At the end of my last post, I asked a question about the structure of verbs, in relation to the events they portray. More specifically, what is the difference between internal and external arguments. Do our grammatical notions such as Subject and Object relate to our notions of Internal and External Arguments? The sentence Jeff ate the fish means something distinct from The fish ate Jeff, and so it seems that subjects and objects correspond to internal and external arguments. Consider the following:
(18a) The door opened
(18b) The key opened the door
(18c) Jeff opened the door with the key
The fact that you can get multiple sentences from the same event structure seems to suggest that event structure is a doomed concept. After all, the door’s relationship to the verb is the same in (18a) as it is in (18c), and so the fact that it can occupy a different position is problematic. An even more troublesome comparison comes from the verb meet.
(19a) Sue met *(Jeff)
(19b) Sue and Jeff met.
(19c) Sue and Jeff met each other.
Some linguistic literature will doubtless tell you that (18b) is simply a shortened version of (18c), and thus we do not need any appeal to the conceptual participants of a verb to determine its subcategorization, and nor can we use them. However, I see (18b) as posing no violation to the idea of conceptual participants. It has two arguments, even if they are both locked into the agent role. I would caution against treating (18d) John and Sue ate similarly. I do not mean to propose that all conjoined subjects are acting on each other in a similar way. But whereas (18d) John and Sue ate is equivalent to the combination of John ate and Sue ate, (18b) John and Sue met is not equivalent to the combination of John met and Sue met. As we can see with prepositions like between (being between two things is not equivalent to being between one and between the other), conjunction does not always function to merge two arguments into one.
What differentiates a verb like eat from a verb like meet? Why can the object of meet move up to the subject, but not the object of eat? Because those objects play different conceptual roles in the event, or thematic roles, or theta roles for short(perhaps because somebody thought the Greek sounded sciencey). The object of eat is a food, and has no agency. It is the thing that undergoes the action and is affected by it, and gets the patient theta role. On the other hand, the object of have would get the theme role in I have a car. Meet, on the other hand, seems a different case, one with two agents. And this, I propose, is why it allows the objects movement up to the subject position.
Looking back up at sentences (18a-c), we can see that door can only be the subject so long as key or John are not present. Something about these theta roles pushes door out of the sentence role. From this data (and many more), we get the idea of a thematic hierarchy, the idea that certain theta roles are structurally ‘higher’ than others. At the top would be the agent, below it, the instrument, and below it, the Patient and Theme. Even with a weaker hypothesis (perhaps that only two nouns with the same theta role can be conjoined) explains the data in (18a-c). Furthermore, it provides more reason that meet differs from eat. The object of eat could never move up to occupy the same spot as the subject, without violating theta role principles, whereas meet, involving two agents, could.
Now we have a theory which explains subcategorizations, one which appeals to event structure by the descriptive terms provided by the theta role. Chomskian linguistics (for the most part) has accepted theta roles as primitives of the grammar. I accept theta roles as descriptive stand-ins for the event structure of the predicate, but not as things in and of themselves (I do not want to deny their existence as primitives, just to remain agnostic for now).
This theory explains why the following sentences are ungrammatical
(20a) *Sven put the microwave
(20b) *I threw
(20c) *Aeneas defeated
Each sentence is missing one of the nouns which is to receive a theta role (that is to say that the event structure is incomplete). The following sentences have the opposite problem: they each have an unsaturated noun, one which has no theta role assigned to it. This can be solved by the addition of a preposition, which ill assign a theta role to the unsaturated nouns.
(21a) *Sven put the microwave in the toilet care
(22a) Sven put the microwave in the toilet with care
(21b) *I threw the ball the goalpost
(22b) I threw the ball through the goalpost
(21c) *Aeneas defeated Turnus Italy
(22c) Aeneas defeated Turnus in Italy
(21d) *I ate food knife
(22d) I ate food with a knife
(21e) *I slept the bed
(22e) I slept in the bed
All this talk of theta roles has rested on an implicit principle, albeit an intuitive one. A noun with no theta role is one which cannot fit into the event structure, and since sentences should fit the event structure, all nouns should receive a theta role. As well, each noun should only receive one theta role (because otherwise, there is no reason that (20c), for example, would be ungrammatical). These combine to make an identity rule, a one-to-one mapping between nouns and theta roles. This principle is called The Theta Criterion.
I will close with a quick point about theta roles: they are not limited to verbs: adjectives, nouns and prepositions can assign them as well, something assumed by sentences (22 a-e). However, this diversion aside, I will return to phrase structure rules.