The founding evolutionary theories of ageing indicate that the force of mortality imposed by environmental factors should influence the strength of natural selection against actuarial senescence and its evolution. To rigorously test this idea, field biologists need methods that yield estimates of age-specific mortality according to cause of death. Here, we present existing methods commonly applied in studies of human health that could be used to accomplish these goals in studies of wild species for which fate can be determined with certainty. We further present a new application of hidden Markov models for capture-reencounter studies of wild animals that can be used to estimate age-specific trajectories of cause-specific mortality when detection is imperfect. By applying our new hidden Markov model with the E-SURGE and MARK softwares to capture-reencounter data sets for long-lived species, we demonstrate that senescence can be severe for natural causes of mortality in the wild, while being largely non-existent for anthropogenic causes.Moreover, we show that conflation of mortality causes in commonly used survival analyses can induce an underestimation of the intensity of senescence and overestimation of mortality for pre-senescent adults. These biases have important implications for both age-structured population modelling used to guide conservation and comparative analyses of senescence across species. Similar to frailty, individual differences in causes of death can generate individual heterogeneity that needs to be accounted for when estimating age-specific mortality patterns. The proposed hidden Markov method and other competing risk estimators can nevertheless be used to formally account for these confounding effects, and we additionally discuss how our new method can be used to gain insight into the mechanisms that drive variation in ageing across the tree of life.