The paper to which this commentary responds (Goldrick, Putnam & Schwarz, Reference Goldrick, Putnam and Schwartz2016) represents a big step forward in a field which was showing signs of stasis – the study of the grammatical properties of intra-sentential code-switching or code-mixing – and this for two reasons. First of all, it explicitly links the insights from the grammatical study of code-mixing to the rich array of results from psycholinguistic research in the domain of bilingual language processing, making use of the Gradient Symbolic Computation framework (GSC; Smolensky, Goldrick & Mathis, Reference Smolensky, Goldrick and Mathis2014). Second, it provides a set of tools to handle problems in the domain of code-mixing having to do with simultaneous representations in the domain of bilingual complexes.

I briefly discussed this phenomenon, which is actually quite common in many bilingual corpora, in Muysken (Reference Muysken2000: 193–219, 253–261), but without really providing a solution. The reason I could not provide a solution is that almost all code-mixing models presuppose some kind of integrity of the component language sections, either sequential (as in Poplack, Reference Poplack1980) or in terms of ‘language islands’, to use Myers-Scotton's (Reference Myers-Scotton1993) metaphor. Neither in sequential or alternation models nor in island or insertional models is it possible to represent complex dependencies as found in the following examples (1)–(2), picked at random from a large set. What these examples have in common is the fact that a collocational expression from one language ( haar mee naar de bioscoop vragen ‘to ask her along to the movies’ in (1) and ‘play a negative role’ in (2)) is embedded, often in discontinuous fashion, in a frame derived from another language.

1. (1):

   

2. (2):

   

Notice that the functional elements associated with the embedded expression are either entirely absent or have been replaced by elements from the language providing the overall syntactic frame, in short the matrix language.

The following example illustrates that even meaningful functional elements (here the negation) can be left out:

1. (3):

   

The idiomatic negative polarity expression me geen reet interesseren ‘not interest me’ (reet is a vulgar term) is reduced to its lexical content.

While many examples in the literature involve bilingual compound verbs, it is not limited to these, as can be seen in the following example:

1. (4):

   

Here the complex expression satisfaire ses besoins ‘satisfy his needs’ is embedded into a Shaba Swahili infinitival complement.

The structures cited have in common that a complex expression, a collocation, from one language is inserted into the frame of another language. In this process, functional elements of the inserted language disappear or are replaced by matrix language elements. Crucially, the components of the collocation may be separated by morphemes from the matrix language, showing that it is not a single constituent that is inserted. The examples supporting the analysis of a single complex insertion involve a collocation of a verb + noun. These examples can well be analyzed in terms of simultaneous access within a Gradient Symbolic Computation framework. However, this requires a way of separating the functional frame from the lexical content: somehow there must be a weighing procedure favoring certain types of elements from a particular language.