Two approaches have been used in the monitoring of local populations of cereal aphids: direct counting of aphids on a number of tillers, and the less time-consuming assessment of incidence levels (the proportion of tillers infested). This paper compares the accuracy of these two methods, using a fixed sample size, as is desirable in a practical monitoring system. It is shown that the two methods are equally reliable over a range of incidence levels, but that the confidence intervals for the population density are very broad. Reduction of these sampling error requires a considerable increase in sample size, so they should instead be taken into account in the management decision-making process.