Koineization in Medieval Spanish - (2. Koines and Koineization)

Chapter 2
Koines and koineization
The model of koineization, of fairly recent development, is based on earlier

metaphorical use of the term koine. We therefore begin with an overview
of the origins and modern use of the term in the linguistic literature, and of
the confusion that its varied meanings have sometimes provoked. The pri-
mary aim will then be to define, as thoroughly as possible, what koineiza-
tion is, and what it is not. Several scholars have sought to answer these
questions, though their responses do not agree in all respects, so I have
organized the bulk of this chapter as a critical review of previous discus-
sion of koineization, with the goal of synthesizing this earlier work and my
own views. Throughout, the various facets of koineization are put in rela-
tion to other theories of language use and change, but a special section
focuses on the differences between koineization and other processes with
which it may interact in real cases of change, or be confused in scholars'
discussion of change. The chapter concludes with the definition of a proto-
typical model of koineization, and the proposal of methodological guide-
lines for application of the model.
1. Koine and koines
Koine is a term with a long history and a wide variety of interpretations. It

Copyright © 2003. De Gruyter. All rights reserved.
has its origin in the name of a variety of ancient Greek that became the
common language of the eastern Mediterranean. Subsequent metaphorical
or technical use of the term has referred to a broad range of language varie-
ties that share some or all of the characteristics of the original Greek
Koine.
1.1. The Greek Koine
The κοινή (from koiné dialektos or koinè glòssa 'common tongue') was a
mixed dialect based largely on the prestigious Attic dialect of Athens.
From the middle of the fifth century B.C., when Pericles converted the
Tuten, Donald N.. Koineization in Medieval Spanish, De Gruyter, 2003. ProQuest Ebook Central,
http://ebookcentral.proquest.com/lib/qub/detail.action?docID=3042784.
Created from qub on 2017-10-27 05:55:26.
10 Koines and koineization
Confederacy of Delos into an Athenian empire, the influence of Attic

spread rapidly throughout the Aegean. Most of the other city-states in this
empire spoke Ionic dialects (to which the comparatively archaic Attic was
closely related)9 and resented the control of Athens, but the emerging
Koine, usually referred to as Great Attic at this early stage (Bubenik 1993:
12; Horrocks 1997: 29), was useful for commerce and general intercourse
and was also employed as the (written) language of administration (Hor-
rocks 1997: 33). It has been suggested (Thomson 1960: 34; Hock 1986:
486) that a likely birthplace for the Koine was the Peiraieus, or port of
Athens, where Attic speakers and Ionic speakers from other parts of the
empire interacted, along with Doric speakers from the neighboring Pelo-
ponnesus. However, its use as written "standard" and spoken vernacular
was never restricted to the Peiraieus, since contact between Attic and Ionic
speakers occurred in a variety of contexts. The city-states in the Attic
League had to provide soldiers for the Athenian armies, as well as deal
with Athenian officials in their territories and Athenian administrative
documents composed in official Attic (Horrocks 1997: 31).10 Athens also
sent out numerous Attic-speaking colonists to the colonial territories,
where they interacted with Ionic speakers. Many speakers of Ionic also
took up residence in Athens, and through their interaction with Athenians
may have contributed to changes in the speech of "middle-class" residents
of the city. Great Attic thus developed in part as a second dialect of Ionic
speakers, but it became the native dialect for following generations in some
of the Ionic cities. Eventually, Philip of Macedón adopted Great Attic as
his language of administration and it later spread throughout the eastern
Mediterranean as a result of the conquests of his son, Alexander the Great.
The early Koine may have benefited from its ambiguous relationship to
traditional Attic; its difference from Attic may have made it more accept-
able to the dominated Ionic speakers of the empire (Hock 1986: 486),
while its similarity perhaps lent it prestige and made it acceptable to Philip
of Macedón. It has been characterized as a "de-Atticized Attic" (Hock
1986: 486) and as a "de-Atticized Ionicized Attic" (Bubenik 1993: 13). It is
interesting to note that this mixed and simplified form of Attic was decried
from the beginning as being impure and corrupt (Palmer 1980: 175), and
centuries later, under the Romans, a campaign of "Atticization" was
launched to improve it (Buck 1933: 22). The following features have often
been identified as typical of the mixed and simplified nature of the original
Koine:
Koine and koines 11
1. Highly distinctive Attic -it- was largely replaced by the more widespread
(Ionic) equivalent -ss-, thus:
Attic Koine
glötta glossa 'tongue'
phulattö phulassô 'guard, watch'
tettares tessares 'four' (Hock 1986:486)
2. Distinctive Attic -rr- was replaced by more widespread (Ionic) -rs-:
Attic Koine
arrën arsën 'male' (Hock 1986: 486)
3. Attic -ä- (<*-ayw) was replaced by more widespread -ai-:
Attic Koine
eläa elaia 'olive' (Hock 1986: 4)
4. Dual number, a feature of Attic, was abandoned in the Koine, as in most
other Greek dialects (Hock 1986: 486).
5. Attic -eös and Ionic -êos were replaced by Doric -äos in läos 'people' and
nâos 'temple', leading to a more regular declension for these nouns (Hock
1986: 487).11
6. Pitch accent was lost, replaced by a stress accent (Thomson 1960: 35).
7. Phonemic vowel quantity was abandoned (Thomson 1960: 35) and distinc-
tive consonant length was lost (Horrocks 1997: 113);
8. The number of vowels was reduced; diphthongs became monophthongs
(Palmer 1980: 176-177).
9. Final -n was regularized in the accusative (Thomson 1960: 35).
10. The optative disappeared (merged with the subjunctive); the infinitive be-
came common in use with prepositions; the imperfect and aorist were reor-
ganized on a new uniform basis; numerous irregular verb forms were regu-
larized (Thomson 1960: 35).
11. The particle äv was replaced by a more transparent periphrasis (Thomson
1960: 36).
12. In some cases new words replaced both Ionic and Attic equivalents:
Attic Ionic Koine
naûs nëûs ploîon 'ship'
(Bubenik 1993: 15)
This list of characteristics is attractively simple and clear - deceptively so

- but not all who have used the concept and the term koine have agreed on
the features that characterized the original Koine. This has led to varying
and problematic interpretations of the term's meaning. Indeed, the great
distance between the present and the period in which these social and lin-
guistic changes occurred has made it difficult to define the features of the
Koine, much less a clear notion of how the Koine was produced. One prob-
lem has been that this temporal distance (and lists like the one above) has
tended to give a falsely static impression of the Koine. Many scholars ap-
pear to have conceived of it as a finite state, but in reality the Koine was
constantly developing. Palmer (1980: 177) points out that precise dating of
attestations of these changes shows that they did not all occur concurrently,
but rather appeared and spread at different times over the course of centu-
ries, along with the social and geographical spread of the Koine. For ex-
ample, Horrocks (1997: 35, 27) discusses the replacement of -tt- by -ss-
and the loss of dual number as a feature of early Great Attic (presumably
lost even earlier in a prehistoric Ionic phase of dialect mixing), but believes
the loss of the pitch accent (and with it the resultant loss of distinctive
vowel and consonant quantity) to have begun in classical times and only to
have reached completion in the (Egyptian) Koine by 150 B.C. (Horrocks
1997: 109). Indeed, many of these phenomena were attested in one or sev-
eral contributing dialects prior to the formation of the Koine itself.
Another assumption, not unrelated to the view of the Koine as a static
entity, has been that the Koine was uniform across the Hellenistic world.
However, this seems to have been true primarily of a conservative and
standardized Koine which was employed in official documents. Horrocks
(1997: 61) observes that the "very high grammatical and orthographic stan-
dards of even very ordinary 'official' papyrus documents from Egypt"
suggests that even low-ranking officials must have received rigorous train-
ing in this formal variety. On the other hand, more private documents re-
veal significant regional diversity, and there exist features of Egyptian
Koine which distinguish it from the Koine of Asia Minor, or that of Pales-
tine and Syria (Bubenik 1989: 175-252; Horrocks 1997: 60-64).
With regard to the causes of these changes, Thomson (1960: 35)
seemed to assume that the extension of Greek to non-native speakers
played a role, but he offered no further details. Others have seen the
changes that resulted in the Koine as examples of "normal" development.
Indeed, Buck dismissed out of hand the possibility that the changes in the
Koine were in any way unique:
But mixture in vocabulary is common to most of the present European lan-
guages. There were also changes in pronunciation, in syntax, and in the
meaning of words, similar to the changes that have taken place in the other
European languages. (Buck 1933: 22)
Buck was partially correct in making these assertions, but, as will be dis-
cussed below, there is reason to believe that there are distinct though gen-
eralizable processes which led to the formation not only of the original
Koine but also of many other language varieties that share similar histories
Koine and koines 13
of dialect mixing and demographic movement. More recently, scholars

such as Bubenik (1993) and Horrocks (1997: 41) have come to view the
changes which characterize the Hellenistic Koine, especially in such new
urban centers as Alexandria, Antioch, and Pergamum, as arising from the
combined effects of top-down imposition of the Koine by the ruling dynas-
ties (favoring overall uniformity), dialect leveling resulting from the mix-
ing of the dialectally heterogeneous immigrant masses from old Greece,
and imperfect acquisition by indigenous populations of the language un-
dergoing koineization (favoring interregional diversity).12 Horrocks makes
the following useful observation on the issue of uniformity and variation in
the Koine:
It is essential, then, to see the Koine not only as the standard written and
spoken language of the upper classes (periodically subject to influences
from belletristic classical Attic), but also more abstractly as a superordinate
variety standing at the pinnacle of a pyramid comprising an array of lower-
register varieties, spoken and occasionally written, which, in rather different
ways in the old and the new Greek worlds, evolved under its influence and
thereafter derived their identity through their subordinate relationship to it.
(Horrocks 1997: 37)
1.2. Modern use of the term koine
Modern metaphorical or technical use of the term koine has grown as

scholars have attempted to identify commonalities between (the develop-
ment of) the original Koine and other language varieties. In fact, this has
been only too easy to do, as different scholars have identified different
features of the original koine as being key to its nature. The meanings as-
signed to metaphorical uses of the term koine became increasingly diverse
as use of the term grew during the 20th century. According to Cardona
(1990: 26), modern use dates from Meillet's (1913) discussion and analysis
of the original Koine. Meillet reported three meanings for the term: for
Hellenistic Greeks, the language of everyday use; for Hellenistic gram-
marians such as Apollonius Dyscolus, the language of reference for use in
grammars, and possibly the base from which new dialects arose; for mod-
ern Hellenists, the base for modern Greek (Meillet [1913] 1975: 253-275).
Meillet suggested that it was easier to define the structure of a koine by
what is was not (the dialectal features it lacked) than by what it was,
thereby establishing the problematic notion that koines are merely the
"least common denominator" of contributing varieties (see below). He also

emphasized the long and apparently punctuated development of the Koine:
la κοινή n'est pas une langue fixée, ce n'est pas non plus une langue qui
évolue en obéissant régulièrement à certaines tendences; c'est une langue où
il y a une sorte d'équilibre, constamment variable, entre fixation et évolu-
tion. (Meillet 1975: 256)
Most importantly, however, Meillet argued that the features of the Koine
were not unique to it, and suggested that Vulgar Latin, among other lan-
guages, showed a similar history of social expansion and structural reduc-
tion (Meillet 1975: 257).13 Meillet's discussion thus identified the useful-
ness of "ce terme commode et nécessaire", as he calls it, and thereby
initiated its more general use as a means of categorizing language varieties.
Jakobson ([1929] 1962: 82) was another early user of the term, and ob-
served that dialects which serve as vehicles of communication in large
areas and gravitate towards the role of koine (by which he seemed to mean
lingua franca; see below) tend to develop simpler systems than dialects
which are restricted to local use (these ideas were further explored in An-
dersen 1988). Despite such early use, the term apparently remained highly
specialized and rarely used until the second half of the 20th century
(Cardona 1990: 27). Cardona offers as another early example the following
passage from Tagliavini's Origini delle lingue neolatine·.
Probabilmente il francone, parlato alle corti dei re merovingi e carolingi, era
una lingua mista, una specie di koiné formato da elementi franchi salì e
franchi ripuarì, nonché da elementi romanzi e germanici assai vari.
(Tagliavini 1949: 206)
Talgiavini uses the term to refer to a variety that results from the mixing of
not only related but also unrelated languages, thus employing it in a way
that seems justified only in the broadest sense (i.e., if the feature of mixing
is the only one picked out by the metaphor; but see below for discussion of
the potential impact of non-native speakers).
Although not all scholars would use the term with such liberty, it has
nevertheless received a tremendous variety of interpretations in the linguis-
tic literature. Siegel (1985) argues that this is so because the original Koine
had six different features which scholars could highlight (or ignore) in
making comparisons. According to Siegel, the Koine:
— was based primarily on one dialect

— had features of several dialects
Koine and koines 15
— was reduced and simplified14

— was used as a regional lingua franca
— was a standard
— was nativized to some extent (Siegel 1985: 358-9, 362)
In order to determine the dominant interpretations of the term, Siegel ana-

lyzed references to some 36 language varieties as koines (Siegel 1985:
359):
1. Literary Italian (Pei 1966: 139)

2. Church Kikongo [Congo] (Nida and Fehderau 1970: 152)
3. Standard Yoruba (Bamgbose 1966: 2)
4. Bahasa Indonesian (Pei 1966: 139)
5. High German (Germanic Review 1 (4): 297 [ 1926])
6. Bühnenaussprache [Stage German] (Dillard 1972: 302)
7. Hindi (Hartmann and Stork 1973: 123)
8. Latin in the Roman Empire (Hill 1958: 444)
9. Belgrade-based Serbo-Croatian (Bidwell 1964: 532)
10. Mid-Atlantic koine [England] (Times Literary Supplement, 22 April 1965)
11. Network Standard English [U.S.A.] (Dillard 1972: 302)
12. Melanesian Pidgin (Ervin-Tripp 1968: 197)
13. Fourteenth-Century Italian of Naples (Samarin 1971: 134)
14. Town Bemba (Samarin 1971: 135)
15. Fogny [Senegal] (Manessy 1977: 130)
16. Kasa [Senegal] (Manessy 1977: 130)
17. Congo Swahili (Nida and Fehderau 1970: 152)
18. Lingala [Congo] (Nida and Fehderau 1970: 153)
19. 'Interdialects' of Macedonian (Lunt 1959: 23)
20. Koineized colloquial Arabic (Samarin 1971: 134)

21. Ancestor of modern Arabic dialects (Ferguson 1959a: 616)
22. Vernacular of north China, seventh to tenth centuries (Karlgren 1949: 45)
23. Calcutta Bazaar Hindustani (Gambhir 1983)
24. Israeli Hebrew (Blanc 1968: 237-51 )
25. Eighteenth-century American English (Traugott 1977: 89)
26. Fiji Hindustani (Siegel 1975: 136; Moag 1979: 116)
27. Trinidad Bhojpuri (Mohan 1978)
28. Guyanese Bhojpuri (Gambhir 1981)
29. Surinam Bhojpuri (Gambhir 1981: 184)
30. Mauritian Bhojpuri (Gambhir 1981: 184)
31. Slavish [U.S.A.] (Bailey 1980: 156)
32. Italian-American (Haller 1981: 184)
33. Slave languages [Caribbean] (Dillard 1964: 38)
34. English-based nautical jargon (Hancock 1971: 290n)

35. Black Vernacular English (Mühlhäusler 1985: 8)
36. Canadian French (Gambhir 1981)
Siegel reports that very few of these language varieties could be said to
have all the properties of the original Koine, and he found wide variation in
the meanings assigned to the term itself. Studies 1-22 used the term to
refer to a lingua franca (any variety used for intergroup communication);
studies 1-11 used it to refer to regional standards. A majority of the studies
indicated that several dialects must contribute to the formation of a koine.
Only a few studies included reference to a base dialect, reduction and sim-
plification, or to nativization (Siegel 1985: 362).
Though Siegel restricted himself to studies published in English, his
general conclusions appear valid for studies published in other languages
as well. Still, further variation in meaning does crop up. For example, Ro-
mance philologists have long used the term koine to describe certain me-
dieval literary varieties, such as the Provençal of the Troubadours and the
"Sicilian" dialect of the court of Frederick Π, praised by Dante in De Vul-
gari Eloquentia (Elcock 1960: 399, 459). These varieties certainly show
mixing and the elimination of dialect features, but they appear to have been
the result of conscious selection and limited to use in writing by a tiny
elite. They have also been labeled, perhaps more appropriately, literary
standards (Elcock 1960: 455).15 In Italian linguistics, the term has also been
used to describe certain (probably spoken) regional varieties that arose
from the Middle Ages around principal urban centers (e.g., Venice, Turin,
Milan, Genoa, Naples, Palermo). This use follows those that emphasize
dialect mixing, use as lingua franca and/or regional standard. More re-
cently, koine has also come to be used as a sociolinguistic label for a cer-
tain level in the dialect continua that characterize most regions of Italy
(Berruto 1989: 13). Pellegrini ([I960] 1975: 37) divided these continua
into four levels: dialect, regional koine, regional Italian, Italian standard.
The regional koines are thus seen as distinct from the regional standards,
but their lingua franca function remains significant, as does, at least for
some authors (e.g., Cardona 1990), the mixing, reduction, and simplifica-
tion of dialect features.
Given such wide variation in actual usage, it is unsurprising that explicit
definitions of the term have also varied widely. The following give some
idea of this variation (some of these are quoted in Siegel 1985):
Koine and koines 17
— "a form of language resulting from a compromise between various dialects

and used as a common means of communication over an area covering all
the contributing dialects." (Graff 1932: xxxvii)
— "a compromise among several dialects" used "by a unified group in a self-
contained area within a larger linguistic area". Pei also considers a koine to
be a planned language: "a deliberately sought sublimation of the constituent
dialects rather than an unconscious and accidental merger". (Pei 1966: 139)
— "Koine is the term for a 'common' dialect which lacks prominent features
of the more conventional dialects of a language. It is the end result of dia-
lect levelling." A koine is often considered "good" speech in the language
and is most often a standard dialect. (Dillard 1972: 302)
— "KOINES. A standard normally has its origin in the dialect of some particu-
lar territory, which comes to enjoy superiority over those of neighboring
regions, for non-linguistic reasons (usually political, less often economic or
social, never purely literary). Such a favored dialect comes to be the com-
mon language or KOINÉ . . . used throughout its region, where it is usually
comprehensible to most of the speakers of the neighbouring dialects. In the
course of its spread, the koiné retains its basic relationship to the dialect on
which it is based, but takes in features from related dialects, as in the in-
stance of Span, /xuérga/ juerga 'spree' from Andalusian . . . or French fa-
bliau 'animal-fable' from Picard (Φ ONFr. fablel 'little fable')."
(Hall 1974: 104)16
— "The spoken language of a locality which has become a standard language
or lingua franca." (Crystal 1992)
Though a more precise definition of the term has been developing since the
publication of Ferguson (1959a), widely varying interpretations still
abound, even in more recent studies such as those in Sanga (1990) and
Knecht and Marzys (1993), where, for example, the terms koine and stan-
dard are frequently conflated.
The different interpretations given to the term have produced a situation
in which its use often produces more confusion than clarity. Siegel (1985:
363) sets out to resolve this problem by specifying a technical meaning for
the term. He claims that the concept of dialect mixing is fundamental, and
specifies that the contributing varieties must be language varieties that are
either a) mutually intelligible or b) share the same genetically-related su-
perposed language (1985: 375-376). 17 These may include regional dialects,
sociolects, and "literary dialects". For the last category, Siegel based his
claim on the development of Israeli Hebrew, which Blanc describes as a
result of the mixing of "a variety of literary dialects, several substrata, and
several traditional pronunciations" (Blanc 1968: 238-239). But this defini-
tion raises the problem of "non-native" speakers in the demographic mix:
should learner interlanguages be included among the contributing varieties
of a koine? The impact of non-native speakers has also been identified as
important to the development of the Hellenistic Koine (e.g., Horrocks
[1997] reports Coptic substrate features in Egyptian koine texts) and the
early Arabic koine (Ferguson 1959a). Mesthrie (1994: 1865) defends their
potential importance in the development of any koine, since the variants of
native speakers of unrelated languages are less likely to be perceived as
"foreign" in the mixed linguistic pool of the prekoine (cf. LePage 1992).
However, certain constraints need to be placed on this broad view of con-
tributing varieties, at least for prototypical cases. First, adult interlanguage
features may form part of the pool, but these speaker-learners must have
easy access to input and interaction with native speakers. This in turn im-
plies that such "foreign" speakers do not form a majority in the commu-
nity, since their dominance would reduce the likelihood of their obtaining
sufficient access to the language (varied though it may be). Thus, the range
of contributing varieties or subsystems must be expanded to include inter-
language varieties of second language learners.18
Siegel also warns that many of the definitions given to the term koine
are either too broad or too narrow. Thus, using koine as a synonym of lin-
gua franca or common language robs it of usefulness, as does restricting
koine to the meaning of "planned, standard, regional, secondary" variety or
one based primarily one dialect. Perhaps more controversially, Siegel's
explanation could be read as favoring a close identification between koines
and standards:
unplanned, nativized, or transported languages may be koines if they exhibit

the mixing of any linguistic subsystems such as regional dialects, literary
dialects, and sociolects. However, although a koine may or may not be a
formal standard, it is implicit in all definitions that a koine has stabilized
enough to be considered at least informally standardized. (Siegel 1985: 363)
In reality, Siegel meant socially-based language norms rather than the codi-
fied language norms that characterize standard languages, and Siegel
(1987: 201) clarifies this issue by abandoning use of the term "informal
standardization". The definition might also be improved by emphasizing
that prototypical koines not only may be but necessarily are unplanned,
nativized, and transported varieties (see below).
Koine and koines 19
According to Siegel, most koines are characterized to some extent by

reduction and simplification, though he comments:
requiring a koine by definition to exhibit these features would be too restric-
tive, as the amount of reduction or simplification may differ between koines
according to both the conditions under which they developed and their cur-
rent developmental stage. (Siegel 1985: 363)
Recent research (e.g., Kerswill and Williams 2000) shows that there are
cases of koineization without obvious examples of simplification; this is
due to the pre-existing similarity between the contributing varieties, in
which most variation is allophonic. Mohan (unpublished paper; reported in
Siegel 1985: 361-2) points out that koines are of two types: those based on
dialects with great structural similarity (such as that studied by Kerswill
and Williams), and those based on more highly differentiated dialects.
While I think these "types" have to be viewed as extremes on a scale,
greater difference between the contributing dialects can be expected to lead
to greater perceived simplification in the resultant koine. On the other
hand, Siegel's reference to the "current developmental stage" is problem-
atic, since it implies that a koine, once formed, continues to be in some
way identifiable as a koine; as will be emphasized below, koines are only
identifiable in a historical sense.
Siegel concluded his discussion of koines with the following definition:
a koine is the stabilized result of mixing of linguistic subsystems such as re-
gional or literary dialects. It usually serves as a lingua franca among speak-
ers of the different contributing varieties and is characterized by a mixture
of features of these varieties and most often by reduction and simplification
in comparison. (Siegel 1985: 363)
The claim that a koine normally serves as a lingua franca requires some
qualification. A koine would only serve as a lingua franca for non-native
speakers, since for native speakers it would serve as a primary (perhaps
even sole) means of communication. The function of lingua franca may be
important in the development of regional koines. Siegel explains that:
a regional koine usually results from the contact between regional dialects of
what is considered to be a single language. This type of koine remains in the
region where the contributing dialects are spoken. (Siegel 1985: 363)
Petrini (1988: 34, 42) points out that regional koines with no native speak-
ers can be extremely unstable, varying from speaker to speaker and from
situation to situation, and may be no more than an abstract perception of
the linguist who observes the frequently similar results of multiple

accommodations by speakers. Petrini claims too that a regional koine as a
clearly distinct variety is likely to arise only as it gains native speakers,
most often in urban centers, who serve to stabilize the norms of the koine.
This regional koine is then used as a lingua franca by speakers of rural
dialects, but such use is secondary to its use by native speakers.19
The notion of koine as lingua franca is more problematic in the case of
immigrant or colonial koines. According to Siegel, an immigrant koine:
may also result from contact between regional dialects; however, the contact
takes place not in the region where the dialects originate, but in another lo-
cation where large numbers of speakers of different regional dialects have
migrated. Furthermore, it often becomes the primary language of the immi-
grant community and eventually supersedes the contributing dialects.
(Siegel 1985: 364)
In this case, it seems that the lingua franca function would only exist for a
short time, until the speakers of the contributing dialects die off. After that,
all or most speakers of the koine are native speakers. However, there is a
larger issue here: emphasis on the use of koines as lingua franca may re-
veal an assumption that koines develop primarily in order to facilitate clear
communication. This is a partly valid assumption in the case of language
subsystems that are sufficiently different to impede mutual comprehensibil-
ity (as seems to have been the case in many socially subordinate koines,
such as those used by workers in the Bhojpuri-Hindi diaspora), but most
dialects are in fact mutually comprehensible (or become so quickly with
interaction), so effective communication cannot be identified as the only or
even the most important factor in koine formation. This issue is discussed
in greater detail below.

