As our understandings of stereotyping and discrimination have developed, a range of interventions to reduce these social problems have also come about. When discrimination is understood as a consequence of ignorance and unfamiliarity, researchers often focus on intergroup contact (Brown and Hewstone, 2005; Pettigrew and Tropp, 2006). When discrimination is understood as a consequence of unconscious bias, then researchers have focused on motivating participants to change using compunction (Devine et al., 1991; Monteith, 1993). Researchers have also explored the cognitive demands involved in avoiding discrimination, and developed an intervention that attempts to address this phenomenon using promotion (Trawalter and Richeson, 2006). In practice, however, stereotyping and discrimination are driven by a number of processes operating at different levels, and so it makes sense to explore how these interventions work in combination. We suggest that it is particularly important to explore how interventions may impact on intergroup contact: intergroup contact is not only the longest standing and most comprehensively researched intervention in the literature, but it is also a part of ordinary social experience, and a key context in which stereotyping and discrimination are played out.

In this paper, we tested the separate effects of promotion (Study 1) and then compunction (Study 2) on the success of intergroup contact. In both studies, participants interacted with a confederate whom they believed to have a history of schizophrenia. Research shows that attitudes to this group are associated with strong emotional responses that include anxiety and fear (Crisp et al., 2000; Corrigan, 2002). People with a history of schizophrenia are widely seen as dangerous, unpredictable, and unlikely to recover (Crisp et al., 2000; Foster, 2001; see also West et al., 2010).

The promotion intervention (Trawalter and Richeson, 2006) builds on evidence that avoiding discrimination is cognitively demanding. People are often strongly motivated to avoid prejudice, but may feel that they lack the skills to do so (Plant and Devine, 1998, 2003). According to Richeson et al. (Richeson and Shelton, 2003; Richeson and Trawalter, 2005; Trawalter and Richeson, 2006), this means that people often enter intergroup interactions with an avoidance focus (because they are motivated to avoid being prejudiced). Unfortunately, however, Higgins's (1998) Regulatory Focus Theory predicts that an avoidance focus is more cognitively demanding (compared to promotion focus) in that it requires more vigilance and hence more executive attentional capacity. Compared to a promotion focus, avoidance focus can therefore be associated with avoidance behavior, vigilant cognitive processing, and ultimately cognitive depletion and exhaustion (Higgins et al., 1994; see also Plant and Butz, 2006). Theoretically, therefore, an avoidance focus could have (ironically) negative effects on actual intergroup interactions.

Trawalter and Richeson (2006) designed and tested a promotion intervention to replace avoidance with a promotion focus (Higgins et al., 1994). White participants were asked to approach an upcoming interaction with a Black confederate as an “opportunity to have an enjoyable intercultural dialogue” (promotion focus), or to “avoid appearing prejudiced” (avoidance focus), or were given no instructions (control condition). Results indicated that participants in the promotion condition appeared to be more comfortable and experienced less cognitive depletion. Trawalter and Richeson's design did not involve an actual intergroup interaction, but we can extrapolate from their data to make the following predictions.

An alternative (and widely researched) intergroup intervention is personal compunction (Devine et al., 1991; Monteith, 1993). Compunction was designed in response to research on the Implicit Association Test (IAT) and unconscious bias (Greenwald et al., 1998; Dovidio et al., 2002; Correll et al., 2007). It involves making participants aware of the gap between their personal standards (in which they are motivated to avoid prejudice) and actual behavior (which may be influenced by unconscious bias). This awareness can be achieved through pencil and paper tasks such as the would/should paradigm (Devine et al., 1991; Monteith and Voils, 1998; Monteith et al., 2010); through false feedback (Monteith, 1993; Monteith et al., 2002); or genuine feedback on a task that is likely to reveal a prejudiced response (Monteith, 1993; Fazio and Hilden, 2001; Gill and Andreychik, 2007; Devine et al., 2012). The IAT is an example of the latter: completing an IAT generates compunction, because participants experience difficultly in controlling their responses and are aware of that difficulty (Monteith et al., 2001; Frantz et al., 2004; Plant and Devine, 2009; Vorauer, 2012).

The compunction intervention therefore makes participants aware of discrepancies between their personal standards and actual behavior, and these discrepancies cause them to experience negative self-directed affect. Theoretically, this negative affect then motivates participants to regulate their subsequent behavior. The dependent measures used in these designs have frequently included changes in implicit or explicit attitudes (Devine et al., 2012), responses to jokes (Monteith, 1993; Monteith and Voils, 1998), or measures of information processing and search (Monteith, 1993; Monteith et al., 2002; Plant and Devine, 2009). Results using these measures have indicated that compunction increased participants' self-monitoring and thereby decreased the impact of implicit associations. However, they have rarely been extended into intergroup interactions. This is important because (as we have outlined above), if self-monitoring is characterized by an avoidance focus, then it may have negative effects during an intergroup contact.

