A matter of time: novel nouns mark object categories when delays are imposed

In previous work, preschool-aged children have revealed a specific expectation that novel count nouns (but not adjectives) will refer to object categories (but not to object properties or to thematic relations). However, in these tasks, children have been permitted to extend a novel word immediately after it has been introduced. The current experiments test the hypothesis that the noun-category bias is sufficiently robust to hold up even when a delay is imposed between a naming episode and the child's opportunity for extension. To capture this phenomenon experimentally, we imposed delays of 30 s (Experiment 1) and 1 h (Experiment 2). Across both delay conditions, children in a `novel noun' condition revealed a strong inclination to consistently choose other members of the same object category as the (previously named) target. Children in either a `novel adjective' or `no word' condition revealed no taxonomic inclination. These results establish the power of the noun-category bias in situations that more closely approximate some of the very real challenges that children face in mapping and extending novel words. Children's natural word learning abilities have been the focus of considerable research interest. The intensity of interest stems, at least in part, from our fascination with children's rapid and successful solutions to the task of mapping words to their meaning. Children's mastery of this complex task certainly involves an intricate interplay between expectations or biases held by the child and support from the environment. In this paper, we focus primarily on the child's expectations in extending novel words, applied ostensively to discrete objects. A review of the literature reveals a sharp distinction between children's expectations for novel words presented as count nouns versus those presented as members of other syntactic categories. For example, preschoolers expect that a novel count noun (e.g. flamingo), applied ostensively to an individual object (e.g. a flamingo), will refer to the whole named object and can be extended to include other members of the same basic or superordinate level object category (e.g. other flamingos) (Markman & Hutchinson, 1984; Waxman & Gelman, 1986; Waxman, 1990, 1991; Waxman & Kosowski, 1990; D'Entremont & Dunham, 1992; Waxman & Markow, 1995). This expectation has been dubbed the nouncategory bias because by 2 or 3 years of age it appears to be specific to count nouns. Although words from other syntactic categories (e.g. proper nouns, adjectives) can also be applied sensibly to individual objects, children's expectations of novel words in these syntactic contexts are quite different. They interpret a novel adjective as referring to a property of the named individual or to a salient subordinate level distinction within an existing basic level kind (Taylor & Gelman, 1988; Waxman, 1990; Smith, Jones & Landau, 1992; Hall, Waxman & Hurwitz, 1993; Waxman & Markow, 1998); they interpret a novel proper noun as referring specifically to the named individual, and do not extend the term to include other members of its kind (Katz, Baker & Macnamara, 1974; Gelman & Taylor, 1984; Hall, 1994, 1996). # Blackwell Publishers Ltd. 1999, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA. Address for correspondence: Sandra R. Waxman, Department of Psychology, Northwestern University, 2029 Sheridan Road, Evanston, IL 60208-2710, USA; e-mail: [email protected] We describe the phenomenon as the noun-category bias rather than the taxonomic bias (e.g. Markman, 1994) to highlight the fact that the tendency to focus on categories of objects is evident specifically with novel words presented as count nouns. Results like these demonstrate that children as young as 2 or 3 years of age expect that particular linguistic forms (e.g. count nouns, proper nouns, adjectives) will refer to particular types of conceptual relations among objects and events in the world (e.g. categories of objects, individual objects, and properties of objects, respectively) (Brown, 1957; Waxman, 1990, 1994; Waxman & Kosowski, 1990; Hall, 1993, 1994, 1996; Hall, Waxman & Hurwitz, 1993; Waxman, Stote & Philippe, 1997b). Moreover, there is now evidence for the nouncategory bias in infants on the brink of producing language (Waxman & Markow, 1995; Balaban & Waxman, 1996, 1997) and in children acquiring languages other than English as their native language (Waxman, Senghas & Benveniste, 1997a). Thus, the noun-category bias is evident across languages and is available from infancy. This is consistent with the view that children are biased in their interpretation of novel nouns, and that this bias facilitates the rapid acquisition of both the lexicon and conceptual systems of organization (cf. Markman, 1994; Waxman, 1994). In this paper, we take as our point of departure the following observation: under naturally occurring circumstances, there are often delays and distractions between the introduction of a novel word at one moment (`Look at the flamingo!') and the child's opportunity to extend it. This delay can range from minutes to hours, and perhaps longer. If the nouncategory bias is to facilitate acquisition in these more naturalistic contexts, then it must be sufficiently robust to hold up even when delays are imposed. Unfortunately, in virtually all of the research assessing the nouncategory bias, children have been instructed to extend a novel word immediately after it has been introduced. We redress this limitation by adapting an established paradigm in which children `read' through a picture book with an experimenter (Waxman & Kosowski, 1990; Waxman et al., 1997a). Each page included five pictures: a target object (e.g. cow), two objects from the same superordinate category as the target (e.g. fox, zebra) and two objects that are thematically related to the target (e.g. barn, milk carton). For children in the `novel noun' condition, the experimenter pointed to the target and said, for example, `See this ``fopin''? Can you find another ``fopin''?' In the `novel adjective' condition, she said, for example, `See this ``fopish'' one? Can you find another one that is ``fopish''?' In the `no word' condition, she said `See this? Can you find another one?' The child and experimenter `read' through the book twice. On the second reading, the experimenter reminded the children of their first choices and asked them to select another from the remaining (three) alternatives. Notice that this design incorporates two distinct types of control for the `novel noun' condition. Performance in the `no word' control reveals whether children are more likely to direct their attention to object categories in a word-learning task than in a non-linguistic control task. Performance in the `novel adjective' condition tests whether the phenomenon is specific to novel words presented as count nouns. In the current experiments, we modified this paradigm in two ways. First, to discover the power of the nouncategory bias under circumstances that approximate some challenges of natural word learning, we imposed delays between the experimenter's presentation of the novel name and the child's opportunity for extension. We predicted that, even in the face of such delays, novel nouns (but not adjectives) would direct children's attention to the commonalities between a named object and other category members, but that children hearing novel adjectives or no novel words would fail to reveal this preference. We predicted that providing a novel count noun for an individual object at one moment would highlight category membership and would facilitate identification of other category members at a later point in time. Second, we tinted the stimuli (which had been presented as black-and-white line-drawings in previous research) as follows. The thematic alternatives were lightly shaded with color to match the target; the taxonomic alternatives were shaded using colors that differed from the target. We predicted that children in the `novel adjective' condition would prefer the colormatched (thematic) alternatives (Gelman & Taylor, 1984; Prasada, 1992; Waxman, 1995; Hall & Moore, 1997; Waxman et al., 1997b; Waxman & Markow, 1998). We suspected that children in the `no word' condition would reveal this same preference, for these alternatives are related to the targets in two distinct ways (color and thematic relatedness). This modification provided a strong test of the noun-category bias, for if children in the `novel noun' condition were to select the taxonomic alternatives, they would have to do so to the exclusion of these color-based thematic alternatives.