Another valuable effort to define koine as a technical term is that of
Mesthrie (1994). Mesthrie, like Siegel, analyzes modern use of the term in
relation to the original Koine, for which he identifies four key features:
— (a) its development as a new, common variety based on existing dialects of

the language (common is taken in the sense of "shared");
— (b) its use as a common (or "vulgar") medium of communication between
speakers with different first languages or speakers from different dialect ar-
eas;
— (c) its use as the standard/official language of a politically unified region;
Koine and koines 21
— (d) changes in its structure on account of its wide use as both first and sec-
ond language (involving a synthesis of these at some stage).
(Mesthrie 1994: 1864)
For modern uses, Mesthrie explains that in one stream of thinking, the for-
mal criteria of (a) and (d) are considered primary, and in another, the
functional properties of (b) and (c) are considered primary. Mesthrie re-
jects (b), (c) and (d) as criteria for definition of koine:
The major objection to (b), (c), or (d) alone as a defining criterion is that on
its own each defines a language variety or linguistic process that has a well-
established label: (b) is synonymous with lingua franca (and the process of
language spread); (c) is better described as 'standardization'; and (d) de-
scribes the phenomenon of substrate influence in second language acquisi-
tion or in language shift. (Mesthrie 1994: 1864-1865)
Mesthrie identifies (a), or the incorporation of features from several (re-
gional) varieties of a single language, as the only necessary feature of a
koine (however, see below for consideration of the impact of language
acquisition). In effect, Mesthrie rejects the synchronic functions - lingua
franca or standard - as defining features of a koine, and accepts only those
aspects that are essentially diachronic in nature, resulting from the process
of dialect mixing:
While the processes involved in koineization are of considerable interest to
the linguist, once a koine has formed there may be nothing to distinguish it
from older dialects of the language. (However, subordinate immigrant
koines do often show a significant reduction in inflections.) Generally, the
designation koine might be appropriate at a particular stage in the history of
the language, but loses significance once the variety becomes established as
the first language of a new generation. Like any other natural language a
koine may in time develop new regional subdialects, as shown by the history
of Greek. (Mesthrie 1994: 1865).
Hence, koine has become, in its technical sense, merely a convenient label
for those language varieties and states that result from the social and lin-
guistic processes of koineization.20
2. Models of koineization
Most recent discussions of koines have shifted from a focus on the resul-
tant state to a focus on the processes of koine formation. Though Samarin
(1971) was the first to use the term koineization, others before him had
already begun to shift focus to the diachronic study of koine formation.
Ferguson's (1959a) study of the Arabic koine, which he claimed was the
common base for modern spoken dialects of Arabic, was essentially an
exercise in reconstruction of a stage of the language. He attributes the for-
mation of this variety (perceived as uniform) to "a complex process of
mutual borrowing and leveling among various dialects", while most of the
14 features he discusses show some sort of loss, reduction or simplifica-
tion. Given the time depth of this study and the lack of documentary evi-
dence, no further study of processes was possible. Blanc (1968) argued that
modern Israeli Hebrew was "gradually given a definite shape by a slow
'koineizing' process drawing on several pre-existing sources . . . Usage had
to be established by a gradual and complex process of selection and ac-
commodation which is, in part, still going on, but which now has reached
some degree of stabilization" (Blanc 1968: 238-239). Samarin (1971) was
only indirectly concerned with koineization, but he suggested use of the
term as a means of differentiating a unique process, distinct from dialect
leveling or borrowing, that leads to the formation of a new dialect. Samarin
(and Dillard 1972: 300) also emphasized that koineization involves the
suppression of localisms or prominent stereotypable features as speakers of
different dialects mix together in new social contexts, particularly in cases
of migration.
None of these studies engaged in detailed discussion of the process or
model of koineization. However, the growth of studies of pidginization and

creolization also drew scholars' attention to other types of colonial/post-
colonial languages, among them the numerous varieties that arose as a
consequence of the Bhojpuri-Hindi diaspora. In the aftermath of the aboli-
tion of slavery, European colonial powers shipped hundreds of thousands
of Indian peasants on indentured contracts to other colonies. The immi-
grants spoke primarily genetically-related Indie languages from the north,
but in some cases there were also speakers of Dravidian languages from the
south. The Indie varieties included dialects of Bhojpuri, Avadhi, other
eastern and western varieties of Hindi, Bengali, Rajasthani, Panjabi and
Calcutta Bazaar Hindustani, with widely varying degrees of mutual com-
prehensibility between the different varieties. In each colony, a compro-
Models of koineization 23
mise variety arose that was used as an in-group language among the Indian
laborers. The areas where these new dialects have been identified and stud-
ied include Fiji, Surinam, Natal (South Africa), Trinidad, Mauritius and
Guyana (Siegel 1988a; Mesthrie 1993: 26-29).
2.1. Siegel ' s stage-based model
As these different varieties received more scholarly attention, efforts to

define common principles of koineization began to appear. An early effort
is that of Gambhir (1981), but the most well-known and influential in this
tradition is that of Siegel (1985), who was investigating the development of
Fiji Hindi (or Hindustani; e.g., 1975, 1987, 1988b). In his (1985) study, he
synthesizes notions of koines and koineization from other studies in order
to arrive at a technical definition of koine (reported above) and a more
precisely defined model of koineization, based on his own findings and
that of others. His model is based on a sequence of four possible stages of
koineization:
1. Prekoine. "This is the unstabilized stage at the beginning of koineization.

A continuum exists in which various forms of the varieties in contact are
used concurrently and inconsistently. Levelling and some mixing has be-
gun to occur, and there may be various degrees of reduction, but few forms
have emerged as the accepted compromise."
2. Stabilized Koine. "Lexical, phonological, and morphological norms have
been distilled from the various subsystems in contact, and a new compro-
mise subsystem has emerged. The result, however, is often reduced in mor-
phological complexity compared to the contributing subsystems."

3. Expanded Koine. A stabilized koine "may become a literary language or
the standard language of a country. This extension of use is often accom-
panied by linguistic expansion, for example, in greater morphological
complexity and stylistic options."
4. Nativized Koine. "A koine may become the first language for a group of
speakers . . . This stage may also be characterized by further linguistic ex-
pansion (or elaboration), but here some of it may be the result of innova-
tions which cannot be traced back to the original koineized varieties."
(Siegel 1985: 373-374)
Siegel emphasizes that not all these stages need necessarily occur in any
particular case of koineization, and provides examples of such variable
development (see Table 1). Siegel consciously modeled this presentation
on then-current approaches to the study of pidginization and creolization,

and borrowed his paradigm (Table 1) from the developmental continuum
for pidgins and creóles (Table 2) constructed by Mühlhäusler (1980: 32).
At first glance, the proposed relationship seems eminently reasonable,
since pidgins, creóles, and koines all result from language contact and
demographic mixing, and they are often found in colonial or post-colonial
regions. However, as Siegel himself has argued in later work (see below),
there are significant differences between pidginization/creolization and
koineization, and these differences underlie some problematic implications
of the (1985) stage-based model.
Table 1. Developmental contunua of koines. Source: Siegel (1985: 375).
prekoine prekoine prekoine

4 i ι
i stabilized koine stabilized koine
4 4 I
J, J. expanded koine
4 I 4
nativized koine nativized koine nativized koine
(Fiji Hindustani) (Guyanese Bhojpuri) (Greek Koine)
Table 2. Developmental contunua of creóles. Source: Siegel (1985: 375), based on

Muhlhausier (1980: 32).21
jargon jargon jargon

4 4 4
4 stabilized pidgin stabilized pidgin

4 4 4
4 4 expanded pidgin
4 4 4
creole creole creole
(West Indian English (Torres Strait Creole) (TokPisin)
Creole)
Pidgins are generally understood to result from contact between typologi-

cally distant varieties, while prototypical koines (such as Spanish) result
from contact between linguistic subsystems that show high degrees of mu-
tual intelligibility. Since speakers in a koineizing context can usually un-
derstand each other, the need to communicate information - which plays

the key role in pidginization - cannot be a generalizable primary motive for
alterations in speaker production (though it may play a greater role when
contributing varieties show greater structural differences, as in the case of
Fiji Hindi). More importantly, prototypical pidgins and creóles arise in
very specific social circumstances in which speaker-learners are separated
from models - native speakers of the target variety - at the same time that
they must communicate with the socially-distant model speakers or, more
importantly, other speakers with whom a common language is not shared.
The social context of koineization could not be more different, for the
speaker-learners, be they native speakers of a related dialect, second lan-
guage learners, or children, must be assumed to have easy access to abun-
dant, if highly variable, input. Indeed, within the Thomason and Kaufman
(1988) model, koineization is properly categorized as change with lan-
guage maintenance (but see below).
As a result, it becomes difficult to accept an unintended implication of
the stage-based model: that both pidginization and koineization are charac-
terized not only by mixing (which remains undefined but which we may
assume means the appearance of features from several source dialects in a
resultant koine) and simplification, but also by reduction/impoverishment.
Siegel borrows Mühlhäusler's (1980: 21) definitions of simplification ("an
increase in regularity or a decrease in markedness") and reduction ("a de-
crease in the referential or non-referential potential of the language"). But
while simplification is indeed a linguistic process of koineization, reduc-
tion of this sort cannot be, for reduction as defined here includes the ex-
treme structural/lexical reduction of pidgins, which makes full comprehen-
sion difficult or impossible outside of contexts of direct oral
communication, where gestures, intonation, and the possibility of clarifica-

tion substitute for structural complexity. Since learners in a koineizing
situation are not deprived of input, there is no reason for extreme reduction
to occur.22
If extreme reduction does not occur, then there is no need for a stage of
structural expansion, which in pidgins is associated with the expansion of
functions and/or creolization of an existing pidgin; this occurs as the pidgin
is extended to use in new communicative contexts, and it therefore requires
new vocabulary and more systematic marking of grammatical relations to
make it less context dependent and more fully functional as a primary
means of communication. Since no radical reduction such as that affecting
pidgins is present in koineization, expansion must be reinterpreted in the
context of koineization. It is certainly true that any expansion of contexts

of use is likely to require an expansion of the lexicon, but this is true of any
language that acquires new functions. In fact, Siegel exemplifies this third
stage with the use of the koine as a literary or standard language. The in-
clusion of standardization is not entirely unwarranted, for it reflects a fre-
quent reality: koines tend to be selected as standards, since standards also
require the minimal variation in form that characterizes koines. However,
while standardization does include a process of elaboration of the lexicon
and syntax, particularly of written language (Haugen 1966: 933; Lodge
1993: 26), this is not quite the same as expansion in pidginiza-
tion/creolization, which includes especially an increase in morphological
complexity. Moreover, standardization may enter into competition with
koineization. For instance, Fontanella (1992: 42-54) argues that in the
history of American varieties of Spanish, standardization has sometimes
(partially) impeded koineization, as in the interior of Mexico, and some-
times reversed its effects, as in Buenos Aires (see below). Rather than in-
cluding processes such as lexical expansion or standardization within a
model of koineization, it is probably best to see them as interacting with
koineization.
Another problem with the parallel stages of pidginization and koineiza-
tion is the timing and significance of nativization. In pidginization, relative
stabilization of grammar and lexicon may occur before nativization (as in
the well-known case of New Guinea Tok Pisin), but pidgins are not native
languages, and structurally they are very simple and therefore easily
learned by adults; nevertheless, they are relatively unstable with regard to
phonology, since each speaker's version will be affected by his/her native
language phonology. In large measure, it is the nativization stage of creoli-
zation that leads to full stabilization of a pidgin (although the question of

how many and what types of features need to be stabilized in order to con-
sider the variety stabilized will depend largely on the perspective taken).
Here then, there is a problem with Mühlhäusler's original proposal. In fact,
while adults do play important roles in the selection of features, nativiza-
tion by children is probably key to full stabilization or focusing of a koine
(Mesthrie 1994: 1866; Kerswill 1996). Petrini (1988: 42), as mentioned
above, argues that the developing koine or prekoine of the Italian region of
Ticino has so far failed to stabilize because there are no native speakers of
this variety, and Kerswill (1996) argues that in a koineizing community the
first signs of the new koine will become evident among the older members
of the first generation of children (see below).
In later publications, Siegel has discussed the similarities and differ-

ences between koineization and creolization that I highlight here (e.g.,
Siegel 2001), and he has accepted many aspects of Trudgill's work into his
own approach (Siegel 1993b), but such reformulations have not led to a re-
evaluation of the original stage-based model, which is still frequently cited
in the literature (e.g., Kerswill and Williams [2000] situate their own dis-
cussion within this framework). The adherence to the pidginization para-
digm which underpins the four-stage model may reflect the particular his-
tory of Fiji Hindi and other post-colonial language varieties, since some of
these varieties show the interaction of koineization with other processes.
For example, Siegel himself (1987: 196) points out that one of the contrib-
uting varieties of Fiji Hindi was itself a pidgin, so we should not be sur-
prised to find some effects of pidginization in the resultant koine (also
suggested by Trudgill 1986: 106). Of course, the challenge to defining
koineization is the necessity of distilling a prototypical model from com-
plex and varied cases of real change.
Siegel's early model has helped to draw the attention of scholars to the
study of koineization and from it we can retain useful insights, including
the concept of prekoine (the highly variable, diffuse, even chaotic, initial
stage of demographic and dialect mixing); the importance of stabilization
or focusing of new norms; the definition of contributing varieties as mutu-
ally intelligible linguistic subsystems (which allows inclusion of child and
adult learner language); and the recognition of the interaction of koineiza-
tion with other processes. The definition of koineization in terms of these
four stages does not appear, however, to be the best means of furthering
our understanding of the process. In fact, the generalizable stages of
koineization can probably be limited to just two: the prekoine, character-
ized by the co-existence of numerous varieties and variants, and the stabi-
lized (or focused) koine, which has become the native language variety of
at least some speakers; these two stages can of course be separated by a
period of variable length of norm selection and enforcement (e.g., Trudgill
[1998] and Kerswill [2002] suggest three stages based on progressive fo-
cusing over three generations; see below). Nevertheless, the restriction to
two basic stages (or three generational stages) does not preclude the de-
scription of stages of development in a particular language variety, in
which case one could refer to the interaction of koineization with processes
such as pidginization or standardization. Siegel makes an observation that
has important implications for application of the model to actual language
histories, and in particular to that of Spanish:
It should be stressed that the developmental continuum of a koine is not

necessarily linear. At any stage, for example, 'rekoineization' can take place
if there is continued contact with the original closely related varieties, or ad-
ditional contact with different ones. (Siegel 1985: 375)
For those varieties that have seen repeated phases of koineization, the attri-
bution of particular changes to particular phases will become an added
objective of historical sociolinguistic research (see Chapters 3, 4, and 5).
2.2. Trudgill's process-based model
The single most influential discussion of koineization is that of the socio-

linguist Peter Trudgill, who was drawn to the topic by his broader interest
in contact between dialects and how this affects language variation and
change. His primary discussion of koineization appears in his (1986) book
Dialects in Contact, which includes analysis of a range of issues associated
with contact between individual speakers of dialects, contact between sta-
ble dialects, and dialect mixing in koineization. In the final chapters of this
book, Trudgill defines a model of koineization and tests it on the history of
certain varieties of colonial English, particularly Australian English.
Trudgill's model is based on the isolation of the processes that occur dur-
ing koineization rather than stages of development. The following is his
summary of what happens during koineization:
In a dialect mixture situation, large numbers of variants will abound, and,
through the process of accommodation in face-to-face interaction, interdia-
lect phenomena will begin to occur. As time passes and focusing begins to
take place, particularly as the new town, colony, or whatever begins to ac-
quire an independent identity, the variants present in the mixture begin to be
subject to reduction. Again this presumably occurs via accommodation, es-
pecially of salient forms. This does not take place in a haphazard manner,
however. In determining who accommodates to whom, and which forms are
therefore lost, demographic factors involving proportions of different dialect
speakers present will clearly be vital. More importantly, though, more
purely linguistic forces are also at work. The reduction of variants that ac-
companies focusing, in the course of new-dialect formation,23 takes place
via the process of koinéization. This comprises the process of levelling,
which involves the loss of marked and/or minority variants; and the process
of simplification, by means of which even minority forms may be the ones to
survive if they are linguistically simpler, in the technical sense, and through
which even forms and distinctions present in all the contributory dialects
may be lost. Even after koinéization, however, some variants left over from
the original mixture may survive. Where this occurs, reallocation may oc-
cur, such that variants originally from different regional dialects may in the
new dialect become social-class dialect variants, stylistic variants, areal
variants, or, in the case of phonology, allophonic variants.
(Trudgill 1986: 126; italics in original)
An important first observation must be made here: Trudgill explicitly
equates koineization only with the "more purely linguistic forces" of level-
ing, simplification, and reallocation. Though he recognizes these as signifi-
cantly related to speaker activity, we will see that his early view of koinei-
zation is both enhanced and limited by this focus on linguistic outcomes.
What follows is a discussion of each of the particular features of koineiza-
tion highlighted in the above passage.
2.2.1. Accommodation and salience
Trudgill borrows the concept of accommodation from the work of the so-
cial psychologist Howard Giles, who developed what is known as Speech
(or Communication) Accommodation Theory. According to Giles (1973:
90), "if a sender in a dyadic situation wishes to gain the receiver's ap-
proval, then he may adapt his accent patterns towards that of this person,
i.e., reduce pronunciation dissimilarities".24 This process is known as ac-
cent convergence. Its opposite is likely to occur when a speaker wants to
dissociate or signal disapproval.25 Accommodation may affect any linguis-
tic level (e.g., lexicon, syntax, morphology, phonology, as well as speaking
rate and style), and it is hypothesized to be a universal tendency of human
behavior (Trudgill 1986: 2).26

