Listeners use distributed spectral information but not all of it

Evidence

Profile analysis, a technique developed in the 1980s by David Green and his colleagues (Green, 1988; Spiegel, Green, & Picardi), provided answers to some of these questions. Profile analysis measured the ability to discriminate between sounds with different spectral profiles, sounds with different shapes in the power spectrum. For example, two vowels have very different spectral profiles because they have formants at different frequencies. In particular, Green’s experiments showed that we can discriminate spectral profiles across a very wide frequency range (effectively the entire audible bandwidth) and that our perception benefits from distributed spectral information across frequency. Yet, Berg and Green (1990) expanded on the profile-analysis technique and showed that while human listeners indeed benefited from added information across the audible bandwidth, they did not use all of the spectral information for discrimination. In general terms, they found that listeners performed more poorly than “bestperformance” models, suggesting loss of some spectral information in the perceptual process. — Jennifer Lentz, Psychoacoustics- Perception of Normal…, 523

Categorization

• DEFINE: Defines profile analysis as measuring discrimination of spectral envelope shapes across frequency.
• XREF: The finding that listeners underperform best-performance models parallels informational-masking and channel-capacity results from other auditory literature.