Review: Myers-Scotton (2003)

[Harold F. Schiffman]

Forwarded from: LINGUIST List <linguist at linguistlist.org>

Date:  Sun, 21 Sep 2003 12:30:46 +0000
From:  Alexander Rusakov <rusakov at AR2015.spb.edu>

2nd review

Myers-Scotton, Carol (2002) Contact Linguistics: Bilingual Encounters
and Grammatical Outcomes, Oxford University Press.


Alexander Yu. Rusakov, St. Petersburg State University

The book under review contains a detailed account of Myers-Scotton's
theory in its current state. This theory was first proposed in her
landmark classic ''Duelling languages'' (Myers-Scotton 1993a); further
developments in this theory can be traced in numerous follow-up
studies written by either Myers-Scotton alone (e.g. 1998, 2001 ) or in
collaboration with colleagues (e.g. Myers-Scotton & Jake 1995,
2000). Although the cornerstone assumptions remain unchanged, the
theory has significantly changed since its appearance. It may be
observed that a general trend of that development was a shift from a
theory of code-switching with special stress on its grammatical aspect
to a broader theory of language contacts. The phenomena viewed in this
theory are different kinds of structural outcome in the languages
involved in the contacts, ranging from borrowing to the formation of
pidgin and creole languages. It is repeatedly pointed out that there
is a fundamental unanimity between the phenomena at issue (cf. ''[t]he
same set of principles and processes explains all contact phenomena'',
xii), as well as between bilingual and monolingual speech
(cf. ''[t]hese principles and processes are apparent in language in
general'', xii).

Along with the discussion of the ideas put forward by Myers-Scotton
and numerous linguistic facts in support of those, the book contains
an elaborate and expedient survey of the up-to-date literature for
each of the raised topics. The text of the book is extremely dense,
which poses certain problems for reviewing it. Thus, in the synopsis,
I will confine myself to the indication at the basic issues raised in
each chapter. It is equally impossible to touch upon all the
theoretical problems discussed by Myers-Scotton in the evaluative part
of the review. Thus I am forced to concentrate on a small range of
issues, mostly on those of particular interest to me personally.

SYNOPSIS

The monograph is aptly organized from a didactic point of view. In the
first chapter, a short outline of a theoretical model or, rather, of
several models proposed by Myers-Scotton is offered, ''[c]hapter 2 is
the only one that does not focus on grammatical structures in specific
contact phenomena; instead, it offers an overview of the
sociolinguistic factors that promote bilingualism across societies and
in individuals'' (28). The third chapter contains a detailed
description of the theoretical approach advocated in the book, while
the following three ones show how this approach ''works'' with respect
to the various types of contact data. In particular, chapter 4 focuses
on the ''problematic'' cases of code-switching, chapter 5 on the
problems of convergence and attrition and chapter 6 on lexical
borrowings, mixed languages and creoles. ''The final chapter (Chapter
7) offers a summary in the form of a set of hypotheses based on
discussions in the earlier chapters'' (29).

1. INTRODUCTION (1-29) briefly outlines the subject of investigation
(see above) and introduces the general theoretical base of the
study. This base includes four general principles: - The Matrix
Language principle - The Uniform Structure principle, cf. ''[a] given
constituent type in any language has a uniform abstract structure and
the requirements of well-formedness for this constituent type must be
observed whenever the constituent appears'' (8) - The Asymmetry
Principle for bilingual frames (asymmetry of the participation of the
languages involved in the bilingual speech) - and The Morpheme-Sorting
Principle (''[a]t the abstract level of linguistic competence and
production, there are different types of morphemes. In bilingual
speech, the outcome of these abstract differences is that all the
morphemes from the participating varieties do not have equal
possibilities of occurrence'' (9).

Based on these principles, three models are put forward: the main
Matrix Language Frame model (MLF), that was originally proposed in
(Myers-Scotton 1993a) and then amended almost to its current state in
Myers-Scotton 1997, and two supplementary models developed in
collaboration with Jan Jake - The 4-M model and the Abstract Level
model. These models are thoroughly described in Chapter 3.

A number of questions essential for further argument are tackled in
the Introduction. In particular, ''implications for a model of
language production'' are discussed; an approach adopted by
Myers-Scotton ''presupposes the model of language production'' that is
generally in accordance with (Levelt 1989) although in a modified
version (basically, as a result of putting forward the 4-M model).

