You are on page 1of 12

Studies in Hispanic and Lusophone Linguistics

Volume 3, Issue 2
Fall 2010
Naturalistic and Elicited Data in Grammatical Studies of Codeswitching

Jeff MacSwan & Kara T. McAlister
Arizona State University


Abstract

The authors discuss the merits of naturalistic and elicited data in the study of
grammatical aspects of codeswitching. Three limitations of naturalistic data are
discussed, including the problems of negative evidence, induction, and unidentified
performance error. The authors recommend the use of language surveys as a tool for
overcoming limitations of elicited grammaticality judgment data.

1. Introduction

Codeswitching involves the alternate use of multiple languages within a single
sentence or discourse. Unlike borrowing, which involves the full phonological and
morphological integration of a word from one language (for example, English type)
into another (as Spanish typiar), codeswitching involves the mixing of
phonologically distinctive elements into a single utterance, as illustrated in (1a),
where the Spanish lexical DP mi hermano is mixed into an otherwise English
sentence.

(1) a. Mi hermano bought some ice cream.
My brother bought some ice cream

As is now well established, some codeswitching sequences are well-formed and
some are not. Note that (1a) contrasts with (1b).

b. *El bought some ice cream.
He bought some ice cream

The contrast between (1a) and (1b) is empirical, part of the factual basis of any
theory about the underlying grammatical basis of bilingual codeswitching.
However, the question remains as to the best way to determine the grammatical
facts of codeswitching. In the current paper, we address to what degree naturalistic
observation or grammaticality judgments are more appropriate methods of data
collection in research on grammatical aspects of codeswitching.
522 Jeff MacSwan & Kara T. McAlister
2. Historical development of codeswitching research

At the Thirteenth Annual Round Table Meeting on Linguistics and Languages at
Georgetown University, held in 1962, Einar Haugen claimed to have been the first
to coin the term codeswitching. The word first appeared in print in H. Vogts (1954)
review of Uriel Weinreichs (1953) Languages in Contact and two years later in
Haugen (1956) (Benson 2001). Haugen was one of several pioneers of
sociolinguistics, where the term code had emerged in the discussion of distinctions
in registers and speech styles revolving around social variables, notably by
Bernstein (1971), who had formulated a controversial distinction between
restricted code and elaborated code.
Although the phenomenon of codeswitching was noted peripherally prior to
Weinreich, Vogt and Haugen, an actual codeswitching research literature did not
emerge until the late 1960s and early 1970s, when work focusing on both social and
grammatical aspects of language mixing began steadily appearing. By the 1970s,
codeswitching had already begun to separate along two discrete paths one directed
toward the grammatical study of codeswitching, another toward its social
motivations and use in bilingual communities. However, its origins within the field
of sociolinguistics influenced scholarly views about appropriate data collection,
which may be in large part responsible for the current preference among some
codeswitching researchers for naturalistic methods.
Among the most influential papers on social dimensions of codeswitching was
Blom & Gumperz (1972). Myers-Scotton (1993) credits this paper with sparking
widespread interest in sociolinguistic dimensions of codeswitching; in addition to its
intellectual merits, the work appeared in Gumperz & Hymes (1972)
sociolinguistics reader, which was widely used as a textbook in newly created
sociolinguistics courses throughout the United States, and was therefore part of the
standard curriculum for new scholars and graduate students in linguistics and
sociolinguistics.
Blom & Gumperz (1972) studied codeswitching between dialects of Norwegian
in Hemnesberget, a Norwegian fishing village. The analysis focused on micro-level
sociolinguistic analysis, using naturally occurring data from small-group
interactions as the core data source. Whereas Labov and other sociolinguists tied
language use to sociological variables, Gumperz considered language use to be a
function of the dynamics of interaction. Thus, for Gumperz, language choice
conveys intentional meaning of a socio-pragmatic sort:

