It’s not my consensus: Motivated reasoning and the sources of scientific illiteracy

Abstract

Individuals who provide incorrect answers to scientific knowledge questions have long been considered scientifically illiterate. Yet, increasing evidence suggests that motivated reasoning, rather than ignorance, may explain many of these incorrect answers. This article uses novel survey measures to assess two processes by which motivated reasoning might lead to incorrect personal beliefs: motivated individuals may fail to identify the presence of a scientific consensus on some issue or they may recognize a consensus while questioning its veracity. Simultaneously looking at perceptions of what most scientists say and personal beliefs, this study reveals that religiosity and partisanship moderate the extent to which Americans identify scientific consensuses and assert beliefs that contradict their perceptions of consensus. Although these pathways predict the scope of disagreement with science for each of 11 issues, the relative prevalence of each process depends on both the scientific issue and motivational pathway under examination.

Keywords

motivated reasoning rejectionism scientific consensus scientific literacy

A basic level of scientific literacy has long been viewed as critical for a functioning democratic society (see Snow et al., 2016). This is particularly true in an era when issues like climate change and vaccine safety dominate newspaper headlines and Congressional hearings. Yet, numerous studies reveal that many Americans report beliefs that are inconsistent with foundational scientific knowledge. In the 2012 General Social Survey, scarcely half of Americans knew that it takes 1 year for the earth to go around the sun or that electrons are smaller than atoms (National Science Board, 2014). These data have often been used to argue that the American public is scientifically uninformed.

Recently, scholars have noted that some Americans who incorrectly answer factual questions about science are not necessarily ignorant. Rather, many more people can identify what scientists say than report corresponding personal beliefs (Kahan, 2015; National Science Board, 2014). Presumably, many individuals who assert that humans did not evolve or that there was no big bang nonetheless recognize scientific status of these theories. In Science and Engineering Indicators 2010, discrepancies between people’s identification of the scientific consensus and their personal beliefs were regarded as sufficient cause for removing questions about evolution and the big bang from the science knowledge index (National Science Board, 2014). But it is unclear what leads people to answer in ways that dismiss their knowledge of scientific consensus. Are these discrepancies a type of ignorance or might there be knowledgeable disbelievers (Roos, 2014)?

This study aims to better understand the rejection of science by (1) examining which individual attributes were associated with a willingness to disagree with scientists, (2) assessing how willingness to disagree with scientists differs from misperceptions about the nature of scientific consensus, and (3) assessing whether willingness to disagree with the scientific consensus depends on the domain of science being considered.

1. Scientific literacy

The centrality of science to modern society is difficult to overstate. Many principal social gains in the last two centuries have been scientific in nature. Scientific advances have skirted Malthusian population bottlenecks by increasing the efficiency of agriculture (Kögel and Prskawetz, 2001), extended human life spans through vaccinations (Bonanni, 1999) and reductions in infant mortality (Singh and Yu, 1995), and enabled nearly instantaneous communication and transportation across our globe (Standage, 2014). Although many individuals do not use their understanding of science on a daily basis, scientific literacy is important if citizens are to participate in a society so reliant on scientifically driven technological developments (Feinstein, 2010).

Furthermore, looming challenges facing both the United States and the world depend on a sophisticated understanding of science. These range from policies about how homosexuality should be regarded—is sexual preference innate or “a choice”—to mandatory vaccinations and measures to combat climate change. There is little doubt that scientifically relevant policies and scientific funding are based, at least in part, on public perceptions (cf. Etchegary et al., 2010; Gauchat, 2011; Weingart, 1999). Hence, public awareness of dominant scientific theories should influence the extent to which scientific research in these areas is incorporated into public deliberation and, subsequently, policy.

Popular scientific literacy has also been purported to accrue economic benefits to scientifically innovative societies, personal benefits to knowledgeable individuals, democratic benefits when deliberation is informed by science, and cultural benefits as they inform the popular understanding of the world (Snow et al., 2016). And though the jury remains out on the extent to which science knowledge elicits support for research funding (Allum et al., 2008), scientific progress may itself depend on public understanding.

For all these reasons, researchers have long been concerned with the state of scientific awareness (Durant et al., 1989; Withey, 1959). In a series of studies, Miller (1983, 1998, 2004) found that less than one-third of Americans appeared to possess the information necessary to read the Science section of the New York Times. And although questions remain about what constitutes “sufficient” scientific knowledge and whether knowing about science is really the key to supporting it (Roberts et al., 2013; Stocklmayer and Bryant, 2011; Sturgis and Allum, 2004), researchers largely agree that some level of public knowledge is important (Evans and Durant, 1999; Miller, 2004; Osborne et al., 2003).

2. Rejection of the scientific consensus

Increasingly, research on knowledge of basic scientific facts has revealed systematic differences between individuals who accept particular scientific conclusions and those who reject them (Lewandowsky et al., 2013b; Solomon, 1993). In particular, certain forms of religiosity have been linked to reduced propensity to believe in evolution and the big bang, and partisanship has been proposed as an important motivation for accepting evidence about climate change (Sherkat, 2011). These findings have led to the claim that questions about evolution and the big bang may “measure a specific conservative religious worldview rather than science knowledge” (Roos, 2014: 808). Similarly, researchers have found that American conservatives are more resistant to information about risks of climate change than liberals (Dunlap and McCright, 2008; Hart and Nisbet, 2012; Kahan et al., 2012; Nisbet et al., 2015).

