Is science for atheists? Perceived threat to religious cultural authority explains U.S. Christians’ distrust in secularized science

Abstract

A history of perceived conflict between religion and science persists in the U.S. “culture wars” that juxtapose religious and secular worldviews. As modern societies grow increasingly secular, religion is often deemed an impediment to science-based policy-making, whereas science is increasingly associated with atheism. In the present research, we addressed how perceived science–atheism associations affect U.S. Christians’ attitudes toward science. In study 1, participants’ own estimates of atheists’ prevalence in science uniquely predicted distrust in science, and study 2 revealed a causal effect of perceived prevalence of atheists on distrust in science. Studies 3–4 (the latter preregistered) manipulated the concept of science–atheism associations more generally, revealing the same effects. These effects were mediated by the belief that scientists are anti-religion, not by moral distrust toward atheists or fundamentalist religious beliefs. Hence, Christians’ institutional distrust in secularized science may derive largely from a perceived threat to the cultural status of religion.

Keywords

atheism attitudes toward science intergroup threat religion

“The cultural authority of science has become so totalitarian . . . that everybody has to have science on their side in order to win a debate.”

Historian of science, Mark Largent (in Green, 18 January 2018)

“All scientists should be militant atheists.”

Astrophysicist, Lawrence Krauss (8 August 2015)

Are science and religion perceived to be in conflict? And if so, how might this perception affect science’s cultural status in a world in which most people are religious believers (Pew Research Center, 2015a)? Such questions are especially concerning given the science–religion debate in the United States, a nation at the forefront of scientific research, yet also among the most religious nations in the developed world (Pew Research Center, 2015a). Indeed, various measures of religious belief strongly correlate with distrust in science among U.S. respondents (e.g. Simpson, Rios, & Cowgill, 2016).

Such findings warrant further attention partly because distrust toward science in the United States appears to be rising (e.g. Gauchat, 2012), while scientific illiteracy is pervasive (e.g. the National Science Foundation found evidence that 26% of Americans believe the sun revolves around the Earth, and only 39% believe in the Big Bang; Neuman, 2014). Moreover, among industrialized nations, the United States exhibits low rates of belief in Darwinian evolution and high rates of belief in creationism (e.g. Ipsos Global @dvisory, 2011). Importantly, low scientific literacy is associated with negative attitudes toward science (Gauchat, 2011).

Here, we address U.S. believers’ trust in science amid apparent tensions between religion and science. We investigated how religious believers’ attitudes toward science are shaped by perceived science–atheism associations (which we broadly refer to with the term secularized science). We focused specifically on attitudes toward atheism–science associations because atheism (more than, for example, Judaism, Hinduism) is increasingly associated with science in Western culture, and science often conflicts with religion regarding explanations of major phenomena (e.g. the origins of the universe).

1. Introduction

Science’s uphill battle

Science stands in sharp distinction from some of the most deep-seated themes of human history and psychology. Most, if not all, civilizations began with some form of religious temple (Norenzayan, 2013), and subsequently, religious institutions and shared theistic beliefs may have been adaptive for interpersonal trust and in-group identity (Norenzayan et al., 2016). Moreover, ample evidence suggests that religious beliefs have evolved even at the individual, psychological level. Supernatural beliefs likely emerged as byproducts of cognitive adaptations for survival in our ancestral past, such as false alarms in perceiving agency (e.g. perceiving agency not just in snakes but also inanimate objects that look like snakes), that led to perceiving supernatural agency behind natural phenomena (Atran and Norenzayan, 2004).

Religious belief, therefore, runs deep in both culture and psychology. Scientific thinking, however, faces numerous cognitive obstacles and requires skills difficult to develop. This presents an uphill battle when science competes with religion for cultural status. Indeed, perceived conflict between religion and science is widespread (Pew Research Center, 2015c). Moreover, recent evidence suggests that such perceived conflict may emerge from basic psychology, not just culture. Preston and Epley (2009) found that scientific and religious explanations for phenomena such as the universe’s origins compete in a zero-sum manner: experimentally discrediting scientific explanations decreased (increased) automatic positive evaluations of science (religion), and experimentally elevating theistic explanations yielded similar results.

Science, religion, and atheism

Amid U.S. culture wars pitting secular values against religious values, and in a society that is mostly religious (Pew Research Center (2015b) estimated ~92%; Gervais and Nalje (2018) estimated ~74%), an increasing cultural association between science and atheism may undermine society’s trust in the institution of science. And concerns about such atheism–science associations may indeed be warranted, as atheists are far more common in science than in the general public: roughly 41% of the American Association for the Advancement of Science (AAAS) members do not believe in a god, universal spirit, or higher power (Pew Research Center, 2014). Furthermore, atheism is increasingly associated with science in public discourse (e.g. Dawkins, 2006; Krauss, 2015), and recent experimental evidence suggests that perceptions of an atheism–science association are common: in a large, representative sample, Scheitle and Ecklund (2017) found that reading about an atheist scientist (Richard Dawkins, world-renowned biologist and “militant atheist”) did not affect beliefs about religion–science compatibility (relative to a control), but reading about a religious scientist (Francis Collins, National Institutes of Health (NIH) director and devout Christian) did affect such beliefs. Relatedly, distrust in science positively correlates with anti-atheist prejudice among U.S. Christians, r(291) = .48 (Simpson et al., 2016).

Nevertheless, little to no research has addressed the role of a perceived atheism–science link in attitudes toward science. Might such perceptions increase religious believers’ distrust in science? If so, then why? We focused particularly on two mechanisms: (1) the perception that science poses a threat to religious cultural authority and (2) moral distrust toward atheists.

Secularized science threatens religious cultural authority

According to intergroup threat theory (Rios, Sosa, & Osborn, 2018), distrust toward outgroups is exacerbated by perceptions of threat to the ingroup’s power, status, values, and identity. Importantly, recent research implicates intergroup threat in attitudes toward science. Evidence suggests that believers’ distrust in science emerges not simply from scientific ignorance or anti-science beliefs but rather from perceptions of biased scientific agendas that undermine religious values (Evans, 2011; Evans and Evans, 2008; Pew Research Center, 2015c). Moreover, such perceptions may be common: U.S. Christians seem well aware of the stereotype of Christians’ scientific distrust and incompetence, despite personally disavowing such stereotypes (Rios et al., 2015). Christians, therefore, may often perceive general anti-religion sentiments and threats among scientists.

