Conceived and designed the experiments: SH JTK MI MSC. Performed the experiments: JTK. Analyzed the data: SH JTK MI MSC. Contributed reagents/materials/analysis tools: MI MSC. Wrote the paper: SH JTK. Performed all subject recruitment, telephone screenings, and psychometric assessments prior to scanning: AC. Supervised our psychological assessment procedures and consulted on subject exclusions: SB. Gave extensive notes on the manuscript: MSC MI.
The authors have declared that no competing interests exist.
While religious faith remains one of the most significant features of human life, little is known about its relationship to ordinary belief at the level of the brain. Nor is it known whether religious believers and nonbelievers differ in how they evaluate statements of fact. Our lab previously has used functional neuroimaging to study belief as a general mode of cognition
We used functional magnetic resonance imaging (fMRI) to measure signal changes in the brains of thirty subjects—fifteen committed Christians and fifteen nonbelievers—as they evaluated the truth and falsity of religious and nonreligious propositions. For both groups, and in both categories of stimuli, belief (judgments of “true” vs judgments of “false”) was associated with greater signal in the ventromedial prefrontal cortex, an area important for self-representation
While religious and nonreligious thinking differentially engage broad regions of the frontal, parietal, and medial temporal lobes, the difference between belief and disbelief appears to be content-independent. Our study compares religious thinking with ordinary cognition and, as such, constitutes a step toward developing a neuropsychology of religion. However, these findings may also further our understanding of how the brain accepts statements of all kinds to be valid descriptions of the world.
Since the 19th century, it has been widely assumed that the spread of industrialized society would spell the end of religion. Marx
Given the importance of religion in human life, surprisingly little is known about its basis in the brain. The relevance of the brain's ventromedial dopaminergic systems to religious experience, belief and behavior is suggested by several lines of evidence, including the fact that a variety of clinical conditions related to dopaminergic dysfunction—mania, obsessive-compulsive disorder (OCD), schizophrenia, and temporal-lobe epilepsy—are regularly associated with hyperreligiosity
There have been a number of neuroimaging and EEG studies done on religious practice and experience—primarily focusing on meditation
As many have noted, religion cannot be reduced to a mere concatenation of religious beliefs. Every religion consists of rites, rituals, prayers, social institutions, holidays, etc., that serve a wide variety of purposes, explicit or otherwise
Our lab published the first neuroimaging study of belief as a general mode of cognition
We used functional magnetic resonance imaging (fMRI) to measure signal changes in the brains of thirty subjects—fifteen committed Christians and fifteen nonbelievers—as they evaluated the truth and falsity of religious and nonreligious propositions. For each trial either a religious statement (
Response time data were submitted to a repeated-measures ANOVA with belief (true, false) and statement content (religious, nonreligious) as within-subject variables, and group (nonbeliever, Christian) as a between-subject variable. Response times were significantly longer for false (3.95 s) compared to true (3.70 s) responses (F (1,28) = 33.4, p<.001), and also significantly longer for religious (3.99 s) compared with nonreligious (3.66 s) stimuli (F (1,28) = 18, p<.001). The two-way interaction between belief and content type did not reach significance, but there was a three-way interaction between belief, content type, and group (F (1,28) = 6.06, p<.05). While both groups were quicker to respond “true” than “false” on both categories of stimuli, the effect of truth was especially pronounced for nonbelievers when responding to religious statements (see Supplementary Information:
For both groups, and in both categories of stimuli, belief was associated with greater blood-oxygen-level-dependent (BOLD) signal in the ventromedial prefrontal cortex (VMPFC, see
Greater signal for belief compared with disbelief appeared in the ventromedial prefrontal cortex, lateral occipital cortex, and superior frontal gyrus. The bottom panel shows percent signal change from baseline in each of the clusters (vmpc = ventromedial prefrontal cortex; log = lateral occipital gyrus; sfg = superior frontal gyrus). Error bars represent standard error of the mean.
