Examining Christians’ Reactions to Reminders of Religion–Science Conflict: Stereotype Threat versus Disengagement

Abstract

Stereotypes of religion (particularly Christianity) as incompatible with science are widespread, and prior findings show that Christians perform worse than non-Christians on scientific reasoning tasks following reminders of such stereotypes. The present studies (N = 1,456) examine whether these reminders elicit stereotype threat (i.e., fear of confirming negative societal stereotypes about one’s group), disengagement (i.e., distancing oneself from a domain perceived as incongruent with the values of one’s group), or both. In Studies 1 and 2, Christians demonstrated lower task performance and greater subjective feelings of stereotype threat (but did not spend less time on the task) relative to non-Christians when beliefs about Christianity–science incompatibility were chronic or made salient. Furthermore, the effects of incompatibility stereotypes on performance were most pronounced among Christians who identified strongly with science and hence worried most about confirming negative stereotypes (Studies 3–4). Implications for Christians’ responses to religion–science conflict narratives and participation in science are discussed.

Keywords

stereotype threat social identity threat religion and science religious stereotypes religion–science conflict

Are religion and science compatible? Is it possible for a religious person to be a good scientist? These questions have arisen among the American populace in the wake of debates about issues such as embryonic stem-cell research and teaching evolution in public schools. Indeed, 59% of American adults report thinking that religion and science are “often in conflict,” compared with the 38% who report thinking that religion and science are “mostly compatible” (Pew Research Center, 2015b). Given the underrepresentation of religious individuals in science relative to their numbers in the overall population, the prevalence of beliefs about religion–science incompatibility is hardly surprising. For instance, although 63% of Americans in general claim they are absolutely certain God exists (Pew Research Center, 2015a), this proportion drops to 10% among American biologists and physicists at universities and research institutes (Ecklund et al., 2016).

The objective of the present research is to investigate Christians’ behavioral (e.g., scientific task performance and persistence) and subjective/affective (e.g., feelings of stereotype threat; Pennington et al., 2016) reactions following reminders of the religion–science conflict narrative. Given that Christians are the religious majority in the United States (75% according to Gallup, 2017), it is critical to better understand the psychological factors that may deter a large proportion of the population from pursuing and succeeding in science. Christians are the focus of these studies because, as prior work has shown, negative stereotypes about religious individuals’ scientific abilities are targeted at Christians and do not extend to groups such as Jews or Muslims (Rios et al., 2015). Furthermore, perceptions of the religion–science conflict are less pronounced among American Jews and Muslims (Vaidyanathan et al., 2016) and in non-Christian societies (e.g., the UAE; Rios & Aveyard, 2019) than among Christians.

Stereotypes of Religion–Science Incompatibility

Social psychological research demonstrates that religion and science are perceived as conflicting with one another, even implicitly. For example, people from majority-Christian countries who are induced to consider strong scientific explanations for events (e.g., the origin of the universe) subsequently exhibit less positive automatic evaluations of God relative to science, whereas those induced to consider strong faith-based explanations of events subsequently exhibit less positive automatic evaluations of science relative to God (Preston & Epley, 2009). Some studies also suggest that the cognitive processes involved in religious belief (e.g., faith, intuition) may be at odds with the cognitive processes involved in scientific reasoning (e.g., logic, analytical thinking). In a meta-analysis, religiosity correlated negatively with measures of general intelligence (Zuckerman et al., 2013), perhaps because many general intelligence measures are at least in part relevant to analytical reasoning skills. Indeed, analytical task performance is negatively associated with belief in God, a relationship that holds after controlling for measures of intelligence, cognitive ability, education, and other relevant individual differences (Pennycook et al., 2016). However, this association is much less robust in non-Western countries where Christians are in the minority (Gervais et al., 2018).

Thus, there appears to be a lay perception of a religion–science conflict in predominantly Christian societies (Preston & Epley, 2009). The negative relationship between religiosity and analytical thinking style is sometimes interpreted as evidence in favor of such a conflict. According to the dual-process model of information processing, for instance, the System 1 (intuitive) mode of thought implicated in religious belief can be overridden or inhibited by the System 2 (analytical) mode of thought (Pennycook et al., 2016). But regardless of the actual conflict (or lack thereof) between religion and science, the present research investigates whether and why reminders of the religion–science conflict narrative can affect Christians, both in terms of their scientific task performance and their subjective feelings of stereotype threat.

Stereotype Threat versus Disengagement

The central premise of stereotype threat theory is that stigmatized group members underperform in a particular domain due to concerns about confirming a negative societal stereotype as self-characteristic. For example, African Americans perform worse than White Americans on standardized tests when they indicate their race/ethnicity beforehand, but these performance differences disappear when participants are not asked to indicate their race/ethnicity (Steele & Aronson, 1995). Furthermore, women perform worse than men on a difficult math test when they are told there are gender differences in test performance, but not when they are told there are no gender differences in test performance (Spencer et al., 1999). Importantly, a group need not be stigmatized in general in order for its members to be susceptible to stereotype threat effects. White American men perform worse on athletic tasks when reminded of the stereotype that African Americans have superior athletic ability (Stone et al., 1999) and perform worse on math tests when reminded of the “model minority” stereotype of Asian Americans in math (Aronson et al., 2001). Therefore, even dominant group members can experience stereotype threat, so long as a negative stereotype about their group relative to another group is activated.

Research on social identity threat, a broader theoretical perspective under which stereotype threat is subsumed (Steele et al., 2002), also sheds light on the psychological processes responsible for such effects. Social identity threat involves a general sense that one’s group is devalued, even if a specific stereotype (e.g., about performance) is not activated. According to this perspective, reduced interest and engagement in a field can stem from a perceived lack of belonging. For instance, women may feel out of place in science, technology, engineering, and mathematics (STEM) fields not necessarily because they believe others hold negative perceptions of their group’s abilities, but rather because their values, interests, and experiences differ from those of their male colleagues (Cheryan et al., 2009). Importantly, although subjective feelings of stereotype threat by nature reflect a low sense of belonging (e.g., being stigmatized in a particular domain means that one’s group is perceived as not suited for that domain), a low sense of belonging can arise from reasons other than stereotype threat (e.g., general feelings of “invisibility”).

Stereotype and social identity threat theories have recently begun to be applied to religious groups. For example, religious minorities (e.g., Jews, Muslims) and highly religious Protestants in the United States often report feelings of threat to their religious identity, and such feelings are associated with a reduced sense of belonging and a greater likelihood of concealing one’s religion (Pasek & Cook, 2019). Similar results have been obtained among religious believers in science specifically. In a survey of biologists and physicists at American universities, 40.4% of Protestant scientists claimed that they faced religious discrimination at work, the second-highest proportion after Muslims (of whom 63.6% claimed to face discrimination; Scheitle & Ecklund, 2018).

Most relevant to the present research, one set of studies demonstrated that Christian participants underperformed relative to non-Christian participants on tasks described as measuring scientific reasoning ability when negative stereotypes about Christians in science were made salient, but not when negative stereotypes of Christians were debunked (Rios et al., 2015). However, the sample sizes in these studies were small by contemporary standards in social/personality psychology, ranging from 17 to 30 participants per cell in all but one study. It is therefore critical to conduct higher-powered tests of these effects, especially because the replicability of stereotype threat research has recently been questioned (Flore et al., 2018; Flore & Wicherts, 2015; Jussim et al., 2016).

Equally important, in these seminal studies, it is unknown whether Christians actually experienced subjective feelings of stereotype threat in the face of negative stereotypes, or alternatively, whether they disengaged even further from a domain (science) that they already considered peripheral or irrelevant to their identity. One theoretical perspective related to the “disengagement” phenomenon is identity-based motivation. According to this theory, members of stigmatized groups (e.g., racial/ethnic minorities) may come to see behaviors that are common among members of non-stigmatized groups (e.g., White Americans) as unimportant to “people like them.” For example, racial/ethnic minorities often associate health-promoting activities, such as diet and exercise, with White middle-class individuals and as a result, are less likely to pursue such activities (Oyserman et al., 2007). Rather than feeling as though their group is devalued or negatively stereotyped in health domains, as stereotype or social identity threat theories might suggest, racial/ethnic minorities do not consider health-promoting activities a central part of their group’s identity in the first place.

