Atheists Rule and Faithful Drool: A Meta-Meta-Analysis on Religiosity and Intelligence

A recent paper from Miron Zuckerman, Jordan Silberman, and Judith Hall in Personality and Social Psychology Review has received a lot of attention in the atheist blogosphere. It is a meta-analysis of 63 studies conducted between 1928 and 2012 on religiosity and intelligence. The authors examine the methods, experimental designs, and conclusions of these studies, both qualitatively and statistically. The punchline? A significant negative correlation exists between religiosity and intelligence.

Now don’t get me wrong; I love the findings. But I’ve seen a lot of comments mis-representing what is in the paper and I want to clarify what it does and does not say.

The Facts

The analysis defines intelligence as “ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.” The studies surveyed encompass many measures of intelligence, including university entrance exams (e.g., SAT, GRE), school performance (GPA), IQ, and IQ proxies (e.g., synonym tests, working memory tests, Mensa membership).

Religiosity is defined as “the degree of involvement in some or all facets of religion, [including] beliefs in supernatural agents, costly commitment to these agents (e.g., offering of property), using beliefs in those agents to lower existential anxieties such as anxiety over death, and communal rituals that validate and affirm religious beliefs.” The measures of religiosity included self-assessments, frequency of religious behavior (e.g., church attendance, prayer), and participation in religious organizations/denominations.

In all, 63 studies were surveyed, with 53 studies showing negative correlations (37 statistically significant) and 10 studies showing positive correlations (2 statistically significant). The strengths of correlation from these studies were tabulated along with many aspects of the samples, such as size, education level, age, religion, race, gender, and primary variable bias. The studies were evaluated for 3 types of primary variable bias: restriction of intelligence range (to high- or low-intelligence individuals), restriction of religiosity range (to high- or low-religiosity individuals), or restrictions of both. Unweighted random-effects analyses and sample-size-weighted fixed-effects analyses were done. All analyses showed significant evidence that the higher a person’s intelligence, the lower the person scored on the religiosity measures.

Secondary analyses indicated that:

  1. The correlation was stronger in older individuals. The authors suggested this was because religiosity at younger ages was more indicative of the religiosity of the family than the religiosity of the individual.
  2. Education level also showed a significant negative correlation with religiosity, though the correlation between intelligence and religion was much stronger.
  3. Controlling for education did not have a large effect on the intelligence-religiosity relation.

The Caveats

First, the type of intelligence this paper deals with is commonly called analytic intelligence. Recent theories tell us there may be other types of intelligence, including creative and emotional intelligence. This is an important point, as analytic intelligence is an effective predictor for some forms of achievement but not others.

The second caveat, related to the first, is the use of IQ in many of the studies. The media has overused and sensationalized IQ so much that just the mention of it is enough to send many people packing. Frequent criticisms are that it does not measure alternative forms of intelligence (see Caveat #1) and that it does not scale well to the extremes of intelligence or for those with mental disorders. Does this mean we should throw out the whole analysis? No. For one, most of the studies tracked multiple intelligence metrics. But also because, despite its flaws, IQ is the most studied intelligence metric. It has been correlated to countless factors and outcomes, such as morbidity, mortality, school performance, job performance, income, and crime. It has a high statistical reliability (scores generally agree with one another and across time) and in 1995 the the American Psychological Association specifically issued a report indicating that “the view that general intelligence factor (g) is a statistical artifact is a minority one” and reaffirming the correlations with school and job performance. Still, it’s important to keep in mind the limitations of IQ.

The third caveat is summed up best in the paper itself: “The present findings are correlational and cannot support any causal relation.” Basically, there’s not enough data to say that religion makes you stupid or that atheism makes you intelligent. The analysis presents two classes of studies that are consistent with the hypothesis that on average intelligent people tend to reject religion, but neither is that compelling to me. More research into this area would be needed to draw such a conclusion.

Lastly, it should go without saying, but I’ll say it anyway. The correlations found in these studies rely on large data sets. You cannot reasonably interpret this data to indicate that a given religious person is any less intelligent than a given atheist.


In the discussion section, the paper provides a number of theories as to the observed relationship between intelligence and religiosity. The three main ones are:

  1. Atheism is a type of nonconformity, and intelligent people are less likely to conform.
  2. Intelligent people are, on average, more empirically minded and less likely to accept beliefs not subject to test or reason. There is some evidence to support this, especially given that the intelligence measures used are precisely those on which more analytically-minded individuals would perform well. Another component of this theory is that rejection of theism in a largely religious community requires a long-term cognitive effort which intelligent individuals may be more likely to be able to sustain.
  3. Intelligence and religion are functionally equivalent. That is, religion satisfies a number of needs which are more likely to be met via other means in more intelligent individuals. The specific needs identified are self-control, self-regulation, self-enhancement, and a sense of security.

Conventional wisdom (and conceit) would side with Theory #2, but after thinking about it honestly for a while I can’t see any evidence to suggest that theory over the others. In fact, it seems possible that all three mechanisms might contribute.