Last updated 9/27/00
Comments, Questions about this page to Colin
Phillips
A number of psycholinguistic experiments have shown that conjoined clauses (e.g. 1) are easier to process than relative clauses (2-5) and that subject-gap relatives (2,3) are easier to process than object gap relatives (4,5). Several explanations have been proposed to explain these differences in terms of working memory (Caplan and Waters 1999, Gibson 1998, King and Just 1991). In the context of these explanations, we tested conjoined clauses and these two types of relative clauses in center-embedded and right branching contexts in two languages: Brazilian Portuguese (BP) and English.
One goal was to examine whether type of embedding (right x center) causes differences in processing in both languages. We also investigated whether subject-gap relatives are easier than object-gap relatives in BP, as in English. A final question was whether all types of relative clauses are more difficult to process than conjoined clauses. The difference between relative (2) and conjoined clauses (1) was reduced to the presence or absence of the conjunction "and" versus the relative pronoun "that". This differs from previous experiments (e.g. King and Just 1991) in which the relative sentences were center-embedded. Some theories of working memory (Gibson 1998) predict that sentence (1) is more difficult than sentence (2), since the distance between the gap and its antecedent (the child) in (1) is longer than the distance between the gap and its antecedent (the man) in (2). Other working memory metrics (King and Just 1991) predict that (1) is easier than (2). Thus, the comparison between these two sentences can shed light onto the factors that modulate processing difficulty in these sentences.
The methodology was RSVP (Rapid Serial Visual Presentation). Based on previous work (Forster 1971, Miyake et al. 1994) showing that word presentation duration affects performance, the present experiment presented words for 200ms/word (0ms inter-word interval). The dependent variables were grammaticality judgments and reaction time.
The performance data were the most informative. In particular, in BP, center-embedded relative clauses were more difficult to process than right-branching relatives. In English, by contrast, there was no effect of embedding. In English, object relative clauses were more difficult to parse than subject relative clauses independent of the type of embedding. BP also showed a tendency for object relative sentences to be more difficult than subject relative sentences, as in English. However, relative clauses were no more difficult than conjoined clauses. In BP, conjoined sentences were more difficult than right branching relatives, and in English the difference in performance between subject right-branching and conjoined sentences was not significant. Crucially, this suggests that not all types of relative clauses are more difficult to process than conjoined clauses. These results are compared to the working memory theories mentioned before in order to evaluate which metric accounts better for the data.
The cross-linguistic difference is explained examining the syntactic properties of both languages. We propose that the lack of difficulty in processing right branching relative clauses in BP is due to the fact that this language does not allow extraposition from relative clauses. English allows extraposition from relative clauses (6), creating an ambiguity during the processing of right branching relatives where the relative clause can modify the NP in object position or in subject position of the matrix clause, increasing the complexity of these sentences.
A working memory metric that respects grammatical principles is proposed in order to explain the differences in processing right branching versus center-embedded relative clauses and subject versus object relative clauses.
|
|
Last updated 9/27/00 |