This study examined the changing levels of air pollution at major racetracks in the United States over the past 35 years of Grade I Stakes Thoroughbred racing (n ¼ 675) to evaluate the impact on race performance. Information on winning race times, race speed, temperature, track conditions, ozone and particulate matter was analysed using multiple regression and ANOVA. The 8, 9 and 12 furlong races showed slight improvement in winning times over the 35-year time span. Although there was no significant difference in mean race speed in terms of particulate matter (F 4,112 ¼ 0.514, a , 0.05), the few races run under 'hazardous' ozone levels, based on the Pollutants Standards Index, were markedly slower (F 4,251 ¼ 12.365, a , 0.05). A multiple regression analysis of six additional pollutants in the Californian races revealed no significant effects beyond a slight improvement in the 7 furlong races, with decreased nitrogen dioxide levels. Of the nine racetracks with 9 furlong races, winning times in Pimlico and Belmont were approximately 4 s slower than average. Belmont also had the greatest percentage of races with hazardous ozone conditions. Other correlations between year, pollutants and improvements in race time were of small magnitude and may be an artefact of the sample size rather than of true predictive value. While racing under high levels of certain pollutants appears detrimental to equine athletes, very few races are run under these conditions and other factors such as breeding, training and individual variation may play a more significant role than this study could document with respect to pollution.