The argument Structure of WH-Operators
Linguistics 322
To
understand the meaning of WH-operators, perhaps a look at the
semantics or logic of WH-operators might be helpful. Let us start
with the following sentence:
(1) What did John buy?
This sentence implies that John bought something, but the speaker does not know what it is. He is eliciting the information. Putting tense aside for the moment, the basic lexical form of (1) is the following. (2) is old information:
(2) [John buy something] = [VP BUY <theme: SOME> <agent: JOHN>].
'SOME' is the lexemic form which underlies something, somebody, somewhere, sometime, somehow, someplace, and so forth. These forms are almost but not quite pronominals--pronominals need an antecedent, but these forms do not. The feature [+humanlike] spells out somebody, [-humanlike] something, and so forth.
Since (1) is an information seeking question, let's add the [+Q] operator to this proposition; [+Q] is an operator:
(3) [VP [C +Q] <[VP BUY <SOME+[-HL]> <JOHN>]>]].
Unmodified, (3) underlies a yes/no question: the speaker would be trying to determine whether John bought something. Questions can be negated; in this case it isn't; the sentence is [-Neg]:
(4) [VP [-Neg] <[VP [C +Q] <[VP BUY <SOMETHING> <JOHN>]>]>].
WH-questions
are linked to something in the pragmatic world or in the discourse.
In (4) the speaker of a WH-clause already knows or assumes that
John has bought something. He desires to find out what the object
or objects are that John has bought. The speaker uses a pragmatic
operator to request this information. In more common terms, it
is called a WH-operator. It builds a link between the old
information in the discourse or situation and copies it to the
question he is about to ask. The WH-operator logically takes into
its scope all of the old information.
In some sense of the term, there are two propositions here. The first proposition is (1) and its logical form (2). The second is the WH-operator itself. The second propositoin the speaker wants to know more about the minimal information given in the first proposition. The first proposition (also called an eventuality) comes from the discourse or from a pragmatic situation: Someone one knocks on the door; you respond "Who is there?" The first proposition is "Someone is knocking on the door", though this propostion has not been uttered.
Let us assuame a case where it has been uttered:
A. John bought something today.
B. What did John buy?
Let us called the initial proposition (A.) VP1. The Wh-clause conflates the second proposition VP2 (B.) into one operator plus the question operator:
(5a) [VP2 WH +Q [VP1 Neg C Prom T R Asp Vc V NP1 NP2 ]].
(5b) [VP2 WH +Q [VP1 Neg C Prom T R Asp Vc BUY SOME JOHN ]].
Structurally, WH is not a clause; +Q replaces C in VP1. This kind of thing Chomsky calls a Generalized Transformation. In the L-structure (5b) (we won't label the structure in (5a) at this time), the [+WH] operator takes scope over VP1. The WH-operator must target SOME:
(6) [VP2 [+WH] <[VP [-Neg] <[VP1 [C +Q] <[VP BUY <SOME+[-HL]> <JOHN>]>]>]>].
The WH-operator modifies (5). [+Q] is now associated with the elicitation of determining what 'something' is: what John bought. Clearly, (5) is not a statement. Similar to the prominence target, the WH-operator needs a target. The target must be something unidentified. The only possible target in (5) is something. The PT operator and the WH-operator may each target a different NP. Here, the WH-operator targets something. We now have the A-structure for (1). 'WHT' stands for the WH-target:
(6) [VP [+WH] <[VP [-Neg] <[VP [C +Q] <[VP BUY <SOMETHING+WHT> <JOHN+PT>]>]>]>].
First, let's draw a reduced structure for (6) continuing to omit the verbal operators:
Let's first discuss the targeted nodes. The NP marked with the prominence target feature needs S as a host; hence the features of the NP are copied to S which then becomes written NP. Similarly, the NP targeted for WH needs a WH host. Therefore, the features of the WH targeted NP are copied to WH. As in apparently all cases of targeted forms requiring a host, the tail of the chain fails to receive a phonetic form, while the head of the chain does receive one:
(9)
To complete the derivation, the verbal operators are included:
[+Past] is copied to the required verbal host. The feature [+Strong] blocks the features of [+Q] from being copied to T; hence the features of T are copied to Q. The remaining operators are realized as phonetic null in the phonological component.
Double
WH words or more.
More than one WH-word may occur in a clause:
(11) What did John give to who(m)?
(12) Who did John give what to?
