Variable binding, reconstruction and attitudes de re Keir Moulton McGill University MOSAIC 2 – McGill – 1 June 2010

1 Co-varying 3rd person pronouns in attitude complements Co-varying interpretations are available for the pronouns in (1-a) and (1-b): (1)

a. b.

I don’t think any boy1 ’s mother has a clue that he1 is skipping school. That he1 is skipping school, I don’t think any boy1 ’s mother has a clue.

Why this is interesting: I am going to give (essentially) syntactic evidence that the “bound” interpretations here are not the result of the quantifier binding the pronoun; rather, I’ll argue that the semantics of de re attitude ascription gives rise to co-varying de re pronouns in an indirect way. The data I present bear on the following questions about the semantic representation of 3rd person pronouns in attitude complements: • How do co-varying de re pronouns co-vary? There are options of binding the pronoun directly by its quantificational binder or, I’ll argue, more indirectly via res arguments. • As is often an option for 3rd person pronouns in general, can/should we treat these a D-type pronouns? And I’ll suggest the following answers: • the data I offer suggests an approach to co-varying de re pronouns that involves indirect binding, via res arguments in the matrix clause. • Then I’l speculate that the ultimate refinement of this account may be one in which co-varying 3rd person pronouns in attitude ascriptions are really D-type pronouns after all.

1

2 Apparent binding Quick note: I am going to focus on 3rd person pronouns that are not interpreted de se; ask how these can be accommodated under the present approach if you’re interested.

2.1 Fronted clauses Co-varying interpretations for pronouns are available in (2).1,2 (2)

It was clear that something bad was going to happen . . . a. . . .But that he1 was in real danger, I don’t think any boy1 had any clue. b. . . .But that he1 was putting the entire economy at risk, I don’t think any banker1 was aware. c. . . .But that it1 was in real trouble, each board1 had no apparent suspicion. d. . . .But that her1 company was in real danger, many CEOs1 had no idea.

The usual conclusion: the pronoun is c-commanded by its quantificational binder. And so that would mean that the the sentences in (2) involve movement of the fronted or left-dislocated clause (if this requires syntactic reconstruction). But these left-dislocated CPs haven’t moved, because . . . • they violate the DP Requirement on fronted clauses (Section 2.2) • they don’t show island effects (Section 2.3) • their gaps have the distribution of Null Complement Anaphors (Section 2.4)

2.2 The DP Requirement The examples in (2) counter-exemplify the DP Requirement (Williams 1981; Grimshaw 1982; Postal 1986; Webelhuth 1992; Alrenga 2005): (3)

The DP Requirement: The gap of a fronted CP must be a position where DPs are licensed.

(4)

a. This assumption accounts for that claim. b. *This assumption accounts for that these nouns behave differently. c. That these nouns behave differently is accounted for by this assumption. (Alrenga 2005:185-6)

DP *CP Complement Fronted CP

1 In many of these and the examples that follow, the quantificational binder is an NPI, which helps rule out the possibility that quantifier itself QRs to bind the pronoun. And if the quantifier did QR to bind the pronoun, this would create a (very) weak crossover violation. 2 The availability of the bound variable interpretation appears to be determined by whether the pronoun is in a topic clause. Pronouns in independent clauses that antecede NCA do not give rise to bound variable interpretations quite as easily: ?He1 was in danger, but no boy1 had any clue.

2

(5)

a. b. c.

John complained/boasted that he could lift 100 pounds. *John complained/boasted that. *That he could lift 100 pounds, John complained/boasted.

CP Complement *DP *CP Topic

The adjective and nouns in (2) do not take DP complements, making the fronting of CP a mystery. (6)

a. *I had no clue/suspicion/idea/was aware that. b. I had no clue/suspicion/idea/was aware that he left.

2.3 Islands Sentential topics related to DP gaps do show island effects: wh-island violations are given in (7-b)/(8-b) and Complex NP island violations in (7-c)/(8-c) (cf. (7-a)/(8-a)). (7)

a. b. c.

That John left, (that) I think that Mary is not happy about. *That John left, (that) I wondered whether Mary is happy about. *That John left, I made the claim that Mary is happy about.

(8)

a. b. c.

That John left, I think that Mary said that John believed. *That John left, I wondered whether Mary believed. *That he was in danger, I made the claim that no boy believed.

In contrast, the left-dislocated CPs exhibited in (2) can be related to their gaps across islands: (9)

a. b. c.

. . .But that he1 was actually in danger, I think that Mary said that no boy1 had any clue. . . .But that he1 was actually in danger, I wondered whether any boy1 had any clue at all. . . .But that he1 was actually in danger, I made the claim that no boy1 had any clue.

