23830409

Upload: anderson-barreto

Post on 14-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 23830409

    1/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    The role of in-group identification, religiousgroup membership and intergroup conflictin moderating in-group and out-group affect

    Ed Cairns1, Jared Kenworthy2, Andrea Campbell1*and Miles Hewstone2

    1University of Ulster, UK2University of Oxford, UK

    We conducted secondary analyses of data from two random samples of the populationof Northern Ireland, involving 1046 participants in 2000, and 1000 participants in 2001,

    to explore the role of in-group identity and religious group membership in moderating

    the relationship between in-group and out-group affect. In both surveys the results

    indicated a general in-group bias - feeling thermometer ratings (affect) for the in-group

    exceeded those for the out-group. This effect was moderated by participants in-group

    identification and religious group (Catholic or Protestant), but these moderations also

    varied as a function of differential sectarian tension between 2000 and 2001. In both

    years, high identifiers and Protestants exhibited more in-group bias than low identifiers

    and Catholics, respectively.

    In-group bias, or in-group favouritism, is a well-established phenomenon in the field ofintergroup relations (see Hewstone, Rubin, & Willis, 2002). This bias, or preference for

    ones in-group over the out-group, is typically expressed in evaluation, liking or in theallocation of resources and rewards (Brewer, 1979; Brewer & Kramer, 1985; Tajfel, 1982;Wilder, 1986). Often concomitant with the occurrence of in-group favouritism is thepresumption of out-group negativity. This presumption has its origins as far back asSumners (1906) theory of ethnocentrism, which postulated an inverse relationshipbetween in-group positivity and out-group negativity.

    However, the assumption that in-group favouritism and out-group discrimination can

    be viewed as if they were two sides of the same coin (Brewer, 1999) was challenged asearly as 1954, when Allport (1954) argued that whereas out-group hostility maystrengthen in-group belongingness, such negativity is not a necessary requisite for

    positive feelings towards the in-group. Instead, Allport suggested that in-group bias isassociated with a variety of attitudes towards out-groups ranging from hatred at one

    * Correspondence should be addressed to Andrea Campbell, School of Psychology, University of Ulster, Cromore Road,Coleraine, BT52 1SA (e-mail: [email protected]).

    The

    British

    Psychological

    Society

    701

    British Journal of Social Psychology (2006), 45, 701716

    q 2006 The British Psychological Society

    www.bpsjournals.co.uk

  • 7/30/2019 23830409

    2/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    extreme to tolerance and appreciation at the other extreme, depending on the out-groups involved (Allport, 1954).

    More recently, Brewer (2001), in support of Allport, noted that some twenty years ofresearch in this area has called into question the possibility of a functional or evenstatistical relationship between reported feelings, or affect, towards in-groups and out-groups. Further, Brewer has also pointed out that not only is a reliable negativerelationship between in-group and out-group affect missing in the research literature(viz., Brewer, 1979; Hinkle & Brown, 1990), in many studies the relationship is actually

    positive. That is to say that sometimes, contrary to Sumner (1906), as in-group positiveaffect increases so does out-group positive affect.

    Brewer (1999) has suggested that a negative relationship between in-group and out-group evaluations may only be found in societies where factors such as strong in-group

    identification and loyalty, social comparison processes and sensitivity to threat andpower politics all encourage and promote out-group hostility. Similarly, Hewstone et al.(2002) argue that out-group derogation in addition to in-group favouritism possibly onlyoccurs in situations of extreme intergroup conflict (p. 579).

    Experimental research by Mummendey, Klink and Brown (2001) has providedsome evidence that helps to integrate these ideas. Their research was designed toinvestigate the suggestion that it is possible to use social identity theory todifferentiate between the processes involved in nationalism (i.e. feelings of in-groupsuperiority and dominance) and patriotism (i.e. positive feelings towards ones

    country without out-group derogation; see Feschbach, 1994). Measures usedincluded items to assess in-group evaluations, out-group derogation and strength ofin-group identification under three conditions. Participants were primed to think of

    three possible contexts: (a) why they preferred living in their own country withoutany other reference being made (control), (b) why they preferred living in theirown country rather than in any other country (intergroup comparison) or (c) whythey preferred living in their own country now, rather than in the past (temporalcomparison).

    Mummendey et al. (2001) only found a statistically reliable (positive) correlationbetween in-group evaluations and out-group derogation for the intergroup comparisoncondition. That is, only when participants were primed to think about the dependentrelational orientation of the intergroup context did they exhibit in-group bias, such that

    out-group derogation increased as in-group evaluations increased. Further, in line withpredictions from social identity theory, in-group identification correlated significantlyand positively with in-group evaluations such that high identifiers evaluated the in-groupmore positively. Based on their findings, Mummendey et al. concluded that the resultssupported the hypothesis that in-group identification is an important moderator of thecorrelation between positive in-group affect and out-group derogation underintergroup comparison conditions only (2001, p 168). In other words, when anintergroup comparison is made salient, those who identify strongly with their nationalgroup exhibit more in-group bias in the form of nationalism (or feelings of superiority

    and dominance) rather than a more benign patriotism.Levin and Sidanius (1999) provided data on the relationships between these same

    variables (identification, in-group evaluations and out-group evaluations; also socialdominance orientation) in societies that have experienced mild to extreme intergroup

    conflict: the USA and Israel. They found that across and within a variety of high-statusgroups (U.S. Whites, Israeli Ashkenazim and Israeli Jews), higher in-group identificationwas associated with more positive in-group affect as well as with more negative

    702 Ed Cairns et al.

  • 7/30/2019 23830409

    3/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    out-group affect. Unlike Mummendey et al. (2001), Levin and Sidanius measuredin-group and out-group affect on identical evaluative scales (how positively or negatively

    they felt towards each group), rather than conceptually distinct scales, and so therelative comparison between in-group affect and out-group affect is clearer in thisinstance.

