I have just come across a delightful experimental finding in Ben Goldacre's "Bad Science" (required reading for anyone who aspires to be a sceptic.) This was a study on our tendency to see cause-effect relationships that aren't there. Subjects were presented every day with a figure purportedly representing a real-world event, and asked to recommend an intervention to make it better next day. By the end of the trial most of them reckoned they had found how to control the event concerned. The catch was, there was no event: the figures were a random string generated before the trial began. Nothing they did could possibly affect it, but they convinced themselves that they could and did.

So far, so interesting in general, but by accident the researchers also discovered something about our prejudices of another kind. The pseudo-variable they offered the subjects was supposedly a child's times of arriving at school, and they were asked to try whether rewards or punishments would do better at making the child arrive on time. By the end of the trial 70 percent of the subjects believed they had proved that punishments were more effective.

There was no child, just a random number table: any effect had to be in the subjects' minds. Yet a solid majority saw a benefit from punishment. Remember that next time someone says "I don't need research to tell me if caning works: I know from experience."