Detailed observation of verbal strategies revealed that an individuals
choice of speech style has symbolic value and interpretive consequences
that cannot be explained simply by correlating the incidence of linguistic
variants with independently determined social and contextual categories.
(Gumperz 1982, p.vii)
Naturalistic and Elicited Data in Grammatical Studies of Codeswitching 523
At its origin, then, codeswitching converged with a set of specific research aims that
seemed well suited to naturalistic data collection.
Considerably much has been learned about social aspects of language use in
bilingual speech communities, much of it following up on Gumperzs pioneering
work (see Ritchie & Bhatia 2004 for a review). The aim of this research is to
provide insights into the nature of language use and bilingual community through
linguistic analysis, yielding a valuable description of bilingual speech. Conversely,
researchers focused on grammatical dimensions of codeswitching face very
different epistemological issues; for them, the task has been to define a predictive
and definitive algorithm accounting not only for the grammatical patterns in their
existing data set but also for any codeswitched utterance a bilingual might say.
However, there may be little hope that a set of explanatory principles of the sort
one sees in the theory of grammar can be developed to predict precisely when and
how people will codeswitch. As Lipski (1978) emphasized,

Indeed, since the role of individual idiosyncratic factors seems to be an
important aspect of code-switching, in that among groups of approximately
equal bilingual abilities, some code-switch more than others, a complete
determination of the sufficient conditions for code-switching probably lies
beyond the reach of the behavioral sciences. With regard to the linguistic
constraints, however, the path toward an eventual model seems more
clearly indicated. (p.261)

Faced with the task of describing and analyzing social motivations for
codeswitching, researchers show a clear and appropriate preference for naturalistic
data, which bring along rich and indispensible aspects of the social context for
language use. Clearly, no meaningful analysis of conversational turn taking,
discourse, or even relative frequency of codeswitching can be undertaken in the
absence of clear and detailed knowledge of the context in which speakers find
themselves. As we will stress in the next section, however, the preferred method of
sociolinguists focused on these aspects of codeswitching does not serve us well
when the research question takes a grammatical turn.

3. The task of a grammatical theory

A grammatical theory arises within the context of a specific research program (such
as generative grammar) to account for patterns of linguistic structure observed in
nature. Formally, we think of a grammar as a definition of a language, obliged to
detail necessary and sufficient conditions for the realization of structure. As such, a
grammar (or grammatical theory) specifies the (infinite) class of well-formed
expressions of the language under analysis. A grammar of English not only
describes all the English sentences we have heard so far; it tells us which English
524 Jeff MacSwan & Kara T. McAlister
sentences are both possible and impossible, that is, which sentences could occur at
any time, under any circumstances. As Lipski and others point out, the scope of this
task must not be overly ambitious, whether one is focused on monolingual or
bilingual language data: a theory of grammar tells us only about grammatical
structure, and not about which sentences a speaker might be motivated to say in
some specific context. As Chomsky (1959) noted in his classic rebuttal of Skinners
(1957) Verbal Behavior, what a person might chose to say in response to a specific
environmental stimulus cannot be directly predicted from the stimulus:

A typical example of stimulus control for Skinner would be the response to
a piece of music with the utterance Mozart or to a painting with the
response Dutch. These responses are asserted to be "under the control of
extremely subtle properties" of the physical object or event (108). Suppose
instead of saying Dutch we had said Clashes with the wallpaper, I thought
you liked abstract work, Never saw it before, Tilted, Hanging too low,
Beautiful, Hideous, Remember our camping trip last summer?, or whatever
else might come into our minds when looking at a picture (in Skinnerian
translation, whatever other responses exist in sufficient strength). (p.146)

However, grammatical structure is an aspect of language which is indeed
subject to predictive analysis. Although we cannot know whether a speaker might
say hanging too low or clashes with the wallpaper (or something else) in
response to a painting, we can predict whether too will precede or follow low, or
where with will be placed in relation to other elements within the example.
More precisely:

The fundamental aim in the linguistic analysis of a language L is to
separate the grammatical sequences which are the sentences of L from the
ungrammatical sequences which are not sentences of L and to study the
structure of the grammatical sequences. The grammar of L will thus be a
device which generates all of the grammatical sequences of L and none of
the ungrammatical ones. (Chomsky 1957, p.2, italics in original)

Similarly, a grammatical theory of codeswitching will generate (predict) all of the
grammatical bilingual expressions, and none of the ungrammatical ones.
In the next section, we discuss some limitations of naturalistic data for the
study of the grammar of codeswitching.

