# Regression to the mean explains perception of action consequences

## TL;DR

The study shows that people perceive sounds based on average loudness, not just self-generated actions, suggesting a new explanation for sensory changes.

## Contribution

The paper introduces regression to the mean sound statistics as a novel explanation for sensory attenuation and enhancement.

## Key findings

- Perceived loudness regressed to the mean sound level regardless of self-generation.
- Predictable sounds, not just self-produced ones, influenced perception through mean statistics.
- Changing the mean context altered perception of rare tones, supporting the regression hypothesis.

## Abstract

Predictions shape the perceptual consequences of our own actions such that self-generated events appear less intense to us. However, recent studies also reported sensory enhancement of self-produced sounds. Here, we tested whether sensory attenuation and enhancement are signatures of an adaptation to the mean sound statistics. In 330 human participants, we tested the idea that predictions about upcoming sounds shift auditory processing to the average sound context. Participants produced sounds between 40 and 80 decibels (dB) and rated their loudness. Estimates of perceived loudness followed a regression to the mean sound level. The effect was similar for self-produced and passively observed but temporally predictable tones, suggesting predictability alone drives perceptual changes. We then artificially created a new mean sound level by presenting sessions in which subjects mostly (80% of trials) produced either loud (80 dB) or quiet (40 dB) tones. In loud contexts, rarely presented quiet tones were enhanced, and in quiet contexts, loud tones were attenuated. Our results challenge the dominant forward model explanation, which attributes sensory attenuation to predictive suppression of self-generated stimuli, and instead open the door for alternative explanations. Our findings point to regression towards the mean sound level as the most plausible account for predictable sounds.