Spatial generalization of serial dependence in visual duration perception

doi:10.3724/SP.J.1041.2024.00394

Abstract

Abstract:

To establish a stable and sensitive experience of the world, the brain tends to use recent history when forming perceptual decisions. This results in serial dependence in perception, by which previous trials affect the current perception. The serial dependence effect can be divided into (at least) two categories: the effect of previous stimuli (i.e., the stimulus serial dependence effect) and the effect of previous decisions (i.e., the decisional serial dependence effect) on the current perception. Although separate stimulus and decisional serial dependence effects have been demonstrated in duration perception, their spatial selectivity is unclear. In the present study, we investigated whether and how serial dependence in duration perception generalizes across different visual positions of stimuli.
The modified temporal bisection task was used in three experiments. During the Experiment 1, the visual stimulus (a white Gaussian blob) was pseudorandomly presented in the central or peripheral (10° from the left fixation) visual field. Participants were asked to judge whether the duration of the test stimulus (i.e., 300, 395, 520, 684, or 900 ms) was shorter or longer than a reference stimulus of intermediate duration (i.e., 520 ms) once the test stimulus disappeared. To investigate the separate effects of previous stimuli and previous choices on the current duration perception, one-sample t-tests were conducted on the β_{prev_stimulus} and β_{prev_choice} coefficients. The results found that the β_{prev_stimulus} coefficients in all conditions were significantly smaller than zero (CC: t(23) = −4.88, p < 0.001, Cohen’s d = −1.00, 95% CI [−1.28, −0.52]; PP: t(23) = −4.61, p < 0.001, Cohen’s d = −0.94, 95% CI [−0.83, −0.32]; PC: t(23) = −5.57, p < 0.001, Cohen’s d = −1.14, 95% CI [−0.96, −0.44]; CP: t(23) = −3.71, p = 0.001, Cohen’s d = −0.76, 95% CI [−0.87, −0.25], see Figure 1A), and the β_{prev_choice} coefficients were significantly larger than zero (CC: t(23) = 12.82, p < 0.001, Cohen’s d = 2.62, 95% CI [1.06, 1.47]; PP: t(23) = 17.49, p < 0.001, Cohen’s d = 3.57, 95% CI [1.18, 1.50]; PC: t(23) = 3.08, p = 0.005, Cohen’s d = 0.63, 95% CI [0.09, 0.47]; CP: t(23) = 4.22, p < 0.001, Cohen’s d = 0.86, 95% CI [0.15, 0.45], see Figure 1B). The 2 (position: central vision, peripheral vision) × 2 (context: consistent, inconsistent) repeated-measures ANOVAs were conducted on the β_{prev_stimulus} and β_{prev_choice} coefficients, respectively. Results based on β_{prev_stimulus} showed that the main effect of position was marginally significant (F(1, 23) = 3.93, p = 0.060, η_p²= 0.15), while the main effect of context (F(1, 23) = 0.69, p = 0.415) and their interaction (F(1, 23) = 0.83, p = 0.372) were not significant. The main effect of context was significant (F(1, 23) = 198.87, p < 0.001, η_p²= 0.90), while the main effect of position (F(1, 23) = 0.54, p = 0.468) and their interaction (F(1, 23) = 0.34, p = 0.567) were not significant, for the β_{prev_choice}. These results showed that previous stimulus duration and previous choice exerted opposing effects on serial dependence of duration perception: specifically, a repulsive stimulus serial dependence and an attractive decisional serial dependence. In other words, current duration estimates were repelled away from the previous trial’s stimulus duration but attracted toward the previous choice. Moreover, these results showed that the stimulus serial dependence fully transferred across the central and peripheral visual fields, while the decisional serial dependence effect could only be partially transferred across the central and peripheral visual fields.
