Abstract

Ormazabal and Romero (2012) take issue with the arguments in Bruening 2010b that certain instances of prepositional dative constructions are really double object constructions. I show that their criticisms, when examined closely, actually support this claim. I also show that their alternative explanation for the facts is not successful. I also expand on the arguments from Bruening 2010b, and further argue that double object constructions must be distinct syntactically and semantically from prepositional dative constructions. The two cannot be derived from the same source.

1 Introduction

An important issue in the literature on ditransitive constructions is whether double object constructions (1a) and prepositional dative constructions (1b) should be derived from the same source.

(1)

  • a.

    Ronald gave his sister a frog. (double object construction)

  • b.

    Ronald gave a frog to his sister. (prepositional dative construction)

Those who argue that they are distinct and should not be analyzed as derivationally related frequently point to semantic differences between the two constructions to motivate this position (e.g., Green 1974, Oehrle 1976, Gropen et al. 1989). For instance, the double object construction requires that the first object be a recipient or possessor, while the prepositional dative construction permits other interpretations—for instance, an endpoint-of-motion interpretation.

(2)

  • a.

    I kicked {her/*the goal line} the ball. (recipient or possessor)

  • b.

    I kicked the ball to {her/the goal line}. (endpoint of motion)

(3)

  • a.

    I took {him/*the windowsill} a cup of coffee. (recipient or possessor)

  • b.

    I took a cup of coffee to {him/the windowsill}. (endpoint of motion)

Similarly, certain choices of NPs are said to be allowed only in the double object construction, not in the prepositional dative.

(4)

  • a.

    That smell gave Bill a headache.

  • b.

    *That smell gave a headache to Bill.

In addition, there are some nonalternating verbs like deny.

(5)

  • a.

    The boss denied George his pay.

  • b.

    *The boss denied his pay to George.

These are not expected, if the two alternants are semantically identical and derived from the same source.

Recently, however, these facts have been contested. Several works have argued that double object constructions and prepositional dative constructions are not distinct (Bresnan 2007, Bresnan et al. 2007, Rappaport Hovav and Levin 2008, Bresnan and Nikitina 2009). Some of this literature points out that sentences corresponding to (4b) and (5b) are actually attested.1

(6)

(7)

In Bruening 2010b, I argued that such instances of apparent prepositional dative constructions are actually double object constructions. The idea is that give a headache (for instance) can only be a double object construction; it can never be a prepositional dative construction. Attested examples that appear to be prepositional datives, like (6) and (7), are actually double object constructions, but with the first object projected on the right, rather than the left. I referred to this rightward projection as R-dative shift. R-dative shift is permitted just when the first object is extracted, for instance by heavy shift, as in these examples where the recipient is a complex NP and quite heavy. The evidence that I presented for the existence of R-dative shift included scope interactions and locative inversion. In scope possibilities and in interaction with locative inversion, examples like (6) and (7) pattern not with prepositional dative constructions, but with double object constructions.

Ormazabal and Romero (2012) (henceforth “O&R”) take issue with this R-dative shift account, claiming that a simpler explanation is available. Their reply is couched within the viewpoint that double object constructions and prepositional dative constructions should be derived from the same source (see their other work, in particular Ormazabal and Romero 2010). If their reply were successful, they would have bolstered the claim that double object constructions and prepositional dative constructions are not distinct semantically. This would pave the way toward an analysis that relates the two derivationally, as in their own account (Ormazabal and Romero 2010, based on the preposition incorporation analysis of Baker 1988).

However, O&R’s reply is not successful. O&R neither dispute the facts nor provide a new way of deriving them. Instead, they propose an analysis of the facts presented in Bruening 2010b that does not account for the full range of data. I show this here, and also address all of their criticisms of the R-dative shift analysis. I show that these criticisms, when examined carefully, actually support the R-dative shift analysis. I also expand on the facts that motivated the R-dative shift analysis and discuss in more detail differences between double object constructions and prepositional dative constructions. I show that these differences strongly favor an analysis that treats them as distinct.

I begin by explaining the R-dative shift analysis and O&R’s alternative (section 2). This section also shows that their alternative does not succeed in accounting for the facts. I then go through their other criticisms of the R-dative shift account, involving locative inversion (section 3), structural asymmetries (section 4), pair-list readings with idiomatic phrases (section 5), and semantics and crosslinguistic facts (section 6).

The end result is that the R-dative shift analysis is very well motivated. The significance of this is that it indicates that double object constructions and prepositional dative constructions are derivationally and semantically distinct. In section 6, I discuss the issue of semantics in detail, and also bring in some crosslinguistic facts. The discussion leads to the conclusion that we need different analyses for prepositional dative constructions and double object constructions. The hope is that this note, beyond simply responding to O&R, also helps to bring some clarity to the issues and facts concerning ditransitive constructions.

2 The R-Dative Shift Analysis and Ormazabal and Romero’s Alternative

In Bruening 2010b, I argued that instances of apparent prepositional dative constructions like the one in (6), repeated here, are actually double object constructions.

(8)

The idea is that give a headache (for instance) can only be a double object construction; it can never be a prepositional dative construction. Attested examples that appear to be prepositional datives, like (6)/(8), are actually double object constructions, but with the first object projected on the right, rather than the left. The structure I proposed was (9) (to be revised slightly below).

(9)

graphic

In the double object construction, the indirect object is projected by an Appl(icative) head, as first proposed by Marantz (1993). Usually the specifier of Appl is projected on the left, but in examples like (6) it is on the right (and I return below to the presence of the preposition just when it is on the right).2

In a real prepositional dative construction, in contrast, the PP is projected by the verb, as in (10), the structure for (1b) (see Bruening 2010a).3

(10)

graphic

(The verb generally moves through any projections to Voice, both in the prepositional dative construction and in the double object construction.)

The rightward projection of the specifier of Appl in (9) must not be freely available; otherwise, examples like (4b) would always be acceptable. I therefore proposed the following restriction on rightward projection (Bruening 2010b:291, (9)):

(11)

  • The Extraction Constraint on Rightward Specifiers

  • The specifier of ApplP may be ordered to the right of its sister only if the NP that occupies it undergoes Ā-extraction.

Ā-extraction includes wh-movement, relativization, and, importantly, heavy shift to the right. Heavy shift is what licenses the order in (6). The phrase to the most athletic constitution undergoes Ā-movement to the right, following its initial projection in a rightward specifier of Appl (below, I will spell out where heavy shift moves the phrase to).4 Since (4b) does not involve a heavy NP, heavy shift is not licensed, and (4b) violates the Extraction Constraint on Rightward Specifiers.5

In Bruening 2010b, I referred to the rightward projection of the specifier of Appl as R-dative shift. The arguments that I presented for the existence of R-dative shift included locative inversion and scope interactions. For instance, a sentence where R-dative shift has applied patterns in scope possibilities not with a standard prepositional dative construction, but with a double object construction. Consider the following examples, using the typically nonalternating verb spare:6

(12)

  • a.

    Let’s spare everyone some ordeal or other. (every > some)

  • b.

    Let’s spare someone or other every ordeal. (*every > some)

  • c.

    Let’s spare some ordeal or other to every prisoner that comes before us. (every > some)

  • d.

    Let’s spare every ordeal to someone or other (who comes before us today). (*every > some)

    (Bruening 2010b:295, (17))

The shifted variants in (12c–d) pattern like the double object constructions in (12a–b). Surprisingly, the scope reading that would correspond to the linear order of the quantifiers is unavailable in (12d), just as the same scope reading is unavailable in (12b). The latter, by hypothesis, has the same structure as (12d). In contrast, a simple prepositional dative construction allows the surface scope reading, of course.

