Si-Taek Yu

The traditional description of the phonem-graphem-relations in graphematics is not compatible with a phonological theory, in which an underlying vowel schwa is not assumed. If schwa is not a phoneme, it remains unclear, to which level of phonological representation the grapheme is related in words with a schwa-0-alternation. In the present article it is argued that the phonetic representation is the appropriate input for the orthography. A precise description of GPK-rules in words with a schwa-0-alternation is not possible without the assumption of a threefold distinction of vowels in German. In analyses with only one or two vowel series no graphematic condition is related to short tense vowels. The writing of schwa and the vowels alternating with schwa together with the regularities of double consonant writing are not explainable without reference to this vowel group. Therefore the phonetic representation is the base for orthography.

This study proposes a novel theoretical interpretation of the so-called Dehnungs- in Ger-man orthography. In contrast to the traditional view that the grapheme merely marks vowel length, the analysis suggests that serves a structural purpose: it signals the non-syllabic realization of the following sonorant. This function accounts for the restriction that only one sonorant may follow and explains why sequences such as are inadmissible within monomorphemic words. The study argues that functions as a grapho-tactic boundary, blocking morphological extensions and preventing homonym formation be-tween monosyllabic and morphologically complex bisyllabic forms. Only when and the sonorant belong to different morphemes is a sequence like permissible. Thus, the study reveals a previously overlooked function of , aligning graphematic, phonologi-cal, and morphological constraints within a unified structural framework.

Degemination in German applies not only in regular but also in irregular verbal inflexion. According to Wiese's (1996) stratal organization for German, these two morphological processes will be placed on two discontinuous levels of lexicon, posing the problem that a rule of degemination disobeys the general restriction in Lexical Phonology (Mohanan 1982, 1986); the domain of a rule may not contain nonadjacent strata. In this paper, I argue that Optimality Theory (and in its more recent versions like Correspondence Theory McCarthy & Prince 1995) provides the solution to this problem in terms of constraints on correspondence relations. More importantly the presented analysis offers a paradigm-based account for the two different strategies of anti- geminates (degemination and schwa-epenthesis) revealing that the otherwise unmotivated deletion and epenthesis within the serial operational theory is due to the uniformity in paradigms between 2. and 3. ps. sg. verbal form. The central role of the constraint Uniform Exponence within the paradigm-based theory is further supported by the contrast between the prefixes in- and un-.