Assignment 1

The Partisan Mind: Is Extreme Political Partisanship Related to Cognitive Inflexibility?

Leor Zmigrod, Peter Jason Rentfrow, and Trevor W. Robbins University of Cambridge

The rise of partisan animosity, ideological polarization, and political dogmatism has reignited important questions about the relationship between psychological rigidity and political partisanship. Two compet- ing hypotheses have been proposed: 1 hypothesis argues that mental rigidity is related to a conservative political orientation, and the other suggests that it reflects partisan extremity across the political spectrum. In a sample of over 700 U.S. citizens, partisan extremity was related to lower levels of cognitive flexibility, regardless of political orientation, across 3 independent cognitive assessments of cognitive flexibility. This was evident across multiple statistical analyses, including quadratic regressions, Bayes factor analysis, and interrupted regressions. These findings suggest that the rigidity with which individ- uals process and respond to nonpolitical information may be related to the extremity of their partisan identities.

Keywords: cognitive flexibility, political psychology, partisanship, cognitive testing, Bayesian analysis

Supplemental materials: http://dx.doi.org/10.1037/xge0000661.supp

In The True Believer (Hoffer, 1951), the thinker Eric Hoffer wrote about crowds and mass movements, “All movements, how- ever different in doctrine and aspiration, draw their early adherents from the same types of humanity; they all appeal to the same types of mind.” Hoffer tapped into an important idea: there may be a certain “type of mind” that is particularly drawn to adopting an ideology or doctrine, almost regardless of its content or ambition. Indeed, scholars and scientists have long sought to identify the psychological underpinnings of rigid ideological adherence, par- ticularly since the atrocities committed in the name of political ideologies at the start of the 20th century.

Many psychologists have theorized that rigid adherence to a polit- ical ideology may reflect underlying cognitive rigidity (Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950; Rokeach, 1954). Nonetheless, there have been two competing hypotheses regarding the precise relationship between cognitive rigidity and political identity. One hypothesis, the ideological extremity hypothesis, has suggested

that individuals on the political extremes— of both liberal and con- servative ideologies—are more cognitively rigid than moderates (Greenberg & Jonas, 2003). This hypothesis emerges from the notion that extreme political partisanship may be linked to inflexible belief systems that capture the world in black-and-white terms and offer certainty and simplicity (Fernbach, Rogers, Fox, & Sloman, 2013; van Prooijen & Krouwel, 2017; see also Crawford, 2019). Indirect evi- dence has illustrated that political extremists on both the political left and right are more dogmatically intolerant (van Prooijen & Krouwel, 2017), more likely to feel superior about their beliefs (Toner, Leary, Asher, & Jongman-Sereno, 2013), and to avoid exposure to each other’s opinions (Frimer, Skitka, & Motyl, 2017) relative to political moderates. Consequently, the rigidity-of-the-extreme hypothesis sug- gests that political extremism is underpinned by a cognitive rigidity that facilitates the attitudinal rigidity that often characterizes political extremists.

The second hypothesis, the so-called rigidity-of-the-right hypothesis, has argued that conservatives perceive the world in a more inflexible, categorical way than liberals (Altemeyer, 1998; Jost, Glaser, Kruglanski, & Sulloway, 2003a). This hypothesis has received empirical support through studies demonstrating a rela- tionship between political conservatism and psychological prefer- ences for traditionalism, familiarity, and certainty, and between political liberalism and acceptance of uncertainty and ambiguity (for reviews, see Hibbing, Smith, & Alford, 2014; Jost, 2017; Van Hiel, Onraet, & De Pauw, 2010).

Notably, empirical research on this debate has almost exclu- sively relied on self-report questionnaires as proxies for cognitive rigidity, as opposed to the use of objective cognitive assessments to quantify cognitive inflexibility directly. Self-report measures of cognitive style tend to suffer from considerable limitations, such as failure in self-assessment of cognitive abilities (Kruger & Dun- ning, 1999), inclusion of political content in nonpolitical measures

This article was published Online First August 5, 2019. Leor Zmigrod, Department of Psychology, and Behavioural and Clinical

Neuroscience Institute, University of Cambridge; Peter Jason Rentfrow, Department of Psychology, University of Cambridge; Trevor W. Robbins, Department of Psychology, and Behavioural and Clinical Neuroscience Institute, University of Cambridge.

Authors declare no competing interests. All data, code, and materials will be made available on a public repository upon publication. This research was supported by a Gates Cambridge Trust scholarship to Leor Zmigrod. The authors thank Jildou Talman and Angelica Fischer for their assistance in coding the Alternative Uses Test responses.

Correspondence concerning this article should be addressed to Leor Zmigrod, Department of Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, United Kingdom. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Journal of Experimental Psychology: General © 2019 American Psychological Association 2020, Vol. 149, No. 3, 407– 418 0096-3445/20/$12.00 http://dx.doi.org/10.1037/xge0000661

407

and nonpolitical content in political measures (Malka, Lelkes, & Holzer, 2017), and poor psychometric properties (Kirton, 1981). Additionally, a recent meta-analysis has shown that correlations between cognitive style and political attitudes are inflated when relying on self-report questionnaires to quantify cognitive rigidity rather than behavioral assessments (Van Hiel, Onraet, Crowson, & Roets, 2016). Furthermore, most measures of “cognitive and per- ceptual rigidity” used thus far in political psychology have not corresponded to those used in the cognitive psychology literature. Consequently, there is a need to test these hypotheses using es- tablished objective and ideologically neutral measures of cognitive rigidity.

Defining Cognitive Flexibility and Rigidity

In the neuropsychological literature, cognitive flexibility is de- fined as the ability to adapt to novel or changing environments and a capacity to switch between modes of thinking (Cools & Robbins, 2004). Specifically, it can be defined as “the ability to flexibly switch perspectives, focus of attention, or response mappings” (Diamond, 2006, p. 70). Cognitive inflexibility is therefore repre- sented by perseveration, “the tendency of an individual not to change” (Schultz & Searleman, 2002, p. 166), and maladaptive continuation of unrewarded behaviors and thought patterns.

There is significant individual variation in cognitive flexibility within the general population (Braver, Cole, & Yarkoni, 2010; Miyake & Friedman, 2012), which has been linked to dopaminer- gic functioning (Barnett, Jones, Robbins, & Müller, 2007). Fur- thermore, from a clinical perspective, some populations manifest a deficit in cognitive flexibility by persisting with previously estab- lished rules or behavioral patterns even when this is maladaptive, as evident in patients with obsessive– compulsive disorder (Cham- berlain et al., 2006), addiction (Verdejo-García, Pérez-García, & Bechara, 2006), and frontal lobe damage (Anderson, Damasio, Jones, & Tranel, 1991).

As reviewed by Ionescu (2012), there are several behavioral tasks that are classically used to operationalize cognitive flexibility in adults, including the Wisconsin Card Sorting Test (Grant & Berg, 1948), task switching and optional shift paradigms (Miyake & Friedman, 2012; Monsell, 2003), the Alternative Uses Task (AUT; Guilford, 1967), insight problems, and induction tasks (Shafto, Coley, & Vitkin, 2007). In these tasks, performance is measured via participants’ accuracy rates, reaction times (RTs), or the number and variety of provided responses to open-ended problems. Given the extensive use and validation of behavioral tasks for assessing cognitive flexibility in neuropsychology and cognitive science, the present study will rely on three validated and independent behavioral measures of cognitive flexibility in order to investigate the relationship between inflexibility and political ideology.

Rigidity-of-the-Extreme: The Psychology of Extreme Partisanship

The two hypotheses—the ideological extremity hypothesis and the rigidity-of-the-right hypothesis—tap into two orthogonal as- pects of political identity. As noted by Settle, Dawes, and Fowler (2009), “Partisanship is typically evaluated along two dimen- sions—the strength of reported partisan attachment and the direc-

tion of that attachment” (p. 601). While a rich literature has emerged on the psychology of partisan direction (i.e., left vs. right political orientation), there has been little empirical investigation of the cognitive origins of partisan intensity, that is, the strength of a person’s partisan identity and attachment (Van Bavel & Pereira, 2018).

