Tritium and He isotopes have been measured at a site where groundwater flow is nearly vertical for a travel time of 100 years and where recharge rates are spatially variable. Because the mid-1960s 3H peak (arising from aboveground testing of thermonuclear devices) is well-defined, the vertical groundwater velocity is known with unusual accuracy at this site. Utilizing 3H and its stable daughter 3He to determine groundwater ages, we compute a recharge rate of 0.16 m/yr, which agrees to within about 5% of the value based on the depth of the 3H peak (measured both in 1986 and 1991) and two-dimensional modeling in an area of high recharge. Zero 3H/3He age occurs at a depth that is approximately equal to the average depth of the annual low water table, even though the capillary fringe extends to land surface during most of the year at the study site. In an area of low recharge (0.05 m/yr) where the 3H peak (and hence the vertical velocity) is also well-defined, the 3H/3He results could not be used to compute recharge because samples were not collected sufficiently far above the 3H peak; however, modeling indicates that the 3H/3He age gradient near the water table is an accurate measure of vertical velocities in the low-recharge area. Because 3H and 3He have different diffusion coefficients, and because the amount of mechanical mixing is different in the area of high recharge than in the low-recharge area, we have separated the dispersive effects of mechanical mixing from molecular diffusion. We estimate a longitudinal dispersivity of 0.07 m and effective diffusion coefficients for 3H (3HHO) and 3He of 2.4¿10-5 and 1.3¿10-4 m2/day, respectively. Although the 3H/3He age gradient is an excellent indicator of vertical groundwater velocities above the mid-1960s 3H peak, dispersive mixing and diffusive loss of 3He perturb the age gradient near and below the 3H peak. ¿ American Geophysical Union 1993