You are on page 1of 22

6.

5 Subject questions
Thus far, our discussion has mainly centred on complement questions – i.e.
structures in which a the question word functions as the complement of a
verb or preposition. In this section, we turn instead to look at subject
questions – i.e. structures like (38) below in which the question word
functions as the subject of the clause containing it:
(38)
(a) Who found it?

(b) Who has the keys?

(c) What happened to him?

(d) What upset you?

Two inter-related issues arise in relation to subject questions: one is where


the wh-word is positioned; and the second is why subject questions like those
in (38) don’t require D O -support (except in emphatic contexts like Who DID

find it? where use of DO is associated with emphasis rather than


interrogativeness, as we see from the use of emphatic DO in non-
interrogatives like He D I D find it).
In relation to the position of the wh-word in subject questions, there are
two conflicting perspectives found in the research literature. One is to
suppose that the key requirement for a (non-echoic) wh-question is that it
begin with an interrogative wh-constituent, and that this requirement can be
met if the wh-subject occupies the same position as other subjects, namely
spec-TP. On this (spec-TP) analysis, (38a) would have the syntactic structure
in (39) below, where Af is a past tense affix:

(39)
The key point to note here is that who (under the analysis in 39) is taken to be
positioned in spec-TP, not in spec-CP. The past tense affix in T will lower
onto the verb find in the PF component so that the verb is ultimately spelled
out as the past tense form found in (38a) Who found it? If C triggers
Auxiliary Inversion only when there is a question operator in spec-CP (in
conformity with the Inversion Licensing Condition 4ii), the spec-TP analysis
will correctly account for why there is no Auxiliary Inversion in subject
questions. The reason is that the question word who in (39) is in spec-TP, and
hence (if only operators in spec-CP trigger Inversion), it follows that there
will be no Auxiliary Inversion in subject questions.
However, an alternative perspective on subject questions is the
following. Since who in (38a) Who found it? is a question operator, it should
behave like other operators which originate below C (e.g. who in 31) and
move into the canonical spec-CP position occupied by question operators.
Under this second (spec-CP) analysis, (38a) Who found it? will not have the
structure (39) above but will instead have the structure (40) below. Under the
analysis in (40), who originates in spec-TP and from there moves to spec-CP
(via the arrowed Wh-Movement operation), leaving behind a copy of who in
spec-TP which is subsequently given a silent spellout in the PF component by
the Default Spellout Rule (10):

(40)

Subsequently the past tense affix in T will lower onto the verb in the PF
component, with the result that the verb is spelled out as the past tense form
found in (38a) Who found it? What then remains to be accounted for is why
there is no D O -support and Auxiliary Inversion in a subject question structure
like (40). Pending a more detailed discussion of this issue below, let’s
suppose that when C attracts an interrogative operator to move to spec-CP in
a root/main clause, C only triggers Auxiliary Inversion if the attracted
operator moves from a position below T: since who moves to spec-CP from a
position above T (i.e. from spec-TP) in (40), it follows that there will be no
Auxiliary Inversion.
As the discussion in the two previous paragraphs illustrates, there are
two (seemingly equally plausible) competing analyses of subject questions:
one supposes that the question word remains in spec-TP, the other that it
moves to spec-CP. But which is the right analysis? As we will see, there are
theoretical and empirical considerations favouring the spec-CP analysis in
(40) over the spec-TP analysis in (39). For one thing, the spec-CP analysis is
more consistent with our existing theoretical principles and assumptions. For
example, the Clause Typing Condition (3) favours the spec-CP analysis,
since it specifies that a clause will only be typed as an operator question if it
contains an operator in spec-CP. Likewise, our account of the feature
composition of C in questions in (26i) claims that C always has a Q-feature in
operator questions, and if so an interrogative C will always require an
interrogative specifier.
There is also empirical evidence which favours the spec-CP analysis. As
we see from the examples in (41) below, a wh-word like who in a subject
question like (38a) Who found it? can be substituted by who on earth or who
the hell:
(41)
(a) Who on earth found it?

(b) Who the hell found it?

Examples like those in (42) below show that wh-expressions like who on
earth and who the hell have the property that they cannot remain in situ in an
echo question like (42b), but rather must move to spec-CP as in (42a):
(42)
(a) Who on earth/Who the hell is she going out with?

