Argument structure and event structure: the case of Korean imperfective constructions

Abstract This paper uncovers a systematic correlation between semantics ofaspect and syntactic argument structure as manifested in the difference betweentwo imperfective aspect markers -ko iss and -a iss in Korean. Unlike thecommon assumption that the -ko iss form is a progressive marker, while the-a iss form is a resultative marker, this paper argues that the difference betweenthe two derives from their different argument structure: -ko iss selects transitiveand unergative verbs, which have an external argument, while -a iss selectsunaccusative and passive verbs, which only have an internal theme argument. Itis argued that the difference in argument structure is determined by semanticevent structure depending on agentivity in Korean. The results of the paperhave broader implications for the issues of syntax and semantics interface andunaccusativity.

Keywords Korean imperfective aspect Æ Argument structure Æ Agentivityversus dynamicity Æ Unaccusativity Æ Syntax/semantics interface

1 Introduction

This paper aims to uncover a systematic correlation between semantics ofaspect and syntactic argument structure as manifested in the difference betweentwo imperfective aspect markers in Korean. The Korean tense and aspectsystem includes two imperfective periphrastic constructions: -ko iss and -a/e iss.

E. H. Lee (&)Department of Linguistics, University at Buffalo, The State University of New York,623 Baldy Hall, Buffalo, NY 14260-1030, USAe-mail: [email protected]

123

J East Asian Linguist (2008) 17:117–139DOI 10.1007/s10831-008-9022-z

Argument structure and event structure:

the case of Korean imperfective constructions

EunHee Lee

Received: 24 July 2006 /Accepted: 30 January 2008 / Published online: 16 April 2008� Springer Science+Business Media B.V. 2008

-A iss is often compared with -ko iss because these two forms contain thesame auxiliary verb iss ‘be, exist’ and differ only with respect to the connectives,-a/e ‘have -en’ and -ko ‘and’, which are attached to the main verb. It is com-monly assumed in the literature that the -ko iss form denotes actions inprogress, i.e., progressive (Choe 1971; Lee 1993; Kim 1986; Martin 1992), andthe -a/e iss form denotes resultant states, i.e., resultative (Lee 1993; Martin1992; Ahn 1995). (1) contains contrasting examples of the two durative(imperfective) markers (Lee 1991).

(1)a. Haksayng-tul-i motwu uica-ey anc-ko iss-ta.student-PL-NM all chair-LC sit-CN exist-DC1

‘The students are all taking a seat.’b. Haksayng-tul-i motwu uica-ey anc-a iss-ta.2

student-PL-NM all chair-LC sit-CN exist-DC‘The students are all in their seats.’

When complemented by the same predicate anc- ‘to sit down’, -ko iss in (1a)expresses an action-in-progress, i.e., the action of taking a seat, while -a iss in(1b) expresses the state of being in one’s seat, which resulted from the action ofsitting down. Although many Korean linguists claim that -a iss denotes theperfective resultative, in line with Lee (1991) and Shirai (1998), I take both-ko iss and -a iss to express durative situations which encode the imperfectiveaspect. In other words, the main assertion of (1b) is the continuous state of thestudents being in their seat, viz., the completion of a previous event of sittingdown is not the focus.

The descriptive dichotomy of process versus result state, however, is prob-lematic since -ko iss, like -a iss, can refer to a resultant state when it is affixed tocertain telic transitive verbs, i.e., verbs of wearing and body postures. (2) isambiguous between the two readings ‘He is now in the process of putting onclothes’, or ‘He has already put on clothes and is now in the state of wearingthem’ (Lee 1991; Kim 1986, 1993; Ahn 1995). Therefore, the semantic feature‘dynamic’ cannot serve as a distinguishing factor between the two. Given thatthe progressive describes an on-going process, without entailing that thedescribed event has ended, the result state reading in (2) is unexpected. -A isscannot be used in (2) because it exclusively takes intransitive verbs. As I willargue later, this complementation restriction does not follow from the analysisof -a iss as the result state marker, and therefore calls for an explanation.

