Event segmentation—continuous experiences divided into distinct memory units. But what fuels this process? Our study compares two theories: prediction error and contextual stability. We found that transitions across stable contexts matter more. osf.io/preprints/ps...

We had two experiments. Both had prediction errors. One had contextual stability. Event segmentation was present only around stable contexts. Prediction errors were not sufficient to drive segmentation. This is against the influential event segmentation theory.