Journal of Occupational Psychology, I986,59,273-286.

Printed in Great Britain

0 1986 The British Psychological Society

Explaining intergroup differentiation in

an industrial organization
R U P E R T BROWN,* S U S A N C O N D O R , A U D R E Y M A T H E W S ,
GILLIAN WADE AND JENNIFER WILLIAMS
Social Psychology Research Unit, University of Kent

We report a study of intergroup relations in a paper factory in which we examine the

utility of three social-psychological approaches: realistic conflict theory, the contact
hypothesis and social identity theory. A sample of 177 shop floor workers from five
different departments was interviewed. From them, measures of intergroup differen-
tiation, perceived intergroup conflict, amount of intergroup contact, and strength of
workgroup identification were obtained. The latter was assessed using a new scale of
group identification developed for this study. Reliability and validity data for this scale
are reported. Using multiple regression analyses we attempt to explain variance in
respondents' intergroup differentiation using the other measures as predictor variables.
The most powerful and reliable predictor was perceived conflict which, as expected,
was positively correlated with differentiation. Less consistent was amount of contact
which was negatively but only weakly associated with differentiation. Strength of
group identification, while generally showing a positive correlation with differen-
tiation as predicted, was also only a weak and inconsistent predictor variable. Noting
that these results confirm findings from other studies we discuss their theoretical
implications.

Intergroup rivalry within organizations is a common phenomenon. Whether it takes the

form of outright conflict between management and union, or merely some biased percep-
tions and judgements on the part of different departments or workgroups, anecdotal and
research evidence as to its prevalence abounds (e.g. Blake et al., 1964; Strauss, 1964;Nichols
& Armstrong, 1976; Alderfer & Smith, 1982). These intergroup relationships can usefully
be studied at several different levels-the economic, political, organizational and social-
psychological-each of which may impinge to varying degrees on the others. Our concern in
this paper is with the last level of analysis. We report a study of a medium-sized industrial
organization comprising a number of subgroups among which intergroup differentiation
was clearly visible. We endeavour to explain that differentiation by reference to three
different but complementary social-psychological approaches.
The first approach to be considered is the realistic group conflict theory (RCT:
Campbell, 1965), most forcibly espoused within social psychology by Sherif (1966). This
theory proposes that intergroup attitudes and behaviour reflect the state of objective
intergroup goal relations. Thus, when these are antagonistic-for example, in conflicts
between union demands for higher wages and shareholders' desire for increased dividends-
intergroup behaviour is likely to be discriminatory and hostile. When, however, groups'
goals coincide-for example, two unions' common interest in securing improved safety
standards in a factory-then their members' behaviour is thought to become more concili-
atory and amicable. An ambiguity in the theory is whether the objective relationship of
interdependence is the critical variable here, or merely perceived interdependence (Turner,

*Requests for reprints should be addressed to Rupert Brown, Social Psychology Research Unit, Beverley
Farm, University of Kent, Canterbury, Kent, UK.

273
274 RUPERT B R O W N ETAL.

1981; Condor & Brown, in press). RCT seems to assume a one-to-one correspondence
between objective and perceived interdependence(Sherif, 1966, p. 1 9 , but clearly there can
sometimes be a mismatch. For instance, in capitalist economies there is objectively a con-
flict between management and the workforce, and yet the workforce's perception of that
relationship may not always correspond to this. Nevertheless, within organizational con-
texts the theory has received fairly consistent support, most notably from the work of Blake
and Mouton (e.g. Blake et af.,1964; Blake & Mouton, 1979). They have found that groups
which adopt win-lose orientations in intergroup encounters invariably manifest marked
in-group bias in evaluativejudgements and show affectively negative attitudes towards the
out-group [see also Brown & Williams (1984) for further supportive evidence]. This inter-
group differentiation was sometimes, though not invariably, reduced when group members
perceived common or superordinate goals linking the groups (Blake et af., 1964). The
prevalence of competitive intergroup relationships and the occasional ineffectiveness of
superordinate goal strategies for reducing conflict has been corroborated in other studies
(Brown, 1978; Skevington, 1980), and suggests that, by itself, the RCT approach may not
be sufficient to explain intergroup behaviour in organizations. It seems that additional
processes may be at work, over and above the purely instrumental factors implicated by
RCT.
Some of these processes have been held to stem from the amount of contact which
groups have with each other. This view derives from a theoretical tradition within ethnic
relations research known as the contact hypothesis (Allport, 1954; Pettigrew, 1971), and is
the second main approach considered here. Put simply, the contact hypothesis holds that
contact between members of different groups, in the appropriate conditions, can lessen
intergroup discrimination and hostility. One reason suggested for this is that contact allows
the discovery of similarities of value and belief which are generally found to lead to attrac-
tion (Pettigrew, 1986) [but see Brown & Turner (1981), Hewstone & Brown (1986), and
Pettigrew (in press) for recent analyses and critiques of this hypothesis]. Although this
approach is most commonly associated with ethnic relations research, Allen (1986)
persuasively argues that similar assumptions are implicit in some organizational research
concerned with job satisfaction and industrial conflict. For example, versions of the
so-called size-militancy hypothesis (e.g. Ingham, 1967) offer an explanation of industrial
behaviour essentially in terms of the amount and quality of worker-worker and worker-
employer contact. In large organizations there is usually more bureaucracy which is
thought to lessen the opportunities for informal contact between management and workers
and result in a more militant workforce (see also Allen, 1986). A similar rationale can
be discerned behind Kerr & Siegel's (1954) analysis of the propensity to strike in terms
of the ' isolation ' of different industries (Parkin, 1967; Allen, 1986). Although these
approaches have not won uncritical acceptance within the industrial relations literature
(e.g. Edwards 1977, 1980; Allen, in press), they do broadly suggest that there should be a
positive correlation between the amount of contact between groups and the favourability of
intergroup attitudes.
A third social-psychological perspective is offered by social identity theory (SIT:
Tajfel, 1978). This theory proposes that a potentially important component of people's
identities derives from their group memberships and that these social identities are main-
tained primarily by means of intergroup comparisons. The theory suggests that people
often seek to establish a positive identity in these comparisons, and that this search
expressesitself in the form of positive differentiationin perceptions, attitudes and behaviour
(Tajfel, 1978). A good example of this is provided by the concern of some highly skilled
industrial workers to preserve (often quite small) wage differentialsover other workgroups,
sometimes even at the expense of their own absolute wage levels (Brown, 1978). The social
identity approach can thus complement RCT by helping to explain the widespread occur-
rence of intergroup differentiation,even in the absence of conflicting group interests (Tajfel
 E X P L A I N I N G I N T E R G R O U P DIFFERENTIATION 275

