Fig 1.
Examples of face stimuli and tasks used across the three studies.
In Study 1, participants evaluated 32 faces from the Chicago Face Database for attractiveness (Panel A). Faces within each attractiveness-sex-ethnic category were classified into lighter and darker variants based on forehead luminance estimates. Panel B illustrates examples of lighter (CFD-ID: AF255; MdnLumens = 0.74) and darker (CFD-ID: AF242; MdnLumens = 0.71) highly attractive Asian female targets, with forehead regions from which skin tone was extracted enclosed in red. In Study 2, unprocessed faces were differentially mapped with positive attributes across brief implicit association tests (Panel C). Panel D shows processed faces as they appeared during continuous flash suppression across Study 3. A demo video of the CFS task is available online. Displayed faces are used with permission from Ma et al., 2015 (online).
Fig 2.
Boxplot-violin summaries of mean-centered attractiveness ratings (Panel A), implicit evaluation difference scores (Panel B), and breaking times following continuous flash suppression (Panel C). Asterisks indicate significant (**—p < .01; ****—p < .0001) differences following two-sample (Panels A, C) and one-sample (Panel B) tests. Facet labels across Panels A and C and x-axis ticks across Panel B refer to High-Attractive Female (HAF), High-Attractive Male (HAM), Low-Attractive Female (LAF), and Low-Attractive Male (LAM) target categories, respectively. Within each category, ‘lighter’ and ‘darker’ variants were equally split along target race.
Fig 3.
Response frequencies (y-axis) to six iterations of the survey item. Could you see yourself being attracted to a [gender-sex] label.
Responses are split by participant gender (x-axis). Across studies (rows), males and females primarily responded with “Definitely not” when considering attraction to feminine-males, masculine-females, and non-binary individuals (columns 2, 3, 5, and 6). Male participants predominantly reported a strong attraction to feminine-females, and female participants expressed a similar sentiment towards masculine-males (columns 1 and 4).
Fig 4.
Standardized effect sizes (y-axes) with 95% bootstrapped CIs following Lighter–Darker contrasts across measures (panels) for female (left column) and male (right column) participants.
Effects larger than 0 indicate a Lighter > Darker variant preference. Shaded regions indicate symmetrical ±.1 equivalence bounds. CIs overlapping equivalence bounds and/or null dotted (grey) intercepts show statistically different and/or equivalent outcomes concerning a negligible effect (-.1 < d < .1–49). Dashed (red) intercepts indicate the smallest effect sizes of interest at 60% power following sample-specific sensitivity analyses. Sensitivity thresholds for male respondents were higher as they represent fewer observations (~25% of the total sample identified as male).
Fig 5.
Ridge density plots of mean-centered ratings (x-axes) for individual ‘darker’ (row 1) and ‘lighter’ (row 2) faces. Ratings provided by female, male, and non-binary/other participants are provided across individual columns. The vertical dashed intercept indicates the grand mean (μ = 2.80).