Trudgill's emphasis on accommodation reveals rather novel assump-
tions about why dialect contact leads to change. Given that most contribut-
ing varieties in a prekoine linguistic pool are mutually intelligible at least
to some degree, many of the alterations in speech that take place are not
strictly speaking necessary to fulfill communicative needs, although some
comprehension difficulties may occur (Trudgill 1986: 1). Rather, speakers
accommodate to the speech of their interlocutors in order to promote a
sense of common identity. This focus on the identity-marking function of
language (cf. Milroy 1992, 1993) is critical to an understanding of how and
why koineization occurs; it will be further discussed below.
Though accommodation is generally studied as a phenomenon affecting

the immediate performance of individuals (or short-term accommodation),
Trudgill identifies and analyzes what he terms long-term accommodation,
which results in a relatively permanent effect on the speech patterns of an
individual. Trudgill sees this as the basic mechanism underlying linguistic
change in two-dialect contact situations and in koineization resulting from
multidialectal contact. Making long-term accommodation the driving force
for linguistic change in situations of dialect contact has further implica-
tions. First, there are important differences between short-term and long-
term accommodation. Most social psychologists up to the time of
Trudgill's writing had focused on speech accommodation as stylistic varia-
tion, where the speakers presumably have pre-existing knowledge of the
variants they use. For example, Trudgill himself adduced data from his
Norwich studies which showed that his informants accommodated with
markers (variables subject both to social class and stylistic variation) but
did not do so with indicators (variables subject only to social class varia-
tion) which were not characteristic of their own speech; presumably they
lacked the knowledge to produce these forms. However, long-term accom-
modation places a greater burden on the speaker:
accommodation beyond the speech community will often be a rather differ-

ent process from accommodation within it. Accommodation within the
speech community, as in my Norwich interviews, involves altering the fre-
quency of the usage of particular variants of variables over which the
speaker already has control. Accommodation beyond the speech community,
on the other hand, may well involve the adoption of totally new features of
pronunciation. (Trudgill 1986: 12)
At this point speakers appear to enter into a process of language or dialect

acquisition, the significance of which is explained by Beebe and Giles:
All speakers have, to some extent, a limited linguistic repertoire, both in
their native languages and in any second languages they might speak . . .
Limitations in repertoire are central to research within SLA [second lan-
guage acquisition], whereas they are peripheral in studies of native speakers
. . . It is important, when extending social pyschological theories to SLA
data, that limitations in repertoire be considered. For it is the tension be-
tween limitations in ability to converge toward a native-speaking interlocu-
tor and motivation to converge that makes second-language data unique.
(Beebe and Giles 1984: 22-23)
Nevertheless, Trudgill hesitates to associate long-term accommodation

with second language acquisition as such. In part, this may be because of a
possible tendency on the part of speakers to approach dialect and language

learning differently (particularly in cases of contact between standardized
languages). If the dialect to be learned is viewed by speakers as simply a
different variety of the same language, then they may take a more piece-
meal approach to adopting or rejecting features of the target variety. If, on
the other hand, they believe the target to be a separate system, they are
likely to adopt strategies more typical of second language acquisition, and
attempt to learn the new variety as a whole. In Nordenstam's (1979: 24)
study of Swedish women in Norway, it was found that some of the women
perceived Swedish and Norwegian as separate languages and tried to keep
them apart, while others adopted features on a one-by-one basis, believing
Swedish and Norwegian to be varieties of one language. Still, no matter
what strategy a learner of a similar variety adopts (piecemeal or whole),
errors due to constraints imposed by the acquisition process are likely to
crop up, so a useful distinction is unlikely to stand on this criterion.27
Chambers (1992: 675-676) also prefers to maintain a distinction be-
tween long-term accommodation and second dialect acquisition. Chambers
bases his distinction on the behavior of the informants in his study of sec-
ond dialect acquisition. These were Canadian children/adolescents (aged
9-17) who had moved to England and were adopting features of Southern
English English. Though Chambers is himself Canadian, the children did
not accommodate toward him by using fewer British variants and more
Canadian ones. For Chambers then, the innovations in the children's
speech represented "irrepressible acquisitions rather than ephemeral ac-
commodations". This may be true, but the important point is that the chil-
dren had extended the newly adopted English features from their most
careful style to even their most colloquial vernacular style (Ellis 1985: 95).
Chambers seems to want to restrict the use of long-term accommodation to

those cases in which the speaker continues to shift between different ac-
cents in different situations and styles. This may be a useful distinction, but
it is a distinction of degree rather than kind, and Ellis makes the point that
acquisition of a feature in a learner's careful style is probably the first step
toward acquisition in all styles. Moreover, the examples of long-term ac-
commodation given in Kerswill (2002: 682-685) would seem to confirm
this analysis, since they show that the speech of adults engaged in long-
term accommodation is characterized by simplifications and great intra-
speaker variability, as is predicted in Ellis' model. I return to the issue of
acquisition below, but from here on I assume that long-term accommoda-
tion forms part of a process of dialect acquisition (although, in the koineiz-
ing context, what is being acquired is generally a set of the most fre-
quently-occurring features, rather than a pre-existing variety; see below).
If second dialect learners (those engaged in long-term accommodation)
do not learn all features, which features do they modify and/or learn? First,
it has been shown that (adult) speakers in more stable dialect contact situa-
tions tend to accommodate (to the extent that they can or want to) to the
most salient features. Following Nordenstam (1979), who was studying
contact between speakers of two different varieties, Trudgill (1986: 11-27)
argued that the most salient features are those which represent differences
in the lexicon and morphology (no mention is made of syntax; see below).
But, aside from the lexicon, it is in the phonology that dialect differences
are most consistently found, and this is where Trudgill focuses his atten-
tion. After analyzing evidence of accommodation from a variety of studies,
he suggests that the following factors contribute to salience (these are
summarized and listed as follows in Kerswill 1994: 154):
1. Phonological contrast. For example, English English speakers who relocate

to the US often adopt the contrasting phoneme /ae/ for their native /a:/ in
words of the dance class.
2. Great phonetic difference. English English speakers in the US tend to
adopt the American [A] for their native [D] in words of the hot class, and
Americans in England do just the reverse.
3. Naturalness. English English speakers in the US begin to flap intervocalic
Ν more quickly than Americans in England acquire the intervocalic [t] for
their flap, since flapping is a kind of "natural" weakening.
However, Trudgill also indicates a number of factors which appear to im-

pede accommodation and either reduce or override salience:
4. Extra-strong stereotyping. For example, accommodating Northern English

English speakers tend not to adopt the Southern English English /a:/ for
their native /ae/ in dance-class words, despite the phonological difference
(cf. Number 1 above).28
5. Phonotactic constraints. For example, English English speakers in the US
tend not to acquire American non-prevocalic /-r/ as in cart, bar (even
though the phonological difference is evident) since in most Southern Eng-
lish English dialects syllables of the shape VrC or Vr# are not allowed.
6. Homonymie clash. For example, English English speakers may avoid the
American pronunciation [hat] since it sounds too much like their own pro-
nunciation of heart.
These few examples make it clear that long-term accommodation cannot be

seen as an entirely predictable process, since accommodation sometimes
occurs or fails to occur for no obvious reason, and is also influenced by
cultural factors, such as in cases of stereotyping. Kerswill (1994b: 154-
155) is quick to point out the circularity of the notion of salience, for it is
defined precisely in terms of those features which are or are not adopted.29
Still, Kerswill too is unwilling to abandon salience, since it seems that
speaker-learners do find some features more "striking" than others. In a
recent and very detailed discussion of this very issue, Kerswill and Wil-
liams (2002) consider the value of numerous factors that have been pro-
posed as contributing to salience, and test several of those proposed by
Trudgill (1986) for their value in explaining the spread of features in dia-
lect leveling in southeast England. In the end, they conclude that language-
internal factors are an essential pre-condition for salience; these include
phonological contrast, great phonetic difference, internally-defined natu-
ralness (which favors weakening or loss of distinctions), semantic trans-
parency, and particular syntactic and prosodie environments (2002: 105;
see below). All these factors are likely to draw speakers' attention to par-
ticular features. However, no single linguistic factor, or set of factors, is
capable of guaranteeing salience. They suggest therefore that it is extra-
linguistic conditioning which is crucial in determining salience: cognitive,
pragmatic, interactional, social psychological, and sociodemographic fac-
tors. As such, salience cannot be predicted, and particular reasons for its
existence must be sought in each case, with special attention to relations
between speakers, their attitudes, and their sociocultural contexts of inter-
action.
In the discussions of Trudgill (1986) and Kerswill and Williams (2002),
salience is seen as the primary factor determining which features are ac-
commodated to and, therefore, which features become more frequent. This
of course is absolutely necessary in non-koineizing situations (how else
can one explain changes in frequency of particular variants?). However, in
nearly all discussions of koineization, it is argued that majority forms win
out in the final dialect mix (e.g., Trudgill 1986; Kerswill and Williams
2000). This argument rests on the assumption that frequency and consis-
tency of use of particular forms and form/function combinations play the
fundamental role in determining what forms are acquired, and this fre-
quency is determined primarily by the original demographic mix (Siegel
1997: 139). Indeed, the importance of frequency as a factor distinct from
salience is shown by Trudgill, Gordon, Lewis, and Maclagan (2000), who
provide clear evidence that most features of New Zealand English selected
during koineization are features which represented majority use among the
sum total of original settlers (a claim which is further supported by the
selection of similar majority forms in other Southern-Hemisphere Eng-
lishes which saw similar population mixes).30 When frequency and consis-
tency are seen as primary, salience is invoked as an explanation for those
features which cannot be explained as the result of greater frequency alone
(in some cases salience can plausibly be viewed as contributing to a fea-
ture's perceived frequency). It would seem therefore that older children
and adolescents - those responsible for the formation of a new koine (see
below) - do accommodate to each other and learn from each other, but that
over time it is the most frequent forms that are consistently favored in this
process (rather than any particular target variety). In a koineizing context,
salience as discussed by Kerswill and Williams (2002) is probably most
important for exceptional adaptations in adult, adolescent, or child speech,
which can then alter the frequency of certain variants and thereby affect the
learning of children and adolescents.
Nevertheless, the potential cumulative effects of perceptual and cogni-
tive salience should not be discounted. For example, Siegel (1997: 139)
emphasizes that stressed words are both phonetically and cognitively sali-
ent. Of the perpectives on perceptual salience reviewed by Kerswill and
Williams (most post-date Trudgill's early work), the most important is that
of Yaeger-Dror (1993: 203-206), who, following a comprehensive review
of studies of cognitive and phonetic factors which contribute to salience,
comes to the following conclusions about universally salient positions:
the beginning of a syllable, word, or sentence is most salient. A vowel nu-
cleus or intervocalic position is also salient (because more acoustically

prominent), as is the semantic nucleus (the focus) of the sentence. The coda
(of a syllable, of a word, or of a sentence) is the most redundant and least sa-
lient. (Yaeger-Dror 1993: 206)
Although, as Kerswill and Williams (2002) indicate, it is impossible to

make predictions about outcomes in particular cases, it may be that percep-
tual salience shows its influence in long-term trends. If speaker-learners are
most likely to learn perceptually salient items (other things being equal)
and to follow "natural" (ease-of-articulation) tendencies, then we would
expect a language variety which has undergone (repeated) koineization to
show a preference for CV syllable structure. In fact, this is exactly what we
find in the history of Spanish. Similarly, Ohala, in a review of acoustic
phonetic studies, finds that "place cues are generally less salient in VC
rather than in CV transitions and are even less salient in VN transitions"

(Ohala 1993: 159). If the effects of such "CV dominance" are combined
with articulatory weakening in syllable-final position, then we might also
expect a heavily koineized variety to show weakening of syllable-final
nasal consonants, with assimilation to the place of articulation of a follow-
ing consonant. In fact, Spanish (or most varieties of Spanish) is literally a
textbook-case of the regular assimilation of nasals to following consonants.
These same factors may have influenced the loss of non-prevocalic /-r/
in New Zealand (as in car, cart). Trudgill, Gordon, Lewis and Maclagan
(2002: 120-124) claim that a majority of immigrants to New Zealand were
rhotic (though many were probably variably so), while those of the second
generation were also rhotic (30% completely so, while most others were
variably so). However, New Zealand is today almost completely non-
rhotic. This case can be partly explained by a certain frequency of loss of/-
r/ in the prekoine linguistic pool, but the complete loss of non-prevocalic /-
r/ must certainly have been favored by its perceptually non-salient position.
Moreover, as these authors point out, the same phenomenon occurred and
was carried to completion at roughly the same time in Australia, South
Africa, and London itself, areas which were all characterized by both dia-
lect mixing, variable use of this feature, and this structural/cognitive factor.
As these authors argue, it is the repeated combination of similar factors and
structures which underlies the common "drift" of these dialects.
Finally, it is important to note that Kerswill and Williams (2002: 103)
extend the concept of salience (and the potential impact of perceptual sali-
ence) to discussion of syntax and discourse. Specifically, they analyze rec-
ognition of the (often commented upon) like, used as a focus marker and
quotative before key segments of utterances (e.g., Like Wow!), and com-
pare it to the clause-final discourse marker like (e.g., I did that when I got
home, like), which shares some of the functions of the more notorious like.
While participants in their study showed a uniformly high recognition rate
of the first, they showed a corresponding low recognition rate for the sec-
ond. While the first is spreading around the world, the second is showing a
decline in use. Following Yaegor-Dror, Kerswill and Williams suggest that
one reason for the low recognition rates and decline of clause-final like is
its lack of prosodie prominence.31
Finally, it is important to discuss briefly one last aspect of Trudgill's
(1986) presentation of accommodation: its limitation to adults. In later
work, Trudgill highlights the importance of adult speakers in situations of
dialect contact and mixing (particularly in regard to simplification; see
below). While certainly not ignoring the importance of child language ac-
quisition and use, to which he attributes focusing, he does not pay much
attention to child language when discussing the sources of innovations that
become established as changes in a koine. While it seems unquestionable
that accommodation by adults is important in the process of koineization,
in purely logical terms it need not be so: adults could refuse to alter their
speech patterns in a situation where speakers of different dialects were
mixed together, but the process of koineization (with all its features) would
occur unhindered so long as children learned from their parents and older
siblings, and then accommodated/learned among each other, as they gener-
ally do (there is evidence of this in the Milton Keynes study; see below).
Indeed, Mesthrie has argued that the primary importance of accommoda-
tion by adults during koineization is that it leads to a neutralization of the
social meaning attached to the linguistic variants affected, since variation
ceases to correlate clearly with non-linguistic factors such as region, social
status, or style (Mesthrie 1994: 1866). The question of the relative impact
of adults and children will be returned to below.
2.2.2. Interdialect
Features may appear in new dialects arising from dialect contact that are
not present in any of the contributing dialects; Trudgill (1986: 62) identi-
fies these as interdialect features. His thorough discussion of interdialect
focuses on the result of contact between two stable dialects, and he pro-
vides numerous examples of new phonetic and lexical interdialectal forms
as well as novel grammatical form-function reanalyses (presumably new
functions for established forms would also be interdialectal, but these can
also be analyzed in the overlapping category of reallocation). With regard
to the first type, Trudgill claims that there are two principal types of inter-
dialect phonetic features: fudged and mixed. In a fudged dialect, speakers
exposed to two different pronunciations opt for a phonetically intermediate
sound; in a mixed dialect, speakers substitute one sound for another, gen-
erally through the process of lexical diffusion, in which words are trans-
ferred one-by-one from one lexical set (with a characteristic pronunciation)
to another. It should be pointed out that fudging and mixing as contrasting
concepts are uncontroversial in the phonetic component of grammar: the
blending or analog nature of the phonetic system allows fudging, but the
phonology, morphology, and syntax are generally considered discrete com-

binatorial systems where only mixing is an option.32
Other types of mixed interdialect features result from "incomplete ac-
commodation". For example, Trudgill identifies the hybrid lexical form
/juba/ 'to work', used by residents of Oslo who immigrated from Sunndal,
Norway, as a cross between the older Oslo Norwegian form /jaba/ and the
Sunndal form /jub/ (Trudgill 1986: 63). Form-function reanalyses (cf. Croft
2000) lead to interdialect forms that may not be intermediate in any simple
or straightforward way. A prime example of this is the development of
forms (and corresponding functions) of the verb to do in the contemporary
English of younger residents of Reading, England (Cheshire 1982, reported
in Trudgill 1986: 65). Here, interaction has occurred between the tradi-
tional Reading dialect, in which different do forms were used to distinguish
between main and auxiliary do, and Standard English, where do forms are
used to distinguish person and number (see Table 3).
Table 3. Form-function reanalysis of Reading do. Source: Trudgill (1986: 65).

Standard English Original Reading Younger Reading
I do it, do I? I dos/does it, do I? I does/do it, do I?
He does it, does he? He dos/does it, do he?, He do/dos/does it,
do/does he?
In the process of shifting towards Standard English usage, some younger

speakers have created the novel construction he do it. In the traditional
Reading dialect, do is restricted to use as an auxiliary form, but as speakers
accommodate towards the standard, they begin to extend its use to mark
the 1st and 2nd persons singular and plural, as well as 3rd person plural. In
a typical case of overgeneralization, they extend it to the third person sin-
gular, thereby using it in a way that it is never used in either the standard or
in the original Reading dialect (however, in this case it remains a minority
usage). It is clear from this example that interdialectisms are often simply
hypercorrections. Trudgill gives a straightforward example of Northern
British English hypercorrection toward the perceived (largely southern)
standard; for the word but, northern speakers "correctly" shift their /but/ to
southern-standard /bAt/, but they hypercorrect when they shift butcher
from northern and southern /but/a/ to /bAt/a/.
Interdialect can also include statistical Labov-hypercorrection, in which
speakers use a variant more frequently than they normally would (Trudgill
1986: 123). Such frequency-enhancing changes in production are probably

the most significant interdialect influence on outcomes in koineization.
Among such changes are hyperdialectalisms, yet another category within
Trudgill's taxonomy of interdialect. These are forms (or form-function
reanalyses) that come to be seen as markers of community identity and
whose use is therefore altered or exaggerated (which in turn may further
contribute to the identity-marking function). As an ideal example of hyper-
dialectalism, Trudgill refers to Labov's (1963) well-known study of Mar-
tha's Vineyard. Labov showed that the islanders who identified strongly
with the local community and wished to remain there had more centralized
realizations of the diphthongs /ai/ and /au/, already centralized in the local
dialect, than those who did not so identify. These islanders took a pre-
existing feature of the local dialect and exaggerated the difference between
their articulations and the uncentralized ones of the mainland as they re-
acted to and resisted the penetration of mainland norms and identity.
Trudgill's identification of interdialectal features in the prekoine lin-
guistic pool is an important addition to the model. Still, his discussion of
interdialect is limited almost entirely to the effects of contact between sta-
ble dialects. Though on some occasions he claims to see the development
of interdialect features as important to multidialectal contact situations
(and gives one example from Fiji Hindi in his discussion of koineization,
as well as a few potential cases of phonetic fudging and mixing), he also
states that "genuinely intermediate forms are . . . less likely" in "dialect
mixtures where more than two contact varieties are involved" (Trudgill
1986: 65). This may be partly true, but it has the unfortunate effect of rein-
forcing the traditional belief that koineization is nothing more than a kind
of reduction to a least common denominator (in Chapters 3 and 4 I explore
the origins of several features of Castilian which show that the introduction
and propagation of interdialectalisms may on occasion lead to the survival
of apparently marked forms in a koine).
Finally, we should note that the concept of interdialect is adapted from
Selinker's (1972) concept of interlanguage. This is a fundamental construct
for understanding of second language acquisition, and its integration into
Trudgill's model also calls into question any meaningful distinction be-
tween long-term accommodation and dialect acquisition. In addition, I
would argue that there is reason to maintain the original term and concept
of interlanguage in discussion of koineization. When Trudgill uses the term
interdialect to refer to novel forms and uses of speakers, these speakers are
necessarily assumed to be native speakers of one of the dialects in contact.
This is very likely a reasonable supposition for contact between two stable
dialects spoken by communities with close-knit social networks. However,
if we accept the definition of contributing varieties given above, it will be
necessary to include the features of interlanguages of "non-native" learners
of the "language" (i.e., the collection of dominant forms and features in
contributing varieties), since, in the mixed linguistic pool and social situa-
tion of the prekoine, the features of their speech may be less identifiable as
"foreign" and thus find easier acceptance in the developing linguistic
norms of the new community (cf. LePage 1992; see below for further dis-
cussion of interlanguage). Interlanguage can thus be used as a cover term
to refer to both interdialect and interlanguage per se.
2.2.3. Focusing
Trudgill discusses Siegel's concept of stabilization within the framework

of focusing, a concept closely associated with the research of LePage and
Tabouret-Keller (e.g., 1985):
Le Page and Tabouret-Keller have pointed out. .. that speech communities,
and therefore language varieties, vary from the relatively focused to the rela-
tively diffuse. The better known European languages tend to be of the fo-
cused type: the language is felt to be clearly distinct from other languages;
its 'boundaries' are clearly delineated; and members of the speech commu-
nity show a high level of agreement as to what does and does not constitute
'the language'. In other parts of the world, however, this may not be so at
all, and we may have instead a relatively diffuse situation: speakers may
have no very clear idea about what language they are speaking; and what
does and does not constitute the language will be perceived as an issue of no
great importance. (Trudgill 1986: 85-86)
Many factors influence the degree to which a language variety is consid-
ered focused or diffuse. Primary among them is a common cultural identity
shared by the members of a speech community. Numerous factors can con-
tribute to the development of a common identity and thus to focusing. In
modern nation-states, standardization (with nationalism) is clearly an
important cause of focusing, as the highly focused European languages
attest. Isolation from outside influences and competing national or ethnic
identities will also aid in the definition of a common identity and
corresponding linguistic norms. On the other hand, the effects of
continuing contact without the creation of a new clearly-defined common
identity are found in the classic diffuse language situation of Belize (cf.
classic diffuse language situation of Belize (cf. LePage and Tabouret-

Keller 1985), long a colony with little national heritage, and where an un-
standardized Creole competes with highly focused English and Spanish.
In cases of dialect mixing (at least), Trudgill also sees isolation as an
important factor, along with the effects of child language acquisition. He
suggests that focusing of koineized varieties will be more likely and there-
fore easy to study in cases of language transplantation to new towns, re-
gions or colonies which are distant from the source communities. One of
the clearest and best known examples is offered by the new town of
Hayanger (Omdal 1977; Kerswill [2002] describes two similar cases stud-
ied by Sandve 1976). Hoyanger is a Norwegian town that in 1916 had a
population of only 120 inhabitants, but through industrialization grew to
have 950 in 1920 (but only 3000 by 1986). In 1920, 28% of the population
came from the immediate vicinity of H0yanger, 32% from other parts of
the surrounding county (split into two dialect zones), and the other 40%
came from other parts of Norway. Over the course of three generations
(first = adult immigrants), the Norwegian spoken in the town moved from
non-systematic variation that reflected the origins of the original immi-
grants, to more limited variation among the second generation, to a new
relatively unified and distinctive set of norms that characterize the speech
of the grandchildren (see below for further discussion). Similarly, Bortoni
(1991) includes a detailed description of the speech of young people of the
new Brazilian capital of Brasilia. Brasilia was founded in 1960 (in its sec-
ond generation at the time of Bortoni's study), and the speech of young
people there shows indications of incipient focusing and stabilization of
new norms (see below for more recent discussion of focusing).
It would thus appear that child language acquisition must play a signifi-
cant if not critical role in focusing or stabilization, but Trudgill (1986) does
not direct further attention to this issue.33 In fact, his only mention of the
social causes of this focusing is the following:
The focusing process may have been aided by the fact that, while workers
and managers originally lived in different parts of the town, this is today no
longer the case. As a consequence, there is little social dialect differentiation
amongst the youngest generation. (Trudgill 1986: 96).
Of course, as Trudgill himself pointed out, he was primarily interested in

the linguistic aspects of focusing, rather than the particular social factors
that lead to it. Nevertheless, it has become abundantly clear in subsequent
years that focusing and rate of focusing can only be understood if attention
is paid to the importance of child learning in focusing/stabilization as well
as the development and solidification of peer groups and social networks

(as also suggested in Trudgill's later publications; see below).
2.2.4. Mixing, leveling, reduction
Trudgill uses these terms to refer in different ways to the effects of what is
essentially the primary linguistic process that accompanies focusing of the
koine. Mixing emphasizes the selection and incorporation of linguistic
features from different dialects into the resultant koine. Leveling, on the
other hand, is used to focus on the exclusion of certain features, and
Trudgill defines it as the "reduction or attrition of marked variants" present
in the initial dialect mixture (where marked is used to refer to forms that
are in the minority in the set of contributing varieties). It is reduction not
with reference to any one contributing variety, but rather to the sum total of
all variants in all contributing varieties, including not only the pre-existing
features of established dialects but also the novel features of interdia-
lects/interlanguages (Trudgill 1986: 97).34 He identifies the primary source
of mixing/leveling/reduction as the accommodation by adults who suppress
marked features of their speech. Clear examples of this can be found in Fiji
Hindi (Table 4), where the surviving forms are those which were shared by
a majority of contributing dialects (or, rather, used by a majority of speak-
ers).
Table 4. Lexical mixing in Fiji Hindi. Source: Trudgill (1986: 101).