In potential support for the notion that negative stereotypes may lead Christians to disengage from science, a recent study found that the relationship between religiosity and lower scientific literacy was partially mediated by religious individuals’ greater negativity toward science (McPhetres & Zuckerman, 2018). One interpretation of these findings is that religious believers tend to perceive science as incongruent with their faith and thus are unmotivated to persist at science or acquire scientific knowledge. In the context of the studies in which negative stereotypes about Christians’ scientific abilities were activated (Rios et al., 2015), Christian participants could have simply interpreted the stereotypes as reinforcing their perception that science was “not for them.” To address this possibility in the present research, time spent on the task was measured as a potential indicator of persistence; a disengagement account would predict less time spent on the task as well as lower task performance in the face of negative stereotypes.

Overview of Studies

Four studies investigated whether and under what conditions stereotypes about the Christianity–science conflict would lead Christians to underperform on scientific reasoning tasks and experience feelings of stereotype threat (vs. disengagement). Data and syntax from all studies are publicly available on Open Science Framework (https://osf.io/j67tc/). In Studies 1 and 2, participants’ perceptions of the compatibility between Christianity and science were tested as a predictor—measured in Study 1, manipulated in Study 2—of religious differences in scientific task performance, time spent on the task, and subjective feelings of stereotype threat. In Studies 3 and 4, identification with science was assessed as a moderator to further test whether reminders of the Christianity–science conflict narrative would induce feelings of stereotype threat or disengagement among Christians. Based on stereotype threat theory, strongly science-identified Christians should be more likely than weakly science-identified Christians to underperform in the incompatible condition (Aronson et al., 2001), whereas a disengagement explanation would predict lower performance among weakly identified Christians in the incompatible condition. Importantly, the performance tasks in each study were described as assessing scientific reasoning; previous research on negative stereotypes of Christians in science has found that describing an otherwise identical task as irrelevant to science (e.g., as assessing “intuitive thought”) eliminates performance differences between Christians and non-Christians (Rios et al., 2015).

Studies 2 to 4 activated beliefs about the conflict between Christianity and science, rather than explicitly activating negative stereotypes about Christians’ scientific abilities (see Rios et al., 2015). Given the widely held perception among Western public figures (Scheitle & Ecklund, 2017) and scientists (Ecklund et al., 2016) that Christianity and science are incompatible, as well as the growing social unacceptability of expressing negative stereotypes against any group (Crandall et al., 2002), the former manipulations could be argued to have greater external validity. That is, it may be more likely for stereotypes of the Christianity–science conflict (which do not directly implicate Christians’ scientific abilities), relative to negative stereotypes of Christians’ aptitude for science, to be mentioned in STEM classrooms and organizations.

Study 1¹

Method

Participants

A total of 407 workers from Amazon’s Mechanical Turk (mTurk) platform participated in exchange for monetary compensation; 400 timeslots total were posted to obtain a sample size of approximately 100 per “cell” with one factor (compatibility perceptions) being a continuous moderator. Four participants with duplicate IP addresses and three non-U.S. residents were excluded from analyses; there were no statistical outliers. The final sample thus consisted of 400 individuals (181 men, 219 women; M _age = 39.16, SD = 12.55), 214 of whom self-identified as Christian (77 Catholic, 100 Protestant, 37 Other), and 186 as non-Christian (4 Jewish, 3 Muslim, 6 Hindu, 3 Buddhist, 50 Atheist, 58 Agnostic, 6 Other Religious Affiliation, 56 No Affiliation).² A sensitivity analysis using G*Power 3.1 (Faul et al., 2009) revealed that a sample of 400 in a 2 × 2 design was sufficient to detect a small-to-medium effect (f = .14) with 80% power and an alpha of .05.

Procedure and materials

In this and all subsequent studies, participants were told that the purpose of the study was to examine the relationship between identity and scientific ability. After providing informed consent, participants answered four questions on 1 to 7 scales assessing perceived compatibility between Christianity and science (e.g., “How incompatible vs. compatible do you personally believe science and Christianity are?”; M = 4.71, SD = 1.52, α = .89).

Next, participants completed an 11-question, multiple-choice measure of their basic knowledge about science (e.g., “Oil, natural gas, and coal are examples of…biofuels, fossil fuels, renewable resources, or geothermal resources”; Pew Research Center, 2019). The total number of correct answers served as the performance measure (M = 8.22, SD = 2.90). Time taken to complete the task was recorded and subjected to a square root transformation to reduce skew (M = 220.48 s, SD = 159.73, skewness = 2.81 before and .90 after transformation, SE = 0.22).

After indicating their religious affiliation, participants responded to four items assessing subjective feelings of stereotype threat. The items, administered on 7-point scales (1 = strongly disagree, 7 = strongly agree), were adapted directly from a previously validated measure of felt stereotype threat among African Americans in academics (Marx & Goff, 2005). The items were worded slightly differently depending on whether participants reported affiliating with a religion (i.e., Christian, Jewish, Muslim, Buddhist, Hindu, Other) or not (i.e., Atheist, Agnostic, No Religion). An example item for religious (non-religious) individuals was “I worry that my ability to perform well in science is affected by my religion (lack of religion).” Participants’ responses were averaged into a composite (M = 2.15, SD = 1.46, α = .96).

At the end of the study, participants responded to additional demographic questions, including strength of belief in God (Preston & Epley, 2005), age, gender, highest education completed, and Biblical literalism (e.g., “Should the holy book associated with your religion [e.g., Bible, Torah, Koran] be interpreted literally, word for word?”). The study results did not change after adding any of these measures as covariates, so they will not be discussed further.

Results

See Table 1 for zero-order correlations among the key variables in this study.

Table 1.

Zero-Order Correlations, Study 1.

Religious affiliation (0 = non- Christian)	Religious affiliation (0 = non- Christian)	Compatibility perceptions	Stereotype threat	Performance
Religious affiliation (0 = non-Christian)		.44***	.31***	–.26***
Compatibility perceptions	.44***		.09⁺	–.01
Stereotype threat	.31***	.09⁺		–.53***
Performance	–.26***	–.01	–.53***

***p < .001. ⁺ p < .10.

Performance

It was predicted that Christians would underperform on the task relative to non-Christians, especially when they believed Christianity and science were incompatible. To test this prediction, participants’ scientific literacy scores were submitted to a multiple regression using PROCESS Model 1 (Hayes, 2013), with religious affiliation as the predictor and compatibility perceptions as the moderator. Overall, non-Christians scored higher than did Christians, b = –1.82, SE = 0.31, 95% CI = [–2.43, –1.21], t(397) = –5.87, p < .001, and greater compatibility perceptions predicted higher scores, b = 0.23, SE = 0.10, 95% CI = [0.03, 0.43], t(397) = 2.29, p = .023. Critically, the two-way interaction was also significant, b = 0.79, SE = 0.20, 95% CI = [0.39, 1.19], t(396) = 3.87, p < .001 (see Figure 1). Among participants who perceived Christianity and science as relatively incompatible (1 SD below the mean), non-Christians scored higher than Christians, b = –3.09, SE = 0.45, 95% CI = [–3.97, –2.21], t(396) = –6.91, p < .001. There were no differences in scores among Christian and non-Christian participants who perceived Christianity and science as relatively compatible (1 SD above the mean), b = –0.68, SE = 0.42, 95% CI = [–1.52, 0.15], t(396) = –1.61, p = .109. Furthermore, compatibility beliefs positively predicted Christian participants’ scores, b = 0.70, SE = 0.16, 95% CI = [0.39, 1.01], t(396) = 4.47, p < .001, and were unrelated to non-Christian participants’ scores, b = –0.09, SE = 0.13, 95% CI =[ –0.34, 0.16], t(396) = –0.68, p = .498.

