This is the fourth installment of my five part series on what we know (and don’t know) about intelligence and religion. These were inspired by questions about and challenges to a video I was in, particularly questions about my statement that the religious are not more or less intelligent than the non-religious. In part one, I introduced the series by briefly discussing some of the very diverse areas of one’s life that are impacted by your religious belief system. Part two focused on the relationship between religion and educational attainment. The religious beliefs of scientists were examined in part three.
This post will review what we know, based on research studies, about level of intelligence and religiosity. This post truly hones in on the central question of this series – are non-theists smarter than theists?
Level of intelligence is a variable that is rarely studied in connection with religion. As such, the literature connecting the two is sparse, with much of it being incredibly old. This post will begin with studies that evaluate college entrance exam scores rather than a direct measure of intelligence. Then, studies that correlate various measures of IQ with religious measures will be addressed.
College Entrance Exams & Religiosity
Several older studies have correlated religious measures with tests that were related to intelligence but did not directly assess it, such as college entrance exams. One of the earliest of these studies found a correlation of -.36 between religious conservatism and intelligence in college students, as determined by grades and tests of memory, problem-solving, and direction-following (Howells, 1929). A few years later, Carlson (1934) found similar results in 215 college seniors using Thurstone’s Attitude toward God scales and scores from college entrance exams, concluding that “there is a tendency for the more intelligent undergraduate to be sympathetic toward… atheism” (p. 208).
Gilliland (1940) conducted a similar study using Thurstone’s Attitude toward God scales and an unnamed intelligence measure. He found no significant correlation between intelligence and attitude toward God in the 326 college students surveyed. Aimed at supplementing Gilliland’s (1940) work, Gragg (1942) ran a study using 100 students attending denominational colleges in the South. Gragg also used Thurstone’s Attitude toward God scales as the religious measure, and he obtained scores from American Council on Education’s (ACE) Psychological Examination for College Freshmen as the intelligence measure. A negative but statistically non-significant correlation was found between ACE scores and attitude toward God.
Brown and Lowe (1951) evaluated students enrolled in either a state university or a Bible college to determine if there was a relationship between ACE college entrance exam and the students’ acceptance of Christian dogma. The degree to which students accepted fundamentalist Christian dogma was assessed using the Inventory of Religious Belief. Students were labeled as Bible students, Believers, or Non-Believers. Results showed that students who reported a strong belief in and acceptance of Christian dogma scored significantly lower on their ACE than those who rejected it. In fact, Non-Believers (ACE = 119) scored an average of 30 percentile points higher than did Believers or Bible students (ACE = 98 and 100, respectively).
Plant and Minium (1967) reported on a set of five different longitudinal studies correlating college entrance exams with religiosity measures. The Dogmatism scale, which assesses a person’s openness to other belief systems, and the Religious subscale of the Allport-Vernon-Lindzey (AVL) Study of Values were used as the religion measures. Depending on the college, intelligence scores were based on either the ACE college entrance exam or the School and College Ability Test (SCAT). Students who scored in the top 25% of participants at their school were placed in the high-aptitude subgroup, and those who scored in the bottom 25% were place in the low-aptitude subgroup. With the exception of a group of females in one study, those in the high-aptitude subgroups consistently scored lower on the Dogmatism scale and Religious subscale. In each group tested, re-administration two or four years later resulted in mean scores that were lower than initial test scores for both scales of interest, meaning that dogmatism and religiosity scores decreased during the respondent’s time in college. Among male and female participants, three of the eight groups administered these scales showed a significantly greater decrease over time in the high- as opposed to low-aptitude groups. Not only did the high-aptitude groups score lower on the dogmatism and religious measures than the low-aptitude groups, their scores tended to decrease more dramatically over time as well.
Those who consider themselves religious can vary greatly in their beliefs and attitudes, making classification of people as either religious or non-religious difficult. In order to make more accurate classifications among religious and non-religious people, Poythress (1975) evaluated 201 college students using the Dogmatism scale and a Likert format of the LAM scales. The LAM scales, which have strong validity and reliability, were developed to classify respondents’ beliefs as Literal (fundamentalist), Antiliteral (non-religious), or Mythological (interpreting religious word and doctrine as symbolic rather than literal). This was done to prevent religious respondents with non-fundamentalist beliefs from being misclassified as non-religious. Students’ SAT scores were used as measures of intelligence. All religious respondents, including both the Literal and the Mythological, were combined to form the “believers” group and were compared to the religious “skeptics,” revealing significant differences. Skeptics had significantly higher Verbal SAT scores (p < .01), Quantitative SAT scores (p < .05), and Total SAT scores (p < .01) than believers did. Skeptics also scored significantly lower on the Dogmatism scale (p < .005) than did believers. This study demonstrates that religious skeptics score higher on measures of academic achievement and have a more open, less authoritarian belief system than believers.
