Alternative noun declension hierarchies

  In this section we discuss how our noun declension hierarchy compares to other inheritance-based accounts of German noun inflection. In the figures, we give the noun class labels together with the corresponding class in our analysis. Where the classes are equivalent, this is indicated by the `=' symbol and where a class is a proper subset of one of our classes, this is indicated by the `<' symbol.

Reinhard (1990) gives the noun inflection hierarchy shown below but her analysis is primarily concerned with umlaut, and so does not attempt a comprehensive account of those classes where umlaut is irrelevant.

  Reinhard's classes E_marked, E, ER and EN broadly match our classes A, B, C and D respectively, although she subdivides A and B according to whether the -e suffix is realised or not. We avoid the need for this subdivision by allowing the realization of the suffix to depend on the phonological properties of the root. Gibbon (1992) uses the same hierarchy as Reinhard:

  Bleiching (1994) provides an extensive hierarchy which appears, at first glance, to be very different from ours:

  However, closer inspection reveals that most of the differences follow from the fact that we postulate suffixes whose form is determined by the gender or by phonological properties of the root. Thus, of the classes she gives, VII, VIII, IX, and X are distinct from I_u, II_u, III and V respectively only in the gender (the former classes representing feminine nouns which have zero suffix in the singular genitive). Classes II, II_u and IX* are distinct from I, I_u and IX only in the presence of a final schwa syllable in the root. Class I* is distinct from II only in phonological form, being disyllabic with a non-schwa (heavy) final syllable. Classes IV and IV_u are not distinct in our account, since IV consists only of roots with front vowels, which are not affected by the umlaut function. Finally, the distinction between classes III and III_@ and classes IX and IX_@ appears to be motivated by compounding facts.

Allowing for these distinctions, her classes map to ours in the following way:

V $\rightarrow$ Noun
I $\rightarrow$ A
I_u $\rightarrow$ B
IV $\rightarrow$ C
III $\rightarrow$ D
VI_@ $\rightarrow$ J
VI $\rightarrow$ K
III* $\rightarrow$ L

The main difference in the hierarchical organisation of these two sets of classes stems from our assumption (defended in the section on declensions) that the class that defines the -s plural subclass (our Noun class) stands at the root of the tree and defines the suffixation of genitive singulars and dative plurals.

Bleiching et al. (1996) give a hierarchy, shown below, which is essentially the same as that given in Bleiching (1994), but with rather fewer subclasses. However, two new subclasses are distinguished. Without seeing their full definitions, we are not able to determine what motivates them. One subclass has Treffen as an instance (which appears to differ from Fehler only in its gender), and the other has Matrose as an instance (which appears to differ from Mensch only in being disyllabic).

  The declension hierarchy presented by Kilbury (1995) comes closest to the one we propose and ours largely subsumes his:

  The main difference between Kilbury's hierarchy and ours is that his umlaut classes inherit the umlaut and override the suffix, whereas ours inherit the suffix and override the umlaut.

All of the hierarchies discussed above cover our classes A, B, C and D, but none cover all of the minor subclasses which we include. Differences in the hierarchies can be accounted for largely by reference to different assumptions about the roles played by gender and phonology and by whether or not -s is taken as the default plural for German. We have shown that by allowing gender and phonology to determine the inflectional realisation of noun forms, we can reduce the declensional classes to just those required for defining the plural alternations. As noted earlier, this yields PEP-compliance.