Brain region | hemi | x | y | z | Peak Z score |
Ventromedial prefrontal | L | −4 | 50 | −12 | 4.42 |
Superior frontal gyrus | L | −22 | 32 | 50 | 4.06 |
Lateral occipital cortex | L | −30 | −82 | 16 | 3.19 |
R | 44 | −88 | 12 | 3.61 |
The differences in VMPFC signal were due to a greater relative decrease in activation from baseline for the disbelief condition. Our finding of greater signal in VMPFC for belief compared to disbelief was significant in both Christians and nonbelievers for both religious and nonreligious stimuli, supporting a role for this brain region in the acceptance of truth-claims across content domains. A direct comparison of
The opposite contrast,
Brain region | hemi | x | y | z | Peak Z score |
Postcentral gyrus | L | −44 | −24 | 54 | 4.41 |
Superior frontal sulcus | R | 24 | 14 | 48 | 3.36 |
While the contrast of
(A) The MRI signal was greater when subjects evaluated religious statements compared with nonreligious statements in areas throughout the brain, including the precuneus, anterior cingulate, insula, and ventral striatum. (B) Increased signal was found for nonreligious statements compared with religious statements in several left hemisphere regions including the parahippocampal gyrus, retrosplenial cortex, temporal pole, middle temporal gyrus and hippocampus.
Brain region | hemi | x | y | z | Peak Z score |
Posterior cingulate | −2 | −22 | 30 | 6.27 | |
Precuneus | L | −10 | −72 | 36 | 6.38 |
Anterior cingulate | 0 | 30 | 26 | 5.09 | |
Frontal pole | L | −32 | 56 | 8 | 5.3 |
R | 30 | 60 | 10 | 4.78 | |
Anterior insula | L | −36 | 10 | −4 | 4.09 |
R | 34 | 12 | −8 | 3.59 | |
Middle frontal gyrus | R | 42 | 40 | 26 | 4.14 |
Lateral occipital gyrus | L | −32 | −62 | 48 | 5.03 |
R | 32 | −54 | 38 | 3.93 | |
Intraparietal sulcus | L | −32 | −56 | 40 | 5.09 |
R | 32 | −54 | 38 | 3.93 | |
Ventral striatum | L | −16 | 20 | 0 | 3.3 |
L | −14 | 12 | −6 | 3.53 | |
Inferior frontal gyrus | L | −50 | 10 | 2 | 3.9 |
Superior frontal gyrus | R | 12 | 16 | 64 | 4.38 |
Thalamus | 2 | −24 | 6 | 3.42 | |
Cerebellum | 2 | −72 | −12 | 3.23 |
Another key region that appears to be preferentially engaged by religious thinking is the posterior medial cortex. This area is part of the previously described resting state network that shows greater activity during both rest and self-referential tasks
The opposite contrast,
Brain region | hemi | x | y | z | Peak Z score |
Ventromedial prefrontal cortex | L | −4 | 22 | −18 | 4.1 |
Superior frontal gyrus | L | −20 | 34 | 52 | 3.54 |
Middle temporal gyrus | L | −56 | −6 | −16 | 4.52 |
Parahippocampal gyrus | L | −26 | −40 | −14 | 5.16 |
Retrosplenial cortex | L | −14 | −52 | 4 | 4.62 |
Orbital frontal | L | −40 | 38 | −16 | 4.35 |
Temporal pole | L | −48 | 16 | −34 | 3.59 |
Hippocampus | L | −22 | −10 | −22 | 3.39 |
Finally, among our religious stimuli, the subset of statements that ran counter to Christian doctrine yielded greater signal for both groups in several brain regions, including the ventral striatum, paracingulate cortex, middle frontal gyrus, the frontal poles, and inferior parietal cortex (see
There were significant differences between blasphemous and non-blasphemous statements in both groups. These are regions that show greater signal both when Christians reject stimuli contrary to their doctrine (e.g. “The Biblical god is a myth”) and when nonbelievers affirm their belief in those same statements (pc = paracingulate gyrus; mf = middle frontal gyrus; vs = ventral striatum; ip = inferior parietal lobe; fp = frontal pole). Error bars represent standard error of the mean.