Scholars have interpreted the rejection of scientific claims by members of particular groups as evidence of motivated reasoning. Individuals who learn that scientific findings challenge preexisting beliefs/affiliations would be expected to encounter cognitive dissonance (Festinger, 1957). Whether these beliefs/affiliations are religious or partisan in nature, individuals are then motivated to avoid incorporating evidence that contradicts preexisting views into their understandings of the world (cf. Kunda, 1990; Lodge and Taber, 2013). Lodge and Taber (2013) contend that individuals who encounter such information activate their prior attitudes and beliefs and experience a negative affect toward the new piece of information. Through a combination of implicit biases and counterargument, new information is rejected and thereby not integrated into respondents’ beliefs. And group affiliations appear to activate similar processes (Kahan et al., 2011).¹

3. Two motivational pathways

Although a considerable literature indicates that partisanship and religiosity moderate the acceptance of scientific beliefs in the United States (Kahan, 2015), motivational processes present two possible pathways for maintaining scientifically unsupported views. When individuals are presented with evidence of a scientific consensus, they could resolve dissonance either by rejecting information asserting the consensus (Kahan et al., 2011) or by downplaying the relevance of the consensus as a basis for their own beliefs (cf. Ellison and Musick, 1995). In public debates, religious proponents of Intelligent Design often employ the first of these strategies, propagating the view that evolutionary science is unsettled (Jones, 2005). And political opponents of climate change challenge the conclusions of the Intergovernmental Panel on Climate Change (IPCC) by deriding its evidence as a “hoax” (Inhofe, 2012). Both approaches allow individuals to retain preexisting beliefs in the face of seemingly contradictory evidence.

The difference between denying the presence of a scientific consensus and accepting its existence while disagreeing with it has not been well explored in previous literature. Among the few who have addressed it, Lewandowsky et al. (2012a, 2013a) have argued that there is a close connection between personal beliefs and perceptions of scientific consensus. And this model seems to underlie the bulk of research to date (cf. Sturgis and Allum, 2004). Those who criticize the traditional measurement paradigm, on the other hand, have noted that motivational forces can introduce a rift between scientific perceptions and personal beliefs, but have not explored whether the correlates of that rift differ from those that predict misperceptions in the first place (Kahan, 2015; Roos, 2014). Yet, evidence of heterogeneity in the effects of scientific information across individuals implies that such a distinction may aid explanation (Hobson and Niemeyer, 2013).

If we hope to improve public scientific literacy, it is important to disentangle these two motivational processes. When motivated individuals fail to recognize the views of scientists, information about the nature and prevalence of a scientific consensus may shift public opinion (Lewandowsky et al., 2012b); if these individuals instead find the presence of scientific consensus unpersuasive, their beliefs are more intractable (Ecker et al., 2011; Lewandowsky et al., 2012a). And since both a motivated lack of awareness and the willful disavowal of scientific evidence likely explain some partisan and religious gaps in scientific literacy, the ideal remedy may depend on the relative prevalence of these processes.

Notably, if motivations moderate the extent to which individuals accept scientific consensuses and incorporate them into their own beliefs, we should expect that both identification of the scientific consensus and acceptance of that consensus would depend on the particular issues under examination. Among commonly used knowledge measures, some questions—such as those about continental drift and evolution—should be sensitive to religious motivations (Roos, 2014), others—such as measures of climate change perceptions—should be sensitive to political motivations (Hart and Nisbet, 2012; Kahan, 2015; Lewandowsky et al., 2013a), and yet others—such as the relative sizes of electrons and atoms—should be immune to these motivational factors. This reasoning implies that we should see heterogeneity in predictors of both identification of the scientific consensus and respondents’ acceptance of those beliefs.

4. The current study

This study examines the extent to which respondents identify the scientific consensus, personally believe in the scientific consensus, and are willing to disagree with what they perceive to be the scientific consensus across scientific issues. In line with the theory of motivated reasoning, religiosity and partisanship are expected to uniquely predict all three outcomes. Furthermore, motivational processes are projected to be concentrated in domains where religiosity or partisanship might lead to dissonance between respondents’ views and those of the scientific consensus. That is, we expect Republican partisans to hold views contradicting the scientific consensus when queried on climate change and religious individuals to hold contrarian views on evolution, the big bang, human–dinosaur coexistence, and—to a lesser extent—plate tectonics (which depends on old-earth timescales).

How those motivational processes play out, however, should depend on the relative ease with which respondents can either reject the presence of a scientific consensus or disavow its relevance. For instance, it is presumably easier to ignore the scientific consensus about human–dinosaur coexistence than about evolution, as the latter is a more consistent component of school curricula (Mead and Mates, 2009). Hence, although more religious individuals are expected to hold less accurate views on both measures, the first should elicit misperceptions about the scientific consensus, whereas the latter should yield rejectionism. Disaggregating knowledge about science should therefore reveal that both predictors operate in some domains but not others. Although motivation should result in both misperceptions and rejectionism, the extent to which each process occurs for any given issue is a research question.

5. Methods

Data

Data for this study come from a four-wave panel of a broad national sample of Americans. Data were collected by Qualtrics using respondents sourced from ClearVoice surveys. Individuals who completed each wave were subsequently invited to additional waves. Attrition was built into the design, and the final sample had 3729, 2454, 2267, and 1046 complete responses for the waves, respectively. Wave 1 was conducted between August 29 and September 8, 2014; wave 2 was conducted between September 26 and October 2, 2014; wave 3 was conducted between October 28 and November 3, 2014; and wave 4 was conducted between November 14 and 19, 2014.

Because the panel was not a probability sample, results are presented across all respondents but were not weighted to match the population.² In general, the sample obtained was considerably more Female and White than the public; it was also slightly more educated, somewhat older, and a bit more liberal. Despite these differences, the sample contained a moderate number of individuals across all demographic groups. Question wordings and sample demographics can be found in online Supplementary Appendix A, additional information on sampling as well as a comparison to benchmark data can be found in online Supplementary Appendix B.

Outcomes

Half of respondents in each wave were asked a series of questions about scientific knowledge.³ For each of 10 issues, respondents answered dichotomous forced choice questions about their perceptions of what most scientists would say about a particular matter followed by their personal beliefs. To avoid “don’t know” options, respondents were asked how certain they were about answers to each question. In addition, respondents were randomly assigned either a question about whether the earth was warming or whether humans were warming the earth. Questions were selected from Miller (1998) and are presented in Table 1. Full wording and a discussion of design decisions can be found in online Supplementary Appendix A.

Table 1.

Distribution of personal beliefs, consensus identification, and disagreement by topic.

