23 Contact as a Source of Language Change

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 1 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

Subject

Key-Topics

DOI:

23. Contact as a Source of Language Change

SARAH GREY THOMASON

Linguistics

Historical Linguistics

language

10.1111/b.9781405127479.2004.00025.x

In a sense, most of what historical linguists study under the designation “language change” is due to
contact. An individual speaker's innovation typically becomes part of the database of historical linguistics
only after other speakers have adopted it - both because the likelihood that any historical linguist will
become aware of one person's innovation is minute and because the innovation may well be ephemeral even
for the single innovator. The changes we investigate therefore tend to be those that have spread throughout
a speech-(sub)community, and the process of spread is a function of contact between speakers.
Nevertheless, the spread of innovations within a speech community has traditionally been considered
separately from the diffusion of features across dialect and especially language boundaries.

One reason for this separation is that quite different methodologies for studying these two kinds of contact
have developed (compare, for instance, this chapter and the following one, by Wolfram and Schilling-Estes).
Another reason, which is of course related to the first, is a commonly perceived difference in the nature of
the processes: dialect borrowing (and indeed diffusion of features from any one speaker of a given dialect to
any other speaker of the same dialect) has generally been considered, at least implicitly, to be primarily a
social process, mostly or entirely unconstrained by linguistic factors. The transfer of features from one
language to another, by contrast, has been the subject of numerous proposed linguistic constraints, and
social factors have often been treated as secondary. The intuition underlying this distinction is that two
dialects of the same language, and certainly any two speakers of the same dialect, share most of their
lexicon and grammatical structures, so that a neighbor's innovation will be easy to learn (and adopt) and
unlikely to disrupt the original linguistic system seriously; but separate languages, since they differ in more
fundamental respects than dialects of the same language, would risk undergoing disruptive change if their
speakers adopted features promiscuously from other languages, and in addition such features would be
harder to learn (and thus harder to adopt). There is also a common assumption that speakers of dialects of
the same language are more likely to talk to each other than to speakers of different languages; so social
contact is often assumed as a given in dialectology, whereas it must be established by argument and
evidence to support a claim of change induced by contact with another language.

These intuitions and assumptions are not wholly mistaken, but the differences turn out to be a matter of
degree, not of kind. Dialects of the same language may have particular structure points that are more
different than analogous structures in related or even unrelated languages; in many speech-communities,
contact with other languages is more frequent than contact with geographically distant dialects of the same
language; and so forth. This means (among other things) that both linguistic and social factors must be
considered in any full account of contact-induced change, regardless of whether the contact is between
dialects or separate languages. More generally, both social and linguistic factors must in principle be
considered in any full account of any linguistic change, although in practice we have little or no social
information about the vast majority of changes we know about. For this reason, and also because the
following chapter covers dialectological aspects of the general topic, I will focus here primarily on what has

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 2 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

following chapter covers dialectological aspects of the general topic, I will focus here primarily on what has
traditionally been studied as contact-induced language change - namely, the linguistic results of contact
between two (or more) languages.

Before turning to particular aspects of the general topic, I should clarify my use of terms. In my view, contact
between languages (or dialects) is a source of linguistic change whenever a change occurs that would have
been unlikely, or at least less likely, to occur outside a specific contact situation. This definition is broad
enough to include both the transfer of linguistic features from one language to another and innovations
which, though not direct interference features, nevertheless have their origin in a particular contact situation.
The most obvious examples in the second category are those changes in a dying language that do not make
the dying language more similar to the dominant language (see section 3 for discussion). I usually use the
terms “(linguistic) interference” and “contact-induced change” interchangeably; but this usage requires a
caveat, because non-convergent simplifying innovations in a dying language are certainly contact-induced,
though they are not interference features. Less obvious but still important examples are innovations that
appear at a late stage of a chain-reaction process in which an initial instance of structural transfer sets off a
series of other changes. In such cases the end result may well be more radical structural change than the
first step, but the ultimate source of the drastic result nevertheless lies in that initial transfer. The late-stage
innovations are therefore still contact-induced changes, but they are not interference features per se.

The notion of feature transfer should also be construed broadly: it includes innovations based on
reinterpretation of source-language features by the speakers who implement the changes as well as the
introduction of features actually present in the source language. All these are interference features in the
receiving language.

A final observation, though not a definition, is also needed to set the stage for the discussions that follow.
On occasion I will refer to the sociolinguistic notion “intensity of contact” as a requirement for certain
degrees and kinds of interference. This is a vague notion, but it is difficult to pin down more precisely in a
way that applies to a wide range of contact situations; among the factors that contribute to greater intensity
of contact are a high level of bilingualism, socioeconomic and/or political pressure on one speaker group in
a two-language contact situation to shift to the other language, length of contact, and relative sizes of
speaker populations. The point I wish to make in connection with intensity of contact is this: great intensity
of contact is a necessary condition for certain kinds of interference, especially structural interference, but it
is by no means a sufficient condition. It is easy to find contact situations in which, despite (for instance)
great pressure on and universal bilingualism among speakers of one language, very little contact-induced
change of any kind has occurred. One such example is Montana Salish (also called Flathead), a Salishan
language spoken in northwestern Montana. Of the several thousand tribal members, fewer than 70 fluent
speakers of the language remain, and all of them have native fluency in English as well as in Montana Salish.
Nevertheless, the English intrusion into Montana Salish is minimal: a few loanwords - some of them dating
back to the nineteenth century, when few if any tribal members spoke English - and no detectable
grammatical influence of any kind. Nor are there any visible signs of language attrition; in particular, all the
elaborate morphological structure of the language is intact. The general conclusion is obvious: as with
internally motivated change, predicting when contact-induced change will occur is at best risky.

In section 1 I discuss various types of linguistic interference: a classification of the changes in terms of their
effects on the receiving system; a fundamental dichotomy between changes in which imperfect learning plays
a role and changes in which it doesn't; and linguistic factors that affect the likelihood that a feature will be
borrowed. Section 2 surveys mechanisms of interference, section 3 concerns the relationship between
linguistic interference and changes that occur in language death, and section 4 compares contact-language
genesis with contact-induced language change. I conclude in section 5 with a brief discussion of
retrospective issues - in particular, how can one “prove” that a particular linguistic change is due to
language contact? - and a summing up of the entire chapter. Throughout this chapter the discussion and
examples will focus on two-language contact situations rather than on multilanguage situations such as
those characteristic of Sprachbund contexts. The reason for this focus is that basic processes and results, as
well as their correlations with social factors, are much easier to isolate in less complex situations (see
section 2 for further discussion of this point).

1 Types of Linguistic Interference

There are of course many possible ways of classifying the results of linguistic interference. In this section I

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 3 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

will discuss the three classifications that are most generally relevant to understanding the nature of contact-
induced change -differing results in terms of effects on the receiving system, differing types of interference
resulting from different social conditions, and differing types of interference resulting from the influence of
various linguistic factors.

1.1 Systematic effects on the recipient language

First, changes may be categorized according to their general effects on the receiving language's structure:
old features may be lost from the system, new features may be added to the system, or old features may be
replaced by new ones. Not all changes fit neatly into one category or another, since some involve partial loss
with partial replacement and others involve partial addition with partial replacement; but these three
categories cover the basic possibilities.

Here are typical examples of the three types. Romansh has lost gender agreement in predicate adjectives
under German influence (Weinreich 1953: 39), and both Kupwar Marathi and Kupwar Urdu have lost gender
agreement in noun modifiers under Kannada influence (Gumperz and Wilson 1971; see Thomason and
Kaufman 1988: 86–7 for discussion). The simplest examples of added features are lexical borrowings where
both form and content are new to the borrowing language, such as English bok choy, but structural features
are also often added via borrowing. For instance, Kupwar Urdu has acquired an inclusive/exclusive “we”
distinction from Marathi (Gumperz and Wilson 1971), and vowel harmony has been introduced into Greek
suffixes in some Asia Minor Greek dialects (Dawkins 1916: 47, 68). An example of replacement is the
appearance, in an Asia Minor Greek dialect of Cappadocia, of the Turkish inflectional suffixes -ik ‘lpl.’ and -
iniz ‘2pl.,’ replacing the corresponding Greek suffixes on Greek verbs (Dawkins 1916: 144).