1 The abbreviations used in the glosses are as follows: NM: nominative case marker, AC: accusativecase marker, PL: plural marker, CN: connectives, DC: declarative sentence ending, MP: imperativesentence ending, PST: past tense marker, PRS: present tense marker, PS: passive, CS: causative,LC: locative marker, TP: topic marker.2 Unlike the English perfect, (1b) does not have an experiential reading but instead describes anon-going current state. The variation between a and e in the -a/e iss construction is phonologicallydetermined and depends on the last vowel of the verb stem to which it is attached. I will represent ithereafter as -a iss.

118 E. H. Lee

123

(2) Ku-ka os-ul ip-ko iss-ta.he-NM clothes-AC wear-CN exist-DC‘He is getting dressed.’ or ‘He is wearing clothes.’

Faced with this difficulty, Lee (1993) proposes that the result states described by-ko iss are consciously and volitionally initiated and maintained by an agentwhereas those described by -a iss are not. Ahn (1995) claims that the distinctionbetween -ko iss and -a iss lies in highly transitive versus intransitive situations, a laHopper and Thomson (1980). Lee (1991) argues that -ko iss is a dynamic durativeand -a iss is a static durative, and that the result states expressed with -ko iss aredynamic in that they indicate a temporary state that inherently assumes anupcoming change of state. Agentivity, stativity, and transitivity all have beenproposed as explanatory features that play a role in distinguishing -ko iss and-a iss, and one of the main concerns of this paper is to determine which feature isrelevant. Although I take agentivity to be the distinguishing property between thetwo forms, in this paper I will argue against the view that it necessarily involvesconscious or volitional agents.

It is interesting to observe that a typologically similar language, Japanese, hasa single form -te i for the meaning covered by both -ko iss and -a iss. Thedistinction between progressive -te i and resultative -te i is also controversialamong Japanese scholars. Some scholars claim that it is an aspectual distinction,that is, -te i triggers a progressive meaning when complemented by durative verbs(activities and accomplishments) while it gives rise to a result state meaning whencomplemented by punctual verbs (achievements) (Kindaichi 1950). Others haveargued that the difference is syntactic, rather than aspectual (Kudo 1995; Kinsui1995): transitive verbs in the -te i form have an on-going action reading whileintransitive verbs in the -te i form have a resultative meaning.3 Shirai (1998, 2000)proposes that, since transitive verbs describe inherently dynamic situationsbecause an action is initiated by the agent and transferred to and ended at thepatient (Smith 1991), aspectual analysis based on duration and syntacticapproach based on (in)transitivity are not necessarily exclusive of each other. Infact, many other researchers have also pointed out that there is a significantcorrelation between aspectual properties of duration and syntactic phenomena:a strong association exists between punctuality (achievements) and intransitivechange-of-state, on the one hand, and between duration (accomplishments) and

3 The duration analysis cannot explain the correlation between transitive verbs and progressivemeaning on the one hand, and intransitive verbs and resultative meaning on the other. For example,even though the situations described are the same, the transitive verb akeru ‘open’ yields progressivesense with -te i- whereas its intransitive counterpart aku ‘be open’ yields resultative sense (Ogihara1998, p. 94).

(i)a. John-wa doa-o ake-te iru.John-Top door-Acc open-TE IRU-Pres�John is opening the door.�

b. Doa-wa ai-te iru.Door-Top open-TE IRU-Pres�The door is open.�

Argument structure and event structure 119

123

transitive change-of-state, on the other (Shirai 1998, 2000; Van Valin 1990, 1993;Dowty 1991; among others).

In this paper, I will argue that the difference between -ko iss and -a iss derivespartly from their different syntactic argument structure rather than fromsemantic aspect alone: -ko iss selects transitive and unergative verbs, which havean external argument, while -a iss selects unaccusative and passive verbs, whichhave no (or suppressed) external argument and only have an internal themeargument. Therefore, it is argued that the distinction between -ko iss and -a iss isa reflex of unaccusativity in Korean and that agentivity is the semantic basis forit (Kim 1990), with a revised notion of agents. I will also discuss the semanticcorrelation between agentivity and internal temporal structure of events. Due tothe fact that -ko iss selects verbs with an external agent argument, which is thefirst argument in an activity predicate (do), it typically expresses a middle phaseof a temporary situation, as argued by Lee (1991, 1999) and Lee (2004, 2006).This paper builds on these previous analyses by making the link between thesemantics of aspect and argument structure more transparent. I will present aformal semantic analysis of the markers, providing a more precise character-ization of their semantic contributions to the sentences in which they occur.

