Computational Linguistics Colloquium

Friday, 17 July, 12:15
Conference Room, Building C7 4

Overt and covert anticipation of verb complements in the visual-world paradigm: Evidence from flickering-cake detection

Altmann & Kamide (1999; AK99) showed that listeners anticipate verb complements before they are available in the spoken input: while looking at scenes comprising, e.g., a boy, a cake, and some toys, participants listened to "the boy will eat [vs. move] the cake". Shortly after hearing "eat" (and before hearing "cake"), participants were more likely (ca. 30% of the time) to launch anticipatory eye-movements to the critical target object (cake) than after hearing "move" (ca. 20%).
The present experiments replicated this design with an additional gaze-contingent picture change manipulation: when subjects were NOT fixating the target during a critical probing point in time (200 ms before or 200 ms after verb-offset), the target would "flicker" for two screen refreshes (16.6 ms) in half of those trials (the remaining trials were controls). Eye-movements in response to this manipulation were taken as an index of covert attention deployment (attending to the target without looking at it; cf. Posner, 1980). We shall report two eye-tracking experiments using this paradigm. In the first experiment, participants were not informed about the gaze-contingent display change manipulation, while in the second experiment, they were asked to "help identify erroneous presentation trials" by pressing a button in case they detected a flicker.
The results can be summarized as follows. In the late probing condition (200 ms after verb-offset and 280 ms before noun-onset), participants were generally more likely to look at the target in the "eat" rather than "move" condition (thus replicating AK99), while no verb-specific differences in flicker detection were established. In the early probing condition (200 ms before verb-offset and 680 ms before noun-onset), no general verb effect in looks to the target was found; however, perceivers were more likely (and faster!) to respond to the flicker manipulation in the "eat" rather than "move" condition, suggesting that they were already covertly attending to the target while still processing "eat".
The between-experiment manipulation (implicit vs. informed change detection) had no significant effect on the observed eye-movement patterns. Button responses in the second experiment indicated rather poor "conscious detection" performance overall, and there were no systematic cross-condition differences in the distance between current gaze location and the location of the display change. We conclude that anticipatory eye-movements are preceded by covert attention shifts, and that "standard" visual-world experiments may therefore underestimate the speed (and possibly even the likelihood) of object-anticipation.

If you would like to meet with the speaker, please contact Pirita Pyykkönen.