(13) Let’s give every toy to some child or other. (every > some)

In Bruening 2001, 2010b, I explained frozen scope in double object constructions through the existence of the Appl projection. In (9), the two NPs are in different projections. The higher one, in Spec,ApplP, will always be closer to any higher scope position, and general economy principles will therefore rule out the lower one crossing over the higher one to take scope (see Bruening 2001 for details). In contrast, in (10), the NP and the PP are dominated by the same projection (VP), and so neither is closer than the other to any higher scope position. Economy principles therefore permit either to move first, resulting in scope ambiguity.7

O&R suggest instead that all examples of apparent prepositional dative constructions are such, with the following small clause structure (the inconsistency in DP and NP labels is theirs; this diagram is based on Ormazabal and Romero 2012:462, (14)).8

(14)

graphic

The lack of a distributive reading in (12d) is due to heavy shift, which takes P′ and adjoins it to VoiceP on the right (Ormazabal and Romero 2012:462, (14)). According to O&R, this heavy shift fixes the scope of the shifted NP over the NP in Spec,PP.

There are a number of problems with this account of the scope facts. First, it leaves unexplained the very contrast that it is supposed to explain: namely, the contrast between heavy shift with nonalternating verbs like spare and an equally heavy NP in a real prepositional dative construction. Let us put the two side by side.

(15)

  • a.

    Let’s spare every ordeal to someone or other who comes before us today. (*every > some)

  • b.

    Let’s give every toy to someone or other who comes before us today. (every > some)

Only with spare is the distributive reading unavailable. With give, the distributive reading is available, even when the object of to is heavy. O&R’s analysis has no way to account for this contrast: either heavy shift is obligatory, and the distributive reading should be unavailable with both verbs, or it is not obligatory, and the distributive reading should be available with both verbs.

A second problem is that O&R’s account relies on scope being fixed by heavy shift, but heavy shift does not generally fix scope. It is true that some instances of extraposition may fix scope high (in particular, extraposition from NP; see Guéron and May 1984, Fox and Nissenbaum 1999), but heavy shift of an argument NP or PP does not fix scope relative to another argument in the way that O&R’s analysis requires. Consider the following examples, both of which clearly do involve heavy shift of a PP:

(16)

  • a.

    The nurse placed every scalpel yesterday on a different sterile tray. (every > a)

  • b.

    Those boys put every part1 without me noticing in a place where it1 didn’t belong. (every > a)

Heavy shift does not fix the scope of the indefinite above the nonshifted universal quantifier; note in particular the licit variable binding in (16b). (See also Johnson 2001 for some discussion of heavy shift and scope interactions in prepositional dative constructions.)

O&R’s alternative account, then, is not successful. It has no way of accounting for the contrast between R-dative shift and a real prepositional dative construction, since it treats them both the same.

3 Locative Inversion

O&R’s alternative runs into the same problem with locative inversion. In Bruening 2010b, I showed that ordinary prepositional dative constructions can undergo locative inversion when passivized, but sentences that involve R-dative shift cannot.

(17)

  • a.

    To the generals that lost the battle were given helicopters.

  • b.

    *To the generals that lost the battle were/was given hell. (Bruening 2010b:297, (27a–b))

In this, R-dative shift patterns with double object constructions, which cannot participate in locative inversion (Bresnan 1994:79n9, Postal 2004:47).

(18)

  • a.

    *At that time was/were given the generals who lost hell. (double object construction)

  • b.

    *At that time was given to the generals who lost hell. (R-dative shift)

Double object constructions and prepositional datives contrast in this respect.9

(19)

  • a.

    At that time was given to the Son of Man dominion and glory and the kingdom that all peoples, nations, and tongues should serve him.

  • b.

    *At that time was given the Son of Man dominion and glory and the kingdom that all peoples, nations, and tongues should serve him.

(20)

  • a.

    At that time was sent to the embattled forces the all-powerful demon Raktabija.

  • b.

    *At that time was sent the embattled forces the all-powerful demon Raktabija.

Although I did not propose an account of the failure of locative inversion with double object constructions in Bruening 2010b, I did reduce the ungrammaticality of (17b) and (18b) to the same factor that rules out locative inversion with double object constructions (18a)/(19b)/(20b). Sentences with R-dative shift are double object constructions, in my analysis.

O&R also attempt to reduce the ungrammaticality of (17b) and (18b) to something else. They note that locative inversion often requires the passive, and passives are not allowed with VPs like those in (17b) and (18b).

(21)

So, both analyses attempt to explain the ungrammaticality of locative inversion by relating it to something else. Let us see which analysis leads to an explanation of the facts.

In O&R’s analysis, (21) is simply a prepositional dative construction. Minimally different prepositional dative constructions can undergo passivization.

(22) Helicopters were given to the generals that lost the battle.

It is not clear why the passive would not be allowed in (21) if it is simply a standard prepositional dative construction.10 It is then also not clear why locative inversion in (17b) and (18b) is not allowed.

In contrast, the R-dative shift analysis explains why (21) is ungrammatical. It is ungrammatical for the same reason that the following is ungrammatical:

(23) *Hell was given the generals that lost the battle.

In many dialects of English, the second object of a double object construction cannot passivize across the first object, especially if the first object is a full NP and not a pronoun. In the R-dative shift analysis, (21) is also a double object construction, and so passivization of the lower object across the higher is not permitted (and since the higher NP must be heavy to license the rightward specifier, it cannot be a pronoun).11

In addition, the R-dative shift analysis can be combined with an existing account to explain why locative inversion is ungrammatical with double object constructions, including R-dative-shifted ones. This same explanation is not available to O&R’s alternative.

First, note that locative inversion does not always require the passive; it is compatible with intransitive verbs as well.12

(24)

  • a.

    Under the bridge was living a troll.

  • b.

    In the center of the elegant ballroom danced several couples.

  • c.

    On the siding puffed a little train engine.

Transitive versions of these same sentences are ungrammatical with locative inversion, regardless of the word order.

(25)

  • a.

    *Under the bridge was living the good life a troll.

  • b.

    *In the center of the elegant ballroom danced a waltz several couples.

  • c.

    *On the siding puffed big puffs of steam a little train engine.

(26)

  • a.

    *Under the bridge was living a troll the good life.

  • b.

    *In the center of the elegant ballroom danced several couples a waltz.

  • c.

    *On the siding puffed a little train engine big puffs of steam.

The generalization about locative inversion is that there can be only one NP argument in the sentence. This is what is wrong with locative inversion involving passive double object constructions, as in (18a) and (19b): there are too many NP arguments. Intransitive clauses and passive clauses permit locative inversion because they have only one NP argument.

A theoretical account of this restriction has been proposed by Alexiadou and Anagnostopoulou (2001, 2007). These authors motivate a generalization they call the Subject-in-Situ Generalization, which says that no more than one NP that requires structural case may remain inside the VP (VoiceP in the analysis here). If the external argument remains in its base-generated position in Spec,VoiceP, then there may be no object inside the VoiceP at the same time.

To explain the facts of double object constructions, I will adopt the theoretical account of the Subject-in-Situ Generalization proposed in Alexiadou and Anagnostopoulou 2007:sec. 5.2, modified to fit the structures here. In this account, case checking under Agree (Chomsky 2000) must wait until Voice, the head that checks accusative case, enters into an Agree relation (i.e., Agrees) with T, the head that checks nominative case. Once Voice Agrees with T, these two heads can proceed to Agree with NPs and check their case features. A problem arises, however, if there are two NPs and two case features: the grammar faces an indeterminacy or ambiguity problem, because it does not have clear instructions on which case to check first. The derivation crashes because of this indeterminacy. Now, if either the subject or the object moved, this must have been accomplished by an EPP feature on either T or Voice. In Alexiadou and Anagnostopoulou’s account, EPP features have the property of resolving the indeterminacy. An EPP feature on T tells the system to first check case by setting up an Agree relation between T and the NP attracted by T’s EPP feature (the subject). An EPP feature on Voice tells the system to first check case by setting up an Agree relation between Voice and the NP attracted by Voice’s EPP feature (the object). If both T and Voice have EPP features, then Agree proceeds strictly cyclically. The indeterminacy problem only arises if there are no EPP features, and hence neither the subject nor the object moves.