It is worth noting that these three results are also basic categories in internally motivated linguistic change;
here, as in certain other important respects, the main difference between the two lies in their sources, not in
anything special about the change processes themselves. So, for instance, we find feature loss in cases
where there appears to be no external motivation, as in the loss of the distinction between the dative and
locative cases in some Serbo-Croatian dialects. Internally motivated feature addition includes, among other
things, discourse markers such as like in She's like, “What are you doing here?”

In both internally and externally motivated change, feature replacements are always complex processes,
involving competition between the original form or construction and the innovative feature. For internal
changes this competition is explored in (for instance) Bloomfield's chapter on “Fluctuation in the frequency of
forms” (1933: 392–403) and embodied in Kurylowicz's fourth “law” of analogy (1945–9: 30), according to
which competing forms that arise through analogic split undergo semantic differentiation if they both remain
in the language (with the innovative form taking over the basic function, thus justifying calling it a partial
replacement). Somewhat parallel examples with internal and external sources can be adduced. Compare, for
instance, the partial internal analogic replacements in hanged versus hung (with semantic differentiation)
and dived versus dove (without semantic differentiation) with partial replacements through borrowing, such
as the borrowed/native words animal versus deer (with semantic differentiation) and the competition
between borrowed and native inflectional material in Cappadocian Greek, where Turkish suffixes are not
used on all Greek verbs all the time.

1.2 Interference with and without imperfect learning

The second especially important classification of types of linguistic interference focuses on a robust
correlation between one prominent sociolinguistic variable and divergent sets of linguistic results. The
sociolinguistic variable is the presence versus the absence of full, or at least extensive, fluency in the
recipient language. That is, the crucial factor is whether the people who introduce the interference features
speak the language into which the features are introduced - or, in other words, whether imperfect learning
plays a role in the interference process. When fluent speakers of language A incorporate features into A from
another language, B, the first and most common interference features will be non-basic lexical items,
followed (if contact is sufficiently intense) by structural features and perhaps also basic vocabulary. This
pattern - (non-basic) vocabulary first, structure later if at all - is at the foundation of most of the borrowing
scales that have been proposed in the literature (e.g., Moravcsik 1978: 110 and Comrie 1981: 202ff; see
Thomason and Kaufman 1988 for discussion). By contrast, if people who are not fluent speakers of A
introduce features into A from another language, B, the first interference features (and usually the most
common ones overall) will not be lexical, but rather phonological and syntactic. Morphological features may
also be introduced under this condition; the likelihood that lexical items from B will be incorporated into A

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 4 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

also be introduced under this condition; the likelihood that lexical items from B will be incorporated into A
depends on other social factors, such as the relative prestige of A and B speakers.

These two types of interference were characterized in Thomason and Kaufman (1988) as borrowing, in which
features are incorporated into A by native (L1) speakers of A, versus shift-induced interference, in which a
group of L2 learners of A carry over features from B (their L1) into A during a process of shift from B to A.
Independently, in the same year, van Coetsem (1988) proposed a nearly identical distinction, labeling the
two types recipient language agentivity (or borrowing) versus source language agentivity (or imposition).
These two formulations are adequate for most cases, but not for all: in particular, borrowing may be carried
out by fluent L2 speakers of A; “shift-induced interference” sometimes occurs when no shift takes place at
all - in cases where, as with English in many parts of the world, local varieties of a language arise and
stabilize but remain second languages; and part of an “imposition” process may involve the participation of
A speakers (see below for discussion).

Moreover, the role played by imperfect group learning of a target language (TL) is more complex than the
definition of shift-induced interference allows for. If shifting speakers do not learn the TL perfectly, their
version of the TL (TL

) will differ from the native speakers’ version (TL

) in two ways: first, the learners will

fail to learn some features of the TL, usually features that are hard to learn for reasons of universal
markedness or typological distance from the structure of their L1, or both; and second, they will carry over
features from their L1 into the TL. In the latter case the term “imperfect learning” may be misleading: in
some instances the learners may well know that the particular L1 features do not exist in the TL, but they
may nevertheless introduce them into their TL

in order to maintain an L1 distinction that is lacking in the

. The learners’ TL

may stabilize as a variety exclusive to the shifting group and their descendants; this

happens if there is sufficient social and/or geographic isolation from the main TL

community to permit,

encourage, or necessitate maintenance of the TL

without linguistic integration. But sometimes TL

speakers

become part of the TL

speech-community, with linguistic integration. In such a case TL

speakers may

borrow features from TL

, thus producing an integrated variety, TL

; this process is of course borrowing in

the narrow sense of Thomason and Kaufman (1988), but the interference features are nevertheless those
characteristic of shift-induced interference - both because the innovations (from the perspective of TL

) will

be a subset of the innovations of TL

and because TL

and TL

already share a common lexicon.

For all these reasons, the formulation of the distinction in Thomason and Kaufman (1988) needs two
revisions: the crucial sociolinguistic factor is not whether or not shift takes place, but whether or not there is
imperfect learning by a group of people;

and one half of the linguistic prediction must be hedged - in

borrowing, interference always begins with non-basic vocabulary unless languages A and B have mostly or
entirely identical lexicons. Unfortunately, the first revision leaves us with no convenient and fully accurate
term for what has been called shift-induced interference; to avoid proliferating terms I will continue to use
it, in the hope that readers will not find its literal inaccuracy too jarring.

Finally, it must be emphasized that the picture presented in this section takes the simplest case as basic.
Many cases are considerably more complicated. One complication is that the two types of interference often
co-occur in the same contact situation: shift-induced interference may be implemented (in the first instance)
by shifting speakers even while original TL speakers are borrowing features directly from the shifting group's
language (not from TL

). The other obvious complication is that more than two languages may be involved,

with varying mixes of borrowing and shift-induced interference going on at more or less the same time.
These are the cases usually labeled as Sprachbund situations, where a number of languages in a particular
region share a set of features that distinguish them from their respective sister languages in other regions.
Probably the most famous Sprachbund areas are the Balkans (see, e.g., Sandfeld 1930; Schaller 1975; Joseph
1983a) and India (see, e.g., Emeneau 1956), but these are by no means the only examples; among the
others that have been discussed in the literature are Arnhem Land in Australia (Heath 1978), the Pacific
Northwest of North America (e.g., Jacobs 1954; Thompson and Kinkade 1990), and Meso-America (Campbell
et al. 1986).

As a result of such complications - and in particular because even shift-induced interference often includes
lexicon, and contact situations in which group shift is taking place sometimes include simultaneous two-way
interference - the retrospective picture may be difficult or impossible to unravel. Only two safe predictions
can be made. If, for a past contact situation, it can be established that contact-induced change occurred,

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 5 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

can be made. If, for a past contact situation, it can be established that contact-induced change occurred,
and if phonological and syntactic interference predominate, then imperfect learning must have been a major
factor in the process of interference. In contrast, if mainly or only lexicon has been transferred from B to A,
then imperfect learning is unlikely to have played any significant role in the process. But if sizable amounts
of both lexical and structural interference can be demonstrated, it is likely to be impossible to tell, from the
linguistic evidence alone, whether or not imperfect learning played a role.

