| dc.description.abstract | Sentence processing has been increasingly considered as a strongly incremental and grammatically guided process in which the comprehender builds a syntactic structure and computes semantic meanings as the sentence unfolds. However, strong incrementality is challenged by linguistic phenomena such as head-final constructions where unambiguous markers of a structure only appear at the end of the structure in the bottom-up input. In these cases, the parser has no indication in the bottom-up input confirming that a particular structure is present and thus might initially commit to an incorrect structural analysis, only having to reanalyze when this information becomes available in the bottom-up input and experiencing a processing disruption known as a garden-path effect. Alternatively, it is possible that particularly in contexts which would appear to engender widespread garden-paths such as head-final constructions, the parser does not pursue incremental processing, instead delaying structural commitments until unambiguous evidence for a phrase (such as its head or other markers) becomes available. This dissertation thus investigates the processing of a head-final structure, Mandarin relative clauses, examining whether strongly incremental processing may indeed be possible within these structures, made possible by engaging in structural prediction using cues which appear early in the sentence and may allow the parser to generate an expectation that a relative clause is present prior to encountering unambiguous bottom-up information marking the relative clause. The previous literature examining whether local linguistic cues before the relative clause marker might facilitate predicting these structures has largely focused on testing one particular cue (classifier-noun mismatch) and has shown mixed findings regarding whether this cue enables the parser to predict relative clause structures in Mandarin, possibly due to the flexibility of classifier-noun relations in Mandarin. This dissertation thus examines whether a new and potentially stronger cue, temporal mismatch, would be an effective cue for relative clause prediction. The present study uses a manipulation where Mandarin relative clauses are preceded by initial mismatch between temporal expressions (such as “tomorrow … used to”), such that the parser might posit a relative clause structure before encountering the Mandarin relative clause marker de, since the only way to resolve that temporal mismatch in Mandarin is by positing a relative clause downstream. This dissertation utilizes the event-related potentials technique in a large-scale study (N=74 participants) to track brain responses for detecting the mismatch cue and generating structural prediction with millisecond-level timing accuracy during the dynamics of moment-by-moment sentence processing. In doing so, this dissertation also addresses two other open questions: (1) whether processing of temporal mismatch itself depends on the type of temporal markers involved, by including two kinds of temporal markers, a temporal adverb (cengjing “used to”) and an aspect marker (-guo), in Mandarin; and (2) the extent to which detecting temporal mismatch and engaging in prediction varies at the level of individuals, by independently assessing participants’ linguistic and non-linguistic cognitive abilities and examining the extent to which they modulate brain responses during the processing of sentences with temporal mismatches and relative clauses. Results show that temporal mismatch overall facilitates the prediction of relative clause structures, suggesting that the parser is indeed able to utilize predictive cues to facilitate predicting head-final structures. In addition, processing at the temporal mismatch itself differed based on the kind of temporal marker involved. While the aspect marker –guo yielded P600 across all participants without significant modulation by individual differences in linguistic or non-linguistic cognitive abilities, the processing of the temporal adverb cengjing was strongly modulated by individuals’ language abilities as assessed by a vocabulary measure. Overall, this dissertation presents strong evidence for syntactic prediction, demonstrating the parser’s ability to utilize implicit linguistic cues to engage in structural prediction and achieve strong incrementality. | |
