Nov 30/04, CSC 247/447 Scoping preference examples ~~~~~~~~~~~~~~~~~~~~~~~~~~~ (a) Everyone loves someone every(1) > some(2) ~~~(1)~~ ~~(2)~~ subj(1) > obj(2) So both preferences favor the "every-some" reading, (Every x: (one x) (some y: (one y) (loves x y))) (b) Someone loves everyone every(2) > some(1) ~~(1)~~ ~~~(2)~~ subj(1) > obj(2) Conflicting tendencies, so both readings should be plausible. (c) A dog entered with every man every(2) > a(1) ~(1)~ ~~~(2)~~~ subj(1) > post-V-advl(2) Conflicting preferences, so forall-exists and exists-forall are viable (d) A fat dog always loses the race the(3) > a(1), all(2) ~~~(1)~~~ ~(2)~~ ~~(3)~~~ subj(1) > obj(3) Conflicting preferences, so we expect at least (3)(1)(2) and (1)(3)(2). But these 2 possibilities are logically indistinguishable, and they give (1) wider scope than (2), while there seems to be a reading where (2) has wider scope than (1) (different fat dogs at various times). Perhaps "always" should be treated like a pre-S modifier, or more like "each" and "every" (rather than "all") in its wide-scoping tendency. (e) Every man lifted the piano the(2) > every(1) ~~~(1)~~~ ~~~(2)~~~ subj(1) > obj(2) Conflicting; however there are two more pertinent points here: "Every man" can just barely get a *collective* reading (though it would work better with "everyone"; cf., everyone gathered in the hallway); and second, the scoping of "the" can in general be narrow relative to a universal only if there is some semantic or pragmatic reason to think the definite *depends* on the universal (e.g., "Every school sent the principal as a representative") ~~~~~~~~~~~~ ~~~~~~~~~~~~~ (f) A representative of every company was present every(2) > a(1) ~~~~~~(1)~~~~~~~ ~~~~~(2)~~~~~ post-nom.PP(2) > nom(1) The preferences are consistent, so we expect only one natural reading, the forall-exists reading. (g) The representative of every company was present the(1) > every(2) ~~~~~~~(1)~~~~~~~~ ~~~~~(2)~~~~~ post-nom.PP(2) > nom(1) The preferences conflict, so both readings should be possible -- *provided* that the representative can reasonably be thought of as dependent on the company, which of course s/he can. (h) Every man in a boat was serenading a woman every(1) > a(2), a(3) ~~~(1)~~~ ~~(2)~ ~~(3)~~ subj(1) > obj(3) post-nom.PP(2) > nom(1) Conflicting preferences, predicting (1)(2,3), (2)(1)(3) readings (order or (2,3) doesn't matter if adjacent). I also get (2,3)(1), and (3)(1)(2) readings, because the corresponding situations are quite imaginable: a woman serenaded by a floating orchestra, in one boat or in multiple boats (a balcony scene in Venice?) (i) When does each plane fly? each(2) > wh(1) ~(1)~ ~~~(2)~~~~ subj(2) > post-V-advl(1) I'm assuming that "when" is a post-V-advl, i.e., it has been extracted from that position. This is consistent, and correctly predicts that we prefer the forall-when reading (asking for various departure times). (j) Several hikers were surprised when a bear appeared. several(1) = a(2) ~~~~~~(1)~~~~~ ~~(2)~ subj(1) > post-V-advl(2) Note that "when a bear appeared" is a scope island -- and this also predicts narrow scope for "a" -- though "a" easily escapes from scope islands. So the preferences predict a preference for the reading where a different bear may be involved for different hikers. (k) Everyone thought that Fido or Fifi would win every(1) = or(2) ~~~(1)~~ (2) subj(1) > obj(2) Again there's also a weak scope island effect ("or" can escape from islands). The predicted wide-scoping of (1) seems correct. For tense, which we've ignored, assume that it takes wide scope in both the main and the subordinate clause, but that the tense of the subordinate clause cannot escape from the scope island. (l) Report the flights that each man took the(1) > each(2) ~~~~(1)~~~~ ~~(2)~~~ post-nom-mod(2) > nom(1)?? Oddly (2) can take wide scope despite the preference the(1) > each(2), and *despite the scope island* formed by the relative clause. So I have indicated a generalization of the "post-nom-PP > nom" rule that might account for this. (m) A man representing each company showed up each(2) > a(1) ~(1)~ ~~~~~(2)~~~~ post-nom-mod(2) > nom(1)?? Here, in contrast with (l), a wide-scope reading for "each" seems hard to get, even though it is the most natural in terms of world knowledge, and the above preferences favor it. For reasons unknown, there seems to be a powerful scope island effect here. (n) He will feed the hungriest dog. pres(1) = the(2) ~(1)~ ~~~~~~(2)~~~~~~~~ Note: "will" is present tense (logically, ). If modals are sentence-level operators (as seems reasonable to assume in the semantics), then they ought to form scope islands. Then (2) lies in this scope island, but being definite, can escape. So we get a rading where we're talking about "the presently hungriest dog", in addition to the (island-bound) reading where we're talking about the hungriest dog at a certain future time. (o) Everyone didn't leave LF: ~(past (leave ) ~~~(1)~~ Negation is semantically a sentence-level operator and forms a scope island, so we prefer a "not-every" reading to an "every-not" reading (i.e., no-one left). (p) Someone didn't leave LF: ~(past (leave ) ~~(1)~~ "Some" can escape from the scope island. This explains why the "some-not" reading is possible -- though it doesn't explain why it's rather strongly preferred!