Solution spillover: Pairing issues with ideological solutions fuels polarization

Significance

Issue salience polarization, in which people with differing political ideologies disagree about the relative importance and urgency of issues, is dangerous as it can undermine collective problem-solving, deepen societal divides, and obstruct effective governance. Yet, the emergence of issue salience polarization, especially regarding seemingly nonideological issues, is poorly understood. This paper offers an account of this phenomenon, suggesting that introducing ideologically divisive solutions to nonpolarized issues triggers “spillover polarization.” Consistent with this account, four experiments demonstrate that pairing a nonpolarized issue with an ideologically divisive solution can cause individuals whose political beliefs align with the solution to exaggerate the issue’s importance, while those whose beliefs conflict with the solution downplay its importance, thus polarizing views of the previously neutral issue.

Abstract

Recent research has focused on understanding ideological and affective political polarization in the United States; however, the polarization of issue salience—defined as personal beliefs about how important or concerning an issue is, and how much attention or priority it deserves—has garnered less attention, especially regarding its underlying causes. We propose that issue salience polarization, particularly that of ideologically neutral issues, is driven in part by two psychological processes: solution aversion (denying problem severity when solutions are introduced that conflict with political beliefs) and solution attraction (exaggerating problem severity when solutions are introduced that align with political beliefs). If this is so, we should be able to observe issues that are nonpolarized at baseline become suddenly polarized when paired with solutions that are consistent with a specific ideological position (liberal or conservative). Consistent with this prediction, four experimental studies demonstrate the emergence of issue salience polarization arising from solution aversion and solution attraction (Studies 1–4). Results also demonstrate support for the role of zero-sum beliefs, or the belief that one political party’s gains must come at the expense of the other party, in driving this process (Study 3).

Perceptions of the relative importance and severity of many issues that do not appear to be inherently ideological have become increasingly divided along partisan lines. For example, although at the onset of the COVID-19 pandemic, Democrats and Republicans were relatively unified in their concerns about catching COVID-19—with approximately a 10- to 20-point gap through April 2020—this gap quickly widened (1). By early June 2020, approximately 80% of Democrats, compared to just 40% of Republicans, reported being concerned (1). Similarly, while in the late 1980s and early 1990s, Democrats and Republicans showed minimal differences in the degree to which they worried “a great deal” about climate change, partisan attitudes sharply diverged over time; by 2021, approximately 90% of Democrats, compared to only about 30% of Republicans, worried a great deal about climate change (2). An analogous pattern is evident in perceptions of whether Americans’ level of agreement on basic facts and violent crime are “very big problems” in the United States. In 2019, there was only approximately a 7-point and 3-point gap, respectively (3, 4). However, these gaps widened; by 2025, more Democrats (58%) than Republicans (32%) perceived disagreement over basic facts as a very big issue, whereas for violent crime, Republicans (55%) were more likely than Democrats (37%) to view it as such (3, 4). These differences are particularly striking given that, when disconnected from broader policy implications, it is hard to see how the perceived severity of threats such as viruses, the changing climate, disagreement over basic facts, and violent crime is inherently linked to traditional conservative or liberal values—a point reflected by initial bipartisan alignment on these issues.

Perhaps even more surprising is the polarization of concerns about issues which are seemingly even further removed from the core values of liberal and conservative ideologies. These include concerns about remote worker (un)productivity (5, 6), the dangers of raw milk (7), and the ethics of various dog training strategies (8, 9)—issues that, to our knowledge, have not historically been systematically tracked in terms of political polarization within the mass public due to their apolitical nature.

In this paper, we propose that an overlooked cause of these divisions in beliefs about whether an issue involves a serious, concerning, or legitimate problem and how much attention or priority it deserves—a phenomenon which we hereon refer to as issue salience polarization (10–12)—is the solutions that issues get paired with. When nonpolarizing issues are linked with ideologically divisive solutions, long-standing ideological splits may create a “spillover” effect, in which people become polarized over the legitimacy and urgency of the underlying problem necessitating the solution. We suggest this can occur via two psychological processes: solution aversion (in which people downplay or deny the severity or legitimacy of a problem because the solution conflicts with their political beliefs) and solution attraction (in which people exaggerate the severity or legitimacy of a problem because the solution aligns with their political beliefs).

Solution Aversion

Solution aversion occurs when a dislike for solutions associated with a scientifically identified problem leads to skepticism or denial of that problem (13). Initially identified in the context of climate change, recent research has expanded application of this phenomenon to other areas, including dismissal of the risks of COVID-19 and denial that inequality is harmful (14, 15).

The motivated reasoning and social identity literature are central to understanding solution aversion. Psychological motivations (e.g., not wanting a specific solution) influence reasoning, leading individuals to evaluate evidence (e.g., denying the problem necessitating that solution) in ways that align with their desires and motivations, which are often closely tied to social identities, including political beliefs (16–20). Solution aversion is theorized to be partly driven by a broader cognitive tendency to seek coherence in beliefs (13, 21, 22). When a solution contradicts one’s ideological values, this drive for coherence can lead to rejection of the legitimacy and urgency of the underlying problem itself, rather than embracing an ideologically incompatible solution. This, in turn, can reduce the salience of the broader issue it represents—shaping perceptions of how deserving it is of attention and resources.

In this paper, we shift the lens from how solution aversion can explain one group’s noted skepticism of a problem to whether solution aversion—along with an entirely unexamined process, solution attraction—can drive polarization of issue salience, in which two groups become divided over whether they think an issue should be prioritized due to differing perceptions of how severe and legitimate its underlying problem is.

