The evidence on astrology

What it is

The empirical question: does astrology produce results better than chance under conditions where the astrologer cannot obtain the answer any other way?

This is narrower than "is astrology true", deliberately. Astrologers make many kinds of claim, some of which — that a chart is a language for reflection, that it structures self-examination — are not empirical claims at all and are not addressed here. What is addressed is the testable part: matching charts to people, predicting outcomes, and correlating placements with characteristics.

Where it came from

Serious testing began in the twentieth century, and one thing should be said before the results: much of the useful work was done by people who started sympathetic. Michel Gauquelin set out to validate astrology. Geoffrey Dean was a practising astrologer for years before becoming its most thorough researcher and critic. This is not a literature produced entirely by hostile outsiders.

How it is used

The main study designs are matching (given a chart and several personality profiles, identify the right one), correlation (do sun signs predict traits across large samples), and prediction (do forecast outcomes occur).

What we can and cannot say

The controlled matching tests have been negative. Shawn Carlson's double-blind study, published in Nature in 1985, gave participating astrologers natal charts and asked them to select which of three California Psychological Inventory profiles belonged to the chart's owner. The astrologers had helped design the protocol and agreed in advance it was a fair test. They performed at chance. Astrologers subsequently disputed the analysis, and some of the methodological criticism is fair — the design has been picked over for decades — but no better-designed replication has produced a positive result.

The time twins study is the hardest single result to argue with. Geoffrey Dean and Ivan Kelly used a British cohort of over two thousand people born in London within minutes of each other in March 1958, followed for decades and measured on more than a hundred variables — intelligence, anxiety, aggression, sociability, occupation, marital status, ability in art, music, sport and mathematics. If closely similar charts produce closely similar people, these subjects should resemble each other more than chance predicts. They did not, on any variable.

Sun sign and personality shows no correlation in large samples. A study by Hartmann, Reuter and Nyborg using around four thousand participants found no relationship between sun sign and personality or intelligence measures.

And then there is Gauquelin, which is messier than either side likes. From the 1950s Michel Gauquelin found a statistical excess of Mars in particular sectors of the sky at the births of eminent sports champions — the "Mars effect". The effect was small but persistent across his samples, and he found related patterns for other professions and planets while finding nothing at all for the standard zodiac, which he concluded was worthless.

Attempts to replicate went badly for everyone involved. The Belgian Comité Para tested it, obtained data resembling Gauquelin's, and disputed his interpretation. A committee associated with the American sceptics' organisation CSICOP conducted its own test and reported a negative result — and then an insider, Dennis Rawlins, published an account in 1981 alleging that the committee had mishandled the statistics and concealed data that initially favoured Gauquelin. That episode, which sceptics themselves refer to by the title of Rawlins's article, remains an embarrassment in the literature and is regularly cited by astrologers, fairly.

The later French CFEPP study, designed with Gauquelin's input and completed after his death, found no effect. The current position is that the Mars effect did not replicate and is most plausibly explained by selection effects in how sports champions were chosen for the samples — but that the sceptical organisation's handling of its own test was poor enough to have done real damage to its credibility on the question.

One confound is worth knowing, because it produces genuine effects that look astrological. Season of birth does correlate with a few real outcomes: slightly elevated schizophrenia risk for winter and spring births, and the relative age effect, whereby children born early in a school or sporting selection year are systematically over-represented at elite level. These are seasonal and administrative, not planetary. They are also a reason to be careful with any study that finds a birth-date pattern and concludes the planets did it.

What cannot be said is that astrology has been shown to be impossible, because that is not the kind of thing evidence shows. What can be said is that where it has made testable claims, under conditions preventing other routes to the answer, it has not performed — and that this has been tested repeatedly, sometimes by people who wanted the opposite result.

The honest closing note is that most practising astrologers today do not claim what these studies tested. The dominant modern framing is psychological and interpretive rather than predictive, which sidesteps the evidence by declining to make the kind of claim evidence could reach. Whether that is a legitimate refinement or a retreat is a question this site will leave with the reader.

Further reading

• Shawn Carlson, "A double-blind test of astrology" (Nature, 1985).
• Geoffrey Dean and Ivan Kelly, "Is Astrology Relevant to Consciousness and Psi?" (Journal of Consciousness Studies, 2003) — the time twins study.
• Dennis Rawlins, "sTARBABY" (Fate, 1981) — the insider account of the CSICOP episode, and necessary reading for anyone who thinks sceptical organisations are automatically the careful party.