Many perennial plants in semiarid rangelands have experienced population declines, and understanding the ecological and demographic processes behind these declines is important to slowing or reversing them. Although anthropogenic disturbances drive many declines, other sorts of environmental variability, such as the differences in solar radiation with aspect, may impact population success locally. We experimentally assessed the role of solar radiation in driving an apparent decline in a common perennial bunchgrass, bluebunch wheatgrass (Pseudoroegneria spicata), on south-facing slopes at a site in the Columbia Basin of North America. Across three separate experiments on south-facing slopes, we observed dramatically (5−36 ×) higher seedling success in plots shaded to approximate the solar radiation of north-facing slopes relative to adjacent open (unshaded) plots. When we applied the rates of seedling success from these experiments to demographic models from this site, we found that seedling success in unshaded plots was often too low to allow a stable population on south-facing slopes, but that seedling success in shaded plots was often high enough to allow a stable or increasing population on north-facing slopes. We originally hypothesized that the primary mechanism driving this effect was water stress caused by greater evaporation from hotter open plots. However, despite soils in open plots being consistently hotter than soils in shaded plots, we did not observe a clear pattern of greater soil moisture in shaded plots. Therefore, it appears that higher solar radiation or higher temperatures on south-facing slopes may be sufficient to dramatically reduce seedling survival, and that the higher density of bluebunch wheatgrass on north-facing slopes relative to south-facings slopes may be driven primarily by this low survival of seedlings. As climate warms, the reduced seedling survival that threatens our bluebunch wheatgrass population may be expected to threaten many other species of perennials in similar rangelands.