Bad Communication: Strength and Attractiveness

measured strength/non-facial physical attractiveness is just r = .32 in men

I am referring to this study in the article:

• Sell, Aaron; Lukazsweski, Aaron W.; Townsley, Michael (2017). Cues of upper body strength account for most of the variance in men's bodily attractiveness.

Evolution equips sexually reproducing species with mate choice mechanisms that function to evaluate the reproductive consequences of mating with different individuals. Indeed, evolutionary psychologists have shown that women's mate choice mechanisms track many cues of men's genetic quality and ability to invest resources in the woman and her offspring. One variable that predicted both a man's genetic quality and his ability to invest is the man's formidability (i.e. fighting ability or resource holding power/potential). Modern women, therefore, should have mate choice mechanisms that respond to ancestral cues of a man's fighting ability. One crucial component of a man's ability to fight is his upper body strength. Here, we test how important physical strength is to men's bodily attractiveness. Three sets of photographs of men's bodies were shown to raters who estimated either their physical strength or their attractiveness. Estimates of physical strength determined over 70% of men's bodily attractiveness. Additional analyses showed that tallness and leanness were also favoured, and, along with estimates of physical strength, accounted for 80% of men's bodily attractiveness. Contrary to popular theories of men's physical attractiveness, there was no evidence of a nonlinear effect; the strongest men were the most attractive in all samples.

I have seen this promoted in manosphere circles and even more respectable circles - which is surprising given how badly communicated the results are. Ultimately, the study found that 70% of the variance in bodily attractiveness in men can be predicted by estimated strength - that’s a pretty bold claim! For reference, intelligence can only explain 31.3% of the variance in educational attainment. This article has multiple issues, unfortunately.

First of all, the study claims that the men’s faces were obscured when the women were doing the ratings:

The men whose bodies were photographed (henceforth subjects) were taken from two databases of young college students at US universities reported in Sell, Tooby & Cosmides [17] (herein Set 1) and Lukaszweski & Roney [33] (herein Set 2). In both sets of photographs, the subjects’ faces were obscured. See figure 1 for sample photographs. Set 1 photos included both front and side views; Set 2 photos had only front views.

I guess you could defend this based on the fact they are examining “bodily” attractiveness, but it is still rather misleading. I prefer the term “non-facial physical attractiveness” - yes it sounds autistic and pedantic, but it is much more accurate and honest.

Second, this study is doing the classic causation=correlation fallacy, when there are clearly two ways you can interpret these results:

A. a halo effect where attractive men are judged as strong as well

B. strength itself is a component of attractiveness.

Given our very strong priors about the halo effect, you should expect it to be a casual variable even without evidence that it applies in this particular situation. Fortunately, the study itself contains evidence that both A and B are true to a degree.

The researchers measure two variables - rated strength and measured strength. These correlate very weakly - only at .52 or so according to their own study. Notably, if you try to predict actual strength using strength ratings and attractiveness, attractiveness negatively predicts measured strength when you control for strength ratings. This is because attractiveness itself confers a halo effect which causes people to rate them as stronger independent of their true level of strength.

In addition, measured strength did not correlate that strongly with attractiveness at all (pooled correlation = .32):

attractiveness correlated with actual measured strength at r = 0.38 (Set 1: front), 0.39 (Set 1: side) and 0.25 (Set 2: front), all p , 0.01 (compare these numbers with those of rated strength in table 2). In other words, despite the extremely high correlation between attractiveness and ratings of strength, actual strength was still better predicted by ratings of strength rather than attractiveness

The strength measurements used (four weight-lifting machines measuring upper body strength in set 1, chest compression and grip strength in set 2) aren’t great, but even if you make some generous assumptions regarding validity, you’re looking at an increase from .32 to .45 or so, which isn’t particularly impressive when you take into account that the face of the men wasn’t even shown.

Try to think of it this way - if an average man worked his way up to being in the 99th percentile of strength - he would reach the 77th percentile in overall attractiveness if you take the .32 correlation at face value, while if you take the .84 correlation at face value, the average man would find himself in the 97th percentile of attractiveness instead. To me, the former conversion is much more plausible than the latter.