This is an open access article distributed under the terms of the Creative Commons
Attribution License (Political scientists and psychologists have long debated the nature of political ideology and the relationship between political identification and associated attitudes and beliefs. Fifty years ago, Converse (1964) suggested that a majority of Americans did not have a coherent, constrained political belief system, arguing that ideology only truly existed among political elites. In the years that followed, scholars debated whether ideology is a useful concept for understanding the political preferences and beliefs of the mass public. In large part, this debate hinged on whether or not political knowledge and consistency were essential components of ideology, as research on political knowledge has shown that the American public lacks a coherent understanding of politics (e.g., Delli Carpini & Keeter, 1996) and that the public is often misinformed about important issues (e.g., Kuklinski, Quirk, Jerit, Schwieder, & Rich, 2000). Over time, this debate has shifted from whether or not people have formal ideologies, toward understanding what organizes individuals’ idiosyncratic worldviews, and a desire to understand the basic social cognitive processes that give rise to political decisions.
From a psychological perspective, the ideological character of policies should be relevant to the extent that it relates to one’s identity, though that perceived ideological character need not be objectively correct. The present research addresses the extent to which people use ideological categories when evaluating policy issues by examining the underlying cognitive processes. Consistent with prior work in political psychology, we argue that ideology can function as a useful schema that guides political decision making, especially for individuals who are more politically sophisticated or highly identified (Fuchs & Klingemann, 1990; Hamill, Lodge, & Blake, 1985). Political self-identification rather than adherence to doctrine characterizes popular ideology and explains when (and for whom) ideological categorization influences policy evaluation (Devine, 2015; Green, Palmquist, & Schickler, 2002; Greene, 1999; Malka & Lelkes, 2010).
To understand the psychology of ideology, it must be recognized as more than a label for a formal political philosophy. As Jost (2006) has suggested, ideology is inherently social in that it is tied to our social groups, and it helps to organize and motivate behavior. If we remove the requirement that ideology be defined as adherence to a formal belief system requiring high political knowledge, then we can consider an alternative—that ideology, defined as an organizing principle for social behavior and judgment, typically reflects social identity (Devine, 2015; Green et al., 2002; Greene, 1999; Malka & Lelkes, 2010). This approach to ideology may lead to a broader theoretical understanding of how ideology impacts the mass public’s policy evaluations.
The relation between ideology and political attitudes is dynamic. Sometimes ideological self-identification is primary. Through some combination of political socialization and biological influence (see e.g., Hibbing, Smith, Peterson, & Feher, 2014), individuals come to identify as liberals or conservatives and derive a sense of meaning and self-esteem from these identities (see Brewer, 2001). This allows political parties or ideological groups to serve an agenda setting function, suggesting to group members which issues they should support or oppose (Zaller, 1992). People do sometimes adopt the beliefs purportedly held by groups they identify with, regardless of the content of those beliefs and even if the beliefs run counter to what they previously believed (Cohen, 2003).i Individuals may also first develop specific policy attitudes, perhaps related to values and personal experience, and then come to identify with groups that share those beliefs. If people hold a strong belief, and then find that others in a group share that belief, identification with that group is strengthened (Huddy, 2003). Regardless of its genesis, the intertwining of self-identification and policy attitudes has important implications for political cognition.
Ideology can be thought of as a social category, while a specific policy issue is an item related to that category. From this perspective, multiple aspects of the underlying knowledge structure are likely to be important when considering how people will use this social category information to inform their policy attitudes (Fuchs & Klingemann, 1990; Hamill et al., 1985). Namely, it is important to consider how accessible these concepts are in memory (i.e., how quickly or easily they can be retrieved), as well as the overall strength of the relationship between them. If social category information (i.e., ideology, partisanship) is easily accessible, people should be more likely to automatically activate that information and use it to make political judgments. Accessibility is one of several known indicators (e.g., extremity, certainty) of attitude strength—the tendency for attitudes to be stable and impact judgment and behavior (Petty & Krosnick, 1995). During the 1984 U.S. Presidential election, for example, more accessible attitudes about the candidates, Walter Mondale and Ronald Reagan, were more predictive of both perception of debate performance and voting behavior (Fazio & Williams, 1986). There is evidence that accessibility also matters when examining the link between social categories and attitudes. Research conducted by Huckfeldt and colleagues (1999) showed that more accessible political self-identifications were more predictive of political attitudes. They found that partisanship was more strongly related to evaluations of political figures than ideology, while ideology was more predictive of support for ideologically-relevant issues than party. If more accessible self-identifications are more likely to influence attitudes, then it should also be the case that priming or activating these identifications should increase the degree to which people use or rely on the category information to make decisions. Further, construing policy positions in terms of such categories should lead to their use in the evaluation of those policy positions when those categories correspond to individuals’ social identities.
