Introduction

In our everyday life, we are endlessly confronted with examples of how liberals and conservatives embrace different views of the world and prioritize different goals and values [1]. Recent research has shown that these differences are skin deep and involve personality traits (e.g. [2], [3]), facial features [4], cognitive styles [5], and even brain structures [6]. In particular, conservatives, as compared to liberals, appear to more carefully process negative than positive environmental information. In a key study, Oxley and colleagues [7] showed that the exposure to threatening stimuli (e.g., a bloody face) gave rise to higher changes in skin conductance among conservative participants, as compared to liberal participants. Similarly, basic attention processes differ as a function of political ideology. For instance, it has been shown that negative information automatically attracts the attention of conservatives [8], [9]. In experimental tasks, such as a Dot-Probe Task, conservatives preferentially shift their attention towards threatening rather than positive stimuli [8], indicating that they are personally far more relevant than positive stimuli.

The asymmetrical processing of positive and negative information has important implications for how liberals and conservatives form impressions about novel social objects. In an intriguing study, Shook and Fazio [10] demonstrated that conservatives are more likely to display a learning asymmetry, namely a tendency to learn negative items relatively better than positive items, suggesting that the former might actually weigh more on the impressions formed by conservatives. People embracing conservative views of the world are also more susceptible to conditioning with negative stimuli than conditioning with positive stimuli [11], and this result points again to the presence of ideology-based asymmetries in the formation of social attitudes.

In a different line of research, Castelli and Carraro [12] explored impression formation processes relying on an illusory correlation paradigm. Illusory correlation is the tendency to misperceive the covariation between two events, and, more specifically, the tendency to believe that two relatively infrequent events are associated with each other even though no such association is actually present ([13], for a review see [14]). In the seminal study by Hamilton and Gifford [13], participants were presented with negative and positive actions performed by various members of two social groups (for instance, Group A and Group B). In the classic paradigm, one group is numerally larger (i.e., the majority group) as compared to the other (i.e., the minority group). In addition, the relative frequency of positive and negative actions is also manipulated, so that positive actions are more frequent than negative actions. Importantly, however, the ratio between negative and positive behaviors is exactly the same within the two groups, and thus there is no factual basis for evaluating one group as better than the other; nonetheless, participants tend to perceive a (illusory) correlation between the two infrequent information (e.g., between being a member of the minority group and performing negative behaviors). Moreover, Castelli and Carraro [12] found that illusory correlation effects were modulated by participants' political ideology, so that conservatives were even more likely to associate the minority group with negative behaviors and the authors reasoned that this was due to the increased distinctiveness of negative information in the case of conservative participants.

In the present study we first attempted to replicate the effects of political conservatism on illusory correlation and, most notably, explored whether the effects are limited to self-reported attitudes or extend to implicit attitudes. To our knowledge, only one experimental work has so far investigated the formation of both implicit and explicit attitudes within an illusory correlation paradigm [15]. Across two studies, the well-established illusory correlation effects were found on the explicit attitude measures but no effect emerged on the implicit attitude measure (i.e., an IAT) [16]. The authors interpreted the findings as the outcome of two largely independent learning processes (i.e., belief-based and contingency-based, respectively), so that implicit evaluations end up being unbiased, reflecting the actual contingencies in the observed information. Other lines of research, however, would lead one to expect illusory correlation effects to emerge on implicit attitudes as well [17]. According to Kruschke's attention theory of category learning [18]–[20], people first learn the attributes of the category that, within one context, is relatively larger than the other, and then learn the properties of the second category, giving more emphasis to what enables to distinguish it from the larger category. This implies that in the typical illusory correlation paradigm, participants first develop an impression about Group A and, because positive behaviors are indeed more frequent than negative behaviors, the final impression will be largely positive. Afterwards, the impression of Group B is formed and negative behaviors will attract the attention because they allow to better distinguish Group B from Group A. After the association between Group A and positive features is established, in relation to Group B there would be an attention-shifting mechanism that would lead perceivers to focus on the group-attribute pairs that maximize intergroup differentiation and, in that specific case, the focus would be on the association between Group B and negative behaviors. An important assumption of the model is that these processes occur while participants read the behavioral information about the two groups [17], [18]. Participants rapidly identify that one group is numerically larger than the other – at least when the ratio between the two groups is high - and set in motion the processes described above that imply a preliminary focus on the majority group followed by an attention-shifting mechanism toward the features of the minority group that allow one to maximize intergroup differentiation. As said, these processes occur at encoding during the initial learning phase, while participants are presented with the behavioral information, and therefore they could potentially impact onto the formation of specific associations between Group A (vs. Group B) and positive (vs. negative) features. In other words, the greater attention that is devoted to the instances in which members of Group B perform negative behaviors might strengthen the associative link between such group and negative features. According to this rationale, illusory correlation effects should be detected on both explicit and implicit attitude measures. In the present study we will thus further explore how illusory correlation may affect the formation of explicit and implicit attitudes, as well as how political ideology may modulate the effects. Overall, we expect that conservatives will display stronger illusory correlation effects on both implicit and explicit attitudes.