Nonetheless, new findings have demonstrated several ways in which political extremists, on both sides of the political spectrum, differ psychologically from political moderates. Relative to mod- erates, political extremists tend to experience more negative emo- tions about politics and to derogate outgroups (van Prooijen, Krouwel, Boiten, & Eendebak, 2015) and view politics in more simplistic terms (Lammers, Koch, Conway, & Brandt, 2017). Political extremists’ tendency to believe in simple political solu- tions to complex societal challenges also predicts their greater likelihood of believing in conspiracy theories (van Prooijen, Krou- wel, & Pollet, 2015). Furthermore, research with cognitive anchor- ing tasks has suggested that political extremists exhibit greater belief superiority and are more likely to reject external information than moderates (Brandt, Evans, & Crawford, 2015). In the case of the EU refugee crisis, political extremists possessed greater judg- mental certainty about their knowledge of the crisis, independently of their actual knowledge, than political moderates (van Prooijen, Krouwel, & Emmer, 2018). This demonstrates that both partisan direction and intensity matter for how individuals evaluate ideo- logical arguments and intergroup conflict.

Recently, Jost (2017) pointed out several key questions cur- rently faced by political psychology, including: (a) What does political psychology have to say about left-wing rigidity? and (b) Are ideological differences only evident in subjective, self-report measures and not behaviorally? With regards to the question of left-wing rigidity, while recent evidence has shown that left-wing authoritarianism exists and is predictive of voting behavior (Con- way III, Houck, Gornick, & Repke, 2018; Conway III & McFar- land, 2019), there is a lack of research directly examining the relationship between partisan intensity and cognitive rigidity in- dependently of partisan direction. When the rigidity-of-the-right hypothesis and the ideological extremity hypothesis have been pitted against each other, the psychological variables of interest have consisted of self-reported intolerance of ambiguity, need for cognitive closure, dogmatism, and integrative complexity (Jost, Glaser, Kruglanski, & Sulloway, 2003b); none have assessed cog- nitive rigidity directly and objectively. This is related to the second question posed by Jost (2017) about the contrast between self- report and behavioral methods—perhaps the use of behavioral methodologies has the potential to illuminate ideological differ- ences (and similarities), which are absent or obscured in self-report questionnaires.

Rigidity-of-the-Right: Evidence and Measurement Problems

Two recent meta-analyses have evaluated the state of the evi- dence in favor of the rigidity-of-the-right hypothesis. Jost (2017) identified 16 studies that investigated whether conservatives score more highly on tests of “perceptual or cognitive rigidity” than liberals. Nine out of 16 studies supported the rigidity-of-the-right hypothesis, such that the overall unweighted (r � .32) and weighted (r � .38) effect sizes were large and significant. Notably,

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

408 ZMIGROD, RENTFROW, AND ROBBINS

however, out of the nine studies demonstrating a significant effect, six studies used self-report rather than behavioral measures: five studies used a self-report measure of cognitive rigidity (Gough & Sanford, 1952), and one study used a self-report measure of intolerance of trait inconsistency (Steiner & Johnson, 1963). Two studies used unvalidated behavioral measures that do not clearly tap into cognitive flexibility: Kidd and Kidd’s (1972) study asked participants to identify changes between visual objects drawn on cards. However, recognizing changes or discrepancies between visual stimuli does not imply adaptability to change or flexibility of thought, and the author-designed task was not validated or used as a measure of cognitive flexibility in later studies. The second study by Neuringer (1964) used the Rokeach Map Test as a test of rigidity, in which participants were presented with sequential street maps and asked to find the shortest distance between two points on the map. After being presented with seven maps in which the shortest route was identical across maps, participants were shown five maps in which a new shortcut was possible. If participants chose the shortcut on any of the five test maps, they were char- acterized as nonrigid. This study’s small number of test trials (i.e., five), low threshold for being classified as nonrigid (scoring above 0 out of 5), small sample (N � 15), along with the problematic nature of binary assessments of continuous cognitive constructs, should raise serious doubts regarding the validity of Neuringer’s (1964) findings. Consequently, the significant findings from this meta-analysis of “cognitive and perceptual rigidity” were based largely on self-report measures and a couple of problematic be- havioral tests.

The second meta-analysis, by Van Hiel and colleagues (2016), identified a larger number of studies in which a measure of rigidity was administered (N � 46), but separated the studies according to whether rigidity was operationalized with a behavioral task (N � 31) or a self-report scale (N � 15). Overall, the findings corrob- orate Jost’s (2017) meta-analysis, as conservatism was signifi- cantly related to self-reported rigidity (r � .47) and, to a lesser extent, behaviorally assessed rigidity (r � .11). Even in this expanded sample of behavioral studies, there is significant varia- tion in how cognitive rigidity was operationalized; for instance, while Rokeach (1954) utilized a set of perceptual problem solving tasks (i.e., the Einstellung problems; which have received criti- cisms: Levitt, 1956), Sidanius (1985) used the Political Prediction Test, in which participants are asked to make judgments within the political domain and cognitive flexibility is extracted as an index from how these judgments are made. Consequently, some behav- ioral measures of cognitive rigidity possess ideological content and so are not inherently politically neutral.

These two meta-analyses clearly demonstrate that no behavioral study has been conducted for over 20 years on disentangling the rigidity-of-the-right and ideological extremity hypotheses (since 1997; see Van Hiel et al., 2016), and the publication year for a behavioral study up to 1997 is on average 1959 (range: 1948 to 1997). Moreover, the debate has been plagued by relatively small sample sizes (M � 65, range � 15 to 225; calculated from Van Hiel et al., 2016) and diverse methodologies and conceptualiza- tions of both conservatism and rigidity. The present study therefore sought to rely on established, ideologically neutral cognitive tasks from the neuropsychological literature and recruit a large sample that would be adequately powered to detect these effects.

Materials and Method

Participants and Procedure

We sought to recruit 648 participants via Amazon’s Mechanical Turk (https://www.mturk.com) to achieve greater than 90% power to detect a small effect of Cohen’s f � .1 at � � .005 in our primary one-way ANCOVAs of the cognitive variables and three groups of political affiliation and conservatism. We anticipated a small effect due to previous work done by Sidanius (1985) and reviewed by Jost and colleagues (2003), which found an average correlation of .15 between general conservatism and conceptually similar cognitive flexibility constructs. We oversampled by 15% and recruited 750 participants, seven of which were excluded due to providing incomplete responses, yielding a total sample of 743 participants. Data collection was not continued after data analysis. All participants were U.S. citizens. Participants were redirected from Amazon’s Mechanical Turk to an online survey hosted by Qualtrics Survey Software for completion of all the self-reported items and the Remote Associates Test (RAT), and later redirected again to Inquisit 5 by Millisecond Software in order to temporarily download software that allows for accurate measure of perfor- mance and RTs in the Wisconsin Card Sorting Test (WCST). Participants were asked about demographic variables such as age, gender, and educational attainment (ranging from high school graduate, some college but no degree, associate degree [2-year], bachelor’s degree [4-year], Master’s degree, professional degree [JD, MD], or doctoral degree). The research was conducted with the ethical approval of the institution’s Department of Psychology Ethics Committee.

Analysis was conducted using the R packages BayesFactor (Morey, Rouder, Jamil, & Morey, 2015), ggstatsplot (Patil & Powell, 2018), visreg (Breheny & Burchett, 2017), and jtools (Long, 2019).

Measures

Political identity measures. Political party affiliation. To indicate their political party

affiliation, participants were asked: “Generally speaking, do you usually think of yourself as a Republican, a Democrat, an Inde- pendent, or something else?,” with the response options being Republican, Democrat, Independent, No preference, or the option to indicate another affiliation.

Political partisanship. To measure participants’ feeling of attachment to their political party, participants were presented with a validated measure of identity fusion, the Dynamic Identity Fu- sion Index (DIFI; Jiménez et al., 2016), which consists of a continuous pictorial representation that allows participants to move a small circle representing “the self” by clicking and drag- ging it toward or away from a large circle representing “the group/ideology.” The amount of overlap between the two circles has been shown to indicate the extent to which individuals feel their personal identity is fused with a collective identity (e.g., Jiménez et al., 2016; Jong, Whitehouse, Kavanagh, & Lane, 2015; Kapitány, Kavanagh, Buhrmester, Newson, & Whitehouse, 2019; Misch, Fergusson, & Dunham, 2018). It has temporal stability, as well as convergent and discriminant validity (Jiménez et al., 2016). In this study, participants were presented with the DIFI twice; once

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

409POLITICAL PARTISANSHIP AND COGNITIVE INFLEXIBILITY

where the group was the “Republican Party” and another when the group was “Democratic Party”. Hence, to compute Republican Party Identity Fusion, the size of the overlapping area between the “Me” circle and the “Republican Party” circle was calculated. The same methodology was applied to the calculation of Democratic Party Identity Fusion.

