Abstract
Detecting and identifying faces is crucially important for social functioning and decision making. Previous studies demonstrated that the perception of a face in the present moment is biased toward previously seen faces, due to serial dependence (e.g., Liberman, Fischer, & Whitney, 2014). This suggests that the visual system takes into account temporal redundancy inherent in the environment to compensate for noisy and ever-changing sensory inputs. However, each observer might have their own manner or pace with which they interact with others, which could lead to different temporal statistics of faces that they encounter. In the present study, we investigated whether the serial dependence in face perception is idiosyncratic for each observer. In the experiment, a target face drawn from a continuous morph of facial identities was briefly presented, and then participants were required to adjust a response face to match the target face. We tested a total of 55 participants online over two different sessions, and each participant completed 400 trials in each session. To investigate the stability of serial dependence, for each session, we plotted the response error as a function of the difference between faces in the previous and current trials. Correlations of response error patterns across sessions were higher within each participant than between different participants (p=.004, permutation test), suggesting stable serial dependence of face perception within each observer. This result is consistent with high test-retest reliability of serial dependence in orientation perception (Kondo, Murai, & Whitney, 2022). Furthermore, this correlation measure is consistent with traditional measures of serial dependence such as fitting data to derivative of von Mises distribution (r=.51; p<.001). Together, the present study provides evidence for individual differences in serial dependence of face perception, as well as a novel method to reliably estimate the serial dependence from noisy online behavioral data.