1.3 Linguistic factors in linguistic interference

Interference features can easily be found in all linguistic subsystems -phonetics, phonology, morphology,
syntax, lexical semantics, discourse, and even narrative structure - under the appropriate social
circumstances. The appropriate social circumstances include, besides the presence or absence of imperfect
learning, the crucial but hard-to-define factor of intensity of contact: the more intense the contact, the more
kinds of linguistic features can turn up as interference features. The probabilities are not the same for all
subsystems, however; the linguistic factors of universal markedness and typological distance between source
and recipient language are important in predicting what kinds of features will be transferred from one
language to another. This is especially obvious in shift-induced interference because, as noted above, these
factors contribute to the learnability of particular features in particular contact situations: TL

features that

are harder to learn are less likely to be learned by shifting speakers, and TL

features that are harder to

learn are less likely to be learned and borrowed by original TL speakers.

The usefulness of borrowing scales attests to the relevance of linguistic factors in borrowing (in my narrow
sense) as well, but here the focus shifts from learnability per se to the degree of integratedness into a
linguistic system, as emphasized by Heath (1978) and Comrie (1981), among others. Features that are
deeply embedded in elaborate interlocking structures are in general less likely to be borrowed, because they
are less likely to fit into the recipient language's structures; that is why the lexicon, which for all its structure
is less highly organized than other grammatical subsystems, is borrowed first, and it is why inflectional
morphology tends to be borrowed last. But highly integrated features may be borrowed readily between
systems that are typologically very similar; that is why, in dialect borrowing, even inflectional morphology is
quite easily transferred. Other factors also enter in, though they are harder to specify; for instance, the
significant difference in borrowing probability between basic vocabulary (less often borrowed) and non-basic
vocabulary (more often borrowed) must depend on something other than degree of internal organization.

And when contact is intense enough, there appear to be no absolute linguistic barriers at all to borrowing
(see section 2.7 below for discussion).

2 Mechanisms of Interference

If we ask how contact-induced change comes about, we find that the actual processes parallel processes of
internally motivated change to a considerable extent. In both types we must consider the competition
between old and new variants, the role of markedness (or, more generally, ease of learning) in helping or
hindering the spread of an innovation, the effects of analogic leveling and extension, and the role of
speakers’ creativity in producing and stabilizing innovations. I will not emphasize these parallels here, but a
fuller treatment of processes of linguistic change would necessarily explore them.

Mechanisms of contact-induced change fall into four categories. Two of them correlate with the distinction
between borrowing and shift-induced interference: one set of mechanisms comes into play when the
implementers of a change are bilingual in both source and recipient language (sections 2.1–2.3, and in a
sense section 2.6), while the other set comprises second language acquisition strategies (section 2.5). A
third category, “negotiation,” seems to overlap with both of these types (section 2.4), and the fourth category
has to do with more or less conscious and deliberate decisions by speakers to implement language change
(section 2.7).

Before beginning the survey of mechanisms, I should make two background assumptions explicit. First, any
feature that can be code-switched from language A into language B can turn into a permanent interference
feature in B, and the same is true for all the other mechanisms. More generally, any feature that can appear
in a single person's speech at any time - for example, in speech errors caused by fatigue or drunkenness or
mere carelessness - can turn into a permanent change in the entire language; it is ultimately irrelevant
whether the source of the feature is internal or external. In other words, the question of linguistic possibility

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 6 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

is settled as soon as a feature appears for the first time anywhere. This assumption is exploited below in
various examples of individuals’ linguistic behavior, which I take to be valid illustrations of what can happen
in language change. Predicting whether a given one-time innovation will become a permanent part of a
person's speech or of an entire language is of course a different question entirely, a matter of probabilities
in the complex interplay of linguistic and social factors. If correct, this assumption means that any theory of
how interference comes about must be compatible with all the evidence from the study of completed
contact-induced changes, and vice versa. Especially in view of the wild-card mechanism - speakers’
deliberate decision to change - my second background assumption is in effect a corollary of this first one.

The second assumption is that there are no absolute linguistic constraints on the kinds and degrees of
linguistic interference that can occur. None of the constraints that have been proposed in the literature is
valid; it is possible, and usually quite easy, to find counterexamples to all of them (see Thomason and
Kaufman 1988: ch. 2 and section 2.7 below for discussion). But if there are no valid constraints, then either
each mechanism of interference is in principle unconstrained or different mechanisms have different
constraints, such that no particular type of change is ruled out by all of them. It is therefore relevant to note
that although various constraints have been proposed, on code-switching in particular, there are in fact no
well-established or generally accepted constraints on this or any other mechanism. As with the first
background assumption, different probabilities can be established for different kinds of changes,
probabilities based both on social factors (e.g., intensity of contact) and on linguistic factors (e.g.,
markedness). But in this domain everything appears to be possible, although some things are improbable.

2.1 Code-switching

Code-switching, as used in this section, includes both intrasentential switching (sometimes called code
mixing) and intersentential switching. I do not mean to suggest by this usage that there are no interesting
differences between the two, but at the level of this discussion they appear to be a single mechanism. This is
by far the most-studied and most-discussed mechanism of contact-induced language change; indeed, it has
sometimes been claimed to be the main or only mechanism (see, e.g., Myers-Scotton 1993: 174). It has
attracted an enormous amount of attention, and the body of empirical studies is growing rapidly, ranging
from earlier works like Hasselmo (1970), Blom and Gumperz (1972), and Pfaff (1979) to recent studies like
Eliasson (1995) and Backus (1996).

Code-switching is a (perhaps the) major route by which loanwords enter a language. It surely plays a role in
at least some kinds of structural borrowing as well, although the more dramatic kinds of structural
interference are probably likely to result from code alternation instead (see section 2.2 below). Some authors
have denied that code-switching can lead to, or become, borrowing, but the possibility seems to be rejected
on a priori grounds rather than on the basis of empirical evidence, and some of the most perceptive case
studies (e.g., Heath 1989) have provided evidence to the contrary. In particular, the notion of “nonce
borrowing” - the occurrence of foreign morphemes in a language just once or a few times, but as
borrowings, not as code-switches - seems to serve as a device for dealing with apparent counterexamples to
theoretical predictions about what can and can't be code-switched (e.g., the purported impossibility of
code-switches that combine a root from one language with affixes from the other). Eliasson (1995), among
others, has presented convincing evidence that at least some elements claimed as nonce borrowings are
actually code-switches.

In fact, I believe that is impossible in principle and in practice to draw an absolute boundary between code-
switching and borrowing. They are indeed two separate phenomena, but they are linked by a continuum: as
in so many other areas of historical linguistics, the dividing line between them is fuzzy, not sharp. A code-
switched word or other morpheme becomes a borrowing if it is used more and more frequently - with or
without phonological adaptation - until it is a regular part of the recipient language, learned as such by new
learners. It seems likely, for instance, that the English pronouns I and you appeared first in Thai as code-
switched elements (a conveniently neutral pronominal usage in place of the elaborate honorific system
encoded in native Thai pronouns) and only later as borrowings, as they are characterized by Foley (1986:
210); these borrowed pronouns are additions to the Thai pronominal system, not (at least not yet)
replacements. Introduction of these pronouns into Thai through code-switching would parallel Eliasson's
example of English and as a code-switched element in Maori - an element that provides speakers with a
short cut, since the native Maori coordinating patterns are much more complex than English coordination.

A strong motivation for treating code-switching and borrowing as points on a continuum is that this
treatment shows borrowing to be an aspect of ordinary linguistic behavior rather than an improbably exotic

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 7 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

treatment shows borrowing to be an aspect of ordinary linguistic behavior rather than an improbably exotic
phenomenon that is wholly unlike everyday usage. The addition by borrowing of a new word for a new
concept, like bok choy in English, must begin with a single use and continue with increasing usage by the
innovating speaker(s) and by other speakers, and the addition by invention of a new word, like photocopy,
must follow the same path. Similarly, an internally motivated replacement of one form or construction by
another begins with the introduction of a new feature and proceeds by competition between the new and old
features; in many replacements through borrowing, a code-switched element would be the innovation, and
its spread would proceed in much the same way as with added features. (This is as true for partial
replacements, e.g., native brothers versus brethren and borrowed animal versus deer, as for total
replacements.)

