The classic study of Social Identity Theory, carried out by Tajfel (1970), had two key findings. First, the study showed the power of the minimal group paradigm. In this procedure, participants are first randomly and anonymously divided into two groups on the basis of trivial criteria (e.g., "you like Kandinsky" and "you like Klee").  Then participants take part in a resource distribution task where participants distribute a valuable resource (e.g., money or points) between other participants who are only identified by code number and group membership (e.g., "participant number 34 of Group A"). Participants are told that, after the task is finished, they will receive the total amount of the resource that has been allocated to them by the other participants.

The second key finding was that groups demonstrated an in-group bias. the tendency to favour one’s own group, its members, and its characteristics, in reference to other groups.  This leads to behaviours in which one's own group is rewarded and there be discrimination against members of an out-group.

Bagby and Rector (1992) criticized the minimal group paradigm as too artificial and wanted to investigate in-group bias in groups that already exist.  They also wanted to explore the role of in-group bias in a less artificial setting - a court-room.

The sample was made up of 102 psychology students from the University of Quebec in Montreal. 27 per cent of the sample was male.  All were bilingual, French-speaking Canadians.

The researchers carried out a true experiment to see if one's social identify would influence their objectivity in determining the guilt of a defendant accused of rape.

The participants were asked to read a transcript of a rape trial which varied the ethnicity of the defendant and the victim – between Anglophone (English-speaking) and Francophone (French-speaking) Québécois. In some cases, the defendant was a native French speaker and the victim was a native French speaker.  In other cases, the defendant was a native English speaker and the victim was a native French speaker.  And so on.

The transcript was in French.

The participants were randomly allocated to four conditions as jurors: French defendant/French victim; French defendant/English victim; English defendant/English victim; English defendant/French victim.  

In each condition, the participants were asked to read a transcript of a rape trial. The only difference was the social identity of the defendant and victim.  They were asked to complete a questionnaire. The first section of the questionnaire asked them to rate the personality traits of the defendant.  Then they were asked to determine the guilt of the defendant on a 7-point scale, with 7 being "extremely guilty." 

The final question asked about the ethnicity of the defendant and the victim.  If they got this wrong, they were omitted from the sample.  This was an important control to make sure that they had actually noted the social identities of the defendant and the victim.

A Chi-squared was used and found that there was no gender difference in the results.  In addition, there was no significant difference in the personality ratings in the first part of the questionnaire.

However, the researchers found that the French Canadians rated the out-group (English) defendant more guilty when the victim was from their in-group (French).  The average scores are in the table below.

The study is done with groups that occur naturally, showing that the phenomenon of in-group bias is not limited to artificially created groups.  The study also gives us a potential insight into why conviction rates of minority groups tend to be higher in countries like the US, McMurdo and Australia.

However, the study is still artificial.  Although the sample may believe that this is a legitimate case, they are not actual jurors and they did not see the actual testimonies; they only read a transcript.

There was a control to make sure that the social identity of the defendant and victim was actually noted by the participants. 

There was a sampling bias as the sample was made up of university students.  The level of intelligence, social skills and age may have played a role in the results of the experiment. In addition, being French Canadian in Quebec means that they are in the majority position. French Canadians outside of Quebec may see themselves as disenfranchised and this may affect their responses.

Only one crime was used in the study - and rape is a rather emotional topic.  It is unclear to what extent the nature of the crime may have influenced the participants' responses.