Perception of voiced-only and noise-vocoded speech by a language-trained chimpanzee ( Pan troglodytes )

The ability of human listeners to understand speech even in altered or synthetic forms is argued to be evidence of uniquely human processing abilities. However, extensive early experience with speech may also contribute to this capability. To investigate this issue two experiments were designed to test the ability of Panzee, a 22-year-old language-trained chimpanzee, to recognize words in synthetic form. Like a human child, she was reared from infancy by human caregivers who routinely spoke to her. She communicates with humans by identifying words using graphical symbols (lexigrams). Experimental training and testing were conducted with two different sets of 24 familiar words presented one-per-trial in natural, voiced-only, or noise-vocoded forms, with Panzee choosing one of four lexigrams presented on a computer monitor. Experiment 1 showed equivalent performance with words heard in natural form versus voiced-only versions, resynthesized from only voiced components of a word. Noise-vocoded words presented in Experiment 2 simulated effects of hearing using a cochlear implant and were based on amplitude-modulated noise bands. Performance with these sounds was significantly higher than chance, but also lower than with natural words. Results suggest specialized processing mechanisms are not necessary to speech perception in the absence of traditional acoustic cues.