Comprehending Mandarin relative clauses: Ambiguity, locality and expectation

This dissertation examines sentence comprehension in Mandarin with the goal of contributing to the development of a clearer picture of how language processing works across languages.;The first goal was to find out which type of relative clause is more difficult for Mandarin speakers to process. In English, object relatives (where the modified head noun plays the object role in the relative clause) are more difficult to understand than subject relatives. There is some evidence from head-final languages like Japanese and Korean that the same is true for them. However, studies investigating Mandarin relative clause processing have produced mixed results so far. The studies reported here uniformly (with one interesting exception) show object relatives to be easier to understand in Mandarin, which replicates some and contradicts others of the previous studies of Mandarin. This pattern of results also rules out some of the theoretical explanations that have been developed to account for English and provides support for others, specifically those that invoke the relative frequency of occurrence of different word order patterns in a language (MacDonald & Christiansen, 2002) as well as those based on memory and integration costs when dependencies between words span long distances in sentences, such as Dependency Locality Theory (Gibson, 1998, 2000). Dependency theory was further tested in a study (Experiment 2) comparing gapped and gapless relative clauses. Relative clauses are usually posited to contain a gap and to require integration of a filler with that gap, which is an example of a long-distance dependency. Mandarin has a structure that looks like a subject relative clause but that does not contain a gap and is thus sometimes called a gapless relative. Gapless relatives were found to be easier to process than true subject relatives containing gaps, which can be explained in terms of the processing cost of having to integrate the gap and its filler, providing further support for dependency theories.;The second goal of these experiments was to find out what kinds of cues help Mandarin speakers disambiguate sentences that are temporarily ambiguous. Previous studies have found that Mandarin speakers make rapid use of animacy cues (Lin & Garnsey, 2010), and the studies reported here show that they also use verb-based structural biases (Experiment 1) and classifier- based semantic restrictions (Experiments 3 and 4) to help resolve temporary ambiguity. Relative clauses were also embedded in constructions that alter the default SVO order such as the BA (object focusing) and BEI (passive) constructions to investigate how word order changes would influence relative clause processing (Experiments 4 and 5). The final experiment (Experiment 5) investigated the role of locality in Mandarin sentence processing. Long-distance dependencies between words in a sentence usually become more difficult when the distance increases. This is called a locality effect and is generally explained in terms of increasing memory load with distance. However, a few previous studies of Hindi, German, and English have found the opposite, i.e. that the difficulty decreases with increasing distance. This has been called an anti-locality effect and has been explained in terms of increasing anticipation of the upcoming element. The head-final property of Mandarin relative clauses provides another opportunity to examine locality. In the previous studies finding anti-locality effects, the anticipated upcoming word was a verb, while in most studies showing locality effects, it was instead a noun. In order to test whether it is the class of the predicted word (and the differences in the types of dependencies that go along with that) that explains why there are sometimes locality and sometimes anti- locality effects. Head-final relative clauses were used to make a noun be the anticipated word and the verb-final BA construction was used to make a verb be the anticipated word. A locality effect (reading times increased as distance increased) was observed when the anticipated word was a noun but an anti-locality effect (reading times decreased as distance increased) was observed when it was a verb, suggesting that it is the type of dependency and the nature of the anticipated information that determine the direction of locality effects in sentence comprehension. (Abstract shortened by UMI.)...