These parallels are hardly surprising, given that all speakers draw on a variety of repertoires, typically
characterized as styles, registers, and dialects, in using a single language. If there is no evidence to the
contrary - and certainly no convincing evidence has been presented - then it is surely most reasonable to
assume that speakers whose repertoires include more than one language will employ the same strategies in
deploying their linguistic resources. Change resulting from code-switching between different languages
does, of course, differ from change via diffusion from another register and dialect borrowing; but, as noted
at the start of this chapter, the differences are a matter of degree, not of kind.

A final question on this topic: is code-switching a major factor in shift-induced interference? Probably not:
lexical items predominate in code-switching, while phonological and syntactic features predominate in shift-
induced interference. Of course lexical transfer does occur in L2 acquisition, but it is not the main transfer
phenomenon; for instance, Backus (1996) found that Turkish immigrants in Holland do use Turkish words in
their Dutch, but not many, and then generally to fill lexical gaps or to give an “ethnic ring” to their Dutch
(Ad Backus, pers. comm., 1996). So, although code-switching undoubtedly occurs in contexts in which
groups are shifting to a TL, other mechanisms - especially L2 acquisition strategies - are more likely to be
responsible for most shift-induced interference.

2.2 Code alternation

It is by no means the case that all bilingual speakers who use both of their languages regularly engage in
code-switching. In many cases the two languages are used by the same speaker with different interlocutors,
often monolinguals; a typical example is the use of one language at home and another language at work.
The structural effects of this type of bilingual usage, called code alternation, have received much less
attention than code-switching, perhaps in part because they are much harder to study directly. But they can
be profound.

The main kinds of interference that come about through code alternation seem likely to differ from the main
kinds of interference that come about through code-switching. Intuitively, the two seem to have different
primary effects: anecdotal evidence about borrowing through code alternation focuses on structural rather
than lexical interference, whereas lexical borrowing is the most prominent direct result of code-switching.
There is too little evidence about code alternation to be confident about this as a systematic difference, but
here are two suggestive examples.

The first is from Michela Shigley-Giusti (1993), a native speaker of Italian (and French) who told this story
after spending 12 years as a graduate student in the United States, speaking English almost exclusively. She
spoke Italian only with occasional visitors and on occasional trips home to Italy. She reported no code-
switching; at most it would have been very rare. Her use of the two languages was therefore code
alternation: she spoke English always and only with Americans, and Italian only with Italians. To her surprise
(and distress), she was complimented during trips to Italy for “speaking Italian decently for an American”:
English had influenced her Italian in phonology (e.g., English intonation patterns, aveolar stops, and
aspirated initial voiceless stops), lexical semantics, and perhaps also syntax. The second example is from Ad
Backus (pers. comm., 1996), who reports that - at weekly social occasions that include one American friend
and two Dutch friends - his code alternation between Dutch and English leads to severe, though temporary,
interference in both languages, including problems with lexical access as well as grammatical interference.

These anecdotes, which can easily be multiplied, are significant for two reasons. First, although the type of
interference is certainly borrowing in my narrow sense, there is no hint of conscious intent. On the contrary:
Shigley-Giusti, for instance, was hardly gratified to be taken for a non-native speaker of her own language.
This point is worth emphasizing because of the lingering belief that borrowing is usually a deliberate choice
carried out for reasons of prestige. Second, and more importantly for understanding what is happening, the

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 8 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

carried out for reasons of prestige. Second, and more importantly for understanding what is happening, the
unconscious and involuntary incorporation of foreign structural features into one of a bilingual's languages
fits very well with psycholinguists’ finding that “bilinguals rarely deactivate the other language totally”
(Grosjean and Soares 1986: 146). Moreover, as Ad Backus (pers. comm., 1996) observes, “deactivation of
entrenched nonsalient elements of speech (e.g. most syntactic elements) is probably much harder to do.” I
would add only that phonetic and phonological elements are also likely to be non-salient and entrenched.
Lexicon, by contrast, is more likely to be salient, and that may account for its apparent lesser role in
interference via code alternation.

Even if it is true that structural interference happens covertly, in the course of partial activation of one
language while speaking another, it may of course still be the case that code-switching produces the same
kinds of grammatical interference as code alternation. But the interference would then, by hypothesis, be
indirect, by way of partial activation of the one language during the production of sentences or parts of
sentences in the other, instead of direct incorporation of one language's structural features into the other
language as part of code-switched sequences. That is, the division might be as follows: morphemes, both
lexical and grammatical, would be introduced directly via code-switching, changing from code-switches to
borrowings through increasingly frequent usage by code-switching speakers and then (if not all members of
the speech community engage in code-switching) by adoption by other speakers; morphemes, but especially
other features, notably phonological and syntactic structures, would be introduced by bilingual speakers
who “inadvertently” access bits of one language while speaking another - and features introduced in this
way too would become permanent additions or replacements in the recipient language if they became
entrenched in the bilinguals’ speech and adopted by other speakers as well. This is clearly not a pure
dichotomy: some foreign morphemes are surely introduced through code alternation; grammatical
morphemes introduced via code-switching might set in motion a series of changes that ultimately produces
significant structural change; and, most obviously, code-switched morphemes will often include foreign
phonological and perhaps also morphosyntactic features, which may or may not be replaced by native
structure when a code-switched morpheme becomes a borrowing.

In any case, the main purpose of this discussion is to emphasize the fact that code alternation is likely to be
as important as code-switching as a mechanism of contact-induced change - and very likely a more
important mechanism when the focus is on structural, as opposed to lexical, interference.

2.3 Passive familiarity

Sometimes interference features are introduced by speakers whose competence in the source language is
strictly passive - that is, a speaker may borrow a feature from a dialect or language that she or he does not
speak actively at all. The mechanism presumably involves partial activation of a foreign system, as in
changes via code alternation. This activation must occur because of frequent exposure to the features; but
of course a speaker whose language changes through this mechanism must at least understand the source
language, so it is likely that most such transfers occur between systems that are very similar, often dialects
of the same language.

Many of the features transferred by this means are lexical. For instance, many Americans who don't speak
African American Vernacular English (AAVE) nevertheless adopt lexical items from it; an example is the use
of cold-blooded as an intensifier and to mean “exceptionally good.”

Structural features may also be

transferred through passive familiarity. The fate of sentence-initial whom in my own speech is a typical
example. My native rule for the use of who and whom in this position - in all formal situations, both written
and spoken - was that of (at least one variant of) Standard English, with who used for subjects of clauses
and whom used for objects: Who is going? and Who did you say was going?, but Whom did you see? and
Whom did you say he saw? Informally, I used only who sentence-initially. But for years I heard people
saying sentences like Whom did you say was going? in formal contexts, and although the construction
struck me as an irritating hypercorrection, I found myself using it occasionally. At this point, in order to
avoid hypercorrection, I dropped sentence-initial whom entirely in speech. My point, however, is that I had
borrowed - certainly unwillingly - a use of whom that was foreign to my native dialect, without being able to
speak the source dialect itself.

It is likely that passive familiarity is the mechanism by which TL

features contribute to the emergence of TL

as the language of a community comprising both original TL speakers and members of a group that has
shifted to the TL: original TL speakers probably never speak the TL

itself, but passive familiarity with the

leads to the introduction of some of its features into their speech, ultimately producing the TL

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 9 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

leads to the introduction of some of its features into their speech, ultimately producing the TL

2.4 “Negotiation”

This mechanism, which resembles the concept of accommodation that has received much attention in
sociolinguistics (and sociology), overlaps with - or, perhaps better, is a component of - at least two other
mechanisms, code alternation (section 2.2) and L2 acquisition strategies (section 2.5). Shudder quotes
surround the term because deliberate, conscious negotiation is unlikely (though see section 2.7 below);
rather, speakers change their speech patterns (language A) to approximate what they believe to be the
patterns of another language or dialect (B).

