Abstract

We argue for the existence of covert focus movement in English focus association with only. Our evidence comes from Tanglewood configurations of the form in Kratzer 1991. We show that Tanglewood configurations are sensitive to syntactic islands, contrary to Kratzer’s claims and predictions. We propose that Tanglewood configurations always involve covert movement of the focused constituent—possibly with covert pied-piping—to bind a bound variable in the ellipsis site. This availability of covert pied-piping explains examples such as Kratzer’s where the Tanglewood construction appears to be island-insensitive. We show that covert focus movement is long-distance and not simply Quantifier Raising. Kratzer’s proposal that ellipsis enforces the identity of focus indices and several other previous approaches are shown to overgenerate Tanglewood readings.

This article studies the mechanism of association with focus in English: in particular, the relationship between the focus-sensitive adverb only and the associating focused constituent in its scope. We begin by briefly introducing the influential analysis of association with focus in Rooth 1985, 1992 and discussing the problem posed by Kratzer’s (1991) famous Tanglewood constructions. We then present our proposal for explaining these constructions via covert focus movement, and we offer evidence supporting our proposal from island sensitivity and Tanglewood readings with overt bound variables. We discuss previous alternative analyses of Tanglewood readings and argue that they cannot explain the facts we present here. Finally, we show that covert focus movement can be long-distance, arguing that its effects cannot be reduced to Quantifier Raising.

1 The Problem of Tanglewood

Focused constituents, indicated by F-marking, are pronounced with prosodic prominence. Semantically, they introduce a set of alternatives into the computation. Focus-sensitive operators such as only then quantify over those alternatives.

(1) I only wear [red]F shirts.

  • a.

    Alternatives to red: green, blue, . . .

  • b.

    Presupposition: I wear red shirts.

  • c.

    Assertion: I do not wear green shirts, I do not wear blue shirts, . . .

Under the Alternative Semantics theory of focus in Rooth 1985, 1992—which continues to be the most widely adopted theory of association with focus—each syntactic node α has two “dimensions” of meaning: an ordinary semantic value

graphic
α
graphic
o as well as a focus semantic value
graphic
α
graphic
f, which can be thought of as a set of alternative denotations and which includes
graphic
α
graphic
o as a member. Focus semantic values for complex phrases are computed compositionally using the meanings of their parts, just as ordinary semantic values are.

(2) Recursive definition for focus semantic values (Rooth 1985:14)1 The focus semantic value of node α,

graphic
α
graphic
f, is
  • a.

    the set of objects in the model matching

    graphic
    α
    graphic
    o in type, if α bears the feature F;

  • b.

    the unit set {

    graphic
    α
    graphic
    o}, if α is a non-focused non-complex phrase;

  • c.

    the set of objects which can be obtained by picking one element from each of the focus semantic values corresponding to the component phrases of α, and applying the semantic rule for α to this sequence of elements, if α is a non-focused complex phrase.

Rooth proposes that focused constituents such as red in (1) are interpreted in their pronounced position at LF. Following the procedure in (2), the alternatives introduced locally (3a) will be reflected in the focus semantic values of all dominating phrases, resulting in a corresponding set of propositional alternatives (3b) in the complement of focus-sensitive operators.2Only α then asserts the negation of all alternatives in

graphic
α
graphic
f that do not entail the prejacent proposition
graphic
α
graphic
o (see, e.g., Horn 1969); this results in the correct assertive content as in (1).

(3) LF: only [VP I wear [red]F shirts]

  • a.

    graphic
    red
    graphic
    f = {red, green, blue, . . . }

  • b.

    graphic
    VP
    graphic
    f = {I wear red shirts, I wear green shirts, I wear blue shirts, . . . }

Empirically, this article centers around the Tanglewood phenomenon first discussed in Kratzer 1991. Tanglewood examples were introduced as a challenge to the basic Roothian theory sketched above, motivating a minor but significant refinement to the theory. Kratzer’s original example is given in (4).

(4)

  • Tanglewood (Kratzer 1991:830)

  • Context: Imagine now you are angry at me and start voicing the following accusations. “What a copycat you are! You went to Block Island because I did. You went to Elk Lake Lodge because I did. And you went to Tanglewood because I did.” I feel you exaggerate and reply:

  • I only went to [Tanglewood]F because you did

    graphic
    .

(5) Paraphrase: Tanglewood is the only place x such that I went to x because you went to x.

Kratzer’s observation is as follows. Considering the interpretation of the ellipsis site in (4), indicated by

graphic
, let us assume the LF structure for (4) to be as in (6). Now notice that (6) includes two instances of the F-marked constituent Tanglewood. According to Rooth’s definition for focus semantic values in (2), the result will include all combinations of different values for the two positions of focus, as in (7a). The assertion of only in (4) is then predicted to be as in (7b).

(6)

graphic

(7)

graphic

Kratzer argues that this predicted assertion in (7c) does not reflect the actual interpretation of example (4). As the paraphrase in (5) indicates, the correct interpretation asserts only that it is not the case that I went to Block Island because you went to Block Island and it is not the case that I went to Elk Lake Lodge because you went to Elk Lake Lodge. In other words, the set of alternatives must be computed so that the alternatives in the two positions of focus covary across the alternatives, as in (8). We will refer to such interpretations that require such covarying alternatives under an in-situ approach to focus as Tanglewood constructions or Tanglewood readings.

(8)

graphic

Kratzer proposes an amendment to Rooth’s theory that allows for the natural derivation of covarying alternatives as in (8). In brief, Kratzer proposes that focused constituents bear distinguished focus indices and that ellipsis ensures their identity (9), resulting in the LF structure in (10a). Focused constituents are then interpreted as distinguished variables in the focus semantic value, ranging over different assignment functions h (10b).3 This yields the desired covarying alternatives in (10c).4

(9)

  • Assumed LF identity condition on VP-ellipsis with focus indices5

  • A VP (VPE) can elide just in case the linguistic context provides an antecedent VP (VPA), such that for any ordinary and distinguished variable assignments g and h,

    graphic
    VPE
    graphic
    g,h =
    graphic
    VPA
    graphic
    g,h

(10)

graphic

We make two notes here regarding Kratzer’s (1991) theory. First, Kratzer retains from Rooth’s work (a) the idea of a multidimensional semantics, with ordinary and focus semantic values, and (b) the claim that foci are interpreted in situ at LF. Her Tanglewood argument challenges how focus semantic values are computed, motivating her focus index approach over Rooth’s recursive procedure in (2). She also briefly considers and argues against an alternative account where the focused constituent covertly moves; we will detail this approach and her argument against it in section 2.