Standard Bhojpuri Avadhi Fiji Hindi
Hindi
'what' kja: ka: ka: ka:
'who' kaun kaun~ke kaun~ko kaun
'one's own' apna: a:pan a:pan apan
Trudgill supposes that "commonness" is the primary reason for the survival
of any linguistic feature. In fact, the organization of the above chart reflects
a perhaps too simple metric:
As far as levelling is concerned, we can note that. . . forms that occur in a
majority of the contributing dialects win out and survive in the emerging fo-
cused dialect. (Trudgill 1986: 143)
Siegel (1993b: 116) agrees with this general idea but criticizes Trudgill ' s
exact conceptualization, and argues that it is the overall frequency of oc-
currence that most matters, with this being initially dependent on the total
number of speakers who use any particular form, be they from the same or
different dialects. In more recent work (Trudgill, Gordon, Lewis, and
Maclagan 2000), Trudgill and his colleagues follow Siegel's suggestion,
and even find that the large size and cohesiveness of any particular group
(its possible use as a target) is unimportant in determining the final results
of koineization: those features that are used by the maximum number of
speakers are those that are generally selected.
Trudgill's original emphasis on number of dialects, as opposed to
speakers, may have reflected a earlier preference for emphasizing systems
over speakers (but note that it sometimes represents a necessity in histori-
cal work at great time depths, where more specific information may not be
available). Trudgill (1986) does recognize, however, that speakers must
have the opportunity to use variants they know. He points out, for example,
that socially marked variants, even very common ones, are likely to be
eliminated along with the contexts of use for which they were reserved. A
clear example of this phenomenon is found in Guyanese Bhojpuri (Gamb-
hir 1981: 255). Whereas Indian Bhojpuri is characterized by a complex
pronoun and suffixation system that is used to indicate respect for the inter-
locutor, in the new social environment of Guyana, the immigrants, largely
of the same caste (and in a situation requiring mutual support and solidar-
ity), no longer had need of the inter-caste cultural conventions of the caste
system or the linguistic forms that were associated with that system. As a
result, forms marking the respect feature were lost in modern Guyanese
Bhojpuri.
When minority forms survive, Trudgill suggests that this is because

other factors interfere, but these exceptions are often difficult to explain.
Sometimes interdialect forms are the ones chosen, particularly when there
is no clear majority form (Trudgill 1986: 101). In other cases, ad hoc ex-
planations become the only recourse; Trudgill provides an example where
the shortest of several competing forms was chosen; in another case, forms
with oral vowels were selected over forms with phonologically-marked
nasal vowels (Trudgill 1986: 102). In many cases, however, there is no
clear explanation of why a particular form was chosen, though perhaps
further analysis would reveal such factors in these cases. Siegel (1993b)
also considers Trudgill's inability to deal with such exceptional cases:
it seems that there would have been greater chance for success if he had
based his generalizations not primarily on linguistic characteristics (compar-
ing forms from different dialects) but also more on the demographic charac-
teristics he mentions, including information about the social context.
(Siegel 1993b: 116)
Attention to social context has in fact been the component of Trudgill's
original model most in need of further development, although Trudgill
does not completely ignore it (see below for further discussion).
2.2.5. Reallocation
Sometimes not all the variants (presumably only pre-existing variants in

this case) in the prekoine linguistic pool are leveled or reduced to just one.
Trudgill (1986) makes the important observation that variants that had the
same function/meaning in different established contributing varieties may
survive, but the function/meaning of each variant is altered, or reallocated,
so that they no longer compete directly. Thus, in socio-stylistic reallocation
(Britain and Trudgill 1999), originally regional, social or stylistic variants
may acquire new functions as social, stylistic or areal variants. Trudgill
also suggests that different phonetic variants in the prekoine linguistic pool
may be retained as allophonic variants in a koine, in a process known as
phonological reallocation (Britain 1997a; Britain and Trudgill 1999).
Particularly frequent is the reallocation of regional variants to a stylistic
function; the examples in Table 5 are taken from Mauritian Bhojpuri.
Table 5. Stylistic reallocation in Mauritian Bhojpuri. Source: Domingue (1981),

reported in Trudgill (1986: 65).
eastern central western high Mau- low Mauri-

India India India ritius tius
'big' /bara:/ /ba^a:/ /ba^a:/ /bara:/
'temple' /mandir/ /mandil/ /mandil/ /mandir/
'road' /ra:hta/ /ra:sta/ /ra:sta/ /ra:hta/
Siegel (1997: 127) provides evidence of a similar occurrence in Fiji Hindi,

in which the different variants appear to mark social class. In this case,
there exist two forms of the third person possessive, /okar/ from the Bho-
jpuri dialects, and /uske/ from Hindi varieties. The Hindi form has become
the norm in Fiji Hindi, while the Bhojpuri form is considered rustic and
employed in drama and comedic routines to characterize rustic characters.
Socio-stylistic variation can affect phonological variants as well. For ex-
ample, Britain and Trudgill (1999) report that in Australia the original Brit-
ish pronunciations of the vowel in the lexical set dance, sample, plant have
been reallocated with new stylistic functions. Original northern British /ae/
has become the lower-status norm in Australian English, while original
southern English /a:/ is retained as a high-status variant in formal registers.
Reallocation can also lead to the creation of novel areal variants in a koine.
Siegel (1997: 127) points to a case in which a dialect difference in India
becomes a marker of island residence in Fiji Hindi. Speakers who live on
the northern islands of Vanualevu and Taveuni insert a back glide before
the perfective suffix -ä while speakers from the main island of Vitilevu
insert a front glide, so that 'sang' is gäwä on the northern islands and gäyä
on the main island.
Cases of phonological reallocation are less easy to identify, but Trudgill
(1986), Britain (1997a), and Britain and Trudgill (1999: 251-254) have
argued that Canadian Raising (and similar phenomena found in the English
Fens and other English dialects that have arisen through dialect mixing) is
the result of phonological reallocation during koineization. In Canadian
Raising, the /ai/ diphthong is subject to allophonic variation. It is pro-
nounced with an open onset before voiced consonants and word-finally
(e.g., time, tie), but with a centralized onset (e.g., [ai]) before voiceless
consonants (e.g., night). The wide diphthongs are and were typical of
southern English dialects, while the narrow diphthongs are and were typi-
cal of Scottish dialects (among others). According to Trudgill and Britain,
as speakers from these areas mixed in Canada, their allophones were re-
tained, but according to regular phonetic principles of English: the shorter

narrow diphthongs were confined to the pre-voiceless consonant environ-
ment where all English vowels have shorter allophones; the broad diph-
thongs were retained elsewhere.
Trudgill (1986) also adduces as evidence of phonological reallocation
the wide range of phonetic variants that characterize the articulation of /a/
for many working-class speakers in Belfast, many of whom have rural ori-
gins and moved to the city to find work. He argues:
the very wide range of allophones of /a/ in working-class Belfast English,
found also in the case of some other vowels, is typical not so much of ver-
nacular varieties [as James Milroy argues] as of mixed varieties. My sugges-
tion in fact is that different phonological variants present in the dialect mix-
ture situation, and not leveled out during focusing, may be retained as allo-
phonic variants. (Trudgill 1986: 125)
Milroy (1982, 1992) holds very different views on the matter, and suggests
that it is the close-knit social networks of the inner-city neighborhoods that
allow them to retain and add variants in their speech, which serve as mark-
ers of identity against other groups and against middle class speakers, who
have much more loose-knit social networks and who generally show far
more restricted allophonic variation for /a/. In fact, Trudgill (1986) treats
the case of Belfast and its developing city-wide vernacular(s) as if it were
the same type of process as that which occurred in Heyanger (or Milton
Keynes), but, as the Milroys' research shows, this is not quite the case. For
koineization to occur, there must be a massive influx of immigrants at
roughly the same time, with a simultaneous breakdown in social networks.
This is not what occurred in the history of Belfast, where the in-migration
was more gradual. Moreover, in mixed, koineized dialects, we would ex-
pect the general tendencies toward reduction and simplification to operate
against the survival of high numbers of allophones.35 Thus, although
Trudgill (1986) views reallocation as a process distinct from leveling, it
may be better to understand it as a sub-process of mixing or leveling (or of
acquisition; see below), since in general, as the above examples show,
there is still reduction in the number of regional and social variants, but
more than one survives.
2.2.6. Simplification
Trudgill considers simplification to be an alternative process of variant

reduction (as against leveling), and it is frequently cited as a primary result
of koineization.36 Simplification, however, is a term that has been used by
many scholars with a variety of meanings. Trudgill provides the following
brief review and definitions:
Simplification is . . . a difficult and perhaps dangerous notion. Mühlhäusler
has argued (197[4]) that simplification can be taken to refer to 'an increase
in regularity', and that it is a term that should be used relatively, with refer-
ence to some earlier stage of the variety or the varieties in question. There
are, according to Mühlhäusler, two main types of simplification. The first
type involves an increase in morphophonemic regularity, and would include
the loss of inflections and an increase in invariable word forms. Mühlhäusler
also points to Ferguson's (1959[b]) discussion, where the following are
listed as indications of simplification: symmetrical paradigms; fewer obliga-

tory categories marked by morphemes of concord, and simpler morphopho-
nemics. The second aspect of simplification involves an increase in the
'regular correspondence between content and expression', which is intended
to refer to an increase in morphological and lexical transparency.
(Trudgill 1986: 103)
Though we will see cases in which it is necessary to distinguish between
regularity and transparency (at least when different levels are involved; see
below), they most often work in tandem, and the range of surface effects
listed above can generally be seen as part of a more general phenomenon:
the reduction in inventories of units and rules within the most systematic
components of the language (phonology, morphology, and syntax). The
elimination of irregular forms and exceptional rules can easily be seen to
increase both regularity and transparency.37 For example, the loss during
the Middle Ages of Spanish strong preterite forms and their replacement by
predictable forms results in greater regularity and transparency: e.g., es-
crise Ί wrote' has become regular escribí, cf. infinitive escribir, mise Ί
put (in)' has become regular metí, cf. infinitive meter (Penny 2000: 52
gives numerous other examples). Nevertheless, there also exist cases in
which regularity in one component, such as phonology, can lead to greater
complexity (opacity) in another, such as the morphology (this is a well-
known Neogrammarian principle), and this seems to be what Mühlhäusler
and Trudgill are referring to when they specify the two types of simplifica-
tion. For example, Kerswill and Williams (2000) show that the Ro-
land/rolling distinction, a case of relatively complex allophonic variation
(not found in all contributing dialects) has been maintained during koinei-
zation in Milton Keynes (the vowel of the open syllable of each word is
pronounced with a different allophone), in part because its maintenance

allows for a more transparent relationship to be maintained between the
morpheme roll and its dimorphemic derivative rolling (pronounced with
the same vowel; Britain [1999] presents a similar kind of argument; see
below for detailed discussion).
The reduction in inventories of units and rules is clearly the result of
imperfect language learning (see below). It is favored by discrepancies
between contributing dialects, which lower frequency and consistency of
input, but the simplified forms themselves, generated through overexten-
sion of dominant patterns, must arise in learner language. If this is so, sim-
plification should be understood as part of the leveling process, but with
Conditions ofspeaker activity 47
the inclusion of novel (learner-generated) simplified variants in the linguis-

tic pool.
Trudgill (1996: 10-11) also extends simplification to include the proc-
ess of lenition, which he finds is frequent in looseknit societies affected by
dialect contact and mixing. He bases this claim on Harris' (1990) formal
definition of lenition, made within the framework of Government Phonol-
ogy. According to Harris, all lenition phenomena can be characterized as
"segmental decomposition", the subtraction of one or more elements from
the internal structure of a segment. This approach to lenition would support
Fontanella's (1992) inclusion of weakening of syllable-final /-s/ as a typi-
cal result of koineization in American Spanish (but note too that final /-s/
lacks perceptual salience; nevertheless, the variable aspiration and loss of
this consonant could be seen as not well justified, since such weakening
can compromise transparency and regularity). Thus, [s] can be character-
ized as composed of a coronal articulation and a narrowed articulatory
stricture; the move from [s] to [h] is a loss of coronal articulation, and loss
of narrowed stricture leads to loss of the segment. Nevertheless, lenition
itself is not always so easily characterized. A more generalizable analysis
is possible: when frequently-produced "natural" phonetic variants are
added to the prekoine linguistic pool, they may survive the process of lev-
eling as additive features of the resultant koine.
3. Elaborating the process model: Conditions of speaker activity
Trudgill's (1986) presentation is most noteworthy for its detailed specifica-

tion of the macro-level linguistic processes or outcomes that characterize
koineization. For the micro-level, he focused his discussion almost exclu-

sively on the process of accommodation by adults, but largely ignored de-
tails of the social context. As an explanatory model of language change, it
requires further elaboration. Siegel (1993b) discusses this problem in his
review of Trudgill (1986). He points out, for instance, that various dialects
can be in contact for extended periods without any mixing or leveling tak-
ing place:
the contact status quo may end . . . with certain political, social, economic,
or demographic changes . . . This may bring about either increased interac-
tion among speakers of various dialects or decreased inclination to maintain
linguistic distinctions. (Siegel 1993b: 117)38
This is, in fact, an allusion to the Actuation problem (see Chapter 1). Ac-
cording to Siegel, most treatments of the issue have been cursory.
Domingue (1981: 150), for example, vaguely ascribes koineization to "the
need for unification among speakers of different dialects in a new envi-
ronment". Others simply assume, without further comment, that it occurs
most frequently in cases of migration and colonization. About this Siegel
comments:
But social and demographic changes, as in migration, do not necessarily
bring about koineization. For the processes to begin people have to give up
(consciously or unconsciously) their own speech distinctions and conform to
the speech of others. Part of a theory of dialect contact would need to be
able to predict when this would occur . . . Trudgill uses the accommodation
of social psychology to show that people may modify their speech in face-
to-face dialect contact situations. But . . . [h]e focuses on the linguistic
changes themselves rather than on the aspects of the theory that attempt to
explain why sometimes these changes occur and why sometimes they do not.
(Siegel 1993b: 117)
The gaps outlined by Siegel were noticed by Trudgill himself, and in a

series of later articles (e.g., Trudgill 1992, 1994, 1996, 2002) he shifts his
attention to the influence of social structure, particularly in terms of social
networks, on speaker activity (and on language structure), and the con-
straints imposed on this activity by the language acquisition process. His
repeated claim in these publications is that smaller societies characterized
by closeknit social networks, overall stability, and isolation, are more ca-
pable of promoting the learning of a language variety with minimal change,
and that over time speakers of such varieties are more likely to retain and
add complex features. Conversely, (usually larger) societies characterized
by more looseknit social networks and high levels of contact are more
likely to tend toward simplification, hybridization, and higher rates of
change. These factors, and how they affect koineization, will be discussed
below. First, it is necessary to take into account the importance of norms,
norm enforcement and social network structure - or lack thereof - at the
micro-level of koineization. Even more importantly, one must understand
how the specific learning context of koineization affects language/dialect
acquisition by children and adults.
Conditions ofspeaker activity 49
3.1. Norms, norm enforcement, and social networks
An understanding of norms is essential to the development of an integrated

model of koineization, since they define the social conditions that limit
speaker actions as well as the end-products of collective speaker activity on
the macro-level. Norms (or conventions as they are sometimes known, cf.
Croft 2000) are socially-determined limits on behavior; in the case of lan-
guage, they define correct linguistic products for a particular community.
Norms exist and are enforced because they are socially useful. Bartsch
(1987: 126) outlines two primary functions of norms. First, they ensure
communication by restricting the choice of actions, making them predict-
able, and thereby supporting the information-bearing function of language.
Second, they serve to indicate group identity:
From the point of view of group identity, adherence to the norms is watched
closely, much more than would be necessary for the mere functioning of a
language as a means of communication. This is especially true of acts of
correction that apply to those linguistic forms that symbolize the internal so-
cial order of a group or society: special morphological and syntactic forms,
as well as vocabulary and styles that are connected to roles, position and so-
cial status. (Bartsch 1987: 87)
At the level of language change, these two functions of norms enter into
conflict. The communicative function is conservative and promotes stabil-
ity and uniformity. The identity function, on the other hand, often leads to
the development of conflicting norms and overall fragmentation (Milroy
1992: 38), at least in more stable societies.
Norms, which are natural to all human societies, need to be distin-
guished from prescriptions, with which they are often confused. A pre-
scription may arise when a norm is codified and then extended beyond the
community where the norm originated, or when a norm changes but an
earlier codification of the norm does not. Norms also need to be distin-
guished from rules. Norms describe the linguistic system of a community,
the langue of Saussure or the Ε-language of Chomsky, and though they are
clearly abstractions based on dominant patterns of behavior, speakers (and
linguists) are aware of them and influenced by them. Linguistic rules de-
scribe the internalized systematizations (or competence) that speaker-
learners construct based on the norm-governed output of other speakers (=
learner input). Speaker-learners are exposed to norm-governed behavior
and abstract the best they can from this behavior, but their rule systems
often fail to match the norms exactly. This can be seen in the fact that these
abstract rule systems tend to overgenerate, producing patterns (or innova-

tions) that are unacceptable (see discussion on acquisition below). Con-
straints on pattern generation - or learning - are thus determined by norms.
Speakers constantly deviate from norms (for varied reasons), so norms
must be enforced, but how is this done? Bartsch sees this as an active proc-
ess and suggests that norms are enforced through correction, criticism,
ridicule, stigmatization or even exclusion. Such emphasis on overt correc-
tion may be valid in the case of salient and socially-stereotyped features,
but indirect correction through constant and consistent presentation of
"correct" models is certainly fundamental for less salient features. The
issue of norm enforcement has been a principal topic in the research of
Lesley and James Milroy, who, in numerous sociolinguistic studies of the
speech of inner-city Belfast, focused on the issue of language maintenance
in non-standard working-class speech communities (many of the results of
which are summarized in Milroy 1987). The Milroys have made certain
fundamental and useful recommendations that allow for clearer demarca-
tion between the micro-level conditions of speaker activity and macro-level
linguistic processes or outcomes. Namely, speakers introduce innovations
(through fluctuation in pronunciation, accommodation and mixing of
codes, incorrect learning, processing errors, etc.), but changes are seen to
occur only at the level of the linguistic system, understood as the product
of speaker activity in social contexts. They too view the linguistic system
as synonymous with the consensus norms of the community, so linguistic
change is change in agreement on the norms of usage. Like Bartsch, the
Milroys emphasize the functionality of norms, but the norms of dialects
depend far less on communicative needs:
There would be little point in having these different norms (which are arbi-
trary in linguistic terms) if they did not carry social meaning, distinguishing
between one community and another and carrying a sense of community
identity for speakers. (Milroy 1992: 83)
The conservation of different norms in low-status urban dialects in the face

of pressure from the standard and other varieties is inexplicable unless one
takes identity into account; the mechanism which allows such conservation
of minority identity and norms is the close-knit social network (Milroy
1987).
For the individual, a social network is the sum of the relationships
which he or she has contracted with others. At the societal level, social
network amounts to a boundless web of ties that permeate an entire society,
and generally extend beyond it as well. Any particular social network will
Conditions of speaker activity 51
have particular structures and properties. In sociolinguistic research, Mil-

roy (1987) has defined social networks in quantifiable terms of density and
multiplexity. A social network is maximally dense if everyone knows eve-
ryone else, and multiplex if each person knows each other in a variety of
capacities (e.g., as family member, friend, neighbor, colleague, schoolmate,
fellow parishioner). Multiplexity is a useful measure of the strength of any
particular tie because it can be measured, but it is only an indirect indica-
tion, and is most useful because it reflects other potentially more signifi-
cant factors:
The strength of a tie is a (probably linear) combination of the amount of
time, the emotional intensity, the intimacy (mutual confiding) and the recip-
rocal services which characterise a tie. (Granovetter 1973: 1361)
It is clear, too, that strong ties must be limited in number. In contrast, weak
ties are characterized by their uniplexity and high number. All social net-
works are built on a combination of strong and weak ties, but close-knit
social networks characterized by a high number of strong ties serve as
norm enforcement mechanisms that tend to impede change of all types,
including linguistic change. The obvious corollary to this is that loose-knit
social networks with numerous weak ties permit change (Milroy and Mil-
roy 1985: 370-375; Milroy 1993: 227; Trudgill 1992). Indeed, the Milroys,
following Granovetter (1982), assert that weak ties serve as the very chan-
nels of transmission for all types of innovations.
According to the Milroys, the weak-tie model is sufficiently abstract to
be applied to other situations besides the close-knit urban communities
investigated in Belfast. In fact, the greatest advantage of the social network
model is its universality:
The fundamental [advantage] in an investigation of this kind is that we do