Recent experiments illustrate how perceived threats to the ingroup can moderate trust in science. Colombo et al. (2016) found that participants supported scientific evidence less if it offended their moral values, even when participants were monetarily incentivized to evaluate the evidence accurately. In addition, Nauroth et al. (2015) found that participants who identified as video-gamers were especially likely to publicly discredit scientific research that threatened their social identity, but not when their group identity was experimentally bolstered or when the research was unrelated to their identity. In short, therefore, perceived intergroup threat (here: threat from science to religious cultural authority) may explain why perceived atheism–science associations undermine religious believers’ trust in science.

Anti-atheist distrust

Alternatively, perceived atheism–science associations might affect trust in science due to anti-atheist prejudice. Atheists are particularly loathed both in the United States and worldwide (International Humanist and Ethical Union (IHEU), 2012). Underlying anti-atheist prejudice is the belief that atheists are morally untrustworthy, as they lack fear of a divine observer and connection to God’s moral commands (Norenzayan et al., 2016). Ample research by Gervais and colleagues has found that, among several target groups, atheists are intuitively associated with immorality at above-average rates; a strong effect replicated in 10 of 12 different cultures, with various types of moral violation, and when contrasting atheists against various social groups (Gervais et al., 2017). Atheism, therefore, may be automatically associated with moral distrust, which may help explain the hypothesized link between believers’ distrusting science and perceiving atheism–science associations. For example, atheist scientists might be judged more likely than theist scientists to harm participants, distort data, or embezzle research funding.

Present research

We investigated whether a perceived atheism–science association would influence religious believers’ attitudes toward science. We then tested two competing accounts of statistical mediation: belief that secularized science threatens religious cultural authority, and anti-atheist distrust.

We recruited Christian participants because Christians constitute the majority of U.S. theists (~92%) and citizens in general (~71%) (Pew Research Center, 2015b). We focused on the United States because, compared to other secularized and scientifically advanced regions (e.g. China, Europe), it exhibits higher rates of theism and arguably more hostile atheism–theism and religion–science conflict (Pew Research Center, 2014).

Study 1 investigated correlations between distrust in science and perceived atheist/Christian prevalence among science professors. In study 2, we manipulated beliefs that either atheists or Christians are prevalent in science. Studies 3–4 featured an activity (presented as a questionnaire) designed to prime a perceived atheism–science association. In all studies, we tested for mediation via perceived threat from science to religion, and in studies 3–4, we also tested for mediation via moral distrust toward atheist scientists.

In all studies, we pre-screened for self-identified Christian theists and used the exact same criteria across studies regarding participant retention. Participants were retained for analyses if they (1) completed the survey’s key measures, (2) took the survey only once, (3) passed attention-checks, and (4) identified as Christian theists. We aimed for sample sizes that would allow at least 80% power to detect as significant (at p < .05) small-to-medium-sized effects (r = .22; $η_{p}^{2}$ = .048; d = .45).

We report all measures, data exclusions, and manipulations either in the article or in the supplemental material, and provide data files and preregistration information (https://osf.io/w52gx/). We also open our file drawer for meta-analysis (see supplemental materials).

2. Study 1

We first investigated correlations between (1) attitudes toward science/scientists and (2) estimated rates of atheists/Christians among university science faculty. We tested the hypothesis that perceived threat from science to religion mediates these correlations, and also explored additional candidate mediators.

Method

Participants

We recruited 211 participants via Amazon’s MTurk after first pre-screening for self-identified Christian theists. We first removed 12 who provided insufficient data, then four cases with duplicate IP addresses, 19 who failed two attention-check items, and five atheists. This left 171 Christian theists (69 females; M_age = 37.7, SD = 12.9; 71 Protestants, 44 Catholics, 27 Baptists, 29 “Christian-other”). For two-tailed correlation tests, this provided ~83% power to detect as significant (p < .05) small-to-medium-sized effects (r = .22).

Materials and procedure

Participants first completed two (order-randomized) sets of measures: attitudes toward science and prevalence estimates of atheists in science.¹

Attitudes toward science

(This set also included measures of potential mediators; see below.) Three scales measured institutional distrust toward science/scientists. One included four items from (McCright et al., 2013) each, asking, “How much do you trust or distrust scientists to. . .” (e.g. “. . . create knowledge that is unbiased and accurate”; 1 = strongly distrust, 6 = strongly trust; α = .84, M = 4.28, SD = 1.05). The second scale was “Nisbet et al.’s (2015) five-item scale” (e.g. “I have very little confidence in the scientific community”; 1 = strongly disagree, 7 = strongly agree; α = .91, M = 3.47, SD = 1.46). The third measure (derived from Gauchat, 2015; McCright et al., 2013) more directly addressed institutional distrust in science, using four items (e.g. “Science is too concerned with theory/speculation to be of much use in making concrete government policy decisions that will affect the way we live”; 1 = strongly disagree, 6 = strongly agree; α = .81, M = 3.31, SD = 1.08). After reverse-scoring the first measure, these three scales strongly correlated, rs = .66–.79 (α = .94), and were thus standardized and collapsed (hereafter, called Distrust Science).

Perceived atheist prevalence in science

Across two items, participants were asked to “estimate (0%–100%) the percentage of (natural) (social) science professors at American colleges/universities who are atheist.” We defined natural science (M = 53.4%, SD = 22.9) as including biology, chemistry, and physics, and social science (M = 50.5%, SD = 22.6) as including economics, political science, psychology, and sociology. These items strongly correlated, r(169) = .67, p < .0001, and were thus collapsed (hereafter, called Atheist Prevalence).

Later in the survey (immediately preceding demographic measures), we measured perceived Christian prevalence in science. Half the sample was reminded of their Atheist Prevalence responses, but this did not affect any prevalence estimate measure (all ts < .82). Participants on average estimated 40.9% (SD = 22.5) natural and 44.0% (SD = 22.5) social science faculty as Christian. These two items strongly correlated, r(169) = .73, p < .0001, and were thus collapsed (hereafter, Christian Prevalence).

