A prospective cohort study was used to estimate the incidence of West Nile virus (WNV) infection in a group of unvaccinated horses (n = 37) in California and compare the effects of natural WNV infection in these unvaccinated horses to a group of co-mingled vaccinated horses (n = 155). Horses initially were vaccinated with either inactivated whole virus (n = 87) or canarypox recombinant (n = 68) WNV vaccines during 2003 or 2004, prior to emergence of WNV in the region. Unvaccinated horses were serologically tested for antibodies to WNV by microsphere immunoassay incorporating recombinant WNV E protein (rE MIA) in December 2003, December 2004, and every two months thereafter until November 2005. Clinical neurologic disease attributable to WNV infection (West Nile disease (WND)) developed in 2 (5.4%) of 37 unvaccinated horses and in 0 of 155 vaccinated horses. One affected horse died. Twenty one (67.7%) of 31 unvaccinated horses that were seronegative to WNV in December, 2004 seroconverted to WNV before the end of the study in November, 2005. Findings from the study indicate that currently-available commercial vaccines are effective in preventing WND and their use is financially justified because clinical disease only occurred in unvaccinated horses and the mean cost of each clinical case of WND was approximately 45 times the cost of a 2-dose WNV vaccination program.