A natural question is how an active double object construction can be allowed in this system, since there are two objects, both of which seem to remain inside VoiceP in English. I propose that Alexiadou and Anagnostopoulou’s indeterminacy problem only arises when there are two cases to be checked that are distinct. In the constructions where the problem arises, we have nominative case that is checked by T, and accusative case that is checked by Voice. In double object constructions, however, we have two objects that seem to receive the same case. I propose that active Voice, once it Agrees with T, is capable of checking more than one object case, in an instance of multiple Agree (e.g., Hiraiwa 2001). Since the two object cases are identical, indeterminacy is not a problem. The grammar can check them in either order. In the passive of a double object construction, however, the two cases will be distinct again: nominative Agreeing with T, and accusative Agreeing with Voice. The indeterminacy problem will arise again if neither object moves out of VoiceP.

In other work (Bruening 2010c), I proposed that in locative inversion, the NP that agrees with the finite verb is in a rightward specifier of VoiceP. If this NP was an underlying object, it moved to this position. If it was an underlying subject, as in the sentences with unergative verbs above, it was base-generated in this position. Either way, it is inside the VoiceP. The constraint that is behind the Subject-in-Situ Generalization then forbids there to be an object NP inside the VoiceP as well.

As an illustration, take the case of (18a), the passive of a double object construction. This would have the VoiceP in (27) (see section 4 on vP, which is irrelevant here).

(27)

graphic

One of the two NPs must move to the empty specifier of VoiceP; given economy, it must be the higher one. I assume that in the passive, Voice always has an EPP feature. This is visible in the obligatory shift observed with expletive passives (the thematization/extraction rule of Chomsky 2001).

(28)

  • a.

    There was a study done in 1979.

  • b.

    *There was done a study in 1979.

I propose that, since this EPP feature is a required property of the passive Voice head, it is not used as an instruction in Alexiadou and Anagnostopoulou’s case-checking algorithm presented above. It is simply a lexical property of passive Voice. Hence, only a second EPP feature on Voice in the passive is interpreted as an instruction on how to proceed with case checking (a second EPP feature would trigger movement of a lower object, as in, for instance, wh-movement).

So, in tree (27), the generals who lost will move to Spec,VoiceP. In locative inversion, this NP moves no further. The other object also does not move (so the word order derived would actually be *At that time was/were given hell the generals who lost, which is also ungrammatical). The EPP feature on Voice in the passive is not an instruction for how to proceed with case checking, so the derivation crashes. T and Voice, having Agreed, now have two cases they need to check, but no instructions on how to do it.

Locative inversion with an active transitive verb would be similar, except that this time the NP in the rightward specifier of Voice would have been base-generated there. Again, it moves no further, nor does the object. Again, indeterminacy of case checking results and the derivation crashes, as for example in (25)–(26).

This accounts for the failure of locative inversion with transitive verbs and with passives of double object verbs. What about R-dative-shifted double object verbs? They will also fall under the account, if the NP that is marked with to is actually an NP that requires structural case. This is an important part of the analysis based on the Subject-in-Situ Generalization: it only governs NPs that require structural case. PPs are irrelevant. We are then led to conclude that the apparent to-PP is not actually a PP; instead, it is an NP that requires structural case, and to is perhaps a marker of this structural case, or is there to perform some other function. It is not a preposition that itself assigns case to its complement. The structure that underlies (18b), then, is exactly the same as (27), except that the specifier of ApplP is projected on the right, and the case marker to appears as part of the NP in Spec,ApplP. The derivation crashes for exactly the same reason that (27) did.

The R-dative shift analysis, then, is able to explain the failure of locative inversion with R-dative shift by appeal to an analysis that has been independently proposed. This analysis explains far more than just this fact (see Alexiadou and Anagnostopoulou 2001, 2007) and so is independently justified. We have also concluded something about the nature of the preposition to in R-dative shift: it is a case marker, not a full-fledged preposition.

Note that this explanation in terms of the Subject-in-Situ Generalization is not available to O&R’s alternative analysis. In their analysis, the sentences that I am analyzing in terms of R-dative shift are just regular prepositional dative constructions. When passivized, these have only one NP needing structural case. They should therefore be able to form a locative inversion sentence. O&R simply have to stipulate that certain apparent prepositional dative constructions cannot undergo passivization, and hence also cannot undergo locative inversion. But this is only a stipulation, not an explanation.

To summarize so far, in the R-dative shift analysis, apparent prepositional dative constructions actually divide into two types. There are real prepositional dative constructions, which can passivize and undergo locative inversion, and also admit scope ambiguities between the internal NP and PP arguments. Then there are R-dative-shifted double object constructions, which cannot passivize or undergo locative inversion, and which exhibit frozen scope between the internal NP and PP. These empirical differences are real, and are explained by the R-dative shift analysis combined with independently motivated proposals about economy of movement and case checking. O&R’s alternative, which treats all apparent prepositional dative constructions as actual prepositional dative constructions, cannot explain why they would divide into two distinct types exhibiting different behavior.

The two different classes of apparent prepositional dative constructions can also be observed in ditransitive idioms, which O&R imply are inconsistent with the R-dative shift analysis. There are idioms that can undergo passivization and locative inversion: these are true prepositional dative constructions. (Most of these also alternate with double object constructions.) Then there are idioms that cannot undergo passivization or locative inversion in the apparent prepositional frame. These I analyze as being limited to double object constructions, but like all double object constructions, they can sometimes undergo R-dative shift and appear to be prepositional dative constructions. The two categories are given in (29) and (30).

(29) Idioms that can be true prepositional datives

  • a.

    throw NP a bone~throw a bone to NP

  • b.

    read NP the riot act~read the riot act to NP

  • c.

    lend NP a hand~lend a hand to NP

  • d.

    lend NP a sympathetic ear~lend a sympathetic ear to NP

  • e.

    give NP a wide berth~give a wide berth to NP

  • f.

    give NP the cold shoulder~give the cold shoulder to NP

  • g.

    lend NP color~lend color to NP

(30) Idioms that can only be double object constructions

  • a.

    give NP the boot

  • b.

    give NP the sack

  • c.

    give NP the creeps

  • d.

    give NP pause

  • e.

    give NP a piece of one’s mind

  • f.

    give NP hell

  • g.

    promise NP the moon

To exemplify, read the riot act to NP can be passivized, as can read NP the riot act.

(31)

  • a.

    Essentially, the riot act was read to the tone-deaf Mormon church by prominent, influential interests in behalf of the faith and families of Jewish Holocaust victims . . .

    (http://exmormon.org/phorum/read.php?2,443418,443422#msg-443422)

  • b.

    The editors of these three papers were summoned to Downing Street and were read the riot act.

    (Geoffrey Morgan, A Call to Arms, page 117, accessed by Google Books)

In principle, read the riot act to NP ought to be able to undergo locative inversion, as well. My own intuition is that this is correct.

(32) To anyone who disobeys these instructions will be read the proverbial riot act.

However, at least one speaker disagrees, finding (32) ungrammatical. This is not surprising, though, as both idioms and locative inversion are known to be delicate.

In contrast, as shown above, give NP hell, although it does sometimes appear as give hell to NP, cannot undergo passivization or locative inversion when it does (see, e.g., (17b), (18), (21)). Similarly, although examples of give the boot to NP and give the creeps to NP occur on the Internet, no examples of *the boot was given to NP or *the creeps were given to NP can be found, and native speakers judge them ungrammatical. Attempts at locative inversion also fail.

(33)

  • a.

    *To Bill was given the boot.

  • b.

