In this article, we present the preliminary results of a study on sexual networks in a homosexual community in Iceland. The network data are transformed into a mixing matrix, which suggests that the pattern of sexual contact within the community is disassortative (i.e., sexual contacts occur largely between individuals of disparate sexual activity levels). The implications for the spread of human immunodeficiency virus (HIV) in the community are assessed by comparison with epidemics generated by proportionate and assortative (like with like) mixing, using a mathematical model of HIV transmission. The potential magnitude of the epidemic associated with the data is shown to be larger than both proportionate and assortative alternatives, although the early rate of growth is lower for the mixing matrix implied by the data. The assumptions made concerning behavioural changes triggered by alterations in population structure as a consequence of AIDS-induced mortality do not significantly alter the shape of the predicted epidemic. The problems in data acquisition and interpretation are discussed.