Two Bacillus subtilis biocontrol strains (GUS2000 and GUS376 - Gustafson Inc., Dallas, TX) were tested on cotton seed (SV453[1992-93] and SVLA887[1993 only]) for suppression of Fusarium oxysporum f.sp. vasinfectum, a causal agent of seedling disease and Fusarium wilt of cotton. Two field tests were conducted in 1992-1993 at the USDA Fusarium Wilt Screening Nursery in Tallassee, AL. All seed, except the untreated control, were pretreated with Carboxin, PCNB, and Metalaxyl: standard industry treatments. GUS2000 (Kodiak^®) was superior to GUS376 and chemical use alone for Fusarium wilt suppression. In 1992, disease incidence was reduced in plots treated with intermediate (5.74 log CFU/seed) and high rates (6.74 log CFU/seed) of GUS2OOO (p<0.075), as determined from season-long areas under the disease progress curve (AUDPC). The GUS2000 intermediate rate was better than any chemical treatment used alone (p<0.10). When results from both years were combined, GUS2000 produced lower AUDPC's than other treatments on SV453, a Fusarium susceptible cultivar. In 1993, on SVLA887, a Fusarium resistant cultivar, the use of additional Metalaxyl-PCNB as a hopper box treatment provided statistically equivalent results to GUS2000 (p<0.10), and both were better (p<0.05) than the standard seed treatment alone. Though not statistically significant, GUS2000 provided the lowest absolute rate of Fusarium wilt disease progress, provided the lowest percentage loss of plant stand, and had a higher end-of-season vigor. Absolute disease rate, vigor, and AUDPC were all highly correlated with yield in both years for the SV453 cultivar. These correlations explain higher yields observed with intermediate and high rates of GUS2000 in 1992. However, significant yield increases were not observed in 1993 with GUS2000, even with better disease control. However, GUS376 provided significantly higher yields than all other treatments in 1993 (P<0.05), with the exception of a GUS2000/GUS376 combination treatment. As reported by others, this data supports the idea that yield increases with PGPR-type organisms cannot always be easily correlated to disease control.