Survival is a fundamental parameter in population dynamics with increasing importance in the management and conservation strategies of wildlife populations. Survival probability in vertebrates is usually estimated by live-encounter data obtained by means of physical mark–capture–recapture protocols. Non-invasive acoustic marking relying on individual-specific features of signals has been alternatively applied as a marking technique, especially in secretive species. Nevertheless, to date no research has compared survival rate estimates obtained by acoustic and physical marking. We estimated half-yearly and annual survival and recapture rates of a secretive and threatened passerine, the Dupont's lark Chersophilus duponti, using two separate live-encounter data sets of males collected simultaneously by physical and acoustic marking in the same study area. The separate analysis of both methods led to different model structures, since transient individuals had to be accounted for in the acoustic marking but not in the physical marking data set. Furthermore, while reencounter probabilities did not differ between methods, survival estimates employing physical marking were lower than those obtained acoustically, especially between the postbreeding and the breeding period when the apparent survival of colour-banded birds was twice as low as for acoustic marking. The combination of marking methods suggested the existence of different subsets of individuals differentially sampled within the population: whereas colour-banded males seemed to represent the territorial fraction of the population, both resident and floater individuals were probably detected by acoustic marking. Using traditional mark–recapture methods exclusively could have misled our estimates of survival rates, potentially affecting prospective predictions of population dynamics. Acoustic marking has been poorly applied in mark–recapture studies, but might be a powerful complement to obtain accurate estimates of fundamental demographic parameters such as survival and dispersal.