On a Scope Reconstruction Paradox
Howard Lasnik
University of Connecticut
Zubizarreta (1982) attributes to Noam Chomsky the observation that while a universal quantifier in subject position can be understood inside the scope of clausal negation, as in (1), this possibility disappears when the subject undergoes raising to subject position, as in (2).
(1) (it seems that) everyone isn't there yet
(2) everyone seems [t not to be there yet]
Zubizarreta takes this as a potential argument against Quantifier Lowering, and, in fact, takes Chomsky to be making such an argument. Quantifier Lowering, discussed especially by May (1977) and May (1985), accounts for the apparent ambiguity of such examples as (3).
(3) Some politician is likely to address John's constituency
May (1977) observes that "[(3)] may be taken as asserting either (i) that there is a politician, e.g., Rockefeller, who is likely to address John's constituency, or (ii) that it is likely that there is some politician (or other) who will address John's constituency." Since May's ground-breaking work, that ambiguity has generally been taken, following May, to stem from whether the surface subject 'lowers' or not.
On May's account, this lowering is a variant of Quantifier Raising (QR). Like QR, it moves a quantifier and attaches it to S. QL differs from QR only it that it attaches the quantifier to a lower clause. Another view of QL has also been considered in the literature. Under this alternative view, recently discussed by Hornstein (1995), QL is a reconstruction effect. The quantifier is interpreted as if it were in a position it occupied at any earlier point in the derivation. In cases like (3), that position is embedded subject.
Chomsky (1995), in effect, distinguishes these two theories. He argues, partly on the basis of (1)-(2), that there are no reconstruction effects at all with A-movement, but he allows for the possibility of QL. Chomsky suggests that 'lowered' readings
"... could result from adjunction of the matrix quantifier to the lower IP (c-commanding the trace of raising and yielding a well-formed structure if the trace of quantifier lowering is deleted, along the lines of May's original proposal). But reconstruction in the A-chain does not take place..." [p. 327]
Under this approach, QL stands in sharp contrast to other reconstruction-type phenomena, since the others are all taken to result from copy-movement and complementary deletion. As Chomsky says, "The basic assumption here is that there is no process of reconstruction..." [p.326] For QL, on the other hand, there does seem to be a literal operation of reconstruction. The question then arises why it should not be more generally available.
With this in mind, I would like to consider the possibility that there is no actual process of QL and thus that apparent lowered readings of examples like (3) would have to arise from reconstruction in the minimalist sense: interpretation of a trace. On this understanding, as on Zubizarreta's, the fact in (2) is what Zubizarreta (and, much later, Hornstein) took it to be, namely, a potential argument against QL. Both Zubizarreta and Hornstein maintain that the QL phenomenon is real. Zubizarreta notes the apparent counter-argument, but puts it aside. Hornstein argues that the lack of narrow scope for everyone in (2) is independent of raising, hence irrelevant to the issue of A-movement reconstruction. Hornstein (personal communication) suggests that the crucial property is contraction - the wide scope for negation is possible only when negation has contracted. However, for my informants (and myself) examples with uncontracted negation (but without raising) seem to allow wide scope negation rather readily. Consider the following examples, in a situation where a teacher is being reprimanded for giving all the students A's:
(4) School policy requires that everyone not get an A
(5) It is important for everyone not to get an A
Both examples seem reasonably appropriate to the situation, indicating that the reading in question, with the universal understood under negation, is available.(1)
If we therefore adopt Chomsky's point of view that the absence of a reading for (2) with the universal under negation does, indeed, reflect absence of 'reconstruction' with A-movement, the question now arises whether that point of view can be reconciled with Hornstein's natural analysis of standard QL phenomena, as in (3), as A-movement reconstruction. On the face of it, that seems an impossible task: How can one reconcile a contradiction? If there is no A-movement reconstruction, and if QL is A-movement reconstruction, then we are led to the completely bizarre conclusion that QL doesn't exist.
