Sex and the size effect – why reporting on gender is often hyped

Not as simple as black and white. Fred Bchx, CC BY-NC-SA

Evolutionary psychology, the field that uses the process of natural selection to provide a theoretical framework to explain human behaviour, gets in the press a lot. And if there’s one thing that gets web hits, it’s stories about sex. But just how relevant are such sexy facts to the reader? How does a scientific truth relate to an individual life? The answer is not as simple as it might seem at first.

One recent study considered the combined effects of using the contraceptive pill and facial attractiveness of a partner on a woman’s marital satisfaction. After following 118 heterosexual women over a period of several years, they reported that among those who stopped using the pill, those who had less attractive partners reported lower marital satisfaction than those with more attractive partners. The opposite relationship was found among women who carried on using the pill.

You may have already heard researchers complaining about poor science reporting misleading the general public. But responsibility can also lie with us – the researchers and the journals that publish our work. In the field of evolution and human behaviour, our frequent sin is a failure to contextualise our findings with some measure of what is called “effect size”. This can effectively take many forms, but it is essentially a number describing the strength of a relationship being reported. This means that the potential influence of a causal factor on a particular measure can be compared to that of other factors.

Using the example of the recent contraceptive pill study, some statement of effect size could help us appreciate the potential strength of any physiological effect of the pill on marital satisfaction, when placed among all the other joys and strife that exists in real marriages, such as children, the bills and the in-laws.

This matters because readers are especially drawn to stories about subjects to which they can relate. Time covered the item with the headline: Going off the pill could affect who you’re attracted to, study finds. Earlier this year a piece in The Huffington Post on related research was entitled: What you didn’t know about how the pill is affecting your relationship.

Context is crucial

Although no researcher would advocate that readers apply the results of such research blindly to their own life, this distance between research and practice isn’t necessarily clear to readers. There is a very strong case to be made that researchers have a responsibility to do their best to ensure that their work is placed in its proper context.

In this case it is necessary to study the research paper in detail to find some indication of effect size. Close inspection of the graphs in the paper suggest that in the extreme groups of partner attractiveness (those with the most or least attractive partners), there were differences of about 15-20% in scores of the “Quality Marriage Index”. This was a score derived from answers to a set of “how much do you agree?” questions. These percentage changes might seem next to meaningless in the context of the ups and downs of real personal relationships, but at least such figures give us some hint of perspective. But this information wasn’t available in the accessible part of the research or published in an easily digestible form in the article.

This is the norm in our field. Many journals don’t demand researchers place any indication of effect size centre stage, for all to see. I surveyed the abstracts of research papers from the most recent published issues of four journals dedicated to the field of evolution and human behaviour. Out of 29 papers I evaluated as having been potentially able to report at least some indication of the magnitude of the principal results reported, only four (13%) did.

So why is this information not more widely available? The answer might be because in most empirical studies in “theory-driven” fields such as this, researchers are often primarily interested in determining whether or not there is evidence to support or falsify a particular hypothesis.

When reporting findings, the key piece of information that researchers in the field of evolution and human behaviour want to communicate is the statistical significance of an effect, which shows that an effect is “real” and not due to random chance.

In my survey of abstracts, a verbal statement confirming statistical significance was always present. However, the problem arguably gets more serious in the area of gender differences. Statistically significant average differences between men and women (in attitudes to personal infidelity, for example) get more attention than you can wave a copy of Men Are From Mars, Women Are From Venus at. This is true even if the effect sizes are small, meaning that the difference between any random man and woman may be absent, or even reversed.

The essential difference

Such findings get so much press because they play on and and reinforce our expectations of the irreconcilability of the genders. So it’s no wonder that we often continue to think that men and women are as fundamentally dissimilar in evolved behavioural tendencies as the the peacock and the peahen are in appearance. But in reality, as Steve Stewart-Williams and Andrew Thomas have compellingly argued, human males and females might be psychologically far more similar than many of us imagine.

Those of us who work in the field of evolution and human behaviour should follow the example set by other fields. In epidemiology, not only statistical significance, but also the real world differences between groups of individuals in metrics such as body mass index, disease risk and mortality, are routinely laid out in the abstract, in black and white

There is public interest in what we do, and therefore we have responsibility to do our utmost to make sure that the results of our research are conveyed with accuracy. That means taking simple steps to place them in the proper context. Putting some indicators of effect size in the abstract of papers – where most people get their sense of the research – would enable a better assessment of how important the findings are as a predictor of behaviour.

If researchers and journals made this a matter of routine, it would serve academia and society well, and do a great deal to improve reporting and credibility of the field.