Feedback on Assignment 3 This wide-ranging survey contains an impressive amount of investigation, and I have no doubt that your collective understanding of anaphora and ellipsis is immeasurably greater than it was just a couple of weeks ago. Rather than comment section by section, this time I will highlight several areas where there is a pervasive issue that seems to need clarification or further thought. The missing antecedent phenomenon: This phenomenon, originally discovered by Grinder and Postal, and used by them as an argument for the transformational deletion analysis of certain anaphoric processes, involves an anaphor (usually a pronoun) whose antecedent is "missing" because it can only be found inside a deletion site: I've never ridden a camel, but Ivan has [ridden a camel], and it stank. In the elided version (where "ridden a camel" in [] is elided via VPE) the antecedent of the pronoun 'it' is "missing". "Hidden" might be a better term. In certain places in the writeup, it appears that your understanding of this phenomenon is not perfect. (Top of p.4, bottom of p.7, top of p.8) Note that the missing antecedent phenomenon is used to argue that the empty part [] (the *container* of the missing antecedent) is derived via deletion; it is not telling us anything about the 'it', whose antecedent is missing. The Backward Anaphora Constraint: (In surface structure) An anaphor cannot both precede and command its antecedent. While this principle is stated early on in the writeup, there is evidence in several places that it is not understood. For example p.12: ""Since we are assuming the BAC to be true, we know that in order for Sluicing to take place the anaphor must be commanded by its antecedent." The BAC does not say that; and if it did, it would be plain wrong: We know that Harvey was talking with someone; we just don't know who. p.13: "... our assumption that the antecedent is preceding and also c-commanding the DP anaphor 'who' in deep structure." (accompanied by a reference to the BAC. But the BAC doesn't say anything at all about deep structure. p.17: "We've said so far that anaphora have to follow and be c-commanded by the antecedent and not the other way around." The BAC doesn't say that. Let's make sure we understand what it does say. Even more important, let's see if we can test it to destruction. You come tantalizingly close at the bottom of p.17, where you tentatively conclude that, for VPE at least, the linear order doesn't matter at all, but only command. In effect you are testing whether a configuration like (6) on p.2 results in grammaticality, and you find that it does not. These configurations are devilishly hard to construct, but your examples (11)-(14) are about as close as you can get. When we do succeed in constructing a case where an anaphor commands a preceding potential antecedent, the result is that the anaphoric relation is impossible. How about this for an improved BAC? (In surface structure) An anaphor cannot command its antecedent. We'll talk about preceding a little bit later. (Un)boundedness: There is really only one test for (un)boundedness: A relation between A and B is unbounded if there can be indefinitely many clause boundaries between them. Since here we are interested in anaphora, we just want to know whether the anaphor and the antecedent can be separated by any number of clause boundaries. So the following does not make any sense: p.12: "Sluicing, by definition, ought to be unbounded, as it relies on WH-movement, which we know to be an unbounded movement." What is of interest to us at this point is whether the elided TP (out of which the WH element moved) can be unboundedly far away from the identical TP that serves as its antecedent. You see that has nothing to do with the unboundedness of the WH-movement. On p.34, (vii) "Philip loves to try to imply that Jacob seems shady, although very rarely" is given as evidence that Stripping is unbounded; but look at where the antecedent is, and where the anaphor is. How many clause boundaries are there in between? The same question can be posed about (xii) on p.35. If you want to see whether Stripping is unbounded, you have to put the stripped clause in an embedded clause inside another embedded clause: Mary drinks gin, but never on Sundays. *Mary drinks gin, but it looks like she would like us to believe (that) never on Sundays. Comparative Ellipsis can also be tested in the same way: Sue has collected more bottle caps than Mary. Sue has collected more