Incremental Processing in Japanese Experimental evidence from as early as the 1970s indicates that English readers process sentences incrementally, without delay, as each incoming word is immediately interpreted in relation to the previous words read (Marslen-Wilson, 1975). There is also mounting evidence that the syntactic representation of sentences is built incrementally in head-final languages, and in Japanese in particular (Miyamoto, 2006, for a summary). However, it is still not clear how exactly case-marked NPs can be semantically associated before a predicate is read. In order to clarify some of the factors involved, I will report data on the effects of internominal relations (e.g., possessor-possessee relations) on the interpretation of intranominal properties such as animacy (Iguchi and Miyamoto, 2006). For example, nominative-marked NPs are in general more likely to be interpreted as subjects if they are animate compared to when they are inanimate; however, the reverse is true if the NP is preceded by a possessor NP. I will conclude by arguing that the importance of determining how
incremental processing takes place should not be underestimated.
Proposals that ascribe processing complexity based exclusively on the
representation that linguists assume for the entire sentence often
fail to take into consideration that this representation is just the
final product and it does not necessarily reflect the difficulty
involved in the parsing process (see Fodor, Bever and Garrett, 1974,
on an extensive critique of one such a proposal --- the derivational
theory of complexity). In particular, the final representation of a
sentence does not reflect the ambiguities and temporary
representations that are typical of incremental processing. To
illustrate this point, I will report data comparing the incremental
processing of OSV sentences to SOV sentences (Miyamoto et al., 2005;
Miyamoto, 2006). Unsurprisingly, the OSV order is harder to process;
however, the difficulty cannot be traced back to one single factor in
the final syntactic representation as slow reading times are observed
at various points in the OSV order. I will suggest that these
slowdowns only make sense within incremental models of sentence
processing. |