Figure 1.

Graphical summary of two neurocomputational mechanisms through which mindfulness may work. First, mindfulness increases the precision of likelihood (i.e., the precision of incoming sensory information), by decreasing the threshold of conscious access through attentional amplification. Thus, at time t, belief is updated by integrating the more precise likelihood and the prior (i.e., the internal model prior to receiving new sensory evidence), which leads to a more precise posterior. Consequently, the newly updated prior at time t+1 (green area), which is equivalent to the posterior at time t, shows an increased precision. Second, the attentional pattern of mindfulness, which is executed “moment to moment,” enables the prior belief to be optimally adjusted to the context of present experience. Thus, prior beliefs are iteratively updated as changes in environment occur over time. This dynamic process makes it possible to continuously minimize prediction error, which is the discrepancy between the prior and the likelihood, as the former is the best suited to the present context.