This article seeks to establish connections between two different language contact phenomena, interlanguage and codeswitching. The data used in this analysis come from two sources: an interlanguage corpus which has English as the target language, but also contains material from the speaker's two first languages, Spanish and German; and a German-English codeswitching corpus containing data from 20 relatively balanced bilinguals. A comparison of these two corpora indicates that similar patterns appear in both types of bilingual output, and that the differences can be explained in terms of proficiency. The analysis is framed within the Matrix Language Frame model and relies on the concepts of complex lexical structure and a composite of features from the speaker's previously learned languages as well as the target language. The composite arises when levels of complex lexical structure come from different languages. This perspective on the study of languages in contact sheds light on the issue of L1 transfer and the structure of interlanguage grammar.