Tuesday, 2 October 2007

How wrong is it to speculate?

Just some thoughts that have been festering since the Hurlbert and Ling paper in Current Biology, the one that showed that women preferred pink hues when compared with men. I wrote about it, Ben Goldacre wrote about it, and I think more or less everyone with an internet connection wrote something about it somewhere.

The general gist of opinion on the article seemed to be that it was an example of bad science, because of the author's speculation that their results had some sort of evolutionary psychology explanation, perhaps that women needed to pick out red berries against a green background in the days when the men were out killing stuff and hauling it back to the cave. I would agree that the data presented in the paper don't support that interpretation. On the other hand, the authors clearly identified that hypothesis as speculative. I think it's worth asking whether speculation can have a place in scientific literature.

In my view, the answer is yes. In the case of the Hurlbert and Ling paper, data was presented showing that women prefer pink hues. The authors indulged in some speculation as to why this might be the case. It should now be the task of scientists to try and devise studies that could refute that speculation. It follows the pattern make observations, formulate hypothesis to explain observations, make more observations to test hypothesis. So I think the speculation in the paper is scientifically defensible. In this context, Kaj Sand-Jensen's famous paper on "How to write consistently boring scientific literature" is always worth a look, especially the section headed "remove most implications and every speculation". Scientific caution is generally sensible, but if taken too far it can mean that possible leads are not followed up.

The problem with that approach is the way such studies get reported in the media. As we've seen, most of the stories on the paper suggested that the authors had confirmed an evolutionary psychological explanation for gender-based differences in colour preference. In fact, they had simply showed that gender-based differences in colour preference exist. With scientists getting brownie points for media engagement, it is inevitable that the differences between results and speculation get lost.

4 comments:

apgaylard said...

I've been thinking a bit about this one. I have included speculation in some of my papers, so my answer must be yes.

However, i'd like to temper that. I would say that a speculation (or speculative hypothesis) should be testable and falisfyable to count as scientific speculation. The author should also make it clear that it is a speculation.

I could be tempted to believe that there are some disciplines within science where speculation is probably best avoided. Medical research may be one of them, given the credance people sometimes put in individual studies.

Paul Wilson said...

Yes, clearly there have to be some ground rules. Ideally, any speculative hypothesis would also have a reasonable 'prior probability' of being correct. That is, it wouldn't simply be made up, like the silica hypothesis for homeopathy.

Also agree it needs to be absolutely clear what is speculation and what is not. In the Hurlbert and Ling paper, the authors did make it clear that their evolutionary psychology explanation for their observations was speculative.

By the way, I'm very much enjoying your blog. I've been reading it, though I haven't commented yet...

bengoldacre said...

speculation is great, to be heartily encouraged, and exactly what you'd hope to see in the discussion section of a paper: but the press release, the reported content of interviews, and the press coverage, went a million miles beyond speculation in this case.

Paul Wilson said...

Yes, I agree entirely. From reading the press reports, you would have thought all the speculative evolutionary psychology stuff represented the actual results of the study.

It's not a bad study, but the way it was spun to the media was not good. While the media are somewhat at fault for apparently not actually reading the paper, most of the blame must fall on the press office at the University of Newcastle. I'm not sure how these things work, but I expect the press release would have been approved by the authors of the study, in which case they have to shoulder some of the blame. And of course, they're entirely responsible for what they said in interviews.

I wonder if this partly comes down to unrealistic expectations of scientific studies? If you put out a press release saying 'A Newcastle University study has found that women prefer pink hues when compared with men. There may be several reasons for this, and we need further research to understand why it happens', would people just think it was a useless study that hadn't really found anything out?