Professional Documents
Culture Documents
WILLEM DOISE
Centre National de la Recherche
Scientifique. Paris
ANNE SINCLAIR
Ecole de Psychologie de I’Universit6
de Genhve
Abstract
In the first two parts of the paper a distinction is made between a ‘conflict or
convergence of interests’ approach (Sherif)and a ‘categorisation’ approach (Tajfel)
in the area of the experimental study of intergroup relations. Some recent experi-
mental findings are mentioned, and a theoretical development of the categorisation
approach is proposed. In the third part a new experiment illustrating the relevance
of the categorisation approach is described.
Introduction
tion’ approach; and in the third one, we will illustrate this approach with an ex-
periment.
One can summarize Sherif‘s theory of intergroup relations in the following way.
To understand the psychological processes that occur during intergroup relations,
one must understand the relationship between the goals of the interacting groups.
When these goals are competive - that is, one group only can achieve its goal to
the detriment of the other group - hostility becomes evident; negative images of
the outgroup and negative affects towards the outgroup arise. On the other hand,
when groups need each other to achieve an important common goal, ‘a super-
ordinate goal’, positive feelings towards outgroup members and a favourable
image of the outgroup spring up. This idea of conflict or convergence of interests
is obviously not peculiar to Sherif. If one accepts that the relationship between
different interests, embedded in an objective situation, is a kind of substructure,
there is an analogy between Sherif‘s conception and the Marxist view of the
relationship, between substructure and superstructure. When LeVine challenges
‘describe to me the economic intergroup situation and I shall predict the content
of stereotypes’, he is not far away from the Sherif-type approach ( c f . Tajfel’s
report of a round-table discussion in Cannes, 1970).
Let us briefly describe some more recent results of similar tendency. Avigdor
(1953) and Wilson, Chun and Kayatani (1965) have shown that in a competitive
or conflictual situation, the unfavourable image of the outgroup is elaborated on
specific dimensions. Group members selectively attribute to outgroup members
characteristics which justify hostile behaviour towards the outgroup. In several
experiments, W. Doise (1969, 1972) has been able to show that the processes of
‘ideologisation’ takes place even if competitive interaction is only expected to take
place. Rabbie and Wilkens (1971) obtained similar results.
Blake and Mouton’s experiment (1962) shows that in a competitive situation
members of one group find it singularly difficult to ‘learn’ solutions proposed by
the other group and that they do not recognize what the ingroup and outgroup
performances have in common. But according to us, this has something to do with
the accentuation of difference and thus brings us to the categorisation approach.
Let us conclude this short description of the convergence or conflict of interest
approach by saying that, although we think there is still a lot to be done in this
direction, many findings indicate that the conception which links intergroup re-
presentation to the relationship between group goals is too restrictive. Even when
The categorisation process in intergroup relations 147
We now come to the categorisation approach. Superficially, one might ask whether
the Tajfel and Wilkes (1963) results have much to do with social psychology.
Let us briefly describe their results without going into the details of the experi-
mental procedure. Using eight lines of different lengths, four of which were
labelled with the letter A and the four others with the letter B, they verified that
when Ss had to judge the length of these lines, they saw the difference between
the A’s and the B’s as larger when A systematically accompanied the four shorter
lines and B the four longest. When there was no relationship between the length
of the line and the label, that is, when the !abellinp, was done at random or when
there was no labelling at all, Ss’ judgments showed no accentuation of the dif-
ferences. One can generalize: When stimuli belong to a continuum (e.g. increasing
length) and this continuum is divided into adjacent categories (e.g. ‘short’ and
‘long’), judgments bearing on the relevant seriated characteristics accentuate the
inter-category differences and the intra-category similarities.
In a later paper, Tajfel (1972) describes this phenomenon as the deductive
categorisation error: When it is known that the stimuli belong to different cate-
gories, they will be judged as more different, as far as certain characteristics are
concerned, than when Ss do not know that the stimuli belong to different cate-
gories.
