This paper expands on Lieber’s (1980) claim that certain instances of allomorphy must be determined before word-formation processes apply. Her notion of `morpholexical rule’ is reinterpreted as a structure-building lexical-redundancy rule. Such rules are triggered by morpholexical or morphosyntactic features assigned to lexical entries. The morpholexical rules specify phonological structure which is predictable, either from general principles or from morpholexical class membership. It is argued that in certain cases the lexical representations so constructed resemble the structures of nonconcatenative morphologies, in that they are multiplanar, with separate tiers corresponding to different morphosyntactic or morphophonological categories. Linguistic and psycholinguistic evidence is presented in support of these claims.