Phonologically Conditioned Allomorphy

Underlying allomorphy poses a challenge to theories of the morphology–phonology interface when the conditioning factors include phonological information. Three basic types of cases are identified with respect to the nature of allomorph selection: (i) phonologically fully optimizing selection; (ii) phonologically partially optimizing selection; and (iii) phonologically arbitrary selection. The approaches that cover these three cases fall into two groups, ones where the selection is left completely to the morphology, with Vocabulary Insertion (in Distributed Morphology terms), and ones where the selection is left to the phonology (within some version of Optimality Theory), for at least fully optimizing and partially optimizing patterns; argumentation is given in favour of the latter approach. A discussion is also included on when and how cases of alleged underlying allomorphy can be reduced to surface-only allomorphy and hence be assimilated to the regular phonological alternations of the language. For the cases discussed, it is shown that capturing the differences between the alternants in terms of a single representation that includes floating segments is feasible especially when one of the alternants surfaces with a subset of the segments of the other alternant, not with most instances of suppletion. One of the examples of suppletion is still reanalysed as a case of surface-only allomorphy with the inclusion of a restricted version of the constraint OCP.