“tight links” make convenient metaphors but loose explanations: replying to a reply
TRANSCRIPT
Language Sciences 42 (2014) 15–29
Contents lists available at ScienceDirect
Language Sciences
journal homepage: www.elsevier .com/locate / langsci
‘‘Tight links’’ make convenient metaphors but looseexplanations: replying to a reply
0388-0001/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.langsci.2013.11.002
⇑ Tel.: +81 666052428.E-mail address: [email protected]
Seizi Iwata ⇑English Department, Graduate School of Literature and Human Sciences, Osaka City University, 3-3-138, Sugimoto, Sumiyoshi-ku, Osaka 558-8585, Japan
a r t i c l e i n f o
Article history:Received 26 May 2013Received in revised form 17 November 2013Accepted 17 November 2013
Keywords:Unique Path ConstraintType A resultativesType B resultativesTight linkConstruction grammar
� 2013 Elsevier Ltd. All rights reserved.
1. Introduction
In Goldberg’s (1995) constructional approach to resultatives, resultatives are uniformly accounted for by means ofconstructions, irrespective of whether the ‘‘base’’ verb is transitive as in (1a) or intransitive as in (1b).
(1)
a. He hammered the metal flat. b. They yelled themselves hoarse.Iwata (2006), agreeing that Goldberg’s approach is quite attractive, nevertheless argues that not all resultatives are to behandled that way. Specifically, Iwata (2006) contends that while resultatives like (1) (=Type A) may be handled along thelines of Goldberg (1995), those like (2) and (3) (=Type B) are to be analyzed by means of a result phrase construction.
(2)
a. He painted the wall red. b. The lake froze solid.(3)
a. He cut the meat thin. b. He tied his shoelaces tight.Iwata (2006, appendix) also argues that the Type A/Type B distinction cannot be accommodated in Broccias’ (2003) theory,by pointing out a number of problems, in an appendix.
Now Broccias (2013), as a reply to Iwata (2006), claims that ‘‘Iwata’s theory is based on questionable assumptions andthat Broccias’ theory is compatible with Iwata’s observations without the need to invoke any of its conceptual machinery.’’
16 S. Iwata / Language Sciences 42 (2014) 15–29
The aim of this paper is to examine whether Broccias (2013) indeed achieves what he claims to do. This paper is organizedas follows. Section 2 briefly reviews Iwata (2006) and then Broccias (2013). Section 3 examines Broccias’ arguments to showthat Type B resultatives can be analyzed in the same way as Type A resultatives, and Section 4 examines his claims about‘‘tight links.’’ Section 5 further examines the validity of his ‘‘tight link’’ approach as applied to sentences involving to one’sdeath. A conclusion is given in Section 6.
2. Two types of resultatives
2.1. From verbs to constructions
Let us start by clarifying the significance of Iwata’s (2006) proposal in the context of research history on resultatives.Probably the earliest study of resultatives in the generative tradition is Green (1970). Green drew attention to resultativesby citing examples like (4).
(4)
a. Jesse shot him dead. b. She painted the house red. c. She kicked the door open.(Green, 1970, p. 271)
While Green herself did not advance a detailed analysis, a specific view of resultatives was implicit in her exposition. That is,Green discussed resultatives on a par with depictives like (5).
(5)
a. He ate the apple whole. b. He ate the meat raw.(Green, 1970, p. 272)
This suggests that Green was implicitly supposing that resultatives are formed by adding an AP to a VP, in the same way thata depicitive phrase is added to a VP. Thus, according to this view, shoot him dead is obtained by somehow adding the AP deadto the VP shoot him, as in (6).
(6)
[shoot him] + [dead] = [shoot him dead]For the sake of convenience, this view of resultatives will be referred to as an AP-addition analysis. The AP-addition analysishas been assumed by most of the subsequent studies on resultatives in the generative tradition.
Eventually, it turned out that the AP-addition analysis may not be correct, however. Rivière (1982) pointed out thatcertain verbs that occur in resultative sentences violate selectional restrictions. In (7), the verb drink is followed by animateNPs, despite the fact that this verb should select for an inanimate, in fact fluid, object.
(7)
a. They drank him under the table. b. He drank himself senseless.(Rivière, 1982, p. 686)
Also, subcategorization restrictions may be violated. The verb laugh is strictly intransitive, and normally laugh cannot be fol-lowed by direct object NPs, as shown in (8).
(8)
a. He laughed himself sick. b. ⁄He laughed himself.All these facts strongly argue against the AP-addition analysis. After all, laugh himself sick cannot possibly be formed byadding the AP sick to laugh himself.
(9)
⁄[laugh himself] + [sick] => [laugh himself sick]Rather, the set of data which Rivière (1982) observes can be elucidated only under an analysis which assumes that thesyntactic frame [NP V NP XP] is crucial, irrespective of whether the putative base verb is transitive or intransitive. But thispossibility has been rarely, if ever, explored among generativists.
More than a decade later, a solution to the problem raised by Rivière (1982) was proposed outside the generative frame-work. Specifically, Goldberg (1995) argues that it is constructions, rather than verbs, that account for the phrasal patterns ofcertain linguistic phenomena, including resultatives.
On a constructional approach to argument structure, systematic differences in meaning between the same verb in differ-ent constructions are attributed directly to the particular constructions.
(Goldberg, 1995, p. 4)
Fig. 1. Composite structure: Resultative + talk (Goldberg, 1995, p. 190).
S. Iwata / Language Sciences 42 (2014) 15–29 17
Constructions are form-meaning pairings which exist independently of particular verbs, carry meaning, and specify the syn-tactic frame. In each construction, the constructional meaning is integrated with the verb meaning.
Goldberg argues that constructions can override the syntax and semantics of lexical verbs by superimposing their ownsyntax and semantics on those verbs. Resultatives are a prime example demonstrating how effective her constructionalaccount is. Thus (10b) can be accounted for on a par with (10a) by claiming that the verb talk is integrated with the transitiveresultative construction, as in Fig. 1.