(b) *She is going out with who on earth/who the hell?

We find a similar restriction in multiple wh-questions like (43) below,


(43)
(a) I’ve no idea [what the hell she gave to who at Christmas]

(b) *I’ve no idea [what she gave to who the hell at Christmas]

If wh-on-earth/wh-the-hell constituents like those italicised in (42,43) always


move to spec-CP, it follows that the italicised subjects in (41) must likewise
have moved to spec-CP; by the same token, it is plausible to suppose that the
same is true of the subject who in (38a) Who found it?
A second piece of empirical evidence favouring the spec-CP analysis
comes from speakers of English who allow wh+that structures – i.e. CPs in
which a wh-word or wh-phrase is positioned in front of the complementiser
that. Such speakers allow a subject wh-constituent to be positioned before the
complementiser that – as the bracketed clauses in the examples below
(recorded from radio and TV programmes) illustrate:
(44)
(a) It’ll probably be evident from the field [which of the players that are
feeling the heat most] (Jimmy Hill, BBC1 TV; Radford 1988)

(b) Regardless of [which version of the FEC bill that is passed], …


(Interviewee on NPR’s All Things Considered, 31 August 1994; Zwicky
2002)

(c) … until late in the week, when we see [how many people that were
arrested] (Interviewee on NPR’s All Things Considered, 26 May 1998;
Zwicky 2002)

(d) Yes, I fully understand [what uproar that would come about]
(Public radio interview of Miami Chief of Police, 2 May 2000; Zwicky
2002)

(e) … and definitions vary as to [which of these types of criteria that are
used] (Member of Göteborg University’s English Department; Seppänen
1994)

(f) You’re talking about [what players that might be coming into that
huge football club] (Micky Gray, Talk Sport Radio)

(g) I have to weigh up [which of those promises that are more important
than others] (Stephen Williams, BBC Radio 5)

(h) We’ll have to see [what that happens] (Eddie Irvine, BBC1 TV)

(i) They’re not worried about [what potential injuries that might come]
(Roy Keane, ITV)
The bracketed clauses in (44) are subject questions containing an italicized
interrogative wh-subject followed by the complementiser that. If the wh-
subject were in spec-TP, it would wrongly be predicted to follow the
complementiser that. The fact that the wh-subject precedes that in sentences
like (44) is consistent with the view that it moves from spec-TP to spec-CP.
More generally, sentences like (44) lend empirical support to the spec-CP
analysis.
Since theoretical and empirical considerations like those outlined above
favour the spec-CP analysis of subject questions, this is the analysis which I
will adopt from this point on. In the light of this assumption, let’s take a
closer look at the derivation of (38a) Who found it? The verb find merges
with the pronoun it to form the VP find it. A past tense affix (Af) is merged
with this VP to form the T-bar Af find it, and this T-bar is in turn merged with
the pronoun who to form the TP who Af find it. The resulting TP is merged
with a null interrogative C constituent which carries a Q-feature in
accordance with (26i) and a T-feature in accordance with (26ii), so forming
the structure below:

(45)
At this point, we might expect the derivation to continue as it did in (31),
with the Q-feature on C attracting the question operator who to move to spec-
CP, and the T-feature on C attracting the tense affix in T to adjoin to C (as
shown by the two different arrows below), so deriving the following
structure:

(46)

However, this would leave the inverted tense affix stranded in C without an
overt host, so causing the derivation to crash.
The affix-stranding problem can be overcome if we make use of DO-

support and merge the auxiliary did in T, and then raise it from T to C in the
manner shown below (where did can be taken to comprise an irregular stem
form of D O with the past tense affix -d adjoined to it):

(47)

And yet the resulting sentence *Who did find it? is ungrammatical if did is
unstressed (as becomes clearer from a sentence like *Who’d find it
yesterday? where unstressed did is contracted to’d). Why should this be?
The likely answer is that a constraint of some kind rules out the dual
movement in (47). But what kind of constraint? One possibility would be the
following:

(48) Edge Constraint


No more than one constituent can be extracted from the edge of any
given projection.

Such a constraint would prevent C from extracting both the T constituent did
and its specifier who out of the edge of TP. The same constraint would also
account for contrasts such as the following:
(49)
(a) The spare parts, how long will [it will take to order the spare parts]?

