Contact Linguistics. Chap

Exercise: Consider the following examples of code switching from various language pairs and determine whether they comply with the equivalence constraint. Do the same for other examples of code switching that you find.


Arabic/French. (Bentahila & Davies 1983).
a. il croyait bi?ana je faisais ça exprès. (p. 310)
“He thought that I was doing that on purpose.”
b. tatbqa tatgratter (p. 315)
You keep DUR-scratch. “You keep scratching.”


Spanish/English.(Poplack 1980)
c. But I wanted to fight her con los puños, you know.
“But I wanted to fight her with my fists, you know.”
d. I got a lotta blanquito friends.
“I have a lot of white friends.”


2.2. Government-based approaches to code switching.

Another approach to constraints on intra-clausal code switching appeals to the notion of government as formulated within earlier X-bar theory. Government refers to the relation between the head of a construction and its complement; thus, for example a preposition is the head of a PP and governs its complement, a noun or NP. Similarly, a verb is the head of VP, while COMP is the head of a clause, and so on. An early attempt to formulate a government-based constraint was made by DiSciullo et al (1986), who suggested that "when a government relation holds between elements, there can be no mixing; when that relation is absent, mixing is possible" (1986:4). In other words, elements related to each other by government must be drawn from the same lexicon, or, as DiSciullo et al put it, must have the same language index.

The constraint on code switching is formulated as follows (Muysken 1995:185):

(13) *[Xp Yq] where X governs Y, and p and q are language indices.

However, some definitions of government were extremely broad, stipulating in effect that heads governed their entire complement or maximal projection (Aoun & Sportiche 1983). As Muysken (1995:186) points out, this broad definition was inappropriate for the government-based constraint in two ways. First, the class of governors it identified was too large, including not just lexical governors such as verbs and prepositions, but also functional categories such as inflections, complementizers etc. A constraint based on this definition would rule out switches that are in fact frequently attested, such as between INFL and the subject, or between a complementizer and its complement clause. The following are examples of such switches in French/Italian code switching (DiSciullo et al 1986:14-15):

Between INFL and the subject:

(14) La plupart des canadiens scrivono 'c'
"Most Canadians write 'c'"

Between COMP and S'

(15) E l'altro dice come s'appelle?
"And the other says how is it called?"

Second, the domain of government identified in earlier approaches was too broad, including in principle the whole maximal projection. This would rule out the switches attested above, as well as others that occur, such as between verb and adverb, or between determiners or quantifiers and the nouns they modify.

Between determiner and noun:

(16) Io posso fare i cheques. (DiSciullo et al. 1986:13).
"I can do cheques."

For these reasons, DiSciullo et al. adopt a modified definition of government in terms of immediate c-command, as follows:

(17) X governs Y if the first node dominating X also dominates Y, where X is a major category N, V, A, P, and no maximal boundary intervenes between X and Y.

In other words, government is minimal, and switching is permitted only when "the highest element in the governed maximal projection is in the same language as the governor" (Pandit 1990:52). Note also that DiSciullo et al. restrict the class of governors to major categories alone, excluding INFL and COMP, contra Chomsky (1981). Muysken (1990:124) reformulated the constraint as follows:

(18) *[Xp Yq] where X L-marks Y, and p and q are language indices.

The revision is intended to allow for attested cases of code switching involving government by functional categories, which the broader definition of government would rule out.
The constraint makes various predictions about possible switches. For example, it predicts that, within a VP, the verb and the immediately adjacent element in its complement will be in the same language. However, the rest of the complement may be switched, as in the following made-up examples of English/Spanish code switching. (italicized items are in the same language, non-italicized items are in the other language.

(19) a. I saw that él se fué.

he left
b. I saw the hombre (‘man’).

Similarly, within a PP, the preposition and an immediately adjacent determiner must be in the same language, but the rest of the NP complement may be switched. DiSciullo et al (1986:12) list various other predictions of their constraint.

Researchers have presented numerous counter-examples to these predictions. Pandit (1990:53) provides examples of switches between the verb and an adjacent complementizer in Hindi/English code switching, as in the following (Hindi in italics):

(20) a. We can't generalize ki [that] love marriage as such is bad.

b. Sudhas Kaa kahanna hai that one should face life as it comes.

Sudha POSS saying is
"Sudha says that one should face life as it comes."

Other researchers have provided examples of switches between P and its complement, (21); between V and its object NP (22)¹¹; between indirect and direct object (23); and between a copula and its complement (24).

