Pylkkänen (2002, 2008) and Bruening (2010a) present several arguments against the ‘‘small clause’’ approach to the double object construction in English, building on the predictions that that proposal makes with respect to the transfer-of-possession entailment, Goal-oriented depictives, nominalizations, subextraction, quantifier scope, and idioms. We argue that the small clause analysis proposed by Harley (1995, 2002) in fact makes correct predictions in all these cases. In addition, we point out the existence of previously overlooked parallels between double object structures and have-sentences with respect to depictives, eventive DP complements, and quantifier scope. This motivates an analysis that links these different behaviors to the properties of a single PHAVE element common to both.
Pylkkänen (2002, 2008) presents two arguments against the ‘‘small clause’’ approach to double object constructions in English, emphasizing the predictions that that proposal makes with respect to two key properties: the entailment of transfer of possession, and the behavior of Goal-oriented depictive secondary predicates. Bruening (2010a) also argues for a different analysis of ditransitive constructions, on the basis of data from nominalizations, subextraction, quantifier scope, and idiomatic interpretations.
In this reply, we address both sets of objections within the framework of Harley’s (1995, 2002) small clause treatment of the double object construction. We argue that it makes correct predictions in all the domains mentioned above, although Bruening (2010a) is correct in his rejection of Harley’s small clause approach to the to-dative alternant. Consequently, we propose a hybrid account, endorsing Bruening’s (2010a) approach to to-datives but arguing for Harley’s (1995, 2002) small clause view of the double object structure.1 In particular, parallels between the behavior of double object structures and have-sentences support the idea that the two share a key subconstituent. Sentences containing have behave the same as double object structures with respect to depictive modification, the interpretive effect of eventive DP complements, and quantifier scope. These facts (and several others previously discussed) motivate an analysis that links all these different behaviors to the properties of a single PHAVE element.
2 Background: PHAVE in Double Object Constructions
Harley (1995, 2002) argues in favor of a treatment of double object give that revives an insight from much earlier work, including that of Green (1974), McCawley (1974), and Oehrle (1976), among others, who observed the existence of a possessive meaning in double object structures that is not necessarily present in to-dative structures. They highlighted a number of phenomena that show a parallelism between sentences with have and double object give that receive a natural account if have is a part of the structure of double object give. For example, adverbial modification of double object structures seems to pick out a have result state, animacy restrictions on the possessor also apply to the Goal in a double object construction, and certain existence presupposition effects are similar between possessors and double object Goals as well. We will not review these and other parallel effects observed in the literature cited above in detail (see, e.g., Harley 2002 or Beck and Johnson 2004 for a more in-depth review), but they form the core of the argument in favor of a decompositional approach to double object give. The phenomena mentioned above are illustrated in (1)–(4); we will also discuss previously overlooked parallels in sections 3 and 4.
(2) Alienable possession requires an animate possessor2
John has a book.
Brenda gave John a book.
#The car has a flyer.
#The advertiser gave the car a flyer.
(3) Inalienable possession does not require animacy
John has blue eyes.
The house has a new coat of paint.
His mother gave John his blue eyes.
The painter gave the house a new coat of paint.
Following Kayne (1984), Guéron (1986), and Hoekstra (1988), Harley (1995, 2002) captured these and similar effects by proposing that give is underlyingly represented as containing a possessive small clause—that is, as something roughly like ‘cause X to have Y’. Her (2002) analysis is summarized by the structure in (5).
The possessive predicate embedded by the causative light verb is treated by Harley (2002) as prepositional, following Kayne’s (1993) conclusion that verbal have involves incorporation of an abstract P/D element HAVE into an abstract verb BE. Without incorporation of PHAVE, vBE is realized as be; with incorporation of PHAVE, however, it surfaces as have. On this account, the internal structure of have is as shown in (6).
Harley (1995, 2002) argued that the possessive small clause embedded by vCAUSE in double object constructions is headed by this abstract PHAVE element, not by verbal have. The incorporation of PHAVE into vCAUSE is, she proposed, realized as the verb give. Harley (2002) assumed that Manner modification of the complex v head by other verbal predicates produces double object structures with other verbs, such as send, hit, and throw; later (2007, 2012), she extended this to give as well. Manner modification will play a significant role in the present treatment (see section 4). Other work supporting this approach to double object give includes Richards 2001 and Beck and Johnson 2004.
3 Low Applicative vs. PHAVE: Pylkkänen 2002, 2008
Pylkkänen (2002, 2008) presents two arguments against the small clause proposal. The first reprises a common objection to the account, namely, the failure of the result possessive state to be strictly entailed by double object structures.4 Consider the contrast between the reported failure of a possession entailment in the double object structure (see (7)) and the entailments of a true lexical causative (see (8)).5
I threw John the ball but he didn’t catch it.
I sent John a letter but he never got it.
I wrote Sue a letter but she never got it.
#I flew the kite over the field but it didn’t fly.
#I broke the vase but it didn’t break.
#I cooked the meat but it didn’t cook.
Pylkkänen argues that if the sentences in (7) really involved causation of a possessive state, then the entailment ‘‘Goal has Theme’’ should be just as nondefeasible as the corresponding result states in the true lexical causatives in (8). The fact that possession is not strictly entailed in (7) leads her to conclude that the small clause analysis is problematic.
Pylkkänen’s second argument involves the behavior of double object give with depictive secondary predicates. Consider the contrast between the ill-formedness of Goal-oriented depictives in double object constructions, illustrated in (9a), and the well-formedness of subject-oriented depictives in a small clause construction, illustrated in (9b).
I told Johni the news (*drunki).
I saw Johni drive his car (drunki).
In (9b), which contains an uncontroversial small clause, [John drive his car], a depictive secondary predicate modifying the small clause subject John is well-formed, while in (9a), a similar depictive secondary predicate modifying Goal John is impossible. Pylkkänen argues that if the Goal were truly the subject of a small clause, the two cases should be entirely parallel.
On the basis of these two arguments, Pylkkänen concludes that the small clause account does not work. We respond to each argument in turn, and we introduce a new parallel in the way depictives interact with the Goal when the Theme is an event-denoting nominal.
3.1 The Entailments of PHAVE
Recall Pylkkänen’s (2002, 2008) first objection to the small clause approach: double object variants of verbs like send or throw do not always entail actual possession, as shown in (7). To address this objection, we must consider what the caused small clause [SC John [PHAVE a book]] does entail, and how such entailments might differ from those of John has a book. Unless they do differ in some respects, Pylkkänen’s objection is significant.
