<< >> TOC
4. Phonological evidence for the featural characterization
Khumalo (1987) makes two interesting observations: Zulu exhibits limited forms of both consonant and vowel harmony, and depressor consonants participate in both. For previous accounts that identified depressor consonants either via a tonal specification for L (cf. Laughren 1981) or via an abstract laryngeal feature [depressed] (cf. Khumalo 1987), the different aspects of depressor behavior remain unintegrated. Given the above feature specifications, however, the interactions between depressor consonants and particular autosegments are not only explainable, but are even expected.
Following Khumalo (1987: 74), Zulu has 7 surface vowels: [ i, e (tense & lax), u, o (tense & lax), a ]. The tense/lax pairs for [e] and [o] are in allophonic variation: tense [e] and [o] occur only when the following syllable contains a high vowel, i.e. they are conditioned surface variants.
Khumalo (1987: 75) provides the following examples with several alternative pronunciations (those marked with a question mark sound strange to Khumalo, the check mark indicates acceptable forms, while the ones marked with an arrow are claimed to be most common):
Since, according to Archangeli and Pulleyblank (1994: 176), [high] is predictable from underlying [ATR], the following feature matrix suffices for the characterization of Zulu vowels:
Given these underlying specifications, the data in (16) can be explained as ATR harmony, where the autosegment ATR spreads from right to left. Speakers may apparently differ with respect to whether or not ATR spreading applies iteratively, which would explain the variation between (16 b1&2, c1&2, d1&2). In the absence of an ATR trigger, lax e/o surface unchanged (16a).
As Khumalo (1987: 76) indicates, depressor consonants block ATR spreading to vowels in their domain:
It thus seems that, at least for some speakers, depressor consonants involve an articulatory gesture that is phonologically describable as [RTR] (or [-ATR], depending on the theory), which would explain an interaction with ATR harmony. For present purposes, it seems sufficient to point out that the articulatory gestures associated with depressor consonants are not limited to F0 control, but have predictable consequences in the segmental phonology.
Notice that even though depressors and L tones correlate with [RTR]/[-ATR], H tones do not correlate with [ATR], i.e. [-ATR] vowels are compatible with H tone:
The fact that H tone assignment does not cause a [-ATR] vowel to be raised suggests that F0 control is independent of [RTR/ATR] per se. This provides phonological evidence for the assumption of an additional laryngeal feature (i.e. [lax vf]) as the trigger of tonal depression and against any assumptions of an exclusively tonal specification on depressor consonants. Instead, segmental features are the underlying cause for tonal effects of these consonants, and interaction with other segmental features thus does not come as a surprise.1
It may be possible that the consonantal features [tense/lax vf] stipulated here are, in fact, the same as the traditional vowel features tense and lax, i.e. the old question of whether and how tongue root phenomena are related to tense/lax is reopened. While it may well be the case that the tense/lax distinction is overshadowed by height/backness settings in Germanic languages on the one hand and tongue root settings in e.g. Igbo and Akan on the other hand (cf. Ladefoged/Maddieson 1996: 302-306), Zulu seems to suggest that there are languages where tense/lax is not only an active but possibly even a dominating feature. As a laryngeal setting, it is, of course, independent of vowel/consonant distinctions.
In a different place, I also hope to show that so-called 'erratic depression' is not a strong argument against the assumption of a natural class of depressor consonants. In short, most cases of 'erratic depression' can either be dismissed as lexical exceptions or derived from particular morphosyntactic constellations.