Contact Linguistics. Chap

égalité entre hommes et femmes.
equality between men and women.
"Our ideas that we have, you know, equality, fraternity, equality between men and women."

(2) ma ñ´w fii, degg affaire u égalité ay hommes ak femmes yooyu.

I come LOC understand thing of equality IND men and women DEM
"When I came here, I heard about the equality thing between men and women."

Multi-word switches are the main focus of Poplack and her associates' theory of constraints on code switching. The approach they use will be compared with that of Myers-Scotton below.

Sankoff & Poplack (1981) suggest that code switching is produced by a (context-free phrase structure) grammar which is subject to two constraints:

I. The free morpheme constraint, which predicts that a switch will not occur between a lexical stem and a bound morpheme. As Poplack (1980:585) puts it, "Codes may be switched after any constituent provided that constituent is not a bound morpheme." The only exceptions are when a stem has been phonologically integrated into the language of the morpheme. Thus, flipeando "flipping" is possible in Spanish/English code switching, but not runeando "running".

II. The equivalence constraint, which predicts that switches can occur only at points where the surface structures of the two languages coincide. As Pfaff (1979:314) puts it, "Surface structures common to both languages are favored for switches."

The equivalence constraint is illustrated in Figure 1, adapted from Sankoff & Poplack (1981:6). The lines indicate permissible switch points, and the arrows indicate the surface relationships of constituents in the two languages. Switches may occur at, but not between the lines.

Figure 1

A . English I seen everything 'cause I didn't take anything.

B. Spanish Yo ví todo porque yo no cogí nada.
The position Poplack and her associates adopt is essentially that there is no code switching involving single morphemes, since the latter are not distinct from borrowings, and behave to all intents and purposes just like other lexical items in the recipient language. Though they acknowledge that, in theory, single morpheme switches are possible, they discuss no examples of these. Hence the primary focus of their model of code switching is on equivalence-based switching involving multi-word fragments.
Sankoff & Poplack (1981) argue that an independent code switching grammar, incorporating rules from the two monolingual grammars but distinct from either, produces code switching utterances. Ability to code-switch results from knowledge of the rules of both grammars, their similarities and differences (1981:11). Evidence for this is that switching involves no hesitations, pauses, corrections or other interruptions or disruptions in the rhythm of speech. This position opposes the idea that one or the other of the monolingual grammars involved in code switching acts as the "base" or "matrix" language of code switching utterances. They argue that "no algorithm to determine 'base language' so far proposed applies consistently and convincingly to performance data containing multiply switched sentences" (1981:12). This can be seen in the following examples from Spanish/English code switching, (the first already cited as example (4) in Chapter 4) where constituents change several times from one language to the other (Spanish items are in italics):

(3) a. There was a guy, you know, que [that] he se montó [got up]. He started playing with congas, you know, and se montó y empezó a brincar [got up and started to jump]. Sankoff & Poplack (1981:11)

b. Bueno, in other words, el flight que sale de Chicago around three o'clock (Pfaff 1976)

"Good, in other words, the flight that leaves Chicago around three o'clock."

Muysken (1997c:362) refers to this type of code switching as "congruent lexicalization", a situation "where the two languages share a grammatical structure which can be filled lexically with elements from either language." Examples like these suggest that the rules used to construct code-switching utterances may be drawn at times from one language and at times from the other. Sankoff & Poplack's code switching grammar, then, contains the combined lexicon as well as grammatical categories of the two monolingual grammars. Their model appears essentially the same as that of Woolford (1983:523), who also investigated Spanish/English code switching. Figure 1 illustrates how the model generates a code switching sentence:

Figure 1: Model of a code switching grammar (Adapted from Woolford 1983:523-525)

Spanish lexicon

Spanish phrase

s tructure rules






Hybrid constituent
structure tree
English phrase
structure rules S

NP Aux VP

English lexicon

I V NP PP

put
Det N' P NP

the forks en

Det N'

las mesas
For Poplack and her associates, the output of such a grammar is controlled primarily by the equivalence constraint, which ensures that switching occurs only between sentence elements that are normally ordered in the same way by the monolingual grammars. Thus, possible switch sites for Spanish/English code switching include (among others), the following boundaries:

7. 
   Between subject NP and VP;
   
8. 
   Between copula and predicate adjective;
   
9. 
   Between verb and object NP;
   
10. 
    Between verb and complement clause;
    
11. 
    Between auxiliary and verb;
    
12. 
    Between preposition and NP;
    
13. 
    Between subordinating conjunctions and following clauses;
    
14. 
    Between a noun and a following relative clause.
    

On the other hand, switching is blocked where there is a mismatch in constituency between the two languages. A simple example involves the difference in constituency between NP's in Spanish and English. The Spanish rule NP ---> Det N Adj has the English counterpart NP ---> Det Adj N. Since the ordering of nouns and adjectives differs, the equivalence constraint predicts that no switch is possible at the boundary between them. This rules out cases like the casa white or the blanca house, since these combinations are ungrammatical in one or the other language. Other prohibited switch sites in Spanish/English code switching include the following:

• Constructions involving NEG placement, which directly precedes the main verb in Spanish, but follows an auxiliary or modal in English (See figure 1).

• Constructions involving reflexive and object pronoun placement, which precede the verb in Spanish, and follow it in English. (Hence *She wants lo [it]; *Lo she wants.)