A crucial terminological opposition is introduced here between classic
codeswitching (codeswitching in which both the matrix language and
embedded language are preserved more or less intact, and ''the
speakers ... can produce well-formed monolingual utterances in the
variety which becomes the source of... ML'' - 8) and composite
codeswitching (matrix language has gone through a convergence with the
embedded language).

2. THE ROOTS OF LANGUAGE CONTACT (30-52) views language contact
phenomena from a sociolinguistic point of view.  Some factors favoring
bilingualism are revealed, as well as ''the costs and rewards of
bilingualism in the international area'' and the ''motivations to
become bilingual''.  A separate section is devoted to language-use
patterns, here lexical borrowing are dealt with (to be discussed in
more detail in chapter 6) along with the use of language in various
functional domains and sociolinguistic aspects of
codeswitching. Besides, Rational Choice Model (cf. in detail Myers-
Scotton & Bolonyai 2001) is briefly outlined, which is an up-to-date
variant of Myers-Scotton's earlier Markedness model (cf. Myers-
Scotton 1993b). The most important innovation in this model is
assuming ''that choices are best explained as cognitive based
calculations that depend on their estimations of what choices offer
them the greatest rewards... [t]hat a bilingual may see switching
languages at some point in a conversation as a way to optimize
rewards'' (46). Further on, one may find a short section devoted to
language shift; finally, in the end of the chapter and as a kind of
transitory part to the essential part of the monograph, six structural
results of bilingualism are listed which are the topics of the book.
These are (i) lexical borrowing, (ii) codeswitching, (iii)
convergence, (iv) attrition, that goes hand in hand with language
shift, (v) mixed (split) languages, and (vi) creoles (52).

3. EXPLAINING THE MODELS AND THEIR USES (53-107) contains a detailed
description of the three basic models, with special stress on the
innovations as compared to the theory outlined in Myers-Scotton
1993a. Some points are highlighted:

- CP (projection of complementizer) and not sentence is used as unit
of analysis (an argument for that has been already proposed in
Myers-Scotton 1997). Codeswitching addressed to in the monograph is
codeswitching within the CP exclusively. Such a preference is first of
all due to the vagueness of the notion of sentence and, contrariwise,
to the clearness of the notion of CP.  - There are some amendments
with respect to the concept of Matrix Language (ML) if compared to the
1993 model. In particular, it is indicated that, although ML may
change within an utterance, it happens very rarely and, most
importantly, ML does not change within the CP. A discussion follows on
the relations between ML and ''the source variety that the Matrix
Language frame so closely resembles'' (66). In order to demonstrate
the distinction, Myers- Scotton points to the fact that there are two
types of elements that are built into the ML frame (bare forms from
Embedded Language and Embedded Language islands) ''that are not
completely integrated into the morphosyntax of the source of the
Matrix Language'' (67).  Admittedly, however, '''Matrix Language' may
be used as a label for the source language as a short cut'' (67). It
is curious in this respect that on the following page one reads that
''[t]he Matrix Language is an abstract construct... . The Matrix
language is an abstract frame... [i]t does not include actual
morphemes nor is it isomorphic with any fully fleshed-out linguistic
variety'' (68). It seems that the relations (or even a controversy)
between the two understandings of ML, viz. 1) a language form that is
near to, although probably distinct from, the source language (this
distinction is in fact determined by the ML's role in Codeswitching)
and 2) ML as an abstract frame remain somewhat unexplicated (see also
Boussofara Omar 2003).

- An opposition between content and system morphemes yields its place
to a more sophisticated 4-M model. The need for such a model was
called for by the fact that there were system morphemes of Embedded
language that did not meet one of the basic principles of the model,
viz. not to appear in mixed constituents. The crucial point of the new
model is a more detailed classification of morphemes that is based on
the parameters that are in no way related to contact phenomena. The
cornerstone opposition of this new classification is [+/- conceptually
activated] distinction of morphemes. The first group of morphemes
embraces those morphemes that ''are salient at the level of the mental
lexicon''. Lemmas underlying these ''types of morpheme are more
directly linked to speaker's intention'' (74); in other words, such
elements have semantic content'' (76). Content morphemes and early
system morphemes belong to this group, the lemmas underlying the
latter kind of morphemes are, as it were, extracted by the lemmas of
underlying content morphemes, as they are activated on earlier stages
of sentence production. The other group encompasses two types of late
system morphemes that serve syntactic relations, within and outside
the Maximal Projection of Head, correspondingly. These morphemes are
activated at the later stages of utterance production. One of the main
objectives of the book is to demonstrate that these two groups of
morphemes behave differently in contact situations.