Figure 1.

Science literacy as a function of compatibility beliefs (plotted at 1 SD above and below the sample mean) and religious affiliation (Christian vs. non-Christian), Study 1.

Subjective feelings of stereotype threat

Overall, Christians reported greater subjective feelings of stereotype threat than did non-Christians, b = 0.97, SE = 0.16, 95% CI = [0.67, 1.28], t(396) = 6.30, p < .001. There was no relationship between compatibility perceptions and subjective feelings of stereotype threat (p = .256). However, the religious affiliation × compatibility perceptions interaction was significant, b = –0.39, SE = 0.10, 95% CI = [–.59, –.19], t(396) = –3.88, p < .001 (see Figure 2). Simple slopes analyses indicated that Christians reported significantly greater subjective feelings stereotype threat than non-Christians when they perceived Christianity and science as relatively incompatible, b = 1.60, SE = 0.22, 95% CI = [1.17, 2.04], t(396) = 7.21, p < .001; this effect was marginal among those who perceived Christianity and science as relatively compatible, b = 0.40, SE = 0.21, 95% CI = [–0.01, 0.82], t(396) = 1.91, p = .056. Furthermore, compatibility beliefs predicted lower subjective feelings of stereotype threat among Christians, b = –0.29, SE = 0.08, 95% CI = [–0.44, –0.14], t(396) = –3.73, p < .001, but not among non-Christians, b = 0.10, SE = 0.06, 95% CI = [–0.02, 0.23], t(396) = 1.59, p = .114.

Figure 2.

Subjective feelings of stereotype threat as a function of compatibility beliefs (plotted at 1 SD above and below the sample mean) and religious affiliation (Christian vs. non-Christian), Study 1.

Time spent on task

The religious affiliation × compatibility perceptions interaction on this measure was nonsignificant, b = 0.44, SE = 0.35, 95% CI = [–0.25, 1.13], t(396) = 1.26, p = .209. There were also no significant or marginal main effects (ps > .104).

Mediated moderation analysis

To test whether subjective feelings of stereotype threat mediated the relationship between religious affiliation (predictor) and performance (outcome) for participants who perceived Christianity and science as incompatible (moderator), a mediated moderation analysis was conducted using PROCESS Model 8 with 5,000 bootstrapping estimates (Hayes, 2013). This analysis revealed an overall indirect effect (b = 0.37, SE = 0.10, 95% CI = [0.19, 0.59]), which was significant among those who perceived Christianity and science as relatively incompatible (b = –1.51, SE = 0.28, 95% CI = [ –2.09, –1.00]), but not among those who perceived Christianity and science as relatively compatible (b = –0.38, SE = 0.20, 95% CI = [–0.79, 0.02]). This mediated moderation is depicted in Figure 3. The overall indirect effect was also significant for the reverse model, with scientific literacy scores as the mediator and subjective feelings of stereotype threat as the outcome (b = –0.18, SE = 0.05, 95% CI = [–0.30, –0.08]), as was the conditional indirect effect for those who perceived Christianity and science as relatively incompatible (b = 0.72, SE = 0.15, 95% CI = [0.45, 1.03]). The reverse mediation path will be revisited in the General Discussion.

Figure 3.

Mediated moderation analysis, Study 1 (standardized betas).

Because tests of mediation are based on the assumption that the mediator has the same relationship with the outcome at each level of the predictor, a multiple regression analysis was conducted using PROCESS Model 3 (Hayes, 2013), with religious affiliation as the predictor (X), compatibility perceptions as the first moderator (W), subjective feelings of stereotype threat as the second moderator (Z), and performance as the outcome. The three-way interaction was not significant, b = 0.03, SE = 0.14, 95% CI = [–0.26, 0.30], t(392) = 0.18, p = .854, supporting this homogeneity assumption.

Discussion

Study 1 demonstrated that among participants who believed Christianity and science were incompatible, Christians underperformed on a scientific literacy task and reported greater subjective feelings of stereotype threat (i.e., beliefs that they would be negatively evaluated in scientific domains based on their religion) relative to non-Christians. Critically, there were no effects on time taken to complete the task, suggesting that Christians who endorsed incompatibility beliefs did not underperform relative to non-Christians simply due to a lack of engagement. Nevertheless, the design of Study 1 was correlational, so a series of subsequent studies was conducted to experimentally manipulate perceptions of compatibility.

Study 2

Given recent concerns raised about the replicability of stereotype threat research (e.g., Flore & Wicherts, 2015), Study 2 was preregistered to increase confidence in the robustness of the results (https://osf.io/54rkq).³

Method

Participants

A total of 517 mTurk workers participated in exchange for monetary compensation. As stated in the preregistration, the targeted sample size in this study was 250 (∼50 participants per cell in a 2 × 2 design after exclusions), given that experiments typically produce larger effects than correlational studies. However, preliminary inspection of the data revealed that one cell contained only 31 participants, so the sample size was doubled. Of the total participants, 34 with duplicate IP addresses, 15 participants who suspected that the article was not real or meant to influence their scientific reasoning performance, 35 participants who failed all three comprehension check items,⁴ four non-U.S. residents, and one statistical outlier with a very high standardized residual score on the dependent measure (–3.10; see Figure S1 in the Supplemental Materials for a boxplot of values) were omitted, leaving 428 individuals (210 men, 218 women; M _age = 37.19, SD = 11.99) in the final sample. A sensitivity analysis revealed that a sample of 428 in a 2 × 2 design was sufficient to detect a small to medium-sized effect (f = .14) with an alpha of .05 and 80% power.

Of the 428 participants retained, 240 self-identified as Christian (104 Catholic, 104 Protestant, 32 Other) and 188 as non-Christian (5 Jewish, 4 Muslim, 3 Hindu, 9 Buddhist, 50 Atheist, 54 Agnostic, 15 Other, 48 No Religion). Participants were randomly assigned to either the incompatible (n = 220) or compatible (n = 208) condition.

Procedure and materials

Participants were told they would read an op-ed article from a major news source. Similar manipulations have been used in prior stereotype threat research (e.g., Aronson et al., 2001). In the compatible (incompatible) condition, the article was titled “AAAS Speaks Out on Perceived Compatibility (Incompatibility) Between Christianity and Science.” It reported the results of a purportedly real survey demonstrating that most Americans, in general and within the scientific community, believed Christianity and science were compatible (incompatible). Participants then answered three questions about the content of the article (i.e., the organization that conducted the survey, the name of the primary investigator, the main finding of the survey). As noted above, participants who answered all three questions incorrectly were excluded from analyses.

After reading the article, participants completed 15 nonsense syllogisms—described as a measure of scientific reasoning—in which they indicated whether a conclusion (e.g., “Therefore, no ghost is a cat”) followed logically from two premises (e.g., “All cats are electrified. No ghosts are electrified”) (Markman et al., 2007). The total number of correct answers served as the dependent measure (M = 9.80, SD = 2.55). Although time spent on this task was not recorded, time spent on the entire study and the article manipulation were both recorded, so the difference between the two served as an approximate measure of persistence (M = 530.11 s, SD = 340.75) and was subjected to a square root transformation (skewness = 2.82 before and 1.34 after transformation, SE = 0.12).

Next, participants completed a subjective feelings of stereotype threat measure on a 7-point scale (1 = strongly disagree, 7 = strongly agree; M = 2.51, SD = 1.52, α = .93), administered after rather than before the syllogisms task to avoid the possibility that participants’ performance would be influenced by their exposure to the items. Unlike in Study 1, all items referred to “my religion” and therefore were not relevant to non-religious participants. However, in Study 1 for which the measure was worded to pertain to all individuals, there were still no effects among non-Christian participants.⁵

As a manipulation check, the four compatibility items from Study 1 were administered at the end of the study on 7-point scales (M = 4.62, SD = 1.59, α = .72), along with the demographic survey. Finally, participants were probed for suspicion and debriefed.

