Fear conditioning prompts sparser representations of conditioned threat in primary visual cortex

Threat detection and valuation is an important function of the visual system. It has been proposed that repeated exposure to threatening stimuli alters sensory responses. Our previous study has found that neural representations of conditioned threat in the primary visual cortex become sparser with aversive learning. We examined this issue further in this study. Simultaneous fMRI and EEG data were recorded from 18 participants viewing the random appearance of two Gabor patches with the 45-degree Gabor patch (CS+) occasionally paired with a loud scream (US) (25% reinforcement rate) and the 135-degree Gabor patch (CS-) never paired with the US. Applying the MVPA decoding method to fMRI data in a sliding trial window fashion, we found that (1) decoding accuracy between CS+ and CS- was significantly higher than chance level in all visual areas throughout the conditioning session, (2) as aversive learning progressed, the number of voxels contributing to the representation of CS+ decreased in primary visual cortex but not in other visual areas, and (3) the number of voxels contributing to the representation of CS- remained the same. Analyzing the concomitantly recorded EEG data, we found that (1) the latency of the event-related potential evoked by CS+ became progressively shorter with aversive learning and (2) the latency of the event-related potential evoked by CS- became progressively longer. These results confirmed that the neural representation of conditioned threat became sparser in the primary visual cortex and suggested that the sparsified representation facilitated the detection and evaluation of threat.