Disfluencies and Eye MOvements during Speech: what can they reveal about language production?

Summary of the project: "Speech...hm...is difficult to to protest- uh process online"... We all produce disfluencies, which includes various phenomena such as pauses, repeated words, and self-corrections. According to some estimates, the rate of disfluencies in spontaneous speech is about 6 per 100 words; however it is still not well understood why speakers are not totally fluent. To reveal the underlying causes of disfluency, most approaches attempt to relate the pattern of disfluencies to difficulties at specific levels of language production. The current project investigates disfluency by manipulating different levels of production. In addition, we bring together different accounts of disfluency by 1) capturing monitoring or stalling strategy using eye-tracking and 2) identifying the role of non-linguistic cognitive functions and individual differences. 

The first objective of this project was to understand where disfluency takes place in the language system. In other words, we analysed the pattern of disfluency related to lexical selection, grammatical selection, and conceptual generation difficulties in a description task. The second objective was to analyse which other problems (than difficulties in speech encoding) cause disfluency.

In a first study, we analysed the pattern of disfluencies related lexical and grammatical selection difficulty, using a Network Task. Given that disfluencies are multifactorial, we combined this paradigm with eye-tracking to disentangle disfluency related to word preparation difficulties from others (e.g. stalling strategies). In Experiment 1, lines connecting the pictures varied in length, which led participants to use a strategy and inspect other areas than the upcoming picture when pictures were preceded by long lines. Experiment 2 only used short lines. In both experiments, lexical selection difficulty promoted self-corrections, pauses and longer fixation latency prior to naming. Multivariate Pattern Analyses demonstrated that disfluency and eye-movement data patterns can predict lexical selection difficulty. Eye-tracking could provide complementary information about network tasks, by disentangling disfluencies related to picture naming from disfluencies related to self-monitoring or stalling strategies.

In a second study, we asked whether delays in the earliest stages of picture naming elicit disfluency. To address this question, we used visual blurring, which hinders visual identification of the items and thereby slows down selection of a lexical concept. We tested the effect of this manipulation on disfluency production and visual attention. Blurriness did not lead to more disfluency on average and viewing times decreased with blurred pictures. However, multivariate pattern analyses revealed that a classifier could predict, from the pattern of disfluency, whether each participant was about to name blurred or control pictures. Impeding the conceptual generation of a message therefore affected the pattern of disfluencies of each participant individually, but this pattern was not consistent from one participant to another. Additionally, some of the disfluency and eye movement variables correlated with individual cognitive differences.