Roy D. Patterson
MRC Appl. Psych. Unit, 15 Chaucer Rd., Cambridge CB2 2EF, UK
Fletcher [Rev. Mod. Phys. 12, 47--65 (1940)] summarized his observations on pitch, loudness, and masking in terms of ``auditory patterns---spiral lines representing the cochlea with shaded regions showing neural responses to sinusoids. The auditory filterbank was a set of overlapping auditory patterns spanning the frequency range of hearing---a concept that has served as a functional model of auditory frequency analysis ever since. It has four main components: the filter shape, its bandwidth, the distribution of filters across frequency, and the detection criterion at the filter output. Fletcher identified these components and focused attention on them with his famous ``band-widening'' experiment in which a tone is masked by a variable width noise centered on the tone. The current paper reviews four areas of research spawned by the components of Fletcher's filterbank concept. The research has shown that the band-widening experiment is actually rather insensitive, and subject to a confounding which led to underestimation of filter bandwidth and overestimation of the detection criterion. But the more important assumptions about filter shape and distribution were proven essentially correct. As a result, current, computational, auditory filterbanks are surprisingly similar to Fletcher's original conception.