I thank D & B for their stimulating review that concentrates on the MLF model and code-switching (CS), even though Contact linguistics (CL) discusses other contact phenomena, too. This is a model of “classic code-switching,” which is defined several times in CL (8, 105, 297). This scope was always intended, but not explicit in Duelling languages (Myers-Scotton 1993a). The nub of the definition is that only one of the languages contributing surface morphemes to a bilingual CP (i.e. clause) is the source of that clause's morphosyntactic frame. The Morpheme Order and System Morpheme Principles of the MLF model identify this language as the Matrix Language (ML). Classic CS contrasts with composite code-switching. The difference is that in Composite CS, part of the abstract morphosyntactic structure comes from more than one of the participating languages. Composite CS may be more prevalent than classic CS, but has yet to be studied systematically (CL, 298).
D&B state that CL applies the MLF model “to a much broader range of language contact phenomena.” Rather, it's more accurate to say that two new supporting models, the 4-M model and the Abstract Level model, are applied more widely than CS in CL. The MLF model remains a model of CS; the 4-M model is a model of morpheme classification; and the Abstract Level model refers to the levels that underlie surface structures.
When the 4-M model is applied to CS, it allows the original MLF model to explain classic CS with more precision and clarity for three reasons: (a) The 4-M model divides system morphemes into three types based on their syntactic properties and when they are accessed in language production; (b) it names them; and (c) it provides a Differential Access hypothesis to explain differences in the empirical distribution of the three types. Most critically, the 4-M model gives a name to the one type of system morpheme that the System Morpheme Principle of the MLF model states must come from only one of the participating languages. This is the outsider late system morpheme. The syntactic property of these “outsiders” is that they depend on information for their form that comes from outside the phrase in which outsiders occur. Affixes or clitics that co-index objects and subjects are clear examples of outsiders.
D&B wish the MLF model were something that it is not; it relies on only grammatical constraints to explain basic grammatical structure in bilingual clauses. After DL (Myers-Scotton 1993a), I abandoned counting morphemes in a discourse sample larger than the clause as a way to identify the ML (cf. the second edition of DL, Myers-Scotton 1997, 246). This change seems to be why D&B argue that the model has “undergone substantial changes.” I won't argue whether this is a substantial change; but I have always thought the Morpheme Order and System Morpheme Principles are the main ways to identify the Matrix Language.
D&B believe that the discourse level is important in discussions of CS. I agree! The evidence is my extensive publications on the social side of CS and my Markedness Model, a social model that accommodates CS very explicitly (e.g. Myers-Scotton 1993b, 2002). Myers-Scotton 1993c refers to parallels between social factors and types of CS patterns, using Bentahila & Davies's (1992) discussion of generational differences in CS patterns as an example. In CL I conclude the full chapter on language contact's social side by stating that “many of the ways in which these contact phenomena differ [including CS] have to do with differences in the sociopsychological milieux in which they develop.” But after recognizing that options exist, I continue, “However, this volume argues that a single set of structural principles is behind all the options that differentiate the outcomes, whatever their sociopsychological setting” (52). That is, my research leads me to argue that there are universally present grammatical constraints on CS. Limiting grammatical analysis to the clause, which is where languages are truly in contact, doesn't mean that sentence structure, discourse context, or sociopsychological factors aren't worth studying, but for other purposes. The aim of the MLF is to provide the basis for constraints on any grammatical patterns in a CP (clause) any speakers can employ whatever their purposes.
D&B also wish that in CL I offered a better definition of cross-language structural congruence and its role. I wish I had a better definition; but what congruence means cross-linguistically is not well understood. I do assume that sufficient congruence plays a part when a particular ML + EL sequence is possible, but I still only rely on the principles of the MLF model, including the possibility of EL islands to constrain that sequence.
A related criticism from D&B is that I envision the ML as “a constant property of whole conversations.” No, I make two points: (a) I add a new hypothesis that disallows changing the ML within a clause; (b) but at the same time I provide for changes across clauses, saying “Yes, the ML can – and does – change from one CP to the next for some speakers in some corpora, even though there are not many examples in the codeswitching literature” (CL, 66). Both of these points follow from the Uniform Structure Principle (CL, 8, 120).
Finally, the counterexamples to the MLF model that D&B offer do not always stand up. One example, building high-rise in À côté il y en un autre gros building high-rise ‘Next door there is another big high-rise building’ (CL, 39 from Poplack et al.1987:59) is not “dismissed.” Rather, it is discussed as an EL sequence that is not an example of an EL island. An EL island is a phrase that shows EL morpheme order. English is the EL here, but building high-rise follows ML (French) order in the example cited. It is integrated into the ML; it meets French's well-formedness requirements (N + A order).
In another example, D&B misidentify the ML; it is Arabic, not French. I add brackets around the French NPs, which are typical EL islands: [les gens] kaj-interpreter-w [les choses] mal ‘the people interpret the things wrongly’. The System Morpheme principle identifies the ML as the language supplying what are now called outsider system morphemes. The outsiders here are the Arabic verbal affixes that co-index the verb with its subject for subject-verb agreement (kaj for third person present and –w for subject-verb agreement as plural). Thus, even though most of the morphemes come from French, Arabic as the ML provides overall structure.
D&B's example (3) is somewhat problematic. I repeat it for clarity: ils contredisent had la théorie de Darwin pour trois raisons ‘they contradict this the theory of Darwin for three reasons’. The problem is the Arabic demonstrative had ‘this’, because obviously French is the source of the frame for this clause (as the source of the clitic ils and the plural affix on the verb). But had is not much of a problem because it is an early system morpheme, not a late system morpheme. The MLF model says nothing about the required source of these morphemes, although the new Uniform Structure Principle mentioned above states that ML structure is preferred for all structural elements. One explanation for why such EL earlies occur occasionally is “mistiming” (CL, 91–93), and such singly occurring counterexamples, even if they are frequent, are not sufficient to make us question the validity of the MLF model across many corpora.