not need to accept any prior assumption about how society at large is organ-
ized or structured . . . it is a universal that all individuals in all societies have
contact with other individuals. (Milroy 1992: 85)
Contact situations are one area where an understanding of network rela-
tions is crucial. Milroy suggests that where speakers of different varieties
come into contact but also maintain strong social networks, stable bilin-
gualism may result (1992: 200; cf. Siegel's comments quoted above). How-
ever, mixing of speakers of different varieties with the breakdown of social
networks, as we would expect in koineization, will strongly favor linguistic
change, since the number, frequency and relative proportion of weak ties
increases dramatically.
The model of weak ties provides an important support to the model of

koineization. When there is great social or geographic mobility, weak ties
will predominate. Whereas strong ties promote local cohesion and overall
fragmentation, weak ties promote greater overall uniformity and, in certain
contexts, simplification (Milroy and Milroy 1985; Milroy 1992: 178). In
the inner city, the Milroys found that complex allomorphy functioned to
mark identity in close-knit social networks. However, middle-class speak-
ers in outlying suburbs of Belfast generally had low network indices (i.e.,
loose-knit social networks), and it was found that the speech of many of
these informants was characterized by a radical reduction in phonetic in-
ventory (Milroy 1992: 98). For example, the six inner-city allophones of
the phoneme /a/ (mentioned above) were often reduced to just one or two
in the middle-class and more socially-mobile outer city (Milroy 1992: 102).
Initially, this reduction occurs as a response to the change in speakers'
social context and the concomitant loss of functionality of certain variants.
This is demonstrated in another example given by Milroy:
one man of 27 commented that during his years away from Northern Ireland,
he had stopped using what I have called the in-group alternants of the (pulí)
set. The most obvious explanation for this is that during his years away from
home, he would develop a large number of relatively weak ties: in such cir-
cumstances, the in-group alternant would cease to have any function for him
and so could be abandoned. This functional explanation seems to be more
satisfactory than, and logically prior to, one based on prestige: as it happens,
this individual's activities away from home had not been upwardly mobile.
(Milroy 1992: 99-100)
The breakdown of consensus (found in outlying middle-class neighbor-

hoods) may lead to patterns of usage that are less predictable by linguistic
rule or by sociolinguistic pattern, with some apparently random variation,

or some lack of agreement on norms of usage (Milroy 1992: 104-105).
This reduction of variants and apparent randomness resembles in some
respects that of the sociolinguistically chaotic prekoine, and this resem-
blance is not casual:
resistance to merger, lexical transfer and restructuring is promoted by the
existence of strong network ties, and . . . types of change that result in sim-
plification are encouraged by weakening of ties. Complex patterns, which
are functional, are maintained by the norm-enforcement function of dense
and multiplex networks. For those whose ties are uniplex and (relatively)
open-ended, these patterns are no longer functional and it is for this reason,
Conditions of speaker activity 5 3
and not primarily because of speakers' desire for 'prestige', that they disap-
pear. (Milroy 1992: 107-108)
This loss of social functionality of linguistic features is especially impor-
tant in language varieties used by larger and more mixed communities:
As Jakobson perceived long ago, those varieties that have supra-local func-
tions and that tend to develop in the direction of koines display simpler pho-
nemic systems than varieties that have purely local functions . . . It may well
be that, in varieties that have supra-local functions, a high degree of com-
plexity (at any level) is indeed dysfunctional. (Milroy 1992: 108)
In koineization, speakers of different dialectal origin come together, leav-
ing behind their communities and established social networks. In this con-
text, many features of their speech lose their functionality, and, of course,
may become counter-functional in the new context to the degree that they
mark a speaker as different (through accommodation they may be sup-
pressed). Norms not shared by most speakers cannot be consistently en-
forced, and new norms can only be created as new social networks crystal-
lize. This is, however, a multigenerational process. When immigrants to a
new mixed community first come together, their social relations will be
based primarily on weak ties. After one or two generations, however, new
more close-knit social networks will often arise (particularly in more tradi-
tional societies, where such networks were probably more necessary for
survival) with the development of new social and linguistic norms. Why is
this the case? According to Mitchell (1986: 74), individuals create personal
communities, or social networks, which provide them with a meaningful
framework for solving the problems of their day-to-day existence. Indi-
viduals in a new context are likely to work to set up new social networks in
order to meet their problem-solving needs. These new networks will of

necessity begin as loose-knit networks based on weak ties (and may remain
as such, as in the case of many middle-class persons in mobile post-
industrial societies), but with time - measured in generations - may de-
velop into close-knit social networks. The development of new social net-
works and the focusing of new norms has been of primary interest in the
research of Kerswill and Williams, and it will be discussed below.
Another important aspect of the Milroys' research has been their at-
tempt to determine which persons act as innovators in close-knit social
networks, and they suggest that this is likely to be a person who is only
related to the group through weak links. Significantly, the innovator cannot
be someone who is thoroughly integrated into a close-knit social network,
although an early adopter must be (Milroy and Milroy 1985: 367-370;

Milroy 1993: 225-226). In a koineizing social situation, however, all the
speakers become innovators vis-à-vis speakers of other dialects. As a re-
sult, the number of relative innovations peaks at the same time that the
strength of norm enforcement mechanisms (social networks) declines to a
minimum. It is clear that one way in which the variation is reduced is
through accommodation in the strict sense: the simple avoidance by speak-
ers of features that mark them as different. However, as we have seen, ac-
commodation, by both adults and children, can also imply the learning of
new features, and this process can itself lead to the introduction of new or
additive features. In fact, language acquisition also plays a significant role
in the selection of innovations and the stabilization of new norms.
3.2. Language acquisition: Adults and children
Speakers in contact with speakers of different dialects do not continue to

speak the same way; at times they have to learn new features that are alien
to their normal ways of speaking. Trudgill refers to this as long-term ac-
commodation when it is successful, and as interdialect when the learning is
incomplete. This incomplete learning may lead to the creation of new
forms that are not present in any of the contributing dialects. This, in fact,
is true of all types of adult second language learning, and it is probably not
necessary to distinguish between second dialect learning and second lan-
guage learning, except as a difference of degree: the number of features to
be learned in the acquisition of a second dialect (and thus of potential in-
novations) will necessarily be smaller than in the case of a typologically
distinct language, and the dialect learner may not have to employ strategies
such as paraphrase with the same frequency as a prototypical foreign lan-
guage learner. Otherwise the similarity between second dialect acquisition
and second language acquisition is neatly captured in Selinker's concept of
interlanguage (1972, 1992).39
Interlanguage describes the range of intermediate systems that charac-
terize the second language speech of individual adults. These systems,
which in the early stages of acquisition are progressively restructured, of-
ten stabilize permanently or fossilize before the learner system reaches full
identity with the target language. Thus, interlanguage systems are unique
systems which are different from both the native language and the target
language. There are five psycholinguistic processes that shape interlan-
Conditions ofspeaker activity 5 5
guage. The first is native language transfer. Selinker, following Weinreich

(1968: 7), suggests that this happens as speakers make "interlingual identi-
fications" between the native language, their interlanguage, and the target
language. In other words, units are not only significant within any one sys-
tem, but also across systems. For example, a speaker may perceive a taxo-
nomic phoneme as being the same in all three systems, even though it pat-
terns differently or has different phonetic realizations in each system. The
second process is overgeneralization. In this case, the learner masters a
general rule, but fails to learn exceptions to the rule. For example, the fre-
quent production among learners of English of *goed, *drinked, and
*hitted for went, drank and hit shows overgeneralization of the rule which
specifies suffixation of -ed (in any of various phonetic realizations) when
the meaning 'past' is intended. A third process is transfer of training,
which occurs when the second language learner applies rules learned from
instructors or textbooks. This is significant only in cases where schooling
is available to a sizable proportion of the population, and in the discussion
of language change is particularly relevant to the process of standardization
and the impact of literacy. However, transfer of training essentially means
that the normal inferencing process is undercut and the learner is provided
with rules (whether correct or not) that can be deductively applied. This
may occur in non-classroom contexts when speakers engage in metalin-
guistic commentary on what is acceptable language practice in their com-
munity.
Communication strategies are used by the learner when the interlan-
guage system seems unequal to the task; thus, unlike transfer and overgen-
eralization, such strategies are less likely to affect the interlanguage of
second dialect learners, where mutual comprehensibility is high. Con-
versely, they will be especially frequent in the speech of learners of typo-

logically distant languages. Communication strategies include: avoidance,
paraphrase (approximation, word coinage, circumlocution), conscious
transfer, appeals for assistance, and mime (Tarone 1977). Certain of these
strategies overlap with those of accommodation. Thus, avoidance and
paraphrase remove tokens of an item or feature from the linguistic pool
(and sometimes replace them with new alternatives, though the reason for
doing so may be different in the koineizing context). Conscious transfer, as
in literal translations or code-switching, might also lead to greater overall
influence of certain speakers' native language variety in the prekoine mix.
Finally, learning strategies are, by and large, of little interest in discus-
sions of language change, with one exception: the cognitive strategy of
analysis. In order to produce forms and structures, the learner must cor-
rectly analyze them in the input he or she receives. If this input is incor-
rectly analyzed, the speaker may produce "incorrect" forms. These nor-
mally will be eliminated in the face of contrasting evidence, but when this
is absent or not perceived, or the speaker-learner has no further need to
continue altering output, then the errors may become fossilized in the
learner's speech (see below).
Second language learning can be quite difficult and almost never "per-
fect" for adults, even when they have plentiful input and interaction with
native speakers of a stable target variety (i.e., there is a difference between
input and intake). However, in a koineizing community with great variation
(i.e., without a clear target) and low norm enforcement, "errors" in learn-
ers' speech will be even harder to correct. Their incorrect production will
add further variation to the linguistic pool, and can also become input for
others. Of course, extreme departures from the shared basic system of all
speakers will probably not survive, but where the transfers, misanalyses,
and, especially, overgeneralizations are matched by many speakers (or
exist in one of the established contributing varieties), then the interlan-
guage forms/features stand a chance of survival in the koine.
The interlanguage model allows us to predict the introduction by adults
of innovative mixed forms, novel form-function analyses, as well as the
overgeneralized forms and uses that, when adopted by other speakers, lead
to simplification of the linguistic system. It also underlies an important and
controversial claim made by Trudgill. In his later publications (1989, 1992,
1994, 1996, 2002), he argues repeatedly that simplification (in all types of
language contact) is exclusively the result of adult second language (or
dialect) acquisition. For example:
simplification is due to the imperfect learning...by non-native adults and

post-adolescents. Everything we know about young children indicates that in
general they are such good language learners that they normally learn per-
fectly any language variety that they have sufficient exposure to. Imperfect
learning, and thus simplification, does not result from non-native language
learning as such but from adult non-native language learning.
(Trudgill 1992: 197)
Interestingly, he makes this claim in the midst of a discussion on the rela-

tion between the structure of social networks and the structural develop-
ment of dialects and languages. However, if we consider more closely the
tendencies of first language acquisition and how these are influenced by
the koineizing social context, we will see that Trudgill's conclusion must
be rejected.
To begin, adults are not the only ones who must learn in a koineizing
environment: children play a doubly significant role, in that they can pro-
duce innovations and it is child learning in conjunction with their estab-
lishment of new social networks that leads to eventual focusing (see be-
low). Indeed, as was argued earlier, were it not for the potential impact of
transfer, it would not be logically necessary to include adult accommoda-
tion and learning in the model of koineization: most of the same effects
would occur even if children were the only ones learning and introducing
innovations. Innovations in child language are often similar to those of
adult interlanguages. This is especially true of overgeneralizations, which
are likely to survive in child speech if children encounter inconsistencies in
adult usage and that of their peers.
The similarities between child and adult learning-based innovations
have been the object of both experimental and historical study. Bybee and
Slobin (1982) studied documented historical changes in the past forms of
English verbs and compared these to child and adult innovations in the
formation of past tenses of irregular verbs. For both adults and children
they found that the verbs most frequently regularized were those with the
lowest frequency of occurrence. Differences between child and adult over-
generalizations were minor. For instance, they found that about half of the
adult innovations were in the choice of the vowel in strong past tenses of
the ring, sing, swim class (e.g., shrunk for normal shrank), while nearly all
the child innovations for strong pasts in this class were of this kind, which
indicates that children are regularizing the change (or turning similar, fre-
quent adult innovations into changes).40 In a situation of weak norm en-
forcement, the combined tendencies of both adults and children to over-

generalize can lead to significant linguistic changes.
The fundamental similarity between child and adult acquisition is also
discussed by Ravid (1995), though she emphasizes that it is children and
non-literate adults - "naive speakers" - who are most prone to overgeneral-
ize or regularize (Ravid 1995: 91).41 More generally, however, child speak-
ers seek regularity and transparency in language far more consistently than
do adults, and this tendency causes them to produce "errors" that enhance
such regularity and transparency (Ravid 1995: 25). Children and non-
literate adults are seen to react against structural opacity in the sense of
Kiparsky (1982) and Lightfoot (1979). In so doing they rely on Operating
Principles of acquisition and processing (Slobin 1985):
— formal simplicity (leading to preservation of base forms when modified by

affixes, and back formation)
— formal consistency (Kiparsky's allomorphy-reducing trend or trend to
grammatical regularity)
— semantic transparency (seeking one-to-one correspondence between mean-
ing and form)
— saliency (perceptual distinctiveness of a feature makes it less subject to the
other principles, since salient structures are more easily learned and re-
tained; also marking information through the use of analytic structures, or
separate words which are by nature salient).
It is implicit in Ravid's discussion that the innovative results of such

strategies - "errors" - are likely to be eliminated only in the face of fre-
quent and consistent norm-governed input with which the child can com-
pare his or her output. When such frequent and consistent input is lacking,
then children will continue to produce "errors" that enhance regularity and
transparency. Of course, such behavior will either produce innovations and
lead to change, or, more likely, complete changes that have been initiated
in the adult community (which in the case of Israeli Hebrew is directly
linked to its status as a "resurrected" language whose first speakers were
second language speaker-learners; these speakers reduced the phonemic
inventory and in so doing made input for the complex morphology more
opaque for learners).
Simplification then is the result of both child and adult overgeneraliza-
tions which go uncorrected. This is also true of the mixed forms and form-
function reanalyses that Trudgill refers to as interdialectalisms. Like adult
second language learners, children must analyze the linguistic uses they
hear around them as they attempt to construct a grammar that corresponds

to community norms; how they go about this is explored in Andersen's
(1973) discussion of abduction and deduction (the flaws in this model have
recently been pointed, but it is worth considering Andersen's proposals
before discussing the problems). In the process of grammar construction, a
learner must construct a grammar based on the surface output of other
speakers (other speakers' output is by definition the input for learning).
According to Andersen, this is done through applications of the three
Peircean modes of logical inference: abduction, induction and deduction.
Deduction and induction are better known, but abduction is hypothesized
to be most important for understanding the relation between learning and
language change. The general relation can be shown in the following illus-
tration (from McMahon 1994: 94). Given a particular result (e.g., Socrates
is dead), and a law (e.g., All men are mortal), one abduces that something
may have been the case (e.g., Socrates may have been a man). Abduction is
notoriously unreliable (consider the consequences of invoking the law 'all
dogs are mortal'), but Andersen considers it immensely important because
it is the only one of the three modes of logical inference that introduces
innovations. In language learning, abduction operates on the output of
other speakers, and the other two modes are used as testing devices (pre-
sumably these abductions are made in accordance with something like
Slobin's Operating Principles). Induction occurs as the initial abduction is
tested against further examples in the output of other speakers. If this out-
put is perceived as inconsistent with the initial abduction, then it must be
revised with further abductions. Deduction occurs when the learner at-
tempts to produce utterances based on the abduction, testing her own out-
put (based on the abduction) on other speakers; if listeners misunderstand,
correct or reject the speaker-learner, then the initial abduction may (or may
not) be revised. This can be illustrated with a hypothetical case based on
Trudgill's example of the Northern English English speaker who attempts
to accommodate to Southern English English usage. Knowing that northern
/but/ is equivalent to southern the speaker-learner abduces "incor-
P O A X I ,
rectly" (through an overgeneralization of a pre-existing pattern) that north-

ern /but/a/ should be substituted with a previously non-existent /bAt/a/. By
hearing others use this word with the correct pronunciation, the learner
may be able to correct herself, but this will depend on both the frequency
and salience with which she hears the word. In addition, she may begin to
use the pronunciation /bAtJa/, and she may be overtly corrected or simply
receive less than positive feedback and therefore begin to determine that
there is some problem with her production.

However, Deutscher (2002) has recently pointed to certain flaws in An-
dersen's conceptualization of learner analysis. First, he points out that An-
dersen conflated earlier and later proposals in Peirce's discussion of the
modes of inference, and therefore the meanings of the term abduction shift
in this discussion. This, in itself, might not be a problem, but Deutscher
also shows that abduction as discussed by Andersen only explains how
learners match surface forms/features to a "wrong" pre-existing rule; it
does not recognize that learners invent new rules and attach them to exist-
ing forms (which they often do). He provides the following example:
The surprising fact is observed: an instrument is called 'fork'. But if it were
true that there was a rule: 'An N-pronged instrument is called N-k', the form
'fork' would be a matter of course. Conclusion: Hence, I have reason to sus-

pect that there is such a rule (so a three-pronged instrument is a 'threek').
(Deutscher 2002: 483).
Crucially, the production of the form threek is dependent on the invention
of a new rule (based in turn on pre-existing general knowledge, general
cognitive tendencies, surface and schematic analogies, and specific con-
texts). Deutscher goes on to point out that there is no useful distinction
between abduction and induction (both are induction). Finally, he suggests
that linguists have better terms to describe the inferencing processes of
induction and deduction: reanalysis (for induction, or association of an old
or new rule/explanation with a surface form) and extension or overgener-
alization (for deduction, which through analogy extends an existing rule -
whether old or new - to the production of new surface forms).
Nevertheless, one aspect of Andersen's model which remains useful is
its emphasis on the testing process that allows confirmation or correction
of initial analyses (abductions). This occurs through what Andersen re-
ferred to as induction and deduction (of course, Andersen's abductions are
simply incorrect inductions, or inductions based only on partial evidence or
input), in which the speaker-learner continues to receive input and try out
innovative forms/features when interacting with other members of the
speech community. This testing process, which allows restructuring of
learners' grammars and replication of grammars of previous generations,
cannot be successfully completed without the stable norms that are en-
forced through close-knit social networks (or the institutions of standardi-
zation). When these norms are lacking, as in the prekoine linguistic pool,
learners will learn enough to match majority (or salient) forms most of the
time, but such learning may well be "imperfect" and leave room for multi-
ple reanalyses and overgeneralizations (which, if produced by enough

speakers, can lead to change). Indeed, this was observed by Henry, who
studied how children handle syntactic input from parents and the surround-
ing community: "What is clear from this [situation] is that language learn-
ing does not involve selecting a grammar which fits all the data. Rather, it
must involve selecting . . . the grammar which best fits the majority of the
data" (Henry 1995: 79). Significantly, Henry's observation is supported by
more general research in cognitive and functional linguistics, which has
begun to emphasize the key role that frequency of use has in the develop-
ment - or emergence - of the grammars of all speaker-learners. Such stud-
ies have shown the impact of frequency effects on phenomena such as
phonetic reduction, functional reanalysis and grammaticalization, hyper-
corrections, and preservation of conservative properties (Bybee and Hop-

per 2001). More fundamentally, such studies show that frequency is key to
the construction and design (or emergence) of all speaker's grammars.
Learning, then, is the source of innovations such as form-function re-
analyses and overgeneralizations, and when these are sufficiently frequent,
and are propagated through the community, they become systemic-level
simplifications and "interdialect" features. They are especially important in
koineization, for the high degree of variation and low level of norm en-
forcement impedes the learning of features that are inconsistently produced
and/or perceived as infrequent.
The preceding models of learning are particularly useful because they
highlight similarities between child and adult language acquisition (though
young children generally are not conscious of the process of grammar con-
struction), as well as the importance of the social context to successful
learning. However, we must also ask whether there are any differences
between child and adult learning that could impact on language change
during koineization. It is widely recognized that children are ultimately far
more successful than adults in language learning, and it is often assumed
that only children can learn a language "perfectly" (cf. Lenneberg's Criti-
cal Period Hypothesis). It is now frequently claimed that children lose the
ability to learn "perfectly" from between the ages of 8-14 (Chambers
1992; Kerswill 1996).42 Kerswill (1996) includes a review of the literature
on this topic, and concludes that adults and children can learn some fea-
tures with equal ease (e.g., new vocabulary). However, other features can
be acquired only by children. The features vary according to age of ac-
quirer, the linguistic level concerned, and relative complexity. Kerswill
summarizes his findings in a hierarchy of learning difficulty (Table 6).
The ranking shows that both adults and children can learn certain kinds
of features equally well, assuming they have access to sufficient input and
interaction. However, even with such input and interaction, only children
can acquire the complex features at the top of the hierarchy. What is the
significance of this ranking to koineization? When simplification occurs in
koineization, we can be sure that it is partly attributable to adults, and
when complex features or patterns survive, we can be sure it is because
children have received sufficient input. However, this does not mean that
children can learn "perfectly" in the koineizing context. Children also sim-
plify, but they do so not because of any lack of inherent abilities, but rather
because the social context of learning often does not provide the sort of
consistent input that they need in order to construct (and restructure) inter-
Table 6. Hierarchy of learning difficulty for children and adults. Source: Kerswill
(1996:200). Rank of 1 represents most difficult features to learn.
Rank Feature type Age acquired
1 i lexically unpredictable phonological rules, by3(?)
which may reflect lexical diffusion nearing
completion and which are not socio-
linguistically salient
ii new phonological oppositions by 3-13
iii grammatical oppositions by 8 (?)
2 iv prosodie systems by 12-15
3 V grammatical change: new morphological peaks in adoles-
classes (in creóles, may be tied to lexical cent years
acquisition)
4 vi morphologically conditioned changes not before 4-7,
then lifespan
5 vii reassignment of words or lexical sets to other lifespan
morphological classes
6 viii mergers lifespan
7 ix Neogrammarian changes (exceptionless lifespan
shifts, easier if they are connected speech
processes)
8 χ lexical diffusion of phonological changes, lifespan
especially those which involve an existing
opposition and are salient
xi borrowing: new lexical forms of old words; lifespan
new phonetic forms of existing morphologi-
cal categories
9 xii borrowing: vocabulary lifespan
nal grammars that (eventually) allow the reproduction of complex patterns

