Is astrology scientific? Reading the evidence

The short answer

Evidence for an astrological claim means observations that support that specific claim under a stated test. Astronomical accuracy, a meaningful reading and demonstrated predictive accuracy are different questions.

Meaning, calculation and evidence

A chart can use accurately calculated astronomical positions without its interpretations becoming scientifically validated. Calculating where Mars was and demonstrating that its position predicts a personality trait are separate tasks. Likewise, finding a reading useful for reflection does not establish its ability to identify an unknown person or forecast an event.

The studies selected below do not establish astrology as a reliable predictive method. They examine particular claims with different methods; none tests every astrological tradition or every possible use. This is a selected reading guide, not a systematic review of all research.

Throughout this library, symbolic interpretations are presented as interpretations. Our editorial approach explains how we distinguish those from calculations, history and research findings.

Birth charts and personality: Carlson’s test

Carlson’s 1985 double-blind experiment included a task in which astrologers matched natal charts to personality profiles from the California Personality Inventory (CPI). Each chart came with three profiles, one belonging to its owner. Random selection would average one correct first choice in three.

The paper reports 40 correct first choices in 116 matching trials, a result Carlson interpreted as consistent with chance. These are trials, not 116 astrologers. A separate self-recognition part was inconclusive: participants also struggled to recognise their own CPI profiles.

The scope matters. This was a particular chart-to-profile task, not a test of whether a consultation feels meaningful. Read Carlson’s original paper.

A published disagreement

In 2009, Suitbert Ertel challenged the design and analysis, including the suitability of the personality measure and the handling of rankings and ratings. His reanalysis argued that some results were more favourable to astrology than Carlson’s conclusion suggested, while explicitly stopping short of calling astrology empirically verified.

This is a reanalysis of existing data, not an independent replication with new participants. Read the critique alongside the original study; neither a familiar negative headline nor a favourable reanalysis settles every claim. Read Ertel’s critique.

Birth date is not a complete natal chart

Hartmann, Reuter and Nyborg’s 2006 study analysed two datasets containing more than 15,000 people in total. One covered middle-aged male veterans; the other covered younger adults. They investigated birth timing and measured intelligence, with personality data available only in the veterans’ dataset.

The authors found no support for the tested Sun-sign associations. Small seasonal intelligence differences did not survive correction for multiple comparisons. The younger sample had birth-month information rather than exact birth dates, and was not a second personality test.

It would be misleading to describe this as 15,000 complete birth-chart readings. It did not test individual charts using birth time, place, houses and planetary configurations. Its findings concern the birth-date relationships actually analysed. Read the methods and results.

Sun-sign compatibility and marriage

Helgertz and Scott’s 2020 study used Swedish register data covering 1968–2001. It compared observed marriages and divorce risks with Sun-sign compatibility classifications drawn from six websites.

Across those classifications, the authors found no consistent evidence that supposedly favourable pairings were more common among marriages or less likely to divorce. The pairing analysis included roughly 66,000 marriages; the divorce analysis followed a smaller set of roughly 46,000 marriages observed from their beginning.

The sample was restricted to Nordic-born, opposite-sex spouses within the study’s age limits. This was not a full synastry test using both partners’ complete charts. Marriage formation and divorce also do not measure every aspect of relationship quality. Read the open-access study.

Why “that sounds like me” is a different test

In Forer’s 1949 classroom demonstration, 39 students received the same personality description while believing it was individual feedback from a test. They generally rated it as a good fit. The demonstration illustrates how a shared description can feel personally accurate.

This is often discussed as the Forer or Barnum effect. It does not show that every personal insight is false, and the classroom exercise was not a test of complete natal charts. It shows why agreement with a description alone cannot establish that a method distinguishes one person from another. Read the original demonstration.

When reading an interpretation, try a second question after “Does this resonate?”: “Would this also fit many other people?” The first explores meaning; the second begins to examine specificity.

Four questions to ask of a study

1. What exactly is the claim? Name the input and outcome: Sun-sign pairing and divorce, for example, or a complete chart and a particular personality score. Do not swap one claim for another after seeing the result.
2. What is the comparison? Look for a chance baseline, shuffled pairings or another appropriate control. A success count needs something to be compared with.
3. How were clues and scoring handled? Were identifying details concealed? Were outcomes and analysis rules chosen before the results were known? Trying many comparisons makes isolated positive findings harder to interpret.
4. Does the result hold up? Ask about uncertainty, sample selection and independent replication. Reworking one dataset may reveal a useful issue; repeating a test with new data asks a further question.

A null result limits support for the tested claim under those conditions. A positive result needs scrutiny too: an association alone does not identify its cause. Keep both standards visible.

A small exercise: better than what?

Invented teaching example: a reader is given 12 independent matching tasks. Each has three equally likely profiles and exactly one correct answer. Under random guessing, the expected number correct is 12 × ⅓ = 4.

The reader scores five. Is that enough to establish a reliable ability?

For your next research note, write four lines: the claim tested, the comparison used, the result reported and the main limitation. Keep a fifth line for any personal meaning you find. That separates two worthwhile but different questions.