Processing Chinese relative clauses in context

Abstract

This paper presents a self-paced reading experiment comparing the processing of subject-extracted relative clauses (SRCs) and object-extracted relative clauses (ORCs) in supportive contexts in Chinese. It is argued that lack of a consistent pattern in the literature for the comparison between Chinese SRCs and ORCs is due to potential temporary ambiguity in these constructions in null contexts. By placing the materials in contexts biased towards a relative clause (RC) interpretation, we limit the effects of temporary ambiguity. The results of the experiment demonstrate that SRCs are read more slowly than ORCs in supportive contexts. These results provide evidence for working memory-based sentence processing theories whereby processing difficulty increases for connecting sentence elements that are further apart. Some convergent evidence that strengthens these conclusions comes from recent research on aphasic populations where a dissociation between English and Chinese RC processing has been revealed: whereas English aphasic patients have more difficulty with ORCs and Chinese aphasic patients have more difficulty with SRCs (Su, Lee, & Chung, 2007). Taken together, these results support the idea that sentence processing is constrained by working memory limitations.

Keywords:

• Sentence processing
• Syntactic complexity
• Relative clause
• Working memory

Acknowledgments

We would like to thank the following people for helpful discussions of and comments on this work: Ev Fedorenko, Tomoko Ishizuka, Arthur Wang, the audience at the CUNY conference on sentence processing at Chapel Hill, NC, 2008, and two anonymous reviewers for their comments on an earlier draft of this paper. The research conducted here was supported in part by the National Science Foundation under Grant No. 0844472, “Bayesian Cue Integration in Probability-Sensitive Language Processing”. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Notes

¹Following standard conventions in the sentence processing and linguistics literature, we will indicate the grammatical position in the embedded clause which the relativiser is co-indexed with (e.g., subject or object position) as an underscore (__), indicating that it is a phonologically empty position. The existence of the empty element is not critical to the accounts that are discussed here.

²To the extent that there are complexity differences across positions, these differences may be due to differences in temporary ambiguity across positions, which may be especially relevant to languages without relative pronouns, to be discussed below.

³According to a third class of explanations of RC complexity, the difficulty in understanding an embedded clause depends on whether or not the same perspective is used in the embedded clause as in the main clause. The perspective of a clause is defined to be its subject (MacWhinney, 1982; MacWhinney & Pleh, 1988). According to this hypothesis, ORCs modifying subject NPs are more complex than SRCs modifying subject NPs, because it is necessary to switch the perspective from the subject of the main clause to the subject of the embedded clause in the case of ORCs. In contrast, no switch is needed for an SRC modifying a subject NP.

Although this hypothesis can account for some aspects of the pattern of results for RCs modifying subject NPs, it does not account for results in which RCs modify other positions in sentences. For example, the perspective theory predicts that SRCs and ORCs modifying object NPs should not differ in complexity, because both involve a similar perspective shift away from the subject of the main clause, and should both be more complex than SRCs modifying subject NPs. But this pattern of data is not observed. In fact, the complexity difference that is observed for RCs modifying subject NPs is also present for RCs modifying object NPs (Gibson et al., 2005), contrary to the predictions of the perspective account. Consequently, the perspective account has not recently been promoted as a viable alternative to experience-based and working-memory-based theories.

⁴One additional retrieval-based account of RC complexity is the proposal by O'Grady (1997), who hypothesises that retrieval difficulty is affected by the hierarchical distance between the empty element and its associated wh-element in the syntactic tree. Under this proposal, an SRC is easier to process than an ORC because the wh-element to be retrieved from memory when processing the empty position in an RC is hierarchically closer to its corresponding wh-element for SRCs than for ORCs. Although this is an interesting hypothesis, to the best of our knowledge there is no independent evidence for it in the language processing literature, other than the SRC/ORC complexity difference. Furthermore, unlike linear distance-based memory theories which were motivated by a large body of literature on memory for sequences of elements, the hierarchical distance-based proposal lacks such a grounding in the cognitive psychology literature.

⁵When a language has fewer morphological cues to the target structure as compared with another language, then there will sometimes be more temporary ambiguity in that language. Thus, languages like Japanese and Korean, which lack a relative clause marker (such as a complementiser or a relative pronoun), will sometimes contain temporarily ambiguous relative clauses which are difficult for comprehenders to resolve. Similarly, although Chinese contains a morphological relative clause marker, it lacks case-marking on its nouns. This lack of case-marking increases temporary ambiguity relative to other languages with head-final relative clauses, like Japanese and Korean.

⁶It is not clear what the syntactic category of the Chinese word “de” is. It may be a complementiser (corresponding roughly to the word “that” in English), as suggested by e.g., Cheng (1986, 1997) and Paul (2006), or it may be a more general linker of a modifier to a head, as suggested by e.g., Den Dikken (2006) and Den Dikken and Singhapreecha (2004) (see also Aoun & Li, 2003; Huang, Li, & Li, 2009). Evidence consistent with the second approach consists of the observation that the word “de” can mark modifiers other than clauses. No matter which approach turns out to be correct, the head noun for the RC needs to be interpreted as coindexed with the empty position within the RC: either the subject or the object of the RC in these examples. This is the critical integration that differs between head-first and head-final languages.

⁷Item 6 was intended to be unique-referent first item, but an error was made in constructing this item. In any case, the items are still almost equally balanced with respect to the presentation order of the unique referent. In fact, as discussed in the results section, due to a script error, item 12 was not presented to the experimental participants. Consequently, the items were as well balanced as possible with respect to the order of presentation of the unique referent.