Information structure of complex sentences: an empirical investigation into at-issueness

Abstract

This PhD project investigates the sentence-structural and information-structural features of complex sentences. Specifically, this thesis investigates the influence of sentence- structural features (clause type and clause position) on at-issue status, to then probe the effects of these sentence-structural features as well as information-structural features (new/givenness of information) on the processing of complex sentences.

At-issueness refers to the status of a clause within a sentence. Does its content express the main point of a sentence (at-issue content), or a point that is more peripheral (not- at-issue content)? A clause’s at-issue status is relevant for interlocutors’ understanding of how to build up a representation of unfolding discourse, what emphasis and attention to place on the content of a clause, and how to make predictions about what content will be picked up going forward in subsequent sentences. There are three main theories of at-issueness. These theories agree that matrix clauses can always achieve at-issue status and that temporal adverbial clauses (TACs) are not-at-issue. However, they make contrasting predictions about the at-issue status of Appositive Relative Clauses (ARCs). Q(uestion under discussion)-at-issueness, considers content at-issue if it can felicitously answer the question under discussion. Under this view, at-issue status is an immutable property of clauses resulting from their type, and ARC content can never achieve at-issue status. P(roposal)-at-issueness considers content at-issue if it proposes an update to the common ground and C(oherence)-at-issueness considers content at-issue if subsequent discourse can coherently connect to it. Both P- and C-at-issueness theories predict ARCs to have at-issue potential in sentence-final position. Under both of these views, at-issue status is a variable property that is influenced by clause type but also clause position. This thesis is an empirical investigation into C-at-issueness divided into two main parts. The first part (Chapter 2) investigates the C-at-issue status of clauses in sentences with an ARC and sentences with a TAC [experiments 1–6]. The second part (Chapter 3) investigates effects of C-at-issue status and other information-structural and sentence-structural features on sentence processing [Experiment 7].

Chapter 2 of this thesis investigates the C-at-issue status of TACs, ARCs and the matrix clauses that subordinate these.

Sentences that consist of multiple clauses are generally expected to contain one (matrix) clause that carries the main point of the utterance, the at-issue content, and other (subordinate) clauses whose content is more peripheral and therefore not-at-issue.

Under C-at-issueness, ARCs are expected to achieve at-issue status in sentence-final position, but TACs are not. The C-at-issue content then in such sentences is expected to be more likely to connect to subsequent discourse than content that is not C-at-issue. In this chapter, the results of the first six self-paced reading experiments are reported in which reading times when an ambiguous pronoun It is disambiguated to a referent in a subordinate clause or a matrix clause in varying positions were measured. We find that not only ARCs, but also TACs can be C-at-issue in sentence-final position, and even more so than the matrix clause that precedes them, suggesting clause position is the most important factor in determining C-at-issue status. In sentence-early position, however, we do observe differences between matrix clauses and TACs/ARCs: only matrix clauses in this position are C-at-issue, suggesting that clause type does play a role in sentence-early position. This study highlights the im- portance of distinguishing between different theories of at-issueness, but it also observes an interdependence between the three theories: ARCs and TACs might never achieve Q-at-issue status, but when they are C-at-issue and/or P-at-issue, they can contain the question under discussion for content which subsequently is Q-at-issue. As such, research carried out within any of the three theories could benefit from complementary inclusion of one (or both) of the other theories.

In Chapter 3 of this thesis, the processing of complex sentences with an ARC is inves- tigated [experiment 7]. If the ARC in these sentences is in early position, it does not compete for C-at-issue status with the matrix clause: only the matrix clause is expected to be C-at-issue. Consequently, only the matrix clause needs to be held in memory to make a potential connection to subsequent discourse. When the ARC is in sentence-final position, however, this ARC competes with the matrix clause preceding it for C-at-issue status. As such, both clauses need to be held in memory to create a potential discourse connection, which would lead to processing difficulty compared to the other situation in which only one clause is likely to carry the at-issue content. We refer to this as the at-issueness principle. This principle, in addition to three other ordering principles – the given-new principle (given before new information ordering facilitates processing), the clause structure principle (matrix clause before subordinate clause ordering facilitates processing) and the clause-type mapping of information principle (given information in subordinate clause and new information in matrix clause facilitates processing) – are investigated in this chapter. While these three ordering principles are well-established, it is not clear if they apply to sentences with an ARC. ARCs stand out among sub- ordinate clauses for their matrix clause-like characteristics from both a syntactic and an information-structural perspective. While we find no evidence for the at-issueness principle, we replicate previous studies in finding evidence for predictions made by the given-new principle and the clause structure principle in sentences containing an ARC. In addition we find indirect evidence for the special status of ARCs through the observed behaviour of matrix clauses in sentence-early position: These seem to have a ground- ing function here: the matrix clause provides the context which is necessary to support understanding of the ARC. This grounding function is something which has previously been attributed to subordinate clauses.

This project sheds more light on the C-at-issue status of clauses in sentences with a TAC and in sentences with an ARC through a series of self-paced reading experiments. Results suggests that C-at-issue status is more flexible than has previously been found, and is distinct from both Q- and P-at-issue status. However, it also observes an in- terdependence between the three theories, suggesting that they should be investigated in a complementary fashion. While the study in Chapter 3 does not reveal C-at-issue potential to affect the processing of sentences with an ARC, it highlights the relevance of established ordering principles. When the at-issueness principle was investigated in tandem with these, effects of C-at-issue status might have been obscured through greater effects of clause order and information order. This leaves open questions about how and if at-issue status can be observed through the processing of at-issue content, and by extension, if a relation between at-issue status and processing time can even be assumed.