What Is Priming? Earlier Exposure Changing a Later Response

Presenting a stimulus speeds up the processing of a related one shortly afterwards, which is measured in reaction times and is highly reliable. Seeing a word makes it faster to recognise that word again, which is repetition priming. Seeing a related word speeds recognition of a target, so bread speeds up butter, which is semantic priming and is one of the most replicated results in cognitive psychology. Presenting part of a word or a degraded image speeds later identification of the complete version. These effects are short-lived, typically operating over seconds, they occur without any awareness of the connection, and they persist in people with severe amnesia who have no conscious memory of the earlier exposure at all, which was a key finding establishing that memory is not a single system. That clinical evidence is a large part of why the basic effect is considered secure.

The effect has been used as a tool as much as studied as a phenomenon:

• •It demonstrates memory without awareness, since the effect appears in people who deny having seen the material
• •It maps how concepts are related, since the strength of the effect between two words indicates how closely connected they are
• •It distinguishes perceptual from conceptual processing, since the two produce different patterns
• •It tracks language processing in real time, which is central to psycholinguistics
• •It reveals automatic activation, since the effect occurs faster than any deliberate strategy could operate
• •It provides a measure that does not depend on asking people what they think, which is valuable given how poor introspective report can be

A different set of studies claimed that exposure to a concept changes complex behaviour minutes or hours later, and this is where the trouble lies. Reported findings included that reading words associated with old age made people walk more slowly afterwards, that holding a warm drink made people judge others as warmer, and that exposure to money-related material made people less helpful. These results were widely publicised, entered popular books and were treated as established. Large replication attempts subsequently failed to reproduce several of them, and the failures were substantial rather than marginal, involving multiple laboratories, large samples and preregistered designs. The episode became a central case in the replication crisis, and it prompted serious examination of the statistical and publication practices that had allowed such findings to accumulate, including small samples, flexible analysis and the near-impossibility of publishing a failed replication at the time.

The experimental designs are worth understanding because the interpretation depends on them. The standard approach presents a prime followed by a target and measures how quickly a participant responds to the target, comparing related and unrelated pairs, with the difference in milliseconds being the effect. Timing between the two matters enormously, since very short intervals capture automatic processes while longer ones allow deliberate strategies that change what is being measured. Masked designs present the prime too briefly to be consciously perceived and then cover it, which isolates processing that occurs without awareness and requires careful checking that participants genuinely could not see it. Control conditions have to be chosen carefully, since a comparison against an unrelated word and a comparison against a neutral string test different things. These design details are where many disputes about specific findings actually live.

The reasonable position separates the layers rather than dismissing the subject. Short-term processing effects on recognition and reading are robust, replicated across decades and supported by converging evidence from clinical populations and brain imaging, and nothing in the replication difficulties touches them. Effects on complex social behaviour at longer delays are not established, and several specific claims should be regarded as unsupported. In between sit findings of uncertain status that require better evidence. The episode also produced genuine improvements, since preregistration, larger samples and multi-laboratory replication projects are now common in the field and were rare beforehand. The general lesson has application well outside psychology, since the pattern of a solid narrow finding being extended into dramatic broad claims, publicised heavily and then failing, is not confined to one discipline.

A stimulus speeds processing of a related one seconds later, which is highly replicated and appears even in people with no conscious memory of the exposure. Claims that exposure changes complex behaviour minutes later are a different matter, and several prominent ones failed large replication attempts. The episode drove preregistration and larger samples, and the narrow effect remains secure.