    Further, Levin and Sidanius (1999) found that for high-status Jews in the Jewish-Arab comparison sample, the relationship between (identical) in-group and out-groupaffect measures was statistically significant and positive, but weak. That is, as in-

    group affect increased, so did corresponding out-group affect measures. In theirother samples, where the regional conflict between the groups compared was lessintense and less violent (viz., U.S. Black-White relations, Israeli Ashkenazim-Sephardim relations), the same positive correlation between in-group and out-group

    affect was far stronger. Indeed, Levin and Sidanius suggested that the absence of astronger relationship between in-group and out-group affect in the Jewish-Arabsample in their study may be related to the intensity of the regional conflictbetween Arabs and Jews (1999, p. 116). Given their methodology with respect tothe affect measures wherein higher scores indicate more positivity towards both thein-group and the out-group, these results suggest that as intergroup conflictincreases, the normally positive correlation between in-group and out-group affectwill weaken.

    Our research was stimulated by both the empirical findings discussed above (Levin &

    Sidanius, 1999; Mummendey et al., 2001) and Brewers (1999) conceptual work.Specifically, Brewer predicted that a negative correlation between in-group and out-group evaluations may be expected in highly segmented societies that are differentiated

    along a single primary categorization, such as ethnicity or religion when thecategorization is dichotomous, dividing the category into two significant subgroups(p. 439). The findings of Levin and Sidanius, like those of Mummendeyet al., suggestthat in-group bias will be greater for those who identify strongly with their in-group;Levin and Sidanius findings also suggest that conflict intensity will weaken the normallypositive correlation between in-group and out-group evaluations. We set out to replicateand extend these prior findings and test Brewers ideas in another deeply dividedsociety: Northern Ireland.

    Social identity and Northern Ireland

    The motivational foundation of social identity theory (Tajfel & Turner, 1986) is the drivefor group members to positively differentiate their own group from relevant out-groupsin order to achieve a sense of positive identity. The conflict in Northern Ireland, whichhas existed for decades if not centuries, is, in the main, a struggle between two large

    groups. On the one hand, there are those who wish to see Northern Ireland remain partof the United Kingdom (the Protestants/Unionists) and, on the other hand, those whowish to see the reunification of the island of Ireland (the Catholics/Nationalists; seeCairns & Darby, 1998, for a review). Furthermore, Catholic or Protestant identity istypically imposed from birth onwards. Thus, ones membership in either of these groupsis virtually inescapable; the group boundaries are relatively impermeable in socialidentity theory terms.

    In survey after survey, when people in Northern Ireland are asked to statewhether they are Catholic or Protestant, the vast majority are willing to answer thequestion regardless of whether they ever attend church or not (Niens & Cairns,

    In-group and out-group affect 703

  • 7/30/2019 23830409

    4/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    2001), although Northern Ireland does have a much higher rate of church attendance

    than the rest of the UK. Nevertheless, the labels Catholic and Protestant refer to the

    two communities whose struggle is essentially political and not theological or

    religious. In this sense, the groups can be seen as dividing along an ethnopolitical or

    ethnoreligious dimension, rather than a strictly religious dimension per se. Most

    relevant to the present context, however, is that Northern Irish politics are often said

    to be a zero-sum game with one side fearing that any out-group gains automatically

    translate into in-group losses. This parallels what Mummendey et al. (2001) called the

    intergroup comparison condition, in which the orientation of an individual toher/his in-group typically involves some simultaneous psychological relationship to

    the out-group. With respect to status differentials, most empirical evidence suggests

    that historically and traditionally higher-status Protestants are less tolerant of their

    out-group than Catholics (see Whyte, 1991, for a review). Thus, Protestants typically

    exhibit stronger in-group favouritism than Catholics.Northern Ireland therefore ought to be the sort of society in which differences

    between in-group and out-group affect can be examined in the ways described thus

    far and according to Brewer (1999), Hewstone et al. (2002) and social identity

    theory (Tajfel & Turner, 1987) in general (see related comments above). Surprisingly,

    no study appears to have been carried out in Northern Ireland to test the

    relationship between the intensity of conflict, in-group identification and in-group

    versus out-group affect.Given the lack of clarity on this topic in the extant literature, we set out to investigate

    in-group and out-group affect in Northern Ireland and to determine if differences

    between these variables would be moderated by the level of in-group identification and

    the varying level of the conflict. We were also interested to see if these phenomena

    applied equally to both groups involved the historically dominant Protestants and the

    historically lower-status Catholics. We note here that ours will be the first study to

    examine varying levels of intergroup tension within the same population; typically,

    researchers can only compare across different societies with differential tension levels

    (e.g. Levin & Sidanius, 1999). We will also test these important theoretical questions

    using large, random samples of the Northern Irish adult population, rather than

    unrepresentative samples of university students.The present study therefore involved secondary analyses of existing random sample

    survey data for two different and consecutive years. Because the political climate wasconsiderably worse in the second-year (2001) that in the first-year (2000), it was possible

    to examine the hypothesis that the intensity of intergroup conflict exacerbates the

    difference between similar measures of in-group and out-group affect (viz., in-group

    bias; see Levin & Sidanius, 1999) and whether in-group identification and religious

    group membership will moderate such an effect.

    Predictions

    Our hypotheses can be classified into two types. To test the foregoing ideas weperformed a series of ANOVA that examined group differences and interactions based onvariability along the factors of religious group, year, in-group identification and target

    (in-group or out-group, within subjects). In addition, as the preceding discussionimplies, we will also conduct a separate set of analyses to examine the correlational

    relationship between in-group and out-group affect.

    704 Ed Cairns et al.

  • 7/30/2019 23830409

    5/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    ANOVA hypothesesWith respect to affect measures, we expected to find (1) a general in-group bias,

    measured as a difference between rated in-group affect and out-group affect (feelingthermometer scales; see Haddock, Zanna, & Esses, 1993). This main effect of in-groupversus out-group target affect (within subjects) is expected to be moderated by in-groupidentification, as well as by the religious denomination of the respondents. Specifically,(2) those who identify highly with their religious in-group should exhibit more in-groupbias than should those who identify more weakly. We also expected that (3) in-group

    bias would be exhibited more strongly by historically higher-status Protestants than bylower-status Catholics.

