<<

Comments on ‘Explaining crossover and superiority as left-to-right evaluation’ by Chung-Chieh Shan and Chris Barker The disunity of crossover phenomena Christopher Potts UC Santa Cruz and UMass, Amherst [email protected] (1) “Our analysis rules out strong and weak crossover in a single stroke.” (Shan and Barker 2003:5) These notes gather together evidence that crossover is not a unified phenomenon. I conclude from this that it shouldn’t be analyzed uniformly. I then highlight some adjustments that seem necessary to account for the disunity in Barker and Shan’s terms.

1 A crossover typology

The boxed/unboxed division is one that should, I argue, be reflected in the grammar. All the examples are deviant to some degree; I provide no judgments in (2). (2) primary secondary extracted element extracted element is the binder properly contains the binder

Whosei teacher did strong Whoi did hei admire most? hei admire most? crossed not embedded in an NP Hei admires everyonei. Hei admires every studenti’s professor.

Whoi did hisi teacher Whosei teacher did weak admire most? hisi friend admire most? crossed pronoun embedded in an NP Hisi assistant contacted Hisi dog bit every professori. every studenti’s roommate.

Weakest crossover? Lasnik and Stowell 1991 + Postal 1993 ⇒ weakest crossover is where no crossover effect exists even though the surface order might have led you to think there would be one.

1.1 Summary of Barker and Shan’s unified analysis of (2) strong = weak In derivations, his mother is (MB(his))(L(mother)), which is of the same category as he, namely, (e ∼ A) → (e B A). Hence, he and his mother interact with binders in the same way. primary Wh = secondary Wh In derivations, whose mother is (MB(BIND(whose)))(L(friend)) when whose is a binder, which is of the same category as BIND(who), namely, (e ∼ (e B A)) → (e ? A). Hence, both who and whose mother act as binders in the same way (via Q). primary quantified = secondary quantified In everyone’s mother, everyone has the same possibilities as everyone when it is a clausal argument.

1 Brown Workshop on Direct Compositionality, June 19–21, 2003 Comments on Barker (and Shan)

2 Disunity

2.1 Contrasting judgments (3) Wasow 1979:158: “The sentences in (7), however, demonstrate that [. . . ] a differentiation is necessary, for al- though weakly crossed sentences would violate a cross-over constraint, they are far less deviant that strongly crossed sentences. ((7a) is from Remembered Death by Agatha Christine, Pocket Books, p. 58; [. . . ]).

(7a) He was the type of man with whom his work would always come first.”

2.2 Repair effects The crossover repair particles even, only, own, and emphatic reflexives can repair both kinds of weak crossover as well as secondary strong crossover. They cannot repair primary strong crossover:

∗ (4) a. the lawyer who1 his1 clients hate t1

b. the lawyer who1 (even) his1 (own) clients hate t1

∗ (5) a. Sally is the artist [whose1 work]2 she1 has criticized t2.

b. Sally is the artist [whose1 work]2 she1 herself1 has criticized t2.

∗ (6) Sally is the artist who1 they say she1 (herself) criticized. The repair effects also point up differences between overt and covert instances of crossover: (7) Postal 1993:549–550, footnote 14 Repair effects “partially undermine a view about the WCO effect that is widespread in the GB literature. This assumes that there is a genuine parallelism between say, (ia) and (ib), so that (ib) also instantiates the WCO effect [. . . ].

∗ (i) a. Which man1 did his1 children dislike t1? ∗ b. His1 children dislike no/every man1. [. . . ] However, the parallelism breaks down when even or only is added:

(iii) a. Which man1 did even/only his1 children dislike t1? ∗ b. Even/Only his1 children dislike no/every man1. This argues that the similarity between (ia) and (b) fails to represent a genuine generalization.”

(8) Shan and Barker 2003:2, footnote 1: “Postal (1993) (see also the discussion in Potts 2001) points out that crossover interpretations sometimes become better if own or only is added: ?A student of hisi own advisor contacted every professori. This amelioration effect is poorly understood, and we will not discuss it further here.”