Overview of Present Research
Research in political psychology on the relationship between ideology and attitudes has begun to incorporate measurement of accessibility (e.g., Huckfeldt et al., 1999), most often in correlational or survey-based designs. While these approaches can demonstrate a relationship between accessible categories and related attitudes, without using an experimental approach it is impossible to conclude that category accessibility is actually causing a change in the processes underlying attitude expression. In a laboratory setting, the role of accessibility can be examined experimentally through use of a priming task or a task facilitation paradigm (see Fazio, Williams, & Powell, 2000, for discussion of accessibility measures in political psychology research). In these research designs, paired judgments are typically employed in which the extent to which the first judgment facilitates the second provides evidence for the usage of information from the first judgment while making the second judgment.
In the present research, we examine the influence of ideological categories on policy evaluation by experimentally manipulating whether or not those categories are activated prior to policy evaluation. If ideology is closely linked to policy attitudes, then activating the ideological category related to a given policy should facilitate (i.e., speed up) evaluative responses relative to a control condition where ideology is not activated. This facilitation effect should only occur when the link between the ideological category and the individual’s policy attitude is relatively strong, and unencumbered by ambivalence or conflict. When individuals hold ambivalent attitudes about policy issues, this link between category and attitude may be undermined (Huckfeldt & Sprague, 2000). So, we expect that facilitation should only occur when participants are making ideologically consistent judgments (e.g., a liberal participant agreeing with a liberal policy), as opposed to inconsistent judgments. Any amount of conflict or inconsistency should slow responses, and may actually have the opposite effect—leading to response interference. However, we should note that an interference effect is also consistent with the view that individuals are utilizing ideological categorization in their evaluation of policy positions, such that departing from the ideological group gives pause.
We postulate that these facilitation and interference effects should be stronger or weaker for individuals depending on their own political identification. More specifically, we expect that accessibility of self-identification, as well as its extremity, should increase the degree to which people rely on these categories in everyday life, and thus, the links between ideological categories and specific policy positions. For these reasons, we expect that these strength-related variables (i.e., accessibility, extremity) regarding self-identification will moderate the influence of our priming manipulation, increasing the facilitation and interference effects. We examine this in three experiments, investigating the role of accessibility of ideological self-identification (Experiment 1), extremity of ideological self-identification (Experiment 2), and extremity of partisan identification (Experiment 3).
Experiment 1
In Experiment 1, we set out to examine the hypothesis that activating social category information (i.e., political ideology) should influence evaluation of policy issues. Importantly, we expect that facilitation should only occur when participants are making responses consistent with their own ideological self-identification. Inconsistent responses, by contrast, should be more likely to lead to an interference effect. We also examined the moderating role of self-identification accessibility, expecting that individuals who were faster to identify as liberal or conservative (indicating stronger political identification) should show stronger facilitation and interference effects. In other words, immediately after categorizing the statement “Same sex marriage should be illegal” as “conservative” in nature, individuals may be faster or slower to express their agreement or disagreement with the statement. The extent to which any such facilitation or interference occurs, however, should depend on how strongly the individual’s own identity corresponds to the classification of the policy statement’s ideological stance. Hence, an individual who strongly identifies as a conservative should express agreement more quickly than an individual who weakly identifies as conservative. Likewise, an individual who strongly identifies as liberal should express disagreement more quickly relative to one who weakly identifies as liberal. In Experiment 1, participants completed a political identification task designed to measure accessibility, followed by a task facilitation paradigm in which they made paired sets of judgments about political policy statements.
