The purpose of this paper is to clarify the grammatical constraints on discourse particles in Chinese–Japanese intra-sentential code switching in light of the general framework of the Matrix Language Frame (MLF) model augmented by the 4-M model.

This study retrieves data collected for three years from three Chinese–Japanese bilingual children aged between 2;1 and 5;0.

The database consists of nearly 300 hours of spontaneous conversations that are audio-recorded from the families of the three bilingual children, as well as diary entries. It shows that a large number of code switching utterances involve discourse particles.

Qualitative analyses of the data indicate that discourse particles are generally constrained by the MLF, yet they do not fit into any category of the 4-M model. Morphologically bound, discourse particles represent the information structure of a sentence (as in the Japanese topic marker -wa) or encode constraints on the inferential processes (as in the Japanese complementizer -kara) rather than truth-conditional information. They manifest some idiosyncrasy at the interface of syntax and pragmatics, and set up the MLF at a discourse level. Thus, the MLF model is extended from a merely syntactic level to the syntax–discourse interface.

The present work has contributed empirical evidence from a hitherto undocumented language pair of Chinese and Japanese, and made theoretical explorations on the linguistic constraints of discourse particles.

On one hand, it is work that provides support for the robust nature of universality of the MLF constraints on code switching. On the other hand, discourse particles exhibit typological features that need further theoretical exploration in order to make a more comprehensive account for the grammatical constraints on Chinese–Japanese code switching.