Individual differences in human gaze behavior generalize from faces to objects

Abstract

Individuals differ in where they fixate on a face, with some looking closer to the eyes while others prefer the mouth region. These individual biases are highly robust, generalize from the lab to the outside world, and have been associated with social cognition and associated disorders. However, it is unclear, whether these biases are specific to faces or influenced by domain-general mechanisms of vision. Here, we juxtaposed these hypotheses by testing whether individual face fixation biases generalize to inanimate objects. We analyzed >1.8 million fixations toward faces and objects in complex natural scenes from 405 participants tested in multiple labs. Consistent interindividual differences in fixation positions were highly inter-correlated across faces and objects in all samples. Observers who fixated closer to the eye region also fixated higher on inanimate objects and vice versa. Furthermore, the inter-individual spread of fixation positions scaled with target size in precisely the same, non-linear manner for faces and objects. These findings contradict a purely domain-specific account of individual face gaze. Instead, they suggest significant domain-general contributions to the individual way we look at faces, a finding with potential relevance for basic vision, face perception, social cognition, and associated clinical conditions.

Keywords: domain-general; face; gaze; individual differences; object.

Fig. 1.

Vertical fixation positions in faces and objects. (A and B) Scatter plots for consistency correlations for vertical (A) face (r = 0.97, P < 0.001) and (B) object (r = 0.94, P < 0.001) fixation positions. (A and B) Each scatter point represents the average vertical fixation position of a single observer. Vertical fixation position was computed relative to the extent of the target along the vertical image axis, which coincided with the vertical meridian of the observer (chin and forehead rest; top-most pixel of the object corresponding to 1 and bottom-most to 0). For faces, only the inner face region was considered, and 0 and 1 corresponded to the bottom of the chin and hairline, respectively. (C) Correlation between the relative vertical positions of fixations landing in faces and objects (r = 0.69, P < 0.001). Each scatter point shows the average vertical fixation position of one observer within faces and objects. (A–C) The black lines depict the linear least-squares fit for each scatter plot. The color of scatter points always corresponds to the individual height of fixations in faces (red-blue from bottom to top). Note that the axis scales correspond to the relative height of fixations. In absolute terms, the spread on faces and objects was highly similar when considering target size (Fig. 3). (D) Examples of fixations landing on faces and objects for two observers (blue and red, respectively). Fixations in blue and red originate from participants who tended to fixate higher up and lower down, respectively. Blurring of the eye region for publication purposes only.

Fig. 2.

Scaling of individual differences with target size. (A and B) Each line shows the individual vertical deviation from the group median in degrees visual angle for one observer and across different (A) face and (B) object size bins. Faces and objects were divided into five bins containing an equal number of targets, based on their vertical extent. The individual deviation of vertical fixation positions from the group median was calculated separately for each target and then averaged for each observer and bin. (C) The heatmap depicts all intercorrelations between face and object size bins (all r > 0.4, all P < 0.001). Colors correspond to correlation values as indicated by the color bar to the right.

Fig. 3.

Scaling of inter-individual spread with target size. Each scatter point shows the estimated spread of individual vertical fixation positions for a given (A) object or (B) face as a function of its vertical extent. This analysis was limited to targets with a vertical extent <8 degrees visual angle and fixated by at least 30% of participants. Interindividual spread was estimated as five times the median absolute deviation of vertical fixation position, roughly corresponding to the 95% interval. (A) The black line shows the best-fitting second-degree polynomial to the size-spread relationship for objects (R² = 52%). (B) The size-spread relationship of faces (orange) was remarkably similar to that of objects (blue) and well described by the same, object-fitted polynomial (R² = 58%).