In fact, a number of proposals have been made concerning the interpretive differences between verbal have and the HAVE relation in the double object construction, all aimed at capturing the failure of the latter to strictly entail the former. Harley (2012), adopting Kayne’s (1993) original treatment, proposes that the semantics of PHAVE are those of possessive ’s. The entailments of possessive ’s are more flexible than those of verbal have, and a much closer fit to the entailments of double object structures than the entailments of verbal have are.6 Alternatively, Beck and Johnson (2004) suggest that a progressive operator is included in double object syntax. This intensional operator, like that of Landman (1992), cancels the result entailment of change-of-state predicates, and consequently has the desired effect when added to the denotation of a constituent such as [John [HAVE the map]].7
Perhaps the most extended treatment of this problem is provided by Beavers (2011), also invoking intensionality in certain double object structures. Following Koenig and Davis (2001), he argues that the possession relation in many double object structures contains a modal component. For some double object verbs, where actual possession is entailed (e.g., give, pass), the verb lexically requires the possessive result, and an actual entailment of possession is the result. However, for most double object verbs, where possession is cancelable, Beavers proposes that the possessive relationship is modified by Koenig and Davis’s ‘‘sublexical’’ modality, requiring only that possession be achieved in some possible worlds, not in all. He terms such potential possession prospective possession. All double object constructions entail at least prospective possession, and some, lexically, have the stronger possession entailment as well. We adopt Beavers’s account here, attaching the prospective possession entailment to PHAVE. This is enough to predict the classic Oehrle’s Generalization effects, as well as the adverbial modification facts and other parallels between have and double object give illustrated in (1)–(4) and elsewhere. However, it does not produce a possession entailment unless strengthened by a particular root’s lexical semantic contribution. Pylkkänen’s first objection, then, does not apply.
3.2 Depictives and Stative Small Clauses
Pylkkänen’s (2002, 2008) second argument against the PHAVE small clause analysis hinges on the behavior of depictive secondary predicates. If depictive secondary predicates can target the subject of regular small clauses, they should be able to target the subject of the small clause headed by PHAVE. In fact, as shown in (9a), the Goal argument of a double object construction is ineligible to serve as the antecedent of a depictive secondary predicate.
However, there is a key difference between small clauses whose subjects accept depictive modification, as in (9b), and the small clause headed by PHAVE in (9a): the former are eventive. Subjects of stative predicates do not accept depictive modification,8 whether they occur on their own (10) or are embedded below a matrix verb (11).
I consider Johni sick (#drunki).
I made Billi confused (#tiredi).
The PHAVE small clause must itself be a stative predicate, since it is an ingredient in verbal have, and run-of-the-mill verbal have-sentences are stative. Consequently, PHAVE cannot be eventive.9 The failure of depictive modification, then, can be ascribed to the fact that small clauses like [SC John [PHAVE a book]] are stative, rather than eventive.
3.3 ‘‘Light verb’’ Give vs. Double Object Give: The Parallel with Have
Pylkkänen’s (2002, 2008) investigation of depictives leads her to propose distinct structures for double object configurations, depending on whether the Theme DP denotes an event or an entity. In the former case (e.g., gave the patient the medication), she treats give as a light verb, forming a complex predicate with the Theme as sister prior to composing with the Goal DP (see (12)).10 In the latter case (e.g., gave John the book), she proposes a ‘‘low applicative’’ structure, in which the two entities are related by a functional Appl(icative) head prior to composing with the verb give (see (13)).
For Pylkkänen, this structural difference is motivated by the behavior of the light-verb-type construction with respect to depictives. When the Theme DP describes an event, depictive modification of the Goal is possible, in contrast to the situation illustrated in (9a), with an entity-denoting Theme. This contrast is illustrated in (14).
The nurse gave the patienti the medication half-asleepi.
*Mary gave Johni the book half-asleepi.
In (14a), the intended reading is one where the nurse induced the patient to actually consume the medication—that is, one in which she medicated him, not simply one in which she handed the physical medicine to him for later self-administration. The structure in (12) represents that reading of (14a); the ‘handing’ reading would have the structure in (13). Pylkkänen’s point is that the Goal-oriented depictive half-asleep is compatible only with this event-denoting reading of giveDP medication, not the physical-transfer interpretation.
In Pylkkänen’s structure for ‘‘light verb’’ give in (12), the V´ denotes an event owing to the presence of the V head, and hence may be modified by a depictive prior to composing with the Goal DP in its specifier (Pylkkänen 2008:35). In contrast, the Appl´ constituent in Pylkkänen’s transfer-of-possession structure (13) does not have the right type to compose with a depictive. The depictive predicate in (14b) may only merge higher up, to the VP projection of (13), and thus a Goal-oriented interpretation is impossible (Pylkkänen 2008:27).
In contrast, we suggest that the ‘‘light-verb’’ double object structure and the transfer-of-possession double object structure are identical; the different semantics of the Theme is the sole trigger for the distinct behavior of the depictive predicate. PHAVE is independently known to inherit the eventiveness quality of its complement. Belvin (1993, 1996) and Ritter and Rosen (1993) show that verbal have varies between eventive and stative interpretations depending on whether its complement is eventive or stative. Consider the following data (adapted from Harley 1998):
*What John did was have a book/a bike/a car.
*John is having a book/a bike/a car.
Look! John has a book/a bike/a car!
What John did was have a party/a massage/a fit.
John is having a party/a massage/a fit.
*Look! John has a party/a massage/a fit!
In (15), we see that with simple possessive have, with an entity-denoting Theme, the resulting clause behaves as a state, according to standard tests for eventiveness: it is ill-formed in the whatX did was . . . frame (Cruse 1973) and in the progressive, and well-formed with the true present tense. In contrast, when the Theme complement denotes an event, as in (16), the resulting clause behaves as an event with respect to these tests. Thus, have-structures inherit eventiveness (or lack thereof ) from their complement DP.11
Depictives, on Pylkkänen’s analysis, can only modify predicates of events. We can then test whether examples like (16a–c) are eventive by composing them with a depictive. As predicted, eventive have allows a depictive to appear, modifying the subject of have, as in (17a). In contrast, stative have does not, as shown in (17b) (repeated from (10a)).12
Johni had a massage drunki.
*Johni had a book drunki.
The relevance of these cases to the (un)grammaticality of depictives with double object give in (14) should be clear. Within the small clause analysis, the denotation of the PHAVE´ constituent below vCAUSE will depend on the eventiveness of the Theme complement DP. When the Theme denotes an event, PHAVE´ relates entities and events, and hence should be a legitimate target of depictive modification.