2 Brown Workshop on Direct Compositionality, June 19–21, 2003 Comments on Barker (and Shan)

3 Towards a more fine-grained theory of crossover

The analysis of Shan and Barker 2003 delivers complete ungrammaticality for all crossover cases. This seems appropriate only for primary Wh- strong crossover and for the full range of covert movement cases. For the boxed parts of (2) (= Wh- secondary strong crossover and Wh- primary and secondary weak crossover), we have shades of deviance, suggesting that a less stringent account is in order.

Two approaches: (9) a. We want to generate the boxed sections of (2) and explain why they sound iffy (to some) in the absence of repair items. All other kinds of crossover should be blocked. b. We want to block all of (2), and then use other mechanisms to explain (i) why the unboxed sections are perceived as worse (Shan and Barker 2003:5); and also (ii) how the repair items save the boxed parts. Since Wh- weak crossover cases appear in naturally occurring texts, and since the judgments for them have always been iffy (one finds all of ∗, ??, and ? in the literature), it might be wise to favor (9a).

But the repair particles are special: (10) a. own and emphatic reflexives create , i.e., pronouns that appear mainly in in- direct and must correspond to indexicals in direct quotation (see, e.g., Schlenker 2003:§4.1.1). b. even and only turn their -associates into quantifiers (see, e.g., Buring¨ and Hartmann 2001). These factors might indicate that we should block all the cases in (2), but then assign the repair particles a special that lets them avoid crossover violations. How?

A suggestion: Controlled W The repair particles’ categories could be lexically-controlled versions of the type- rule W that enforces right-to-left evaluation instead of left-to-right evaluation (see Shan and Barker 2003:§7.1). It might prove hard to do this locally, though. One might be forced to say that a repair particles just flips us from M(ETA) to W for the whole of the derivation. That kind of condition requires a whole new style of grammar, though.

Another suggestion: Another BINDer Shan and Barker (2003:22) write that Wh- primary weak crossover cases have no derivations because

(11) “(BIND who) has category (e ∼ (e B A)) → (e ? A). Note that the e corresponding to the trace is outside the e corresponding to the pronoun (i.e., the ∼ connective is outside the B connective). This means that the trace must always outscope the pronoun [. . . ]”. Thus, we could define a second bind operator, one that flips the order of the trace and the pronoun (with no change to the semantics). This would have to be accompanied by some other adjustments to the type shifting rules: for instance, where ∼ appears in META, we would have to allow ∼ or B. In the end, this might undermine the distinguished status that pronouns have in the theory of categories.

3 Brown Workshop on Direct Compositionality, June 19–21, 2003 Comments on Barker (and Shan)

But Both solutions still leave the problem of controlling the repair particles when neither the pro- noun nor the binder is embedded — i.e., we need to avoid predicting that primary strong crossover is repairable. I believe that the accounts of quantification and Wh- extraction are sufficiently different from each other that we can allow Wh- extraction some leeway without losing the ability to block all the quantified cases.

In sum At a minimum, we need to distinguish who from whose mother and he from his mother in the theory of categories. Once we’ve done this, we’ll have the tools to get a grip on the contrasting intuitions that Wasow (1979) first noted. We’ll also be well positioned to understand how and where the repair particles work.

References

Buring,¨ Daniel and Katharina Hartmann. 2001. The and semantics of focus-sensitive particles in German. Natural Language and Linguistic Theory 19(2):229–281.

Lasnik, Howard and Tim Stowell. 1991. Weakest crossover. Linguistic Inquiry 22(4):687–720.

Postal, Paul M. 1993. Remarks on weak crossover effects. Linguistic Inquiry 24(3):539–556.

Potts, Christopher. 2001. (Only) some weak crossover effects repaired. Snippets 1(3).

Schlenker, Philippe. 2003. A plea for monsters. Linguistics and Philosophy 26(1):29–120.

Shan, Chung-chieh and Chris Barker. 2003. Explaining crossover and superiority as left-to-right evaluation, Ms., Harvard and UC San Diego.

Wasow, Thomas. 1979. in . Ghent: E. Story-Scientia.

4