Mediation

Perceived threat to religion from science (Threat) was measured with the item, “Scientists are typically anti-religion” (1 = strongly disagree, 6 = strongly agree; M = 3.85, SD = 1.31).² To explore additional possible mediators, we measured perceived religion–science Compatibility with three items (e.g. “To what extent are science and religion compatible?” 1 = very incompatible, 7 = very compatible; α = .86, M = 5.24, SD = 1.32). We also measured perceived Political Bias in science (0 = extreme liberal bias, 50 = no bias, 100 = extreme conservative bias; M = 44.3, SD = 25.0). Finally, we measured Biblical Literalism (0 = Bible is ancient book of fables. . ., 100 = Bible is actual word of God. . ., 50 = unsure). This allowed us to investigate a possible role of beliefs about religion’s epistemological superiority over science (e.g. when explaining the origins of the universe).

Potential covariates

We included several control variables, addressing traditional religious values and right-wing authoritarian values, political ideology, religious belief and behavior, and demographics (income: M = $40,000, SD = $30,000, mode = $40,000–$50,000; education: M = 2-year college/university degree, mode = 4-year degree). To avoid excess clutter, we describe these in the supplemental materials.

Results

We first compared participants who completed the estimates measures before versus after the trust measures. There was no significant order effect on any dependent measures or potential mediators (ts < 1.62, ps > .108), including, most importantly, Distrust Science and Threat (ts < 0.44, ps > .664). In additional preliminary analyses, participants estimated atheists (M = 51.7%, SD = 21.1) as more common in American science than Christians (M = 42.5%, SD = 20.9), t(170) = 3.36, p = .001, $η_{p}^{2}$ = .062. These two estimates measures were negatively associated, r(169) = –.47, p < .0001.

Zero-order correlations between prevalence estimates, Distrust Science, and the candidate mediators are presented in Table 1. As expected, Atheist (Christian) Prevalence correlated positively (negatively) with Distrust Science. Both prevalence measures correlated with Threat (but Atheist Prevalence yielded a larger correlation than Christian prevalence: z = 3.36, p < .001).

Table 1.

Zero-order correlations (study 1) (N = 171).

	2.	3.	4.	5.	6.	7.
1. Perceived atheist prevalence	−.47***	.20**	.55***	−.08	−.14^	.10
2. Perceived Christian prevalence		−.22**	−.32***	.25***	.06	−.06
3. Distrust in science			.42***	−.15^	−.26***	.36***
4. Perceived threat from science to religion				−.14^	−.23**	.23**
5. Perceived ideological bias in science (0 = Left-wing bias, 100 = Right-wing bias)					−.12	−.13^
6. Perceived religion–science compatibility						.07
7. Biblical literalism

^p < .087; *p < .05; **p < .01; ***p ⩽ .001.

Both Atheist and Christian Prevalence variables remained significant predictors of Distrust Science when controlling for all demographic and religiosity measures, Biblical Literalism, and/or Compatibility/Ideological Bias (ps < .042). We also ran regression models (one per predictor)³ controlling for presentation order (i.e. Atheist Prevalence measures presented before vs. after the distrust measures) and its interaction with the pertinent estimates measure. In one model, Atheist Prevalence remained significant and actually increased in size (β = .29, p = .018), but neither Order (β = –.002, p = .982) nor the interaction (β = –.11, p = .344) was significant. In the other model, Christian Prevalence (β = –.12, p = .297), Order (β = .011, p = .881), and the interaction (β = –.12, p = .291) were not significant.

Mediation and moderation

We then tested whether the abovementioned associations were mediated by Threat. Following guidelines by Hayes (2013), we ran two analyses (one per predictor—Atheist or Christian Prevalence) using the PROCESS macro (model 4) with 5000 bootstrapped samples after first standardizing all variables. These analyses revealed an overall indirect effect of Atheist Prevalence (boot = .232, SE = .052, 95% confidence interval (CI) = .137–.343) and Christian Prevalence (boot = –.121, SE = .048, 95% CI = –.228 to –.045) on distrust science via Threat (direct effects were reduced to non-significance, βs < .09, ps > .227).

We repeated these analyses but included all candidate mediators (Threat, Compatibility, Ideological Bias, and Biblical Literalism). Threat remained a significant mediator of the Atheist Prevalence effect (boot = .172, SE = .048, 95% CI = .082–.278) and the Christian Prevalence effect (boot = –.085, SE = .041, 95% CI = –.190 to –.023). Otherwise, all candidate mediators were non-significant in both analyses (boots < .032).

We also tested for statistical moderation to investigate the possibility that fundamentalist sub-cultures (represented here by participants high in Biblical Literalism) may be particularly likely to distrust atheist-dominated science. That is, literalists may show stronger correlations between Distrust and perceived atheist prevalence in science, because excess atheist influence may be perceived to render science factually untrustworthy. We used Hayes (2013) PROCESS macro (model 1) and 5000 bootstrapped samples. The Literalism–Atheist Prevalence interaction did not significantly predict Distrust Science (β = .10, p = .132). However, whereas both middling and high scores on Literalism revealed significant Atheist Prevalence–Distrust Science associations (boots = .144 and .240, ps = .028 and .007, respectively), low scores did not (boot = .048, p = .613). Hence, lacking fundamentalist religious beliefs may help prevent perceived atheist dominance in science from increasing Christians’ distrust in science, but stronger evidence is needed to warrant further interpretation, as the interaction effect was non-significant. This is consistent with prior research suggesting that relatively fundamentalist Christians are only subtly different from non-fundamentalist Christians in their relationship with science. Zigerell (2012) found that Christians who read the Bible as literal, factual truth, compared to those who read the Bible through either an interpretive or “uninspired” lens, averaged only one less correct answer out of 16 on a test of science knowledge—a significant difference, but very small.

Discussion

Study 1 provides preliminary evidence that perceived prevalence of atheists (Christians) in science is uniquely associated with more (less) distrust in science among Christian participants, and that this is mediated by a stronger (weaker) perception that scientists are “anti-religion.” We also found evidence to rule out the notion that associations between distrust toward science and perceived Atheist/Christian prevalence in science are explained by perceived political bias among scientists, perceived religion–science (in)compatibility, or fundamentalist religious beliefs.

