Parallelism and prosody in the processing of ellipsis sentences

This thesis investigates the processing of ellipsis sentences, focusing on the following questions: (i) are ellipsis sentences processed using special routines employed only for ellipsis or are they processed using the same principles needed for unelided sentences? (ii) does parallelism influence sentence processing? if so, what kinds of similarities matter? The interpretation of ambiguous gapping sentences (e.g., Janie asked my dad about careers and Sharon about politics) is explored first, finding that lexical and prosodic similarities between the DP remnant (Sharon) and either DP in the first clause raise the rate of analyses placing Sharon in a syntactic position corresponding to that of the most similar DP, supporting (1). (1) DP Parallelism Hypothesis. The processor favors analyses in which DPs that share internal properties (have similar syntactic, prosodic, and semantic features) share external properties (appear in similar structural positions within their respective clauses or phrases), and vice versa. The availability of a smaller syntactic structure for the object interpretation of Sharon, however, leads to an overall bias towards that analysis. These results show that parallelism between DPs is indeed favored by the processor, but it modulates the general preference for minimal structure (e.g., Frazier, 1978, 1987). Further experiments explore whether parallelism is only effective in structures containing and, or whether it has a broader domain of application. Experiments on comparative, stripping, and replacive ellipsis sentences show that (1) applies generally in a range of ellipsis types. The relationship between focus and prosodic parallelism is explored to investigate whether prosodic similarity of elements in the elided and antecedent clauses is due entirely to their focus structure. An experiment manipulating parallelism of pitch range shows that prosodic properties unrelated to focus can also affect processing. Finally, a production experiment finds that prosodic renditions of ellipsis sentences can be quite similar to those of full conjoined sentences. The overall conclusion is that parallelism between DPs can affect the processing of a range of ellipsis structures, as well as unelided structures (e.g., Black, Coltheart, & Byng, 1985; Frazier, Munn, & Clifton, 2000; Henstra, 1996), but that there is no need for construction-specific mechanisms in processing theory