8c Turner, 1979). A positive identity is, of course, easier to achieve by higher status groups
since they are consensually more favourably regarded. But lower status groups may not
inevitably confer a negative identity on their members, particularly if the status hierarchy is
in a state of flux or is being challenged in some way (Tajfel & Turner, 1979). Research in
organizational contexts using the SIT framework has been rather sparse. Apart from some
studies of intergroup pay and status comparisons (Brown, 1978; Skevington, 1981; Bourhis
& Hill, 1982), The authors have found only one study to date which has directly examined
the link postulated within the theory between group identification and intergroup differen-
tiation (Brown & Williams, 1984). According to the theory, group identification and the
tendency to perceive the in-group as positively distinct should covary. However, Brown 8c
Williams (1984) found only limited support for this hypotheses since the predicted positive
association between identification and differentiation was visible in only some of the work-
groups they studied. However, their findings must be considered as tentative in view of the
methodological difficulties associated with their rather small sample and their measure of
group identity. One of the objectives of the present study was to build on this preliminary
study by developing a more robust measure of group identification and to use this with a
substantially larger sample.
In summary, therefore, we set out to explore the utility of these three approaches in
making sense of one particular intergroup context. According to RCT there should be a
positive relationship between perceptions of conflictual goals and discriminatory inter-
group attitudes. According to the contact hypothesis greater contact should be associated
with less differentiation. And, finally, SIT suggests that the strength of a person’s affiliation
to a social group should predict the amount of positive intergroup differentiation.

DEVELOPMENT OF A MEASURE OF G R O U P I D E N T I F I C A T I O N
As noted above, the third approach implies a relationship between social identification
and the tendency to differentiate the in-group positively from other groups. Within SIT
social identity is defined as:

. . . that part of an individual’s self concept which derives from his [or her] knowledge of his [or her]
membership of a social group (or groups) together with the value and emotional significance attached to
that membership (Tajfel, 1978. p. 63).

In this definition it is possible to pick out three facets: awareness of group membership
(which contributes to self-definition), evaluation (which relates to self-esteem) and affect.
Although this tripartite division is not completely clear cut it did provide us with a blueprint
to search for an appropriate measure of identification. Existing techniques for assessing
identity (e.g. Kuhn & McPartland, 1954; Zavalloni, 1971; Weinreich, 1975; Liebkind, 1982)
seemed unsatisfactory for several reasons: they tended to focus rather exclusively on the
first aspect of identity (awareness) at the expense of the other two; they do not yield very
simple quantitative indices; and they can be impractically long to administer.
We therefore developed a new instrument, based on an original scale of ethnic identity
devised by Driedger (1976).This is a 10-item inventory which consists of five items affirming
group identification in various ways and five items denying it (see Table 1). The items also
attempt to tap the three aspects of identity referred to above. (Awareness is tapped by items
2 and 5; evaluation by items 1,6,7 and 10; affect by items 3,4,8 and 9.)
The scale was piloted on a sample of 137 department store employees. On the basis of
an item analysis some of the original items were modified or replaced to yield the final
selection shown in Table 1. In a later section we report on the scale’s reliability and validity.
276 RUPERT BROWN E T A L .

Table 1. Items in group identification scale (IDENTIFY)

1. I am a person who considers the - group important.

2. I am a person who identifieswith the- group.
3. I am a person who feels strong ties with the - group.
4. I am a person who is glad to belong to the - group.
5. I am a person who sees myself as belonging to the- group.
6. I am a person who makes excuses for belonging to the- group.
7. I am a person who tries to hide belongingto the- group.
8. I am a person who feels held back by the - group.
9. I am a person who is annoyed to say I'm a member of the- group.
10. I am a person who criticizes the - group.

Notes. Items are presented in random order.

Each item is answered on the following five-point scale:
Never Seldom Sometimes Often Very often
The scores for the last five items are reversedto give a possible range of 10-60 for the whole scale.

THE RESEARCH SITE

The study was carried out in the winter of 1983 in a medium-sized paper mill in the
South of England. Access was obtained to the factory initially through the Personnel
Manager and subsequently through the trade union committee representing the workforce.
The factory made a wide variety of paper products for distribution to the packaging and
print industry. It was owned by a large multinational company and at the time of the study
employed some 600 manual staff and a further 100 white-collar and managerial staff. Three
months prior to the start of the study the company announced a large redundancy pro-
gramme which resulted in the loss of some 300 jobs mainly through early retirement and
voluntary redundancy. The company's justification for these redundancies was the general
economic recession. This had particularly acute effects in the paper industry where the
company was already under strong pressure from its competitors overseas. According to
half-yearly figures published just before the study, the company was reported to be making
a loss on its British paper-making operations, with a 14 per cent drop in sales over the
period. This, then formed the economic background to our study.
Organizationally, the factory was rather complex. There were five main departments
which are described in the order in which they occurred in the various stages of the paper-
making process. Paper making was concerned with the initial production from pulp making
to the stage where the rough quality paper is ' coated ' with a fine white clay. This department
was the largest of the five and its employees tended to be the highest paid, particularly the
highly skilled ' machine men ' controlling the main paper-making machines (basic wage
El 50 per week at the time of the study). Other grades in this department earned in the region
of €100 to El20 per week. Finishing polished the paper using large roller machines called
calenders. This was the third largest section and workers here earned between E90 and E l 10
per week. Conversion transformed the large reels of paper into the required sizes for the
customer. This was somewhat larger than Finishing but wages were roughly comparable.
The Salle, a quality control department, was the second smallest in size, and employed
mainly women, many of them on a part-time basis. Wages here tended to be somewhat
lower, in the region of E85 to El00 per week. The department known as the Wharf was
responsible for loading the packaged paper on to lorries for dispatch to the customer.
Geographically separated from the other four departments, it was the smallest in size.
Wages here were similar to those in Conversion and Finishing. Because of the linear nature
of the paper-making process these five groups can be regarded as functionally and positively
 EXPLAINING INTERGROUP DIFFERENTIATION 217