Method
Participants in Experiment 1 were 59 undergraduate students at The Ohio State University in the United States. The data from four participants were removed prior to analysis—one due to technical issues, one who admitted to falling asleep during the study, and two due to a coding error that made their responses impossible to separate (the same participant identification code was used for both). This left 55 participants for analysis (25 female, 29 male; age range: 18-21, M = 18.74, SD = .85). The target sample size was 60 participants, based on the fact that our primary analyses relied on a within-subject design, such that statistical power is enhanced as a function of the number of trials (as opposed to just the number of participants). Upon arrival at the lab, participants first completed a political identification task, followed by the task facilitation paradigm. All experimental stimuli were presented on individual computers using MediaLab and DirectRT software (Jarvis, 2004a, 2004b).
Political Identification Task
Participants expressed their identification with a variety of groups by answering yes or no to the following question stem: “Do you identify as a member of the group XX?” The target stem completions were related to political identification, including the words liberal and conservative. Participants also completed some filler items consisting of non-political groups they may or may not belong to (e.g., fraternity, sorority, psychology major). Item order was randomized for each participant. Participants were instructed to respond quickly and response latencies were recorded. This data was used to categorize participants as liberal or conservative, and to calculate the accessibility of political identification in terms of response latency.
Participants were categorized as liberal, conservative, or neither on the basis of the identification task. Only participants who showed consistent responses were categorized as liberal or conservative and included in the primary analyses. Liberal participants were those who responded yes to liberal and no to conservative (n = 16), and conservative participants were those who responded yes to conservative and no to liberal (n = 17). All other participants either said yes to both items or no to both and were categorized as neither (n = 21). Ideology was coded as -1 = liberal, 0 = neither, and 1 = conservative. To index the accessibility of ideological self-identification, we used the relevant response latencies. So, to discriminate among those who identified as liberal, we examined the speed with which they responded “yes” to the liberal question. To control for individual differences in baseline speed of responding, we subtracted mean response latency for each participant on the control self-identification items from the target response latency for the liberal item. We followed the same procedure to discriminate among those who identified as conservative, with the speed with which they responded “yes” to the conservative question serving as the target latency.ii To account for differences in accessibility between liberal and conservative participants, these variables were independently standardized such that we created separate Z-scores for liberal accessibility and conservative accessibility and then combined those into a single variable representing self-identification accessibility. Negative values indicate that participants were faster to identify with that ideology (relative to control items), whereas positive values indicate that participants were slower to identify with that ideology (again, relative to control items).
Task Facilitation Paradigm
The task facilitation paradigm was modeled after procedures used in prior research in social and cognitive psychology (e.g., Collins & Quillian, 1970; Klein, Loftus, & Burton, 1989; Van Bavel, Packer, Haas, & Cunningham, 2012). In this paradigm, participants are presented with paired sets of judgments, allowing for measurement of the extent to which two tasks or judgments rely on the same information. Facilitation (and interference) can be examined by manipulating the information presented in the first task, and measuring response latency on the second task. For example, Van Bavel and colleagues (2012) used a task facilitation paradigm to show that moral versus pragmatic action evaluations were more likely to facilitate judgments of universality, given that morality tends to be associated more strongly with normative preferences. Klein and colleagues (1989) used a similar paradigm to examine the extent to which traits overlap with autobiographical memories.
The task used in this experiment consisted of three trial types: experimental trials, control trials, and filler trials. Each trial type consisted of a paired set of judgments about the same stimulus (i.e., political policy statements), each including both an initial trial and a target trial. In other words, participants made two separate judgments about each policy issue, in quick succession. On experimental trials (n = 16), participants first categorized the policy statement as liberal or conservative (ideology), then indicated whether they agreed or disagreed (evaluation) with the statement. On control trials (n = 16), participants first categorized the statement as either growing or declining in importance (Control Condition 1) or national versus global in scope (Control Condition 2), and then indicated their agreement or disagreement with the statement. Filler trials (n = 88) consisted of all other possible combinations of these judgment types, so that participants could not easily discern the purpose of the experiment. For example, some trials had participants first indicate whether an issues was national or global in scope (Control Condition 2), and then categorize the issue as liberal or conservative (ideology). The inclusion of filler trials meant that participants were unable to anticipate the type of decision they would be asked next at any given time—the only indication about the decision type came from the cue that was presented prior to the decision. Trial order was also fully randomized for each participant, such that participants had no expectation about which trial type would come next.