Second, we note that Kratzer’s proposal that ellipsis can enforce the identity of focus indices is quite powerful. In particular, it predicts no locality restrictions between the focus-sensitive operator (only), the pronounced focus, and the ellipsis site. As long as the pronounced focus and its interpreted copy in the ellipsis site are both in the scope of the focus-sensitive operator, the Tanglewood effect is predicted: the operator will quantify over alternatives where the two focused positions covary.

In this article, we present previously unobserved restrictions on the distribution of Tanglewood readings that earlier accounts of the phenomenon do not predict. We concentrate first on Kratzer’s account, as the most widely known account of Tanglewood readings, and discuss alternative accounts in section 5. Of particular importance are two findings: (a) that Tanglewood readings exhibit sensitivity to syntactic islands, and (b) that Tanglewood readings are possible in the absence of ellipsis.

In section 2, we present our own proposal for Tanglewood constructions. We maintain the Roothian multidimensional semantics for the computation of alternatives, but diverge from Rooth and Kratzer in arguing that foci are not interpreted in situ when associating with only: they move covertly to the higher operator, and it is this movement that makes Tanglewood readings possible. In sections 36, we present our new evidence that motivates this approach, discuss the nature of the movement, and examine alternative accounts.

2 Proposal

We propose that Tanglewood constructions such as (4) always involve covert movement of the focused constituent to a position from which it binds a bound variable in the ellipsis site.

We first illustrate a basic example of association with the English adverb only using covert focus movement; see (11).6 For concreteness, we adopt the form of covert focus movement discussed briefly in Rooth 1985:31–32 and used in Wagner 2006.7 This involves covert movement of a constituent containing the focus to a complement position of the attracting operator—also called Undermerge by Pesetsky (2007, 2013)—together with adjunction of the associated λ-binder to the complement from which the focused constituent is moved out.8

(11)

graphic

Let (12) be the semantics of this two-place only, based on Horn’s (1969) classic description of only’s meaning: only presupposes the truth of its prejacent (the combination of its first and second arguments) and asserts that, for all alternatives to the first argument in set C, if it is not equal to the stated (prejacent) value of the first argument, its combination with the second argument must be false.9 The variable C must be fixed contextually to be equal to (or a subset of) the focus semantic value of the first argument of only at LF (see, e.g., discussion in Rooth 1992, Tancredi 2004, Wagner 2006). Here we let C =

graphic
[Tanglewood]F
graphic
f = {Tanglewood, Block Island, Elk Lake Lodge}. The resulting interpretation of (11) is given in (13).

(12)

graphic

(13)

graphic

We now demonstrate how this covert focus movement helps derive the Tanglewood reading in Kratzer’s original example, (4). We propose that the overt focus Tanglewood moves covertly to become the first argument of only, leaving the variable x in its trace position with a corresponding λ-binder. In the ellipsis site, we have a matching bound variable therex , which will also be bound by the same λ-binder. Ellipsis is licensed by the LF identity condition in (15). This yields the correct interpretation for the Tanglewood example (4).10

(14)

graphic

(15)

  • LF identity condition on VP-ellipsis

  • A VP (VPE) can elide just in case the linguistic context provides an antecedent VP (VPA), such that for any variable assignment g,

    graphic
    VPE
    graphic
    g =
    graphic
    VPA
    graphic
    g.

This approach ensures quantification over propositions with the same values in the position of pronounced focus and within the ellipsis site through general mechanisms of movement, abstraction, and variable binding. This takes away the need to generate alternatives that covary in two positions of focus, discussed in section 1. Note that, under this approach, ellipsis is not a crucial ingredient of Tanglewood readings; we discuss this point in section 4.

Note that there is an asymmetry between the two positions of x in the LF structure in (14a). The first variable x is a trace position of movement, and therefore the relationship between the LF position of Tanglewood and the λ-binder and the position of the variable x in the trace position should be subject to constraints on syntactic movement. The second variable therex , however, is simply base-generated as a variable; it is not the product of movement and therefore should have no constraints beyond being in the scope of the matching λ-binder derived by movement. This asymmetry underlies the novel evidence we present in section 3: in brief, we will show that the position of overt focus (corresponding to the first variable x in (14a)) is sensitive to syntactic islands, whereas the hypothesized bound variable position, within the ellipsis site, is not.

Kratzer (1991:831) briefly considers this type of movement approach to Tanglewood readings but dismisses it, citing the availability of examples such as (16). As the paraphrase in (17) makes clear, this example has a Tanglewood reading where only quantifies over the possibilities that I contacted the person who chairs a certain group before you contacted the person who chairs that same group. It does not assert, for example, that it’s not the case that I contacted the person who chairs the Zoning Board before you contacted the person who chairs the Planning Board.

(16)

  • A Tanglewood construction with the focus inside an island (Kratzer 1991:831) Context: “You always contact every responsible person before me.”

  • No, I only contacted [island the person who chairs [the Zoning Board]F] before you did

    graphic
    .

(17) Paraphrase: The Zoning Board is the only x such that I contacted the person who chairs x before you contacted the person who chairs x.

What is important about example (16) is that the focus the Zoning Board is contained within a relative clause island. If Tanglewood readings require movement of the focus to a position to bind a variable in the ellipsis site, we might expect (16) to have the LF structure in (18). But this would be an island violation. To wit, corresponding overt movement of the focus as in (19) is ungrammatical.

(18)

graphic

(19)

  • Corresponding overt focus movement of the Zoning Board (Kratzer 1991:831)

  • *It was [the Zoning Board]F that I contacted [island the person who chairs ____ ].

Therefore, Kratzer claims, the grammaticality of (16) with its intended Tanglewood interpretation shows that Tanglewood readings do not depend on covert movement of the focus. This then motivates Kratzer’s proposal whereby ellipsis enforces identity of focus indices, briefly introduced in section 1.

What Kratzer did not consider is the possibility of covert focus movement triggering pied-piping (Drubig 1994, Tancredi 1997/2008, 2004, Krifka 2006, Wagner 2006, Erlewine and Kotek 2014).11 Under the approach presented in this article—based on the work of the authors listed here—the first argument of only at LF, derived by covert movement, need only contain the focused constituent. In this case, we can covertly move the island containing the focus, the person . . . , leaving a variable and an associated λ-binder that is restricted to different persons chairing organizations. This binder will also bind the matching variable in the ellipsis site.

