Background

Young children face a serious problem of induction in word learning, as there are virtually an unlimited number of candidates for the meaning of a word that can be induced from a single referent (Quine, 1960). However, researchers have converged on the view that young children do not go astray in the labyrinth of the induction problem. Children possess a certain set of principles or biases about how words are mapped onto their meanings, and these principles/biases enable them to map a word to its meaning even at the first exposure to the word (e.g. Gleitman, 1990; Markman & Hutchinson, 1984). For example, young children assume that a new word (noun) refers to the entirety of the referred entity rather than its part, color, texture, or material (Markman, 1989; Landau, Smith & Jones, 1988). They also assume that the word denotes a category, and hence that it should be generalized to other objects of like kind (Hall, 1991; Markman & Hutchinson, 1984). It has been also shown that children use shape similarity as a basis for determining what objects are of “like kind” and what are not (e.g. Golinkoff et al., 1995; Imai, Gentner & Uchida, 1994).

At the same time, these so-called constraint theories are not sufficient to fully explain young children's lexical development, because while word learning constraints/biases help the learning of basic-level object category names, they do not help the learning of other types of nouns such as substance names, proper names, subordinate and superordinate category names. In other words, the word-learning constraint theories say very little about how children infer meanings of words that are not basic-level object category names.