or features (of course, older children and adolescents may simplify because
of a combination of decreasing ability/motivation and lack of appropriate
input).
Child language acquisition thus plays a unique role in the propagation
of certain complex features. But child learning and use of language must
also play a primordial role in the establishment of all new linguistic norms.
This is because it is only the features that are consistently learned by suc-
ceeding generations of children that will survive to characterize a new
koine. Adults, through their accommodation and acquisition, merely alter
Observing koineizatìon in Milton Keynes 63
the output that serves as the starting point for child learning. And it must
be assumed that accommodation takes place between children, too, who
leam from each other as much as from adults. This process, which occurs
over the course of one or two generations of children, is what allows the
eventual focusing of new norms, though it is inextricably linked to the
solidification of social networks. In fact, these questions are at the core of
the Milton Keynes research project organized by Paul Kerswill and Ann
Williams, to which we now turn.
4. Observing koineization: The Milton Keynes project
Most studies of koineization (including this one) are post hoc studies of
change in previously stabilized dialects. As a result, it has been difficult to
do more than theorize about how adults and children behave and learn in a
koineizing environment, and how they fashion new norms. In order to un-
derstand better the micro-level processes and conditions of koineization,
Kerswill and Williams initiated a quantitative sociolinguistic research pro-
ject to study the developing dialect of the British New Town of Milton
Keynes, located in southeastern England (just to the northwest of London
in Buckinghamshire). They have reported on this research in numerous
publications throughout the 1990s (e.g., Kerswill and Williams 1992, 1999,
2000; Kerswill 1994a, 1996), one of the most recent of which (2000) in-
cludes their own overview of koineization, expressed in terms of eight
"Principles of Koineization", and exemplified with the results of their re-
search in Milton Keynes.
To discuss Kerswill and Williams' Principles and the examples support-
ing them, it is necessary to know something of Milton Keynes and the

sample used in the project. Milton Keynes was officially designated a Brit-
ish New Town in 1967, and its population grew rapidly, reaching 145,000
by 1990, when the project was initiated (Kerswill and Williams 2000: 78).
Migration accounts for 80% of the city's growth, while a relatively high
birth rate accounts for the other 20%. The proportion of residents over 60
is very low in comparison with other English towns, and the proportion of
children relatively high (as might be expected in most immigrant communi-
ties). The incoming population included persons from all parts of the UK,
but 76.2% were from the southeast of England, with some 35.2% from
London alone. The great majority of inhabitants work in manual occupa-
tions, as do all the families in the sample used.
Kerswill and Williams chose to investigate the development of nine

phonetic variables among Milton Keynes children (some of which are dis-
cussed below). To do so, they recorded (first in 1991, then again in late
1992) a socially homogeneous group of 48 children (eight boys and eight
girls from each of three age groups - 4, 8, and 12), who had been born in
Milton Keynes or who had moved there within the first two years of life.
They also recorded the principal caregiver of each child (in most cases the
mother). An analysis of the migration patterns of the sample families re-
vealed that they were representative of the Milton Keynes population as a
whole.
From their research in Milton Keynes and the work of other scholars,
Kerswill and Williams derived their eight Principles of Koineization.
These are grouped according to the aspect of koineization they describe.
The first group has to do with the outcomes in post-contact varieties:
1. Majority forms found in the mix, rather than minority forms, win out.
2. Marked regional forms are disfavored.
3. Phonologically and lexically simple features are more often adopted than
complex ones.
This group of principles recapitulates (some of) the conclusions of Trudgill

about macro-level linguistic processes and outcomes, so I will not discuss
them further here, though it is important to note that in Milton Keynes
there are no obvious cases of simplification, so the data are most useful for
studying leveling. The other two groups of principles, however, refer to
domains in which the Milton Keynes data have proved useful for elaborat-
ing our understanding of koineization: how acquisition and social networks
interact, and, more importantly, focusing and the time scale of koineiza-
tion. These are discussed below in separate sections.
4.1. Interaction of acquisition and social networks
The second group of principles is intended to describe micro-level proc-

esses and conditions, and more specifically the factors that affect the lin-
guistic behavior of the migrants and the first generation of native-born
children:
4. Adults, adolescents, and children influence the outcome of dialect contact

differently.
Observing koineization in Milton Keynes 65
5. The adoption of features by a speaker depends on his or her network char-

acteristics.
Principle 4 is important for two reasons. First, children and adolescents

play central roles in defining new norms; this aspect relates to the third
group of principles, so it is discussed below. Second, children must play a
role in the learning of complex features which differs from that of adults.
Above I argued that both children and adults favor and are responsible for
simplifications during koineization, but Kerswill and Williams also draw
attention to the fact that (minority) simplified features in the prekoine lin-
guistic pool are sometimes rejected in favor of (majority) complex features.
Adults are far less capable of acquiring certain more complex features (cf.
Kerswill 1996: 200), so we must assume that even in the koineizing envi-
ronment some children can and do learn features that their parents cannot
or do not use. The example that Kerswill and Williams provide from their
data is the acquisition (and propagation) of the morphologically-
conditioned Roland/rolling vowel split, a pattern found throughout south-
eastern England and therefore dominant among the adult immigrants to
Milton Keynes. In these persons' speech, monomorphemic words such as
Roland, pola, cola, sola are syllabified with the syllable division between
the first vowel /au/ and the following lateral (e.g., Ro-land). These are
realized with a front allophone [aeu]. On the other hand, dimorphemic
words such as roll+ing are articulated with a back allophone [OU], even
though they share the same syllable division as Roland (i.e., ro-lling). Cor-
responding monomorphemic and monosyllabic forms, such as roll, are
realized with the back allophone [DU]. It was found that 35 of 41 children
recorded had acquired the distinction, even though many of their parents
did not use it. The exceptions tended to be the very youngest children who
remained oriented towards their parents and had not yet acquired the dis-
tinction from their peers. Does this mean that a complex majority variant
will be acquired and become a norm before a simple minority variant? Not
necessarily, for Kerswill and Williams also point out that this alternation is
favored by the principle of morpheme invariance, itself a relation of trans-
parency (and salience) that highlights the semantic relation between related
items such as roll/rolling. The model of koineization may allow no sure
predictions about outcomes when a simple minority variant and a complex
majority variant enter into competition; secondary factors relating to sali-
ence will need to be adduced to explain the preference for one or the other.
In this case, contact with surrounding dialects, in which the split is fre-
quent, seems to have been the deciding factor.43
Principle 5 relates to findings about the impact of individuals' social
network characteristics on the adoption or acquisition of features. Kerswill
and Williams use quantitative findings on the articulation of (ou) variable
(phoneme = /auf) to discuss this principle (see Table 7).
Table 7. Distribution of variants of (ou) across sample (%). Data for children
obtained in elicitation tasks, data for adults in interviews. Source:
Kerswill and Williams (2000: 93).
[ei], [aei] [BY], [asY] [au], [ay], [eu],
l'evi» Γο:1, [oui
4-year-olds 13.5 30.2 55.7
8-year-olds 12.9 53.6 33.3
12-year-olds 3.0 68.6 28.2
Caregivers 3.5 37.3 60.0
It can be seen that the younger children have both the widest range of vari-
ants and that the 4-year-olds are the most likely to use one of the variants
(in the righthand column) that is dominant among the adult caregivers.
However, the 8- and 12-year-olds are already focusing on the fronted vari-
ants in the middle column, and the 12-year-olds have abandoned both the
extreme fronted and unrounded variants of the 4-year-olds and, more
slowly, the conservative variants preferred by the adults. An essential point
to be gleaned from these data is that it is impossible to lump children into
one group. Children of different ages have different social orientations as
they grow older, and these differing social orientations have a significant
effect on their use and acquisition of language. Young children, repre-
sented by the 4-year-olds, tend to be oriented towards their parents, par-
ticularly the primary caregiver (Kerswill 1996), and they tend more than
others to reproduce the features of their parents' speech (though their
speech may also be affected by older siblings and other children). Older
children, represented by the 8- and 12-year olds, are already well integrated
into mainly school-centered peer groups, and as they grow older they tend
to abandon features of their parents' speech and adopt those of their peers,
to the degree that they can.44 This is especially true of the most sociable
and peer-oriented children, who have the most social contacts.
4.2. Focusing and the time scale of koineization
The last group of principles is not unrelated to the second, but all concern
the time scale of koineization and the process of focusing. Trudgill (1986)
ascribes the rise of a new system of norms to a process of linguistic focus-
ing, but does not address the social roots of this phenomenon. It is here,
however, where the Milton Keynes Project has proved most informative.
The principles are:
6. There is no normal historical continuity with the locality, either socially or

linguistically. Most first and second generation speakers are oriented to-
ward language varieties that originate elsewhere.
7. From initial diffusion, focusing takes place over one or two generations.
8. Because of sociolinguistic maturation, the structure of the new speech
community is first discernible in the speech of native-bom adolescents, not
young children. (Kerswill and Williams 2000: 84-85)
Principle 6 deals with the question of normal vs. abnormal transmission,

and is discussed below with pidginization/creolization and dialect leveling.
Principle 7 confirms what has been supposed about focusing, and Principle
8 elaborates on how this process occurs. The central claim is that focusing
of new norms can occur within one generation of settlement, even more
rapidly than the H0yanger data suggested. As the preceding table for the
(ou) variable shows, the oldest children (pre-adolescents) are already fo-
cusing on a new norm that differs substantially from that of their parent's
generation. In this case the new norm is not the majority articulation
among the adult migrants, but rather a fronted articulation that is a feature
of regional dialect leveling. This tendency is probably communicated

through weak ties between adolescents in southeast England, and Kerswill
and Williams suggest that it may also be favored as a sign of adolescent
autonomy. In this scenario, the first generation of children negotiates the
new norm among themselves, since they generally have no obvious target,
and their norms become most clearly defined when they reach adolescence
and their orientation toward their peer group is strongest. Their speech then
becomes a target for younger members in the speech community, such as
the 8- and 12- year-olds in the Milton Keynes study, who further focus the
developing norms. Young children are primarily oriented toward their par-
ents, but they may also be influenced by older children (as well as show
developmental features), and their speech tends to show the lowest degree
of focusing.
It is clear then that older children and adolescents play key roles in de-
termining new norms in the koineizing community. It is assumed in
Kerswill and Williams' approach that even though both children and adults
simplify, only children will be able to accommodate sufficiently well and
in sufficiently large numbers to focus the most difficult items. Still, as we
have seen, complex features are likely to survive only if they are majority
features, and even then only when other factors intervene to favor their
selection and production by speakers. However, Kerswill and Williams
leave open the possibility that adults might be responsible for focusing of
some types of features. These are likely to be those that Kerswill (1996:
200) identified as the easiest features for all speakers to learn (and the most
salient), such as the introduction of new lexical items and new forms for
existing morphological categories. Unfortunately, these are precisely the
items that have suffered the greatest degree of dialect leveling in the Mil-
ton Keynes contributing varieties, so there is no evidence of simplification
in Milton Keynes, at least in comparison to southeastern varieties (in Chap-
ter 5, I provide textual evidence of a possible adult-sponsored norm: the
rejection of leísmo [a morphosyntactic feature] in early Andalusia, which
appears to have happened very quickly).
The few (salient) norms negotiated by adults will be the first to appear
in a new community, but the set of norms negotiated by the first generation
of older children and adolescents may be the first to define a new relatively
stabilized variety. Kerswill and Williams claim that focusing is nearly
complete within this first generation of children in Milton Keynes, and
they originally claimed that this might be typical of koines (but Kerswill
[2002] adopts Trudgill's model). However, Trudgill (1998) and Trudgill,
Gordon, Lewis, and Maclagan (2000) have argued (based partly on their
study of New Zealand English) that koineization typically occurs over the
course of (at least) three generations in which: 1) the first generation of
adult migrants shows rudimentary leveling (loss of some salient minority
features); 2) the second generation (first of children) continues to show
extreme variability but also further leveling (loss of still more minority
variants); and 3) the third and subsequent generations realize focusing,
leveling, and reallocation.45 Still, this leaves two problems: 1) why do some
features resist focusing even when most other features in the variety have
focused (resistance to focusing of isolated features), and 2) why does fo-
cusing of all features occur more slowly in some cases of koineization,
such as in H0yanger, than in others, such as in Milton Keynes (general
resistance to focusing)? I discuss each of these problems below.
Kerswill and Williams identify a few isolated phonetic features that

have resisted focusing in Milton Keynes. For instance, the vowel of the
thought class still shows considerable variation between a more common
monophthong [o:] and a diphthong [ou] (Kerswill and Williams 2000:
100). The diphthongal articulation is a well-established London variant,
and over one third of adults in Milton Keynes are from London. Children
with London parents tend to show higher percentages of this articulation:
52.4% for children with both parents from London, 29.1% for those with
one parent from London and one from southeast England, 20% for those
with one parent from London and one from outside southeast England, and
only 10% for those with parents who are not from London or southeast
England. The fact that a very large minority of adult speakers use this vari-
ant has certainly favored its continued use among the children, but it is also
likely that weak-tie contacts with London and surrounding regions have
contributed to the survival of each variant. In cases such as this there is no
way to predict how the competition will be settled, though Kerswill and
Williams argue that the third generation will probably decide the matter. In
that case, it may be that one variant will win out, but it is also possible that
the variants will both survive but suffer socio-stylistic or phonological
reallocation.46 It has also been suggested that a feature may resist focusing
over a long period; this possibility is discussed below.
There are several reported cases of general resistance to focusing
(though this depends in part on what the normal time scale is considered to
be). Kerswill and Williams (2000) point to the well-known example of
Heyanger, where it was not until the third generation (the grandchildren of
the first migrants, or second generation of children) that focusing could be
discerned (Omdal 1977: 7, reported in Kerswill and Williams 2000: 73).
Kerswill and Williams argue that two factors were largely responsible for
the slower focusing in Heyanger. First, they suggest that the relatively
great differences between contributing varieties may have slowed accom-
modation, presumably because a) such differences might themselves help
reinforce differences in identity between children, and b) the greater differ-
ence between varieties may have encouraged children to favor a communi-
cation strategy of neutrality such as code-switching.47 Much more impor-
tant, however, was the early social separation of the town's population.
Some neighborhoods and schools were settled predominantly by the east-
ern-dialect-speaking managerial class, while others were settled by a west-
ern-dialect-speaking working class.48 This social separation did not break
down until the third generation, when children from both communities
began to form new mixed peer groups and with them a new koine.
Kerswill and Williams also discuss the case of the Israeli Hebrew koine
(or quasi-koine). According to both Blanc (1968) and Ravid (1995), Israeli
Hebrew has been slow to focus, even though the majority of Hebrew chil-
dren now have native-speaking parents (Yaegor-Dror [1993], however,
believes the variety is focused enough to be clearly defined for most speak-
ers). Perhaps the major factor slowing focusing in this case is the competi-
tion that exists between the spoken Hebrew koine and the officially-
sanctioned standard language, which retains a relatively complex phonol-
ogy and morphology (Ravid 1995; Kerswill and Williams 2000: 71).
Kerswill and Williams also suggest that this slow focusing is partly due to
the nature of the first generation adult input, which was in every case an
adult interlanguage. Supposedly such interlanguage input would have im-
peded the children's learning, but this seems unlikely since children must
always contend with great variation during koineization, and the differ-
ences between these varieties tended not to impede communication. On the
other hand, the tendency among Mid-Eastern Jews to maintain a distinct
ethnic identity certainly retarded their accommodation toward the Euro-
pean-dominated koine. The maintenance of their varieties of Hebrew, heav-
ily influenced by an Arabic substrate, has helped maintain a high degree of
variation. The constant arrival of new immigrants for many years may also
have slowed focusing.
Kerswill and Williams identify other factors that may retard or affect
the direction of focusing. The first relates to the relative proportions of
adults and children in the original mix:
Where there is a high proportion of adults, simplification and reduction will

occur more readily than otherwise, and focusing will not take place until the
third generation. Where there is an unusually high proportion of children (as
in most cases of migration), there MAY be a lack of simplification, as well as
the presence of focusing in the second generation.
(Kerswill and Williams 2000: 75)
I have already argued above that children and adults are probably both
responsible for simplification and that the key factors in determining
whether or not simplification occurs are the nature of the differences be-
tween contributing varieties and the degree to which learners are denied
frequent and consistent input for any particular feature. Without access to
such input, accurate learning of a feature is impossible. It does seem likely
that a higher proportion of children will favor faster focusing, since it in-
creases the proportion of the population actually engaged in negotiation of

new norms, and it increases the possibility of forming new social networks
among children and young people. If population density is low, or univer-
sal schooling is absent, then focusing may be retarded. But the presence of
more children will facilitate contact between them and increase overall the
number of speakers with greatest peer-group orientation. Schooling can
certainly facilitate the formation of peer groups among children and with
them the development of new norms (cf. Amery 1993; Britain 1997b), but
we need to keep in mind that in other societies different traditions may
exist that allow regular contact among young people. If the tendency for
older children and adolescents to be oriented more towards peers than par-
ents is indeed universal, then we would expect them to take advantage of
any possibilities for contact, particularly in immigrant societies where ex-
tended families (which are themselves social networks) are less likely to be
a source of support/restraint, at least initially. Of course, the exact ages at
which such peer groups are allowed to form (if at all) will certainly be
affected by particular cultural factors.
Finally, we should consider the evidence of the development of New
Zealand English, which has heavily influenced the three-generation model
proposed by Trudgill (1998) and his colleagues (e.g., Trudgill, Gordon,
Lewis, and Maclagan 2000). These researchers have managed to assemble
an unusual corpus of recordings of elderly people made in 1946-1948 by
the National Broadcasting Corporation of New Zealand. The persons inter-
viewed were residents of small towns and rural areas, and they were inter-
viewed as part of an oral history of pioneer reminiscences. Significantly,
all were children of adult immigrants to New Zealand (= members of the
second generation). Trudgill and his colleagues have closely analyzed
these recordings as part of their Origins of New Zealand English project

(ONZE). They have found that this group of speakers was characterized by
tremendous inter-speaker and intra-speaker variability (although, in gen-
eral, the most frequent features manifested in their collective speech were
later selected as norms of New Zealand English). From this they conclude
that focusing had not occurred in the second generation of New Zealand-
ers. This is clearly the case among the persons recorded, but it is worth
remembering that the recordings were of people living in small towns and
rural areas (some of them quite isolated), and do not directly reflect devel-
opments in urban areas. Presumably, focusing would have occurred first in
those urban areas, with rural areas following urban trends only later.
The above discussion on focusing forces one last question: is there an

upper limit on the time scale of koineization? Most scholars seem to con-
sider one or two child generations as the bottom and usual limit for this
time scale, but an upper limit on focusing has not been set. If no limit is
set, then one could conceivably relate changes that are occurring now with
cases of population movement that occurred centuries before. But there is a
methodological danger here, for the model could then be used to "explain"
just about any change in a variety that has at some time undergone dialect
and demographic mixing. Unfortunately, a model which explains every-
thing will in fact explain nothing. Given the cases of general resistance to
focusing we have seen above, there may be no simple answer to this ques-
tion, but it may be worth exploring one particular case of proposed centu-
ries-long focusing in order to better appreciate the methodological implica-
tions of such a strategy. I refer specifically to the case of koineization in
Fenland English studied by Britain (1997a, 1997b). The Fens, located in
eastern England, were once a wet, marshy area that was largely unpopu-
lated, but in the 17th century the area was drained and settlers moved in
from both east and west. The Fens happen to lie between two major dialect
zones (to the east and west), so settlement of the region led to the mixing
of speakers of different dialects. Britain analyzed the development of two
features: (ai) and /Λ/. The first feature is essentially the phenomenon of
Canadian Raising (see above), and Britain argues that the different variants
were brought by speakers of the neighboring dialects and then suffered
phonological reallocation as focusing occurred. There is evidence that "Ca-
nadian Raising" of this sort has existed for at least 200 years in the Fens,
and Britain argues convincingly that it can be attributed to 17th-century
koineization (though, as I will argue below, such claims are more
convincing if more than one such koineized feature is identified).