We are aware of only one study that has examined the role of compunction during intergroup contact. Vorauer (2012) explored the effect of completing a race relevant IAT (as opposed to an irrelevant IAT or an explicit measure of racism) on an interracial interaction, and specifically on the metaperceptions of minority participants. Results indicated that when majorities completed a race relevant IAT, they were perceived by their minority partners as less self-disclosing and having less self-efficacy (compared to when participants had not completed an IAT). This study therefore provides some early evidence that the positive effects of compunction may not extend into intergroup contact.

Contact is probably the most intuitive and widely researched intervention to improve intergroup relations (Brown and Hewstone, 2005; Pettigrew and Tropp, 2006). It works on the assumption that prejudice is a consequence of ignorance and unfamiliarity that can be dispelled during intergroup dialogue, but perhaps underestimates the ways in which contact can go wrong and the damage that negative contact can have (Vorauer and Sakamoto, 2006; Paolini et al., 2010; Barlow et al., 2012; Graf et al., 2014; but see also Paolini et al., 2014). As we have already outlined, even people who are strongly motivated can inadvertently undermine the success of intergroup contact, either through avoidance motivations or negative self-directed affect (Shelton, 2003; Shelton and Richeson, 2005; see also Vorauer et al., 2009).

We explored two key factors that can undermine the success of intergroup contact: expectancies and intergroup anxiety. Plant et al. (Plant and Devine, 2003; Plant, 2004; Butz and Plant, 2006; Plant et al., 2008) have explored two expectancies that are relevant to intergroup contact: expectancies relating to self-efficacy (i.e., the belief that you can perform in a way that is not prejudiced) and expectancies related to the other (i.e., that your interaction partner is open to meeting you and does not assume that you will be prejudiced). These expectancies can have a positive effect participants' affective and behavioral responses to contact.

In contrast, intergroup anxiety has negative consequences that include stereotyping, negative affect, and a desire to avoid future interactions (e.g., Islam and Hewstone, 1993; Greenland and Brown, 1999; Greenland et al., 2001; Brown and Hewstone, 2005; Turner et al., 2007; but see also Paolini et al., 2016). It can be experienced as “self-anxiety” (a fear that the participant will act or think in a way that makes them appear to be prejudiced) and as “other-anxiety” (a fear that the other represents a risk or a threat; Greenland et al., 2012; see also Plant and Devine, 1998; Stephan and Stephan, 2000). However, it is often measured in the form of “generic anxiety” (measuring the intensity of participants' anxiety without exploring its content, e.g., Stephan and Stephan, 1985; Plant and Devine, 2003).

How might a promotion (i.e., approach) intervention affect expectancies and intergroup anxiety? There is some evidence that expectancies affect approach/avoidance motivations (Butz and Plant, 2006; Plant et al., 2008), but (to our knowledge) no research that has explored the inverse. Similarly, there is some correlational (but no causal) data that links approach/avoidance to self- and other-anxiety. Greenland et al. (Study 4, 2012) reported a positive relation between approach and self-anxiety and a negative relation between approach and other-anxiety. However, there was also a positive relation between self-anxiety and avoidance (which was interpreted by the authors as indicative of freezing; see also Vorauer and Turpie, 2004; Trawalter et al., 2009).

In contrast, we might expect a more direct effect of the compunction intervention on both expectancies and intergroup anxiety. The compunction intervention is designed to deliver apparently diagnostic information about participants' prejudice toward the target group. We would expect this to affect participants' expectancies of self-efficacy and self-anxiety.

In this paper, we tested the separate effects of promotion (Study 1) and then compunction (Study 2) on participants' interactions with a confederate whom they believed to have a history of schizophrenia. We selected these particular interventions for two reasons. First, because they are well-established in the academic literature; and second, because we had good reason to expect that promotion and compunction might impact on the success of contact. In Study 1, participants received a promotion intervention, in which they were instructed to try to relax and enjoy the interaction. In Study 2, participants received a high compunction intervention, in which they received false feedback indicating that they were high in prejudice toward people with mental health difficulties.

Drawing on Trawalter and Richeson (2006), we evaluated the effects of a promotion focus intervention on an actual interaction with a confederate who participants believed had a history of schizophrenia. Participants engaged in a real interaction in the context of a shared task. Consistent with Trawalter and Richeson (2006), we expected that participants in the promotion condition would have a more positive experience compared to participants in the control condition (Hypothesis 1) but made no specific predictions for expectancies and intergroup anxiety (Hypothesis 3).

The results of Study 1 revealed positive effects of a promotion manipulation during intergroup contact. Study 2 used the same context (interacting with a confederate who apparently had a history of schizophrenia) and the same dependent measures. However, we replaced the promotion intervention with compunction. Participants in both conditions completed the Single Category Implicit Attitude Test (SC-IAT; Karpinski and Steinman, 2006), an adaption of the IAT. They were then given feedback to indicate that they were either high in prejudice toward people with mental health difficulties (high compunction condition) or low in prejudice toward people with mental health difficulties (low compunction condition).