Many experiments have been done on the categorisation effect (cj. Lilli, 1970
and Marchand, 1970). Some of these experiments have used lines and squares as
stimuli, but others have used ‘social’ stimuli, such as attitude statements (Eiser,
1971). It is clear that the categorisation effect is relevant to the study of inter-
group phenomena: It is one of the processes underlying stereotyping (cf. Tajfel,
Sheikh and Gardner, 1964).
We believe, however, that the categorisation process has wider interpretative
value. Two well-established experimental findings on the consequences of dividing
people into groups have to be reformulated to fit the categorisation framework.
The first experiment (Tajfel, Flament et al., 1971) shows that ingroup members,
when asked to distribute points (for which they will later get money), system-
atically adopt a strategy that results in a positive difference between them and the
outgroup members, even if, by doing so, they reduce their gain (in absolute value).
148 Willern Doise and Anne Sinclair
3. The experiment
3.1 Hypotheses
The goal of the experiment was to compare the images group members form of
their ingroup and an outgroup in situations where categorisation can take place
and in situations where it is less likely to occur.
Suppose in one condition we ask group members to describe themselves by
filling out a questionnaire without mentioning any other group; in another con-
dition we ask members of the same group to describe themselves, but we evoke
an outgroup. It is quite clear that in the second condition the categorisation
process must play a larger role than in the first, where no mention is made of
any other group.
If the members of the two groups actually come into contact, one would sup-
pose that the categorisation process would also occur, but not in the same way
in an individual encounter (one member from one group, one member from
another) as in a collective encounter (at least two members from each group).
If we consider the studies made on collective judgments, that is, judgments made
in the presence of others or judgments made after discussion with others, we can
say that their most general finding has been a convergence effect: Sometimes
convergence on a moderate position, sometimes convergence on an extreme posi-
tion (cj. Doise and Moscovici, 1972). We would predict that in an individual
encounter the convergence effect would make the individuals’ opinions more
similar and thereby lessen the categorisation effect.
However, this is not the case in the collective encounter where convergence
is not limited to convergence towards a member of the other group. In this situa-
tion, convergence should at least be as strong towards a member of one’s own
group as that towards a member of the other group. Therefore, the categorisation
effect should be stronger in the second situation.
One should not forget that an experiment in which Ss describe real groups
cannot take place in a social vacuum. Ss will reproduce or may transform their
pre-existing ideological representations. Ss who participated in this experiment
came from two different groups. Half of them were coZl&gierzs,male students in
secondary school preparing for university entrance; the other half were boys of
the same age undergoing vocational training in factories and shops, apprentis.
It is clear that these two groups occupy very different positions in their society.
The collkgiens are far more privileged in many respects. Obviously this has reper-
cussions on a psychological level, and so one cannot set aside this variable under
the pretext that it pertains more to the socio-economic field.
150 Willem Doise and Anne Sinclair
3.2 Predictions
Let us sum up our predictions concerning the categorisation of inter-group images
in different encounter conditions:
1 . When one asks group members to describe their own group without mention-
ing any other group, and then asks for a description of the other group (no
encounter condition), the difference between these two descriptions will be
smaller than the same difference in the condition where the other group is
mentioned right from the start (symbolic encounter condition).
2. When two people, each from a different group, actually meet (individual
encounter condition), the difference between the descriptions they make of
the two groups will be smaller than that between the descriptions in the sym-
bolic encounter condition.
3. When several members of two groups meet (collective encounter condition),
the difference between the descriptions of the two groups will be greater than
in the individual encounter condition.
4. The effect of the different conditions will not be the same for the two groups,
as one group is socially more privileged.
The categorisation process in intergroup relations 15 1
3.3 Method
3.3.1 Subjects
Fifty-six collkgiens and forty-nine upprentis, all living in Geneva, Switzerland, and
aged 15, 16 or 17.
3.3.2 Questionnaire
The items used were obtained from content analyses of fifteen lengthy interviews
with collkgiens and apprentis. There were twelve items in the questionnaire, six
of which were clearly judged as favourable (two items) or unfavourable (four
items) by a control group of forty-five collkgiens and apprentis.
