REVIEWS

Schutz, Alfred (1967). The phenomenology of the social world. Trans. by G. Walsh and F. Lehnert.
Evanston: Northwestern University Press.
Willis, Paul. 1990. Common culture. Milton Keynes: Open University Press.
(Received 11 February 2002)

Language in Society 31 (2002). Printed in the United States of America

DOI: 10.1017.S0047404502314088

Pieter Muysken, Bilingual speech: A typology of code-mixing. Cambridge: Cam-

bridge University Press, 2000. Pp. xvi, 306. Hb $ 59.95.
Reviewed by Melissa G. Moyer
Dept. de Filologia Anglesa, Edifici B
Universitat Autònoma de Barcelona
08193 Bellaterra, Spain
[email protected]

Bilingual speech takes research on code-mixing a step further toward achieving a

better understanding of the differences in what in the past has been referred to sim-
ply as the mixing of two languages in the sentence (or intrasentential code-
switching). In addition, Muysken presents the state of the discipline of language
contact in the year 2000 from the perspective of the grammar and structure of lan-
guage contact phenomena. He brings together and analyzes an extensive set of lan-
guage pairs from a wide variety of communities and social contexts. Good
familiarity with such varied multilingual data provides the author with a strong base
on which to support his three-way classification (insertion, alternation, and
congruent lexicalization) of code-mixing phenomena at the sentence level.
The book is organized in nine chapters. Chap. 1 summarizes the main propos-
als put forth in the code-mixing literature, along with a brief introduction to
Muysken’s three-way classification of code-mixing data. Chap. 2 discusses, with
data, several issues related to the interaction of typologically different languages
with distinct grammars and lexicons. Chaps. 3, 4, and 5 present each of his three
code-mixing types with reference to specific questions raised by different data
sets. Chap. 6 takes up the role of a specific set of closed class items or functional
categories, and the non-equivalence between languages of syntactic categories
such as nouns, verbs, and adjectives. Chap. 7 examines the often special behavior
of bilingual verbs. Chap. 8 seeks to link the different mixing patterns with dif-
ferent bilingual communities and also with the various sorts of extra-linguistic
factors that may influence the mixing types proposed. Finally, chap. 9 looks at the
way these patterns of mixing contribute to language shift. According to Muysken,
no single explanation of code-mixing accounts for the variety of mixed structures
that have been described in bilingual settings. The shortcomings of the three
models in the field that are perhaps best known – Poplack’s (1980) variationist
perspective, Myers-Scotton’s (1993, 1995) 4-M model, and DiSciullo et al.’s (1986)
generative grammar government constraint – are discussed in light of concrete
problems raised by the code-mixing data.
Language in Society 31:4 (2002) 621
 M E L I S S A G . M OY E R

The three processes of mixing are constrained by different structural condi-

tions tied to paradigmatic and syntagmatic relations. Code-mixing at the sen-
tence level is limited by the grammars of the participating languages, but to
determine a grammar for a sentence is not always a straightforward task. This
is a special problem because researchers in the field do not agree on what
sentence-level syntax is in specific terms, nor on whether sentence-level syntax
is manifested at a surface or abstract level. Some of the criteria that have been
put forth for assigning a grammar to a code-mixed sentence are (i) the lan-
guage with the greatest number of lexical items; (ii) the language of the major
sentential functional categories (i.e. tense, infl); (iii) the surface word order of
the sentence; and finally (iv), the language of the main verb that determines the
argument structure of the sentence. Diagnosing the matrix-language grammar
of a mixed sentence becomes more complicated when the languages participat-
ing in code-mixing share these key grammatical features. On a syntagmatic
plane, code-mixing may develop more complex interactions between two gram-
mars depending on the categorial0grammatical equivalence construed by bilin-
gual speakers. Different communities tend to adopt one of the mixing processes
proposed, but the mixing practices of a community are not fixed and absolute.
Rather, they may undergo change in processes of shift or convergence. There
are also communities that use more than one strategy.
Insertion involves the incorporation of lexical items or entire constituents
from one language into a structure of another language. This is a form of uni-
directional language influence. In order to explain code-mixing, there is no
need for Poplack’s distinction between “borrowings” and “nonce-borrowings”:
both phenomena can be considered instances of insertion. A generative syntac-
tic approach that assumes categorial equivalence and functional elements as
heads of syntactic constituents explains cases of insertion. Muysken centers on
noun constructions in order to illustrate this notion, although he explicitly rec-
ognizes that insertion can be extended to include other categorial construc-
tions. At the level of observation, code-mixing of noun constructions is the
most frequent kind of switching observed in most language pairs, but in a for-
mal syntactic analysis, not all noun constructions can be analyzed in the same
way. This is illustrated in the discussion of the difference between Dutch0
Moroccan Arabic and French0Moroccan Arabic mixing.
Alternation is defined as the switching between structures from separate
languages. The boundary of the switches may be a clause, or some peripheral
element such as a discourse marker or tag form. The grammars of two languages
are being used in an autonomous or independent way. The syntactic connections
between alternating constructions involve constituents that have been adjoined;
thus, they are not basic to the clause structure. The absence of bounded syntactic
relations or selection (i.e. argument, syntactic role or subcategorization) makes
code-mixing a strong candidate for being classified as alternation. This contrasts
with insertion in that alternating forms have not been incorporated in a nested
622 Language in Society 31:4 (2002)
 REVIEWS

