ABSTRACT

The seemingly homogeneous class of one-participant verbs, i.e., verbs with a single argument in subject position, splits up into two distinct subclasses when one considers their syntactic behavior more carefully. 1 This is the phenomenon of unaccusativity. Unaccusativity poses an intriguing acquisition problem: how can the child find out that there are two different subclasses when they look so similar at the surface? Moreover, how does she determine to which subclass a new one-participant verb belongs? After defining what unaccusativity involves in the chess model, I will discuss these acquisition questions. Approaching them from two opposite development perspectives, the Strong Continuity hypothesis and the Maturation hypothesis, I will outline two scenarios for learning unaccusativity. Different predictions follow from these two scenarios. They have been tested in an experiment that aimed at finding out whether or not two semantic factors, felicity and agentivity, determine unaccusativity in Dutch. The results from two groups of children (4 and 5 year-olds, 7 and 8 year-olds) and a group of adults show that subjects split up novel intransitive verbs in two subclasses. They do so from early on: even the youngest children make a distinction. This is evidence in favor of the Strong Continuity hypothesis. Furthermore, the split is defined by telicity alone; agentivity does not play a role. This result suggests that mapping must be defined on the event-semantic properties of verbs, as in the CHESS model; it does not support mapping theories defined on the semantic properties of the arguments (for example, in terms of θ-roles).