Purchase this article with an account.
Charles Or, Talia Retter, Bruno Rossion; The effect of head orientation on face detection in natural images as evidenced by fast periodic visual stimulation. Journal of Vision 2015;15(12):690. doi: https://doi.org/10.1167/15.12.690.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Face detection is typically fast and accurate, but some faces are detected faster and better than others in natural images. While some studies have shown that low-level visual cues contribute to rapid face detection, there is less information concerning the contribution of high-level factors. Here, we tested whether face detection differs for full-front vs. 3/4 head views using fast periodic visual stimulation (FPVS). This approach can isolate cortical responses for the detection of faces – as compared to other object categories – objectively, rapidly and implicitly in natural images, without contribution from low-level visual cues (Rossion et al., 2015). High-density electroencephalogram (EEG) was recorded from 16 observers presented with 40-s sequences containing natural images of objects sinusoidally contrast-modulated at a fixed rate of 12.0 Hz (F). Natural face images were introduced at fixed intervals of F/9 (‘oddballs’, 1.33 Hz). In Condition 1, all faces were posed full-front. In Condition 2, they were at 3/4 views (either left or right, but only in one orientation within a sequence). In Condition 3, the oddball alternated between full-front and 3/4 views throughout the sequence (frequency of each view, F/18 = 0.67 Hz). In all conditions, significant responses were recorded at 1.33 Hz and its harmonics, mainly over the right occipito-temporal region, indicating a high-level face-detection response. Interestingly, Condition 3 also showed significant responses at 0.67 Hz and its harmonics over the same cortical region, implying a differentiable face-detection response to full-front vs. 3/4 views. Analysis in the time domain revealed a sequence of face-selective components, with peak latencies approximately 12 ms earlier for full-front than 3/4 views emerging as early as the first face-selective component (130-150 ms). These findings indicate that a full-front view presents an advantage in face detection, arising in part from a faster high-level brain response.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only