The surface forms of NC
markers are largely parallel in different morphophonological environments:
The few differences in
surface form seem to be derivable by rule:
The verbal noun class markers seem
to be closest to the /CV-/ base form, since their shape reoccurs in all forms
with slight modifications.
The class markers on nouns are
combinations of the class marking proper and an 'initial vowel' (Canonici 1995,
Doke 1954 refers to it as a 'pre-prefix'), which does not seem to be more than a
mere reduplication of the class marker vowel,
which is presumably prosodically motivated. In her analysis of Swahili, Carstens
1993 subsumes the first 10 noun classes under 5 genders with singular/plural
Gender A: NC1
Gender B: NC3
Gender C: NC5
Gender D: NC7
Gender E: NC9
further assumes that noun roots are lexically marked for gender. The noun class
prefixes are analysed as heads of DP which provide gender specific number
marking: [DP NC [NP N ] ].
Noun class markers on adjectives
are similar combinations of a noun class marker and an initial vowel and seem to
be derivative of the nominal markers. The derivation seems to follow a
straightforward phonological rule that lowers initial [+ high] vowels in the
nominal prefix. Adjectival class markers only occur in their bound form, never
Demonstrative marking seems to be
derivative from the adjectival marking by addition of /l-/. The demonstrative
forms given in (6) typically occur as prefixed elements, but they can also stand
alone, i.e. they can have full pronominal functionality. Of course, they have a
similar emphatic interpretation as absolute pronouns, but this is already
expected from the demonstrative semantics.
The absolute pronouns mostly
substitute the vowel of the base form with /o/. They can surface with their
basic CV form or optionally take the suffix /-na/. Since there appear to be good
reasons to analyse the base form as D in the typical nominal context, one may be
tempted to conjecture that /-na/ represents the actual pronominal element which
may be replaced by pro when the base form surfaces alone. However, Zulu
overall seems to comply to a (violable?) constraint against prosodic words of
less than two syllables (except for the frequently occuring cha ‘no’
and na ‘question marker’, the occurrence of CV words seems to be
rather restricted). Since absolute pronouns are in their semantics emphatic,
they are presumably required to have prosodic word status, which is achieved by
adding a default suffix /-na/.
Prosodically deficient /-C-/ verbal roots in the imperative are rescued with a
similar strategy: since the imperative is formed without noun class or tense
inflection, the root would merely be expanded by the final vowel, yielding /Ca/.
Since verbs have to gain prosodic word status, /-na/ is affixed here as well,
resulting in /Cana/. The analysis of /-na/ as a prosodic filler is supported by
the fact that deficient imperative verbs can alternatively be fixed with a
default prefix /yi-/:
|[ lw ]imperative||>||yi -lw-a||or||lw-a-na|
|def root FV||root-FV-def|
that both /yi-/ and /-na/ are phonologically inconspicuous insofar as their
consonants require only minimal underlying specification. Both [coronal] and
[voice] are features that are supplied by default mechanisms in many languages.
If this analysis is extended to Zulu, /y/ would only require [liquid] and /n/
only [nasal] underlyingly. Both /yi-/ and /-na/ are thus prototypical default
syllables in their phonology, and the fact that they are freely interchangeable
suggests that they do not carry distinctive meaning either. It is thus unlikely
that /-na/ supplies the pronominal element in absolute pronouns, which seems to
force an analysis along the line of
[ [ NC ]D pro ]DP
are a few more environments in which NC morphology occurs (e.g. on numerals,
possessive). These manifestations are related to the base form in a similarly
systematic fashion as in the instances described above. I thus propose that we
are really not dealing with a whole collection of different NC markers which all
accidentally look alike (as the literature implicitly seems to suggest), but
rather with a single paradigm with different surface forms. These surface forms
are predictable from the morphophonological and/or syntactic environment. The
most contentious part of this claim is that subject and object class markers
belong to one and the same category, so I will devote a large part of this paper
to the discussion of verbal morphology.
systematicity of the NC morphology is not particular to Zulu, but quite likely
represents a general characteristic of the Bantu language family as a whole. In
spot-checks, I have verified that similar paradigms exist in
geographically close (Southern African) languages (Tsonga, Southern
Sotho, Northern Sotho, Venda, Tswana, Ndebele, Siswati), and also the
geographically distant Oshindonga (from Cameroon).
 'Absolute pronouns' are free morphemes and thus contrast with otherwise bound NC morphology.
 Thwala (1996: 25) assumes that the 'initial vowel' is an epenthetic underspecified V-slot, to which the features of the noun class marker vowel spread.
 cf. a parallel analysis for Siswati by Thwala (1996: 45)
 Radford (1988: 141) provides a similar line of argument for conflating lexical categories: 'Adjectives and Adverbs are in systematic complementary distribution, and should therefore be regarded as positional variants of each other, and assigned to the same overall category'.
 In fact, many languages with similarly rich morphology exhibit the same systematic relationship between subject and object agreement with independent pronouns, cf. Amele (Gum, Papua-New Guinea), Hausa (Chadic, Africa), Navajo (Athapaskan, USA).