argument linking and the commercial transaction frame in english, russian and japanese

Argument linking and the commercialtransaction frame in English, Russian

and Japanese

William Croft a,*, Chiaki Taoka a, Esther J. Wood b

aDepartment of Linguistics, University of Manchester, Oxford Road, Manchester, M13 9PL, UKbUniversity of California at Berkeley, Berkeley, CA, USA

Abstract

We present an analysis of the argument linking of the commercial transaction frame inEnglish, Russian and Japanese. The commercial transaction frame is semantically quite com-

plex, in that there are two transfers in opposite directions (money goes from buyer to sellerand goods go from seller to buyer). Also, the three languages examined have verbs linkingeither buyer or seller to subject position, and either money or goods to object position; theremaining two participant roles are linked to obliques. We use Croft's analysis of causal

structure in argument linking to model the variation and constraints on linking. The modeluses an underlying uniform representation of the commercial transaction frame, built in turnon the transfer of possession frame, and allows for di�erent construals of the event in di�erent

languages. The choice of oblique cases or adpositions for the oblique participant roles aremotivated by the use of the case/adposition in other semantic frames. English and Russianconstrue the commercial transaction frame in essentially the same way. Japanese construes

one aspect of the frame di�erently, leading to di�erent argument linking patterns in that lan-guage. # 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Argument structure; Lexical semantics (verbal semantics) (semantics); Syntax; Russian; Japanese;

English

1. Introduction

This paper describes the argument linking of exchange and commercial transac-tion verbs such as English buy, sell and pay, based on the framework developed by

Language Sciences 23 (2001) 579±602

www.elsevier.com/locate/langsci

0388-0001/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.

PI I : S0388-0001(00 )00037 -1

* Corresponding author.

E-mail address: [email protected] (W. Croft).

Croft (Croft, 1990, 1991, 1993, 1994, 1998). A brief comparison will be made tocommerical transaction verbs in Russian, which appear to behave in essentially thesame way as the English verbs. A somewhat more detailed comparison will be madewith commercial transaction verbs in Japanese, which behave somewhat di�erentlythan English. It will turn out, however, that both English and Japanese draw on thesame conceptual representation of the commercial transaction frame, and most ofthe di�erences are due to a relatively simple di�erence in the construal of that con-ceptual representation.Croft argues for a model of verbal semantics in which causal relationships among

participants in events play a primary role in argument linking. In particular, thedirection of the transmission of force (Talmy, 1976, 1988) determines the order oflinking of arguments in terms of grammatical relations. This observation is crucial inproviding a general characterization of the semantics of subject and object roles.Subject and object roles are polysemous, that is, they are each used for a wide vari-ety of participant semantic roles, depending on the event class of the verb (agent,patient, instrument, recipient, and more narrowly de®ned participant roles). But onecan generalize that whatever the participant roles of the arguments are, the subjectwill antecede the object in the causal chain denoted by the event class (Croft, 1998,p. 4, Linking Rule 2). To take a simple and straightforward example, in the proto-typical event type with a volitional agent and a physically a�ected patient, as in 1,the argument linked to subject position acts on the argument linked to object posi-tion:

(1) Bill broke the egg.

The directness of the transmission of force can vary to some extent, as in 2,although lexicalization patterns generally imply a relatively direct causal relation-ship between subject and object of a simple verb (see e.g. Haiman, 1983, pp. 783±787):

(2) Bill broke the egg with a spoon.

What matters is that the subject is antecedent in the causal chain relative to theobject.Oblique (prepositional) phrases may be either antecedent to the object in the

causal chain, as in 2, or subsequent to it, as in 3:

(3) Bill broke the egg for Greg.

Like subject and object roles, most prepositions are massively polysemous, that is,each preposition encodes a wide variety of participant semantic roles depending onthe event class of the verb. In virtually all cases, however, each preposition can beuniformly described as specifying either only antecedent participant roles (e.g. with)or subsequent participant roles (e.g. for; Croft, 1998, p. 4, Linking Rule 3). (In lan-guages with only one preposition, and languages where the case system is breaking

580 W. Croft et al. / Language Sciences 23 (2001) 579±602

down, the antecedent-subsequent distinction also breaks down Ð Croft, 1991, pp.188±189).There are two other signi®cant properties of the semantic representation used in

this model, both of which are characteristic of cognitive linguistic semantic repre-sentations. First, a semantic representation speci®es not only the concept denoted bythe linguistic element, but also the semantic frame in which it occurs (Fillmore, 1982,1985; Langacker, 1987, p. 183 uses the terms `pro®le' and `base' for `concept' and`frame' respectively). In a sentence like 3, the whole scene includes the causal rela-tions between agent, patient and bene®ciary (participant roles will be capitalized inthis paper); but the verb pro®les only the part of the causal chain extending fromagent to patient (indicated in boldface in 4):

(4) Agent! Patient! Bene®ciary

In 4, the verb pro®les the ®rst part of the causal chain, while the preposition forpro®les the latter part. In an example such as 3, where the obliquely encoded parti-cipant is between the subject and object in the causal chain, it is included in the verbpro®le:

