Abstract
Many studies have examined the role of spatial frequencies (SFs) in facial expression perception. However, although their detection and recognition have been proposed to rely on different perceptual mechanisms (Sweeny et al., 2013; Smith & Rossit, 2018), the SFs underlying these two tasks have never been compared. Thus, the present study aimed to compare the SFs underlying the detection and recognition of facial expressions of basic emotions and pain. Here, we asked 10 participants (1400 trials per participant) to decide if a stimulus randomly sampled with SF Bubbles (Willenbockel et al., 2010) corresponded to an emotion or a neutral face. Classification vectors for each emotion were computed using a weighted sum of SFs sampled on each trial, with accuracies transformed in z-scores as weights. We then compared the SFs used in this task to those obtained in a previous study using the same stimuli and method but during a recognition task (Charbonneau et al., 2018). Overall, accurate detection of emotions was significantly associated with the use of low-SFs (ranging from 3.33 to 6 cycles per face (cpf); Zcrit=3.45, p< 0.05). Happiness was the only emotion relying on similar low-SFs for both tasks. Other emotions were associated with the use of higher SFs in the recognition task. Interestingly, the detection of fear (ranging from 1.67 to 7 cpf, peaking at 4 cpf) and surprise (ranging from 1.33 and 6.33 cpf, peaking at 3.33 cpf) was associated with the lowest SF information. These results are consistent with the idea that low-SF represent potent information for the detection of emotions, especially those with a survival value such as fear. However, the contribution of higher SFs is needed to discriminate between emotions for their accurate recognition.