A critical difference between Davis' representation and Jackendoff's representation is that Davis does not make an explicit distinction between events and states. In fact, as is clear from examination of which of Jackendoff's functions have no good correlates in Davis' hierarchy, states are generally neglected in his representation. This event/state distinction is left implicit in the entailments associated with proto-roles in the semantic relations. However, it seems critical to explicitly represent this difference in the ontology due to the linguistic influence of this distinction.
In Section 2.2.1, I made reference to Jackendoff's arguments in favor of the state/event distinction, in particular the constrast between how states and events can be referred to in a discourse. There are in addition further semantic and syntactic distinctions between them, and they have different effects on the temporal relations which can be established in a narrative discourse (Kamp and Reyle 1993). For instance, when a state is expressed with the present simple tense in English, the sentence is interpreted as a referring to a situation in which the state holds, while events expressed in this tense receive a habitual interpretation or have a `newspaper headline' quality (Jackendoff 1983). Consider the data in dav6.
John loves Mary. (state) John walks. (activity, habitual) John reaches the summit. (achievement, habitual) John walks to the station. (accomplishment, habitual)
States also differ from events in that they cannot normally occur in the progressive form in English, as show in dav7.
*John is loving Mary. John is walking. John is reaching the summit. John is walking to the station.
partee:84 and hinrichs:86 provide proposals for updating time in narrative discourse. Roughly put, events move narrative time forward and states do not. This is shown in dav8, where the state of the music being very loud in the pub is interpreted (by default) as being true before, during, and after the entering event.
Peter entered the pub. The music was very loud.
There are, however, instances in which a state is interpreted as moving time forward, and/or as having an explicit starting point. Consider the discourse in dav9, for example.
Mary switched off the light. The room was pitch black.
These cases would seem to require the reinterpretation of a state as a bounded event, but schilder:97 has argued that instead the interpretation is the result of discourse processing which forces the introduction of a boundary for the interval characterised by the state. The state does not shift aspectual class, its associated interval merely acquires a starting point. This analysis follows from the fact that the conceptualisation of the state and its aspectual properties do not change when a discourse adds a boundary. So for my purposes these cases do not provide a counter-example to the need for a lexical distinction between states and events.
These examples show clearly that the state/event distinction has implications for linguistic processing at the syntactic, semantic, and discourse levels. This evidence points to the need for explicit representation of the distinction. I will therefore incorporate this distinction into the semantic relations hierarchy.