(20)

graphic

No islands are violated in this LF structure. We note that parallel pied-piping is possible in overt focus movement, as in (21). Such structures have previously been described simply as clefts where a subpart of the pivot is focused (Chomsky 1970:91ff., summarized in Jackendoff 1972: 232–234; see also Velleman et al. 2012, Erlewine and Kotek 2014).

(21)

graphic

For completeness, we demonstrate the interpretation of Kratzer’s Zoning Board example (16) under our approach to Tanglewood constructions, using the LF structure in (20) involving covert focus movement with pied-piping. Following the context discussed by Kratzer (1991:829), we let

graphic
[the Zoning Board]F
graphic
f = {the Zoning Board, the Planning Board, the Rent Control Board, the Conservation Commission}. Using the simple Roothian procedure for the interpretation of focus semantic values (2), we yield the focus semantic values for the moved constituent—labeled island here—in (22b).

(22)

graphic

The semantics for only here correctly reflects sensitivity to the placement of focus, even though the focus is a proper subpart of the constituent moved to be the first argument of only. This is because the set of alternatives C quantified over in (22d) is constrained by the focus semantic value of the first argument of only: here, C =

graphic
island
graphic
f. This is the domain restriction mechanism of Rooth (1985, 1992), adapted for the two-place only in (12), also used by Wagner (2006). The need for such sensitivity may be clearer in Merchant’s (2008:150) example shown here, which can be similarly analyzed.

(23) I only played [island a song that [Ringo]F wrote] because you did

graphic
.

The demonstration in (22) shows that Kratzer’s one argument against a covert movement account of Tanglewood readings is easily defeated by the possibility of pied-piping in covert focus movement, which has since been independently developed and argued for in works such as Drubig 1994, Krifka 1996, 2006, Tancredi 1997/2008, 2004, Wagner 2006, and Erlewine and Kotek 2014. At the same time, this discussion reflects the difficulty of testing for reflexes of movement such as island sensitivity in Tanglewood constructions, because of the possibility of covert pied-piping; hence, we have not yet given an argument for covert movement. In the next section, we present new evidence that focus association in Tanglewood constructions is island-sensitive in a manner predicted by our movement-based proposal but not by Kratzer’s approach.

3 New Evidence from Island (In)sensitivity

Our proposal for Tanglewood constructions, presented above, involves covert movement of the overt focus—or a constituent properly containing it—which then binds a corresponding bound variable in the ellipsis site. This predicts an asymmetric pattern of island sensitivity: covert movement of the focus (possibly with pied-piping) is subject to island constraints, but variable binding is not. In this section, we will show that Tanglewood constructions exhibit precisely this pattern of island sensitivity, predicted by our covert focus movement account but unpredicted by alternative proposals, including Kratzer’s.

In particular, in the examples in this section we isolate island sensitivity by controlling for pied-piping. We do this by constructing contexts in which the island-sized constituent that we predict moves to only is an inappropriate object for the verb in the ellipsis site, whereas the likely object is embedded inside an island and hence cannot be moved to only. When these conditions are met, island effects are revealed. We then show that, in contrast to the position of overt focus, the bound variable in the ellipsis site is not island-sensitive.

We begin with example (24). The context is designed to make the intended Tanglewood reading natural; nonetheless, the sentence does not have the intended Tanglewood reading, which we indicate with *TW. We note that this sentence does have a number of other possible readings.12

(24)

graphic

Why is this intended reading unavailable? Under our approach, the intended Tanglewood reading requires covert movement of Spanish or a phrase properly containing Spanish to only, binding a corresponding bound variable within the ellipsis site. Consider first the LF structure in (25a): although movement of the focus Spanish would arrive at the intended Tanglewood reading, movement of Spanish alone violates the relative clause island. In (25b), we also consider movement of the entire island containing the focus. The problem here is that the bound variable in the ellipsis site is the object of speak and therefore should correspond to a language, but in order to yield the Tanglewood reading, this variable will be bound by the λ-binder introduced by covert focus movement, and this λ-binder ranges over different nannies, not languages.

(25)

graphic

The unavailability of the Tanglewood reading in (24) is unpredicted by Kratzer’s account. Recall that under Kratzer’s proposal, foci are interpreted in situ at LF (following Rooth 1985) with distinguished focus indices, and ellipsis enforces their identity. Focus association through focus indices and ellipsis are both insensitive to syntactic islands, as Kratzer explicitly claims, predicting the availability of a Tanglewood reading here.

(26)

graphic

In (24), we placed the overt focus inside an island, without a corresponding island in the intended ellipsis site, and as a result the Tanglewood reading became unavailable. However, as we have shown, Kratzer’s (1991) proposal of enforcing the identity of focus indices under ellipsis predicts this reading to be available. This is not a coincidence—this mechanism was explicitly designed to generate Tanglewood readings without locality restrictions or island sensitivity. As a result, this mechanism will systematically overgenerate such Tanglewood readings and will not predict any island sensitivity.

We can similarly demonstrate sensitivity to adjunct islands, as in (27). Example (27a) is a grammatical Tanglewood baseline example, which is then modified in (27b) so that the focus is in an adjunct clause. The intended Tanglewood reading is then judged as unavailable in (27b).

(27)

graphic

Tanglewood readings are similarly unavailable with the overt focus in one conjunct and the ellipsis site in another conjunct (28). This reflects the fact that covert movement is subject to the Coordinate Structure Constraint, as has been independently argued by Bošković and Franks (2000).13

(28)

  • Tanglewood reading blocked by coordination

  • Context: I am under investigation by the Real Estate Board. Sarah and Rebecca claim that I advised them both to bid on many of the same houses, to raise their prices. I reply:

  • *TW I only advised Sarah to bid on [the Elm St. house]F and (told) Rebecca to

    graphic
    as well.

  • Intended Tanglewood reading: The Elm St. house is the only house x such that I advised Sarah to bid on x and (told) Rebecca to bid on x as well. (

    graphic
    = “bid on . . . ”)

Note that there is no independent subject of tell or advise in the right conjunct in (28), ensuring that the entire conjunction is below only as would be required for the Tanglewood reading both by our approach and by Kratzer’s.

Kratzer’s approach predicts no difficulty in deriving the intended Tanglewood readings in (27b) and (28), in the same way that it would overgenerate a Tanglewood reading without island sensitivity in (26). The examples above show that Tanglewood configurations are instead sensitive to syntactic islands in the manner our proposal predicts. Tanglewood readings involve covert focus movement and variable binding in the ellipsis site. The placement of an island around the overt focus will block covert movement of the focus and consequent variable binding, though movement of the island itself is possible, as in (20).