Variable	Full wording	Scientific consensus	Personal beliefs, %	Consensus identification, %	Mean difference, %	Total disagreement, %	N
Humans share DNA with mice	That more than half of human genes are identical to those of mice	TRUE	51	53	2	13	3130
Universe began with explosion	That the universe began with a huge explosion	TRUE	54	75	21***	28	3130
Humans lived with dinosaurs	That the earliest humans lived at the same time as dinosaurs	FALSE	56	61	5***	19	3130
Humans evolved	That human beings, as we know them today, developed from earlier species of animals	TRUE	56	80	24***	30	3130
Sound faster than light	That sound travels faster than light	FALSE	59	64	6***	13	3130
Electrons smaller than atoms	That electrons are smaller than atoms	TRUE	67	67	0	15	3130
Humans warming the earth	That humans have been causing the earth’s climate to warm in recent years	TRUE	68	77	9***	19	1603
GMOs modify your genes	That eating a genetically modified fruit could lead to a person’s own genes being modified	FALSE	69	75	6***	19	3130
Earth is Warming	That the earth’s climate has been warming in recent years	TRUE	75	83	8***	16	1527
Earth’s plates moving	That the continents on which we live have been moving their locations for millions of years	TRUE	80	85	5***	11	3130
Plants produce oxygen	That plants produce the oxygen that we breathe	TRUE	89	90	1	6	3130
Overall	–	–	65	73	8***	17	31300

GMO: genetically modified organism.

Each respondent received only one of the two climate change measures. Averages were calculated at the respondent level, and thus differed slightly depending on which of these measures the respondent received.

***

p < .001 two-tailed. Exactly 240 individuals switched their answers in each direction between the identify consensus and respondent belief measures.

Responses to these questions were used to generate three dummy-coded measures. A Consensus Identification measure indicated whether respondents correctly identified the scientific consensus. A Personal Beliefs measure indicated whether respondents’ beliefs mirrored the true scientific consensus. And a Disagreement measure indicated whether the Consensus Identification and Personal Belief items differed from one-another.

Predictors

Religiosity

A measure of the sort of religiosity that might motivate scientific rejectionism was created by averaging responses to six questions about the importance of religion, religious attendance, closeness to their religious community, closeness to others that share their religion, status as born again, and biblical literalism. The variable was coded to range from 0 for the least religious answers, to 1 for the most religious (α = .86, M = 0.39, SD = 0.28, 960 cases missing).⁴

Party identification

Respondents in waves 1 and 3 were asked about their political partisanship. A branched set of questions was recoded to produce a 7-category scale ranging from “Strong Republican” (0) to “Strong Democrat” (1). To deal with possible nonlinearities in this scale, linear and quadratic versions of the scale were used in all analyses (M = 0.54, SD = 0.33, 0 cases missing).

Covariates

Education, sex, age, race, marital status, and income were also included in all regressions to ensure that differences observed were not due to demographic variability across levels of religiosity or partisan orientation. Full wording and descriptive statistics are in online Supplementary Appendix A.

Missing data

Because the study consisted of multiple waves with planned missingness and included different measures across waves, ignoring missing data would drastically reduce the sample size compared to the information available in the data set; doing so might also bias results toward the individuals who were least likely to suffer from attrition. Hence, all analyses were produced using multiple imputation. Ten data sets with all predictors and outcomes were imputed using multiple imputation via chained equations (mice) and analyses were run separately across these data sets and pooled to produce the results shown. Imputations were conducted for all individuals who answered the science information battery at least once and had nonmissing values for at least one predictor (n = 3,130).⁵ Additional information is provided in online Supplementary Appendix C.⁶

6. Results

Distributions

The traditional measure of scientific literacy examines whether individuals hold beliefs in line with the true scientific consensus (Miller, 1998). Across all 11 measures, a majority of respondents reported personal beliefs in line with most scientists (Table 1, column 2). The strength of these majorities varied widely: 89% of respondents correctly asserted that plants produce the oxygen we breathe, whereas merely 51% recognized that humans share most genes with mice, a number that did not significantly differ from chance (p = .37). On average, 65% of personal beliefs echoed corresponding scientific conclusions.

As others have noted, the presence of unscientific personal beliefs does not imply that respondents are ignorant of the scientific consensus (Roos, 2014). For all but two measures, respondents’ personal beliefs tended to be further from the true consensus than respondents’ perceptions of most scientists’ beliefs (Table 1, column 3). Overall, respondents answered 73% of the scientific majority questions correctly, performing an average of 7.8 percentage points better on these measures than on the personal belief items (p < .001 difference). The scale of these differences varied widely. Identification of the scientific consensus was more than 20 percentage points higher than the personal belief measure when respondents were asked about evolution or the big bang (Table 1, column 4). For whether plants produce oxygen, however, personal beliefs almost perfectly matched perceptions of scientific consensus (a difference of 0.9 percentage points, t = 1.1, p = .025). This implies that some responses to personal beliefs measures came from individuals who knew the consensus but rejected it.

Respondents reported personal beliefs that differed from their perceptions of the scientific majority 17% of the time (Table 1, column 6). That this number was larger than the net difference between perceptions of scientists’ beliefs and personal beliefs is to be expected, as this includes cases where respondents mischaracterized scientists’ beliefs and nonetheless answered correctly themselves. For 5% of questions, respondents disagreed with a misidentified scientific consensus; this occurred about half as frequently as respondents correctly characterizing the scientific majority and reporting contradictory personal beliefs (12%). Disagreement with perceptions of the scientific consensus also varied across questions, ranging from 6% for the measure about plants producing oxygen to 30% for the evolution question. Hence, respondents were more prone to disagreeing with scientists on some items than others.

Most respondents disagreed with scientists at least some of the time. Only 32% of respondents did not diverge from their perceptions of the scientific majority on any question. But when respondents disagreed with scientists, it was typically on only a few measures. Around half of respondents diverged from what they thought was the scientific majority two or more times (46%) and only 8% of respondents disagreed for five or more questions of the ten. Again, disagreement appears to have been selective rather than universal.