    Because cross-community tension was greater in the second sample year than in thefirst, however, we expected a further moderation of the effect of in-group identification

    by sample year. That is, (4) we expected that the interaction between in-groupidentification and target would be stronger in the peaceful-year sample than in the moreconflictual-year sample. Thus, under relatively peaceful conditions, low identifiersshould exhibit little in-group bias as compared with high identifiers; under greaterintergroup tension, this moderation of identification should be weakened. This latterprediction is based on the fact that religious group membership is relativelyimpermeable and, thus, social mobility for low identifiers (i.e. leaving the group; seeTajfel & Turner, 1987) is not an option. We expect that even low identifiers, whosegroup membership is inescapable, will exhibit in-group bias when intergroup conflict is

    high compared with when it is low.

    Correlational hypothesesBased on theoretical statements by Levin and Sidanius (1999) and Hewstone et al.(2002) that increased intergroup tension weakens or even reverses the normallypositive correlational relationship between in-group and out-group affect, we expected(A) the correlation between these two variables to be positive and strong in 2000 ascompared with 2001 when intergroup relations were worsening. More important,however, we expected that the correlation between in-group affect and out-group affectwould var y as a function of the level of in-group identification and also as a function ofreligious group membership. Specifically, the positive correlation between the twoaffect measures should be (B) stronger for low identifiers than for high identifiers and

    (C) stronger for Catholics than for Protestants.

    Method

    Survey 1: 2000

    Context, fieldwork, and sampling

    All interviews were conducted between mid-April to mid-June of 2000 in therespondents homes by professional interviewers. During this period the NorthernIreland peace process appeared to be making progress. For example, in May of 2000 theIrish Republican Army (IRA) agreed to open some of its arms dumps for inspection andto begin a process to put its weapons beyond use. This was accompanied bystatements that British Army troop numbers in Northern Ireland were to be reduced to

    their lowest level since 1970 and that a number of security installations were to close.The (Protestant) Unionist leader (David Trimble) accepted the IRA declaration and wenton (in May) to win his partys backing to re-enter the power-sharing local government.

    In-group and out-group affect 705

  • 7/30/2019 23830409

    6/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    Further, confirming the view of 2000 as being a relatively benign period of cross-community relations, in a Northern Ireland life and times (NILT) survey conducted in

    2000, 42% of respondents who were asked whether things are better now than 5 yearsago reported that they were better.

    Achieved sampleInterviews resulted in 1,046 completed surveys. The final sample had a mean age of

    48.01 years (SD 18:

    79) and comprised 38% Catholics (42% male, 58% female) and 62%Protestants (43% male, 57% female).

    Survey 2: 2001

    Context, fieldwork, and samplingAll interviews were conducted between early April and the end of May 2001 in therespondents homes by professional interviewers. During this period, political events

    were notably more negative than the prior year. At least three bomb exploded in Londonbetween early March and early May and a large bomb was defused in Northern Ireland in

    early April. All these incidents were attributed to a dissident IRA faction.Decommissioning of IRA arms appeared to be stalled (nothing had happened sinceJune 2000). As a result, the Unionist leader stated his intention to resign on 1 July 2001 if

    decommissioning did not begin. Finally, this period included a UK-wide census and therun up to a UK-wide general election which was held on 7 June 2001. In this election,Sinn Fein made notable gains replacing the Social Democratic and Labour Party as thelargest Nationalist/Catholic party. In the NILT survey for 2001, only 28% of respondentsreported that things were better now than 5 years ago.

    Achieved sampleInterviews resulted in 1,000 completed surveys. The final sample had a mean age of49.26 years (SD 18:80) and comprised 41% Catholics (42% male, 58% female) and 48%Protestants (40% male, 60% female).1

    MeasuresBoth surveys contained questions to assess basic demographic data including age,gender, socio-economic status and religious denomination. Socio-economic status (SES)was measured by asking participants to state their fathers occupation from a choice ofsix categories ranging from professional managerial to unskilled manual.

    Respondents were asked to indicate which specific Christian denomination they

    were affiliated with. This information was then recoded into a dichotomous variablerepresenting either Catholic or Protestant group membership.

    Other key variables included measures of in-group affect and out-group affect as wellas items assessing in-group identification. The measures of in-group and out-group affectinvolved the use of a single-item feeling thermometer; respondents were asked to ratetheir own community and the other community (regionally understood as religious

    1 In 2000, the data that were available contained only those who had classified themselves as either Catholic or Protestant,whereas in 2001 the data contained Catholics, Protestants and others.

    706 Ed Cairns et al.

  • 7/30/2019 23830409

    7/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    in-group and out-group, respectively) using a thermometer scale that ran from zero toone hundred. Participants were told that lower numbers indicated less favourable

    feelings and that higher numbers indicated more favourable feelings. In both surveys thein-group affect measure was always obtained first, followed by the out-group affectmeasure.

    In both surveys, items assessing in-group identification were adapted and shortenedfrom Brown, Condor, Mathews, Wade, and Williams (1986) group identification scale.The psychometric properties of the original 10-item scale are adequate. It has good

    reliability with item analysis of the scale yielding a Cronbachs alpha of .71. The validityof the scale has also been established (Brown et al., 1986).

    However, the way in which identity was measured varied slightly from survey tosurvey. In survey 1, in-group identity was measured using three 4-point items (I identify

    strongly with my community, My community is an important part of who I am and Mycommunity is an important group to me; a .90) with end-points labelled as Agreestrongly to Disagree strongly. In survey 2, identity was measured using five 5-pointitems (viz., Would you say that you are a person who identifies with theCatholic/Protestant community?, Would you say that you are a person who feelsstrong ties to the Catholic/Protestant community?, Would you say that you are a personwho is glad to be a member of the Catholic/Protestant community?, Would you say thatyou are a person who sees yourself as belonging to the Catholic/Protestant community?,Would you say that you are a person who considers the Catholic/Protestant

    community to be important?; a .96) with end-points labelled as No-never to Yes-very often.

    Before measuring or dichotomizing the in-group identification scales within a

    common, merged dataset, those of survey 1 were reverse-coded to be directionallysimilar to survey 2. Thus, for both surveys, higher scores indicated greater in-groupidentification. Each surveys identification index was standardized and dichotomized(along a median split) for use in the analyses reported below.

