What is to be learned?
This paper is an effort to define the target of the language learner: asking, what are the data that the child attends to in the process of becoming a native speaker? This question necessarily engages the definition of language in the largest sense. It is argued that the human language learning capacity is directed to the acquisition of the general pattern used in the speech community. Supporting this view is overwhelming evidence that children do not acquire the non-native features of their parents’ speech, whether these are dialect differences or foreign accents. The end result is a high degree of uniformity in both the categorical and variable aspects of language production, where individual variation is reduced below the level of linguistic significance. Studies of changes now in progress typically show uniform patterns across metropolitan areas and even larger regions involving 30 to 40 million speakers. The sharp dialect boundaries that separate such regions confront us with the obverse problem of accounting for what children do not learn from closely neighboring speech communities.