The parallel effect of eventiveness of the Theme DP in clauses with have and in clauses with double object give strongly suggests that a unified treatment is motivated. In Pylkkänen’s system, there is no reason to expect to find the same effect with have and give; the parallel behavior requires a similar structural distinction, but her analysis makes no prediction that have and give should behave in similar ways.13
In contrast, in the small clause approach, have and give share a key substructure, namely, the constituent headed by PHAVE. If this variable behavior in eventiveness properties, conditioned by the eventiveness of the direct object, is a property of PHAVE, then we predict that have and give should exhibit similar behavior, as indeed they do. It is not necessary or desirable to ascribe the distinction to two different positions for the verb give when it is clearly dependent on the semantics of the Theme DP and its interaction with PHAVE, both when it occurs as a subpart of give and when it occurs as a subpart of have.
Assuming that PHAVE with an eventive complement produces a PHAVE´ constituent that has an eventive denotation, such a PHAVE´ constituent will be a legitimate target for depictive modification ( just as, e.g., v´ is in a normal eventive predicate). We illustrate the locus of depictive modification in (17a) here, in (18); the locus of depictive modification in (14a) is exactly analogous, with a different embedding v head.
Exactly how the ‘‘inheritance’’ of eventiveness from an event-denoting complement works remains an open question. Ideally, PHAVE would have a denotation that undergoes coercion in a natural way when presented with an argument of type 〈s〉 (for state) instead of type 〈v〉 (event). John Beavers ( pers. comm.) suggests that PHAVE may simply be ambiguous between an event-selecting and an entity-selecting meaning. However, the experimental results reported by Piñango, Mack, and Jackendoff (2006) and Wittenberg and Piñango (2011) suggest that the event-selecting meaning must be coerced online (see footnotes 11 and 13), not lexically available. With the usual inventory of types, coercion is not straightforward to implement. We need to distinguish events from states type-theoretically, as argued by Katz (2003) and Copley and Harley (2015). The data tell us that when PHAVE composes with an eventive complement, PHAVE´ becomes a function of type 〈e,〈vt〉〉, and hence an appropriate target for depictive modification, while with an entity-denoting complement, PHAVE´ is of type 〈e,〈st〉〉, and not modifiable by a depictive. Coercion could then involve changing the domain of arguments of PHAVE. One possibility might be a subsortal analysis, with an overarching eventuality type of which events and states are sorts.14
We now turn to a separate set of objections to the PHAVE proposal raised by Bruening (2010a). Like Pylkkänen, Bruening includes an applicative projection in double object structures. We will show that the arguments he presents for his analysis can for the most part also be accommodated in the PHAVE approach, and that with respect to certain properties, the PHAVE approach fares better than Bruening’s applicative analysis.
4 High Applicative vs. PHAVE: Bruening 2010a
Bruening (2010a) provides an extensive discussion of the properties of to-dative vs. double object structures, arguing against the ‘‘symmetric’’ analysis of the two proposed by Harley (1995, 2002). In particular, he shows, convincingly, that in the case of the to-dative the verb give is base-generated low in the structure. He argues that a sentence like Maria gave the bottle to the baby has the underlying representation shown in (19).
The main verb give head-moves to Voice to derive the surface linear order.
Harley (2007) adopts a similar structure for the to-dative construction, abandoning her previous proposal that the lower portion of the verbal projection in a to-dative was a small clause headed by PLOC, a locative counterpart to PHAVE that inverted the positions of the Goal and Theme arguments.15 Bruening shows that the verb must be base-generated low in the structure in these cases, basing his arguments on nominalizations, subextraction from the Theme, opaque domains for anaphora, quantifier scope, locative inversion, and idioms.
In his analysis of the double object variant, Bruening also proposes that the verb is base-generated low. The difference between the two variants is that in the double object structure, the Goal argument is introduced by an Appl head that selects for the verb’s projection (20) (Bruening 2010a:526).16 In this, Bruening follows Marantz (1993).
Bruening shows that the proposal in (20) captures many of the properties of the double object construction, but we argue below that it does not fare as well as the PHAVE approach. We argue that the hybrid proposal advocated in Harley 2007 is more successful. There, the double object structure is based on PHAVE rather than an Appl head. Give (like other double object verbs) is adjoined to the upper v head by Manner adjunction, specifying the manner in which the caused-transfer event occurs. This process is the same as that available to other change-of-state and change-of-location structures in English, and yields resultative adjectives, verb-particle constructions, and manner-of-motion constructions (see also Snyder 1995 on the Compounding Parameter).
Just as in Harley 1995, 2002, PHAVE heads the lower projection in the argument structure. The verb root give, however, is adjoined to the v head via Manner adjunction. (For further discussion of the Manner adjunction operation, see Tomioka 2006, Levinson 2007, Embick 2010, and Harley and Folli 2011.)18 We now turn to a comparison of this proposal with Bruening’s ‘‘high applicative’’ analysis of the double object construction.
As in Bruening’s (2010a) analysis, in the double object construction the Goal is not an argument of the verb give, which is simply adjoined higher up in the structure, to the vCAUSE. Nominalizations involve only the lower portion of the extended verbal projection, excluding the external-argument-introducing head (see Harley and Noyer 1998, Alexiadou 2001, among many others). In our hybrid approach, in the to-dative the main verb is low in the structure as in (19), and nominalization can apply, producing examples like (22a). However, nominalization of give will not be able to apply to the PHAVEP small clause in (21), because give is base-generated too high in the structure, explaining the ungrammaticality of (22b).
the gift of a statue to Mary
*the gift of Mary (of ) a statue
There is a contrast between the to-dative construction and the double object construction with regard to subextraction from the first DP argument. It is possible with the Theme in the to-dative but not with the Goal/Possessor in a double object structure.
Whoi did you send pictures of ti to all the season ticket holders?
*Whoi did you send relatives of ti invitations to the wedding?
Bruening (2010a:524) uses these data to argue against Harley’s (1995, 2002) PLOC account in which the Theme of a to-dative is the subject of a small clause. True subjects of small clauses cannot undergo subextraction, as shown by Postal (1974:195).
(24) *Whoi do you consider supporters of ti beneath our notice?
We now follow Bruening (2010a) in not treating to-datives as small clauses. However, the failure of subextraction from the Goal in double object constructions supports a small clause approach. Under the PHAVE account, their patterning with other small clauses is straightforwardly predicted. The explanation for the ungrammaticality of (24) carries over to that of (23b) if double object constructions contain a small clause. John Beavers ( pers. comm.) notes that the Subject Island Constraint rules out subextraction from the Goal/Possessor on the PHAVE account. It does not do so on Bruening’s applicative-based account.20
4.3 Opaque Domains for Anaphora
Bruening (2010a:524n5) also mentions that a small clause is typically an opaque domain for anaphora in English, as in (25a). This fact is raised by Pesetsky (1995) against the small clause approach to double object constructions. Pesetsky attributes the ungrammaticality of such examples to Chomsky’s (1986) notion of a complete functional complex (CFC) as the relevant formal characterization of a binding domain; small clauses qualify as CFCs because they have a subject. The PHAVE analysis fares less well on this diagnostic, as the Theme is accessible to binding by the subject in a double object structure, as in (25b).