interdependent. A blockage in any one department has consequences for groups both
earlier and later in the production process, most immediately, of course, for the adjacent
group(s). Within each department, and especially the first four, there were also important
divisions according to which machine or process a person worked on. Cutting across this
classification by job function were a variety of shift patterns which differed across depart-
ments. In all there were five different shift systems each of which contained two or more shift
groups. Finally, to complete the complexity, the factory was virtually 100 per cent
unionized with the vast majority of workers belonging to the Society of Graphical and
Allied Trades (SOGAT 82). This categorization therefore subsumed all the other divisions,
with one or two exceptions-particularly those who belonged to the craft unions.
Faced with this multiplicity of groupings we eventually selected the departmental
group to be the focus of the study. This decision was influenced by a number of consider-
ations. One was to maintain continuity with an earlier study (Brown & Williams, 1984),
which had studied workgroup identity, and with our pilot investigation in which the inter-
view schedule and instruments had been developed. Although in preliminary discussions
with both management and shop floor workers it was apparent that people’s shift member-
ship was also an important source of identity, the variety of shift arrangements meant that
cross-factory comparisons would be impossible. The possibility of studying union affili-
ations was also considered but then rejected once it became apparent that the union was not
very salient for most of the workforce.

METHOD

Respondents
A sample of 177 respondents was randomly selected from the five departments. Over
three-quarters of those originally selected from the company’s employment roll agreed to
participate in the study. The remainder were unable to either because of sickness, incompat-
ible shift patterns, or (in a few instances only) because of outright refusal. These gaps in the
sample were filled with a further random selection from the roll. The distribution of the
sample across, and the proportion within, departments was as follows: Paper making
(n= 39,23 per cent), Finishing (n = 37,37 per cent), Conversion (n = 40,32 per cent), Salle
(n= 37,52 per cent), Wharf (n = 24,83 per cent). With the exception of the Salle which was
all female, the respondents were male. The mean age of respondents was 35 years (range
2CL6 I years), and the median length of service at the factory was 4 years (range one month to
35 years).

Interviews
The data were collected at the factory over a period of four months by one of the
research team (A. M.). The interviews, which were tape-recorded, were conducted during
working time and lasted around 30 minutes. After an introduction to the project by the
interviewer, who explained that we were ‘ looking at groups and group behaviour ’, the
interview began with some factual questions about the respondent’s personal history (age,
length of service, etc.). There were then some questions designed to explore which groupings
within the factory were perceived by the respondent, and which he/she felt most a part of,
and why. This was followed by both open-ended and structured questions assessing the
nature of people’s identifications with these groups, including our identification scale. We
then enquired about people’s perceptions and feelings about other groups in the factory,
first in an open-ended way, and then in a series of rating scales tapping particular dimen-
sions. The stimulus groups used were the five departments under study, and in addition we
included the management and the engineers. In the different sets of ratings the stimulus
278 R U P E R T B R O W N ETAL.

groups were randomly ordered for each respondent. Finally, some questions were asked
about the redundancies.*

Reliability and validity of group identification measure

Item analysis of the scale yielded a Cronbach's alpha of 0.71, indicating a fair measure
of inter-item correlation. Factor analysis yielded an oblique solution with three intercorre-
lated factors ( r , = -0.22, r , = - 0.48, r23 =0*20)-0ne loading on the five affirmative
items (nos. 1-5 in Table 1). and the others loading on the five denial items (the second
loading on nos. 6-8; the third on 10 and 8-9 respectively)--so that it seems reasonable to
summate across items (reversing nos. 6-10) to produce a single scale score. A check on the
validity of the scale was provided by comparing the scores from respondents whose spon-
taneous comments about their group had been classified as positive, with those whose
comments were neutral/negative. The two groups had significantly different scale scores
(42.9 vs. 38.7: t = 5.4, d.f. = 174, P<O-Ool), suggesting that the inventory was indeed
tapping aspects of people's affinity to their group. The distribution of the scores was
approximately normal, although slightly skewed towards the upper end of the scale.

RESULTS

The results are presented in three sections. First we present some of the findingsrelating
to intergroup differentiation in the factory. It was this which we aimed to explain by means
of our three main independent variables derived from the theoretical approaches discussed
earlier: perceived conflict, intergroup contact and in-group identification. It is this final
variable which is dealt with in the second section. Finally, in the third section we report the
regression analyses which we used to explore the utility of our three independent variables.

