# Serial dependence in face-gender classification revealed in low-beta frequency EEG

## TL;DR

This study shows that past face-gender perceptions influence current ones through beta-frequency brain activity, with specific frequencies linked to male and female biases.

## Contribution

The study identifies specific beta-frequency EEG components (14 Hz and 18 Hz) associated with serial dependence in face-gender perception.

## Key findings

- Serial dependence in face-gender perception is strongest in the beta frequency range (14–20 Hz).
- EEG classification accuracy correlates with the strength of serial dependence in individual participants.
- Beta frequencies of 14 Hz and 18 Hz are specifically linked to serial biases for female and male faces, respectively.

## Abstract

Perception depends not only on current sensory input but is also heavily influenced by the immediate past perceptual experience, a phenomenon known as “serial dependence,” particularly robust in face perception.

We measured discrimination of face-gender in participants to a sequence of intermingled male, female, and androgynous images, while recording EEG responses. The discriminations showed strong serial dependence (androgynous images biased towards male when preceded by male and female when preceded by female). The strength of the bias oscillated over time in the beta range, at 14 Hz for female prior stimuli, 18 Hz for male. Using classification techniques, we were able to successfully classify the previous stimulus from current EEG activity. Classification accuracy correlated well with the strength of serial dependence across individual participants, confirming that the neural signal from the past trial biased face perception. Bandpass filtering of the signal within the beta range showed that the most useful information to classify gender was around 14 Hz when the previous response was “female,” and around 18 Hz when it was “male,” reinforcing the psychophysical results showing serial dependence to be carried at those frequencies.

Overall, the results suggest that recent experience of face-gender is selectively represented in beta-frequency (14–20 Hz) spectral components of intrinsic neural oscillations.

The neurophysiological mechanisms of how past perceptual experience affects current perception are poorly understood. Using classification techniques, we demonstrate that the response to gender of the previous face image of a sequence can be decoded from the neural activity of the current EEG response, showing that relevant neural signals are maintained over trials. Classification accuracy was higher for participants with strong serial dependence, strongly implicating these signals as the neural substrate for serial dependence. The best information to classify gender was around 14 Hz for “female” faces, and around 18 Hz for “male,”, reinforcing the psychophysical results showing serial dependence to be carried at those beta -frequencies.