STATS ARTICLES 2006

2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003


Fingering the Media

September 28, 2006
Rebecca Goldin Ph.D.
What is it about finger length that journalists find irresistible?

The media just loves a finger story. Last year, the Express Newspaper of the U.K. reported on a study finding that male aggression is linked to finger length – the longer the ring finger is as compared to the index finger, the more likely the male is to be aggressive. The report at the time also claimed that finger length is correlated to testosterone exposure during labor.

Now MSNBC reports (via Reuters) on a new study which claims that women’s athletic ability is also related to the same ratio – the longer the ring finger is, the more likely the woman will excel at running and running-related sports like soccer. Apparently previous studies showed the same is true for men. In a related study this year, however, finger length was not shown to be correlated to testosterone exposure before birth.

The problem with these kinds of studies is two fold. First, these researchers don’t have a plausible mechanism to justify their claim. The proposed explanation last year was that “testosterone in the womb” made the difference between long and short index-fingered people, but that has now been dismissed. Perhaps all the correlations are just flukes in the same way.

Secondly, there are rarely studies published about what finger-length is not correlated with. Some might argue this is irrelevant, but it actually is extremely relevant from a statistical point of view.

Any time you test for a correlation, there is a small possibility (about five percent) that you will measure a relationship (such as finger-length and athletic talent) when it isn’t actually present in the whole population. These “findings” will seem as “true” as any other statistical finding, such as the fact that lung cancer and smoking are correlated. But this means that if you try to correlate twenty different traits to finger length –even if none of them are actually related to finger length – you will likely find some correlation, just by chance.

If there’s no reason for a correlation to be in place, and it hasn’t undergone repeat testing under strict “rules” (such as being forced to report how many checks against other traits came out negative), we have no reason to believe it.