Optimal eye-gaze fixation position for face-related neural responses

Abstract

It is generally agreed that some features of a face, namely the eyes, are more salient than others as indexed by behavioral diagnosticity, gaze-fixation patterns and evoked-neural responses. However, because previous studies used unnatural stimuli, there is no evidence so far that the early encoding of a whole face in the human brain is based on the eyes or other facial features. To address this issue, scalp electroencephalogram (EEG) and eye gaze-fixations were recorded simultaneously in a gaze-contingent paradigm while observers viewed faces. We found that the N170 indexing the earliest face-sensitive response in the human brain was the largest when the fixation position is located around the nasion. Interestingly, for inverted faces, this optimal fixation position was more variable, but mainly clustered in the upper part of the visual field (around the mouth). These observations extend the findings of recent behavioral studies, suggesting that the early encoding of a face, as indexed by the N170, is not driven by the eyes per se, but rather arises from a general perceptual setting (upper-visual field advantage) coupled with the alignment of a face stimulus to a stored face template.

Figure 1. Illustration of the experimental protocol.

(a) Schematic view of a trial progression during the eye-gaze contingent paradigm used in this study. On the beginning of each trial, a fixation-cross is presented randomly on the screen, and is replaced by a face image as soon as the eye-tracker detects that the observer's line of sight lays on the fixation-cross. The yellow circles represent the observer's gaze position on the screen. Eye-fixation landing position on the seven predefined fROI (b) is controlled by the presentation of a fixation-cross located randomly in one of nine invisible quadrants on the screen, that were spatially mapped with the location of the fROIs (top). Actual gaze positions compiled over all subjects and trials for each face orientation (bottom). Gaze position for each trial is depicted with a black dot. (c) Saliency map and spatial resolution of face images and visual field map coverage when the centre of gravity of the face is fixated.

Figure 2. N170 ERP responses over PO7 and PO8 scalp sites are shown as a function of eye-gaze landing positions on face stimuli.

(a.) N170 responses recorded at fixated face regions of interest (fROIs) in upright and inverted face images. (b.) N170 responses elicited by fixated face regions along the vertical meridian (upper in green, middle in red, and lower in blue) in upright and inverted face images. (c.) N170 responses elicited by fixated face regions along the horizontal meridian (left and right visual fields).

Figure 3. Heat maps for upright and inverted faces.

Values of the heat maps indicate the relative strength of N170 response as a function of viewpoint (left). Bootstrap analysis was applied on the heat maps up to a p = 0.05 significance threshold (right).

Figure 4. Sensitivity of the P100 waveform to fixation position.

(a) P100 ERP responses to specific eye-gaze landing positions (fROI) are shown at occipital scalp sites (O1/O2) separately for upright and inverted faces. (b) P100 ERP waveforms recorded over O1 (in blue) and O2 (in red) electrodes are shown for fixated face regions along the horizontal meridian (left and right visual fields). (c) P100 ERP waveforms elicited by fixated face regions along the vertical meridian (upper in green, middle in red, and lower in blue).

Figure 5. ERP face inversion effects.

(a.). Pooled ERP responses elicited by all fROIs show the typical effect of face inversion on the N170. (b.). Scalp ERP difference (upright minus inverted faces) and ERP waveforms over POz are shown at P100 ERP latency range to foveated nasion (left) and mouth (right) features. (c.) N170 Face inversion effects for fixated face regions along the vertical meridian (upper, middle, and lower visual fields). (d.) N170 Face inversion effects for fixated face regions along the horizontal meridian (left and right visual fields).

Figure 6. Reconstruction of the time course of sources active during the N170 evoked potential using the MEM technique.

Bilateral sources found in the inferior temporal gyrus exhibit higher amplitude in response to fixations at the eyes of upright faces, and at the mouth of inverted faces.