Results

Performance

A 2 (religious affiliation: Christian vs. non-Christian) × 2 (condition: compatible vs. incompatible) between-subjects analysis of variance (ANOVA) revealed a marginal main effect of religious affiliation, such that non-Christians (M = 10.05, SD = 2.51, 95% CI = [9.65, 10.37]) tended to perform better than Christians (M = 9.61, SD = 2.57, 95% CI = [9.28, 9.93]), F(1, 424) = 2.71, p = .100, $η_{p}^{2} = .006$ . There was no main effect of condition: F(1, 424) = 0.58, p = .446, $η_{p}^{2} = .001$ .

Critically, the two-way interaction was significant, F(1, 424) = 11.83, p = .001, $η_{p}^{2} = .027$ . Decomposition of the interaction indicated that, as predicted, Christians (M = 9.28, SD = 2.58, 95% CI = [8.82, 9.73]) underperformed relative to non-Christians (M = 10.52, SD = 2.48, 95% CI = [10.03, 11.02]) in the incompatible condition, F(1, 424) = 13.45, p < .001, $η_{p}^{2} = .031$ , but Christians (M = 9.93, SD = 2.53, 95% CI = [9.49, 10.38]) and non-Christians (M = 9.49, SD = 2.44, 95% CI = [8.96, 10.02]) performed equally well in the compatible condition, F(1, 424) = 1.55, p = .214, $η_{p}^{2} = .004$ . Furthermore, Christians performed significantly better in the compatible than incompatible condition, F(1, 424) = 4.09, p = .044, $η_{p}^{2} = .010$ , whereas the reverse was true for non-Christians, F(1, 424) = 7.85, p = .005, $η_{p}^{2} = .018$ .

Subjective feelings of stereotype threat

A 2 × 2 between-subjects ANOVA revealed that overall, Christian participants (M = 2.92, SD = 1.57, 95% CI = [2.74, 3.10]) reported greater feelings of stereotype threat than did non-Christian participants (M = 1.98, SD = 1.29, 95% CI = [1.78, 2.20]), F(1, 424) = 43.79, p < .001, $η_{p}^{2} = .094$ . There was no main effect of condition, F(1, 424) = 1.75, p = .187, $η_{p}^{2} = .004$ . More germane to hypotheses, the two-way interaction was significant, F(1, 424) = 5.61, p = .018, $η_{p}^{2} = .013$ . Simple effects tests revealed that Christians (M = 3.18, SD = 1.63, 95% CI = [2.92, 3.44]) reported greater feelings of stereotype threat than did non-Christians (M = 1.92, SD = 1.25, 95% CI = [1.63, 2.20]) in the incompatible condition, F(1, 424) = 41.98, p < .001, $η_{p}^{2} = .090$ . This difference between Christians (M = 2.66, SD = 1.47, 95% CI = [2.40, 2.92]) and non-Christians (M = 2.06, SD = 1.33, 95% CI = [1.76, 2.37]) was weaker but still significant in the compatible condition, F(1, 424) = 8.69, p = .003, $η_{p}^{2} = .020$ . Furthermore, Christians reported greater feelings of stereotype threat in the incompatible than compatible condition, F(1, 424) = 7.78, p = .006, $η_{p}^{2} = .018$ , but non-Christians did not differ in reported feelings of stereotype threat across conditions, F(1, 424) = 0.49, p = .485, $η_{p}^{2} = .001$ .

Time spent on task

There were no main effects (ps > .274). The two-way condition × religious affiliation interaction was marginal, F(1, 424) = 3.21, p = .074, $η_{p}^{2} = .008$ , and the pattern suggested that Christians spent less time on the task than non-Christians in the incompatible condition, but more time on the task than non-Christians in the compatible condition. However, these simple effects did not reach significance in either the incompatible condition, F(1, 424) = 0.81, p = .368, $η_{p}^{2} = .002$ , or the compatible condition, F(1, 424) = 2.62, p = .106, $η_{p}^{2} = .006$ .

Mediated moderation analysis

A mediated moderation analysis was conducted using PROCESS Model 8 (Hayes, 2013). Religious affiliation (0 = non-Christian, 1 = Christian) was entered as the predictor, condition (0 = incompatible, 1 = compatible) as the moderator, subjective feelings of stereotype threat as the mediator, and performance as the outcome. The overall indirect effect with 5,000 bootstrapping estimates was significant (b = 0.31, SE = 0.14, 95% CI = [0.05, 0.60]). Furthermore, the conditional indirect effect was more pronounced in the incompatible condition (b = –0.58, SE = 0.14, 95% CI = [–0.88, –0.33]) than in the compatible condition (b = –0.28, SE = 0.11, 95% CI = [–0.51, –0.10]). Thus, especially among those induced to believe Christianity and science were incompatible, being Christian (vs. non-Christian) negatively predicted subjective feelings of stereotype threat and in turn task performance (see Figure 4). The overall indirect effect (b = –0.25, SE = 0.09, 95% CI = [–0.45, –0.11]) and the conditional indirect effect in the incompatible condition (b = 0.19, SE = 0.06, 95% CI = [0.08, 0.33]) were also both significant for the reverse model (i.e., with performance as the mediator and subjective feelings of stereotype threat as the outcome); these findings are addressed in the General Discussion.

Figure 4.

Mediated moderation analysis, Study 2 (standardized betas).

To test the homogeneity assumption of mediation, a moderated multiple regression analysis was conducted using PROCESS Model 3 (Hayes, 2013), as in Study 1. The three-way interaction was not significant, supporting this assumption, b = 0.32, SE = 0.34, 95% CI = [–0.36, 0.99], t(420) = 0.93, p = .355.

Manipulation check

Participants in the compatible condition (M = 5.01, SD = 1.49, 95% CI = [4.74, 5.14]) perceived Christianity and science as more compatible than did participants in the incompatible condition (M = 4.24, SD = 1.58, 95% CI = [4.01, 4.39]), F(1, 426) = 26.79, p < .001, $η_{p}^{2} = .059$ , thus validating the manipulation.

Discussion

Study 2 demonstrated that Christian participants reminded of the perceived conflict between their religion and science both performed worse on a scientific reasoning task and reported greater subjective feelings of stereotype threat than did non-Christian participants; the amount of time they spent on the task was not affected. Although the results of Studies 1 and 2 suggest that subjective feelings of stereotype threat may help explain Christians’ underperformance in science following reminders of the conflict narrative, the limited utility of cross-sectional mediation analyses rendered it necessary in Studies 3 and 4 to conduct an alternative investigation of stereotype threat as a mechanism.

Study 3

The main goal of Studies 3 and 4 was to measure the strength of participants’ identification with science as a moderator of the hypothesized effects; indeed, moderation effects can often speak to underlying psychological processes (Spencer et al., 2005). If stereotype threat is at play, then Christian participants higher in identification with science should be most susceptible to performance decrements in the face of information about incompatibility, as such participants should be especially concerned about their task performance. Supporting this possibility, prior research has demonstrated that stereotype threat effects are strongest among individuals who identify with the domain relevant to the task—in this case, science (Aronson et al., 2001). However, if disengagement is at play, then Christian participants lower in identification with science should be most susceptible to performance decrements when they receive incompatibility-related information, as such participants should be most likely to dismiss the task as inconsistent with their preexisting values and thus be unmotivated to perform well.

