This paper deals with the Affix Ordering Generalisation (AOG), an integral part of stratal frameworks, which postulates that all stem-level affixes are positioned closer to the root than all word-level affixes. Tigrinya, a Semitic language spoken in Eritrea and Ethiopia, seems to exhibit a configuration where a word-level affix is closer to the root that a stem-level affix; a violation of the AOG. I argue that this AOG violation is illusory and show that there is no phonological evidence to assume that stem-level phonology is triggered after word-level phonology. There is therefore no reason abandon the AOG or Strict Layering of strata. I propose that the stem-level suffix that follows the word-level suffix changes its stratal affiliation to word-level in this position. This paper provides a detailed, novel analysis of this stratal change using global, morphological optimisation that interacts with weak and strong stratal indices modelled using gradient symbolic representations to determine stratal domains. The analysis remains restrictive by predicting possible and impossible patterns of stratal changes.
Katie McCann (Fri,) studied this question.