On each trial (see Figure 1), response labels for the initial task appeared first for 1000 milliseconds (e.g., participants were instructed to press the F key for a liberal statement, or the J key for a conservative statement). Next, the policy statement appeared on screen until the participant responded. Following the response, the response labels disappeared and there was another 1000 ms delay before the second target task labels appeared (e.g., F = agree, J = disagree). The policy statement remained on screen during this delay. Once the target response labels appeared, the program again waited for the participant’s response. After the participant responded, a fixation string (*****) appeared on screen for 2000 ms to indicate a break between trials.
Figure 1
Example trial from Experiments 1 and 2. Trials were similar in Experiment 3 but the labels liberal and conservative were replaced with Democrat and Republican.
Policy Statements
The policy statements used in these experiments were drawn from a separate pilot study. Data was collected from 171 undergraduate students at The Ohio State University in a laboratory setting using a survey (non-experimental) design. The mean age of the sample was 19.2 (SD = 2.22) and the sample was approximately 70% female. Policy statements were selected from this pilot data based on the strength of relationship between the policy statement and political ideology. Sample policy statements included: “Same sex marriage should be illegal” and “Abortion should be a legal right.” Policy statements were divided into two roughly equivalent sets (Set A and Set B; see Appendix). Selected policy statements showed statistically significant (p < .05) correlations with political ideology and items in Set A and Set B had equivalent mean correlations with ideology (Set A: M = .257, SD = .070; Set B: M = .260, SD = .067; t(30) = -.116, p = .908). Set A and Set B showed similar mean response latencies (Set A: M = 5112 ms, SD = 1547; Set B: M = 5842 ms, SD = 1754; t(30) = -1.249, p = .221) in pilot data. Assignment of set (A/B) to trials (experimental/control) was counterbalanced between subjects. Filler trials utilized additional policy statements that were not used for experimental or control trials. In order to avoid repetition, we minimized similarity across statements.
Coding of Task Facilitation Paradigm
The experimental trials were coded as consistent or inconsistent on the basis of participant ideology and evaluative judgment (see Table 1). For a liberal participant, consistent trials were those categorized as liberal and subsequently agreed with, or trials categorized as conservative and subsequently disagreed with. Inconsistent trials were those where they indicated conservative and agreed, or indicated liberal and disagreed.iii For conservative participants, these were reversed. Raw latency data was trimmed to remove trials where participants responded very quickly (faster than 300 ms) or very slowly (slower than 15000 ms; removing 2% of trials) and then log transformed to correct for skewness (see Fazio, 1990).
Table 1
Coding of Consistent and Inconsistent Trials in the Task Facilitation Paradigm
| Participant Ideology | Response Type
|
|
|---|---|---|
| Consistent Trials | Inconsistent Trials | |
| Liberal | ||
| Liberal-Agree | Liberal-Disagree | |
| Conservative-Disagree | Conservative-Agree | |
| Conservative | ||
| Liberal-Disagree | Liberal-Agree | |
| Conservative-Agree | Conservative-Disagree | |
Results
Natural log transformed response latencies were subjected to a repeated measures ANOVA across trials where trial type (ideology-consistent, ideology-inconsistent, control) and participant were modeled as within-subject factors. There was a significant difference in response latency as a function of condition, F(2,64) = 6.36, p = .003, η2p = .17; see Figure 2. Mean latency was significantly faster for ideology-consistent trials, M = 1679 milliseconds, SE = 107 ms, than for ideology-inconsistent trials, M = 2151, SE = 189, F(1,32) = 11.3, p = .002, η2p = .26. Latency on ideology-consistent trials was also significantly faster than latency on control trials, M = 1944, SE = 93, F(1,32) = 6.18, p = .018, η2p = .16, and latency on ideology-inconsistent trials was marginally slower than latency on control trials, F(1,32) = 3.384, p = .075, η2p = .10.iv This pattern supports the overall expectation that relative to a control condition, consistent responses should be facilitated, while inconsistent responses are interfered with.
