Who do you choose to work with?

We all know that the ideal experiment consists of drawing up a hypothesis and then testing it to see if your hypothesis stands up. If the experiment is well designed you will get a clear answer, ‘yes’ or ‘no’, and if it’s a ‘no’ it’s a case of going back to the drawing board. If you know what answer you want to find you have to be careful that the design really does distinguish between alternative explanations and that you don’t cherry pick results, stop before you have enough data to be convincing or fall into the familiar trap of confusing correlation with causation. Any work submitted for publication should ensure it has done all this thoroughly before it gets published. If not, holes will and should be picked in it.

I am reminded of these facts after reading a curious letter that appeared last week in Current Biology claiming to examine the “tendency for men to cooperate more than women with same-sex individuals of differing rank”. Although written in a suitably dry academic style, the sub-text was that many women are queen bee, Maggie Thatcher-like individuals who go out of their way not to cooperate with less senior women – and this behaviour was in contrast to men. Indeed, the Harvard press release said explicitly: “It’s long been a popular stereotype: Men are hugely competitive, meaning cooperative effort is the exception rather than the norm, while women have a tendency to nurture relationships with others, making them much more likely to cooperate with one another. A new Harvard study, however, is turning that cliché on its head.”

Does the evidence really back that statement up? I am not convinced based on the limited information provided in the letter.

The study, by Benenson, Markovitz and Wrangham, examined papers published in the discipline of psychology by pairs of authors working in the same department across 50 US departments. They found that when the co-authors were both of the same gender and either of the same high professorial rank or when there was a professor and a student, the probabilities of pairing were identical for men and women. But where the co-authors were a high status (full) and a lower status (assistant) professor, women were less likely to publish with another woman than the men were with another man. Based on this finding they conclude that there is less cooperation between women than men.

Does this stack up? I found the evidence less than compelling. Firstly, as Hannah Devlin pointed out in her own article (£) in the Times deconstructing the piece, the number of pairs of women who fitted into the ‘high status-low status’ pairing was only 14, a very small number on which to base their conclusion (there were 76 equivalent male pairs). However, what worries me more is that it is clear that the authors set out to confirm their hypothesis and did not properly explore other explanations for their findings, even assuming the numbers were robust. I can come up with several alternatives which strike me as being at least as plausible as the one they propose.

Maybe the men actively seek out the ‘lower status’ women for reasons of their own: predatory, power-based or entirely thoughtful reasons of support. Who knows? If a young female professor is approached by such a high status male she may well see them as likely to be useful to them in the future given plausible power structures. This fact may make them likely to accept the approach, thereby rendering them less available to a senior female professor in the same department. The junior women may themselves actively seek out the high status men, for this same reason of regarding a collaboration with them as beneficial to their careers. According to the Times article, Benenson accepted this idea was “equally possible” as an interpretation as their own hypothesis, but you wouldn’t know that from the published letter.

I would also suggest that the low numbers of women involved in the study may lead to a separate set of problems, beyond the mere statistics of the probability of the dyad interactions. If a department has only a few women it is not sufficient simply to work out the frequency of such interactions, since one can’t collaborate with people randomly. Your areas of work have to have something in common. The numbers given in the letter suggest that, whereas there are approximately equal average numbers of men and women assistant professors in a given department, there are nearly twice as many senior men as women. So, the probability of finding an appropriate collaborating pair involving a senior man will be equivalently higher than one involving a senior woman. We are not told how many individuals don’t collaborate at all or collaborated with individuals of the opposite gender. Nor are we told anything about collaborations existing outside a single department. If I were a woman wanting to collaborate with another woman, and there wasn’t one close to my field of research in my own department, maybe I’d look further afield. Or maybe I wouldn’t. But without all these other possibilities being explored, how can I have confidence in the conclusion drawn in the letter?

There was another curious aspect of the paper: the authors took as their starting point the fact that even young children exhibit differences in how they interact. The evidence is apparently that boys are likely to work/play in loosely structured groups, while girls tend to focus on one-to-one interactions. But the analysis in the paper is solely to do with pairwise interactions and tells us nothing about these loose groups. Thus they used as a basis for their initial hypothesis the difference in size of group and ended up with a conclusion about the hierarchy of pairs of collaborators.

This is one of those papers that gets air-time, as I am giving it here, because it conveys a controversial message. Again, to refer to the press release, we are told: “That’s not to suggest women are inherently flawed when it comes to cooperation”, a sentence which immediately makes one sense that it is exactly what they’re suggesting. Norwegian champion for gender equality Curt Rice, who has also taken this paper to pieces on his own blog, has a field day with the meaning of cooperation and how the authors are conflating different senses of the word. He also comes up with other explanations of his own such as “the absence of co-authorship cannot be taken to imply that two people have not worked together, nor does it demonstrate that they have not cooperated and it certainly does not demonstrate a lack of friendliness. For all we know, the women professors are so friendly and cooperative that they promote their junior women colleagues by letting them take single authorship on the papers.”

So, see what you think. Are all senior women a bunch of prima donnas who go out of their way to avoid their junior colleagues? Or can you come up with some other explanation for the data of your own I haven’t described here? I suspect most female physicists of my acquaintance would simply be delighted if there were some local female colleagues they could even share a cup of tea with, let alone with whom they could collaborate.

Athene Donald

Athene Donald

Professor Dame Athene Donald is a Professor of Experimental Physics and Master of Churchill College. Her research is in the general field of soft matter and physics at the interface of biology; she has published over 250 papers in these fields. In recent years she has become heavily involved with policy issues, ranging from chairing the Royal Society’s Education Committee to chairing the Scientific Advisory Council of DCMS.She served on University Council (2009-14) and was the University’s Gender Equality Champion (2010-14). She is currently a member of the Scientific Council of the ERC and a former Trustee of the Science Museum. As well as various prizes from the IOP and Royal Society, she won the 2009 L'Oreal/Unesco Laureate for Europe award.
Athene Donald

Comment via Facebook

Comment via Disqus


Comment via Google+