To calculate political partisanship, we computed each partici- pant’s maximum identity fusion to either political party (Demo- cratic/Republican Party), and retrieved the value for the party with which the participant was most fused. To capture the left–right spectrum, if participants’ maximum fusion was with the Demo- cratic Party, their fusion score was multiplied by �1, to create a spectrum from �100 to 100. This was taken as an index of participants’ political partisanship, weighted by partisan direction.

PPmax � ��FusionD if FusionD � FusionRFusionR otherwise where FusionR reflects Republican Party Identity Fusion and Fu- sionD reflects Democratic Party Identity Fusion.

As a robustness check, we also computed the difference in participants’ fusion to the two parties:

PPdif f � FusionR � FusionD

Furthermore, in order to obtain an unweighted measure of political partisanship, for cases when it is necessary to examine participants’ fusion with their favored party regardless of the party’s political direction, we collapsed PPmax into:

PP| max | � �FusionD if FusionD � FusionRFusionR otherwise Consequently, the PP| max | scale reflects political party identity

fusion between 0 and 100, rather than between �100 and 100 as in PPmax.

Political conservatism. The 12-item Social and Economic Conservatism Scale (SECS; Everett, 2013) was administered, which asks participants about how positively or negatively they feel toward 12 social and economic political issues on a scale of 0 –100 (with increments of 10). As detailed by Everett (2013), the social issues consisted of abortion, religion, the family unit, tradi- tional marriage, traditional values, patriotism, and military and national security. The economic issues included welfare benefits, limited government, business, gun ownership, and fiscal respon- sibility. By summing the participant’s score in response to social issues and separately in response to economic issues, we could compute a score of issue-specific social conservatism and eco- nomic conservatism respectively. The reliability of the SECS was high (Cronbach’s alpha for overall scale � .883, social conserva- tism subscale � .872, economic conservatism subscale � .736).

Cognitive flexibility tests. Remote Associates Test (RAT). The Remote Associates Test

(RAT; Mednick, 1962) consisted of 15 compound remote associate problems, in which participants are presented with three cue words (e.g., cottage, swiss, and cake), and must generate the compound word solution that connects these three words (e.g., cheese). Items of varying difficulty levels were selected from established remote associate problems (Bowden & Jung-Beeman, 2003). Participants were given 20 s to respond to each item. RAT performance was computed in terms of the proportion of correct words.

Wisconsin Card Sorting Test (WCST). The WCST (Grant & Berg, 1948) was administered with Inquisit 5 by Millisecond Software in standard fashion (Heaton, 1981). Participants are presented with four key cards and a deck of response cards that vary on three dimensions (color, shape, and number of geometric figures) and are asked to match a fifth card from the sequentially presented response cards to one of the four key cards. Participants need to identify the correct classification rule (out of three poten- tial rules: matching by color, shape, or number) according to the feedback they receive after each trial. They are informed that the classification rule may change without warning, and indeed the rule alternates after participants correctly respond to 10 consecu- tive trials, requiring a flexible set shift. The task ends after partic- ipants complete six categories (twice for each of the three rules) or after 128 trials. To index participants’ performance, the number of categories completed and the accuracy rate were computed.

Alternative Uses Test (AUT). In the AUT (Guilford, 1967), participants are asked to generate as many possible uses for two common household items (brick and newspaper) for 2 min. Par- ticipants’ responses were recorded and scored along four compo- nents by two independent raters in accordance with previous guidelines (Cronbach’s alpha � .994; Chermahini & Hommel, 2010; Madore, Addis, & Schacter, 2015; Roberts et al., 2017). Flexibility was scored according to the number of distinct catego- ries that participants’ responses for a given item could be clustered into (e.g., using a newspaper for making origami and making paper dolls are uses that would fall under the same category of arts and crafts, while using a newspaper for swatting a fly would fall under a separate category). The total flexibility score comprised the sum from all trials. Fluency constituted the total number of appropriate responses. Elaboration reflected the amount of detail provided by the participants (for brick, “build” would receive a score of 0; “build a house” would receive a score of 1; and “a weapon to protect family when your house is robbed” would be awarded two points for specifying detailed use and context). To score original- ity, each response was compared to the responses from the rest of the participants, such that responses to a given object that were only provided by 5% of the sample received an originality point. The total originality score reflected the sum of original scores per participant across all trials. In accordance with convention, non- sensical uses were excluded prior to coding of responses. To establish interrater reliability for appropriate categories, level of detail for the elaboration scoring, and originality, the raters sepa- rately scored 25 random participants’ responses, and once high interrater reliability was established with this set on all AUT measures (Cronbach’s alpha � .91 on all measures); the raters independently scored the rest of the participants. Each AUT mea- sure reflects the mean score given by the two independent raters.

Results

The sample consisted of 743 participants (Age: M � 36.58, SD � 13.46; 55.6% female), including 323 self-affiliated Demo- crats, 161 Republicans, and 203 Independents (56 participants with responded as “Other” and so were excluded for the analyses of political affiliation but were included in all other analyses). We first validated known associations between political affiliation, political partisanship, and conservatism. Univariate ANOVAs on participants’ conservatism with respect to their party affiliation

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

410 ZMIGROD, RENTFROW, AND ROBBINS

demonstrated significant differences between Democrats, Indepen- dents, and Republicans in social conservatism, F(2, 596) � 111.01, p � 1.046 � 10�41, �p2 � .271, and economic conserva- tism, F(2, 621) � 169.268, p � 2.113 � 10�59, �p2 � .353. After post hoc Bonferroni correction, across both conservatism mea- sures, Democrats (social conservatism: M � 338.85, SD � 135.65; economic conservatism: M � 225.73, SD � 67.17) were signifi- cantly more liberal than Independents (ps � .001; social conser- vatism: M � 392.64, SD � 148.57; economic conservatism: M � 284.56, SD � 83.49), and Independents were significantly more liberal than Conservatives (ps � .001; social conservatism: M � 540.52, SD � 117.09; economic conservatism: M � 361.88, SD � 76.12).

Democrats were no more fused to their favored party—that is, were no more partisan—than Republicans, t(451) � �.237, p � .813 (see Figure 1). Participants who identified as Independents were significantly less partisan (to the party with which they felt greater identity fusion) than either Democrats or Republicans, F(2, 617) � 38.582, p � .001, �p2 � .111.

Pearson’s correlations among the cognitive flexibility variables were modest, replicating past research: AUT Flexibility was pos- itively related to RAT, r � .20, p � .001 and WCST, r � .19, p � .001, and the correlation between RAT and WCST approached significance, r � .10, p � .066. There was manifest individual variability in each of the cognitive flexibility measures (RAT: M � 68.02%, SD � 23.85%, range � 0 to 100%; WCST: M � 71.61%,

SD � 15.31%, range � 20.31%; AUT Flexibility: M � 3.745, SD � 1.636, range � 0 to 9).

In terms of the demographic variables, there were no gender differences in performance on any of the cognitive flexibility mea- sures: RAT, t(639) � 1.346, p � .176, WCST, t(396) � �.172, p � .863, and AUT Flexibility, t(716) � .416, p � .678. There was a significant gender difference in absolute political partisanship (PP|max|), t(636) � 2.410, p � .016, with women (M � 56.09, SD � 31.99) exhibiting heightened identity fusion to their favored political party relative to men (M � 49.94, SD � 31.95). With regards to conservatism, there was no significant gender difference in social conservatism, t(622) � .112, p � .911 but there was a significant difference in economic conservatism, t(659) � �3.777, p � .001, with men (M � 293.24, SD � 91.80) reporting heightened economic conservatism relative to women (M � 266.13, SD � 90.69). Educa- tional attainment was not significantly related to any of the variables of interest (ps � .124). Furthermore, age was positively related to RAT performance, r � .136, p � .001, but not WCST, r � �.042, p � .403 or AUT Flexibility, r � �.021, p � .573. Age was also positively related to social conservatism, r � .217, p � .001 and economic conservatism, r � .083, p � .034 but not to absolute political partisanship (PP|max|, r � �.003, p � .943). However, there was no significant age difference between Republicans (M � 38.32, SD � 12.831), Democrats (M � 35.69, SD � 14.059), and Indepen- dents (M � 36.23, SD � 12.913; F(2, 676) � 2.076, p � .126), nor a significant difference in educational attainment (F(2, 682) � .884,

Figure 1. Political partisanship according to political affiliation. Democrats and Republicans were equally partisan in their identity fusion with their favored political party (PP| max |). See the online article for the color version of this figure.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

411POLITICAL PARTISANSHIP AND COGNITIVE INFLEXIBILITY

p � .414; Republicans: M � 2.24, SD � 1.373; Democrats: M � 2.41, SD � 1.289; Independents: M � 2.38, SD � 1.261). Conse- quently, and in accordance with convention, the variables of age, educational attainment, and gender were consistently included as covariates in all statistical analyses unless otherwise specified.

