SummaryIn three field experiments in 1985 and 1986, we studied the effect of the date of primary infection on the spread of beet yellows closterovirus (BYV) and beet mild yellowing luteovirus (BMW) from artificially inoculated sugar beet plants. Laboratory‐reared vector aphids,Myzus persicae, were placed on these sources of virus. There was no substantial natural immigration of vectors or viruses. In two experiments, one with BMYV in 1985 and the other in BYV in 1986, populations of vector aphids remained low and there was little virus spread, i.e.c.50 infected plants from one primarily infected source. The cause of this small amount of spread was the low number of vector aphids. In the third experiment, with BYV in 1986, large populations ofM. persicaedeveloped and there was substantial virus spread: c. 2000 infected plants in the plots which were inoculated before canopy closure. In later‐inoculated plots in the same experiment, there was much less spread:c.100 infected plants per virus source plant. Differences between fields in predator impact are implicated as the most probable factor causing differences in vector establishment and virus spread between these three experiments. Virus spread decreased with later inoculation in all three experiments.A mathematical model of virus spread incorporating results from our work has been used to calculate how the initial proportion of infected plants in a crop affects the final virus incidence. This model takes into account the effect of predation on the development of the aphid populations. The processes underlying the spread and its timing are discus