Figure 2
Effect of trial type on response latency in Experiment 1.
Because the facilitation task measures the strength of the relation between ideological categories and policy evaluation, we expected that individuals with more accessible identification as liberal or conservative should show greater facilitation and interference effects. In order to examine moderation by identification accessibility, we ran a mixed effects ANOVA including condition as a within-subject factor and self-identification accessibility as a between-subjects factor and modeled the interaction of trial type by accessibility (see Table 1). Self-identification accessibility did not influence overall response latency, F(1,27) = .108, p = .745, η2p = .004, and the interaction between trial type and ideological accessibility was not significant, F(2,62) = 1.43, p = .247, η2p = .04. The interaction effect was marginal when comparing only consistent with inconsistent trials, F(1,31) = 3.117, p = .087, η2p = .09. When accessibility was high (all cases below the mean), the difference between consistent and inconsistent trials averaged 592 ms, F(1,16) = 3.34, p = .086, in contrast to 286 ms when accessibility was low (all cases above the mean), F(1,13) = .945, p = .349, suggesting that the difference between consistent and inconsistent trials was larger for participants with more accessible self-identifications. The interaction effect was not significant when examining the difference between consistent and control trials, F(1,31) = .097, p = .758, η2p = .003, or inconsistent and control trials, F(1,31) = 2.561, p = .120, η2p = .08. So, in this data we did not find conclusive support for the idea that accessibility of self-identification moderates the speed of evaluative responses, although the pattern of data was in the expected direction.
Discussion
The results of Experiment 1 show that participants utilize the perceived ideological character of policy statements in judging them. They were faster to make evaluative judgments about policy statements after first categorizing those statements as liberal or conservative. Importantly, this is only true when participants’ judgments are consistent with his or her own ideology. In other words, for a liberal participant, categorizing a policy statement as liberal facilitates subsequent agreement, whereas categorizing a policy statement as conservative facilitates subsequent disagreement, and vice versa for conservative participants. By contrast, when evaluative judgments are inconsistent with participants’ ideology, those judgments are slowed or interfered with. Thus, the results established the utility of the task facilitation paradigm for examining the role of ideology in policy stances. However, the results provided only weak support for the idea that the facilitation effect should be moderated by the accessibility of political identification. However, the examination of this hypothesis in Experiment 1 was somewhat limited by the small sample size that remained after removing participants who were not clearly identified as liberal or conservative. The goal of Experiment 2 was to replicate the pattern of results found in Experiment 1 for consistent versus inconsistent judgments and see if moderation by identification strength would be observed with a different indicator of identification strength—identification extremity—and a larger sample size.
Experiment 2
Experiment 1 showed that participants relied on ideology when making evaluative responses to policy statements, and suggested that this may be especially true for those individuals with highly accessible political identification. In Experiment 2, we wanted to examine whether this effect would generalize to self-reported extremity of ideological self-identification (which we could not examine in Experiment 1 due to the use of a dichotomous identification measure). If ideology is functioning as a social identity, then those individuals who identify as very liberal or very conservative should be especially likely to evaluate social and political issues in terms of ideological categories.
Method
Participants in Experiment 2 were 122 undergraduate students at The Ohio State University in the United States. Two participants were omitted from analyses for failing to complete the study and a third for admitting to random responding, leaving 119 participants for analysis (55 female, 64 male; age range: 18-36, M = 19.69, SD = 2.27). Participants completed the same task facilitation paradigm as in Experiment 1 and provided their political identification on a seven-point scale (ranging from 1 = very liberal to 7 = very conservative). All experimental stimuli were presented on individual computers using MediaLab and DirectRT software (Jarvis, 2004a, 2004b).
