The scientific method

As much as we like to pretend that science is a rationally objective endeavor, sometimes looking at data is like doing a Rorschach test — people see what they want to see. Some of the commenters to the post below do not quite draw the same conclusions that I do from the AIP study. So let’s go through the exercise slowly.

President Summers presented three hypotheses for why there are fewer professional women scientists than men. (Picking on Summers as an individual is certainly not the point, and it’s getting kind of tiresome, but he did try to provoke people, after all.) They are:

  • “80-hour work weeks” — women have family responsibilities, and don’t want to devote the huge effort required to being a science professor.
  • “Intrinsic aptitude” — something in women’s brains makes them just not as good at science, at least at the upper levels.
  • Systematic biases — women are discriminated against, or at least pressured away from, becoming scientists.

These are in decreasing order of importance, in Summers’ estimation. (Although, sympathetic as he is, he’d love to be proven wrong.) They are good solid hypotheses, in that they make predictions that can be compared with the data. So let’s do it.

The AIP study considered the representation of women in science at different levels up the ladder, from high school to full professors. It found that the biggest leakage of women from the pipeline was between high school and college; once women got their bachelor’s, their representation at higher levels is consistent with what we expect from a gender-independent rate of success, given the obvious time lags it takes for people to progress through the various stages. What do our hypotheses predict? The 80-hour-work-week idea makes a pretty clean prediction: as we travel up the ladder and the competing pressures of work and family become more real and more evident, women should preferentially be dropping out. And this prediction is — false. If the workload and childrearing pressures are to blame, why would the effect be localized during high school and college? Of course, a good theorist can wriggle out of any experimental finding. For example, we could imagine that female undergraduates very effectively anticipate the upcoming work/family squeeze, and get out while the getting is good. Except that, they don’t. Of all the reasons why college students have given me why they wouldn’t become physics majors, “I’m worried that some day I won’t have time to both be a good science professor and also raise children, and therefore I’m going to medical school instead” has never been one of them.

So let’s consider “intrinsic aptitude.” The idea here is that there is a bell-curve distribution of cognitive abilities, and that the curves are different for men and women. Either the mean for women is simply lower, or the standard deviation is smaller. In either case, as we get far out along the exponential tail at high levels of achievement, there is a very clear prediction: the ratio of successful women to successful men should become dramatically smaller and smaller. So as we look up the pipeline, women should be dropping out more and more as we climb up the ladder. And again the prediction is — false.

What about the idea of systematic biases? Unfortunately, this hypothesis doesn’t really make any good predictions for this particular test. Until you tell me what the biases are, I can’t predict when they will operate most strongly. Of course I can come up with glib stories after the fact, suggesting that the biases are most pronounced at the point when students are making choices about what major to pick — and indeed I did come up with that story, and I think it’s likely to be true. But these data don’t really give us much evidence one way or the other.

The “biases” hypothesis does make predictions for other experiments, of course. For example, it would predict that women would suffer subjective biases in blind experiments where people are asked to judge work by men and women. And indeed, they do.

Likewise, the intrinsic-aptitude hypothesis makes other predictions. For example, it would predict that the fraction of women is basically the same everywhere, since it’s intrinsic rather than due to social factors. That’s wrong. It would also predict that the number of women in the field is remaining approximately constant, for the same reason. That’s wrong, too. Of course, you could claim that the true, unbiased fraction of women receiving Ph.D.’s should be about five percent, and is only 18% now because of the pressures of political correctness forcing unqualified women into this role. You would, to be sure, be implicitly admitting that social factors can easily trump intrinsic differences, except that you’d be thinking that these factors work in women’s favor. You should also look into loosening the elastic band on your tinfoil hat.

Scroll to Top