Directly Measured IQ & Religion
Few studies have directly assessed the relationship between intellectual ability and religious beliefs. One early study (Franzblau, 1934) used the Terman Group Test of Mental Ability to find the IQs of 378 Jewish adolescents enrolled in religious schools. Terman’s Mental Ability test yielded a “mental age” score, which was used to calculate an IQ score. He also administered the Religious Ideas Test, which was created for this study as a means of measuring the children’s tendency to accept traditional religious beliefs. Higher scores indicated greater acceptance of traditional religious beliefs. Analyses showed that participants’ Religious Ideas Test scores had a correlation of -.349 with “mental age” and a correlation of -.147 with IQ. The religious measures also correlated negatively (i.e., as one goes up the other goes down) with the adolescents’ chronological age and with grade and acceleration in school. Franzblau (1934) concluded that “lower scores on the Religious Ideas Test tend to be associated with higher intelligence, greater grade progress, and greater maturity” (p. 39).
In a more recent study of the relationship between intelligence and religiosity in adolescents, Francis (1998) evaluated a group of 711 fifteen- and sixteen-year old British students. The religious measure was the internally reliable Francis Scale of Attitude toward Christianity, which contains Likert scale items asking the students how they feel about God, Jesus, the Bible, prayer, and church. Raven’s Standard Progressive Matrices was the measure of intelligence used in this study (it should be noted that this test is a measure of nonverbal cognitive ability, not overall cognitive ability). The only significant relationship found in this study with attitude toward Christianity was its correlation with participant’s sex. Females reported a much more favorable attitude toward Christianity than did males, as is typically found in religious research (Stark, 2003). There was no significant relationship found between intelligence and attitude toward Christianity in this group of British adolescents.
Foy (1976) hypothesized that more intelligent participants would be less likely to subscribe to church dogma. To test this hypothesis, he administered the Wechsler Adult Intelligence Scale (WAIS) to 36 Caucasian adults living in Mississippi. Based on their IQ scores, participants were placed in the Low (less than 111), Middle (111-124), or High (greater than 124) intelligence group. The Religious Attitude Inventory was also administered, with higher overall scores indicating greater religious orthodoxy. The Inventory consisted of two subscales measuring “general religiosity” and “church attitude,” the multi-dimensionality of which Foy (1976) believed to be important in measuring religiosity. Results showed that intelligence groups differed significantly in their “church attitude” scores, with those in the High intelligence group scoring lower than those in the Middle and Low intelligence groups. Participants in the High intelligence group also scored significantly lower on “general religiosity” than those in the Middle or Low groups. Overall Religious Attitude Inventory scores showed a correlation of -.498 (p < .01) with participants’ full scale IQ. Results of this study demonstrate that adults with IQs greater than 125 are significantly less likely than those with lower IQs to hold orthodox religious beliefs, to dogmatically follow traditional church teachings, or to have high levels of religiosity.
Clark (2004) also evaluated intelligence and its relationship with religion in a multidimensional manner, differentiating between religiousness and spirituality. Participants were 77 undergraduates from the University of California, Davis. The Wechsler Adult Intelligence Scale-Third Edition (WAIS-III) was used to determine IQ. Self-reported SAT scores, which Clark (2004) found to be significantly correlated with IQ scores, were also included in analysis. Spirituality was measured using the Spiritual Transcendence Scale (STS), which assesses participants’ feelings of connectedness with humanity, belief that all life is unified, and satisfaction attained from prayer. Religiousness was measured based on responses to how participants classified themselves religiously and to questions concerning religious background and behaviors, with lower scores indicating that religion played less of a role in their lives. Several significant relationships emerged between intelligence and religiosity. The frequency of participants’ religious behaviors and their self-reported Quantitative SAT scores had a correlation of -.31 (p<.05). The Prayer Fulfillment subscale of the STS was the most significantly related to intelligence, correlating significantly and negatively with Verbal IQ (r = -.27), Verbal SAT (r = -.25), and Quantitative SAT (r = -.30). This study indicates that people who perform religious activities frequently tend to have lower scores on intelligence measures than their less religiously active counterparts.
A more recent but small-scale study was designed to assess the relationship between Protestant fundamentalist beliefs and level of intelligence by directly measuring these factors in college students (White, Scott, & Lack, 2010). They used the Weschler Abbreviated Scale of Intelligence (WASI) and measures of both overall religiosity and fundamentalism. These results found no difference between those high in fundamentalism (HF) and those low in fundamentalism (LF). The HF group obtained an average Full-Scale IQ (FSIQ) of 98.75, while the LF group scored an average FSIQ of 101.40. Furthermore, no significant relationships were found between any of the measures of religiosity and intelligence. Results from this study indicate that, as a group, college students who hold fundamentalist religious beliefs do not differ in intellectual ability from college students who do not hold fundamentalist religious beliefs.