(b) *The spare parts, how long will [the spare parts will take to arrive]?

In (49a), only one constituent (= the auxiliary will) is moved out of the edge
of the bracketed TP, and the resulting sentence is grammatical. By contrast in
(49b) both the auxiliary will and the subject the spare parts are moved out of
the edge of the bracketed TP (the latter perhaps moving to the specifier
position in an outer CP, under a CP recursion analysis), and the resulting
structure violates the Edge Constraint and so is ungrammatical.
An alternative way of accounting for the ungrammaticality of structures
like (47) is to suppose that they violate a constraint against crossing
movement paths which was formulated by Pesetsky (1982b: 309) in the
following terms:

(50) Path Containment Condition


If two paths overlap, one must contain the other.

The effect of this constraint can be illustrated by comparing the movement


paths for Wh-Movement and Auxiliary Inversion in the two structures below:

(51)
In (51a), the movement path/chain for did and that for who cross each other,
leading to ungrammaticality. By contrast, in the grammatical structure (51b),
the did-chain is contained inside the who-chain, and this leads to a
grammatical outcome (Who did you find?). (As you will no doubt have noted,
the contrast in 51 could equally be handled in terms of the Edge Constraint,
48.)
So, the situation we face is the following. If both Wh-Movement and
Auxiliary Inversion apply in root subject questions, the resulting structure
will be ungrammatical by virtue of violating the Path Containment Condition
and/or the Edge Constraint. But given the feature composition of C outlined
in (26) above, root questions (whether subject questions or not) will always
carry both a Q-feature and a T-feature. We would therefore expect the Q-
feature to trigger Wh-Movement and the T-feature to trigger Auxiliary
Inversion in all types of root question, including root subject questions. But
(as we have seen), this will induce a constraint violation in root subject
questions. How can we overcome this problem?
One way is as follows. Let us suppose that the T-feature on C (contrary
to what we have assumed so far) does not necessarily require C to attract the
head T of TP to adjoin to the head C of CP. Instead, let us suppose that the T-
feature on C works in the manner specified by the generalisation below:
(52) T-Feature Generalisation
A T-feature on C attracts a constituent on the edge of TP to move to
the edge of CP.

Assuming (52), consider what happens when we reach the stage of derivation
in (45) above, repeated as (53) below:

(53)

Let us suppose that a head can only move into another head position, and
likewise a specifier can only move into another specifier position (below, we
will see that this is the consequence of a constraint called the Chain
Uniformity Condition). This being so, it follows from (52) that the T-feature
on C in (53) can either attract the head T of TP to move to adjoin to the head
C of CP or attract the specifier of T to move to become the specifier of C. In
the light of this, let’s return to consider what the Q-feature and the T-feature
on C do in a structure like (53). As we have already seen, the Q-feature on C
triggers Wh-Movement of the question operator who from spec-TP to spec-
CP. But what of the T-feature on C? In accordance with the generalisation in
(52), the requirements of the T-feature on C could in principle be satisfied by
adjoining T to C (i.e. by T-to-C Movement/Auxiliary Inversion), since this
results in movement of a head from the edge of TP to the edge of CP.
However, the requirements of the T-feature on C could alternatively be
satisfied by Wh-Movement of who from spec-TP to spec-CP, since this
results in movement of a specifier from the edge of TP to the edge of CP – as
shown by the arrow in (54) below:

(54)

