Can information theory explain early word choice.

One problem of interest in child language study has been specifying what features of the speech situation determine what children will say. Several cognitively based approaches to child language development have proposed principles of 'informativeness' to explain the child's choice of word(s). These principles predict that the child will choose the 'most informative' element of a situation and encode it in speech. Detailed inspection of this view reveals that the notion of ' informativeness ' is not rigorously defined as in information theory, and would require the development of formal semantic and pragmatic information theories. Simpler accounts of available data do not require the notion of 'information'. One of the first applications of behaviourism to complex human behaviour was in the domain of language (Skinner 1957), with the explicit aim of enabling predictions of future speech events on the basis of current environmental conditions and known previous events. This enterprise has resurfaced, with little notice, in the modern garb of cognitively based approaches to child language development which putatively apply information theory to child speech. have proposed specific principles as explaining the child's choice of words in thesingle-word utterance period. The child is said to choose the 'most informative' element of a situation and encode it in speech. An analysis of this position reveals that the notion of 'informativeness' could not be derived from information theory (as the authors suggest) and as used is inapplicable to either spontaneous or experimentally elicited child speech. This failure is critical because these accounts of single-word use and child cognition depend on this theoretical conception.