Adaptations for the visual assessment of formidability: Part I

In the last couple of years there has been an explosion in research on faces and what can be inferred from them. It turns out, for example, that you can predict electoral outcomes from rapid and unreflective facial judgments, that women can (partially) determine a man's level of interest in infants from his face alone, that the facial expression of fear enhances sensory acquisition, and much, much else. A particularly interesting addition to this literature is Aaron Sell and colleagues' paper, "Human adaptations for the visual assessment of strength and fighting ability from the body and face". Sell et. al. hypothesized that human beings possess evolved psychological mechanisms 'designed' to estimate the fighting ability (or physical formidability) of conspecifics - i.e. other Homo sapiens sapiens - from minimal visual information. An ancillary, but important, claim the authors also make is that formidability is largely a function of upper-body strength and thus the latter is a suitable proxy for the former. To summarize for clarity, Sell et. al. claim that: 

• (i) people can estimate the formidability of others from visual cues of their bodies and faces, 
• (ii) this ability is an adaptation, and thus evolved by natural selection, and
• (iii) upper-body strength is the single most important determining factor of fighting ability. 

The authors’ rationale for the first two hypotheses stems from the observation that in social species such as humans, ‘the magnitude of the costs an individual can inflict on competitors largely determines its negotiating position’ (p. 575). That is, formidability is often an important component of an organism’s ability to compete in zero-sum games (notably, access to limiting resources). Given the dangers of physical confrontation, a rapid visual assessment of the formidability of an opponent could be extremely beneficial because it would allow an individual to weigh up its chances of success, and thus choose to fight only when there is a reasonable prospect of victory. Indeed, Sell et. al. note that the widespread so-called ritualized animal contests are best interpreted as joint demonstrations and assessments of formidability, with physical violence usually ensuing only when individuals are closely matched. If the ability to visually estimate a competitor’s formidability was indeed adaptive, and if violence was frequent and recurrent throughout human evolutionary history (as is likely the case), it is not unreasonable to expect natural selection to have forged mechanisms to make such estimates. Sell and his colleagues tested hypothesis (i) empirically in a number of studies and the evidence seems to bear it out overall. While the truth of (ii) is more doubtful, I will argue that, pending further research, it is reasonable to accept it preliminarily for a number of reasons. Finally, I will argue the lack of empirical evidence in the study for (iii) is problematic but not decisively so: it is clear that there is a correlation between upper-body strength and formidability, but we do not know how strong this correlation is so it is difficult to judge how good a proxy the one is for the other.


After the jump, I summarize Sell et. al.'s primary findings (though I leave out one of their experiments). In Part II - coming later in the week - I evaluate their paper.


Broadly speaking, Sell et. al. divided their subjects into two groups: the stimulus subjects (who provided the target photographs and strength measurements) and the judgment subjects (who rated the strength of the men in the target photos). The first group filled out a questionnaire, posed for photographs, had their body measurements taken (e.g. weight, height) and then had their strength measured in a number of ways. The photographs were then standardized and edited, and the target individuals’ upper-body strength measurements were combined to create a composite score for each. The judgment group was then presented with various versions of the photographs and asked to rate the physical strength of the target individuals. Specifically, these subjects were asked to: “Please rate the following [men/bodies/faces] on how physically strong you think the man is compared to other men of his age” on a 7-point Likert scale (1 = very weak, 7 = very strong). In the first study, 59 male undergraduates at the University of California Santa Barbara (UCSB) were recruited at a campus gym, their photos were taken and so on as described above. Each participant’s upper-body strength was then measured on four weight-lifting machines in random order (arm curl, abdominal crunch, chest press, and super long pull) and his lower-body strength was measured with a leg-press. The subjects also posed for two color photographs: (a) full-person, without a shirt, standing next to a male experimenter for scale, and (b) face-only. (Dress was standardized and subjects were asked to keep a neutral expression). The images were then edited to create three sets of photographs: face-only, cropped below the jaw-line (as in the top row of the picture below), full-person (not pictured), and body-only with the face removed (as in the bottom row).

After these stimulus materials were created, an additional 142 UCSB undergraduates (59 female) were asked to rate one set of 59 images (full person, body only or face only). The findings robustly supported the authors’ hypothesis: the average ratings of the men’s strength correlated with their actual upper-body strength at r=0.71 (p=10^-10) for photographs of the whole person, at r=0.66 (p=10^-8) for the body alone, and at r=0.45 (p=0.0003) for the face alone. These results, though, relate to perceived strength, not to perceived formidability itself. So to test whether judgments of strength track judgments of fighting ability, the researchers asked another group of 37 subjects (25 female) to “Please look at the following photographs of men and rate them on how tough each would be in a physical fight – how likely he would be to beat his opponent”. The correlation between the two types of ratings was nearly perfect: r=0.96 (p = 10^-32).