(5) Agent! Instrument! Patient

The distinction between the pro®led and unpro®led portion of the event semanticrepresentation is also crucial for describing argument linking: subject and object arethe initiator and endpoint respectively of the pro®led part of the causal chain (Croft,1998, p. 4, Linking Rule 1). The three linking rules referred to above are intended toaccount for argument linking of event semantic representations to subject, objectand (both kinds of) oblique.The second salient feature of the semantic representation is the pervasive presence

of conceptualization (construal). Not all events are as simple and straightforward todescribe in terms of linear causal chains as volitional agent-physically a�ectedpatient event types, as all researchers on argument linking know. This is becausemany events are more complex in their causal structure than a simple linear causalchain model suggests. Also, events include noncausal relations among participantsas well as causal ones. Languages (that is, their speakers) conceptualize or construeevents in systematic ways, coercing them so to speak in order to ®t them into thesemantic patterns implied by their grammatical constructions.A good example of a more complex causal structure is found with verbs denoting

mental events, that is, states and processes involving emotion, cognition and per-ception (Croft, 1993). The two participants in this event are usually described asexperiencer and stimulus. In a mental event, there are in fact two mutually opposedcausal relations between experiencer and stimulus: the experiencer directs her atten-tion to the stimulus, and the stimulus causes a change in mental state in the experi-encer. Verbs describing mental processes highlight (that is, pro®le) either theattending-to relation (as in 6) or the alter-mental-state relation (as in 7):

W. Croft et al. / Language Sciences 23 (2001) 579±602 581

(6) a. Jack watched the children.b. Melissa thought about the problem.

(7) a. The dog frightened my daughter.b. Janet's behavior puzzled him.

Stative mental verbs do not involve a transmission of force. The linking rules donot specify a single linking pattern when there is no transmission of force (includingalso noncausal relations). Therefore the linking rules allow for either experiencer orstimulus to be linked to subject, and in fact that is what is found, both in English (asin 8) and across languages (see Croft, 1993).

(8) a. I can hear the orchestra.b. The music is barely audible to me.

The full conceptual structure of mental events involves both directions of causalrelations, attending-to and alter-mental-state, as in the diagram in (9):

(9) Experiencerÿ! ÿ Stimulus

The verbs in 6 pro®le only the upper causal relation and the verbs in 7 only the lowerrelation.One of the motivations for the model of semantic representation presented here is

the hypothesis that there is a single underlying conceptual representation for a par-ticular event class, such as the transfer of possession event class described in the nextsection. The di�erent verbs in the event class and the di�erent argument structureconstructions associated with them represent alternative construals of the sameevent (cf. Langacker, 1976, who argues for a distinction between a universal con-ceptual representation and a language-speci®c semantic representation whose con-strual is determined by the grammar of the language).The construal operations we have posited for argument linking are independently

motivated as well. They chie¯y involve the process of selection (Langacker, 1987,pp. 117±120; Croft and Wood, 2000), whereby a speaker defocuses certain aspects ofa phenomenon. Examples of selection in a di�erent domain are found in 10a±b:

(10) a. a red appleb. I sat on the chair.

In 10a, red is normally taken to describe the color of the skin of the apple, not theentire apple. That is, the interlocutors construe the denotation of apple to select onlythe skin. (In other contexts, such as solid plastic apples, the entire apple may be red.)In 10b, it is assumed that I sat on the seat of the chair, that is, the denotation ofchair is construed to select the seat of the chair. In argument linking, shifting the


pro®le of the causal chain represents a di�erent selection of the salient part of theevent by the speaker.The construal operations also can be motivated by speaker intuitions. In parti-

cular, if there is a semantic contrast between object and oblique coding of a partici-pant, object coding implies more complete or direct a�ectedness of the participant;and if there is a semantic contrast between subject and oblique coding of a partici-pant, subject coding implies more direct instigation or responsibility for the outcomeof the event (Croft, 1994). Also, encoding as a direct argument (subject or object) Ðthat is, encoding as a salient part of the event pro®le Ð results in that participantbeing more highly focused upon, as various studies of the discourse function ofpassive and antipassive voice have demonstrated (Croft, 1994). These cross-linguisticgeneralizations, along with the cross-linguistic generalizations that subject-objectorientation follows transmission of force and that oblique case markers/adpositionsgenerally subsume either only subsequent roles or only antecedent roles, provideevidence for the reality of the event structures as mental representations of speaker'sknowledge of verbs, clauses and events.

2. The representation of possession and transfer of possession

Noncausal relations, such as possession, also lack transmission of force. So inprinciple the simple predication of possession allows linking of either argument tosubject position, and in fact it does, in English (see 11a and b) and other languages(Croft, 1991, p. 210):

(11) a. I have a car.b. This car belongs to me.

In more complex event structures, there is typically a conventionalized construalof one participant as antecedent to the other participant even if the relation betweenthe participants is not causal. For example, the possessum is conventionally con-strued as antecedent to the possessor, as can be observed in the following examples(Croft, 1991, p. 207):

(12) The dean presented an award to the valedictorian.dean! award! valedictorian

(13) The dean presented the valedictorian with an award.dean! award! valedictorian

The event type underlying 12 and 13 is that of a donor causing a possession rela-tion to come about between recipient and a gift.The conceptual structure of the transfer of possession frame without the construal

of gift (possessum) as antecedent to recipient (possessor) is represented in 14:


(14)

The ditransitive construction found with verbs such as give, in which both the giftand recipient are coded as direct objects (that is, prepositionless postverbal NPs),pro®les only the causal relation between the donor and the other two participants,and does not impose an antecedent-subsequent construal on the resulting possessionrelation, as represented in 15:

(15) Tess gave Hannah a book.