Method

Participants

A total of 305 mTurk workers participated in exchange for monetary compensation. Because this study was conducted in 2015, before norms for larger sample sizes were well-established in social psychology, the targeted number of participants was 300, or 25 per “cell” in a 3 × 2 × 2 design (with one factor, science identification, being a continuous moderator). Of the total participants, 24 with duplicate IP addresses, 17 participants who failed a comprehension check question about the article’s content, eight participants who suspected that the article may have influenced their subsequent responses, and one statistical outlier with a high studentized deleted residual score on the dependent measure (–3.06; see Figure S2 in the Supplemental Materials for a boxplot of the values) were omitted, leaving 255 individuals in the final sample (150 women, 99 men, 3 other, 3 unspecified; M _age = 35.88, SD = 12.25). Of them, 108 self-identified as Christian (29 Catholic, 45 Protestant, 34 Other) and 147 as non-Christian (6 Jewish, 3 Muslim, 3 Hindu, 3 Buddhist, 36 Atheist, 40 Agnostic, 7 Other, 49 No Religion). A sensitivity power analysis revealed that a sample of 255 in a 3 × 2 × 2 design was sufficient to detect a small-to-medium effect (f = .15) with 80% power and an alpha of .05 (Faul et al., 2009). Participants were randomly assigned to one of three conditions: compatible (n = 86), incompatible (n = 78), or control (n = 91).

Procedure and materials

Participants in the compatible (incompatible) condition read a short passage, ostensibly an op-ed from a major news source, arguing that science and Christianity can coexist (are in conflict). The passage cited evidence from across cultures and over time to support the central argument. To bolster the cover story, participants in the control condition were told that they would read an article at the end of the study; in reality, they did not read an article at all.

Next, participants completed the 15 nonsense syllogisms from Study 2, as well as 10 items from the Remote Associates Test (see Rios et al., 2015) in which they were presented with three words (e.g., envy, golf, beans) and generated a fourth word related to all three of those words (i.e., green). The total number of correct solutions across both tasks served as the dependent measure (M = 16.99, SD = 4.03). Time spent on the task was also recorded (M = 482.50 s, SD = 271.02) and subjected to a square root transformation (skewness = 1.81 before and .81 after transformation, SE = 0.15).

Participants then indicated the extent to which five science-related traits described them (1 = not at all, 5 = extremely): rational, logical, analytical, scientifically minded, and research oriented. Their responses were averaged into a science identification composite (M = 3.56, SD = .76, α = .84); scores on this measure were not affected by condition, religious affiliation, or their interaction (ps > .349). To minimize suspicion, participants also indicated the extent to which general competence-related traits (e.g., competent, intelligent) and four warmth-related traits (e.g., warm, sincere) described them (see Fiske et al., 2002).

As a manipulation check, participants answered the following questions: (a) To what extent do you personally believe that science and Christianity are incompatible versus compatible? and (b) To what extent do you personally believe that it is impossible versus possible to be both Christian and a scientist? (1 = definitely incompatible/impossible, 7 = definitely compatible/possible; M = 4.45, SD = 1.73, α = .75). They then completed a demographic survey and suspicion probe, prior to being debriefed.

Results

Performance

To test whether identification with science would moderate the effects of religious affiliation (Christian vs. non-Christian) and condition (compatible vs. incompatible vs. control) on scientific task performance, participants’ performance scores on the syllogisms and remote associates test were submitted to a multiple regression using PROCESS Model 3 (Hayes, 2013). In this model, religious affiliation was the predictor, condition (indicator coded with the incompatible condition as the comparison group) was the first moderator (W), and science identification was the second moderator (Z). Overall, non-Christians performed marginally better than Christians (b = –0.99, SE = 0.50, 95% CI = [–1.98, 0.002]), t(250) = –1.97, p = .050, and participants in the incompatible condition performed marginally better than did participants in the compatible condition (b = –1.12, SE = 0.62, 95% CI = [–2.34, 0.11]), t(250) = –1.79, p = .074. There were no other lower-order effects.

The three-way interactions were significant both when comparing the incompatible and compatible conditions (b = 5.12, SE = 1.74, 95% CI = [1.70, 8.55]), t(243) = 2.95, p = .004, and when comparing the incompatible and control conditions (b = 3.28, SE = 1.63, 95% CI = [0.06, 6.50]), t(243) = 2.01, p = .046 (see Figure 5). The two-way condition (incompatible vs. compatible) × religious affiliation (Christian vs. non-Christian) interaction was significant for high science identifiers (1 SD above the mean; b = 4.36, SE = 1.86, 95% CI = [0.69, 8.02]), t(243) = 2.34, p = .020, and marginal for low science identifiers (1 SD below the mean; b = –3.43, SE = 1.80, 95% CI = [–6.96, 0.11]), t(243) = –1.91, p = .058.⁶ Simple slope analyses revealed that highly science-identified non-Christians outperformed Christians in the incompatible condition (b = –4.46, SE = 1.34, 95% CI = [–7.24, –1.69]), t(243) = –3.17, p = .002, but not in the compatible condition (b = –0.10, SE = 1.22, 95% CI = [–2.50, 2.29]), t(243) = –0.09, p = .932, or control condition (b = –1.24, SE = 1.13, 95% CI = [–3.47, 0.99]), t(243) = –1.10, p = .273. No simple slopes were significant or marginal among low science identifiers (ps > .119).

Figure 5.

Scientific reasoning task performance as a function of science identification (plotted at 1 SD above and below the sample mean), condition (incompatible vs. compatible vs. control), and religious affiliation (Christian vs. non-Christian), Study 3.

Time spent on task

The three-way interaction was significant when comparing the incompatible and compatible conditions (b = 5.34, SE = 2.49, 95% CI = [0.44, 10.24]), t(243) = 2.15, p = .033, but not when comparing the incompatible and control conditions (b = 3.40, SE = 2.34, 95% CI = [–1.20, 8.01]), t(243) = 1.46, p = .147. Notably, none of the simple slopes reached significance (ps > .110), nor did the two-way interactions for high and low science identifiers (ps > .210). However, the pattern of results suggested a trend for highly science-identified Christians to spend less time on the task than non-Christians in the incompatible condition, but not in the other conditions.

Manipulation check

The omnibus ANOVA was marginal, F(2, 252) = 3.00, p = .052, $η_{p}^{2} = .023$ . As predicted, however, participants in the incompatible condition (M = 4.05, SD = 1.78) perceived Christianity and science as less compatible than did participants in the compatible condition (M = 4.59, SD = 1.65, b = –0.54, SE = 0.27, 95% CI = [–1.07, –.012], p = .045). Participants in the incompatible condition also perceived Christianity and science as less compatible than did participants in the control condition (M = 4.65, SD = 1.74, b = –0.60, SE = 0.27, 95% CI = [–1.12, –0.07], p = .025), who did not differ from those in the compatible condition (b = –0.06, SE = 0.26, 95% CI = [–0.57, 0.45], p = .831).

Discussion

Using a measure of science identification as a moderator, the results of Study 3 were consistent with a stereotype threat account. Specifically, being induced to believe that Christianity and science are incompatible (vs. incompatible or receiving no information at all) led highly science-identified Christians to underperform on a scientific reasoning task relative to non-Christians. This is because stereotype threat is especially pronounced among people who care about performing well on the task at hand (Aronson et al., 2001). By contrast, Christians who considered science unimportant to their self-definition were unaffected by the manipulation, suggesting that Christians’ underperformance did not arise from a lack of motivation (e.g., because science was not aligned with their values in the first place). Although the pattern of the results on time taken to complete the task potentially supports the disengagement account, the effects were weak and will be subject to replication in Study 4.

Study 4

The central objectives of Study 4 were to conceptually replicate Study 3 with a larger sample size, a more subtle manipulation of exposure to stereotypes of the Christianity–science conflict, and the science identification measure (the critical moderator) at the beginning rather than end of the experiment.

