Using network models of absolute pitch to compare frequency-range discriminations across avian species

The spectral frequency ranges of song notes are important for recognition in avian species tested in the field. Frequency-range discriminations in both the field and laboratory require absolute pitch (AP). AP is the ability to perceive pitches without an external referent. The authors provided a network model designed to account for differences in AP among avian species and evaluated it against discriminative performance in eight-frequency-range laboratory tests of AP for five species of songbirds and two species of nonsongbirds. The model's sensory component describes the neural substrate of avian auditory perception, and its associative component handles learning of the discrimination. Using only two free parameters to describe the selectivity and the sensitivity of each species’ auditory sensory filters, the model provided highly accurate predictions of frequency-range discrimination in songbirds and in a parrot species, but performance and its prediction were less accurate in pigeons: the only species tested that does not learn its vocalizations. Here for the first time, the authors present a model that predicted individual species’ performance in frequency-range discriminations and predicted differences in discrimination among avian species with high accuracy.