REQUERIMIENTOS HARDWARE

There are some empirical problems with Wagner’s analysis which cast doubt on whether de-accenting should be treated as a unified phenomenon. First, and this is perhaps a tech- nical point, the caveat that accent shift at the sentence level is constrained by mutual ex- clusivity under an exhaustivity operator breaks down in cases where the antecedent that is taken to license the accent shift is entailed by the sentence in question. Secondly, and more importantly, the assumption that objects move to a propositional node at LF, effectively weakening the contrast requirement for de-accenting, is problematic for three reasons: 1) de-accenting within DPs with no propositional node is acceptable in contexts with no ex- plicitly contrastive antecedent, 2) apparent Givenness-marking does not obey island con- straints, and 3) no specific predicate of a de-accented direct object needs to be mentioned or implied in discourse.

Let’s address these issues in turn, beginning with mutual exclusivity under exhaustivity. The Exh operator is semantically the same as only—Exh(a)(b) is true so long as any

statement[ab�_{]that is also true is entailed by [ab]. For example, the sentenceonly John and} Mary went swimmingis true when, modulodomain restriction, any other true statements about who was swimming (i.e.John went swimmingandMary went swimming) are entailed by the operand of [[only]]. Consider again example (6), repeated below.

(8) Mary wentSWIMMING. After that, JANEwent swimming.

The application of this operator to a proposition like Jane went swimming certainly does exclude the antecedent Mary went swimming; however, we can manipulate the example

so that the antecedent is entailed by the accent-shifting sentence, in which case mutual exclusivity underExhno longer holds.

(9) Mary wentSWIMMING. In fact,EVERYBODYwent swimming.

If we apply Exh to the de-accenting sentence we get a denotation that is true as long as any contextually relevant statement of the form ‘x went swimming’ is entailed byev- erybody went swimming. This is true of all possible alternatives, and thus this condition is vacuous—the application ofExhto ‘everybody went swimming’ is truth-conditionally equivalent to ‘everybody went swimming’. Of course, everybody going swimming cannot entail that there is somebody (Mary) who is not swimming. Therefore, applyingExhdoes not exclude the antecedent in (9) or in any other similar example.

Is there some other way in which an example like (9) represents MEC? It is worth considering the possibility of contrast between events in a neo-Davidsonian event seman- tics (Davidson, 1967; Parsons, 1990). If the subject everybody can be interpreted as a collective Agent of a swimming event, then the two sentences in (9) encode different events with different Agents, in which case the UniClo of an event-based denotation, ∀e.swim(e) &past(e) &Agent(everybody, e), excludes its antecedent without any refer- ence toExh. However, this is not fully generalizable, as evidenced by the following. (10) I don’t think the city is safe anymore. My COUSIN lives in the city, and she says

it’s deteriorating fast.

Though mentioning unsafe conditions in a city makes it salient that there are people who live in that city, allowing Givenness, there is no specific event or state evoked by the con- text, and thus no Experiencer to be excluded by the neo-Davidsonian UniClo ofmy cousin lives in the city. In other words, (10) cannot be an instance of contrast between events be- cause no specific event is invoked in the discourse which is distinct from and excluded by λe.live(e) &pres(e) &in(e, city) &Experiencer(cousin, e). The easiest way to account for examples of this type is simple Givenness with no contrast requirement.

There is another problem with unifying Givenness, Focus and Contrast: one must posit LF-movement of G-marked objects, and this makes false predictions regarding de- accenting within DPs and islands. Recall example (7), repeated below.

(11) Smith got away from the scene of the crime in Mary’s cheap convertible. Q: Then what happened?

A: The car broke down, and a detectiveARRESTEDSmith.

Here, it is said that Smith moves to a propositional node at LF, partitioning the semantic structure into the moved objectSmithandλx.past(arrest(detective, x)). Wagner gives

the following example as further support for the LF movement approach, reasoning that DPs should disallow the broader pragmatic license for de-accenting in (12) because there is no propositional node within the DP for the de-accented element to move to.

(12) You should hire a DJ. #The PRESENCE of a DJ makes a big DIFFERENCE at a party.

The mere salience of DJs in the context is not a sufficient license for accent shift, and thus the example seems not to be parallel to (11). However, there is an issue with this particular case. The nounpresenceis semantically rather empty, and in almost any context meets the criteria of Schwarzschildian Givenness. We expect such a word to resist accent a priori. A better test case is one where the accent shifts within a DP to a word with non-Given semantic content. Below is a such an example, showing that accent shift within a DP can be acceptable without a mutually exclusive antecedent.