    Results

    Before proceeding with the main analyses we tested for differences between the twosamples (from 2000 to 2001) in relation to gender, age and SES. There were nosignificant differences between the 2000 and 2001 samples in terms of gender, females

    57% in 2000, 59% in 2001; F1; 1846 0:874, p :35, or age, M 48 in 2000,M 49:1 in 2001; F1; 2044 1:69, p :19. There was a difference in SES,F1; 1846 39:10, p , :001, indicating that the overall sample in 2000 had a slightlyhigher SES (M 3:65) than in 2001 (M 3:22). For all analyses reported below, gender,age and SES will be entered as covariates in order to control for their respectivecontributions to the variance.

    To test our hypotheses we submitted the data to a 2 (target: in-group affect, out-group affect) 2 (year: 2000, 2001) 2 (religion: Catholic, Protestant) 2(identification: high, low) Analysis of Covariance (ANCOVA; with gender, age and SES ascovariates) with the first factor measured within subjects and the remaining threefactors as between subjects. As a group, Catholic respondents exhibited greater in-group

    identification (z 1:

    65, SD 0:

    49) than did Protestants (z 1:

    53, SD 0:

    48),F1; 1898 26:16, p , :001; religion did not interact with sample year,F1; 1898 0:532, p :47.

    In-group and out-group affect 707

  • 7/30/2019 23830409

    8/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    In the analyses reported below, partial beta-squared (02) values will be presented as

    measures of effect size, representing the percentage of variance in the model accounted

    for by the associated effect. Based on the number of post hoccomparisons, we have set

    the alpha level to a corrected p , :005.In-group bias. Hypothesis (1) that we would observe a general in-group bias

    effect was supported when a within-subjects main effect of target emerged,

    F1; 1586 52:15, p , :001, 02 :032. Specifically, in-group affect ratings(M 73:66, SD 19:67) exceeded out-group affect ratings (M 63:84, SD 21:61).

    In-group identification. Hypothesis (2) that those who identify highly withtheir religious in-group should exhibit more in-group bias was also supported. There

    was an interaction between target and identification, F1; 1586 75:81, p , :001,02 :046, such that the within subjects target effect was far stronger2 for highidentifiers, F1; 1586 348:57, p , :001, 02 :18, than it was for low identifiers,

    F1; 1586 14:64, p , :001, 02 :009 (see Figure 1). High and low identifiersdiffered for in-group affect ratings, F1; 1586 115:48, p , :001, 02 .068, whereasthey did not differ for out-group affect ratings, F1; 1586 0:123, p :726.

    Religious group. Target also interacted with religion, as predicted by hypothesis (3),

    F1; 1586 16:89, p , :001, 02 :011, such that the target effect was stronger fortraditionally higher-status Protestant respondents, F1; 1586 219:61, p , :001,02 :122, than it was for traditionally lower-status Catholic respondents,

    F1;

    1586 47:

    87, p ,:

    001, 02

    :

    029 (see Figure 2). Protestants and Catholicsdiffered for out-group ratings, F1; 1586 42:42, p , :001, 02 :026, whereas theydid not differ for in-group ratings, F1; 1586 4:62, p :032, 02 :003.

    Identification and sample year. Hypothesis (4), that there would be a three-way

    interaction between target, identification and year, was supported, F1; 1586 6:77,p , :01, although this interaction was associated with a relatively small effect size,

    02 :009. Decomposing this interaction, we found, as expected, that the two-wayinteraction between target and identification was stronger in 2000, F1; 1586 73:85,

    p , :001, 02 :077, than it was in 2001, F1; 1586 15:06, p , :001, 02 :021(see Figure 3). Specifically, and as expected, no bias was observed for the low identifiers

    in 2000, F1; 882 0:913, p :34, as compared with 2001, F1; 701 18:85,p , :001, 02 .026. High identifiers exhibited strong in-group bias in both 2000,

    F1; 882 236:61, p , :001, 02 :212, and in 2001, F1; 701 125:96, p , :001,

    02 :152.Identification, sample year, and religious group. Although it was not expected, a

    four-way interaction emerged between target, year, identification and religion,

    F1; 1586 8:04, p , :001, 02 :005 (means and standard errors can be found inTable 1). Follow-up analyses revealed that the three-way interaction between target,

    identification and religion emerged for the 2001 sample, F1; 701 16:13, p , :001,02 .022, but not for the 2000 sample,F1; 882 0:052,p :82. In the 2000 sample,only a target identification interaction was obtained, F1; 882 73:85, p , :001,02 :077, indicating that the pattern of interaction was parallel for Protestant andCatholic respondents. By contrast, in the 2001 sample, a target identification

    interaction was found for Protestant respondents, F1; 361 24:49, p , :001,02 .064, but not for Catholic respondents, F1; 337 :002, p :97.

    2 We refer to the respective effect sizes as one measure of their relative strength.

    708 Ed Cairns et al.

  • 7/30/2019 23830409

    9/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    60

    65

    70

    75

    80

    85

    Ingroup affect Outgroup affect

    Target

    Thermometerratings

    Low identifiers

    High identifiers

    Figure 1. Means (;SE) representing the interaction between target and in-group identification,

    collapsing across religion and year.

    55

    60

    65

    70

    75

    80

    Ingroup affect Outgroup affectTarget

    Thermometerratings

    Protestant

    Catholic

    Figure 2. Means (;SE) representing the interaction between target and religion, collapsing across

    in-group identification and year.

    50

    55

    60

    65

    70

    75

    8085

    90

    Ingroupaffect

    Outgroupaffect

    Ingroupaffect

    Outgroupaffect

    2000 2001

    Year and target

    Thermometerratings

    Low identifiers

    High identifiers

    Figure 3. Means (;SE) representing the three-way interaction between target, religion and year,

    collapsing across religion.