Two other points must be emphasized. 1) The very term 'morpheme' is
used to convey two different meaning in Myers-Scotton's book, namely,
for the actual surface-level morphemes, but also for the lemmas that
support them, abstract entities in the mental lexicon
(106). Accordingly, several 'underlying' morphemes may correspond to a
single 'surface' one. This is of particular importance when dealing
with inflexional languages (see below). 2) 'Early' and 'late'
morphemes may be mixed within one grammatical category;
e.g. 'semantic' case morphemes (such as locative and the like) are
'early' morphemes, while syntax-oriented case morphemes belong to the
'late' type of morphemes.

- Another important achievement is an introduction of the Abstract
Level theory claiming ''that there are three levels of abstract
grammatical structure in any lexical item... [:] (i) the level of
lexical- conceptual structure...; (ii) the level of predicate-argument
structure..; (iii) the level of morphological realization
patterns...'' (96). Two domains in which this model is at work are
discussed at some length. On the one hand, in classic codeswitching
(see above for the term) a morpheme of the embedded language that
'pretend' to be uttered must be checked for congruence with its
''Matrix language counterparts''. If this congruence fails at a
certain level, the elements of the embedded language are included in a
not fully integrated form (bare forms or Embedded Language islands;
see Chapter 4 for these problems). On the other hand, the Abstract
Level model neatly accounts for the convergence phenomena (to be
discussed in Chapter 5).

4. CONSIDERING PROBLEMATIC CODESWITCHING DATA AND OTHER APPROACHES
(108-163) views the 'behavior' of morphemes of Embedded Language, when
they do not meet the requirement of congruence (imposed by the
Abstract Level model). One option is the incorporation of bare
forms. It is shown that the incongruence of the NP structures in
Embedded and Matrix Languages leads to the intrusion of a lexical
morpheme in its bare form; on the contrary, if the early system
morphemes of the NP (e.g., determiners) show the full congruency with
the corresponding elements of the Matrix language they may be used
with their content morphemes.

Another topic of this chapter is Embedded Language Islands. A number
of important theoretical issues are touched upon here. These include
triggering (Myers-Scotton is rather skeptical with respect to the role
of this phenomenon), pragmatic and grammatical motivation of Embedded
Language Islands use, Embedded Language islands and proficiency. As
regards the latter, Myers-Scotton makes a rather witty remark: a wide
use of Embedded Language Islands is indicative of high proficiency in
Embedded Language. On the other hand, ''when speakers are nearly
equally at home in both languages, almost ironically, Embedded
Language Islands lose their importance.  Instead, switching between
CPs becomes very frequent as well as switching between sentences''
(149). (Could not it be the case that in this situation one may rather
speak about a short-term poise between the two languages without clear
domination of one of them?)

Finally, in the last section of the chapter, Myers-Scotton tackles the
question of distinguishing between borrowing and codeswitching, the
topic that has been already discussed in some detail as early as in
Myers-Scotton 1993a. This section is primarily based on the dispute
with those researchers whose views go contrary to those of Myers-
Scotton. These are first of all Susanne Polack and her associates as
well the adherents of the Government and Binding theory or Minimalist
Program. Myers-Scotton points at fundamental similarity between
borrowing and codeswitching, at least on a synchronic level.

There are some other crucial points in this chapter that seem to be
relevant for the whole theory of Myers-Scotton.  1) The problem of
different patterns of behavior of verbs resp. nouns in language
contacts. This problem has been attracting researchers' attention for
many years (see e.g. a special section on bilingual verbs in Muysken
2000). This topic is prevailing throughout the book.  There are
several remarks that are worth mentioning in this respect.  - Unlike
nouns, verbs ''are [+thematic role assigner] and therefore carry more
'syntactic baggage' than nouns, meaning their fit with the recipient
language may be harder to make'' (76); - The reason for the frequently
attested use of Embedded Language verbs in do-constructions in Matrix
Language could be ''a conflict of branching requirements between the
Matrix language and the Embedded Language'' (162).  - Infrequency of
adapted verb forms' use may be accounted for by the ''lack of
congruence across tense/aspect systems'' (138).