The other variable is more problematic. To the (north)west of the Fens,
all words of the butcher, cushion, luck, up class are (and were at the time
of Fenland koineization) realized with [u] (=/u/). To the east, this class is
and was split into two different sets, with the set butcher, cushion articu-
lated with [u] (=/u/), and the set luck, up articulated with [Λ] (=/Λ/). How-
ever, it appears that koineization did not lead to a clearly focused new
norm in the Fens. Britain shows that at least since the 19th century no sta-
ble realization of /Λ/ has existed in the central Fens. Rather, ItJ is articu-
lated with a range of phonetically intermediate articulations between [Λ]
and [u] (at least), including intermediate or fudged [γ]. Only recently have
younger speakers in the town of Wisbech begun to focus a norm by regu-
Koineization and other contact phenomena 13
larly using the interdialectal [r]. Britain (1997b: 160-165) identifies a

constellation of factors that may have colluded to impede focusing of this
feature: lack of salience for local speakers, an originally sparse population,
lack of schooling and thus contact during initial koineization, the complex-
ity of the /U~A/split, the continued contact with the neighboring dialect
regions, ongoing change in neighboring southeastern regions (with increase
in number of allophones). This approach to the data seems quite reason-
able, though we cannot be certain that an earlier norm did not exist, for this
variation may have arisen subsequent to koineization through contact be-
tween the neighboring dialects. However, there is a problem in the larger
interpretation of this phenomenon. After showing that in Wisbech younger
speakers are focusing a new norm, Britain concludes that we are observing
"koineisation-in-progress in the Fens, despite the fact that the original con-
tact began over 300 years ago" and that this demonstrates a differential rate
of focusing for different features (Britain 1997b: 165). There does seem to
be evidence for a differential rate of focusing for individual features, but it
is not prudent to assume that the focusing that is now occurring in Wisbech
is in any way explicable as a result of koineization. This would simply
overextend the explanatory usefulness of the model, since these speakers
are in no way directly affected by the original social circumstances of some
300 years ago. We should rather view this as a case in which complete
focusing was prevented (or complex allophony was maintained) by a com-
bination of factors, not least of which was the continued contact with
neighboring varieties (as discussed in Britain 2002). In fact, it is normal for
varieties that have undergone or are undergoing koineization to be affected
by other processes of language change and contact.
5. Koineization and other contact phenomena
The preceding sections have attempted to define what koineization is and

how it works. However, given the uncertain history of meanings of the
terms koine and koineization, it is also necessary to consider what koinei-
zation is not, for it has been confused or conflated with numerous other
phenomena, particularly other sorts of language and dialect contact. In the
following sections, I distinguish koineization from other kinds of language
contact and change, but also explore some of the ways in which these dif-
ferent processes might interact with it.
5.1. Pidginization and creolization
Koines, pidgins, and creóles have not always been clearly distinguished, in
part because each process was not adequately defined, but also because
similarities do exist between them with a certain degree of overlap (Siegel
1985, 1997, 2001). Koines, pidgins, and creóles are new varieties of lan-
guage used by new speech communities, and they result from the mixing of
speakers of different pre-existing language varieties, and the need of such
speakers to negotiate and acquire new structures and lexicon. The results in
each case also show the results of mixing and leveling, but there are impor-
tant differences.
All observers (e.g., Siegel 1988a; Thomason and Kaufman 1988; Muf-
wene 1997) agree that pidgins arise rapidly as "emergency" varieties when
adult speakers of typologically-distant languages need a simple but conven-
tionalized means of communication. In prototypical cases (Thomason
1997), communication is limited to a few contexts and functions (e.g., bar-
gaining), and can be supplemented with non-verbal communication (e.g.,
gesturing), but is frequent enough to require a conventionalized language
variety. The language of the dominant group, or a simplified version of it
known as foreigner talk, serves as the target for other groups. Contact be-
tween groups, often more than two, is limited, and contact may not even be
with the native speakers of the target variety. As a result, the input, interac-
tion, and motivation needed to learn the full target variety, or even a sim-
plified version, is lacking. The resulting pidgin is therefore characterized
by a radically reduced lexicon (that needed for its few functions) and a
"no-frills grammatical system" lacking in inflectional morphology and
complex syntactic structures. The radical structural and lexical reduction
makes it possible for adults to learn, stabilize, and even expand significant
components of the pidgin (e.g., New Guinea Tok Pisin before it gained
native speakers). The lack of native speakers implies, however, that the
phonology in particular is likely to remain less stable across the commu-
nity, reflecting rather the native language background of each individual.
Prototypical creóles, like pidgins, develop when "no group has the need,
the desire, and/or the opportunity to learn the other groups' languages"
(Thomason 1997: 78), but a conventionalized means of communication is
still needed for both inter- and intra-group communication. Creoles draw
the bulk of their lexicon from a dominant (though inaccessible) target or
lexifier language, which is usually typologically distant from other varie-
ties in the mix. However, unlike pidgins, creóles are learned as native lan-
guages and come to be used as the main language of a speech community.

The greater number of contexts and functions for which creóles are used
leads to the creation of an extensive lexicon and richer structural resources
than those of pidgins. As such, they cannot on structural grounds be distin-
guished from other natural languages (though they do generally lack com-
plex morphology).
Koines differ from both pidgins and creóles primarily in terms of the
social situation in which they develop. Contact (or interaction) among
speakers is frequent and unfettered and input is easily accessible. More-
over, that input, though highly variable, is generally intelligible, since the
contributing varieties share much linguistic structure and lexicon. Speak-
ers, adults, adolescents, and (older) children, usually alter their speech not
in response to an urgent need to communicate information, but rather to
accommodate and project altered identities in the new community. Most
importantly, the resultant koine shows no radical structural discontinuities
with source varieties, and is mutually intelligible with them (though this
may be asymmetrical, with the koine being more immediately and com-
pletely intelligible to speakers of conservative varieties than the reverse).
This last characteristic implies that koineization must be considered as
normal transmission within the Thomason and Kaufman (1988) frame-
work. In this view, normal transmission, the object of much traditional
historical linguistic research, contrasts radically with change through con-
tact, best exemplified in the processes of pidginization and koineization. It
is characterized by "complete and successful transmissions, by native
speakers [either parents or peer groups] to child or adult learners, of an
entire language, i.e., a complex interlocking set of phonological, morpho-
logical, syntactic, semantic, and lexical systems" (Thomason 1997: 74).
Croft sees the ideal case of such normal transmission as dependent on a

very specific social structure, in which:
There are no significant communicative interactions with individuals outside
the society, and the only significant changes in the membership of the popu-
lation are through biological reproduction and death, not through individuals
from other societies entering the population. If these social conditions hold,
then the lingueme lineages [forms and features] will be traced back in time
to that society alone, and back through that society's unique parent society,
and so on. Any new linguemes will arise only through altered replication of
the existing linguemes, and hence will belong to lineages confined to that
society of speakers. (Croft 2000: 200)
This description matches exactly the isolated societies with close-knit so-
cial networks that Trudgill (e.g., 1996, 2002) sees as most capable of suc-
cessfully transmitting a variety with a minimum of change (or of success-
fully completing a change once implemented). According to Croft, such
communities are assumed in the traditional family tree model of genetic
relationships (even though no pure examples exist), in which parent lan-
guages are followed in time by diverging daughter languages, and each
daughter language has one and only one parent. As argued by Penny (2000)
and Croft (2000), the tree model does not allow for the convergence and
hybridization between dialects that characterizes koineization. But since
koineization most certainly leads to the successful transmission of "a com-
plex interlocking set of phonological, morphological, syntactic, semantic,
and lexical systems", it must be included within the Thomason and Kauf-
man category of normal transmission. In fact, koineization is distinguished
from pidginization and creolization by allowing normal transmission con-
currently with a degree of hybridization, and either can be emphasized. For
example, it is the rapid (but limited) hybridization of koineization which is
emphasized in Kerswill and Williams' (2000) Principle 6, which denies
continuity between the koine and contributing varieties to the prekoine
linguistic pool (cf. also Kerswill 2002: 695-698). Thomason herself (1997:
85) sees koineization as an example of a borderline case between normal
transmission and contact, but, as she also argues, categories and models
such as those discussed here are abstract and to some degree arbitrary. I
will discuss this issue in relation to Mufwene's (1997) critique of the value
of koine as a technical term.
Mufwene has argued quite forcefully that there is no useful distinction
to be made between the terms koine and pidgin/creole. 49 This is based in
part on his questionable redefinition of creole as any variety which arises

through intimate contact (1997: 56; an issue which I will not address di-
rectly here). If we maintain the more regular criterion that pidgins and cre-
óles arise in social situations characterized by relatively inaccessible input
and interaction with native speakers, then koines can be safely distin-
guished from pidgins and creóles on this basis. But Mufwene also claims
that the results of contact in creolization and koineization are in no way
distinct:
In much of the literature, koineization is . . . talked about as involving "lev-
eling" of dialectal features. As is suggested in Siegel's work, those who de-
velop varieties called "koine" did not do this by simply dropping from their
varieties structural features that distinguish them from other varieties.
Rather, they select from a pool of competing features those that are less, or
the least, marked relative to the contact setting. They restructure one target
variety, as in the prototypical case of Attic Greek, subject to the influence of
the varieties in contact. (Mufwene 1997: 48)
According to Mufwene, the exact same thing can be said of creolization.
There are indeed similarities between these phenomena, but Mufwene's
claim is not completely true. As we have seen above, koineization leads to
the passing on of "a complex set of interlocking systems" (dependent, ad-
mittedly, on pre-existing correspondences between the contributing varie-
ties), but this is never true of pidginization and creolization, in which the
lexicon is derived largely from the dominant target language (the lexifier)
and the grammar is often not clearly relatable to any one language in the
input. Moreover, as Siegel (2001: 182) points out, there may be no target
variety in koineization (though there may be a dominant or majority vari-
ety, and in certain social situations salient features of particular varieties
may become a target), since the changes in speech are not generally made
in order to enhance communication, but rather to project and define new
identities.
Mufwene's third argument, however, is potentially more convincing. He
claims that the distinction between language and dialect, on which the
model rests, cannot be maintained systematically enough for it to justify
unique categories such as koine and koineization (Mufwene 1997: 44-48).
The problems with making the distinction between dialect and language are
well known, and Mufwene reviews them all: structural differences do not
suffice for making the distinction; mutual intelligibility is often difficult to
determine; political and cultural factors always play a role in determining
what is same and different. He also points to many cases of language con-
tact in Africa that are not easily classifiable as either koineization or creo-
lization on this basis, since the contributing varieties are genetically related
but are a) not mutually comprehensible for most or all speakers, and/or b)
not considered the same language by the speakers.
Siegel (2001: 180-184, following Thomason [1997]; see below), ac-
knowledges that the boundaries between dialect and language are "fuzzy",
but he does not accept that the categories are therefore meaningless or use-
less. Siegel argues that mutual intelligibility is still a useful criterion for
distinguishing dialects (or related subsystems, as he prefers), though in
some cases it will be difficult to apply the criterion. Significantly, he ac-
knowledges that mutual intelligibility also depends on factors such as atti-
tudes, beliefs and goodwill. One must ask what speakers, both adults and
children, believe about the varieties in contact and about their own chang-
ing identity, for this may determine whether they maintain their own varie-
ties, attempt to shift to another established variety, or (inadvertently) move
towards creating a new hybrid variety (through code-mixing and accom-
modation). It will also be affected by the particular situation in which
speakers find themselves, and in the koineizing environment mutual intel-
ligibility will be favored with accommodation and continuing interaction.
Nevertheless, Siegel emphasizes that there are structural constraints on
mutual intelligibility: only varieties with enough similarity have the poten-
tial to be mutually intelligible. Such constraints are lacking in pidginization
and creolization. Finally, it is worth pointing out that in many cases there is
no doubt about the status of the contributing varieties, as in Milton Keynes,
or, as I will argue, most of the contributing dialects to the medieval phases
of koineization in Spain.
It is clear, however, that in many cases we will have to struggle with the
"fuzzy" boundaries between dialects and languages, koines and creóles (or
other contact varieties). Mufwene (1997) questions the existence of such
boundaries and emphasizes the commonalities between these contact phe-
nomena. He is not wrong to do so; Milroy (e.g., 1997) too has argued co-
gently that all language change is a contact phenomenon, since a change
(in his framework) only occurs when an innovation of one person has been
adopted by at least one other, and this propagation is dependent on so-
cially-mediated contact. But this does not mean that all contact situations
are exactly alike, or that no recurrent patterns can be found that will allow
the definition (and better understanding) of different types of contact.
Thomason (1997) responds to this question by emphasizing that catego-
ries and models such as pidgin/pidginization, creole/creolization (and pre-
sumably koine/koineization) are by definition abstract and arbitrary, as are

all the other classifications that linguists make, but that they are based on
observable regularities and patterns of co-occurrence. Models of change
are derived from study of particular cases of language contact and change,
but this is not to say that they are easily applicable to all cases. This is not
a problem with the classificatory criteria per se, but results from the "fluid
nature of language history, and indeed of human language behavior as a
whole" (Thomason 1997: 71). Such categories and models of linguistic
contact are best conceived of as prototypes, and the boundaries between
them will be fuzzy, but they are useful because it makes it easier to talk
about observed phenomena and to compare them. Real cases of change can
be measured against the models. Some will match the categories and mod-
els more closely and thus be considered prototypical; others will be less
easy to classify, but the different models will still serve to help pinpoint
their unique characteristics. A final question, of course, is this: how many
models of language contact do we need? Thomason, like Mufwene, argues
for parsimony. I would agree, but, for cultural and political reasons, mixing
between speakers of mutually intelligible dialects has been a frequent phe-
nomenon in world history. I would thus argue that a model of koineization
is a particularly useful prototype.
5.2. Contact between stable dialects
The results of dialect mixing as in koineization and those of contact be-

tween stable dialect communities have not always been clearly distin-
guished in earlier work. For example, Thomason and Kaufman (1988) cite
and criticize Mühlhäusler's assertion that "mixing of linguistic subsystems
tends to lead to leveling or a kind of common-core grammar" (Mühlhäusler
1980: 28). They comment:
Now, dialect leveling is one of the oldest traditional notions in historical lin-
guistics. But in its traditional sense 'leveling' merely refers to change to-
ward greater similarity of dialects and not, as Mühlhäusler's use of the term
suggests, to change toward a less marked overall system. All of the evidence
we have indicates that the traditional notion is correct and Mühlhäusler's
more constrained prediction is not. Dialect interference, like cross-language
interference, very often leads to replacement of a structure in one dialect by
a partially or entirely corresponding structure from another. This replace-
ment may of course constitute an overall change [more marked] > [less
marked], but it is just as likely to result in a change [less marked] > [more
marked], or one that is equivalent in markedness. Most changes that are not
simple replacements are partial reinterpretations, with the same variation in
their effects on the markedness of the system. Many of the clearest (recent,
well documented) examples involve pressure from the standard on the non-
standard dialect. (Thomason and Kaufinan 1988: 30)50
The confusion here rests on the use of the term "dialect leveling". Thoma-
son and Kaufman seem to have in mind contact between stable dialects (or
dialect leveling as discussed in the next section), while Mühlhäusler ap-
pears to refer to mixing of dialects as in koineization. The results of con-
tact between established dialects with stable and close-knit social networks
are likely to differ significantly from those of koineization, since such
communities are more likely to promote the successful learning of rela-

tively complex features. Though they may resist the introduction of innova-
tions, when they do accept them, such communities will be far more suc-
cessful in propagating them (Trudgill 1992, 1996; Andersen 1988). On the
other hand, it would not be correct to claim that koines can have no marked
or additive features. As I will argue in Chapters 3 and 4, these are usually
due to incorrect inductions (reanalyses) based on conflicting input in the
linguistic pool (i.e., they represent attempts of speaker-learners to make
sense of very complex input, and are in effect simplifications from the
speaker-learners' perspective).
5.3. Dialect leveling, homogenization, uniformity
Dialect leveling in its broadest sense occurs when two dialects become
more alike by sharing features. However, it is now frequently used to refer
to a more specific process which occurs when communication networks
over entire regions become dominated by weak ties. Each looseknit com-
munity of speakers is therefore incapable of resisting the combined impact
of numerous weak ties with other communities, which are themselves char-
acterized by weak ties. Such leveling is frequent in modern urban and sub-
urban middle-class society, and is due to both social mobility and cultural
changes which underscore the importance of personal autonomy. Kerswill
and Williams (2000) show that dialect leveling has a pervasive influence
on both new koineizing and old towns in southeast England. It is difficult
to distinguish dialect leveling from koineization because they share fea-
tures and because koineization over large areas is generally accompanied
by dialect leveling. Dialect leveling can easily be considered a weak form

of koineization, but it does show some distinctions. First, it does not lead to
the formation of a clearly-distinguished new dialect, but rather seems to
favor a general spread of features and the loss of peripheral dialects. Sec-
ond, it does not necessarily favor simplification. Third, dialect leveling
leads to gradual rather than abrupt changes of use of features between gen-
erations (Kerswill and Williams 2000).51
When koineization occurs simultaneously (or nearly so) in many places
over a wide geographical region, it must be supposed that dialect leveling
between each koineizing center (towns and cities, with surrounding areas)
also occurs. As evidence, Trudgill (1986: 145) pointed to the case of Aus-
tralian English, which shows clear social and stylistic variation, but which
is also noted for its general lack of geographical diversity. The differences
that do exist are limited to certain lexical items and fine-grained phonetic
differences. Trudgill attributed such geographical leveling to similar com-
binations and proportions of speakers of different varieties in the input of
different regions (with a predominance of speakers from southeast Eng-
land), which tended to lead to similar results. He highlights the fact that a
small number of seaports received the bulk of immigration (which would
become urban epicenters of leveling), and that these ports stayed in close
contact with each other and the hinterland. Trudgill (1986) also considers
the roughly similar case of Canadian English, and concludes that "uniform-
ity appears to be quite typical of the initial stages of mixed, colonial varie-
ties... with degree of uniformity being in inverse proportion to historical
depth" (Trudgill 1986: 145). Still, the evidence from the Origins of New
Zealand English project would seem to indicate that in the earliest stages a
good deal of interspeaker and intraspeaker variation is maintained in rural
areas, but that the similar combinations of settlers and contact between
communities then lead to similar results (Trudgill, Gordon, Lewis, and
Maclagan 2000).
Dillard (1985: 51-72) attempts to argue exactly along these lines in his
discussion of the origins of American (U.S.) English. He claims that, from
the founding of the first colonies to around the middle of the 18th century,
American English developed as a koine which was remarkably uniform
from New England to Georgia; it is only in the 19th century that signs of
diversity appear. Unfortunately, he bases this argument not on textual evi-
dence but on comments about colonial American English made by contem-
porary British travelers. The comments are quite positive and praise the
uniformity and "purity" of American English up to the end of the 18th cen-
tury, but turn quite negative afterwards. As Montgomery (1996) argues,

this probably tells us more about the culture and politics of the observers
than it does about the language observed, though he acknowledges that
differences in the colonies probably struck British observers as minimal,
given the high degree of dialect variation in the British Isles. In fact, most
scholars seem to agree that regional dialects formed fairly quickly, and that
this differentiation resulted from the differing origins of the immigrants,
their immigration patterns, and patterns of contact with Britain and be-
tween different colonies. Wolfram and Schilling-Estes (1998: 92-113),
who summarize earlier research, explain that the heavy immigration of the
Scotch-Irish to the Midland regions and Appalachia (fanning out from
Philadelphia) contributed greatly to the survival of non-prevocalic /-r/
(pronunciation of /-r/ in car, cart) in inland varieties of American English.

Other (older) areas such as eastern New England and Tidewater Virginia
were populated primarily by persons from southeastern England (with
some admixture of persons from southwestern England), and the dialects of
these areas became characterized by the typically southeast-of-England
loss of non-prevocalic /-r/, which was probably present in the original im-
migrant mix and reinforced by continued contact with London (as well as
factors discussed above for other varieties). Other dialect regions were also
clearly established during the colonial period, and it is supposed that initial
lack of contact contributed to the distinction between the dialects of east-
ern and western New England. Hence, it seems we should speak not of
koineization in American English but rather of multiple regional koineiza-
tions, some of which happened concurrently (e.g., Virginia and New Eng-
land), and others which occurred later (e.g., the Midwest, followed by the
American West, with a progressive decline in dialect variation from east to
west).52
There are then factors which favor koineization with regional dialect
leveling, sometimes over great expanses, as in Australia and the American
West, and factors which favor the creation of local differentiated koines.
Widespread homogeneity will be favored by:
1. A similar mix of immigrants in different locales: the same simple and ma-
jority solutions will tend to win out in each case.
2. Concurrent population of the zones in question: the contributing varieties
will also change over time, so a later mix with similar proportions of im-
migrants may lead to different results.
3. Extensive contact throughout the region and looseknit social networks in
all areas: numerous weak ties between speakers occupying urban centers
and between speakers occupying other urban and rural areas will help
spread changes, and the changes will not be easily resisted even if they
come from outside the community, since norm enforcement mechanisms
will initially be weak everywhere. In Milton Keynes, for example, children
are not developing a variety based simply on the majority forms among
their parents, but are rather developing a variety which is following the
dominant tendencies of widespread dialect leveling in southeast England
(favored not by koineization per se but by social mobility and autonomy).
In cases of regional koineization, we must assume that speakers of each
urban center affect speakers of other urban centers through weak ties, with
the largest centers of population being the most influential (cf. Callary
1975; Wolfram and Schilling-Estes 2003: 725). Especially in pre-modem
societies, we must also assume that contact and weak ties will be facilitated
by smaller size of the region undergoing koineization.
4. Development of transregional identity: to the extent that new settlers in dif-
ferent cities, towns and rural areas believe that they share a common iden-
tity, they will be more likely to resist innovations from outside the immedi-
ate community (cf. LePage and Tabouret-Keller 1985). Common culture
and political unity will favor the development of a common identity, as
will the perception of opposing identities outside the group.
Resistance to regional dialect leveling also occurs. This has been reported
for some areas of Canada:
Inland rural dialects often differ sharply from standard speech, especially in
the numerous rural communities in which the founders were Irish and Scots,
and inland working-class dialects differ not only from standard speech but
also from one another, with the ethnic origins of the founders cutting across
social class. (Chambers and Hardwick 1985, cited in Trudgill 1986: 146)
The factors favoring local koineization with resistance to regional dialect
leveling tend to be corollaries of those listed above:
1. A different mix of immigrants: this will tend to lead to different results.

