Since the 1980s, the occurrence of pertussis cases in developed countries has increased and shifted towards older age groups. This resurgence follows 30 years of intense mass vaccination, and has been attributed primarily to three factors: (1) more effective diagnosis of the disease, (2) waning of vaccine-induced immunity, and (3) loss of vaccine efficacy due to the emergence of new Bordetella pertussis strains. Here we develop and analyse a mathematical model to assess the plausibility of these hypotheses. We consider that exposure to B pertussis through natural infection or vaccination induces an immune response that prevents severe disease but does not fully prevent mild infections. We also assume that these protective effects are temporary due to waning of immunity. These assumptions, describing the mode of action of adaptive immunity, are combined with a standard transmission model. Two distinct epidemiological scenarios are detected: under low transmission, most infections lead to severe disease; under high transmission, mild infections are frequent, boosting clinical immunity and maintaining low levels of severe disease. The two behaviours are separated by a reinfection threshold in transmission. As a result, the highest incidence of severe disease is expected to occur at intermediate transmission intensities--near the reinfection threshold--suggesting that pertussis resurgence may be induced by a reduction in transmission, independently of vaccination. The model is extended to interpret the outcomes of current control measures and explore scenarios for future interventions.