3. Study 2

In study 2, we manipulated the perceived prevalence of both atheists and Christians in science. We presented participants with results from a bogus, nationwide survey into scientists’ religious beliefs, and then measured attitudes toward science as in study 1.

Method

Participants

We recruited 341 participants via MTurk after first pre-screening for Christian theists. We first removed 21 participants who provided insufficient data and six cases with duplicate IP addresses. Then, to ensure the final sample included only attentive participants who believed the manipulation, we removed 12 who did not believe the findings presented in the manipulation, suspected the findings were fake, or whose text response was non-sensical (e.g. “rewrrewrewrew”). We also removed five who failed two attention-check items, and two self-identifying as non-Christian or non-theist. This left 297 participants (114 males, 178 females, five unreported; 102 Protestants, 83 Catholics, 44 Baptists, 63 “Christian-other,” five unreported; M_age = 38.6, SD = 13.8). This provided ~94% power to detect as significant (p < .05) small-to-medium-sized effects ( $η_{p}^{2}$ = .048) in a one-way analysis of variance (ANOVA) with three conditions.

Materials and procedure

Participants were randomly assigned to one of three conditions: Atheist Prevalence (n = 93), Christian Prevalence (n = 101), or control (n = 108). No information was presented in the control; participants began by completing the focal measures. In the experimental conditions, participants first read a summary (~240 words) of a bogus 2014 Pew Research Report titled “Science in the U.S.” This described Pew’s efforts to comprehensively survey U.S. scientists across several scientific fields and institutions (e.g. science education; private industry), concluding that “Pew found that ~65% of scientists [claim to not believe in God/a higher power] [identify as Christian/believe in God]” and presented five pie charts depicting the rates of (atheists) (Christians) in science (Figure 1). Then, to conceal the manipulation and provide means to detect inattentive or suspicious participants, we asked participants to type their thoughts about the report. To ensure that participants processed the prime, the page (which included all the above details) was programmed so that participants could proceed only after 120 seconds.

Figure 1.

Pie charts used in the manipulation of study 2.

We then presented participants with the following measures from study 1. First, we measured perceived Threat from science to religion. Then, we included the three measures of distrust toward science/scientists (the dependent variables (DVs); all αs > .80); after reverse-coding the trust-framed scale, these measures were collapsed (Distrust Science; α = .93), as in study 1. Finally, we included most of the demographic and religiosity measures from study 1.

Results and discussion

A one-way ANOVA revealed that Condition significantly predicted Distrust Science, F(2, 294) = 4.31, p = .014, $η_{p}^{2}$ = .029. The contrast comparing the atheist condition (2) against the Christian and control conditions (–1 –1) was significant—F(1, 294) = 8.52, p = .004, $η_{p}^{2}$ = .028—but the contrast comparing the latter two conditions was not, F < 1 (Figure 2).

Figure 2.

Effects of condition on attitudes toward science (bars represent SEs). Scores converted to 0–5 scale for display (study 2).

Condition also significantly predicted Threat—F(2, 294) = 3.13, p = .045, $η_{p}^{2}$ = .021. The contrast comparing the atheist condition against the Christian and control conditions was significant—F(1, 294) = 5.08, p = .025, $η_{p}^{2}$ = .025—but the contrast comparing the latter two conditions was not, F(1, 294) = 1.25, p = .264, $η_{p}^{2}$ = .004.

We then tested whether Threat mediated Condition’s effect on distrust science. We first converted condition into two contrast-coded variables, one comparing Atheist Prevalence (2) against Christian Prevalence and Control conditions (–1 –1), and the other comparing the latter two (0 1 –1). A mediation analysis using PROCESS model 4 (Hayes, 2013) revealed a significant overall indirect effect of Condition (2 –1 –1) on Distrust Science via Threat (boot = .028, SE = .014, 95% CI = .005–.060).

Hence, as in study 1, perceived atheist prevalence in science predicted distrust in science largely because of the perception that scientists are anti-religion. Unlike study 1, perceived Christian prevalence in science did not predict distrust in science, suggesting that effects of perceived atheist prevalence are more reliable.⁴

4. Studies 3–4: Overview

Next, we sought further evidence of causality with a novel experimental design. Participants estimated the rates of four sub-types of non-believer (private atheists; evangelical atheists; anti-theists; agnostics) in both natural and social science, either at the beginning or end of the survey. We reasoned that this faux “measurement” task would prime participants to perceive an atheism–science association because it forces participants to think about science and atheism simultaneously and implies that atheist prevalence in science is an important research topic. This design is advantageous over study 2’s design as the manipulation is not explicit. Also, because it was placed seemingly randomly among several measures in a study with the broad title “attitudes toward social groups and social institutions,” it more effectively precludes participant awareness of our focus on atheists and science.

Although this design slightly resembles study 1’s design, study 1 revealed no effect of presentation order (i.e. estimates measures presented before vs. after trust measures). However, study 1 included only two measures (without preamble) of estimated rates of atheists in science, after which participants could immediately proceed. In contrast, studies 3–4 included a lengthy preamble, followed by eight “measures” of estimated rates of atheists in science, and precluded participants from proceeding to the next screen until 90 seconds had elapsed. Hence, despite similar designs, study 1 was purely correlational, whereas studies 3–4 were experimental.

5. Study 3

To now address the moral distrust hypothesis, we included a measure of trust in scientists’ morality. In addition, to aid generalizability, we recruited participants from an alternative participant pool (not MTurk). We also sought to replicate study 1’s correlations between perceived Christian prevalence in science and distrust toward science.

Method

Participants and design

We recruited participants using the Qualtrics panel service. This draws from multiple participant pools (e.g. market research panels; specialized recruitment campaigns), can target specific demographic characteristics, and aims for samples that are as close to nationally representative as possible. Qualtrics also uses IP address checks and digital fingerprinting to ensure data integrity.

For study 3, we requested that Qualtrics collect data from Christian theists, with equal representation across all age groups and males/females. Four attention-checks were placed at various points of the survey, and Qualtrics immediately removed participants who failed any one attention-check. We recruited 148 participants, 24 of whom were rejected for incomplete responses and/or a failed attention-check. There were no duplicate IP addresses. This left 124 participants (60 males, 64 females; 46 Protestants, 40 Catholics, 24 Baptists, 14 “Christian-other”; M_age = 46.5, SD = 17.0). This provided ~80% (~70%) power to detect small-to-medium-sized effects (d = .45, p < .05) in a one-tailed (two-tailed) t test. Participants were randomly assigned to provided estimates of atheist groups in science either before (n = 65) or after (n = 59) the dependent measures.

