cbostrum@watdaisy.UUCP (Calvin Bruce Ostrum) (07/21/83)
Yes, there have been a number of attempts to test astrology's claims in a scientific manner. Yes, the skeptical inquirer has covered many of these and even sponsered one of them. It has been a while since I read their coverage, however. The crackpot (ahem, controversial) psychologist Hans Eysenck was drawn into it for a while. He specialises in finding really wierd explanations for correlations; for example, he used to claim that rather than smoking causing cancer, there was one geneotype that disposed individuals to both otherwise unconnected phenotypes of becoming a smoker and developing cancer. He has since withdrawn this claim, as well as his astrology claim, the latter when he discovered that the subjects knowledge of astrology was sufficient to explain the correlation (these subjects judged themselves as to which horoscope they fit. When he took people who professed no knowedge of typical astrological relationships, the correlations vanished). The major investigators are the French couple Gaugelin. From the reports in the SI, I think it is clear that their study is very suspect. But SI itself has to be taken very carefully since it has a venomous hatred of the paranormal (I suspect at least in the case of Gardner and Randi this is due to their "traditional" religious beliefs that do not jibe with astrology, does anyone know anything about this possibility?). The Gaugelin test used something called the rising sign, which varies very quickly and therefore is not subject to the criticism that Tim Maroney gave. (By the way, his suggestion for a study has very clear flaws in it: a skilled "astrologer" could quite easily lead the conversation to his advantage. Better to let friends of the people involved attempt to match up the horoscopes). The Gaugelins took a number of champion athletes and found a statistically signifigant deviation from the chance distributions in comfirmation of the predictions of astrology. Admidst the paranoid and venomous attacks of SI it is possible to discern problems with the result. The central one is a variation on the following theme: when the results are not signifigant, the G's maintain that the athletes included are not good enough, and that the thing only works for true champions. Thus, the lessor champions have to be dropped. It is not at all clear how they do this, and it is clear that there is no principled dividing line between the true champions and the lessor champions. This insistence of the G's gives them a much better chance (certain?) of finding their correlations. On one occasion they actually went so far as to exclude all basketball players (who didnt show the "effect")! If we imagine that the correlation is random, using a random walk principle or something similar I imagine (I dont know this type of math myself, maybe someone can enlighten us) that with some simple idealisation one could prove that it is always possible to find a "statistically signifigant" correlation by continuously tightening the definition of the independent variable until it was found, and stopping just at that point. It seems that this is what the G's have done. There was a bit of a scandal in SI about intestine strife between members of the SI team investgating this. One of them claims to have been kicked out of CSICOP (the org pusblishing SI) when he criticised the investigation of SI as being shoddy statistics arguing from a foregone conclusion point of view. From the contentious info I have, I side with him at this point. I do not like the attitude of Randi for one, altho he does provide a useful service. Calvin Bruce Ostrum, Computer Science, University of Waterloo ...{decvax,allegra,utzoo}!watmath!watdaisy!cbostrum