2. Homogeneity: a relative lack of mixing will favor the successful transmis-
sion of a particular variety, and will reinforce ethnic identity.
3. Strong ethnic identity: this will favor resistance to accepting a broader re-
gional identity and speech.
4. Maintenance of closeknit social networks: if a whole community moves at
once, then its members may be able to maintain stronger ties and resist al-
tering identity and speech.
5. Political divisions within the region: this will tend to create distinct identi-
ties within the region
6. Different date of arrival: if a particular locale is settled either a generation
or more before or after other areas in the region, then the settlers in that lo-
cale may resist adopting changes current in other areas.
7. Isolation: if the population of the community is relatively isolated from
other koineizing communities, then a unique local koine is most likely to
arise (this underlies Trudgill's identification of new towns, frontiers, and
colonies as prototypical sites of koineization).
8. Continued contact with source dialects: if some locales in the koineizing
region remain more oriented towards the source varieties, then their con-
tacts will differ and they will develop differently.
These lists highlight the primary linguistic, social, and cultural factors that
contribute to resistance to koineization with dialect leveling, but it is im-
possible to list all the factors, and the researcher will need to consider
purely local factors as well.
While it is useful to have a conceptual distinction between dialect level-
ing and koineization, and to see that they interact, there remain cases where
no clear distinction may be possible. I refer to so-called "regional koines".
Are these the result of dialect leveling or prototypical koineization? No
general answer to this question is possible, since "regional koine" has re-
ceived no technical definition, and regional koines, whatever they are,
probably result from the interaction of both processes. The problem is
made more difficult by the fact that immigration to regional urban centers
is often (but not necessarily) slow, and in-migrants tend to come from sur-
rounding territories and thus speak similar dialects. Again, one possibility
for making a distinction is to use Kerswill and Williams' Principle 6,
which suggests that in koineization changes are rapid and abrupt, with
speakers of the oldest and youngest generation showing little overlap (for
changing features), while dialect leveling leads to more gradual change
across generations.
5.4. Standardization and standards
There exists long-standing confusion between the terms koine and standard
and a tendency to conflate or integrate the processes of koineization and
standardization. This is not surprising, as there are, indeed, similarities
between the two processes. Haugen (1966), in his classic essay, explains
that standardization occurs for functional and ideological reasons. Stan-

dards arise as supra-local language varieties in nation-states in order to
assure internal cohesion and external distinction. Socially, the process of
standardization begins with the selection of a dialect to serve as the base of
the standard. Following this, measures are taken to ensure this dialect's
acceptance by the population. Linguistically, standardization is character-
ized by codification, leading to minimal variation in form, and elaboration,
which is tied to maximal variation in function. Moreover, codification gen-
erally begins and is most effective in written versions of the language, so
literacy is a necessary concomitant of standardization. Unsurprisingly,
phonetic variation tends to be most resistant to standardization, even when
phonographic writing systems are employed.
An analysis of each of these aspects reveals a major source of confusion

between koineization and standardization. First, the process of selection is
a socially and politically thorny issue in any community. The group whose
variety is selected is automatically favored. If a widespread, stable koine
has already come into existence (and it is used by the most powerful
group), and no previous standard exists (or is rejected because of its struc-
tural distance and difficulty, as in some cases of diglossia), the choice of
the koine for the standard is a fairly easy one to make. Moreover, since the
bulk of the population will already use the koine, or will find learning it
relatively less difficult, choice of a koine will aid acceptance by the popu-
lation. In addition, a koine will already include a mixture of features from
other dialects and the elimination of much variation, which will aid the
task of codification. Finally, the koine is a structurally complete language,
so it will already be used for a wide range of functions and can be easily
extended through lexical and syntactic expansion to meet new ones. Obvi-
ously, if a koine has arisen spontaneously before standardization has be-
gun, the process of standardization (whether evolutionary or planned) will
be greatly facilitated. Nevertheless, the processes are conceptually distinct.
The distinction between standardization and koineization is highlighted
by Fontanella de Weinberg (1992: 42-54), who analyzes cases in which a
pre-existing standard or written prestige norm is present in the mix. Fon-
tanella argues that standardization can impede or erase the results of
koineization, and bases this claim on the unique developments in three
varieties of American Spanish: Mexican Spanish of the interior (relatively
conservative, standard), Paraguayan Spanish (highly non-standard), and
Argentine Spanish (an intermediate variety where later standardization
erased some of the effects of an earlier koineization). In the case of Mex-
ico, the pre-existent prestige norm of central northern Spain was a domi-
nant component in the prekoine linguistic pool of the colonial capital,
home to large numbers of prestige-oriented representatives of the vice-
regal government, church, and university. As a result, many (though not
all) features of other American koines (such as weakening and loss of syl-
lable-final /-s/, and confusion of liquids) were rejected in Mexico. On the
other hand, the early Buenos Aires koine was characterized by a strong
tendency to neutralize syllable-final /-r/ and /-l/ (found also in Andalusian
and Caribbean Spanish), but increasing literacy and influence of the stan-
dard (whose spelling was based on the pronunciation of northern central
Spain) during the 19th century led to the suppression of this simplified
feature in Argentine Spanish.53
Fontanella's claims about the interaction of koineization and standardi-

zation in Mexico find some support in Scholtmeijer's research on the de-
velopment of Dutch in the new polders of the reclaimed Zuider Zee,
founded between 1930 and the 1950s (Scholtmeijer 1992, 1997, reported in
Kerswill and Williams 2000: 74). In these areas, children speak very dif-
ferently from their immigrant parents and grandparents, but no new variety
or koine has developed in these areas (though Kerswill and Williams note
that Scholtmeijer may be overstating the case). Rather, children speak a
highly standardized form of Dutch, and Scholtmeijer ascribes this to the
combined influence of schooling in the standard and a need for a language
variety for external communication (1992: 145). On the other hand,
Kerswill and Williams note that the inhabitants of the polders have re-
mained in continued contact with surrounding regions, so dialect leveling
has probably played a role in the development of speech in the polders.
Significantly, move toward the standard is part of the more general process
of dialect leveling in the Netherlands.54
5.5. Language shift and language death
Language shift occurs when an entire speech community acquires a new

language variety and abandons its former community language. Substrate
theory has focused on the effect of shift on the target language into which
features from the source are accepted. According to Thomason and Kauf-
man (1988: 119-119), substratum interference tends to lead to alteration of
phonological and abstract syntactic patterns, but not to the borrowing of
morpholexical items. Presumably, dialect shift is not essentially different
from language shift, and koineization can certainly be seen as a particular

instance of complex language shift. However, koineization involves the
mixing of vocabulary from different sources, which is not predicted, and,
unlike prototypical cases of shift, there may be no one target. As we have
seen, this tends to favor processes of leveling and simplification, and ulti-
mately the creation of a new variety. On the other hand, since majority
features tend to be selected from the linguistic pool, it is essential to con-
sider the effects of shift (acquisition) by large groups of speakers in the
demographic mix, particularly when this involves language shift in the
strict sense, i.e., the learning of the "language" by speakers of unrelated
languages.
When the speakers of an unrelated language in the mix form a signifi-

cant but minority presence (e.g., Basque learners of Romance in medieval
Castile), it is appropriate to include their production in the prekoine pool,
for they (or, rather, their children) will have adequate access to input and
features of their speech will not necessarily be more marked than those of
speakers of established dialects. Nevertheless, clear influence of a minority
of "non-native" speakers will tend to be washed out. The unique social
context of koineization might seem to provide ideal circumstances for sub-
strate influence by weakening potential social resistance to substrate-based
innovations, but it also weakens the possibility of selection of substrate
features unless a) they are also found in the speech of others, and b) the
substrate features are themselves simplifications that are favored by other
learners in the community, particularly (older) children. I explore this issue
with reference to Basque influence on the development of Castilian in
Chapters 3 and 4.
Language death studies have looked at shift from the opposite perspec-
tive: the effects of shift on the minority language during its abandonment
by a speech community (Dorian 1981). In the best known version of this
process, the language suffers a progressive reduction in its functional do-
main as parents cease to use certain forms and children do not learn them;
extensive simplification and even impoverishment occurs. Silva-Corvalán
(1994) argues that contact between English and Spanish in Los Angeles
(where the language is maintained at the community level thanks to new
immigration, but dies within families) has led to an exaggeration of certain
"internal" leveling and simplifying tendencies of Spanish, rather than
wholesale transfer of English forms and structures to Spanish. The effects
of language death can therefore be difficult to distinguish from those of
koineization. Moreover, In Los Angeles, immigrants have come from many

parts of Mexico and other Latin American countries, and it is probable that
koineization has occurred at the same time as language death. Gambhir
(1981) indicates that Guyanese Bhojpuri, as it is gradually replaced by
Guyanese Creole, is also undergoing language death, making it impossible
to clearly distinguish (in the contemporary language) the original effects of
koineization from the subsequent effects of language death.
5.6. Convergence
Koineization is often considered to be an example of linguistic conver-

gence, but only in the broad sense of the term. Convergence as a technical
term is now often used to refer to the effects of stable bilingualism or mul-
tilingualism on the structures of clearly distinct languages. The classic
example is the case of Kupwar, India (Gumperz and Wilson 1971). In this
multilingual community, the languages of Marathi, Urdu and Kannada
(unrelated to the first two) each retain a distinct lexicon and morphological
markers, but their use by multilingual speakers over generations has led to
complete parallelism in the abstract syntactic structure, so that in general
sentences can be translated word for word. The motivation for this conver-
gence is strikingly different from that of koineization, since accommoda-
tion appears to have played little part. Rather, a conscious desire to retain
separate identities and languages was reinforced by close-knit social net-
works, while patterns of intergroup relations forced multilingualism on the
members of the community, and in their interlanguages they (probably
unconsciously) transferred patterns across the different languages despite
their efforts to keep them apart.
5.7. Borrowing
Thomason and Kaufman (1988: 74-75) define borrowing on a five-level

scale from 1, "casual contact" with only lexical borrowing, to 5, "very
strong cultural pressure" characterized by heavy structural borrowing. The
case of Kupwar, for example, would be placed on this scale at 4, "strong
cultural pressure and moderate structural borrowing" (due to what is seen

as the relatively low typological distance between the languages involved).
Normally, however, the term borrowing has a more limited meaning close
to "1" on the Thomason and Kaufman scale. In a language that has under-
gone koineization, the results of mixing could easily be confused with ex-
tensive borrowing, although both will be present. Any study of koineiza-
tion will need to distinguish, to the degree possible, between the results of
mixing and the results of borrowing. For many items or features this may
be impossible, but for others knowledge of the time of their appearance and
the contemporary social circumstances may allow a determination of their
origin as borrowings or as items that survived the process of mixing in
koineization.
The model and its use 89
5.8. Interacting processes and the study of koineization
From the above discussion we can conclude that koineization must be un-
derstood as a prototype of change, and that we should not expect to find
"pure" examples of koineization in real cases of linguistic change. We will
find koineization being affected by a range of varying linguistic, social,
and cultural factors (aside from more stable cognitive factors), and it will
interact with other processes of contact-induced change. A central chal-
lenge will therefore be to distinguish effects of koineization from those of
other processes, or, at least, to show in a given case why this cannot be
done (i.e., how two or more processes might all favor a similar outcome).
This is a problem in any investigation of causes, but, as we have seen, there
are places and situations in which the effects of koineization are likely to
be particularly clear-cut. Still, the researcher is only too likely to find in-
teraction with other factors. Siegel highlights the influence of a pidginized
variety of Hindi on the development of Fiji Hindi, Blanc the effects of
multiple language shifts on Israeli Hebrew, Fontanella the impact of stan-
dardization on the development of American varieties of Spanish, Gambhir
the effects of language death on Guyanese Bhojpuri, and Kerswill and
Williams the influence of regional dialect leveling on Milton Keynes
speech. The frequency with which we find such combinations of influence
means that they probably represent the messy norm, but the model is de-
signed not to exactly simulate reality, but to help us tease out patterned
threads of cause and effect that might otherwise remain hidden.
6. The model and its use

In the preceding I have sought to review previous conceptualizations of

koineization, and to elaborate those views with clarifications and sugges-
tions of my own. The intention, of course, has been to arrive at a thorough
definition of koineization and an understanding of how koineization works.
This is particularly important for the purposes of this study, for my goal is
to apply the model at relatively great time-depths of between 1100 and 700
years, to reconstruct the sociolinguistic contexts of these periods (to the
degree necessary and possible), and to reinterpret previous scholarly dis-
cussion of the changes and periods under study here. Still, it seems that
there may exist potential abuses or hazards in applying the model to the
historical record, so I have also sought to define basic methodological
guidelines for its use. Below I present the definition and a discussion of
methodological guidelines.
6.1. Definition of koineization
Koineization is best conceived as a two-tiered model or process. On the

micro-level speakers of different linguistic subsystems interact in a new
community. Such mixing is especially likely during large population
movements to new towns, frontiers, and colonies, and is accompanied by a
sudden breakdown in social networks and, consequently, norm enforce-
ment. The language varieties are mutually intelligible (or quickly become
so through exposure), and include regional varieties, sociolects, adult inter-
languages, and child language. In this mixed social environment, variation
in the linguistic pool peaks at the very time that social networks are weak-
ened and norm enforcement declines. As adult speakers attempt to interact
in the new community and establish social ties, they accommodate to the
speech of their interlocutors, often eliminating non-functional or minority
variants from their speech; they may also seek to reproduce new features
that are perceived as frequent and/or salient in the prekoine, thereby neu-
tralizing the original social value of many variants, and possibly introduc-
ing interlanguage (interdialect) variants (developing frequency of variants
can be favored by high salience, both perceptual and, especially, sociolin-
guistic). In this contact- and input-rich environment, children generally
learn what they perceive to be the most frequent and consistently produced
features, including those found in their own learner language (such as sim-
plified, transparent, regular forms and/or those based on incorrect induc-
tions or analyses). As older children shift their social-psychological orien-

tation from parents to peers, they accommodate their speech to that of their
peers to the degree that they can, introducing more interdialectal forms as
they learn and generally favoring majority (and/or salient) features, what-
ever their origin may be. It is among the older children and adolescents of
the first generation that regularities of usage begin to appear, but it may
take another generation of children for all norms to focus.
Over time, therefore, language acquisition, accommodation, and re-
solidification of social networks (beginning with adolescent peer groups)
lead to the establishment of new linguistic norms, or systemic changes at
the macro- or linguistic level. These include mixing/leveling and simplifi-
cation. Mixing refers to the survival in the resultant koine of variants from
different contributing varieties. Leveling refers to the survival of one vari-

ant from a pool of competing variants. Reallocation occurs when more than
one competing variant survives but each with a different function (social,
stylistic, geographical, phonological) or meaning. Simplification is under-
stood as limited reduction in inventories of units and rules, generally seen
as greater regularity or transparency in the grammar. These macro-level
phenomena are explained as the unintended results of micro-level speaker
activity, such as accommodation, constrained by the process of acquisition
and the particular social context. It is the unique social context that allows
for the formation, in only one or two generations, of a relatively mixed,
simplified (though socio-stylistically variable), and focused koine.
6.2. Methodological guidelines
A well-defined model is of great use for reconstruction of the social condi-

tions of past changes. However, there remain potential problems. First, any
and all of the particular types of changes that characterize koineization can
occur in language varieties which clearly have not been affected by it. For
example, the effects of mixing and borrowing could easily be confused,
and examples of simplifying changes can certainly be found even in con-
servative varieties used by isolated closeknit communities. The only way to
be (relatively more) certain that a particular change is causally related to
koineization is by situating it within a larger analysis of both macro-level
linguistic phenomena and micro-level social conditions and constraints on
speaker activity.
The researcher must therefore begin with an accurate description of the
linguistic changes themselves, specifying (minimally) that:
1. the changes reflect mixing/leveling and simplification;

2. several (preferably numerous) such changes co-occur (and changes which
add opacity and complexity are few, and explicable within the constraints
of theframeworkdescribed above);
3. the features are selected and focus rapidly, normally over the course of just
2 or 3 generations (=1 or 2 generations of children).
The primary prediction of the model is that all changes will tend toward
mixing and simplification. Therefore it is necessary to show that numerous
changes correspond with this prediction. Analysis of an isolated case of
change in terms of koineization is not sufficient to show that the change in
question is causally related to koineization. Of course, such focused analy-

sis may be necessary on a provisional basis (it is impossible to do every-
thing at once), but the goal should be to identify broad patterns of change.
Only this will lend validity to the argument that any particular change is
the outcome of koineization. Moreover, if numerous counterexamples to
predicted patterns are found, then it will be difficult to analyze even sim-
plifying changes in terms of koineization (though relative chronologies and
interaction with other processes of change may be considered in such
cases). Finally, if a change is not complete within the prototypical time
scale of koineization, then it will be far more difficult to establish a causal
relation between social changes and linguistic changes. If a change does
not occur (or a feature focus) until hundreds of years after the original
demographic mixing, it will be difficult if not impossible to establish a
convincing causal link between the events. If such extensions in the time
scale are allowed, then any recent changes in any modern variety of
American Spanish and English might be linked to koineization, but this, of
course, is of no real explanatory value (but see Chapter 6 for discussion of
chain reactions or drift).
Clearly, however, the core of an explanation based on koineization is
the specification of micro-level conditions of speaker activity and the link-
ing of (collective) speaker activity to the linguistic results. This must in-
clude an understanding of the constraints imposed on speakers by the ac-
quisition process (which remains constant), but the explanation of
linguistic change will depend on how speaker-learners are affected by the
changing social context. Thus, the variable aspects of the micro-level that
must be demonstrated include the following:
1. an increase in variation due to demographic and dialect mixing;

2. a breakdown in social networks and consequently in norm enforcement;
3. specification of the contributing varieties and the relative proportions of
speakers of each;
4. a specification of the lexicon and structures of the contributing varieties,
including possible learner varieties.
But as the above review has shown, numerous other factors may have to be
taken into account, such as the isolation of the koineizing community from
the source communities, contacts with new communities, the potential
(cultural) impact of a "foreign" threat that might contribute to community
solidarity, and, in general, any factor which affects the development of
community identity. Kerswill and Williams also assume that an exact un-
derstanding of the social and linguistic history of the koineizing commu-

nity is necessary; for them, an ideal social history will include the follow-
ing information:
1. The original population of the area: its size, its social characteristics, and
its speech forms.
2. The size of the incoming population in relation to the original population.
3. The abruptness of the settlement: Was it sudden and fmite, or did it con-
tinue over a long period? Did it continue after koineization had taken
place?
4. The proportion of children to adults among the incomers and the original
population, and the rate at which children were bom to the incomers after
migration.
5. The continued contacts of the incomers with their place of origin: Did they
break off relations with their original home completely, or did they main-
tain links with it to the exclusion of new, local contacts?
6. The social characteristics and ethnicity of the incomers: Did they come to
take up specific jobs, e.g., in a new industry? Were they socially mixed?
Were they an ethnically distinct group?
7. The speech of the incomers: Was it diverse or homogeneous? Was it simi-
lar to that of the native population? Were some social dialects better repre-
sented than others?
(Kerswill and Williams 2000: 70)
Kerswill and Williams acknowledge that finding precise answers to all

these questions is probably not possible, even in contemporary cases of
koineization. However, even when they cannot be answered, it is important
to keep such questions in mind when exploring the impact of the social
context on language use and change during koineization. Indeed, the same
can be said for several of the methodological guidelines listed above. For
example, in following chapters I analyze three periods of koineization that
affected medieval Castilian (or varieties of Castilian). For each case or
stage, I analyze and reconstruct several groups of changes that (I believe)
were linked to koineization, as well as some which were not (or only indi-
rectly). Such groups of changes are investigated in accordance with the
methodological restriction that particular koineizing changes must occur as
part of a broad pattern of koineizing change.

Koineization in Medieval Spanish - (2. Koines and Koineization)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Koineization in Medieval Spanish - (2. Koines and Koineization)

Uploaded by

Copyright:

Available Formats

Chapter 2

Koines and koineization

The model of koineization, of fairly recent development, is based on earlier

1. Koine and koines

Koine is a term with a long history and a wide variety of interpretations. It

1.1. The Greek Koine

Confederacy of Delos into an Athenian empire, the influence of Attic

This list of characteristics is attractively simple and clear - deceptively so

of dialect mixing and demographic movement. More recently, scholars

1.2. Modern use of the term koine

Modern metaphorical or technical use of the term koine has grown as

"least common denominator" of contributing varieties (see below). He also

— was based primarily on one dialect

— was reduced and simplified14

In order to determine the dominant interpretations of the term, Siegel ana-

1. Literary Italian (Pei 1966: 139)

20. Koineized colloquial Arabic (Samarin 1971: 134)

34. English-based nautical jargon (Hancock 1971: 290n)

— "a form of language resulting from a compromise between various dialects

unplanned, nativized, or transported languages may be koines if they exhibit

According to Siegel, most koines are characterized to some extent by

the linguist who observes the frequently similar results of multiple

in greater detail below.

— (a) its development as a new, common variety based on existing dialects of

model of koineization. However, the growth of studies of pidginization and

2.1. Siegel ' s stage-based model

As these different varieties received more scholarly attention, efforts to

1. Prekoine. "This is the unstabilized stage at the beginning of koineization.

phological complexity compared to the contributing subsystems."

on then-current approaches to the study of pidginization and creolization,

Table 1. Developmental contunua of koines. Source: Siegel (1985: 375).

prekoine prekoine prekoine

Table 2. Developmental contunua of creóles. Source: Siegel (1985: 375), based on

jargon jargon jargon

4 stabilized pidgin stabilized pidgin

Pidgins are generally understood to result from contact between typologi-

derstand each other, the need to communicate information - which plays

communication, where gestures, intonation, and the possibility of clarifica-

context of koineization. It is certainly true that any expansion of contexts

zation that leads to full stabilization of a pidgin (although the question of

In later publications, Siegel has discussed the similarities and differ-

It should be stressed that the developmental continuum of a koine is not

2.2. Trudgill's process-based model

The single most influential discussion of koineization is that of the socio-

2.2.1. Accommodation and salience

behavior (Trudgill 1986: 2).26

Though accommodation is generally studied as a phenomenon affecting

accommodation beyond the speech community will often be a rather differ-

At this point speakers appear to enter into a process of language or dialect

Nevertheless, Trudgill hesitates to associate long-term accommodation

possible tendency on the part of speakers to approach dialect and language

Chambers seems to want to restrict the use of long-term accommodation to

1. Phonological contrast. For example, English English speakers who relocate

However, Trudgill also indicates a number of factors which appear to im-

pede accommodation and either reduce or override salience:

4. Extra-strong stereotyping. For example, accommodating Northern English

These few examples make it clear that long-term accommodation cannot be

cleus or intervocalic position is also salient (because more acoustically

Although, as Kerswill and Williams (2002) indicate, it is impossible to

rather than in CV transitions and are even less salient in VN transitions"

phonology, morphology, and syntax are generally considered discrete com-

Table 3. Form-function reanalysis of Reading do. Source: Trudgill (1986: 65).

In the process of shifting towards Standard English usage, some younger