*John considered her angry at himself.
John showed her himself.
The CFC view predicts that the Goal, as the subject of a small clause, should form an opaque domain in (25b), which it does not.
However, if the modern Minimalist correlate of the 1986 CFC is a phase, as argued by Lee Schoenfeld (2004), then the PHAVE analysis in fact makes the correct prediction for give, since we do not posit a phasal boundary between vCAUSE and the PHAVEP. On that view, small clauses embedded under verbs like consider would include a phasal projection—something like Moro’s (2000) PredP. Other resultative structures that are commonly analyzed as involving small clauses without phasal boundaries also permit nonopacity. Consider the well-formedness of reflexives in the small clause complement of causative have in examples like John had honey poured on himself (though see Harley 1998 for a different view of these anaphors), in verb-particle constructions such as John showed the medal off to himself (analyzed as small clauses in, e.g., Basilico 2008), or in caused-motion structures such as John pushed Mary away from himself (treated as small clauses in Hoekstra and Mulder 1990, Folli and Harley 2006).21 The claim that the complements of such causative predicates are small clauses allows the caused situation to be represented in the syntax, leading to a more uniform syntax-semantics mapping; however, the difference in behavior between the different types of small clauses with respect to opacity (and other properties) requires creating a typology of small clauses, differentiating phasal from nonphasal small clauses, and perhaps other subtypes also. Basilico (2003) articulates the beginnings of such a typology, contrasting the properties of the subjects of adjectival small clauses (categorial, topical, fixed-scope) with those of verbal small clauses (thetic, nontopical, variable-scope). Developing Basilico’s typology to include the opacity contrasts at issue here would be the desirable next step.
4.4 Quantifier Scope
Bruening (2001, 2010a,b) provides extensive discussion of the difference between double object and to-dative structures with respect to variable quantifier scope in the Theme and Goal arguments. In the to-dative construction, a universal quantifier in the lower Goal can take wide scope with respect to an existential quantifier in the higher Theme (see (26a)). In contrast, in the double object construction, scope is fixed to the surface order: an existential quantifier in the higher Goal must take wide scope with respect to a universal quantifier in the Theme (see (26b)).
Bruening (2010a) takes an equidistance-based view of this interaction. Hierarchical scope must be preserved, because of locality constraints on quantifier movement. However, Bruening considers two quantifiers that m-command each other to be equally local to (i.e., equidistant from) any c-commanding head. Hence, in the to-dative, in which both arguments are contained in the projection of the V, either is eligible to undergo quantifier movement in either order, and can take wide scope. In contrast, the Goal and Theme arguments in the double object construction are in different maximal projections; the Goal in the higher ApplP and the Theme in the lower VP. Because they do not m-command each other, they are not equidistant, and the Goal must outscope the Theme.
Let us consider the implications of this contrast for the current proposal. Given that it is a property of double object structures that scope must be fixed, we might expect to see the same effect with verbal have. As pointed out by Satoshi Tomioka (see Bruening 2010a:547), this expectation is borne out: inverse scope is surprisingly impossible with verbal have, as in (27a), while both scopes are possible with a normal transitive verb, as in (27b).
As with the cases discussed in section 3.3, there is a surprising parallel between the double object structure and verbal have. Such a parallel is expected under the PHAVE analysis but not under an ApplP analysis, and we take it to support the former.22
That leaves the question open, however, of why PHAVE does not permit inverse scope, in either structure. This fact must be due to some key property of PHAVE that is common to both structures. One avenue of attack would be to appeal to an equidistance-based formulation of Superiority effects like that of Chomsky (1993, 1995) (updated in Den Dikken 2007 and Gallego 2010). The idea would be that PHAVE behaves like a particle and is stranded within PHAVEP.23 Without head movement to the c-commanding domain, PHAVE’s arguments are not equidistant and their scope is fixed. This would contrast with the behavior of the two arguments of to-dative give, which exhibit scope flexibility due to head movement of (give to the c-commanding v. Such an account would need to be supplemented, however, to explain why scope flexibility is possible in other PHAVE-containing verbs, such as get, passive be given, and want. Want introduces a control structure (see Harley 2004:264n8) and hence may permit inverse scope because of the multiple argument positions involved. Get and be given are more difficult to understand, but differ significantly from have in denoting dynamic events rather than a state. If eventiveness is connected to the presence of a particular type of VoiceP projection, an account building on the proposals by McGinnis (1998) and Bruening (2001) could be given, according to which the A-Merge of an argument permits reconstruction for scope interpretation below an Ā-merged scope-taking movement of another argument (thus circumventing the Superiority requirement). Spec,VoiceP, present in dynamic predicates, would allow A-Merge of the Goal and consequent inverse scope for an Ā-merged Theme, but would forbid it for active give, since the external argument position is independently filled there, and for have, since it is stative. Fully exploring the implications of this hypothesis, however, would take us too far afield here.
The key contrast we wish to emphasize in comparing the PHAVE proposal with Bruening’s ApplP proposal is that the latter does not fare any better in accounting for this particular array of facts. With ApplP in double object give, the (quite general) puzzle is why it should show any parallels with have at all, both the broad array of similarities described in sections 2 and 3 and the fixed-scope properties at issue here. If, in order to account for these parallels, the ApplP account of give and get were extended to have, then parallel predictions about scope for have, get, and be given would be made, producing the same conundrums described for the PHAVE proposal above; on the other hand, if the ApplP account is not extended to have and have’s scope properties are explained independently (making it a coincidence that they match the properties of double object give), then the other parallels concerning interpretation, animacy effects, and so on, also remain unexplained. We suggest that the PHAVE account does better in explaining the general parallelism between have and give, though the details of the scope parallelism remain to be hammered out.