Various researchers have pointed out many exceptions to the claims of the equivalence constraint. Thus, Bentahila & Davies (1983) show that this constraint would rule out switches that occur quite commonly in Morrocan Arabic/French code switching. For example, we would expect that switches within the NP would be possible only where the adjective follows the noun, since this is the only ordering shared by the two languages. Yet switches involving pre-nominal adjectives are found, as in the following example (Arabic in italics)

(4) j'ai vu un ancien tilmid djali

"I saw an old student of mine" (Bentahila & Davies 1983:319)

Similar examples are provided by Myers-Scotton for Swahili/English code switching. Example (5) illustrates a noun plus adjective ordering following Swahili rules, though both noun and adjective are English (in italics).

(5) ni -ka -i- taste ni-ka-ona i-na taste lousy sana

1s-consec-obj.cl.9 1s-consec-perceive it-with taste lousy very
"And I thought it had [was with] a very lousy taste" (Scotton 1988:74)

Sankoff & Poplack (1981) in fact acknowledged that even in the case of Spanish/English code switching, the switch site between adjective and noun represents the majority of the few attested violations of the equivalence constraint in their data.

Romaine (1989:124) provides other counter-examples to the equivalence constraint from Panjabi/English code switching. For instance, even though Panjabi has left-branching structure and postpositions, while English has right-branching structure and prepositions, switches can occur within the PP, as in the following example (English in italics):

(6) Parents te depend hona E$

parents post depend be Aux
"It depends on the parents"

Similarly, Pandit (1990:45) provides examples like the following from English/Hindi code switching, where an entire Hindi postpositional phrase is switched for an English prepositional phrase:

(7) John gave a book ek larakii ko

a girl to
"John gave a book to a girl"

Other studies which provide counterexamples to the equivalence constraint include Berk-Seligson 1986 (Spanish/Hebrew); Bokamba 1988 (Lingala/French); Forson 1979 (Akan/English); Nartey 1982 (Adangme/English). The wealth of counterexamples led Poplack & Sankoff (1988) to acknowledge that equivalence-based code switching (as exemplified by Spanish/English code switching in New York city) may be an extreme case. At any rate, the equivalence constraint fails to account for the common occurrence of switching under conditions of structural mismatch.

It follows that equivalence-based models also have little to say about the strategies bilinguals resort to so as to facilitate switching between typologically different languages. As Romaine (1989:117) has pointed out, mismatches in constituent order or placement can often lead to omission or repositioning of elements. Thus, in Dutch/English code switching, subject pronouns can be omitted when they occur at points where the two languages require different word order. Clyne (1987:752) gives the example in (8), which can be compared with the Dutch equivalent in (9).

(8) Dan make the beds and then I do the washing up.

(9) Dan maak ik de bedden (op) and dan doe ik de afwas.

The mismatch between Dutch and English placement of the pronoun is resolved here simply by omitting it in the first clause of (8). Another strategy Clyne records is to reposition a Dutch pronoun so that it conforms to the English placement rule, thus violating Dutch word order, as in (10):

(10) en dan je realize dat this, dat farmleven.....

"And then you realize that this, that farm-life...."
Compare St. Du: en dan besefje dat....

Another weakness of approaches based on surface equivalence is that they pay less attention to categorial equivalence than to equivalence in syntagmatic relations across categories. They seem to assume that the two languages involved in code switching share the same categories, when in fact categories in different languages often fail to match each other closely. Muysken (1995:193) points to several examples of such categorial mismatch - for instance between clitic and non-clitic pronouns; between different types of determiners and demonstratives, different types of auxiliaries, and so on. These types of mismatch often result in omissions of constituents and other kinds of restructuring in code switching discourse. For example, Berk-Seligson (1986) shows that switches from a Spanish base to Hebrew often result in omissions of articles, which Hebrew lacks. There are also omissions of prepositions and prepositional phrases, no doubt motivated by the fact that Spanish prepositions are free forms while those in Hebrew are often bound.

Amuda (1986:411) provides an interesting example of restructuring in Yoruba/English bilingual discourse, which is due to a mismatch in copula structure between the two languages. In code-switched discourse, Yoruba copula wa often occurs before an English predicate adjective, as in the following example:

(11) ò wa very nice. "It's very nice."

Yet Yoruba never uses wa in this kind of construction; predicate adjective structures lack a copula, as in the following example:

(12) ò dara pupo "It's very nice."

Of course, these kinds of innovation and restructuring in bilingual discourse pose problems for all models of code switching, not just those based on equivalence.

Among the strengths of the equivalence-based model is its recognition of congruence (typological fit) as a key factor in regulating code-switching patterns. The model has achieved some success in identifying the structural conditions under which code switching is more likely to occur, and in identifying various types of structural mismatch which inhibit code switching. For instance, its prediction that switches are rare between clitic and non-clitic pronouns, or in cases of mismatch in negative placement, etc., have generally been borne out. Unfortunately, there are still too many exceptions that the model fails to account for.
Other earlier models of code switching explain these exceptions by appealing to structural relations rather than just linear sequence. For instance, Woolford (1983) suggests that switching can occur as long as there is a match in the subcategorization frames of the relevant heads in the two languages. She explains the unacceptability of switches between object pronouns and verbs in Spanish/English code switching by noting that Spanish verbs subcategorize for a preceding object clitic, while English verbs subcategorize for a following free pronominal object. A similar view of constraints on switching is also found in Azuma (1991b:7), who says that "the subcategorization of the main verb is always preserved" and in Bentahila & Davies (1983:329), who suggest that "switching is constrained by the requirement that there be no violation of the subcategorization rules of either language." These approaches go some way toward explaining why switches ruled out by the equivalence constraint are in fact possible. By focussing on grammatical relations rather than linear order, they provide a new perspective on how constraints on code switching operate in different language pairs.