Tid: Fredag 15 januari 2016, kl. 13:00-15:00
Plats: C389, Södra huset, Frescati

Skuggopponent är docent Mikael Roll vid Språk-och litteraturcentrum, Lunds universitet.

Abstract

This thesis investigates the comprehension of grammatical relations of transitive sentences (i.e., subject and object) from a functional and usage-based perspective. It is based upon and aims to test the following assumptions and claims about grammatical relations and the process of comprehending grammatical relations:

  • Grammatical relations express role-semantic (e.g., Actor and Undergoer) and discourse-related (e.g., topic and focus) functions that are encoded on the basis of a systematic interplay between morphosyntactic (e.g., word order and case) and prominence based (e.g., animacy, volitionality and definitetness) information.
  • The comprehension of grammatical relations involves the assignment of functions to argument NPs. This is a highly incremental process that draws upon both morphosyntactic and prominence based information.  
  • These information types therefore function as cues during grammatical function assignment. The weighting, interplay and availability of these cues vary across languages but do so in systematic ways that are reflected in their distribution in language use.
  • The weighting and interplay between morphosyntactic and prominence based cues in a given language can therefore be quantified on the basis of corpus data.
  • The distribution of these cues in corpora can further be used to derive predictions regarding the process of incremental grammatical function assignment, and these predictions can be empirically tested using psycholinguistic experiments. The aim of this thesis is to do that.

The thesis consists of the following 7 chapters:

Chapter 1 is an introduction to the thesis that provides an overview of the fundamental ideas and concepts that it is based upon.

Chapter 2 provides the theoretical background to the thesis. It presents the theories and previous empirical findings that motivate the underlying assumptions of the studies in the thesis. It gives an overview of grammatical relations in the typological perspective, arguing that grammatical relations are best conceived as functional (as opposed to syntactic) categories whose encoding is conditioned on an interplay between prominence based and morphosyntactic information types. Chapter 1 then gives an overview of some of the studies that have investigated the distribution of grammatical relations with respect to prominence based features in corpora, showing that distributions of prominence features within particular languages correspond to the distributions found across languages. Chapter 1 finally gives an overview of some psycho- and neurolinguistic theories of grammatical function assignment that assumes that this process is guided by the availability of prominence features as well as their interplay with morphosyntactic features. It also gives an overview of studies that have found evidence for the view that prominence based information is of importance during grammatical function assignment.

Chapter 3 presents an ERP-experiment that empirically tests the claim that grammatical relations should be conceived as functional categories rather than syntactic. This is done on the basis of investigating the ERP-correlate to grammatical function reanalysis, the process by which an initial assumption regarding the grammatical function of an NP argument is revised. More specifically, the experiment finds that grammatical function reanalysis in Swedish engenders a N400 effect rather than a P600 effect, indicating that grammatical function reanalysis involves a remapping of thematic roles to the NP arguments rather than a revision of the syntactic structure of the sentence.

Chapter 4 presents a corpus study of the distribution of prominence based, verb semantic and morphosyntactic features in transitive sentences in written Swedish texts. The aim of this study is two-fold. Firstly, it investigates the functional motivations for an object-before-subject word order in Swedish transitive sentences, with respect to earlier accounts of the object-before-subject word order in Swedish. Secondly, and most importantly, it quantifies the strength of prominence based and verb semantic cues and their interactions in terms of their ability to predict the sentence word order. This is done using logistic mixed effects modeling. This study shows that Animacy by far is the strongest prominence cue, and, further that this is particularly the case in sentences with volitional or causative verbs, which in many cases require an animate subject argument. The study also shows that the case marking / person of the initial NP is of great importance in sentences with experiencer verbs.

Chapter 5 presents three statistical models of incremental grammatical function assignment. These models aim to estimate processing difficulties during grammatical function assignment by modeling the on-line change in the expectation of a object-before-subject word order as a function of the subsequent presentation of the sentence constituents (i.e., NP1, the verb, and NP2) over time. This is done in terms of the surprisal of encountering the features of the current constituent, given the features of the previous constituents. As such, the models quantify the change in the expectation of an object-before-subject word order at constituent Ci with respect to constituent Ci-1 as the relative entropy between p(Object-before-subject | Ci) and p(Object-before-subject | Ci-1). The fundamental characteristics of the three models are the same, but they differ with respect to how the conditional probabilites for the Object-before-subject word order are estimated and integrated. All three models make rather similar predictions regarding the processing difficulties in locally ambiguous transitive sentences which are disambiguated towards the object-before-subject word order at the position of the final NP (such as, e.g., “Killen sparkar jag på smalbenet”). In particular, the models predict a high surprisal effect at the position of the disambiguating NP of such sentences, and further, that this surprisal effect can be highly mitigated when the features of the initial NP and the verb speaks in favor of the object-before-subject word order. This is in particular the case when the initial NP is inanimate and the verb is volitional or causative, or when the initial NP is allophoric and the verb is an experiencer verb.

Chapter 6 presents a self-paced reading experiment that sets out to test the most prominent predictions of the incremental models. More specifically, the experiment tests whether sentence processing is impaired, as reflected in slower reading times, at the position of the disambiguating NP in locally ambiguous object-initial sentences. It also tests whether such a processing difficulty can be mitigated when the animacy and the verb class speaks in favor of the object-before-subject word order, as predicted by the model. The results of the experiment are at large in line with the predictions of the incremental models, but indicate that the models might underestimate the effect of egophoricity in sentences with experiencer verbs.

Chapter 7 finally concludes the thesis by summarizing the results of the individual studies and relating them to theoretical assumptions presented in Chapter 1 and 2. It is concluded that the results of the ERP-experiment speaks in favor of the assumption that grammatical relations are best conceived as functional categories rather than syntactic. It is also concluded that the statistical models presented in chapter 5 together with the self-paced reading experiment presented in chapter 6 provides evidence for the assumption that the process of grammatical function assignment is based upon statistical regularities in the distribution of morphosyntactic and prominence based information in language use. Processing difficulties during the comprehension of Swedish transitive sentences can therefore be predicted on the basis of the distributions of morphosyntactic and prominence based information in corpora.

Hjärtligt välkomna

Maria Koptejvskaja-Tamm