Intergroup differentiation
In this organization, as in many others, intergroup differentiation in people's attitudes
was very easy to elicit. This was clear from the unstructured comments on the different
groups which, although they were fairly favourable about other departments, were even
more favourable when referring to their own group. Furthermore, comments about other
groups not infrequently included a mildly disparaging remark (e.g. ' I think they damage
too many reels down there '; ' I think they're a bit out of touch '; ' Some of the paper they
turn out is a bit rubbishy '). The one out-group target which most consistently elicited these
negative sentiments was Management.
These impressionswere confirmed by an analysis of our principal dependent measure:
each subject's estimation of the contribution made by each of the stimulus gr0ups.f As
Table 2 and the associated analysis of variance reveal there were some very clear patterns in
these ratings. The most powerful effect was clearly the main effect due to stimulus target. As
just noted, all departments rated Management least favourably, followed by the Salle and
the Engineers. On the whole, the Paper makers received the highest ratings which was a
reflection of their consensually high status position in the factory. However, despite some
broad agreement between departments over these ratings, the existence of a highly signifi-
cant interaction effect indicates that this consensus was far from perfect. The single biggest
contributor to this interaction was the rating given by the respondents to their own group.
These were uniformly positive (all 2 6 on the seven-point scale) and were always the highest
*Not all the data collected are reported here. Details of the interview schedule and their findingsare available
from the authors on request.
tThis measure has been profitably used in a previous investigation (Brown & Williams, 1984) and in the pilot
work showed itself to be easily comprehended by respondents. Essentially, it provides a value-laden but
slightly ambiguousdimension along which subjectscan make intergroupcomparisons.
 E X P L A I N I N G I N T E R G R O U P DIFFERENTIATION 279

Table 2. Estimated contribution to the company of each out-group

Stimulus group being rated

Respondents Paper Conv. Finish Salle Wharf Man. Eng. Mean

Paper makers 6.3, 5.3, 6.0,, 4.ECd 5.5, 4.7, 5.8.b 5.5
-
(n=35)
Conversion 6.2. 6.0,
- 5.8, 5.9. 5.7. 4.9, 5.8, 5.8
(n=40)
Finishing 6.1, 4.7, 6.1, 4.6, 5.3b 4.4, 5.0, 5.2
(n=34)
-
Salle 5.8.k 6.1,, 5.2. -
64, 5.4, 5.2, 5.4, 5.6
(n= 29)
Wharf 6.5, 6.0.b 6.l,, 5.5, 6.7,
- 4.7, 5,8,, 5.9
(n=19)
Mean 6.2, 5.6, 5*8b 5.4, 5.6, 4.Ed 5.5, 5.6
(n=157)

Notes. Contributions were estimatedon a 1-7 scale anchored by the statements ' not much at all ' to ' a
great deal '.
Some respondentscould not be included in this analysis because of a large amount of missing data on
this measure: Where any respondent had >2 missing observations he/she was excluded. Other
occasional missing data points were estimated using the group mean for that cell.
ANOVA summary (5 x 7 mixed design with unequal ns):
Department F=3.2, d.f. =4,152, P<0.05
MSE between = 5.2
Stimulus F=24.9,d.f.=6,912, P<O.OOl
DxS F=4.9.d.f.=24.912, PiO.001
MSE within = 1.1
Means in the same row not sharing a subscript are significantly different from one another by Tukey's
HSD procedure (P<0.05). Means in italics and underlinedwere ratings given by a group to itself.

or equal highest in each row. Furthermore, there were considerable variations in the depart-
ments' intergroup ratings. Thus, Paper makers and Finishing particularly derogated the
Salle, whilst other groups were more favourable towards this target. In addition, the
range between the highest and lowest ratings from the Wharf respondents was a full two
scale points, whilst for Conversion it was only 1.3. It was these variations in intergroup
differentiation which we now endeavour to explain.

Group identification
Since one of our three explanatory theories invokes group identification as a causal
factor, in this section we examine the nature of people's group identifications in the factory.
We report first on data which substantiate our initial decision to focus on the workgroup
as a way of identification. We then describe what that workgroup identification meant to
our respondents.

The perceived group structure within the organization. One of the first questions in the
interview asked the respondent to analyse the factory in terms of the groupings she/he
perceived. The modal response to this question, made by just under a third of the sample
(30.5 per cent), was to describe the organization in terms of the five main departments. A
further 29 per cent described groupings which were either subdivisions within these larger
categories or constituted just a subset of the five departments. It is apparent, therefore, that
some 60 per cent of the sample organized the factory psychologically along similar category
280 RUPERT B R O W N E T A L .

lines as we had chosen to do ourselves. Of the remainder, a quarter (25.4 per cent) either
could give no answer to the question* or simply saw the factory as one single unit, 9 per cent
identified shift groupings,? and 6 per cent mentioned ' staff' and ' workers ' as the salient
division.
The meaning of group identijcation. Immediately following this question respondents were
asked which group they felt most a part of and what were the most important things about
that group membership. Invariably, respondents located themselves in the relevant depart-
mental or subdepartmental grouping. (Those who had answered ' Don't know ' to the
previous question were asked to answer this in relation to their departmental group.) The
most common sentiment associated with this affiliation, mentioned by over half (53.6 per
cent) of the sample, was the importance they attached to the interpersonal friendships in
their group (e.g. ' Just a friendly bunch of people '; ' I suppose, in a sense, they're like a
second family '; ' A certain amount of cameraderie with the lads '). This was the modal
response for every group but was particularly evident in Conversion, Finishing and the
Salle. The next most common answer was a null or negative response. Seventeen per cent
replied that nothing was important about their workgroup (e.g. ' It's just a place of work
really '), and a further 9 per cent expressed clearly negative sentiments (e.g. ' My group is
miserable. Me and my mate don't get on with some of 'em in there '). Finally, 14 per cent
expressed views indicating that they derived some individual job satisfaction from the group
membership, an answer particularly in evidence in the Wharf (e.g. ' Job satisfaction-my
job looks after the whole mill more or less '). The remaining small minority of responses
mentioned either aspects of workgroup interdependence and cohesion, or the reverse, i.e. a
feeling of isolation.
These responses were further analysed to assess whether they expressed a broadly
positive view of the in-group, or a negative/neutral evaluation. Two independent judges
coded each response into one of these two broad categories, agreeing on 8 1 per cent of the
cases (Cohen's kappa =0.64). Disagreements were resolved by the first author. Just over
half (53 per cent) of the statements were classified as clearly positive, and this proportion did
not differ noticeably among the five departments k2= 1 '2, n.s.). This broadly positive social
identification was confirmed by an analysis of respondents' scores on the identification
scale. The overall mean was 41.0 (SD = 5.4)with some small variations among departments
x, x,
(2,= 40.0, =4.29, =40.5, = 39.4,% , ' =42.4; F= 2.9, d.f. =4, 17 1, P <0.05; no single
pairwise comparison was significant by Tukey's HSD procedure).
Explaining intergroup diflerentiation
For each respondent six indices of intergroup differentiation were computed. This was
accomplished by subtracting each respondent's rating of the out-group targets from the
rating of his/her in-group on the dimension of estimated contribution to the factory (see
Table 2). These were used as dependent measures in multiple regression analyses. The
independent variables used to explain variation in these indices were:
(1) Perceived conflict with the target group (CONFLICT). Analysis of these conflict
ratings indicated considerable variation among the groups (see Table 3). Although the
perceptions were highly consistent within departments (as indicated by the significant analy-
sis of variance effects), the departments revealed rather different patterns. It is interesting to
*This was despite repeated attempts by the interviewer by rephrasing it if necessary.The initial question was
phrased as follows: ' Thinking of [nameof factory]as a whole, how would you ' break it down '?Are there any
groups of people within the factory that you can think of?' We suspect that, for a variety of cultural and
linguistic reasons, the concept of ' group ' per se is not widely used by the public. This, of course, does not
mean to say that particular social groups, including the workgroups studied here, cannot have a profound
influenceon people.
tFrom other remarks made during the interviews and our own observations, we believe that shift groupings
were more important for our respondentsthan this low frequency would imply.
 EXPLAINING INTERGROUP DIFFERENTIATION 28 1