    *To our intrepid explorers were given the creeps.

So, idioms that appear to be prepositional dative constructions also divide into two categories, just like the nonidiomatic ditransitives discussed above. The R-dative shift analysis explains why these two classes of apparent prepositional dative constructions—idiomatic (give hell to NP ) or not (give a headache to NP)—pattern differently in passivization, locative inversion, and scope. O&R’s alternative does not, since it analyzes them in exactly the same way.

4 Structural Asymmetries

O&R also criticize the R-dative shift analysis on the basis of its predictions regarding structural asymmetries. In the R-dative shift analysis, the NP marked with to in an R-dative-shifted extended VP is structurally higher than the NP direct object (the theme), although it occurs to the right of that NP. According to O&R, this incorrectly predicts that phenomena like anaphor binding, variable binding, weak crossover, superiority, NPI (negative polarity item) licensing, and the each . . . the other construction should go backward, from right to left. This issue is not as simple as O&R make it out to be, since R-dative shift is always necessarily followed by Ā-movement, because of the Extraction Constraint on Rightward Specifiers. Nevertheless, weak crossover with wh-movement makes the point nicely. Wh-movement licenses R-dative shift, so many instances of extraction out of a to-PP should be ambiguous. Since the R-dative shift parse should be available, there should be no weak crossover, according to O&R, but this is false.

(34)

  • a.

    Which check1 did you send t1 to its1 owner?

  • b.

    *Which worker1 did you send his1 check to t1 ?

The sentence in (34b) should have a parse where the base position of which worker c-commands his (i.e., in a rightward specifier of ApplP), and so there should be no weak crossover, according to O&R.

O&R follow most of the literature in assuming that c-command is the relevant structural notion for all structural asymmetries. However, this is false, as other authors and I have documented extensively (Hoeksema 2000, Barker 2012, Bruening 2014). Precedence is actually of prime importance for all of these phenomena, and c-command is irrelevant. This is easy to show for weak crossover. Consider the following example, from Reinhart 1976. The lack of a Principle C effect here shows that the adjunct CP is high, outside the c-command domain of the object.

(35)

This conclusion is further confirmed by the fact that this type of adjunct cannot be fronted along with the VP in VP-fronting, but must be stranded.

(36)

  • a.

    *. . . and write to Brando [that he couldn’t answer them all], so many people did.

  • b.

    . . . and write to Brando, so many people did [that he couldn’t answer them all].

The adjunct in this construction is obligatorily high, then, outside of VP, and definitely outside of the c-command domain of the object. If c-command were necessary to avoid weak crossover, we would expect a weak crossover violation in this construction when the object is an extracted wh-phrase and Brando is replaced with a pronoun to be bound by the wh-phrase. But this is false.

(37) Who1 did so many people write to t1 [CP that he1 couldn’t answer them all]?

Clearly, it is not necessary for the trace of a wh-phrase to c-command a pronoun in order for the wh-phrase to bind the pronoun.

C-command is not relevant to weak crossover, then. Precedence is. In (37), the base position of the wh-phrase precedes the pronoun, and binding is licit. In all cases of weak crossover, the base position of the wh-phrase follows the pronoun.

(38) *Who1 are [his1 many fans] always writing to t1 ?

The right generalization about weak crossover is that the base position of the wh-phrase must precede the pronoun to be bound (cf. Shan and Barker 2006). This means that O&R’s criticism of the R-dative shift analysis is off the mark, because c-command is not relevant to any of the structural asymmetries they discuss. Precedence is what matters. Since the theme NP precedes the R-dative-shifted NP in the rightward specifier of ApplP, we see all of these asymmetries going left to right rather than right to left. The R-dative shift analysis in fact makes the correct predictions, once we understand the role of precedence and the irrelevance of c-command.

Moreover, O&R’s own analysis runs into a paradox if c-command is the relevant notion, as they assume. As described above, they posit (apparently obligatory) heavy shift of the to-PP in examples like the following:

(39) The boss denied that position [to every employee who requested it].

This heavy shift puts the PP high, adjoined to VoiceP. It should therefore be outside the c-command domain of the theme, here that position. We should then expect that switching that position and the pronoun it should not lead to a Principle C violation. This is false.

(40) *The boss denied it1 to every employee who requested that position1 .

Note that it will not do to say that heavy shift must reconstruct, for two reasons. First, this would nullify O&R’s account of the scope facts; and second, reconstruction does not seem to be possible in other cases of heavy shift.

(41)

If reconstruction were possible in (41), there would be no Principle C violation, because there is none in the base position.

(42) We gave John1’s brand-new toy to him1 on Friday.

In sum, O&R’s analysis plus c-command makes the wrong predictions concerning structural asymmetries like Principle C. In contrast, the R-dative shift analysis, using precedence, gets the facts exactly right, as I now show. First, although c-command is not involved in any syntactic phenomenon, there is an important structural relation that is involved in coreference in addition to precedence. This is a much coarser relation than c-command, which I call phase-command.13

(43)

  • Phase-command

  • X phase-commands Y if and only if neither X nor Y dominates the other and there is no ZP, ZP a phasal node, such that ZP dominates X but does not dominate Y.

(44)

  • Phasal nodes

  • CP, VoiceP, NP

C-command is hypersensitivity to structure: every node counts. Phase-command says that only certain nodes count: the phasal nodes. In Bruening 2014, I show that phase-command achieves much better empirical coverage than c-command.14

Importantly here, Principle C is formulated in terms of phase-command and precedence.

(45)

  • Principle C

  • An R-expression may not be coindexed with an NP that precedes and phase-commands it.

I update the analysis of double object constructions in Bruening 2010b to that in Bruening 2010a, where there is an additional projection, vP, between ApplP and VoiceP. Heavy shift, I propose, moves the shifted NP (marked with the case marker to) from the rightward Spec,ApplP and adjoins it to vP as in (46).

(46)

graphic

The pronoun it phase-commands the R-expression that position in this structure, because there is no phasal node that dominates it but does not dominate that position. The first phasal node that dominates it is VoiceP, which also dominates that position. Since it also precedes that position, Principle C is violated.

This theory also accounts for (41) vs. (42). In (42), the pronoun does not precede the R-expression, but in (41), the pronoun precedes and phase-commands the R-expression (which is contained in a phrase adjoined to vP).

To summarize this section, the structural asymmetries that O&R bring up actually raise problems for their analysis, if their assumptions are adopted. Once their assumptions are set aside, and we understand the role of precedence, we can see that these asymmetries are in no way problematic for the R-dative shift analysis. In fact, the R-dative shift analysis gets the facts exactly right.

5 Scope and Idioms

As described above, O&R try to account for fixed scope as an effect of heavy shift. However, they also suggest a role for idiomatic interpretations in fixing scope, as revealed in pair-list readings in particular. For instance, they point out, correctly, that pair-list readings are not available in examples like the following (when they are interpreted idiomatically):

(47)

These examples do not involve potential double object constructions, so there is no way to say that they involve R-dative shift. This means that some other factor(s) besides R-dative shift can also fix scope.

O&R suggest that one such factor is idiomatic interpretations: many idiomatic expressions do not lend themselves to scope interactions. In this they are surely correct, and this is probably the right account of their examples in (47). However, they also appear to be suggesting that all of the data presented in Bruening 2010b can be accounted for by this factor, given that many of the cases that I analyzed as instances of R-dative shift involve idiomatic interpretations. However, in presenting the scope facts, I deliberately avoided idiomatic expressions, precisely because they do not lend themselves to scopal interactions. In fact, none of the scope examples presented in Bruening 2010b involved idiomatic interpretations. All of them are interpreted in a wholly literal way. I repeat all of the crucial examples here:

(48)

  • a.

    This lighting gives every kind of headache to a different (type of) person. (*every > a)

  • b.

    Let’s spare every ordeal to someone or other (who comes before us today). (*every > a)

  • c.

