The most common method for identifying individual amphibians is toe-clipping (TC), whereby captured individuals are marked by a unique combination of amputated phalanges that corresponds to a unique alphanumeric code. However, ethical and methodological objections to this method have been raised and there is broad interest in developing alternative methods. One alternative is to use photo-identification methods (PIMs) to identify individuals based on their natural markings. We tested the efficacy of TC and two PIMs — visual matching (VM) and computer-assisted matching (CAM) using the software Wild-ID — in identifying individual adults of the endangered southern red-bellied toad, Melanophryniscus cambaraensis. We collected data over 5 mo at Floresta Nacional de São Francisco de Paula, Rio Grande do Sul, Brazil. All specimens were toe-clipped and photographed. The total dataset included 492 captures of 147 individuals. VM was most accurate (99.4%), followed by TC (95.3%) and CAM (90.9%); VM was significantly more accurate than TC and CAM and TC was significantly more accurate than CAM. CAM accuracy diminished as dataset size increased but was considerably faster than VM. All CAM and VM errors were false negatives but involved different images; all TC errors were cross-identifications. Given that misidentifications occurred using both PIMs and TC, our results suggest that studies that require high accuracy should employ at least two methods to allow cross-validation. The performance of each method and the impacts of different kinds and rates of error on inferences depend on the organisms, field conditions, dataset sizes, and study questions. As such, researchers must carefully evaluate the trade-offs of each method before investing significant time and resources in collecting field data.