Political Affiliation and Cognitive Flexibility

A univariate ANCOVA on RAT performance according to self-reported political affiliation (Democrats, Republicans, vs. In- dependents) was computed, with age, gender, and educational attainment as covariates. This revealed a main effect of political affiliation, F(2, 595) � 5.345, p � .005, �p2 � .018, such that Independents performed significantly better on the RAT (M � 72.38%, SD � 22.26%) than Democrats (M � 66.67%, SD � 24.19%, Independent–Democrat estimated marginal mean differ- ence � .065, p � .016, 95% CI [.009, .121]) and Republicans (M � 65.16%, SD � 25.38%, Independent–Republican estimated marginal mean difference � .078, p � .010, 95% CI [.014, .142]) after post hoc Bonferroni correction. There was no significant difference between Democrats and Republicans. There were no effects of gender or educational attainment, but there was a sig- nificant effect of age, F(1, 595) � 10.131, p � .002, �p2 � .017 (see Figure 2).

Similarly, a univariate ANCOVA on WCST performance with age, gender, and educational attainment as covariates demon- strated a main effect of political affiliation, F(2, 363) � 5.794, p �

.003, �p2 � .031, whereby Independents performed significantly better on the WCST (M � 75.53%, SD � 11.83%) than Democrats (M � 71.27%, SD � 15.67%, Independent–Democrat estimated marginal mean difference � .045, p � .047, 95% CI [.000, .089]) and Republicans (M � 68.28%, SD � 17.06%, Independent–Republican estimated marginal mean difference � .072, p � .003, 95% CI [.019, .125]) after post hoc Bonferroni correction. There was no effect of age, gender, or educational attainment. There were no significant differences between Democrats and Republicans (see Figure 2).

Finally, a univariate ANCOVA on the AUT Flexibility score, with age, gender, and educational attainment as covariates re- vealed a main effect of political affiliation, F(2, 664) � 5.872, p � .003, �p2 � .017 (see Figure 2), such that Democrats (M � 3.903, SD � 1.533) and Independents (M � 3.987, SD � 1.481) per- formed significantly better than Republicans (M � 3.456, SD � 1.637), after post hoc Bonferroni correction (Republican– Democrat estimated marginal mean difference � �.443, p � .014, 95% CI [�.797, �.068]; Republican–Independent estimated mar- ginal mean difference � �.535, p � .004, 95% CI [�.932, �.137]). There was no difference in AUT Flexibility between Democrats and Independents, and there was no effect of age, gender, or educational attainment.

Partisanship and Cognitive Flexibility

Bayesian quadratic regressions. Given the inverted U-shaped relationship evident in Figure 2, we tested for quadratic associations

Figure 2. Cognitive flexibility according to political affiliation for the three cognitive flexibility tests: Remote Associates Task, Wisconsin Card Sorting Test, and Alternative Uses Test Flexibility score. The shaded areas reflect 95% confidence intervals. Comparisons indicate significant differences after Bonferroni correction, accounting for age, gender, and educational attainment. Data points reflect partial residuals. � p � .05, �� p � .01. See the online article for the color version of this figure.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

412 ZMIGROD, RENTFROW, AND ROBBINS

between cognitive flexibility and political partisanship across the political spectrum (PPmax). For each cognitive flexibility measure, we conducted two regression models: one that assumed a linear relation- ship and another that assumed a quadratic relationship between cog- nitive flexibility and political partisanship. We computed Bayes fac- tors for the linear and quadratic regressions to compare the strength of the evidence for a quadratic (subscript q) over a linear (subscript l) model specification (BFql).

The Bayes factors in this case express the relative likelihood of a quadratic model versus a linear model given the data and certain prior expectations. To calculate Bayes factors using Bayesian regression, we relied on a default Bayesian approach promoted by Wetzels and colleagues (2011), Rouder and Morey (2012), and Liang, Paulo, Molina, Clyde, and Berger (2008), and computation- ally specified in the R package BayesFactor (Morey et al., 2015). As evident in Table 1, across all three measures, there was evi- dence in favor of quadratic regressions over linear regressions. This was especially pronounced for the RAT and AUT Flexibility. For instance, in the case of the RAT results, the data are more than 2000 times more likely to have occurred under a quadratic model than a linear model, reflecting decisive evidence (for more details on evidence categories see: Wetzels et al., 2011; Jeffreys, 1961) in favor of a quadratic relative to a linear model. Moreover, the coefficients of the quadratic regressions, provided in Table 1, demonstrate a significant quadratic effect of political partisanship after adjusting for age, gender, and educational attainment, for all three tests of cognitive flexibility.

This was corroborated by a robustness check with the alternative measure of political partisanship, PPdiff. Regressing political par- tisanship (PPdiff) on RAT revealed a significant quadratic term ( � �.125, t � 02.861, p � .004) and greater evidence for a quadratic than a linear association, BFql � 10.329. Similarly, there was a quadratic effect when regressing political partisanship (PPdiff) on AUT Flexibility ( � �.114, t � �2.731, p � .007), with strong evidence for a quadratic regression model BFql � 7.493. The quadratic effect on WCST was not statistically signif- icant ( � �.097, t � �1.665, p � .097), with anecdotal evidence for a quadratic regression model BFql � 1.018.

Interrupted regressions. To validate the existence of in- verted U-shaped relationships, we use an interrupted regression model (the two-lines test; Simonsohn, 2018). This method simul-

taneously estimates two regression lines— one for low x-values and one for high x-values. This facilitates the testing of a sign change, that is, whether the average effect of x on y is of opposite sign for high versus low values of x. The two-lines test also identifies a data-driven change point where the two lines split using the “Robin Hood” algorithm, which has been demonstrated to obtain higher statistical power for detecting U-shaped relation- ships than other algorithmic alternatives (Simonsohn, 2018). An inverted U-shaped relationship exists if the explanatory variable positively correlates with the outcome variable at low values, and negatively correlates with the outcome variable at high values.

The results are presented in Figure 3 (see also Figure S2 in the online supplemental materials for figures with data points). The slope between cognitive flexibility and political partisan identity on the left of the political spectrum is positive across all three measures of cognitive flexibility, suggesting that greater party partisanship on the political left is related to reduced cognitive flexibility. Symmetrically, the slope between cognitive flexibility and political partisan identity on the right of the political spectrum is negative, indicating that greater political partisanship on the political right is also related to reduced cognitive flexibility on the RAT and AUT. In the case of the WCST, the results indicated that the relationship between flexibility and partisanship was present on the political left but not political right.

The Specificity of Flexibility

Additionally, since the AUT assesses multiple aspects of cog- nition, including flexibility, fluency, elaboration, and originality, we were able to investigate whether cognitive flexibility is specif- ically implicated in political partisanship relative to other facets of cognition. We conducted a multiple linear regression predicting political partisanship, as operationalized in PP| max |, with the four AUT measures as predictors. This regression was significant, F(4, 619) � 5.353, p � .0003, and revealed that cognitive flexibility was the only significant predictor out of the four measurements ( � �.166, t � �3.284, p � .001), and elaboration ( � �.043, p � .319), fluency ( � .054, p � .492), and originality ( � �.056, p � .440) were not related to political partisanship.

