Although dramatic amphibian declines have been documented worldwide, only few of such events have been quantitatively documented for the tropical forests of South America. This is due partly to the fact that tropical amphibians are patchily distributed and difficult to detect. We tested three methods often used to monitor population trends in amphibian species in a remote lowland tropical forest of French Guiana. These methods are capture-mark-recapture (CMR), estimation of the number of calling males with repeated counts data and distance sampling, and rates of occupancy inferred by presence/absence data. We monitored eight diurnal, terrestrial amphibian species including five Dendrobatidae and three Bufonidae. We found that CMR, the most precise way of estimating population size, can be used only with two species in high density patches where the recapture rate is high enough. Only for one of the species (Dendrobates tinctorius), a low coefficient of variation (CV = 0.19) can be achieved with 15 to 20 capture events. For dendrobatid species with day-calling males, audio surveys yield a better probability of detection with only 8 audio surveys needed; quantitative estimates can be achieved by computing the number of calling males inferred from audio counts or distance sampling analysis. We therefore suggest that an efficient monitoring protocol for Neotropical amphibian species should include a combination of sighting and audio techniques, and we discuss the need of implementing a large-scale monitoring in order to provide a baseline for comparison with future changes.