Fig 1.
The African face (left) and the European face (right) have the same average luminance and are closely matched in size (e.g., in Experiment 2: mean pixels’ brightness or Lmean = 112.5, SD = 41, in HSL/RGB coordinates).
Because the perceived difference in skin tone persists even after the observer is informed that the two face images deliver the same average intensity of light to the eyes, such a visual demonstration has been included in the “library” of optical illusions and it is often referred to as the Face Race Illusion (source: “Distortions in the perceived lightness of faces: the role of race categories,” by Daniel T. Levin and Mahzarin R. Banaji, 2006, Journal of Experimental Psychology: General, vol. 1358:4).
Fig 2.
Stimuli used in the crowdsourcing, internet-based, Experiment 1.
The first, from top, row corresponds to the standard stimuli (Lmean = 112.7, SD = 41), one version with African face on the left or right side of the pair on the right side there are the same stimuli rotated 180° or ‘upside-down’ faces. The second and third rows show the European face paired with a 50% face morph of the European and African face and the African face paired with the same 50% face morph, plus the inverted stimuli. All face stimuli were equiluminant except the face pairs in the fourth bottom row which shows “catch trials”, with the face morph presented twice (as the first and last trial) but one version being visibly darker than the other stimulus (Lmean = 91.8, SD = 32).
Table 1.
Analyses of choices in the lightness comparison task.
Fig 3.
Local regions of brightness difference between the African and European face stimuli used in Experiment 2.
Brighter regions indicate more luminance in the African’s than the European’s facial image.
Fig 4.
Timeline of one trial in Experiment 2: Participants maintained gaze within the red ellipse at all times.
The red ellipse worked as a cue for keeping gaze (every 4 secs) within one of the regions containing only one of the two faces. A trial began with a neutral blank gray image, equiluminant to the average brightness of the successive faces image, and serving as a baseline image to compute pupil changes in an event-related manner (by subtracting the mean pupil diameter during viewing of each baseline image from the mean pupil diameter when subsequently attending a face stimulus).
Fig 5.
Mean % fixation durations (Y axis) within different Faces regions (AOI) for the three groups of participants when viewing each upright face: Africans (blue columns), Asians (red columns), and Europeans (green columns).
Error bars represent Standard Errors. In the inlay the face regions corresponding to each AOI (the colors are arbitrary) are shown superimposed to the “African” face.
Fig 6.
Mean % fixation durations (Y axis) or dwell time within different faces parts (Eyes, Nose) when viewing the upright African face (blue) and European face (red).
Error bars represent Standard Errors.
Fig 7.
Left panels: Mean baseline-corrected pupillary diameters of Asian participants (Top panel; N = 24), European participants (Middle panel; N = 38) and African participants (Bottom panel; N = 22), evolving over a time period of 4 seconds from onset of the cue around a face, i.e. while attending the “African” face (red line) or the “European” face (blue line). The colored stripes indicate 95% confidence intervals for within-subject comparisons. Right panels: Functional t -tests of the difference between pupils when attending to the African versus European face. The solid (red) horizontal lines represent the two-tailed critical value for t.
Fig 8.
The top centered pair of faces corresponds to the typical FRL illusory stimuli, where the African and European faces are equiluminant (in this case: Lmean = 110.7, SD = 42). Below, the left column shows from the top Step -1, -2, and -3 in luminance of the African face (i.e., Lmean = 105, SD = 42; Lmean = 101, SD = 41; Lmean = 96, SD = 42, respectively) while the luminance of the European face is kept unchanged. The right column shows from the top Step +1, +2, and +3 in luminance also of the African face only (i.e., Lmean = 118, SD = 46; Lmean = 125, SD = 48; Lmean = 133, SD = 51, respectively). In the experiment, there were an equal number of trials in which the European face’s luminance was similarly manipulated while the African face was not and horizontally-flipped version were shown, so that each ethnicity would be equally likely to be either lighter/darker than the other or appearing to the left/right.
Fig 9.
Experiment 3: Mean % accuracy in judging the African face as either brighter or darker or of same brightness as the European face.
Bars represent standard errors.