Naturalistic and Elicited Data in Grammatical Studies of Codeswitching 525
4. Limitations of naturalistic data for the study of grammatical aspects of
codeswitching

In light of the nature of a grammatical theory as just discussed, three specific
limitations appear to face naturalistic data in the context of research on grammatical
aspects of codeswitching, including (1) the problem of negative evidence, (2) the
problem of induction, and (3) the problem of unidentified performance error.
Importantly, these limitations appear specifically in the context of research on
grammatical aspects of codeswitching, and do not extend to the study of the social
motivation of codeswitching.

4.1 The problem of negative evidence

A grammatical theory projects a class of grammatical expressions, infinite in range,
based on a finite number of observations of well-formed and ill-formed expressions.
Note that the set of observations which makes up this body of evidence has a
specific character namely, it must provide samples of what can occur (the
grammatical expressions) as well as what cannot occur (the ungrammatical
expressions). While positive evidence will allow the linguist to theorize about the
grammatical sequences of L, as Chomsky (1957) noted, only negative evidence
can provide confirmation of the ungrammatical ones. Thus, without both kinds of
evidence, it will not be possible for the linguist to verify whether a proposed theory
is empirically adequate.
To illustrate the point, consider (2), a Spanish-Nahuatl example (MacSwan
1999).

(2) Amo estoy trabajando
not be/3SS work-DUR
Im not working

Based on (2), one might hypothesize that (3) holds as a descriptive generalization.

(3) Codeswitching between negation and a verb is licit.

The disconfirmation of (3) requires negative evidence, that is, data consisting of ill-
formed expressions in which negation precedes a verb. Such data could not be
obtained through naturalist observation, as it is non-occurring. Indeed, on further
investigation probing bilinguals grammatical intuitions through grammaticality
judgment tasks, we discover the facts in (4).
526 Jeff MacSwan & Kara T. McAlister

(4) a. *El no wants to go
he not want to go
He doesnt want to go
b. *He doesn't quiere ir
He doesn't want/3SS go/INF
He doesnt want to go
c. *No nitekititoc
no ni-tekiti-toc
not 1S-work-DUR
Im not working

Our bilingual consultants confirm the indicated grammaticality judgments in
(2) as well as (4). The facts now suggest that codeswitching in the context of
negation is much more complex. Rather than posit a general ban on codeswitching
in this context, along the lines of (3), we must construct a theory which tells us why
Spanish and English negation, but not Nahuatl negation, results in ill-formedness in
this context. (For an analysis and further discussion on the negation data, see
MacSwan 2009.)
One sees that a grammatical theory of codeswitching, like any grammatical
theory, must meet necessary and sufficient conditions it must generate all of the
grammatical cases and none of the ungrammatical cases. The only way we have of
knowing whether a theory might be over-generative, sanctioning both grammatical
and ungrammatical examples, is to inspect the predictions of the theory vis--vis the
ill-formed examples. Naturalistic data collection does not include ill-formed
examples (or if it does, they are not identified, as we will discuss below). The
absence of negative evidence makes theory construction in the tradition of
generation grammar impossible, and codeswitching theories which rely
overwhelmingly on naturalistic data have been shown to be dramatically over-
generative (for discussion, see MacSwan 2005). Thus, while naturalistic data may
be useful evidence of what can occur in codeswitching (positive evidence), it is
inadequate alone, as a grammatical theory cannot be properly tested without
evidence of what cannot occur (negative evidence).

