In the history of English, two diametrically opposite processes took place: voiced obstruents underwent devoicing and voiceless fricatives were voiced in word-final positions. While the former process is entirely natural in terms of its phonetic motivation, the latter must be regarded, by the simplest logic, as unnatural. This state of affairs presents a challenge to 'naturalist' approaches which have come to play an important role in linguistics. The traditional account of the two opposite processes is that the devoicing rule applied to stressed and the voicing rule to unstressed words. Because this hypothesis runs into problems, an alternative account is developed which relates the issue of voicing vs. devoicing to lexical frequency. Only the ends of the most common words are susceptible to voicing, all others exhibit devoicing. As voiced fricatives are phonetically much shorter than voiceless ones, voicing is construed as one facet of the shortening process to which high-frequency words are subject. This effect may be so strong that it can bring about an 'unnatural' phonological process. Voicing and devoicing emanate from two natural principles, a semiotic and a phonetic one, which have opposite effects. Thus, the historical data are fully congruent with the tenets of Natural Phonology.