Introduction

At one time there was a simple picture of the role of the linguistic environment in child language acquisition. The classic motherese hypothesis – as embodied, for example, in many of the papers in the inaugural Snow and Ferguson (1977) volume – went something like this. Just as children cannot learn to ride a bicycle by simply being given one, they cannot learn to speak a language by simply being exposed to one. In both cases adults must provide some training wheels. In the case of language this means addressing children directly, in a high pitch, with child-friendly utterances – morphologically and semantically simple words put together into short and simple sentences – that are tuned in to the child's current focus of attention or conversational topic. In general, empirical research with Western middle-class mothers has confirmed that how well adults do these things is reliably associated with such outcomes as children's early vocabulary size and mean length of utterance (MLU); and there are even some experimental results that suggest a causal role for motherese (see Bohannon and Warren-Leubecker 1988 for an overall review).

The complications with this hypothesis that have emerged in recent years are twofold. First, mothers in some other cultures do not interact with their children in the same way that Western middle-class mothers do.