The idea that constraints on well-formedness play a role in determining phonological alternations, which dates back at least to Kisseberth’s (1970) pioneering work, has by now achieved almost universal acceptance. A tacit assumption of this program, largely unquestioned even in recent research, is the notion that valid constraints must state true generalizations about surface structure or some other level of phonological representation. Anything different would seem antithetical to the very idea of a well-formedness constraint.
No takes yet. Share an insight, caveat, or question.
John J. McCarthy (1993) studied this question.