What visual illusions teach us about schizophrenia

Abstract

Illusion, namely a mismatch between the objective and perceived properties of an object present in the environment, is a common feature of visual perception, both in normal and pathological conditions. This makes illusion a valuable tool with which to explore normal perception and its impairments. Although still debated, the hypothesis of a modified, and typically diminished, susceptibility to illusions in schizophrenia patients is supported by a growing number of studies. The current paper aimed to review how illusions have been used to explore and reveal the core features of visual perception in schizophrenia from a psychophysical, neurophysiological and functional point of view. We propose an integration of these findings into a common hierarchical Bayesian inference framework. The Bayesian formalism considers perception as the optimal combination between sensory evidence and prior knowledge, thereby highlighting the interweaving of perceptions and beliefs. Notably, it offers a holistic and convincing explanation for the perceptual changes observed in schizophrenia that might be ideally tested using illusory paradigms, as well as potential paths to explore neural mechanisms. Implications for psychopathology (in terms of positive symptoms, subjective experience or behavior disruptions) are critically discussed.

Keywords: Bayesian inference; delusions; hallucinations; illusions; predictive coding; psychosis; schizophrenia; visual perception.

Figure 1

Main classical illusions. In the Ebbinghaus (A), Ponzo (B) and Müller-Lyer (C) illusions, same-sized patterns are misevaluated because of the context. In the Poggendorff illusion (D), the context disrupts the impression of continuity. Herman's grid (E) generates illusory gray points at each intersection of the white lines. The Boring wife/mother-in-law (F), the Necker Cube (G) and Rubin's Maltese Cross (H) are ambiguous figures that result in different interpretations.

Figure 2

An example of a context suppression effect: the Ebbinghaus illusion. Depending on whether the peripheral circles are large or small, the central targets appear smaller or larger, respectively. When comparing two targets, the actual difference (here, left one smaller compared with right one) (A) appears reduced with the misleading context (B) or increased with the facilitating context (C).

Figure 3

Necker Cube. (A) The ambiguity results in the subjective impression of two interpretations switching: the phenomenon of bistability. (B) The shadow introduces a cue that is supposed to bias perception toward one interpretation. The fact that bistability persists despite the presence of the cue ensures this cube responds to the definition of a visual illusion, i.e., a dissociation between perception and the physical characteristics of its support.

Figure 4

Hierarchical inference with Gaussian variables. In this toy example (left part), the inference that corresponds to the hidden cause (x) could be understood as the probability that the green color that I am observing (sensory evidence, represented by the magenta arrows) is due to the presence of a leaf, given my knowledge of the existence of a tree (prior expectation, represented by the violet arrows). Blue and yellow lines fit for the feedback and feedforward connections that enable the inferential process. Green and black circles fit for the controlling inhibitory system. The right part of the figure represents the probability distribution of each variable and the resulting posterior probability (in red).

Figure 5

Circular inference and relationship with predictive coding. If both descending and climbing loops are impaired (A), both sensory evidence and prior knowledge are reverberated and over counted (multiplication of the pink and violet arrows). In the Predictive coding framework, this results in an overconfident (narrowing of the posterior distribution) but not biased (unchanged K) inferred belief. In contrast, when the impairment only affects climbing loops (B), sensory evidence, but not prior knowledge, is reverberated. The prediction error is emphasized (K is too large), and the inferred belief is biased toward sensory evidence. If, in contrast, the inhibitory disequilibrium disfavors the descending loops (C), only the prior knowledge is over counted because of its reverberation. The prediction error is then minimized (K is too small), and the resulting posterior is biased toward expectations. Note that case (B,C), the inferred belief is associated with an excessive degree of confidence.