We tested whether socially anxious individuals perform better in processing facial information with low spatial frequencies (LSFs). For this, we presented socially anxious and nonanxious participants with hybrid face stimuli that contained independent facial expressions in high (HSF) and LSF bands. In two tasks, participants either rated the images according to "angriness" or had to learn how hybrid facial expressions predicted the location of an upcoming target. We found mostly additive effects of LSF and HSF information in the rating task for both groups. In contrast, socially anxious participants showed better prediction performance for LSF expressions in the implicit learning task. We conclude that socially anxious participants are more sensitive to facial information within LSFs, but this higher sensitivity may become mostly evident in indirect tasks.