a1 University of Arizona
We have presented two different analyses of stress in Macedonian and Polish: one in terms of CGs and the other in terms of QS/ROB footing. Further, we showed how there were a number of problems with the CG account. We now consider those problems again in light of the latter analyses above.
The first problem was that the emergence of unmarked stress under suffixation did not find a uniform explanation in Macedonian and Polish. This is not true of the revised analyses. Both are couched in terms of accent and the reemergence of regular stress in both languages follows as a consequence of distribution of lexical accent. In fact, aside from the contribution of secondary stresses in Polish and their absence in Macedonian, the analyses differ minimally. Footing in Macedonian is QS; footing in Polish is ROB.
A second problem for the CG was that the reemergence of unmarked stress in Polish did not find a uniform explanation. Some cases were handled with lexical extrametricality, while others were handled with the special extrametricality rule.(44)
This is also not a problem for the revised analysis. Class I and class II now differ minimally in the placement of accent.
The third problem discussed above was the ad hoc character of the special extrametricality rule (44). This too is no longer a problem since this rule is no longer a part of the analysis. In fact, perhaps we can make the strong claim that rules of this type are to.be excluded in principle. An even stronger claim which might be supportable, though it would require we reconsider a number of analyses, is that lexical accent can never be marked with extrametricality. We leave this latter hypothesis open.
Summarising, the best analysis of Macedonian and Polish requires a theory that makes use of two different kinds of footing – quantity-sensitive (QS) and revised obligatory-branching (ROB) – and that marks accent in terms of a diacritic feature rather than in terms of grid marks without constituency. Only by adopting such a theory can we capture in a satisfying fashion the generalisation exhibited in Macedonian and Polish, that regular stress reemerges when a word with exceptional stress undergoes sufficient affixation.
* Thanks to D. Archangeli, S. Franks, D. Pulleyblank, three anonymous reviewers, the editors, and especially M. Mikoś for useful feedback. Thanks also to F. Eckman for institutional support during the summer of 1988. All errors of data and analysis are the author's.