In recent work, Mielke (2004) argues against universal distinctive features, in favor of emergent features, with language acquirers relying both on the phonetic properties of the sounds they hear and on the phonological patterning to determine the relevant sound categories in the language being learned. Blevins (2004) presents a model of how phonological systems change over time, which relies heavily on perception and production in determining how sound systems evolve.

This presentation supports the emergence hypothesis with respect to features, and explores the implications of this hypothesis for phonological theory, arguing in favor of a three-part strategy:

1. determining what must be said in order to account for a given language pattern;

2. determining which parts can be a result of human cognitive capabilities and human physiology;

3. examining the (possibly null) residue -- properties of human language systems that cannot be derived from general human cognitive strategies nor from physical properties of humans -- for organizational principles specific to language.

Examples illustrate (i) crazy classes, where groups of segments that are not definable in terms of standard definitions of distinctive features; (ii) ambivalent segments, that is, segments that pattern as [+F] in one language and as [-F] in another language; and (iii) covert patterning: categorical patterning that is not shared in the speech community.