/i/. The fact that there are no surface alternations which call for a rule changing
fa can appear in phonological representations after the application · 2013
An adequate solution to this problem will undoubtedly require an appeal to a more structured representation of the segment. For example, if we follow Clements (1985) in postulating a structure for the segment in which features branch from a root node which is distinct from the timing slots of the CV tier, then we can analyze E-Mutation as a process by which the features of fy/ spread onto the root node of a preceding /e/ at the same time as the associated V-slot is deleted. I will not attempt to settle this matter here, however, and note only that a feature-spreading account of this type would be consistent with the two claims stated above.