Despite the high productivity in English of denominal verb formation by conversion, a constraint on possible interpretations is identified: intransitive denominal conversion verbs are infelicitous where the source nominal is intended to be an incremental theme (#apple 'eat apple'), a patient (#shirt 'wear shirt'), or the holder of a result state (#window 'open window'). This little-studied constraint is compared to a well-known constraint on denominal conversion verbs in which the source nominal is intended as an agent or cause (#it cowed a calf 'the cow had a calf', Hale and Keyser, 1993), accounted for in previous literature by a prohibition on incorporation in sublexical syntax from specifier position. To account for both constraints it is proposed that nominal roots in English denominal conversion verbs do not originate in argument position, but adjoin directly to subeventual heads, and are interpreted as predicates of events. A prediction is made that English denominal conversion verbs should exhibit the same argument and event structure possibilities as verbs built from verbal roots, which is confirmed by a corpus study of 250 verbs from each class. Nominal roots, however, have a significantly higher likelihood of being lexicalized as transitive verbs than verbal roots do. A study is further undertaken of the semantics of English denominal conversion verbs, resulting in a reformulation of Kiparsky's (1997) Canonical Use Constraint; specifically, it is proposed that a canonical event type associated with a nominal root can be a default interpretation for a denominal conversion verb if and only if the argument structure template of the canonical event type can be unified with the arguments of the denominal verb. The role of pragmatics in the interpretion of English denominal conversion verbs is considered, and it is suggested that apparent exceptions to the constraint on themes, patients, and result states can be attributed to re-interpretation of the verb as an agent-oriented manner verb.