The task of Experiment 2 was similar to that of Experiment 1, except that the visual stimulus was pseudorandomly located at either 5° to the left or 5° to the right of the central fixation. The one-sample t-tests showed that the β_{prev_stimulus} coefficients in all conditions were significantly smaller than zero (LL: t(22) = −2.48, p = 0.021, Cohen’s d = −0.52, 95% CI [−0.65, −0.06]; RR: t(22) = −2.43, p = 0.024, Cohen’s d = −0.51, 95% CI [−0.67, −0.05]; RL: t(22) = −3.17, p = 0.004, Cohen’s d = −0.66, 95% CI [−0.84, −0.17]; LR: t(22) = −2.97, p = 0.007, Cohen’s d = −0.62, 95% CI [−0.81, −0.14], see Figure 2A), and the β_{prev_choice} coefficients were significantly larger than zero (LL: t(22) = 11.50, p < 0.001, Cohen’s d = 2.40, 95% CI [0.92, 1.33]; RR: t(22) = 8.49, p < 0.001, Cohen’s d = 1.77, 95% CI [0.79, 1.30]; RL: t(22) = 4.50, p < 0.001, Cohen’s d = 0.94, 95% CI [0.28, 0.77]; LR: t(22) = 4.66, p < 0.001, Cohen’s d = 0.97, 95% CI [0.30, 0.78], see Figure 2B). The 2 (position: left, right) × 2 (context: consistent, inconsistent) repeated-measures ANOVAs were conducted on the β_{prev_stimulus} and β_{prev_choice} coefficients, respectively. Results of β_{prev_stimulus} showed that none of the main effects of position (F(1, 22) = 0.01, p = 0.941), context (F(1, 22) = 0.96, p = 0.339), and their interaction (F(1, 22) = 0.02, p = 0.878) was significant. The results of β_{prev_choice} showed that the main effect of context was significant (F(1, 22) = 68.36, p < 0.001, η_p²= 0.76), while the main effect of position (F(1, 22) = 0.26, p = 0.616) and their interaction (F(1, 22) = 0.60, p = 0.449) were not significant. Additionally, the stimulus serial dependence could be fully transferred across the left and right visual fields, while the decisional serial dependence effect could only be partially transferred across the left and right visual fields. These results further indicate that the serial dependences could occur in the peripheral visual field.
In Experiment 3, the positions of both the fixation and the visual stimulus were changed; there were thus four types of positional relationships between stimuli across trials (i.e., identity, retinal position change, external position change, and both changes). The one-sample t-tests showed that the β_{prev_stimulus} coefficients in all conditions were significantly smaller than zero (SC: t(23) = −5.84, p < 0.001, Cohen’s d = −1.19, 95% CI [−1.66, −0.79]; RC: t(23) = −6.38, p < 0.001, Cohen’s d = −1.30, 95% CI [−1.34, −0.68]; BC: t(23) = −6.21, p < 0.001, Cohen’s d = −1.27, 95% CI [−1.58, −0.79]; NC: t(23) = −5.24, p < 0.001, Cohen’s d = −1.07, 95% CI [−1.56, −0.68], see Figure 3A), and the β_{prev_choice} coefficients were significantly larger than zero (SC: t(23) = 6.60, p < 0.001, Cohen’s d = 1.35, 95% CI [0.45, 0.85]; RC: t(23) =5.76, p < 0.001, Cohen’s d = 1.18, 95% CI [0.34, 0.72]; BC: t(23) = 7.18, p < 0.001, Cohen’s d = 1.47, 95% CI [0.50, 0.90]; NC: t(23) = 9.90, p < 0.001, Cohen’s d = 2.02, 95% CI [0.93, 1.43], see Figure 3B). The one-way repeated-measures ANOVA was conducted separately on the β_{prev_stimulus} and β_{prev_choice} coefficients. Results showed that the main effect of position on β_{prev_stimulus} was not significant (F(3, 69) = 0.25, p = 0.861), while the main effect on β_{prev_choice} was significant (F(3, 69) = 8.54, p < 0.001, η_p²= 0.27). The post-hoc test (Bonferroni) revealed that β_{prev_choice} coefficient in the NC condition was significantly larger than those in other conditions (SC: p_bonf < 0.001, Cohen’s d = −1.30; RC: p_bonf = 0.002, Cohen’s d = −1.07; BC: p_bonf = 0.006, Cohen’s d = −0.96). These results demonstrated that the stimulus serial dependence fully transferred across different spatiotopic and retinotopic positions. However, we found that the decisional serial dependence effect was larger in the position-consistent context than in the position-inconsistent context. This indicates that the decisional serial dependence effect could only be partially transferred across different visual positions regardless of the types of positions (i.e., spatiotopic vs. retinotopic).
To sum up, both previous stimuli duration and previous choices affect subsequent perceptual decisions about duration, resulting in repulsive and attractive serial dependence effects, respectively. The repulsive stimulus serial dependence effect fully generalizes across different visual positions, suggesting it occurs primarily in higher-level visual areas. This also implies the existence of fast-duration adaptation. The attractive decisional serial dependence effect suggests that there is decision inertia in perceptual choices. Moreover, this effect is partly contingent on the visual position, which may result from the category organization function of higher-order brain areas. This suggests that the brain takes advantage of the visual position context when forming the decisional prior. These findings are helpful for understanding the plasticity of duration perception.