(10)
a. He wiped the table clean. b. He talked himself blue in the face.According to this view, the resultative semantics resides not in the lexical verb but in the string as a whole. This view will bereferred to as a ‘fixed string’ analysis.
2.2. Iwata (2006)
Thus Goldberg’s (1995) theory, as one instantiation of the ‘fixed string’ analysis, can solve the long-standing puzzle ofresultatives. Logically, though, just because certain resultative sentences can be properly handled by means of the ‘fixedstring’ analysis does not necessarily guarantee that all resultatives should be so handled. This is exactly what Iwata(2006) claims.
Iwata (2006) concedes that resultatives with non-subcategorized objects like (11a) necessitate a ‘fixed string’ analysis, asGoldberg (1995) argues.
(11)
a. The joggers ran the pavement thin. b. ⁄The joggers ran the pavement.Iwata (2006) observes, however, that resultatives like (12a) and (12b) do not necessarily warrant a ‘fixed string’ analysis, inthat the result phrase can be deleted without affecting the grammaticality as in (13).
(12)
a. The river froze solid. b. He painted the wall red.(13)
a. The river froze. b. He painted the wall.If anything, resultatives like (12a) and (12b) are better accounted for in terms of an AP-addition analysis.Iwata (2006) advances several arguments to favor an AP-addition analysis, one of which has to do with the Unique Path
Constraint, as proposed by Goldberg (1991, 1995).
Unique Path (UP) Constraint: If an argument X refers to a physical object, then more than one distinct path cannot bepredicated of X within a single clause. The notion of a single path entails two things:
(1) X cannot be predicated to move to two distinct locations at any given time t,(2) the motion must trace a path within a single landscape.
(Goldberg, 1991, p. 368)
As is well-known among scholars working on resultatives, the proposed constraint allows one to capture the fact that in (14)black and blue and out of the room cannot occur simultaneously.
(14)
a. ⁄Sam kicked Bill black and blue out of the room. b. ⁄Sam kicked Bill out of the room black and blue.(Goldberg, 1995, p. 81)
18 S. Iwata / Language Sciences 42 (2014) 15–29
Remarkably, however, some resultatives are not subject to this constraint. Thus in (15a) and (15b) the result phrase and thepath phrase co-occur.
1
unbe
(15)
In theconstraidroppe
a.
initianed ld, fol
He spread the butter thin/thick on the bread.
) (Iwata, 2006, p. 463b.
Amelia rolled up Nina’s torn sleeve and tied the tourniquet tight around her upper arm (BNC, cited in Iwata, 2006, p. 464)Iwata (2006) observes that the difference between (14) and (15) can be accounted for by assuming that in (15) the resultphrase is accommodated by a result phrase construction (=an AP-addition analysis), while in (14) it is part of the (transitive)argument structure construction for resultatives (=a fixed string analysis).
Consequently, there are now (at least) two types of resultatives: those that are to be handled by means of a ‘fixed string’analysis like (11a) (=Type A) and those that are better handled by means of an AP-addition analysis like (12a) and (12b)(=Type B).
2.3. Broccias (2013)
Now Broccias (2013) challenges Iwata (2006), claiming that Type B resultatives can be analyzed in essentially the sameway as Type A resultatives in his theory, and that no special mechanism is necessary for Type B resultatives. Rather, Brocciasclaims that the occurrence of multiple result phrases as seen in (15) can be accounted for in terms of his ‘‘tight links.’’
In what follows, let us examine whether Broccias (2013) really succeeds in proving his point. For reasons of space, theensuing discussion is limited to these two issues, which are central to Broccias (2013): (1) whether Type B resultatives donot call for a special mechanism; and (2) whether his ‘‘tight links’’ really account for a number of linguistic phenomena,including the violation of the Unique Path Constraint.1
3. Property path or not
In Iwata (2006) the observed difference between Type A and Type B resultatives as to the Unique Path Constraint is attrib-uted to the difference in the conceptualization of change: The change of state is conceptualized in terms of a traversal of aproperty path with Type A, but not with Type B, resultatives.
In contrast, with Type B resultatives change of state is not conceptualized in terms of translational motion over a propertypath.
(Iwata, 2006, p. 464)
Broccias (2013) states three reasons why he does not believe this is the case. So let us examine Broccias’ arguments in turn.First, Broccias (2013) cites (16) and claims as follows:
(16)
a. She pounded the metal thin.lly submitted version of the present papinking, incoherent diagrams, numerouslowing the suggestion by the Editor.
(Type A)
b. He cut the salami thin. (Type B)It is not clear why e.g. thin should be related to a property path in [(16a)] but not in [(16b)]. To claim that this is sobecause (some) Type B resultatives allow for multiple RPs is of course fallacious.
(Broccias, 2013, p. 44)
A moment’s reflection tells us that the property of being thin is achieved differently. In (16a) clearly the metal was not thinbefore being pounded. But as the pounding progresses, the metal becomes thinner, and finally it attains thinness. So themetal can be said to have traversed a property path, i.e. from being not thin (probably thick) to being thin.
By contrast, in (16b) nothing tells us about the former state that obtained before the cutting event. After all, the entitywhich thin is predicated of (=slice) did not exist before the salami was cut. Rather, the property of a slice being thin‘‘emerges’’ exactly when the salami gets cut. Broccias fails to see this obvious difference.
Second, Broccias (2013) cites (17) and claims as follows:
(17)
a. He cut the salami thin. b. He cut the salami thin into slices. c. He cut the salami into thin slices.Since thin and slices evoke the same object . . . one may want to conclude that the change referred to by the PP in [(17b)] isconceptualized in the same way as the change referred to by the AP.
(Broccias, 2013, p. 44)
er, various other problems with Broccias (2013) are discussed as well, ranging from the absent syntax,misquotes, to the misunderstanding of lower-level constructions. But for reasons of space, these had to
Fig. 2. The FCS for He hammered the metal flat (Broccias, 2013, p. 38).
Fig. 3. The FCS for He cut the meat thin (Broccias, 2013, p. 39).