On the other hand, the constructions in 12 and 13 and the construction based ondirected motion (Goldberg, 1995), illustrated in 16, represent a construal of thepossession relation as antecedent-subsequent (cf. Croft, 1991, p. 206).

(16) Tess gave a book to Hannah.

The representation in 15 indicates that the gift±recipient relation is out of pro®le(light gray), while the representation in 16 indicates that the donor±recipient relationis out of pro®le, and the noncausal gift±recipient relation is construed such that the giftis antecedent to the recipient. (The dark gray shading further indicates that this part ofthe derived causal chain is pro®led by the preposition to.) The resulting linear con-strual of the causal chain in 16 is presented in simpler form in 17:

(17) Donor! Gift! Recipient


The causal structure model, combined with the construal of complex events aslinear causal chains via selection (see above), allows us to represent the seeminglycontrary linking pattern with receive (cf. Croft, 1991, p. 252) as yet another con-strual of the transfer of possession frame:

(18) Hannah received a book from Tess.

(19) Donor! Recipient! Gift

The representation is interpretable as ``[donor acts such that] recipient comes topossess gift'' (from is an antecedent oblique).It seems reasonable to us to say that give focuses on the donor's action, while

receive focuses on the recipient's relation to the gift. These intuitions motivate thepro®ling of the relevant segments of the causal chain in 16 and 18 respectively.Moreover, the existence of a verb like receive, with recipient as subject, can bemotivated by the fact that the recipient is a volitional being who is in a position toaccept or reject the gift. If the ``recipient'' is not a volitional being, then a sentencewith receive is unacceptable, since (as noted above) initiators of transitive eventsusually have some degree of control or responsibility for the outcome of the action:

(20) a. I received a package from Bill.b. *My pigeonhole received a package from Bill.

In fact, this motivation is only moderately strong for receive; receive can be usedto describe inanimate undergoers in certain metaphorical uses, which will not bediscussed here:

(21) The building received a direct hit from the shelling.

However, the motivation for alternative subject codings is much stronger for theexchange/commercial transaction frame, since there both the buyer and the sellermust act volitionally and reciprocally for the event to successfully come about.The possible construals of the transfer of possession frame appear to be remark-

ably unconstrained. In fact, the linguistic encoding of possession and transfer ofpossession must be liberal enough to allow the di�erent argument linkings of have,belong, give (both the ditransitive and direct-motion constructions) and receive. A


too restrictive model would be simply empirically inadequate. Nevertheless, we canestablish three constraints on the coercion of a complex event structure into a linearcausal chain for argument linking:

(i) The direction of transmission of force (literal or metaphorical) must notcontradict any directionality in the conceptual structure. For example, thetransfer of possession frame in 14 implies that there is no verb of transfer ofpossession which encodes the recipient as subject and the donor as direct object.

(ii) When three or more participants are involved in an event, the causal chainderived from the conceptual structure must represent a single connected paththrough the links in the conceptual structure. There is one exception to thisconstraint, namely the ditransitive construction, since it has two direct objectsin coding terms. Its representation is in fact not linear in the analysis in 15.

(iii) The resulting causal chain should use appropriate antecedent or subsequentprepositions to link the participants to oblique positions. For example, to is asubsequent preposition and from is an antecedent preposition, a hypothesisbased on examples such as the following (Croft, 1991, pp. 194±195):

(22) The house burned to cinders.

(23) The rabbit died from thirst.

In 22, the object of to describes the resulting state of the house, and in 23, the objectof from describes the cause of the death of the rabbit. We, therefore, conclude that toshould be subsequent and from should be antecedent in the exchange/commercialtransaction frame as well.Constraints (i) and (iii) follow from the universals of argument linking given in the

last section, namely the orientation of subject and object coding with transmissionof force and the division of oblique markers into antecedent and subsequent types.Constraint (ii) follows from the hypothesis, also well supported across languages,that single verbs represent causally uni®ed events (Croft, 1990, 1994). Needless tosay, verbs encoding events in the exchange/commercial transaction frame are pre-dicted to conform to the same universals as verbs encoding events of other types.In the next section, we analyze the meaning of the preposition for, which is also

used in the exchange/commercial transaction frame, and then turn to the conceptualstructure of the frame itself and its manifestation in English, Russian and Japanese.

3. The semantics of for

Jackendo� (1990, pp. 183±184) argues that for is (at least) three ways polysemous.We largely agree with Jackendo�'s basic distinctions, but modify his analysis incertain ways.


The ®rst meaning of for we describe here is bene®ciary for. Bene®ciary for indi-cates that the event pro®led by the verb bene®cially a�ects its object. The examplesbelow are taken in part from Jackendo�:

(24) a. Bill sang a song for Mary.b. Bill sold a book to Harry for Mary.c. Bill emptied the garbage for Harry.

If the person encoded by the object of for is intended to come to have (literally ormetaphorically) the direct object of the verb, then the ditransitive construction isallowable:

(25) a. Bill sang Mary a song.b. *Bill sold Mary a book to Harry.c. *Bill emptied Harry the garbage.

