Dynamic competition between sensory and decisional biases in perceptual decision making

Abstract

Bias in perceptual decision making can have both sensory and decisional origins. These competing sources of bias are typically seen as static and stable over time. However, human behavior is dynamic and constantly adapting. Yet, it remains unclear if the competition between sensory and decisional biases changes over time too. We addressed this question by tracking the dynamics of sensory and decisional biases during a task that involves a visual illusion. Observers saw multiple pairs of peripherally presented faces that induce a strong illusion making the faces appear distorted and grotesque. The task was to judge whether one of the last two faces had true physical distortion (experimentally introduced in half of the trials). Initially, subjects classified most faces as distorted as exemplified by a liberal response bias. However, over the course of the experiment, this response bias gradually disappeared despite the fact that the distortion illusion remained equally strong, as demonstrated by a separate subjective rating task without artificially distorted faces. The results suggest that the sensory bias was progressively countered by an opposite decisional bias. This transition was accompanied by an increase in reaction times and a decrease in confidence relative to a condition that does not induce the visual illusion. All results were replicated in a second experiment with inverted faces. These findings demonstrate the existence of dynamic competition between sensory and decisional biases that together determine how humans make perceptual decisions.

How to objectively quantify a subjective phenomenon: The case of Flashed Face Distortion Effect

“…We re-analyzed data from two previous experiments (Gao et al, 2023) and further collected data for a new experiment (N = 27 in each). Experiments 1 and 2 tested the effects of temporal frequency and eccentricity on FFDE, whereas Experiment 3 tested the effects of inserting temporal gaps between successive faces.…”

“…We used the same stimuli from our previous work (Gao et al, 2023). Briefly, the stimuli consisted of 15 female and 15 male faces from the Karolinska Directed Emotional Faces database (Lundqvist et al, 1998).…”

“…We analyzed data from three experiments (N=27 for each). Experiments 1 and 2 were taken from a previous manuscript (Gao et al, 2023). Experiment 3 was a new experiment and comprised 17 females and 10 males with a mean age of 20.04 (SD = 22.61,.…”

“…For Experiments 1 and 2, we performed 2-way ANOVA (analysis of variance) with repeated measures on the response bias, distortion rating and sensitivity to examine the effect of temporal frequency and eccentricity of face presentation (Gao et al, 2023). For Experiment 3, we performed paired t-tests on the response bias and distortion rating between the control condition and the illusion condition with different temporal gaps.…”

Testing Location Invariance of the Flashed Face Distortion Effect

“…Later research, however, showed that the effect of inversion is rather small (Balas & Pearson, 2019; see also Exp. 2 of Gao et al, 2023), suggesting that holistic face processes might play less of a role than originally thought. Instead, studies found that image eccentricity and presentation rates are more important factors in determining the strength of the illusion: faces presented more peripherally and at a slower rate created the strongest illusion (though local contrast does not seem to matter, see Balas & Pearson, 2019; see also Bowden et al, 2019).…”