S. Iwata / Language Sciences 42 (2014) 15–29 19
Here an identificational change, in the sense of Jackendoff (1983, 1990), coded by into slices, and a property predicated of theentity having undergone the identificational change, coded by thin, are confused.
This confusion finds its way into the diagrams. Thus along with the diagram for He hammered the meta flat in Fig. 2,Broccias (2013) draws the diagram for He cut the meat thin as in Fig. 3.
Notice that in the change component in Fig. 3, TH traverses a property path, ending with the property of thin, exactlyparallel to the change component in Fig. 2. But the change component in Fig. 3 should express an identificational changefrom the meat into slices, not a change from the meat to (being) thin. After all, the property of being thin is not somethingthat is achieved through path-traversal (from being not thin to being thin), as pointed out above. Rather, T in the changecomponent in Fig. 3 should be linked to an element representing the newly identified state of the meat (i.e. newly cut slices),in which the property of thinness then accurately obtains. But it is mistakenly (and conveniently) linked to thin.
In this connection, Broccias (2013, p. 38, fn 12) responds to a comment by a reviewer, who points out exactly this problemof arbitrary linking, by saying that thin in He cut the meat thin can be analyzed as involving the transition from a ‘‘non-thin’’state of the meat to the ‘‘thin’’ state of the slices. But this is merely a quibble of words. The state of the slices before the cuttingcannot be characterized as the ‘‘non-thin’’ state of the meat. After all, the slices are non-existent before the cutting, as notedabove; it is impossible to predicate a property of a non-existent entity. Besides, the meat as such may have been thin beforebeing cut. Logically it is possible to cut thick slices (‘‘thick’’ as slices go) off the thin meat (‘‘thin’’ as meat goes). Thus it isdoubly mistaken to refer to the non-existence of the slices as ‘‘the non-thin state of the meat.’’
Rather, the diagram should be drawn in such a way as to place a property of being thick by the side of the change com-ponent and have that property be predicated of the result of an identificational change (i.e. slices) as in Fig. 4.
In order to give a coherent account, therefore, it is necessary to introduce a result state that comes to obtain when theverbal event of cutting is achieved, in addition to the change component that is entailed by the verb cut.
This analysis naturally extends to resultatives like paint the wall red. After all, the state of being red comes to obtain whenthe verbal event of painting is achieved. Logically, however, an alternative analysis is possible: With these resultatives, theverbal events do not result in the creation of a new entity, and accordingly a property-path traversal interpretation (e.g. fromnon-red to red) may be pragmatically inferred (i.e. before the wall becomes red, it must have been non-red).2 But these
2 But an anonymous reviewer observes that ‘‘He painted the wall red’’ does not imply that before the wall becomes red it must have been non-red, becausethere is nothing wrong with ‘‘He painted the wall red, as it had been before.’’ This observation may be taken to be evidence arguing against a property-pathanalysis of red in Broccias’ account.
Fig. 4. Alternative diagram for He cut the meat thin.
Fig. 5. A possible diagram of He painted the wall red.
20 S. Iwata / Language Sciences 42 (2014) 15–29
changes unfold simultaneously with the changes encoded by the verbs (i.e. from non-painted to painted). For these Type Bresultatives, then, a second change component, rather than a simple property, may be placed by the side of the first change com-ponent, which represents the change as directly resulting from the verbal event, as in Fig. 5.3
Thus logically there are two possibilities: The first is to apply the analysis as embodied in Fig. 4 to all Type B resultatives;the second is to use both of the analyses as embodied in Figs. 4 and 5. But the second possibility does not seem to be veryattractive, as it requires two different mechanisms when those two can be collapsed into a single one.
Either way, these diagrams do not suffer from the problem noted above. Significantly, however, they are clearly merenotational variants of the analysis of Iwata (2006). Thus because of their inconsistency, Broccias’ ‘‘arguments’’ highlight,rather than refute, the necessity to introduce a mechanism for accommodating a property which holds in addition to theresult state as entailed by the verbal event itself.
Now let us return to our original discussion and examine the third claim of Broccias. Broccias (2013) cites (18) and claimsas follows:
3
ph
(18)
Noticerase red
a.
thatis pr
He punched him to death.
b. He punched him dead.Since the change of state is construed as punctual in [(18b)], one may want to claim that the change of state is not con-ceptualized as a property path, after all. But the problem with this view for Iwata’s theory is that (i) [(18b)] is an instance
in order to capture the fact that the verb paint entails a change, the change component must be present, irrespective of whether the resultesent or not (i.e. He painted the wall).
S. Iwata / Language Sciences 42 (2014) 15–29 21
of the Type A resultative – no change is entailed by He punched him – and (ii) punch can be construed as either iterative orpunctual.
(Broccias, 2013, p. 45)
Nowhere in Iwata (2006, 2008) is the Type A/Type B distinction correlated with the aspectual contrast. The reason is simple:There is no such correlation. Thus Type A resultatives are possible whether the verb is durative (e.g. shake and beat) as in (19)or punctual (e.g. kick and shoot) as in (20).
(19)
a. He shook her awake. ) (BNCb.
. . . until he eventually beat her senseless and killed her. ) (BNC(20)
a. Without compunction, Fox kicked him awake. ) (BNCb.
Some were shot dead when thousands of women, protesting against the killings, tried to march on thePresident’s residence.(BNC)
So Broccias is ‘‘criticizing’’ Iwata for what Iwata has not said. Thus none of the three ‘‘arguments’’ hold up to scrutiny.
4. Tight link
Now is the time to assess the validity of the notion of ‘‘tight link,’’ which should be central to Broccias’ alternative anal-ysis. Broccias (2013) argues as follows:
I would like to contend that the occurrence of multiple RPs here, unlike impossible examples such as ⁄Sam kicked Bill blackand blue out of the room (. . .) is quite trivially motivated by the fact that the two RPs are tightly linked, i.e. they refer to (orspecify) two intrinsic facets of the same common event.