If the speakers who implement the change are fully bilingual, their beliefs about B's structure will be
accurate, and the resulting change will make A more similar to B; this is what has been called convergence,
a kind of borrowing in the framework used here.

A typical example is reported by Sandalo for Kadiwéu, a

Waikurúan language of Brazil (1995: 7). Kadiwéu has six word order patterns, OVS, VOS, SOV, OSV, VSO, and
SVO; but when translating sentences from Portuguese, bilingual speakers usually use SVO order, thus
making their sentences more similar to Portuguese. (Of course this example does not mean that Kadiwéu
has changed, or even that it is changing; examination of untranslated spontaneous discourse would be
needed to find out whether permanent change is in progress.)

But if the speakers who implement the change are not fully bilingual, the resulting change may not match
the original B pattern at all. This is extremely common in shift-induced interference. A particularly clear
example is the stress pattern that emerged in a dialect of Serbo-Croatian spoken in northern former
Yugoslavia near the Hungarian border. Local Hungarian speakers, in shifting to Serbo-Croatian, apparently
perceived correctly that it lacked fixed initial stress, the Hungarian pattern; but they maintained their L1-
based expectation of fixed stress - missing the actual Serbo-Croatian accent pattern, which has free stress
(with some restrictions) - and their resulting TL

had fixed penultimate stress. Original Serbo-Croatian

speakers adopted this feature from TL

, so the entire dialect, TL

, now has fixed penultimate stress.

In the most extreme cases of non-bilingual contact situations, namely the new contact situations that give
rise to pidgins (or creoles), members of all speaker groups (usually more than two) employ “negotiation” in
making guesses about what their interlocutors will understand; the guesses that facilitate comprehension will
become part of the emerging contact language (see Thomason and Kaufman 1988: ch. 6 for discussion).

2.5 Strategies of second language acquisition

One major strategy of second language acquisition has just been discussed: as noted above, “negotiation” is
an important mechanism - possibly the most important one - in shift-induced interference, accounting for
L2 learners’ efforts to make sense of the input they receive from speakers of the TL. But another important
strategy also comes into play in some cases of shift-induced interference, perhaps even in most cases of
natural (as opposed to classroom) L2 acquisition. This is the carryover of features from the L1 into the TL as
a way of filling gaps in the learner's knowledge of the TL, making expression of certain meanings possible
when the speaker lacks adequate knowledge of the TL and/or when the TL lacks the particular features. Both
lexical and grammatical features may be carried over in this way; the grammatical features may include both
actual transferred morphemes and transferred structure without morphemes, but the latter is probably more
common. It isn't clear that this is an entirely independent mechanism; it overlaps with the final one,
deliberate decision (section 2.7), though it is probably less conscious than the examples given under that
heading.

Here's a typical example. During a period in which speakers of Uralic (specifically Finnic) languages were
shifting to Russian, Russian was undergoing extensive internally motivated changes in noun inflection. One
aspect of this inflectional change was a merging of two noun classes inherited from Proto-Indo-European,
the o-stems and the u-stems, whose original semantic basis (if any) was already completely opaque in PIE.
In most case/number combinations the o-stem endings replaced the u-stem endings - not surprisingly,
because the o-stem class was by far the larger. In the genitive singular, however, both the o-stem ending -a
and the u-stem ending -u remained side by side. The reason for the retention of both endings seems to be
that shifting Uralic speakers, whose native languages lacked (and lack) noun classes entirely, assigned
separate meanings to the two competing genitive singular suffixes, meanings that were already encoded in
separate cases in their native language(s). Their usage was then adopted by original Russian speakers
(through borrowing; see section 2.3), so that modern Russian, in the genitive singular of this noun class

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 10 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

(through borrowing; see section 2.3), so that modern Russian, in the genitive singular of this noun class
alone, now has a distinction between partitives (marked by -u) and general genitives (marked by -a). The
semantic differentiation is reminiscent of pairs like English hanged and hung, and in both instances a
semantic distinction was introduced to differentiate between competing forms resulting from analogic
innovations; but the source of the semantic difference was internal in the English example and external in
the Russian one.

2.6 First language acquisition

The role of first language acquisition in language change is a topic that has received much attention in
theoretical discussions of internally motivated change, especially since the rise of generative grammar in the
1960s. The results of generative research on L

acquisition and language change are still controversial

among historical linguists, but it remains a lively research area. By contrast, the question of L

acquisition as

a potential or actual mechanism of contact-induced language change has received (virtually?) no attention in
the literature - until 1996, when Robin Queen completed a study that provides solid evidence that language
change can result directly from the (near-?) simultaneous acquisition of two first languages.

Queen's evidence comes from her analysis of the speech of 31 children in Germany, 15 of them Turkish-
German bilinguals and 16 of them German monolinguals. She investigated a specific phrase-final intonation
pattern in which German and Turkish differ, and she found that the bilingual children employed both
patterns in speaking both of their languages, but in different pragmatic contexts. That is, they learned both
patterns, but instead of confining each one to its source language they developed a mixed system in which
each pattern had its own function - and they incorporated this system into both their Turkish and their
German. Their innovative system is apparently not stigmatized, because (for instance) German-speaking
teachers interviewed by Queen did not seem to be aware of it in the children's German.

Queen's results have several interesting implications. First, and most obviously, it is difficult to describe
these children's mixed intonational system as the result of interference per se if the children, starting from
scratch with both languages, never developed separate German and Turkish intonational systems which could
then merge (though this is hard to establish definitely; see discussion below). That is, the mechanism of
change was the creation of two L1s by children, not the incorporation into an existing system of features
from a different one (with reinterpretation of the competing features) - even though the end result, from the
viewpoint of either Turkish or German, is partial replacement of one intonational system with another, under
the influence of the other language.

Second, her results raise the question of the extent to which L1 acquisition is operative as a general
mechanism of contact-induced change - in other words, the extent to which it overlaps with other
mechanisms of contact-induced change. Clearly this is an area for future research, but a few preliminary
comments can be made now. Both code-switching and code alternation could be relevant to the
development of such a pattern during bilingual L1 acquisition: as children achieve facility with their two
emerging languages, some switching back and forth may occur, and this could contribute to the fixing of
the two patterns with their pragmatic distinction. If this is what happened with the bilinguals in Queen's
study, it means that the children first learned both systems and then mixed them, so the developmental
pattern would not be primary L1 acquisition but rather something closer to (though not identical with)
processes of change in adult speech that result from these two mechanisms.

Another possibility is that the children never learned and used the two separate intonational systems, but
instead acquired both to a sufficient degree that they understood their functional identity, and then sorted
out the competing features by assigning them different functions. This is similar to the Russian change
described in section 2.5, and it is of course also potentially relevant, at least indirectly, to “negotiation.” It
differs from the Russian example in that the implementers of the Russian change, by hypothesis, already
knew a complete Uralic language; this inference is justified by the comparable behavior of individual L2
learners, but Queen's results show that it may be difficult or impossible to distinguish, in retrospect,
between changes effected by adult L2 learners and changes effected by young L1 learners.

Without much more information about the actual developmental process (for instance, information about
whether the bilingual children ever used the actual Turkish pattern in speaking Turkish and/or the actual
German pattern in speaking German), we can't decide between the two possible routes of development for
Queen's results: acquisition of two patterns followed by mixing or partial acquisition of competing patterns
with creation of a new mixed system in the first instance. The most I can say is that the outcome resembles

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 11 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

with creation of a new mixed system in the first instance. The most I can say is that the outcome resembles
the type of change exemplified in section 2.5 more closely than any other type of contact-induced change
that I know of, and that this favors the second route.

2.7 Deliberate decision

Theories of language change rarely or never allow for the possibility of deliberate change, except in such
trivial cases as the conscious adoption of loanwords and even new sounds in words of foreign origin, for
example, when an English speaker pronounces the name Bach as [bar]. A major reason for including the
present section in this chapter is to underline the fact that a sizable body of evidence now attests to the
possibility of large-scale changes that, if not all due to actual conscious decision, nevertheless reflect a
community's more or less deliberate manipulation of its linguistic resources in order to create something
new.

