The contact hypothesis posits that having diverse neighbours may reduce one’s intergroup prejudice. This hypothesis is difficult to test as individuals self-select into neighbourhoods. Using a slum relocation programme in India that randomly assigned neighbours, I examine the effects of exposure to other-caste neighbours on trust and attitudes towards members of other castes. Combining administrative data on housing assignment with original survey data on attitudes, I find evidence corroborating the contact hypothesis. Exposure to more neighbours of other castes increases inter-caste trust, support for inter-caste marriage, and the belief that caste injustice is growing. I explore the role of friendships in facilitating these favourable attitudes. The findings shed light on the positive effects of exposure to diverse social groups through close proximity in neighbourhoods.