Materials and procedure

Manipulation

Participants first read, “we are interested in how common you believe various social groups are among the elite science institutions of America.” This helped conceal our focus on non-believer targets, who were presented as just one of many social groups of interest. We then presented definitions of the following: private atheist; evangelical atheist; anti-theist; agnostic; natural science; social science. Participants then estimated (0%–100%) the rate of each of the four groups among elite natural scientists and social scientists (order-randomized; eight estimate “measures,” total) (see Figure 3). To ensure that participants processed the prime, we programmed this activity so that participants could proceed only after 90 seconds. Experimental (control) condition participants completed this at the beginning (end) of the survey.

Figure 3.

The “measurement” activity used for manipulation in studies 3–4. This display was presented twice: one inquired about natural scientists, the other about social scientists. All participants spent at least 90 seconds on this page.

DVs/mediators

We presented five measures in randomized order. This included the three measures of distrust in science from studies 1–2; as in all prior studies, these were highly correlated (α = .83, after reverse-coding the trust-framed scale), so we standardized and collapsed them (hereafter, Distrust Science). The five randomized measures also included the single-item Threat measure from studies 1–2 (“Scientists are typically anti-religion”), and a new, four-item measure of moral distrust toward scientists (hereafter, Scientists Neglect Morality) derived from Rutjens and Heine (2016; for example, “Scientists mostly value acquiring knowledge over preventing harm”; 1 = strongly disagree, 7 = strongly agree; α = .62).

Participants then estimated the prevalence of Christians in both natural (M = 48.6%, SD = 22.7%) and social (M = 49.2%, SD = 22.9%) science (as in study 1), and we collapsed these measures (r(112) = .73, p < .0001). (These were strategically placed in the middle of the survey to avoid affecting or being affected by the distrust or atheist prevalence measures.) Finally, we measured demographics as per study 2.

Results

There were no effects of condition on any prevalence estimate “measure” used in the manipulation (ts < 1.59, ps > .115), except estimated prevalence of agnostics in social science, t(122) = 1.90, p = .060. Controlling for any of these “measures” in the focal analyses did not reduce the effect of condition (see supplemental materials).⁵

As hypothesized, experimental-condition participants expressed greater distrust in science (M_standardized = 0.16, SD = 0.72) than control participants (M_standardized = –0.18, SD = 0.97), t(122) = 2.24, p = .027, d = .40. The effect of Condition was stronger when controlling for the eight estimates “measures” in an analysis of covariance (ANCOVA), F(1,114) = 6.95, p = .010, $η_{p}^{2}$ = .057. Experimental-condition participants (M = 3.98, SD = 1.36) also perceived greater threat than control participants (M = 3.32, SD = 1.57), t(122) = 2.52, p = .013, d = .45. Finally, experimental-condition participants (M = 4.90, SD = 1.02) judged scientists as more morally neglectful than control participants (M = 4.50, SD = 1.09), t(122) = 2.07, p = .041, d = .37. Hence, we tested both Threat and Scientists Neglect Morality for mediation of the Condition–Distrust effect.

We used PROCESS model 4 (Hayes, 2013), with one model per mediator. Both models revealed a significant overall indirect effect of Condition (0 = control, 1 = atheism–science prime) on Distrust Science (via Threat: boot = .209, SE = .090, 95% CI = .057–.428; via Scientists Neglect Morality: boot = .108, SE = .063, 95% CI = .010–.259). When entered simultaneously as mediators, only Threat (boot = .189, SE = .088, 95% CI = .044–.412), but not Scientists Neglect Morality (boot = .035, SE = .034, 95% CI = –.010 to .134), significantly mediated the effect.

Correlations with Christian prevalence estimates

Perceived Christian prevalence yielded the following zero-order associations: Distrust Science, r(112) = –.38, p < .001; Threat, r(112) = –.35, p < .001; Scientists Neglect Morality, r(112) = .04, p = .657. The first two correlations are consistent with those found in study 1 (Scientists Neglect Morality was absent in study 1), and remained significant βs = –.35 to –.40, ps < .001) when controlling for demographics, religiosity, and political ideology. As in study 1, the association with distrust science was mediated by Threat (boot = –.167, SE = .054 95% CI = –.285 to –.073); it was not mediated by Scientists Neglect Morality (boot = .016, SE = .038, 95% CI = –.058 to .098).

Discussion

Study 3 provided further evidence that a perceived atheism–science association causes increased distrust in science, and that perceived threat from science to religion mediates this effect (meanwhile, the hypothesis that moral distrust toward atheists/scientists mediates this effect received much weaker support). Hence, this study further suggests that a perceived atheism–science association undermines religious believers’ trust in science because it threatens religious cultural authority.

Finally, we replicated some findings of study 1: perceived Christian prevalence in science negatively predicted distrust in science, and perceived (lack of) threat from science to religion, but not moral distrust toward scientists, mediated this effect.

6. Study 4

Study 4 was identical to study 3, but with some important changes. First, the research design, target sample size, all measures, and all hypotheses were preregistered before data collection (https://osf.io/k97ea). Second, we included an alternative measure of moral distrust toward scientists to further investigate this as a potential mediator. Third, we included an expanded, eight-item measure of perceived threat from science to religion, derived from previous research on intergroup threat, and we included a measure of biblical literalism to re-explore questions touched on in study 1.

Method

Participants and design

We recruited 231 participants via MTurk after first pre-screening for Christian theists. We first removed 38 who provided insufficient data. There were no duplicate IP addresses. We then removed nine participants who identified as non-Christian or non-theist and 15 who failed an attention-check. This left 169 participants (68 males, 101 females; 68 Protestants, 51 Catholics, 20 Baptists, 30 “Christian-Other”; M_age = 37.8, SD = 13.0). This provided ~90% (~83%) power to detect small-to-medium-sized effects (d = .45) as significant (at p < .05) in a one-tailed (two-tailed) t test. Participants were randomly assigned to provided estimates of atheist groups in science either before (n = 85) or after (n = 84) the dependent measures.

