Departmental Colloquium: Brent Strickland, Jean Nicod Institut
Automaticity in the perception of causality
For many years, following Michotte researchers believed that in simple events, like billiard ball collisions, were in some cases "directly" seen as causal. In other words these events were postulated to be automatically categorized as involving causality in a way that may divorced from higher level judgment. In studying this phenomenon however, one major problem has been the use of direct as opposed to indirect measures. Since the 1950's, researchers interested in this topic have typically shown a causal or non-causal event to participants and asked them to assess the extent to which that event looks causal. This leaves open the possibility that any factors that are hypothesized to affect the perception of causality could in fact merely be affecting judgments about causality (Rips, 2011). Here I discuss two new sets of results involving indirect measures in the perception of causality and which help strengthen the argument that causal perception is an automatic perceptual mechanism. The first involves a novel visual search task in which we show that physically impossible accelerations "pop-out" for causal launching events but accelerations do not do so for closely matched but non-causal events (Kominsky*, Strickland*, Wertz, & Keil, under review). We further show that similar effects obtain in pre-verbal infants from 10 months of age. A second demonstration of the automaticity of causal perception involves a novel "switch cost" paradigm in which participants are asked to make a judgment about an orthogonal property (such as shirt color) on images involving an agent (i.e. the actor performing an action) and a patient (i.e. the actor undergoing an action) in a causal interaction (Hafri, Trueswell, & Strickland, under review). Participants are faster in making orthogonal judgments on trials in which they were asked about actors with the same role on the previous trial (e.g. Agent-Agent trial pairs) than when they are asked about actors with different roles (e.g. Patient-Agent trial pairs). Collectively these findings help demonstrate that causality is detected rapidly and automatically during on-line perception, and this can have surprising down stream effects.