The most dramatic cases involve two-language mixtures created by bilinguals, apparently to serve as a
symbol of a new ethnic identity; see section 4 for examples. Less dramatic but still important cases are the
various reports, from widely scattered locations, of speaker groups deliberately withholding their “real”
language from outsiders, using instead a distorted and simplified foreigner talk version that, in some cases,
forms the basis for a trade pidgin (see Thomason and Kaufman 1988: 175–7 for examples). Other cases
involve the creation of a secret language, either by phonological distortion (as in Pig Latin, to which parallels
can be found in languages used by entire communities -see Thomason 1995 for discussion) or lexical
replacement. And in still other cases the motive for making a particular change has to do with emphasizing
in-group status, or differentness from other groups; an example is the introduction of the voiceless lateral
fricative into Bantu words by speakers of Ma'a (see below), an extension of this non-Bantu phoneme that
serves to make their speech less Bantu-like (Mous 1994: 199). More spectacular examples are the changes
made in New Guinea communities that have “purposely fostered linguistic diversity because they have seen
language as a highly salient mark of group identity” (Kulick 1990: 1–2). Kulick cites the startling case of a
reversal of all anaphoric gender agreements in one language, with the result that masculine elements
correspond systematically to feminine elements in neighboring dialects, and vice versa.

My conclusion from these and other similar examples is not that historical linguists should abandon the
search for regularities in language change and theoretical explanations for them, but rather that much more
caution is needed in predicting what can and especially what cannot happen to the language(s) of a
community over time. Speakers’ creativity is certainly not limited to trivial lexical innovations; given the right
social circumstances, it can have enormous linguistic consequences. In the domain of contact-induced
change, it is the main reason for the claim at the beginning of section 2 that anything goes.

3 Sources of Change in Language Attrition

Attrition - the overall simplification and reduction of a language's linguistic structures, without concomitant
complication elsewhere in the system - occurs only as a prelude to language death, but the reverse is not
true: language death is not always preceded by attrition. In some cases, as is well known, language shift
occurs so soon after initial contact that there is no time for attrition to occur. More rarely, speakers will
maintain their language under great pressure from a dominant group, borrowing so much lexicon and
structure that eventually only fragments of the original language survive (especially the basic vocabulary; see
Thomason 1995 for discussion of several such cases, notably Ma'a, a language of Tanzania with Bantu
grammar and primarily non-Bantu basic vocabulary). In still other cases, like that of Montana Salish
(mentioned near the beginning of this chapter), speaker attitudes or other poorly understood social factors
seem to block large-scale borrowing; and in some tragic cases the language vanishes abruptly when all its
speakers die as a result of massacre, disease, or natural disaster. Regardless of whether or not there is
interference from the dominant language, all these linguistic results, including attrition, are the direct result
of language contact - except for the last two, since a killer disease need not be of foreign origin and an
earthquake has no human agent. All attrition is thus contact-induced change by the definition given near the
start of the chapter, though only some of it involves interference.

In the literature on language attrition, two competing views of the sources of the various simplifying changes
have been debated. According to one view, characterized by Woolard (1989: 356) as the “loans to loss”
model, extensive borrowing leads eventually to language loss; the implication is that attrition is always
accompanied by, or perhaps actually comprises, the adoption of lexical and structural features from the
dominant language - that is, convergence toward the dominant language.

The most extreme version of the

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 12 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

dominant language - that is, convergence toward the dominant language.

The most extreme version of the

opposing view is that, as Cook claims (1995: 226), “convergence does not occur in a dying language,” and
“the simplification in the speech of semispeakers is internally motivated” (ibid.: 218). Most linguists who
have studied attrition processes would probably argue for some intermediate position, that borrowing does
occur in dying languages but is not all that goes on; see, for instance, most of the papers in Dorian (1989).

It seems to me that this debate rests on a false dichotomy. The underlying assumption appears to be that a
given change must have one and only one source, either borrowing from the dominant language or
simplification resulting from forgetting or never properly learning one's ethnic-group language. Few authors
have considered the possibility of mulitiple causation, and fewer still have investigated it empirically. A
systematic study of the sources of the changes must first identify as many changes as possible in a dying
language that is undergoing, or has undergone, significant attrition. The next step is to ask, for each
change, whether it is best accounted for by borrowing alone, or by simplification alone (governed by such
factors as universal markedness), or by a combination of the two. If either borrowing or simplification could
be expected to produce the observed effect, then surely the most reasonable conclusion is that both are
responsible for it: multiple causation is well known, though relatively rarely discussed, in historical linguistics
(e.g., in internal analogic changes), and in general a change is more likely to occur if independent forces are
pushing in the same direction.

One recent case study (one of only two that I know of - see Joseph 1983: ch. 7 for the other) in which the
issue of multiple causation is addressed systematically is Fenyvesi (1995), an investigation of changes in the
Hungarian spoken by first-generation immigrants and second-generation US-born members of an American
Hungarian community in McKeesport, Pennsylvania. McKeesport Hungarian is a short generation away from
extinction: almost all the speakers are over 60 years old, and the language is now used only rarely in
everyday conversation. Fenyvesi identified several dozen changes - phonetic, phonological, morphological,
syntactic, and lexical - and classified them according to the three source categories just mentioned.

For first-generation speakers, the immigrants, she found just one change that could be attributed to
simplification alone and 13 changes in each of the other two categories, borrowing-only and both borrowing
and simplification. Any attrition in these subjects’ speech would be due to forgetting, not to incomplete
acquisition; nevertheless, these results seem comparable to those of the second-generation speakers, for
whom imperfect learning is a real possibility. For this group Fenyvesi found four simplification-only changes,
20 borrowing-only changes, and 28 changes attributable to both borrowing and simplification.

Here are a few illustrations of changes in the three categories. All these examples are found in second-
generation speakers’ speech, and some also occur in the first-generation speakers’ samples. The
replacement of one pre-verb by another in pre-verb-verb constructions, though simplificatory (in tending to
reduce the number of pre-verbs), does not make Hungarian more like English; the same is true of the
regularization of irregular stems in inflection. But the change from fixed initial Hungarian stress to stress on
a non-initial syllable (namely, on the verb) in pre-verb-verb combinations is certainly not a simplification,
and is therefore attributable to English influence alone - the pre-verbs are comparable to English prefixes,
and prefixes are normally unstressed in English. The same is true of the appearance of such phonetic
features as allophonic aspiration of voiceless stops, a retroflex vocoid realization of /r/, and lengthening of
short vowels under stress; the last is a potential complication in Hungarian, since it has phonemic vowel
length (which is preserved in McKeesport Hungarian). Similarly, the partial replacement of SOV by SVO word
order and the replacement of quantifier + singular noun by English-style quantifier + plural noun certainly
make the language's structure more like English, but they don't obviously simplify it. In the last category,
changes attributable both to borrowing and to simplification, we find such changes as the loss of
morphophonemic rules of coalescence and voicing assimilation in consonant clusters, partial collapse of the
distinction between definite and indefinite conjugations, dramatic increase in the use of Hungarian generic
sibling terms for “male sibling” and “female sibling” instead of specific terms for elder and younger siblings,
and loss of the semantically redundant pronominal possessive suffixes in the dative possessive construction.
All these changes bring the structure of the language closer to English, and all of them arguably simplify the
language.

It is of course quite possible that McKeesport Hungarian is an idiosyncratic case for some reason, or (more
plausibly) that language attrition in certain types of immigrant languages differs from language attrition
under other social circumstances. And it is certainly true, as various authors have noted, that a complete
picture of the attrition process would have to take into account the input received by successive generations

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 13 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

picture of the attrition process would have to take into account the input received by successive generations
of learners of the dying language, because that input will surely differ from one generation to the next if the
attrition takes place gradually over two or more generations. The point I would like to emphasize here,
however, is that the search for sources of change in a process of global attrition must not be narrowly
restricted by a belief that there is necessarily an either/or choice between potential contributing factors.