4.5 V + Theme Idioms
Bruening (2010a) argues that his ApplP structure for the double object construction predicts the existence of V + Theme idioms with double object structures and the impossibility of V + Goal idioms in the same structure, since Goal arguments in double object structures are selected by the Appl head, not the V. In the to-dative frame, he claims, both V + Theme and V + Goal idioms are possible, since both Theme and to-marked Goal are arguments of the verb. He lays out the four logical possibilities as follows:
Surveying ditransitive idiomatic phrases in English, Bruening finds an asymmetry between Class 4 (nonexistent) and Class 1 (robustly attested). He attributes this to the fact that, in his proposed structure in (20), the Theme is directly selected for by the V, while the Goal is introduced by an Appl head. Thus, the Theme and the verb together can receive a special interpretation (Class 1), since in his account such special interpretations depend on a selectional relationship.24 However, in Class 4, the Goal and the verb do not stand in a selectional relationship, since the Goal is introduced by the Appl head. Bruening’s analysis predicts that Class 2 and Class 3 idioms should both be possible, since the verb selects both Theme and to-marked Goal in the to-dative construction.
Let us consider the implications of Bruening’s idiom typology for the PHAVE proposal. It might appear that the existence of Class 1, V + Theme idioms in the double object frame constitutes an objection to the proposed high merger of the verb. If the main verb is Manner-adjoined to v, it never forms a constituent with the Theme (or selects for it); hence, the existence of a special interpretation for V + Theme would be problematic.
However, as Richards (2001) shows, such V + Theme idioms in the double object construction actually involve PHAVE, not the verb root itself. (These are the idioms that Bruening lists as nonalternating, ‘‘pure’’ Class 1 idioms.) The idiomatic interpretation remains even when the verbal content is replaced with other PHAVE-containing verbal elements. Typical cases involve give alternating with get or take, as in (29).
His boss gave John the boot.
John got the boot.
Mary’s support gave John heart.
John took heart (from Mary’s support).
The idiom, in other words, is PHAVE + Theme, not V + Theme, as in Bruening’s account.
In the current proposal, a double object construction involves no selectional relationship between the verb and either internal argument. A selection-based theory of idiomatic interpretation therefore predicts that any instances of idiomatization of the Theme in a double object construction must involve PHAVE, not the verb. That is, they should all find interpretable counterparts in some other PHAVE-containing expression. This is certainly the case for the true Class 1 idioms, those discussed by Richards.25
Bruening also lists several other idioms as members of his Class 1 ( p. 541); in his classification, this group alternates between Class 1 and Class 2, as he gives examples showing they also occur in the to-dative frame. The relevant examples are listed in (30).
In every case, we are able to establish via Internet searches that these idioms can be well-formed even in the absence of the relevant main verb. In the first set of naturally occurring examples, presented in (31), the idiomatic chunk occurs in a PHAVE-containing frame.
‘‘Forgive me if I’m incredibly skeptical that Katie Couric and the CBS gang are pulling for McCain and tried to hand him a bone.’’26
‘‘Cholly didn’t get the riot act from the commissioner’s office like Rube did.’’27
give NP a hand; also:
‘‘Military spouses get a hand.’’28
‘‘At-home businesses get a sympathetic ear on Capitol Hill.’’29
‘‘It feels like sometimes I get a wide berth because I am a very tall waegukin.’’30
‘‘Rogue sites get the cold shoulder from advertisers.’’31
g. ‘‘The infusion of the sheriff ’s humor also added color to the story.’’
A reviewer suggests the possibility that these idioms are simply NPs, able to occur with their special interpretation in the absence of any particular verb or verb frame, and provides the Google results in (32).
‘‘As a tank barb I just can’t seem to get him down in time, a hand would be greatly appreciated.’’32
‘‘Either way, a sympathetic ear can make all the difference.’’33
‘‘And while the motivations for such wariness have obviously changed . . . the cold shoulder certainly remains.’’34
‘‘Below is the current travel advice from DFAT – if you are going, take the east side! Even then, a wide berth would be prudent.’’35
In all these cases, the same idiomatic interpretation is present in the absence of the verb in (30). The idiom in these cases, then, is not dependent on the particular verb—thus, it is not surprising that they ‘‘alternate’’ between Class 1 and Class 2.36
We do not present a comprehensive theory of idiomatization in place of Bruening’s (2010a) selection-based proposal. However, in considering Bruening’s classification of ditransitive idioms, two things stand out: double object idioms involving V + Goal do not exist (Bruening’s Class 4 idioms), and to-dative idioms involving V + Theme are extremely rare (Bruening’s Class 2 idioms). Indeed, Bruening gives only two examples that he considers clear cases: give it to NP ‘beat NP up’ and give rise to NP ‘make NP happen’. The give it to NP idiom is plausibly analyzed as PHAVE+ it idiomatization, given the well-formedness of You really got it that time! or He reallylet you have it!, with the same interpretation (‘You really were beaten up!’/‘He really beat you up!’). We hypothesize that since it is a pronominal clitic, the idiom mandatorily undergoes Bruening’s (2010b) ‘‘R-dative shift’’ rule, postposing the Goal NP from Spec,PHAVE and right-adjoining it to VoiceP/vP, and adding to.
What about the Class 2 idiom give rise to X, and the other three examples of idioms that Bruening (2010a:541) suggests can be base-generated in the Class 2 frame, as well as Class 1?
give the lie to X
give birth to X
give way to X
Bruening (2010b) provides diagnoses for whether a V + Theme idiom is base-generated in the to-dative frame or appears in the to-dative frame as the result of R-dative shift. If it is base-generated as a Class 2 idiom (i.e., in the to-dative frame), it should appear in the locative inversion construction following passivization (Bruening 2010a:539, 2010b:297). Applying this test to (33), we find that all four putative Class 2 idioms are cases of the ( possibly obligatory) R-dative shift rule, rather than to-datives; that is, they are like the give a headache idioms in Bruening’s (2010a) (42), rather than the lend an ear cases in his (41).
*To the complaint was given rise by poor customer service.
*To her words was given the lie by the arrival of Bill.
*To the baby Jesus was given birth in a manger.
*To the professor was given way (by the students).
Thus, none of these putative Class 2 idioms behave as expected for the to-dative frame, indicating that they also are base-generated as Class 1 idioms and subject to R-dative shift. If they are actually Class 1 idioms, then we predict that it should be possible to passivize the Goal argument from the double object frame. This seems to be the case: in an Internet search, all occur with the Goal argument as the subject of a passive derived from the double object frame, as the following examples show (including give rise to in (35d), though this example is marginal in the first author’s judgment):
‘‘Weedon Scott looked across at his companion with eyes that almost pleaded, though this was given the lie by his words.’’37
‘‘John was given birth on February 21, 1929 in Harlow, England.’’38
‘‘Zulkifli is one of two Perkasa leaders contesting with BN’s backing, the other being Ibrahim Ali, who was given way in Pasir Mas by the BN nominee.’’39
‘‘ . . . the judgment of victim-victimizer was given rise and made real in our mind . . . ’’40
If this conclusion is valid, then neither Class 2 nor Class 4 idioms exist—that is, in no case does the head of the lower projection idiomatize with its specifier. This suggests a sisterhood-based constraint on idiomatization: the head of an idiomatic constituent relates to its sister before idiomatizing with other constituents.