Method

Results

Illusory correlation as a function of ideology

In order to analyze the role of political ideology in the illusory correlations effects, a series of regression analyses were performed. As for implicit attitudes, a positive relation was found, β = .21, t(115)  = 2.28, p = .024, indicating that conservatives were more likely to display automatic associations between Group B (vs. Group A) members and negative (vs. positive) words as compared to liberals.

From further regression analyses political ideology emerged as a significant predictor of the differential evaluation of the two groups both along the 11 traits, β = .27, t(110)  = 2.99, p = .004, and along the 6 semantic differentials, β = .23, t(110)  = 2.50, p = .014. Also the phi coefficient calculated in relation to the cued recall task was positively associated to political ideology, β = .23, t(110)  = 2.45, p = .016. Finally, from responses to the last explicit measure a difference between the proportion of negative behaviors attributed to Group B and Group A was calculated. Again, political ideology emerged as a significant predictor, β = .20, t(110)  = 2.12, p = .036. Overall, the regression analyses consistently showed that stronger illusory correlation effects emerged for conservatives as compared to liberals both at the level of explicit attitudes as well as at the level of implicit attitudes.

The pattern of results described here has been confirmed also by a series of mixed models including “Group Type” (majority vs. minority; D in the models below) in the first model (i.e., m1 below). Next, in the second model (i.e., m2), we added the political ideology (X in the models below), and then in the third and last model (i.e., m3) we added the interaction between Group Type (majority vs. minority) and political ideology. The equations of the models are reported below:where i ( =  participants); j ( =  group); Y indicates the dependent variable (Explicit evaluation/Semantic differential/Ratio); D indicates the dummy variable associated to “Group Type” (0  =  Group A, majority group; 1  =  Group B, minority group); X stands for political ideology; β are the coefficients of the fixed part of the model; λ represents the random effects of participants.

For all the variables, the model including the interaction between type of group and political ideology (m3) provided the best fit to the data. Moreover, these analyses indicated that political ideology was related both to the evaluation of the majority group and to the evaluation of the minority group. Results are reported in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Results emerged by a series of mixed-models.

https://doi.org/10.1371/journal.pone.0096312.t001

Finally, in order to explore whether illusory correlation can be detected among liberals, we performed a series of additional analyses centering the predictor of political ideology at one standard deviation above and below the ideology mean, and then we examined whether the intercept was significant in order to understand whether a bias occurs for both conservatives and liberals, respectively. As for the IAT, results indicated that a bias emerged both for conservatives, β  = 1.13, t(114)  = 14.42, p<.001, and liberals, β = .88, t(114)  = 11.03, p<.001. As for the cued recall task, results from a similar analysis indicated an illusory correlation bias both for conservatives, β = .31, t(109)  = 5.31, p<.001, and for liberals, β = .11, t(109)  = 1.98, p = .05. This pattern of results emerged also for the other explicit variables, namely the explicit evaluation of the two groups (difference score) [β = .81, t(109)  = 7.76, p<.001, for conservatives, β = .38, t(109)  = 3.78, p<.001 for liberals], the semantic differentials (difference score) [β = 1.26, t(109)  = 6.07, p<.001 for conservatives, β = 0.54, t(109)  = 2.70, p = .008 for liberals] and the estimation of the ratio of negative behaviors performed by the two groups [β = .26, t(109)  = 5.92, p<.001,for conservatives, β = .13, t(109)  = 3.09, p = .003 for liberals]. Current results seem to indicate that an implicit and explicit illusory correlation bias emerged both in the case of liberals and in the case of conservatives, although remarkably stronger in the latter case.