2) As regards Embedded Language Islands, Myers-Scotton dwells on the
notion of ''internal Embedded Language Island'', that is, a
constituent which is ''part of larger constituent in which they
constitute a sister to a Matrix language element under N-bar...''
(149). In some cases, such an island is in fact just an inflected
wordform of Embedded Language, e.g. a plural forms (ghost-s).
Elsewhere, arguing against (although partially agreeing with) Ad
Backus, Myers-Scotton advances an important observation, according to
which ''[i]dioms, like irregular plurals and irregular past tenses in
English (and other languages), may well be contained in single lemmas
and therefore are not compositionally assembled'' (141). It is not,
however, completely clear whether the units of this kind that are
reproduced by rote are Embedded Language Islands (probably not?). This
problem is very important for the understanding of the essence of
codeswitching in inflexional languages; I'll touch upon it once again
in the last part of the review.

5. CONVERGENCE AND ATTRITION (164-232) discusses convergence as
outcome as having ''two distinctive features: (i) all surface
morphemes come from one language; (ii) the abstract lexical structure
projecting these morphemes no longer comes from one language, but
includes some abstract structure from another language'' (164). These
features are characteristic for the attrition as well. The difference
between these phenomena has a sociolinguistic rather purely linguistic
sense: the convergence is characteristic for the given speech
community as a whole (or for the part of speech community); the
attrition is an individual feature.  Besides It may be passingly
remarked that the distinction between convergence and attrition is
drawn less straightforwardly than is usually typical of Myers-Scotton.

In this chapter several key notions of contact linguistics are
discussed, such as convergence areas (=Sprachbund). It is emphasized
that ''such areas result from past instances of asymmetrical
relationships'' (230). Existing studies of language attrition are
inquired into in much detail. In this discussion, Myers- Scotton
appears to be rather skeptical towards the notion of markedness
(following Thomason and Kaufman 1988 in this respect).

Central for this chapter are theoretical assumptions of Myers-Scotton
herself. Being based on the studies of individual attrition (belonging
to both Myers-Scotton and other researchers), these assumptions are,
of course, valid with respect to convergence as well.

An essential notion of composite matrix language is introduced, i.e.
of a language that has undergone convergence (''[b]oth convergence and
codeswitching necessarily involve a composite Matrix language'',
165). It is noticed below, however, that convergence merely ''often
involves codeswitching''. The major part of the section is devoted to
the discussion of whether Abstract Level model and 4- M model are
applicable in the analysis of attrition and convergence.  A number of
hypotheses are put forward; these could be briefly summarized in the
two following hierarchies of susceptibility of alteration under
attrition: (i) Predicate-argument structure < morphological
realization patterns < lexical-conceptual structure (ii) Late system
morphemes < early system morphemes < content morphemes

It remains unclear, however, with respect to the first of these two
clines, whether it is arrived at deductively or based on a
quantitative analysis of empirical data. In the latter case, it must
be noticed that statistical data reported in the monograph (p. 200)
are not themselves convincing enough for the hierarchy proposed.

6. LEXICAL BORROWING, SPLIT (MIXED) LANGUAGES, AND CREOLE FORMATION
(233-294) is one of the most substantial chapters in the book; it
concentrates on a topic, which is in fact essential for the whole
monograph, namely, on the discrepant behavior of lexical elements and
''those signaling grammatical relations''.

Speaking about lexical borrowings, Myers-Scotton traditionally
distinguishes between cultural borrowed forms and core borrowed forms;
the former may ''begin life'' ''in the monolingual speech of either
bilinguals or monolinguals .... [as well as] in the codeswitching of
bilinguals'' (239), while the latter may do so as code switches
only. It is further argued that there is a crucial difference in the
mechanism of borrowing content and system morphemes (first of all,
late system morphemes), the latter may ''come into a language when its
morphosyntactic frame undergoes a reconfiguration'', that is, after
convergence has come into play.  Borrowing of such morphemes ''is a
sign of a Matrix Language Turnover''. It is noteworthy that
Myers-Scotton does not comment on the borrowing hierarchy of the
different structural types of grammatical morphemes (auxiliaries,
agglutinative affixes, flexions), although this topic is quite popular
in the literature on language contacts.

