In this study, we investigated whether the parsing was performed in an incremental manner or a delyaed manner. To address this issue, we examined two types of sentences which are identical except their main-clause predicates. The Full Attachment Model (an incremental model) and the Head Driven Model (a delay model) predict that reanalysis will occur in one of two different types of sentences respectively. Event-Related brain Potentials (ERPs) were recorded while subjects read these sentences. The results revealed that the ERPs elicited in one type of sentence were significantly more positive-going than those in the other. This ERP effect was comparable to the P600 which was assumed to be the electrophysiological marker of reanalysis. The result was compatible with the prediciton of the Full Attachment Model, but contradicts the prediction of the Head Driven Model. We, therefore, concluded that the parser pursued syntactic analyses not in a delayed manner but rather in an incremental manner.
Recently, many researchers have reported about how children acquire a word meaning. Especially some of them reported that the verb meaning development is delayed to that of noun. In these reports, they used the ANOVA (the ANalysis Of VAriance) for the detection of a difference in the behavioral data. But ANOVA is not a sufficient way of analysis in the sense that we can't know the detailed mechanism of the meaning acquisition in children. So, in this paper, we developed a model based analysis of behavioral data that enable more detailed structure estimation, and analyzed the children's word acquisition data with the method. From the analysis, we found a hierarchical mixture of binomial distribution with three sub-modules is a suitable model for the data. The parameter change in the model indicated that what is changing between three to five years infant is a choice of proper learning action from the recognition of verb/noun situation.