Psychology
Statisticians may be the butt of mean jokes, but they exercise hidden powers.
The way you deal with statistics can be very important.
As well as looking a like an extra in a Peter Jackson movie – I mean that fondly – Stuart Middleton has made a long and distinguished contribution to education. He is also a very funny after-dinner speaker.
At one such occasion, I remember him getting a good-humoured heckling from a statistician in the audience. Giving as good as he got, he replied along the lines of: “What would you know? A statistician is someone who likes numbers but doesn’t have the personality to be an accountant.” Well, I laughed.
I teach statistics. Indeed, I enjoy a good statistical back-and-forth as much as the next numbercruncher, because the way you do your statistics can be very important.
Ask Daryl Bem, an emeritus professor of psychology at Cornell University, whose recent research has sparked a debate that is less about what he says he has found than how he’s gone about testing it.
You don’t get a professorship at Cornell for nothing. Bem has been a professor there for almost 40 years and his self-perception theory has been cited more than 6000 times. It presents the rather counter-intuitive argument that people infer their attitudes – usually towards people or things they don’t already have a fixed view of – from their own behaviours.
My favourite illustration of this notion is an experiment published in 1967 by Stuart Valins and Alice Ray that involved showing snake-phobic subjects pictures of snakes and then giving them the chance to pet a real snake. Subjects who are given false feedback as they’re looking at snake pictures (“see, your heart isn’t beating as fast as you think it is!”) were able to approach the real snake more than the poor so-and-so’s who got to see their true heart rate.
Bem has become interested in the quest – some would say fool’s errand – to test the existence of “psi” phenomena: extrasensory perception (ESP), clairvoyance and other stuff of which parapsychology is made. In 1994, he published an analysis of 11 experiments using what is known as the ganzfeld procedure.
In this experiment, you lie down and relax with ping pong balls over your eyes in a soft-lit room and describe what comes into your head while someone in another room concentrates hard on one of four randomly selected pictures. Evidence for ESP would be finding that an independent rater picks out the right picture based on your narrative at a rate statistically higher than chance.
Bem’s review showed a small but statistically significant “anomalous information transfer” effect (32% hits against 25% expected by chance).
If you’re wondering why this hasn’t been settled once and for all – that some of those times you were thinking of your mother-in-law only to have her call on the phone aren’t just coincidence – it’s partly down to statistics. A few years after Bem’s work, another analysis came out that added more experiments and this time the results were non-significant. And around and around it went.
More recently, Bem has moved to another procedure. His new experiment involved asking participants to look at a computer screen showing two curtains and then predict behind which curtain there was a picture (as opposed to a blank wall). Some of the pictures were erotic in nature, but all of them were randomly selected only after the participant had made his or her guess. Bem found erotic pictures were picked 53% of the time, versus the 50% expected by chance.
It’s not a big effect, but as well as triggering a new wave of parapsychology research, this project has made people care about statistics in a way that more conventional subject matter just wouldn’t inspire.
And I could have been an accountant.
Some of the pictures were erotic in nature, but all of them were randomly selected.