The following section of the chapter is devoted to the mixed
languages, for which Myers-Scotton prefers the label of 'split
languages'. Two definitions of split languages are given, a strong one
(''[a] split language shows all - or almost all - of its
morphosyntactic frame from a different source language from large
portions of its lexicon; this frame includes all - or almost all - of
its late system morphemes from the language of the morphosyntactic
frame'') and a less stringent one (''[a] split language shows a major
constituent with its system morphemes and major parts of the
morphosyntactic frame from a different source language from that of
most of the lexicon and the morphosyntactic frame of other
constituents'', 249).

The three most commonly known cases of split languages are discussed,
namely, Michif, Mednyj Aleut and Ma'a (Mbugu). The basic mechanism
giving rise to split languages is the Matrix Language
turnover. Convergence is a prerequisite for this mechanism, while
codeswitching is a favored although not completely obligatory
requirement. The Matrix Language turnover may trigger a number of
different scenarios of development: (i) it might be arrested at a
certain stage; such a scenario is assumed for Mednyj Aleut, based
largely on Golovko's (1996, 1999) point of view; (ii) it might be
(almost) completed: ''a complete turnover of all the late system
morphemes with or without a turnover in at least some of the lexicon''
(248); an example of such development is Ma'a; (iii) finally, Matrix
Language turnover may lead to a language shift. As far as Michif is
concerned, Matrix language turnover is not, as far as I can
understand, postulated in its development; rather we deal with a
peculiar combination of fossilized codeswitching and convergence. A
distinct pattern of behavior of lexical and grammatical elements is
typical of all instances of split languages.

It is worth emphasizing that while Myers-Scotton assumes common
structural pattern of development for all split languages, she
acknowledges the difference in direction of such development in
individual languages depending on particular sociolinguistic
circumstances. Discussing conscious effort on speakers' part
(constructing, inventing) as a possible factor in the formation of
split languages, Myers-Scotton notices that ''[s]peakers can
consciously decide they want to change the way they speak, but this is
not the same thing as deciding how to change it'' (253).

The last section of the chapter deals with creoles. Myers-Scotton does
not draw a distinction between pidgins and creoles assuming that
nativization (or its absence) is not related to the structure of these
languages and thus is not relevant for the approach adopted in the
monograph. Generally, Myers-Scotton adheres to ''the subtratist''
position and substantiates this position within her general
theoretical framework. Myers-Scotton's views on the structural
development of creoles are represented in five basic hypotheses that
can be briefly summarized as follows: (i) ''[t]he substrate varieties
contribute to creole formation by supplying the 'invisible'
morphosyntactic frame of the creole'' (277); (ii) ... [s]uperstrate-
content morphemes are much more frequent in the creole than substrate
ones'' (281); (iii) ''[c]ontent morphemes from the superstrate can be
reconfigured as system morphemes'' in creoles (283; curiously,
Myers-Scotton does not mention the notion of grammaticalization at any
point of the discussion of such examples); (iv) ''[e]arly system
morphemes from the superstrate are only available to satisfy creole
requirements when they are accessed along with their heads'' (286,
numerous cases of the use of superstrate words with their determiners
are meant; it seems more appropriate to view these as single units
resulting from 'incorrect analysis' of the noun phrases of superstrate
language); (v) ''[l]ate system morphemes from the superstrate are not
available to satisfy the requirements of the creole morphosyntactic
frame'' due to the difficulty of access to the frame of the
superstrate language (287).

7. CONCLUDING REMARKS: THE OUT OF SIGHT IN CONTACT LINGUISTICS
(295-310), which concludes this monograph, elegantly recapitulates its
main ideas as a set of 'hypotheses for further testing'. These
hypotheses are grouped around two basic theoretical themes: (i)
''[t]he asymmetry between participating languages in contact phenomena
and the press forward of the abstract frame of one language to
prevail'' (297), and (ii) ''[t]he inherent lack of parity between
different types of morphemes within the abstract frame of all
languages in terms of their patterns of distribution'' (297). Another
crucial point is putting forward some basic assumptions that link
together the various contact phenomena discussed in the book. A
summary of basic assumptions reads as follows: ''If there is language
shift, the mechanisms involved follow this hierarchy: Classic
codeswitching < convergence < composite codeswitching (i.e. shift most
likely with composite codeswitching)'' (299).