Method

Participants

A total of 408 mTurk workers participated in exchange for monetary compensation. The targeted sample size in this study was 400, or 50 per “cell” in a 2 × 2 × 2 design (with science identification as a continuous moderator). Six participants with duplicate IP addresses, 21 participants who failed a basic attention check (i.e., “Please select ‘somewhat disagree’ for this item if you are paying attention”), seven non-U.S. citizens, and one outlier with a high studentized deleted residual score on the dependent measure (–3.35; see Figure S3 in the Supplemental Materials for a boxplot of values) were omitted from analyses. The final sample thus consisted of 373 individuals (173 men, 200 women; M _age = 36.26, SD = 12.76), of whom 186 identified as Christian (76 Protestant, 60 Catholic, 50 Other) and 187 as non-Christian (9 Jewish, 3 Muslim, 1 Hindu, 6 Buddhist, 54 Atheist, 62 Agnostic, 40 No Affiliation, 12 Other Affiliation). A sensitivity power analysis revealed that a sample of 373 in a 2 × 2 × 2 design was sufficient to detect a small-to-medium effect (f = .15) with 80% power and an alpha of .05 (Faul et al., 2009). Participants were randomly assigned to either the compatible (n = 177) or incompatible (n = 196) condition. The control condition was not included to maximize power.

Procedure and materials

Participants first completed a 7-item measure of their identification with science. Two items (It is important to me that I am good at science and I am good at science) were taken from prior stereotype threat research (1 = strongly disagree, 7 = strongly agree; Steele & Aronson, 1995). The five remaining items were the same as in Study 3 and once again were interspersed with items about general competence and warmth to reduce suspicion. Because the science identification items were administered on different scales, responses to each item were standardized before averaging them into a composite (α = .83).

Next, participants completed the compatibility manipulation, which took the form of biased statements. Specifically, participants viewed five items related to the compatibility between Christianity and science, administered on 7-point scales (1 = strongly disagree, 7 = strongly agree), but these items either included the word “sometimes” in the compatible condition (e.g., Christianity and science are sometimes compatible) or “always” in the incompatible condition (e.g., Christianity and science are always compatible). Because it is easier for most people to “sometimes” than “always” agree with a statement, the “sometimes” items were expected to elicit greater perceptions of compatibility. That is, participants should score higher on the “sometimes” relative to “always” statements and should thus shift their mindsets toward perceiving Christianity and science as more compatible in the former condition. This manipulation was first introduced by Salancik and Conway (1975) and has been used in recent research to manipulate self-perceptions of attitude certainty (e.g., people who respond to statements about “sometimes,” vs. “always,” being certain about their attitudes subsequently behave in ways that reflect greater attitude certainty; Rios et al., 2014).

After the compatibility manipulation, participants completed the 15 syllogisms from Studies 2 and 3. The total number of correct answers served as the dependent measure (M = 10.14, SD = 2.42). Time spent on the task was also recorded and subject to a square root transformation (M = 253.73 s, SD = 222.22, skewness = 5.16 before and 1.11 after transformation, SE = 0.13).

Finally, participants provided demographic information, completed a suspicion probe, and were debriefed. No participants suspected that their beliefs about Christianity–science compatibility were meant to influence task performance.

Results

Performance

To test whether identification with science would moderate the effects of religious affiliation (Christian vs. non-Christian) and condition (compatible vs. incompatible) on scientific task performance, participants’ syllogisms scores were submitted to a multiple regression using PROCESS Model 3, as in Study 3 (Hayes, 2013). There were no lower-order effects except for a positive overall relationship between science identification and performance (b = 0.55, SE = 0.18, 95% CI = [0.20, 0.89]), t(369) = 3.08, p = .002, and an overall effect of condition such that performance was higher among participants who read “sometimes” (compatible) than “always” (incompatible) statements (b = 0.52, SE = 0.25, 95% CI = [0.03, 1.00]), t(369) = 2.09, p = .037.

The three-way interaction between condition, religious affiliation, and science identification, although only marginal, was in the predicted direction (b = 1.31, SE = 0.73, 95% CI = [–0.12, 2.74]), t(365) = 1.81, p = .072 (see Figure 6). The two-way condition × religious affiliation interaction was significant among high science identifiers (1 SD above the M; b = 1.84, SE = 0.88, 95% CI = [0.11, 3.57]), t(365) = 2.10, p = .037, but not low science identifiers (1 SD below the mean; b = –0.78, SE = 0.88, 95% CI = [–2.51, 0.95]), t(365) = –0.89, p = .376. Decomposition of the former interaction revealed that highly science-identified Christians underperformed relative to non-Christians in the incompatible condition (b = –1.07, SE = 0.48, 95% CI = [–2.02, –0.12]), t(365) = –2.20, p = .028, but not in the compatible condition (b = 0.70, SE = 0.38, 95% CI = [–0.64, 1.39]), t(365) = 0.73, p = .467. In addition, highly science-identified Christians performed better in the compatible than incompatible condition (b = 1.24, SE = 0.54, 95% CI = [0.18, 2.30]), t(365) = 2.30, p = .022, whereas highly science-identified non-Christians did not differ in performance according to condition (b = –0.20, SE = 0.46, 95% CI = [–1.10, 0.70]), t(365) = –0.44, p = .660. No simple slopes among low science identifiers reached significance (ps > .339).

Figure 6.

Scientific reasoning task performance as a function of science identification (plotted at 1 SD above and below the sample mean), condition (incompatible vs. compatible), and religious affiliation (Christian vs. non-Christian), Study 4.

Time spent on task

Besides an overall positive relationship with science identification (b = 1.08, SE = 0.43, 95% CI = [0.24, 1.91]), t(368) = 2.54, p = .012, there were no effects on this variable. Most critically, the three-way interaction was not significant (b = 0.82, SE = 1.75, 95% CI = [–2.62, 4.26]), t(364) = 0.47, p = .640. Thus, the significant interaction from Study 3 did not replicate.⁷

Agreement with items

Participants agreed more strongly with the “sometimes” (compatible) statements (M = 5.24, SD = 1.19, 95% CI = [5.05, 5.43]) than the “always” (incompatible) statements (M = 3.82, SD = 1.37, 95% CI = [3.64, 4.00]), F(1, 371) = 114.35, p < .001, $η_{p}^{2} = .24$ . Thus, the manipulation was successful.

Discussion

The results of Study 4 provide further support for stereotype threat as an explanation for Christians’ scientific underperformance by demonstrating that only those who care about doing well at science are susceptible to performance decrements. Although the three-way interaction on performance was marginal, the simple slopes of interest (and the two-way interaction among high science identifiers) were significant, and the results replicated those of Study 3.

General Discussion

The present studies investigated how Christians react to reminders of the perceived Christianity–science conflict. In Studies 1 and 2, Christian participants who chronically believed (Study 1) or were induced to believe (Study 2) that Christianity and science were incompatible exhibited poorer scientific performance and greater subjective feelings of stereotype threat than did non-Christian participants. Although subjective feelings of stereotype threat mediated the relationship between religious affiliation and performance among participants led to perceive Christianity and science as incompatible, the cross-sectional nature of the mediated moderation analyses and the significance of the reverse path (with performance as the mediator and stereotype threat as the outcome) indicate these results should be interpreted with caution. In Studies 3 and 4, the effects of religious affiliation and compatibility beliefs on scientific performance were most pronounced for participants who identified strongly with science.

Collectively, the present studies investigated two potential mechanisms underlying Christians’ underperformance in science, which were briefly described but not tested in Rios et al. (2015). The first mechanism (stereotype threat) is that Christians underperform when induced to believe their religion is incompatible with science because they worry about confirming negative stereotypes of their group’s scientific abilities. In other words, they care about performing well on the task, and this concern leads them to “choke” under pressure and inhibits their performance. The second mechanism (disengagement) is that Christians underperform when induced to believe their religion and science are incompatible because such a belief confirms their initial perception of science as clashing with their personal value system. That is, they do not care about performing well on the task, and as a result, information about Christianity–science incompatibility causes them to disengage from and underperform on the task. The fact that, in Studies 3 and 4, the effects of the incompatibility manipulation were most pronounced among Christians who identified strongly with science suggests that Christians underperformed because they did not want to confirm societal stereotypes, not because they were unmotivated to perform well in science (i.e., perceived science as “not for them”) in the first place.

