In this note, we explore the nature of the missing constituent in the comparative clause in examples like (1).

(1)	Jones published more papers than Smith {knew / said / thought / predicted / admitted / expected / wanted}.

Though the varieties of missing constituents and ellipses found in comparative clauses have long attracted attention, comparatives of the kind illustrated in (1) have mostly escaped notice. In this note, we argue that despite appearances, what is missing from the comparative clause in (1) and similar examples is not a CP, but rather a nominal element, an anaphoric comparative operator that contains a free variable over propositions. Positing a DP element in such cases solves certain syntactic puzzles for the distribution of such gaps in an elegant way, as we show below.¹

1 Missing CP?

At first glance, the verbs in (1) seem to be missing their propositional arguments; compare their counterparts in (2), in which the propositional argument (here, a complement) is a tensed CP, and in (3), in which it is an untensed IP.

(2)	Smith {knew / said / thought / predicted / admitted } that Jones published five papers.
(3)	Smith {expected / wanted} Jones to publish five papers.

The question at hand is whether this prima facie identification of the missing material is correct, that is, whether or not an example like (1) is to be generated by the operation of comparative deletion (CD) in conjunction with null complement anaphora (as pursued in Napoli 1983) or 'CP/IP ellipsis'.

The most compelling argument against assimilating the cases in (1) to null complement anaphora (NCA; see Hankamer and Sag 1976) is that the set of verbs that occur in comparatives like (1) is not coextensive with those that license NCA, as pointed out in Moltmann 1993, and illustrated by the difference between (4) and (5).

(4)		The missile test had failed, but only the brass knew.
(5)	*	The missile test had failed, but only Prof. Hicks {said / thought / expected / predicted / admitted / wanted}.

Neither is there any independent reason to believe that an operation of ellipsis targeting CPs in particular exists in English (see Merchant to appear, sec. 4.2.1; Hankamer 1971).

So far, our conclusions parallel those of Hendriks 1995:229–231, who argues that the category of the than that selects such 'missing CP' comparatives must be unique to such predicates. Such a solution, however, is only partial: it still leaves mysterious the fact that the verbs in (1) seem to exhibit this strange selectional behavior (missing their usual arguments) only in comparative clauses.

Luckily, there is a set of facts that have been previously overlooked that show the way out of this dilemma: observe that if the verbs in (1) are passivized and an expletive is inserted in subject position, the examples become ungrammatical.

(6)	*	The committee took much longer to decide than it was expected.
(7)	*	The mission turned out to be more expensive than it was originally predicted.
(8)	*	The storm inflicted an even greater amount of damage than it was first reported.

This effect extends as well to adjectives that take CP complements:

(9)	*	Jones published more papers than it was necessary.
(10)	*	Sally had a more serious problem than it was {evident / apparent}.

The ungrammaticality of these examples would be surprising if it were simply a matter of a CP being missing, all the more so given that when a (full or reduced) CP is present, the examples are fine:

(11)	The committee took much longer to decide than it was expected that they would (take to (decide)).
(12)	The mission turned out to be more expensive than it was originally predicted that it would be.
(13)	The storm inflicted an even greater amount of damage than it was first reported that it would (inflict).
(14)	Jones published more papers than it was necessary for him to publish.
(15)	Sally had a more serious problem than it was {evident / apparent} that she had.

The contrast between the examples in (6)–(10) and those in (11)–(15) is unexpected if the former are simply elliptical versions of the latter. A further mystery for the 'missing CP' view is the fact that if the expletive subject it is omitted in (6)–(10), the resulting sentences are grammatical:

(16)	The committee took much longer to decide than was expected.
(17)	The mission turned out to be more expensive than was originally predicted.
(18)	The storm inflicted an even greater amount of damage than was first reported.
(19)	Jones published more papers than was necessary.
(20)	Sally had a more serious problem than was {evident / apparent}.

Instead, these contrasts point to the conclusion that the gap in examples like (1) is nominal in nature; as such it must be assigned Case, either accusative as in (1), or nominative in (16)–(20).

