Bark beetles (Coleoptera: Curculionidae: Scolytinae) play important roles in temperate conifer ecosystems, and also cause substantial economic losses. Although their general life histories are relatively similar, different species vary markedly in the physiological condition of the hosts they select. Most of ∼6000 known species colonise dead or stressed trees, a resource they share with a large diversity of insects and other organisms. A small number of bark beetle species kill healthy, live trees. These few are of particular interest as they compete directly with humans for resources. We propose that tree killing evolved when intense interspecific competition in the ephemeral, scarce resource of defence-impaired trees selected for genotypes that allowed them to escape this limitation by attacking relatively healthy trees. These transitions were uncommon, and we suggest they were facilitated by (a) genetically and phenotypically flexible host selection behaviours, (b) biochemical adaptations for detoxifying a wide range of defence compounds, and (c) associations with symbionts, which together aided bark beetles in overcoming formidable constitutive and induced host defences. The ability to detoxify terpenes influenced the evolutionary course of pheromonal communication. Specifically, a mate attraction system, which was exploited by intraspecific competitors in locating poorly defended hosts, became a system of cooperative attack in which emitters benefit from the contributions responders make in overcoming defence. This functional shift in communication was driven in part by linkage of beetle semiochemistry to host defence chemistry. Behavioural and phenological adaptations also improved the beetles’ abilities to detect when tree defences are impaired, and, where compatible with life history adaptations to other selective forces, for flight to coincide with seasonally predictable host stress agents. We propose a conceptual model, whereby the above mechanisms enable beetles to concentrate on those trees that offer an optimal trade-off between host defence and interspecific competition, along dynamic gradients of tree vigour and stand-level beetle density. We offer suggestions for future research on testing elements of this model.