This paper examines the influence of sexual contact patterns (mixing matrices) on the pattern of the AIDS epidemic in a male homosexual community via numerical studies of a mathematical model of the transmission dynamics of HIV. A discussion is presented of the range of possible structures of networks of sexual contacts with extremes of assortative (within sexual activity groups) and disassortative (between sexual activity groups) mixing. The assortative mixing extreme is shown to generate the most rapid growth in the incidence of infection in the early stages of the epidemic while the disassortative extreme is shown to generate the epidemic of the largest magnitude over a long period. High within-group mixing (assortative) may generate multi-peak epidemics. The results are discussed in the context of both the interpretation of observed patterns of the spread of HIV and the acquisition of data on sexual contact patterns.