Key words: duration perception, serial dependence effect, duration adaptation, decision inertia, spatial generalization

WANG Biyao, CHEN Chen, HU Xiaofei, WANG Di, LI Baolin. (2024). Spatial generalization of serial dependence in visual duration perception. Acta Psychologica Sinica, 56(4), 394-411.

Figures/Tables 3

Figure 1. Model results of Experiment 1. Beta coefficients for two predictors: (A) previous stimulus and (B) previous choice in different conditions. The gray solid bar of CC represents the previous stimulus presented in the central field, and the current stimulus also presented in the central field; the black solid bar of PP represents the previous stimulus presented in the peripheral field, and the current stimulus also presented in the peripheral field; the light striped bar PC represents the previous stimulus presented in the peripheral field, while the current stimulus presented in the central field; the dark striped bar CP represents the previous stimulus presented in the central field, while the current stimulus presented in the peripheral field. Scatter points represent the individual data, and the error line represents the standard error.
***p < 0.001, **p < 0.01.

Figure 2. Model results of Experiment 2. Beta coefficients for two predictors: (A) previous stimulus and (B) previous choice in different conditions. The gray solid bar of LL represents the previous stimulus presented in the left field, and the current stimulus also presented in the left field; the black solid bar of RR represents the previous stimulus presented in the right field, and the current stimulus also presented in the right field; the light striped bar RL represents the previous stimulus presented in the right field, while the current stimulus presented in the left field; the dark striped bar LR represents the previous stimulus presented in the left field, while the current stimulus presented in the right field. Scatter points represent the individual data, and the error line represents the standard error.
***p < 0.001, **p < 0.01, *p < 0.05.

Figure 3. Model results of Experiment 3. Beta coefficients for two predictors: (A) previous stimulus and (B) previous choice in different conditions. The gray solid bar of SC represents that the positions of the fixation and the stimulus in the current trial are different from the previous trial, but their relative positions remain consistent (external position change); the black solid bar of RC represents that the position of the fixation in the current trial is different from the previous trial, but the position of the stimulus remains unchanged (retinal position change); the light striped bar BC represents that in the current trial, the position of the stimulus and the relative position between the fixation and the stimulus are different from the previous trial, but the position of the fixation remains unchanged (both external and retinal position changes); the dark striped bar NC represents that the positions of the fixation and the stimulus in the current trial are the same as in the previous trial (identity). Scatter points represent the individual data, and the error line represents the standard error.
***p < 0.001, **p < 0.01.