    In-group and out-group affect 709

  • 7/30/2019 23830409

    10/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    Further post hoc analyses (again, using the corrected alpha level of .005) revealed

    that in the 2000 sample, no simple target effect was found for either Protestant low

    identifiers, F1; 552 5:89, p :015, or Catholic low identifiers, F1; 327 0:199,p :66. In the 2001 sample, however, Catholic low identifiers exhibited a muchstronger target effect, F1; 337 18:46, p , :001, 02 :052, than did Protestant lowidentifiers, F1; 361 4:55, p :034, 02 :012. High identifiers exhibited a strongtarget effect regardless of religion or sample year; means and standard errors can be

    found in Table 1.

    Correlational analyses

    In the preceding sections we used ANCOVA to examine the within subjects differencesbetween in-group affect and out-group affect as a function of the measured factors.Another way to test the ideas set forth in the introduction (see Brewer, 1999; Hewstoneet al., 2002) concerning the strength and direction of correlations between in-group andout-group affect is to compare the correlations between in-group and out-group affectscores across different levels of the theoretically important factors. That is, instead of

    examining mean levels across factors, one can also test the correlational relationships

    between variables across factors. This is a different kind of prediction altogether.As a reminder, we expected (A) the correlation between in- and out-group affect to

    be positive and strong in 2000 as compared with 2001 when intergroup relations were

    worsening. More importantly, however, we expected that the positive correlation

    between the two affect measures should be (B) stronger for low identifiers than for high

    identifiers, and (C) stronger for Catholics than for Protestants. Unfortunately, as

    mentioned, this method of analysis precludes a test of the complex interactive effects

    between year, identification and religion found in the ANCOVA results above. Therefore,

    we are limited to only testing these three simple differences. We tested these three

    simple hypotheses by transforming the correlations to z scores (Fishers

    z-transformation) and then testing their magnitudes.Hypothesis (A) was not supported. The correlation between in-group affect and out-

    group affect was not weaker in 2001, r843 :

    282, p,

    :

    001, as compared with 2000,r1011 :235, p , :001; the difference between these was not reliable, z 1:08,p :28. Next, we found that hypothesis (B) was supported. As expected, the

    Table 1. Mean thermometer ratings (and standard errors) for in-group and out-group targets as a

    function of identification, religion and sample year

    Affect In-group affect Out-group affect

    High identifiers

    Protestant, 2000 (N 300) 72.62a (1.03) 55.31b (1.01)

    Protestant, 2001 (N 188) 83.62a (1.34) 65.47b (1.72)

    Catholic, 2000 (N 223) 74.05a (1.22) 61.36b (1.20)

    Catholic, 2001 (N 219) 84.65a (1.07) 76.70b (1.47)

    Low identifiers

    Protestant, 2000 (N 257) 59.06a (1.11) 55.61a (1.01)

    Protestant, 2001 (N 178) 75.44a (1.38) 71.23a (1.77)

    Catholic, 2000 (N 109) 60.92a (1.76) 61.79a (1.72)Catholic, 2001 (N 123) 79.96a (1.44) 72.10b (1.97)

    Note. Means with different superscripts (within rows) differ at a post hoccorrected level ofp , :005.

    710 Ed Cairns et al.

  • 7/30/2019 23830409

    11/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    correlation between in-group affect and out-group affect was stronger for lowidentifiers, r727 :588,p , :001, than for high identifiers, r1031 :252,p , :001;

    z 8:59, p , :001. Supporting hypothesis (C), we found that the correlation betweenin-group affect and out-group affect was stronger for Catholics, r752 :742,p , :001,than for Protestants, r1026 :271, p , :001; z 14:08, p , :001.

    We note that, regardless of whether or not these correlational findings yield supportfor the respective hypotheses, caution must be exercised in their interpretation. This isbecause, as mentioned above, the analytical method cannot explore interactive effects

    and, thus, moderation of these reported differences is clearly not ruled out. We also notethat a high correlation between these measures says nothing about their respective meandifferences, which was examined more appropriately using the ANCOVA approachabove. Thus, a high correlation is not indicative of less in-group favouritism because the

    means could differ vastly yet still be correlated; nor is a weak correlation indicative ofrelatively stronger in-group favouritism because mean differences might be at similarlevels yet still be uncorrelated. In this context, only a reliable negative correlation wouldyield any useful, theoretical information: positivity towards one group corresponds toantipathy towards the other and vice versa. Thus, more demanding hypotheses wouldpredict that not only will high identifiers and Protestants in-group-out-group affectcorrelations be weaker than those of low identifiers and Catholics, respectively, but thattheir correlations should also be reliably negative, instead of positive. However, nonegative correlations were found in either sample year or for any level of anyother factor.

    Stronger versions of the correlational hypotheses were clearly not supported.

    Discussion

    The major purpose of the present study was to explore in-group bias or favouritism, asmeasured by a mean difference between in-group and out-group affect scores, in a

    highly segmented society and to provide insight into the factors influencing this bias.The study suggests several major factors, namely, variable sectarian tension, levels of in-group identification and religious group status.

    First, as we had predicted, group members showed a clear preference for their in-group, with in-group affect ratings consistently and significantly higher than out-groupaffect ratings in each year. This phenomenon of in-group bias is well established (Hinkle& Brown, 1990; Mullen, Brown, & Smith, 1992) and is consistent with social identitytheory (Tajfel & Turner, 1987).

    We also found the expected interaction between in-group identification and targetratings. In-group bias was stronger for high identifiers than it was for low identifiers (seeFigure 1). This interaction was expected based on a fairly straightforward interpretationof social identity theory, as well as on its empirical history with respect to the differential

    behaviour of low and high in-group identifiers (see Jetten & Spears, 2003).Because the Protestant group in Northern Ireland is the de facto higher-status group

    and because higher-status groups tend to exhibit more bias than lower-status groups(Ellemers, Kortekaas, & Ouwerkerk, 1999; Guimond, Dif, & Aupy, 2002), we expectedthat Protestants would show more in-group bias than Catholics. This expectation wassupported; collapsing across sample year and level of in-group identification, the withinsubjects difference between in-group and out-group affect scores was greater in the

    Protestant sample than in the Catholic sample (see Figure 2).There is a lack of consensus in the literature regarding the effect of status differenceson in-group favouritism. It has been argued that higher status leads to stronger in-group

    In-group and out-group affect 711

  • 7/30/2019 23830409

    12/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    favouritism and in-group identification (Ellemers et al., 1999; Guimond et al., 2002). Incontrast, other researchers have claimed that lower status should lead to stronger in-

    group favouritism and identification in order to counteract a negative social identity(Brewer, 1979; Cameron & Lalonde, 2001; Verkuyten, 2003). Hinkle and Brown (1990)have pointed to empirical evidence in support of both positions. Evidence from meta-analytic studies suggests that higher status leads to stronger in-group bias in artificialgroups, whereas, in the context of real groups, lower status is associated with more in-group bias (Mullen et al., 1992). Our results with real groups in Northern Ireland do not

    support the meta-analytic findings; instead, they support the assertion that higher statusis more associated with in-group favouritism than lower status.

