Do These Points Form a Curve?

I was interested to browse through a paper by Buunk et al. in the most recent issue of Evolution and Human Behavior in which the authors report the results of psychological experiments exploring the differential relationship between height and sexual jealousy in women and men. The authors predicted that (self-reported) sexual jealousy would decline with increasing height in men and that women of average height would report the lowest levels of sexual jealousy. The theory driving these predictions is that higher-status, more attractive individuals should be less jealous on average because they are better able to prevail over would-be competitors and, presumably, if they experience partner infidelity, they can always find another partner. The authors cite the abundant evidence for increased social dominance in taller men and suggest the relationship between women's attractiveness and height is quadratic, with women of average height being most attractive. One hundred women and 100 men were asked question, “In general, how jealous are you in your current relationship?” Responses fell on a six point scale ranging from (1) "not jealous" to (6) "morbidly jealous". The authors' results apparently support their hypotheses. So here are the two figures that they use to show that (1) jealously declines linearly with height in men and (2) is quadratic for women, with average-height women least jealous. The first figure is for men:

Buunk et al. (2008) Figure 1

The second figure is for women:

Buunk et al. (2008) Figure 2

Hmmm. I don't know if I would rest much on the interpretation of that figure as being "quadratic." It seems entirely possible that the curve is driven simply by the sparseness of the tails. There are fewer women of extreme height, either tall or short and this allows a few influential points to leverage the line up at the ends. Think about the upper 95% confidence interval of a linear regression line. Doesn't look that different from their figure 2, no? This makes me wonder how robust the relationship is. For example, if we were to bootstrap replicate samples (with replacement) and re-fit the quadratic form, how many would have a significant at some conventional level (e.g., p<0.05)? There is also the question of whether this quadratic curve fits better than a linear relationship. One could test the two nested models using a likelihood ratio test.

Then there is the question of confounding variables. At the very least, it seems that one would want to control for age of the actors, duration of relationship, and quite possibly other measures of wealth or status. It seems reasonable to posit that being extremely wealthy would modify the degree of sexual jealousy experienced by a man of average height, for instance.

This is why I remain a skeptic of evolutionary psychology...


Buunk, A.P. J. H. Park, R. Zurriaga, L. Klavina and K. Massar (2008) Height predicts jealousy differently for men and women. Evolution and Human Behavior. 29(2):133-139.

2 thoughts on “Do These Points Form a Curve?”

  1. As one of the authors of the height and jealousy article, I've found people's enthusiastic reactions to the research far more interesting than the results themselves (which are, let's face it, mostly a bunch of fairly weak correlations). I don't pay attention to the rather common "hey I'm short, but I'm not jealous!" denials of the results, as these people appear not to understand what an imperfect correlation implies (there are MANY exceptions). But when more serious issues are raised--such as the one about the quadratic effect raised here--then I am more inclined to follow it up.

    I found the comments regarding the quadratic effect helpful, because it led me to realize that the figure doesn't show the dataset in is entirety. There were actually many overlapping data points (for instance, there were four women who were 172 cm tall, with a jealousy score of 1, but these just show up as a single dot on the scatterplot). While I don't disagree that the tails are sparse, there are many data points in the average-height/low-jealousy part of the graph that are not visible in the figure.

    How robust is this relationship? Not very. This is what Study 2 of the article showed. There were three additional figures for women that show rather convincingly that the quadratic effect of the type found in Study 1 is not robust. (So it's rather unfortunate that the Internet newspapers latched on to the Study 1 finding and declared that average-height women are least jealous, when our Study 2 results showed that the story is more complicated.)

    What about all the confounding variables? Yes, it would be good to control for some obvious confounds, but not because they would weaken the evolutionary account. There is no rule that evolved psychological tendencies must show up robustly in spite of all other cultural and social influences. "It seems reasonable to posit that being extremely wealthy would modify the degree of sexual jealousy experienced by a man of average height, for instance." That's not merely reasonable--I would be surprised if wealth didn't also influence jealousy.

Leave a Reply

Your email address will not be published. Required fields are marked *