Exploring the role of serial dependence in visual time perception

Abstract

Serial dependence biases current perception based on recent experiences, creating continuity in subjective experience. Although extensively studied in vision across tasks such as numerosity, orientation discrimination, and attractiveness, its effect on visual time perception remains partially unexplored. Here, we investigated serial dependence in visual temporal perception, using two common tasks: temporal interval duration discrimination and temporal reproduction. In the discrimination task, participants judged whether the second of three visual stimuli was longer or shorter than the third, with the first stimulus being irrelevant, to induce serial dependence on the second stimulus. In the temporal reproduction task, participants were asked to reproduce an interval presented between two visual stimuli by pressing a button. Given the debate concerning the origin of the serial dependence effect and possible relation with memory processing, we also investigated the relationship between serial dependence and working memory capacity using a Corsi test. Our results showed that serial dependence does occur in visual time perception, but no relationship was found between the effect of the two tasks and memory retention capacity. The lack of correlation between serial dependence effects suggests that different processes may be involved in serial dependence across the two types of tasks.

Figure 1.

(A) Experimental procedure for the duration discrimination task. The sequence of stimuli presented was composed of, first, a task-irrelevant inducer stimulus (duration 199 or 481 ms, randomized within trials), followed by a reference in the same position as the inducer (duration 310 ms) and test, positioned in the opposite hemifield stimulus (with variable duration from 199 to 600 ms). Participants had to compare the reference and test stimuli and determine whether the reference or the test lasted longer. (B) Temporal reproduction task procedure. Two white dots were presented on the screen (the first one on the left of a black fixation dot and the second on the right opposite position) for 20 ms, with a delay between the presentations that varied in a range of 494 to 847 ms. Participants were asked to reproduce the perceived duration that passed between the two dots. (C) Example of Corsi block test in physical form. The experiment was classically conducted with physical wooden blocks; the experimenter touched the blocks sequentially with one hand, and the participant was then asked to repeat the sequence. (D) PEBL version of Corsi block test. In this case, some of the squares would change color from blue to yellow sequentially, and participants were asked to point to the squares that changed color in the correct positions, both forward and backward.

Figure 2.

(A) Representative single-subject psychometric curve for the temporal discrimination task condition for the two levels of inducer duration. When the inducer was smaller than the reference, the curve shifted leftward compared with when it was larger than the reference, indicating an under- or overestimation of the reference magnitude in the direction of the inducer. (B) Proportion between the PSE at 199 ms and PSE at 481 ms. The red dot and error bars refer to the average values for both distributions.

Figure 3.

(A) Pre- and post-residualization of subjects’ error in response was performed to account for the regression-to-the-mean effect. In the biased data (left plot), it is evident that the data exhibit a central tendency bias (regression curve with a negative coefficient). (B) Results of the LMEM used to estimate serial dependence both before (blue) and after (red) the residualization process. Following residualization, the effect of the N-stimuli was markedly reduced, approaching non-significance (with an estimate near zero). In contrast, the negative effect of the previous stimuli and the positive effect for the response one remained significant after residualization, indicating that its association with serial dependence is robust and unaffected by confounding influences.