Contact Linguistics. Chap

Exercise: Examine the following examples of code switching from various language pairs and determine whether they are cases of inter-clause or intra-clause code switching. In the latter case, determine whether single morphemes or phrasal constituents are switched. Find other examples of each type of code switching on your own.


Swahili/English. (Myers-Scotton, 1993b)
a. Si- ku-come na books z-angu
1s-NEG-PAST come with books cl.10-my
“I didn’t come with my books.”
b. Tu-na-m-let-e-a our brother wa Thika.
1PL-PROG-him-take-APPL-INDIC our brother of Thika.
“We’re taking it to our brother of Thika.


Japanese/English (Nishimura 1986)
c. Mannaka ni they’re growing (p. 132)
middle in
“They’re growing in the middle.”
d. Camp-seikatsu ga made him rough. (p. 136)
Camp-life NOM
“(That) camp-life made him rough.”
e. He never moratta from anybody. (p. 137)
get-PAST
“He never got (it) from anybody.”


3.1. Hypotheses of the MLF model.

The claims of the MLF model are formulated as a set of interrelated hypotheses, as follows (Myers-Scotton 1993b:7):

I. The Matrix Language Hypothesis.

The ML sets the morphosyntactic frame for ML+EL constituents.

From this follow two related principles:

a. The morpheme order principle:

Morpheme order must not violate ML morpheme order.

b. The system morpheme principle:

All syntactically relevant system morphemes must come from the ML.

II. The blocking hypothesis.

The ML blocks the appearance of any EL content morphemes which do not meet certain congruency conditions with ML counterparts.

III. The EL island trigger hypothesis.

Whenever an EL morpheme appears which is not permitted under either the ML hypothesis or the blocking hypothesis, the constituent containing it must be completed as an obligatory EL island.

IV. The EL implicational hierarchy hypothesis.

Optional EL islands occur; generally they are only those constituents which are either formulaic or idiomatic or peripheral to the main grammatical arguments of the sentence.

To understand why these particular hypotheses and principles are proposed, it is helpful to clarify some of the theoretical underpinnings of the MLF model. It draws on psycholinguistic research which postulates three major components of speech production - the conceptualizer, which creates the preverbal message; the formulator, which generates a linguistic representation of the message; and the articulator, which produces actual speech. The model further assumes that sentence production by the formulator involves two stages: first, a syntactic frame is set, and second, the segmental and prosodic structure of words and their positional arrangements are established (Myers-Scotton 1993c:487). The MLF model is primarily concerned with the first stage - how the sentential form is set and filled in.

Further motivation for the model comes from research on speech errors (Garrett 1988; 1990) and aphasic production by monolinguals (Zurif 1990), as well as psycholinguistic studies of bilingual speech (e.g., Grosjean 1988). What these studies suggest is that:
a. One language is dominant in bilingual language processing.
b. Content and system morphemes behave differently in speech errors and in aphasic language production.
c. Each language may have its own direct access to a common conceptual system in code switching production. (Myers-Scotton 1993b:46-47).

The MLF model draws also on Levelt's view that (morpho-) syntactic processes are directed by information contained in the mental lexicon. Syntactic directions come from so-called "lemma information" or "lemmas" for short. Levelt (1989:6) defines a lemma as the non-phonological part of an item's lexical entry, including its syntactic and semantic properties as well as certain aspects of its morphology. Lemma information regarding meaning and grammatical relations is included within the formulator, along with information concerning word order and phonological form. The MLF model follows Levelt in distinguishing three levels of representation for abstract lemmas: lexical-conceptual structure; predicate argument structure and morphological realization patterns (Myers-Scotton & Jake, 2001:85). These distinctions are crucial to explaining congruence-related constraints on code switching (see section 4 below).

We can now examine how the model attempts to account for actual cases of intra-sentential code switching.


3.1. Establishing the Matrix Language.

There is no clearly established set of criteria for identifying the ML in code switching utterances. Some researchers (e.g., Doron 1983; Joshi 1985) define it as the language of the first word in a sentence. Others (e.g., Klavans 1983; Treffers-Daller 1991) suggest that the ML is the language of the verb or INFL. But these definitions have proven inappropriate in many cases. Nortier (1990:159) provides the following examples from Dutch/Arabic code switching illustrating that neither the first word nor the INFL category of a code switching utterance determines the ML. (Arabic in italics)

(29) emma krijgen ze veel meer

there get they much more
"There they get much more."

(30) xes¡s¡-hu*m anders omgaan met hun.

they-must differently deal with them
"They must deal with them differently."

Myers-Scotton (1993b:68) defines the ML as "the language of more morphemes in interaction types including intra-sentential code switching." She further suggests that frequency counts must be based on a discourse sample rather than single sentences, and that cultural borrowings from the EL must not be counted as part of the latter, since they might skew the comparison. Nortier, however, prefers to distinguish the macro-level of discourse from the micro-level of individual sentences, since "it is possible that (parts of) a conversation have language A as the base language while individual sentences in that particular conversation can have either base language A or base language B" (1990:158). Both Myers-Scotton and Nortier acknowledge that there are individual code switching sentences for which the ML is difficult to identify purely on the basis of a frequency count. Bentahila & Davies (1983:309) provide the following example from Arabic/French code switching (French in italics).

(31) tajbPiw js&ufu s&iÌaz&a qui est different.

they enjoy looking at something which is different.

Nortier (1990:159) adds the following example from her Arabic/Dutch data (Dutch in italics)

(32) Pend-na bezzaf bezzaf moeilijkheden u problemen met, eh, met eh, nou

we-have much much troubles and problems with, uh, with, uh, well,