[PS-2.47] Multiple Visual Components of Thematic Role Identification

Jessop, A. 1, 2 & Chang, F. 1, 2

1 The University of Liverpool
2 ESRC LuCiD Centre

Thematic roles linguistically code who did what to whom. Here we examine how role identification is related to different components of visual processing. We used scenes where identical circles moved randomly (Gao, Newman & Scholl, 2009), interrupted by events where one circle pushed another. We manipulated the number of times such push events occurred (1, 2, 3). At test, three circles were coloured (red, blue, green) and participants described which objects were involved in a pushing action together using an English transitive frame to mark the agent/patient relations (?red pushed blue?).

The dependent measure was the mean correct labelling of agent and patient for each participant in each condition, with 0.33 subtracted out to make this measure relative to chance. Focal attention enhanced the tracking of 1-push events over the 2/3-push events (t = 2.50), which were not different from each other (t = -0.22). However, overall performance was above chance (intercept t = 4.82), suggesting that some parallel non-attentive process was also available that helped participants to guess correctly in the 3-push trial, where 6 circles had to be tracked. Our work suggests that multiple attentive and non-attentive components of the visual system are used in thematic role identification.