MITECS: Semantics, Acquisition of

One area of research in the acquisition of semantics investigates children's use of logical connectives such as and and or, and QUANTIFIERS such as every and some. The main goal of this research is to determine the extent to which children assign a semantics to logical words that conforms to classical LOGIC. We will return to this topic. Another area of research investigates children's knowledge of semantic universals (see LINGUISTIC UNIVERSALS AND UNIVERSAL GRAMMAR). Semantic universals are often cast as constraints against certain linguistic forms or meanings. For example, a constraint on form prevents negative polarity items such as any from appearing in certain linguistic environments (the asterisk indicates deviance):

(1) * Every linguist fed any squirrel. Cf. No linguist fed any squirrel.

A second example is a constraint on meaning, called closure. Closure prevents pronouns from referring back to particular kinds of quantificational NPs that have appeared earlier in a discourse (e.g., Chierchia 1995). Thus, the pronoun he in (2) cannot be linked to the quantificational NP every linguist -- the pronoun can refer to Chomsky, or it can refer to someone who is not mentioned in the DISCOURSE.

(2) Every linguist went to Chomsky's party. He was happy.

Suppose that children's grammars lack these constraints at some stage of development. If so, the language generated by their grammars would produce sentence forms that are illicit for adults, such as (1), and their grammars would permit illicit links between pronouns and quantificational NPs, as in (2). In the absence of systematically available negative semantic evidence (e.g., parental correction), it is difficult therefore to see how children could learn constraints. Embracing the conclusion of the argument from the- poverty-of-the-stimulus, we are led to consider an alternative source: innate specification (see POVERTY OF THE STIMULUS ARGUMENTS).

Among the hallmarks of innate specification are universality and the early emergence of a linguistic principle despite the absence of decisive evidence from experience (see INNATENESS OF LANGUAGE and NATIVISM). Research in the acquisition of semantics is directed at the early emergence hallmark of innateness. For example, 4- to 5-year-old children's understanding of negative polarity items was examined using an elicited production task (O'Leary and Crain 1994). One experimenter acted out stories with toys and props; a second experimenter manipulated a puppet, Kermit the Frog. Following each story, Kermit told the child what he thought happened in the story. The child's task was to decide whether or not Kermit "said the right thing" and, if not, to explain "what really happened." One of Kermit's (false) statements was (3).

In light of what actually happened in the story, children consistently corrected Kermit's statement, producing sentences like "No, every squirrel got some food." Despite Kermit's own use of any in examples like this, children never (re)produced a negative polarity item. This is evidence of children's early mastery of the semantic constraint that regulates the appearance of these items.

Children's adherence to the closure constraint was examined by comparing their responses to discourse sequences like (2) and ones like (4) (Conway and Crain 1995; Conway 1997).

Example (4) shows that a pronoun can refer back to the indefinite NP a linguist, which appears earlier in the discourse. Using a comprehension task, it was found that 3- to 5-year-old children adhere to the closure constraint. They accepted coreference between pronouns and indefinite NPs in discourses like (4) but not between pronouns and other quantificational NPs (e.g., every linguist) in discourses like (2).

This brings us to the acquisition of logical reasoning. An extensive literature on this topic has led to the view that children's reasoning is not the same as that of adults. For example, studies of children's understanding of the universal quantifier, every, has resulted in the widespread belief that children younger than 7 lack adult-like competence. This belief is based in part on the finding that younger children reject sentences like (5) in circumstances with "extra squirrels", that is, ones that were not fed by a linguist (e.g., Bucci 1978; Inhelder and Piaget 1964).

Children's nonadult responses have been interpreted as evidence that their grammars authorize nonadult domain restrictions for the universal quantifier, extending the domain of quantification beyond the boundary conditions established in the adult grammar (e.g., Barwise and Cooper 1981). By contrast, other research presented sentences like (5) in contexts that are arguably more felicitous, yielding consistent adult-like responses from children as young as 4 (Crain et al. 1996; Brooks and Braine 1996).

Studies of other logical words have led some researchers to conclude that children"s logical reasoning does not conform to that of classical logic. A study by Neimark and Chapman (1975) found that children as old as 10 do not tolerate "Some A are B" as a description of a situation in which A and B are equivalent sets. Research findings have also led to the view that children lack the inclusive sense of *or* (see Braine and Rumain 1983). In making judgments about sentences with logical words, however, people are not only influenced by their semantic knowledge (the literal meanings of words) but also by the pragmatic norms they follow (how words are used in conversational contexts; see PRAGMATICS). For example, the statement "Some boy is swimming" is true in situations where every boy is swimming, but this fact is often difficult even for adults to see, because speakers tend to avoid using some N when the use of every N provides a more accurate description of the situation. Similarly, statements of the form A or B do not logically entail exclusivity, but this reading is often inferred (e.g., Chierchia and McConnell-Ginet 1990).

The presence of pragmatic implicatures masks the semantic contribution of logical words in ordinary circumstances, but these implicatures are canceled in certain contexts, such as when a speaker is making a bet or a prediction and, hence, does not know what the actual outcome will be. The conclusion that children"s logical reasoning is not consistent with classical logic is based on research that has not systematically investigated children"s use of logical expressions by manipulating the pragmatic context, so as to cancel the relevant implicatures (see IMPLICATURE). Therefore, it is premature to infer that children lack proper understanding of the truth conditions associated with logical words in standard logic. The critical investigations of children"s logical reasoning are now underway (e.g., Chierchia et al. 1997). When the findings are in we will be in a better position to say how well the acquisition of semantics comports with conclusions based on linguistic research.

Semantics, Acquisition of

See also

References