    The bosses denied every position to some applicant or other from within the bureau. (*every > a)

    (Bruening 2010b:294, (14c); 295, (17d), (18d))

(49)

  • a.

    #Let’s make a list of who he denied every request to. (*pair-list)

  • b.

    #Let’s make a list of who this lighting gives every kind of headache to. (*pair-list)

    (Bruening 2010b:296, (24b); 297, (25b))

As can be seen, none of these examples involve idiomatic interpretations. Rather, they are all nonalternating but literally interpreted phrases. One could perhaps argue that give X a headache has some special interpretation that might be relevant to scope, but there is a more general point to be made here, as well: namely, it is important to establish that the NPs involved can in principle engage in scope interactions. This point was in fact made in Bruening 2010b. Consider the first set of examples presented there, involving give X a headache.

(50)

  • a.

    This lighting gives everyone a different kind of headache. (every > a)

  • b.

    This lighting gives a different person every kind of headache. (*every > a)

  • c.

    This lighting gives every kind of headache to a different (type of) person. (*every > a)

  • d.

    This lighting gives a different kind of headache to everyone who enters the room. (every > a)

    (Bruening 2010b:294, (14))

In (50a) and (50d), the indirect object can be a universal quantifier that can distribute over the direct object. There is simply no way to claim that one of these NPs is idiomatic in some way that prevents it from participating in distributive readings. The NPs clearly can interact in a distributive fashion. Note also that the direct object can be a universal quantifier that can distribute over something else—for instance, an adjunct.15

(51) (Context: There are four types of headaches that I regularly suffer from: tension headaches, cluster headaches, sinus headaches, and migraine headaches.)

  • a.

    I was given every kind of headache by a different electronic device.

  • b.

    I was given every kind of headache at a different point during the day.

These examples show that there is nothing about the phrase give X a headache that stops a quantified version of headache from participating in scope interactions.

In contrast, the idiomatic expressions that O&R point to seem to be scopally inert (when interpreted idiomatically).

(52)

  • a.

    These photographs brought everything home to a (#different) person. (no distributivity)

  • b.

    The journalist put every word in a (#different) source’s mouth. (no distributivity)

(53)

  • a.

    These photographs brought everything home to me at a (#different) point during the day.

  • b.

    The journalist put every word in that source’s mouth at a (#different) point during the interview.

O&R’s examples of idiomatic expressions, then, are completely different from the examples of fixed scope presented in Bruening 2010b. The latter cannot be explained in the same way, and they require a syntactic analysis like the one I proposed there.

6 On Semantics, Verb Sensitivity, and Crosslinguistic Variation

As described in section 1, a large body of work has tried to deny that there are semantic differences between the double object construction and the prepositional dative construction (e.g., Rappaport Hovav and Levin 2008, Bresnan and Nikitina 2009), including some of O&R’s other work (Ormazabal and Romero 2010). All of this work argues that the two are simply two alternating syntactic expressions of the same semantics. Ormazabal and Romero (2010), in particular, argue that the double object construction is derived from the prepositional dative construction by a process of preposition incorporation (following Baker 1988).

One thing that this literature has shown is that certain instances of the prepositional dative construction, with certain verbs (chief among them give), may have a semantics that is very similar to that of the double object construction. In particular, give has a meaning of caused possession even in the prepositional dative construction, and does not simply involve change of location, as some work has claimed (e.g., Harley 2002, Krifka 2004).

However, it is a leap from that small observation to the claim that double object constructions and prepositional dative constructions are identical. What all of this literature downplays, and what I wish to emphasize here, is that the double object construction, in contrast with the prepositional dative construction, has a completely uniform semantics, regardless of the verbs involved (even when that meaning is negated, as with spare and other nonalternating verbs).16 This is tacitly acknowledged in the following table from Rappaport Hovav and Levin 2008:132, (6); note the last column in particular.

(54)

graphic

That is, double object constructions, unlike prepositional dative constructions, have a uniform caused-possession meaning. This uniformity is unexpected in a theory like that of Ormazabal and Romero (2010), or any theory that derives one construction from the other. In a preposition incorporation theory, the double object construction should have exactly the same meaning as the prepositional dative construction from which it is derived. This is not true, however. In the following examples (with Rappaport Hovav and Levin’s “throw-type verbs”), the double object construction only has the meaning of (intended) caused possession, while the prepositional dative correspondent has other meanings available to it:

(55)

  • a.

    I kicked {her/*the goal line} the ball.

  • b.

    I kicked the ball to {her/the goal line}. (endpoint of motion)

(56)

  • a.

    I took {him/*the windowsill} a cup of coffee.

  • b.

    I took a cup of coffee to {him/the windowsill}. (endpoint of motion)

Semantic differences are even clearer when the preposition is not to, but for (not discussed in Rappaport Hovav and Levin 2008).

(57)

  • a.

    Melinda melted me some ice cream.

  • b.

    Melinda melted some ice cream for me.

(58)

  • a.

    He brought me some beer.

  • b.

    He brought some beer for me.

The (a) examples here only have an intended-caused-possession meaning. In contrast, the (b) examples have (at least) two possible meanings: (1) a proxy reading, where the surface subject performs the action in place of the object of for; (2) a vague benefactive reading, where the object of for receives some benefit from the action. Reading (2) is compatible with caused possession, but I would argue that reading (2) is simply vague, and there is no distinct caused-possession reading with for. The proxy reading is clearly distinct, because it disappears when the PP is fronted.

(59) For me, she melted some ice cream. (*proxy reading)

I know of no context that teases apart caused possession and vague benefit. This suggests that for-phrases simply have a vague benefactive meaning, which caused possession is compatible with.

Additionally, where caused possession is not compatible with the meaning of the verbs and NPs involved, no double object variant is possible.

(60)

  • a.

    She shoveled the sidewalk for me.

  • b.

    *She shoveled me the sidewalk.

(61)

  • a.

    She answered the phone for me.

  • b.

    *She answered me the phone.

(62)

At the same time, the for-phrase does not seem to differ in its semantics with these verbs as opposed to the verbs that allow the double object construction. In both cases, it can have a proxy reading or a vague benefactive reading. I see no reason to think that the vague benefactive reading with these verbs differs in any way from the vague benefactive reading that we see with verbs that do permit the double object construction. If this is correct, then for-phrases do not actually have a caused-possession meaning; they only have a vague benefactive meaning.

In contrast, the double object variants only have a caused-possession meaning. If I am correct that for-phrases only have a vague benefactive meaning, then the meaning of the double object construction is a meaning that the prepositional dative correspondent does not have.17

All of the above raises two problems for a derivational approach: (1) how to explain the complete uniformity of the double object construction, when that construction is derived from something that does not exhibit this uniformity (to-datives); and (2) how to derive a construction with one meaning from another construction that lacks that meaning (for-benefactives). It might be possible to overcome the first problem, but the second is intractable.

Yet another problem with such approaches is the existence of nonalternating verbs that only appear in the double object construction. Some examples are the following:

(63)

  • a.

    That word lost me the fifth-grade spelling bee.

  • b.

    *That word lost the fifth-grade spelling bee to/for/at/(up)on me.

(64)

  • a.

    I envy you your equanimity.

  • b.

    *I envy your equanimity to/for/at/on you.

Most of these have a negative meaning, negating the caused possession. (For an analysis, see Bruening 2010a.) The problem here is that there is no source for the double object construction, if double object constructions are all derived from prepositional dative constructions (but see Larson 1988 for a derivational analysis that attempts to grapple with such facts).

These three problems render a derivational analysis quite implausible. The balance of the evidence tilts in the direction of an analysis where double object constructions and prepositional dative constructions are distinct and derivationally unrelated.