Brain region | hemi | x | y | z | Peak Z score |
Paracingulate gyrus | R | 2 | 40 | 34 | 3.5 |
Ventral striatum | R | 14 | 16 | 0 | 3.52 |
R | 16 | 14 | −8 | 3.61 | |
Middle frontal gyrus | R | 46 | 30 | 34 | 4.32 |
L | −48 | 36 | 22 | 3.07 | |
Frontal pole | L | −36 | 64 | 2 | 3.77 |
R | 32 | 64 | 4 | 4.5 | |
Inferior parietal lobe | L | −42 | −48 | 46 | 3.58 |
R | 48 | −48 | 46 | 3.39 |
Nearly a century of opinion polling attests that 70–85 percent of Americans profess not merely a belief in a generic God, but a belief in highly specific, religious propositions: that the Bible is the word of God (whether literal or “inspired”), that Jesus Christ will physically return to earth at some point in the future, that Satan exists and leads people to sin, that prayers actually get answered, etc. The failure to subject such beliefs to rational criticism may be one reason for their survival. But, as Boyer
Boyer may be correct in saying that we have cognitive templates for religious ideas that run deeper than culture (in the same way that we appear to have deep, abstract concepts like “animal” and “tool”). We may, in fact, be what Bloom
A variety of experiments suggest that children are predisposed to assume both design and intention behind natural events—leaving many psychologists and anthropologists to believe that children, left entirely to their own devices, would invent some conception of God
Because our minds have evolved to detect patterns in the world, we may tend to detect patterns that aren't actually there—ranging from faces in the clouds to a divine hand in the workings of Nature. Hood
And yet, however predisposed the human mind may be to harboring religious beliefs, it remains a fact that each new generation receives a religious worldview, at least in part, in the form of linguistic propositions—far more so in some societies than in others. Whatever the evolutionary underpinnings of religion, it seems unlikely that there is a genetic explanation for the why the French, Swedes, and Japanese tend not to believe in the God of Abraham while Americans, Saudis, and Somalis do. The importance of religious doctrines that purport to be true, and their subsequent acceptance as true by great numbers of human beings, seems indisputable.
Recent attempts to study the neural correlates of religious belief have either suffered from a lack of a nonreligious control condition
The data reported above present statistical tests of the reliability of signal changes occurring throughout the brain as a function of the stimuli and their associated behavioral responses. However, these data are of greater value when interpreted against related results in the neuroscientific literature. Such a discussion necessarily entails “reverse inference” of a sort often considered problematic in the field of neuroimaging
The contrast,
Our study was designed to produce high concordance on nonreligious stimuli (
In our earlier study of belief, we found anterior insula signal to be associated with the contrast
Our previous study of belief, in which we explicitly modeled uncertainty, revealed greater signal in the ACC and adjacent regions of the superior frontal gyrus in the uncertainty condition. Given that our signal maps in the contrast
The contrast
The opposite contrast,
Finally, there were several regions that showed greater signal in both groups in response to “blasphemous” statements (i.e. those that ran counter to Christian doctrine). The ventral striatum signal in this contrast suggests that decisions about these stimuli may have been more rewarding for both groups: Nonbelievers may take special pleasure in making assertions that explicitly negate religious doctrine, while Christians may enjoy rejecting such statements as false.