Although stereotype threat (relative to disengagement) emerged as the more compelling mechanism behind Christians’ and non-Christians’ performance differences in these studies, disengagement may be a stronger explanation in other contexts or populations. For example, Christian fundamentalists (e.g., who believe there is “one true religion”; Altemeyer & Hunsberger, 2004) may be less susceptible than non-fundamentalists to feelings of stereotype threat in science. Instead, Christian fundamentalists may underperform on scientific tasks because they perceive science as encroaching upon their core values and, consequently, care less about succeeding in science. In addition, Christians in predominantly religious environments (e.g., the “Bible Belt,” religious universities) may be less aware of or less affected by negative stereotypes about Christians’ scientific abilities. Like fundamentalists, if they do underperform on scientific tasks following reminders of such stereotypes, it could be due to their seeing science as antithetical to their beliefs, rather than to stereotype threat. Alternatively, Christians in such environments could be buffered against feelings of stereotype threat because they encounter more religious scientists in their everyday lives.

Notably, the manipulation in Study 4 (statements with biased wording) was subtle, whereas the manipulations in Studies 2 and 3 (supposedly real news articles) were more blatant. Prior stereotype threat research has similarly used a mix of subtle (e.g., having participants report their race/ethnicity prior to a test; Steele & Aronson, 1995) and blatant (e.g., having participants read fabricated articles; Aronson et al., 2001) manipulations. According to suspicion probes, many participants in the current studies realized the research was generally about religion and science. Indeed, consistent with some stereotype threat studies (e.g., Aronson et al., 2001; Spencer et al., 1999), the cover stories directly mentioned identity and scientific reasoning. However, only a small number of participants in Studies 2 and 3—and none in Study 4—reported awareness that the purpose of the studies was to examine the relationship between religion–science conflict perceptions and task performance, which minimizes concerns about demand effects.

As discussed earlier, the reproducibility of some stereotype threat effects has been questioned in recent years (Flore & Wicherts, 2015; Jussim et al., 2016). Critically, only one other set of studies has tested Christians’ responses to negative stereotypes about their group’s scientific abilities, though the underlying mechanisms were not directly investigated (Rios et al., 2015). It will be important for additional research to both attempt to replicate these findings and uncover critical boundary conditions. Whether or not stereotype threat effects in other domains (e.g., women in mathematics, racial/ethnic minorities in academia) are replicable is beyond the scope of the present work. However, it may be fruitful to examine if the robustness of stereotype threat depends on the social acceptability of expressing the negative stereotype at hand. In the Rios et al. (2015) studies, non-Christian participants were willing to explicitly report that they believed Christians’ scientific abilities were worse than both those of the average American and those of other religious groups (e.g., Jews, Muslims). By contrast, endorsing negative stereotypes about women and racial/ethnic minorities in academia is likely far less acceptable, and given the growing prevalence of interventions to broaden participation of these groups in STEM and other disciplines, the content of such stereotypes may have also changed over the years. Thus, stereotype threat effects may be especially pronounced the more strongly and openly the societal stereotype is held.

Limitations and Future Directions

A few limitations of the present research should be noted. First, data collection for the present studies occurred online. Yet, the fact that online settings are by nature more anonymous may have made the present studies a more conservative test of hypotheses. If, for instance, Christian participants were told face-to-face by an experimenter that most people believed religion and science were incompatible, their fear of confirming negative societal stereotypes may have been even more pronounced. Furthermore, recent investigations of stereotype threat in various domains and groups have relied primarily on online populations (e.g., Alquist et al., 2019; Kulik et al., 2016; Rios et al., 2015).

Second, the small numbers of non-Christian religious participants in the present studies makes it difficult to draw meaningful conclusions about such individuals. Nevertheless, some work suggests stereotype threat effects of this kind are unlikely to emerge among religious minorities in the United States (e.g., Jews, Muslims), both because negative stereotypes about religious individuals in science are targeted at Christians (Rios et al., 2015) and because the religion–science conflict is considered less relevant to Judaism and Islam than to Christianity (Rios & Aveyard, 2019; Vaidyanathan et al., 2016). Examining whether and how stereotype threat operates both among religious minorities in the United States and in countries with non-Christian religious majority groups would be a worthy direction for future research. One possibility is that non-Christian religious groups in the United States do not experience stereotype threat about their scientific abilities, but instead experience a more general social identity threat related to their sense of belonging in American society (see Pasek & Cook, 2019).

Finally, as noted above, the reverse mediation in Studies 1 and 2 (i.e., religious affiliation—performance—subjective feelings of stereotype threat in the incompatible condition) was also significant. Studies 1 and 2 thus do not provide clear evidence that subjective feelings of stereotype threat explain Christians’ scientific underperformance following reminders of the conflict narrative. Instead, it could be that among Christians exposed to stereotypes about their religion conflicting with science, poorer performance causes them to experience heightened feelings of stereotype threat. Even if this alternative is the case, however, the effects in Studies 1 and 2 are still informative because they show that Christians, rather than simply not being invested in science, actually are concerned with being negatively evaluated based on their group’s scientific abilities. Whether these concerns come before or after scientific task performance, the fact that they emerge at all suggests that identifying as Christian and caring about science need not be mutually exclusive. Indeed, much extant research has focused on Christians’ presumed lower scientific interest and literacy (relative to non-Christians’ or non-religious individuals’), rather than on Christians’ subjective/affective responses to stereotypes about their group (e.g., McPhetres & Zuckerman, 2018; Sherkat, 2011).

Conclusion

In the face of widespread beliefs about the religion–science conflict in American society (Pew Research Center, 2015b), it may be tempting to conclude that these studies present a bleak picture of Christians’ performance and engagement in science. Encouragingly, however, the findings show that Christians perform just as well as non-Christians on tasks described as assessing scientific reasoning ability when induced to believe that most others perceive Christianity and science as compatible. Thus, one implication of the present findings is that although religion–science conflict stereotypes may have the adverse effect of deterring Christians from scientific fields, thinking about how religion and science can coexist has the potential to bolster Christians’ performance in these fields. For instance, STEM instructors and employers may be well served to highlight religious in addition to other forms of diversity in their retention and recruitment efforts (e.g., providing exemplars of successful religious scientists; see Scheitle & Ecklund, 2017). Doing so has the potential to heighten scientific interest, performance, and literacy among a large majority of the American population.

Supplemental Material

Supplemental Material, Rios_Online_Appendix - Examining Christians’ Reactions to Reminders of Religion–Science Conflict: Stereotype Threat versus Disengagement

Supplemental Material, Rios_Online_Appendix for Examining Christians’ Reactions to Reminders of Religion–Science Conflict: Stereotype Threat versus Disengagement by Kimberly Rios in Personality and Social Psychology Bulletin

Footnotes

Acknowledgment

The author thanks Cameron Mackey and Zhen Hadassah Cheng for providing feedback on an earlier version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Funding statement should read: This research was supported in part by the John Templeton Foundation (Award #13397). Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the John Templeton Foundation.

ORCID iD

Kimberly Rios

Supplemental Material

Supplemental material is available online with this article.

Notes

References

Alquist

J. L.

Price

M. M.

Hancock

Talley

A. E.

Cukrowicz

(2019). Exposure to negative stereotypes impairs older adults’ self-control. Self and Identity, 18(2), 217–226.

Altemeyer

Hunsberger

(2004). A revised religious fundamentalism scale: The short and sweet of it. The International Journal for the Psychology of Religion, 14(1), 47–54.

Aronson

Lustina

M. J.

Good

Keough

Steele

C. M.

Brown

(2001). When White men can’t do math: Necessary and sufficient factors in stereotype threat. Journal of Experimental Social Psychology, 35(1), 29–46.

Cheryan

Plaut

V. C.

Davies

P. G.

Steele

C. M.

(2009). Ambient belonging: How stereotypical cues impact gender participation in computer science. Journal of Personality and Social Psychology, 97(6), 1045–1060.

