Understanding host–pathogen evolutionary dynamics needs characterisation and quantification of processes occurring at many spatiotemporal scales. With this aim, the effects of smut on a naturally infected population of the summer annual Digitaria sanguinalis were followed for 4 years in an uncropped field. The main purpose of the study was to quantify the effects of within-population density on the infectivity and the aggressiveness of the pathogen in a range of densities that occurred naturally. The infectivity-related variable measured was the proportion of smutted plants at the end of each growing season; proportions were analysed using a generalised linear model with a binomial distribution considering the year, the density and their interaction as effects. The aggressiveness-related variables chosen were the number of smutted inflorescences per plant and per area, obtained over the last 2 years; they were analysed by means of ancova considering disease status (seeded or smutted), year, density and all the interactions between them. Although the disease is monocyclic, results showed clearly that infectivity increased with plant density. The number of inflorescences per plant was 1.5 times higher in smutted plants than in healthy plants throughout the range of densities. This variable declined when density increased, but as the infectivity increased at a higher rate, the aggressiveness also increased with density. The surprising results on infectivity are discussed in the context of current knowledge of plant–pathogen interaction dynamics, as well as neighbour effects on pathogen aggressiveness. Moreover, the results could be useful to develop weed biological control strategies.