Coding of Ideology
In order to code responses as consistent or inconsistent, participants were categorized as liberal or conservative on the basis of the political identification item. Liberal participants were those who responded that they were either very liberal, liberal, or slightly liberal (n = 41), and conservative participants were those who indicated that they were either very conservative, conservative, or slightly conservative (n = 31). Participants who selected the midpoint of the scale or declined to answer were coded as moderates (n = 47) and, as in Experiment 1, were not included in the primary analyses. We also created a variable for ideological extremity by first centering the ideology variable on the scale midpoint (4) and then squaring that value (ranging from 1 = weakly to 9 = strongly identified; see e.g., Haas, 2016; Jost et al., 2007). We examined whether ideological extremity differed as a function of whether participants were liberal (coded -1) or conservative (coded 1), finding that extremity did not differ as a function of ideological direction (b = -.405, SE = .273, t = -1.481, p = .143). As in Experiment 1, trials were coded as consistent or inconsistent on the basis of participant ideology and evaluative judgment. Latency data was processed as in Experiment 1 (trimmed and natural log transformed).
Results
Natural log transformed mean response latencies were subjected to a repeated measures ANOVA where condition was modeled as a within-subject factor with three levels (ideology-consistent, ideology-inconsistent, control). The overall effect of condition was not significant, F(2,138) = 2.05, p = .133, η2p = .03; see Figure 3. Latencies were faster for ideology-consistent trials (M = 2116 ms, SE = 79 ms) than for ideology-inconsistent trials (M = 2401, SE = 137), although this effect was marginal, F(1,69) = 3.64, p = .061, η2p = .05. Latency on ideology-consistent trials did not significantly differ from latency on control trials, M = 2183, SE = 65; F(1,69) = 1.08, p = .30, η2p = .02, and latency on ideology-inconsistent trials was marginally slower than latency on control trials, F(1,69) = 2.25, p = .138, η2p = .03.v Our primary focus here was on the moderating role of ideological extremity, so we describe that model below.
Figure 3
Effect of trial type on response latency in Experiment 2.
We expected that individuals who were more strongly identified as liberal or conservative should show the facilitation effect to a greater degree than those who were less strongly identified. To examine moderation by ideological extremity, we ran a mixed effects ANOVA with extremity added as a between-subject factor. As expected, ideological extremity moderated the effect of condition on response latency, F(2,136) = 8.32, p < .001, η2p = .11. The main effect of ideological extremity on response latency was not significant, F(1,64) = .122, p = .728, η2p = .002. Ideological extremity moderated the size of the difference between consistent and inconsistent trials, F(1,68) = 11.10, p = .001, η2p = .14, control versus consistent trials, F(1,68) = 5.76, p = .019, η2p = .08, and control versus inconsistent trials, F(1,68) = 8.03, p = .006, η2p = .11. In order to visualize this interaction, we subsetted the data for individuals who were weakly (extremity = 1), moderately (extremity = 4), and strongly (extremity = 9) identified with a political ideology. As shown in Figure 4, for weakly identified individuals, condition had no effect on response latency, but the pattern for moderately and strongly identified individuals mirrors the overall pattern described above, suggesting that the facilitation and inhibition effects are driven by those individuals who expressed stronger identification with an ideological group. It is worth noting that the standard errors are larger for individuals who are strongly identified given that we have fewer observations in those cells, especially for inconsistent trial types (see Appendix B for more information). This supports the idea that people who are more highly identified as liberal or conservative are more likely to show facilitation of (and interference with) evaluative judgments after categorizing policy statements as liberal or conservative, and shows that a second attitude strength-related variable (i.e., extremity) moderated the facilitation effect.
Figure 4
Effect of trial type on response latency as a function of ideological extremity (weak, moderate, strong) in Experiment 2.
Discussion
Experiment 2 shows that individuals higher in ideological extremity rely more on ideological categories (i.e., liberal, conservative) when evaluating policy statements. This finding supports the pattern found in Experiment 1, where accessibility of political identification had a similar (although statistically weaker) effect. In two studies, we have shown support for this overall pattern, demonstrating that attitude strength-related variables such as accessibility and extremity are indicators of the extent to which the self is associated with the political ideology categories and, hence, moderate the influence of ideology on the evaluation process.