Motivations

To assess how religiosity and partisanship related to scientific beliefs, 11 logistic regressions—one for each issue—predicted each of three outcomes: (1) whether respondents’ beliefs matched the true scientific consensus, (2) whether respondents correctly identified the true scientific consensus, and (3) whether respondents’ beliefs differed from their perceptions of the scientific consensus. Each regression included the religiosity measure, both a linear and a quadratic version of the partisanship scale, and demographic covariates. Table 2 presents logistic regression coefficients for the religiosity measure in predicting each outcome and pooled likelihood tests comparing the inclusion of both partisanship measures in each regression to identical regressions lacking those predictors. Full results are given in online Supplementary Appendix E. Replications using only respondents who were at least moderately certain of their answers, treating religiosity for Christians and non-Christians separately, using liberal–conservative ideology instead of partisanship, and listwise deletion instead of multiple imputation provided substantively identical results and are presented in online Supplementary Appendices F through I, respectively.⁷

Table 2.

Religiosity coefficients and pooled likelihood tests for partisanship from logistic regressions predicting outcomes for each question.

	Religiosity regression coefficients						Partisanship pooled likelihood tests
	Personal beliefs		Consensus identification		Disagreement		Personal beliefs		Consensus identification		Disagreement
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	df	Coef.	df	Coef.	df
Humans share DNA with mice	−0.79***	.16	−0.54**	.17	1.02***	.24	3.0	2	2.7	2	1.9	2
Universe began with explosion	−2.67***	.20	−0.67***	.18	2.42***	.22	14.6***	2	2.0	2	15.5***	2
Humans lived with dinosaurs	−1.01***	.19	−1.49***	.18	1.28***	.19	.1	2	.4	2	2.5	2
Humans evolved	−3.54***	.22	−1.22***	.23	2.77***	.23	31.4***	2	1.1	2	22.9***	2
Sound faster than light	−0.69***	.19	−1.15***	.20	−.04	.23	4.6*	2	6.7**	2	.8	2
Electrons smaller than atoms	0.13	.18	0.19	.21	.02	.22	.5	2	1.9	2	.0	2
Humans warming the earth	−0.81**	.25	−0.10	.26	1.22***	.32	60.4***	2	27.1***	2	30.1***	2
GMOs modify your genes	−0.93***	.21	−1.05***	.18	.51*	.22	2.4	2	1.4	2	.4	2
Earth is warming	−0.77**	.29	−0.70*	.33	.61	.34	41.0***	2	25.6***	2	14.5***	2
Earth’s plates moving	−1.52***	.22	−0.80***	.22	2.14***	.31	5.7**	2	.5	2	3.2*	2
Plants produce oxygen	−0.35	.27	−0.07	.28	.49	.34	.0	2	.5	2	.0	2

GMO: genetically modified organism.

Pooled likelihood tests consider the joint inclusion of linear and quadratic 7-point partisanship scales. They differ from traditional likelihood ratio tests to conform to multiply imputed data.

p < .05; **p < .01; ***p < .001 two-tailed. Full regressions in online Supplementary Appendix E.

Religiosity

Religiosity was related to personal beliefs about 8 of the 11 scientific questions when demographics and partisanship were controlled. Compared to the least religious respondents, more religious individuals were significantly less likely to hold personal beliefs in line with the true scientific consensuses about humans and mice-sharing genes, the big bang, evolution, the relative speeds of sound and light, global warming, genetically modified foods, and plate tectonics (Table 2, column 1).

Correspondence between religiosity and personal beliefs can be illustrated clearly by the case of evolution, which was the most closely related to religiosity (b = −3.54, SE = .22, p < .001; Table 2). The solid line in Figure 1d shows how the likelihood that a typical individual⁸ would report that humans evolved from earlier species of animals depended on that individual’s religiosity. Holding all covariates constant, a typical individual who scored a zero on the religiosity scale would be expected to have an 81% chance of asserting that humans evolved; that same individual would be projected to have only an 11% chance of believing the true scientific consensus at the highest observed level of religiosity.

Figure 1.

Relations between religiosity measures and scientific belief outcomes. Lines present predicted personal beliefs (solid lines), consensus identification (dashed lines), and disagreement with perceived scientific consensus (dotted lines) by religiosity, controlling for demographics and partisanship. 95% confidence intervals shown.

An otherwise typical individual at the highest observed level of religiosity was predicted to believe the true scientific consensus less often than would be expected by chance on questions about the humans and mice-sharing genes, the big bang, human–dinosaur coexistence, and evolution and not significantly differently from chance for the question about whether sound is faster than light (see Figure 1). That same individual, at the lowest levels of religiosity, would be predicted to hold personal beliefs in line with the scientific consensus for all questions at rates greater than chance.

Part of the belief gap for religiosity appears to stem from a failure to identify the scientific consensus. Compared to less religious respondents, more religious individuals were less likely to know the consensus on all but three of the scientific matters examined (Table 2, column 2). Projected consensus identification on evolution, for instance, dropped from 87% for a typical respondent at the lowest level of religiosity to 66% at the highest level of religiosity (dashed line in Figure 1d; b = −1.22, SE = .23, p < .001). The absolute level of awareness was much higher using consensus identification measures than personal beliefs measures. Hence, although more religious individuals were often less likely to correctly identify a consensus than the least religious individuals, even the most religious respondents never performed significantly worse than chance on these measures (Figure 1).

Willingness to disagree with scientists also appears to explain some of respondents’ inaccurate beliefs. More religious individuals were more likely than less religious individuals to express beliefs contradicting their perceptions of scientific consensus for questions about sharing genes with mice, the big bang, human–dinosaur coexistence, evolution, anthropogenic warming, genetically modified organisms (GMOs), and plate tectonics (Table 2, column 3). For evolution, a typical individual at the lowest levels of religiosity would be expected to have a 14% chance of providing personal beliefs that contradicted their perceptions of what most scientists would say; for the most religious individuals, this was fully 72% (dotted line in Figure 1d; b = 2.77, SE = .23, p < .001).

Partisanship

Like religiosity, partisanship accounted for variations in personal beliefs for six issues when other variables were controlled. The right side of Table 2 shows pooled likelihood tests assessing differences in variance explained by a model with all predictors compared to one where both linear and quadratic partisanship measures were removed. Partisanship helped to explain personal beliefs about the big bang, evolution, the relative speeds of sound and light, global warming, and plate tectonics (Table 2, column 4). For the scientific issues where partisanship explained variations in personal beliefs, the strongest Republicans typically tended to hold views that most contradicted the true scientific consensus, with the sole exception of the speeds of sound and light (solid lines in Figure 2).

Figure 2.