Moreover, even when a prepositional dative construction may have a caused-possession meaning, there appear to be significant differences between this meaning and the caused-possession meaning that double object constructions have. This is what is behind the frequently expressed intuition that certain phrases only permit the double object construction, like give NP a headache. Much of the work cited above, in particular Rappaport Hovav and Levin 2008 and Bresnan and Nikitina 2009, has tried to deny these differences, primarily by citing corpus data indicating that such phrases do in fact appear as an apparent prepositional dative. The significance of the facts that support an R-dative shift analysis is that they show that the intuition was correct after all, and apparent instances of prepositional datives with these expressions are actually double object constructions. In passivization, locative inversion, and scope interactions they pattern exactly like double object constructions, and do not pattern with real prepositional dative constructions.

This means that Rappaport Hovav and Levin’s (2008) table in (54) is not quite accurate: the “caused-possession” meaning that prepositional dative constructions may have is not the same as the “caused-possession” meaning that double object constructions have. In Bruening 2010a, I tried to capture this by giving certain prepositional datives a “recipient” semantics, while the double object construction involves causation of a having eventuality, where the indirect object is a “possessor” in that eventuality (see also work by Harley on identifying a “have” component in double object constructions; e.g., Harley 1997, 2002). Of course, the success of this approach depends on having precise definitions of recipient and possessor roles, and independent ways of identifying and distinguishing them, but I leave that to future research (and there may well be better ways of capturing the difference). What is important here is that such differences are real and must be accounted for, and it is incorrect to view the two constructions as having the same semantics.

Additionally, many English speakers have the intuition that R-dative shift is marked and requires special licensing, an intuition that I formalized as the Extraction Constraint on Rightward Specifiers. O&R claim that examples that I analyze as cases of R-dative shift do not need special licensing, but their examples do not support this claim. One (their (18c)) involves heavy shift; the other, which I repeat here, requires heavy focal stress on the NP after to, and the context makes it contrastive.

(65)

I contend that the special licensing that is required, plus the locative inversion, passivization, and scope facts, all indicate that R-dative shift is real. Certain verb phrases can only be expressed as double object constructions; when they appear to be prepositional datives, the NP in Spec,ApplP has actually been projected to the right and undergone Ā-extraction.

Finally, O&R voice one other criticism of the R-dative shift analysis. This is that, according to them, a theory with two different structures cannot account for double object constructions in numerous other languages besides English. However, this criticism is based on a conflation of very different constructions. Much other work of mine and my colleagues (Bosse and Bruening 2011, Bosse, Bruening, and Yamada 2012, Bosse 2015; see also Cuervo 2003) has shown that there are several different types of applied arguments with very different properties that cannot be conflated: for instance, benefactives vs. affected experiencers vs. ethical datives and so on. The syntactic and semantic properties of these different classes of applied argument are very different, and they are all different from the English-type caused-possession applicative. O&R treat them as a single thing, a monolithic double object construction. This conflation is not justified by the facts.

As illustration, consider examples like (66), discussed in Christian 1991, Webelhuth and Dannenberg 2006, Conroy 2007, Horn 2008, and Bosse, Bruening, and Yamada 2012.

(66) Davy Crockett killed him a bear when he was only three.

Superficially, this is a double object construction: the first object is him, and the second object is a bear. According to O&R’s logic, we should analyze this type of example in the same way as a caused-possession double object construction. However, there are numerous differences between them: (1) the semantics are entirely distinct (in (66), some kind of extra involvement or satisfaction/success on the part of the subject); (2) there is no prepositional dative corresponding to (66); (3) the first object is obligatorily coreferential with the subject in (66); (4) nevertheless, it typically takes the form of a pronoun, not a reflexive; (5) the first object in (66) cannot be questioned and cannot be a quantifier (it is obligatorily a coreferential pronoun). Additionally, examples like (66) differ from caused-possession double object constructions in ways that make it clear that the first object in (66) contributes its meaning as something like an implicature, and not as at-issue content (Horn 2008). For instance, the first object does not add a condition to a conditional.

(67)

There is no difference between the two members of this pair in the conditions for receiving the coffee. So long as the speaker sits, he or she can expect some coffee (assuming the interlocutor agrees). The sentence with the additional pronoun only adds the implicature that the speaker will be more involved in, or get some kind of enjoyment out of, sitting, but that makes no difference to the conditional.

In contrast, the first object in a caused-possession double object construction does add a condition to a conditional.

(68)

  • a.

    If you melt me some ice cream, I’ll give you a thousand dollars.

  • b.

    If you melt some ice cream, I’ll give you a thousand dollars.

In (68a), if the addressee melts some ice cream, but does not do so with the intention that the speaker receive it, the addressee will not be entitled to the thousand dollars. In (68b), he or she will be.

To give one more example, the additional meaning in sentences like (66) cannot be negated with sentential negation (Horn 2008).

(69) #Davy Crockett killed a bear when he was only three, but he didn’t kill him a bear when he was only three. (contradiction)

Sentential negation can only negate the at-issue content of the second clause in (69), which makes the entire sentence a contradiction. In contrast, sentential negation can target the caused-possession meaning of a caused-possession double object construction.

(70) Davy Crockett whittled a flute, but he didn’t whittle me a flute. (not a contradiction)

These differences, and numerous others (see especially Horn 2008), indicate that in a caused-possession double object construction, the first object (and the meaning of caused possession) is part of the at-issue content of the sentence. In examples like (66), it is not; it is something like an implicature. These facts indicate that these two types of sentences, both of which are superficially “double object constructions,” are quite distinct and should be analyzed in very different ways.

Though less dramatic, similar differences can be shown for benefactives, affected experiencers, ethical datives, and so on; see the works cited above, especially Bosse, Bruening, and Yamada 2012 and Bosse 2015. To the extent that such “double object constructions” can be shown to differ in significant ways, they should be analyzed differently. It is therefore not a disadvantage of a particular analysis of caused-possession double object constructions that it cannot cover other types of double object constructions. It is actually an advantage, since the differences between them are legion. (Of course, if another language does have a caused-possession double object construction, that construction should be analyzed in the same way as the English one, unless significant differences indicate otherwise.)

7 Conclusion and Discussion

To conclude, O&R’s criticisms of R-dative shift actually support the R-dative shift analysis when they are considered more carefully. Their own analysis stumbles because it treats all apparent prepositional dative constructions in the same way. But the empirical facts indicate that apparent prepositional dative constructions divide into two types: real prepositional dative constructions, and R-dative-shifted double object constructions. The latter pattern in every way with double object constructions. They are double object constructions, in the R-dative shift analysis, and we have seen reasons to think that the to that appears with R-dative shift is not a preposition at all but some kind of case marker.

Additionally, the analysis that I have presented interacts with the right theory of command phenomena like weak crossover and coreference (Principle C) to predict exactly the right array of facts. The theoretical account of the Subject-in-Situ Generalization (Alexiadou and Anagnostopoulou 2001, 2007) explains why R-dative shift is incompatible with locative inversion, on the assumption that the NP marked with to is an NP argument that requires structural case, exactly like the first object of a double object construction. All of this supports the conclusion that R-dative shift is a valid, internally consistent, and explanatory account of the data under analysis. For any proposed alternative to be viable, it will have to match the R-dative shift analysis on all of these counts.

Finally, I have presented arguments that double object constructions and prepositional dative constructions have distinct underlying forms and are not derivationally related. The facts that motivate the R-dative shift analysis further bolster this conclusion.

Notes

1 Note that similar examples involving idioms were presented as early as Larson 1988; Larson’s examples, like those from Bresnan et al. 2007, seem to involve heavy shift and so are amenable to the R-dative shift analysis.

2 Note that there is crosslinguistic precedent for distinguishing prepositional dative constructions from double object constructions that are marked with a preposition: Anagnostopoulou (2005) argues that certain apparent prepositional dative constructions in Greek, and possibly also French and Spanish, actually have properties of double object constructions and must be distinguished from prepositional datives.