4.2 The problem of induction

Induction involves an inference of a general law or principle from the observation
of particular instances. The problem of induction notes that while a set of
observations o
1

i
may confirm a generalization G, observation o
i+1
may disconfirm
it. Codeswitching researchers committed to naturalistic data may assume that the
absence of a specific pattern implies that it is grammatically impossible; however,
we cannot confidently assume that the absence of a form in naturalistic data means
Naturalistic and Elicited Data in Grammatical Studies of Codeswitching 527
that the structure is not permitted. It may be absent because it cannot occur, or it
may be absent because it has not yet occurred during data collection. Whereas the
problem of negative evidence raises concerns over the absence of data indicating
which constructions are disallowed, the problem of induction raises concerns about
the sufficiency of observations leading to firm beliefs about what is allowed.
To illustrate the point more concretely, consider the widely reported absence of
codeswitching between a pronominal subject and a verb (Gumperz 1976, Lipski
1978, Timm 1975 and van Gelderen & MacSwan 2008). In Spanish-Nahuatl
codeswitching, as in numerous other language pairs, one notes a striking absence of
codeswitching in this context; however, grammaticality judgments revealed in (5)
and (6) show a slightly more complex result (MacSwan 1999).

(5) a. *Yo nikoas tlakemetl
yo ni-k-koa-s tlake-me-tl
I 1S-3Os-buy-fut garment-pl-nsf
I will buy clothes
b. *T tikoas tlakemetl
tu ti-k-koa-s tlake-me-tl
you/sing 2S-3Os-buy-fut garment-pl-nsf
You will buy clothes

(6) a. l kikoas tlakemetl
el 0-ki-koa-s tlak-eme-tl
he 3S-3Os-buy-fut garment-pl-nsf
He will buy clothes
b. Ella kikoas tlakemetl
ella 0-ki-koa-s tlake-me-tl
she 3S-3Os-buy-fut garment-pl-nsf
She will buy clothes

The examples in (6) suggest that a more nuanced account of these data will be
needed, one which can analyze the codeswitching facts in the context of the
asymmetries revealed in (5) and (6). Noting that the subject agreement morpheme is
null for the third person in Nahuatl, but is ni- for first person and ti- for second
person, van Gelderen & MacSwan (2008) relate the data to Pollocks (1994)
argument that agreement morphemes on bare verbs are not merely phonetically null
affixes but rather do not exist at all, motivating the claim that English verbs undergo
feature checking only if they are marked with an agreement morpheme whose
features require checking, and otherwise remains in situ. For details about the
analysis, see van Gelderen & MacSwan (2008).
Thus, one cannot infer from the absence of a structure that such structures
cannot occur, and waiting for relevant data to confirm or disconfirm theories
528 Jeff MacSwan & Kara T. McAlister
narrowly focused on specific structures may lead to perpetual disappointment.
Elicitation of data provides a convenient and efficient shortcut: we can simply ask
bilinguals who codeswitch whether a particular structure is one they might say in
some context or another.

4.3 The problem of unidentified performance error

Linguists and child language acquisition researchers assume that typically-
developing mature speakers will evidence some degree of error in naturally
occurring speech due to such factors as slips of the tongue and fatigue. So-called
performance errors of this nature are believed to result from the failure of language-
related performance systems to execute grammatical instructions due to the
interference of non-linguistic factors. These are errors of the sort we all make every
day, errors we often recognize ourselves as inconsistent with our knowledge of
language immediately after producing them. Researchers have estimated the range
of error rate in typically developing mature speakers to be as high as 10% (Brown
1973, Goodluck 1991 and Reilly, Marchman & Bates 1998).
Clearly, we expect bilinguals to be as susceptible to speech errors as
monolinguals, and might expect some proportion of data collected through
naturalistic observations to be erroneous in precisely this way. It is a significant
problem for codeswitching researchers concerned with the grammatical analysis of
bilingual speech that some proportion of their data are ill-formed, but the specific
ill-formed cases are not identified. Hence, theory construction proceeds with no
knowledge regarding the quality of data collected, and significant conclusions may
turn on examples which the speakers themselves might identify as ill-formed.
One can raise objections regarding grammaticality judgment data as well
(Mahootian & Santorini 1996 and Myers-Scotton 2006). We will briefly address
these in the context of our conclusions and recommendations below.