The distribution of such gaps tracks the availability of Case across a range of environments (again, something unexpected if the gap is categorially a CP). First, note that all the verbs in (1) can assign Case to nominal objects, as in Smith {knew / said / thought / predicted / expected / wanted} the answer or Smith admitted his error. Second, if we test a predicate which can take a CP argument, but cannot take a DP argument, we have a clear prediction: if we are dealing with a 'missing CP' (either by virtue of NCA or ellipsis), an example parallel to (1) should be grammatical. If, on the other hand, we are in fact dealing with a nominal element with its concomitant Case requirement, as proposed here, such an example should be ruled out. The test case, using verbs such as bet or wager (Smith {bet / wagered} that he'd win but *Smith {bet / wagered} the race²), bears out our conjecture:

(21)	Jones published more papers than Smith {bet / wagered} *(she would).

We conclude that there is no 'missing CP', and that the gaps in question are the traces of phonologically null nominal expressions that have moved into specCP within the comparative clause.

2 The syntax and semantics of the operator

The proposed operator is a variant of the overt wh–phrase found in questions like (22) and comparatives like (23), which shows the same sensitivity to Case.

(22)	What was (*it) {necessary / expected / predicted / reported}?
(23)	The committee took much longer to decide than what was (*it) expected.

We will therefore represent the resulting structures as follows:

(24)	Jones published more papers than [CP what [IP Smith thought t]]
(25)	The committee took much longer to decide than [CP what [IP t' was expected t]]
(26)	The mission turned out to be more expensive than [CP what [IP t' was originally predicted t]]
(27)	The storm inflicted a greater amount of damage than [CP what [IP t' was first reported t]]
(28)	Jones published more papers than [CP what [IP t' was necessary t]]
(29)	Sally had a more serious problem than [CP what [IP t' was evident t]]

Like other nominal expressions, the operator what must receive Case. This accounts for the Case sensitivity noted above. In particular, the novel data in (6)–(10) fall into place: because an expletive occupies the subject position in these examples, what fails to get Case, with the consequence that the sentences are ill-formed.

The semantic analysis of this operator is similarly straightforward. Here we will illustrate using the examples of nominal amount comparison as in (24) and (28). We assume that more denotes a relation between two degrees/amounts (cf. Postal 1974, Cresswell 1976, von Stechow 1984, Heim 1985, Moltmann 1992, Hendriks 1995, Kennedy 1999):

(30)	||more'(i d1 [...d1...])(i d2 [...d2...])|| = 1 iff d1 > d2

In the comparatives at hand, what must be a propositional anaphor (cf. Lerner and Pinkal 1995), not a degree term (as in e.g. Moltmann 1992). Assuming that the comparative clause denotes a definite description of a (maximal) degree (von Stechow 1984, Rullmann 1995, Kennedy 1997), the semantic value of what— ||what|| = Op in (31b)— must be a propositional expression that contains a free variable over degrees (or amounts).

(31)	a.	Jones published more papers than [CP what [IP Smith thought t]]
	b.	than [CP what [IP Smith thought t]] = i d [l p[thought'(smith',p)](Op)]

Therefore, the value of Op must be contextually determined, anaphoric to the open degree term given by the main clause (we leave open here the mechanism by which this anaphoric dependency should be resolved; presumably the approach developed in Rooth 1985 for propositional anaphors is extendible to the present cases):

(32)	Op = $ X[papers'(X) & |X|=d & published'(jones',X)]]

Lambda conversion on (31b) with (32) as the value of Op yields (33):

(33)	i d[thought'(smith',($ X[papers'(X) & |X|=d & published'(jones',X)]))]

This term then serves as the second argument to ||more||, yielding the full desired translation of (24):

(34)	more'(i d[$ X[papers'(X) & |X|=d & published'(jones',X)]]) (i d[thought'(smith',($ X[papers'(X) & |X|=d & published'(jones',X)]))])

Similarly for (28), given below:

(35)	a.	Jones published more papers than [CP what [IP t' was necessary t]]
	b.	than [CP what [IP t' was necessary t]] = i d[l p[necessary'(p)](Op)]
(36)	a.	more'(i d[$ X[papers'(X) & |X|=d & published'(jones',X)]]) (i d[l p[was-necessary'(p)](Op)]])
	b.	Op = $ X[papers'(X) & published'(jones',X) & |X|=d]
	c.	more'(i d[$ X[papers'(X) & |X|=d & published'(jones',X)]]) (i d[was-necessary'($ X[papers'(X) & |X|=d & published'(jones',X)]))])

Thus it is possible, and desirable, to combine the syntax of a nominal with the semantics of a proposition.

