In plant-virus disease epidemiology, dynamical models have invariably incorporated a bilinear inoculation rate that is directly proportional to both the abundance of healthy (susceptible) hosts and the abundance of infective vectors. Similarly, the acquisition rate is usually assumed to be directly proportional to the abundance of nonviruliferous vectors and that of infectious hosts. These bilinear assumptions have been questioned for certain human diseases, and infection rates that incorporate power parameters of the variables have been proposed. Here, infection rates for plant-virus diseases that are of a more general form than the familiar bilinear terms are examined. For such diseases, the power parameter can be regarded as a measure of the spatial aggregation of the vectors or as a coefficient of interference between them, depending on the context.Field data of cassava mosaic virus disease (CMD) incidence were examined. When vector population density and disease incidence were high, disease progress curves over the first 6 months from planting could not be explained using models with bilinear infection rates. Incorporation of the new infection terms allowed the range of observed disease progress curve types to be described. New evidence of a mutually beneficial interaction between the viruses causing CMD and the whitefly vector, Bemisia tabaci, has shown that spatial aggregation of the vectors is an inevitable consequence of infection, particularly with a severe virus strain or a sensitive host. Virus infection increases both vector fecundity and the density of vectors on diseased plants. It is postulated that this enhances disease spread by causing an increased emigration rate of infective vectors to other crops. Paradoxically, within the infected crop, vector aggregation reduces the effective contact rate between vector and host and therefore the predicted disease incidence is less than when a bilinear contact rate is used.