You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B4, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Prediction under Load: The Effects of Cognitive Load Presence and Type on Anticipation and Competition in Spoken Language Processing

Kate Pirog Revill1;1Emory University

There is an ongoing debate about whether prediction is fundamental to language processing or whether it is a cognitively costly strategy that is deployed only under likely circumstances. Further, there are questions about the level and specificity of the predictions that are made, and what mechanisms or brain systems are used to generate these predictions. Previous research has shown that participants make anticipatory fixations to likely target objects following a biasing verb. Fixations to phonological ‘cohort’ competitors are reduced when an incompatible biasing context is present. Here, we examine the effect of a resource-taxing memory load task on participants’ abilities to use biasing verb information to make anticipatory fixations to likely targets or to reduce competition from inconsistent cohort competitors. Participants heard sentences like ‘The person will sell/drive the truck’ while viewing displays containing a truck and several distracter objects, one of which could be a cohort competitor (trunk) that was a plausible completion following one verb but not the other. Before half of the trials, participants were given a list of spatial locations or a list of consonants to remember; they reported the list back following the selection of the sentence’s referent. Participants remembered a three-item list of locations (squares in a 4x4 grid) equally as well as participants remembered a five-item list of consonants (from among 16 possible choices), correctly reporting back 91.7% and 92.0% of the memory items respectively. Main task accuracy also remained high (> 99%) despite the presence or type of cognitive load. Analysis of anticipatory eye movements showed a small but significant effect of cognitive load, as target curves were shifted 45ms later when a cognitive load was present. Despite the delay, participants made anticipatory fixations to potential target objects following a constraining verb in both load conditions, with target fixation curves shifted 205ms earlier relative to a nonconstraining verb. This effect did not interact with load presence and was not affected by whether the load was spatial or verbal. However, load presence and type did affect the amount of lexical competition from a displayed cohort competitor. As expected, there was a significant effect of verb constraint on the size of the competitor effect; participants fixated cohort competitors less when the verb was constraining than when it was not. Unlike anticipatory fixations, this effect interacted with load: the difference in cohort fixations between a biasing and neutral context was smaller when the participant had to maintain a memory load. Furthermore, load type may matter; participants were still able to use context from a biasing verb to reduce competitor fixations while maintaining a spatial memory load, but context did not significantly reduce competitor fixations for participants asked to maintain a verbal memory load. Taken together, this data suggests that while at least some forms of prediction in language processing are resistant to the presence of a secondary, resource-demanding task, other consequences of prediction, like the modulation of lexical competition, may be more demanding and may rely on domain-specific language production mechanisms.

Topic Area: Control, Selection, and Executive Processes

Back to Poster Schedule