The effect of motivation and working memory on perspective-use: Evidence from eye-tracking

Language comprehension frequently requires the listener to adopt a speaker’s perspective in order to narrow down an intended target referent. For example, to understand a speaker who asks for “the ball” when more than one ball is available, one must infer that speaker’s knowledge of, or access to, the available objects. Over the last decade, a number of psycholinguistics researchers have used the visual-world paradigm to investigate how such inferences evolve over time during communication (see Huettig, Rommers, & Meyer, 2011). Recently, research has demonstrated a link between the ability to use perspective during language processing and working memory (WM; e.g. Lin, Keysar, & Epley, 2010). Across two eye-tracking experiments we examine whether motivation modulates the effects of WM load on the time-course of perspective-taking in a referential communication task.

In the task participants moved target objects (e.g. a glass with a straw in) around a grid based on instructions from an avatar (e.g. “Move the glass with the straw in left”). In a ‘listener privileged’ condition, an alternative target object (e.g. a glass with a lemon in) could only be seen by the participant. In a ‘speaker privileged’ condition, there was an indication of a potential target object only available to the speaker. In a control condition, both target and alternative target objects were available to both participant and speaker. Working memory load was manipulated within each condition. In Experiment 2 (N=31), but not Experiment 1 (N=36), participants received financial reward for quick and accurate responses.

In Experiment 1 - where there was no reward or time-pressure - listeners did not use perspective cues to disambiguate the target object from the distractor under either the high or low WM load. In Experiment 2 - where there was a reward for speed and accuracy - listeners used perspective cues to disambiguate the target object from the distractor object from the earliest moments of processing, but only under low load. Under high load, responses were similar to that of the control condition, where both objects were in common ground. Furthermore, attempts to initiate perspective relevant responses under high load led to reduced recall on the concurrent WM task – indicating that inferring perspective from a speaker’s utterance was drawing on limited cognitive resources. Across both experiments listeners’ fixations were not influenced by the speaker’s potential knowledge about an alternative target.