Echolocation calls of bats are highly flexible and have been shown to vary within species due to habitat, foraging ecology, individual identity, family affiliation, sex, and age. Calls also vary between species due to morphology, habitat preference, and foraging ecology. Such plasticity in call design has made the development of species identification systems extremely difficult. Techniques employed previously include microphones linked to event recorders, manually tuning heterodyne bat detectors, zero-crossing analysis, discriminant function analysis (DFA) of time and frequency characteristics of calls, and synergetic analysis of spectrograms. To date, all have met with limited success. Research at the University of Bristol has combined estimates of curvature with ‘‘traditional’’ measures of echolocation calls (duration, start frequency, end frequency, frequency with most energy) to increase the precision of species identification. Call parameters were analyzed using a backpropagation neural network and DFA. Results indicate that a correct identification rate of 96% was possible for 14 species of British bats using the neural network. This compares with a success rate of 80% using the same data analyzed by DFA and represents the highest identification rate yet achieved.