2.4 Diagnosis: It’s Null complement anaphora (NCA) The constructions in (2) involve a base-generated CP in the left-periphery with a null complement anaphor in the gap position—no movement dependency. NCA is a null pro-form anaphor (Hankamer and Sag 1976; Grimshaw 1979; Depiante 2001). NCAs can be CPs (10); this explains the obviation of the DP Requirement.3 (10)

Fred moved to Paris but Mary didn’t know/was not aware/had no idea.

There are no island constraints regulating the relation between NCA and its antecedent (because its not movement): 3 The NCA could be a PP here (i.e. no clue about that). Nothing requires the null pro-form to be of the same category as its antecedent.

3

(11)

a. b. c.

When Jack asked Hector to drive, I wondered whether he refused enca . Mary has to work late nights, and her father made the prediction that her mother wouldn’t approve enca . Anna did the dishes. The fact that no one else volunteered enca , made her upset.

And we saw that the constructions exhibited by (2) showed no island constraints either (see (9)).

Correlation! Just those Ns and As that allow NCA also allow left-dislocated CPs (topicalized CP antecedents given in the (a) examples and the general availability of NCA shown in the (b) examples). (12)

a. b.

. . .But that John was in danger, I had no clue. John was in danger, but I had no clue.

(13)

a. b.

. . . But that John was in danger, I was unaware. John was in danger and we were completely unaware.

(14)

a. b.

. . .That John was in danger, I might have had some idea. No one told me that John was in danger, but I might have had some idea.

(15)

a. b.

. . .But that John was in trouble, I had a suspicion. John didn’t tell me he was in trouble, but I had a suspicion.

Where NCA is independently not allowed, neither are fronted CPs: (16)

a. *. . .That John was in danger, I had a belief. b. *John was not actually in danger, but we had a belief.

(17)

a. *. . .That John was in danger, I didnt hear any report. b. *John was in danger but I didnt hear any report.

(18)

a. *. . ..That Rita was stealing, I overheard a rumor. b. *They didnt know Rita was stealing, but I overheard a rumor.

2.5 Summary No movement, so no reconstruction, but co-varying interpretations are available nonetheless.

4

3 No “reconstruction conflicts” The following argument shows that even those co-varying 3rd person pronouns in left-dislocated CPs that obey the DP Requirement don’t require syntactic reconstruction.

Creating a reconstruction conflict: Syntacitc reconstruction is diagnosed by the following design which creates a binding theory conflict (Lebeaux 1990, Romero 1998 and Fox 1999). (19)

a. b.

[ XP . . . pronoun1 . . . r-expression2 . . . ] *tXP . . . pronoun2 . . . QP1 . . . *tXP [ XP . . . pronoun1 . . . r-expression2 . . . ] ok t XP . . . QP1 . . . pronoun2 . . . *tXP (Fox 2000: 149(9))

A claim in the literature: reconstruction for variable binding feeds Condition C (Lebeaux 1990; Fox 1999).4 (20)

a. *Which part that he1 played with Madonna2 did she2 think that every aspiring actor1 had failed at? b. Which part that he1 played with Madonna2 did every aspiring actor1 wish that she2 would support? (Lebeaux 1991: 223(43))

(21)

a. *But the paper that he1 gave to Mrs. Brown2 , I don’t think she2 would want any man1 to read. b.

But the paper that he1 gave to Mrs. Brown2 , I don’t think any man1 would want her1 to read.

Fronted CPs don’t exhibit reconstruction conflicts (22)

. . . But that he1 might be too old for Mrs. Brown2 , I don’t think she2 would want any man1 ’s mother to believe.

The controls aren’t appreciably better in allowing variable binding: (23)

Copy location satisfying variable binding and Cond C . . . But that he1 might be too old for Mrs. Brown2 , I don’t think any man1 would want her1 to believe.

(24)

No variable binding needed . . . But that John1 might be too old for Mrs. Brown2 , I don’t think he1 would want her1 to believe.

(25)

No possible Condition C violation . . . But that he1 might be too old for her2 , I don’t think any man1 would want Mrs. Brown1 to believe.

4 Sidenote: Late-merger needed to account for (20-b)/(21-b) (Lebeaux 1988; Chomsky 1995; Fox 2002); the relative can be latemerged to the intermediate copy, allowing for variable binding but bleeding condition C:

(i)

hWhich parti did every aspiring actor1 hwhich part that he1 played with Madonna2 i wish that she2 would support hwhich parti?

