This paper investigates “mixed” early two-word utterances by bilinguals, in order to determine whether function words match the language context less frequently than content words. Data collected in two language contexts from a child acquiring English and Spanish from birth were used to identify those two-word utterances occurring in the first two months of two-word utterances, between the ages of 1;7 and 1;9. Those utterances containing one word from each language, where one word was a function and the other a content word, were analysed quantitatively to determine whether the function word was more or less likely to match the context than the content word. The results showed that function words matched the context considerably less than content words. This finding is interpreted as suggesting that function words may not be treated as language-specific by early bilinguals, whereas content words are. It reinforces the significance of the well-established function/content distinction in language acquisition theory in a way which would not be possible with monolingual data.