Vocal individuality varies between species and/or ontogenesis stages depending on needs in the vocal recognition, but also estimation of individual differences depends on the method of analysis. We studied pair-specific differences of duets elicited by mating pairs of Siberian crane Grus leucogeranus. We quantitatively described the duet structure and compared visual and statistical classification methods of pair identification by duet. Three methods were used: discriminant analysis, method of classification trees and visual classification of spectrogram. We found significant interpair differences. The pairs differ by duet structure that is by the ratio of male- and female-initiated duets and by the ratio of the number of male to female calls; temporal-frequency duet characteristics are pair-specific, too. All methods showed high interpair differences, which exceeded random values significantly. Discriminant analysis stepwise procedure based on 11 parameters resulted in 97.3% of correctly assigned duets. Human observers correctly assigned 80.7% of spectrograms. Our data provide a basis for remote monitoring of this endangered species with a wild population of only 3,000 birds.