Solution Attraction

Logically, if one’s identities and accompanying motivations can lead them to evaluate evidence in a biased way such that they deny the realities of certain problems due to aversion to proposed solutions, high attraction to proposed solutions could lead to the opposite: exaggeration and emphasis of the severity of problems. We call this solution attraction. Like solution aversion, we explore whether one group engaging in solution attraction can lead to polarization of issue salience.

While motivated cognition work has largely focused on problem denial (e.g., 13, 15, and 23), exaggeration of the urgency of certain issues is potentially just as important and dangerous. Exaggerating problem severity to justify solutions may lead to unnecessary panic, conspiracy theories, misdirection of resources, and misinformation (24, 25). Thus, though the focus of this paper is the role of solution attraction (and aversion) in the creation of polarization, introducing solution attraction to the literature offers a substantial contribution in itself, with many potential implications and avenues for future research.

Zero-Sum Beliefs and the Co-Occurrence of Solution Aversion and Solution Attraction

Polarization of issue salience arising from the co-occurrence of solution aversion and attraction should generally be more extreme than when one process occurs in isolation, as two groups move further away from each other regarding their beliefs about an issue’s importance and urgency. In theory, this should occur when an issue is paired with a solution that appeals to one political party while simultaneously repelling the other, provoking opposing motivated cognition processes. This may be especially likely in a two-party, hypercompetitive political landscape that can engender zero-sum thinking, or the belief that one political party’s gains come at the expense of the other party, and vice versa—as observed in the current U.S. sociopolitical environment (26–29). For those high in zero-sum thinking, any policy prescription offered as a solution that aligns with a particular ideology should be seen as a loss for the other side. If so, as zero-sum beliefs increase (decrease), solution attraction and aversion from the same solution should be more (less) likely, making polarization stronger (weaker).

Overview of Studies

Across three experimental studies, our primary objective was to investigate whether the proposal of solutions manipulated to align with either liberal or conservative ideals would lead to the emergence of polarization of issue salience. Thus, in Studies 1–3, we present participants with a problem (a different one in each study) that is (presumably) not currently polarized and then randomly assign half of the participants to learn about a proposed ideologically divisive solution^* that aligns or contrasts with their political beliefs while the other half see a control solution that lacks any ideological slant. In each, we compare beliefs regarding problem severity, urgency, and/or priority (which we label: issue salience) across the two conditions. Specifically, in each study, we assess whether the ideologically divisive solution condition produced polarization, operationalized as a shift from no association between left–right political orientation (i.e., ideology) and issue salience in the control solution condition to a significant association in the ideologically divisive solution condition—reflecting whether issue salience differed meaningfully across the ideological spectrum only when an ideologically divisive solution was introduced (30, 31). Then, we follow up that primary test of polarization with specific tests of solution aversion (decreased issue salience compared to control when the solution is ideologically inconsistent) and solution attraction (increased issue salience compared to control when the solution is ideologically consistent). In Study 3, we also test the role of zero-sum thinking in exacerbating this process. Then, in Study 4, we more stringently explore whether polarization driven by solution aversion and attraction can emerge in the absence of any source/partisan cues in a nonpolitical context.

Research Transparency Statement

For each study, all data, analysis code, study materials, and preregistrations are accessible on the Open Science Framework (OSF): https://osf.io/mgs4q/?view_only=c1fd1e2d903044eca5d345a20a796fb0 (32). All studies were preregistered and approved by the Institutional Review Board at Duke University. All participants provided informed consent. All analyses presented in the main text are preregistered, and all preregistered analyses are provided unless otherwise noted (e.g., they were moved to SI Appendix). We did not recruit additional participants once we obtained our target sample sizes, nor did we analyze data before reaching our target sample sizes. All data were collected between March 1, 2024, and May 1, 2025, and each study consists of a unique participant sample. In the paper, we sometimes report only shortened versions of our stimuli; full-length stimuli can be found in SI Appendix and on OSF.

Study 1

In Study 1, we varied whether a liberal-leaning solution was or was not paired with a presumably nonpolarizing issue: America lagging behind other countries in innovation and development of advanced and emerging technologies. We predicted that issue salience polarization would arise as a result of the proposed ideologically divisive solution, evidenced by a shift from no association between political orientation (i.e., ideology) and issue salience in the control solution condition to a significant association in the liberal-leaning solution condition; specifically, we predicted that in the liberal-leaning solution condition, as participants’ views became more liberal (less conservative), they would rate the issue as more salient. We also predicted that this polarization would be driven by solution aversion and attraction, such that more conservative participants would rate the issue as less salient in the liberal-leaning solution condition compared to the control, while more liberal participants would rate the issue as more salient in the liberal-leaning solution condition compared to the control.

Method

Participants.

Using Prolific’s custom prescreening, we targeted recruitment of 1,050 participants: ½ self-identified Democrat and ½ self-identified Republican to ensure variability of political beliefs. After removing participants who failed attention checks, data were analyzed with the remaining 1,031 participants (Mage = 44.59, SDage = 14.22, Gender: 557 women, 437 men, 37 other, Political Orientation: 211 very liberal, 196 somewhat liberal, 76 slightly liberal, 63 middle of the road, 138 slightly conservative, 226 somewhat conservative, and 121 very conservative, Political Affiliation: 509 Republicans, 522 Democrats, Race: 689 White, 189 Black, 65 Asian, 45 Hispanic/Latino, 43 other). An a priori power analysis in G*Power indicated that we would be adequately powered to detect the predicted interaction effect (F test, Linear multiple regression: Fixed model, R² increase, Effect size f² = 0.02, α = 0.05; power = 0.80, Number of tested predictors = 1, Number of predictors = 3).

