Adventures in Vernerization: Voiced and Voiceless Verb Endings in Germanic
Endings: 2 sg., 3 sg., 2 pl., 3 pl.
Old English / Old Frisian: s(t), þ, -, (*n)þ
Ancient Nordic (that is, the language of the earliest Scandinavian runic inscriptions, c. 150-550, ancestor of Old Norse, whatever its relationship to the theoretical entity North-West Germanic): z, þ, ?, *n(þ)
Old Saxon: s, d, -, (*n)þ
Old High German (before 2nd consonant shift): s(t), *d, *d, *nd
Gothic (before final devoicing of fricatives, see below): z, ?, d, nd
Summary of data: According to Verner's Law, inherited voiceless endings in Proto-Germanic would have become voiced in the unstressed endings of strong verbs. But the resulting distinction of voiced endings in strong verbs and voiceless in weak verbs hasn't survived in any of those daughter languages which are sufficently attested for us to tell. Instead the early Germanic world was divided by a splayed set of isoglosses. At its western edge, Old English and Old Frisian generalised the voiceless endings to both types of verb. In the east, Gothic probably generalised all voiced endings (although evidence for the fricatives is limited to just a couple of examples where an enclitic particle covers the ending, preventing the devoicing that otherwise applies to all final fricatives in this language). At the northern and southern extremes, Ancient Nordic and Old High German generalised a mix of voiced and voiceless endings, each taking the exact opposite set of options to the another. Old Saxon, in the middle, agrees with the Ingvaeonic languages (Old English and Old Frisian), except in the 3rd pers. sg. where it agrees with Old High German.
Suggestion: These generalisations were completed at a time when the most periferal dialects were still in reasonably close contact with their nearest Germanic neighbours, and were still essentially the same language. This would explain why likeness is proportional to nearness, and why the isoglosses rarely coincide [ http://listserv.linguistlist.org/cgi-bin/wa?A2=ind9805c&L=histling&D=1&F=&S=&P=1083 ]. If the levellings took place individually after the periferal dialects were isolated, their selection of endings ought to be random with no connection to that of their neighbours.
Complications and alternatives: The original pattern may have been disrupted by a later devoicing rule in the northern dialects (cf. gaf 'gave' in the Stentoften stone inscription). There is also a possibility that some endings--especially the 2nd pers. sg.--could have been devoiced by assimilation to the initial consonant of following pronouns. Old Saxon is thought to have been an Ingvaeonic dialect that was later influenced by more southerly forms of German; it could be that the isoglosses were more bundled at an earlier date.