Simultaneous improvement of forage traits and seed yield in orchardgrass (Dactylis glomerata L.) has been problematic because of geographic separation of forage and seed production locations. Previous work has shown that a complex multilocation selection program in forage production environments can increase forage yield as well as seed yield in Oregon. The objective of this experiment was to compare target‐environment (TE) and nontarget–environment (NTE) selection approaches for increasing seed yield of orchardgrass in Oregon. Two cycles of recurrent phenotypic selection for panicle seed mass (PSM) and agronomic traits were conducted on four populationsin four eastern USA locations (NTE) and one Oregon location (TE). Seed yield was increased in three of four orchardgrass populations by TE selection, averaging 5.1% cycle−1, but was improved by NTE selection in only one of four populations. Conversely, TE selection for PSM and agronomic traits resulted in no changes to forage yield in the eastern USA and Canada, while NTE selection for PSM and agronomic traits increased forage yield in two of four populations, confirming results of a previous study. It appears that the most efficient system for simultaneously improving forage and seed traits of orchardgrass would be to practice selection for forage traits in forage production environments and seed traits in seed production environments, with sufficiently large populations to allow multitrait selection.