Relations between partisanship measures and scientific belief outcomes. Lines present predicted personal beliefs (solid lines), consensus identification (dashed lines), and disagreement with perceived scientific consensus (dotted lines) by partisanship, controlling for demographics and religiosity. 95% confidence intervals shown.

The strongest variation in personal beliefs attributable to partisanship was for perceptions of anthropogenic climate change (pooled likelihood = 60.4 (2), p < .001; Table 2, column 4). Moving from strong Republicans to strong Democrats, belief in anthropogenic climate change increased monotonically (linear b = 1.84, SE = .64, p = .004; quadratic b = .18, SE = .62, p = .77). The solid line in Figure 2g shows that a typical respondent who identified as a strong Republican had a 46% chance of believing in anthropogenic warming, whereas an otherwise identical strong Democrat had an 86% chance of holding these beliefs.

Partisanship was associated with consensus identification less frequently than for personal beliefs. Partisan differences were observed for perceptions of the relative speeds of sound and light and both climate change measures (Table 2, column 5). Here, even the most partisan Republicans overwhelmingly stated that anthropogenic warming constituted the scientific consensus; strong Republican status led a typical respondent to have a 72% chance of correct identification (dashed line in Figure 2g). This was still significantly lower than consensus identification among strong Democrats; a typical respondent with strong Democrat status was projected to have a 91% chance of correct identification (pooled likelihood = 25.6 (2), p < .001).

Respondents’ willingness to disagree with scientists also explained some of the correspondence between partisanship and personal beliefs. Partisanship moderated the extent to which individuals expressed personal beliefs that differed from their perceptions of the scientific consensus for the big bang, evolution, climate change, and plate tectonics (Table 2 column 6); and Republicans were more willing to express disagreement for all these issues (Figure 2). For anthropogenic warming, projected disagreement ranged from a high of 36% for strong Republicans to a mere 10% among strong Democrats (pooled likelihood = 30.1 (2), p < .001).

Issues and motivational processes

Collectively, the influence of motivational processes on personal beliefs can be seen as a combination of the impact of those motivations on both consensus identification and willingness to disagree with scientists. For many issues, these motivational processes appear to operate in tandem. Religiously motivated disbelief over whether humans share DNA with mice, the big bang, human–dinosaur coexistence, evolution, GMOs, and plate tectonics corresponded with both processes. The most religious individuals were less likely to identify scientific consensuses and more prone to disagreeing with them than the least religious (Figure 1; Table 2). Similarly, illiteracy about climate change among partisans appears to stem from both a failure to identify the consensus and a willingness to disagree with it (Figure 2; Table 2). In these cases, motivated reasoning seems to be happening for every step of the process.

For some issues, however, only one of the processes appears to explain a discrepancy in personal beliefs. Compared to nonreligious individuals, religious respondents were not significantly less likely to identify the scientific consensus that humans are warming the earth; they were, however, much more willing to disagree with scientists on this matter (Figure 1g). In contrast, more religious individuals were less likely to identify the scientific consensuses on the speeds of sound and light, and that the earth is warming even though they were no more likely to disagree with scientists on these matters (Figure 1).

Similar discrepancies were observed across levels of partisanship. For questions about the big bang, evolution, and plate tectonics, Republicans were no more likely than Democrats to misidentify the scientific consensus, though they were more likely to disagree with the perceived consensus (Figure 2). In contrast, partisans differed in consensus identification but not disagreement on whether sound travels faster than light (Figure 2e). Thus, the presence of a motivator influencing the scope of consensus identification did not necessarily imply that the same motivator would operate for disagreement and vice versa.

Indeed, across topics, correlations between belief in the scientific consensus and awareness of the consensus ranged from as high as .73 for the belief that humans and mice shared DNA to as low as .41 for evolution. Correspondence between personal beliefs and disagreement with a known consensus ranged from −.08 for humans and mice-sharing DNA to as large as −.61 for evolution (Table 3). Despite the sometimes high correlations between personal beliefs and scientific consensus items, the predictors of these often differed, suggesting that they were quite distinct (see online Supplementary Appendix J).

Table 3.

Pearson’s correlations between personal beliefs, consensus identification, and disagreement by topic.

	Personal beliefs with correct identification	Personal beliefs with disagreement	Consensus identification with disagreement
Humans share DNA with mice	.73	−.08	.05
Universe began with explosion	.47	−.52	.18
Humans lived with dinosaurs	.61	−.19	.02
Humans evolved	.41	−.61	.16
Sound faster than light	.72	−.24	.06
Electrons smaller than atoms	.66	−.15	−.15
Humans warming the earth	.55	−.43	−.03
GMOs modify your genes	.54	−.36	−.10
Earth is warming	.53	−.50	−.09
Earth’s plates moving	.63	−.45	−.13
Plants produce oxygen	.69	−.37	−.26

GMO: genetically modified organism.

7. Discussion

This study represents an exploration of how religiosity and partisanship relate to variations in individuals’ scientific perceptions and beliefs. Overall, the evidence suggests that personal scientific beliefs can be thought of as a function of individuals’ likelihood of identifying the scientific consensus and their propensity to disagree with scientists on an issue. Hence, when religious or partisan identifications motivate individuals to reject the scientific consensus, either or both of these processes may be at work.

As expected, individuals varied considerably in whether their personal beliefs aligned with the true scientific consensus and whether they correctly identified the consensus. For many measures, much of this variation could be explained by religious and partisan motivations. As an example, more religious individuals were the least likely to believe that the universe began with an explosion; and this disbelief is likely a product of direct conflict between deific creation myths and the big bang explanation. Similarly, partisan debates about climate change seem to have induced disbelief about the existence and nature of global warming among Republicans.

Although evidence for religious and partisan motivation is far from novel (Blank and Shaw, 2015), current results suggest that motivations operate through two unique pathways: some motivated individuals appeared ignorant about the state of the scientific consensus and some identified the presence of consensus but maintained contradictory beliefs. For the big bang, both pathways appeared to operate, although rejection of the scientific consensus accounted for most of the personal belief gap among religious individuals. For the question about whether the earth was warming, among Republicans, much of the inaccuracy in personal beliefs appeared to stem from misperceptions about the scientific consensus.

