Phonology

 

We assume that each lexeme has a phonological structure consisting of a sequence of syllables, each syllable having a structure.

As in Cahill (1990b) and Bleiching (1992), we define the tree-structure of the syllable by means of simple context-free phrase structure rules:

root --> syllable
syllable --> onset rhyme
rhyme --> peak coda

A root thus consists by default of a single syllable. A syllable consists of an onset and a rhyme. And a rhyme consists of a peak and a coda. We can use DATR to express these context-free phrase structure rules as follows:

Syllable:
    <phn root>  == "<phn syll>"
    <phn syll>  == "<phn onset>"  "<phn rhyme>"
    <phn rhyme> == "<phn peak>"   "<phn coda>"
    <> == Null.
Null:
    <> == .

Here the path-initial attribute phn contrasts with the path-initial mor (see section 4). These serve to partition the feature space into phonological and morphological domains, respectively. We have also made one important substantive addition above: the maximally unspecified path (<>) is defined by reference to Null which always returns the empty sequence as its value. A <phn onset>, <phn peak> or <phn coda> which is left undefined at lower levels of the hierarchy will, as a consequence, end up as null.

Given this general definition for Syllable, we can now use it to define particular concrete syllables. Here, for example, is a node definition for the monosyllabic -en suffix, realised as @ n:

Suffix_en:
    <> == Syllable
    <phn peak> == @
    <phn coda> == n.

Our interest in phonology in the present paper is restricted to those aspects of phonological structure that are relevant to the description of German inflection. That includes syllable structure as characterised above but does not include any structure above the level of the syllable, such as metrical structure. Even the simple notion of primary lexical stress, which is arguably the only aspect of suprasyllabic phonology that is required for morphological description, is unnecessary for the current fragment. We also restrict ourselves to a segmental representation of the phonology. As one of us has shown in earlier work (Cahill 1993), the step from representing structures with segments to representing the same structures with full feature sets at each point in the tree is relatively simple. We have not taken that step here because it would not add anything to the present analysis but it would make our DATR code much harder to read. A featural level is helpful in defining other alternations in German, e.g. umlaut, but none of the alternations that occur in the present fragment benefit from such an analysis. For the same reason, we omit discussion of purely phonological matters such as final devoicing (Wiese 1995, 00-00).

We assume below that a fully inflected form is simply a string of phonological segments. For our present purposes, there is no need for the implicit tree structure to be made manifest in the output.

In our analysis, a root is a sequence of syllables. Exactly one of these syllables is the focussed syllable. The focussed syllable of words whose inflection involves a stem alternation will normally be the syllable in which the alternation occurs. This will usually coincide with the stressed syllable, which will be the focussed one in words where no stem alternation occurs. Again, these are default assumptions which do not have to hold for all words. Typically, the segmental components of the focussed syllable get explicitly defined at the (phonological) entry for the lexeme in question. There are few polysyllabic roots in the fragment of German presented in this paper and so, for simplicity of exposition, we will specify unfocussed syllables just as segment strings rather than by reference to their internal structure. Indeed, German roots are rarely more than three or four syllables long and the vast majority of underived roots are monosyllabic. Since most of the roots considered below are monosyllabic, patterns do not really emerge, but, in languages that are predominantly suffixing, as German is, the final syllable of the root is usually the focussed syllable. However, in German there is significant variation, with the focussed syllable occurring as both initial and final in disyllabic roots. We can provide for those disyllabic roots where the focus is on the initial syllable as follows:

Disyllable:
    <> == Syllable
    <phn root> == "<phn syll>" "<phn final>".

Given this abstract definition, a disyllabic root such as irgend can be specified in terms of the individual components of its initial syllable (onset, peak, coda) and, given the expositional simplification noted above, by the segment string that corresponds to its final syllable:

Irgend:
    <> == Disyllable
    <phn peak>  == i:
    <phn coda>  == r
    <phn final> == g @ n t.

From the definitions given above, we can now infer that:

Irgend:
    <phn root> = i: r g @ n t.

Thus, the phonological root emerges as the result of concatenating the <phn syll> value, which is determined by the structure definition at the Syllable node, and the <phn final> value, which is given explicitly at the node as the string g @ n t. The structure of the <phn syll> is spelt out by concatenating the values for the onset (which here defaults to the empty string), the peak (i:) and the coda (r).

Crucially for our approach, the realisation of any component of the phonological structure can be determined by morphosyntactic features. This can be illustrated by the core component of our treatment of the inflectional phonology of the definite article and relative pronoun forms:

Der:
    <> == Strong_2
    <phn onset> == d
    <phn peak>  == e
    <phn peak nom sing femn> == i:
    <phn peak nom sing neut> == a
    <phn peak gen sing masc> == E.

This says that the phonology of Der is determined (by default) by inheritance from the Strong_2 declension; that Der's onset is d; that its (default) peak is e; but that the nominative singular feminine and neuter peaks are i: and a respectively and the genitive singular masculine peak is E.

The definitions given earlier in this section thus provide the phonological skeleton for our analysis of inflection. The flesh on these bones can be filled in at any point in the hierarchy, and can be shaped by information from any domain of lexical representation, be it syntactic, morphological or phonological. As we shall see in the definitions below, the determining factors in the present analysis are always morphosyntactic attributes such as gender, case and number, but examples of phonological determinants in German occur in verb and noun inflection, and examples of the way in which they can be dealt with in the present framework can be found in Cahill (1993) in relation to English morphology.