The potential of an experimental threshold for reducing the number of insecticide applications for control of the cotton aphid, Aphis gossypii Glover, was demonstrated in cotton. A 3-yr field study at Clarkedale, AR, compared the current economic threshold for the cotton aphid to an experimental threshold that incorporates predaceous coccinellids, aphid parasitoids and the aphid fungus, Neozygites fresenii (Nowakowski) Batko. Treatments consisted of untreated plots, conventional treatment threshold plots, experimental threshold plots, and a fungicide treatment designed to disrupt aphid fungus epizootics. This fungicide treatment failed to influence fungal epizootics, so data for these plots were removed. Twice weekly, aphids were counted on 20 plants per plot and predators were counted using a dislodgment method. An application of 0.033 kg ai/ha of imidacloprid was made when aphids reached treatment level according to conventional or experimental threshold levels. Insecticide applications were triggered by the conventional threshold on 18 and 28 June 1999, on 28 June and 3 July 2000, and on 7 and 12 July 2001. The experimental threshold called for applications on 28 June 1999, 3 July 2000, and 19 July 2001. Aphid densities peaked in untreated plots at 142, 76, and 27 aphids per leaf and in treated plots at 49, 34, and 29 in 1999, 2000 and 2001, respectively. Larval coccinellid densities in untreated plots peaked at 9.0, 3.7, and 0.6 larvae per row-m and in treated plots 3.0, 1.4, and 0.5 in 1999, 2000 and 2001, respectively. Adult coccinellid densities peaked in untreated plots at 3.0, 2.7, and 0.5 adults/row-m and in treated plots at 0.8, 2.1, and 0.5 in 1999, 2000, and 2001, respectively. During each year, an epizootic of N. fresenii caused a rapid decrease in aphid numbers in mid to late July, eliminating the need for additional treatments. In 1999, significantly lower yield of lint occurred in the untreated plots than in conventional or experimental plots. No significant yield differences among treatments occurred in 2000 or 2001. Use of the experimental threshold delayed the initial insecticide application by 1 wk and eliminated one insecticide application on the experimental plots whereas maintaining cotton yields.