Accounting for the phonetics of German r without processes Adrian P. Simpson Institut f ¨ ur Phonetik und digitale Sprachverarbeitung der Christian-Albrechts-Universit¨ at zu Kiel 1 Introduction A rich variety of phonetic patterns are associated with German . Leaving aside interdialec- tal and interstylistic differences, these patterns are determined by the place in the syllabic and rhythmic structure. For a typical North German speaker the correlates of can range from a voiceless uvular fricative in the initial consonant cluster of a word such as trat (“stepped”) to apparent absence of anything at all following the open vowel in a word such as Bart (“beard”). In particular, the patterns observed for in postvocalic position are particularly rich. In combi- nation with short quantity vowels (e.g., wird “becomes, will”, Wurst “sausage”, Korb “basket”, Erna proper name) we can find phonetically long monophthongs, whose quality is opener and more central than the quality of their -less congeners. In combination with non-open long quantity vowels (e.g., ihr “her”, Uhr “clock”, wer “who”) we can find phonetically long diph- thongs which end centrally somewhere between and . Both in extensive descriptive surveys (Ulbrich 1972; Graf and Meißner 1996) as well as phonological analyses (Hall 1993), these patterns are accounted for in terms of generative pro- cesses. The names of the descriptive categories ‘vokalisiert’ and ‘elidiert’ used in Ulbrich (1972) and Graf and Meißner (1996) are process-oriented. In Hall’s lexical phonological ac- count all allophonic variants are derived from the consonantal specification of a voiced uvular trill employing rules such as ‘ -vocalisation’ (Hall 1993: 88). There would appear to be both formal linguistic as well as phonetic grounds why a gener- ative phonological account of the phonetic patterns associated with German is inappropriate. In non-linear approaches to accounting for phonetic patterns such as Firthian phonology (Firth 1948) and more recently in related declarative frameworks (Coleman 1994; Local and Ogden 1997) as well as articulatory phonology (Browman and Goldstein 1989), it has been succes- sively shown that differences in the phonetic appearance of the same phonological objects can be accounted for using a combination of rich phonological structure and non-linear phonetic exponency, avoiding the need for the destructive might of rewrite rules. From a phonetic point of view a different interpretation of the cases of elision reported in Ulbrich (1972) and Graf and Meißner (1996) suggest that the phonetic correlates of the phonological object are not absent. Using a declarative phonological analysis and non-linear phonetic exponency this paper demonstrates that the complex set of consonantal and vocalic patterns associated with can be reduced to two phonetic exponency statements, without using processes such as vocalization and elision. One exponency statement describes the consonantal correlates associated with at syllable onset. The second exponency statement describes the vocalic correlates of at coda. The monophthongal vocalic qualities found in connection with short quantity vowels as opposed to diphthongal patterns found with long quantity vowels are seen as the product of differences in co-temporality of the correlates of and those of the vowel. However, these differences are not treated as being specific to , but rather as part of the more general observation that consonantal strictures following short vowels are often longer than those following long vowels. Verification is provided using synthesis examples produced by a computer implementation of the phonological abstractions and their phonetic exponents. 91