Amphibian monitoring programs rarely question the quality of data obtained by observers and often ignore observer bias. In order to test for bias in amphibian call surveys, we sampled 29 clusters of wetlands from the Rainwater Basin, Nebraska, USA, totaling 228 functionally connected wetlands. Sampling consisted of 3-minute stops where volunteers recorded species heard and made digital recordings. Based on 627 samples, we examined 3 types of observer bias: omission, false inclusion (commission), and incorrect identification. Misidentification rates ranged from 4.2% to 18.3%. Relatively high and unquantified error rates can negatively affect the ability of monitoring programs to accurately detect the population or abundance trends for which most were designed.