Introduction

Serial derivations have been a central idea in the theory of generative phonology throughout its history, but scant attention has been paid to a key question: is any serial derivation possible in human languages? More precisely, can any independently licit rule coexist with any other licit rule, and can the rules apply in any order? The rule coexistence question has, to my knowledge, never been raised in the literature. The rule ordering question was investigated intensively during the early 1970s (see Iverson 1995 for a review), but often just a pair of rules was studied in isolation from the broader derivational context. Since about 1975, research in generative phonology has mostly dealt with the form of rules and the nature of representations – subjects that are interesting in themselves but do not help advance the theory of derivations.

A rare contribution to this neglected topic is Pullum's (1976) study of the “Duke-of-York gambit.” Duke-of-York (DY) derivations have the general form A → B → A, where underlying A passes through a B stage before returning to surface A again. For example, in some analyses of r-dropping and r-intrusion in various English dialects, final r is first deleted and then reinserted before a vowel: Homer is → Hom[ǝ] is → Homer is (cf. Hom[ǝ] saw).