| Caroline Heycock |
Syntactic Theory |
Exam Guide |
5. How can the following set of data concerning wh-movement be accounted for within GB/Principles and Parameters Theory? If there are any problems (for example, if any judgement cannot be accounted for), note these also:
| i. | a. | Who did Kim think that Jo said her brother admired? | |
| b. | * | Who do you despise people who admire? | |
| ii. | a. | Who did they say that Alex gave the box to? | |
| b. | * | Who did they ask which box Alex gave to? (Compare: They asked which box Alex gave to Kim) | |
| iii. | Who would you like a picture of? |
It has been noticed for some time that wh-movement is "blocked" in certain circumstances: for example, movement out of an indirect question (as in (iib)) is ungrammatical, as is movement out of a relative clause (as in (ib)). Within P&P Theory, these "islands" for movement are explained by the proposal that movement is subject to Subjacency . This is a constraint which states that movement can only be out of a maximum of one bounding node, where the "bounding nodes" are NP and IP.
It might seem that Subjacency would predict (incorrectly) that (ia) should be ungrammatical, as the wh-phrase "who" moves from its position as complement to "admire" out of 3 IPs:
i. a. who(i) did [IP Kim think [CP that [IP Jo said [CP [IP her brother admired t(i) ]]]]]
However, it is assumed that wh-phrases need not move in one sweep, but can "stop off" in any intervening CP-Specifier positions that are empty. The Subjacency Condition will treat each movement "between stops" separately from each other movement. In (ia) there is an empty CP-Specifier position above each IP, so that the wh-phrase can "stop off" twice on its way to the initial position; hence each movement respects Subjacency:
i. a. who(i) did [IP Kim think [CP t(i) that [IP Jo said [CP t(i) [IP her brother admired t(i) ]]]]]
In (ib), on the other hand, there is no way for the initial wh-phrase "who" to reach its final position without violating subjacency: the CP-Specifier position of the relative clause is already filled by the relative pronoun "who", and the relative clause is itself an NP. Hence movement to the beginning of hte sentence will always have to be out of at least three bounding nodes, indicated by bold face:
i.b. who(i) do [IP you despiise [NP people [CP who(j) [IP t(j) admire t(i)]]]]
(iia) is accounted for by Subjacency since although the movement is out of two IPs, there is a "stopping" off place in the empty CP-Specifier position (as for ia, it is assumed that that is a head (Comp) and therefore does not occupy the CP-Specifier position itself, unlike relative pronouns such as who, which, etc):
ii.a. Who(i) did [IP they say [CP [C' that [IP Alex gave the box to t(i)]]]]?
ii.b. is ungrammatical, in contrast to iia, since here the CP "which box Alex gave to ___" has a filled specifier (the wh-phrase "which box" has already moved to this position). As a result, the other wh-phrase has to cross two bounding nodes to reach the initial position (again, these are indicated by bold face):
ii.b. Who(i) did [IP they ask [CP which box(j) [C' [IP Alex gave t(j) to t(i)]]]]?
The only problematic example here is the last one, (iii). It appears that movement must have taken place from a position inside the NP "a picture of t":
iii who(i) would [IP you like [NP a picture of t]]
But this violates the principle of Subjacency, since the wh-phrase moves out of two bounding nodes (IP and NP). Yet the sentence is grammatical. This is not explained by the Principle of Subjacency as defined above.
[Extra question] A traditional description of passive is that the object of the sentence becomes the subject of the sentence (where the former subject can optionally be added using a "by"-phrase). How is passive described in GB/P&P Theory and how does it differ from the traditional description?
Note: this answer is an edited and slightly amended version of an answer kindly supplied by a member of class.
