Typically, German inflectional affixes are monosyllabic. We can capture this fact by attaching the Affix node to the (mono-) Syllable node. This affix node then forms the root of a directed graph of suffix nodes. The simplest suffixes are those that have a constant phonological form, independent of phonological context or morphosyntactic variation. For such suffixes, all we need to specify are the phonemes that make up their peak, body and tail, as in the nodes defined below:
Affix:
<> == Syllable.
Suffix_i:
<> == Affix
<phn syl1 peak> == i:.
Suffix_es:
<> == Affix
<phn syl1 peak> == @
<phn syl1 body> == s.
Suffix_us:
<> == Suffix_es
<phn syl1 peak> == U.
Suffix_is:
<> == Suffix_es
<phn syl1 peak> == I.
Other suffixes exhibit morphosyntactic variance. Thus, for example, Suffix_um is realized as -ums on singular genitives and Suffix_er is realized as -ern on plural datives.
Suffix_um:
<> == Affix
<phn syl1 peak> == U
<phn syl1 body> == m
<phn syl1 tail sing gen> == s.
Because morphosyntactic attributes are visible in the phonology in the framework adopted here, it is completely straightforward to capture such variance. In particular, how both Suffix_e1 and Suffix_er inherit a component of their plural dative forms from Suffix_n (cf. note Zwicky 1985, 383):
Suffix_e1:
<> == Suffix_n
<phn syl1 peak> == @.
Suffix_er:
<> == Suffix_e1
<phn syl1 body> == r.
Suffix_n:
<> == Affix
<phn syl1 tail plur dat> == IF:<AND:<POLYSYLLABLE:<>
EQ:<n "Root:<phn syl1 tail->">>
THEN Null
ELSE n>.
The definition of Suffix_n provides /n/ as the tail consonant of the suffix in plural datives unless the root is a polysyllable that itself ends in /n/. We have not provided a definition of the Root node crucially invoked here and in the definitions of various other suffix nodes below. This is because Root is not a standard node name (like Noun, Affix, Umlaut, Tutor, etc.), but rather an alias for the DATR `query node' (commonly called Qnode) which is essentially an indexical expression (like here or now in English) that always refers to the node at which the current query originated. Thus, for example, if one is engaged in deriving an inflected form of Tutor, any occurrence of Root will be treated exactly as if it was a reference to Tutor.
As every undergraduate linguist eventually learns, the English -s suffix that is used to mark third person singular present tense on regular verbs and plural on regular nouns appears in three different phonological guises depending on the phonological context provided by the end of the root to which it is attached. Similarly, German has a suffix (Suffix_S, below) that requires the insertion of schwa if the root ends in a sibilant. It also has a couple of suffixes whose realization is determined by whether the final syllable of the root has a schwa peak. Zwicky suggests that the first of these, Suffix_e2, needs to check whether the root ends ``in schwa plus a sonorant'' (1992, 352) but the reference to sonorant appears to be redundant since every German noun in the relevant declension that has schwa as the peak of the final syllable of its root has a sonorant following schwa. The only nouns listed in CELEX as having a final syllable with a schwa followed by a non-sonorant consonant are five English loan words ending in /@t/ (e.g., Kricket) which all take the -s plural suffix, and the archaic Sammet, which is given as Samt (`velvet') in most modern dictionaries and, as a mass noun, arguably has no plural forms in any case. In connection with the second, Suffix_en1, Köpcke notes that ``the schwa will be deleted in exactly those cases where the stem of a noun already ends in schwa or in schwa + consonant'' (1988, 307) (whereas Kloeke's statement of the rule is stress dependent (1982, 174)).
Suffix_S:
<> == Affix
<phn syl1 tail> == s
<phn syl1 peak> == IF:<SIBILANT:<"Root:<phn syl1 tail->">
THEN @
ELSE Null>.
Suffix_e2:
<> == Affix
<phn syl1 rhyme> == IF:<EQ:<@ "Root:<phn syl1 peak->">
THEN Null
ELSE "Suffix_e1:<phn syl1 rhyme>">.
Suffix_en1:
<> == Affix
<phn syl1 body> == n
<phn syl1 peak> == IF:<EQ:<@ "Root:<phn syl1 peak->">
THEN Null
ELSE @>.
Suffix_e2 differs from Suffix_S and Suffix_en1 in that one of its two allomorphs is the empty sequence. It would be technically possible to treat the allomorphy of Suffix_e2 in the same kind of way that Corbett & Fraser employ in their analysis of Russian genitive plurals (1993, 125; Fraser & Corbett 1995, 127). But their monadic morpheme style of analysis would not readily extend to the peak alternation found in Suffix_S and Suffix_en1. By contrast, the style of analysis we adopt here treats all three cases of phonologically conditioned allomorphy in an identical manner.
The most interesting of the German suffixes are those whose form is partly determined by an inherent morphosyntactic property of the root - specifically, the gender of the root. Thus, for example, there is a suffix (Suffix_s, below) that marks singular genitives and plurals. When it is realized this suffix is realized phonologically by Suffix_S considered above (thus a schwa gets inserted after sibilants) but it has no phonological realization as a singular genitive on feminine nouns (readers already familiar with the DATR language will note that our definition of the Suffix_s node invokes Suffix_S as a quoted node). There are also a couple of other suffixes that inherit this sensitivity to the gender of the root (cf. Carstairs-McCarthy 1992, 231).
Suffix_s:
<> == Affix
<phn syl1 peak> == <phn syl1 tail>
<phn syl1 tail sing gen> == IF:<EQ:<femn "Root:<syn gender>">
THEN Null
ELSE "Suffix_S">
<phn syl1 tail plur> == "Suffix_S".
Suffix_a:
<> == Suffix_s
<phn syl1 peak> == a:.
Suffix_en2:
<> == Suffix_en1
<phn syl1 tail> == Suffix_s.
Suffix_o:
<> == Suffix_s
<phn syl1 peak> == o:.
