When participant structure and argument structure do not match: Participant structure construction in adults and prelinguistic infants

On one theory, verb-learning complies with Participant-Argument-Match (PAM): each participant in a verb’s event representation corresponds to an argument NP in its clause. Yet sometimes it seems that a participant corresponds to no argument NP: Anne robbed Betty has no argument for what was stolen, and Anne jimmied the box has no argument for the implied lever. Are such cases problematic for PAM? To answer this, we need to understand what participant structure learners likely construct for events, independently of language. We assess this with experiments examining the perception of meaningful differences in participant structure in 10-month-olds (using a dishabituation paradigm) and adults (measuring reaction time in a similarity judgment task).

Exps.1-2 target events described by verbs whose natural usage shows participant-argument matching: give (three participants/three arguments) and hug (two participants/two arguments). Exp.1 replicates Gordon (2003)’s finding that infants distinguish give-type scenes depending on whether a teddy-bear is given or not, but do not distinguish hug-type scenes based on the teddy’s participation, consistent with PAM. Exp.2 demonstrates adults make the same distinction.

Exps.3-6 target events described by verbs whose natural usage shows participant-argument mismatching: jimmy (three participants/two arguments) and rob (three participants/two arguments). Exp.3 shows that adults treat a lever as a participant in open-type events only when it is used to facilitate the opening. Exp.4 shows that adults view an individual as a participant in take-type events when the toy is taken from them as opposed to near them (take-pickup), and distinguish if they are active or resistant (take-rob). Trends in ongoing data collection for Exps.5-6 suggest infants parallel adults.

If Exps.3-4 predict how infants construe similar scenes in Exps.5-6, this poses a challenge for PAM: represented event/participant structures do not always align one-to-one with the argument structure of their most natural linguistic description.