The question (presumed to be legitimate) is asked: which of the three allomorphs of the Welsh article is 'basic' or 'underlying'? Its surface from may depend left or rightwards on its phonological context, in some cases despite the fact that the form of the context is dependent on the presence of the article. All the logically-possible ways of deriving the surface forms from each candidate underlier are examined. Of those ways which yield descriptively adequate solutions, it is found that each requires general mechanisms and at least on ad hoc stipulation or other theoretically undesirable feature. The issue of what constitutes a 'best' or 'correct' solution is then discussed, and it is argued that no optimal solution is achievable within the framework of classical generative phonology. A face-saver borrowed from autosegmental/3-D phonology is offered. The paper is by now something of a period piece (on the then current background see Coates 1982), but it has been extended to show how the insoluble dilemmas presented in the original can be elegantly sidestepped using more modern phonological techniques.
This paper is not available online