4 Contact-Language Genesis versus Contact-Induced Change

At first glance the genesis of a contact language seems quite distinct from contact-induced language
change, especially in the case of contact languages that are created and stabilized in a relatively few years.
But the mechanisms through which contact languages arise are essentially the same as those which operate
in ordinary contact-induced change - and, as argued in section 2, to a considerable extent in internally
motivated change as well. Contact languages are therefore extreme results of quite ordinary processes.
Moreover, the boundaries between contact languages and cases of heavy borrowing or extensive shift-
induced interference are fuzzy, though there are many clear cases on both sides. Because there are
borderline cases, and because the same mechanisms are common to both, the differences are best
characterized as ones of degree, not of kind (see Thomason 1995, 1997d for arguments in support of this
position). In this section I will support this conclusion by surveying briefly the processes by which the various
types of contact languages arise.

There are, in my view, just three basic types of contact language. The two best-known and most-studied
types are pidgins and creoles, which on my analysis followed the same basic route of development.

Prototypical pidgins and abrupt creoles (that is, creoles that do not emerge from a stable pidgin stage) are
crystallized through “negotiation” from ad hoc efforts to communicate in a new, usually multilingual, contact
situation; this mechanism was discussed in section 2.4, where the point was made that the degree of
bilingualism is the main difference between the results of this process in pidgin/ creole genesis situations
and the results in cases of shift to a TL that is more fully available to the learners.

In sharp contrast to pidgins and creoles, bilingual mixed languages are created by bilinguals: the evidence
for this assertion lies in the fact that material from both component languages is intact, undistorted by the
kinds of processes that operate under conditions of imperfect learning. One type of bilingual mixed
language is created, more or less deliberately, as a symbol of a new ethnic identity. A prominent example is
Michif, spoken by the Métis of Canada and the northwestern US, which has French noun phrases embedded
in a Cree matrix (see Thomason and Kaufman 1988: ch. 9.3; Bakker 1992; Bakker and Papen 1997; for
historical and descriptive information). The most likely initial mechanism for the emergence of this language
is code-switching (section 2.1) -the split between noun phrases and other material is reported from a
number of code-switching contexts. But of course code-switching alone cannot account for its stabilization
as the main language of certain Métis communities; to achieve this result at least a quasi-deliberate decision
must have been taken (see section 2.7), and the new sociopolitical identity of the Métis people provides a
reasonable motive for such a decision. Another example is Media Lengua of Ecuador (Spanish lexicon,
Quechua grammar; see Muysken 1981, 1997). In this case code-switching is a possible mechanism, but
code alternation (section 2.2) is at least as likely. And here again deliberate decision, at some level of
consciousness, is required to account for the outcome - the replacement, by native Quechua speakers, of
almost the entire Quechua lexicon by Spanish lexicon, and stabilization of the resulting mixture for use as
an in-group language. A third example is Mednyj Aleut, the language of one of the Commander Islands, with
Russian finite verb inflection and Russian loanwords in an Aleut matrix (see Thomason and Kaufman, 1988:
ch. 9.4; Golovko and Vakhtin 1990; Sekerina 1994; Thomason 1997a). The major mechanism through which
this language arose is likely to have been code alternation: unlike the mix in Michif, the type of mixture in
Mednyj Aleut is not typical of code-switching.

The other type of bilingual mixed language arises through borrowing carried to an extreme, with gradual
incorporation of lexical and structural features from another language until nothing is left of the original
language but lexicon, including most of the basic vocabulary; the mechanisms involved are the same as in
ordinary borrowing situations. As noted in section 3, this is one way in which language death occurs
gradually, though it is apparently a rare process. In its final stage, as in the case of Ma'a (see Mous 1994),
its effects are indistinguishable from those of language attrition, which, as we have seen, involves extensive
simplification as well as borrowing. This probably accounts for the fact that the idea of a long-term process
of eventually massive borrowing is still controversial (see, e.g., Brenzinger 1987 and Sasse 1992 for a
skeptical view and Thomason 1997b for discussion). Ma'a, which still had bits of non-Bantu grammar as late

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 14 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

skeptical view and Thomason 1997b for discussion). Ma'a, which still had bits of non-Bantu grammar as late
as 1960, now has almost entirely Bantu grammar (fully elaborated, with no simplification or distortion), and
even its non-Bantu lexicon is used only as a secondary code: all its non-Bantu lexical items (mainly of
Cushitic origin) have Bantu counterparts that are also used regularly by Ma'a speakers (Mous 1994). The
mechanisms involved in the creation of Ma'a probably included both code-switching and code alternation,
though there is no way to be certain; another longstanding mixture of this same general type, Kormakiti
Arabic (a mixture of Arabic and Greek), surely involved code-switching, at least.

Both the abruptly created bilingual mixtures and the gradually developing bilingual mixtures serve as ethnic-
group symbols. But while the former are the languages of new groups (a French/Cree mixed-blood group in
the case of Michif, for instance), the latter are the languages of groups that have doggedly maintained their
ethnic identity in the face of overwhelming cultural pressure from speakers of another language (or, in the
case of Ma'a, two other languages, both Bantu).

5 Conclusion: Evidence for Contact as a Source of Language Change

In this chapter I have attempted to give a comprehensive sketch (the emphasis is on “sketch”) of the vast
topic of language change that occurs as a result of language contact. The survey has included classifications
of contact-induced changes from three different perspectives - effects on the receiving system, correlations
of linguistic results with one social variable (presence versus absence of imperfect learning), and differing
results in different linguistic subsystems - as well as seven mechanisms through which contact-induced
change occurs. Briefer sections touch on the important topic of changes that occur in dying languages and
the essential unity in processes of contact-induced change and of contact-language genesis.

In emphasizing mechanisms of change and their differing results under different social circumstances, my
goal has been to give an idea of both possibilities and probabilities. I have argued that anything is possible,
from a strictly linguistic perspective; assessing the probabilities that any particular (type of) contact-induced
change will occur requires careful sifting of both social and linguistic factors. Among the more useful
predictors are degree of bilingualism, degree of linguistic integratedness into a system, and typological
distance between the source and receiving languages. But the most interesting examples are those which,
like the switch of anaphoric gender-agreement patterns in a New Guinea language, highlight the
idiosyncratic nature of speakers’ creativity. This factor alone guarantees the continuing failure of all attempts
to construct a neat hierarchical ordering valid for all types of contact-induced change.

One topic that has hardly been touched on in the preceding sections might be thought to be the intended
heart of the matter, given the inclusion of this chapter in a part entitled “Explaining linguistic change”: what
about predicting when contact will cause change? This topic has been slighted not because it is
unimportant, but because, as noted near the beginning of the chapter, no global predictions can be made.
Investigations of contact-induced language change do not permit any confident predictions about when such
changes will occur, any more than investigations of internally motivated change permit such predictions.

One question to which some useful answers can be given, however, concerns the retrospective identification
of contact-induced changes. That is, how can we tell when we should claim language contact as a, or the,
source of a particular change? The easy cases are those in which both form and function have been adopted
from another language. When Han Chinese conjunctions turn up in the Kadai language Mulam, together with
un-Kadai-like Chinese syntactic patterns (Zheng 1988: 174), no one is likely to deny that Han Chinese is the
source of both the morphemes and the grammatical change. Even when no source language can be
identified, the shape of a loanword will often betray its foreign origin - for instance, it may (like English
asparagus) be too long for a single native morpheme, or it may violate the phonotactics of native words.

