Priming: The Cue You Never Noticed and Its Limits

This chapter is here for two reasons, and the second one matters more than the first.

In 1996 John Bargh, Mark Chen and Lara Burrows reported that participants who unscrambled sentences containing words associated with the elderly — Florida, bingo, wrinkle, grey — subsequently walked more slowly down the corridor away from the laboratory than participants given neutral words. None of them noticed the theme. The finding, called elderly priming, became one of the most cited results in social psychology and one of the most repeated in popular science writing, because it appeared to show that a concept could be activated outside awareness and then drive behavior directly.

In 2012 Stéphane Doyen and colleagues tried to replicate it. With walking speed measured by infrared beams rather than by a stopwatch held by an experimenter who knew the hypothesis, the effect disappeared. Doyen then ran a further condition in which experimenters were told which participants had been primed and led to expect an effect. The effect reappeared — in the group whose experimenters expected it.

That is not a small correction. It relocates the cause from the participant's mind to the experimenter's expectations, which is the Clever Hans problem covered in chapter 59, and it was one of the episodes that opened the replication crisis in psychology. Bargh's response was heated. Subsequent meta-analytic work on behavioral priming has found the literature heavily affected by publication bias, with effect sizes shrinking as study quality rises.

So what survives?

Semantic priming — the ordinary, robust finding that processing a word makes related words faster to recognize — survives comfortably. It is one of the best replicated effects in cognitive psychology and it is the basis of most models of how semantic memory is organized. Perceptual and conceptual accessibility effects are real. What is doubtful is the strong claim: that ambient cues reliably produce meaningful downstream changes in complex social behavior, in the wild, at effect sizes worth acting on.

This distinction has direct consequences for anyone using this guide. A great deal of commercial persuasion is sold on the strong claim. Consultancies sell scent priming, color priming, and subliminal brand exposure. Political operators are advised that ballot design and ambient imagery will move votes at scale. Some of these have real effects for other reasons — ballot order effects are well documented and have nothing to do with priming — and some are the strong claim being resold after it failed.

The epistemic lesson generalizes past this one experiment, and it is the reason the chapter earns its place in a book about manipulation. Findings about influence are unusually attractive to people who want to sell influence, which means this literature is subject to more motivated amplification than most. Before you build a defense on a mechanism, or fear an attack based on one, ask three questions: has it been replicated by people who were not invested in the original result, does the effect survive when the measurement is automated and the experimenter is blind, and how large is it in absolute terms.

Several mechanisms in this guide fail one or more of those tests, and they are included with the failure attached rather than quietly dropped. A guide that listed only the findings that held up would give you a false picture of how much this field actually knows.