This paper is organized as follows. In Sect. 2, I show that the aspectualdistinction of progressive and result state is not appropriate for defining thenature of -ko iss and -a iss. I then propose that the difference between the twoderives from variations in their argument structures. In Sect. 3, I discuss thesemantic basis of unaccusativity in Korean. In Sect. 4, I provide a semanticanalysis of the two constructions. Section 5 summarizes the main points of thispaper and discusses their implications.

2 The syntactic and semantic differences between -ko iss and -a iss

2.1 Problems with the progressive-versus-resultative analysis

There are many cases in which -ko iss and -a iss cannot be distinguished interms of the aspectual distinction of progressive versus result state markers.First, the following examples show that the -a iss form cannot be exclusivelycharacterized as expressing result states (Lee 1991).

(3)a. Ku-nun acik sal-a iss-ta.4

he-TP still live-CN exist-DC‘He is still alive.’

b. Hankwuke Program-un enehak kwa-ey sokhay-e iss-ta.Korean Program-TP linguistics department-LC belong-CN exist-DC‘The Korean Program belongs to the Linguistics Department.’

4 Sal- is glossed as ‘live’ in this paper for consistency. When it occurs with -ko iss, it means ‘to live(in a place)’ or ‘make a living’ while it means ‘to be alive’ when it occurs with -a iss.

120 E. H. Lee

123

Examples in (3) do not express result states but simply describe the durativestates of one’s being alive and a Korean Program’s being part of the LinguisticsDepartment. They cannot be construed as result states because the durativestates that are described cannot be understood as the end-result of any actions.Furthermore, in order to qualify as a ‘‘result’’ state, the main verb must have aclear culmination, after which the result state obtains; however, the verbs in (3),sal-ta ‘to live’ and sokha-ta ‘to belong to’, are not telic verbs, refuting the claimof many scholars that -a iss exclusively takes telic intransitive verbs (Lee 1993;Kim 1990; Ahn 1995). For example, they cannot be modified by maney‘(with)in’ or -nun tey (time) kellita ‘it takes (time) to’.

Second, as we have already observed in (2) above, which is repeated as (4a),expressing result states does not distinguish between the -a iss form and the -ko issform because there are cases in which the latter may describe a result state as well.

(4)a. Ku-ka os-ul ip-ko iss-ta.he-NM clothes-AC wear-CN exist-DC‘He is getting dressed.’ or ‘He is wearing clothes.’

b. Ku-ka cacenke-lul tha-ko iss-ta.he-NM bike-AC ride-CN exist-DC‘He is getting on a bike.’ or ‘He is on a bike.’

The second readings of (4) with the -ko iss form do not express progressivenessin a typical sense but rather appear to describe result states.

Third, unlike in English where stative verbs such as know, love, believe, have,etc., are not normally used with the progressive -ing, the corresponding verbs inKorean naturally occur with the -ko iss form.

(5) Ku-ka ku sasil-ul al-ko iss-ta.he-NM the fact-AC know-CN exist-DC‘He is knowing – now aware of – the fact.’

(5) describes the mental state of the subject as opposed to any action-in-progress. If -ko iss is a progressive construction, then further explanation isneeded to elucidate in what sense the above examples can be considered asexpressing progressiveness. The fact that -ko iss is compatible with cognitive/emotive verbs does not mean that it is a general imperfective marker althoughsome scholars (Ahn 1995) make this claim. Unlike imperfectives in Romanceand Slavic languages, -ko iss cannot occur with stative predicates in general,and its use is not restricted to the past tense.