Crandall

C. S.

Eshleman

O’Brien

(2002). Social norms and the expression and suppression of prejudice: The struggle for internalization. Journal of Personality and Social Psychology, 82(3), 359–378.

Ecklund

E. H.

Johnson

D. R.

Scheitle

C. P.

Matthews

K. R.

Lewis

S. W.

(2016). Religion among scientists in international context: A new study of scientists in eight regions. Socius, 2, 1–9.

Faul

Erdfelder

Buchner

Lang

A. G.

(2009). Statistical power analyses using G* Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160.

Fiske

S. T.

Cuddy

A. J. C.

Glick

(2002). A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82(6), 878–902.

Flore

P. C.

Mulder

Wicherts

J. M.

(2018). The influence of gender stereotype threat on mathematics test scores of Dutch high school students: A registered report. Comprehensive Results in Social Psychology, 3(2), 140–174.

10.

Flore

P. C.

Wicherts

J. M.

(2015). Does stereotype threat influence performance of girls in stereotyped domains? A meta-analysis. Journal of School Psychology, 53(1), 25–44.

11.

Gallup. (2017, 22 December). 2017 update on Americans and religion. Gallup. https://news.gallup.com/poll/224642/2017-update-americans-religion.aspx

12.

Gervais

W. M.

van Elk

Xygalatas

McKay

Aveyard

Buchtel

E. E.

Dar-Nimrod

Klocova

E. K.

Ramsay

J. E.

Riekki

Svedholm-Hakkinen

A. M.

Bulbulia

(2018). Analytic atheism: A cross-culturally weak and fickle phenomenon? Judgment & Decision Making, 13(3), 268–274.

13.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. Guilford Press.

14.

Jussim

Crawford

J. T.

Anglin

S. M.

Stevens

S. T.

Duarte

J. L.

(2016). Interpretations and methods: Towards a more effectively self-correcting social psychology. Journal of Experimental Social Psychology, 66, 116–133.

15.

Kulik

C. T.

Perera

Cregan

(2016). Engage me: The mature-age worker and stereotype threat. Academy of Management Journal, 59(6), 2132–2156.

16.

Markman

K. D.

Lindberg

M. J.

Kray

L. J.

Galinsky

A. D.

(2007). Implications of counterfactual structure for creative generation and analytical problem solving. Personality and Social Psychology Bulletin, 33(3), 312–324.

17.

Marx

D. M.

Goff

P. A.

(2005). Clearing the air: The effect of experimenter race on target’s test performance and subjective experience. British Journal of Social Psychology, 44(4), 645–657.

18.

McPhetres

Zuckerman

(2018). Religiosity predicts negative attitudes towards science and lower levels of science literacy. PLOS ONE, 13(11), Article e0207125.

19.

Oyserman

Fryberg

S. A.

Yoder

(2007). Identity-based motivation and health. Journal of Personality and Social Psychology, 93(6), 1011–1027.

20.

Pasek

M. H.

Cook

J. E.

(2019). Religion from the target’s perspective: A portrait of religious threat and its consequences in the United States. Social Psychological and Personality Science, 10(1), 82–93.

21.

Pennington

C. R.

Heim

Levy

A. R.

Larkin

D. T.

(2016). Twenty years of stereotype threat research: A review of psychological mediators. PLOS ONE, 11(1), Article e0146487.

22.

Pennycook

Ross

R. M.

Koehler

D. J.

Fugelsang

J. A.

(2016). Atheists and agnostics are more reflective than religious believers: Four empirical studies and a meta-analysis. PLOS ONE, 11(4), Article e0153039.

23.

Pew Research Center. (2015a, 4 November). Americans’ faith in God may be eroding. http://www.pewresearch.org/fact-tank/2015/11/04/americans-faith-in-god-may-be-eroding/

24.

Pew Research Center. (2015b, 22 October). Religion and science: Highly religious Americans are less likely than others to see conflict between faith and science. http://www.pewinternet.org/2015/10/22/science-and-religion/

25.

Pew Research Center. (2019). How much do you know about science topics? https://www.pewresearch.org/science/quiz/science-knowledge-quiz/

26.

Preston

Epley

(2005). Explanations versus applications: The explanatory power of valuable beliefs. Psychological Science, 16(10), 826–832.

27.

Preston

Epley

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

28.

Rios

Aveyard

(2019). Science-religion compatibility beliefs across Middle Eastern and American young adult samples: The role of cross-cultural exposure. Public Understanding of Science, 28(8), 949–957.

29.

Rios

Cheng

Z. H.

Totton

R. R.

Shariff

A. F.

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

30.

Rios

DeMarree

K. G.

Statzer

(2014). Attitude certainty and conflict style: Divergent effects of correctness and clarity. Personality and Social Psychology Bulletin, 40(7), 819–830.

31.

Salancik

G. R.

Conway

(1975). Attitude inferences from salient and relevant cognitive content about behavior. Journal of Personality and Social Psychology, 32(5), 829–840.

32.

Scheitle

C. P.

Ecklund

E. H.

(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.

33.

Scheitle

C. P.

Ecklund

E. H.

(2018). Perceptions of religious discrimination among U.S. scientists. Journal for the Scientific Study of Religion, 57(1), 139–155.

34.

Sherkat

D. E.

(2011). Religion and scientific literacy in the United States. Social Science Quarterly, 92(5), 1134–1150.

35.

Spencer

S. J.

Steele

C. M.

Quinn

D. M.

(1999). Stereotype threat and women’s math performance. Journal of Experimental Social Psychology, 35(1), 4–28.

36.

Spencer

S. J.

Zanna

M. P.

Fong

G. T.

(2005). Establishing a causal chain: Why experiments are often more effective than mediational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89(6), 845–851.

37.

Steele

C. M.

Aronson

(1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69(5), 797–811.

38.

Steele

C. M.

Spencer

S. J.

Aronson

(2002). Contending with group image: The psychology of stereotype and social identity threat. In Zanna

M. P.

(Ed.), Advances in experimental social psychology (Vol. 34, pp. 379–440). Academic Press.

39.

Stone

Lynch

C. I.

Sjomeling

Darley

J. M.

(1999). Stereotype threat effects on Black and White athletic performance. Journal of Personality and Social Psychology, 77(6), 1213–1227.

40.

Vaidyanathan

Johnson

D. R.

Prickett

P. J.

Ecklund

E. H.

(2016). Rejecting the conflict narrative: American Jewish and Muslim views on science and religion. Social Compass, 63(4), 478–496.

41.

Zuckerman

Silberman

Hall

J. A.

(2013). The relation between intelligence and religiosity: A meta-analysis and some proposed explanations. Personality and Social Psychology Review, 17(4), 325–354.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.30 MB

Examining Christians’ Reactions to Reminders of Religion–Science Conflict: Stereotype Threat versus Disengagement

Abstract

Keywords

Stereotypes of Religion–Science Incompatibility

Stereotype Threat versus Disengagement

Overview of Studies

Study 1 1

Method

Participants

Procedure and materials

Results

Performance

Subjective feelings of stereotype threat

Time spent on task

Mediated moderation analysis

Discussion

Study 2

Method

Participants

Procedure and materials

Results

Performance

Subjective feelings of stereotype threat

Time spent on task

Mediated moderation analysis

Manipulation check

Discussion

Study 3

Method

Participants

Procedure and materials

Results

Performance

Time spent on task

Manipulation check

Discussion

Study 4

Method

Participants

Procedure and materials

Results

Performance

Time spent on task

Agreement with items

Discussion

General Discussion

Limitations and Future Directions

Conclusion

Supplemental Material

Supplemental Material, Rios_Online_Appendix - Examining Christians’ Reactions to Reminders of Religion–Science Conflict: Stereotype Threat versus Disengagement

Footnotes

Acknowledgment

Declaration of Conflicting Interests

Funding

ORCID iD

Supplemental Material

Notes

References

Supplementary Material

Study 1¹