Issues in Linguistic Teory 282. Amsterdam: John Benjamins.
Norris, D., McQueen, J. & Cutler, A. (1995). Competition and segmentation in spoken word
recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition 21,
12091228.
Palsgrave, J. (1530). L claircissement de la langue franaise. Translation S. Baddeley, 2003. Paris:
Honor Champion.
Peperkamp, S. (2005). A psycholinguistic theory of loanword adaptation. In M. Ettlinger, N.
Fleischer & M. Park-Doob (eds.), Proceedings of the 30th Annual Meeting of the Berkeley
Linguistics Society (341352). Berkeley, CA: Te Society.
Pulgram, E. (1970). Syllable, word, nexus, cursus. Te Hague: Mouton.
Saunders, G. (1988). Te structure of errors in the perception of French speech. Revue de
Phontique Applique 86, 43100.
Schane, S. (1968). French phonology and morphology. Cambridge, MA: Te MIT Press.
Shoemaker, E. (2010). Te exploitation of fne phonetic detail in the processing of L2 French.
In B. VanPatten & Jill Jegerski (eds.), Research in second language processing and parsing
(259279). Amsterdam: John Benjamins.
Shoemaker, E. (In press). Durational cues to word recognition in spoken French. Applied psy-
cholinguistics. Available online at: http://journals.cambridge.org/action/displayAbstract?fr
omPage=online&aid=8756297.
Shoemaker, E. & Birdsong, D. (2008). La rsolution de la liaison par des locuteurs natifs et non-
natifs. Acquisition et Interaction en Langue Etrangre 27, 4362.
Soum, C. (1987). Lapprentissage de l criture ; contraintes orthographiques ; contraintes orales.
Unpublished doctoral dissertation, University of Toulouse 2.
Spinelli, E., McQueen J.G. & Cutler, A. (2003). Processing resyllabifed words in French. Journal
of Memory and Language 48, 233254.
Stridfeldt, M. (2005). La perception du franais oral par des apprenants sudois. Unpublished
doctoral dissertation, Ume University, Sweden.
Tomas, A. (2004). Phonetic norm versus usage in advanced French as a second language.
International Review of Applied Linguistics 42, 365382.
Tranel, B. (1995). Current issues in French phonology: Liaison and position theories, in In
J. Goldsmith (ed.), Te handbook of phonological theory (798816). Cambridge, MA:
Blackwell.
Tranel, B. (2000). Aspects de la phonologie du franais et la thorie de loptimalit. Langue
franaise 126, 3972.
Tremblay, A. (2011). Learning to parse liaison-initial words: An eye-tracking study. Bilingualism:
Language and Cognition 14, 257279.
Tremblay, A. & Spinelli, E. (2013). Segmenting liaison-initial words: Te role of predictive de-
pendencies. Language and Cognitive Processes 28, 10931113.
Tremblay, A. & Spinelli, E. (In press). English listeners use of distributional and acoustic-pho-
netic cues to liaison in French: Evidence from eye movements. Language and Speech.
Valdman, A. & Rottet, K. (eds.) (2010). Te dictionary of Louisiana French: As spoken in Cajun,
Creole, and American Indian Communities. Jackson: University Press of Mississippi.
Vroomen, J. & de Gelder, B. (1995). Metrical segmentation and lexical inhibition in spoken word
recognition. Journal of Experimental Psychology: Human Perception and Performance 21,
98108.
Te acquisition of French liaison by native and non-native speakers 189
Wauquier, S. (2009). Acquisition de la liaison en L1 et L2 : stratgies phonologiques ou lexicales ?
AileLia 2, 93130.
Wauquier, S. (2010). Acquisition de la phonologie du franais, des usages la structure. In
M. Barra (ed.), Le(s) franais : formaliser la variation , Langue franaise 168 (127144).
Wauquier-Gravelines, S. (1996). Organisation phonologique et traitement de la parole continue.
Unpublished doctoral dissertation, University of Paris 7.
Wauquier-Gravelines, S. (2005). Statut des reprsentations phonologiques en acquisition, traite-
ment de la parole continue et dysphasie dveloppemental. Unpublished HDR dissertation,
EHESS, Paris.
Wauquier-Gravelines, S. & Braud, V. (2005). Proto-dterminant et acquisition de la liaison obli-
gatoire en franais. Langages 158, 5365.
Rsum
Cet article prsente une synthse rendant compte de ltat actuel des recherches sur lacquisition
de la liaison en langue maternelle et en langue seconde (respectivement L1 et L2). Nous pr-
sentons lensemble des rsultats des tudes en production et les travaux en psycholinguistique
exprimentale sans exclusivit thorique. Notre but est de mettre en lumire les similarits et les
difrences observables chez ces deux groupes dapprenants tant en termes de stratgies dap-
prentissage que de dcours dveloppementaux et dexaminer tout particulirement comment les
reprsentations de la liaison se mettent en place dans la composante phonologique de la gram-
maire de lapprenant et se modifent ventuellement sous linfuence de diverses dynamiques de
dveloppement. Pour conclure, nous discutons des cas o les donnes empiriques sont encore
lacunaires et ouvrons vers dautres questions de recherche.
Mots clefs : liaison, phonologie, acquisition de la phonologie en langue seconde, acquisition de
la phonologie en langue maternelle

Authors addresses
Sophie Wauquier (frst author)
Laboratoire Structures Formelles du Langage
Centre CNRS Pouchet,
59 rue Pouchet
75017 Paris
France
sophie.wauquier@orange.fr
Ellenor Shoemaker
Universit Sorbonne Nouvelle Paris 3
Dpartement Monde Anglophone
13, rue de Santeuil,
75005 Paris
France
ellenor.shoemaker@univ-paris3.fr

You might also like