5. Conclusions and recommendations

The reliance on naturalistic data in codeswitching research arose from its historical
roots in the sociology of language and related topics. Such data are well suited to
linguistic analysis focused on conversational turn taking, discourse, and social
motivation of codeswitching, and would indeed seem to be an essential element of
any effort aimed at understanding bilingual language use in its social context.
However, while naturalistic data can be informative and suggestive in research on
grammatical aspects of codeswitching, grammatical theories of codeswitching
cannot be constructed using such data alone.
In particular, there are three specific problems associated with relying
exclusively on naturalistic data for grammatical theory construction. First,
naturalistic data provide us with positive evidence alone; however, negative
Naturalistic and Elicited Data in Grammatical Studies of Codeswitching 529
evidence is an indispensible resource in verification of grammatical theories.
Without it, we cannot detect when our theories over-generate, erroneously
sanctioning ill-formed sentences as well-formed. Second, naturalistic data cannot
provide us with an exhaustive list of even the well-formed sentences, so we cannot
conclude from the absence of a form that it is inconsistent with the underlying
grammar; it may be absent because the grammar does not permit it, or it may be
absent because the speaker has not had occasion to use it. Third, some proportion of
the data will likely be speech errors, but the errors are not identified. Attempting to
construct generalizations about grammatical patterns based on erroneous data will
no doubt lead to incorrect generalizations.
A number of researchers have developed or begun to develop large corpora of
codeswitching data. For instance, Hlavac (2000, 2003) collected an impressively
large corpus of 148,000 words from 100 Croatian-English bilinguals, and the ESRC
Centre for Research on Bilingualism in Theory and Practice (2010) has developed a
large corpus of English-Welsh codeswitching data deemed the Siarad corpus. Large
corpora such as these might be collected as part of the BilingBank associated with
TalkBank (see http://talkbank.org/BilingBank/ for discussion). An effective,
appropriate and comprehensive tagging system, such as those described in
MacWhinney (2000) or Curtiss, MacSwan, Schaeffer, Kural & Sano (2004) for
similar research purposes, has not yet been developed for the specific purpose of
coding codeswitching corpora. Such corpora provide useful insights into bilingual
language behavior, and should be encouraged, but they cannot serve as the
exclusive source of information when the specific research aim is to discover
whether a grammatical theory generates all and only the (infinite) set of
grammatical sequences permitted in bilingual speech, for reasons discussed.
Of course, objections may be made about grammaticality judgment data too,
and with some merit. For example, Mahootian & Santorini (1996) argue that
grammaticality judgment data cannot be trusted due to the highly stigmatized nature
of codeswitching. While codeswitching may be stigmatized for some speakers
(professional translators, for instance), it also serves as a source of linguistic pride
and as a maker of identity in numerous instances (English in the Philippines or
Japan, Chicanos in a bilingual community in the United States). Moreover, linguists
have successfully conducted research on stigmatized languages and language
varieties for many years, such as Nahuatl in Mexico or African-American
Vernacular English in the U.S., and they have used traditional grammaticality
judgment tasks to do so. However, it does indeed seem appropriate to worry that
attitudes about codeswitching, which vary considerably among bilinguals, may
negatively affect the quality of elicited data.
Another concern, emphasized by Schtze (1996), is that grammaticality
judgment data from single consultants, or even very small samples, may be
misleading. Schtze argues that the use of grammaticality judgments alone is not
sufficient, but rather that such data should be part of a well-designed methodology
530 Jeff MacSwan & Kara T. McAlister
that includes particular attention to informants and analysis. Informants, for
instance, should be members of the community of the language(s) being studied,
and preferably should not themselves be linguists or the actual researcher(s) (p.186).
In addition, how the tasks are presented and the details of the rating scale can also
affect the quality of the data. Schtze (1996) notes that individual language
consultants may vary in their judgments for reasons unrelated to the task, such as
age, education, and metalinguistic processing.
Interference from non-linguistic factors, such as attitudes about codeswitching
and those enumerated by Schtze, may be ameliorated by scaling up the number of
respondents to the grammaticality judgment survey. Researchers will then be able to
identify trends within the linguistic data, and might regard anomalies as error
introduced by non-linguistic factors. These methods have been put to use in
Cantone & MacSwan (2009) and McAlister (2010), and permit researchers to focus
bilinguals attention narrowly on a set of constructions and determine their
grammaticality based on data trends rather than the intuitions of an individual
bilingual.
Another approach which may be useful when very few speakers are available is
the focus group, used in MacSwan (1999) for Spanish-Nahuatl. The focus group
format allows speakers to openly discuss their shared intuitions, imagining whether
an utterance could be used in some or another context, and placing less stress on
individual consultants. The opportunity for consultants to share their intuitions with
other bilinguals and discuss their reactions helps to clarify the facts for participants
and researchers alike. Finally, in any approach involving elicited judgment data,
researchers should train consultants carefully with respect to the nature of the
questions being asked, perhaps by using a clear case contrast (Labov 1975).
In any case, researchers should ask questions about their speakers specific
language background. Codeswitching researchers generally focus on simultaneous
bilinguals who are reported to be comfortable switching languages. In recruiting
participants for a survey or focus group, researchers will want to make sure that
participants have the appropriate language background and attitudes before data
collection begins.

