This study compared the reliability of population estimates of adult horn fly, Haematobia irritans (L.), obtained using different sample units. Mean-variance relationships were similar for abundance estimates obtained by counting flies on the sunny sides of cattle, on the upper body, and on the whole animal. Precision varied among the sample units, and was lowest for estimates obtained using the sunny side. Abundance estimates obtained using the sunny side and upper body sample units were related to estimates obtained using the whole body sample unit. However, the proportion of flies in the upper body and sunny side sample units declined with increasing fly density. Seasonal movement toward the belly accounted for this decline. This movement resulted in bias in estimating rates of change based on counting flies on the sunny side and upper body sample units. Rates of change based on sampling the sunny side were more biased than estimates based on the upper body sampling unit. Bias in estimating rates of change was examined using an analytical model compared with field data, and resulted from changes in the proportion of flies occupying the sample unit. Bias also increased with increasing actual rates of change. The implication of these findings for studying horn fly populations are discussed.