Experiment 3
Experiments 1 and 2 suggest that ideology can be usefully conceptualized as a social identity, and past work has shown that ideology is linked to political issue preferences (e.g., Huckfeldt et al., 1999). However, political scientists have often focused on the link between partisan identification and issue preferences (e.g., Campbell, Converse, Miller, & Stokes, 1960; Goren, Federico, & Kittilson, 2009). For that reason, we conducted a third experiment to examine whether the effects found in Experiments 1 and 2 were specific to ideology, or would generalize to partisan identification. While there may be important differences between ideology and party in some domains, in this context we did not expect a difference. From a social identity perspective, both organized social groups such as political party and more abstract, symbolic groups such as political ideology can meet identity goals (see Brewer, 2001). In other words, we expect that both ideological and partisan identities should be tied to political attitudes in memory, and we test the effects of partisan identity in Experiment 3.
Method
Participants in Experiment 3 were 70 undergraduate students at the University of Nebraska-Lincoln in the United States. Four participants were omitted from analyses due to duplicate use of subject numbers and one for incomplete data, leaving 65 participants for analysis (26 female, 39 male; age range: 18-38, mean age = 20.26, SD = 2.81). Participants completed a task facilitation paradigm as in Experiments 1 and 2, but the labels liberal and conservative were replaced with Democrat and Republican. All other aspects of the task remained the same. Participants also provided their party identification on a single-item measure (ranging from 1 = Strong Democrat, 4 = Independent, 7 = Strong Republican). All experimental stimuli were presented on individual computers using MediaLab and DirectRT software (Jarvis, 2004a, 2004b).
Coding of Party
In order to code consistency versus inconsistency of responses, participants were categorized as Democrat or Republican on the basis of the political identification item. Participants who selected Strong Democrat, Democrat, or Weak Democrat were categorized as Democrats (n = 11), and participants who selected Strong Republican, Republican, or Weak Republican were categorized as Republicans (n = 46).vi Participants who selected Independent were coded as a third category (n = 12). We also created a variable for partisan extremity by first centering the party variable on the scale midpoint (4) and then squaring that value (ranging from 1 = weakly to 9 = strongly identified). We examined whether partisan extremity differed as a function of whether participants identified as Democrat (coded -1) or Republican (coded 1), finding that extremity did not differ as a function of partisan affiliation (b = -.475, SE = .355, t = -1.337, p = .187). As in the earlier experiments, trials were coded as consistent or inconsistent on the basis of party identification and evaluative judgment. Latency data was trimmed to retain data between 300 and 15000 ms and natural log transformed to correct for skewness.
Results
Natural log transformed mean response latencies were subjected to a repeated measures ANOVA where subject and condition (party-consistent, party-inconsistent, control) were modeled as within-subject factors. There was a significant effect of condition, F(2,103) = 5.33, p = .006, η2p = .09; see Figure 5.
Latencies were faster for party-consistent trials (M = 2005 ms, SE = 86 ms) than for party-inconsistent trials, M = 2275, SE = 131; F(1,51) = 10.54, p = .002, η2p = .17. Latency on party-consistent trials did not significantly differ from latency on control trials, M = 2145, SE = 77; F(1,52) = 1.15, p = .289, η2p = .02, but latency on party-inconsistent trials was significantly slower than latency on control trials, F(1,51) = 6.57, p = .01, η2p = .11.vii As in the earlier experiments, our primary focus is on moderation of these effects by partisan extremity, which we examine below.
Figure 5
Effect of trial type on response latency in Experiment 3.
