neur.ro

Priming

type: effect · origin: study · evidence: contested · themes: judgements and probabilities

Claim assessed: A word or image seen in passing changes your behaviour without you noticing.

Published · how we work

In short

Priming means that a stimulus you meet now influences how you process what comes next. The term actually covers two very different ideas. The first, from cognitive psychology, is that a word makes you recognise related words faster, for a fraction of a second. That is well established. The second, from social psychology, is that a word or image met in passing can change your behaviour for minutes on end, without your noticing. The famous studies in the second category did not show up again in careful replications.

What it says

Semantic priming. Meyer and Schvaneveldt (1971) showed participants pairs of letter strings and asked them to say as quickly as possible whether both were words. The answer came faster for pairs of associated words than for unrelated words. Shanks et al. (2013) note that effects of this kind are well established in cognitive psychology, but they generalise very narrowly, and unconscious influences, when they occur, last under a second.

Priming behaviour. Bargh, Chen and Burrows (1996) had students form sentences from scrambled words. For some, the words included ones related to old age. The authors reported that these students then walked more slowly down the corridor on their way out than the others. In another experiment, those who had formed sentences with words about rudeness interrupted the experimenter’s conversation sooner. The study became one of the most cited in social psychology. We haven’t read the original; we describe it from Doyen et al. and Weingarten et al.

Slow walking, repeated. Doyen et al. (2012) repeated the experiment with 120 students. Unlike the original, where time was measured with a hand-held stopwatch, they used infrared sensors. They found no difference. In a second experiment, they led half the experimenters to expect participants to walk more slowly and the other half to expect them to walk faster. The effect appeared only for participants tested by experimenters who expected slow walking. What’s more, the hand-held stopwatches of those who expected fast walking showed a reverse effect, which the sensors did not confirm. And most participants exposed to words about old age could later say which category of people had been suggested to them.

Top: in the replication by Doyen and colleagues, the difference in corridor time between those exposed to words about old age and the others was minus 0.12 seconds in the first experiment, plus 0.43 seconds when the experimenter expected slow walking and plus 0.06 seconds when the experimenter expected fast walking. Bottom: in Many Labs, the effect of flag priming fell from 0.50 to 0.03, and that of money priming from 0.80 to minus 0.02. The priming that didn't replicate Doyen et al. (2012) and Many Labs (Klein et al., 2014) Difference in corridor time, in seconds positive: those exposed to words about old age walked more slowly exp. 1: 120 participants −0.12 exp. 2: experimenter expects slow walking 0.43 exp. 2: experimenter expects fast walking 0.06 Effect size (d): original study and Many Labs American flag, original study 0.50 American flag, 4,896 participants 0.03 money, original study 0.80 money, 6,333 participants −0.02 It appeared only where the experimenter expected it, and in Many Labs it all but disappeared.
Top: Doyen et al. (2012), times measured with infrared sensors, the difference between the mean of those exposed and the mean of the others. Bottom: Klein et al. (2014), Table 2, the weighted effect across all samples, against the effect in the original study.

Intelligent behaviour. Another often-cited study, by Dijksterhuis and van Knippenberg (1998), claimed that people answer general knowledge questions better after thinking for a few minutes about a professor, and worse after thinking about a football hooligan. Shanks et al. (2013) tried to obtain the effect in nine experiments with 475 participants, using the original procedures and variants of them. They found it in none. A cash reward, on the other hand, did improve performance.

Many Labs. Klein et al. (2014) repeated 13 classic effects in 36 samples, with over 6,000 participants. Ten replicated clearly. Two priming effects did not appear at all: a photo of the American flag supposedly increasing conservative attitudes, and images of money supposedly increasing acceptance of the existing social system. In both cases the combined effect was practically zero.

What the meta-analyses say. Weingarten et al. (2016) collected 352 effects from 133 studies on words presented in passing before a task. They found a small effect, d ≈ 0.35, and concluded that it isn’t explained by mostly positive results getting published. That conclusion is disputed. Shanks and Vadillo (2021) show, for field studies of goal priming, that the large effects come from the small studies. When they correct for this, the effect is no longer different from zero. They also cite a meta-analysis on money priming: 62% of ordinary studies found the effect, but only 11% of pre-registered ones did.

Example

A business book recommends putting a photo of a runner winning a race on your desk, so that you work better without noticing. The recommendation rests on small goal priming studies. When these are corrected for the fact that negative results are rarely published, the effect is no longer different from zero. The example is an illustration, not a study cited here.

How to apply it

The steps below are a practical way of applying it that we propose, based on the studies cited.

  1. Don’t build plans on behavioural priming. Posters, “motivating” words or images slipped into the environment have no solid evidence that they change what people do.
  2. Watch what the person observing you expects. In Doyen et al.’s study, the experimenter’s expectations mattered more than the words read. In the same way, someone evaluating an effect they want to see can see it more easily than it is.
  3. When you read about a spectacular effect, ask whether it has been replicated with pre-registration. The gap between 62% and 11% in money priming shows how much selective publication can inflate an effect.
  4. Semantic priming is still real, but small and brief. A word you see now helps you recognise a related word in the next fraction of a second. It doesn’t make you slower, smarter or more conservative.

Limits and nuances

Sources

See also: Anchoring, Ego depletion, Power posing, Stereotype threat