EVALUATION

It must be clear from what has been said above that the new book of
Myers-Scotton is one of the most important contributions to the study
of codeswitching and language contacts that has been published in the
recent years both because of the width of issues discussed and
theoretical depth of treating these issues. It seems, however, that
the very striving for an all-embracing and uncontroversial global
theory is the reason why some points remain somewhat unclear. Due to
space limitations, I will dwell at some length on those issues only
that are of special interest to me.

1. Composite codeswitching and congruent lexicalisation. According to
Myers-Scotton, composite matrix language, that is, matrix language
that has undergone convergence, is a prerequisite for composite
codeswitching. Grammatical frame incorporates elements of the abstract
structure of embedded language. Thus, incorporation of the elements
from the embedded language is facilitated, as it becomes easier for
these elements to be checked for congruence on all of the three levels
of the Abstract Level model. Basically, the phenomenon at issue is
reminiscent of what Pieter Muysken calls congruent lexicalisation
(Muysken 2000: 153). There is, however, an important distinction
between the two. Muysken does not confine himself to the instances of
assimilation between grammatical structures of two languages as a
result of convergence, treating under the same label those cases when
''[t]he languages share the grammatical structure of the sentence''
(Muysken 2000: 122) due to other reasons (e.g., by virtue of their
close cognation). In other words, Muysken approaches this phenomenon
from a purely synchronic point of view, not taking into consideration
its potentially different origins, while Myers-Scotton, on the
contrary, views it, as it were, from a developmental point of view;
however, she sticks to discussing one particular scenario of
development not inquiring other possible variants. The question
arises, whether Myers-Scotton views as composite codeswitching any
codeswitching occurring between the languages whose grammatical and
conceptual structures are similar. Most likely, the answer is
negative.

A possible cue to this problem could have been a comparison of
codeswitching in those situations when grammatical structures of two
languages are similar due to convergence and in those situations when
there is a pre-established structural affinity; however, such a
comparison is not undertaken.

Another remarks pertains to Myers-Scotton's discussion of composite
matrix language, which is characterized as a deviation from ''desired
target language'', that is accounted for by the lack of ''sufficient
access'' to the latter. This explanation is suitable, it seems, for
the cases of individual attrition.  Things get more complicated if the
history of individual languages is concerned, such as e.g. the
development of Romani dialects almost all of which have undergone
convergence, although with different languages and to a different
extent. On the one hand, the very existence of target language is
doubtful in this case. On the other hand, a study of the real
functioning of Romani dialects shows that speakers of these dialects
have sufficient access to their native language in every particular
moment of time. It's quite another matter that this very language has
changed and the functional range of its use has narrowed under the
influence of the surrounding population (which usually dominates).

2. A tendency to follow rigidly the System morpheme principle leads to
unnecessary, as it seems, formalization of the notion of Embedded
Language island. It is particularly so with respect to the so-called
internal Embedded Language islands (see above), that are sometimes in
fact just isolated word forms of embedded language. As an example of
this the ''unadapted'' Russian verb forms in the North Russian Romani
dialect (NRRD, cf. Rusakov 2001) may be given, that are used in this
dialect along with the adapted nominal forms.  (Curiously enough,
these Russian verb forms and nominal forms are treated uniformly in
metalinguistic comments of the speakers of this dialect; see Rusakov
2001 for more detail). From a purely formal point of view, nothing
prevents treating these forms as instances of Embedded Language
islands. Moreover, the reasons of the use of these forms fit nicely
into Myers-Scotton's theoretical assumptions, namely, the lack of
congruence with the corresponding Russian forms (most likely, on the
level of morphological realization patterns, according to the Abstract
Level model). It appears, however, that under such a purely
formalistic approach, the difference between isolated ''islands'' of
this type and those islands that are relatively long stretches of text
gets unnoticed or underestimated. And still, this difference, however
difficult it is to formalize it, seems to be crucial, and crucial for
the processes of speech production that are generally essential for
Myers-Scotton's approach. Here a reference to Pieter Muysken's
conception could be of some help as he distinguishes between the two
mechanisms o codeswitching (code-mixing, in Muysken's terminology),
that is, between insertion and alternation. In Muysken's view,
internal Embedded Language islands would be undoubted instances of
insertion, while a vast majority of 'longer' Embedded Language islands
would be treated as alternation. It must be admitted for the sake of
objectivity, however, that the difference between alternation and
insertion is less amenable to formalization.