Jackendo�, following Green (1974) and Oehrle (1975), argues that the ditransitiveconstruction with a bene®ciary is distinct from the argument construction with forfor two main reasons (Jackendo�, 1990, pp. 195±196). First, the ditransitive versionis restricted in the verbal event classes to which it can apply. Jackendo� describes theother di�erence as follows: in the ditransitive, `the object NP is intended for thebene®t of the bene®ciary NP. The for-bene®ciary carries no such restriction: rather,the action as a whole is intended for the bene®t of the bene®ciary NP' (Jackendo�,1990, p. 195). It is more accurate to say that in the ditransitive, the action includesintended transfer of possession. These two di�erences can be accounted for by thesemantics of the ditransitive construction. The addition of transfer of possession(intended or actual) is a semantic contribution of the ditransitive construction(Goldberg, 1995, pp. 31±39). Also, the ditransitive construction is lexically restricted(albeit partially productive Ð ibid., chapter 5): for example, it includes somesemantically idiosyncratic verbs such as envy and forgive (ibid., pp. 131±32).Bene®ciary for is clearly a subsequent preposition: the action denoted by the main

verb causes a bene®t to accrue to the object of for.The second meaning of for is purpose for, a subtype of which Jackendo� calls the

for of bene®t. Jackendo� gives the example in 26 and describes the for of bene®tsuch that `the object of for is the bene®t that the subject intends to receive by per-forming the action' (Jackendo�, 1990, p. 183).

(26) a. Bill sang a song for fun.

Jackendo� de®nes the for of bene®t as having the subject ``receive'' in some sensethe object of for (in 26, fun).Jackendo� lists a range of other uses of for which he describes as involving

intended goals or purposes (ibid., p. 184):


(27) a. Bill headed for home.b. Bill looked for Harry.c. Bill aimed for the target.d. What is this machine for?e. What do you do for a sti� neck?

In our view, the for of bene®t and the unclassi®ed examples of Jackendo�'s allrepresent the purpose or goal of the action. In 26 for example, the purpose of theaction is (to have) fun. In general, we may describe purpose for as implying theexistence of an action ultimately brought about by the action denoted by the verb,whose endpoint is the object of for. This de®nition will apply even to 27e, whichba�ed Jackendo�: the action (what you do) will result in a change of state a�ectingthe sti� neck, namely eliminating it. This is what we call purpose for.Purpose for is also a subsequent preposition type, as should be clear from the

description of its semantics in the preceding paragraph. We would predict this sincea single preposition should describe roles that are all of the same type. In fact, ben-e®ciary for and purpose for could be grouped together under a general subsequentpreposition for. More precisely, all the individual participant roles in 24, 26, and 27are of equal status under the umbrella of subsequent for. Jackendo� gives the fol-lowing example to show that there is an amibiguity between bene®ciary for and (forus) purpose for:

(28) Bill would do anything for a pretty face.

Example 28 is ambiguous between Bill doing anything to bene®t someone with apretty face vs. Bill doing anything to have a pretty face himself. In our analysis, foris indeed ambiguous, but all of the possible readings of for involve a participantsubsequent in the causal chain from the event denoted by the main verb.The last type of for that Jackendo� describes is the for of exchange, which he

illustrates with the following examples:

(29) Bill bought/rented a lawnmower from Harry for $25.

(30) Bill sold/rented a lawnmower to Harry for $25.

(31) Bill paid $25 to Harry for a lawnmower.

(32) Bill traded a lawnmower to Steve for a weedeater.

Jackendo� de®nes the for of exchange as follows: `the object of for denotes theTheme of the countertransfer' (Jackendo�, 1990, p. 183). But there are a number of


problems with this analysis. In 29 and 30, the so-called countertransfer in thecommercial transaction is the Money, whereas in 31 it is the Goods. In 30 and 32,the subject receives the item denoted by the object of for (Money or Goods), butin 29 and 31, the subject gives up the item denoted by the object of for. And in29 and 31, the subject is the Recipient, while in 30 the subject is the Donor.Although it is clear that some of the di�erences in semantic relations among theparticipants must be abstracted away from in order to analyze for in these sen-tences, we believe that one should not subsume all of these examples under a singletype of for.There is another type of for which is involved in some of the exchange/transaction

patterns given above, but is also found outside the exchange/transaction frame. Thisis what we call substitution for, which is illustrated below:

(33) The director substituted Mary for Janet in the lead role.

(34) John stood in for Mary.

(35) I feel for you.

The object of substitution for is the former or original occupier of a some desig-nated role. In 33, Janet is the original person in the lead role, now taken by Mary. In34, Mary is the occupier of the role now taked by John. Example 35 is probablyanother example of substitution for, in that I am empathetically taking on the role ofsu�erer of which you are the original occupier. It is also possible that 24c is anexample of substitution for. If it is normally Harry's role to take out the garbage,then Bill is doing it in Harry's stead. Presumably, Harry also bene®ts from Bill tak-ing out the garbage (it frees him from that onerous duty). But this is not necessaryfor substitution for: in 33, Janet may not bene®t from being replaced by Mary.Nevertheless, the overlap between being substituted for and bene®ting from thesubstitution is probably the semantic motivation for substitution and bene®t beingdescribed by the same preposition.There is also evidence that substitution for is construed as a subsequent partici-

pant role. First, in 33±35, the new occupier of the role is linked to subject (34 and 35)or object (33), both of which positions denote participants antecedent to a sub-sequent prepositional object. Second, substitute alternates with replace in the expec-ted manner:

(36) The director replaced Janet with Mary in the lead role.

