old displayed receptive vocabularies (i.e., word comprehension, as distinct from the ability to produce words) that ranged from the level of a typical 1 year, 9-month-old to the level of a 10 year, 8-month-old (Morrison et al., 1997, 1998). These individual differences not only emerge early, but they also appear to be stable over time. It is hard to imagine that such striking differences would not affect how children fare and are treated during their early years of school, in ways that perpetuate the initial differences. In fact, children 's scores on early literacy tasks at kindergarten entry consistently predict academic performance throughout the first three years of formal schooling and beyond (Morrison et al., 1995; Stevenson et al., 1976). Similar patterns have also been reported for early mathematical abilities (National Research Council, 2000).

That these early emerging and quite stable individual differences in language skill are consistently linked to the social class of children's families lends them even greater importance in a society that established its educational system in part to promote equity of opportunity. There is some evidence to suggest that socioeconomic factors exert their most powerful effects on children's achievement during early childhood and that these early influences contribute to sustaining socioeconomic effects on achievement throughout the school years and beyond.

It is also important to note that these aspects of early language development (e.g., vocabulary, semantics), unlike morphology, grammar, and phonology, do not show critical or sensitive periods. In these domains, children can, in principle, catch up given appropriate and sufficient exposure. As Hart and Risley point out, however, the amount of additional exposure a child needs to catch up increases over time. With each passing year, the gap widens and, at some point, may become insurmountable for all practical purposes.

The studies just described explore the effects of linguistic input on child output by examining the natural range of variation found in mother talk to children. But what would happen if one were to augment the amount of input children typically receive? As an example, Nelson (1977) enriched the input children received in forming questions and found that this enriched experience selectively increased the children's production of this type of construction. However, it is not clear from such studies whether the enriched input is actually teaching children a new construction or merely teaching them to produce an already known construction in a particular context. Thus enriched input may be important, not to establish a particular construction in a child's linguistic repertoire, but to influence the production of that construction in a given context. Given that production of language is what teachers hear and base their judgments of competence on, strategies that improve production warrant substantial attention in early invention programs.