Monday, 9 September 2013

What a gold earring can tell us about the paranormal

Seriously StrangeThe task was easy - find a hidden gold earring. The materials provided to assist with the search were a map of the area where the earring was hidden, a dowsing pendulum and the other earring of the pair! Should be easy, yes? Well, it was a good thing that the owner of the earring remembered where it was because no one found it on the map.

For those who weren't there, I'm referring to one of the tasks at the Paranormal Games event at ASSAP's Seriously Strange conference last weekend. A question that intrigued me, and others, was what were the odds of locating the earring by chance. The map was laid with a regular grid of squares. To 'win' someone had to name the correct grid square. So what are the odds?

Let's say, for the sake of argument, that it was a grid of 20 by 20 squares. That would give a total of 400 grid squares to choose from. So you'd have a 1 in 400 chance of getting the right answer by pure chance. Well, maybe if you were a random number generator you might. But people don't actually think randomly, even if it can feel like it sometimes. So what ARE the real odds, then?

To find out, we decided to plot people's actual answers on the map itself. Assuming people were not using any paranormal ability then the plotted answers ought to reflect any general bias among the group of 65 or so participants. Given that no one got even near the right answer, it's probably safe to say that they weren't using any paranormal ability, though this could be tested with a control experiment which I'll discuss later.

Anyway, there were two fairly obvious biases in the distribution of answers. Firstly, most were in or around the buildings on campus, general avoiding the parkland areas. Secondly, most answers were away from the edges of the map with a noticeable concentration in the central area. What might cause these apparent biases? Obviously, it's only speculation but I wonder if, unconsciously, people are asking themselves, if I was hiding the earring, where would I put it? And the answer is somewhere near or in a building where it be easy to remember and retrieve. Out in the parkland there would be fewer useful landmarks available. And the areas near the edge of the map may just be too far to bother going! But that's just speculation, of course. It would be interesting to research reasons for any biases!

So back to the odds question. Given the observed biases, it is clear that the odds of finding the earring in any particular square will vary depending on its location. If the earring was in a central location near a building, the odds of getting a correct answer would be dramatically lower than if it was near the edge of the campus in parkland. The earring was, in this case, near the edge of the campus, though not far from a building. Had it been placed centrally, no doubt at least one person, possibly several, would have got the answer right. And remember, all of this assumes no paranormal ability is used!

The 'target dependence' is a problem with some psi experiments. We had a picture test which also illustrated the problem. Participants had to guess the contents of a concealed picture. Many of the answers received were of boats, houses, trees or people. Had the target actually been any of those, we would have had many correct answers by pure chance. A target of, say, the moon would, by contrast, get few hits by blind chance.

Does this matter? Yes, because experimenters use odds to calculate how likely it is that someone got a correct answer by pure chance. And if the odds are millions to one against, many people would say it was impressive evidence of the use of psi. So anything the alters those odds can dramatically affect whether a result appears paranormal or not. One obvious way to correct such biases is to repeat the experiment a huge number of times with a vast variety of different targets. But this could still leave a bias present, depending on the details of how the experiment is run. However, the gold earring test suggests another possible way forward - a control experiment.

With a control experiment you might get each subject to do a number of identical tests, some with and some without any actual target. Suppose we do the map dowsing test, for instance. Each subject would be presented with a map to dowse twice with a gap of, say, an hour, between the two trials. However, only in one of the trials would there actually be a target present in the physical area depicted by the map. And nobody, apart from the person hiding the target, would know in which hour the target was present or where it was. You could use the distribution of guesses on the 'non-target trial', where only chance is assumed to operate, to work out the odds of someone guessing where the target was by pure chance.

I haven't thought this through in practical detail and I'm sure there are problems with the scheme that would need ironing out but you get the idea. You are no longer assuming humans guess randomly, which they obviously don't, but seeing where they might actually tend to guess without any target being around. It might well be possible to produce realistic odds for any particular target square taking into account the observed distribution of guesses in the control trial. I suspect the odds for many squares in the test described here would be significantly less than 1 in 400, thereby making hits far less dramatic in those locations.

No comments:

Post a Comment