In GB/P&P Theory, the passive construction is explained as an instance of NP-movement (other instances being the "Raising" construction exemplified by examples like "The cat seems to be out of the bag.") The parts of the theory that are relevant to the desciption of passive are a) subcategorisation, b) theta role assignment, c) the Case filter and d) the extended projection principle (EPP). To start, let's look at the differences and similarities of active and passive constructions:
The dog ate the cake
The cake was eaten (by the dog)
In the above example, there are two similarities. As can be seen from the interpretation of the sentences, both the active and passive verb assign their internal theta role (such as "patient" or "affected" to the NP "the cake". The second similarity is not so evident: According to the GB/P&P theory, the NP "the cake" fills the subcategorisation frame as a complement NP for both the active and passive verbs "eat/be eaten", despite the fact that in the passive sentence "the cake" appears to be in subject position.
There are differences in terms of case and external theta role assignment. In the active sentence, the verb "eat" assigns accusative to "the cake" (via head-complement government); the passive "be eaten" does not assign accusative case (there is no complement NP to assign it to). Also, "the dog" functions as the external theta role of active "eat", whereas passive "be eaten" does not need an external theta role (the "by-phrase" is optional in the passive).
There are several problems that still have to be explained: Firstly, how can "be eaten" assign its internal theta role to "the cake", when the verb does not govern the NP? Similarly, how can we say that "the cake" satisfies the subcategorisation frame for "be eaten" in the same way as it does for "eat", since in the passive sentence it does not appear to be a complement (and only complements appear in the subcategorization frame). Both of these problems can be explained by NP-movement: If we assume that the NP "the cake" originates from a complement position, to which an internal theta role can be assigned, then the deep structure of the passive example above would look like this:
[IP [I was [VP eaten [NP the cake]]]]
Once the NP has moved, it leaves a trace in its original position, so that the internal theta role assignment is not lost completely through movement. The surface structure would look like this:
[IP The cake (i) [I was [VP eaten [NP (i)]]]]
Another question to be answered is why is this NP movement necessary, i.e. why is it ungrammatical to say simply: *was eaten the cake? There are two possible answers for that. One is the EPP, which says that all sentences must have a subject (this is fairly uncontroversial for English; whether this is a universal principle is more controversial). Because in the passive sentence the subject position is empty at DS, the only other available NP must move. This is not quite true as English allows expletives in subjectless cases such as "it is raining" or "it is thought that car fumes cause asthma". However, an EPP-compliant sentence such as: "*it was eaten the cake" is still ungrammatical. A second possibility is the Case filter, which says that all overt NPs need to be assigned case. If a passive verb construction cannot assign case, the NP has to move to receive case, and the only available position is [Spec, IP], where it can receive nominative by INFL.
The necessary assumptions for this explanation to work are that there are two differences in the lexical entry of an active verb and its corresponding passive. While the subcategorisation of a passive verb is the same as that of the active verb, (a) the passive verb can no longer assign accusative case, and (b) the thematic role that the active verb assigns to the external argument (the D-Structure subject) is not assigned at all by the passive verb (so that at D-Structure the subject position is not assigned any thematic role and is therefore empty).*
The traditional description for passive constructions fails to take into account that in terms of subcategorisation and internal thematic roles, passive and active are the same. There is no attempt to explain how these two are related and how one can be derived from the other. GB/P&P theory tries to account for just that and how a multitude of different structures can be simplified to a basic (deep) structure from which all others can be determined.
Further, there are sentences where a verb assigns accusative Case to an NP that is not its "object"; as for example in the following sentence:
I believe [IP Malcolm to be a fool].
Since "him" is not the object of the verb "believe", the traditional description of passive would not predict that "Malcolm" would become the subject if "believe" is passivised. Under the GB/P&P account, however, we do expect this to happen, since the passive of "believe" would no longer be able to assign accusative case, so that "Malcolm" would have to move into the matrix subject position to be assigned nominative. And in fact it is this prediction that is borne out by the facts:
Malcolm is believed to be a fool.
*It does have to be noted that the status of the "by-phrase", which, when it appears, does seem to have the same thematic role as the corresponding subject of the active verb, is not easily explained in this theory.
| Caroline Heycock |
Syntactic Theory |
Exam Guide |
 |
12th May 2000 |