In 36, the new occupier of the role is governed by the antecedent preposition with,showing that the new occupier is treated as antecedent to the former occupier (whichis in direct object position).


Substitution events involve a noncausal relation between the former and the newoccupier of the role: they are equated with each other as alternative occupiers of thesame role. In sentences like 33 and 36, there is also an agent who brings about theswitch in roles. The event structure for 33 and 36 and the conventional construal ofnew occupier as antecedent to the old occupier are given in 37 and 38 respectively:

(37)

(38) Agent!New Occupier!Former Occupier

In the next section, we use the analysis of for as purpose (including bene®ciary) forvs. substitution for to account for the patterns of argument linking found in Englishexchange and commercial transaction verbs.

4. Exchange and commercial transaction verbs in English

Exchange and commercial transaction verbs in English can be divided into twotypes. The ®rst type we will call give type verbs-the subject of the verb comes to havethe object of for:

(39) Bill sold/rented a lawnmower to Harry for $25.

(40) Bill paid $25 to Harry for a lawnmower.

(41) Bill exchanged/traded (in) his weedeater for a lawnmower.

The for in give type verbs is purpose for (in fact, it is Jackendo�'s narrower categoryof the for of bene®t): there is an action causally subsequent to the one denoted bythe main verb which has an e�ect on the object of for, namely, the subject of themain verb obtains the object of for.The second type of exchange and commercial transaction verbs are the get type

verbs-the subject of the verb comes to have the direct object of the verb:

(42) Bill bought/rented a lawnmower from Harry for $25.


(43) Bill got a lawnmower for $25.

(44) Bill got/received $25 from Harry for his lawnmower.

(45) Jacques accepted a sixpack of Red Tail Ale for mowing around our cabin.

The for in get type verbs is substitutive for: the object of for is the former occupierof the role of ``good owned by subject'' (in 45, the ``good'' is a more abstract bene-®cial value which is made up for by the donation of the beer).There is at least one verb in which there is an ambiguity between purpose for

(construed broadly to include bene®ciary for) and substitution for:

(46) Randy obtained some food for his dog.

In the purpose for reading of 46, Randy's action was intended to cause a furtheraction which would (bene®cially) a�ect his dog, namely the food will be given to thedog. In the substition for reading of 46, Randy traded his dog for some food.The analysis of exchange/commercial transaction verbs into two types, the give

type and the get type, is motivated by the polysemy of for in verbs of other eventclasses, and the ambiguity of for in 46. The division into give type and get type verbsallows us to give a consistent mapping of the participant roles into argument posi-tions for the English exchange/commercial transaction verbs.The exchange/commercial transaction frame essentially involves two transfers of

possession: a transfer of money from the buyer to the seller, and a transfer of thegoods from the seller to the buyer. (In nonmonetary exchange, it is another goodthat is being transferred from the ``buyer'' to the ``seller''.) These two transfers arerepresented independently in 47 and 48:

(47)

(48)

In addition, there is an equivalence (substitution) relation between the money andthe goods Ð the whole idea behind a commercial transaction is that the money andthe goods are equal in value (likewise for exchane). If we combine the diagrams in 47


and 48 for the buyer's transfer and the seller's transfer, and the equivalence relationbetween money and goods, we get the following conceptual structure for the com-mercial transaction frame:

(49)

All of the exchange/commercial transaction verbs can be analyzed as construals ofthe conceptual structure in 49. None of the construals violates the constraints onconstruals given above (with one possible exception; see below). We should alsoexpect that if substitution for is found (i.e. in get type verbs), it will pro®le theequivalence relation between goods and money in 49, without violating the con-straint that a single connected linear causal chain is selected by the construal. Ifpurpose for is found (i.e. in give type verbs), it will pro®le the last event in the linearcausal chain selected from the conceptual structure in 49.We will now analyze each of the main commercial transaction verbs. The argu-

ment structure construction is given in square brackets. Prepositions are given, withfor speci®ed as substitution [sub] or purpose [purp]. As before, a solid arrow indi-cates the pro®le of the main verb, dark grey the pro®le of the preposition(s), andlight grey the unselected part of the commercial transaction frame. Finally, theselected linearized causal chain is presented below each diagram. (We ignoreditransitive variants in the examples; they do not signi®cantly alter the analysis.)Our ®rst example is buy:

(50) a. Bill bought a lawnmower from Harry for $25. [= 42]b. buy: [Subject, Object, from Oblique, for Oblique]

from Seller!Buyer!Goods!for [sub] Money

Buy is a get type verb. Hence, it uses substitution for, which pro®les the equivalencerelation between the money and the goods. The Seller is found with the antecedentpreposition from (cf. examples 17, 20a and 23), as expected.Buy contrasts with pay:

(51) a. Bill paid $25 to Harry for a lawnmower. [=40]


b. pay: [Subject, Object, to Oblique, for Oblique]

Buyer!Money!to Seller!for [purp] Goods

Pay is a give type verb. Hence, it uses purpose for, which is the last segment in thecausal chain selected by the verb. The seller is found with the subsequent prepositionto (cf. examples 15 and 22), as expected.Sell is also a give type verb, and is analyzed the same way as pay, except for the

reversal of the buyer and seller:

(52) a. Bill sold a lawnmower to Harry for $25. [= 39]b. sell: [Subject, Object, to Oblique, for Oblique]