(Broccias, 2013, p. 47)
There are a number of problems. First and foremost, this idea is obviously a re-wording of Iwata’s (2006, 2008) observa-tion: It is a fact that before Iwata (2006), the very notion that with some resultatives the verbal event and the change of stateare co-extensive and co-temporal had not been duly recognized by Goldberg (1995), Boas (2003), or Goldberg and Jackendoff(2004), let alone Broccias (2003) (Levin being the sole exception). And obviously, Broccias’ development of his ‘‘tight link’’ is‘‘inspired’’ by a number of Iwata’s (2006, 2008) original observations. But incredibly, Broccias pretends as if he was the firstto come up with this idea.
Interestingly, Iwata himself is aware that the problem with these examples is that the conceptual distance between thetelic predicate and the hypothetical result phrase is too great:
(Broccias, 2013, p. 48, my emphasis)
But no one can fail to notice that Broccias owes the idea of ‘‘tight link’’ to Iwata (2006, 2008). Thus Broccias quotes thefollowing passage from Iwata (2008), which is about the sentences in (21) and (22).
(21)
a. The door swung open inwards into the bedroom. b. Her robe fell open to the waist, her small jutting breasts exposed, heaving withindignation.
c. Kicked the radiator and screamed as the bonnet fell shut on his fingers.(22)
The dog, which had followed its master onto the roof, jumped down towards the lifeless body, missed, and felldead on the stones below.(B
NC)In contrast, in [(21)] and [(22)] a change of state is a direct consequence of a physical motion, be it translational or inter-nal.
(Iwata, 2008, p. 1077)
The above quote from Iwata (2008) is immediately followed by the following passage:
This close relationship or short distance between multiple facets of a complex event is what I mean by ‘‘tight links.’’(Broccias, 2013, p. 49, my emphasis)
This sentence clearly shows that Broccias is simply rephrasing what is argued in Iwata (2008).Rather surprisingly, however, Broccias (2013, p. 47, fn 24) responds to a comment by a reviewer, who points out exactly
this problem, by saying that ‘‘Iwata’s wording reproduces Broccias’ (2003).’’ As evidence, Broccias cites the following passage,which is about The mansion burned down:
22 S. Iwata / Language Sciences 42 (2014) 15–29
In other words, the two events in question are temporally co-extensive (see Rappaport Hovav and Levin for a similarpoint) because the causal relationship existing between them has been compressed into (the vital relation of) identity. Theburning of the mansion and its collapse are different facets of the same event.
(Broccias, 2003, p. 222, emphasis in the original)
But it is not a matter of whether a similar wording happens to be found in Broccias (2003) or not. The wording in question islimited to this part alone and does not play any significant role in Broccias (2003): It is not found in the discussion of The riverfroze solid (p. 34), in the discussion of the Unique Path Constraint (p. 117), or anywhere that has some bearing on Type Bresultatives.
As a matter of fact, Broccias (2003) employs ‘‘tight link’’ in an entirely different sense. Thus Broccias (2003) claims that theresultatives in (23) are well-formed because of the ‘‘tight link.’’
4
to
(23)
Thus tbe acco
a.
he clemmo
They drank me under the table.
b. Alice cooked Tom and Bill to death. c. Penny surfed the night away.(Broccias, 2003, p. 136)
we observe that the verbal events and the change events coded by the change complexes are tightly linked.(Broccias, 2003, p. 136, my emphasis)
Broccias (2003) goes on to say that certain resultative sentences are unacceptable because of the ‘‘loose link’’ between thetwo subevents:
More in general, case 2) change constructions show tight links between the verbal event and the change event, which doesnot seem to be the case with the impossible sentence ⁄The bears frightened the campground empty.
(Broccias, 2003, p. 137, my emphasis)
The resultatives in (23) are all instances of Type A resultative, and the ‘‘tight link’’ in Broccias (2013) is intended to accountfor Type B resultatives. So the ‘‘tight link’’ in Broccias (2013) cannot possibly find its origin in Broccias (2003).
Now back to the notion of ‘‘tight link.’’ The second problem is that it is nothing more than a metaphor. After all, eventscannot be literally ‘‘tied.’’ In order to be a viable account, the notion must be defined. But Broccias does not even begin to tryto give a definition.
Third, this is circular: Broccias is saying that multiple result phrases are possible because they are ‘‘tightly’’ linked. But tothe question of when the links are ‘‘tight,’’ one can only deduce from Broccias’ (paucity of) criteria that ‘Whenever multipleresult phrases are possible’ there must be tightness.
Fourth, Broccias’ ‘‘tight link’’ actually fails to offer a viable account of when and why multiple result phrases are allowed.Broccias (2013) claims as follows:
Still, it should be observed that the ties between the causing event and the caused event are tighter when the causingevent is telic (i.e. with Iwata’s Type B verbs) than they are when non-telic predicates (i.e. Iwata’s Type A verbs) are used.
(Broccias, 2013, p. 49)
Here Broccias equates the possibility of multiple result phrases with the degree of ‘‘tightness.’’ Essentially, Broccias is claim-ing that multiple result phrases are possible with Type B but not with Type A, because the link is tighter with Type B verbsthan with Type A verbs. This is another way of saying that verbs that occur in the Type A resultative (i.e. verbs which do notlexically encode a change) never allow for multiple occurrences of a result phrase.
Actually, however, multiple result phrases are observable even with verbs that are found to occur in the Type Aresultative.4
(24)
a. He wiped the blade clean on his skin coat and walked out. b. Afterwards he washed the tin out in the stream, splashed water over his face and hands and wiped them dryon a handkerchief.
(both from BNC)This is possible even with the prepositional result phrase to death.
(25)
The lad on its back, dolled up as King William, had been crushed to death against the wall. (BNC)To sum up this subsection, it won’t do to simply say that multiple result phrases are possible when those facets of anevent are ‘‘tightly linked’’: first, this is simply a rephrasing of what Iwata (2006, 2008) observes; second, this is nothing more
an and dry in He wiped it {clean/dry} are the result phrases that are to be accommodated by means of a fixed string analysis, but those in (24) aredated by means of an AP-addition analysis, as noted in Section 2.
S. Iwata / Language Sciences 42 (2014) 15–29 23
than a metaphor, and not a theory; third, it is circular; fourth, multiple result phrases are allowed even with verbs that ap-pear in the Type A resultative, contrary to Broccias.