Materials and procedure

After the prime (or—in the control condition—at the beginning of the survey), we presented four order-randomized scales. Distrust science was again measured with the two distrust-framed scales from study 1, which we standardized and collapsed (α = .93) (the four-item, trust-framed scale was excluded for sake of brevity). To assess moral distrust toward scientists, we used Rutjens and Heine’s (2016) three-item (α = .78) measure of beliefs about scientists’ inclinations to trade off morality for scientific discovery. Participants responded to the preamble, “what do you think scientists typically value more?” on three items (e.g. “scientific knowledge vs doing the right thing”) using 0-100 scales (50 = scientists value both options equally). As high scores reflect beliefs that Scientists Prioritize Morality, we refer to this measure as such. To measure Threat, we added to the previous item (“Scientists are typically anti-religion”) seven items derived from Stephan and Stephan’s (2001) measure of racial intergroup threat, which were amenable to rewording (e.g. “Scientists want policies to disregard religious principles in favor of scientific principles”; “The basic values that science and religion promote often come into conflict”; 1 = strongly disagree, 7 = strongly agree; α = .82). We again included the measure of Biblical Literalism from study 1 to further rule out the possibility that the perceived epistemological superiority of religion, not perceived threat from science to religion, explains the effects of interest, and to explore whether relatively fundamentalist Christians are more likely to distrust atheist-dominated science. Finally, we measured demographics as per study 3.

Results

There were no differences across conditions on any prevalence estimate “measure” (ts < 1.13, ps > .260), except for estimated prevalence of agnostics in social science, t(167) = 1.99, p = .049. Controlling for any of these “measures” in the focal analyses did not influence results (also, see Footnote 5).

Distrust Science was higher in the Atheist Prevalence condition (M_standardized = 0.15, SD = 0.93) than in the control (M_standardized = –0.15, SD = 0.91), t(167) = 2.13, p = .034, d = .33, and this effect was similar when controlling for the eight estimate “measures” in an ANCOVA, F(1,159) = 4.17, p = .043, $η_{p}^{2}$ = .026. In contrast, there was no effect of Condition on Scientists Prioritize Morality (M_experimental = 33.0, SD = 21.8; M_control = 34.9, SD = 19.5, t < 1), thus ruling out moral distrust toward scientists as mediator of the above effect.

Mediation analyses

Regarding the new, eight-item measure of threat, the experimental condition (M = 4.66, SD = 1.05) yielded higher perceived threat than the control (M = 4.32, SD = 1.08), t(167) = 2.08, p = .039, d = .32. Using the same mediation process as in the previous studies, we found an overall indirect effect of Condition on Distrust Science via Threat (boot = .180, SE = .088, 95% CI = .022–.364), but not via Scientists Prioritize Morality (boot = .004, SE = .014, 95% CI = –.014 to .057). (The exact same pattern of results emerged, with similar coefficients, when both mediators were entered into the model together.)

Finally, we explored the possible role of fundamentalist beliefs, as in study 1. After standardizing Biblical Literalism scores, we first tested its mediation role. It was a non-significant mediator when entered alone (boot = .097, SE = .060, 95% CI = –.019 to .222) or alongside both Threat and Scientists Prioritize Morality (boot = .034, SE = .025, 95% CI = –.003 to .106; mediation significant via Threat but not Scientists Prioritize Morality). When entered alongside only threat, its mediation effect was significant but small (boot = .038, SE = .027, 95% CI = .002–.122), overshadowed by mediation via Threat (boot = .158, SE = .082, 95% CI = .016–.337). We then tested for moderation. The Condition × Literalism interaction did not significantly predict distrust science (β = .118, p = .369); hence, participants of varying degrees of fundamentalism were generally similar in how a perceived atheism–science association affected their trust in science.

Discussion

Study 4 replicated study 3, as a manipulated atheism–science association significantly increased distrust in science. Unlike study 3, however, the manipulation did not affect judgments of scientists’ morality, suggesting that effects on perceptions of scientists’ morality (e.g. beliefs about scientists valuing knowledge over ethics) are less reliable than effects on institutional distrust in science (e.g. beliefs about science being too concerned with theory and speculation and not helpful for societal progress). As in study 3, the effect on distrust in science was mediated not by moral distrust toward scientists but by perceived threat from science to religion: when science seems somewhat interwoven with atheism, this leads to an average increase in beliefs that science poses a threat to religious institution and cultural influence, and this brings with it an increased distrust in science as an institution of societal value.

In addition, as in study 1, there were suggestive yet small and non-significant differences in the effects of the manipulation when comparing different levels of biblical literalism. Therefore, and again echoing Zigerell (2012), fundamentalists appear to show at most only small divergence from non-fundamentalists in their relationship with science.

This study goes beyond study 3 in some important ways. First, it was preregistered: all measures, manipulations, hypotheses, and decisions regarding sampling and statistical power were registered before data analysis. Second, it included an expanded, eight-item measure of perceived threat from science to religion, suggesting that the mediational effects of perceived threat are not limited to the specific belief that “scientists are typically anti-religion” (the measure used in studies 1–3).

7. General discussion

In one correlational and three experimental studies, a perceived association between atheism and science predicted numerous measures of institutional distrust in science. We sampled from two distinct sources (MTurk and Qualtrics) and preregistered study 4 before collecting data. In each study, we measured or manipulated perceived atheist prevalence in a variety of elite scientific roles (study 1: university professors; studies 2–4: various scientific fields/institutions). Hence, our results speak to distrust in science in general, not simply distrust toward scientists in academia, industry, or other domains.

Although atheists are widely distrusted morally (Gervais et al., 2017), studies 3–4 found that moral distrust toward scientists did not help explain why participants showed greater distrust in science after the atheism–science prime. Moreover, the manipulated atheism–science association increased institutional distrust in science but was inconsistent in its effect on moral distrust toward scientists. Study 1 also found little to no evidence that believers’ distrust in secularized science was explained by perceived religion–science incompatibility, perceived political bias of scientists, or fundamentalist/literalist religious beliefs. In contrast, perceived threat from science to religion consistently mediated the distrust–atheist prevalence associations. There was also mediation in studies 1 and 3 such that perceived Christian prevalence in science predicted greater institutional trust in science because of lower perceived threat from science to religion. Together, these results suggest that, among U.S. Christians, atheism-dominated science is not necessarily morally dubious, but it potentially threatens the capacity for religion to serve as cultural authority. This is largely consistent with prior evidence that, beyond Christians specifically, religious believers’ distrust in science emerges primarily from fear that biased scientific agendas will violate or at least challenge religious values and principles (Evans, 2011; Evans and Evans, 2008).

