Listening between the lines: the roles of social reasoning and disfluency in the comprehension of ambiguous utterances

Abstract

Everyday speech is rarely fluent and often contains disfluencies, such as filled pauses uh and um. Disfluency is typically related to a speaker’s production difficulties and it may therefore be regarded as an impediment to effective communication. However, disfluency can actually be a source of information, whereby a speaker conveys non-literal information, or paralinguistic cues (e.g., regarding the speaker’s certainty or veracity), that can facilitate listeners’ language comprehension in a given context. But exactly how does a speaker’s manner of delivery exert its pragmatic effects on listeners’ language processing? This thesis takes a perspective to the study of disfluency by focusing on the influence of the speaker’s manner of delivery on listeners’ pragmatic comprehension, specifically the influence of context and disfluency on listeners’ real-time interpretations of ambiguous statements. Four mouse-tracking experiments and one eye-tracking experiments are reported here, reflecting an innovative attempt at investigating the pragmatic roles of disfluencies on online language comprehension.

The experiments focus on the effects of filled-pause disfluencies like uh on listeners’ interpretations of utterances. The mouse-tracking studies aimed to test how this type of disfluency influences listeners’ interpretations of the ambiguous scalar quantifier some. The interpretation of some is known to depend on context, which makes this ambiguous word a good test case for exploring how pragmatic cues informed by disfluencies shape listeners’ understanding. For example, how do listeners interpret statements like “I got, uh, some ‘A’s in my Psychology courses” in an interview context, where a speaker might be trying to present their few number of ‘A’s more positively, or deceptively?

The results – both in participants’ final interpretations as well as in their online mouse trajectories – show that disfluency influences listeners’ understanding of some via a rapid integration of the social situation and the speaker’s goals, and that this effect emerges in the early stages of listeners’ comprehension. These results suggest that listeners engage in speaker-modelling behaviour rather than relying on a heuristic relation between disfluency and deception. Also, the current findings clarify prior work whose findings were compatible with either a social reasoning account or an account in which disfluency merely heightens listener attention. Further experiments explore whether listeners’ interpretations reflect beliefs that the speaker has either stretched or even overwritten the meaning of some. Results from these experiments lead this work to broader questions about speakers’ strategic use of language. Given the role of social reasoning, these findings open up new questions about possible differences in this kind of reasoning across individuals with varying pragmatic abilities or across contexts that evoke different conversational goals or speaker agendas.

A final eye-tracking study was conducted to test whether listeners’ pragmatic reasoning varies systematically across individuals. The eye-tracking study was conducted with participants whose pragmatic abilities were posited to vary. The study elaborated on prior work that has suggested that similar behavioural outcomes might be achieved via distinct cognitive processes. People’s eye movements were recorded as they responded to potentially misleading instructions to click on one of two objects which might conceal a hidden treasure. Eye movements analysis showed that listeners were less likely to click on the named object following disfluent instructions. Critically, when hearing disfluent utterances, a tendency to make early fixations to the named object increased with individuals’ Autism Quotient scores. This finding provides further evidence that even where utterances are equivalently understood, the process by which interpretations are achieved varies across individuals with different pragmatic abilities.

Taken together, the results from mouse-tracking and eye-tracking studies reported in this thesis support an account of disfluency processing whereby listeners rapidly integrate the pragmatic cues of disfluency into a social reasoning process in a given context, a process that in turn varies with individuals’ pragmatic abilities.