Testing Astrology

Astrology makes claims that are, in places, testable: that a natal chart predicts personality, or that a practitioner reading a chart blind can match it to the person it belongs to better than chance. Those specific, operationalised claims have been tested, more than once, and have not replicated. The evidence that carries the weight of that conclusion is a different study from the one most often reached for, and getting that ordering right matters more here than in almost any other entry on this site.

The result that carries the weight

The strongest single result is a time-twins study. Geoffrey Dean and Ivan W. Kelly compared people born less than five minutes apart, across a large battery of traits, and found no hint of the similarities astrology predicts for people born that close together. The same paper reports a meta-analysis of more than forty controlled studies in which astrologers, working from charts, performed no better than chance even at a task as coarse as judging which of several people was the more extraverted ¹. Both findings are stated in the paper’s own abstract, and together they are this entry’s load-bearing evidence. Kenneth McRitchie published a reply arguing the case against astrological research had been closed prematurely, and Dean and Kelly answered it again in turn; omitting that published critique of this entry’s own load-bearing evidence would be exactly the selective citation this project exists to avoid, and a reader following the citation should expect an ongoing exchange rather than a single final word.

The study everyone reaches for

The study most often cited is Shawn Carlson’s double-blind test, published in Nature in 1985, which did not find that astrologers could match natal charts to their owners’ personality profiles better than chance ². It is the result almost every popular account on either side of the question reaches for first, and it is the wrong one to treat as decisive.

Suitbert Ertel’s reappraisal argues that Carlson’s design disadvantaged the astrologers taking part and that the original analysis broke with statistical convention by testing sub-samples rather than the full data set, while disregarding statistical power and effect size. On reanalysis of Carlson’s own data, the astrologers matched personality profiles to charts slightly better than chance: p = .054 on the three-way forced-choice test and p = .04 on the ten-point rating method ³.

Ertel’s own conclusion is the part of this that most often gets dropped by both sides of the argument, and this entry keeps it whole. His abstract reads: “The results are regarded as insufficient to deem astrology as empirically verified, but they are sufficient to regard Carlson’s negative verdict on astrology as untenable” ^3 abstract. Both halves of that sentence are load-bearing. A site whose whole value rests on being able to check its own claims cannot repeat “Carlson disproved astrology,” because the record does not support it; it can say that the specific, operationalised claims tested so far, chiefly the time-twins result and the forty-study meta-analysis above, have not replicated.

What this does not establish

A negative result on an operationalised test does not show that every possible formulation of astrology is false. Many formulations are not specified precisely enough to test at all, and practitioners reasonably object that chart interpretation, as it is practised, is not the mechanical input-output process that a matching test like Carlson’s or Ertel’s reanalysis of it assumes. That objection is fair, and stating it plainly is part of reporting the argument. An untestable claim is not thereby evidence for anything either, which is why this empirical argument, on its own, is the weakest of the reasons a reference of this kind declines to issue a personal reading: see The Barnum Effect for a separate question this site can answer more confidently, why a reading can feel accurate to the person receiving it regardless of whether the system producing it is valid.

The stronger argument against a chart having a single testable content is that the tradition disagrees with itself about which planet governs a given degree, which triplicity a given sign belongs to, and where an exaltation’s degree came from: see Essential Dignity. It also disagrees about a more basic question still, how to divide a chart into houses in the first place, with at least seven defensible answers in current use: see House Division. A body of doctrine that cannot say, with one voice, what a chart claims has no single claim left over for an experiment, however well designed, to test.