3. Explicit rejection of the distinction between codeswitching and
borrowings on the synchronic level highlights the problem of borrowing
of grammatical morphemes (first of all, of late system morphemes
according to Myers-Scotton).  Myers-Scotton convincingly observes that
the mechanism behind such borrowings is quite different from that
behind lexical borrowings. An explanation of these phenomena as
instances of arrested (after having begun) Matrix Language turnover,
that is, as a first step towards a formation of mixed (split) language
seems convincing and far-reaching as well. However, in order to avoid
a possibility of degrading the notion of 'arrested Matrix Language
turnover' to a mere synonym of high level of interference (in the
spirit of Thomason and Kaufmann), it is vital to understand what
properties are indeed shared (if there are such properties) by
relatively infrequent but non-unique cases of borrowing of grammatical
morphemes from another language (e.g. borrowing of the Russian
prefixes to the NRRD, of Slavic verb prefixes and suffixes to
Megleno-Rumanian, of Turkish verbal affixes to Asia Minor Greek).  It
may be noticed that a similar problem arises with respect to the rise
of ''classical'' mixed (split) languages. Existing theories, among
which Myers-Scotton's is one the most deeply grounded, provide
convincing general patterns of their development; however, none of
them is able to explain why has a particular very unusual
configuration of the elements from the two languages emerged in this
or that particular place and in this or that particular period of
time. This situation is generally typical of historical linguistics,
though.

4. The theoretical model advanced by Myers-Scotton is undoubtedly
universal in nature. However, a question arises to what extent are the
properties of the postulated processes dependent on the typological
characteristics of the languages involved. An example of such
dependence is provided by Myers-Scotton herself who relates frequent
use of bare forms in codeswitching with left-branching character of
the Matrix Language. There are, however, some problems of typological
nature. As has been already mentioned above, Myers-Scotton uses the
term 'morpheme' for both ''the actual surface-level morphemes'' and
''the lemmas that support them''.  Thus, several abstract morphemes
may correspond to a single surface one; this situation is first of all
typical of inflexional languages. Myers-Scotton introduces a ''pull
down'' principle for those cases when surface morpheme corresponds to
''abstract'' morphemes of different types (early and late). According
to that principle ''the entire element shows distribution patterns as
if it were a late system morpheme'' (305). Empirical data in support
of this principle are provided in (Myers-Scotton & Jake 2001). The
question, however, arises how does this principle work in case of
inflexional languages.

As mentioned above, Myers-Scotton assumes that some word forms (of
inflexional languages - A.R.) ''may well be contained in single lemmas
and therefore are not compositionally assembled''. It is obvious that
these cases are viewed as rather peripheral. It must be kept in mind,
however, that there is no unanimity among linguists as to what forms
are produced by rote resp. by rule in inflexional languages. The
problem is even more complicated with respect to the second language
(naturally, Embedded language is usually though not necessarily always
a second language of a speaker). In any case, it is clear that the
problem of holistic processing of inflected wordforms in codeswitching
exists and needs further investigation. It may be also mentioned that
wordforms of inflectional languages are not only characterized by the
cumulative character of expressing morphological meanings, but also by
the blurriness of morpheme boundaries, etc. It seems in this
connection that intrusion of the Embedded language elements into the
Matrix Language frame may cause additional troubles when checking on
the level of morphological realization patterns. For instance,
adaptation of the Russian verbs for their intrusion into the NRRD
grammatical frame requires elicitation of the Russian verb stem, which
is an intricate operation itself, especially if one takes into account
the tangled character of the Russian morphonology.

