Linguistic cues enhance the learning of perceptual cues

Abstract

Linguistic Cues Enhance the Learning of Perceptual Cues Hanako Yoshida (hayoshid@indiana.edu) Linda B. Smith (smith4@indiana.edu) Psychology Department, Indiana University, 1101 E. 10 th Street Bloomington, IN 47405-7007 USA Abstract Past research on children’s categorizations has centered on the mechanism of children’s use of multiple cues in categorization. This paper examines correlations between perceptual cues and linguistic cues. The question asked is a classic one in learning theory: given two redundant cues, does the learner learn more about each than when one cue independently predicts the category? This question has special cogency in the context of children’s language learning. We show that when linguistic cues correlated with perceptual cues, children learn more about perceptual cues. Introduction Two- and 3-year-old children learn new object names rapidly, often correctly determining the range of instances to which the name applies from just one experience hearing the word used in a single context. Children do this by exploiting multiple cues to meaning. Past research indicates they use both linguistic cues and perceptual cues to figure out the likely meaning of a novel noun. Much of the relevant evidence in this literature concerns the count-mass distinction in English. Count nouns refer to entities conceptualized as discrete objects and as countable. Count nouns obligatorily take the plural (e.g., cups, hopes). Mass nouns refer to entities conceptualized as continuous substances and do not take the plural, but rather mass quantifiers (e.g., some water, a lot of sand). Children use linguistic cues to the count/mass status of a noun to figure out the category to which a novel noun refers. For example, if an entity named with a novel name is presented in a frame that indicates it is a count noun (e.g., “This is a mel”), English-speaking children interpret the word as referring to a discrete entity and typically extend the object name to a class of similarly shaped things (Soja, Carey & Spelke, 1991; Soja, 1992; Landau, Smith and Jones, 1988; Landau, Smith and Jones, 1998; Imai & Gentner, 1997). When the same noun is presented in a frame indicating it is a mass noun (e.g., “This is some mel”), English-speaking children interpret the word as referring to a substance and extend its meaning to entities of the same material (Soja, Carey & Spelke, 1991; Soja, 1992). Children also use perceptual cues. For example, children extend novel names to new instances by shape when the named entity is solid and rigid (e.g., made from wood) but extend the name to new instances by material when the named entity is nonsolid and non- rigidly shaped. Much previous research has explored which of these kinds of cues dominate by putting them in conflict. In this paper, we ask whether and how they might interact and support children’s learning of object names. This is a relevant question for two reasons. First, linguistic and perceptual cues are highly correlated. This was documented by Samuelson and Smith (1999) who studied the structure of the first 300 nouns commonly learned by English-speaking children. Among these 300 names for common categories, solid things tend to be named by count nouns that refer to things of the same shape, whereas nonsolid things tend to be named by mass nouns that refer to entities of the same material. For learners of English, then, there is a tight correlation between linguistic cues associated with count/mass distinction and perceptual cues that indicates the solidity or non-solidity of an entity. Second, the evidence suggests that children learn the correlations among perceptual cues, linguistic cues, and category structure as they learn names for common object and substance categories. Specifically, the influence of perceptual cues on children’s noun extensions emerges and grows stronger as vocabulary grows. Samuelson and Smith’s (1999) data indicate that children learning English do not extend names for solid and nonsolid things differently until children have over 150 nouns. Similarly, English-speaking children’s sensitivity to count/mass syntax in the novel noun extension task emerges during this same time period (Soja, 1992). Two hypotheses What is the relation between learning about perceptual cues to category organization and linguistic cues to category organization? One possibility is that they are completely independent. Cross-linguistic comparisons of English and Japanese speakers are consistent with this view. Japanese differs from English in that it has no obligatory plural and no counterpart to the count-mass distinction in English. Yet, several studies suggest that