Serial Dependence Scales with Action-binding Depth in Duration Perception

Serial dependence biases perception toward recent history, yet it often appears only when observers respond actively. The critical issue, however, is whether the effect depends on motor output itself or on how strongly stimulus features are built into the response plan. We tested this by comparing active duration reproduction with passive following, where observers simply tracked stimulus onset and offset with a sustained keypress. Twenty-four participants completed both tasks with physically matched sensory input, equating stimulus properties and motor execution, while the degree of coupling between duration and action varied. The sequential bias was strongest for reproduction-after-reproduction trials, weaker for reproduction-after-following trials, and minimal when the current trial required following. Passive following still showed central tendency, with offset latency revealing implicit duration encoding, but it generated little attraction across trials. These findings point to action-binding depth, rather than motor output alone, as the key driver of serial dependence in time perception.