Distinguishing personal beliefs and consensus identification

Most research to date has served to either conflate accurate personal beliefs with identification of the scientific consensus or to ignore this distinction entirely. This conflation undergirded early “deficit model” studies, which presumed that more public information would yield greater public support (Sturgis and Allum, 2004); it also remains a feature of most “deficit model” critiques. In the cultural cognition paradigm, motivated individuals are expected to misperceive expert opinion (due in part to biased determination of who is an expert), but their perceptions of expert attitudes are presumed to largely transfer to personal beliefs and attitudes (Kahan et al., 2011).⁹ Evidence of “backfire” and “boomerang” effects, whereby informational campaigns paradoxically retrench preexisting beliefs (Nyhan and Reifler, 2010; Redlawsk et al., 2010), comes from studies that also largely ignore this distinction (Hart and Nisbet, 2012; Nyhan et al., 2014).

More recently, a few researchers have begun to disentangle personal beliefs from consensus identification; the current study largely replicates their findings. Using between-subjects designs, Roos (2014) and Kahan (2015) illustrated that individuals were more likely to identify the scientific consensus than to hold corresponding personal beliefs; they also noted that the personal beliefs measure was sensitive to religious and partisan views. Lewandowsky et al. (2012a, 2013b) also recognized a close link between perceptions of agreement among scientists and personal beliefs, and Van Der Linden et al. (2015) provided experimental evidence that altering perceptions of consensus could increase personal beliefs. Finally, Blank and Shaw (2015) found that individuals used motivational processes to determine how much policy influence scientists should have when they learn about a scientific consensus.

This study extends this body of evidence in important ways. For one, the current work provides a basis for conceptualizing motivational forces as capable of influencing both consensus identification and the translation from identification to personal beliefs and provides evidence of these on the individual level. It also suggests that the factors leading to both outcomes can be somewhat distinct, with motivational processes sometimes influencing only one of them. And the results indicate, at least for the issues examined herein, that motivations appear to have somewhat more of an influence on the willingness to disagree with scientists than on consensus identification. That motivations are nonetheless related to perceptions of scientific consensus provides support for Kahan’s (2015) suggestion that perceptions of most scientists may not be immune to cultural cognition.

Because both the scientific knowledge and personal belief questions were asked in quick succession to the same individuals, however, it seems unlikely that individuals are reporting on beliefs that they do not realize are related. Instead, individuals appear to be perfectly comfortable asserting that their views differ from those of scientists at least some portion of the time.

Explaining motivational effects

Elite discourse provides one explanation for the differences between issues where individuals fail to identify a consensus and issues where individuals identify and reject it. Claims of “scientific debates” around climate change and evolution—however dubious—present a picture of scientists divided on the issues (Malka et al., 2009). And these have historically been reinforced by norms of presenting equal time for “both sides” of scientific issues (M. T. Boykoff and Boykoff, 2007). It seems likely that the political discourse around particularly contentious scientific matters serves to undermine perceptions of the scientific consensus in addition to individual beliefs (cf. Hart and Nisbet, 2012; Hmielowski et al., 2013; Nisbet et al., 2015). In contrast, issues discussed in a less polarized way in the news may still be rejected by those motivated to disbelieve them, but perhaps only after their detractors accept the presence of a scientific consensus. Additional research is necessary to test these mechanisms.

One intriguing result is the presence of religious and political moderation of scientific disagreement even where there is little reason to theorize these differences (e.g., partisan differences on plate tectonics and religious differences on perceptions about GMOs). To date, the literature implies that trust in scientists may introduce a mechanism for spillover from motivation-relevant issues to scientific beliefs more generally. Lewandowsky et al. (2012 b, 2013 b) demonstrate that latent variables indexing consensus identification and personal beliefs are strongly related to one another even when controlling for the unique correspondence between science knowledge topics across these two types of questions. And multiple studies have identified distinct differences in individuals’ trust in science as a product of motivational factors (e.g., Hamilton et al., 2015; Hmielowski et al., 2013; Nisbet et al., 2015). This evidence is bolstered by an earlier analysis of the current data showing that a lack of trust in scientists was closely related to individuals’ willingness to disagree with what they thought most scientists would say (Pasek, in press).

Unpacking psychological mechanisms

The results presented in this study add weight to the conclusion that personal belief questions about scientific topics elicit a complex psychological interplay of preexisting information stores and group identifications (Kahan, 2015). Presumably, as individuals consider how to respond, they vary in the information they bring to bear, the consistency of that information (both from science and from other sources), and the desire to reach identity-bolstering conclusions.

Furthermore, individuals can engage in motivated reasoning around science in both conscious and subconscious ways. Some individuals may fail to translate perceptions of scientific consensus to personal beliefs because they overtly disagree with or distrust scientists (Hmielowski et al., 2013). Others may ascertain that scientists are wrong by privileging information from their own experiences (e.g., Hamilton and Stampone, 2013). Yet, others may discount the relevance of scientists’ claims on the matter at hand. They could accomplish this by believing that the bible trumps science (and hence, not really considering the new information at all) or by adhering to claims that scientists are themselves reaching motivated conclusions.

To the extent that this study adjudicates between competing psychological mechanisms, it makes three points. First, these results further undermine the notion that incorrect answering of personal belief questions about science is solely evidence of ignorance (Sturgis and Allum, 2004). Second, it provides a strong basis for asserting that group identifications shape how individuals answer knowledge questions (Kahan et al., 2011). And third, it suggests that different issues elicit distinct patterns of acceptance and rejection among motivated individuals and thereby imply unique remedies for the improvement of scientific awareness.

Limitations and future research

This study represents a first cut at understanding how motivational processes may underlie distinctions between perceptions of scientific consensus and individuals’ own beliefs. Relations between motivations and beliefs are tested across regression models, the predictors of which likely combine some factors that precede scientific beliefs and some that succeed it. Temporal relations between the variables observed and thus their causal ordering cannot be assessed due to the cross-sectional nature of the data. This means that personal beliefs may well precede consensus identification causally or that attitudes toward science might influence religiosity or partisanship. If we hope to better understand the distinctions between identification and acceptance of the scientific majority, an important future endeavor would be to test the causal ordering of these measures through the use of prospective designs and experimental manipulations (cf. Van Der Linden et al., 2015).

