Do low spatial frequencies explain the extremely fast saccades towards human faces?

Area of Science:

• Cognitive Neuroscience
• Visual Perception
• Computational Vision

Background:

• Human face recognition is exceptionally rapid and efficient compared to other object categories.
• Previous research indicates saccadic reaction times (SRT) towards faces can be as fast as 100-110ms.
• Classical visual perception models struggle to account for these extremely fast SRTs.

Purpose of the Study:

• To investigate if low spatial frequency content explains the rapid saccadic eye movements towards human faces.
• To compare saccadic responses to faces, animals, and vehicles under different image filtering conditions.
• To elucidate the role of spatial frequencies in the coarse-to-fine visual processing of faces.

Main Methods:

• A saccadic choice task was employed, presenting two images simultaneously to different visual fields.
• Participants made saccades towards a target image (face, animal, or vehicle) while another image served as a distractor.
• Images were presented in color, grayscale, low-pass filtered, or high-pass filtered to analyze spatial frequency effects on SRT.

Main Results:

• The shortest SRTs were consistently observed for saccades directed towards faces, outperforming those towards animals or vehicles.
• Participants initiated correct saccades towards faces within 130-140ms for unfiltered and low-pass filtered images.
• High-pass filtered images significantly increased SRT for faces, while low-pass and high-pass filtering slowed responses to animals and vehicles.

Conclusions:

• Extremely fast saccadic eye movements towards human faces are likely initiated by their prominent low spatial frequency components.
• The findings support a coarse-to-fine visual processing model, where initial rapid responses rely on low-resolution image information.
• Spatial frequency content plays a critical role in modulating the speed of visual categorization and subsequent eye movements.