We expected that individuals who were more strongly identified with a political group should show greater facilitation (and interference) relative to those who were less strongly identified. In Experiment 3, we examine the role of party identification (as opposed to ideological identification in the earlier studies). To examine moderation by partisan extremity, we ran a mixed effects ANOVA with extremity added as a between-subject factor. There was a marginal main effect of partisan extremity on response latency, F(1,47) = 3.79, p = .058, η2p = .07, qualified by the predicted interaction between condition and partisan extremity, F(2,101) = 4.38, p = .015, η2p = .08. Partisan extremity moderated the size of the difference between consistent and inconsistent trials, F(1,50) = 5.85, p = .019, η2p = .10, and between control versus inconsistent trials, F(1,50) = 7.16, p = .01, η2p = .13, but did not influence the size of the difference between consistent and control trials, F(1,51) = .02, p = .884, η2p = .00. In order to visualize this interaction, we subsetted the data for individuals who were weakly (extremity = 1), moderately (extremity = 4), and strongly (extremity = 9) identified with a political party. As shown in Figure 6, for weakly identified individuals, condition had no effect on response latency, but the pattern for moderately and strongly identified individuals mirrors the overall pattern described above, suggesting that the facilitation and inhibition effects are driven by those individuals who express stronger partisan identity. As in Experiment 2, the standard errors are larger for individuals who are strongly identified given that we have fewer trials in those cells (see Appendix B for more information). This supports the idea that people who are more highly identified as Democrat or Republican are more likely to show facilitation of (and interference with) evaluative judgments after categorizing policy statements as Democratic or Republican, and suggests that partisan identity functions similarly to ideological identity, at least in this context.
Figure 6
Effect of trial type on response latency as a function of partisan extremity (weak, moderate, strong) in Experiment 3.
Discussion
Experiment 3 shows that individuals higher in partisan extremity may rely more on party categories (i.e., Democrat, Republican) when evaluating policy statements. This finding is consistent with results from Experiments 1 and 2, which showed that accessibility of ideological identification (Experiment 1) and extremity of ideological identification (Experiment 2) both increased the degree to which people utilize these categories when making policy decisions.
General Discussion
On the basis of conceptualizing ideology as self-identification, we tested in three experiments whether individuals used ideological and partisan categories while evaluating policy statements. We have shown support for the idea that, for those who possess ideological or partisan self-identities, evaluation of policy statements was influenced by these group memberships. In a task facilitation procedure, identifying the perceived ideological or partisan character of policy statements facilitated the evaluation thereof when participants’ evaluations were consistent with their political identities. Evaluations were not, of course, entirely based on the perceived ideological content of policy positions. Sometimes participants made judgments inconsistent with their group identifications, but importantly in these cases categorizing the positions in ideological terms slowed responding. Ideologically inconsistent responses may actually interfere with or slow down evaluative decision making, as respondents contend with the unexpected disparity with their ideological identification. Importantly, individuals differ in the degree to which they show these facilitation and interference effects. Namely, people with more accessible or more extreme self-identifications are more likely to be helped (or hindered) by activation of the social category, suggesting that this information plays a role in decision making. With respect to these effects, across three studies, we found no evidence that the extent to which individuals utilize ideological or partisan character in evaluating policies differed as a function of whether participants’ identities were left- or right-wing. However, it is worth noting that these experiments were not designed to test for left-right differences and our ability to fully examine those here may have been limited by relatively small sample sizes.
The important question for researchers is not whether or not people have ideology, but understanding when, how, and what kinds of people use ideology to make decisions about important political issues. The present work allows us to draw two general conclusions. First, people do use ideological categories to evaluate policy issues, and their use of these categories can be influenced by the surrounding context. In other words, it is possible to prime ideology as a social category and change the cognitive processing involved in evaluating policy issues. Second, the use of ideological categories differs as a function of individuals pre-existing knowledge structure and identification with these groups. Strength of identification is important here—both in terms of accessibility and extremity. People whose ideological or partisan identities come to mind quickly and easily are more likely to use the perceived ideological or partisan character of a policy in its evaluation, as are individuals who think about themselves as highly identified with their ideological group.
While we find that individuals with some political identities utilize them in the evaluation of policy, we are unable to determine whether independents, moderates, or individuals who identify with ideologies other than those we studied engage in similar processing. For political independents or moderates, it may be that ingroup consensus on the issues is more difficult to establish in which case a similar process is unlikely to unfold. It is also possible that the nature of being an independent or moderate does not lead to strong self-identification with these groups. The exact nature of how and when independents and moderates use these group identities in the process of policy evaluation remains to be determined.