Under the analysis in (54), the requirements of two different features on the
same head (the Q- and T-features on C) can both be satisfied by a single
movement operation (Wh-Movement of who from spec-TP to spec-CP).
Since the resulting structure in (54) contains an interrogative operator in
spec-CP, it is typed as a wh-question in accordance with the Clause Typing
Condition (3). The unattached past tense affix in T will then lower onto the
verb F I N D via Affix Lowering in the PF component, with the result that the
verb is spelled out as the past tense form found in (38a) Who found it?
Our discussion here shows that there are (in principle) two different
ways of satisfying the requirements of the Q- and T-features on C in a subject
question structure like (53). One is for the Q-feature on C to trigger Wh-
Movement of who to spec-CP, and the T-feature on C to trigger Auxiliary
Inversion (i.e. T-to-C Movement), as in the dual movement derivation in
(47). However, an alternative possibility is for the Q- and T-features on C to
work in tandem to trigger Wh-Movement of who to spec-CP, as in the single
movement derivation in (54). Since a single movement derivation like (54)
(involving only Wh-Movement) is more economical than a multiple
movement derivation like (47) (involving both Wh-Movement and Auxiliary
Inversion), the Economy Principle (32) requires us to use the former rather
than the latter. Consequently, the single movement derivation in (54) yields a
grammatical outcome, whereas the multiple movement derivation in (47)
does not.
While the analysis in the preceding paragraph accounts for the absence
of Auxiliary Inversion in subject questions, the issue arises of why (as
claimed in 52) a T-feature on C should be able to attract either the head of TP
or its specifier to move to the edge of CP. One conceptually attractive answer
to this question is the following. Let us suppose that the T-feature on C
actually attracts a tensed constituent (i.e. a constituent carrying a tense
feature) to move to the edge of CP. Since T is the locus of tense in clauses, it
is obvious why a T-feature on C should be able to attract T to move to the
edge of CP and adjoin to C (in Auxiliary Inversion structures). But why
should a T-feature on C alternatively be able to attract the subject/specifier of
T? An intriguing answer is to suppose that the subject of T acquires a tense
feature as a result of agreement between T and its subject/specifier.
Reasoning along these lines, let us suppose that agreement between T and its
subject works in the manner specified below:

(55) Subject agreement


When T agrees with its subject/specifier

(i) the person/number features of the subject are copied onto the
auxiliary/affix in T

(ii) the tense feature on T is copied onto the subject.

The idea that a nominal constituent could carry a tense feature might at first
sight seem far-fetched, but it is plausible to suppose that certain nominals do
have temporal properties (e.g. my then wife, the future king, his past
achievements). Let’s also assume that what is traditionally called ‘nominative
case’ is actually a reflex of a subject being marked as tensed, thereby
accounting for why only the subject of a tensed T receives nominative case,
not the subject of an untensed infinitival T.
Given these assumptions, in a sentence like He is lying, agreement
between the subject he and the T auxiliary is results in the third person
singular features of he being copied onto is via (55i); and conversely, it
results in the tense feature of is being copied onto he via (55ii). We can
account for why the tensed subject is spelled out as he (rather than as him or
his) if the dictionary entry for the pronoun H E specifies that it is spelled out at
PF as follows:

(56) H E is spelled out as he if tensed, his if genitive, and him otherwise.

In addition to making the assumptions in (55) and (56), let us also assume
that derivations have to be as economical as possible, in consequence of the
Economy Principle (32).
In the light of the assumptions in the previous paragraph, let’s now
reconsider how the requirements of the Q-feature and T-feature on C are
satisfied in a subject question structure like (54) above. The arrowed Wh-
Movement operation satisfies the Q-feature on C because it results in C
having the question operator who as its specifier (so allowing the sentence to
be typed as a wh-question); it also satisfies the T-feature on C because it
results in the tensed subject who moving to the edge of CP (who being
marked as tensed via agreement with the tense affix in T in accordance with
55ii). Recall from (13) above that the dictionary entry for the pronoun WHO

specifies that it is spelled out as follows:

(57) W H O is spelled out as whose if genitive, whom if accusative (in


formal style), who otherwise.

Via agreement with T, W H O in (54) is tensed (and not genitive or accusative),


and so it is spelled out via (57) as the default (‘otherwise’) form who in (38a)
Who found it? Since the Economy Principle (32) favours derivations
involving fewer operations, it follows that the single-movement derivation in
(54) will be preferred to the dual-movement derivation in (47), as we saw
earlier.
The discussion in this section can be summarised as follows. Root
subject questions (i.e. sentences in which the subject of a root clause is
questioned) have the property that they do not trigger Auxiliary Inversion.
One way of accounting for this is to suppose that subjects remain in spec-TP
in subject questions, and there is no Auxiliary Inversion because only
operators in spec-CP trigger Auxiliary Inversion. However, an alternative
possibility (argued for here) is to suppose that C in all types of root operator
questions has a Q-feature and a T-feature, and that these two can work
together to trigger movement of a tensed interrogative subject from spec-TP
to spec-CP. We have seen that there is some theoretical and empirical
evidence favouring the spec-CP analysis.

You might also like