In many analyses of Bantu reduplication, one puzzling aspect is the absence of correspondence between the final vowel (FV) of the reduplicant (RED) and the FV of the base. In Zulu, the default FV for a verb is the -a found throughout Bantu, but certain forms, such as the recent past and subjunctive, take an FV of -e, and a final -i is correlated with negation, all of which are barred from appearing on RED. This systematic mismatch between the RED and base is difficult to account for within Optimal Theory, where it is necessary to formulate constraints that penalize including ‚Äúinflectional‚Äù material in RED, or have different rankings for RED-Base Faith constraints for root material (high-ranked) vs. non-root material (low-ranked). In Distributed Morphology, the absence of correspondence between the FV of RED and the FV of the base follows straightforwardly from the nature of the derivation, as the FV -a is taken to be an intermediate spell-out of the v (verbalizing) head that attaches to an acategorical root. In the RED+base verb complex as a whole, this -a gets overwritten as the verb moves up to higher syntactic projections (such as mood, aspect, and negation), but at this point, RED is no longer accessible as a privileged constituent, and its -a FV cannot be targeted.
"Explaining the Final Vowel Mismatch in Zulu Reduplication,"
University of Pennsylvania Working Papers in Linguistics:
1, Article 6.
Available at: http://repository.upenn.edu/pwpl/vol19/iss1/6