Contact Linguistics. Chap

ten eerste, de taal
in the first place, the language.

In (31), there is a switch between main and dependent clause, with each containing three words. This would probably be treated as an inter-CP switch in the MLF model. In (32), the number of Dutch morphemes exceeds that of Arabic morphemes, yet the latter language contains the main verb and begins the sentence. Here, one might argue that the syntactic core of the sentence is Arabic, and assign the ML to that language. Nortier suggests that a working definition of the ML must be based on more than one criterion, including the language of the first words of the utterance, the syntax of the sentence, and the frequency of constituents from each language. She concludes that, if the ML is still doubtful, it should be assumed there is none.

The identification of an ML can be especially problematic in cases of code switching arising from recent or ongoing language shift, where the ML can change from turn to turn even for the same speaker. Nishimura (1986) discusses such a case of code switching by Japanese/English bilinguals (whose first language is Japanese) in Toronto and San Francisco. She found that the most reliable basis for ML assignment was the syntactic pattern of the sentence, including the constituent order as well as morphological marking of syntactic functions such as subject, object, topic, etc. For instance, example (33) is assigned to Japanese as the ML, since it displays Japanese SOV (head-final) order. Note also the use of Japanese function morphemes such as topic marker wa and postposition de 'on.' (Japanese in italics).

(33) Kaeri ni wa border de we got stopped, eh? (p. 132)

return on TOP on
"On our return we got stopped at the border."

Nishimura further argues that the English clauses in such sentences are used as equivalents of Japanese predicates, which are inflected verbs, eg tomerareta 'we got stopped'. On the other hand, sentences like (34) would be assigned to English.

(34) I slept with her basement de ['in']. (p. 130)

"I slept with her in the basement."

In this case there is a switch at the boundary of VP and locative PP, with the latter following Japanese constituent order. But the overall syntax of the sentence is English.

Some of Nishimura's examples recall the kinds of mixture found in second language learning, which Hamers & Blanc (1989) referred to as "incompetence switching", to distinguish them from the kinds of code switching practised by fluent bilinguals in stable situations. Sentences (35-36) illustrate.

(35) Only small prizes moratte ne. (Nishimura 1986:128)

get-past TAG
"[We] got only small prizes."

(36) Right in the center grow-shitara (p. 129)

do-CONDIT.
“If (they) grow (it) right in the center (of it)”

Nishimura assigns both of these sentences to Japanese on the basis of their constituent order, despite the fact that most of the words they contain are English.

Sentences similar to these are typical of the varieties of Hawai’ian Pidgin English produced by Japanese immigrants who were born in Japan and who presumably learnt the pidgin as a second language. Bickerton (1981:9) provides examples like the following:

(37) a. mista karsan-no tokoro tu eika sel shite

Mr. Carson-POSS place two acre sell do
“I sold two acres to Mr. Carson’s place.”
b. sore kara kech shite kara pul ap
and then catch do then pull up
“When he had caught it, he pulled it up.”

The pattern of mixture in all of these cases seems to involve the retention of abstract L1 syntactic patterns with insertion of L2 words which we find in cases of SLA (See Chapter 7). This once more calls into question the suggestion that the code-switching behavior of fluent bilinguals is different in kind from that of second language learners.

The opposite seems to occur in examples like the following, where a Japanese tag is added to an otherwise English sentence:

(38) The idea is bound to change NE ['don't you think?'] (p. 134)

Similarly, we find English discourse markers attached to an otherwise Japanese sentence, as in the following (English in italics):

(39) Anyway, atama ga furui yo. (p. 134)

"Anyway, they are old-fashioned."

These cases of tag-switching do not affect language assignment.

Other cases of mixture, however, prove somewhat more problematic. Among these are examples like the following, involving switches between subject and predicate (Japanese in italics):

(40) Hakujin no heitai-san didn't like nihonjin or something. (p. 135)

white soldier Japanese
"The white soldier didn't like Japanese or something."

(41) Ushiro no kao wa looks like Japanese. (p. 135)

behind GEN face TOP
"The face of [the man in] the background looks Japanese."

Nishimura assigns (40) to English on the grounds that the subject NP is not marked by any Japanese case particle. On the other hand, she assigns (41) to Japanese merely because the topic marker wa follows the subject, despite the fact that the VP is English. This seems somewhat arbitrary, as is her argument that "for our bilingual informants, a Japanese NP without a particle is equivalent to an English subject NP" (p. 136).

Another problematic example is the following:

(42) She-wa took her a month to come home yo. (p. 136)

TOP TAG
"Talking about her, it took her a month to come home, you know."

Nishimura apparently assigns this sentence to Japanese, claiming that its structure is identical to that of (41). She analyses she-wa as a Japanese topic, since it can't be the subject of the VP. But if this is so, it might be argued that she-wa is simply a substitute for English as for her, and that (allowing for the missing dummy element it) the rest of the sentence is English. Nishimura herself suggests the possibility that both (41) and (42) consist of a Japanese topic and an English sentence (p. 137).

Finally, Nishimura offers several examples of what she calls "portmanteau sentences" (p. 139) to which a single ML cannot be assigned. These are typically sentences in which an English clause is merged with a Japanese clause because they share an intermediate constituent, as in the following:

(43) You should see his karada kimochi warui n da. (p.139)

body appearance awful is
You should see his body; [his body's] appearance is awful."

Mergers like this are possible because the final element of the English sentence (whether it is switched or not) can also function as the initial element of the Japanese clause. Such sentences may be facilitated by the mirror-image relationship between English and Japanese constituent order, for example, VO/OV; P NP/NP P, etc. These mirror-image sentences resist assignment to any one ML.

All of the examples considered here raise interesting questions about the criteria for identifying the ML in a code-switched utterance. They therefore pose a challenge to models like the MLF, which treat code switching as involving a single ML grammar rather than interacting grammars.
In general, however, there appears to be strong empirical support in code switching data cross-linguistically for the notion of a matrix language in intra-sentential code switching. According to Myers-Scotton (1993b:69), "supporting evidence that all sentences have an ML is that code switching material in all data sets can be classified as one of the three types of code switching constituent [discussed above]." Moreover, an experimental study by Kamwangamalu & Lee (1991) showed that bilingual speakers could easily identify the ML in Mandarin Chinese/English code switching utterances. Treffers-Daller (1994:238) also points out that in most cases it is easy to identify the matrix language in her French/Dutch code switching data.