There are good reasons to believe that the discrepancies observed between identification of the scientific consensus and personal beliefs in the current study are understated. For one, the study relies on a nonprobability sample of respondents. Individuals who volunteer to take surveys online might be less likely to overtly reject science than those who do not regularly assent to scientific surveying. For another, personal belief questions closely followed by consensus identification might lead individuals to answer question in more consistent ways than they would if the measures were asked independently (see, for example, Schuman and Presser, 1981). Effects observed in this study likely differ from what would be expected were similar measures asked of a nationally representative sample, though there is little reason to expect that general patterns of results would not replicate (see, for example, Baker et al., 2013; Pasek, 2016).

Perhaps the biggest limitation of this study is an inability to isolate the mechanisms that render religious and partisan individuals less likely to identify or believe in some matters of scientific consensus. In particular, it remains possible that many people who did not correctly identify the scientific consensus either had never heard that a particular issue was indeed a consensus or received misinformation about the state of the science. Although evidence of widespread ignorance should result in failure to replicate among individuals holding misperceptions with confidence, which was not observed (see online Supplementary Appendix F), misinformation remains a compelling alternative hypothesis. Indeed, individuals visiting the Creation Museum in Petersburg, Kentucky might be forgiven for asserting that scientists believe humans and dinosaurs coexisted. Experimental evidence is necessary to distinguish motivated rejection of information from confident misperceptions induced by inaccurate information. Similarly understanding the extent to which rejectionism and misperception are truly cognitive states as opposed to response patterns would be valuable.

Nonetheless, the evidence presented here suggests that we ignore an important part of the story if we assume that motivated assessments of science are limited either to identification of a consensus or to holding beliefs in line with that consensus. Both processes are presumably capable of resolving cognitive dissonance and are likely at work. We would be well served to test mechanisms by which these relations operate in the future. Future research should also examine additional identifications that might lead to motivated disbelief in science.¹⁰

Conclusion

This study examines how two motivational factors—partisanship and religiosity—relate to identification of the scientific consensus and belief in that consensus across scientific issues. Results suggest that motivational processes are capable of leading individuals to reject either the presence of a scientific consensus or its relevance to their personal beliefs. Whether one or both processes occur appears to vary depending on the issue under examination. Hence, motivational factors appear to operate via both theorized pathways.

Footnotes

Acknowledgements

The author thanks Jon Miller, Santiago Olivella, Brendan Nyhan, Adam Berinsky, and Joe Uscinski for their comments on earlier drafts 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 study was funded in part by a grant from the Marsh endowment of the University of Michigan, Department of Communication Studies.

Notes

Author biography

Josh Pasek is Assistant Professor of Communication Studies, Faculty Associate in the Center for Political Studies, and Core Faculty for the Michigan Institute for Data Science at the University of Michigan. His research explores how new media and psychological processes each shape political attitudes, public opinion, and political behaviors.

References

Allum

Sturgis

Tabourazi

Brunton-Smith

(2008) Science knowledge and attitudes across cultures: A meta-analysis. Public Understanding of Science 17(1): 35–54.

Baker

Brick

Bates

Battaglia

Couper

Dever

et al (2013) Summary report of the AAPOR task force on non-probability sampling. Journal of Survey Statistics and Methodology 1(2): 90–143. Available at: https://doi.org/10.1093/jssam/smt008

Blank

Shaw

(2015) Does partisanship shape attitudes toward science and public policy? The case for ideology and religion. Annals of the American Academy of Political and Social Science 658(1): 18–35.

Bonanni

(1999) Demographic impact of vaccination: A review. Vaccine 17: S120–S125.

Boykoff

(2007) Climate change and journalistic norms: A case-study of US mass-media coverage. Geoforum 38(6): 1190–1204. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0016718507000188

Dunlap

McCright

(2008) A widening gap: Republican and democratic views on climate change. Environment: Science and Policy for Sustainable Development 50(5): 26–35.

Durant

Evans

Thomas

(1989) The public understanding of science. Nature 340(6228): 11–14.

Ecker

UKH

Lewandowsky

Swire

Chang

(2011) Correcting false information in memory: Manipulating the strength of misinformation encoding and its retraction. Political Behavior 18(3): 570–578.

Ellison

Musick

(1995) Conservative protestantism and public opinion toward science. Review of Religious Research 36(3): 245–262.

10.

Etchegary

Cappelli

Potter

Vloet

Graham

Walker

et al (2010) Attitude and knowledge about genetics and genetic testing. Public Health Genomics 13(2): 80–88.

11.

Evans

Durant

(1999) The relationship between knowledge and attitudes in the public understanding of science in Britain. Public Understanding of Science 4(1): 57–74.

12.

Feinstein

(2010) Salvaging science literacy. Science Education 95(1): 168–185.

13.

Festinger

(1957) A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.

14.

Gauchat

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

15.

Hamilton

Stampone

(2013) Blowin’ in the wind: Short-term weather and belief in anthropogenic climate change. Weather, Climate, and Society 5(2): 112–119.

16.

Hamilton

Hartter

Saito

(2015) Trust in scientists on climate change and vaccines. SAGE Open. 5(3): 1–13.

17.

Hart

Nisbet

(2012) Boomerang effects in science communication: How motivated reasoning and identity cues amplify opinion polarization about climate mitigation policies. Communication Research 39(6): 701–723.

18.

Hmielowski

Feldman

Myers

Leiserowitz

Maibach

(2013) An attack on science? Media use, trust in scientists, and perceptions of global warming. Public Understanding of Science 23(7): 866–883. Available at: http://pus.sagepub.com/cgi/doi/10.1177/0963662513480091

19.

Hobson

Niemeyer

(2013) “What sceptics believe”: The effects of information and deliberation on climate change scepticism. Public Understanding of Science 22(4): 396–412.

20.

Inhofe

(2012) The Greatest Hoax: How the Global Warming Conspiracy Threatens Your Future. Washington, DC: WND Books.

21.

Jones

III JE

(ed.) (2005) Kitzmiller V. Dover Area School District. Harrisburg, PA: Federal District Court.

22.

