Anne Springer: Some insights into the simulation machinery: Action semantics and kinematics affect action prediction

Previous studies provided evidence for the claim that predicting others’ actions involves an internal real‐time simulation. In my talk, I want to highlight the representational formats on which internal simulation is based. I will present several experimental studies showing that semantic processes as well as motor processes affect action simulation. Our participants watched videos of familiar actions performed by a point‐light actor, which were interrupted by an occluder, followed by a test pose. The task was to judge whether the test pose depicted a continuation of the occluded action in the same depth angle (‘prediction task’). Using a paradigm proposed by Graf et al., we independently manipulated
the duration of the occluder and the temporal advance of the test pose relative to occlusion onset (occluder time and pose time, respectively). Therefore, the paradigm allows us to assess real‐time simulation, based on prediction performance across different occluder time/pose time combinations (i.e., an increase in task performance with decreasing time distance between occluder time and pose time is taken to indicate real‐time simulation). Prediction performance was measured when the participants had been initially primed with verbs and concrete nouns (Experiment 1) and with verbs describing dynamic versus static actions (e.g., “to catch” vs. “to stretch”; Experiment 2) (i.e., semantic priming effects). Secondly, in order to assess the effects of motor processes, participants performed actions while doing the prediction task. Observed and executed actions were congruent or incongruent regarding the body side (left vs. right) and the movement pattern involved. (upwards vs. sideways) In short, the results support our idea that action prediction involves real‐time simulation. Consistent with our expectations, motor processes clearly affected action prediction performance; the size of their impact seemed to depend on anatomical mappings between executed and observed actions. Moreover, action prediction performance was modulated by action semantics, suggesting action prediction involves not only sensorimotor representations (as highlighted by research on the human mirror system) but also more abstract, knowledge‐based codes of action. Furthermore, our results allow the differentiation of two processes that may mediate prediction performance. In addition to real‐time simulation, we found evidence of a similarity‐based match of internal action representations, implying that the last action pose perceived before the occlusion is statically maintained as an internal reference and then matched to the action pose appearing after the occlusion (‘static matching’ process). The results will be discussed with respect to recent theories of embodied social cognition.

Springer, A. & Prinz, W. (2010). Action semantics modulate action prediction. The Quarterly Journal of Experimental Psychology, 63 (11), 2141 – 2158.

Springer, A., Brandstädter, S., Liepelt, R., Birngruber, T., Giese, M., Mechsner, F., & Prinz, W. (resubmitted). Motor execution affects action prediction. Brain & Cognition.

Stadler, W., Schubotz, R., von Cramon, D. Y., Springer, A., Graf, M., & Prinz, W. (in press). Predicting and Memorizing Observed Action: Differential Premotor Cortex Involvement. Human Brain Mapping.