Seller!Goods!to Buyer!for [purp] Money

Get and receive allow for two mappings from participant roles to its argumentstructure. In one mapping, the seller is linked to subject position:

(53) a. Bill got/received $25 from Harry for his lawnmower.b. get/receive: [Subject, Object, from Oblique, for Oblique]

from Buyer!Seller!Money!for [sub] Goods


As get type verbs, get and receive use substitution for, which pro®les the equiva-lence relation between money and goods. The same is true of the other mapping ofparticipant roles with get/receive, where the buyer is the subject:

(54) a. Harry got a lawnmower from Bill for $25.b. get/receive: [Subject, Object, from Oblique, for Oblique]

from Seller! Buyer! Goods! for [sub] Money

We conclude this section by presenting examples where the verb pro®les theequivalence relation between money and goods. There are argument structure con-structions which allow pro®ling of just the equivalence relation:

(55) a. This lawnmower costs/is worth $25.b. Goods! Money

(56) a. $25 will buy a lawnmower.b. Money! Goods

Both patterns are allowed by the model, because there is no directed transmissionof force between money and goods. It turns out that both patterns also allow theexpression of one of the transactors. The patterns with cost and be worth can beeasily accounted for:

(57) This lawnmower is worth $25 to me. [i.e. I would sell it for $25]

Goods!Money!Seller

(58) This lawnmower will cost you $25. [i.e. you would have to pay $25 for it]


Example 58 involves the ditranstive construction, and so does not result in a simplelinear causal chain (see above).Unfortunately, adding a transactor to 56 leads to an anomaly in the representa-

tion:

(59) $25 will buy you a lawnmower. [i.e. if you had $25, you could buy it]

The ditransitive construction in 59 would lead to an analysis in which the money isantecedent to the buyer, as indicated in the diagram (the money is the subject and thebuyer is the direct object). But the commercial transaction frame in 49 requires that thebuyer be antecedent to the money, since the buyer pays the money for the goods.It it not entirely clear to us how to account for the anomaly. The construction in

59 is restricted in a number of ways, however. It is di�cult to get anything except thegeneric you as recipient, in contrast with the construction in 58 for example:

(60) ??$25 will buy John a lawnmower.

(61) The lawnmower would cost John $25.

Nor does the construction in 59 occur in another tense such as the past, again incontrast with the construction in 58:

(62) *$25 bought you/John a lawnmower.

(63) The lawnmower cost John $25.

The only plausible sentences allowing a recipient other than generic you or the pasttense use a money phrase that describes its source:

(64) a. The money donated to the charity will buy the orphans blankets.b. ?The money you gave me bought me a CD player.


It is possible that the construction in 59 represents a di�erent construal of theevent, something like `Ìf I had $25, it would cause (allow) me to obtain the lawn-mower''. In this construal, there is an enabling causal relation between (possessionof) the money and the Buyer such that the money is antecedent to the buyer. Thishypothesis is corroborated by the wider range of contexts allowed when the moneyphrase describes its source, since source is an antecedent role. Thus, our best expla-nation of 59 is that it represents a somewhat di�erent sort of situation type than theone represented by the commercial transaction frame in 49.

5. Commercial transaction verbs in Russian

Russian encodes commercial transaction verbs in essentially the same way asEnglish does. Examples of Russian counterparts to 50±54 are given below:

(65) Ja zaplatil bratu za knigu rubl0

I paid brother:dat for book:acc ruble:accÌ paid a ruble to my brother for the book.0

(66) On kupil knigu ot sestry za rubl0

he bought book:acc from sister:gen for ruble:acc`He bought the book from his sister for a ruble.'

(67) On poluchil den0gi ot fabriki za rabotuhe got money:acc from factory:gen for work:acc`He got money from the factory for his work.'

(68) Ja prodal knigu sestre za rubl0

I gave book:acc sister:dat for ruble:accÌ sold the book to my sister for a ruble.'

Russian uses the dative case without a preposition analogous to English sub-sequent to. The Russian dative without a preposition governs the objects of certainverbs and is used for the Recipient role in transfer of possession verbs:

(69) Ja dal knigu bratuI gave book:acc brother:datÌ gave the book to my brother.'

The Russian dative is also used with other Goal-type roles such as the person spo-ken to or taught (Pulkina and Zakhava-Nekrasova n.d.:pp. 81±82). On the basis ofthis evidence, we conclude that the dative case is a subsequent oblique case.


Russian uses the preposition ot+genitive case analogous to English antecedentfrom. The preposition ot is used for cause or reason:

(70) On umer ot ranyhe died from wound:gen`He died from a wound0

(71) Ona drozhala ot xolodashe trembled from cold:gen`She was trembling from/with cold.'

The preposition ot is also used for the source of motion, a common metaphoricalsource of antecedent prepositions (Croft, 1991 p. 194). Thus, ot appears to be anantecedent preposition.Russian also uses the preposition za+accusative case which appears to be analo-

gous to the English subsequent preposition for. If the Russian argument linking ofcommercial transaction verbs is essentially like the English pattern, then we wouldexpect to ®nd that za is used for both purpose/bene®ciary and substitution functions.This is in fact what we ®nd. Examples 67 and 68 are examples of purpose za andsubstitution za respectively:

(72) On srazhalsja za rodinuhe fought for country:acc`He fought for his country.0

(73) Segodnja ja rabotaju za tovarishchatoday I work for friend:acc`Today I am working for (instead of) my friend0

In sum, Russian commercial transaction verbs encode the participants with thesame event construals and argument linking patterns as English. In particular,Russian uses a preposition (za+accusative) with both purpose and substitutionmeanings in the same contexts that English uses purpose for and substitution for.

