Field studies were conducted in 2000–2002 to compare foliage-dwelling arthropod populations on Bacillus thuringiensis Berliner (Bt) (Bollgard) cotton and non-Bt (conventional) cotton season-long in South Carolina, Georgia, northern Alabama, and southern Alabama. For each of these four regions, three or four paired fields were sampled weekly in each of the 3 yr. Each pair of fields consisted of a Bt and a non-Bt cotton field, both at least 5 ha in size. The dominant arthropod taxa collected included target pests (heliothine moths and Spodoptera spp.), nontarget pests (stink bugs and plant bugs), and generalist natural enemies [Geocoris spp., Orius spp., Solenopsis invicta (Buren), ladybeetles, and spiders]. Where target pests were present, particularly Helicoverpa zea (Boddie), their numbers were consistently significantly lower in the Bt cotton fields. Natural enemy populations generally were not significantly different between the Bt and the non-Bt cotton fields (50% of all comparisons) and, where significant differences were present, natural enemy abundance usually was higher in the Bt than the non-Bt cotton fields. These differences were correlated with lower insecticide use on the Bt than the non-Bt cotton fields, particularly in South Carolina, where target pest pressure was heaviest. When presented with insect eggs or larvae as prey items, the larger natural enemy populations in Bt cotton fields exhibited significantly higher predation rates. These results show that Bt cotton has no significant adverse impacts on the nontarget arthropod populations studied and, compared with insecticide-treated non-Bt cotton, Bt cotton supports higher natural enemy populations with significant positive impacts on biological control.