structure. As a result, there are no syntactic dependencies. At a discourse level,

alternation also corresponds with what has been described as intersentential code-
switching, where a single speaker may say one utterance in one language and
another in a different one. Several issues are raised by alternating structures. The
doubling of prepositions observed by Poplack in Finnish-English code-mixing
and by Nishimura in Japanese-English code-mixing has not been explained ad-
equately by any code-switching models. Muysken claims that these instances of
doubling are alternation and are thus evidence for an adjustment taking place in
the planning of the sentence. Another type of phenomenon is what Poplack has
identified as “flagged switching.” In this case, Muysken argues that these hesi-
tations (as observed by Poplack with Finnish-English mixing) are ways of draw-
ing attention to the switch in language, and they are simply efforts on the part of
speakers to overcome conflicts in word order or linearization patterns. Regarding
the transition point between two alternating constructions0structures in cases where
the linear order of words in the sentence is parallel, it may be the case that this
equivalence plays a role at the level of sentence processing, allowing for switches
or alternation even when there are grammatical links between the different parts
of the sentence. Finally, an unresolved syntactic question that Muysken raises is
whether (i) the sentence containing alternation should be dominated by a lan-
guage A or B structure, or (ii) whether it is the sum of constituents from language
A and B, or (iii) whether it is simply a non-language-specific node (i.e., sentence
CP or IP), in which case categorial equivalence is assumed.
Congruent lexicalization is the third category of code-mixing distin-
guished. It is defined on a purely observational level as the combination of items
from different lexical inventories into a shared grammatical structure. Both lan-
guages contribute to the grammatical structure of the sentence, which in many
instances is shared. This point raises the question of what aspects of the two
grammars can be different and which must be alike. A definite answer is not
provided, although different data sets are examined with respect to this point.
Congruent lexicalization as opposed to the mixing which fits insertion or alter-
nation types is basically the same as dialect0standard variation and style-shifting
in the Labovian sense. Evidence from Giesber’s study of the dialect in the Dutch
town of Ottersum shows, according to Muysken, that the mixes occurring in this
community are instances of congruent lexicalization. Another important feature
of congruent lexicalization is that it involves what Poplack calls “ragged” or
“non-constituent” mixing, where the elements incorporated from a different lan-
guage do not form any syntactic unit that would permit a formal syntactic account
as with insertion. Multi-word mixing (of both constituents and non-constituents
together) is expected because the two languages being combined for the most part
share the same grammatical structure. Some additional features of congruent
lexicalization include bidirectional code-mixing with frequent back-and-forth
switches, language pairs with homophonous diamorphs, and many mixed collo-
cations and idioms. An important point regarding congruent lexicalization is that
Language in Society 31:4 (2002) 623
 MARK ARONOF F

it is related to language change through processes of structural convergence. The

observable grammatical similarity between the two mixed languages and the cog-
nitive ability of speakers to establish equivalence even when surface structure
may not be identical point to a totally different strategy of mixing than one sees
with insertion and alternation.
A relevant question raised by Muysken is whether bilingual communities can
be characterized by the type of code-mixing pattern they adopt. The author pro-
poses a classification for both stable and immigrant bilingual communities un-
dergoing language shift. While the major part of the book is dedicated to the
processes and constraints on the different types of code-mixing, it is explicitly
recognized that a structural explanation of bilingual data is only one dimension
for understanding code-mixing; one needs to take into account other dimensions,
including the structural resemblance of the languages, the stage in the process of
language shift, level of bilingual proficiency, community attitude toward code-
mixing, and the fixedness of language norms in the community.
Bilingual speech makes an important contribution to the field of language
contact. It is not, however, an introductory text for persons wishing to find out
about code-mixing from a more general perspective. Familiarity with the issues
and debates in the area of code-mixing and bilingual research is helpful in order
to understand variations in synonymous terms adopted throughout the book to
refer to various language contact phenomena. A formal knowledge of generative
linguistics is also needed to follow some of the discussions. Several typograph-
ical errors are present in the text, which sometimes make it difficult to follow
certain examples and lines of argumentation, but these errors do not invalidate
the claims put forth. This is an obligatory reference for those working in the field
of language contact from a grammatical perspective. It is a clear point of depar-
ture that future researchers cannot ignore if they seek to give a comprehensive
explanation of the code-mixing facts that Muysken presents in his book.
(Received 1 February 2002)

Language in Society 31 (2002). Printed in the United States of America

DOI: 10.1017.S004740450231409X

Paul Saenger, Space between words: The origins of silent reading. Stanford,
CA: Stanford University Press, 1997. Pp. xviii, 480. Hb $75, pb $25.95.
Reviewed by Mark Aronoff
Linguistics, Stony Brook University
Stony Brook, NY 11794-4376
[email protected]

This is an impressive, fascinating, and exasperating work of scholarship, based

on an astonishingly exhaustive survey of manuscript codices produced in the
British Isles and western continental Europe between the 7th and the 13th cen-
turies. Saenger traces the transition from continuous to word-divided script, which,
624 Language in Society 31:4 (2002)