Amazingly, there is some reason to believe that that completely bizarre conclusion is correct.(2) Precise semantic characterizations of 'lowered' readings in QL configurations are surprisingly difficult to find in the literature. Standard descriptions generally rely on paraphrase by a similar sentence with pleonastic it as subject and a finite complement, as in the following, where (7) is the paraphrase for the 'lowered' reading of (6):
(6) Someone is likely to solve the problem
(7) It is likely that someone will solve the problem
But for an interesting and rather wide range of raising examples, paraphrase of this sort fails. Consider the following example:
(8) No large Mersenne number was proven to be prime
(8) cannot accurately be paraphrased by (9).
(9) It was proven that no large Mersenne number is prime
Similar paraphrase failure occurs in the following pair.
(10) Noone is certain to solve the problem
(11) It is certain that noone will solve the problem
(10) describes a situation where the problem under discussion is of at least middling difficulty, and the potential problem solvers aren't omniscient. (11), on the other hand, is a sentence about either an impossible problem or a hopelessly inept group of solvers.
This paraphrase failure is not limited to negative contexts. Consider (12) in a situation where there are five fair coins, flipped in a fair way.
(12) Every coin is 3% likely to land heads
Note that the situation strongly biases the sentence towards the lowered reading, but that reading still is not possible. (12) cannot be accurately paraphrased as (13).
(13) It is 3% likely that every coin will land heads
The only reading for (12) is one describing the much less plausible situation in which each coin is weighted in such a way that it is 33 times more likely to land heads than tails.
So far, we have seen that for two scope phenomena, raising to subject position fails to display reconstruction. Interestingly, there is a 'raising-to-object' construction that displays similar lack of reconstruction. On a very plausible analysis of the make...out construction (a construction discussed by Kayne (1984)), John in (14) has raised into the higher clause (since it appears to the left of the particle out, which is associated with the matrix verb).
(14) They made John out to be a fool
Evidence that John is the underlying subject of the lower clause comes from the fact that (15) seems to be a perfect paraphrase of (14).
(15) They made out that John is a fool
Further, the NP can be a pleonastic:
(16) They made there out to be a solution
(17) They made out that there is a solution
Now notice that the scope contrast between Chomsky's (2) and (1) is mirrored in this particle construction:
(18) The mathematician made every even number out not to be the sum of two primes
(19) The mathematician made out that every even number isn't the sum of two primes
In contrast with (19), the only reading available for (18) is the implausible one where the mathematician was engaged in the absurdly futile activity of trying to convince someone that no even number is the sum of two primes (and not the far more plausible one where she is merely trying to convince someone that Goldbach's conjecture is false). Thus, even with strong pragmatic bias towards wide scope for the negation, it still isn't available.
Significantly, though the judgments are subtle, the failure of 'quantifier lowering' seen in classic raising examples like (8) is also mirrored in the raising particle construction. Compare (20) with (21):
(20) The DA made no defense witnesses out to be credible
(21) The DA made out that no defense witnesses were credible
On pragmatic grounds, the only remotely plausible interpretation of (20) would be one synonymous with (21). But that interpretation is very difficult to obtain. Instead, the sentence has a bizarre interpretation where the DA perhaps had the intention of trying to show that (some of) the defense witnesses were credible, but never acted on that intention.
Thus, there is substantial evidence that Chomsky is correct about absence of (one type of) A-movement reconstruction. Further, there is even some reason to believe that the stronger interpretation of Chomsky's claim, whereby even 'quantifier lowering' doesn't exist, is correct. But for the latter, there is, of course, all the standard contradictory evidence that QL does exist, including the famous ambiguity of (3), repeated here as (22).
(22) Some politician is likely to address John's constituency
What is the difference between examples like (22), where lowering seems to obtain, and examples like (10) and (12), repeated here, where it doesn't?
(23) Noone is certain to solve the problem
(24) Every coin is 3% likely to land heads
As far as I can tell, all of the standard examples displaying a QL ambiguity have indefinite subjects, unlike my examples of lowering failure. However, on standard accounts, either of the reconstruction variety or of the lowering variety, it is totally unclear why that should matter.
There is one familiar argument that 'lowered' readings do, indeed, involve some sort of syntactic operation. While compelling on the face of it, the argument actually seems to me inconclusive. The argument, due to May (1985), is that a 'lowered' reading for the quantifier is incompatible with the binding of a pronoun in the upper clause. May gives the following example:
(25) No agenti was believed by hisi superior to be a spy for the other side
(25) cannot be paraphrased as (26).
