Many studies suffer from imperfect detection probability, i.e., species are not detected when individuals may be present. In occupancy studies, detection probability is often treated as a nuisance variable. When used as a primary variable of interest, detection probability can be examined as a function of sampling covariates with the goal of maximizing the probability of encountering target species. Efforts to determine which methods maximize detection probability will benefit monitoring programs, particularly for species that are difficult to detect. We used three sampling methods, leaf litter bag (LLB) surveys, visual encounter surveys (VES), and flip and search (FS) methods to detect larval Spring Salamanders (Gyrinophilus porphyriticus). We estimated occasion-specific estimates of detection and used an analysis of variance to determine if detection probability varied among sampling methods. We found the FS method yielded higher detection estimates than did the LLB and VES. In addition, occupancy estimates derived from FS sampling changed drastically when compared among other single-method models, suggesting that LLB and VES gave biased estimates of occupancy related to a low probability of detecting Spring Salamanders at occupied sites. Furthermore, our results suggest the FS method provided higher detection probability estimates as compared to estimates derived from models that combined all sampling methods. In conclusion, efforts to monitor Spring Salamanders should rely on FS for sampling populations to maximize detection probability to reduce costs and increase effectiveness for large, widespread research projects.