Materials and Procedure.

First, participants responded to measures of political orientation (1 = very liberal, 7 = very conservative) and political affiliation (Democrat vs. Republican). Next, all participants read a brief blurb about a specific problem: “In recent years, there have been concerns that America’s innovation and development of advanced and emerging technologies is lagging behind other countries. While some claim that the strong economy and culture of innovation in the U.S. remains unrivaled, others point to research showing that labor productivity growth has been slowing over the last two decades, making the U.S. less competitive in the global arena.” Then, participants were randomly assigned to the liberal-leaning solution condition or control solution condition (between-subjects). Participants in the liberal-leaning solution condition read the following: “To address this issue, policy experts have put forward new recommendations. A frequently proposed solution that is gaining momentum involves increased support for policies emphasizing diversity, equity, and inclusion (DEI)…,” while participants in the control solution condition read a similarly toned message with only vague mention of a range of solutions: “To address this issue, policy experts have put forward new recommendations. These recommendations include a range of solutions.” In this paper, when we report the language from problem and solution conditions, it is sometimes shortened to preserve space. See SI Appendix and OSF for full-length stimuli.

After reading the assigned material according to their assigned solution condition, all participants responded to three items measuring issue salience, using Likert scales: “How worried should Americans be about the U.S. lagging in innovation and global competitiveness?” (1 = Not at all worried, 7 = Extremely worried); “How much of a priority should the U.S. government make increasing innovation and global competitiveness in their policy-making?” (1 = Not at all a priority, 7 = A top priority); and “What portion of government resources should be directed toward policies that increase U.S. innovation and global competitiveness?” (1 = None at all, 7 = An extremely large portion). We preregistered collapsing the 3 continuous items into one measure if they were strongly correlated, and because they were (α = 0.90), we proceeded as such.^†

The study also included four or five attention checks (depending on assigned solution condition) and a series of demographic questions. See SI Appendix and OSF for full study stimuli.

Results

Test of Polarization.

First, we tested for polarization: we predicted an interaction between political orientation and solution condition such that political orientation would relate to issue salience in the liberal-leaning solution condition (such that as participants became more liberal, they would rate the issue as more salient), but not in the control solution condition. With political orientation mean-centered (M = 3.86), results revealed the predicted interaction, b = −0.14, SE = 0.04, t(1027) = −3.51, P < 0.001, 95% CI [−0.21, −0.06], and follow-up analyses were consistent with the prediction that polarization would emerge in the liberal-leaning solution condition in the hypothesized direction. Specifically, a linear regression revealed that political orientation negatively predicted issue salience in the liberal-leaning solution condition, b = −0.08, SE = 0.03, t(508) = −2.92, P = 0.003, 95% CI [−0.13, −0.03], such that as participants became more liberal (and less conservative), they rated the issue as more salient. While political orientation also significantly predicted issue salience ratings in the control solution condition, it was in the opposite direction as in the liberal-leaning solution condition (as people were more conservative, they rated the issue as more salient), b = 0.06, SE = 0.03, t(519) = 2.04, P = 0.04, 95% CI [0.001, 0.11].

Tests of Solution Aversion and Attraction.

Next, we tested whether this polarization was driven by solution aversion, attraction, or both: we predicted that more liberal participants would rate the issue as more salient in the liberal-leaning solution condition compared to the control and vice versa for more conservative participants. To test for solution aversion and attraction, we used a Johnson–Neyman floodlight analysis to identify the ranges of political orientation (non-mean-centered to increase interpretability) for which the simple effect of the solution manipulation was significant. This analysis revealed evidence of solution aversion: very conservative participants (political orientation > 6.53) rated the issue as significantly less salient (P < 0.05) in the liberal-leaning solution condition compared to the control.^‡ It also revealed evidence of solution attraction: more liberal participants (political orientation < 3.35) rated the issue as significantly more salient (P < 0.05) in the liberal-leaning solution condition compared to the control (Fig. 1).

Fig. 1.

Study 1: Issue salience by political orientation and solution condition. Note: In the liberal-leaning solution condition, political orientation significantly predicted issue salience ratings (p = 0.003). In the control solution condition, political orientation significantly predicted issue salience ratings but in the opposite direction (P = 0.04). (For all line graphs: Trend lines represent model-based predictions. Y-axes have been truncated to highlight meaningful variation.) Dotted lines represent Johnson–Neyman points. Error bands indicate 95% CI.

Together, these findings provide evidence that even an issue that is weakly polarized at baseline can become polarized in the opposite direction when paired with an ideologically divisive (liberal-leaning) solution—driven by solution aversion among more conservative participants and solution attraction among more liberal participants.

Study 2

In Study 2, we varied whether a conservative-leaning solution was or was not paired with a different presumably nonpolarizing issue: increasing strain on emergency services personnel. We again predicted that issue salience polarization would arise as a result of the proposed ideologically divisive solution, evidenced by a shift from no association between political orientation and issue salience in the control solution condition to a significant association in the conservative-leaning solution condition; specifically, we predicted that in the conservative-leaning solution condition, as participants’ views became more conservative (less liberal), they would rate the issue as more salient. We also predicted that this polarization would be driven by solution aversion and attraction, such that more liberal participants would rate the issue as less salient in the conservative-leaning solution condition compared to the control, and more conservative participants would rate the issue as more salient in the conservative-leaning solution condition compared to the control.

Because the liberal-leaning solution in the previous study was also more specific and detailed than the control solution, one alternative explanation for the observed interaction in Study 1 is that the results were driven by differences in specificity rather than ideology. That is, perhaps liberals tend to like detailed solutions, or conservatives tend to dislike them—regardless of the political content. To rule out this alternative, in Study 2, we match the specificity of both the control and ideological solutions and also pivot to a conservative-leaning solution.^§

Method

Participants.