There is, of course, no reason to expect that any regions of the human brain are dedicated solely to belief and disbelief. Nevertheless, our work suggests that these opposing states of cognition can be discriminated by functional neuroimaging and are intimately tied to networks involved in self-representation and reward. Despite vast differences in the underlying processing responsible for religious and nonreligious modes of thought, the distinction between believing and disbelieving a proposition appears to transcend content. These results may have many areas of application—ranging from the neuropsychology of religion, to the use of “belief-detection” as a surrogate for “lie-detection,” to understanding how the practice of science itself, and truth-claims generally, emerge from the biology of the human brain.
We enrolled 54 subjects who were (1) between the ages of 18–30, (2) not taking anti-depressants, (3) neurologically healthy, (4) free of obvious psychiatric illness or suicidal ideation, and (5) native speakers of English as their first language. These inclusion/exclusion criteria sought to remove confounding effects of (1&2) age- or drug-related hypometabolism in the brain, (3) structural and functional anomalies due to illness or injury, (4) differences in psychological health, and (5) differences in linguistic processing. Subjects with implanted metal are routinely excluded from experiments using magnetic resonance imaging (MRI) for reasons of safety. All subjects gave written, informed consent according to the guidelines of the UCLA Human Subjects Protection Committee.
In order to implement these inclusion/exclusion criteria, subjects were screened by means of a telephone questionnaire. This questionnaire allowed us to isolate the variable of religious belief, in an effort to admit only dedicated Christians and nonbelievers into the protocol.
Once we had two groups of subjects (Christians and Nonbelievers), we attempted to balance these groups with respect to 1) general reasoning ability, 2) age, and 3) years of education. We also sought to exclude all subjects who exhibited signs of psychopathology. To this end we assessed subjects' general intelligence using the Weschler Abbreviated Scale of Intelligence (WASI) and screened for psychopathology using the Brief Psychiatric Rating Scale (BPRS). Subjects were not given the results of these tests.
Thirteen subjects were excluded on the basis of these psychological assessments. This left us with 41 subjects (19 female, 22 male; 20 Christians; 21 Nonbelievers). Forty of these participated in the fMRI portion of our study, but ten were later dropped, and their data excluded from subsequent analysis, due to technical difficulties with their scans (2 subjects), or to achieve a gender balance between the two groups (1 subject), or because their responses to our experimental stimuli indicated that they did not actually meet the criteria for inclusion in our study as either nonbelievers or committed Christians (7 subjects).
While gradations of belief are certainly worth investigating, our experiment sought to characterize belief and disbelief in their purest form. It was, therefore, essential that we exclude subjects who could not consistently respond “true” or “false” with conviction. Our decision to exclude data from subjects whose answers were not consistent with our pre-screening criteria was part of our original design and was not made based on any evaluation of the scanning data (the fMRI data from these subjects were never analyzed). While we adopted the criteria of excluding anyone who responded to one category of statements with less than 90% predictability, the 7 subjects who were excluded on this basis had responses that ranged from 22% to 43% discord with the expected responses. (For instance, one subject who passed our initial screening as a nonbeliever actually agreed with 43% of the religious Christian statements once inside the scanner.) Because our telephone questionnaire needed to screen for all relevant variables (age, native language, MRI safety issues, etc.), it contained only a very abbreviated assessment of belief. Thus, the high exclusion rate at this later stage of the experiment represents the failure of our brief screening procedure to accurately assess a person's religious beliefs, rather than a bias in our approach to data analysis. These exclusions ensured that our final group of subjects did, in fact, strongly believe/disbelieve our religious stimuli. We note, however, that the subjects retained in this experiment do not represent the full range of religious commitment found in the general population.