3 In this article, I continue to use the structure from Bruening 2010a. However, given the findings regarding c-command in Bruening 2014, discussed in section 4, an alternative is the following:

  • (i)

    graphic

The structure in (i) is consistent with all of the binding facts and seems to be more compatible with constituency tests than the one in the text.

4 There must be both rightward initial projection of Spec,ApplP and Ā-extraction, because wh-movement can strand to on the right: How many people did this stench give a headache to? The preposition is stranded in the rightward specifier, while the wh-phrase undergoes wh-movement to the left. See Bruening 2010b.

5 At this point, the Extraction Constraint on Rightward Specifiers is simply a stipulation. However, it is possible to relate it to more general facts about English and the workings of syntax. Below, I will present an analysis of locative inversion that also has an NP in a rightward specifier. This rightward specifier also requires special licensing and is not freely available (see Bruening 2010c). It therefore appears that it is a fact about English that projecting a specifier on the left is the default, and projecting it on the right is sometimes permitted but requires special licensing. It is probable that this relates to some of the facts that have been taken to motivate the antisymmetry approach to syntax (Kayne 1994), although I do not subscribe to any of the tenets of antisymmetry. Nevertheless, there are numerous facts indicating that syntax is not symmetric the way classical X-bar theory predicts, and the hope is that the special licensing that is required for rightward specifiers in English will be explained by whatever explains this crosslinguistic lack of symmetry.

6 Note that I use the indefinite some X or other in all of these examples. This indefinite typically brings out the distributive reading quite readily. This helps to ensure that where the distributive reading is unavailable, the failure must be due to the syntax and not to some pragmatic difference between the examples, since they are minimally different. Note also that it is not the case that the first object in the double object construction must be topical and therefore a specific indefinite. A subject quantifier can distribute over the first object quite easily.

  • (i)

    Every judge spared someone or other that ordeal today. (every > some)

The first object clearly does not have to be interpreted with wide scope or as a specific indefinite. This means that there is not some reason other than the syntax that the second object could not take scope and distribute over the first.

A note on the judgments presented here: Every sentence in this article has been checked with several native speakers of English, including anonymous journal reviewers. No one has disputed the judgments, and several speakers have even remarked on how striking they are.

7 Note that this way of describing locality or economy does not need to refer to c-command; it need only refer to dominance. This becomes relevant in section 4.

8 A series of arguments against small clause accounts of prepositional dative constructions can be found in Bruening 2010a:522–525. These provide further reasons to reject O&R’s alternative.

9 Note that the account of the failure of locative inversion with double object constructions based on the Subject-in-Situ Generalization described below predicts that locative inversion will become more acceptable if one of the NPs is heavy and undergoes heavy shift. Survey data bear this out (Bruening 2016). Examples like (19b) and (20b) might therefore be judged marginal rather than fully ungrammatical by some speakers, since the NP on the right is somewhat heavy.

10 A reviewer suggests that bare Ns like hell do not undergo passivization easily. This is simply not true of numerous other bare Ns with give (i), and modifying hell makes the passive no better (ii).

  • (i)

    • a.

      Consideration should be given to issuing a direction under Regulation 25.

    • b.

      Preference at that time was given to a scale of 1:1.000.

    • c.

      Credence was given to Magalotti’s belief that . . .

  • (ii)

    • a.

      They gave her a lot of hell for that.

    • b.

      *A lot of hell was given to her for that.

It is therefore unlikely that the NP’s status as a bare noun has anything to do with the failure of passivization.

11 A reviewer points out that dialects that do allow passivization of the second object should permit examples like (21) (and (23)). Unfortunately, I have not been able to locate speakers of the relevant dialects and so have been unable to check this.

12 Note that it is not the case that locative inversion is only grammatical with unaccusatives. The verbs in (24) in combination with the same PPs pass standard tests for unergatives: they can form pseudopassives (i), and resultatives are only grammatical with a fake reflexive (ii).

  • (i)

    • a.

      This bridge is being lived under!

    • b.

      This elegant ballroom has been danced in on numerous occasions.

  • (ii)

    • a.

      Many couples have danced *(themselves) breathless in this room.

    • b.

      The little train engine puffed *(itself ) breathless on the siding.

This means that locative inversion is acceptable with both unaccusatives and unergatives. It is not the case that there can be no underlying subject when locative inversion takes place.

13 In Bruening 2014, the maximal VP node is called vP. What the node is called makes no difference, so long as the phasal node is the maximal VP node, before any aspectual and tense projections. Here that node is VoiceP.

14 Note that phasal nodes are very similar to the cyclic nodes of much earlier work (e.g., Langacker 1969); see Bruening 2014 for discussion.

15 The examples might work better with each than with every, but that would make the point just as well.

16 I should qualify this statement: some dialects of English also permit benefactive or “dative of interest” interpretations of some double object examples. I will not address this type of interpretation here.

17 This raises the following question, as pointed out by a reviewer: why do we not see R-dative shifted versions of double object constructions like She melted me some ice cream with the preposition to? The fact is that they do not seem to be acceptable.

  • (i)

    • a.

      *Who did she melt some ice cream to?

    • b.

      *She melted some ice cream to every child on the Little League team who had had teeth broken in the fight at the end of the game.

The reviewer suggests that to might still have some of its prepositional semantics even when it is being used as a case marker in R-dative shift, and so is not compatible with some verbs. Another possibility is that, since the caused-possession meaning of the double object construction is compatible with the vague benefactive meaning of for, using for is available and strongly preferred, enough so that speakers feel that to is unacceptable (although in this alternative it might actually be generated by the grammar).

Acknowledgments

I would like to thank the anonymous LI reviewers for helpful comments.