Using Prolific’s custom prescreening, we targeted recruitment of 1,070 participants: ½ self-identified Democrat and ½ self-identified Republican to ensure variability of political beliefs. After removing participants who failed attention checks, data were analyzed with the remaining 982 participants (Mage = 40.98, SDage = 13.63, Gender: 576 women, 395 men, 11 other, Political Orientation: 224 very liberal, 190 somewhat liberal, 55 slightly liberal, 65 middle of the road, 105 slightly conservative, 190 somewhat conservative, and 153 very conservative, Political Affiliation: 485 Republicans, 497 Democrats, Race: 737 White, 127 Black, 37 Asian, 40 Hispanic/Latino, 41 other). Like Study 1, an a priori power analysis in G*Power indicated that we would be adequately powered to detect the predicted interaction effect (F test, Linear multiple regression: Fixed model, R² increase, Effect size f² = 0.02, α = 0.05; power = 0.80, Number of tested predictors = 1, Number of predictors = 3).

Materials and Procedure.

First, participants responded to the same measures of political orientation and political affiliation as in Study 1. Next, all participants read a brief blurb about a specific problem: “Recently, some Americans have expressed concerns about increasing strain on emergency services, such as police, fire departments, and ambulance services, as they are experiencing higher call volumes and longer response times in rapidly growing areas. This is impacting their ability to provide timely assistance and ensure public safety…” Then, participants were randomly assigned to the conservative-leaning solution condition or the control solution condition (between-subjects). Participants in the conservative-leaning solution condition read the following: “To address this issue, policy experts have proposed harsher immigration policies (e.g., stricter visa requirements and border security) to reduce the number of immigrants, which proponents argue will limit population growth and therefore, decrease the strain on emergency services...” Participants in the control solution condition read a similarly toned message with a specific, nonpartisan solution: “To address this issue, policy experts have proposed implementing Lean Six Sigma, a project management approach proven to boost operational efficiency across many sectors…”^¶

After reading the assigned material according to their assigned solution condition, all participants responded to four items measuring issue salience using Likert scales: “How worried should Americans be about increasing strain on emergency services?” (1 = Not at all worried, 7 = Extremely worried); “How would you rate this problem in terms of urgency?” (1 = Not at all urgent, 7 = Extremely urgent); “How much of a priority should the U.S. government make reducing the strain on emergency services in their policy-making?” (1 = Not at all a priority, 7 = A top priority); and “The U.S. government has a set of resources pre-allocated toward public services…What portion of these pre-allocated government resources should be directed toward reducing the strain on emergency services workers, specifically?” (1 = None at all, 7 = An extremely large portion). We preregistered collapsing the 4 continuous items into one measure if they were strongly correlated, and because they were (α = 0.89), we proceeded as such.

Finally, the study included three attention checks and a series of demographic questions. See SI Appendix and OSF for full study stimuli.

Results

Test of Polarization.

First, we tested for polarization: we predicted an interaction between political orientation and solution condition such that political orientation would relate to issue salience in the conservative-leaning solution condition (such that as participants became more conservative, they would rate the issue as more salient), but not in the control solution condition. With political orientation mean-centered (M = 3.83), results revealed the predicted interaction, b = 0.18, SE = 0.03, t(978) = 5.35, P < 0.001, 95% CI [0.12, 0.25], and follow-up analyses were consistent with the prediction that polarization would emerge in the conservative-leaning solution condition in the hypothesized direction. Specifically, a linear regression revealed that political orientation positively predicted issue salience in the conservative-leaning solution condition, b = 0.19, SE = 0.03, t(467) = 7.49, P < 0.001, 95% CI [0.14, 0.24], such that as participants became more conservative (and less liberal), they rated the issue as more salient. As predicted, political orientation had no significant association with issue salience ratings in the control solution condition, b = 0.01, SE = 0.02, t(511) = 0.39, P = 0.70, 95% CI [−0.04, 0.05].

Tests of Solution Aversion and Attraction.

Next, we tested whether this polarization was driven by solution aversion, attraction, or both: we predicted that more conservative participants would rate the issue as more salient in the conservative-leaning solution condition compared to the control and vice versa for more liberal participants. We again used a Johnson–Neyman floodlight analysis to identify the ranges of political orientation (non-mean-centered to increase interpretability) for which the simple effect of the solution manipulation was significant. This analysis revealed evidence of solution aversion: liberal participants (political orientation < 4.40) rated the issue as significantly less salient (P < 0.05) in the conservative-leaning solution condition compared to the control. It also revealed evidence of solution attraction: very conservative participants (political orientation > 6.56) rated the issue as significantly more salient (P < 0.05) in the conservative-leaning solution condition compared to the control (Fig. 2).

Fig. 2.

Study 2: Issue salience by political orientation and solution condition. Note: In the conservative-leaning solution condition, political orientation significantly predicted issue salience ratings (P < 0.001). In the control solution condition, political orientation did not significantly predict issue salience ratings (P = 0.70). Dotted lines represent Johnson–Neyman points. Error bands indicate 95% CI.

In sum, Study 2 replicated the central finding of Study 1: the introduction of an ideologically divisive solution to a nonpolarized issue^# can produce issue salience polarization driven by solution aversion and attraction. This result held even when employing a different issue domain, a conservative-leaning solution, and a specificity-matched control condition.

