While Demers characterizes the root alternations as a process of deletion of unstressed schwa, it could also be analyzed as a process of metathesis. One option is to assume that metathesis is morphologically conditioned. However, Demers argues that the schwa deletion is phonologically predictable, so apparently metathesis would be then as well. Adapting Demer’s bisyllabic analysis, metathesis would occur to when a root was followed by two consonants or more consonants, as exemplified from the (b) and (c) paradigms in the examples section: Non-Metathesized (b) cə́ssən ‘I got hit’ tə́s ‘It got smashed’ Methathesized (c) csə́tŋs ‘Someone hit him’ tsə́tsən ‘I smashed it’ However, the forms in (a) are apparent exceptions to this rule: (a) (c) cə́stŋs ‘He’s getting hit’ csə́tŋs ‘Someone hit him’ On the surface, these two forms are identical except for the difference in whether the root has metathesized or not. Demers argues, however, that the forms in (a) have an underlying glottal stop preceding the second root consonant that is deleted. With an analysis involving metathesis, the forms in (a) could be treated as having the same rule of metathesis as above, except that the glottal stop would metathesize and then subsequently delete. This analysis would be difficult in a surface- constraint driven theory.