Body condition metrics are widely used to infer animal health and to assess costs of parasite infection. Since parasites harm their hosts, ecologists might expect negative relationships between infection and condition in wildlife, but this assumption is challenged by studies showing positive or null condition-infection relationships. Here, we outline common condition metrics used by ecologists in studies of parasitism, and consider mechanisms that cause negative, positive, and null condition-infection relationships in wildlife systems. We then perform a meta-analysis of 553 condition-infection relationships from 187 peer-reviewed studies of animal hosts, analysing observational and experimental records separately, and noting whether authors measured binary infection status or intensity. Our analysis finds substantial heterogeneity in the strength and direction of condition-infection relationships, a small, negative average effect size that is stronger in experimental studies, and evidence for publication bias towards negative relationships. The strongest predictors of variation in study outcomes are host thermoregulation and the methods used to evaluate body condition. We recommend that studies aiming to assess parasite impacts on body condition should consider host-parasite biology, choose condition measures that can change during the course of infection, and employ longitudinal surveys or manipulate infection status when feasible.