This article questions the view that phonetics governs exclusively gradient phenomena. Experimental data from an ultrasound study of English flap allophones show unconditioned categorical variation in the production of everyday words. Individual variation is discussed as an essential area of empirical investigation in uncovering emergent categorical properties of language.

Examination of phonetic data can show when phonologization has or has not occurred. Under the general Economy assumption of generative linguistics, this is important for theoretical phonology, since only cases where it has occurred require a phonological account. A modular view of phonetics and phonology is assumed, as it is clearly consistent with generative theory. A procedure for acoustic phonetic testing for phonologization is illustrated using data from one speaker of Palestinian Arabic and one speaker of Stát'imcets Salish. The testing can confirm phonological accounts or cause their revision. How it can cause revision is illustrated with the Stát'imcets data: the speaker's retraction harmony affects a rightward vowel, which is unexpected from standard descriptions of the harmony in that language. The testing can also identify cases of phonetics-within-language. This is illustrated with the St'at'imcets speaker's schwa lowering.

This article argues for the existence of expressive palatalization (E-Pal) – a phonologically unmotivated process that applies in sound symbolism, diminutive constructions, and babytalk registers. It is proposed that E-Pal is grounded in iconic sound-meaning associations exploiting acoustic properties of palatalized consonants and thus is inherently different from regular phonological palatalization (P-Pal). A cross-linguistic survey of patterns of E-Pal in 37 languages shows that it exhibits a set of properties different from P-Pal. The case study focuses on patterns of palatalization in Japanese mimetic vocabulary and babytalk. Two experiments testing native speaker intuitions of these patterns revealed that both patterns exhibit place and manner asymmetries typical of cross-linguistic patterns of E-Pal. The cross-linguistic survey, the two experiments, and analysis of the origins and structural differences of E-Pal and P-Pal provide strong empirical and theoretical motivation to distinguish the two.

It is argued that proxy relations between feature-defining and enhancing gestures in the phonetics can become phonologized. The article proposes Optimality-theoretic markedness constraints, which are grounded in phonetic enhancement and which can compel a delinked/unparsed/floating feature to be realized with another feature (or prosodic element), giving the impression of featural transmutation. Among the phonological phenomena revisited are incomplete neutralization, devoicing, vocalization, and debuccalization.