bottle caps than marbles. *Sue has collected more bottle caps than Harvey said Mary. *Sue has collected more bottle caps than Harvey said marbles. Alongside Gapping, which you correctly identified as bounded (p.11), these two ellipses (Stripping, Comparative Ellipsis) constitute a distinct type, as opposed to VPE, NPE, and Sluicing, all of which are unbounded. We should talk about this. If you look carefully, you will find that there is another property that distinguishes Gapping, Stripping, and Comparative ellipsis from the long-distance ellipses: these three never go backward, while the others all do. But there is a constraint on backward ellipsis, even for the long-distance ones. Try to do them backward in a coordinate structure. SPECIFIC COMMENTS: p.4: "I've never gone to bed before midnight, but Chloe did __, and she said it was wonderful." This is not exactly the missing antecedent phenomenon of Postal and Grinder, though it looks like an interesting variant. Here the "missing" antecedent is not *contained in* the ellipsis site, it *is* the elided material. This kind of example provides evidence that there is a silent VP, but it does not provide evidence that that silent VP has internal structure. passim: "in regards to" is still a ghastly spelling error. You did it 3 times. p.5: You imply that Sentential 'it' can't be pragmatically controlled. Is that correct? p.8: (2) Andrew is happy and Travis sad. -- is perfectly grammatical. pp 10-11: Nice investigation of Gapping, but you didn't try to find out if it goes backward. (26) is grammatical. Gapping is bounded, but not *clause* bounded. p 14: Wrong punctuation around 'however'. "... all backward sluicing takes place in a but-clause ..." ?? Someone who knows how should teach Harvey to wiggle his ears. p 15: ... the ungrammaticality of (5) suggests that VP ellipsis is *surface* anaphora ... 'elided' has one l. p 16: "Adjuncts can't be part of the antecedent VP ..." I didn't go to the store on a motorcycle yesterday, but Pete did. "English seems to like having its T spelled out in some fashion." What about Gapping? Stripping? Comparative Ellipsis? Could there be an interesting typological difference here? p 18: "The grammaticality judgments are questionable here, especially with (15), which is crucial to this distribution working." Interestingly, there is no (15). In your formulation of VP Deletion, the first requirement is that the VP to be elided must be c-commanded by an identical VP. But in your very tree illustrating VPE in (1), that requirement is not satisfied. p 20: "...'my', ultimately becoming 'my',..." You mean 'mine'? p 21: 'This', 'that', 'these', and 'those' introduce a confound because each is both a D that takes an NP complement and a deictic pronoun that stands on its own. Nice observations. It looks like NPE, like VPE, has licensors. p 24: The comparative construction poses an interesting puzzle: clearly 'more' selects the 'than' CP; but look at the structure: how can 'more' select 'than'? We should talk about what the comparative construction suggests to us about the DegP/AP question. p 29: Nice little excursus on Pseudogapping. At bottom: is 'eat' the kind of V that can move to T? p 30: Excellent basic investigation of the Comparative construction; but the testing of its properties needs work. You haven't tested Comparative Ellipsis for boundedness. Do that. In your tests for backwardness, you're deleting stuff that's not in the comparative clause! Of course that won't work. Let's try More girls than like Harvey like Mike. *More girls than Harvey like Mike. I don't even understand the examples that are supposed to show something about island constraints. Julie ate more pears than Jacob thought Shawna said Phil ate. Julie ate more pears than Jacob thought Shawna said Phil did. Julie ate more pears than Jacob thought Shawna said Phil did apples. *Julie ate more pears than Jacob thought Shawna said Phil. p 31: "...A-B = the intersection of the complement of A with B ..." Surely you mean the intersection of the complement of B with A? -- But it's not really *sets*, is it? "... this transformation cares about both identity *and* coreference." Interesting observation. Are there other ellipsis processes like that? p 32: [Top] Is it possible that 'always' and 'never' are not involved in the ellipsis? p 34: What is neg raising? We have a few things to talk about. If you are interested in a follow-up, take a look at Sag & Hankamer 1984, which I will post on the course web page.