Abstract
Recent studies suggest that low spatial frequencies (SFs) are particularly important for the visual processing of the facial expression of pain (Wang et al., 2015; 2017). However, these studies used arbitrary cut-off to isolate the impact of low (under 8 cycles per faces (cpf)) and high (over 32 cpf) SFs, thus removing any contribution of the mid-SFs. Here we compared the utilization of SFs for pain and other basic emotions in three tasks (20 participants per task), that is 1) a facial expression recognition task with all basic emotions and pain, 2) a facial expression discrimination task where one target expression needed to be discriminated from the others and 3) a facial expression discrimination task with only two choices (i.e. fear vs. pain, pain vs. happy). SF Bubbles were used (Willenbockel et al., 2010), a method which randomly samples SFs on a trial-by-trial basis, enabling us to pinpoint the SFs that are correlated with accuracy. In the first task, accurate categorization of pain was correlated with the presence of a large band of SFs ranging from 4.3 to 52 cpf peaking at 14 cpf (Zcrit=3.45, p< 0.05 for all analysis). In the second task, the correct discrimination of pain was correlated with the presence of a band of SFs ranging from 5 to 20 cpf peaking at 11 cpf. In the third task, we computed the classification vectors for pain-happiness and pain-fear conditions and revealed the overlapping SFs. In this task, SFs ranging from 2.7 to 13 cpf peaking at 7.3 cpf are significantly correlated with pain discrimination. Our results highlight the importance of the mid-SFs in the visual processing of the facial expression of pain and suggest that any method removing these SFs offers an incomplete account of SFs diagnosticity.
Meeting abstract presented at VSS 2018