SUMMARYGrowth, phenology and frost tolerance of seedlings from 50 populations of Douglas-fir (Pseadotsuga men ziesli var. glauca) were compared in 12 environments.Statistical analyses of six variables (bud burst, bud set, 3-year height, spring and fall frost injuries, and deviation from regression of 3-year height on 2-year height) showed that populations not only differed in mean performance, but also reacted differently to the environmental gradient. Most of the populationenvironment interaction was attributable to heterogeneous regressions of population means on environmental means. For all variables except growth rate, the variance of heterogeneous regression coefficients was explained by convergence of regression lines to a common point on the environmental gradient. Consequently, mean values for populations were significantly correlated with regression coefficients. Thus, main effects of populations in those single environments that induced the greatest mean differences reflected the interaction.Multiple regression analyses associated adaptive differentiation of populations with geographic and ecologic characteristics of the seed source. Differentiation was controlled primarily by elevation and secondarily by latitude.Whereas longitude was a minor factor, habitat types accounted for no differentiation beyond that associated with elevation, a factor closely correlated with habitat types. From these results it is recommended that seed for reforestation should not be moved more than 140 m elevation, l6 degrees latitude, or 27 degrees longitude in northern Idaho and eastern Washington.