Sensory attenuation scales with the strength of action-outcome coupling: A psychophysical study

Sensory attenuation refers to the reduced neural or perceptual responses elicited by self-generated stimuli relative to externally generated ones. Yet behavioral findings remain inconsistent. These inconsistencies may reflect the widely differing operationalizations of the strength of the action-outcome coupling (SAOC), that is, how reliably an action predicts its sensory consequence. The present study systematically examined how four factors related to the SAOC influence perceptual sound attenuation. Participants performed an auditory discrimination task in which test tones of varying intensities were generated either after a keypress (overlearned coupling), after an ocular saccade (novel coupling), or following a visual cue (passive listening). We also varied the action-tone delay (150 vs. 1500 ms) and the tone occurrence rate (100% vs. 50% contingency). After each test tone, a reference tone was presented and participants judged which one was louder. In the 50%-contingency condition, trial-wise expectations of tone occurrence were estimated with a Rescorla-Wagner model. Keypress-generated tones were perceived as less loud than passively presented tones, consistent with attenuation typically observed for such overlearned couplings. By contrast, saccade-generated tones were judged quieter when their occurrence was more strongly expected. Overall, this between-task difference indicates that perceptual attenuation scales with SAOC, differently for overlearned and novel couplings. Our findings are compatible with the hybrid account proposed by Dogge et al. (2019), according to which overlearned couplings rely on motor-based forward models, whereas novel couplings engage broader predictive mechanisms.