This paper offers an overview of the 4-M model from its inception, but pays special attention to how the characterization of morpheme types in the model has evolved. A new proposal is that the level at which morpheme types are "elected" in an abstract model of language production is a critical factor in predicting morpheme distribution across languages in bilingual data.

A new addition to the model, the Variable Election Hypothesis, predicts which language is likely to be the source of morphemes in certain structures in bilingual speech, based on how they are elected.

Data from available codeswitching literature illustrate the classification of morphemes according to the 4-M model and test this new hypothesis. Much of the analysis focuses on the role of Embedded Language nonfinite verb forms and the structure of mixed determiner phrases (DPs), as well as subordinators in codeswitching data.

Both the 4-M model and the Variable Election Hypothesis make predictions for the source and relative frequency of morpheme types observed in the codeswitching literature. The paper provides evidence that, in bilingual constituents in codeswitching, the frequency of certain morpheme types depends on how they are elected at an abstract level of language production.

The emphasis on the abstract nature of morphemes constitutes a new approach to studying language contact.

These predictions could also apply to other types of contact data, such as borrowing and creole development.