Kinematic readout of intention primes action prediction

The ability to anticipate what other people are going to do is key to social interaction. Here we explored whether human observers exploit subtle preparatory cues in movement kinematics to anticipate the actions of others. Participants (n = 20) performed a binary categorization of action pictures that showed a person drinking or pouring. On each trial, the action picture (action probe) was preceded by a video of a reach-to-grasp act (kinematic prime), performed either with the intent to pour or drink. The prime-probe relationship varied from trial-to-trial, with matching kinematic prime-action probe pairings shown in 75% of the trials. Participants were faster to respond on trials in which the intention encoded in the kinematic prime matched the action probe. Response time priming was further supported by eye tracking data, with the landing of initial fixations on the action probe picture varying systematically with the intention encoded in the prime. This “kinematic priming” phenomenon varied widely across trials and observers. To quantify whether the amount of kinematic priming in a given trial could be predicted by the amount of intention information extracted by the observer, one hour later, in a second experimental session, we asked participants to judge the intention of the reaching acts used as primes in the action categorization task. Using logistic regressions fitted to intention choices for each participant, we obtained a measure of intention information extracted from each reaching act. We found that the amount of kinematic priming, reflected the amount of kinematic readout with single-subject, single-trial resolution. Results from both response latencies and initial fixations demonstrate that intention-related information encoded in movement kinematics is implicitly read out and spontaneously used to anticipate the actions of others.