References

Benson, Erica. 2001. The neglected early history of codeswitching research in the
United States. Language and Communication 21(1), 23-36.
Bernstein, Basil. 1971. Class, codes and control. Volume 1: Theoretical studies
toward a sociology of education. London: Routledge and Kegan Paul.
Blom, Jan-Petter & John J. Gumperz. 1972. Social meaning in structure: Code-
switching in Norway. In John J. Gumperz & Dell Hymes (eds.), Directions in
sociolingustics, 407-434. New York: Holt, Rinehart and Winston.
Brown, Roger. 1973. A first language. Cambridge: Harvard University Press.
Naturalistic and Elicited Data in Grammatical Studies of Codeswitching 531
Cantone, Katja Francesca & Jeff MacSwan. 2009. The syntax of DP-internal
codeswitching. In Ludmila Isurin, Donald Winford & Kees de Bot (eds.),
Multidisciplinary approaches to codeswitching, 243-278. Amsterdam: John
Benjamins Publishing.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
Chomsky, Noam. 1959. Review of B.F. Skinner, Verbal behavior. Language 35, 26-
57.
Curtiss, Susan, Jeff MacSwan, Jeannette Schaeffer, Mural Kural & Tetsuya Sano.
2004. GCS: A grammatical coding system for natural language data. Behavior
Research Methods, Instruments, and Computers 34(3), 459-480.
ESRC Centre for Research on Bilingualism in Theory and Practice. 2010. The
Siarad Corpus. Available from www.siarad.org.uk/siarad.php
Goodluck, Helen. 1991. Language acquisition: A linguistic introduction.
Cambridge: Blackwell.
Gumperz, John J. 1976. The sociolinguistic significance of conversational code-
switching. In Jenny Cook Gumperz & John J. Gumperz (eds.), Papers on
Language and Context: Working Paper No. 46, 1-46. Berkeley: Language
Behavior Research Laboratory, University of California.
Gumperz, John J. 1982. Discourse strategies. Cambridge, MA: Cambridge
University Press.
Gumperz, John J. & Dell Hymes (eds.). 1972. Directions in sociolinguistics. New
York: Holt, Rinehart, and Winston.
Haugen, Einar. 1956. Bilingualism in the Americas: A bibliography and research
guide. American Dialect Society 26. Alabama: University of Alabama Press.
Hlavac, Jim. 2000. Croatian in Melbourne: Lexicon, switching and morphosyntactic
features in the speech of second-generation bilinguals. Doctoral dissertation,
Monash University.
Hlavac, Jim. 2003. Second-generation speech: Lexicon, code-switching, and
morpho-syntax of Croatian-English bilinguals. Berlin/ New York: Peter Lang.
Labov, William. 1975. What is a linguistic fact? In Robert Austerlitz (ed.),
Empirical foundations of linguistic theory, 77-113. Lisse: The Peter de Ridder
Press.
Lipski, John. 1978. Code-switching and the problem of bilingual competence. In
Michel Paradis (ed.), Aspects of bilingualism, 250-263. Columbia, SC:
Hornbeam Press.
MacSwan, Jeff. 1999. A Minimalist approach to intrasentential code switching.
New York: Garland.
MacSwan, Jeff. 2005. Codeswitching and generative grammar: A critique of the
MLF model and some remarks on modified minimalism. Bilingualism:
Language and Cognition 8(1), 1-22.
532 Jeff MacSwan & Kara T. McAlister
MacSwan, Jeff. 2009. Generative approaches to codeswitching. In Barbara E.
Bullock & Almeida Jacqueline Toribio (eds.), Cambridge handbook of
linguistic codeswitching, 309-335. Cambridge, MA: Cambridge University
Press.
MacWhinney, Brian. 2000. The CHILDES project: Tools for analyzing talk (3
rd

