This study, which draws on substantial knowledge of formal models adopting the hypotheses of generative Grammar, aims to achieve the following goals: to provide a critical account of the analysis based on bare radicals and affixation, by showing its descriptive limitations; to argue that the triliteral radical (mujarrad) is characterized by a morphological complexity, which can be accounted for by the auto-segmental approach based on the morpheme; to expose a formal theory of derivation allowing a distinction between lexical derivation and inflectional derivation on categories independently motivated. The first type does not need to extract a derived item from the radical; rather, it is a relation between representations. The second type (inflectional derivation) is a process that involves deriving an inflected from another inflected one by apophonic ablaut. To raise some issues that concern the nature of prosodic template and the relationship between the morphological level and the phonological level.