Table 3. Perceived conflict between in-group and each out-group'

Stimulus group

Respondents Paper Conv. Finish Salle Wharf Man. Eng. Fvalue

Paper makers 3.4. 2.6, 3.6, 3.1.b 3.6. 2.4, 7.4"'

Conversion 3.5. 3.1, 1.8, 2.2, 3.1, 2.2, 9.4""
Finishing 2.4, 3.7. 3.9. 3.4. 3.7, 3.1,, 6.8"'
Salle 3.3, 1.9, 3.1. 3.1. 3.0. 3.2, 6.5"'
Wharf 3.2, 3.0. 3.0. 3.0. 4.0. 2.8, 2.8'

"P<0,05, "'P<0.005. One-way repeated measure analyses of variance were conducted on each
group separately. Means in each row without a shared subscript are significantly different from one
another using Tukey's HSD procedure (P<0.05).
"As measured by the question, ' How do you see the people in [own group] and the people in [target
group] in the company?'. The response scale was anchored by the descriptions ' A s two teams on
opposite sides (i.e. working against each other) ' ( = 1 ) and ' As two teams pulling together (i.e. working
with each other) ' ( = 7 ) . In the analyses reported here we have reversed the scale so that a high
score= high conflict.

note, however, that the reciprocal ratings of groups often corresponded quite closely (viz.
Conversion rating of Paper makers and vice versa, similarly for Conversion-Salle, Wharf-
Paper making, Wharf-Salle pairings). Finally, note that ratings of Management, a group
with whom there is some objective conflict, whilst higher than some other target groups,
were not invariably so, and were still towards the non-conflictual end of the scale ( <4).
(2) Amount of self-reported contact with each of the target groups (CONTACT). This
was measured by the question, ' How often would you say that you had contact with people
in [target group]? ': ' almost never (i.e. not at all/less than once a month) ' ( = 1) to ' very
often (i.e. every day or more) ' ( = 7).
(3) Strength of identification with the departmental in-group as measured by our scale
(IDENTIFY).
The results of the regression analysis are presented in Table 4. Five of the seven
analyses yielded acceptably high values of R2,as indicated by the highly significant F
statistics for the overall regression equations.
In each of these five analyses perceived conflict proved to be a reliable predictor of
differentiation: as expected from RCT the coefficients for the CONFLICT variable across
the whole sample were positive and significantly different from zero. A similar pattern was
observable in each of the respondent groups analysed separately, with only a few excep-
tions. The amount of contact, on the other hand, explained much less of the variance in the
differentiation measures. In only two of the overall significant regression equations were the
Ps for CONTACT significantly different from zero, in the predicted negative direction.
However, in the two non-significant overall analyses [DIFF (P) and DIFF (F)] the
CONTACT coefficients were also significant. Nevertheless, inspection of the group-by-
group analyses reveals a much less consistent picture than for CONFLICT. The results
from our third explanatory variable, strength of identification, were similarly equivocal.
Although the Ps for IDENTIFY for the whole sample were invariably positive, only two of
these were significantly different from zero. Examining them on a departmental basis
reveals that these positive coefficients could only be reliably observed in the Paper making
and Conversion groups; in the other three groups a mixture of positive and negative
coefficients was found, although for the most part these were not significantly different
from zero.
282 RUPERT B R O W N ETAL.

Table 4. Standardized regression coeficients (Bs) between three explanatory variables and
indices of intergroup diflerentiation

Whole
Independent sample Paper Conv. Finish Salle Wharf Dependent
variables (n=157) (n=35) (n=40) (n-34) (n=29) (n=19) variable

CONFLICT (M) +50**' +34' +50"' +49" +30 +89*** DlFF (M)
CONTACT (M) +01 +18 -08 -04 +14 +18 RZ=0.23
IDENTIFY +09 +28 +12 +16 -33 -02 F=14*4"'
d.f.=3.143

CONFLICT (E) +37"' +48" +29' +44' +28 +49' DlFF (E)
CONTACT (E) -09 -01 +11 -04 -36' -15 R2= 0.16
IDENTIFY +1 9" +40' +19 +16 +41' +05 F=9.3"'
d.f. = 3,143

CONFLICT (P) -02 +00 +25 -04 -08 DlFF (P)

