1. During drought, the tree subpopulations (such as size or vigor classes) that suffer disproportionate mortality can be conceptually arrayed along a continuum defined by the actions of biotic agents, particularly insects. At one extreme, stress dominates: insects are absent or simply kill the most physiologically stressed trees. At the opposite extreme, host selection dominates: outbreaking insects kill trees independently of their stress, instead selecting trees based on size or other traits. Intermediate responses are also possible. Yet for mixed-species forests, we lack a broad understanding of the relative importance of insects in determining exactly which subpopulations of trees suffer disproportionate mortality during drought, and whether these subpopulations differ among co-occurring tree species. 2. During an extreme drought, we documented the roles of native bark beetles in the mortality of five tree species in California’s Sierra Nevada. We analyzed patterns and agents of tree mortality in 12 permanent plots, and patterns of mortality in 89 temporary plots. 3. Most tree mortality was associated with bark beetles. But the growth rates (an indicator of chronic stress) and sizes of trees that suffered greatest bark-beetle-related mortality differed sharply among tree taxa, variously conforming with domination by stress (Abies concolor), domination by host selection (Pinus lambertiana and P. ponderosa), or a mix of the two (Calocedrus decurrens). Quercus kelloggii mortality remained relatively low. Thus, even during extreme drought substantial proportions of stressed trees survived because they were of sizes that mostly avoided fatal insect attack. Conversely, substantial proportions of comparatively unstressed trees died because they were of sizes that were selectively killed by outbreaking insects. 4. Synthesis. Native bark beetles were primarily responsible for determining which subpopulations of trees suffered greatest mortality during drought. However, idiosyncratic host-tree selection by the different bark beetle taxa meant that the tree subpopulations suffering greatest mortality differed strikingly among tree taxa – for example, high mortality of small trees of one species, but of large trees of another. If idiosyncratic host-tree selection by biotic mortality agents proves to be a generally common phenomenon, it could help explain weak broad-scale correlations between tree traits and tree mortality during drought.