We furthermore predict that the strengths of these contrasts will correlate with the strengths of the islands themselves. For example, the availability of the Tanglewood reading in example (29) shows that covert focus movement can violate wh-islands, which have classically been observed to be of variable strength or weaker than many other islands in English (see, e.g., Ross 1967, Pesetsky 1982, Grimshaw 1986).

(29)

graphic

Tanglewood constructions can thus be used to diagnose the sensitivity of covert (focus) movement to different islands. Note that the strength of individual islands is additionally subject to interspeaker variation. We predict a correlation: a Tanglewood sentence with an island should be as acceptable as a baseline example for that same island. For example, one anonymous LI reviewer reports being able to get the intended Tanglewood reading in the relative clause example (24) but also reports allowing exceptionally wide scope of quantifiers out of relative clause islands, such as the every > some reading of Josh knows someone who speaks every Germanic language from Sabbagh 2007:367.

Our analysis for Tanglewood constructions predicts island sensitivity due to the covert movement of the focus, but it does not predict any island sensitivity around the ellipsis site. To test this, we change the position of the island in our test sentences: we place the ellipsis site inside a syntactic island, without a corresponding island around the antecedent.14 The intended Tanglewood reading is grammatical in this configuration in (30), based on (24).

(30)

graphic

The grammaticality of this Tanglewood construction is predicted by our account. Covert movement of the focus Spanish in (30) is not constrained by any syntactic island. This movement introduces a variable and its λ-binder, which in turn binds the matching bound variable in the ellipsis site. This ellipsis site is inside an island, but this is not a problem: variable binding is not sensitive to syntactic islands. This LF structure for (30) is illustrated in (31).

(31)

graphic

The asymmetry in the availability of Tanglewood readings between (30) and (24) is exactly what our account predicts. The overt focus must move covertly—possibly with pied-piping—and is thus sensitive to islands, but the position of the ellipsis site, under our account, simply hosts a bound variable and is thus insensitive to islands. In contrast, Kratzer’s account would predict no difference between these examples, predicting the availability of a Tanglewood reading in (24), as demonstrated in (26).

We conclude that Tanglewood constructions are island-sensitive, contrary to Kratzer’s claim and prediction. The pattern of island sensitivity observed—where the position of overt focus is island-sensitive but the position of the ellipsis site is not—is precisely as predicted by our proposal, where Tanglewood readings involve covert movement of the focus, which then binds a bound variable in the ellipsis site. Kratzer’s approach of enforcing identity of focus indices under ellipsis systematically overgenerates Tanglewood readings, as it is explicitly designed to not require syntactic movement for deriving them, as do other previous accounts of Tanglewood readings. Kratzer’s approach of enforcing the identity of focus indices under ellipsis must not be available to the grammar.15 Other previous approaches to Tanglewood constructions will be discussed in section 5.

4 Tanglewood Readings with Overt Bound Variables

All of the Tanglewood examples we have discussed thus far have involved ellipsis. For Kratzer’s (1991) proposal, the ellipsis is a crucial component of Tanglewood readings: the ellipsis site is interpreted under identity with the antecedent focus, yielding an LF structure with two foci (32). Matching focus indices between the two positions of focus yield covarying alternatives, as demonstrated in (10).

(32)

graphic

In contrast, our proposal derives Tanglewood readings through general mechanisms of (covert) movement and variable binding, and does not depend on ellipsis. This predicts that Tanglewood readings could also involve overt bound variables. Beaver and Clark (2008) observe that this is indeed the case.

(33)

graphic

In the intended reading of (33), there is an overt bound variable and is not focused. The availability of this reading follows immediately from our account. This LF structure is equivalent to the one proposed in (14a) for the original Tanglewood example.

(34)

graphic

If we instead attempt to interpret (33) using Kratzer’s proposal, there will have to be focused with the same focus index as on Tanglewood—but note that there in (33) is not focused. An unfocused there will not yield the intended reading if it is not a bound variable but simply refers to Tanglewood.

Such Tanglewood examples with overt bound variables allow us to explicitly observe the effects of the covert pied-piping proposed here. Recall Kratzer’s original Zoning Board example (16), repeated in (35), which is a grammatical Tanglewood construction even though its focus is within an island. We proposed above that this example is grammatical because the island the person . . . undergoes covert movement, binding a variable over different persons; see (20). This is reflected explicitly by the bound variable in the grammatical ellipsis-less variant (36), which has the truth-conditionally equivalent Tanglewood reading from (35).

(35)

  • Kratzer’s Zoning Board example, repeated from (16)

  • Context: “You always contact every responsible person before me.”

  • TW No, I only contacted [island the person who chairs [the Zoning Board]F] before you did

    graphic
    .

  • Intended Tanglewood reading: The Zoning Board is the only x such that I contacted the person who chairs x before you contacted the person who chairs x.

(36)

  • Overt bound variable paraphrase of Kratzer’s Zoning Board example (16)

  • TW I only contacted [island the person who chairs [the Zoning Board]F] before you contacted her/him/them.16

The example in (36) corresponds to a parse of (35) where the ellipsis site is resolved as

graphic
= “contact her/him/them.” In contrast, there is no grammatical equivalent of (35) that explicitly spells out the ellipsis site as
graphic
= “contact the person who chairs . . . ” This again reflects the fact that the focused constituent the Zoning Board cannot covertly move out of the island to a position where it can bind the bound variable it. Instead, the entire island must move.

(37)

  • Bound variable corresponding to the focus, not the island, is not possible

  • *TW I only contacted [island the person who chairs [the Zoning Board]F] before you contacted [island the person who chairs it].

To conclude, Tanglewood constructions can involve overt bound variables and do not depend on ellipsis, as our account predicts. This was previously observed by Beaver and Clark (2008), but without an explicit account that predicts the island sensitivity observed in section 3. Such data are problematic for Kratzer’s account, which relies on ellipsis for generating Tanglewood constructions. Kratzer’s proposal undergenerates the examples with overt bound variables in this section, while simultaneously overgenerating the island examples in section 3.

5 Alternative Analyses of Tanglewood Readings

Several alternatives to Kratzer’s (1991) analysis of Tanglewood readings can be found in the literature. In this section, we briefly discuss these alternatives and their shortcomings in light of the evidence presented in this article. In particular, two common difficulties these accounts face are (a) the asymmetric island sensitivity we presented in section 3—a fact that has never been previously noted—and (b) the availability of Tanglewood readings with overt bound variables but without ellipsis, discussed in section 4.