Broccias (2013) ‘‘criticizes’’ Iwata (2006) on the grounds that the term ‘‘adjunct’’ is a controversial notion. But at least adefinition is given in Iwata (2006), to the effect that the result phrase in Type B is not licensed by argument structure con-structions. This is certainly far better than no definition at all.
5. To one’s death
Broccias (2013) claims that his ‘‘tight links’’ can account for why to one’s death may co-occur with path PPs. Thus althoughBroccias’ ‘‘tight links’’ have been shown to be untenable in the last section, logically there is still a possibility that this notionmight be valid for handling this phenomenon (alone). It is therefore necessary to examine whether to one’s death can beaccounted for in terms of ‘‘tight links,’’ as Broccias claims.
But we must first make it clear how Broccias (2013) has come to make this proposal: Broccias’ (2013) account of to one’sdeath in terms of his ‘‘tight links’’ is presented as a response to Iwata’s (2006) criticism of Broccias’ (2003) handling of a sen-tence involving to one’s death.
5.1. For what does Iwata (2006) criticize Broccias (2003)?
As already noted above, Goldberg (1991, 1995) proposes a Unique Path Constraint:
Unique Path (UP) Constraint: If an argument X refers to a physical object, then more than one distinct path cannot bepredicated of X within a single clause. The notion of a single path entails two things:
(1) X cannot be predicated to move to two distinct locations at any given time t,(2) the motion must trace a path within a single landscape.
(Goldberg, 1991, p. 368)
The proposed constraint allows one to capture the fact that in (26) black and blue and out of the room cannot occursimultaneously.
(26)
a. ⁄Sam kicked Bill black and blue out of the room. b. ⁄Sam kicked Bill out of the room black and blue.(Goldberg, 1995, p. 81)
Now Broccias (2003) argues as follows:
It is also worth stressing that, whereas condition (1) of the Unique Path Constraint seems to be correct in the sense thattwo endpoints cannot be specified simultaneously, the following example belies the correctness of condition (2):
(27)
The tyrant ordered them to jump to their death off the castle.[(27)] clearly evokes two landscapes, one pertaining to life (cf. jump to their death, where to their death specifies the goal ofmotion) and the other pertaining to physical motion (cf. jump off the castle, where off the castle specifies the source ofmotion).
(Broccias, 2003, pp. 116–117)
As one can easily see, Broccias (2003) cites (27) as a counterexample to the Unique Path Constraint.Iwata (2006, pp. 488–89) points out that this is quite beside the point, since this type of example has already been
handled by Goldberg (1991) in discussing sentences like (28).
(28)
He pushed her through the window to her death. (Goldberg, 1991, p. 372)According to Goldberg (1991, p. 372), to her death, unlike to death, is an idiom which metonymically stands for ‘‘the pathwhere she died.’’ Consequently, sentences like (27) are not counterexamples to the Unique Path Constraint.
Iwata (2006) thus simply points out Broccias’ oversight. Strangely enough, however, the supposed response to this issueby Broccias (2013) is as follows:
The tight link manifests itself overtly by way of the prepositional phrase to her death, which, unlike to death, presumes aphysical path. Hence, Iwata’s objection is not a counterexample to Broccias’ theory but rather provides a further sub-stance to it.
(Broccias, 2013, p. 49, fn 29)
The second part of this quote is entirely faulty, as Iwata (2006) cites (28) as an illustration of Broccias’ failure to properlyunderstand Goldberg’s theory, not as a counterexample to his account. But what is really crucial is the first part: Here
24 S. Iwata / Language Sciences 42 (2014) 15–29
Broccias is repeating Goldberg’s (1991) observation that to her death, unlike to death, expresses a physical path. But to admitthat to her death expresses a physical path is to admit that (27) is not a counterexample to the Unique Path Constraint, namelythat Broccias’ (2003) allegation above is indeed mistaken. So that should be the end of the discussion, at least as far as the issueraised by Iwata (2006) is concerned. Strangely, however, Broccias does not seem to be aware of this obvious fact.
5.2. Two arguments of Broccias (2013)
Anyway, let us go on to see what Broccias (2003) has to say. As seen above, Goldberg argues that (28) is not a counter-example because to her death is an idiom which metonymically stands for ‘‘the place where she died.’’
Broccias (2013) challenges Goldberg’s argument. In discussing (29), he claims that the co-occurrence of to one’s death witha motion verb is accounted for in terms of his ‘‘tight link’’:
5
6
co
(29)
Examp(i) HeThe W
nsists of
The tyrant ordered them to jump to their death off the castle.
(Broccias, 2003, p. 117)Two landscapes are involved, one having to do with physical motion and the other with metaphorical motion from life todeath, but they are tightly linked by virtue of almost unavoidable causality: falling off a tall building is very likely to result indeath. It is a convention of English that to one’s death, rather than to death, is used to code death resulting from e.g. a fall.
(Broccias, 2013, p.49, fn 29, my emphasis)
Broccias (2013) argues against a metonymy analysis on the following two grounds. First, he claims that in (30) the metonym-ic source expression (e.g. on the ground) and the metonymic target expression (i.e. to one’s death) co-occur, contrary to theprediction of the metonymy analysis.
(30)
. . . the bomber instead chose to leap from the balcony’s rail to his death on the ground below. (web)Second, Broccias cites (31) and asks the question: ‘‘Can one really say that the location on the ground reached by the dare-devil stands for a place in the history book?’’
(31)
Wingman daredevil falls from 37,000 feet . . . into the history books. (web)I do not see the relevance of Broccias’ second argument, as (31) does not involve to one’s death. After (31) is set aside asirrelevant, then, Broccias’ claims are summarized as follows: (1) Physical motion and metaphorical motion from life to deathare tightly linked by virtue of almost unavoidable causality; and (2) to one’s death is not a metonymy; it is a convention ofEnglish that to one’s death, rather than to death, is used.
