Modulation of sensory attenuation by intensive meditation practice: an active inference perspective

Active inference describes motor action as a prediction-driven inferential process, whereby ascending proprioceptive prediction errors are attenuated to allow the fulfillment of expected movement. Meditative practices typically involve a heightened attention to bodily sensations, begging the question of whether this could partially offset the normal proprioceptive suppression during a simple motor act. In this study, 42 experienced meditators completed a tactile force-matching task, designed to measure somatosensory attenuation. The active group ($N=19$) performed the task before (T1), during (T2), and three weeks after (T3) an intensive 10-day mindfulness meditation retreat, while a control waiting list group ($N=23$) was also measured three times, but before participating in the retreat. Analysis of T1 data confirmed the presence of a general somatosensory attenuation effect across groups, which correlated negatively with pre-retreat trait measures of mindfulness, as predicted by our hypothesis. Contrary to our expectations, however, longitudinal analyses did not reveal a global reduction in somatosensory attenuation as an effect of intensive meditation practice. We observed instead a subtler regression-to-the-mean effect at T1, which increased with task repetition in control participants (T1>T2>T3), a training-related phenomenon not previously reported for the force-matching task. Interestingly, this habituation behavior was not shown by the active participants, who maintained the level of regression-to-the-mean observed at baseline at T2, suggesting that the formation of prior expectations about the presented force intensity may be affected by the retreat. We discuss how multiple, opposite effects of meditation on proprioceptive active inference mechanisms, and/or an alteration of prior formation and their influence, may explain these findings.