Bayes factor analysis corroborated these findings (see Figure S1 in the online supplemental materials). Computing the Bayesian

Table 1 Standardized Coefficients of Quadratic Regressions Predicting Cognitive Flexibility

Cognitive flexibility test RAT WCST AUT Flexibility

Political Partisanship PPmax Linear effect .00007 �.047 �.091�

Quadratic effect �.189��� �.112� �.186���

Age .149��� �.051 �.018 Gender �.076 �.005 .006 Educational attainment .060 .022 .022

BFql 2117.664 1.782 3712.614 BF Interpretation Decisive evidence Anecdotal evidence Decisive evidence

Note. BFql indicates the evidence in favor of a quadratic regression over a linear regression model. Interpre- tation reflects evidence categories for Bayes factors as suggested by Jeffreys (1961) and Wetzels and colleagues (2011). RAT � Remote Associates Test; WCST � Wisconsin Card Sorting Test; AUT � Alternative Uses Test. � p � .05. ��� p � .001.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

413POLITICAL PARTISANSHIP AND COGNITIVE INFLEXIBILITY

regressions and associated Bayes factors for all possible combina- tions of AUT submeasure predictors allows us to balance predic- tive power and parsimony (by removing redundant predictors). This analysis revealed that the best model is the one that predicts political partisanship only with cognitive flexibility, and no other AUT predictors. The Bayes factor for this model was BF10 � 1,100.362, indicating that the data are 1,100 times more likely under this model than an intercept-only null model (H0). Further- more, the flexibility-only model (subscript x) was over 100 times more likely than the full model (subscript f) with all AUT sub- measures for these data (BFxf � 109.551).

The Role of Political Conservatism

The analyses above demonstrate that the relationship between political partisanship and cognitive inflexibility is evident on both the political right and left, thus giving support for the rigidity-of- the-extreme hypothesis. The rigidity-of-the-right hypothesis also makes a prediction regarding political conservatism specifically: that there would be a linear relationship between conservatism and cognitive inflexibility. Correlational and Bayesian analyses offered inconclusive support for this hypothesis. Social conservatism was significantly correlated with AUT Flexibility (r � �.14, p � .001, BF10 � 3.09 [in favor of relationship]) but not with WCST (r � �.03, p � .567, BF01 � 2.34 [in favor of null]) or RAT (r � �.08, p � .055, but approached significance, BF01 � 0.91 [in favor of null]). Economic conservatism was not significantly cor- related with AUT Flexibility (r � �.07, p � .067, but approached significance, BF01 � 1.15 [in favor of null]), WCST (r � .04, p � .451, BF01 � 2.24 [in favor of null]), or RAT (r � �.03, p � .535, BF01 � 2.57 [in favor of null]).

To ensure that relationships were not attenuated due to the influence of covariates, we conducted a set of multiple hierarchical linear regressions, in which social and economic conservatism were simultaneously included as predictors of cognitive flexibility, while adjusting for the demographic covariates of age, gender, and educational attainment. Social conservatism, economic conserva- tism, and the demographic variables were therefore included in the first step of the hierarchical linear regression for each cognitive flexibility measure. In the second step, we tested whether the relationship demonstrated above between cognitive inflexibility and political partisanship (unweighted by partisan direction, as operationalized with the collapsed PP| max | measure) would persist after accounting for any relationships between cognitive inflexi- bility and political conservatism. Consequently, in the second step of the hierarchical regressions, we included PP| max | as a predictor of cognitive flexibility.

With respect to predicting flexibility on the RAT, social con- servatism was a significant negative predictor ( � �.170, t � �2.842, p � .005), while economic conservatism was not ( � .085, t � 1.452, p � .147) in the first step. In the second step,

Figure 3. Interrupted regression results according to cognitive flexibility test. Political partisanship is operationalized according to PPmax, such that partisan intensity is weighted by partisan direction (below 0 reflects left- leaning partisanship and above 0 reflects right-leaning partisanship). The shaded area reflects 95% confidence intervals.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

414 ZMIGROD, RENTFROW, AND ROBBINS

political partisanship was a significant negative predictor of flex- ibility ( � �.207, t � �4.665, p � .001) as was social conser- vatism ( � �.123, t � �2.076, p � .038). Economic conserva- tism was still not a significant predictor in the second step ( � .051, t � .884, p � .377). Furthermore, predicting flexibility on the AUT demonstrated that economic conservatism was not a signif- icant predictor in the first step ( � �.024, t � �.418, p � .676) and social conservatism approached significance ( � �.108, t � �1.892, p � .059). In the second step, political partisanship was a significant negative predictor ( � �.162, t � �3.762, p � .001) while neither economic conservatism ( � �.048, t � �.852, p � .395) or social conservatism ( � �.073, t � �1.273, p � .204) were statistically significant predictors. WCST was not linearly predicted by either social conservatism ( � �.086, t � �1.096, p � .274) or economic conservatism ( � .090, t � 1.178, p � .240) in the first step, and was also not predicted by political partisanship in the second step ( � �.057, t � �.996, p � .320). This may be due to the asymmetry evident in the interrupted regressions (Figure 3; Figure S2 in the online supplemental materials) whereby there is a significant relationship between WCST flexibility and partisanship with regards to the Democratic Party, but not the Republican Party.

Lastly, we also tested whether quadratic relationships between social conservatism and cognitive flexibility would better reflect the data than linear relationships. Bayesian analyses suggested that a linear relationship was either better than a quadratic relationship or there was insufficient data to conclude (WCST: BFlq � 3.433, AUT: BFlq � 1.64, RAT: BFlq � .501), after controlling for the demographic covariates. Notably, the rigidity-of-the-extreme hy- pothesis does not predict a curvilinear relationship with regards to conservatism specifically—it predicts that extreme ideologues or partisans would exhibit cognitive rigidity relative to moderates (as shown in the sections above). Consequently, we would not expect a curvilinear relationship with respect to conservatism, because it is not a measure that was designed to measure strength of ideol- ogy—rather it focuses on particular policy attitudes.

Discussion

The present investigation sought to address the question: Does mental rigidity reflect one’s partisan intensity or political orienta- tion? The results reveal that strong partisan intensity predicts reduced cognitive flexibility, regardless of the political party’s orientation and doctrine. Quadratic regressions revealed that strong partisan intensity, on both the political left and right, was related to reduced flexibility across all three tests of cognitive flexibility (see Table 1). This was corroborated by Bayes factor analysis, which demonstrated that the relationship between political partisanship across the political spectrum was quadratically—rather than lin- early—related to cognitive flexibility (see Table 1). Bayes factor analysis illustrated that the data were over 2,000 times more likely to occur under a quadratic model than a linear regression model for the RAT and AUT (see Table 1). The inverted U-shaped relation- ship between flexibility and partisanship was further validated with interrupted regressions (the two-lines test; Simonsohn, 2018; Figure 3), underscoring that lower cognitive flexibility is evident among strong political partisans of both liberal and conservative ideologies. This was further corroborated by the finding that par- ticipants who self-described as political Independents exhibited

heightened cognitive flexibility relative to Democrats and Repub- licans on the WCST and RAT (see Figure 2). Consequently, investigating the roots of partisan intensity uncovers important psychological similarities between adherents to opposing political ideologies.

Analysis of participants’ performance on the AUT, which mea- sures multiple aspects of cognition, highlighted that flexibility was specifically implicated as a psychological correlate of partisanship. Other cognitive traits, such as fluency, elaboration, or originality, were not significantly related to partisan intensity. Moreover, when political partisanship was regressed on all possible predictor combinations of the four AUT cognitive measures, the model consisting of cognitive flexibility as the only predictor was the best model (i.e., with the greatest evidential strength). The data were more than 1,000 times more likely to occur under a model pre- dicted only by cognitive flexibility than a null hypothesis model (Figure S1 in the online supplemental materials).

These results have implications for the two dominant hypoth- eses about the nature of mental flexibility and political ideol- ogy. To the best of our knowledge, these findings constitute the first direct objective testing of the ideological extremity hy- pothesis using behavioral assessments of cognitive flexibility rather than self-report questionnaires. The data here support the essential claim of the ideological extremity hypothesis: political extremists were more cognitively rigid than political moderates, across multiple tests of cognitive flexibility (Table 1, Figures 2 and 3). These results suggest that the rigidity-of-the-right hy- pothesis may be incomplete, as it does not account for the presence of the “rigidity-of-the-left.”

