How can sound descriptors arising from studies in music perception be applied to inform composing and hearing? Perceptual research by Wessel and Grey suggest timbre can be organized by listeners into a multi-dimensional spatial representation. Building on this work, we propose an approach to timbre that is based on computer analysis of perceptually-relevant descriptors. Using the recent concept of corpus-based concatenative synthesis (CBCS), a database of samples is plotted in a spatial representation corresponding to any two or three of these descriptors. A musical phrase may be generated by drawing a curve in the space or by closest matches to an external target sound file. While this technique can be used for more traditional sounds, it is especially effective for organizing non-pitched sounds based on their timbral characteristics, both in the contexts of electronic music and computer-assisted composition for acoustic instruments. We implement this technique with the CataRT package for computer programs Max/MSP and OpenMusic (OM). Recent works for instruments, interactive electronics, and sound installation illustrate this approach. Examples include What the Blind See for five instruments and live electronics, as well as a version of the same work for sound installation, in which the the listener is simultaneously the performer and part of the space itself. The title, taken from an article by neurologist Oliver Sacks, suggests perception as the focus of the musical experience. While these tools are already being effectively exploited by composers, they can be adapted for broader uses in improvisation, scholarship, and therapy.