We first discuss analyses in the Structured Meaning (SM) framework, beginning with Krifka 1991. This framework involves “projection” of the focused constituent in a separate “stack” of the computation, without requiring any movement. Krifka applies this approach to Kratzer’s Tanglewood constructions, assuming a complete copy of the focus within the ellipsis site, together with a mechanism to equate the foci “projected” from the antecedent and ellipsis sites. Jäger (1999) then extends this analysis using a theory that directly equates the ellipsis content with its antecedent. Akin to our conclusions regarding Kratzer’s enforcement of identical focus indices through ellipsis, these SM projection approaches fail to predict the island sensitivity of the position of overt focus in Tanglewood constructions, presented in section 3.

Krifka (2006) then proposes that the “projection” mechanism in the SM approach should in fact be thought of as syntactic movement, but this does not improve the situation. Updating the Krifka 1991 account for Tanglewood constructions using the movement approach to projection of Krifka 2006 will predict island sensitivity for both the position of overt focus and the ellipsis site, contrary to what we have shown in section 3.

Charlow (2008) presents an analysis of a subclass of Tanglewood constructions as involving variable binding, which is a component of our analysis as well. However, this Categorial Grammar approach uses the same type-shifting rules to abstract over the antecedent side and the ellipsis site or bound variable. This predicts island sensitivity on either both sides or neither side, thus failing to derive the asymmetric island sensitivity observed in section 3.

Another proposal is Sauerland’s (2007a,b) structure-sharing account. Sauerland briefly relates Tanglewood constructions to more general effects of dependence on contrastive foci that can occur specifically in ellipsis, independently observed by Hardt (1999) and Schwarz (2000). Citing these other, non-Tanglewood examples, Sauerland claims explicitly that this structure-sharing mechanism is not subject to syntactic locality conditions (see Sauerland 2007a:46). Furthermore, his structure-sharing account would also fail to extend to Tanglewood readings with overt bound variables.

Finally, Beaver and Clark (2008) discuss Tanglewood constructions at some length and consider two possible approaches, each of which seems to capture some but not all of the properties of Tanglewood constructions. The first is a movement account, which differs significantly from ours in involving VP-movement as the mechanism for VP-ellipsis and using this movement to derive the Tanglewood effect (pp. 109–111). This approach is challenged by the availability of Tanglewood readings that involve overt bound variables, without ellipsis, which Beaver and Clark, too, observe (p. 112). They then present a proof-of-concept dynamic semantics approach to Alternative Semantics (pp. 111–115), with the aim of accounting for both overt pronouns bound by focused constituents and Kratzer’s Tanglewood examples. However, this account predicts no sensitivity to syntactic islands and hence cannot be extended to model the data we have presented in section 3. Ultimately, Beaver and Clark do not commit themselves to either approach.

6 Covert Focus Movement Is Long-Distance and Is Not Quantifier Raising

In this section, we consider and argue against one final possible alternative analysis for the data we have presented. This is the possibility that Tanglewood readings indeed involve covert movement and variable binding, but that this movement is not covert focus movement but rather reflects a general-purpose operation such as Quantifier Raising (QR). We will show that the covert movement involved in Tanglewood constructions can be long-distance, across finite-clause boundaries, and in particular that this movement can be longer than that of quantifiers undergoing QR.

We first consider example (38), which is a grammatical Tanglewood construction. In the intended reading, the because-clause adjoins to and modifies think. Therefore, for the binder of the moved focus anaphora to bind the bound variable in the ellipsis site, anaphora must necessarily move outside of the embedded finite clause. (39) gives the LF structure that we propose for this sentence.

(38)

  • Tanglewood construction requiring long-distance covert movement

  • Context: John, a first-year grad student, doesn’t quite understand the field yet. He seems to think that everyone works on focus, on ellipsis, and on anaphora. Some people think he is just extrapolating from what his advisor works on. But actually . . .

  • TW He only thinks [CP that everyone works on [anaphora]F] because his advisor does

    graphic
    .

  • Intended Tanglewood reading: Anaphora is the only topic x such that John [thinks that everyone works on x [because his advisor works on x]]. (

    graphic
    = “work on . . . ”)

(39)

graphic

Next, let us compare this with the behavior of variable binding by a quantifier that has undergone QR. Example (40) is a version of (38) with the focus replaced by the quantifier at least one topic and without the associating only. This sentence does not have the intended Tanglewood-like reading, which would involve at least one topic undergoing long-distance QR and binding into the ellipsis site. The baseline example in (41) shows that variable binding into a because-clause by an object quantifier at least one topic is possible, if the because-clause is attached to the local clause.

(40)

  • QR does not move as high as the focus in (38)

  • *TW He thinks [that everyone works on at least one topic] because his advisor does

    graphic
    .

  • Intended Tanglewood-like reading: There is at least one topic x such that he [thinks everyone works on x [because his advisor works on x]]. (

    graphic
    = “work on . . . ”)

(41)

graphic

The contrast between (40), containing a quantifier, and (38), with focus association with only, shows that the covert movement in Tanglewood constructions cannot simply be reduced to QR’s independent ability to covertly move arguments.17 Covert focus movement is long-distance, crossing finite-clause boundaries, in environments where quantifiers cannot. Hence, we argue here for the existence of covert focus movement, which is distinct from QR and must be available to the grammar alongside QR. This focus movement is necessitated in our analysis for simple reasons of semantic composition: the two-place formulation of only in (12) requires a first argument. This argument is supplied to the operator through covert movement of the overt focus—or a constituent properly containing it—as detailed in our proposal above.

7 Conclusion

In this article, we argued for covert focus movement in English focus association with only. Our evidence came from Tanglewood configurations of the form in Kratzer 1991. We showed that Tanglewood configurations are sensitive to syntactic islands, contrary to Kratzer’s claims and predictions. In particular, we showed an asymmetric pattern of island sensitivity: the position of overt focus is island-sensitive but the position of the ellipsis site is not.

We propose that Tanglewood constructions are derived through covert movement of the focused constituent to the focus-sensitive operator, with binding of a bound variable in the ellipsis site. This movement may involve covert pied-piping of a larger constituent properly containing the focus (Drubig 1994, Krifka 1996, 2006, Tancredi 1997/2008, 2004, Wagner 2006, Erlewine and Kotek 2014). This availability of covert pied-piping explains examples such as Kratzer’s, which are apparently island-insensitive: in such examples, the entire island undergoes covert movement to the operator, and hence there is no island violation. It also explains the asymmetric pattern of island sensitivity we describe above, since the ellipsis site contains a base-generated bound variable that does not undergo any movement and hence is not island-sensitive.