When partisan intensity and partisan direction (i.e., conserva- tism) were simultaneously entered into a linear multiple regres- sion, the relationship between partisan intensity and cognitive inflexibility persisted after controlling for conservatism. This adds to an emerging literature illustrating that political extremists across the political spectrum tend to possess and exhibit similar levels of dogmatism and prejudice (for a review, see Brandt, Reyna, Cham- bers, Crawford, & Wetherell, 2014), partisan bias and selective exposure to opposing opinions (Ditto et al., 2019; Frimer et al., 2017), moral motives (Frimer, Biesanz, Walker, & MacKinlay, 2013; Frimer, Tell, & Motyl, 2017), simplicity of political cate- gorization (Lammers et al., 2017), and belief in conspiracy theories (Krouwel, Kutiyski, van Prooijen, Martinsson, & Markstedt, 2017). Notably, social conservatism, but not economic conserva- tism, was a significant predictor of cognitive inflexibility, indicat- ing that there may still be ideological asymmetries that need to be empirically evaluated (Jost, 2017). Note, however, that the regres- sion coefficient of partisan intensity was consistently larger and more statistically significant than that of social conservatism in the multiple linear regression predicting cognitive flexibility (see Re- sults section). Partisan intensity, therefore, appears to be more predictive of mental flexibility than partisan direction. Economic conservatism was consistently statistically insignificant as a pre- dictor of flexibility. This offers nuanced directions for future research on the rigidity-of-the-right hypothesis—perhaps it is spe- cific to social, as opposed to economic, right-wing conservatism (see also Crowson, 2009; Feldman & Johnston, 2014; Malka & Soto, 2015). Moreover, this gives rise to the question: given the cognitive similarity between individuals on both political ex- tremes, what factors determine their political orientation? Future

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

415POLITICAL PARTISANSHIP AND COGNITIVE INFLEXIBILITY

studies should seek to replicate and expand these results, as well as explore ways in which the two hypotheses can be combined and empirically negotiated. Jost and colleagues (2003b) proposed a way in which the ideological extremity and rigidity-of-the-right hypotheses may be combined (see Figure 2 in their original paper), whereby there would be a U-shaped relationship between mental rigidity and political partisanship, but there would be a weaker relationship between partisan intensity and mental rigidity on the political left relative to the political right. This merits future examination.

The present investigation is relevant to other studies in political psychology that have sought to use indicators of cognitive pro- cessing that do not explicitly rely on self-reports. These studies have not directly assessed cognitive flexibility, instead focusing on other cognitive domains, such as confidence and metacognition (Brandt et al., 2015; Rollwage, Dolan, & Fleming, 2018), explor- atory behavior (Shook & Fazio, 2009), integrative complexity (Tetlock, Armor, & Peterson, 1994; Van Hiel & Mervielde, 2003), information evaluation (Ditto et al., 2019), political categorization tendencies (Lammers et al., 2017), and threat sensitivity (Hibbing et al., 2014). Moreover, many of these tasks inherently possess political content (e.g., Ditto et al., 2019; Tetlock et al., 1994; Lammers et al., 2017) and so are not reflections of ideologically neutral cognitive tendencies. Future research will benefit from examining multiple cognitive domains in tandem in order to eval- uate the relative contributions of these individual differences to political ideology and how they may interact (for a review, see van Prooijen & Krouwel, 2019). Furthermore, it is important to ac- knowledge the limitations of obtaining representative samples using online participant samples, such as Amazon Mechanical Turk, and so replicating these effects in nationally representative samples in the United States and other political systems will constitute a valuable validation.

In sum, the present findings signify that the way individuals process neutral stimuli and react to the environment can reveal how they process social and political information, and conse- quently how they form their ideological attitudes and political identities. Moreover, the findings highlight that investigating pro- cesses such as partisan intensity, attachment, and extremism across opposing ideological orientations may be as scientifically fruitful (if not more so) as studying the content of those ideologies. This is consistent with recent research across multiple studies demon- strating that objectively assessed cognitive inflexibility is related to greater ideological thinking in the realms of nationalism (Zmi- grod, Rentfrow, & Robbins, 2018), religiosity (Zmigrod, Rent- frow, Zmigrod, & Robbins, 2018), extremist attitudes (Zmigrod, Rentfrow, & Robbins, 2019), and general dogmatism (Zmigrod, Zmigrod, Rentfrow, & Robbins, 2019). The intensity and strictness of one’s adherence to an ideology, therefore, appears to be more relevant to—and revealing of— one’s mental flexibility than the content of one’s favored ideology. The results of this study, together with those across studies, suggest that the cognitively inflexible mind may be especially susceptible to the clarity, cer- tainty, and safety frequently offered by strong loyalty to collective ideologies and doctrines, regardless of their subject matter and motivation. This is in line with Rokeach’s (1954) argument that adherents of both extreme left-wing and right-wing ideologies would exhibit tendencies toward rigidity. The cognitively flexible mind may be more likely to formulate sociopolitical opinions in a

way that does not fully conform with the particular constellation of beliefs advocated by a political party. These findings nicely echo Hoffer’s early theoretical writings which suggested that “all move- ments, however different in doctrine and aspiration . . . all appeal to the same types of mind” (Hoffer, 1951).

References

Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian personality. New York, NY: Harper Brothers.

Altemeyer, B. (1998). The other “authoritarian personality”. Advances in Experimental Social Psychology, 30. 47–92. http://dx.doi.org/10.1016/ S0065-2601(08)60382-2

Anderson, S. W., Damasio, H., Jones, R. D., & Tranel, D. (1991). Wis- consin Card Sorting Test performance as a measure of frontal lobe damage. Journal of Clinical and Experimental Neuropsychology, 13, 909 –922. http://dx.doi.org/10.1080/01688639108405107

Barnett, J. H., Jones, P. B., Robbins, T. W., & Müller, U. (2007). Effects of the catechol-O-methyltransferase Val158Met polymorphism on exec- utive function: A meta-analysis of the Wisconsin Card Sort Test in schizophrenia and healthy controls. Molecular Psychiatry, 12, 502–509. http://dx.doi.org/10.1038/sj.mp.4001973

Bowden, E. M., & Jung-Beeman, M. (2003). Normative data for 144 compound remote associate problems. Behavior Research Methods, Instruments, & Computers, 35, 634 – 639. http://dx.doi.org/10.3758/ BF03195543

Brandt, M. J., Evans, A. M., & Crawford, J. T. (2015). The unthinking or confident extremist? Political extremists are more likely than moderates to reject experimenter-generated anchors. Psychological Science, 26, 189 –202. http://dx.doi.org/10.1177/0956797614559730

Brandt, M. J., Reyna, C., Chambers, J. R., Crawford, J. T., & Wetherell, G. (2014). The ideological-conflict hypothesis: Intolerance among both liberals and conservatives. Current Directions in Psychological Science, 23, 27–34. http://dx.doi.org/10.1177/0963721413510932

Braver, T. S., Cole, M. W., & Yarkoni, T. (2010). Vive les differences! Individual variation in neural mechanisms of executive control. Current Opinion in Neurobiology, 20, 242–250. http://dx.doi.org/10.1016/j.conb .2010.03.002

Breheny, P., & Burchett, W. (2017). Visualization of regression models using visreg. The R Journal, 9, 56 –71.

Chamberlain, S. R., Müller, U., Blackwell, A. D., Clark, L., Robbins, T. W., & Sahakian, B. J. (2006). Neurochemical modulation of response inhibition and probabilistic learning in humans. Science, 311, 861– 863. http://dx.doi.org/10.1126/science.1121218

Chermahini, S. A., & Hommel, B. (2010). The (b)link between creativity and dopamine: Spontaneous eye blink rates predict and dissociate di- vergent and convergent thinking. Cognition, 115, 458 – 465. http://dx.doi .org/10.1016/j.cognition.2010.03.007

Conway, L. G., III, Houck, S. C., Gornick, L. J., & Repke, M. A. (2018). Finding the Loch Ness monster: Left-wing authoritarianism in the United States. Political Psychology, 39, 1049 –1067. http://dx.doi.org/ 10.1111/pops.12470

Conway, L. G., III, & McFarland, J. D. (2019). Do right-wing and left- wing authoritarianism predict election outcomes?: Support for Obama and Trump across two United States presidential elections. Personality and Individual Differences, 138, 84 – 87. http://dx.doi.org/10.1016/j.paid .2018.09.033

Cools, R., & Robbins, T. W. (2004). Chemistry of the adaptive mind. Philosophical Transactions of the Royal Society of London A: Mathe- matical, Physical and Engineering Sciences, 362, 2871–2888.