Our final study consisted of data acquired from 30 subjects (15 Christians; 15 Nonbelievers; 7 men and 8 women in each group). The mean full-scale WASI scores, years of education, and ages for the groups appear in
GROUP | WASI | EDUCATION | AGE |
125.6 | 15.1 | 22.0 | |
124.7 | 15.1 | 21.6 | |
127.6 | 15.3 | 22.7 | |
123.9 | 14.9 | 21.4 | |
123.7 | 14.6 | 21.3 | |
125.5 | 15.6 | 21.9 |
Once inside the scanner, subjects were presented with a series of short statements through a video-goggle display (Resonance Technology, Inc). After reading each statement, they were asked to evaluate its truth content with the press of a button, indicating “true” (belief), “false” (disbelief), and “undecidable” (uncertainty). The presentation of stimuli was self-paced. Stimuli were drawn from two categories, religious and nonreligious. All statements were designed to be judged easily as “true” or “false” (the response of “undecidable,” while available to subjects, was not expected).
Within each category, we attempted to balance the stimuli with respect to semantic structure and content. Strict balancing across categories was not possible, however, as the two categories differ with respect to content, in principle. For the purposes of stimulus design (not presentation) we generated our statements in groups of four (true and false; religious and nonreligious):
Christians and Nonbelievers were expected to respond identically to nonreligious stimuli and to be discordant for all religious trials. The nature of the questions, along with a telephone screening protocol that selected for nonbelievers and committed Christians, more or less ensured that subjects' responses would segregate in this way (see Supplementary Information:
Prior to scanning, all stimuli were tested to ensure that they would function appropriately in our experiment. For this purpose, we created several sets of candidate stimuli and solicited responses from the nonbelievers and Christians on the Internet. For each statement the number of respondents averaged around 5000, 80–90% of whom were nonbelievers. The numbers of committed Christians responding to each statement ranged from 254–787. Participants were asked to judge the veracity of each statement using a Likert scale (ranging from 1-“strongly disbelieve” to 5-“strongly believe”). In selecting stimuli for this study, we retained only those statements that reliably elicited ratings of 1 or 5 in these surveys. We kept only those religious statements that segregated along the lines of stated belief (Christian v. nonbeliever), and only those nonreligious statements that showed no such interaction.
Each functional scan was balanced with respect to category content (religious/nonreligous) and response valence (true/false). After scanning, subjects were asked to review their recorded responses to all statements to ensure that they reflected their actual beliefs at the time of scanning. Erroneous responses, responses of “undecided,” or those statements which, upon debriefing, could not be clearly judged by subjects to be “true” or “false” were excluded from subsequent data analysis.
The stimuli were presented in an order optimized to produce maximal signal differentiation and to ensure temporal jitter between trials using a genetic optimization algorithm
All scanning was performed on a Siemens Trio 3T scanner. Each subject received three functional scans of approximately 6 to 10 minutes in length. Functional images were acquired in the AC-PC orientation using T2*-weighted echo-planar scans (TR = 2000 ms, TE = 35 ms, flip angle = 80 degrees, FOV = 192×192 mm, slice thickness = 3 mm, number of slices = 29, inter-slice gap = 1 mm, bandwidth = 3256 Hz/pixel). FMRI data processing was carried out using FEAT (FMRI Expert Analysis Tool) Version 5.98, part of FSL (FMRIB's Software Library,
All functional data were analyzed using FSL. We performed standard preprocessing—slice timing correction, motion correction, brain extraction, spatial smoothing (using a 5 mm kernel), high-pass filtering, and pre-whitening—prior to contrast modeling. Individual responses were analyzed in an event-related manner. We modeled four types of trials with separate regressors: nonreligious true, nonreligious false, religious true, and religious false. Since response time varied among conditions, we also included in our model an additional regressor to account for the effects of response time. This regressor had a height equal to the response time for each trial, and was orthogonalized with respect to the other four regressors. The six motion correction parameters were also included as additional regressors. Our maps of blood oxygen level dependant (BOLD) signal changes were the result of pairwise contrasts between each of the task conditions. Statistical images were thresholded using clusters determined by Z >2.3 and a corrected cluster size significance threshold of p = 0.05.
The full set of stimuli used in this experiment.
(0.05 MB DOC)
(0.04 MB DOC)
(0.08 MB TIF)