5.2.4.1 Evidence from passives
There is some evidence that the correct structure should be something like (74) however. This comes from the fact that transitive verbs can undergo passivisation. When a verb is passivised, it loses its agent and the object becomes the subject:
(77) | a | Mark made the bed | – | the bed was made |
b | Harry hit Bill | – | Bill was hit | |
c | Richard write the letter | – | the letter was written |
Even under a less strict view of the UTAH than we are attempting to keep to here, one would like to assume that the object is generated in the same place in active sentences and their passive counterparts. This has been the assumption since the beginning of generative grammar in the 1950s. Thus, the object is generated in object position in the passive, but moves to the subject position. Presumably the only reason it would do this is to get Case. Thus while in the active structure the object gets Case in object position, this ceases to be a Case position in the passive and hence not only does a passive verb lose its subject, it also loses the accusative Case assigned to its object. Again, these are fairly standard assumptions about the analysis of the passive which have been proposed since the 1980s.
Interestingly, the passive is a construction which conforms to Burzio’s generalisation: the verb stops assigning a Θ-role to its subject and loses the ability to assign accusative to its object. But Burzio’s generalisation is a description of a state of affairs, it is not an explanation of that state of affairs. What we need is something that links the two properties. In previous examples we have seen a way to link the Θ-role assignment to the subject and the accusative Case assignment to the object: through the light verb which is assumed to do both:
(78) |
If the light verb ceased to be there, both the agent Θ-role and the accusative Case would be lost in one step. What would be left is the main verb with its patient argument which would lack Case and hence have to move to subject position:
(79) |
Thus, if we analyse the passive construction as involving the loss of the light verb, we readily account for its two most salient properties.
We might extend this analysis to take into consideration other aspects of the passive construction. It has been argued that one of the central aspects of the passive is the appearance of the passive morpheme. The passive morpheme appears in all English passives, no matter what else happens. Thus, not all passives involve an object moving to subject position (80a), and while some passives contain a by phrase reintroducing the missing subject, not all do (80b). Furthermore, most passive constructions involve a passive auxiliary be, but not all (80c):
(80) | a | it is expected that Pete will post the letter |
b | the letter will be posted (by Pete) | |
c | I will expect [the letter posted by noon at the latest] |
In all these examples however, the passive verb has the passive morpheme. We can incorporate this into our analysis of the passive making use of an idea presented above, that certain bound morphemes can be analysed as light verbs. Suppose we assume that the passive morpheme is a light verb which replaces the agentive light verb of the active. As the passive light verb does not assign a Θ-role to its subject, it will not be able to assign a Case to the theme in the specifier of the VP and hence this argument will have to move to subject position. Moreover, the bound morpheme status of the passive light verb will force the main verb to move in order to support it:
(81) |
From this perspective, then, the analysis of the passive involves replacing the agentive light verb with a non-agentive passive light verb, most of the other aspects of the passive construction follow straightforwardly from this.
A crucial point to make at this point is that this analysis of the passive would simply be unavailable if we supposed that the structure of the active were to be (76) and not (74). Inasmuch as this analysis helps us to understand the passivisation process any better, then, it can be used as evidence in favour of the assumption of (74).