How little does it take to turn strangers into "us" and "them"?

Remarkably little. In the early 1970s the psychologist Henri Tajfel set out to find the point at which discrimination begins. His plan was to build groups with as little meaning as possible and then add meaning step by step. The surprising finding was that even in the most minimal group conditions, people favoured their own group.

Here is what a participant went through. First they were randomly and anonymously sorted into "Group A" or "Group B", on grounds as trivial as a preference between paintings or the toss of a coin. Then, in an apparently unrelated task, they shared out a valuable resource such as money or points between other people. These people were known only by a code number and a group label, like "participant number 34 of Group A".

Two details of the design matter. The participant could not give anything to themselves, and nothing linked the in-group's total to their own pay-off. So direct self-interest and real competition were ruled out. Even so, people tended to give more to their own group than to the other. They were quite fair overall, but the tilt was clear, and it appeared even when favouring the in-group meant that the in-group got less in absolute terms, as long as it did better relative to the other group.

That is the core finding of the method: arbitrary labels, such as painting tastes or shirt colour, were enough to trigger a tendency to favour one's own side at the expense of others.