Psychological theories which attempt to explain the origins of prejudice fall into two major categories. Personality theories, which see the source of prejudice as being in the individual and social psychological theories, which see prejudice as a result of group membership.
An example of a personality theory would be Bandura's social learning theory, which argues that attitudes such as prejudices are learned from role models.
Many social psychological theories argue that society may be much more important than personality types in accounting for prejudice. Such theories see prejudice as a result of group membership and group interaction.
Tajfel like Sherif believes that the personality approach is inadequate in explaining prejudice and he also uses a social psychological approach. However, Tajfel et al (1971) argue that 'competition' is not a sufficient condition for inter-group conflict and hostility. Tajfel does not deny the importance of 'competition' between groups, personality types as explanations for the origins of prejudice but argues that mere perception of the existence of another group can itself produce discrimination. Tajfel et al argue that, before any discrimination can occur, people must be categorized as members of an in-group or an out-group, but more significantly the very act of categorization by itself produces conflict and discrimination.
By in-group we mean a group to which a person belongs, or thinks he or she belongs.
By out-group we mean a group to which a person does not belong, or thinks he or she does not belong.
The experiments carried out by Tajfel clearly demonstrate that inter-group discrimination is easy to trigger off. Tajfel demonstrates that the very act of categorization into groups is enough to produce conflict and discrimination.
In making their intergroup choices a large majority of the subjects, in all groups in both conditions, gave more money to members of their own group than to members of the other group. Intergroup discrimination was the strategy used in making intergroup choices.
In contrast the in-group and out-group choices were closely distributed around the point of fairness.
The second experiment also clearly demonstrated that the most important factor in making their choices was maximizing the differences between the two groups.
I think about this one a lot. (I also see the psychology vs sociology thing as a "yes and".)
(original text from here but it's an old website without a secure link)
If everything above this line was tldr, see also @myjetpack: