Spectral contrasts underlying auditory stream segregation in goldfish (Carassius auratus)

Abstract

This study investigates the effects of spectral separation of sounds on the ability of goldfish to acquire independent information about two simultaneous complex sources. Goldfish were conditioned to a complex sound made up of two sets of repeated acoustic pulses: a high-frequency pulse with a spectral envelope centered at 625 Hz, and a low-frequency pulse type centered at 240, 305, 390, or 500 Hz. The pulses were presented with each pulse type alternating with an overall pulse repetition rate of 40 pulses per second (pps), and a 20-pps rate between identical pulses. Two control groups were conditioned to the 625-Hz pulse alone, repeated at 40 and 20 pps, respectively. All groups were tested for generalization to the 625-Hz pulse repeated alone at several rates. If the two pulse types in the complex resulted in independent auditory streams, the animals were expected to generalize to the 625-Hz pulse trains as if they were repeated at 20 pps during conditioning. It was hypothesized that as the center frequency of the low-frequency pulse approached that of the 625-Hz pulse, the alternating trains would be perceived as a single auditory stream with a repetition rate of 40 pps. The group conditioned to alternating 625- and 240-Hz pulses generalized least, with maximum generalization at 20 Hz, suggesting that the animals formed at least one perceptual stream with a repetition rate of 20 pps. The other alternating pulse groups generalized to intermediate degrees. Goldfish can segregate at least one "auditory stream" from a complex mixture of sources. Segregation can be based on spectral envelope and grows more robust with growing spectral separation between the simultaneous sources. Auditory stream segregation and auditory scene analysis are shared among human listeners, European starlings, and goldfish, and may be primitive characteristics of the vertebrate sense of hearing.

Full Text

The Full Text of this article is available as a PDF (134.2 KB).