The criss-cross categorization effect in intergroup discrimination

Publication Year


Journal Article

The effect of criss-crossing two dichotomous social categorizations on intergroup discrimination is studied. One hundred and thirty-nine 12–13 year old schoolchildren were randomly assigned to five categorization conditions: one simple, division into one ingroup and outgroup; and four crossed, half the ingroup members by one division are outgroup members by a second division and vice versa. Simple subjects rated ingroup and outgroup members; crossed subjects rated members of both ingroups and either simultaneous members of one ingroup and outgroup, or one outgroup and ingroup, or both outgroups, or everybody. Ingroup members were rated more favourably than outgroup members (P < 0.025), this effect varying with experimental condition (P < 0.025). In group bias was found with crossed but not simple categorization and was greatest when members of both ingroups were compared to members of both outgroups. Bias was non-significant when crossed subjects rated everybody. The significance of the data for earlier work and the explanation of social categorization effects in intergroup behaviour is discussed.

British Journal of Social and Clinical Psychology
Type of Article
Journal Article
Full text

The following is an excerpt of the intervention methodology. For more information, please see the full text of the article on the publisher's website or through your institution's library.


Subjects Subjects were 139 schoolchildren of both sexes aged 12-13 years, from a Bristol comprehensive school. [...]

Procedure Subjects were randomly assigned to five experimental conditions (n = 28; n = 27 in condition 2). Conditions 2-4 were run simultaneously in three sessions, conditions 1 and 5 were run separately in single sessions. The procedure was in three parts. Firstly, subjects were assigned to a category or categories on the supposed basis of their preferences for a number of photographs. Then they took part in a task designed ostensibly to measure their perceptual ability. Finally, they estimated the performance of themselves and others in the task.

(1) Assignment of categories The study was introduced as being concerned with making decisions on the basis of limited information. The first decision-making task was for subjects to indicate their preferences for one or other of a series of pairs of scenic photographs. [...]

Condition 1: simple categorization. The class was randomly divided into two. Half were shown eight pairs of slides projected onto a screen and depicting scenes (mainly rural) from supposedly just two countries. [...] These subjects privately and individually recorded which picture of each pair they preferred. After the first set of preference sheets had been removed for ‘marking’, the other half of the subjects were shown a different eight pairs of scenic pictures [...]. When the latter subjects’ preference sheets had been removed, the experimenter explained to all subjects that the purpose of this procedure had been to divide them into four separate groups according to their preferences. [...]

Conditions 2-5: crossed categorization. In these conditions subjects were shown one set of eight pairs of slides, supposedly of just two countries and taken by just two photographers, with the same fictitious names as in condition 1. The experimenter then explained that subjects would be assigned to groups according to their preferences. Firstly, depending on their country preferences, they would be equally divided into Pran or Dilt groups. Then, depending on photographer preferences, Pran and Dilt group members would be divided equally into Davis and Evans groups. A blackboard diagram was used to illustrate that these divisions would be perfectly criss-crossed. Thus, it was stated, each subject would belong to two groups simultaneously.

(2) The perceptual ability task A second decision-making task was then described to the subjects; one which would measure their perceptual ability in ambiguous situations. A row of eight digits would be projected onto a screen for about 15 seconds; subjects would gain one mark for every correctly positioned digit they could recall on their answer sheets. There would be 10 trials giving a maximum possible score for each subject of 80. [...] The experimenter then introduced an element of competition by saying that he was interested ‘to see whether the Pran group does better that the Dilt group’ or vice versa, and ‘whether the Davis group does better than the Evans group ’ or vice versa. [...] Subjects were then physically divided into their groups and given new seating arrangements to correspond. Condition 1 subjects, already divided into country and photographer halves, were further divided into Pran or Dilt and Davis or Evans to produce four separate groups (n = 7), one at each corner of the room. In conditions 2-5, subjects were first divided into Pran and Dilt halves and then again into Davis and Evans groups such that the grouping in each corner of the room represented one of the four quadrants (n = 7) of the criss-cross. Once seated, subjects performed the perceptual task privately and individually and their answer sheets were collected.

(3) Evaluations of performance As a final decision-making task, subjects were asked in the following way to estimate performances on the perceptual task [...] In conditions 1-4, subjects were told that they would be estimating only how some people had done, whilst in condition 5 they were asked to estimate how everybody had done [...]. In condition 1, the country groups rated each other and the photographer groups rated each other. In conditions 2-4, subjects rated members of the in-in (same country-same photographer) and one other quadrant (in-out, out-in and out-out respectively); and in condition 5, subjects rated members of all quadrants (see Table 1). [...]

Dependent measures. These were subjects’ estimates of the performance of individuals and groupings of people on the perceptual task, always presented in the following order: (1) Individual ratings: subjects had to assign a score from 0-80 for each person mentioned in the booklet - the members of the relevant groups or quadrants. The order of presentation of ratees was randomized for each subject. (2) Average ratings: subjects then had to estimate the relative average scores obtained by the groups or quadrants whose members they had just rated individually, using a modified form of Tajfel’s choice matrices. In condition 5, subjects rated the two pairs of categories [...].

Type of Prejudice/Bias