If there is such a sisterhood constraint on idiomatization, then what is the source of the to-dative variants of Bruening’s ‘‘alternating’’ idioms in (30) (if any of them are in fact true V + Theme idioms; see the remarks around (32))? Such forms would have to be base-generated as PHAVE + Theme idioms (Class 1), and, like other Class 1 idioms, their to-dative variants would have to be generated by R-dative shift. That is, give a wide berth to NP would involve Theme postposing and to-insertion, rather than base generation in the to-dative structure. This predicts, correctly, that *To John was given a wide berth is unacceptable.41
We have argued that the idiom patterns analyzed by Bruening are consistent with the expectations of the PHAVE account, and the apparent unavailability of Class 2 as well as Class 4 idioms suggests that a new look at constraints on idiom formation is in order.
4.6 Depictives and Denial
Finally, we discuss two remaining issues. First, we return to the depictive facts introduced by Pylkkänen (2002, 2008). In Bruening’s analysis of double object structures in (20), there is an eventive projection (of type 〈e, vt〉) that composes with the Goal argument, namely, the lower vP projection (Bruening 2010a:553, (67)). A Goal-oriented depictive would be expected to be able to adjoin to a projection of this type. However, as shown above, a Goal-oriented depictive is impossible in a double object construction. Bruening’s proposal offers no handle on why depictives of Goal arguments are ill-formed.
On the other hand, Bruening (2010a) discusses a class of verbs for which neither the PHAVE approach nor Pylkkänen’s approach to double object structures provides a clear analysis, namely, deny-class verbs, which typically introduce a negative component in their lexical meaning. Although these verbs occur in the double object construction, they do not have a to-dative alternant. Bruening proposes that these verbs are interpreted in the low position, and that this accounts for several special properties. First, the negative lexical component takes scope only over the Theme, not the Goal, shown by the fact that negative polarity items are licensed as the Theme of such verbs but not the Goal (Bruening 2010a:559).
They denied me any chance of redeeming myself.
*They denied any worker a raise.
If the verb is base-generated (or interpreted) in a high position, adjoined to v, one would expect it to take scope over both Goal and Theme, and license a negative polarity item in both.
With these verbs, the possession relation typically introduced by the double object frame cannot be isolated by any of the usual tests. For example, restitutive again is unavailable, meaning ‘Goal had Theme once before, and is now prevented from having Theme again’. Benjamin Bruening ( pers. comm.) provides the following examples:42
He always gets nasty comments when he speaks up at meetings, but #today they spared him their nasty comments again.
William got satisfaction last week, #but today they denied William satisfaction again.
The verbs cost and envy behave similarly.
One final special property of these verbs is that true V + Theme idioms exist with them, idioms that do not allow verbal substitution in other PHAVE contexts.
That car cost Bill an arm and a leg.
#Bill didn’t have an arm and a leg.
#Bill got an arm and a leg from his uncle.
#The purchase of this car took an arm and a leg.
This set of behaviors suggests that Bruening’s view of these verbs is correct: they are base-generated low in the structure, being possibly the only verbs to lexically select the double object frame.43 Nothing about the PHAVE approach to (alternating) double object structures is inconsistent with the existence of such verbs in the lexicon.
Pylkkänen (2002, 2008) and Bruening (2010a) both describe properties of double object constructions that in their analyses are accounted for by the presence of an applicative projection in the structure. On the basis of these accounts, both reject the notion that a dedicated possessive PHAVE head forms part of these structures, in contrast to the analysis presented by Harley (1995, 2002) on the basis of generalizations from a wide body of literature dating back to Green 1974, McCawley 1974, and Oehrle 1976, among others.
The central result we wish to emphasize here is the parallel behavior between the verb have and double object give with respect to some of the particular properties adduced by Pylkkänen and Bruening in support of their applicative-based analyses. The subject of have, like the Goal in a double object construction, cannot be the target of depictive secondary modification. Similarly, the interpretation of a have VP as an event or as a state depends on the event-denoting properties of its Theme DP, just as the behavior of a double object verb changes when these same event-denoting vs. entity-denoting properties of the Theme are manipulated. Finally, in just the same way that an indefinite quantifier in a Goal must take wide scope with respect to a quantified Theme in a double object construction, the subject of have must take wide scope with respect to a quantified possessed object. The correlations between all of these properties of double object structures and the properties of verbal have, we argue, support the notion that these two representations share a crucial substructure: namely, the P/D relational element identified as a subpart of verbal have by Kayne (1993), supplemented with a modal lexical component as in Beavers 2011. We call this element PHAVE, since, like prepositions, it relates two DPs. With Kayne, we resist the notion that this element is an abstract verb, nor do we think it is reasonable to identify it as a verbal applicative of the kind familiar from agglutinating languages like Swahili or Hiaki. Nonetheless, as long as the key properties we have identified here are ascribed to it, its precise label is possibly immaterial. In our opinion, however, the properties that have been identified here, and those outlined in previous work, including its possessive semantics and the (in)alienability restrictions connected to animacy that it imposes, make it reasonable to maintain that PHAVE is the most appropriate name for this category.
The authors gratefully acknowledge the very valuable feedback of reviewers, Ben Bruening, John Beavers, the members of the Spring 2012 dissertation workshop at the University of Arizona, Yosuke Sato, and many sundry friends and associates for their responses to impromptu judgment requests. All shortcomings remain our responsibility.
2 As observed originally by Oehrle (1976), this effect gives rise to a nice interpretive coercion in cases where an NP that is ambiguous with regard to animacy occurs in Goal position in the double object construction. In the to-dative (i), London can refer either to the city or, metonymically, to members of a company or organization located in the city. However, in the double object construction (ii), only the latter interpretation is possible, since the Goal in that structure must be animate—the ‘‘London office’’ effect. Since have is subject to the same animacy restriction on its subject, as shown in examples (2) and (3), the same disambiguation of London operates there, as illustrated in (iii).
Brenda sent a book to London. (Either regular or office reading OK)
Brenda sent London a book. (London office reading only)
London has the book. (London office reading only)
3 Benjamin Bruening ( pers. comm.) suggests that the parallel here is not complete, in that he finds that (4b) is acceptable when the baby is in utero, but (4a) is not. We disagree, finding that (4a) is fine in that circumstance, although pragmatically unusual. More pragmatically self-evident possession examples such as Our baby has a college fund seem clearly acceptable even when the baby is in utero.
