The present study addresses the question of whether structural analyses of verb-arguments are postponed up until the head verb has been processed (head-driven parsing accounts) or initiated prior to the appearance of the verb (pre-head attachment accounts). To explore this question in relation to a head-final language, a Japanese dative argument attachment ambiguity was examined in both a questionnaire study (Experiment 1) and a self-paced reading test (Experiment 2). The data suggested that the dative argument attachment ambiguity is resolved in the manner predicted by pre-head attachment accounts. The results were incompatible with most variants of the head-driven parsing model, and were not of the form currently predicted by constraint-satisfaction models. We end by discussing the general theoretical implications of the findings.