Using hierarchical linear modeling (HLM) Sell et. al. further established that upper-body strength specifically, and not lower-body strength or other features such as height or age, largely determined perceived strength. In the first HLM, the target variable was rated strength and the predictor variables were upper-body strength and leg strength as measured on the leg press. Again, the results supported the authors’ hypothesis: in all cases, the predictive contribution of upper-body strength was large and highly significant, whereas the contribution of lower-body strength was equivocal, modest and of mixed significance. For face only photographs, upper-body strength was γ=0.31 (p=10^-11) but leg strength γ=-0.09 (p=0.003); for body only, upper-body strength was γ=0.44 (p=10^-19) but leg strength γ=0.007 (p=0.81); and for full person photos, upper-body strength was γ=0.41 (p=10^-12) but leg strength γ=0.06 (p=0.03). Clearly, then, perceived strength is more a function of upper-body than lower body strength. The authors also created a HLM that determined to what extent measured upper-body strength, age, weight or height predicted rated strength when controlling for the other three variables. (That is, this HLM showed the independent effect of each variable). The results are summarized in the following table:

 

	Independent effects of target measurements on ratings
Rated Strength	Γ	P
Full person photos
Strength	.50	10-14
Height	.36	10-8
Weight	-.16	.004
Age	.06	10-4
Body alone photos
Strength	.52	10-19
Height	.36	10-9
Weight	-.21	.001
Age	.10	10-7
Face alone photos
Strength	.22	10-8
Height	.05	.10
Weight	.07	.05
Age	.04	.10

Clearly, the two most important components of perceived strength are measured upper-body strength and height (As Sell et. al. note, height determines reach, which is likely an important component of formidability). In each case, however, the independent effect of upper-body strength is statistically significant and larger than any of the other variables. In other words, (a) raters can accurately estimate a man’s upper-body strength from static visual images (significantly, including images of just faces), and (b) perceived strength is not a function of overall body size, but of upper-body strength and, to a lesser extent, height.

The data cited above, however, are limited by the fact that both the stimulus and judgment groups were recruited from the same culture and thus leave the question of how generalizable the findings are open to question. Sell et. al. partially addressed this concern by recruiting two further groups of stimulus subjects from non-US cultures. In the first case, fifty-three adult Tsimane men (a group of semi-sedentary forager-horticulturists who live in the lowlands of Bolivia) were photographed and had various body measurements taken. The photos were edited to present the face only (see below) and a composite score of upper-body strength was created for each using measurements of their chest, shoulder, flexed bicep circumference and handgrip strength. Thirty-two UCSB undergraduates (17 female) were then asked to rate the men in the same way as in the first study.

The second non-US stimulus group was recruited from an Andean population of herder-horticulturalists from the province of Salta in Argentina. Twenty-eight adult men were photographed, had various measurements taken and then a composite score of their upper-body strength was created from their flexed bicep circumference and a direct measure of chest/arm strength. Another twenty-eight USCB students (19 female) then rated the face only photos, again in the same way as in the other studies. The results from both experiments robustly supported the hypothesis that people can make accurate judgments about the formidability of men even from different cultures. Recall that in the first study the correlation between actual strength and strength estimated from facial photographs was 0.45 (p=0.0003); in the Tsimane study the correlation was even stronger at r=0.52 (p=0.0001), and the correlation was similarly high for the Andean stimulus group with r=0.47 (p=0.01). As in the US study, the researchers created HLMs for the Tsimane and Andean groups and again found perceived strength is largely a function of upper-body strength, and not any of the other variables. The independent effect of measured strength (i.e. controlling for height, weight and age) was γ=0.18 (p=0.0003) for the Tsimane subjects, and γ=0.43 (p=10^-5) for the Andean participants. Interestingly, none of the other variables were significant in the Tsimane study, but in the Andean study age had a very strong negative effect with γ= -0.22 (p=0.0002). The overall results, however, are clear: raters can accurately estimate the upper-body strength of men from different cultures based on nothing more than a photograph of his face.


Part II is up...


-------------------------

Sell, A., Cosmides, L., Tooby, J., Sznycer, D., von Rueden, C., & Gurven, M. (2009). Human adaptations for the visual assessment of strength and fighting ability from the body and face Proceedings of the Royal Society B: Biological Sciences, 276 (1656), 575-584 DOI: 10.1098/rspb.2008.1177.