5

Condition C violation in situ *?I don’t think she2 would want any man1 to believe that he1 might be too old for Mrs. Brown2

(26)

Controlling for “distance effects” (Reinhart 1983) and depth of embedding between pronoun and rexpression: (27)

a. *He1 didn’t expect of any woman2 that she2 would actually ask John1 out on a date. b. That she2 would actually ask John1 out on a date, he1 didn’t expect of any girl in his class2 .

4 Proposal: a res argument is bound We want to let pronouns in clausal arguments (but not just any old DP argument) co-vary without being c-commanded by the quantifier. So the solution should be tied to differences between clausal arguments and other arguments. I am going to argue that a null argument of the embedding verb is bound, and the embedded pronoun only co-varies indirectly. That is, our examples are really of the following form: (28)

That he1 was too old for Mrs. Brown2 , I don’t think she2 would want any boy1 ’s mother to believe

≈ That he1 was too old for Mrs. Brown2 , I don’t think she2 would want any boy1 ’s mother to believe about him1 The argument in this PP expresses the res argument of a de re attitude. And it can be null.

4.1 De Re Attitudes A de re attitude verb, like believe, has a res argument; a subject argument (the attitude holder, att) and a property-type argument P (see e.g.Quine (1956), Kaplan (1968), Stechow (1982) among others.)5 J believe K = λPe,st.λyres .λxatt .λw.∀hx’,w’i ∈ Dox(hx,wi) [P(y)(w’)] ≈ x believes P de re of y ( in w)

(29)

5 (29) hides a crucial semantic component of de re belief. To define de re attitudes, we need the more sufficient formulation in (i), which requires that a suitable “Acquaintance Relation” hold between the subject and the res (Lewis 1979).

(i)

a. b. c. d.

J believe(P)(y)(x)(w) K = 1 iff ∃R such that R(y)(x)(w) R is a suitable Acquaintance Relation ∀ hx’,w’i(hx’,w’i ∈ Dox(x,w) → P(ιz.R(x’)(z)(w’))(w’))

With that in mind, (29) is a shorthand useable for our purposes.

6

A syntactic representation of res arguments and their pronominal associates: #1. Suppose the res argument can be a null pro; it can be bound.6 # 2. Property complement formed by abstracting over the pronoun (see e.g. Chierchia (1989) on de se pronouns, PRO). (30)

Every boy1 ’s mother believed pro1 [CP Cλx [ hex is too old for Mrs. Brown ]]

(31)

a. b. c.

Every boy1 ’s mother believed pro1 that he1 was too old for Mrs. Brown. ∀x[boy(x)(wo )→∀hx’,w’i∈ Dox(x’s mother)(wo )[Too-old-for-Mrs.Brown(x)(w’)] ] ≈ Every boy x’s mother believes in w de re of x that x has the property of being too old for Mrs. Brown.

Base-generated clausal argument: with null operator (Koster 1978; Alrenga 2005). OPP ranges over properties, and the clause is simply lambda’d in. [That he is too old] OPe,st every boy1 ’s father believes pro1 te,st

(32)

You can see in the Appendix that accounts where de re pronouns are bound directly in situ by their “apparent” quantificational binders will not give us what we want here (see, for instance, the “Acquaintance Based Concept Generator” approach in Percus and Sauerland (2003) and possibly the res-phrases used in Anand (2006))).

5 No res, no co-variation 5.1 Clausal complements of nouns We predict that clausal complements of predicates with res arguments, like those in (33) will give rise to co-variation for constituent pronouns without syntactically reconstructing. (33)

a. b. c.

The rumor about John that he is crazy . . . The suspicion about John that he stole the car . . . Any belief about Mary that she has escaped . . .

And indeed, no “reconstruction conflicts” arise in (34-a), which would be expected if the CP argument had to syntactically reconstruct for binding purposes.7 6 Note: You may be familiar with moving terms to the res position. But movement to get transparent interpretations (and what I have been calling de re interpretations of pronouns) is not tenable. That’s why I propose a null res argument. 7 The controls are the following. When the noun and its complement are in situ we find a disjoint reference effect for both of the sentences in (34):

(i)

a. b.

*I don’t think she2 would want any man1 to believe the idea/claim/notion/ suspicion that he1 might be too old for Mrs. Brown2 *I don’t think any man1 would want her1 to believe the idea/claim/notion/ suspicion that he1 might be too old for Mrs. Brown2 .

7

(34)

a. b.

. . . But the idea/claim/notion/suspicion that he1 might be too old for Mrs. Brown2 , I don’t think she2 would want any man1 to believe (that). . . . But the idea/claim/notion/suspicion that he1 might be too old for Mrs. Brown2 , I don’t think any man1 would want her2 to believe (that).