Hindi/English: (Pandit 1990:45)
(21) John gave a book to ek larakii.
“John gave a book to a girl.”
Arabic/French (Bentahila & Davies 1983:313)
(22) ?ateik une enveloppe.
"I gave you an envelope."


(Morroccan Arabic/Dutch: Nortier 1990:131)
(23) z&ib li-ya een glas of water of zo
"Get for me a glass of water or so."

(24) wellit huisman.

"I-became 'houseman'."

The wealth of counterexamples led Muysken (1995:188) to concede that "the government constraint, even in the revised form of Muysken (1990), cannot be maintained." He suggests that a reformulation of the constraint which takes the notion of equivalence into account may better account for code switching phenomena. The revised constraint is as follows:

*[Xp Yq], where X L-marks Y, p and q are language indices, and there is no equivalence between the category Y in one language and the category Y in the other language involved."

In other words, an element that is lexically governed can be switched only when it is equivalent to a corresponding element in the other language. This kind of switching seems characteristic of situations involving contact between typologically similar languages. Treffers-Daller (1994:240) finds some empirical support for this constraint in French/Dutch code switching in Brussells. However, this attempt to combine the constraints based on government and equivalence into an overarching constraint has not so far been tested against a body of code switching data. Its virtue is that it takes account both of the principles constraining constituent structure and the nature of the typological fit between the two languages involved. But Muysken offers no precise definition of "equivalence" that would allow one to predict what switches can or cannot occur. And at any rate, the notion of government has itself become outdated.

3. A production-based model of code switching.

As Treffers-Daller (1994:227) points out, the chief problem with the equivalence and government constraints, as well as other earlier constraints, was that they attempted only to identify points at which switching was blocked, rather than explaining which constituents can be switched, and why. The latter, she suggests, should be the central question in code switching studies. Myers-Scotton (1993b) seeks to answer this question. Unlike the scholars discussed so far, she approaches code switching phenomena from the standpoint of the language production process. In her view, code switching can be accounted for only by examining "how language is accessed and retrieved before it takes its final form" (1993b:45). Her Matrix Language-Frame (MLF) model of code switching is a "production-based model which sees code switching constraints as set by processes which operate well before the positional level at which surface orders and structures are realized" (1993b:6). This is diametrically opposed to earlier models of code switching, which she criticizes for "operating at a level which is too 'purely syntactic', or too close to the surface" (1993b:45).

The MLF model is based on the assumption that one of the languages involved in code switching, referred to as the matrix language (ML) sets the grammatical frame for mixed constituents. The grammatical frame consists of morpheme order and system morphemes (System morphemes are roughly equivalent to grammatical or 'function' morphemes; they are discussed more fully below). The other language involved, from which elements are incorporated into the ML frame, is the embedded language (EL). The following example from Swahili/English code switching illustrates this, with Swahili as the ML and English as the EL (Myers-Scotton 1993b:80). Henceforth I follow Myers-Scotton's convention of marking the EL in italics in all examples involving intra-sentential code switching.

(25) Leo si- -ku- come na books z-angu.

today 1s-neg past-neg with cl-10 my
"Today I didn't come with my books."

Strictly speaking, it is the complement phrase (CP), roughly equivalent to the "clause" of traditional grammar, that is the relevant unit of analysis for this model of code switching. As Myers-Scotton (1997:222) notes, what qualifies as a sentence in discourse may in fact contain one or more CP's. Switches from one clause to another would therefore be treated as instances of inter-sentential (or inter-clausal) switches in this framework. The MLF model is not concerned with such switches. Cases of co-ordination as in (26) and subordination as in (27) are examples of inter-clausal switches (French in italics).

Moroccan Arabic/French (Bentahila & Davies 1983:309):

(26) J'avais faim w xft na:kul

"I was hungry and I was afraid to eat."

(27) Si j'avais la maison, ma¿amri na:kul temma.

"If I had the house, I would never eat there."

Intra-sentential code switching is defined as follows in the MLF framework:

“A CP shows intra-sentential code switching if it contains at least one constituent with morphemes from language X and language Y." Myers-Scotton 1997:222).

Such mixture can result in two different kinds of mixed constituent, one involving single morpheme EL switches, as in (25) above, and the other involving EL phrases or "islands" as in (28), from Myers-Scotton (1993b:131), citing Oloruntoba (pc 1990).

Yoruba/English: (Oloruntoba 1990)

(28) Awon nkan ti o come naturally to me ni molike
those things that PRO is I-like
"I like those things that come naturally to me."