Processing information structure: A case study of Contrastive Topics in Estonian

This dissertation looks at the processing of information structure -- more specifically Contrastive Topics (CTs) -- in Estonian. Estonian is a flexible word order language meaning that information-structural relations like topicality and focus can be optionally expressed through syntactic means, by deviating from its canonical Subject-Verb-Adverb-Object word order. The syntactic encoding of information structure in Estonian raises interesting questions for language processing at the syntax-pragmatics interface. In this dissertation, I touch on the time course of contrast assignment in non-canonical clauses and whether contrast resolution (or pairing a contrastive constituent with a grammatically licensed contrastive alternative) is active and predictive in nature. Further implications for the processing of discourse-configurational languages are discussed throughout this work.In the theoretical portion of this dissertation, I lay out a working hypothesis for the syntactic structure of the left periphery of Estonian V3+ clauses (where the verb occurs in the third or later linear position), presenting grammaticality judgment data in information-structurally controlled contexts.The experimental portion of this dissertation utilizes psycholinguistic research methods -- eye tracking during reading, a speeded acceptability task with rapid serial visual presentation and a sentence completion task. The dissertation includes two experimental chapters. In Chapter 4, I use the principle of information-structural parallelism to disambiguate the information structure of V3+ clauses by following them with contrastive ellipsis. Looking at the processing of ellipsis remnants (which have previously been shown to exhibit a bias towards the most local correlate in the antecedent clause) reveals that the language processor rapidly computes contrast during the incremental processing of non-canonical V3+ structures and uses information-structural representations for processing clausal ellipsis. I also show evidence that information-structural representations may be temporarily underspecified in the absence of biasing context.In Chapter 5, I explore the processing of temporary DP Object / Clausal coordination ambiguity in Estonian, showing that information-structural representations (whether computed based on the preceding discourse context or syntactic marking) lead the processor to anticipate upcoming contrast, which leads to the language processing system's independent preference for syntactically simpler structures (a.k.a. Minimal Attachment) being overridden. In contexts conveying a CT structure, DP Object coordination is penalized compared to the more complex Clausal coordination, as the former leads to a delay in meeting the presupposition for the existence of a salient CT alternative to a previously CT-marked constituent.The experimental work presented here highlights the close connection between syntactic and information-structural processing. Syntactic marking of information structure (i.e. non-canonical word order) feeds the computation of discourse representations, and conversely, information-structural representations influence parsing decisions during incremental processing. The present work raises interesting questions for future research, particularly pertaining to the relative contribution of syntactic and discourse factors in computing information-structural relations, the effects of (implicit) prosody on information-structural processing, and the extent to which information-structural representations can remain underspecified during incremental comprehension, across constructions and across languages