The hard cases are those in which the interference features consist of structure alone, expressed by native
rather than transferred morphemes. This often occurs in borrowing, when there is full bilingualism (e.g., in
the example of English interference in Italian discussed in section 2.2), but it is especially prevalent in shift-
induced interference, when imperfect learning plays a role in the process. In these cases all the evidence for
the interference is circumstantial: a structural feature of language A matches a feature of language B and
diverges from functionally corresponding features in sister languages of A. How did A get that feature?
Considerations of universal markedness have often been invoked to decide whether the feature was a
spontaneous innovation in A or a transfer from B, the premise being that marked features are unlikely to be
innovated spontaneously. But unless some marked features arose through internally motivated change
sometime in history, no languages would have them, so this argument is inadequate - though suggestive -

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 15 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

sometime in history, no languages would have them, so this argument is inadequate - though suggestive -
when applied to any single feature.

The key to a convincing demonstration that the change occurred at least partly because of contact with B is
to look beyond this one change and consider all the changes that have occurred in A but not in its sister
languages. If this feature turns out to be completely isolated in the system, the only innovation that makes A
more like B, then a contact explanation is not promising. But if other innovations in A also match B features,
then contact with B is a likely cause of the whole package of changes. This is especially true if some
innovations involve actual transferred morphemes, or if there are several added marked features that are
independent of each other (that is, universally marked features with no mutual structural links, e.g., because
they occur in separate grammatical subsystems). In such cases, even changes toward less marked structure
can reasonably be attributed at least in part to B influence, since interference offers a unified explanation for
all the changes.

All this of course requires that B be firmly established as the source language. This task is trivially easy
when the package of innovations includes actual morphemes transferred from a language that is still spoken
in the neighborhood of A (or even elsewhere), or from a now-vanished language that either has well-known
sister languages or is well attested itself. In borrowing situations this requirement is usually easier to fulfill
than in shift situations, because the source language for extensive structural borrowings tends to be that of
a dominant group, while the original language of a shifting group often disappears entirely after the shift
(i.e., when it is not maintained in other communities), and it may not have had any close relatives. In
addition to identifying a source language B, one must know what its structure was at the time of the
proposed interference, and what the recipient language's structure was; otherwise the direction of
interference cannot be established. And finally, it must be demonstrated that the innovations in A occurred
at the time of contact, rather than (say) hundreds of years after the putative source language vanished as a
result of shift.

The lesson here is that establishing contact as a cause of language change is possible under favorable
circumstances but impossible under less favorable conditions. In this respect contact-induced language
change is no different from other subfields of historical linguistics: inevitably incomplete information all too
often makes it impossible to tell just what occurred at some distant past time, and why.

1 But not always. See, in particular, Nancy Dorian's important (1993) paper on the difficulty of distinguishing
internally from externally motivated change in contact situations.

2 See Thomason (1997c) for further discussion and examples of these categories.

3 Note that this revision is not needed for van Coetsem's similar distinction, because his definition does not insist
on language shift -only on the agentivity of a speaker for whom the recipient language is a non-dominant foreign
language. However, van Coetsem's formulation raises other questions; in particular, it relies crucially on the notion
of “linguistic dominance,” but a speaker could in principle be quite fluent even in a non-dominant language, and in
such a case imperfect learning would be unlikely to influence the process of linguistic interference.

4 This section is based primarily on Thomason (1997c). There are, however, several important differences in the
discussion here that reflect changes in my views, thanks in significant part to insightful comments from Nancy
Dorian and Ad Backus, to whom I am most grateful. The other main analytic difference between the schema
presented here and the one in Thomason (1997c) results from my learning about Robin Queen's (1996) results
shortly before I completed the present analysis.

5 This example may or may not be current in general American English, but it occurs regularly in at least one
subpopulation - prison inmates at the state penitentiary in Pittsburgh, PA. White inmates do not speak AAVE, and
in fact it would be socially unacceptable (and quite possibly actually dangerous) for them to do so; but the general
prison slang includes a sizable number of words and phrases that originated in AAVE.

6 Convergence has sometimes been claimed as a separate category of contact-induced change, but the
mechanisms and results fit well into the ordinary category of borrowing. Convergence is perhaps (not certainly)
more likely to be mutual than other borrowing, and it may (though it need not) involve changes in frequency of
occurrence of pre-existing patterns in A and/or B rather than the addition of new patterns.

7 This is not Woolard's own view. She suggests instead that interference may be an indicator, rather than a cause,
of language shift (1989: 357).

12/11/2007 03:43 PM

23. Contact as a Source of Language Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 16 of 16

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747925

8 Classifying the changes is not always a straightforward process, of course. In particular, deciding whether or not
a change simplifies the language's overall structure is often difficult or impossible. Nevertheless, Fenyvesi's counts
are based on a sufficient number of clear cases to permit confidence about the overall pattern in the results.

9 Loanwords were not included in the counts; the lexical features that were included concerned lexical semantics
and calquing.

10 See, for example, Bickerton (1981) for an opposing view about creole genesis. But Bickerton's theory that
abrupt creoles are created by children during first language acquisition crucially entails that pidgins and creoles
arise through totally different mechanisms; this makes his theory less general than mine, and since my approach
accounts for the data as well as his does, it seems to me that mine is preferable. I say “mine,” but of course the
view that pidgin and abrupt creole genesis are the same basic process is hardly original with me. Equally of
course, the assertion that the two approaches are empirically equivalent requires argumentation and evidence.
Exploration of these issues is beyond the scope of this chapter, but the literature on pidgins and creoles contains
many discussions of Bickerton's theory, including, for instance, a brief exchange between Bickerton and me on this
topic in the Journal of Pidgin and Creole Languages 7 (1992).

11 Various arguments have appeared in the literature to the effect that shift-induced interference, in particular,
can be established under certain circumstances even if these requirements are relaxed. One especially interesting
proposal is in Johanna Nichols's recent work (e.g., 1994a). Nichols suggests that certain features, for instance a
distinction between inclusive and exclusive “we,” can be safely attributed to interference from a long-vanished,
unattested substratum language because they are so unlikely to emerge spontaneously. Although I would not
accept the evidence of a single feature (for reasons given above), a large enough package of innovated, highly
marked, and mutually independent features in A might eventually convince even skeptics like me that a
substratum explanation is justified when no candidate for B is attested. Unfortunately, given the very common
tendency for marked features to be replaced by unmarked features, the likelihood that a long-vanished, unattested
substrate language B will leave behind a large enough package of highly marked interference features in A is
probably slim.

Cite this article

THOMASON, SARAH GREY. "Contact as a Source of Language Change." The Handbook of Historical Linguistics.
Joseph, Brian D. and Richard D. Janda (eds). Blackwell Publishing, 2004. Blackwell Reference Online. 11 December
2007 <http://www.blackwellreference.com/subscriber/tocnode?
id=g9781405127479_chunk_g978140512747925>

Bibliographic Details

The Handbook of Historical Linguistics

Edited by: Brian D. Joseph And Richard D. Janda
eISBN: 9781405127479
Print publication date: 2004

Wyszukiwarka

Podobne podstrony:
van leare heene Social networks as a source of competitive advantage for the firm
Accounting Recording and Firm Reporting as Source of Information for Users to Take Economic Decision
Emotion Work as a Source of Stress The Concept and Development of an Instrument
Kołodziejczyk, Ewa Literature as a Source of Knowledge Polish Colonization of the United Kingdom in
Migration, Accomodation and Language Change Language at the Intersection of Regional and Ethnic Iden
Attitudes toward Affirmative Action as a Function of Racial ,,,
Pragmatics and the Philosophy of Language
DEATH AS A SUBJECT OF SELECTED AMERICAN
Dying As A Result Of Fraternity Hazing
Lord of the Flies Symbolism as Illustration of Golding's V
Luhmann's Systems Theory as a Theory of Modernity
Oakeley, H D Epistemology And The Logical Syntax Of Language
Cognitive Exploration of Language and Linguistics
[Wirth]Urbanism as a way of life
McNally & Boleda Relational adjectives as properties of kinds
Język angielski Writing as a mean of discovery
General Aspects of Language

więcej podobnych podstron