Fourth, while most action verbs, including all transitive verbs, cannot beused with the -a iss form, a limited set of action verbs, e.g., locomotion andposture verbs, can be used with it.

(6)a. Ku-nun uyca-ey anc-a iss-ta.he-TP chair-LC sit-CN exist-DC‘He is sitting on a chair.’


123

b. Ku-nun hankwuk-ey ka-a iss-ta.he-TP Korea-LC go-CN exist-DC‘He is (staying) in Korea.’

Each of the sentences in (6) describe a durative state of the subject, either in hisposture, as in (6a), or in his location, as in (6b). It is puzzling why the actionverbs in (4) are used with -ko iss to express the result state while the action verbsin (6) are used with -a iss. Note that verbs in (4) are transitive verbs while thosein (6) are intransitive verbs. Since transitive verbs cannot be used with -a iss,it cannot occur in examples like (4), as shown in (7a), while -ko iss has anon-going action interpretation with verbs in (6), as in (7b).

(7)a. *Ku-nun os-ul ip-e iss-ta.he-TP clothes-AC wear-CN exist-DC[Intended] ‘He is wearing clothes.’ (As a result of putting on clothes.)

b. Ku-nun uyca-ey (chenchenhi) anc-ko iss-ta.he-TP chair-LC (slowly) sit-CN exist-DC‘He is taking a seat on a chair slowly.’ (no futurate meaning)

If the observed complementation restriction imposed on the use of -a iss issimply due to its semantics of referring to a result state, there is no logicalreason why it cannot take telic transitive verbs to describe the result state, of acompleted action, e.g., an object coming into being in the case of creation verbs.Note that the transitive verbs given in (4), ip-ta ‘to put on’ or ‘to wear’ andtha-ta ‘to ride’ or ‘to get on’ do express result states, but they do so with -ko iss,not with -a iss. If the -a iss construction is characterized as expressing resultstates, the fact that transitive verbs, even telic ones, cannot be used with -a issneeds to be accounted for.

2.2 Difference in argument structure of the two forms and unaccusativityin Korean

All the puzzles we observed in the previous section can be resolved by lookingat the two constructions from a different perspective, namely in terms ofargument structure. I argue that -ko iss takes verbs which have an externalargument while -a iss selects verbs which have only an internal argument, thatis, unaccusative verbs. The Unaccusativity Hypothesis, as first proposed byPerlmutter (1978) within the context of Relational Grammar and later adoptedby Burzio (1986) within the Government-Binding framework (Chomsky 1981),is a syntactic hypothesis that claims that there are two types of intransitiveverbs, the ‘‘unergative’’ verbs and the ‘‘unaccusative’’ verbs, each associatedwith a different underlying syntactic configuration. The hypothesis states thatunaccusative verbs have an underlying object but lack an underlying subject,while unergative verbs have an underlying subject and no underlying object.In argument-structural terms, the unaccusative predicates are ones which

122 E. H. Lee

123

originally lack an external argument in their argument structure representationwhereas the unergatives have an external argument but no direct internalargument. As reflected in the name given to the class, an unaccusative verb isunable to assign structural case to its object. Examples of typical unergativeverbs are walk, run, laugh, and work; unaccusative verbs typically are ‘‘change-of-state’’ verbs such as melt, die, and arrive.

It has been pointed out that unaccusatives pattern with passives and unerga-tives pattern with transitive verbs in auxiliary selection. For example, in Italian,transitive clauses and unergative clauses select avere ‘have’ whereas unaccusativeclauses and passive clauses take essere ‘be’ in perfect tenses (Burzio 1986).

(8)a. Giovanni ha telefonatohas telephoned

‘Giovanni has telephoned.’b. Giovanni e arrivato

is arrived‘Giovanni has arrived.’

It is interesting to notice that Korean exhibits a similar contrast. First, -a issis exclusively used with intransitive verbs with non-agentive subject. (9) containsmore examples of -a iss sentences.

