This article steps back from the abstractness issue in order to examine it in a new light. An attempt is made to define the issue rather than offer a solution. Kiparsky's "alternation condition" is shown to comprise four distinct constraints, which are not equivalent, as most phonologists have assumed. Others have defended, implicitly or explicitly, a number of other possible conditions. A definition is given of each of these conditions and a number of celebrated "abstract" analyses are compared with respect to them. The absence of a single condition on abstract analyses forces us to reexamine the question of whether such a constraint is necessary and if so, how to construct a principled formulation.
Jensen, John T.
"How Abstract Is Abstract?,"
Colorado Research in Linguistics: Vol. 4.
Available at: https://scholar.colorado.edu/cril/vol4/iss1/2