Gaze direction as a facial cue of memory retrieval state

Abstract

Gaze direction is a powerful social cue that indicates the direction of attention and can be used to decode others' mental states. When an individual looks at an external object, inferring where their attention is focused from their gaze direction is easy. But when people are immersed in memories, their attention is oriented towards their inner world. Is there any specific gaze direction in this situation, and if so, which one? While trying to remember, a common behavior is gaze aversion, which has mostly been reported as an upward-directed gaze. Our primary aim was to evaluate whether gaze direction plays a role in the inference of the orientation of attention-i.e., external vs. internal-in particular, whether an upward direction is considered as an indicator of attention towards the internal world. Our secondary objective was to explore whether different gaze directions are consistently attributed to different types of internal mental states and, more specifically, memory states (autobiographical or semantic memory retrieval, or working memory). Gaze aversion is assumed to play a role in perceptual decoupling, which is supposed to support internal attention. We therefore also tested whether internal attention was associated with high gaze eccentricity because the mismatch between head and eye direction alters visual acuity. We conducted two large-sample (160-163 participants) online experiments. Participants were asked to choose which mental state-among different internal and external attentional states-they would attribute to faces with gazes oriented in different directions. Participants significantly associated internal attention with an upward-averted gaze across experiments, while external attention was mostly associated with a gaze remaining on the horizontal axis. This shows that gaze direction is robustly used by observers to infer others' mental states. Unexpectedly, internal attentional states were not more associated with gaze eccentricity at high (30°) than low (10°) eccentricity and we found that autobiographical memory retrieval, but not the other memory states, was highly associated with 10° downward gaze. This reveals the possible existence of different types of gaze aversion for different types of memories and opens new perspectives.

Keywords: autobiographical memory retrieval; eye movements; facial cues; gaze aversion; internal attention; social cues.

Figure 1

Stimuli presented in Experiments 1a and 1b.

Figure 2

(A) Polar graph in which each angle represents a gaze direction. For each gaze direction, the radii of the bubbles represent the frequency at which a given mental state (see colors) was associated with the given gaze direction during Experiment 1a. The size of the bubbles depends on the frequency of the responses. Note that the position of the bubbles along the y-axis has no meaning and was chosen to avoid overlapping and allow better visualization. The bubbles in the center of the graph (coordinates 0,0) represent the answers for the question where the stimulus was presented with straight-ahead gaze. (B) Polar graph representing the frequencies at which gaze directions (see angles) were associated with mental states (see colors) during Experiment 1b. (C) Example of a stimulus, illustrating the correspondence between angles in polar coordinates and gaze directions. (D) For the Experiment 1a, one bar plot per gaze direction where the bars represent the probability estimates from the multinomial logistic regression for each mental state. The dashed gray lines represent chance level (i.e., 0.11). (E) For experiment 1b, one polar plot per mental state where the probability estimates from the multinomial logistic regression are represented for each gaze direction (chance level at 0.10). The legend shows the correspondence between colors and mental states. Shades of blue were used for internal attention, shades of red for external attention, shades of green for control proposals, and black for “Do not know” responses. Hatching has been added to make the figure color-blind friendly.

Figure 3

Examples of stimuli presented in Experiment 1b. (A) female face with gaze at low eccentricity (direction 45°, up left); (B) female face with gaze at high eccentricity (direction 45°, up left); (C) male face with gaze at low eccentricity (direction 225°, down right); (D) male face with gaze at high eccentricity (direction 225°, down right).

Figure 4

Polar graphs in which each angle represents a gaze direction. (A) For each gaze direction, the radii of the bubbles represent the frequency at which a given mental state (see colors) was associated with the given gaze direction during Experiment 2 for the condition with low gaze eccentricity (for examples of stimuli, see Figure 2); (B) For each gaze direction, the bubble radii represent the frequency at which a given mental state (see colors) was associated with the given gaze direction during Experiment 2 for the condition with high gaze eccentricity. (C) Low eccentricity condition. The bars represent the probability estimates from the multinomial logistic regression for each mental state. The dashed gray lines represent chance level (i.e., 0.11). (D) High eccentricity condition. For graph conventions and legend, see Figure 2.