(9)a. Ku-ka cwuk-e iss-ta.he-NM die-CN exist-DC‘He is dead.’

b. Kwuk-i sik-e iss-ess-ta.soup-NM get cold-CN exist-PST-DC‘The soup was (has become) cold.’

c. Elum-i el-e iss-ta.ice-NM freeze-CN exist-DC‘The ice is (has) frozen.’

d. Ku-ka pang-ey tuleka-a iss-ta.he-NM room-LC enter-CN exist-DC‘He is in the room.’ (He has entered the room)

The verbs which can complement -a iss have the [)agentive] feature. Theseverbs are unaccusatives that have only an internal argument in the subjectposition. The only exceptions are some ambiguous cases like sit, stand, lie, go,come, and arrive, which we observed in (1b) and (6) above. I will explain theseexceptional cases later.

Second, although -a iss takes only intransitive verbs, some intransitive verbs,i.e., unergatives, cannot occur with it, as shown in (10).

(10) *Ku-ka talli-e iss-ta.he-NM run-CN exist-DC[Intended] ‘He has run.’


123

Most scholars assume that (10) is not allowed because -a iss, due to itssemantics of referring to resultant states, takes only telic verbs whose situa-tions have an inherent endpoint and bring about a salient result state (Lee1993; Kim 1990; Ahn 1995). As we have observed in (3), however, thisstatement is not valid because nam-ta ‘remain’, sal-ta ‘live’, sik-ta ‘cool’, andsokha-ta ‘belong to’ are not telic verbs but still accept the -a iss form (Lee1991). Therefore, the ungrammaticality of (10) cannot be traced to a violationof the telicity requirement; rather, it is due to the fact that activities—alsoreferred to as unergatives—only have an external agent argument, and -a issrequires unaccusative verbs with a sole internal argument, which is the themesubject.

Third, -a iss cannot occur with transitive verbs, as in (11).

(11) *Ku-ka cip-ul ci-e iss-ta.he-NM house-AC build-CN exist-DC[Intended] ‘He has built a house.’ (He is in the state of having builta house.)

The transitive verbs are accusative-case assigners and thus are not compatiblewith the -a iss form. In contrast, -ko iss is compatible with the unergative andtransitive verbs, describing an action-in-progress.

(12)a. Ku-ka cip-ul cis-ko iss-ta.he-NM house-AC build-CN exist-DC‘He is building a house.’

b. Ku-ka talli-ko iss-ta.he-NM run-CN exist-DC‘He is running.’

Fourth, -a iss can select passive verbs. As we have observed in (11), which isrepeated in (13a), -a iss never occurs with transitive verbs. However, if thetransitive verb is turned into its passive form, as in (13b), -a iss happily occurswith it.

(13)a. *Ku-ka cip-ul ci-e iss-ta.he-NM house-AC build-CN exist-DC[Intended] ‘He has built a house.’

b. Cip-i ci-e ci-e iss-ta.house-NM build-PS-CN exist-DC‘A house is built.’

(14) shows that lexical passive verbs also accept the -a iss form. In lexicalpassives, passive suffixes -i, -hi, -li, or -ki, are attached directly to the verb stem.These four suffixes are phonologically conditioned allomorphs, and they arederivational, not inflectional (Sohn 1999).

124 E. H. Lee

123

(14)a.*Kyengchal-i totwuk-ul cap-a iss-ta.police-NM thief-AC catch-CN exist-DC[Intended] ‘The thief has caught the thief.’

b. Totwuk-i kyengchal-eykey cap-hi-e iss-ta.thief-NM police-by catch-PS-CN exist-DC‘The thief is caught by the police.’

The acceptability of (13b) and (14b) follows from the fact that passives areadicity reducers, and -a iss occurs solely with one-place predicates selecting onlya theme argument. Furthermore, in passive sentences, the agent role is sup-pressed, and the theme argument becomes the topic (subject) of the sentence.The fact that the -a iss form does not pick out transitive verbs and unergativeverbs while it selects [)agentive] verbs, including passives, suggests that com-patibility with the aspectual form -a iss is sensitive to unaccusativity in Korean.On the other hand, -ko iss selects transitive verbs and unergative verbs, whichhave an external argument.

