Connectionist Account of Object-substance Distinction 1 Running Head: Connectionist Account of Object-substance Distinction a Connectionist Account of the Object-substance Distinction in Early Noun Learning

Young children learning vastly different languages generalize nouns in the same systematic way. Young children extend the name of a solid object to other objects that match the exemplar in shape irrespective of their colors, sizes or materials. However, when generalizing the name of a non-solid substance, children consider material to be more important. Thus, young children seem to know that objects and substances are fundamentally different and classified in different ways. Where does this knowledge come from? This paper presents evidence that this distinction is a product of learned correlations among perceptual properties, syntactic cues, and the lexical category structure of early learned nouns. In nine experiments we show that (1) a connectionist network trained on noun vocabularies that present the same statistical regularities characteristic of early learned nouns form a generalized distinction between solids and non-solids; (2) the networks simulate both global and finer-grained aspects of previous findings on children’s novel noun genralizations; and (3) the simulations generate new predictions that are supported by new experiments investigating children’s novel noun generalizations. Connectionist Account of Object-Substance distinction 3 A Connectionist Account of the Object-Substance distinction in Early Noun Learning An object-substance distinction has played a role in discussions by linguists, philosophers, and psychologists (Prasada, 1993; Pelletier, 1979; Hall, 1996; Quine, 1960; Lucy, 1992; Piaget, 1952; Xu, 1997). At its core, the distinction is about discrete versus continuous quantities. Conceptually, objects are discrete and bounded unitary wholes, whereas substances are continuous unbounded masses. Whether an entity is considered an object or a substance determines its grammatical class in some languages, the categories one forms, the inferences one makes, and how one reasons about amount. This paper concerns a distinction that young children make that is not identical to the object–substance distinction, but that is arguably the developmental foundation for those more mature and more abstract concepts. The distinction children make is between solid and nonsolid things, between things that hold their shape when pushed and prodded and things that do not. These two distinctions one conceptual and the other perhaps more perceptual do not map directly onto each other: there are solid objects (table) and solid substances (wood), and also non-solid objects (bubble) and non-solid substances (water). However, in some ways, solidity seems an ideal expression of objectness because solid things have stable bounded shapes, and similarly, non-solidity seems an ideal expression of substances since for non-solids material is stable and form is transient. The evidence certainly sugests that a solid-nonsolid distinction is made early. In habituation tasks, young infants expect solids and non-solids to behave differently when transformed (Spelke, Breinlinger, Macomber, & Jacobson, 1992; Gibson, Owsley, Walker, & Megaw-Nyce, 1979; J, 1979; Mash, Quinn, Dobson, & Narter, 1998). By the time children are 2 1/2-years-old, they also appear to know that solid and non-solid entities are fundamentally different kinds that are categorized by different properties. Specifically, children expect solid things to be classified by shape and non-solid things to be classified by material (Soja, Carey, & Spelke, 1991). The Connectionist Account of Object-Substance distinction 4 literature indicates that childrens expectations about solid and non-solid categories are more robust in lexical categorization tasks, more robust for solids than for non-solids, emerge during the same period that children learn names for common objects and substances and are different for children learning different languages. The evidence also suggests that this distinction may be closely linked to task and stimulus properties beyond solidity. The goal of this paper is to explain these data and the developmental processes that give rise to childrens ideas about how categories of solids and non-solids are organized, ideas that may be the starting point for a more conceptual understanding of objects and substances (Xu, 1997; Bloom, 1996, 2000).