ed.). Hillsdale, NJ: Erlbaum.
McAlister, Kara. 2010. Age of onset of exposure in code switching. Doctoral
dissertation, Arizona State University.
Mahootian, Shahrzad & Beatrice Santorini. 1996. Code switching and the
complement/adjunct distinction. Linguistic Inquiry 27(3), 464-479.
Myers-Scotton, Carol. 1993. Social motivations for codeswitching: Evidence from
Africa. Oxford: Oxford University Press.
Myers-Scotton, Carol. 2006. Natural codeswitching knocks on the laboratory door.
Bilingualism: Language and Cognition 9(2), 203-212.
Pollock, Jean-Yves. 1994. Checking theory and bare verbs. In Guglielmo Cinque,
Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaella Zanuttini (eds.), Paths
towards universal grammar: Studies in honor of Richard S. Kayne, 293-310.
Washington, DC: Georgetown University Press.
Reilly, Judy S., Virginia Marchman & Elizabeth A. Bates. 1998. Narratives in
children with early focal brain damage. Brain and Language 61(3), 335-37.
Ritchie, William C. & Tej K. Bhatia. 2004. Social and psychological factors in
language mixing. In Tej K. Bhatia & William C. Ritchie (eds.), Handbook of
Bilingualism, 336352. Oxford: Blackwell.
Schtze, Carson. 1996. The empirical base of linguistics: Grammaticality judgments
and linguistic methodology. Chicago, IL: University of Chicago Press.
Skinner, Burrhus Frederic. 1957. Verbal behavior. New York: Appleton-Century-
Crofts.
Timm, Lenora A. 1975. Spanish-English code-switching: El porqu and how-not-to.
Romance Philology 28, 473-482.
van Gelderen, Elly & Jeff MacSwan. 2008. Interface conditions and code-
switching: Pronouns, lexical DPs, and checking theory. Lingua 118(6), 765-
776.
Vogt, Hans. 1954. Review of Languages in contact. Word 10, 79-82.
Weinreich, Uriel. 1953. Languages in contact. The Hague: Mouton.

You might also like