Abstract
Consonant reduction is often treated as an allophonic process at the phonological planning level, with one production target (allophone) being substituted for another. We propose that, alternatively, reduction can be the result of an online process driven by prosodically conditioned durational variability and an invariant production target. We show that this approach can account for patterns of coronal stop (/t/, /d/, and /n/) production in both American English and Spanish. Contrary to effort-driven theories of reduction, we show that reduction does notdepend on changes to gestural stiffness. Moreover, we demonstrate how differences between and within a language in the particular articulatory postures used to produce different coronal stops automatically lead to reduction to what have normally been considered distinct allophones - coronal approximants ([ð̞]) and flaps ([ɾ]). In this way, our approach allows us to understand different outcomes of coronal stop reduction as the dynamic interaction of a single process (durationally driven undershoot) and variable spatial targets. We show that these patterns are reflected across a wide variety of languages, and show how alternative outcomes of reduction may fit within the same general framework.