Discussion

Nowadays there is a renewed interest in the study of the likely differences that characterize conservatives vs. liberals [22], [23]. Literature has largely documented how political orientation and ideology are not only related to different views of the world, but also to more deep cognitive and automatic processes. For instance, conservatives appear to be more sensitive to negative stimuli as compared to positive stimuli [7]–[11], and negative information immediately captures the automatic attention of conservatives [8]. This attentional asymmetry between people who support different ideologies may, at least partially, explain why conservatives often develop more negative impressions about social minorities. Indeed, conservatives appear to weigh negative information more than positive information and this leads to even more biased impressions in the classic illusory correlation paradigm [12]. In the present study we replicated this effect showing that the explicit evaluations and the memory performance of conservatives were more likely affected by an illusory correlation bias. Even though there was no factual basis for considering the minority group as less positive than the majority group, people in general, and conservatives in particular, ended up with such an inaccurate perception. Notably, similar effects were also observed on implicit attitudes, as assessed through the IAT. Conservatives, indeed, showed a stronger tendency to automatically associate the minority group (vs. majority group) to negative features (vs. positive features) suggesting that ideological individual differences can shape, in an illusory correlation paradigm, both explicit and implicit attitudes. This is very much relevant to the extent that explicit and implicit attitudes are related to the regulation of different types of behaviors. Whereas explicit attitudes mainly predict deliberate behavior, implicit attitudes are more strongly related to spontaneous behavior [24]–[27]. For instance, explicit attitudes are correlated with the content of verbal behaviors and implicit attitudes are better predictors of nonverbal and uncontrolled behaviors that are displayed while interacting with the attitude object. In addition, implicit attitudes are more likely to predict behaviors under specific circumstances [28]–[29], such as reduced cognitive capacity and ego depletion, that temporarily weaken central executive control. Implicit and explicit attitudes also differ in relation to how they change over time. Indeed implicit attitudes (relative to explicit attitudes) change more slowly and are more impermeable to disconfirming information that is subsequently encountered [30]. Taken together, it appears that conservatives, once confronted with a minority group that displays relatively infrequent negative behaviors with the same probability of a majority group, develop widespread negative attitudes that prepare them to negatively react toward such minority group.

The findings from the present study are also relevant at a more general level for the understanding of illusory correlation phenomena. Until now, to the best of our knowledge, only one study [15] has investigated illusory correlation effects on both explicit and implicit attitudes concluding that two different processes are involved, leading to distinct attitudes: biased explicit attitudes but unbiased implicit attitudes. In sharp contrast, we here found that the illusory correlation bias was not restricted to explicit evaluations but it also emerged when automatic associations were assessed. Indeed, from responses in the IAT [16], [21], participants revealed more negative implicit attitudes toward the minority group as compared to the majority group. Moreover, implicit and explicit attitudes appeared to be somehow correlated to each other, whereas no order effects emerged to be significant. These results are in line both with the “strategic explanation” proposed by Fiedler [31], [32] and with the “attention theory of category learning” proposed by Kruschke [18]. Indeed, according to the “strategic explanation” [31], [32] while reading evaluative information about the two groups, each positive or negative action performed by a member of a group reinforces the association between that specific group and positive vs. negative evaluations. This rationale may explain both implicit and explicit illusory correlation effects. In addition, the same effects may also be explained according to Kruschke's theory [18]. The theory states that people initially form an impression (in this case positive) about the majority group and then give more emphasis to the features (in this case negative) that distinguish one group from the other. Importantly, this process would occur while participants are engaged into reading the behavioral information about the two groups [17], [18], thus enabling to form biased explicit as well as implicit attitudes. However, further studies will have to better investigate why in some cases illusory correlation effects emerge at both the implicit and explicit level whereas in some other cases illusory correlation effects are restricted to explicit attitudes [15]. The methodology used in the present study differs from that employed by Ratliff and Nosek on several respect, and therefore one may only speculate about the origin of the obtained different findings. It is noteworthy that previous studies [15] made use of fictitious names to designate both the groups and group members and this may have hindered impression formation processes. Interestingly, the effects on explicit measures were also remarkably weaker in Ratcliff and Nosek's studies as compared to those observed here, suggesting that the methodology used in their studies was less likely to trigger illusory correlation phenomena overall, and therefore reducing the likelihood of detecting any effect at the implicit level. Moreover, because of the difference between the present findings and those reported by Ratcliff and Nosek, we performed an additional study (N = 86, F = 68, age M = 19.20, SD = .87) using the very same material and procedure adopted in the present study, with only one difference, namely the name of the two groups (Group Blue and Group Yellow instead of Group A and Group B). The key effect on the implicit measure was replicated and, indeed, a significant implicit preference for the majority group over the minority group emerged (D-IAT  = 1.17, SD  = .70; t(85)  = 2.11, p = .030).

To conclude, the current results confirm and extend previous findings [12] indicating that political orientation can actually impact onto the formation of both explicit and implicit attitudes in an illusory correlation paradigm. Indeed, conservatives developed more negative explicit and implicit attitudes toward the minority group as compared to the majority one. This result is important not only from a theoretical standpoint but also for its practical implications, suggesting that basic cognitive processes, such as those involved in the illusory correlation bias, may contribute to the more likely emergence of negative attitudes toward minorities among conservatives.