In other words, an atheist majority in science may be interpreted as a cultural status threat to religion, particularly regarding the role of religion in shaping how science influences society and public policy. This is evident in some open-ended responses from study 2: “. . . scientists need to caution when publishing their findings that the viewpoint is that of an atheistic worldview”; “Science has attacked the Bible for centuries . . .”; “. . . The whole point of the sciences to begin with was to ‘disprove’ God. . .”; “. . . Atheists are determining what science looks like, and whether they will consider God in the equation.”

A perceived threat of secularized science, bereft of theists, is particularly evident in the debate over abortion. A recent article in The Atlantic (Green, 2018) explored the subject:

[The] desire to harness “objective science” is at the heart of the pro-science bent in the pro-life movement: Science is a source of authority that’s often treated as unimpeachable fact . . . “Both sides tend to use scientific information when it is useful towards making a point that is based on . . . firmly and sincerely held philosophical and religious convictions.” (Quoting Daniel Sulmasy, biomedical ethics professor at Georgetown University and former Franciscan friar.)

In other words, if pro-life religious believers fear they have limited control over scientific research programs, they may in turn fear a limited ability to defend and advance their opposition to abortion. More generally, amid perceptions of a secularized science, with many atheists and few religious believers, theists may fear they are losing a battle against secularists to use science in support of their moral convictions. The present work provides empirical support for this line of argument.

Limitations and future directions

Although we used two rather different techniques to manipulate a perceived atheism–science association, future research could explore additional ways to manipulate it. Attitudes toward specific fields of science may also warrant further attention. For example, McCright et al. (2013) distinguished “impact science” (e.g. science that investigates the impact of human activity on ecosystems, such as environmental science) from “production science” (e.g. science that helps produce advanced technology). The authors found that political conservatives, compared to liberals, trusted impact scientists less but trusted production scientists more, suggesting that this distinction may be important in understanding heated disagreements about scientific research and authority. Future research could also address perceptions of secularized science amid “culture-war” topics, including abortion, euthanasia, and cloning. Finally, we should look beyond U.S. Christians to address other populations. We focused on the United States given its relatively high rate of theism amid an otherwise secularized, scientifically advanced culture, but the effects reported here could look different in other cultures. Indeed, in relatively secular areas such as Scandinavia, or among U.S. atheists, perceived atheist prevalence may predict trust in science.

Implications

Many reacted with concern to President Obama’s nomination of Francis Collins (a devout evangelical) as NIH director (Pew Research Center, 2015c). But given the present findings, such appointments might help buffer the threat of perceived secularized science. Increasingly public criticism of religion by many famous science popularizers (e.g. Dawkins, 2006), and arguments that science is an atheistic enterprise (e.g. Krauss, 2015), must be squared with various empirical observations. First, religious beliefs and institutions may be inalienable not only to cultures but also to human minds. Religions are sophisticated systems of belief that are difficult to challenge logically (Atran and Norenzayan, 2004) and likely formed the historical bedrock of societal expansion from tribes into city-states (Norenzayan, 2013). Indeed, religion likely played a crucial role in facilitating cooperation among non-kin and expanding networks of trust and trade (Norenzayan et al., 2016). Unsurprisingly, therefore, the majority of both the United States and the world currently identifies as religious and believes that religion is necessary to be moral (Pew Research Center, 2014). Second, historical (Harrison, 2015) and contemporary (Ecklund, 2012) analyses of scientists and their religious beliefs suggest that most scientists do not perceive conflict between science and religion and are not as irreligious or hostile to religion as is often assumed. In fact, throughout modern history, science and religion may have co-existed peacefully in a symbiotic relationship (Harrison, 2015). Finally, evidence suggests that religious believers are prone to stereotype threats regarding scientific interest and incompetence (Rios et al., 2015). Hence, strong associations between science and atheism risk bolstering an artificial barrier that at times can separate the scientific enterprise from the interests and concerns of literally billions of believers worldwide. Crucially, therefore, garnering institutional trust in science may not be well served by polarizing discourse that depicts science as incompatible with religious faith.

Some evidence indicates that trust in science has been steadily declining in recent decades (Gauchat, 2011). Moreover, such phenomena may be catalyzed by increasing concerns over “questionable research practices,” data fraud, and the replicability of scientific findings more generally (e.g. Open Science Collaboration, 2015). With so many threats to the cultural status of science, the present findings suggest that, perhaps ironically, the status of science may be further undermined by threats to the cultural status of religion.

In this context, the present results might suggest that in order to avoid increases in general distrust in science, scientists should exercise added caution if they wish to challenge religion on scientific grounds and should avoid rhetoric that alienates religious believers from scientific communities (e.g. Krauss, 2015: “All scientists should be militant atheists”). In addition, in acknowledging just how deep-seated religious belief and customs are in both human culture and psychology (Norenzayan et al., 2016), active steps could be taken to prevent potential damage that anti-religion rhetoric may exert on societal trust in science. This might involve, for example, better attempts to appeal to religious believers who are considering tertiary scientific education.

These suggestions may attract opposition among some fellow scientists. This is understandable, especially amid the ongoing U.S. culture wars, widespread scientific illiteracy, and anti-science sentiment that can seem regressive and dangerous (e.g. anti-vaccination movements dispersing theories that directly conflict with scientific proof; Bonhoeffer and Heininger, 2007). But this is precisely why we argue, based on the present findings, that scientists should take caution when attacking religious institutions and beliefs. Distrust in science can pose a genuine threat to society; it is thus imperative that we consider all of the various factors that contribute to such distrust in our efforts to further promote the positive role of science in the modern world.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Ain Simpson

Kimberly Rios

Supplemental material

Supplemental material for this article is available online.

