[P1] Poster: Day 1

[P1-16] Behavioral Evidence for Precision-Weighted Prediction Updating in the Sub-Second Range: A Pilot Study

*Maki Uraguchi¹, Hideki Ohira¹ (1. Nagoya University (Japan))

Keywords:

predictive processing,precision-weighted updating,temporal generalization task

Predictive processing theory posits that perception involves continuously updating internal models to minimize prediction error, with the rate of updating depending on the relative precision of predictions and sensory inputs. When predictions are highly precise, they are more resistant to change. This study aimed to provide behavioral evidence for this precision-weighted updating hypothesis. We hypothesized that repeated exposure to the standard would enhance the precision of temporal predictions, thereby reducing prediction updating. Participants (120 adults) performed a temporal generalization task using 500 Hz pure tones. After memorizing a 600 ms standard, participants judged whether comparison intervals (420–780 ms) matched the standard. During the learning phase, half of the participants (repetition group) received three additional presentations of the standard, while the rest (control group) encountered it only at the beginning. In the subsequently administered test phase, the longest stimulus was presented more frequently in both groups, encouraging prediction updating toward longer durations. The repetition group exhibited smaller shifts in the weighted mean of the generalization gradient compared to the control group, indicating reduced updating of the internal standard. This supports the idea that greater prediction precision dampens updating, consistent with the principle of precision-weighted inference. We also examined response entropy during the learning phase as a potential marker of prediction uncertainty. The repetition group showed a higher group-mean entropy for the 660 ms stimulus compared to the control group. While this may reflect increased response variability due to a limited number of trials, it could also indicate a dynamic adjustment process—where participants were actively refining their predictions in response to repeated exposure. These findings raise the possibility that entropy may capture transitional stages in the formation of high-precision predictions, though further validation is needed.