Up to this point we have been assuming that the subject of the clause originates fairly low in the clause, inside the VP or a vP just above it. We have said this DP will move from its original position to the specifier of IP to get Case and thus avoid a Case Filter violation which would render the sentence ungrammatical. Two aspects of this analysis are in need of elaboration. First it must be accounted for that the subject’s original position is a Caseless one and second it must be established exactly where in the complex clause structure we have been arguing for the subject moves to and why this is a Case position.
We have said that the light verb which is responsible for assigning the Θ-role to the subject is responsible for assigning Case to the object. This seems to be the locus of Burzio’s generalisation that verbs which assign a subject Θ-role assign an accusative Case. Hence the object is in a Case marked position and need not move away in order to get Case. Consider the subject: why is it not in a Case position? Note that the verbal element above the subject, the tense in this case, does not assign any Θ-roles and hence its specifier position is empty at D-structure. Clearly this is unlike the light verb. We may propose therefore that tense is not an accusative Case assigning head. But why doesn’t the light verb assign Case to its subject? One might attempt to answer this by claiming that the light verb has to assign Case to the object and assuming that accusative Case can only be assigned to one place. While this seems reasonable, it doesn’t explain why the light verb taking an intransitive verb does not assign Case to its subject:
In this case, the light verb assigns agent Θ-role to its specifier and so should be capable of assigning Case and in fact the possible appearance of a cognate object seems to confirm this assumption. But when there is no object, the subject still undergoes the movement, suggesting that it does not get Case from the light verb.
There are a number of possible ways to account for these observations. The simplest is to assume that Case assignment is directional and that accusative Case in English is assigned to the right. Thus, the light verb will be able to assign Case to the object as the object appears to the right. However the light verb will not be able to assign Case to the subject as the subject is in the specifier position and specifiers are to the left of the head.
This is too simple, however, as it is not the case that a light verb can assign Case to any element on its right. Consider a more complicated case in which we have a verb with a clausal complement. The light verb of this verb will be to the left of the complement clause and hence to the left of the subject of that clause. But it cannot assign accusative Case to this subject, allowing it to stay inside its own subject position:
|(61)||a||*they think1-e [VP [IP - will Larry leave] t1]|
|b||they think1-e [VP [IP Larry2 will t2 leave] t1]|
The word order within the embedded clause shows that the subject does not get Case in its original position and that like any other subject, it has to move to get Case in the subject position of the IP. Considering where a light verb can assign Case to, i.e. the specifier of its own VP complement, and where it cannot assign Case to, i.e. the specifier of a VP inside another clause, it is obvious that there is some locality restriction on Case assignment in addition to the directional one. We are not yet in a position to be able to determine the exact nature of this locality condition and so for now we will just assume that a light verb can only Case mark elements within its own clause.
Having put this in place, we can see that the subject will not be able to be case marked in its original position as it is on the wrong side of the local light verb and too far from any other light verb that might have a Case to assign.
This must be a Case position as the sentence is grammatical with the subject sitting in it and therefore the Case Filter must be satisfied. If we assume that it is the inflection which is responsible for assigning the Case we account for why this is the landing site for this movement. Moreover, we also account for the difference between the subjects of finite and infinite clauses. Recall that while the subject of the finite clause has nominative Case, the typical Case for the subject of the non-finite clause is accusative:
|(63)||a||she can sing|
|b||for him to dance|
Presumably the agreement element is different in both these cases. For one thing, modal auxiliaries can appear in the agreement position of finite clauses, but not of non-finite clauses. Moreover, finite inflection selects for a finite tense headed phrase as its complement while the one in (63b) clearly selects for a non-finite complement. I will not have anything to say about the accusative Case of the subject of the non-finite clause at present, leaving this for future discussion. Instead, we will concentrate on the finite inflection, which assigns nominative to the subject.
The Case assigned by the finite agreement element is very different from that assigned to the object. Obviously the former is a nominative Case while the latter is accusative, but the differences extend further than this. For one thing assuming that it is the agreement element which is responsible for nominative Case, this Case must be assigned in a leftward direction:
Furthermore, this case is assigned by a functional element, something that is not involved in Θ-role assignment. For accusative Case the Case assigner necessarily Θ-marks its subject in order to be able to assign accusative Case. But the same is not true for the inflectional assigner of nominative Case.
There are two basic positions linguists take with respect to these observations. One is to assume that nominative and accusative Case are similar and the other assumes that they are fundamentally different. From the first perspective the challenge is to come up with restrictions which define the conditions under which Case can be assigned generally. For example, there are three positions to which we have seen a Case assigned: the specifier of the inflection; the specifier of the complement, and the complement position itself:
|(66)||an element governs everything within its own phrase, but not past a certain point|
The point of the ‘but’ clause in this definition is to impose locality on government. If a head governs everything inside its phrase, then it can govern quite a long way if its phrase happens to be a long one. Yet government is clearly a local relationship if it is restricted to the situations in (65) and there seems to be a point beyond which government cannot hold. How this point is identified and defined is a matter for discussion, with a number of positions possible. But one thing we can conclude at this point is that VP cannot be something that blocks government, otherwise the light verb would not be able to Case-mark the object in the specifier of VP.
The other point of view observes that it does not seem to be mere coincidence that the two instances of rightward Case assignment in (65) happen to involve accusative Case while the leftward Case assignment involves nominative. The assumption is then that there are two different processes at work here. For nominative the relevant relationship is supposed to be specifier–head agreement, something we mentioned in the previous chapter. The idea is that the finite inflectional element can assign nominative Case to whatever it agrees with, and this will be its specifier. For accusative Case however, government is the relevant relationship, though defined in a slightly different way as it is no longer required to extend to the subject. Informally, this version of government can be defined as:
|(67)||a head governs its sister and everything inside its sister, up to a point.|
Again, the ‘point’ imposes locality restrictions on the government relationships. From the present perspective there is very little between the two views that allow us to favour one or the other and therefore we will leave the matter unresolved at this point.