5.3. What the corpus data tell us
5.3.1. To one’s death is not a result phraseIn assessing which analysis is more viable, let us start by examining the attested data. By searching the BNC, I have come
across 171 instances of to one’s death in the relevant sense.5 The list of verbs appearing with to one’s death, along with thenumbers of occurrence, is summarized in Table 1.
Similarly, I have come across 145 instances of to one’s death in the Wordbank Corpus.6 The list of verbs and their numbersof occurrence are summarized in Table 2.
Both corpus data exhibit very similar patterns: The three most frequent verbs are the same (fall, plunge, and send), and theother less frequent verbs are also more or less the same across the two corpora (go, lure, drive, jump, etc.).
What is immediately obvious from these corpus data is that all the verbs necessarily involve motion. That is, to one’s deathnecessarily accompanies physical motion, fully in accordance with Goldberg’s observation.
At the same time, it turns out that to one’s death is not linked with physical motion by virtue of ‘‘almost unavoidable cau-sality,’’ contra Broccias (2013). Consider (32).
(32)
a.les lihad bordbfifty
He walked to his death at Reading on 7 July 1896, the famous hangman Billington in charge of operations.
b. Or did he go to his death in the anguished conviction that he was responsible for his master’s?(both from BN
C)Neither walking nor going results in death. Rather, he walked or went to a place where his capital punishment was to be executed.
ke the following are excluded from this count:een cautious of the 9pm appointment which led to his death.
ank Corpus is a subcorpus of the Bank of English and is made accessible by Shogakukan Corpus Network (http://wordbanks.jkn21.com/). It-six million words drawn from British, American and Australian Englishes.
Table 1BNC counts of V to one’s death.
_____ to one’s death
fall 51plunge 22send 16go 13lure 12drive 7leap 6jump 5bring, sweep 3dance, drag, hand, hunt, sink, swim, throw, walk 2advance, carry, crash, dash, deliver, dive, draw, hound, leave, repatriate, ride, push, skip, suck, thump, topple, tumble 1
Total 171
Table 2Wordbank counts of V to one’s death.
_____ to one’s death
fall 32plunge 22send 16jump 13sweep 9lure 7drive, go, hurl 5carry, drag, throw 4chase, suck 2crash, drop, hound, leap, plummet, push, sail, ship, slide, splash, stumble, take, transfer, transport, wash 1
Total 145
S. Iwata / Language Sciences 42 (2014) 15–29 25
Also, in (33a) Yugoslavs and Cossacks were sent to the battlefield; and in (33b) the Russian prisoners were handed to theconcentration camp where death awaited them.
7
(33)
I am in
a.
debt
. . . over a pamphlet accusing him of deliberately sending 70,000 Yugoslavs and Cossacks to their death in1945.
b.
If we hand the Russian prisoners back to their death, it will be the military authorities who do so, on myinstructions . . .(both from BNC)
Again, therefore, the physical motion in itself does not cause the death.Similar examples abound. Consider (34).
(34)
a. . . . do we, or do we not approve of killing animals for fun do we approve of allowing a pack of dogs to hunt ananimal to its death.b.
A WOMAN is believed to have lured British computer boss Howard Bates to his death and then helped bury hisbody in a swampland grave. Police in Florida believe Magaly Carr, 29, telephoned father-of-three Mr Bates andset up the fateful meeting where her boyfriend carried out the killing. The body of Mr Bates, 44, was foundshot through the head at a derelict factory site in Miami at the weekend.c.
As Gloucester’s knight he earned an ignominious name at Bannockburn in 1314 by leaving his lord to his deathin the mêlée.(all from BNC)
In (34a) the animal is hunted to a place where it will be finally killed by the dogs; in (34b) Howard Bates was lured to a placewhere he was then shot to death by the woman’s boyfriend; and in (34c) even an interpretation involving no movement ispossible. After all, the knight in question may well have simply stood there doing nothing.7
Remarkably, I have found one instance of to one’s death in which the death is not strictly entailed. Consider (35).
(35)
. . . because unless it is cleared up then history will say that we were sent to our deaths by Britain. ) (BNCed to an anonymous reviewer for this observation.
The subject of this sentence is first person plural we, whose referent includes the speaker. So the very fact that this sentenceis uttered indicates that the subject entity did not die.
Clearly, all the sentences in (32)–(35) argue against the ‘‘tight link’’ thesis by Broccias: The physical motion does not causethe death. Rather, all these sentences can be made sense of by understanding that to one’s death means ‘‘to a place whereone’s death is presumed to take place,’’ which is fully in line with Goldberg’s analysis.
Note further that there can be a time lag between reaching the place and dying in the above examples. Thus in (32), onedoes not die the moment one arrives at the chamber. The same is true of (33) and (34). Thus fall to his death is a limiting casewhere one’s death occurs immediately after the physical motion.
In this sense, to one’s death significantly departs from true result phrases: It is well-known in the literature on resultativesthat ‘‘the change of state must occur simultaneously with the endpoint of the action denoted by the verb,’’ a constraintknown as the Aspectual constraint in Goldberg (1995, p. 194).
All these considerations strongly indicate that to one’s death is not a result phrase predicating a property change, fully inaccordance with Goldberg (1991).
26 S. Iwata / Language Sciences 42 (2014) 15–29
5.3.2. The status of co-occurring PPsLet us turn to Broccias’ second claim. As noted above, Broccias challenges the metonymy analysis by citing (30), repeated
here as (36), and saying that the metonymic source expression (e.g. on the ground) and the metonymic target expression (i.e.to one’s death) co-occur.
(36)
. . .the bomber instead chose to leap from the balcony’s rail to his death on the ground below. (web)This type of data is indeed attested in the BNC.
(37)
In the dark, panic-stricken by what she’d done, she ran down the wrong passage, towards the sea instead oftowards the land, slipped and fell to her death on the rocks below.(BNC)
Note, however, that the PP accompanying to one’s death is not a goal PP (e.g. onto the rocks) but a location PP (e.g. on therocks). In fact, most of the PPs found to accompany to one’s death in the corpora are location PPs. Thus in (38), in the riverand on the street, rather than into the river and onto the street, are found to occur.