(26) It was believed by hisi superior that no agenti was a spy for the other side
There are, however, two interfering factors here. First, I don't believe that 'lowering' is possible for (25) even without pronoun binding, as in (27).
(27) No agent was believed by Dulles to be a spy for the other side
In (27), it is not easy to see this, perhaps because the two readings are close. Changing the matrix predicate to known or proved sharpens the difference and makes it clear that the lowered reading doesn't exist:
(28) No agent was known by Dulles to be a spy for the other side
(29) No agent was proved by Dulles to be a spy for the other side
The second interfering factor stems from the very characterization of the lowering phenomenon. Taking paraphrasability by an expletive construction as the defining property, it becomes trivial that lowering isn't available in (25), since the paraphrase is ungrammatical in its own right, constituting a Weak Crossover violation. And, as now expected, this second interfering factor persists even when the first one is controlled for (i.e., by substituting an indefinite subject):
(30) Some agenti was believed by hisi superior to be a spy for the other side
Once again, paraphrase fails (on the indefinite reading), but once again, the expletive example is ungrammatical, a WCO violation, whether as a paraphrase or not:
(31) It was believed by hisi superior that some agenti was a spy for the other side
Thus, it is difficult to draw any firm conclusion from the pronoun binding argument.
What should we make of all of this? As far as I have been able to tell, the standard lowering examples differ from the new 'anti-lowering' ones I have presented in that the truth conditional difference between the hypothesized lowered reading and the non-lowered reading is much less clear in the standard examples. The clearer the truth conditional difference, the less accessible the lowered reading seems to be. This suggests that there is no lowering at all, the apparent lowered reading of raised indefinites having some other source. Conceivably, the basic difference between the two readings of sentences with raised indefinites lies in the speaker's point of view with respect to the raised subject. On one reading, the speaker has a particular individual in mind, but, for some discourse reason or other, does not identify that individual. On the other reading (the 'lowered' one), the speaker does not have any particular individual in mind. The ambiguity might than fall under theme-rheme properties, the 'wide scope' quantifier being a theme or topic.(3) Or perhaps some as yet unidentified semantic property of indefinites is relevant.(4) If either of these possibilities turns out to be correct, the strong interpretation of Chomsky's claim that there is no A-movement reconstruction might yet be viable.(5)
References
Chomsky, Noam. 1995. Categories and transformations. In The minimalist program, 219-394. Cambridge, Mass.: MIT Press
Hornstein, Norbert. 1995. Logical Form: From GB to minimalism. Generative Syntax. Cambridge, Mass. and Oxford: Blackwell Publishers Inc.
Kayne, Richard. 1984. Principles of particle constructions. In Grammatical representation, ed. Jacqueline Guéron et al. Dordrecht: Foris
Kratzer, Angelika. 1997. Scope or pseudoscope? Are there wide-scope indefinites? In Events and grammar, ed. Susan Rothstein. Dordrecht: Kluwer
Lasnik, Howard. In press. Chains of arguments. In Working minimalism, ed. Samuel D. Epstein and Norbert Hornstein. Cambridge, Mass.: MIT Press
May, Robert. 1977. The grammar of quantification. Doctoral dissertation, MIT, Cambridge, Mass.
May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press.
Postal, Paul M. 1974. On raising: One rule of English grammar and its theoretical implications. Cambridge, Mass.: MIT Press.
Zubizarreta, Maria Luisa. 1982. On the relationship of the lexicon to syntax. Doctoral dissertation, MIT, Cambridge, Mass.
Notes
1. In fact, in the examples originally cited by Zubizarreta contraction is not involved:
i) Everyone will not come
ii) Everyone is likely not to come
2. Postal (1974) already claimed that quantifiers that have undergone subject raising have only high scope.
3. 3. Or, alternatively, the specific/non-specific ambiguity of indefinites. These two distinctions overlap to a significant extent.
4. I would like to think that this property would relate to the 'pseudoscope' of indefinites enlighteningly discussed by Kratzer (1997), but at the moment I see no obvious way to connect the phenomena.
5. See Lasnik (In press) for further examination of A-movement reconstruction.