How Is the Aspiration of English /p, t, k/ "Predictable"?

Abstract

Aspiration as a phonetic property of the English stop categories is usually said to be non-distinctive on the ground that its occurrence can be accounted for by context-sensitive rules. The word-pair pin -spin is often cited by way of example. The word-initial voiceless stop is aspirated; the post-/s/ voiceless stop is not. But the presence of aspiration is "predicted" only for some voiceless stops - namely those that are "spelled" phonologically /p/ and are either word-initial or in a position where the next vowel is stressed and in the same word. Initial stops that are spelled /b/, as in bin, may also be voiceless, so that a rule which predicts aspiration from the voicelessness of an initial stop will not work, since bin is never aspirated. Thus the knowledge on which the prediction is based is not the voicelessness of the stop, or indeed on any other ascertainable phonetic property. We know that in some words voiceless initial stops can be freely replaced by voiced stops without semantic effect, and that those voiceless stops are never aspirated, while in other words there are initial voiceless stops that are regularly aspirated, and cannot be freely replaced by voiced stops. In other words, we know whether a voiceless stop is to be aspirated or not if we know how it is spelled phonologically.