Yen-Ju Chen¹ (), Zitang Sun¹, Shin’ya Nishida¹; ¹Graduate School of Informatics, Kyoto University

Temporally asynchronous changes in visual features induce visual spatial segmentation. We investigated the temporal dynamics of this phenomenon, particularly the temporal window for computing temporal correlation, using continuous tracking. Participants tracked by mouse a Brownian motion of the target. The temporal impulse response from perception to action was estimated from the cross-correlogram (CCG) of the trajectory between the target and mouse. The stimulus was a texture made of numerous Gaussian bulbs. Each bulb subtended 0.5 deg in diameter. Within the target area, the luminance or color (Red-Green or Blue-Yellow) of the bulb was temporally modulated with an interelement correlation of ~0.9. The background texture consisted of (1) static bulbs or (2) dynamic bulbs with an inter-element correlation of ~0.9 within the background, while the target-background correlation was ~0.0. Note that the target could be segregated from the background by temporal changes in the static condition, while by temporal asynchrony in the dynamic condition. We hypothesized that the difference in CCG between these two conditions must reflect additional temporal processing for correlation computation. The estimated CCGs (n=3) indicated slower and broader impulse responses for the dynamic condition than for the static condition. For luminance stimuli, the peak latency and half-height bandwidth were 333 ms and 209 ms in the static condition and 704 ms and 733 ms in the dynamic condition, respectively. Similar trends were observed for the color conditions. This temporal difference reflects an additional low-pass filtering in the dynamic CCG, providing an estimate of the temporal correlation window. Assuming a linear cascade processing system, the dynamic CCG is the static CCG convolved with the correlation processing time course. A mixed Gaussian kernel fit indicated that the time window for computing the correlation had a peak latency of 257 ms and a bandwidth of 753 ms.

Acknowledgements: This study is supported by SPRING, Kakenhi 24H00721, 20H05957