An article in the current Economist reopens the debate about ‘priming’ and, indeed, about the validity of much scientific research that is published today. In psychology priming refers to preparing a participant before a task by setting up perceptual links in advance that will affect the outcome of the task. Someone shown in advance the word ‘yellow’ might be somewhat faster to recognize the word ‘banana’ because yellow and banana are associated in memory so the one ‘primes’ the other.
The problem is that the published experimental results are not being replicated by those who repeat the tests. A 1998 study found that those taking IQ tests did better if asked beforehand to think about a professor than did those asked to think about a football hooligan. The journal PLoS ONE reported in April that nine separate experiments had not managed to repeat these results.
Nobel Laureate Daniel Kahneman, a Princeton psychologist, has spoken of “a train wreck looming” in the field because of a storm of doubt about the results. He believes that researchers in the area might find it difficult to gain jobs because it risks being discredited. He wants to establish a circle of researchers who will repeat each other’s results to lend them credibility.
Priming has been used fraudulently by some researchers anxious to achieve preconceived results. They prime their subjects to increase the likelihood of the desired outcome, but without admitting this in their findings. When later researchers attempt to repeat the work without the advance priming, they fail to replicate the results. Priming has also been used fraudulently in legal trials, by priming witnesses in advance of their witness-box appearance. This has been exposed in cases involving recorded evidence from children. In one case when the full recording was later shown, including the priming, the case was instantly dismissed.
The issue has significance in the current attention being given to ‘nudge’ politics. In such cases people retain free choices over their actions but are ‘nudged’ by alterations to the ‘choice architecture’ to make decisions that are thought to be in their best interests. In some cases people are primed in questionnaires by being first given information or questions designed to influence their own decision. Opinion pollsters have long known that brief statements ahead of a question can influence the answers given. But if the published findings of priming research are themselves now doubted, then actions taken on the basis of them lose their validity and justification. The train wreck looms.
Filed under: Updates |
Madsen Pirie 
The old adage ‘ you can’t teach an old dog new tricks ‘ holds an element of truth where priming is ‘ the order of the day ‘. Us older brigade have ‘ been there and done it ‘ and generally are not caught out twice. Having said this there is the saying/adage that ‘ there is no fool like an old fool ‘. I prefer to believe that wisdom comes with age and my decisions are based on experience and not some psychological trickery which will result in regret and despondency.
What I had intended to post was the following, before my confusion with cut and paste.
Is there really anything new here? Be reminded by Yes Prime Minister
As the linked article discussed too briefly, the real problem is the difference between natural science (eg physics and chemistry) and social science (which includes economics and psychology).
Natural science benefits us through our knowledge of its repeatability (ie the time invariance of the laws of physics and its derivatives). So, with effort, experiments in natural science can be conducted in sufficiently narrowly defined circumstances that (given the repeatability – ie time invariance) allow us to define laws that will reliably and usefully predict what will happen.
There is no time invariance in social science. Today is different from yesterday; and yesterday can never be recreated! The wishes of social scientists to obtain authority by reliable prediction of the future cannot be fulfilled in the same way and to the same extent as natural scientists and engineers obtain and keep authority.
Such predictive power as there is in the social sciences is short-lived and subject to considerable approximation. That is: its usefulness fades rapidly with time and is prone to significant error.
Around a year ago, I came across the following book on a related subject: The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice and Lives. It makes an interesting read, but is heavy going in logical analysis and understanding of probability and statistics (though there is a helpful absence of equations): what does it mean rather than how does one compute it.
Best regards