When sighting-based surveys to estimate population densities of large herbivores in tropical dense forests are not practical or affordable, surveys that rely on animal dung are sometimes used. This study tested one such dung-based method by deriving population densities from observed dung densities of six large herbivores (chital, elephant, gaur, muntjac, sambar, and wild pig) in two habitats, dry deciduous forests (DDF) and moist deciduous forests (MDF), within Nagarahole National Park, southern India. Using the program DUNGSURV, dung pile counts, decay rates estimated from field experiments, and defecation rates derived from literature were analyzed together by a model that allows for random events affecting dung decay. Densities of chital were the highest, followed by sambar. Wild pig densities were similar in the two habitats, sambar densities were higher in DDF, and densities of the other species were higher in MDF than in DDF. We compared DUNGSURV estimates with densities estimated using distance sampling in the same season. DUNGSURV estimates were substantially higher for all species in both habitats. These differences highlight the challenges that researchers face in computing unbiased estimates of dung decay rates and in relying on defecation rates from literature. Besides the elephant, this study is the first to rigorously test the efficacy of using a dung-based approach to estimate densities of large herbivore species in Asia, and based on this evaluation, we provide specific recommendations to address issues that require careful consideration before observed dung densities are used to derive animal densities. Our results underline the need for an experimental study of a known population in a fenced reserve to validate the true potential of using dung-based approaches to estimate population densities.