Shape perception enhances perceived contrast: evidence for excitatory predictive feedback?

www.nature.com/scientificreports OPEN received: 11 September 2015 accepted: 24 February 2016 Published: 14 March 2016 Shape perception enhances perceived contrast: evidence for excitatory predictive feedback? Biao Han1,2 & Rufin VanRullen1,2 Predictive coding theory suggests that predictable responses are “explained away” (i.e., reduced) by feedback. Experimental evidence for feedback inhibition, however, is inconsistent: most neuroimaging studies show reduced activity by predictive feedback, while neurophysiology indicates that most interareal cortical feedback is excitatory and targets excitatory neurons. In this study, we asked subjects to judge the luminance of two gray disks containing stimulus outlines: one enabling predictive feedback (a 3D-shape) and one impeding it (random-lines). These outlines were comparable to those used in past neuroimaging studies. All 14 subjects consistently perceived the disk with a 3D-shape stimulus brighter; thus, predictive feedback enhanced perceived contrast. Since early visual cortex activity at the population level has been shown to have a monotonic relationship with subjective contrast perception, we speculate that the perceived contrast enhancement could reflect an increase in neuronal activity. In other words, predictive feedback may have had an excitatory influence on neuronal responses. Control experiments ruled out attention bias, local feature differences and response bias as alternate explanations. Predictive coding is a form of efficient sensory coding1 that relies on the elimination of predictable neuronal responses and thereby the exclusive processing and transmission of unpredicted portions of the sensory input2–5. As such, predictive coding could have important implications for the dynamics of information flow among the different levels of a sensory hierarchy such as the visual cortex. Standard neuronal implementations of predictive coding assume that the feedback connections carry predictions of expected neural activity and the feedforward connections carry the residual activity between the predictions and initial lower area activity. To carry the residual, the feedforward connections are supposed to be excitatory, whereas to produce the residual the feedback connections are supposed to be inhibitory3,4. To simplify, standard neuronal models of predictive coding hold that the different hierarchical levels interact by excitatory feedforward carrying residual activity and inhibitory feedback carrying predictions. Recent implementations of predictive coding have divided neurons in each cortical area into two sub-populations, one coding for predictions/representations and one for prediction errors4,6. These models suggested that only error units would be suppressed through either direct or indirect inhibition from the prediction units; the prediction/representation units, on the other hand, may actually be enhanced by predictive feedback6,7. Since theory must follow fact, it appears important to investigate the overall perceptual effect of feedback in predictive coding: is it excitatory or inhibitory? Neurophysiology and neuroimaging provide converging supporting evidence for the hierarchical structure and excitatory feedforward connections of predictive coding models8–11, but the experimental data are less unanimous regarding the inhibitory or excitatory nature of predictive feedback12: most neuroimaging studies show reduced activity by predictive feedback13–16, while neurophysiology indicates that most inter-areal cortical feedback is excitatory and targets mostly on the lower area excitatory neurons17–22. In summary, the experimental literature does not clearly and unambiguously support the notion of inhibitory feedback, which is nonetheless an integral part of many models of predictive coding. Here, we employed a psychophysical approach to investigate the properties of predictive coding. To produce predictive feedback, we employed similar stimuli as in Murray et al.: 3D-shape outlines and random-lines versions of the same stimuli13. The former can be easily recognized, and should thus normally produce more 1 Centre National de la Recherche Scientifique, Unité Mixte de Recherche 5549, Faculté de Médecine de Purpan, Toulouse Cedex, 31052, France. 2Université de Toulouse, Centre de Recherche Cerveau et Cognition, Université Paul Sabatier, Toulouse, 31062, France. Correspondence and requests for materials should be addressed to R.V. (email: ) Scientific Reports | 6:22944 | DOI: 10.1038/srep22944 1 www.nature.com/scientificreports/ Figure 1. Main experiment and results. (A) Experimental paradigm. Each trial consisted of a 200–800ms blank screen, a 750ms stimulus screen and a response screen that remained visible until the response was provided. The stimulus screen consisted of a fixation point, one circular gray disk with a 3D-shape stimulus and another with a random-lines stimulus (with randomized positions on the left and right of fixation for every trial). One disk had a fixed contrast level and the other a variable contrast value around that level (randomly assigned on every trial). Subjects were instructed to compare the luminance of the two disks. No feedback was given after the response. (B) Comparison of the grand average psychometric functions (when pooling data over all subjects). Each curve represents the selection probability of the variable disk when this disk contained the 3D-shape (green) or the random-lines stimulus (red). Error bars represent standard error of the mean (SEM) across subjects (C) Psychometric shift for each subject and mean across subjects. Psychometric shift was defined as the difference between the two psychometric functions at 50% selection probability. All 14 subjects showed a positive effect, with the disk behind the 3D-shape stimulus perceived brighter against the black background than the one behind the random-lines. Subjects 1 and 4, marked by colored bars, performed the experiment while their eye position was monitored, and any eye movement or break of fixation discarded. Error bar represents SEM. predictive feedback than the latter. The two kinds of stimuli (black 3D shape and black random lines) were displayed on gray disks simultaneously on the left and right of a fixation point on a black background. Subjects were asked to compare the luminance of the two disks (report the side of the brightest disk). The 3D-shape disk was perceived systematically brighter than the random-lines disk (i.e., its contrast relative to the black lines and screen background was higher than the contrast of the random-lines disk). Since there is experimental evidence suggesting a monotonic relationship between perceived contrast and neuronal activity in early visual areas23,24, we speculate that, at least at the moment at which subjects made their perceptual decision about local contrast, predictive feedback was excitatory rather than inhibitory. Results Main Experiment: luminance judgmen (...truncated)