Speech perception by the chinchilla: Discrimination of sustained /a/ and /i/

Four chinchillas were trained to respond differently to sustained /a/ and /i/. The ensemble of vowels included two repetitions of each of the vowels by each of four talkers at each of three pitch levels for a total of 24 /a/'s and 24 /i/'s. The sound levels of the vowels were randomly changed from trial to trial over a 10-dB range. The animals easily transferred the training to a new set of vowels produced by 24 new talkers. Subsequently, the animals similarly transferred to synthetic /a/'s and /i/'s. The only relevant difference between the synthetic vowels was their formant structure, while there were irrelevant differences in their pitch contours. We conclude that the chinchilla can abstract some essential difference (s) between sustained /a/ and /i/ and ignore irrelevant variations of sound level, pitch level, pitch contour, and voice quality. These results are discussed in terms of perceptual learning and auditory concept formation.