Replication failure: No evidence that big smilers live longer

What's in a smile? According to a widely reported 2010 study of US major league baseball players, which we covered here at BPS Research Digest, one important answer is: an indication of how long the smiler will live.

By analysing official individual photos of players from the 1952 baseball season, and then looking at subsequent death records, Ernest Abel and Michael Kruger at Wayne State University, Detroit, concluded that players who'd smiled like they meant it – with full "Duchenne smiles", which involve muscles around the eyes as well as the mouth – lived on average seven years longer than players who'd posed with less convincing grins.

The result was taken to support existing evidence that happier people tend to live longer. It also seemed to show that smiles in posed photos – even on just one occasion – are a fairly reliable signal of people's underlying emotional disposition and therefore their likely longevity.

But a new replication and extension of the Baseball photo study has produced very different results. This is important, because the idea that happier people live longer is widely promoted, and has implications both for individuals and policy-makers.

The original study, published in Psychological Science, involved an analysis of vintage photos of 196 players. They were classed as either Duchenne smilers, non-Duchenne smilers, or non-smilers. By the time of the study, most had already died. Abel and Kruger found that 35 per cent of the variation in lifespan between the players was related to how intensely they had smiled in their photo. Put differently, Duchenne smilers were half as likely to die in any given year after the photos were taken, as compared with non-smilers.

For the new study, also published in Psychological Science, the researchers, led by Michael Dufner at the University of Leipzig, Germany, were unable to confirm exactly which photos were analysed by Abel and Kruger (who said the information was unavailable), but they used the same photo database and the same selection procedures, and came up with a sample of 224 player photos, which they dubbed the "1952 group". To extend their investigation, they also identified 527 more photos of players who were from earlier or later cohorts than the 1952 group. These extra players had ended their career in 1951, or they had debuted between 1953 and 1957, meaning they had played around the time of the original group, but could not have been included in Abel and Kruger's study.

As in the original study, a team of five facial coders evaluated the players' expressions, and classified them as full, partial or non-smilers. But this time, the analysis also went further: another group of human raters also judged the "joy intensity" on each player's face, and, finally, three different emotion-recognition computer systems also assessed the faces for happiness.

For both the 1952 group, and the extension group, when rated by human coders, smile intensity did not predict lifespan. This null result held whether other variables that might be expected to be related to longevity were taken into account or not, such as birth year, age at debut, career length and BMI. It also didn't make any difference whether the researchers compared just partial smiles with non-smiles, or full smiles and non-smiles.

In contrast, the new measures of emotions shown in the photos – greater happiness, as coded by a computer, or greater joy, as rated by human observers – did both correlate with players' longevity. But as soon as the researchers took the other variables into account (players' BMI etc) these correlations were no longer statistically significant. One particular variable turned out to be critical: birth-year. "Thus, when we considered that it was uncommon for players from earlier cohorts to smile in photographs and that these players also had reduced life expectancy [also revealed by the data], smiling ceased to predict mortality," the researchers write in their paper.

Since the researchers did not know exactly which photographs were used in the original study, it is impossible to be certain why they failed to replicate Abel and Kruger's finding, but it "appears to be a false positive result," Dufner and his team write.

While some research has indeed found a link between happiness and longevity, this new paper isn't the only recent study to produce apparently contradictory findings. In 2015, for example, a study of 700,000 women in the UK found that those said they were unhappy had about the same chance of dying during the ten year follow-up period as those who reported being mostly or usually happy.

The new findings don't provide a direct test of whether happiness and longevity are linked. "However," the researchers conclude, they do at least "indicate that the degree to which professional baseball players smile on a photograph taken during their career does not contain any genuine information about whether they are still alive half a century later."