The positively evaluated items were:
a) X like the theatre.
b) X are proud of being X.
The negatively evaluated items were:
c) X prefer to talk about what is happening in their own milieu than to talk
about society in general.
d) Even if they have the necessary equipment, X find it difficult to do odd jobs.
e) X do not express themselves easily.
f) X are not interested in what is going on in the world.
Only these six items were used for the statistical analyses, as the other six items
were evaluated in a far less clear-cut way. For each item there were six answers
possible: Strongly disagree, disagree, mildly disagree, agree mildly, agree, agree
strongly. These answers were respectively rated -3, -2, -1, + 1, +2, + 3 for the
favourable items (a and b above) and the signs were inversed for the unfavourable
items (c to f above).
All the Ss in the different conditions were first asked to describe their own
group by filling in the questionnaire and then asked to describe the other group
by filling in the same questionnaire. ‘X’ stands for collkgiens or upprentis.
3.4 Results
3.4.1 The effect of the symbolic encounter
The description Ss make of the ingroup in the no-encounter condition is not very
favourable; it is not significantly different from a description with an equal pro-
portion of favourable and unfavourable ratings. Both groups - collkgiens and
apprentis - tend to make a slightly more favourable description of the outgroup,
but this is not statistically significant. However, in the symbolic encounter con-
The categorisation process in intergroup relations 153
~.
Subjects
Experimental CollCgiens
Apprentis
conditions
N Ingroup Outgroup Difference N Inproup Outgroup Difference
vergence effect. This is not the case for the apprentis who are in a conflicting
<
situation, as their more varied responses show ( p 0.02 Table 3).
* t-values for the differences: apprentis, colltgiens - 7,79, d.f. 1,28, p <
a and c - 2,38, p < 0.05 0.0 1;
b and c - 2,74, p C 0.02 1 x 1, 2 x 2, x before, after - 8,89, d.f.
Significant F values of the analysis of var- 1,28, p < 0.01.
iance for conditions 1 x 1, 2 x 2
representations (of the ingroup and outgroup) become favourable. The collkgiens
do not discriminate in favour of their group, and the apprentis show a bias in
favour of the other group (collkgiens) ( p < 0.02). As predicted, in this condition
pluralistic ingratiation plays a large role and even provokes the apprentis into
making concessions. In the responses made after the discussion, discrimination
in favour of the ingroup on the part of the collkgiens tends to reappear, as they
are clearly in a position of strength during the discussion. The discussion also has
the effect of accentuating the difference between the symbolic-encounter and real-
encounter conditions for the apprentis ( p : 0.05). The prediction that the individ-
ual-encounter condition will show less categorisation effect than the symbolic-
encounter condition is confirmed, but for the apprentis it is only statistically
significant after the discussion. If we compare the effects of the individual-
encounter condition (before and after discussion, and for the two groups) with the
effects of the collective-encounter condition, we can see that in the collective-
encounter condition, differences between both images of the ingroup and outgroup
become more positive ( p < 0.05, Table 2). This confirms our third prediction;
once again the collkgiens consistently evaluate their own group more favourably
than the outgroup ( p < 0.01). The convergence effect observed in the individual-
encounter condition is overcome by the categorisation effect. The apprentis still
significantly evaluate the other group as superior before the discussion, but after
the discussion the difference is no longer significant.
If we examine the modal response index, we can get useful information con-
cerning the real encounters and the effect of the discussion. First of all, the col-
lkgiens’ responses are less varied than those of the apprentis ( p < 0.01) - the
experimental manipulations create less conflict for them. Secondly, the effect of
convergence towards a member of the same group is not the same in the individ-
ual- and collective-encounter conditions. Before the individual encounter, even
though no member of the same group is physically present, there is an implicit
consensus between members of one group concerning the responses which should
be given. This implicit consensus disappears somewhat after discussion with a
member of the other group. Does discussion with a member of the other group
thus weaken the ingroup consensus? Things happen very differently in the col-
lective encounter. The tacit agreement between same-group members is not very
strong before the encounter but increases during it. The implicit consensus of
same-group members is reinforced during the encounter when there is another
same-group member present; ingroup cohesion is increased. (Significant inter-
action between before, after and encounter conditions: p <O .01.)