    Previous research has tended to confound dominant versus subordinate group statuswith differential group size (see Ellemers & Van Rijswijk, 1997). Although these factors

    often covary, it is necessary to distinguish between them. An alternative explanation forour findings, then, could be offered in terms of the relative size of the minority Catholiccommunity in Northern Ireland. It is generally accepted in the research literature thatminority group size results in stronger in-group identification than does majority groupsize (Brewer, 1991; Ellemers et al., 1999; McGuire, McGuire, Child, & Fujioka, 1978;McGuire & Padawer-Singer, 1976). Catholics constitute a numerical minority withinNorthern Ireland, albeit an increasing one. It could be argued, therefore, that the resultsof the current study are a function of group size rather than status differentials. Thisargument, however, is difficult to uphold because even the relative status of these two

    groups in Northern Ireland is a subject of ongoing debate.Jackson (1971) postulated a double minority model acknowledging that, while

    Catholics are a minority in Northern Ireland, Protestants are a minority in the whole ofIreland. In contrast, Cairns (1982, 1989) suggested a double majority model Catholicsthink of themselves as a majority in an all-Ireland context, whereas Protestants seethemselves as a majority in the context of Northern Ireland. Trew (1992) concluded thatboth may be correct because, as minorities both groups fear being overpowered and asmajorities they question the legitimacy of their position vis-a-vis the other group(p. 344). This is consistent with recent political developments in Northern Ireland

    where it has been suggested that, broadly speaking, Protestants perceive their in-groupto be losing ground politically, economically and numerically to the Catholiccommunity, relative to its historically dominant position (Irwing & Stringer, 2000).Under such threat, Protestants would thus be expected to act in ways that reaffirm orprotect their dominant position within Northern Ireland. Indeed, as discussed,Protestants showed more overall in-group bias than Catholics.

    Of course, this discussion concerning differences as a function of status per se isrelevant but secondary to the main interactions that we observed in the analyses of thedata. We found that in the relatively peaceful year (2000), Protestants and Catholics

    displayed equivalent amounts of in-group bias. That is, irrespective of religious group,high identifiers showed strong in-group bias, whereas low identifiers did not. In themore conflictual year (2001), however, although we expected both high and lowidentifiers to exhibit more bias (which was confirmed when collapsing across religiousgroup; see Figure 3), this was only true for the Catholic group. Although Protestants in-group and out-group affect scores rose in 2001, as compared with 2000, only Protestanthigh identifiers exhibited strong in-group bias; low identifiers did not. By contrast, for

    the Catholic group in 2001, both high and low identifiers showed strong in-group bias.This finding is consistent with social identity theory which states that when onesgroup is under threat, or when in-group negativity is implicated, individuals will attempt

    712 Ed Cairns et al.

  • 7/30/2019 23830409

    13/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    to reassert positive social identity. Further, although low identifiers are normally

    expected to try to leave the group instead of acting in norm-consistent ways (e.g. in-

    group favouritism or discrimination against the out-group), the group boundaries in

    Northern Ireland are essentially impermeable. Thus, regardless of whether or not one

    identifies with ones religious group, religion is an inescapable element of social identity.

    Note also that the intergroup violence that occurred in 2001 was primarily attributed to

    Nationalist/Catholic paramilitary groups, which means that Catholics in general were

    likely to be perceivedas bearing collective responsibility for the bombings, even if only

    peripherally. Thus, even Catholic low identifiers may have felt some motivation torestore positive in-group identity and displaying relatively higher amounts of in-group

    bias than usual may have served that purpose. By contrast, as the violence and thus the

    source of increased tension was not attributed to Protestant groups, Protestant low

    identifiers probably felt no need to display in-group bias. Interestingly, though, the

    Protestant low identifiers in 2001 did have elevated in-group affect scores, relative to

    their counterparts in 2000. Their out-group affect scores, however, were elevated to the

    same degree (see Table 1).

    Another aim of the present study was to explore the correlation between in-group

    and out-group affect. In line with predictions, there was a positive relationship between

    in-group and out-group affect. This is made clear by the simple correlations where,

    whether one calculates the in-group/out-group correlation for all respondents or

    separately for Catholic respondents and Protestant respondents, the outcome in eachyear is always a statistically significant, though modest, positive correlation. This finding

    is consonant with Brewers (1979, 1999 and 2001) and Nesdales (2004) contention that

    in-group favouritism is not inevitably related to out-group derogation (Bennett et al.,

    2004). It does not, however, confirm the prediction of a negative correlation in highly

    conflictual and segmented societies, where factors such as strong in-group identification

    and loyalty, social comparison processes and sensitivity to threat and power politics all

    encourage and promote out-group hostility (Brewer, 1999). It is likely, however, that

    contemporary Northern Ireland is progressing towards being a society where out-group

    hostility is openly and normatively discouraged (see McLernon, Ferguson, & Cairns,

    1997). Thus, further research with implicit measures of bias (e.g. the Implicit

    Association Test; Greenwald, McGhee, & Schwartz, 1998) may be required for a stronger

    test of this hypothesis.