Notes

Author biographies

Ain Simpson is a social psychologist who earned his PhD in psychology at the University of Melbourne, Australia, in 2014. Currently at Miami University, his research focuses on moral judgment, the social psychology of religion, and how these topics overlap.

Kimberly Rios is an associate professor of psychology at Ohio University. Her research focuses on intergroup relations and social identity, including religious identity.

References

Atran

Norenzayan

(2004) Religion’s evolutionary landscape: Counterintuition, commitment, compassion, communion. Behavioral and Brain Sciences 27(6): 713–730.

Bonhoeffer

Heininger

(2007) Adverse events following immunization: Perception and evidence. Current Opinion in Infectious Diseases 20(3): 237–246.

Colombo

Bucher

Inbar

(2016) Explanatory judgment, moral offense and value- free science. Review of Philosophy and Psychology 7(4): 743–763.

Dawkins

(2006) The God Delusion. Boston, MA: Houghton Mifflin.

Ecklund

(2012) Science vs. Religion: What Scientists Really Think. Oxford: Oxford University Press.

Evans

(2011) Epistemological and moral conflict between religion and science. Journal for the Scientific Study of Religion 50(4): 707–727.

Evans

(2008) Religion and science: Beyond the epistemological conflict narrative. Annual Review of Sociology 34: 87–105.

Gauchat

(2011) The cultural authority of science: Public trust and acceptance of organized science. Public Understanding of Science 20(6): 751–770.

Gauchat

(2012) Politicization of science in the public sphere: A study of public trust in the United States, 1974 to 2010. American Sociological Review 77(2): 167–187.

10.

Gauchat

(2015) The political context of science in the United States: Public acceptance of evidence-based policy and science funding. Social Forces 94(2): 723–746.

11.

Gervais

Najle

(2018) How many atheists are there? Social Psychological and Personality Science 9(1): 3–10.

12.

Gervais

Xygalatas

McKay

Van Elk

Buchtel

Aveyard

, et al. (2017) Global evidence of extreme intuitive moral prejudice against atheists. Nature Human Behaviour 1(8): 0151.

13.

Green

(2018) Science is giving the pro-life movement a boost. The Atlantic, 18 January. Available at: https://www.theatlantic.com/politics/archive/2018/01/pro-life-pro-science/549308/

14.

Harrison

(2015) The Territories of Science and Religion. Chicago, IL: University of Chicago Press.

15.

Hayes

(2013) The PROCESS macro for SPSS and SAS (version 2.13). Software. Available at: www.processmacro.org

16.

International Humanist and Ethical Union (IHEU) (2012) Freedom of thought report. Available at: www.humanists.international/

17.

Ipsos Global @dvisory (2011) Supreme being(s), the Afterlife and evolution. Available at: https://www.ipsos.com/en-us/news-polls/ipsos-global-dvisory-supreme-beings-afterlife-and-evolution

18.

Krauss

(2015) All scientists should be militant atheists. The New Yorker, 9 August. Available at: http://www.newyorker.com/news/news-desk/all-scientists-should-be-militant-atheists

19.

McCright

Dentzman

Charters

Dietz

(2013) The influence of political ideology on trust in science. Environmental Research Letters 8(4): 044029.

20.

Nauroth

Gollwitzer

Bender

Rothmund

(2015) Social identity threat motivates science-discrediting online comments. PLoS ONE 10(2): e0117476.

21.

Neuman

(2014) 1 in 4 Americans thinks the sun goes around the earth, survey says. Available at: http://www.npr.org/blogs/thetwoway/2014/02/14/277058739

22.

Nisbet

Cooper

Garrett

(2015) The partisan brain: How dissonant science messages lead conservatives and liberals to (dis)trust science. ANNALS of the American Academy of Political and Social Science 658(1): 36–66.

23.

Norenzayan

(2013) Big Gods: How Religion Transformed Cooperation and Conflict. Princeton, NJ: Princeton University Press.

24.

Norenzayan

Shariff

Gervais

Willard

McNamara

Slingerland

et al. (2016) The cultural evolution of prosocial religions. Behavioral and Brain Sciences 39: e1.

25.

Open Science Collaboration (2015) Estimating the reproducibility of psychological science. Science 349(6251): aac4716.

26.

Pew Research Center (2014) Worldwide, many see belief in god as essential to morality, March. Available at: http://www.pewglobal.org/2014/03/13/worldwide-many-see-belief-in-god-as-essential-to-morality/

27.

Pew Research Center (2015a) Americans are in the middle of the pack when it comes to importance of religion. Available at: http://www.pewresearch.org/fact-tank/2015/12/23/americans-are-in-the-middle-of-the-pack-globally-when-it-comes-to-importance-of-religion/

28.

Pew Research Center (2015b) America’s changing religious landscape. Available at: http://www.pewforum.org/2015/05/12/americas-changing-religious-landscape/

29.

Pew Research Center (2015c) Religion and science. Available at: http://www.pewinternet.org/2015/10/22/science-and-religion/

30.

Preston

Epley

(2009) Science and god: An automatic opposition between ultimate explanations. Journal of Experimental Social Psychology 45(1): 238–241.

31.

Rios

Cheng

Totton

Shariff

(2015) Negative stereotypes cause Christians to underperform in and disidentify with science. Social Psychological and Personality Science 6(8): 959–967.

32.

Rios

Sosa

Osborn

(2018) An experimental approach to Intergroup Threat Theory: Manipulations, moderators, and consequences of realistic vs. symbolic threat. European Review of Social Psychology 29(1): 212–255.

33.

Rutjens

Heine

(2016) The immoral landscape? Scientists are associated with violations of morality. PLoS ONE 11(4): e0152798.

34.

Scheitle

Ecklund

(2017) The influence of science popularizers on the public’s view of religion and science: An experimental assessment. Public Understanding of Science 26(1): 25–39.

35.

Simpson

Piazza

Rios

(2016) Belief in divine moral authority: Validation of a shortened scale with implications for social attitudes and moral cognition. Personality and Individual Differences 94: 256–265.

36.

Stephan

(2001) Improving Intergroup Relations. Thousand Oaks, CA: SAGE.

37.

Zigerell

(2012) Science knowledge and biblical literalism. Public Understanding of Science 21(3): 314–322.