On the impact of syntactic and semantic information when processing words in sentences

We report the results of two experiments that used the Rapid Parallel Visual Presentation (RPVP) procedure to evaluate the contributions of syntactic and semantic information when identifying words within a sentence context. Participants were presented with five-word sequences that could either be grammatically correct (e.g., <i>"the cat ate fresh food"</i>) or an ungrammatical re-ordering of the same words (e.g., <i>"ate cat food the fresh"</i>). Word sequences were displayed for 200 ms, after which participants were asked to classify a word at a post-cued location - either Position 2, 3, or 4, in the 5-word sequence. In Experiment 1, the classification was based on a syntactic category (is this word a noun or an adjective?). In Experiment 2, the classification was based on retrieval of the object associated with the word (is the object referred to by this word bigger or smaller than 1 m?). Analyses tested for effects of grammaticality on word classification accuracy (i.e., a sentence superiority effect - SSE) per target word position in both experiments. For syntactic classifications (Experiment 1), the SSE was significant at all positions. On the other hand, for semantic classifications, the SSE was only significant at Position 4. We conclude that in the RPVP paradigm readers use both syntactic and semantic information to rapidly build a primitive sentence structure, but with semantic constraints having a later influence. This structure then helps them to associate more detailed syntactic and semantic information to word identities at specific locations within the sentence.