Zebra finches (Taeniopygia guttata) can categorize vowel-like sounds on both the fundamental frequency ("pitch") and spectral envelope.

Abstract

Humans can categorize vowels based on spectral quality (vowel identity) or pitch (speaker sex). Songbirds show similarities to humans with respect to speech sound discrimination and categorization, but it is unclear whether they can categorize harmonically structured vowel-like sounds on either spectrum or pitch, while ignoring the other parameter. We trained zebra finches in two experimental conditions to discriminate two sets of harmonic vowel-like sounds that could be distinguished either by spectrum or fundamental frequency (pitch). After the birds reached learning criterion, they were tested on new sounds that were either noise-vocoded versions of the trained sounds (sharing the spectral envelope with the trained sounds but lacking fine spectral detail from which pitch could be extracted) or sounds lacking the amplified harmonics (sharing only pitch with the trained sounds). Zebra finches showed no difference in the number of trials needed to learn each stimulus-response mapping. Birds trained on harmonic spectrum generalized their discrimination to vocoded sounds, and birds trained on pitch generalized their discrimination to harmonic sounds with a flat spectrum. These results demonstrate that, depending on the training requirements, birds can extract either fundamental frequency or spectral envelope of vowel-like sounds and use these parameters to categorize new sounds. (PsycINFO Database Record (c) 2019 APA, all rights reserved).