Crawford, J. (2019). Replication of Fernbach, Rogers, Fox, & Sloman (2013). Retrieved from http://dx.doi.org/10.31234/osf.io/yn64p

Crowson, H. M. (2009). Are all conservatives alike? A study of the psychological correlates of cultural and economic conservatism. The

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

416 ZMIGROD, RENTFROW, AND ROBBINS

Journal of Psychology: Interdisciplinary and Applied, 143, 449 – 463. http://dx.doi.org/10.3200/JRL.143.5.449-463

Diamond, A. (2006). The early development of executive functions. Lifes- pan Cognition: Mechanisms of Change, 210, 70 –95.

Ditto, P. H., Liu, B. S., Clark, C. J., Wojcik, S. P., Chen, E. E., Grady, R. H., . . . Zinger, J. F. (2019). At least bias is bipartisan: A meta-analytic comparison of partisan bias in liberals and conservatives. Perspectives on Psychological Science, 14, 273–291. http://dx.doi.org/10.1177/ 1745691617746796

Everett, J. A. (2013). The 12 item Social and Economic Conservatism Scale (SECS). PLoS ONE, 8(12), e82131. http://dx.doi.org/10.1371/ journal.pone.0082131

Feldman, S., & Johnston, C. (2014). Understanding the determinants of political ideology: Implications of structural complexity. Political Psy- chology, 35, 337–358. http://dx.doi.org/10.1111/pops.12055

Fernbach, P. M., Rogers, T., Fox, C. R., & Sloman, S. A. (2013). Political extremism is supported by an illusion of understanding. Psychological Science, 24, 939 –946. http://dx.doi.org/10.1177/0956797612464058

Frimer, J. A., Biesanz, J. C., Walker, L. J., & MacKinlay, C. W. (2013). Liberals and conservatives rely on common moral foundations when making moral judgments about influential people. Journal of Personal- ity and Social Psychology, 104, 1040 –1059. http://dx.doi.org/10.1037/ a0032277

Frimer, J. A., Skitka, L. J., & Motyl, M. (2017). Liberals and conservatives are similarly motivated to avoid exposure to one another’s opinions. Journal of Experimental Social Psychology, 72, 1–12. http://dx.doi.org/ 10.1016/j.jesp.2017.04.003

Frimer, J. A., Tell, C. E., & Motyl, M. (2017). Sacralizing liberals and fair-minded conservatives: Ideological symmetry in the moral motives in the culture war. Analyses of Social Issues and Public Policy, 17, 33–59. http://dx.doi.org/10.1111/asap.12127

Gough, H. G., & Sanford, R. N. (1952). Rigidity as a psychological variable. Unpublished manuscript, University of California Institute of Personality Assessment and Research, Berkeley, CA.

Grant, D. A., & Berg, E. A. (1948). A behavioral analysis of degree of reinforcement and ease of shifting to new responses in a Weigl-type card-sorting problem. Journal of Experimental Psychology, 38, 404 – 411. http://dx.doi.org/10.1037/h0059831

Greenberg, J., & Jonas, E. (2003). Psychological motives and political orientation—The left, the right, and the rigid: Comment on Jost et al. (2003). Psychological Bulletin, 129, 376 –382.

Guilford, J. P. (1967). The nature of human intelligence. New York, NY: McGraw-Hill.

Heaton, R. K. (1981). A manual for the Wisconsin Card Sorting Test. Odessa, FL: Western Psychological Services.

Hibbing, J. R., Smith, K. B., & Alford, J. R. (2014). Differences in negativity bias underlie variations in political ideology. Behavioral and Brain Sciences, 37, 297–307. http://dx.doi.org/10.1017/S0140525 X13001192

Hoffer, E. (1951). The true believer: Thoughts on the nature of movements. New York, NY: HarperCollins.

Ionescu, T. (2012). Exploring the nature of cognitive flexibility. New Ideas in Psychology, 30, 190 –200. http://dx.doi.org/10.1016/j.newideapsych .2011.11.001

Jeffreys, H. (1961). Theory of probability (3rd ed.). New York, NY: Oxford University Press.

Jiménez, J., Gómez, A., Buhrmester, M. D., Vázquez, A., Whitehouse, H., & Swann, W. B. (2016). The dynamic identity fusion index: A new continuous measure of identity fusion for web-based questionnaires. Social Science Computer Review, 34, 215–228. http://dx.doi.org/10 .1177/0894439314566178

Jong, J., Whitehouse, H., Kavanagh, C., & Lane, J. (2015). Shared negative experiences lead to identity fusion via personal reflection. PLoS ONE, 10(12), e0145611. http://dx.doi.org/10.1371/journal.pone.0145611

Jost, J. T. (2017). Ideological asymmetries and the essence of political psychology. Political Psychology, 38, 167–208. http://dx.doi.org/10 .1111/pops.12407

Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003a). Political conservatism as motivated social cognition. Psychological Bul- letin, 129, 339 –375. http://dx.doi.org/10.1037/0033-2909.129.3.339

Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003b). Exceptions that prove the rule—Using a theory of motivated social cognition to account for ideological incongruities and political anoma- lies: Reply to Greenberg and Jonas (2003). Psychological Bulletin, 129, 383–393. http://dx.doi.org/10.1037/0033-2909.129.3.383

Kapitány, R., Kavanagh, C., Buhrmester, M. D., Newson, M., & White- house, H. (2019). Ritual, identity fusion, and the inauguration of presi- dent Trump: A pseudo-experiment of ritual modes theory. Self and Identity. Advance online publication. http://dx.doi.org/10.1080/ 15298868.2019.1578686

Kidd, A. H., & Kidd, R. M. (1972). Relation of F-test scores to rigidity. Perceptual and Motor Skills, 34, 239 –243. http://dx.doi.org/10.2466/ pms.1972.34.1.239

Kirton, M. J. (1981). A reanalysis of two scales of tolerance of ambiguity. Journal of Personality Assessment, 45, 407– 414. http://dx.doi.org/10 .1207/s15327752jpa4504_10

Krouwel, A., Kutiyski, Y., van Prooijen, J. W., Martinsson, J., & Markst- edt, E. (2017). Does extreme political ideology predict conspiracy be- liefs, economic evaluations and political trust? Evidence from Sweden. Journal of Social and Political Psychology, 5, 435– 462. http://dx.doi .org/10.5964/jspp.v5i2.745

Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one’s own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77, 1121–1134.

Lammers, J., Koch, A., Conway, P., & Brandt, M. J. (2017). The political domain appears simpler to the politically extreme than to political moderates. Social Psychological & Personality Science, 8, 612– 622. http://dx.doi.org/10.1177/1948550616678456

Levitt, E. E. (1956). The water-jar Einstellung test as a measure of rigidity. Psychological Bulletin, 53, 347–370. http://dx.doi.org/10.1037/ h0049245

Liang, F., Paulo, R., Molina, G., Clyde, M. A., & Berger, J. O. (2008). Mixtures of g priors for Bayesian variable selection. Journal of the American Statistical Association, 103, 410 – 423. http://dx.doi.org/10 .1198/016214507000001337

Long, J. A. (2019). jtools: Analysis and presentation of social scientific data (R package version 2.0.1) [Computer software]. Retrieved from https://cran.r-project.org/package�jtools

Madore, K. P., Addis, D. R., & Schacter, D. L. (2015). Creativity and memory: Effects of an episodic-specificity induction on divergent think- ing. Psychological Science, 26, 1461–1468. http://dx.doi.org/10.1177/ 0956797615591863

Malka, A., Lelkes, Y., & Holzer, N. (2017). Rethinking the rigidity of the right model: Three suboptimal methodological practices and their im- plications. In J. T. Crawford & L. Jussim (Eds.), The politics of social psychology (pp. 116 –135). New York, NY: Routledge.

Malka, A., & Soto, C. J. (2015). Rigidity of the economic right? Menu- independent and menu-dependent influences of psychological disposi- tions on political attitudes. Current Directions in Psychological Science, 24, 137–142. http://dx.doi.org/10.1177/0963721414556340

Mednick, S. A. (1962). The associative basis of the creative process. Psychological Review, 69, 220 –232. http://dx.doi.org/10.1037/h00 48850

Misch, A., Fergusson, G., & Dunham, Y. (2018). Temporal dynamics of partisan identity fusion and prosociality during the 2016 U.S. presiden- tial election. Self and Identity, 17, 531–548. http://dx.doi.org/10.1080/ 15298868.2018.1430063

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

417POLITICAL PARTISANSHIP AND COGNITIVE INFLEXIBILITY

Miyake, A., & Friedman, N. P. (2012). The nature and organization of individual differences in executive functions: Four general conclusions. Current Directions in Psychological Science, 21, 8 –14.

Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7, 134 –140. http://dx.doi.org/10.1016/S1364-6613(03)00028-7

Morey, R. D., Rouder, J. N., Jamil, T., & Morey, M. R. D. (2015). Package ‘BayesFactor’: Computation of Bayes Factors for Common Designs (R Package Version 09.12–4.2) [Computer Software]. Retrieved from https:// cran.r-project.org/web/packages/BayesFactor/BayesFactor.pdf

Neuringer, C. (1964). The relationship between authoritarianism, rigidity, and anxiety. The Journal of General Psychology, 71, 169 –175. http:// dx.doi.org/10.1080/00221309.1964.9710303

Patil, I., & Powell, C. (2018). Ggstatsplot:‘Ggplot2=based plots with sta- tistical details [Computer software]. Retrieved from https://CRAN.R- project.org/package�ggstatsplot

Roberts, R. P., Wiebels, K., Sumner, R. L., van Mulukom, V., Grady, C. L., Schacter, D. L., & Addis, D. R. (2017). An fMRI investigation of the relationship between future imagination and cognitive flexibility. Neu- ropsychologia, 95, 156 –172. http://dx.doi.org/10.1016/j.neuropsycholo- gia.2016.11.019

Rokeach, M. (1954). The nature and meaning of dogmatism. Psychological Review, 61, 194 –204. http://dx.doi.org/10.1037/h0060752

Rollwage, M., Dolan, R. J., & Fleming, S. M. (2018). Metacognitive failure as a feature of those holding radical beliefs. Current Biology, 28, 4014 – 4021.

Rouder, J. N., & Morey, R. D. (2012). Default Bayes factors for model selection in regression. Multivariate Behavioral Research, 47, 877–903. http://dx.doi.org/10.1080/00273171.2012.734737

Schultz, P. W., & Searleman, A. (2002). Rigidity of thought and behavior: 100 years of research. Genetic, Social, and General Psychology Mono- graphs, 128, 165–207.

Settle, J. E., Dawes, C. T., & Fowler, J. H. (2009). The heritability of partisan attachment. Political Research Quarterly, 62, 601– 613. http:// dx.doi.org/10.1177/1065912908327607

Shafto, P., Coley, J. D., & Vitkin, A. (2007). Availability in category-based induction. In A. Feeney & E. Heit (Eds.), Inductive reasoning: Experi- mental, developmental, and computational approaches (pp. 114 –136). New York, NY, Cambridge University Press. http://dx.doi.org/10.1017/ CBO9780511619304.006

Shook, N. J., & Fazio, R. H. (2009). Political ideology, exploration of novel stimuli, and attitude formation. Journal of Experimental Social Psychology, 45, 995–998.

Sidanius, J. (1985). Cognitive functioning and sociopolitical ideology revisited. Political Psychology, 6, 637– 661. http://dx.doi.org/10.2307/ 3791021

Simonsohn, U. (2018). Two-lines: A valid alternative to the invalid testing of U-shaped relationships with quadratic regressions. Advances in Meth- ods and Practices in Psychological Science, 1, 538 –555. http://dx.doi .org/10.1177/2515245918805755

Steiner, I. D., & Johnson, H. H. (1963). Authoritarianism and “tolerance of trait inconsistency”. Journal of Abnormal Psychology, 67, 388 –391. http://dx.doi.org/10.1037/h0048897

Tetlock, P. E., Armor, D., & Peterson, R. S. (1994). The slavery debate in antebellum America: Cognitive style, value conflict, and the limits of compromise. Journal of Personality and Social Psychology, 66, 115– 126.

Toner, K., Leary, M. R., Asher, M. W., & Jongman-Sereno, K. P. (2013). Feeling superior is a bipartisan issue: Extremity (not direction) of political views predicts perceived belief superiority. Psychological Sci- ence, 24, 2454 –2462.

Van Bavel, J. J., & Pereira, A. (2018). The partisan brain: An identity- based model of political belief. Trends in Cognitive Sciences, 22, 213– 224. http://dx.doi.org/10.1016/j.tics.2018.01.004

Van Hiel, A., & Mervielde, I. (2003). The measurement of cognitive complexity and its relationship with political extremism. Political Psy- chology, 24, 781– 801.

Van Hiel, A., Onraet, E., Crowson, H. M., & Roets, A. (2016). The relationship between right-wing attitudes and cognitive style: A com- parison of self-report and behavioural measures of rigidity and intoler- ance of ambiguity. European Journal of Personality, 30, 523–531. http://dx.doi.org/10.1002/per.2082

Van Hiel, A., Onraet, E., & De Pauw, S. (2010). The relationship between social-cultural attitudes and behavioral measures of cognitive style: A meta-analytic integration of studies. Journal of Personality, 78, 1765– 1800. http://dx.doi.org/10.1111/j.1467-6494.2010.00669.x

van Prooijen, J. W., & Krouwel, A. P. (2017). Extreme political beliefs predict dogmatic intolerance. Social Psychological & Personality Sci- ence, 8, 292–300. http://dx.doi.org/10.1177/1948550616671403

van Prooijen, J. W., Krouwel, A. P., Boiten, M., & Eendebak, L. (2015). Fear among the extremes: How political ideology predicts negative emotions and outgroup derogation. Personality and Social Psychology Bulletin, 41, 485– 497. http://dx.doi.org/10.1177/0146167215569706

van Prooijen, J. W., & Krouwel, A. P. M. (2019). Psychological features of extreme political ideologies. Current Directions in Psychological Sci- ence. http://dx.doi.org/10.1177/0963721418817755

van Prooijen, J. W., Krouwel, A. P. M., & Emmer, J. (2018). Ideological responses to the EU refugee crisis: The left, the right, and the extremes. Social Psychological & Personality Science, 9, 143–150. http://dx.doi .org/10.1177/1948550617731501

van Prooijen, J. W., Krouwel, A. P. M., & Pollet, T. V. (2015). Political extremism predicts belief in conspiracy theories. Social Psychological & Personality Science, 6, 570 –578. http://dx.doi.org/10.1177/1948 550614567356

Verdejo-García, A., Pérez-García, M., & Bechara, A. (2006). Emotion, decision making and substance dependence: A somatic-marker model of addiction. Current Neuropharmacology, 4, 17–31. http://dx.doi.org/10 .2174/157015906775203057

Wetzels, R., Matzke, D., Lee, M. D., Rouder, J. N., Iverson, G. J., & Wagenmakers, E. J. (2011). Statistical evidence in experimental psy- chology: An empirical comparison using 855 t tests. Perspectives on Psychological Science, 6, 291–298. http://dx.doi.org/10.1177/174 5691611406923

Zmigrod, L., Rentfrow, P. J., & Robbins, T. W. (2018). Cognitive under- pinnings of nationalistic ideology in the context of Brexit. Proceedings of the National Academy of Sciences of the United States of America, 115, E4532–E4540. http://dx.doi.org/10.1073/pnas.1708960115

Zmigrod, L., Rentfrow, P. J., & Robbins, T. W. (2019). Cognitive inflex- ibility predicts extremist attitudes. Frontiers in Psychology, 10, 989. http://dx.doi.org/10.3389/fpsyg.2019.00989

Zmigrod, L., Rentfrow, P. J., Zmigrod, S., & Robbins, T. W. (2018). Cognitive flexibility and religious disbelief. Psychological Research. Advance online publication. http://dx.doi.org/10.1007/s00426-018- 1034-3

Zmigrod, L., Zmigrod, S., Rentfrow, P. J., & Robbins, T. W. (2019). The psychological roots of intellectual humility: The role of intelligence and cognitive flexibility. Personality and Individual Differences, 141, 200 – 208. http://dx.doi.org/10.1016/j.paid.2019.01.016

Received December 2, 2018 Revision received June 18, 2019

Accepted June 22, 2019 �

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

418 ZMIGROD, RENTFROW, AND ROBBINS