(38)
a. . . . by leaping to their deaths in the river. b. And of those scenes my hero, for that is what Kong had quickly become, tumbling head-over-heels from theroof of the skyscraper to his death on the street was the one to dominate my mind.
(both from BNC)Similarly, in (39) and (40) an in PP and an at PP co-occur with to one’s death.
(39)
. . . he was responsible for sending more people to their deaths in the Siberian concentration camps than anyother KGB officer in living memory.)
(BNC (40) a. He walked to his death at Reading on 7 July 1896, the famous hangman Billington in charge of operations.)
(BNC b. They were lured to their deaths at a flat in North Belfast in an IRA ‘‘honey trap’’ operation.(Wordbank)
This casts serious doubt on Broccias’ claim that in (36) on the ground is a metonymic source of to one’s death.This does not mean that a goal PP never appears with to one’s death, however. In the BNC and Wordbank, the following
attested data are found, in which the into PP accompanies to one’s death.
(41)
a. As the sun shrank the ledge, he waited to plunge to his death into a crevasse. ) (BNCb.
Families living in the flats close to the point of impact were burned alive, trapped in their homes. Many othersjumped to their deaths into the bubbling cauldron.)
(Wordbank c. . . . she lured them for a night, only to push them each morning to their death into the swirling cataractbeneath.
(Wordbank)These are true goal PPs. But these goal PPs may precede to one’s death, as in (42).
S. Iwata / Language Sciences 42 (2014) 15–29 27
(42)
‘‘You sent Bob to Trinidad, to his death –’’ (Wordbank)This is true of the into PPs in (41) as well. Thus (43), which is the counterpart of (41a), seems to be perfectly acceptable.
(43)
As the sun shrank the ledge, he waited to plunge into a crevasse, to his death.By contrast, the in PP cannot be similarly placed before to one’s death. (44), which is the counterpart of (38b), seems to beunacceptable (in the relevant reading).
(44)
?⁄And of those scenes my hero, for that is what Kong had quickly become, tumbling head-over-heels from the roofof the skyscraper on the street to his death was the one to dominate my mind.This is quite understandable in that the on PP is a location phrase, not a goal phrase.Given that some goal PPs may accompany to one’s death as in (41), then, does this fact constitute an argument against the
metonymy analysis? Not necessarily. In the literature, it has been well-known since Gruber (1976) that more than one goalphrase may appear as in (45).
(45)
a. John sent the message to New York to Bill. b. John sent the message from New York to Philadelphia to Bill.(Gruber, 1976, p. 80)
Thus not only does the sentence cited by Broccias (2013) fail to prove his point; even true goal PPs accompanying to one’sdeath do not constitute evidence that to one’s death does not express a goal.
Thus all these considerations unambiguously argue against Broccias (2013) in favor of Goldberg (1991).
5.4. Why to one’s death means what it does
5.4.1. To one’s death as a metonymyA natural question that arises is ‘‘Why does to one’s death mean what it does, then?’’ To answer this question, let us first
have a look at (46), in which to one’s death receives a different interpretation.
(46)
From 1960 to his death he was an editor of the Journal of Symbolic Logic. (BNC )Here, his death (rather than to his death) metonymically stands for ‘‘the year when he died.’’This is made possible by the fact that death is a deverbal nominal of die. Accordingly, in death a process (die) is reified.
Being a reified process, death includes in its conception its temporal location. By highlighting this temporal location, wecan understand one’s death to mean the time of one’s death. This is why in (46) his death comes to mean ‘‘the year whenhe died.’’
Note that essentially the same is applicable to his death in He fell to his death: The spatial location of a reified process ishighlighted, yielding the interpretation ‘‘to a place where he is presumed to die.’’ In this sense, one’s death indeed metony-mically stands for the place of his presumed death.
Interestingly enough, there is a further parallel between to one’s death as temporally interpreted as in (46) and to one’sdeath as spatially interpreted. While his death is construed as standing for the time of his death in (46), the exact year ofdeath may be further specified by an in-PP as in (47).
(47)
a. J Edgar Hoover ran the Federal Bureau of Investigation for nearly 50 years, from 1924 to his death in 1972. b. He was a celebrated folk hero in his adopted Darlington from his arrival in 1919 to his death in 1946.(both from BNC)
Clearly in 1972 and in 1946 are modifiers of his death, expressing its temporal locations.Note that the spatial counterparts of (47a) and (47b) are sentences like (37) and (38), repeated here as (48) and (49), in
which location PPs like on the rocks or in the river accompany to one’s death.
(48)
In the dark, panic-stricken by what she’d done, she ran down the wrong passage, towards the sea instead oftowards the land, slipped and fell to her death on the rocks below.)
(BNC (49) a. . . . by leaping to their deaths in the river.b.
And of those scenes my hero, for that is what Kong had quickly become, tumbling head-over-heels from theroof of the skyscraper to his death on the street was the one to dominate my mind.(both from BNC)
28 S. Iwata / Language Sciences 42 (2014) 15–29
Just as in 1972 and in 1946 are modifiers of his death in (47a) and (47b), so on the rocks, in the river and on the street are mod-ifiers of one’s death in (48) and (49).
Thus Goldberg (1991) is basically correct in claiming that to one’s death metonymically stands for the place where onedies. While Broccias (2013) cites the co-occurrence of an on PP with to one’s death as evidence against Goldberg’s metonymyanalysis, the location PP naturally finds its place in a metonymy account.
5.4.2. Contextual modulationLet us consider further how the metonymy in question arises. Just as one’s death is construed as expressing a temporal
location when appearing in the temporal PP as in (46), so one’s death is construed as expressing a spatial location whenappearing in the spatial path PP, like fall to one’s death. This means that both the temporal location interpretation and spatiallocation interpretation are induced by contexts.
What seems to be relevant in this connection is what Cruse (1986, p. 52) calls ‘‘contextual modulation’’: a single sense canbe modified by different contexts, each context emphasizing certain semantic traits, and obscuring or suppressing others.Contextual modulation is to be contrasted with contextual selection, by which different senses are activated by differentcontexts.
