Hidden beliefs in science stereotypes predict size of gender gap across 34 countries

Think of a scientist – not anyone in particular, just a random individual working in the field. Got one? Did you picture a man or a woman? If it’s the former, you’re probably not alone. There have been a few times when I’ve only ever known a scientist through their surname on a citation and automatically assumed that they were a man, only to learn, to my chagrin, that they’re actually a woman. It’s always a galling reminder of how pervasive the stereotype of science as a male endeavour can be, even at an unconscious level.

Now, Brian Nosek from the University of Virgina, together with scientists from over 14 countries, has charted the extent of these implicit associations across the globe, and shown that they predict the size of the gender gap in school-level scientific achievement.

Nosek suggests that these biases and gender gaps feed off each other in a vicious cycle. The sex differences already present in the sciences, especially at the top echelons, are hard to miss and they can make stereotypes feel very real. In one study, women who saw a conference video where three in four attendees were men (a very real situation for many female scientists) felt less belonging and less desire to participate.

Stereotypes can also create themselves. Women who buy into stereotypes are less likely to take up a maths or science degree. Even if they refuse to be pigeonholed, they can be so stressed about conforming to a stereotype that they actually increase the odds of doing by taking a hit to their performance. This phenomenon is called “social identity threat” and it’s evident in research that shows women do more poorly in tests if they have previously been reminded of the supposed male superiority or even, simply, if their gender is highlighted.  

Nosek’s group relied on a powerful tool that provides a standardised measure of the strength of stereotypes across different countries – the Implicit Association Test (IAT). The computer-based tests ask people to group words into pairs categories using assigned keys. One key might represent male words (he, boy) as well as science words (physics, chemistry), while the second key might represent female words or arts-related  ones.

The idea is that people perform the task more quickly and more accurately if the combinations of categories match their unconscious associations. So people who think of science as a game for boys would be quicker at trials where science words were twinned with male ones than those where they were paired with female ones. The power of these tests is that they don’t rely on any conscious reflection – it doesn’t matter what people actually believe about gender roles in science, just what hidden biases they might have.

You can try them yourself on Harvard’s website, and that’s exactly what over half a million people have done from 34 countries around the world and in 17 different languages. It’s these data that Nosek used for his study. Around 70% of those who took the test tended to associate science with men and arts with women more easily than the opposite pairing. It’s important to remember that this doesn’t mean that 70% of the participants actually endorsed these stereotypes – indeed, what they said about their views only weakly matched their implicit biases, as revealed by the test.

The IAT data have proved their worth in other studies that show, for example, that women who more strongly associate men with science are less fond of science, less interested in pursuing it and fare worse on tests or exams. 

This time, Nosek compared the data to a measure of national sex differences calculated from the 2003 results of the Trends in International Mathematics and Science Study (TIMSS). This is a standardised set of maths and science tests run among 8th-graders (usually 13-14 years old) in 34 different countries. For each one, Nosek subtracted the average score for girls from that for boys. This figure varied greatly from country to country, and it was strongly linked to scores on the IATs as the diagram below shows.

So the more a nation believes in the stereotype of the scientific male (even unconsciously), the greater the gap in performance between boys and girls in both science and maths. In fact, these hidden biases were a better predictor of the gender divide than what people actually said about science stereotypes. These explicit opinions accounted for about 2% of the international variation in the science sex gap, while implicit associations accounted for a much larger 19%.

It’s especially interesting that Nosek found such a strong relationship by looking at two different groups of people – those who took the IAT test and 8th-graders. It’s unlikely that there was much overlap or interaction between the two groups; the more probable explanation is that both were influenced by the societies they live in.

The relationship that Nosek found was statistically significant (unlikely to be  a fluke result) and it remained that way after he put it through a gauntlet of challenges. He took out an outlier that was skewing the data but it stayed robust. He adjusted his numbers for various relevant factors such as the proportion of men in the IAT sample, the age of the participants, the countries’ wealth and even a general measure of gender inequality – the Gender Gap Index. Still, the correlation was significant. He even found the same link when he compared implicit associations to TIMSS scores from 1999, before the IAT website had been launched (although with a smaller sample size and fewer countries, the statistics weren’t as strong).

The link that Nosek found is also specific to the sciences. It doesn’t reflect a general tendency to believe in stereotypes, for the Harvard website includes tests for age and racial bias and neither of these prejudices were associated with the TIMSS scores. Nor is the result a reflection of general gender issues, for as I’ve already mentioned, accounting for international differences in the Gender Gap Index didn’t diminish the significance of the statistics.

The big question then is which came first? Do we have stereotypes because of different performances between men and women, or does the stereotype fuel the gap in performance? It’s probably the case that both answers are right, and that stereotypes and gender gaps reinforce each other with small changes spiralling out of control. You can see the same thing at work with racial stereotypes and academic achievement too.

Nosek’s conclusion is that we will fail to battle inequality in the sciences unless both issues are tackled by the same initiative – encouraging women to take up a scientific career will do little without addressing the all-encompassing stereotypes they face. Likewise, you can’t change these stereotypes without addressing the current reality of fewer women in top scientific careers. As Nosek says, these cyclical effects “make it more difficult to jerk the system out of homeostasis” but they can also lead to more positive cascades.  

Change one factor and you influence the other, pushing the whole status quo towards greater equality one nudge at a time. Again, there’s evidence from other areas of prejudice and inequity that this can work. Just three months ago, I wrote about a promising and simple writing exercise designed to break a similar cycle affecting black students, by boosting their sense of self-worth and negating the stereotype of the underachieving black student. It’s exactly the type of intervention that could be transferred to other areas, where hidden prejudices hold back entire sectors of society.

Reference: PNAS 10.1073/pnas.0809921106

More on gender issues:

• Gender gap in maths driven by social factors, not biological differences
• Why are there so few female chess grandmasters?
• Mind your words – how stereotypes affect female performance at maths