CONTACT (P) -16' -06 -18 -01 -14 R2= 0.03
IDENTlFY + 05 +11 +28 +05 +20 F=l.O,
d.f.=3,110

CONFLICT (C) +45"' +49"' +51"' +37' +15 DlFF (C)

CONTACT (C) -17' -04 -34' +16 -27 R2= 0.29
ID ENTlFY +14 + 36' +14 -21 +13 F = 14.2, ***
d.f. = 3,104

CONFLICT (F) +07 -14 -04 +35 +54' DlFF (F)

CONTACT (F) -17' -22 -01 -05 -43' RZ= 0.04
IDENTIFY +07 +24 +16 +03 +03 F= 1.7.
d.f.=3,113

CONFLICT (S) +20' +21 +12 +19 -03 DlFF (S)

CONTACT (S) -24' -34' +10 -23 -22 R2=0.14
IDENTl FY + 04 +15 +17 -25 -15 F=6.3,"'
d.f.=3,114

CONFLICT (W) +33"' +23 +10 +51"' + 29 DlFF (W)

CONTACT (W) -12 -26 -21 +13 -19 Rz=0.13
IDENTIFY +15' +08 +15 +34' +I9 F-6.3,"'
d.f. = 3,127

' P < 0.05; "P< 0.01; * * * P < 0.005 (one-tailed).

Notes. Suffix on independent and dependent variables refers t o the out-group with which they are
associated [e.g. DlFF (M) =differentiation against Management group, etc.].
Decimal points have been omitted from B coefficients.

DISCUSSION

In this paper our aim was to attempt to explain intergroup differentiation in organiz-
ations by three theoretical frameworks within social psychology. Our findings point to the
fruitfulness of this approach. We were able to document a very systematic pattern of
differentiation in people's attitudes and then explain sizable proportions of the variance in
that differentiation via our regression analyses.* On only two of the seven analyses con-
ducted did we fail to obtain a reasonable fit between our regressionmodel and the observed
data.
*Relatively speaking only. As noted in the introduction, social psychology is only one of many disciplines
which can help us to understand intergroup relations. Thus, we should not expect our R2s remotely to
approach unity until and unless these other perspectivesare incorporatedinto our analyses.
 EXPLAINING INTERGROUP DIFFERENTIATION 283

Our major explanatory variable was, quite clearly, the perceived relationship between
our group and the target group in that the more conflict people perceived, the greater the
polarization in their intergroup attitudes. This was generally true, and at high levels of
significance, both for the sample as a whole and in the different subgroups across most of
our indices of differentiation. This result confirms findings from an earlier study (Brown &
Williams, 1984), and is consistent with the premises of RCT. However, an interesting
question remains as to the correspondence of perceived conflict with actual conflict.
Although we did not have the resources to analyse the objective relationships very fully
within the factory, the relationship ofeach of the worker groups with management is prima
facie conflictual on economic grounds. This conflictual relationship was underlined by the
drastic redundancy programme implemented by management just prior to the study. It is
thus of interest to note that, on the whole, the sample did not perceive that relationship as
conflictual (Table 3). Indeed, only a very small minority gave responses which reflected any
clear class-consciousnessin the classical Marxist sense.
By contrast, our second explanatory variable, intergroup contact, was much less
powerful. The relationships between amount of self-reported contact and differentiation
which we found, although mainly negative as predicted, were rather weak and did not hold
up consistently across groups. This suggests that explanations in terms of mere contact with
out-groups are in need of some further elaboration. One promising avenue in this direction
is to distinguish between contact which occurs on an interpersonal level (in which group
memberships are not psychologically salient) and intergroup contact (where they are)
(Brown & Turner, 1981). There is increasing evidence that it is the latter which exerts the
most significant and generalizable influence on intergroup attitudes, while interpersonal
contact may have only short-term and context-specific effects (Hewstone & Brown, 1986).
Unfortunately, pragmatic considerations prevented us from distinguishing between these
two forms of contact in the measure we used, which may, in part, account for the rather
equivocal effects associated with it. In view of the importance attached to contact in several
different theories in both psychology (e.g. Brewer & Miller, 1984) and industrial relations
(e.g. Allen, 1986), it would seem important to pursue this issue further. For example, do
structures which facilitate intergroup contact via group representatives, e.g. negotiating
bodies, collaborative working arrangements, joint committees, have a greater impact on
organizations than simply increasing the amount and number of informal interpersonal
contacts?
Our third theoretical variable, group identification, also proved to be an inconsistent
predictor of differentiation. The correlations were generally positive, as predicted by SIT,
but they tended to be rather weak and only consistently visible in two of the five subgroups.
There are three main issues which need to be discussed here.
The first is whether this result can be attributed to some methodological factor
associated with inappropriate sampling or measurement. We do not think this is tenable for
several reasons. First, the relationship between identification and differentiation has now
been examined in several different contexts. In addition to Brown & Williams’ (1984)
findingsalready noted, two further studies have been completed since the present study was
undertaken, reaching broadly similar conclusions. In the first, Condor et al. (1984), in an
experimental study, found significant but weak positive correlations between those vari-
ables, but they varied considerably across different experimental conditions and on differ-
ent indices of differentiation. More recently, Oaker & Brown (in press) found that nurses
from specialized and general hospital departments actually showed significant negative
relationships. It seems clear, therefore, that the absence of a clear positive correlation
between identification and intergroup differentiation is a genuine phenomenon. Second, the
new measure of identification developed in this study, while still obviously far from perfect,
appears to have acceptable reliability and to correspond reasonably well to people’s spon-
taneous evaluations of their in-group. Thus, we feel reasonably confident that it is tapping
284 R. B R O W N E T A L .