5. Speaking about word order in the chapter devoted to convergence
(Myers-Scotton points out that word order is an early system morpheme
in some cases) Myers-Scotton almost does not touch upon the
possibility of contact-induced changes that take place on superficial
level, that is, changes of analogical character (syntactic
calques). However, some scholars consider these to be a major type of
contact-induced syntactic changes (see e.g. Joseph 1998). In any case,
changes of this type must play a key role in the assimilation of
syntactical frames of languages involved in contacts. Myers-Scotton
notices on p. 202 that ''abstract specifications for word orders at
all levels of syntax also represent the level of morphological
realization patterns''. It remains, however, somewhat unclear what
role do superficial changes play in Myers-Scotton's theory.

All what has been said above is not to be understood as criticism;
rather, it was thought of as pointing out some problems, further
elaboration of which could have been useful in my opinion. Some minor
remarks and considerations are also scattered in Synopsis.

It is worth emphasizing once again that the book under review
represents an extremely significant contribution to the study of
language contacts.

REFERENCES

Boussofara-Omar, Naima (2003) Review of Contact Linguistics: Bilingual
Encounters and Grammatical Outcomes, by Carol Myers-Scotton.
http://linguistlist.org/issues/14/14-1077.html

Golovko, Evgenij (1996) A Case of Nongenetic Development in the Arctic
Area: The Contribution of Aleut and Russian to the Formation of Copper
Island Aleut. In: Ernst H. Jahr and Ingvild Broch (eds.), Language
Contact in the Arctic: Northern Pidgins and Contact Languages,
63-77. Berlin: Mouton de Gruyter.

Golovko, Evgenij (2000) Language and Ethnic Identity: Sociolinguistic
Conditions for the Emergence of Mixed Languages, Paper presented at
the workshop on Mixed Languages, University f Manchester, 12/2000.

Joseph, Brian D. (1998) Is Balkan Comparative Syntax Possible?
http://www.ling.ohio-state.edu/~bjoseph/

Levelt, Willem J.M. (1989) Speaking: From Intention to Articulation.
Cambridge, MA: MIT Press.

Muysken, Pieter (2000) Bilingual Speech. A Typology of Code-
Mixing. Cambridge: CUP.

Myers-Scotton, Carol (1993a [1997]) Duelling Language.  Grammatical
Structure in Codeswitching. Oxford: Clarendon Press.

Myers-Scotton, Carol (1993b) Social Motivations for Codeswitching:
evidence from Africa. Oxford: Clarendon Press.

Myers-Scotton, Carol (1997) Afterword. In: Myer-Scotton (1993b).

Myers-Scotton, Carol (1998) A way to dusty death: the Matrix Language
turnover hypothesis. In: Grenoble, Lenore A. & Lindsay J.  Whaley
(eds.) Endangered Languages. Cambridge: CUP, 289-316.

Myers-Scotton, Carol (2001) The Matrix Language Frame Model:
Developments and Responses. In: Rodolfo Jacobson (ed.), Codeswitching
Worldwide II, 23-58. Berlin: Mouton de Gruyter.

Myers-Scotton, Carol and Jake, Janice (1995) Matching Lemmas in a
Bilingual Language Production Model: Evidence from Intrasentential
Codeswitching. Linguistics, 33: 981-1024.

Myers-Scotton, Carol and Jake, Janice (2000) Four Types of Morpheme:
Evidence from Aphasia, Codeswitching, and Second Language
acquisition. Linguistics, 38: 6, 1053-100.

Myers-Scotton, Carol and Jake, Janice (2001) Explaining Aspects of
Codeswitching and Their Implications. In: Janet Nicol 9ed.), One Mind,
Two Languages: Bilingual Language Processing, 84-116.  Oxford:
Blackwell.

Myers-Scotton, Carol and Bolonyai, Agnes (2001) calculating Speakers:
Codeswitching in a Rational Choice Model. Language in Society, 31/1:
1-28.

Rusakov, Alexander (2001) The North Russian Romani Dialect:
Interference and Code Switching. // O.Dahl & M.Koptjevskaja-Tamm
(eds.).  Circum-Baltic languages. v.1,
313-337. Amsterdam/Philadelphia: John Benjamins.

Thomason, Sarah and Kaufman, Terence (1988) Language Contact,
Creolization, and Genetic Linguistics. Berkeley: University of
California Press.

ABOUT THE REVIEWER

Alexander Yu. Rusakov is assistant professor at the St. Petersburg
State University, Department of General Linguistics. His research
interests include language contacts, historical linguistics, Balkan
linguistics, Albanian language, and Romani.