Study 3

In Study 3, we again varied whether a conservative-leaning solution was or was not paired with yet another presumably nonpolarizing issue: declining plane safety. As before, we predicted that issue salience polarization would arise as a result of the proposed ideologically divisive solution, evidenced by a shift from no association between political orientation and issue salience in the control solution condition to a significant association in the conservative-leaning solution condition; specifically, we predicted that in the conservative-leaning solution condition, as participants’ views became more conservative (less liberal), they would rate the issue as more salient. Once again, we predicted that this polarization would be driven by solution aversion and attraction, such that more liberal participants would rate the issue as less salient in the conservative-leaning solution condition compared to the control, and more conservative participants would rate the issue as more salient in the conservative-leaning solution condition compared to the control. In addition, we included a measure of political zero-sum thinking, predicting a three-way interaction between political orientation, solution condition, and zero-sum beliefs, such that the polarizing effect of the conservative-leaning solution condition would increase (decrease) when zero-sum beliefs were higher (lower), with no corresponding effects in the control solution condition.

Method

Participants.

Using Prolific’s custom prescreening, we targeted recruitment of 1,600 participants: ½ self-identified Democrat and ½ self-identified Republican to ensure variability of political beliefs. After removing participants who failed attention checks, data were analyzed with the remaining 1,572 participants (Mage = 39.20, SDage = 13.02, Gender: 964 women, 588 men, 20 other, Political Orientation: 279 very liberal, 306 somewhat liberal, 119 slightly liberal, 145 middle of the road, 183 slightly conservative, 297 somewhat conservative, and 243 very conservative, Political Affiliation: 781 Republicans, 791 Democrats, Race: 1,045 White, 306 Black, 78 Asian, 80 Hispanic/Latino, 63 other. Again, an a priori power analysis in G*Power indicated that we would be adequately powered to detect the predicted interaction effect (F test, Linear multiple regression: Fixed model, R² increase, Effect size f² = 0.02, α = 0.05; power = 0.80, Number of tested predictors = 1, Number of predictors = 3), although we multiplied our previous sample sizes by ~1.5 to account for the additional complexity in the analysis (three-way interaction with zero-sum beliefs).

Materials and Procedure.

For the most part, procedures mirrored those of the previous studies. First, participants responded to the same measures of political orientation and political affiliation. Additionally, participants responded to a six-item measure of political zero-sum beliefs (e.g. “The more resources the government spends on predominantly liberal [conservative] regions in the U.S. (“Blue” [“Red”] states), the less it can spend on predominantly conservative [liberal] regions (“Red” [“Blue”] states)”) on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree)) (29). Next, all participants read a brief blurb about a specific problem: “In recent years, some people have been worried about the safety of flying in the U.S. However, experts say there’s no proof of declining passenger airplane safety. Instead, they say that the media has focused on a few newsworthy incidents, which has spread fears. In fact, research shows that flying in the U.S. is getting safer every year. For example, in 2023, the International Air Transport Association reported its best safety results ever, with only 1 accident per 1.26 million flights, the lowest rate in over ten years, and just one fatal accident.”^|| Then, participants were randomly assigned to the conservative-leaning solution condition or control solution condition (between-subjects). Participants in the conservative-leaning solution condition read the following: “To address concerns about declining passenger airplane safety in the U.S., some policy and legal experts have put forward new recommendations. A frequently proposed solution that is gaining support involves cutting back on diversity, equity, and inclusion (DEI) initiatives. This recommendation aligns with the general preference of Republicans to ban many DEI programs on the grounds that they are discriminatory and divisive…”^** Participants in the control solution condition read a similarly toned message with only vague mention of a range of solutions: “To address concerns about declining passenger airplane safety in the U.S., some policy and legal experts have put forward new recommendations. These recommendations include a range of solutions that are gaining support.” After reading the assigned material according to their assigned solution condition, all participants responded to one continuous issue salience item^†† using a 7-point Likert scale: “How worried should Americans be about airplane safety declining?” (1 = Not at all worried, 7 = Very worried).

The study also included three to four attention checks (depending on assigned condition) and a series of demographic questions. See SI Appendix and OSF for full study stimuli.

Results

Test of Polarization.

First, we tested for polarization: we predicted an interaction between political orientation and solution condition such that political orientation would relate to issue salience in the conservative-leaning solution condition (such that as participants became more conservative, they would rate the issue as more salient), but not in the control solution condition. With political orientation mean-centered (M = 3.96), results revealed the predicted interaction, b = 0.14, SE = 0.04, t(1568) = 3.69, P < 0.001, 95% CI [0.06, 0.21], and follow-up analyses were consistent with the prediction that polarization would emerge in the conservative-leaning solution condition in the hypothesized direction. Specifically, a linear regression revealed that political orientation positively predicted issue salience in the conservative-leaning solution condition, b = 0.19, SE = 0.03, t(771) = 6.88, P < 0.001, 95% CI [0.13, 0.24], such that as participants became more conservative (and less liberal), they rated the issue as more salient. As predicted, political orientation had no significant association with issue salience ratings in the control solution condition, (b = 0.05, SE = 0.03, t(797) = 1.80, P = 0.07, 95% CI [−0.004, 0.10].

Tests of Solution Aversion and Attraction.

Next, we tested whether this polarization was driven by solution aversion, attraction, or both: we again predicted that more conservative participants would rate the issue as more salient in the conservative-leaning condition compared to the control and vice versa for more liberal participants. We again used a Johnson–Neyman floodlight analysis to identify the ranges of political orientation (non-mean-centered to increase interpretability) for which the simple effect of the solution manipulation was significant. This analysis revealed evidence of solution aversion: very liberal participants (political orientation < 1.28) rated the issue as significantly less salient (P < 0.05) in the conservative-leaning solution condition compared to the control. It also revealed evidence of solution attraction: more conservative participants (political orientation > 4.27) rated the issue as significantly more salient (P < 0.05) in the conservative-leaning solution condition compared to the control (Fig. 3).