References

References
Alexiadou,
Artemis
, and
Elena
Anagnostopoulou
.
2001
.
The Subject-in-Situ Generalization and the role of case in driving computations
.
Linguistic Analysis
1
:
205
245
.
Alexiadou,
Artemis
, and
Elena
Anagnostopoulou
.
2007
. The Subject-in-Situ Generalization revisited. In
Interfaces + Recursion = Language?
, ed. by
Hans-Martin
Gärtner
and
Uli
Sauerland
,
31
60
.
Berlin
:
Mouton de Gruyter
.
Anagnostopoulou,
Elena
.
2005
. Cross-linguistic and cross-categorial distribution of datives. In
Advances in Greek generative syntax
, ed. by
Melita
Stavrou
and
Arhonto
Terzi
,
61
126
.
Amsterdam
:
John Benjamins
.
Baker,
Mark
.
1988
.
Theta theory and the syntax of applicatives in Chichewa
.
Natural Language and Linguistic Theory
6
:
353
389
.
Barker,
Chris
.
2012
.
Quantificational binding does not require c-command
.
Linguistic Inquiry
43
:
614
633
.
Bosse,
Solveig
.
2015
.
Applicative arguments: A syntactic and semantic investigation of German and English
.
New York
:
Peter Lang
.
Bosse,
Solveig
, and
Benjamin
Bruening
.
2011
. Benefactive versus experiencer datives. In
WCCFL 28: Proceedings of the 28th West Coast Conference on Formal Linguistics
, ed. by
Mary Byram
Washburn
,
Katherine
McKinney-Bock
,
Erika
Varis
,
Ann
Sawyer
, and
Barbara
Tomaszewicz
,
69
77
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Bosse,
Solveig
,
Benjamin
Bruening,
and
Masahiro
Yamada
.
2012
.
Affected experiencers
.
Natural Language and Linguistic Theory
30
:
1185
1230
.
Bresnan,
Joan
.
1994
.
Locative inversion and the architecture of Universal Grammar
.
Language
70
:
72
131
.
Bresnan,
Joan
.
2007
. Is syntactic knowledge probabilistic? Experiments with the English dative alternation. In
Roots: Linguistics in search of its evidential base
, ed. by
Sam
Featherston
and
Wolfgang
Sternefeld
,
75
96
.
Berlin
:
Mouton de Gruyter
.
Bresnan,
Joan
,
Anna
Cueni,
Tatiana
Nikitina,
and
Harald
Baayen
.
2007
. Predicting the dative alternation. In
Cognitive foundations of interpretation
, ed. by
Gerlof
Boume
,
Irene
Krämer
, and
Joost
Zwarts
,
69
94
.
Amsterdam
:
Royal Netherlands Academy of Science
.
Bresnan,
Joan
, and
Tatiana
Nikitina
.
2009
. The gradience of the dative alternation. In
Reality exploration and discovery: Pattern interaction in language and life
, ed. by
Linda Ann
Uyechi
and
Lian-Hee
Wee
,
161
184
.
Stanford, CA
:
CSLI Publications
.
Bruening,
Benjamin
.
2001
.
QR obeys Superiority: Frozen scope and ACD
.
Linguistic Inquiry
32
:
233
273
.
Bruening,
Benjamin
.
2010a
.
Ditransitive asymmetries and a theory of idiom formation
.
Linguistic Inquiry
41
:
519
562
.
Bruening,
Benjamin
.
2010b
.
Double object constructions disguised as prepositional datives
.
Linguistic Inquiry
41
:
287
305
.
Bruening,
Benjamin
.
2010c
.
Language-particular syntactic rules and constraints: English locative inversion and do-support
.
Language
86
:
43
84
.
Bruening,
Benjamin
.
2014
.
Precede-and-command revisited
.
Language
90
:
342
388
.
Bruening,
Benjamin
.
2016
.
Alignment in syntax: Quotative inversion in English
.
Syntax
19
:
111
155
.
Chomsky,
Noam
.
2000
. Minimalist inquiries: The framework. In
Step by step: Essays on Minimalist syntax in honor of Howard Lasnik
, ed. by
Roger
Martin
,
David
Michaels
, and
Juan
Uriagereka
,
89
155
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
. Derivation by phase. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Christian,
Donna
.
1991
. The personal dative in Appalachian speech. In
Dialects of English: Studies in grammatical variation
, ed. by
Peter
Trudgill
and
J. K.
Chambers
,
11
19
.
London
:
Longman
.
Conroy,
Anastasia M
.
2007
. The personal dative in Appalachian English as a reflexive pronoun. In
University of Maryland working papers in linguistics 16
, ed. by
Akira
Omaki
,
Ivan
Ortega-Santos
,
Jon
Sprouse
, and
Matthew
Wagers
,
63
88
.
College Park
:
University of Maryland, Department of Linguistics
.
Cuervo,
María Cristina
.
2003
.
Datives at large
.
Doctoral dissertation, MIT, Cambridge, MA
.
Fox,
Danny
, and
Jon
Nissenbaum
.
1999
. Extraposition and scope: A case for overt QR. In
WCCFL 18: Proceedings of the 18th West Coast Conference on Formal Linguistics
, ed. by
Sonya
Bird
,
Andrew
Carnie
,
Jason D.
Haugen
, and
Peter
Norquest
,
132
144
.
Somerville, MA
:
Cascadilla
.
Green,
Georgia M
.
1974
.
Semantics and syntactic regularity
.
Bloomington: Indiana University Press
.
Gropen,
Jess
,
Steven
Pinker
,
Michelle
Hollander
,
Richard
Goldberg
, and
Ronald
Wilson
.
1989
.
The learnability and acquisition of the dative alternation in English
.
Language
65
:
203
257
.
Guéron,
Jacqueline
, and
Robert
May
.
1984
.
Extraposition and Logical Form
.
Linguistic Inquiry
15
:
1
31
.
Harley,
Heidi
.
1997
. If you have, you can give. In
Proceedings of the West Coast Conference on Formal Linguistics XV
, ed. by
Brian
Agbayani
and
Sze-Wing
Tang
,
193
207
.
Stanford, CA
:
CSLI Publications
.
Harley,
Heidi
.
2002
.
Possession and the double object construction
.
Yearbook of Linguistic Variation
2
:
29
68
.
Hiraiwa,
Ken
.
2001
. Multiple Agree and the Defective Intervention Constraint in Japanese. In
The Proceedings of the 1st HUMIT Student Conference in Language Research (HUMIT 2000)
, ed. by
Ora
Matushansky
et al
,
67
80
.
MIT Working Papers in Linguistics 40
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Hoeksema,
Jack
.
2000
. Negative polarity items: Triggering, scope, and c-command. In
Negation and polarity: Syntactic and semantic perspectives
, ed. by
Laurence
Horn
and
Yasuhiko
Kato
,
115
146
.
Oxford
:
Oxford University Press
.
Horn,
Laurence R
.
2008
. “i love me some him”: The landscape of non-argument datives. In
Empirical issues in syntax and semantics 7
, ed. by
Olivier
Bonami
and
Patricia Cabredo
Hofherr
,
169
192
.
Paris
:
CSSP
. Available at http://www.cssp.cnrs.fr/eiss7.
Johnson,
Kyle
.
2001
.
Clausal edges and their effects on scope
.
Ms., University of Massachusetts, Amherst
. Available at http://people.umass.edu/kbj/homepage/Content/peripheries.pdf.
Kayne,
Richard
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
Krifka,
Manfred
.
2004
.
Semantic and pragmatic conditions for the dative alternation
.
Korean Journal of English Language and Linguistics
4
:
1
32
.
Langacker,
Ronald
.
1969
.
On pronominalization and the chain of command
. In
Modern studies in English
, ed. by
David A.
Reibel
and
Sanford A.
Schane
,
160
186
.
Englewood Cliffs, NJ
:
Prentice-Hall
.
Larson,
Richard K
.
1988
.
On the double object construction
.
Linguistic Inquiry
19
:
335
391
.
Marantz,
Alec
.
1993
. Implications of asymmetries in double object constructions. In
Theoretical aspects of Bantu grammar
, ed. by
Sam A.
Mchombo
,
113
150
.
Stanford, CA
:
CSLI Publications
.
Oehrle,
Richard
.
1976
.
The grammatical status of the English dative alternation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Ormazabal,
Javier
, and
Juan
Romero
.
2010
. The derivation of dative alternations. In
Argument structure and syntactic relations: A cross-linguistic perspective
, ed. by
Maia
Duguine
,
Susana
Huidobro
, and
Nerea
Madariaga
,
203
232
.
Amsterdam
:
John Benjamins
.
Ormazabal,
Javier
, and
Juan
Romero
.
2012
. PPs without disguises: Reply to Bruening.
Linguistic Inquiry
43
:
455
474
.
Postal,
Paul M
.
2004
. A paradox in English syntax. In
Skeptical linguistic essays
,
15
82
.
Oxford
:
Oxford University Press
.
Pylkkänen,
Liina
.
2008
.
Introducing arguments
.
Cambridge, MA
:
MIT Press
.
Rappaport Hovav,
Malka
, and
Beth
Levin
.
2008
.
The English dative alternation: The case for verb sensitivity
.
Journal of Linguistics
44
:
129
167
.
Reinhart,
Tanya
.
1976
.
The syntactic domain of anaphora
.
Doctoral dissertation, MIT, Cambridge, MA
.
Shan,
Chung-Chieh
, and
Chris
Barker
.
2006
.
Explaining crossover and superiority as left-to-right evaluation
.
Linguistics and Philosophy
29
:
91
134
.
Takano,
Yuji
.
2003
.
How antisymmetric is syntax?
Linguistic Inquiry
34
:
516
526
.
Webelhuth,
Gert
, and
Clare J.
Dannenberg
.
2006
.
Southern American English personal datives: The theoretical significance of dialectal variation
.
American Speech
81
:
31
55
.