Another social psychology classic bites the dust
Meta-analysis finds little evidence for the Macbeth effect.
16 November 2018
By Jesse Singal
Perhaps no concept has been more important to social psychology in recent years — for good and ill — than “social priming”, or the idea, as the science writer Neuroskeptic once put it, that “subtle cues can exert large, unconscious influences on human behaviour.” This subgenre of research has produced a steady drumbeat of interesting findings, but unfortunately, an increasing number of them are failing to replicate – including modern classics, like the idea that exposure to ageing-related words makes you walk more slowly, or that thinking about money increases your selfishness.
The so-called “Macbeth effect” is another classic example of social priming that gained mainstream recognition and acceptance from psychologists and laypeople alike. The term was first introduced by the psychologists Chen-Bo Zhong and Katie Liljenquist, who reported in a 2006 paper in Science that “a threat to one’s moral purity induces the need to cleanse oneself”.
This claim is such an interesting, provocative example of the connection between body and mind that it’s little wonder it has spread far and wide — there aren’t a lot of social-priming findings with their own Wikipedia page (it was also covered here at the Research Digest). But is it as strong as everyone thinks? For a recent paper in Social Psychology the psychologists Jedediah Siev, Shelby Zuckerman, and Joseph Siev decided to find out by conducting a meta-analysis of the available papers on the Macbeth effect to date.
In their original research, Zhong and Liljenquist found that prompting people to think of a previous bad deed they’d committed led them to exhibit “an increased mental accessibility of cleansing-related concepts, a greater desire for cleansing products, and a greater likelihood of taking antiseptic wipes”.
The basic idea, then, is that someone who feels like they have committed an act of wrongdoing will be more likely, relative to a morally unsullied person, to want to wash their hands. Think Lady Macbeth in Macbeth, ravaged with guilt over the murder he husband has committed: Out, damned spot!
Siev and his colleagues collected all the relevant papers they could find, both published and unpublished (to help with the latter they used the website PsychFileDrawer.org and asked around among the research community). They only included direct replication attempts of Zhong and Liljenquist’s first three studies that “examined the effects of an immoral versus moral prime on cleansing-related preferences or behaviour,” not studies that examined the effect in the other direction, nor broader, so-called “conceptual replications” with somewhat different methodologies. This left them with “15 studies conducted with 1,746 participants from three different continents.”
Above is a graphic from the paper laying out Siev and his colleagues’ results. The bold vertical line represents the point of zero effect size (no relationship between unethical primes and cleansing preferences), with larger effect sizes to the right.
What’s perhaps most noteworthy here is that whereas the effect sizes found in the original three “Macbeth effect” studies were moderate-to-large, in all the 11 independent replication attempts “there was no effect whatsoever”. In light of this, the authors argue, “the evidence suggests either that unethical primes do not generate a greater preference for cleansing-related stimuli than do ethical primes, or they generate a small one.” In other words it looks like the strength of the Macbeth Effect phenomenon has been overhyped, which won’t come as a surprise to those who have been following psychology’s replication crisis. (On the reverse question — of whether “cleansing alleviates moral threat” — the authors suggest there’s stronger evidence, in the form of several independent replications.)
Siev’s team dutifully point out that these failed replications shouldn’t be seen as the end of the story: A “Macbeth Effect may plausibly manifest within certain boundary conditions.” It could depend on the type of moral threat a study participant is faced with, for example, or the type of cleaning involved (say, a floor versus one’s hands).
Maybe the biggest lesson of this meta-analysis, though, has to do with science communications: It’s so often the case that when big, snazzy findings like the Macbeth Effect, or power posing or the implicit association test or any of a handful of other recent darlings of social psychology come down the pipeline, they’re quickly written up in university press releases and taken to be true by many consumers of popular science, as though they had been handed down on Biblical tablets.
Yet subsequent attempts to replicate early, exciting effects have often either failed, or have uncovered something significantly less exciting. For everyone involved in this then — press officers, journalists and researchers themselves — just a bit more caution and humility is probably in order. In the meantime, it’s vital that researchers perform more replication attempts, and when practical, conduct these sorts of meta-analyses of replication attempts as well. It’s the only way to determine just how “real” snazzy ideas in psychology really are.
About the author
Jesse Singal (@JesseSingal) is a contributing writer at New York Magazine. He is working on a book about why shoddy behavioral-science claims sometimes go viral for Farrar, Straus and Giroux.