Neurobiological models of drug abuse propose that drug use is initiated and maintained by rewarding feedback mechanisms. However, the most commonly used drugs are plant neurotoxins that evolved to punish, not reward, consumption by animal herbivores. Reward models therefore implicitly assume an evolutionary mismatch between recent drug-profligate environments and a relatively drug-free past in which a reward centre, incidentally vulnerable to neurotoxins, could evolve. By contrast, emerging insights from plant evolutionary ecology and the genetics of hepatic enzymes, particularly cytochrome P450, indicate that animal and hominid taxa have been exposed to plant toxins throughout their evolution. Specifically, evidence of conserved function, stabilizing selection, and population-specific selection of human cytochrome P450 genes indicate recent evolutionary exposure to plant toxins, including those that affect animal nervous systems. Thus, the human propensity to seek out and consume plant neurotoxins is a paradox with far-reaching implications for current drug-reward theory. We sketch some potential resolutions of the paradox, including the possibility that humans may have evolved to counter-exploit plant neurotoxins. Resolving the paradox of drug reward will require a synthesis of ecological and neurobiological perspectives of drug seeking and use.