Our manipulation of compunction was therefore slightly different to previous designs which compared participants who had completed a relevant IAT condition with a non-IAT (or non-relevant IAT) control condition (e.g., Monteith et al., 2001; Plant and Devine, 2009; Vorauer, 2012). We used the SC-IAT (rather than an IAT) because of the absence of a relevant outgroup to the category “people with a mental illness.” The conventional IAT requires the pairing of ingroup/outgroup items that was not available for our target group.

Participants in both the compunction and control conditions completed the SC-IAT: the manipulation was contained in the feedback that they had to the task. This was because the SC-IAT used the single category “mental illness” (target words included crazy, mental, and psycho) and could therefore potentially prime a negative stereotype associated with mental illness. This would have generated a potential confound in comparing participants who had/had not completed the SC-IAT. Instead, all of our participants completed the SC-IAT, but we then gave them false feedback that would either confirm their compunction (i.e., that the results showed that they were high in prejudice) or disconfirm (and therefore reduce) their compunction (i.e., that they were low in prejudice). Giving participants neutral or no feedback would not have been effective, since simply completing an IAT has been shown to generate compunction (Monteith et al., 2001; Plant and Devine, 2009).

We expected that participants in the high compunction condition would have a more negative contact experience compared to participants in the low compunction condition (Hypothesis 2), more negative expectancies of self-efficacy, and more self-anxiety (Hypothesis 4). They might also have more negative expectancies of the other and more generic intergroup anxiety.

Across two studies, we tested the separate effects of promotion (Study 1) and then compunction (Study 2) on participants' interactions with a confederate whom they believed to have a history of schizophrenia. Prejudice, stereotyping, and discrimination are driven by a number of processes operating at different levels, so it makes sense to explore how an intervention designed to tackle one process (e.g., unconscious bias) can impact on another (e.g., intergroup contact). Any intervention that can improve (or undermine) the success of intergroup contact is particularly relevant, because the success or failure of intergroup contact has profound effects on the experiences of both minorities and majorities (Graf et al., 2014). Contact is therefore a key area on which to focus. As has been noted elsewhere, experimental designs using real interactions are also surprisingly rare (Hebl and Dovidio, 2005). Our data therefore has a real world significance that is often absent from intergroup intervention research.

The results indicated a broadly positive effect of the promotion intervention. Working from the assumption that many participants enter intergroup interactions with avoidance motivations (i.e., to avoid thinking or doing something inappropriate), Trawalter and Richeson (2006) designed a simple intervention that instructed participants to relax and enjoy an interaction. Our data indicate that this intervention had medium to large effect sizes on participants' post-interaction feelings and stereotypes, with ds between 0.55 and 0.98 (see Cohen, 1988).

In contrast, the high compunction intervention had broadly negative effects (compared to low compunction). Monteith (1993) suggested that participants can be motivated to change by alerting them to the gap between how they should behave and how they actually behave. Our results suggest that there is not a simple relation between this motivation and the success of intergroup contact: participants in the high compunction condition reported lower levels of self-efficacy and increased levels of intergroup anxiety, as well as a less positive contact experience in general. These results need to be treated with some caution in the absence of a non-compunction control condition: it is possible that the difference between the high and low compunction conditions were driven by positive effects in the low compunction condition (rather than negative effects in the high compunction condition). Nevertheless, our data suggests that high levels of compunction are at best ineffective, and at worst potentially damaging to intergroup contact. Of course, the compunction intervention was not designed to promote positive intergroup contact, and so there is no reason to expect that it should have positive effects in this specific context. At the same time, however, there is evidence that negative contact experiences can have a disproportionately negative affect on prejudice (Barlow et al., 2012). Our results therefore suggest that compunction researchers should begin to explore how the undoubtedly positive effects of compunction can be extended into positive contact experiences.

One key limitation in our design relates to generalization. Although participants in the promotion condition had more positive contact experiences, and stereotyped their interaction partner less, we have no data on how these effects might generalize to the target group as a whole (i.e., people who have a history of schizophrenia) or on future intergroup contact (Paolini et al., 2014). We also have no data on how enduring these effects are over time (see also Paolini et al., 2016). Pettigrew (1998) suggested that it was important to consider the sequence of interventions over time. We suggest that the same approach be taken to promotion and compunction: low to moderate levels of compunction might be an effective early intervention that motivates participants to engage, while promotion during contact reduces participants' avoidance motivations and thereby facilitates good contact.

This study was carried out in accordance with the recommendations of APA guidelines. Participants gave written informed consent in accordance with the Declaration of Helsinki. They were also given the opportunity to withdraw after debriefing and full disclosure. The protocol was approved by the Cardiff School of Social Sciences Research Ethics Committee.

Conceptualization: KG, GM; Data Curation: KG, DX; Formal Analysis: KG, DX; Funding Acquisition: KG, GM; Investigation: DX; Methodology: KG, GM; Project Administration: DX; Software: DX; Supervision: KG, GM; Writing—Original Draft Preparation: KG, DX, GM; Writing—Review and Editing: KG, GM.

This research was funded by the UK Economic and Social Research Council grant number RES000231106.