important aspects of group identification.Third, our decision to pre-select the groups might
be criticized on the grounds that we imposed categories on respondents. However, the fact
that in the interviews some 60 per cent of the sample spontaneously identified departmental
or subdepartmental groups indicates that our imposition of categories was not totally at
variance with the respondent’s own category systems.
The second main issue concerns the nature of the identification process and the
meaning of social identity for group members in the organization we studied. It is clear, first
of all, that the single most important and recurring aspect of group membership for our
respondents were the interpersonal relationships within the group. This took the form
either of friendships with particular individuals in a group of more general affiliative
tendency. The interest in these findings is that they contrast strongly with the conception
of identity favoured in SIT. For example, Turner (1984) in the most recent theoretical
formulation places most emphasis on the cognitive and socially comparative nature of the
identificationprocess at the expense of interpersonal relationships. It seems, therefore, that
what we have uncovered is that social identification can have quite different meanings in
different group settings and is not the unidemensional process implied by SIT.
This may be illustrated by two contrasting industrial relations examples. In situations
involving union-management bargaining over wages or job demarcation issues it seems
likely that the socio-cognitive processes as originally outlined in SIT are likely to be opera-
tive since what is often at stake are groups’ relative standings or the integrity of their
‘ boundaries ’. On the other hand, in more extreme situations involving a strike or a lock-out
what may be much more important are affiliative processes involved in maintaining
cohesion and solidarity.
The third and related issue concerns the more general implications of our findings for
SIT. It now seems clear that researchers using this approach should pay more attention to
the apparently multidimensionalcharacter of group identificationand comparison, includ-
ing a more detailed consideration of how and why certain group attributes become salient
for people. This research supports other work in widely different contexts in arguing for a
more elaborated theoretical analysis in which the specificity of identity processes is recog-
nized (Condor et al., 1984; Mummendey & Schreiber, 1984; van Knippenberg & van Oers,
1984). What seems to be implicated here is some form of group ideology which determines
which dimensions of evaluation are most central for each group’s identity. This is likely to
be related to, but not reducible to, the material conditions defining group membership
and intergroup interaction (Condor & Abrams, 1984). On this argument, the inconsistent
relationships between identificationand differentiation are attributable to our failure to tap
a wide enough range of meaningful aspects of group members’ identities. An important
contribution which future research can make is therefore to explore further this relationship
between ideology and identity.
Elucidating this relationship is particularly important when considering what sorts of
social-psychological factors are implicated in, for example, union joining and trade union
activity. Although our own findings are admittedly only attitudinal,* on the basis of the
above ideas we may speculate that becoming a trade unionist involves both a change of
identity and a corresponding change in ideological outlook. A change in identity occurs
because one may now define oneself as a member of a collectivity which formerly one was
not a part of. From the perspectives of both RCT and SIT one should expect that this fact
alone should lead to increased differentiation from management-behaviourally as well as
attitudinally. SIT would also predict that one’s views of other trade unions would also
become more differentiatedas the social significanceof belonging to (say) SOGAT 82 rather
than some other union became apparent. Although there is ample experimental-if not, as

*There is, however, evidence to suggest that biased intergroup attitudes and perceptions may not be
completely at variance with discriminatory behaviour (Brewer, 1979; Turner, 1981).
 EXPLAINING INTERGROUP DIFFERENTIATION 285

yet, field-research which substantiates these hypotheses concerning the sequelae ofjoining
a new group (e.g. Turner, 1981), our findings suggest that the extent of this differentiation
would not be predictable merely from the strength of the group, in this case union, attach-
ment. From our argument above, it follows that one must also take account of the ideologi-
cal component of the trade union membership. Although, to be sure, becoming a trade
unionist at all involves at least some shift in ideology-from a symbiotic model of an
individualized worker-manager relationship to a more collective and conflictual one-there
may be a variety of orientations associated with trade union membership and activity.
These include, for example, self-interest, social obligation and political commitment
(Tagliacozzo & Seidman, 1956; Batstone et al., 1977).According to which was predominant
in any particular individual or group of individuals, one would expect very different indus-
trial relations attitudes to be displayed. What we suggest, in other words, is that a cognitive
self-definition as a group member and an emotional attachment to the group are necessary
precursors to intergroup behaviour, but are not in themselves sufficient to produce or
explain the variety of different group responses in any particular context.

ACKNOWLEDGEMENT
This research was conducted under the auspices of an ESRC grant (no. HR 7935)
awarded to the first and last authors. We are grateful to Peter Allen and the editors for their
helpful comments on an earlier version of this paper.

REFERENCES
ALDERFER, C. P. & SMITH,K. P. (1982). Studying intergroup relations embedded in organizations.
Administrative Science Quarterly, 27,3545.
ALLEN, P. T. (1982). Size of workforce, morale, and absenteeism. British Journal of Industrial Relations, 20,
83-100.
ALLEN, P. T. (1986). Contact and conflict in industry. In M. R.C. Hewstone & R. J. Brown (eds), Contact and
Conflict in Intergroup Encounters. Oxford: Blackwell.
Al.LPoRT, G. W. (1954). The Nature of Prejudice. Reading, MA: Addison-Wesley.
BATSTONE, E., BORASTON, 1. & FRENKEL, S. (1977). Shop Steward7 in Action. Oxford: Blackwell.
BLAKE,R. R. & MOUTON,J. S. (1979). Intergroup problem solving in organizations: From theory to
practice. In W. G. Austin & S . Worchel (eds), Social Psychology of Intergroup Relations. Monterey, CA:
Brooks/Cole.
BLAKE,R. R., SHEPARD, H. A. & MOUTON, J. S. (1964). Managing Intergroup Conflict in Industry. Houston,
TX: Gulf Publishing.
BOURHIS, R. & HILL,P. (1982). Intergroup perceptions in British higher education: A field study. In H. Tajfel
(ed.), Social Identity and Intergroup Relations. Cambridge: Cambridge University Press.
BREWER, M. B. (1979). lngroup bias in the minimal intergroup situation: A cognitive-motivational analysis.
Psychological Bulletin. 86,307-324.
BREWER, M. B. & MILLER, N. E. (1984). Beyond the Contact Hypothesis: Theoretical perspectives in desegre-
gation. In N. E. Miller & M. B. Brewer (eds), Groups in Contact: The Psychology of Desegregation. New
York: Academic Press.
BROWN, R. J . ( I 978). Divided we fall: An analysis of relations between sections ofa factory workforce. In H.
Tajfel (ed.), Differentiation between Social Groups: Studies in the Social Psychology of Intergroup Relations.
London: Academic Press.
BROWN,R.J. &TURNER, J. A. (1981). Interpersonal and intergroup behaviour. In J. C. Turner & H. Giles
(eds), Intergroup Behaviour. Oxford: Blackwell.
BROWN, R. J. &WILLIAMS, J. A. (1984).Group identification: The same thing to all people? Human Relations.
37,547-564.
CAMPBELL, D. T. (1969, Ethnocentric and other altruistic motives. Nebraska Symposium on Motivation,
pp. 283-3 I I . Lincoln, NB: University of Nebraska Press.
CONDOR, S. & ABRAMS, D. (1984). Womanhood as an aspect of social identity: Group identification and
ideology. Paper presented to The British Psychological Society International Conference on Self and
Identity, Cardiff, July.
286 RUPERT BROWN ETAL.

