Contribution
It turns Allport’s (1954) hypothesis, a list of the conditions under which contact reduces prejudice, into a theory with processes, stages and levels. It names the four problems with the hypothesis (causal direction, an ever-growing list of conditions, no process, no account of generalization), answers each, and proposes the model the later meta-analyses test (LIT-1413, LIT-1405). It adds cross-group friendship as the potent case and “friendship potential” as a fifth condition.
Key insight
Contact changes more than beliefs about the outgroup. It changes behavior first, and the attitude follows to resolve the dissonance; it reduces anxiety and builds affective ties; and it shows that “ingroup norms and customs turn out not to be the only ways to manage the social world”, which makes one’s own group look provincial. Friendship works because it engages all four processes, so the theory is about long-term relationships, not first acquaintance.
Assumptions
- A review, not a study: the evidence is the literature, a re-analysis of the 1988 Eurobarometer surveys (over 3,800 majority respondents in seven samples from France, Britain, the Netherlands and West Germany), and Cook’s (1984) laboratory experiment.
- Causal claims from cross-sectional data rest on a nonrecursive structural model with mixed-neighbourhood residence as the instrument: living in a mixed neighbourhood is assumed to affect affective prejudice only through friendship.
- Meso level. Contact is a situational phenomenon placed within individual (prior attitudes, anxiety) and societal (norms, conflict) contexts.
Key results
- Selection bias (pp. 69, 72). Prejudiced people avoid contact, but in the European data the path from outgroup friendship to lower affective prejudice (−.210) is significantly stronger than that from prejudice to less friendship (−.137); mixed neighbourhoods raise the chance of an outgroup friend (+.356).
- Conditions (pp. 69–70, 76). The proposed extra conditions (common language, voluntary contact, a prosperous economy, initial views not too negative, stereotype disconfirmation) are mostly facilitating. The reformulated model treats Allport’s four and friendship potential as essential, the rest facilitating.
- Four processes (pp. 70–73). Learning about the outgroup, limited by the cognitive conditions for stereotype change (Rothbart and John); changing behavior, “a benign form of behavior modification. Behavior change is often the precursor of attitude change”, resolved by dissonance and reinforced by repetition; generating affective ties, anxiety reduction and empathy, with friendship strongest; and ingroup reappraisal: the new perspective “can reshape your view of your ingroup and lead to a less provincial view of outgroups in general (‘deprovincialization’)”. Its evidence: outgroup friendship relates to less national pride with education, age and conservatism controlled (Pettigrew 1997c, an unpublished manuscript), and Wilder and Thompson’s (1980) finding that less ingroup contact lowered bias.
- Generalization (pp. 73–75). Three kinds: across situations, from the individual to the outgroup, to uninvolved outgroups. Brewer and Miller’s decategorization and Hewstone and Brown’s salient categorization conflict, since those likeliest to have contact are atypical of their groups and contact generalizes best from typical members; they are reconciled as a sequence, ending, if ever, in the recategorization of Gaertner and colleagues’ common ingroup identity model. Wright and colleagues’ extended contact effect is noted.
- Uninvolved outgroups (p. 75). In the European data, friends of another group relate to more favourable views of eight outgroups, many not in the country, with seven controls; again the friendship-to- prejudice path is the stronger.
- Overall (p. 79). Friendship–prejudice correlations across five measures and seven samples average −.14 to −.32 without controls and −.10 to −.25 with seven controls, consistent with the preliminary meta-analysis then under way.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Contact usually reduces prejudice even without all of Allport’s conditions | moderate | review of field, survey and laboratory work |
| C2 | Contact’s effect outweighs the selection of the tolerant into contact | moderate | nonrecursive models, the European surveys and Powers and Ellison |
| C3 | Contact works through four processes, including changed behavior preceding attitude change | weak | argued from varied studies; no direct test |
| C4 | Contact with an outgroup leads to reappraisal of the ingroup (“deprovincialization”) | weak | one unpublished analysis of national pride, one laboratory study |
| C5 | Generalization is best reached by decategorization, then salient categorization, then recategorization | weak | reconciling two literatures; untested as a sequence |
| C6 | Friendship potential is an essential condition | weak | the European friendship results |
Concepts
- Deprovincialization: the less provincial view of outgroups in general that follows when contact shows one’s own group’s norms and customs are not the only ways to manage the social world. The word occurs once, in parentheses, as the name of the outcome of ingroup reappraisal.
- Ingroup reappraisal: the process by which optimal contact gives insight about the ingroup as well as the outgroup.
- Friendship potential: the situation’s opportunity for close, repeated interaction across varied settings, allowing self-disclosure; Cook’s “acquaintance potential”.
- Facilitating versus essential conditions: conditions that enhance contact’s effects against those it requires.
Connections
The 2006 meta-analysis (LIT-1413) tests this review’s problems and partly overturns its model: it finds Allport’s four conditions facilitating too, where this review’s reformulation kept them, with friendship potential, as essential. The 2008 meta-analysis (LIT-1405) tests three of the four processes’ components (knowledge, anxiety, empathy) and leaves ingroup reappraisal untested. Gaertner and colleagues’ common ingroup identity model is registered (LIT-1709), as is Paluck, Green and Green’s later re-evaluation of the experimental evidence (LIT-1420).
Bearing on the record
- LIT-1408’s flag is resolved. The review uses “deprovincialization”, once (p. 72), for the outcome of ingroup reappraisal. The exchange’s gloss, seeing one’s group’s ways as one way among others, matches the review’s wording. But the review’s deprovincialization is a less provincial view “of outgroups in general” reached through a revised view of the ingroup, and its evidence is thin: one unpublished analysis of national pride and one laboratory study.
- CLAIM-275. The review states the Franklin shape outright: “Behavior change is often the precursor of attitude change”, with dissonance as the mechanism. It is argued, not tested.
- CLAIM-417 (a perspective revised by being made one of several). Ingroup reappraisal is the process nearest the claim. The review offers it as a fourth mechanism with weak evidence; the later meta-analysis of mediators does not test it.
No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Its causal evidence is mostly cross-sectional; the path comparisons depend on the exclusion assumption for neighbourhood.
- Processes are proposed, not tested; the author says they “may well overlap and interact in complex ways”.
- The longitudinal model is a heuristic, “obviously oversimplified” in its three stages, and the review calls for the longitudinal research that would test it.
- Errors and slips. “Having an ingroup friend related to greater acceptance of minorities of many types” (p. 75) must mean an outgroup friend, as the surrounding text does. No other error was found.
Open questions
- Does contact produce ingroup reappraisal, and does that mediate reduced prejudice? A study measuring views of the ingroup before and after contact would test it.
- Does the decategorization–salience–recategorization sequence occur in that order in real contact?