Bayesian priors are encoded independently from likelihoods in human multisensory perception

Abstract

It has been shown that human combination of crossmodal information is highly consistent with an optimal Bayesian model performing causal inference. These findings have shed light on the computational principles governing crossmodal integration/segregation. Intuitively, in a Bayesian framework priors represent a priori information about the environment, i.e., information available prior to encountering the given stimuli, and are thus not dependent on the current stimuli. While this interpretation is considered as a defining characteristic of Bayesian computation by many, the Bayes rule per se does not require that priors remain constant despite significant changes in the stimulus, and therefore, the demonstration of Bayes-optimality of a task does not imply the invariance of priors to varying likelihoods. This issue has not been addressed before, but here we empirically investigated the independence of the priors from the likelihoods by strongly manipulating the presumed likelihoods (by using two drastically different sets of stimuli) and examining whether the estimated priors change or remain the same. The results suggest that the estimated prior probabilities are indeed independent of the immediate input and hence, likelihood.

Causal Model Comparison Shows That Human Representation Learning Is Not Bayesian

“…Despite much work suggesting that the human brain is Bayes-optimal (Körding and Wolpert 2004;Beierholm et al 2009), and in line with our previous findings (Niv et al 2015), our manipulation was only effective when we based it on predictions of the reinforcementlearning model. Our ability to manipulate the learning process both precisely and in real time consists of a proof of concept for the new proposed model-testing tool, and is a step in the right direction in terms of development of individualized tools to improve learning in general.…”

The COGs (context, object, and goals) in multisensory processing

“…As detailed below, statistical learning is known to support a variety of mental functions, both within and across the senses (reviewed in Frost et al 2015), with effects transpiring across multiple temporal scales(e.g. Baier et al 2006;Beierholm et al 2009;Chandrasekaran et al 2009;Barakat et al 2013;Barenholtz et al 2014;Altieri et al 2015;Sarmiento et al 2012Sarmiento et al , 2016.…”

The context-contingent nature of cross-modal activations of the visual cortex

“…Studies using simplistic (but carefully controlled) experimental paradigms have been documenting for de facto decades the benefits of perception in cross-modal settings (reviewed in, e.g., Spence and Driver, 2004). Similarly, if context is understood as 'the immediate situation within which a stimulus is processed by the brain, the literature offers a plethora of examples whereby the brain utilises regularities (short-term as well as long-term) within and across the senses to facilitate the processing of the current inputs that are relevant to the task-at-hand (e.g., von Kriegstein and Giraud, 2006;Rosenblum et al, 2007;Beierholm et al, 2009;Powers et al, 2009;Barakat et al, 2013;Chandrasekaran et al, 2009;Thelen et al, 2014;Matusz et al, 2015c;Altieri et al, 2015). For example, the efficiency of the perceptual processing can be increased, as indexed by early-latency reductions in the brain-response strength, after several days of explicit, object-discrimination training with multisensory stimuli (Altieri et al, 2015).…”

“…More recently, the context has been revealed as an important source of top-down control over processing of multisensory information. While some studies demonstrated the role of long-term experience and learning (e.g., Froyen et al, 2009;Stevenson and Wallace, 2013;Barenholtz et al, 2014;Ten Oever et al, 2014;Matusz et al, 2015b), many focused on effects operating at shorter timescales, such as expectations and/or experiences built over the course of a single experimental session (e.g., Murray et al, 2004Murray et al, , 2005von Kriegstein and Giraud, 2006;Meylan and Murray, 2007;Rosenblum et al, 2007;Beierholm et al, 2009;Powers et al, 2009;Barakat et al, 2013;Chandrasekaran et al, 2009;Thelen et al, 2012Thelen et al, , 2014Matusz et al, 2015c;Altieri et al, 2015), or even across a pair of successive experimental trials Murray et al 2009;King et al 2012;Sarmiento et al, 2015). Considered together, the overwhelming evidence for the importance of context-based factors for stimulus processing across the senses and the concomitant limited existing data on the ACOP makes it plausible that irrelevant sounds activate the visual cortex in some contexts but not in others.…”