CONDOR,S. & BROWN,R. (in press). Psychological processes in intergroup conffict. In W. Stroebe, A.
Kruglanski, D. Bar-Tal & M.Hewstone (4s). The Social Psychology of Intergroup and International
Conflict: Theory, Research and Application. New York Springer.
CONDOR, S.,WILLIAMS,J. A. & BROWN,R. J. (1984). Identity, intragroup relations, and intergroup
behaviour. Paper presented to the BPS Social Psychology Sectionconference, Oxford, September.
DRIEWER,L.(1976). Ethnic self-identity:A comparison of ingroup evaluations. Sociometry, 39,131-141.
EDWARDS, P. K. (1977). A critique of the Kerr-Siege1 hypothesis of strikes and the isolated mass. The
Sociological Review, 25,55 1-574.
EDWARDS, P. K. (1980). Size of plant and strike-proneness. Oxford Bulletin of Economics and Statistics, 42,
145-156.
HEWSTOW- M. R. C. &BROWN, R. J. (1986). Contact is not enough: An intergoup perspectiveon the contact
hypothesis. In M. R. C. Hewstone & R. J. Brown (eds), Contact and Conflict in Inrergroup Encounters.
Oxford: Blackwell.
INGHAM, G. K. (1967). Organization size, orientations to work and industrial behaviour. Sociology, 1,
239-258.
KERR,C. & SIEGEL, A. (1954). The inter-industrypropensity to strike. In A. Kornhauser, R. Dubin & A. M.
Ross (eds), Industrial Conflict. New York: McGraw-Hill.
KUHN,M. H. & MCPARTLAND, T. S. (1954). An empirical investigation of self-attitudes. American
Sociological Review, 19,68-76.
LIEBKIND, K. (1982). The Swedish-speaking Finns: Acase study ofethnolinguisticidentity. In H. Tajfel (ed.),
Social Identity and Intergroup Relations. Cambridge: Cambridge University Press.
MUMMENDEY, A. & SCHREIBER, H. J. (1984). ' Different 'just means ' better ': Some obvious and some hidden
pathways to in-group favouritism. British Journal of Social Psychology, 23,363-367.
NICHOLLS, T. & ARMSTRONG, P. (1976). Workers Diviakd. Glasgow: Fontana/Collins.
OAKER, G. & Brown, R. J. (in press). Intergroup relations in a hospital setting. Human Relations.
PARKIN, F. (1967). Working class conservatives:A theory of political deviance. British Journal of Sociology,
18,278-290.
P E ~ ~ I G RT.
EW F..(1971). Racially Separate or Together? New York: McGraw-Hill.
PETTIGREW, T. F. (1986). The intergroup contact hypothesis reconsidered. In M. Hewstone & R. Brown
(eds), Contact and Conflict in Intergroup Encounters. Oxford: Blackwell.
SHERIF, M. (1966). Group Conflict and Cooperation. London: Routledge & Kegan Paul.
SKEVINGTON, S. (1980). Intergroup relations and social change within a nursing context. British Journal of
Social and Clinical Psychology, 19,201-213.
SKEVINGTON, S.(198 I). Intergroup relations and nursing. European Journal of Social Psychology, 11,43-59.
STRAUSS, G. (1964). Work-flow frictions, interfunctional rivalry, and professionalism: A case study of
purchasing agents. Human Organization, 23,137-149.
TAGLIACOZZO, D. L. & SEIDMAN, J. (1956). A typology of rank and file union members. American Journalof
Sociology, 61,546553.
TAJFEL,H.(ed.) (1978). DifJerentiation between Social Groups: Studies in the Social Psychology of Intergroup
Relations. London: Academic Press.
TAJFEL, H. &TURNER, J. C. (1979). An integrative theory of social conflict. In W. Austin & S.Worchel (eds),
The Social Psychology oflntergroup Relations. Monterey, CA: Brooks/Cole.
TURNER, J. C. (1981). Theexperimentalsocial psychology ofintergroup behaviour. In J. C. Turner& H. Giles
(eds), Intergroup Behaviour. Oxford: Blackwell.
TURNER, J. C. (1984). Social identification and psychological group formation. In H. Tajfel (ed.), The Social
Dimension, vol. 2. Cambridge: Cambridge University Press.
VAN KNIPPENBERG, A. &VAN OERS,H. (1984). Social identity and equity concerns in intergroup perceptions.
British Journal of Social Psychology, 23,351-361.
WEINREICH, P. (1975). Identity diffusion in immigrant and English adolescents. In G. K. Verma & C. Bagley
(eds), Race, Education and Identity. London: Heinemann.
ZAVALLONI, M. (1971). Cognitive processes and social identity through focused introspection. European
Journal of Social Psychology, 1,235-260.

Received 20 August 1985; revised version received 28 January 1986