Nouns that do not have res arguments and clausal complements that describe the content of the noun, like article in (35), do not obviate reconstruction conflicts (these are really just the original Lebeaux sentences): (35)

*The article about binding that it is not syntactic.

(36)

a. *. . . But the article that he1 read for Mrs. Brown2 , I don’t think she2 would want any man1 to believe. b. . . . But the idea that he1 might be too old for Mrs. Brown2 , I don’t think she2 would want any man1 to believe.

5.2 How general a strategy? Subject clauses Sentential subjects of causatives (often with psych-verb complements) allow “backward bindng” (see e.g. Larson and Cheung (2008), Pesetsky (1995)) (37)

That Mary would speak to his1 mother didn’t make any boy1 uncomfortable.

• It is unclear just how the pronoun co-varies here (this is in itself a configuration that looks like “apparent binding”; there are syntactic approaches, too) • the “res solution” is not possible here • But! there may be restrictions here that suggest the res solution shouldn’t be expected; we create a conflict in (38-a) (38)

a. *That Mary2 spoke to his1 mother, (that) I don’t think she2 would want uncomfortable. b. That she2 spoke to his1 mother, (that) I don’t think Mary2 would want uncomfortable

to make any boy1 to make any boy1

A necessary control: (39)

That Mary2 spoke to John1 ’s mother, (that) I don’t think she2 would want to make him1 uncomfortable.

8

6 Conclusion • pronouns in fronted clauses can co-vary with respect to a quantificational term without “syntactic reconstruction” • I suggested that this is due to something about complements of attitude predicates, specifically the way res arguments and 3rd person pronouns in attitude complements work, that gives rise to co-variation. • We encountered cases (such as subject clauses) where the de re analysis could not apply. We asked whether this apparent binding had a different source. Multiple res arguments? (40)

a. b.

That she1 would actually agree to marry him2 , I don’t think any boy2 ’s parents expected of any girl1 . That she1 would actually agree to marry him2 , I don’t think any boy2 ’s parents had a clue about any girl1 .

We would have to posit multiple res arguments and multiple abstractions over pronouns.8

Speculation: maybe the res includes more • Res arguments can be situations—Kratzer (1998): the story of the Butler and the Judge • is it those res situations that co-vary with respect to the quantifier? (41)

Every boy1 ’s mother believes that he1 /the boy is skipping class. For every boy, there is a res situation about which that boy’s mother has a belief.

• might this res furnish a situation of the right size and type to pursue a D-type strategy for the pronoun along the lines of Elbourne (2005) and others (Schwarz (2009))? – a D-type pronoun interpreted with respect to the (appropriate counterparts of) the res situation *

References Aloni, M. 2001. Quantification under conceptual covers. Doctoral Dissertation, University of Amsterdam, Amsterdam. Alrenga, Peter. 2005. A sentential subject asymmetry in English and its implications for complement selection. Syntax 8:175–207. Anand, Pranav. 2006. De de se. Doctoral Dissertation, Massachusetts Institute of Technology. Chierchia, Gennaro. 1989. Anaphors and attitudes de se. In Language in context, ed. Johan van Bentham Bartsch, Renate and Peter van Emde Boas, 1–32. Dordrecht: Foris. 8 See

e.g. Cresswell and von Stechow (1982)