5 We use the term lexical causative as a descriptor here despite subscribing to a nonlexicalist account of such alternations; see Marantz 1997 for discussion.
6 Indeed, it is possible that they are too flexible, as noted by Benjamin Bruening ( pers. comm.). Possessive ’s can be used to express prospective possession (as in This land will be subdivided when its owner dies. The portion on theleft is John’s land), but also agency ( John’s painting is owned by Mary) or even simple association (John suggested Igo to a club in Soho. Mary mentioned one in the meat-packing district. John’s club was lame, so I went to Mary’s). However, a reviewer points out that many such relations can be expressed by verbal have as well, given the right context. For example, Tham (2006:139) shows that prospective possession can be encoded by actual verbal have, in the correct context. She argues that if speaker B has a plan to distribute gifts to particular friends and is discussing this plan with speaker A, the following exchange is acceptable:
7 See also Kratzer 2013 on the failure of ‘‘existential exportation’’ as a general phenomenon applicable to all verbs of transfer of possession. Kratzer argues that the intensional property of verbs such as allow, bequeath, and owe, such that John bequeathed Bill a horse does not necessarily entail that there is a horse that Bill has, is actually available to all verbs in the family.
8 A reviewer mentions that examples like (10)–(11) have a grammatical interpretation (in the reviewer’s judgment) when they are interpreted as habituals or generics, rather than as episodic. The reviewer points to the interpretability of examples such as (i).
I consider her unattractive drunk.
When an episodic reading is enforced by means of a temporal adverbial, the availability of the depictive interpretation again degrades, consistently with our claim in the text.
??I considered her unattractive yesterday drunk.
We set aside the possible interference of this habitual/generic interpretation here.
9 See section 3.3 for evidence that have (and hence PHAVE) inherits eventiveness from its complement.
11 Indeed, results of a processing experiment by Wittenberg and Piñango (2011) support the notion that a coercion process is involved in the composition of give with eventive nominals, in contrast to its composition with straightforward entity-denoting nominals. They showed that there is an additional processing cost involved in give a kiss–type stimuli in contrast to stimuli like give a pen (or other phrases with ‘‘literal’’ interpretations of give). This is consistent with our proposal below that give (and have) are the same verbs with both kinds of complements; the coercive interpretive process (likely involving type-shift) that is required when an event-denoting nominal composes with them accounts for the additional processing time required.
12 The availability of a depictive with any predicate is subject to poorly understood variation. Benjamin Bruening ( pers. comm.) reports finding that examples like John had a party drunk and John had a fit naked are less natural than John had a massage drunk. We find that John had a fit naked is perfect, while John had a party drunk is indeed somewhat degraded. However, the same kind of variation can be found with other eventive predicates (e.g., ?Mary failed the testdrunk or ??John held the bird drunk); we assume that this kind of variation with eventive have and depictives is subsumed as part of the general variability of depictive modification.
13 Benjamin Bruening ( pers. comm.) notes that if have with an eventive DP receives a ‘‘light verb’’ analysis like the one Pylkkänen gives to give with eventive DPs, her account would then extend to predict that they should behave the same with depictives. This may be the case, but such an analysis would require two separate verbs have as well as two separate verbs give (light and heavy versions of each). This particular connection would not have anything to do with the have/give relationship; rather, it would be due to the character of ‘‘light verbs’’ in general.
It is also worth noting that Piñango, Mack, and Jackendoff (2006) suggest that a compositional clash in combining a given verb with an eventive nominal explains a processing slowdown they observed in ‘‘light verb’’ constructions but not their heavy verb counterparts. This is consistent with our line of thinking, in which the single verb has a basic entity-selecting denotation, triggering a type clash that must be resolved when it is composed with an eventive complement (see footnote 11 and discussion below). This approach can also provide insight into the other have/give parallels, while the ‘‘light verb/heavy verb’’ stipulation does not.
14 Benjamin Bruening ( pers. comm.) notes that Pylkkänen’s analysis of depictives accounts for the fact that when the Goal of a double object construction undergoes passivization, depictive modification suddenly becomes possible.
*John told me the news drunk.
I was told the news drunk.
Since passivization in Pylkkänen’s analysis involves abstraction over VoiceP, it results in the type 〈e,〈v,t〉〉, which is the appropriate type for the targets of depictive modification in her account.
Pylkkänen (2008:39) notes a significant difficulty with this approach, however: it cannot distinguish between A-moved Goals and Ā-moved Goals. Ā-extraction of the Goal should produce exactly the same type in the low ApplP, but such Goals cannot be modified by a depictive.
(iii) *Who did John tell the news drunk?
In section 4.4, we allude to a distinction between A-Merge and Ā-Merge that could be appealed to in the syntactic description of depictives to account for this contrast.
15Harley (2007) continued to label the projection below vCAUSESC, for small clause, indicating a predication relationship between the Theme in the specifier of the V projection and the constituent labeled V´´in (19). We do not address this issue directly here, but it seems likely that this projection cannot be straightforwardly treated as a small clause. The element in Spec,VP must be selected for by V directly (rather than introduced by a special predication relation) in order for some of the arguments below to be valid. Depending on one’s assumptions about the typical characteristics of a small clause, SC may or may not be an appropriate label for the VP constituent in (19).
16 In fact, the final structure adopted by Bruening involves another verbal projection, vP, between ApplP and VoiceP (Bruening 2010a:551–552). Bruening assumes this v is verbalizing the structure so far built, but he still maintains the label V for the verbal root. His final conclusion that the lowest projection in a double object structure is not verbal brings his account closer to the PHAVE proposal than might at first be apparent.
17 We use vP here to subsume two projections that in fact we assume are distinguished in the correct, fully explicit representation, VoiceP and vP. For present purposes, this conflation is not significant.
18 Although we do not propose a detailed semantic treatment of the interpretive consequences of Manner adjunction here, the adjunction operation must result in the strengthening (or lack thereof ) of Beavers’s (2011) prospective possession relation that is contributed by PHAVE, depending on the lexical semantics of the particular adjoined root.
19 If the analysis proposed in (21) posited head movement of PHAVE to v, we too could appeal to the null-affixation explanation for the blocking of further derivational morphology in the double object construction, since PHAVE is null.
