Measurements of NO, NOy, CO, and O3 were conducted at Oki Island, 65 km west of the Japanese mainland during September--October 1991. The results show that the Oki Island site is relatively clean with mean (and median) CO and NOy concentrations of 137 (130) ppbv and 713 (505) pptv, respectively. These relatively low concentrations reflect the frequent occurrence of winds from the less populated regions to the north during this campaign. During several periods, elevated CO, NOy, and O3 concentrations were associated with transport of pollutants from Japan and Korea. CO and NOy are significantly correlated in the entire data set, reflecting the overall influence of anthropogenic emissions. Mean CO and NOy concentrations show small diurnal cycles with a maximum in the morning and early evening hours. Hourly mean O3 concentrations exhibit a diurnal cycle of amplitude 3.5 parts per billion by volume, with a maximum in the early afternoon attributed to photochemical O3 production. The source of this afternoon O3 enhancement is analyzed using regression analysis of O3 and NOy and of O3 and CO. Both NOy and CO are significantly correlated with O3 during the daytime but are poorly or insignificantly correlated with O3 at night, indicating that upwind photochemistry operating within the previous ~1 day was responsible for the observed afternoon O3 enhancement. The NOy--O3 regression analysis indicates formation of ~9 O3 molecules per NOy molecule reaching Oki Island. The mean and median midday NO concentrations of 55 and 23 parts per trillion by volume, respectively, were sufficient to support net production of O3. Using the results of these measurements and prior modeling studies (Lin et al., 1988), we estimate a ''lifetime-averaged'' O3 production efficiency of 10 molecules O3 per NOx molecule. Combined with estimated East Asian NOx emissions for 1987, this indicates an annual O3 production of 1.0¿1014 g O3 from photochemistry which is approximately 2 times the annual stratospheric flux in this region. This value must be considered as a rough estimate but is probably accurate to within a factor of 2. ¿ American Geophysical Union 1996