6. Commercial transaction verbs in Japanese

Japanese is rather di�erent from both English and Russian in its linking of parti-cipant roles to argument positions. Japanese is essentially a verb-®nal language, anduses case markers (usually written as separate words) postposed to NPs in order toindicate grammatical argument positions. The case markers corresponding to Eng-lish subject and object in argument linking are ga [nominative] and o [accusative]respectively. In the following examples, ga is replaced with wa [topic] in order to


render the examples slightly more natural. (Of course, one would almost never encoun-ter sentences of four-argument verbs with all four arguments expressed as full NPs.)The oblique case markers corresponding to English subsequent to and antecedent

from are ni and kara respectively. Ni behaves in many respects like a subsequent casemarker, being used for purpose, resulting state and goal roles:

(74) Inu no sampo ni ittadog gen walk for went`He took the dog for a walk.'

(75) kanojo no musume wa isha ni nattashe gen daughter top doctor to became`Her daughter became a teacher.'

(76) watashi wa kanojo ni subete no himitsu o hanashitaI top she to all gen secret:acc toldÌ told her all the secrets.'

Ni indicates direction towards a goal, which is a common metaphorical source ofsubsequent markers (Croft, 1991, p. 194). Ni also has some antecedent uses, such asthe agent of a passive (Martin 1975, p. 47); but ni is also used as a locative, which isa source of some antecedent role markers (Croft, 1991 p. 196).Kara is used to indicate source of motion, and also as a cause (Martin, 1975: 46):

(77) Kisya no tukare kara sugu nemuttatrain gen weariness from right fell.asleepÌ fell right to sleep with weariness from the train.'

Other uses of kara include material genitive, agentive passive and `because' (ibid.,45). Thus, kara appears to be an antecedent preposition.The di�erences between English and Japanese come with the case markers used

for money and goods. As we have seen, English uses substitution for and pur-pose for, both subsequent prepositions, in various constructions. Japanese usesthe instrumental case marker de and two specialized postpositional phrases: nodaikin to shite às a price for' and to hikikae ni in exchange for'. These are abbre-viated as exch in the representation of the causal chains in the following examples.Instrumental de is an antecedent case marker, as in its instrumental function:

(78) Kono naihu de keki o kittathis knife inst cake acc cut

`He cut the cake with this knife.'

It is also used for material genitive, means, means of transportation, medium of


communication, cause and agent (Martin, 1975, p. 42-43), all typical antecedentparticipant roles.We hypothesize that the two exchange postpositional phrases are also antecedent

case markers. Although they do not occur in other contexts, we will see that in oneconstruction, de can be substituted for the exchange postpositional phrases, whichsuggests that all three are antecedent. And the assumption that the exchange post-positional phrase encode antecedent roles allows for a consistent construal of thecommercial transaction frame for each verb, in conformity with the constraintsgiven above. Compared to English (and Russian), Japanese typically reverses theposition of money and goods in the linearized causal chain that results from theconstrual of the commercial transaction frame. The format here will follow the pre-sentation of the English examples. We begin with kau `buy':

(79) Harry wa Bill kara hon o $5 de kat-taHarry top Bill from book acc $5 inst buy-pst`Harry bought the book from Bill for $5.'

Seller kara! Buyer ga! Money de! Goods o

In Japanese, the money is governed by the instrumental, and so occurs antecedent tothe goods (the direct object). If we compare 79 with the representation of the Englishexample of buy in 50, we can see how the conventional construal of the relationbetween money and goods is reversed. This reversal is allowed because there is nocausal transmission of force between the money and the goods: the equivalencerelation is a noncausal, stative relation.We now turn to uru `sell':

(80) Bill wa Harry ni hon o $5 de ut-taBill top Harry dat book acc $5 inst sell-pst`Bill sold the book to Harry for $5.'

Seller ga! Money de! Goods o! Buyer ni


Again, the money is encoded by the instrumental case marker, and hence is ante-cedent to the goods (which is direct object). The buyer takes the subsequent ni, asexpected. If we compare 80 to English sell in 52, we ®nd a somewhat greater di�er-ence between the two. English sell uses purpose for, and so does not pro®le theequivalence relation between money and goods. Japanese de does pro®le theequivalence relation (albeit ``buried'' inside the pro®le of the main verb).There is a similar contrast between English `pay' and Japanese harau `pay':

(81) Harry wa hon nodaikintoshite/tohikikaeni Bill ni $5 o harat-taHarry top book asapricefor/inexchangefor Bill dat $5 acc pay-pst`Harry paid $5 to Bill for the book.'

Buyer ga! Goods exch!Money o! Seller ni

Here, we ®nd the exchange postpositional phrases instead of the instrumental casemarker. If we assume that the exchange phrases are antecedent markers, then theanalysis is unproblematic. Again, the di�erence between Japanese and English isgreater than with buy because English uses purpose for rather than substitution forwith pay.Finally, Japanese eru `obtain'/`receive' allows two mappings of participant roles

onto argument positions, like English get/receive. The ®rst mapping links the Sellerto subject position, as in 53 in English:

(82) Bill wa hon no daikin to shite/to hikikaeni Harry kara $5 o e-taBill top book as a price for/in exchange for Harry from $5 acc obtain-pst`Bill obtained $5 from Harry for the book.'

