ABSTRACT Variation in body size, behavior, feeding habits and habitat use patterns in medium- and large-sized mammals influence the adequacy of sampling methods to register presence and abundance. Moreover, even if methods are similarly adequate, different methodologies result in distinct cost-efficiency relationships (i.e. some may have reduced costs, be less time-consuming and/or require less-skilled technicians). Focusing on three different sampling methods commonly used to monitor medium and large mammals in seasonal tropical forests, we compared the species richness detected by each method and quantified their cost-efficiencies: (1) camera traps; (2) line transects for direct observations of animals; and (3) line transects seeking tracks/footprints. We simultaneously monitored medium and large mammals along five trails between July and August 2009 and January and February 2010, in the Serra do Japi Biological Reserve, São Paulo, Brazil. Data from two distinct seasons demonstrated that significantly higher species richness was achieved by using signs of presence and direct observations detected in transects. Camera traps recorded the fewest species, but represented the lowest cost per species. Direct observations and searches for tracks/footprints required a greater number of field technicians (with more skill and experience) to record the focal species and therefore have a higher cost, but allowed twice as many species to be recorded compared to camera traps. The choice of sampling methodology depends on the study objective, mammal species targeted and/or amount of resources available. We advocate use of camera traps for long-term studies and in conjunction with the other two methods to improve identification accuracy, allow individual identification and permit more accurate abundance estimates.