A model of subjective report and objective discrimination as categorical decisions in a vast representational space

Abstract

Subliminal perception studies have shown that one can objectively discriminate a stimulus without subjectively perceiving it. We show how a minimalist framework based on Signal Detection Theory and Bayesian inference can account for this dissociation, by describing subjective and objective tasks with similar decision-theoretic mechanisms. Each of these tasks relies on distinct response classes, and therefore distinct priors and decision boundaries. As a result, they may reach different conclusions. By formalizing, within the same framework, forced-choice discrimination responses, subjective visibility reports and confidence ratings, we show that this decision model suffices to account for several classical characteristics of conscious and unconscious perception. Furthermore, the model provides a set of original predictions on the nonlinear profiles of discrimination performance obtained at various levels of visibility. We successfully test one such prediction in a novel experiment: when varying continuously the degree of perceptual ambiguity between two visual symbols presented at perceptual threshold, identification performance varies quasi-linearly when the stimulus is unseen and in an ‘all-or-none’ manner when it is seen. The present model highlights how conscious and non-conscious decisions may correspond to distinct categorizations of the same stimulus encoded by a high-dimensional neuronal population vector.

Confidence as Bayesian Probability: From Neural Origins to Behavior

“…First, manipulating the source of the original probabilistic representation should affect the readouts of both choices and confidence levels. Stronger levels of evidence lead, as expected, to higher levels of confidence in humans, monkeys, and rodents across a variety of tasks, such as visual or olfactory discrimination Mamassian, 2010, 2009;Kepecs et al, 2008;Kiani and Shadlen, 2009;King and Dehaene, 2014;Peirce and Jastrow, 1884). Further evidence is provided by manipulation of neural representations in early sensory areas.…”

Heuristic use of perceptual evidence leads to dissociation between performance and metacognitive sensitivity

“…It is relatively straightforward to extend the simple two-dimensional SDT model described above to a Bayesian decision theoretic formulation (see e.g., King & Dehaene, 2014). Notably, with the most basic assumptions (described below), Bayesian and SDT formulations produce identical predictions.…”

“…In sensory psychophysics experiments, this ability is reflected in the fact that subjects’ confidence ratings correlate meaningfully with the likelihood of accurate decisions (Fleming, Huijgen, & Dolan, 2012; Fleming, Weil, Nagy, Dolan, & Rees, 2010). In traditional psychophysics models (Green & Swets, 1966; King & Dehaene, 2014; Macmillan & Creelman, 2004), subjects are assumed to rate their confidence based on the same internal sensory evidence underlying perceptual decisions; this theoretical view is adopted in recent animal studies (Kepecs, Uchida, Zariwala, & Mainen, 2008), and supported by the finding that there are single neurons whose firing rate reflects both confidence and perceptual decision (Kiani & Shadlen, 2009). …”

Perceptual awareness and its neural basis: bridging experimental and theoretical paradigms

“…Several of the contributed articles report new experimental findings supporting theoretical claims, hypotheses and predictions [17,38,47,56,59]. Others emphasize review and theoretical aspects; however, they are systematically and critically related to relevant empirical findings with different experimental paradigms.…”

“…In the past few years, the GNW model has inspired a new trend of theoretical developments in which series of discrete conscious processes, involving GW transitions, are considered as related to mental programmes controlling multiple steps or routines in cognitive tasks ( [48,54 -56], about the role of discreteness in perceptual processing). The GNW [34,47,57] model has led to the emphasis on the 'all-or-none' character of conscious representations, whereas other approaches ( [58]; see also [52]) suggest a graded account. Such positions might, however, be reconciled [25,59].…”