Buyer kara! Seller ga! Goods exch! Money o


If we compare the linearized construed causal chain in 82 with that for the Englishequivalent in 53, we see that the position of money and goods is once again reversedin the chain. This is expected, as English uses (subsequent) substitution for, whileJapanese uses the (antecedent) exchange phrases.In the second mapping, the Buyer is linked to subject position:

(83) Harry wa $5 de/to hikikae ni Bill kara hon o e-taHarry top $5 inst/in exchange for Bill from book acc obtain-pst`Harry obtained the book from Bill for $5.'

Seller kara! Buyer ga!Money de/exch! Goods o

This example allows either the instrumental case marker or the exchange phrase ohikikae ni. This alternation is evidence that the exchange phrases are antecedentobliques. Again, in comparison to the English construal in 54, the position of moneyand goods in the linearized causal chain in 83 is reversed, since English uses sub-stitution for.

7. Conclusion

Verbs denoting exchange and commercial transactions are problematic for mostargument linking theories because there is a wide latitude of linkings of buyer, seller,money and goods to subject, object and oblique argument positions. Most theoriestake a too-restrictive approach to argument linking and cannot account for all of thepatterns. Other theories (such as Jackendo�0s) are more liberal in their possibilitiesfor argument linking, but do not exploit the structure of their verbal semanticrepresentations in order to account for other event classes where argument linkingpatterns are indeed restricted.In this paper, we have used a model of argument linking based on transmission-

of-force relations among participants (rather than highly general thematic roles ormacroroles), combined with semantic representations of event types that use thedistinction between pro®led concept and frame and exploit the fact of semanticconceptualization (construal). This model provides for the restricted argumentlinking possibilities found with other verb classes (see Croft, 1991, 1998 for exam-ples). But it also allows for the wide range of linking possibilities for the exchange/commercial transaction frame, because the commercial transaction frame involves


a complex network of causal and noncausal relations among buyer, seller, moneyand goods. Yet the model of the commercial transaction frame allows us to cap-ture the systematic character of the argument linking patterns found in English,Russian and Japanese, and the similarities and di�erences in those patterns acrossthe three languages.

References

Croft, W., 1990. Possible verbs and event structure. In: Tsohatzidis, S.L. (Ed.), Meanings and Prototypes:

Studies on Linguistic Categorization. Routledge, London, pp. 48±73.

Croft, W., 1991. Syntactic Categories and Grammatical Relations: The Cognitive Organization of Infor-

mation. University of Chicago Press, Chicago.

Croft, W., 1993. Case marking and the semantics of mental verbs. In: Pustejovsky, J. (Ed.), Semantics and

the Lexicon. Kluwer Academic, Dordrecht, pp. 55±72.

Croft, W., 1994. Voice:beyond control and a�ectedness.. In: Hopper, P., Fox, B (Eds.), Voice:Form and

Function. John Benjamins, Amsterdam, pp. 89±117.

Croft, W., 1998. Event structure in argument linking. In: Butt, M., Geuder, W. (Eds.), The Projection of

Arguments:Lexical and Compositional Factors. Center for the Study of Language and Information,

Stanford, pp. 1±43.

Croft, W., Wood, E.J., 2000. Construal operations in linguistics and arti®cial intelligence. In: Albertazzi,

L. (Ed.), Meaning and Cognition. John Benjamins. Amsterdam.

Fillmore, C.J., 1982. Frame semantics. In: The Linguistic Society of Korea. Linguistics in the Morning

Calm. Hanshin, Seoul, pp. 111±137.

Fillmore, C.J., 1985. Frames and the semantics of understanding. Quaderni di Semantica 6, 222±254.

Goldberg, A.E., 1995. Constructions: A Construction Grammar Approach to Argument Structure. Uni-

versity of Chicago Press, Chicago.

Green, G., 1974. Semantics and Syntactic Regularity. Indiana University Press, Bloomington.

Haiman, J., 1983. Iconic and economic motivation. Language 59, 781±819.

Jackendo�, R., 1990. Semantic Structures. MIT Press, Cambridge, Mass.

Langacker, R.W., 1976. Semantic representations and the linguistic relativity hypothesis. Foundations of

Language 14, 307±357.

Langacker, R.W., 1987. Foundations of Cognitive Grammar, Vol I: Theoretical Prerequisites. Stanford

University Press, Stanford.

Martin, S.E., 1975. A Reference Grammar of Japanese. Yale University Press, New Haven and London.

Oehrle, R. 1975. The Grammatical Status of the English Dative Alternation. PhD dissertation, Massa-

chusetts Institute of Technology.

Pulkina, I., Zaxava-Nekrasova, E. (n.d.) Russian. Progress, Moscow.

Talmy, L., 1976. Semantic causative types. In: Shibatani, Masayoshi. (Ed.), The Grammar of Causative

Constructions. Syntax and Semantics, Vol. 6.1. Academic Press, New York, pp. 43±116.

Talmy, L., 1988. Force dynamics in language and cognition. Cognitive Science 12, 49±100.


argument linking and the commercial transaction frame in english, russian and japanese

Documents