Fig. 3.

Study 3: Issue salience by political orientation and solution condition. Note: In the conservative-leaning solution condition, political orientation significantly predicted issue salience ratings (P < 0.001). In the control solution condition, political orientation did not significantly predict issue salience ratings (P = 0.07). Dotted lines represent Johnson–Neyman points. Error bands indicate 95% CI.

Study 3 replicated the central finding of Studies 1 and 2: the introduction of an ideologically divisive solution to a nonpolarized issue can produce issue salience polarization driven by solution aversion and attraction.

Tests of the Moderating Role of Zero-Sum Beliefs.

Next, we examined whether this polarization was driven by high zero-sum beliefs; we predicted a three-way interaction such that the polarizing effect of the conservative-leaning solution condition would be most (least) pronounced as zero-sum beliefs became higher (lower), with no corresponding effect in the control solution condition. With political orientation (M = 3.96) and zero-sum beliefs (M = 5.01) mean-centered and solution condition dummy-coded, results revealed the predicted three-way interaction, b = 0.10, SE = 0.03, t(1564) = 2.94, P = 0.003, 95% CI [0.03, 0.17]. To parse out this three-way interaction, we ran two linear regressions filtering by solution condition. In the conservative-leaning solution condition, there was a significant main effect of political orientation, b = 0.18, SE = 0.03, t(769) = 6.65, P < 0.001, 95% CI [0.13, 0.23] and a significant main effect of zero-sum beliefs, b = 0.11, SE = 0.06, t(769) = 1.99, P = 0.047, 95% CI [0.001, 0.22], both qualified by the predicted interaction between political orientation and zero-sum beliefs, b = 0.05, SE = 0.02, t(769) = 2.13, P = 0.03, 95% CI [0.004, 0.10]. A Johnson–Neyman floodlight analysis revealed that, as predicted, political orientation significantly predicted issue salience ratings (P < 0.05) in the expected direction (more conservative participants rated the issue as more salient than more liberal participants)—or in other words, issue salience was polarized—only when zero-sum beliefs were above 3.43 (non-mean-centered for interpretability). In the control solution condition, a significant main effect of zero-sum beliefs, b = 0.26, SE = 0.05, t(795) = 4.91, P < 0.001, 95% CI [0.16, 0.36] was qualified by an unexpected interaction between political orientation and zero-sum beliefs, b = −0.05, SE = 0.02, t(795) = −2.04, P = 0.04, 95% CI [−0.10, −0.001]. A Johnson–Neyman floodlight analysis revealed that political orientation significantly predicted issue salience ratings (P < 0.05) (more conservative participants rated the issue as more salient than more liberal participants) when zero-sum beliefs were below 4.97 (non-mean-centered for interpretability), suggesting that under lower zero-sum beliefs, more conservative beliefs predicted higher salience ratings. Thus, for some reason, when no specific solution was offered, political orientation predicted issue salience, but only among those low in zero-sum beliefs (Fig. 4).^‡‡

Fig. 4.

Study 3: Issue salience by political orientation and zero-sum beliefs for each solution condition. Note: Dotted lines represent Johnson–Neyman points. Error bands indicate 95% CI.

Although we did observe an unpredicted and robust positive correlation between zero-sum beliefs and issue salience ratings, analyses revealed the predicted interaction: the polarization that emerged between liberals and conservatives upon the introduction of the conservative-leaning solution condition was moderated by zero-sum thinking, such that it strengthened (weakened) when zero-sum beliefs were higher (lower), with no corresponding effect in the control condition.

Study 4

In Study 4, we sought to distinguish the solution spillover process from traditional source cueing accounts of polarization, which suggest that partisan divides can arise from reliance on non-issue-relevant features, specifically the (partisan) identity of those advocating for specific issues to be taken seriously or not (33–37). To do so, we examined whether pairing problems with aversive or attractive solutions can trigger motivated reasoning based solely on the substance of the solutions, even when the source remains constant. While we acknowledge that partisan cueing likely contributed to the effects found in Studies 1–3, since it is reasonable to infer that an ideological solution may have been offered by an ideologically aligned source, our goal in Study 4 was to isolate the content of the solution itself as a driver of polarization, independent of who is proposing it. By testing the impact of solutions in a context completely devoid of political source cues, we also sought to demonstrate that solution aversion and attraction processes can occur beyond political contexts.

Research on solution aversion and climate change denial has tried to rule out source cues as the sole cause of this process by homing in on the content of the solution and its match with the beliefs of the participants, even within a sociopolitical category. For example, Campbell and Kay found that linking a free market solution to climate change reduced skepticism only among Republicans who most strongly held a specific conviction in the value of the free market, suggesting the particulars of one’s beliefs are relevant beyond their broader social identity (13). Even in cases like that, however, source cues are hard to fully eliminate as explanations. For instance, free-market beliefs could simply be potential markers of one’s alignment with the Republican Party. Thus, because of how difficult it would be to entirely unconfound these variables in an ecologically valid way in a political context, and to show the relevance of solution aversion and attraction beyond politics, we pivot away from politics in this study.