20 An alternative approach to licensing subextraction from a left branch might appeal to lexical selection/D-linking in Pesetsky’s (1987) sense. If left branches are generally islands because of, for example, renumeration, as Uriagereka (1999) proposes, then both (23a) and (23b) are predicted to be ungrammatical. The difference could lie in the fact that in (23a), but not (23b), the element that is subextracted from is selected by the verbal head of its containing projection; the idea would be that lexical selection helps the parser with opaque left branches. Such an account would not differentiate between the present analysis and Bruening’s (2010a), as the Theme is lexically selected in both, and the Goal is not. However, John Beavers ( pers. comm.) notes that there is not an explicit theoretical basis for this ameliorating effect of selection, and that Stepanov (2007) has independently argued that Uriagereka’s account of Condition on Extraction Domain effects is untenable.
21 With regard to the notion that the phase boundary is relevant to the definition of binding domain, it is worth considering the cases mentioned in Pesetsky 1995:315n151, attributed to Yvonne Bordelois by Chomsky (1972). Bordelois observed a contrast in grammaticality between (i) and (ii) (which also is observed with reflexives (iii) and possessive reciprocals (iv)).
John and Mary let [the honey drip on each other’s feet].
*John and Mary let [Sue drip honey on each other’s feet].
John let [the honey drip on himself ].
*John let [Sue drip honey on himself ].
These examples suggest that when the embedded subject of a small clause is the subject of an unaccusative, a CFC is not formed; in modern parlance, this may indicate that these complements do not introduce a phasal boundary, as in Chomsky’s (1995) original conception of phase theory. Benjamin Bruening ( pers. comm.) suggests that the contrast here has more to do with logophoricity than unaccusativity, noting that when the intervening subject is inanimate, an improvement results.
(v) ??John let the honeycomb drip honey on himself.
22Bruening (2010a) does not make a proposal concerning the internal structure of verbal have, but a reviewer suggests that if he were to extend his analysis of double object give to have, his account of the scope-fixing facts in the former would extend to the latter. However, this is actually not the case, given Bruening’s treatment of scope flexibility in passives of give. That account in fact predicts scope flexibility for have even if have is analyzed as involving an ApplP, since there would be reconstructing A-movement from the base position in Spec,ApplP to Spec,TP, perhaps via intermediate A-positions. The hypothetical account might instead require some version of the proposal we sketch below.
23 On this approach, the null PHAVE element would resemble the abstract clitic posited by Keyser and Roeper (1992). Another notable property of PHAVE is its failure to assign Case to its complement; in this respect, it is more like a particle than a preposition. See Wechsler 2008 for extensive discussion.
24Bruening’s (2010a) treatment of the possessive relation between the Goal and the Theme of a double object structure is somewhat involved. Because he proposes a ‘‘high’’ applicative structure, where the Appl head takes a VP as its argument, in order to explain how the possessive semantics (excluding the verbal event) arises, Bruening argues that the semantic interpretation of the main verb is put off until after head movement of V to the higher heads in the extended verbal projection. On his theory, V, in the double object structure, is interpreted at the level of v, above ApplP (Bruening 2010a:552), not in its base-generated position next to the Theme. With the verbal semantics not present in the lower projection, the higher Appl head composes directly with the denotation of the Theme, and directly relates the Goal to the Theme (as does Pylkkänen’s (2002, 2008) ‘‘low’’ applicative), giving the possessive interpretation. The Goal and the Theme then raise and adjoin to vP, and are recycled as arguments of the raised V; as in a complex predicate construction, they serve as arguments of two separate predicates. Bruening’s account, then, does not involve a truly ‘‘high’’ applicative in Pylkkänen’s sense.
25Bruening (2010a:538) proposes that the low verbal element in double object structures is an abstract g- element, rather than actual give, in order to accommodate the give~get alternating idioms in (29). Note that some idioms involve other verbs as well, particularly take, as in (29d). This g- element, then, is effectively identical to PHAVE, as it is not linked to the surface form of any particular verb and is common to all three; if Bruening extended it to include have, the two proposals would be identical save for the (superfluous) presence of ApplP in Bruening’s account.
26http://www.mes-blogs.com/top-9/2996836/Kudos-For-Katie-Shortlived.html (accessed 13 June 2012).
27http://www.forums.mlb.com/n/pfx/forum.aspx?tsn=26&nav=messages&webtag=ml-phillies&tid=106797 (accessed 13 June 2012).
28http://www.delawareonline.com/article/20120529/NEWS14/305290043/Military-spouses-get-hand (accessed 13 June 2012).
29http://connection.ebscohost.com/c/articles/9602020163/at-home-businesses-get-sympathetic-ear-capitol-hill (accessed 13 June 2012).
30http://waywardgukin.wordpress.com/tag/waegukin/ (accessed 13 June 2012).
31http://dailycaller.com/2012/05/03/rogue-sites-get-the-cold-shoulder-from-advertisers/ (accessed 13 June 2012).
32http://eu.battle.net/d3/en/forum/topic/4849015263 (accessed 10 July 2013).
34http://crosscurrentworship.com/category/art/ (accessed 10 July 2013).
35http://www.cruisersforum.com/forums/f23/eastern-philippines-still-suspect-74091.html (accessed 10 July 2013).
37http://www.jacklondons.net/writings/WhiteFang/5wf1.html (accessed 10 July 2013).
40http://spiritualguidancecoach.com/intermediate/ego/ (accessed 10 July 2013).
41 Bruening (2010a:539) reports a parallel example with only a single question mark (?To skittish horses should be given a wide berth) and takes its reported near-grammaticality to support his account. We disagree with the grammaticality judgment on that example. The other cases he discusses in that paradigm are good candidates for DP idioms, as suggested by the data in (32).
42 The facts here are not always crystal clear; some restitutive agains are available with these verbs, in the judgment of the first author and other native speakers we have consulted. Note that the judgments here contradict the data reported in Bruening 2010a:558.
They spared him that ordeal again. (He had undergone the ordeal before.)
He denied William the raise again. (He had successfully requested a raise before.)
43 The low interpretive position of the verbal component for these verbs might suggest that Goal-oriented depictives should be well-formed with them, assuming that the verb introduces an event argument; in that case, the lower projection would not be stative. In fact, although some cases seem reasonably interpretable, as in (i) and (ii), others are clearly ungrammatical.
They spared mei that ordeal drunki.
I envy youi your equanimity drunki.
*John denied Billi a raise drunki.
*The traffic violation cost mei my car drunki.
However, as Benjamin Bruening ( pers. comm.) notes, the depictive period under consideration is the period of the ordeal (in (i)) or the period of the equanimity (in (ii)), not the period of the verbal event of sparing or envying. We conclude that while these examples may be telling us more about the licensing of depictives with eventive nominals, they do not conform to the prediction that a low position of base generation for the verb should make Goal-oriented depictives possible. We leave this problem for future work.