Sean Roberts and James Winters have evidently had it with dubious correlational studies of cultural traits. You know the kind: those viral-ready “countries who eat more chocolate win more Nobel Prizes” sorts of findings that regularly fill your news feed.
In a recently published paper on the online peer-reviewed journal PLOS One, the two researchers, from the Max Planck Institute for Psycholinguistics in the Netherlands and the University of Edinburgh, respectively, show just how easy it is to produce these results by demonstrating a correlation between a country’s rate of traffic fatalities and linguistic diversity, even controlling for a wide array of other variables. They then build a kind of infinite loop of cultural correlations, some more dubious than others:
Linguistic diversity is linked with climate. Climate affects the likelihood of cultural siestas. Cultures that take siestas tend to have languages with less morphological complexity (t = 3.47, p = 0.001, see methods). Morphological complexity is linked with group size. Group size is linked to the levels of extra-marital sex in a community (r = −0.54, p = 0.001, see methods). Levels of extra-marital sex have been linked to a language’s phoneme inventory. Phoneme inventories have been linked to patterns of migration. Migration patterns are linked to the level of political collectivism in a culture (r = 0.42, p = 0.004, see methods). Collectivism is predicted by genetic factors. There are also genetic correlates of linguistic tone. Tonal languages co-occur with acacia trees (t = 3.77, p = 0.0002, see methods). To bring the chain full-circle, the presence of Acacia nilotica also predicts a greater number of traffic accident fatalities, controlling for linguistic diversity, length of road network, GDP, distance from the equator, population size and population density.
According to the authors, there are two problems at work here. One is that large computerized databases make these types of correlations extremely easy to find. The other is that “the media typically exaggerate the implications of this type of finding and try to link it to current events rather than emphasize the long-term change implied in most studies.” This could also function as a much broader critique of media coverage of scientific findings of any kind.
I was amused to see one of my own articles cited in the footnotes as an example of this type of coverage, a short piece for Foreign Policy on how the way languages deal with future events affects the economic and health behavior of people who speak them. (In my defense, I also wrote a follow-up noting the critiques of the finding by some linguists, including Roberts himself.)
In general, clickbait headlines aside, I would hope most readers don’t take findings like the chocolate-Nobel prize thing all that seriously. Most correlational findings of this type are meant to raise questions rather than answer them. (See my colleague Daniel Engber on the correlation/causation cliché.)
But I understand why researchers get irritated by media coverage of findings like this, and the paper is a brilliant way of critiquing them. Though to be honest, findings in linguistics don’t have quite the public consequences as dubious correlations put forward by researchers in other fields.