(13) My mother asked if we were moving to the city. I told her that the VIOLENCE in the city is aTURN-OFF.

Here we have a clear case of de-accenting within a DP (citywould bear some accent if its denotation were not salient), and there is no mutually exclusive antecedent forviolence in the city. At the very least, it is not clear that DPs can be used as evidence for LF movement. More likely, cases like (13) are instances of Givenness with no contrast requirement.

It is also possible to de-accent within islands. The following illustrates de-accenting of an element that, were it to move, would violate the Coordinate Structure Constraint (Ross, 1967).

(14) a. *[PP From which store ] did you buy a necklacetPPand a belt from Macy’s? b. Oh, you went to Sak’s? Yesterday I bought a NECKLACE from Sak’s and a

We could construct similar examples for other island constraints, but Wagner (2012, p.27) points out that other islands have propositional nodes, allowing movement within them rather than out of them, making these cases inconclusive under his analysis. The felicity of (14) is conclusive, however, given that the PP should not be able to move at all. Anin situ

analysis of this sentence predicts an antecedent of the form λx.P(x) & f rom(Saks)(x)

whose existential closure is made false by the universal closure (UniClo) of the first con- junct NP,∀x.necklace(x) &f rom(Saks)(x). This prediction is not borne out. Even if we

allow the possibility that only the DPSak’sis de-accented rather than the whole PP—and we should, since prepositions resist accent regardless of information structure—we are still left with a false prediction.

(15) a. *[DP Which store ] did you buy a necklace fromtDPand a belt from Macy’s? b. Oh, you went to Sak’s? Yesterday I bought a NECKLACE from Sak’s and a

BELT from MACY’S.

Under this configuration, the illegality of the DP movement to a propositional node also re- quires anin situanalysis, which should require an element in the prior discourse to contrast with ‘necklace’. Such a contrast does not exist in (15-b).

Finally, let’s take a closer look at examples like (5)/(16). These seem on the surface to suggest that some form of MEC is necessary to shift accent onto the verb.

(16) a. Q: I heard that somebody shot Smith, and that he’s been recovering in the hospital. Is he OK now?

A: Actually, something bad happened again. You’ll never guess: Someone shot SMITH! / #SomeoneSHOTSmith!

b. Q: I heard that somebody shot Smith, and that he’s been recovering in the hospital. Is he OK now?

A: Actually, something bad happened again. You’ll never guess: Someone

STABBEDSmith! / #Someone stabbed SMITH!

(after Wagner 2012, p.15)

In (16-a) there is no contrast under Wagner’s analysis, because the clausal sister to the LF- moved object Smithand its potential antecedent are identical—λx.∃y.past(stab(y, x))is

not mutually exclusive with itself. But this identity relation offers independent explana- tion for the infelicity of de-accenting. In this case, to de-accent would be to shift accent from one Given constituent to another Given constituent. This is not motivated, and since

one cannot de-accent an entire utterance, an entirely Given utterance results in default, right-edge prosody. The following illustrates that in straightforward cases of all-Given utterances, prominence falls in the same place as it would were the utterance entirely new. (17) Q: You got a new computer?

A: Indeed, I got a newCOMPUTER.

The true test case is one where the verb is not Given and also lacks any contrastive an- tecedent. Contra Wagner, felicity in such cases is possible. Recall the phenomenon of de-accenting in conversation starters.

(18) Context: Pat is reading a book about castles in Germany. Chris walks in, sees Pat, and utters the following sentence out of the blue in order to engage Pat.

I’ve neverBEENto Germany. Have you?

(19) Context: June comes home to find her roommate watching a documentary about Leo Tolstoy. June sits down next to her roommate and utters the following sen- tence.

My great-grandfather wasFRIENDSwith Tolstoy.

(20) Context: Driving on the interstate, passing a road sign reading “Dayton, OH”, a passenger utters the following to the driver.

I used toLIVEin Dayton.

The evidence against LF-movement, the inconsistent presence of Contrast at the proposi- tional level, and the ability of constituents to de-accent when salient in the non-linguistic context with no prior discourse, when taken together, all point toward a simple Givenness- based account. This requires a separate notion of Focus to explain cases where some form of contrast is in fact required. The next important question, then, is this: when can de- accenting be the result of Givenness, and when must it be a reflex of Focus? The short answer is that Focus is the only route to de-accenting in English (and thus a wh-question or explicit contrast is required) when accent is shifting onto an adjunct. In 3.3 I analyze this distribution as being a consequence of feature projection, an analysis which I extend to German in Chapter 4. But first, I give evidence for the adjunct generalization in English.