This proposal also severs the link between Tanglewood readings and ellipsis. Indeed, we showed that parallel Tanglewood readings are available in sentences with overt bound variables, which do not involve ellipsis. Additionally, we showed that covert focus movement is long-distance and may cross finite-clause boundaries, unlike QR in the same environment. Kratzer’s influential proposal—that ellipsis enforces identity of focus indices—greatly overgenerates Tanglewood constructions. We also showed that alternative analyses of Tanglewood constructions fail to capture the facts presented here.

Notes

1 In Rooth 1985, focus semantic values were called p-sets, short for presuppositional set from Jackendoff 1972. The definition in (2) is a quotation from Rooth 1985:14 but modified to use the now standard terminology and notation of Rooth 1992. As noted in Rooth 1992:84n7, the recursion step in (2c) is equivalent to that for the compositional interpretation of wh-questions proposed in Hamblin 1973 (see Hamblin 1973:49, esp. n. 8).

2 For convenience, here and elsewhere, we will ignore the contribution of tense and illustrate subjects in their VP-internal base positions. The categories we label VP could also, more precisely, be called vP. The denotations of propositional alternatives as in (3b) and (7a) should also be thought of as standing in for their intensions.

3 Note, however, that this requires a focused constituent to be elided, against the common prohibition on the ellipsis of focused constituents (see, e.g., Tancredi 1992, Heim 1997, Merchant 2001). As we will show, our proposal does not face a similar problem.

4 Other solutions to the Tanglewood problem have also been proposed. See section 5 for discussion.

5 We thank an anonymous LI reviewer for suggesting this phrasing for the LF identity condition assumed in Kratzer 1991, as well as the condition in (15).

6 Again, we do not illustrate tense or movement of the subject out of its predicate-internal base position.

7 Three notes on this movement. First, we present only the corresponding PF and LF representations in (11); for our purposes, this movement could be thought of as taking place in the narrow syntax, with pronunciation at the tail of the chain, or taking place after Spell-Out, feeding only LF.

Second, the movement in (11) can be derived without violating the Extension Condition (Chomsky 1993) by (a) first merging the focused constituent with only, resulting in an independent [only TanglewoodF] tree in the workspace, (b) adjoining the λ-binder to the root of the tree containing the trace of the focused constituent, then (c) merging the results of steps (a) and (b). The necessity of such derivations has been independently claimed for cases of head movement (Bobaljik and Brown 1997) and sideward movement (Nunes 2001, 2004).

Third, the movement in (11) also does not violate the Proper Binding Condition (Fiengo 1977) if we think of this as a semantic condition requiring variables to be bound by their binders. Even though the landing site of movement does not c-command its trace position, the -binder associated with its movement does properly bind the trace position. See Keine and Bhatt 2016 for related discussion supporting this position of the λ-binder. See also Pesetsky 2007, 2013 and Yuan 2017a,b for independent motivation for overt movement of this form.

8 Other movement options are also conceivable. What is important for a two-place only is the position of the λ-binder in (11), so that only takes a derived predicate as an argument: here, λx. I go to x. This would not occur if only attracts the focus-containing phrase to its specifier, together with common Heim and Kratzer 1998 assumptions on the position of λ-binders. A lower placement of the λ-binder as in (11), as independently proposed in Nissenbaum 2000, would however make movement to specifier position viable.

A reviewer notes that this two-place only is similar to the denotation of focus-sensitive particles in Structured Meaning theories of focus (see Jacobs 1983, von Stechow 1991a,b; also section 5). The availability of the movement proposed here is then subject to Rooth’s (1996:278) criticism regarding the hypothetical verb tolf. An alternative would be to use a one-place only that triggers movement of the focus-containing phrase to the edge of only’s complement, resulting in a representation as in (i). This form of covert focus movement is compatible with the analysis here and avoids the tolf problem.

  • (i)

    [only [VP [Tanglewood]Fx [VP I go to x]]]]

We leave further discussion of the precise geometry of covert focus movement for future work.

9 The semantics for (12) here is a naive formulation that blindly negates all nonprejacent alternatives. Formally, this must be modified so that it is all alternatives that are innocently excludable that are negated. See discussion in, for example, von Fintel 1997, Fox 2007, and Spector 2016. In the examples in this article, this issue will not arise, so we will continue to use the naive formulation in (12).

There are also debates in the literature regarding the status of the prejacent inference, which we call a presupposition here. This question is orthogonal to the present discussion.

10 The coindexation in the configurations we propose (e.g., (14)) is allowed by Heim’s (1997:202) No Meaningless Coindexing rule, because the trace and the pronoun in (14) are both bound by the same operator.

  • (i)

    No Meaningless Coindexing

    If an LF contains an occurrence of a variable v that is bound by a node α, then all occurrences of v in this LF must be bound by the same node α.

  • As Charlow (2008) notes, the pronoun in the ellipsis site could also refer directly to the moved phrase, rather than being bound by the λ-binder of covert movement, contra Beaver and Clark 2008:112. Consider the following dialogue, based on an example from Charlow 2008:200:

  • (ii)

    A: You always try to show off how much you can eat by ordering the same thing I do plus something else. When I order soup, you order soup and salad. When I eat udon, you get udon and soba.

    B: In general, I suppose you’re right. But here’s a counterexample: Last week I only ate [brisket]F when you did

    graphic
    . I didn’t eat brisket and ribs or anything like that. I just got the same thing you did.

The ellipsis site of the italicized sentence in (iiB) is successfully interpreted as ‘eat it (= brisket)’, without a bound variable. In this article, we will concentrate on the derivation of Tanglewood readings, narrowly defined, where this variable is bound.

11 None of these previous authors specifically discusses—let alone argues for—the application of covert pied-piping to the problem of Tanglewood readings and Kratzer’s challenge to the movement account. To our knowledge, the closest that anyone has come to this solution in previous literature is the last sentence of footnote 14 in Beaver and Clark 2008: 110–111 and footnote 27 in Griffiths and Lipták 2014:220.

12 In particular, there is another reading that we would call a Tanglewood reading: namely, a reading where the ellipsis is resolved to a higher VP,

graphic
= “hire a nanny that speaks . . . ,” paraphrasable as Spanish is the only language x such that we hired [a nanny that speaks x] because our son hires [a nanny that speaks x]. The availability of this reading is predicted under our account, following a derivation parallel to (22) for Kratzer’s Zoning Board example. This reading differs from our intended reading here, and is not supported by the context in (24).

