The Origin of Argument Structure in Infant Event Representations

When deaf children of hearing parents are not exposed to a language model, they invent their own communication systems using gestures (Goldin-Meadow, 1985). Such " home sign " systems have been shown to be rich in expressive power, and to exhibit many parallels to conventional languages including a primitive form of argument structure. When gesturing actions such as giving, home signers demonstrate that they are sensitive to the fact that giving entails three arguments: A giver, a recipient, and an object being given. This phenomenon suggests that the human conceptual system constructs event representations that possess argument-like structures independent of overt linguistic input. Understanding the nature of this development addresses fundamental issues of the relation between language and thought and their origins. In a set of pioneering studies of the development of linguistic concepts in infancy, Golinkoff (1975) and Golinkoff & Kerr (1978) examined whether 14 to 24 month olds had general concepts of agent and patient, the fundamental elements of transitive structures in language. In studies employing looking time and heart-rate deceleration, they familiarized infants with filmed actions involving people pushing other people or pushing tables and chairs. On test trials, infants saw a new film in which there were reversals of roles and/or positions on the screen. There were effects of role switching in this paradigm although it is difficult to determine whether this implied the existence of semantic roles in infant representations. These and other studies of infant linguistic knowledge (e.g., Golinkoff, Hirsch-Pasek & Gordon, 1987), suggest comprehension of linguistic structure in infants who are actually using language in some way, beginning at about 14 months of age. Questions about the conceptual origins of linguistic abilities, however, require that we look to a period before language is being learned in earnest prior to the end of the first year. To study the possible non-linguistic origins of verb-argument structure, we need a task that provides a reasonable non-linguistic conceptual analogue in the encoding of event structure. Any attempt to provide a one-to-one mapping between prelinguistic event structure and verb-argument structure is surely doomed to failure because of the fact that there is cross-linguistic variation in how any particular verb is expressed in terms of argument structure. What might be an argument in one language might be an adjunct in another. In addition,