Serial dependencies between form orientation and motion direction are asymmetric

Expanding the V1-MT model to the estimation of perceived fluid direction

Humans can readily perceive the direction of liquid flow, yet computational modeling of this process remains challenging due to the complexity of non-rigid motion. Previous models based on neural activities in the primary visual cortex (V1) and the middle temporal area (MT) have been effective in explaining rigid motion perception. In this study, we extend the V1-MT model to address the perception of liquid flow direction. Participants observed video clips of liquid flow and reported the perceived direction, while the V1-MT model was used to predict these perceptions. The winner-take-all approach failed to accurately capture the observed perceptions. In contrast, a weighted mean of directional energies yielded strong predictions, highlighting that the human visual system spatially integrates directional energies from non-rigid motion components. These findings broaden the applicability of the V1-MT model to non-rigid motion and provide insights into how the visual system bridges the gap between computational models of rigid and non-rigid motion perception.

Effects of attention on the asymmetric serial dependences between form and motion patterns and their computational processes

Recent studies have revealed that serial dependences are asymmetric in the estimation of the focus of expansion (FoE) in the global static form and dynamic optic flow displays. In the current study, we conducted two experiments to examine whether and how attention affected the serial dependences between the two displays. The results showed that when all attentional resources are allocated to the FoE estimation task, the serial dependence of the form FoE estimation on the previous flow FoE (SDE flow-form ) still existed even as the flow FoE was 40°, while the serial dependence of the flow FoE estimation on the previous form FoE (SDE form-flow ) disappeared as the form FoE was beyond 30°. When attentional resources are distributed by other tasks, the SDE flow-form tended to be stronger than the SDE form-flow . Therefore, the SDE flow-form and SDE form-flow are asymmetric regardless of observers' attentional states. Finally, we developed two Bayesian models to address the computational mechanism underlying the attentional effects. Both models proposed that attention modulated the certainty of sensory representations of currently presented features. In addition, the effects of working memory on previously presented features were considered in one model. The results showed that the Bayesian inference model that included working memory predicted participants' performances better than the model without considering working memory. In summary, the current study demonstrated that attention and working memory affected the serial dependences between form and flow displays, and the effects could be quantitatively predicted by Bayesian inference models.

Recent, but not long-term, priors induce behavioral oscillations in peri-saccadic vision

Perception of a continuous world relies on our ability to integrate discontinuous sensory signals when we make saccadic eye movements, which abruptly change the retinal image. Here we investigate the role of oscillations in integrating pre-saccadic information with the current sensory signals. We presented to participants (N = 24) a brief pre-saccadic Gabor stimulus (termed the inducer) before voluntary 16° saccades, followed by a test Gabor stimulus at various times before or after saccadic onset. Orientation judgments of the test stimulus were biased towards the orientation of both the inducer and previous (1-back) test stimulus, consistent with serial dependence. In addition to the average bias, judgments oscillated in synchrony with saccadic onset at alpha frequencies (~9.5 Hz) towards the orientation of the inducer or 1-back stimulus. There was also a strong bias towards the mean orientation (central tendency): however, that bias was constant over time, not associated with saccade-synched oscillations. Perceptual oscillations in serial dependence (but not central tendency) suggest that alpha rhythms may be instrumental in communicating short-term (but not long-term) perceptual memory across saccades, helping to preserve stability during saccades. The distinction between the modes of communicating short- and long-term memory suggests that the two phenomena are mediated by distinct neuronal circuitry.

Structural disorder facilitates future memory decisions

It is well known that perception and cognition are systematically biased towards the recent past. That is, a decision about the current state of a perceptual feature (e.g. orientation) can be predicted based on a recent state of the same feature. Such serial dependencies have been demonstrated across perception, memory and cognition, and have been jointly attributed to an adaptive mechanism meant to promote stability in a constantly changing environment. Here, we argue that this adaptive mechanism prioritizes past information on the most basic structural level, such that the strength of the attractive bias is modulated by the amount of structural coherence in stimuli. We presented visual patterns of varied structural disorder (randomness) prior to a recognition memory decision that required discriminating between trained and novel visual patterns. Both highly generic geometrical shapes and completely random patterns failed to elicit an effect on decisional response times. By contrast, we found recognition memory decisions to be significantly faster in trials where the irrelevant probe pattern was 'optimally' random. This result suggests that decision-making is influenced by the past's informational worth. More importantly, it suggests an optimal amount of uncertainty to facilitate future decisions.

Conflicts between short- and long-term experiences affect visual perception through modulating sensory or motor response systems: Evidence from Bayesian inference models

The independent effects of short- and long-term experiences on visual perception have been discussed for decades. However, no study has investigated whether and how these experiences simultaneously affect our visual perception. To address this question, we asked participants to estimate their self-motion directions (i.e., headings) simulated from optic flow, in which a long-term experience learned in everyday life (i.e., straight-forward motion being more common than lateral motion) plays an important role. The headings were selected from three distributions that resembled a peak, a hill, and a flat line, creating different short-term experiences. Importantly, the proportions of headings deviating from the straight-forward motion gradually increased in the peak, hill, and flat distributions, leading to a greater conflict between long- and short-term experiences. The results showed that participants biased their heading estimates towards the straight-ahead direction and previously seen headings, which increased with the growing experience conflict. This suggests that both long- and short-term experiences simultaneously affect visual perception. Finally, we developed two Bayesian models (Model 1 vs. Model 2) based on two assumptions that the experience conflict altered the likelihood distribution of sensory representation or the motor response system. The results showed that both models accurately predicted participants' estimation biases. However, Model 1 predicted a higher variance of serial dependence compared to Model 2, while Model 2 predicted a higher variance of the bias towards the straight-ahead direction compared to Model 1. This suggests that the experience conflict can influence visual perception by affecting both sensory and motor response systems. Taken together, the current study systematically revealed the effects of long- and short-term experiences on visual perception and the underlying Bayesian processing mechanisms.

Attention influences the effects of the previous form orientation on the current motion direction estimation

Recent studies have found that the estimates of motion directions are biased toward the previous form orientations, showing serial dependence, and the serial dependence does not involve cognitive abilities. In the current study, we conducted two experiments to investigate whether and how attention-a cognitive ability-affected the serial dependence. The results showed that serial dependence was present in the current study, reproducing the previous findings. Importantly, when the attentional load reduced the reliability (i.e., estimation accuracy and precision) of previous form orientations (Experiment 1), the serial dependence decreased, meaning that the biases of motion direction estimates toward previous form orientations were reduced; in contrast, when the attentional load reduced the reliability of current motion directions (Experiment 2), the serial dependence increased, meaning that the biases of motion direction estimates toward previous form orientations were increased. These trends were well consistent with the prediction of the Bayesian inference theory. Therefore, the current study revealed the involvement of attention in the serial dependence of current motion direction estimation on the previous form orientation, demonstrating that the serial dependence was cognitive and the attentional effect can be a Bayesian inference process, initially revealing its computational mechanism.