13 As a reviewer notes, the coordinate structure in (28) could conceivably allow for an alternative derivation using across-the-board covert focus movement, schematized in (i).

  • (i)

    graphic

However, the LF structure in (i) is unavailable due to the unavailability of covert across-the-board movement, as Bošković and Franks (2000) argue.

14 Note that the ellipsis site in Tanglewood examples is already inside an adjunct island such as that headed by because in (30), including in Kratzer’s original example (4). However, because of possible differences in island strength, mentioned above, we insert the additional relative clause island here around the ellipsis site, to make (30) parallel to (24) both in the island involved and in overall complexity.

15 A possible stronger conclusion that we might entertain is that Kratzer’s mechanism of computing focus alternatives using focus indices as a whole must not be available to the grammar. Aside from Kratzer’s argument for this mechanism from Tanglewood readings, two additional arguments are found in the literature. The first comes from crossing focus dependencies in Wold 1996 (but see Krifka 1996, 2006, Tancredi 1997/2008, 2004 for arguments that such data are better captured by assuming covert focus movement). A second argument comes from the interaction of focus with the copy theory of movement (Erlewine 2014), in particular in so-called backward association configurations. Absent an alternative account for these data, the interaction of focus with copies discussed there would necessitate the use of Kratzerian focus indices. See Erlewine 2014 for details.

16 In the authors’ English, (36) sounds best with the gender-neutral singular them, but we want to make it clear that this pronoun here is animate. Given sufficient contextual expectations that all persons chairing relevant organizations are female or male, the singular her or him becomes grammatical.

17 We recognize that there is some interspeaker variability in the locality of QR (see, e.g., Wurmbrand 2018). What is important here is that there is a contrast between these two examples in the availability of the intended reading.

Acknowledgments

For valuable comments and discussion, we thank audiences at the University of Edinburgh, Yale University, SALT 26, and the Tokyo Semantics Research Group, Brian Buccola, Jason Merchant, Miriam Nussbaum, Gary Thoms, anonymous reviewers, and especially Aron Hirsch. The first author is supported by the National University of Singapore (FASS start-up) under WBS R-103-000-135-133. Errors are of course each other’s.