Two recent studies have garnered a large amount of attention and dominate the Wikipedia page on this issue. Lynn, Harvey, and Nyborg (2009) reported on a significant correlation between national IQ and level of religiosity, where the less religious nations had significantly higher IQs. This study used estimates of IQ scores from Lynn and Vanhanen’s (2006) book IQ and Global Inequality, which showed national IQs ranged from highs of 110 or so in eastern Asian countries to lows of below 70 in sub-Saharan Africa (for reference, here in the US, IQ scores of 70 or below qualify someone as being mentally retarded). Right away, there are some obvious problems with these ranges,as a vast majority of people living in sub-Saharan Africa could not be mentally retarded, since population estimates for actual mental retardation are between 1-2%. Further concerns, including the misreporting of scores from original studies to the book, have also been raised (Mackintosh, 2006). These types of numbers (compiled from studies across the last 40 years and using a host of different measures) are likely the result of using non-culturally neutral (or culture-specific) IQ tests. Further, religiosity was simply “measured” as being the percentage of the population that reported not believing in god. These results are highly suspect, then, at any sort of individual level, given that they neither directly measured intelligence or religiosity/fundamentalism.
Kanazawa (2010) analyzed data from the American National Longitudinal Study of Adolescent Health, a US sample tested with the PPVT (Peabody Picture Vocabulary Test) and were asked: “To what extent are you a religious person?” The responses were coded “not religious at all”, “slightly religious”, “moderately religious”, and“very religious”. The results showed that the “not religious at all” group had the highest IQ (103.09), followed in descending order by the other three groups (IQs=99.34, 98.28, 97.14). An important issue with this study is that the PPVT only measures verbal aspects of intelligence (which is strongly correlated with overall intelligence, at around .80, but fails to account for other aspects of IQ), and that initial PPVT scores were used to look at religious beliefs seven years later (despite the fact that Kanazawa acknowledges only a moderate correlation between initial and later PPVT scores). Further, using a single question rather than a standardized measure to examine religious beliefs is problematic in that it does not allow for separation of fundamentalist from non-fundamentalist beliefs. Finally, the noted difference of six points in “IQ” scores represents less than half of a standard deviation, and is therefore unlikely to have major real-world impact (i.e., what psychologists refer to as “clinically significant differences” as opposed to “statistically significant differences”).
In part two of this paper, Kanazawa (2010) used General Social Survey (GSS) results to examine the same issue. Similar methodological problems are seen, in that religiosity was measured through the use of a single question and “intelligence” was measured by asking respondents to select a synonym for a word out of five candidates (a total of 10 times). Although this is certainly a measure of an aspect of verbal skills, which are correlated positively with intelligence, to simply rely on these 10 questions as a measure of “intelligence” is stretching. As in part one, the actual differences are also quite small (when examining the regression coefficients) and although statistically significant thanks to a huge sample size, are unlikely to be reflective of meaningful real-world differences. A more accurate title for this paper would be “Why Atheists and Liberals show slightly higher verbal skills in a few limited domains,” but that’s not as sexy as “Why Atheists and Liberals are More Intelligent.”
To address the problems with the above two studies and a number of others, Lewis, Ritchie, and Bates (2011) conducted a study using a large sample of US adults (over 2200 subjects). In it, they measured six domains of religiosity, the personality factor of openness, and “general intelligence” based on results of five sub-tests (measuring word list recall, working memory span, verbal fluency, inductive reasoning, and speed of processing). It should be noted that these were not part of a test of a standardized test of intelligence, but are certainly much better than the measures used in Kanazawa (2010) or Lynn et al. (2009). Their results found that intelligence was negatively associated with five of the six measures of religiosity, most strongly with fundamentalism. It is important to note that the effect sizes (an indicator of the level of strength between two factors) were statistically significant but only added a small amount of predictive validity on top of education and openness (accounting for only 5% of the variance in religiosity). In addition, openness positively predicted the spiritual elements of religion, but was negatively associated with fundamentalism.
Summary / TL;DR
Findings in studies correlating religious beliefs and intelligence are mixed and unclear. There is a clear correlation between lower religiosity and higher intelligence in the studies on a group of 1930s adolescent Jewish boys and for highly intelligent adults, but no relationship found in a more recent studies of adolescent Christian boys or college students. Methodological problems plague some of the major, large scale studies in this area, casting doubt on whether or not they were even measuring religiosity and/or intelligence. Some studies found higher college entrance exam scores for those who were lower measures of religiosity, while others found no relationship between the two. The extreme range of these studies’ measures, variables of interest, subjects, and publication dates makes synthesis of the literature difficult. Well-controlled, modern studies that use reliable and valid measures of both intelligence and various aspects of religiosity have shown either no relationship between the two or a very small relationship. More research on this subject is sorely needed with various populations in order to sort out potential mediating variables.
In the next and final post in this series, I will give a summary of the total literature I have reviewed and a road map for where researchers need to go next in order to more definitively provide an an answer to the question “Who’s more intelligent, theists or non-theists?”
(Many thanks to Julie White for allow the adaptation of her thesis – which I supervised – into this series. For a list of cited works, please contact me.)