    Contrary to expectation, the positive correlation between in-group and out-groupaffect was not weaker in 2001. This finding is also contrary to Levin and Sidanius (1999)

    and Hewstone et al.s (2002) suggestion that increased intensity of conflict should lead

    to a negative correlation between in-group and out-group affect. Nevertheless, we

    should note that, although the political tensions were greater in 2001 than in 2000, both

    years are still subsequent to the peace agreement signed in 1998 and Northern Ireland

    can be considered a post-conflict area in many respects. In effect, the relatively high

    degree of political tension in 2001 was still far lower than in the 1970s and 1980s so the

    prediction regarding the intensity of intergroup tensions may not necessarily apply. It is

    also conceivable, however, that this finding is more reflective of fairness norms or a

    tendency to give positive scores to both groups; consequently, out-group affect will

    correlate with in-group affect (Verkuyten, 2003). If this is the case, out-group affect

    would vary as a function of the increase in in-group affect, rather than reflecting agenuine increase in the relation between these two variables in the two years. As noted,

    the reported correlational results for the effect of sectarian tension on the relationship

    In-group and out-group affect 713

  • 7/30/2019 23830409

    14/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    between in-group and out-group affect cannot be interpreted unequivocally with

    respect to their implication for bias per se.Another key prediction in the correlational analyses was that the relationship

    between in-group and out-group affect would be moderated by both the level of in-

    group identity (weaker for those with high levels of in-group identity) and group status

    (weaker for the historically dominant Protestant group). These predictions received

    strong support. Overall, we have shown support for the hypothesis that the relationship

    between in-group and out-group affect is generally positive but that it is moderated by in-

    group identity and status. This is consistent with the growing amount of researchsuggesting that feelings of in-group positivity do not necessarily imply out-group

    antipathy (Allport, 1954; Brewer, 1979, 1999, 2001; Nesdale, 2004).

    Conclusions

    We believe the results of our study contribute to the ongoing policy debate concerninghow best to fashion government initiatives aimed at improving intergroup relations inNorthern Ireland and elsewhere. This revolves around the need to draw a cleardistinction between bias, which is calculated as the difference between in-group affectand out-group affect, and the correlational relationship between the two variables. Ourfindings demonstrate for the first time that, within the context of an overall positivecorrelation between in-group and out-group affect, bias can and does exist when factors

    such as in-group identification, status and increasing intergroup tension are taken intoaccount and examined appropriately. In other words, an observed positive correlation

    between in-group affect and out-group affect does not guarantee the absence of biaswithin a population. Such a key finding should be taken into account when establishinginterventions aimed at conflict prevention or reduction (Brewer, 2001).

    In contrast to our use of thermometer ratings, there are clearly more sophisticated

    measures of in-group bias or prejudice and future studies could benefit from a richer

    conceptualization of intergroup attitudes. Recent evidence suggests that both in-group

    identification (Ashmore, Deaux, & McLaughlin-Volpe, 2004; Ellemers et al., 1999; Simon

    & Klandermans, 2001) and prejudice (Duckitt, Wagner, Duplessis, & Birum, 2002; Fiske,

    Xu, & Cuddy, 1999; Pettigrew & Meertens, 1995) are multidimensional components

    with different dimensions that are predictive of different responses to out-groups. Also,

    there is the possibility that participants may have been influenced in the present study

    by the fact that in-group ratings were always made before the out-group was evaluated.Previous research has shown that rating the out-group first can increase social

    categorization and intergroup comparison, resulting in greater bias compared with

    methodologies in which the in-group is rated first (see Oakes, Haslam, & Turner, 1994).

    Given the fact that we were analysing secondary data, these potential order effects could

    not be overcome in the present study but should be controlled in future research.

    In sum, this research has examined the role of in-group identification, religious group

    membership and group status in moderating the relationship between in-group and out-

    group affect across two years with different levels of sectarian tension. This study

    further substantiates the view that in-group and out-group affect scores are not

    necessarily negatively correlated and highlights the influence of in-group identification,

    group status and sectarian tension. Future research using alternative measures and

    counterbalancing the order of in-group and out-group ratings by adopting strictercontrol of the relevant variables can provide further insight into the various processes

    by which these variables operate in the context of intergroup conflict.

    714 Ed Cairns et al.

  • 7/30/2019 23830409

    15/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    References

    Allport, G. W. (1954). The nature of prejudice. Cambridge, MA: Addison-Wesley.

    Ashmore, R. D., Deaux, K., & McLaughlin-Volpe, T. (2004). An organizing framework for collective

    identity: Articulation and significance of multidimensionality. Psychological Bulletin, 130,

    80114.

    Bennett, M., Barrett, M., Karakozov, R., Kipiani, G., Lyons, E., Pavlenko, V., & Riazanova, T. (2004).

    Young childrens evaluations of the in-group and of out-groups: A Multi-National Study. Social

    Development, 13, 124139.

    Brewer, M. B. (1979). In-group bias in the minimal intergroup situation: A cognitive motivationalanalysis. Psychological Bulletin, 86, 307324.

    Brewer, M. B. (1991). The social self: On being the same and different at the same time.

    Personality and Social Psychology Bulletin, 17, 475482.

    Brewer, M. B. (1999). The psychology of prejudice: In-group love or out-group hate? Journal of

    Social Issues, 55, 429444.

    Brewer, M. B. (2001). In-group identification and intergroup conflict: When does in-group love

    become out-group hate? In R. Ashmore, L. Jussim, & D. Wilder (Eds.), Social identity,

    intergroup conflict, and conflict reduction (pp. 1741). New York: Oxford University Press.

    Brewer, M. B., & Kramer, R. M. (1985). The psychology of intergroup attitudes and behaviour.

    Annual Review of Psychology, 36, 219243.

    Brown, R., Condor, S., Mathews, A., Wade, G., & Williams, J. (1986). Explaining intergroup diff-

    erentiation in an individual organization. Journal of Occupational Psychology, 59, 273286.

    Cairns, E. (1982). Intergroup conflict in Northern Ireland. In H. Tajfel (Ed.), Social identity and

    intergroup relations (pp. 277297). Cambridge: Cambridge University Press.Cairns, E. (1989). Social identity and intergroup conflict: A developmental perspective.

    In J. Harbison (Ed.), Growing up in Northern Ireland (pp. 115130). Belfast: Stranmillas.