References

References
Beaver,
David Ian
, and
Brady
Clark
.
2008
.
Sense and sensitivity: How focus determines meaning
.
Chichester, West Sussex
:
Wiley-Blackwell
.
Bobaljik,
Jonathan David
, and
Samuel
Brown
.
1997
.
Interarboreal operations: Head movement and the Extension Requirement
.
Linguistic Inquiry
28
:
345
356
.
Bošković,
Željko
, and
Steven
Franks
.
2000
.
Across-the-board movement and LF
.
Syntax
3
:
107
128
.
Charlow,
Simon
.
2008
. Free and bound pro-verbs: A unified treatment of anaphora. In
Proceedings of SALT18
, ed. by
Tova
Friedman
and
Satoshi
Ito
,
198
211
. http://journals.linguisticsociety.org/proceed ings/index.php/SALT/issue/view/91.
Chomsky,
Noam
.
1970
. Deep structure, surface structure and semantic interpretation. In
Studies in general and Oriental linguistics
, ed. by
Roman
Jakobson
and
Shigeo
Kawamoto
,
62
119
.
Tokyo
:
TEC Corporation for Language Research
. Reprinted in
Semantics: An interdisciplinary reader in philosophy, linguistics, and psychology
, ed. by
Danny D.
Steinberg
and
Leon A.
Jakobovits
,
183
216
.
Cambridge
:
Cambridge University Press
(
1971
).
Chomsky,
Noam
.
1993
. A minimalist program for linguistic theory. In
The view from Building 20
, ed. by
Kenneth
Hale
and
Samuel Jay
Keyser
,
1
52
.
Cambridge, MA
:
MIT Press
.
Drubig,
Hans Bernhard
.
1994
.
Island constraints and the syntactic nature of focus and association with focus
.
Arbeitspapiere des Sonderforschungsbereichs 340: Sprachtheoretische Grundlagen der Computerlinguistik 51. University of Stuttgart/Tübingen
.
Erlewine,
Michael Yoshitaka
.
2014
.
Movement out of focus
.
Doctoral dissertation, MIT, Cambridge, MA
.
Erlewine,
Michael Yoshitaka
, and
Hadas
Kotek
.
2014
. Intervention in focus pied-piping. In
NELS 43
, ed. by
Hsin-Lun
Huang
,
Ethan
Poole
, and
Amanda
Rysling
,
1
:
117
130
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Fiengo,
Robert
.
1977
.
On trace theory
.
Linguistic Inquiry
8
:
35
62
.
von Fintel,
Kai
.
1997
.
Bare plurals, bare conditionals, and only
.
Journal of Semantics
14
:
1
56
.
Fox,
Danny
.
2007
. Free choice and the theory of scalar implicatures. In
Presupposition and implicature in compositional semantics
, ed. by
Penka
Stateva
and
Uli
Sauerland
,
537
586
.
New York
:
Palgrave-Macmillan
.
Griffiths,
James
, and
Anikό
Lipták
.
2014
.
Contrast and island sensitivity in clausal ellipsis
.
Syntax
17
:
189
234
.
Grimshaw,
Jane
.
1986
.
Subjacency and the S/S/ʹ parameter
.
Linguistic Inquiry
17
:
364
369
.
Hamblin,
Charles
.
1973
.
Questions in Montague English
.
Foundations of Language
10
:
41
53
.
Hardt,
Daniel
.
1999
.
Dynamic interpretation of verb phrase ellipsis
.
Linguistics and Philosophy
22
:
185
219
.
Heim,
Irene
.
1997
. Predicates or formulas? Evidence from ellipsis. In
Proceedings of SALT 7
, ed. by
Aaron
Lawson
,
197
221
. http://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/102.
Heim,
Irene
, and
Angelika
Kratzer
.
1998
.
Semantics in generative grammar
.
Malden, MA
:
Blackwell
.
Horn,
Laurence Robert
.
1969
. A presuppositional analysis of only and even. In
Papers from the Fifth Regional Meeting of the Chicago Linguistic Society
, ed. by
Robert I.
Binnick
,
Alice
Davison
,
Georgia M.
Green
, and
Jerry L.
Morgan
,
98
107
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Jackendoff,
Ray
.
1972
.
Semantic interpretation in generative grammar
.
Cambridge, MA
:
MIT Press
.
Jacobs,
Joachim
.
1983
.
Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im Deutschen
.
Tübingen
:
Niemeyer
.
Jäger,
Gerhard
.
1999
. Focus without variables: A multi-modal analysis. In
WCCFL 17
, ed. by
Kimary N.
Shahin
,
Susan
Blake
, and
Eun-Sook
Kim
,
317
331
.
Stanford, CA
:
CSLI Publications
.
Keine,
Stefan
, and
Rajesh
Bhatt
.
2016
.
Interpreting verb clusters
.
Natural Language and Linguistic Theory
34
:
1445
1492
.
Kratzer,
Angelika
.
1991
. The representation of focus. In
Semantik: Ein internationales Handbuch der zeitgenössischen Forschung
, ed. by
Arnim
von Stechow
and
Dieter
Wunderlich
,
825
834
.
Berlin
:
Walter de Gruyter
.
Krifka,
Manfred
.
1991
. A compositional semantics for multiple focus constructions. In
Proceedings of SALT 1
, ed. by
Steven K.
Moore
and
Adam Zachary
Wyner
,
127
158
. http://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/136.
Krifka,
Manfred
.
1996
. Frameworks for the representation of focus. In
ESSLLI ’96: Formal grammar
, ed. by
Geert-Jan M.
Kruijff
,
Richard T.
Oehrle
, and
Glyn
Morrill
.
Universität des Saarlandes, Department of Computational Linguistics and Phonetics
.
Krifka,
Manfred
.
2006
. Association with focus phrases. In
The architecture of focus
, ed. by
Valéria
Molnár
and
Susanne
Winkler
,
105
136
.
Berlin
:
Mouton de Gruyter
.
Merchant,
Jason
.
2001
.
The syntax of silence
.
Oxford
:
Oxford University Press
.
Merchant,
Jason
.
2008
. Variable island repair under ellipsis. In
Topics in ellipsis
, ed. by
Kyle
Johnson
,
132
153
.
Cambridge
:
Cambridge University Press
.
Nissenbaum,
Jonathan W
.
2000
.
Investigations of covert phrase movement
.
Doctoral dissertation, MIT, Cambridge, MA
.
Nunes,
Jairo
.
2001
.
Sideward movement
.
Linguistic Inquiry
32
:
303
344
.
Nunes,
Jairo
.
2004
.
Linearization of chains and sideward movement
.
Cambridge, MA
:
MIT Press
.
Pesetsky,
David
.
1982
.
Paths and categories
.
Doctoral dissertation, MIT, Cambridge, MA
.
Pesetsky,
David
.
2007
.
Undermerge, Overmerge, and the secret genitive inside every Russian noun
.
Handout from Formal Approaches to Slavic Linguistics (FASL) 16
.
Pesetsky,
David
.
2013
.
Russian case morphology and the syntactic categories
.
Cambridge, MA
:
MIT Press
.
Rooth,
Mats
.
1985
.
Association with focus
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Rooth,
Mats
.
1992
.
A theory of focus interpretation
.
Natural Language Semantics
1
:
75
116
.
Rooth,
Mats
.
1996
. Focus. In
The handbook of contemporary semantic theory
, ed. by
Shalom
Lappin
,
271
297
.
Malden, MA
:
Blackwell
.
Ross,
John Robert
.
1967
.
Constraints on variables in syntax
.
Doctoral dissertation, MIT, Cambridge, MA
.
Sabbagh,
Joseph
.
2007
.
Ordering and linearizing rightward movement
.
Natural Language and Linguistic Theory
25
:
349
401
.
Sauerland,
Uli
.
2007a
.
Copying vs. structure sharing: A semantic argument
.
Linguistic Variation Yearbook
7
:
27
51
.
Sauerland,
Uli
.
2007b
. Focus dependency as structure sharing. In
Proceedings of Sinn und Bedeutung 11
, ed. by
Estela
Puig-Waldmüller
,
540
554
.
Barcelona
:
Universitat Pompeu Fabra
.
Schwarz,
Bernhard
.
2000
.
Topics in ellipsis
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Spector,
Benjamin
.
2016
.
Comparing exhaustivity operators
.
Semantics and Pragmatics
9
:
1
33
.
Stechow,
Arnim von
.
1991a
. Current issues in the theory of focus. In
Semantik: Ein internationales Handbuch der zeitgenössischen Forschung
, ed. by
Arnim
von Stechow
and
Dieter
Wunderlich
,
804
824
.
Berlin
:
Walter de Gruyter
.
Stechow,
Arnim von
.
1991b
. Focusing and backgrounding operators. In
Discourse particles: Descriptive and theoretical investigations on the logical, syntactic and pragmatic properties of discourse particles in German
, ed. by
Werner
Abraham
,
37
84
.
Amsterdam
:
John Benjamins
.
Tancredi,
Christopher
.
1992
.
Deletion, deaccenting, and presupposition
.
Doctoral dissertation, MIT, Cambridge, MA
.
Tancredi,
Christopher
.
1997/2008
.
Focus, focus types, and associative operators
.
1997 manuscript, printed in
Reports of the Keio Institute of Cultural and Linguistic Studies
39
:
1
62
,
2008
.
Tancredi,
Christopher
.
2004
.
Associative operators
.
Gengo Kenkyu
125
:
31
82
.
Velleman,
Leah
,
David Ian
Beaver
,
Emilie
Destruel
,
Dylan
Bumford
,
Edgar
Onea
, and
Liz
Coppock
.
2012
. It-clefts are IT (inquiry terminating) constructions. In
Proceedings of SALT 22
, ed. by
Anca
Chereches
,
441
460
. http://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/87.
Wagner,
Michael
.
2006
.
Association by movement: Evidence from NPI-licensing
.
Natural Language Semantics
14
:
297
324
.
Wold,
Dag
.
1996
. Long distance selective binding: The case of focus. In
Proceedings of SALT 6
, ed. by
Teresa
Galloway
and
Justin
Spence
,
311
328
. http://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/103.
Wurmbrand,
Susi
.
2018
.
The cost of raising quantifiers
.
Glossa
3
(
1
):
19
.
Yuan,
Michelle
.
2017a
. More on Undermerge: Phrasal and head movement interactions in Kikuyu. In
A pesky set: Papers for David Pesetsky
, ed. by
Claire
Halpert
,
Hadas
Kotek
, and
Coppe
van Urk
,
543
552
.
MIT Working Papers in Linguistics 80
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Yuan,
Michelle
.
2017b
.
Movement to complement in Kikuyu and the syntax of focus association
.
Ms., MIT, Cambridge, MA
.