3 Conclusions

Three conclusions can be drawn from our discussion of 'missing CP' comparatives. First, we have uncovered a new descriptive fact about the grammar of English comparatives: 'missing CP' comparatives are sensitive to Case. This follows if a phonologically null nominal expression is generated in the position of the 'missing' clause. Second, 'missing CP' comparatives do not provide evidence for CP-ellipsis in English, leaving the typology of ellipsis operations in English unchanged. Finally, if the anaylsis presented here is correct, no uniform analysis of comparatives in English can be given. While it may be possible to analyze many different types of comparatives in terms of a single operation plus general principles of ellipsis, the facts discussed here indicate the irreducible need for a more interpretive analysis of some constructions as well.

References

Adger, David. 1994. The licensing of quasi-arguments. In P. Ackema and M. Schoorlemmer (eds.), ConSole I proceedings, 1–18. Holland Academic Graphics: The Hague.

Cresswell, M. J. 1976. The semantics of degree. In B. Partee (ed.), Montague Grammar, 261–292. Academic Press, New York.

Hankamer, Jorge. 1971. Constraints on deletion in syntax. Ph.D. thesis, Yale University.

Hankamer, Jorge, and Ivan Sag. 1976. Surface and deep anaphora. Linguistic Inquiry.

Heim, Irene. 1985. Notes on comparatives and related matters. Ms., University of Texas, Austin.

Hendriks, Petra. 1995. Comparatives and categorial grammar. Ph.D. thesis, University of Groningen.

Kennedy, Chris. 1997. On the quantificational force of the comparative clause. In The Proceedings of ESCOL '97. Cornell Linguistics Club Publications: Ithaca.

Kennedy, Chris. 1999. Projecting the adjective: The syntax and semantics of gradability and comparison. Garland Press: New York.

Lerner, Jean-Yves and Manfred Pinkal. 1995. Comparative ellipsis and variable binding. In The Proceedings of SALT 5, 222–236. Ithaca: Cornell Linguistics Club Publication.

Merchant, Jason. To appear. The syntax of silence: Sluicing, islands, and identity in ellipsis. Oxford University Press: Oxford.

Moltmann, Friederike. 1992. Coordination and comparatives. Ph.D. thesis, MIT.

Moltmann, Friederike. 1993. The empty element in comparatives. In A.J. Shafer (ed.) Proceedings of the North East Lingusitics Society (NELS 23): 319–333. Univ. of Ottawa.

Napoli, Donna Jo. 1983. Comparative ellipsis: A phrase structure analysis. Linguistic Inquiry 14.4:675–694.

Postal, Paul. 1974. On certain ambiguities. Linguistic Inquiry.

Rooth, Mats. 1985. Association with focus. Ph.D. thesis, University of Massachusetts, Amherst.

Rullmann, Hotze. 1995. The syntax and semantics of wh–constructions. Ph.D. thesis, University of Massachusetts, Amherst.

von Stechow, Arnim. 1984. Comparing semantic theories of comparison. Journal of Semantics 3:1–77.

Endnotes

¹ We are very happy to present this paper to Jorge Hankamer on the occasion of his 60th birthday. Jorge's inspiring teaching and scholarship, and his work on comparatives and deletion phenomena in particular, have sparked and guided our own research on these topics. In fact, the central data discussed in this paper may even have been first noticed by Jorge in his (in)famous "Pied-Wiping" paper, though we may never know for sure, since that paper seems to have been the target of an unrecoverable deletion. [Back]

² The fact that these verbs allow a pseudo-object in the form of the amount wagered is immaterial here; see Adger 1994, who argues that structural Case is not assigned to such 'quasi-arguments'. [Back]