Both examples have similar argument structures, except for one feature:
(13) WH Pol C S T ... V SOMETHING+[+WH\___] SOMEBODY+[-WH\___]
(14) WH Pol C S T ... V SOMETHING+[-WH\___] SOMEBODY+[+WH\___]
There are two WH-targets here. Only one of them in each structure needs the WH-operator as a host. The other cannot have one as a host. The situation is more complex than this. There is more than one WH-operator. Only one of them is visible in the syntax; the others are not. The visible one functions as a host for the targetted WH-form. The analysis of the "invisible" WH-operators is a topic for more advanced linguistics. It will suffice here to say that they target remaining NPs and PPs.
WH-subjects
Things begin to get hairy when we consider WH-clauses with WH-subjects:
(15) Who read the book?
(16) What hit the car?
Note that the auxiliary verb do does not occur as is customary with interrogative sentences questions. Do is not possible if it is not emphatic:
(17) *Who did read the book?
(18) Who did read the book?
(19) *What did hit the car?
(20) What did hit the car?
So why is the dummy verb do not possible in the unemphatic forms?
We propose that the reason may have to do with the double targeted subject. First note that the external object in (11) is targeted both for [+PT] and for [+WHT]:
(21) WH Pol C S T ... V[READ] NP[BOOK] NP[SOMEBODY]+PT+WH.
The features of the external argument are copied to the subject position (S):
(22) WH Pol C S[NP[SOMEBODY]+PT+WH] T ... V[READ] NP[BOOK] [TRACE].
If we look at the target sentence (11), it is clear that tense ([+Past]) is copied to the verb:
(23) READ+[+Past] => read+d.
The dummy verb {DO} is not inserted when it is not emphasized as (15) shows. If +Q needs tense as a host and it is strong, why doesn't T raise to Q? One possibility is that that the double targetted subject blocks the raising of T. If so, then T, which needs a verbal host, is copied to V resulting in (19). If so, then what happens to Q? It leaves a mark with the intonation contour associated with WH-questions. Is it deleted leaving just this mark? All these suggestions here have not been observed elsewhere in the language.
On the other hand, perhaps Q is lowered to T, and then T is lowered to V--both find their respective hosts, T and V. This would mean that the feature [+Strong] normally associated with Q is not present in this context. But why would the presence of a doubly targetted NP cause this feature to disappear?
Neither solution is supported by a similar phenomenon elsewhere in the language. However, stripping Q of [+Strong] leaves fewer unanswered questions than attempting to block upwards copying here. Let as temporarily adopt the position that the feature [+Strong] is mysteriously removed from Q when Q c-commands a doubly targetted external NP.
Bracketing
w/ WH
When S targets NP, the features [+PT, S\___] are added externally to NP (i.e., they are adjoined to it. When WH targets NP, the features [+WHT, +WH\___] are adjoined to the previous structure:
24
If any feature in a bracketing structure like the above must be copied by NP copying, then all features in the structure are copied
Feature
Processes
There are two basic processes of feature copying. One is feature matching. This occur between heads only, and is called 'Head-to-Head' movement in the standard literature. The other type is targetting. If a feature needs a host, the host is targetted. The process of copying features here is called targetting. It seems to move all features. The other kind of targetting includes +PT, and +WHT targetting. These involve phrases, and the entire phrase is copied to the source targetting position. This includes the traditional rules of NP, VP, PP, and AP movement.
There is more targetting than what we have discussed so far. We have encountered emphasis. Emphasis is apparently a discourse operator that can target almost any constituent or more likely a phrase in the syntax. We haven't gone into this one much, because emphasis is a silent operator. Silent operators are being deferred until advanced syntax.
In
indirect questions it is the subject of the questioning verb who
is seeking information. Let us start with the following sentence:
(25) Mary asked what John bought.
There are two basic propositions here:
(26) Mary ask X
(27) X = John buy something.
In the first proportion, the speaker is not eliciting information. There is no WH-operator here. Let us mark C as [-Q]. [-Q] is a statement (a non-question). [-Q] is null in matrix sentences and optionally that in embedded sentences.
The speaker is trying to elicit information from the embedded clause. C contains [+Q] to mark the elicitation (question) and [+WH] to mark what, the speaker want to know what it is that John bought:
Now we replace X in (14) with (15):
In tree form (15) is represented as:
We
have been lowering tense to the verb.
To
return to course
outline Click
here.
This page last updated 12 AP 2000