    Cairns, E., & Darby, J. (1998). The conflict in Northern Ireland: Causes, consequences and

    controls. American Psychologist, 53, 754760.

    Cameron, J. E., & Lalonde, R. N. (2001). Social identification and gender-related ideology in women

    and men. British Journal of Social Psychology, 40, 5977.

    Duckitt, J., Wagner, C., DuPlessis, I., & Birum, I. (2002). The psychological bases of ideology and

    prejudice: Testing a dual process model. Journal of Personality and Social Psychology, 83,

    7593.

    Ellemers, N., Kortekaas, P., & Ouwerkerk, J. W. (1999). Self-categorization, commitment to the

    group and group self-esteem as related but distinct aspects of social identity. European

    Journal of Social Psychology, 29, 371389.

    Ellemers, N., & Van Rijswijk, W. (1997). Identity needs versus social opportunities: The use of

    group level and individual level identity management strategies. Social Psychology Quarterly,60, 5265.

    Feschbach, S. (1994). Nationalism, patriotism, and aggression: A clarification of functional

    differences. In L. R. Huesmann (Ed.), Aggressive behavior: Current perspectives

    (pp. 275291). New York: Plenum.

    Fiske, S. T., Xu, J., & Cuddy, A. C. (1999). (Dis)respecting versus (dis)liking: Status and

    interdependence predict ambivalent stereotypes of competence and warmth. Journal of

    Social Issues, 55, 473489.

    Greenwald, A. G., McGhee, D. E., & Schwartz, J. K. L. (1998). Measuring individual differences in

    implicit cognition: The implicit association test. Journal of Personality and Social

    Psychology, 74, 14641480.

    Guimond, S., Dif, S., & Aupy, A. (2002). Social identity, relative group status and intergroup

    attitudes: When favourable outcomes change intergroup relations: : :for the worse. European

    Journal of Social Psychology, 32, 739760.

    Haddock, G., Zanna, M. P., & Esses, V. M. (1993). Assessing the structure of prejudicial attitudes:The case of attitudes toward homosexuals. Journal of Personality and Social Psychology, 65,

    11051118.

    In-group and out-group affect 715

  • 7/30/2019 23830409

    16/17

    Copyright The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

    Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review of Psychology, 53,

    575604.

    Hinkle, S., & Brown, R. (1990). Intergroup comparisons and social identity: Some links and

    lacunae. In D. Abrams & M. Hogg (Eds.), Social identity theory: Construction and critical

    advances (pp. 4870). London: Harvester Wheatsheaf.

    Irwing, P., & Stringer, M. (2000). New measures of political attitudes in Northern Ireland: A social

    identify perspective. Journal of Community and Applied Psychology, 10, 139154.

    Jackson, H. (1971). The two Irelands - A dual study of intergroup tensions . London: Minority

    Rights Group Report, No. 2

    Jetten, J., & Spears, R. (2003). The divisive potential of differences and similarities: The role ofintergroup distinctiveness in intergroup differentiation. European Review of Social

    Psychology, 14, 203241.

    Levin, S., & Sidanius, J. (1999). Social dominance and social identity in the United States and Israel:

    In-group favouritism or out-group derogation? Political Psychology, 20, 99126.

    McGuire, W. J., McGuire, C. V., Child, P., & Fujioka, T. (1978). Salience of ethnicity in the

    spontaneous self-concept as a function of ones ethnic distinctiveness in the social

    environment. Journal of Personality and Social Psychology, 33, 743754.

    McGuire, W. J., & Padawer-Singer, A. (1976). Trait salience in the spontaneous self-concept.

    Journal of Personality and Social Psychology, 33, 743754.

    McLernon, F., Ferguson, N., & Cairns, E. (1997). Comparison of Northern Irish childrens attitudes

    to war and peace before and after the paramilitary ceasefires. International Journal of

    Behavioral Development, 20, 715730.

    Mullen, B., Brown, R. J., & Smith, C. (1992). In-group bias as a function of salience, relevance and

    status: An integration. European Journal of Social Psychology, 22, 103122.Mummendey, A., Klink, A., & Brown, R. (2001). Nationalism and patriotism: National

    identification and out-group rejection. British Journal of Social Psychology, 40, 159172.

    Nesdale, D. (2004). Social identity processes and childrens ethnic prejudice. In M. Bennett &

    F. Sani (Eds.), The development of the social self (pp. 219245). New York: Psychology Press.

    Niens, U., & Cairns, E. (2001). Intrastate violence. In D. J. Christie, R. V. Wagner, & D. D. Winter

    (Eds.), Peace, conflict and violence (pp. 3948). New Jersey: Prentice-Hall.

    Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social reality. Oxford:

    Blackwell.

    Pettigrew, T. F., & Meertens, R. W. (1995). Subtle and blatant prejudice in Western Europe.

    European Journal of Social Psychology, 25, 5775.

    Simon, B., & Klandermans, B. (2001). Politicized collective identity: A social psychological

    analysis. American Psychologist, 56, 319331.

    Sumner, W. G. (1906). Folkways. New York: Ginn.

    Tajfel, H. (1982). Social psychology of intergroup relations. Annual Review of Psychology, 36,

    219243.

    Tajfel, H., & Turner, J. C. (1986). The social identity of intergroup behaviour. In S. Worchel &

    W. G. Austin (Eds.), Psychology of intergroup relations (pp.724). Chicago: Nelson Hall.

    Trew, K. (1992). Social psychological research on the conflict. Psychologist, 5, 342344.

    Verkuyten, M. (2003). Ethnic in-group bias among minority and majority early adolescents: The

    perception of negative peer behaviour. British Journal of Developmental Psychology, 21,

    543564.

    Whyte, J. (1991). Interpreting Northern Ireland. Oxford: Clarendon Press.

    Wilder, D. A. (1986). Social categorization: Implications for creation and reduction of intergroup

    bias. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 19,

    pp. 293355). New York: Academic Press.

    Received 30 January 2004; revised version received 16 August 2005

    716 Ed Cairns et al.

  • 7/30/2019 23830409

    17/17