Goal uncertainty biases memory for observed actions

Similar actions can be taken in different contexts. For example, we may make similar reaching movements towards a glass of water either when water is the only choice in front of us, or when a decision between competing goals must be made (e.g., a glass of water versus a neighboring glass of wine). This raises an intriguing question: Does the context in which an action is observed leave traces in our memory for that action? To address this, we algorithmically generated 10 videos, each depicting an agent’s hand following a unique trajectory to a target. Participants were instructed to closely observe how the agent moved their mouse. After a brief mask, they then replicated the trajectory they observed by moving their own mouse into the same target. Critically, each of the curved trajectories was presented in two distinct contexts: the agent’s cursor either moved to a single target in the workspace (single-target context) or toward one of two possible targets (dual-target context). We reasoned that dual-target movements would be perceived more as “choices” while single-target movements would be perceived more as “demonstrations.” Thus, we hypothesized that the presence of a second goal would repel observers' memories of the movement trajectories toward the chosen target. Analysis of reproduced trajectories supported our prediction: Participants consistently reproduced more efficient trajectories in the dual-target context than the single-target context (as indicated by significantly different initial angles and positional deviations of their movements). A follow-up experiment replicated this finding by having participants reproduce movements without a visible cursor, ruling out the possibility that the effect was driven by feedback during reproduction. The results suggest that the memory of an observed action can be shaped by the context in which that action is executed: uncertainty about an agent’s intent biases our memory of their actions.