Kahan

(2015) Climate-science communication and the measurement problem. Advances in Political Psychology 36(S1): 1–43. Available at: http://doi.wiley.com/10.1111/pops.12244

23.

Kahan

Peters

Ellen

Wittlin

Slovic

Ouellette

Braman

et al (2012) The polarizing impact of science literacy and numeracy on perceived climate change risks. Nature Climate Change 2(10): 732–735.

24.

Kahan

Jenkins-Smith

Braman

(2011) Cultural cognition of scientific consensus. Journal of Risk Research 14(2): 147–174.

25.

King

Honake

Joseph

Scheve

(2001) Analyzing incomplete political science data: An alternative algorithm for multiple imputation. American Political Science Review 95(1): 49–69.

26.

Kögel

Prskawetz

(2001) Agricultural productivity growth and escape from the malthusian trap – Springer. Journal of Economic Growth 6(4): 337–357.

27.

Kunda

(1990) The case for motivated reasoning. Psychological Bulletin 108(3): 480–498.

28.

Lewandowsky

Ecker

UKH

Seifert

Schwarz

Cook

(2012a) Misinformation and its correction: Continued influence and successful debiasing. Psychological Science in the Public Interest 13(3): 106–131.

29.

Lewandowsky

Gignac

Oberauer

(2013a) The role of conspiracist ideation and worldviews in predicting rejection of science. PLoS ONE 8(10): e75637.

30.

Lewandowsky

Gignac

Vaughan

(2012b) The pivotal role of perceived scientific consensus in acceptance of science. Nature Climate Change 3(4): 399–404.

31.

Lewandowsky

Oberauer

Gignac

(2013b) NASA faked the moon landing—Therefore, (climate) science is a hoax: An anatomy of the motivated rejection of science. Psychological Science 24(5): 622–633.

32.

Lodge

Taber

(2013) The Rationalizing Voter. New York, NY: Cambridge University Press.

33.

Malka

Krosnick

Debell

Pasek

Schneider

(2009) Featuring Skeptics in News Media Stories about Global Warming Reduces Public Beliefs in the Seriousness of Global Warming. Stanford, CA: Woods Institute for the Environment, Stanford University.

34.

Mead

Mates

(2009) Why science standards are important to a strong science curriculum and how states measure up. Evolution: Education and Outreach 2(3): 359–371.

35.

Miller

(1983) Scientific literacy: A conceptual and empirical review. Daedalus 112(2): 29–48.

36.

Miller

(1998) The measurement of civic scientific literacy. Public Understanding of Science 7(3): 203–223.

37.

Miller

(2004) Public understanding of, and attitudes toward, scientific research: What we know and what we need to know. Public Understanding of Science 13(3): 273–294.

38.

National Science Board (2014) Science and Engineering Indicators 2014. Arlington, VA: National Science Foundation.

39.

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.

40.

Nyhan

Reifler

(2010) When corrections fail: The persistence of political misperceptions. Political Behavior 32(2): 303–330.

41.

Nyhan

Reifler

Richey

Freed

(2014) Effective messages in vaccine promotion: A randomized trial. Pediatrics 133(4): e835–e842.

42.

Osborne

Simon

Collins

(2003) Attitudes towards science: A review of the literature and its implications. International Journal of Science Education 25(9): 1049–1079.

43.

Pasek

(2016) When will nonprobability surveys mirror probability surveys? Considering types of inference and weighting strategies as criteria for correspondence. International Journal of Public Opinion Research 28(2): 269–291.

44.

Pasek

(in press). Don’t trust the scientists! Rejecting the scientific consensus “conspiracy”. In: Uscinski J (ed.) Conspiracy Theories and the People Who Believe Them. New York, NY: Oxford University Press.

45.

Peytchev

(2012) Multiple imputation for unit nonresponse and measurement error. Public Opinion Quarterly 76(2): 214–237.

46.

Redlawsk

Civettini

AJW

Emmerson

(2010) The affective tipping point: Do motivated reasoners ever “get it”? Political Psychology 31(4): 563–593.

47.

Roberts

Reid

Schroeder

Norris

(2013) Causal or spurious? The relationship of knowledge and attitudes to trust in science and technology. Public Understanding of Science 22(5): 624–641.

48.

Roos

(2014) Measuring science or religion? A measurement analysis of the national science foundation sponsored science literacy scale 2006–2010. Public Understanding of Science 23(7): 797–813.

49.

Schuman

Presser

(1981) Questions and Answers in Attitude Surveys. New York, NY: Academic Press.

50.

Sherkat

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

51.

Singh

(1995) Infant mortality in the United States: Trends, differentials, and projections, 1950 through 2010. American Journal of Public Health 85(7): 957–964.

52.

Snow

Dibner

(eds) (2016) Science Literacy: Concepts, Contexts, and Consequences. Washington, DC: National Academies Press.

53.

Solomon

(1993) Reception and rejection of science knowledge: Choice, style and home culture. Public Understanding of Science 2(2): 111–121.

54.

Standage

(2014) The Victorian Internet. New York, NY: Bloomsbury Publishing.

55.

Stocklmayer

Bryant

(2011) Science and the public—What should people know? International Journal of Science Education, Part B 2(1): 81–101.

56.

Sturgis

Allum

(2004) Science in society: Re-evaluating the deficit model of public attitudes. Public Understanding of Science 13(1): 55–74.

57.

Van Der Linden

Leiserowitz

Feinberg

Maibach

(2015) The scientific consensus on climate change as a gateway belief: Experimental evidence. PLoS ONE 10(2). Available at: http://dx.plos.org/10.1371/journal.pone.0118489

58.

Weingart

(1999) Scientific expertise and political accountability: Paradoxes of science in politics. Science and Public Policy 26(3): 151–161.

59.

Withey

(1959) Public opinion about science and scientists. Public Opinion Quarterly 23(3): 382.

60.

Yeager

Krosnick

Chang

Javitz

Levendusky

Simpser

et al (2011) Comparing the accuracy of RDD telephone surveys and Internet surveys conducted with probability and non-probability samples. Public Opinion Quarterly 75(4): 709–747. Available at: http://poq.oxfordjournals.org/cgi/doi/10.1093/poq/nfr020