9

Chomsky, Noam. 1995. The minimalist program. Cambridge, Massachusetts: MIT Press. Cresswell, M. J., and Arnim von Stechow. 1982. De re belief generalized. Linguistics and Philosophy 5:503–535. Depiante, Marcela A. 2001. On null complement anaphora in Spanish and Italian. Probus 13:193–221. Elbourne, Paul. 2005. On the acquisition of Principle B. Linguistic Inquiry 36:333–365. Fox, Danny. 1999. Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry 30:157– 196. Fox, Danny. 2000. Economy and semantic interpretation. Cambridge, Massachusetts: MIT Press. Fox, Danny. 2002. Antecedent-contained deletion and the copy theory of movement. Linguistic Inquiry 33:63–96. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10:279–326. Grimshaw, Jane. 1982. Subcategorization and grammatical relations. In Subjects and other subjects: Proceedings of the harvard conference on the representation of grammatical relations, ed. Annie Zaenen, 35–55. Indiana University Linguistics Club. Hankamer, Jorge, and Ivan Sag. 1976. Deep and surface anaphora. Linguistic Inquiry 7:391–428. Kaplan, David. 1968. Quantifying in. Synthese XIX:178–214. Koster, Jan. 1978. Why subject sentences don’t exist. In Recent transformational studies in European languages, ed. Samuel Jay Keyser. Cambridge, Massachusetts: MIT Press. Kratzer, Angelika. 1998. Scope or pseudoscope? Are there wide-scope indefinites? In Events and grammar, ed. Susan Rothstein, 163–196. Dordrecht: Kluwer Academic Publishers. Larson, Richard K., and C. Cheung. 2008. “Make” as a triadic unaccusative. Handout of paper presented at the LSA, Chicago, Il, January 2008. Lebeaux, David. 1988. Language acquisition and the form of the grammar. Doctoral Dissertation, University of Massachusetts, Amherst. Lebeaux, David. 1990. Relative clauses, licensing, and the nature of the derivation. In Proceedings of the North East Linguistic Society 20, ed. Juli Carter, Dechaine Rose-Marie, Bill Philip, and Tim Sherer, 318–332. University of Massachusetts at Amherst: Graduate Linguistic Student Association. Lebeaux, David. 1991. Relative clauses, licensing, and the nature of the derivation. In Perspectives on phrase structure, ed. Susan D. Rothstein and Margaret Speas, 209–239. Academic Press. Lewis, David. 1979. Attitudes de dicto and de se. Philosophical Review 88:513–543. Percus, Orin, and Uli Sauerland. 2003. On the lfs of attitude reports. In Proceedings of Sinn and Bedeutung 7, ed. Matthias Weisgerber, 228–242. Konstanz: University of Konstanz, Department of Linguistics. Pesetsky, David. 1995. Zero syntax. Cambridge, Massachusetts: MIT Press. Postal, Paul. 1986. Studies of passive clauses. Albany, N.Y.: SUNY Press. Quine, W. V. O. 1956. Quantifiers and propositional attitudes. The Journal of Philosophy 53:177–187. Reinhart, Tanya. 1983. Anaphora and semantic interpretation. Chicago, Illinois: University of Chicago Press. Romero, Maribel. 1998. Focus and reconstruction effects in wh-phrases. Doctoral Dissertation, University of Massachusetts, Amherst. Schwarz, Florian. 2009. Two types of definites in natural language. Doctoral Dissertation, University of Massachusetts at Amherst. Stechow, Arnim von. 1982. Structured propositions. Report 59 of the Sonderforschungsbereich 99. Webelhuth, Gert. 1992. Principles and parameters of syntactic saturation. Oxford: Oxford University Press. Williams, Edwin. 1981. Argument structure and morphology. The Linguistic Review 1:81–114.

10

7 Appendix: Binding de re pronouns directly One implementation: Percus and Sauerland (2003) (see also Anand (2006), and Aloni (2001)): the res is interpreted in situ, “wrapped” in an acquaintance-based concept-generator. • acquaintance-based conception generator (ABCG): function from individuals to individual concepts (the familiar conditions of suitability are placed on these concepts in being “acquaintance-based”). • Concept generators are represented in the syntax by (null) variables G and these are abstracted over • the complement of an attitude verb is a function from concept generators to propositions.9 (42)

a. b.

Ralph believes that Ortcutt is a spy. that Orctutt is a spy ≈ λG.λw.[ G(Ortcutt)(w) is a spy in w ]

(43)

J believe K = λΓ.λx.λw. there is an acquaintance based concept generator G for x in w s.t. ∀hx’,w’i ∈ Dox(hx,wi) [Γ(G)(w’)]

The crucial piece for our purposes: the res-construed pronoun is bound by its quantificational binder “directly”: Every suitor1 ’s mother thinks that λG.λw’.[G(he1 )(w) is too old in w’ ]

(44)

∀x s.t. x is a suitor in w there is an acquaintance based concept generator G for x’s mother in w s.t. ∀hx’,w’i ∈ Dox(hx’s mother,wi) [(G)(w’) is too old ] PREDICTION: the de re-interpreted pronoun is bound by the quantifier and so we expect the syntax is required to furnish a binding configuration. That prediction is not borne out by the data which shows the existence of co-varying interpretations without reconstruction.

9 Percus

(i)

G is a concept-generator for individual x in w iff a. b.

(ii)

and Sauerland offer the following familiar conditions on Concept Generators.

G is a function from individuals to individual concepts Dom(G) = { z: x is acquainted with z in w}

G is an acquaintance-based concept-generator (AGCG) for an individual x in w iff a. b.

G is a concept-generator for x in w the concepts G yields are “acquaintance-based” in the sense that For all z in Dom(G), there is some acquaintance relation R such that x bears R uniquely to z in w, and for all hx’,w’h in Doxx,w , x’ bear relation R uniquely to G(z)(w’) in w’.

11