To examine whether issue salience polarization driven by solution aversion and attraction can arise even when political identity—or any salient social identity—of the solution proposer is held constant, we presented all participants (i.e., Prolific workers) with the same general issue related to their Prolific work: researchers’ concerns about participant screen fatigue and burnout. Using a between-subjects design, we paired this issue with a control solution, a solution that would be aversive to Prolific workers, or a solution that would be attractive to Prolific workers, all attributed to the same source—researchers who use the Prolific platform. We predicted that issue salience polarization among the Prolific participants would emerge as a function of the content of the proposed solutions: Prolific workers in the attractive solution condition would rate the issue necessitating the solution as more concerning and of higher priority (more salient) than those in the aversive solution condition. We also predicted that this polarization would be driven by solution aversion and attraction relative to the control condition.

Method

Participants.

We targeted recruitment of 1,100 Prolific participants. After removing participants who failed attention checks, data were analyzed with the remaining 1,056 participants (Mage = 40.71, SDage = 14.58, Gender: 582 women, 449 men, 25 other). An a priori power analysis in G*Power indicated that we would be adequately powered to detect our predicted effects (F test; ANOVA: fixed effects, omnibus, one-way; effect size f = 0.10, α = 0.05; power = 0.80; number of groups: 3).

Materials and Procedure.

All participants read a brief blurb about an issue: “Recently, some researchers have raised concerns about Prolific participants experiencing screen fatigue and burnout. They note that completing many surveys—always on screens and with repetitive or demanding tasks—can lead to reduced focus and well-being. These researchers worry that this fatigue may be detrimental to participants’ health and wellbeing while also negatively impacting data quality.” Then, participants were randomly assigned to the control, attractive, or aversive solution condition (between-subjects). Participants in the control solution condition read the following: “To address this issue, researchers have proposed a variety of potential solutions.” Participants in the attractive solution condition read: “To address this issue, researchers have proposed offering PAID 20-min breaks for every five studies completed, provided participants maintain a 90% approval rating or higher. Supporters argue that while this approach may increase researcher costs slightly, it would help prevent screen fatigue and burnout, encourage healthier pacing, and ultimately support better data quality.” Participants in the aversive solution condition read: “To address this issue, researchers have proposed limiting participants to a maximum of three studies per day. Supporters argue that while this approach may limit participant earnings, it would help prevent screen fatigue and burnout, encourage healthier pacing, and ultimately support better data quality.”

After reading the assigned material according to their assigned solution condition, all participants responded to two items measuring issue salience, using Likert scales: “How concerned should researchers be about Prolific workers experiencing screen fatigue and burnout?” (1 = Not at all concerned, 7 = Extremely concerned); “How much of a priority should researchers make addressing screen fatigue and burnout among Prolific workers?” (1 = Not at all a priority, 7 = A top priority). We preregistered collapsing the two continuous items into one measure of issue salience if they were strongly correlated, and because they were (α = 0.93), we proceeded as such.

Finally, the study included one or two attention checks (depending on solution condition) and a series of demographic questions. See SI Appendix and OSF for full study stimuli.^§§

Results

We conducted a one-way ANOVA on issue salience across the three solution conditions: control, aversive, and attractive. There was a significant main effect of condition: F(2, 1053) = 115.51, P < 0.001, η² = 0.18. Pairwise comparisons (Tukey-adjusted) revealed that polarization emerged such that participants in the attractive solution condition (M = 4.23, SE = 0.08, 95% CI [4.06, 4.40]) rated the issue as significantly more salient than participants in the aversive solution condition (M = 2.39, SE = 0.09, 95% CI [2.22, 2.57]), b = 1.84, t = 15.06, P < 0.001.

Additionally, further exploration revealed that this polarization emerged from both solution aversion, such that participants in the aversive solution condition rated the issue as significantly less salient than the control solution condition (M = 3.55, SE = 0.08, 95% CI [3.39, 3.72]), b = −1.16, t = −9.59, P < 0.001, and solution attraction, such that participants in the attractive solution condition rated the issue as significantly more salient than those in the control condition, b = 0.68, t = 5.68, P < 0.001 (Fig. 5).

Fig. 5.

Study 4: Issue salience by solution condition. Note: Participants in the attractive solution condition rated issue salience significantly higher than those in both the aversive (P < 0.001) and control (P < 0.001) solution conditions. Additionally, participants in the control condition rated issue salience significantly higher than those in the aversive condition (P < 0.001).

While partisan cues—a type of source cue—are known to influence issue attitudes (e.g., refs. 33–37), these findings suggest a distinct pathway to polarization rooted in the appeal of the substance of proposed solutions, beyond their source. While source cues may certainly be at play when solutions appear to be proposed by members of a particular political party, leading people who identify with that party to adopt compatible views, in this study, both the attractive and aversive solutions were believably attributed to the same source, suggesting that the proposed solutions had differing impacts on issue salience based on their substance alone.

General Discussion

Across three experimental studies situated in political contexts, each examining issues that are nonpolarized^¶¶ at baseline, we find evidence supporting a spillover account of political polarization, in which issue salience becomes polarized upon the brief introduction of an ideologically divisive solution. We find that this polarization is driven by two psychological processes: solution aversion and solution attraction. This account of spillover polarization is timely, given that many Americans currently view politics as divisive and exhausting (38). In a fourth experimental study, we demonstrate that these same psychological processes—solution aversion and attraction—can drive polarization in entirely apolitical contexts, underscoring the broader relevance of our account.

We recognize that many other factors—including divergent media coverage—shape real-world divides in issue salience. For instance, liberal and conservative media outlets highlight issues with varying frequency and emphasis; CNN, for example, covered COVID-19 more frequently and highlighted its severity more than Fox News, likely contributing to partisan gaps in perceived threat (39–41). These effects are likely amplified by selective exposure, as people tend to consume media aligned with their beliefs about which issues are important (41, 42). However, our controlled approach allows us to isolate the specific impact of solution content on issue salience polarization, independent of other contributing factors, such as media or source cues.