Taylor (2003) illustrates contextual modulation by citing a number of examples, one of which is car.
When we ‘wash a car’ we have in mind the car’s exterior, not all the parts of the car; when we ‘vacuum-clean the car’ wehighlight its upholstered interior; to ‘fill up a car’ is to full up only the petrol tank; while to ‘service a car’ involves itsmechanical parts.
(Taylor, 2003, p. 126)
These different uses can be coordinated without any hint of zeugma.
(50)
They washed, vacuum-cleaned, and serviced the car. (Taylor, 2003, p. 126)As can be easily seen, with contextual modulation, different aspects are simply highlighted, not that one sense is selected tothe exclusion of all the others.
Now back to to one’s death. The fact that one’s death may receive either a temporal interpretation or a spatial interpreta-tion depending on contexts points to contextual modulation, rather than contextual selection. After all, the notion of a pro-cess is still latent. Both the temporal location and the spatial location are inseparable parts of this process.
As a matter of fact, there is evidence that one’s death does not express a location pure and simple. First, death may be plu-ralized as in the following:
(51)
a. The two German tourists who fell to their deaths while on holiday on Skye have been cremated at a ceremonyin Perth.b.
The Valley Bridge, which links the South Cliff with the town centre, became notorious for people jumping totheir deaths and, since 1970, 47 have died there.c.
. . . the incorporation of wrought-iron latticework in the unglazed windows to prevent the lovelorn emulatingtheir Latin counterparts by leaping to their deaths in the river.(all from BNC)
Clearly, their deaths do not express simple locations, as separated from the dying process. Rather, when different peoplereach a place where his/her death is presumed to take place, those people (normally) end up dying. This is why their deathscomes to mean ‘‘multiple occasions’’ of death.
Second, one’s death may be accompanied by modifiers of a reified process, rather than those of a pure location. Consider(52).
(52)
Criminals will be seen going to their deaths in the gas chamber, electric chair and by firing squad in a two-hourdocumentary called The Executioners.(Wordbank)
Note that in the gas chamber, electric chair are coordinated with by firing squad. The first two modify their deaths by specifyingwhere the capital punishment is to be executed, while the third modifier by firing squad modifies their deaths by specifyinghow the capital punishment is to be executed. This latitude in interpretation suggests that to one’s death means what it doesvia contextual modulation.
5.5. Interim conclusion
Broccias’ (2013) claim that the apparent counterexample to the UP Constraint involving to one’s death can be accountedfor in terms of his ‘‘tight link’’ has been shown to be untenable. His account is built on a single example alone (i.e. jump totheir death off the cliff), and a corpus search easily reveals that the motion and the change from life to death are not linked by
S. Iwata / Language Sciences 42 (2014) 15–29 29
virtue of ‘‘almost unavoidable causality.’’ Rather, one’s death metonymically stands for ‘‘the place where one’s death is pre-sumed to take place,’’ fully in accord with Goldberg (1991). The metonymy in question is further shown to be effected viacontextual modulation.
Together with the number of problems seen in the previous section, all this shows that Broccias’ ‘‘tight link’’ is of little usein accounting for actual linguistic phenomena.
6. Conclusion
This paper has examined the validity of Broccias’ (2013) account by focusing on the following two issues: (1) whetherType B resultatives do not call for a special mechanism; and (2) whether his ‘‘tight links’’ really account for a number of lin-guistic phenomena, including the violation of the Unique Path Constraint.
As to the first issue, none of his three arguments for analyzing Type B resultatives in terms of a property path stand up toscrutiny. As to the second issue, his ‘‘tight link’’ analysis suffers from a number of problems: it is a rewording of Iwata’s(2006) observations; it is simply a metaphor; it is circular; and it makes a wrong prediction. Besides, his ‘‘tight link’’ accountof to one’s death is built on a single sentence and is easily shown to be mistaken via a corpus search.
Acknowledgements
I would like to express my gratitude to Tony Higgins, who acted as an informant and suggested stylistic improvements. Iam also grateful to an anonymous reviewer for his/her pertinent observations.
This work is financially supported by Grant-in-Aid for Scientific Research (C), Nos. 22520504, 2010–2012 and 25370558,2013–2015 from Japan Society for the Promotion of Science.
References
Boas, H.C., 2003. A Constructional Approach to Resultatives. CSLI Publications, Stanford.Broccias, C., 2003. The English Change Network: Forcing Changes into Schemas. Mouton de Gruyter, Berlin & New York.Broccias, C., 2013. Tying events tight: a reply to Iwata (2006). Lang. Sci. 38, 32–52.Cruse, D.A., 1986. Lexical Semantics. Cambridge University Press, Cambridge.Goldberg, A., 1991. It can’t go up the chimney down: paths and the English resultative. In: Proceedings of the Seventeenth Annual Meeting of the Berkeley
Linguistics Society, pp. 368–378.Goldberg, A., 1995. Constructions: A Construction Grammar Approach to Argument Structure. The University of Chicago Press, Chicago.Goldberg, A., Jackendoff, R., 2004. The English Resultative as a Family of Constructions. Language 80, 532–568.Green, G., 1970. How abstract is surface structure? In: Proceedings of the Sixth Annual Meeting of the Chicago Linguistics Society, pp. 270–281.Gruber, J., 1976. Lexical Structures in Syntax and Semantics. North-Holland, Amsterdam.Iwata, S., 2006. Argument resultatives and adjunct resultatives in a lexical constructional account: the case of adjectival result phrases. Lang. Sci. 28, 449–
496.Iwata, S., 2008. A door that swings noiselessly open may creak shut: internal motion and concurrent changes of state. Linguistics 46, 1049–1108.Jackendoff, R., 1983. Semantics and Cognition. MIT Press, Cambridge, MA.Jackendoff, R., 1990. Semantic Structures. MIT Press, Cambridge, MA.Rivière, C., 1982. Objectionable objects. Linguist. Inq. 13, 685–689.Taylor, J., 2003. Linguistic Categorization, third ed. Oxford University Press, Oxford.