The new analysis being proposed here explains the problematic cases weobserved in the previous section. First, the verbs in (3), sal-ta ‘to live’ andsokha-ta ‘to belong to’ are unaccusative verbs and thus can complement -a iss.If we assume that -a iss refers to result states, the examples in (3) cannot beexplained because they do not denote any result state caused by a precedingaction. Second, examples in (2) and (4) are also naturally explained. The verbsip-ta ‘to put on’ or ‘to wear’ and tha-ta ‘to get on’ or ‘to ride’ assign a structuralaccusative case to their internal arguments, meaning they are not unaccusativesbut typical transitive verbs.5 If we simply assume that -ko iss expresses anaction-in-progress, then we cannot account for why -ko iss, when combinedwith these verbs, can refer to result states. Third, examples such as (5), in whichstative verbs occur with the -ko iss form, are also naturally explained in myanalysis. Note that these psych-verbs assign structural case to their object, andthe experiencer subject is obviously an external argument. I will discuss thepsych-verbs in more detail in the next section. Finally, given my analysis, thefact that -a iss cannot take telic transitive verbs in order to express a result stateis not curious anymore: since the -a iss form is only compatible with unaccu-sative verbs, it can never occur with transitive verbs.

As noted in the introduction, a different aspect-based account, one whichdoes not make an explicit reference to argument structure, has been proposedby Lee (1991, 1999). He argues that -ko iss is a dynamic durative and -a iss is

5 An anonymous reviewer of JEAL pointed out that -a iss co-occurs with the accusative case in (i)below.

(i) Chang-i tong-ccok-ul hyangha-y iss-ta.window-NM east-side-AC face-CN exist-DC�The window is facing toward the east side.�

In (i), the -lul marker does not seem to mark the internal theme argument selected by the main verband therefore can be replaced by -ulo ‘towards’. In Korean, the accusative case marker -(l)ul cansometimes be attached to non-arguments of the verbs such as temporal and measure terms, whichare adjunct (twu sikan-ul/tongan ttuy-ess-ta. ‘I ran for two hours’).


123

a static durative, and that the result states expressed with -ko iss are dynamic inthat they indicate a temporary state that inherently assumes an upcoming changeof state. According to his analysis, the -a iss construction with posture verbs, e.g.,nwuw-e iss-ta ‘is lying’, focuses on the end results and their persistence and henceis static whereas the -ko iss construction with verbs of wearing, e.g., ip-ko iss-ta ‘iswearing’, concentrates on ongoing states and their temporariness. He furtherargues that the resultant state interpretation in the -ko iss construction with verbsof wearing comes from the lexical properties of these verbs, that is, actions des-ignated by these verbs result in the contact between the agent and the patient (Kim1993). However, in both ip-ko iss-ta ‘is wearing’ and nwuw-e iss-ta ‘is lying’,contact between the agent and the patient exists. Moreover, because -ko iss onlypragmatically implicates that the described eventuality will end, without entailingan actual termination, the assertion that ip-ko iss-ta is temporary, while nwuw-eiss-ta is not, is less persuasive than his other arguments. Nwuw-e iss-ta equallyinvites such an implicature, e.g., one does not plan to lie in bed forever. So while Iagree with Lee (1991) that -ko iss and -a iss should be contrastively interpreted, thedichotomy of dynamic duratives versus stative duratives, especially because hisnotion of ‘‘dynamicity’’ crucially relies on temporariness, should be supplanted bythe argument structure that is put forward in this paper. Based on the discussionof this section, a rather clear generalization regarding the occurrence of -ko iss and-a iss emerges in terms of syntactic argument structure. Compared to the aspectualanalysis of progressive versus resultative, this new analysis provides a morecomprehensive explanation that accounts for a wider range of data. In the nextsection, I will discuss the semantic basis of unaccusativity in Korean and itscorrelation with aspectual phenomenon and event structure.

