Didi Dunin¹, Alon Hafri², Benjamin F. van Buren¹; ¹The New School, NYC, ²University of Delaware

When people see a moving object disappear, they misremember its final position as slightly farther along its trajectory — a phenomenon known as ‘representational momentum’. Such ‘forward’ biases provide evidence that perception is based on predictive processing, which constructs a model of the world to predict upcoming sensory input. While such predictive mechanisms are traditionally discussed in terms of external stimuli, here we ask whether they can also guide the unfolding of internally generated procedures. We often generate step-by-step internal sequences — for example, counting in 2’s or mentally rehearsing the steps needed to navigate from one location to another. If predictive mechanisms support these kinds of processes as well, an effect similar to representational momentum might arise — specifically, people’s memory for where they last were in an internally generated sequence might likewise be biased forward. Participants completed a single trial in which they mentally counted up or down in 2’s, tapping a key on each step. At an unexpected interruption point, they were asked what number they had last counted to. Participants reliably reported numbers ahead of where they actually were: a forward bias for internally generated sequences. In several further experiments, we confirmed that this forward bias emerges regardless of whether participants are counting upward in 1’s or 2’s, whether they make their report immediately or after a delay, and whether they count upward arbitrarily or with the goal of reaching a particular target number. This highly reliable bias indicates that when interrupted mid-sequence, participants' internal computation has already progressed, revealing a form of procedural momentum. We conclude that predictive mechanisms similar to those which anticipate the future in perception also govern the unfolding of internally generated procedures.