156 Willem Doise and Anne Sinelair
It seems to us that the results of this experiment strongly suggest that it is possible
to explain intergroup phenomena and variations in intergroup representation by
using the categorisation approach. It seems difficult to explain the data obtained
by using only explanations within the framework of the conflict or convergence
of interest approach. This does not mean that we claim to have explained all the
characteristics of the intergroup images in the four different encounter conditions.
Despite the fact that the collkgiens’ and the apprentis’ responses consistently go
in the predicted directions, the apprentis often gave a more favourable image of
the outgroup than the ingroup. According to us, this means that the items in the
questionnaire were more pertinent to the ‘social status scale’ than to the ‘ethno-
centric scale’. An experiment in which the two scales are represented by different
items should be carried out. However, it is nevertheless interesting to note that
it is only in the symbolic-encounter condition that the underprivileged apprentis
can elaborate an ingroup image that is as favourable as the image of the outgroup.
The difference observed between the real-encounter conditions - individual and
collective - also called for further research which has been carried out (see Doise
and Weinberger, 1973). The results of the Doise and Weinberger experiment
show that in a collective-encounter situation, when there is a conflict of interest,
the difference between ingroup and outgroup members is enhanced for items which
represent the current social stereotypes. In an individual-encounter situation, these
‘social’ differences are more easily by-passed. According to us, similar processes
explain why during the expectation of the collective encounter the colldgiens
significantly discriminate in favour of their own group and did not do so when
expecting an individual encounter.
REFERENCES
Avigdor, R. (1963) Etude exPCrimentale de Doise, W., Csepeli, G., Dam, H. D., Gou-
la gentke des stCrtotypes. Cahiers I n - ge, C., Larsen, K. and Ostell, A. (1972)
tern. de Sociologie, 14, 154-168. An experimental investigation into the
Blake, R. R. and Mouton, J. S. (1962) Com- formation of intergroup images and
prehension of points of communality intergroup behaviour. Eur. 1. SOC.Psy-
in competing solutions. Sociornefry, 25, chol., 2, 202-204.
56-63. Doise, W. and Moscuvici, S. (1972) Les
Doise, W. (1969) StratCgies de jeu B l’intC- dCcisions collectives. In S. Moscovici:
rieur et entre des groupes de nationa- Introduction d la Phychologie Sociale,
lit6 diff6rente. Bulletin du C.E.R.P., 18, Vol. 2 . Paris, Larousse. Pp. 114-134.
13-26.
The categorisation process in intergroup relations 157
R ksum k Zusammenfassung
Dans les deux premieres parties de l’article, In der ersten beiden Teilen des Artikels wird
nous distinguons deux approches thtoriques differenziert zwischen einem Ansatz auf der
diffkrentes de I’ttude exPCrimentale des rela- Grundlage von ‘Konflikt oder Interessenkon-
tions entre groupes: celle qui est basCe sur vergenz’ (Sherif) und einem ‘Kategorisie-
1’Ctude du conflit et de la convergence des rungs’-Ansatz (Tajfel) im Bereich experimen-
inttrCts (Sherif) et celle basCe sur l’ttude du teller Untersuchungen der Beziehungen zwi-
processus de cattgorisation (Tajfel). Des rt- schen Gruppen. Einige neuere Untersu-
sultats expkrimentaux rtcents sont mention- chungsergebnisse werden dargestellt und es
n6s et une extension thCorique du modkle de wird eine theoretische Fortentwicklung des
la cattgorisation est proposte. Une troisibme Kategorisierungsansatzes vorgeschlagen. Im
partie prtsente une nouvelle experience qui dritten Teil wird ein neues Experiment be-
illustre la portie de l’approche b a d e sur schrieben, das die Relevanz des Kategorisie-
l’itude de la cattgorisation. rungsansatzes aufzeigt.
Copyright of European Journal of Social Psychology is the property of John Wiley & Sons,
Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.