Theoretical Contributions.

First, while many mechanistic accounts of polarization exist (e.g., refs. 43–48), ours contributes to this literature through an explicit examination of how polarization of issue salience, or beliefs about the severity, urgency, and priority of issues, emerges when issues—particularly those that are seemingly ideologically neutral—become linked to specific policy solutions. Understanding the emergence of issue salience polarization, specifically, is crucial because when individuals across political divides disagree about the importance and urgency of issues, this may impede collective problem-solving, exacerbate societal divides, and obstruct effective governance. Additionally, this understanding provides insight into the formation of public priorities, the shaping of political agendas, and divisions over not just what people believe, but which problems they believe are worth engaging with at all (10–12).

Our spillover account may be particularly helpful in explaining issue salience polarization in cases where partisan differences do not necessarily align with people’s direct experiences or personal exposure to the issue itself. For instance, despite living in states more prone to natural disasters (49), Republicans tend to see them as less urgent, while Democrats—who are more likely to live in large U.S. cities that have historically been targets of high-profile international terrorism (e.g., New York City; 50, 51)—tend to see this issue as less worrisome (3). These patterns suggest that experience with a problem is not sufficient to explain how severe people perceive it to be and point to the importance of considering alternative factors, such as the impacts of the policies that are attached to these problems.

Second, while our spillover account shares similarities with partisan cue accounts—both propose that polarization can stem from factors associated with an issue rather than the issue itself (e.g., 35, 36)—it adds to this literature in important ways. First, in Studies 1 and 2, no explicit partisan cue was provided; to the extent that cues were present, they could only be inferred through the substance of the ideological solution itself. This suggests that, at the very least, solutions themselves may operate as a heretofore unacknowledged partisan cue; a finding with significant implications in a political landscape where issues are constantly paired with ideological policy solutions. What’s more, in Study 4, we also demonstrate the emergence of issue salience polarization in a context completely devoid of political partisan identity cueing, a finding that is consistent with past work also ruling out a strict identity-cueing-only account of solution aversion (13). Specifically, by holding the source constant while manipulating the proposed solution in a nonpolitical context, we found that polarization arose solely based on whether the solution aligned with or threatened participants’ self-interests. In other words, polarization was driven by the substance of the solution, not its source, also suggesting that solution attraction and aversion can be driven not only by political identities but also by apolitical, self-relevant motivations. Solution spillover, therefore, like most meaningful sociopolitical phenomena, is most likely multiply determined—driven by both identity-based motivations and epistemic motives to maintain coherence across beliefs.

Third, our results highlight what may be considered an underappreciated form of false polarization, where perceived divisions over the importance and urgency of issues may instead stem from disagreements on how to solve them, given that the issues we tested were not polarized until paired with an ideologically divisive solution. False polarization is typically defined as an overestimation of differences in beliefs between two groups in terms of extremity (52). However, spillover polarization of issue salience highlights a different form: false polarization of breadth, arising from a false reality in which Americans appear divided over the importance of many issues, when in fact, they may be truly divided over only a few longstanding ideological solutions. Solutions that have long been central to politics may become so tightly linked to associated issues that they become difficult to psychologically or politically separate. This form of false reality—in which political opponents seem divided over which problems deserve attention when they may mostly disagree over their proposed solutions—may, similarly to false polarization as traditionally defined, contribute to intensifying affective polarization and ultimately fuel real ideological polarization as party loyalties grow (53).

Limitations and Future Directions.

First, while our studies provide strong empirical evidence that, in both political and non-political contexts, perceptions of the solutions that issues are paired with can spill over into perceptions of the issues themselves, our methods — experimental studies conducted in highly controlled online settings — have limitations. Future research could use alternative approaches, such as longitudinal designs, to examine how and to what extent solutions fuel real-world divisions over time.

We also see promising opportunities for future research to more fully examine the phenomenon of solution attraction—the tendency to emphasize or exaggerate a problem when the proposed solution aligns with one’s political ideology—especially as motivated cognition work has largely focused on problem denial. While our findings introduce this motivated cognition process to the literature, future work should investigate its broader consequences. For instance, how might overstating problems to justify particular ideological solutions affect political priorities, resource allocation, conspiracy beliefs, and fear-based rhetoric? Additionally, to what degree does solution attraction affect everyday behavior and agenda-setting in organizational contexts?

Finally, evidence from Study 3 suggests that issue salience polarization heightens under high zero-sum beliefs, suggesting that a key mechanism underlying solution aversion, attraction, and polarization may be Americans’ view of politics as win-lose. When individuals see a solution as a win (loss) for the opposing party and a loss (win) for their own, motivated reasoning may lead them to evaluate evidence in a way that lessens (heightens) likelihood of implementation of the proposed solution. The consistent polarizing effects observed across Studies 1–3 may reflect that our proposed solutions were perceived as zero-sum—either because they were explicitly tied to one political party (Study 3) or addressed issues (e.g., DEI, immigration) that are frequently framed as zero-sum in public discourse (28, 54). Future research could further explore how zero-sum beliefs contribute to polarization in different contexts. Another productive direction for future research may involve testing ways to mitigate spillover polarization of issue salience. As the absence of proposed solutions is untenable, potential interventions may involve framing solutions to emphasize shared interests while de-emphasizing ideological affiliations, or proposing multiple solutions simultaneously to reduce zero-sum assumptions.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Data, analysis code, study materials, and pre-registrations have been deposited in OSF (https://osf.io/mgs4q/?view_only=c1fd1e2d903044eca5d345a20a796fb0) (32).