3 Semantic basis of unaccusativity in Korean

The issue of whether there is a semantic basis for unaccusativity, and whetherwe can tell whether a certain intransitive verb is unaccusative or unergative onthe basis of meaning, has been much discussed. Following Perlmutter (1978),Perlmutter and Postal (1984), and Levin and Rappaport (1995), I assume thatthe distinction between the two classes of verbs is syntactically encoded, but it isfully semantically determined by event structure.6 Kim (1990, pp. 61–89) pro-vides a detailed and illuminating discussion concerning the semantic basis of

6 There are debates between those who advocate syntactic versus semantic approaches of unaccus-ativity. On the syntactic approach (Rosen 1984), all that unaccusative verbs have in common is aparticular syntactic configuration, and there is no single semantic property common to allunaccusative verbs selected by all diagnostics in all languages. This is because verbs with similarmeanings in and across languages may be classified differently with respect to unaccusativity. Fur-thermore, there are individual verbs that appear to be classified both unaccusative and unergative bythe same diagnostic. The semantic approach (Van Valin 1990), on the other hand, assumes thatunaccusativity is semantically determined but denies that it is syntactically encoded. Most unaccu-sative diagnostics do not single out the sole argument of all unaccusative verbs and the underlyingobject of all passive verbs. Different syntactic diagnostics which supposedly distinguish unaccusativefrom unergative verbs within a single language actually draw the boundary in different places.

126 E. H. Lee

123

unaccusativity in Korean. Let me summarize her main points because I rely onher argument that agentivity, rather than stativity, is the semantic basis forunaccusativity in Korean although I do not adopt her notion of agents and willinstead propose a revised notion.

If we assume Burzio’s (1986) generalization that verbs assign accusative caseif and only if they have an external argument, we would expect to see a dif-ference in case marking patterns between unaccusatives and the verbs that haveexternal arguments. Korean has a class of psych-verbs that show the same case-marking patterns as regular transitive verbs in Korean, as illustrated in (15)below. As we have seen in (5), these verbs are compatible with the -ko issconstruction. As expected, they cannot occur with -a iss.

(15)a. Na-nun umak-ul coh-a ha-n-ta.I-TP music-AC be.likable-do-PRS-DC‘I like music.’

b. Na-nun apeci-lul musew-e ha-n-ta.I-TP father-AC be.scary-do-PRS-DC‘I fear my father.’

Just as in the case of regular transitive verbs, the first NP, which is the exper-iencer, gets nominative marking whereas the theme NP is marked accusative. Ifwe have an unaccusative verb in place of a transitive verb, we would not expectto see the same case-marking pattern. Interestingly, there are classes of predi-cates which take two arguments just like regular transitives but which manifesta different case pattern from the transitive psych-verbs like those in (16).Neither -a iss nor -ko iss co-occurs with these adjectival psych-verbs.

(16)a. Nay-ka/eykey umak-i coh-ta.I-NM/DT music-NM be.likable-DC‘I am fond of music.’

b. Nay-ka/eykey apeci-ka musep-ta.I-NM/DT father-NM be.scary-DC‘I am afraid of my father.’

(15) and (16) are minimal pairs that differ only with respect to the verbal suffixand the case-marking patterns.7 The most immediately observable differencebetween (15a, b) and (16a, b) besides the case-marking pattern is that in (16) theverbs are [+stative] whereas in (15) verbs are [)stative]. Kim (1990) also pointsout that the two are different in terms of agentivity. The agentive property of

7 The transitive psych-verbs are formed by attaching -e ha ‘to do’ to the adjectival psych-verb stem.A semantic difference between the adjectival psych-verbs and the transitive psych-verbs is that theformer describes the inert, internal emotional state of the experiencer whereas the latter implies thatthe experience shows typical (behavioral) signs of a person in the emotional state denoted by theverbal stem (Kim 1990; Sohn 1999). Therefore, the adjectival psych-verbs are awkward with thirdperson subject since it is difficult, if not impossible, to know the internal emotional state of a thirdperson unless she expresses it.


123

the transitive psych-verbs compared with the adjectival psych-verbs is attestedby the fact that they pass the agentivity tests.

(17)a. Ku-uy cwukum-ul sulphu-e ha-la.he-PS death-AC be.grieved-do-MP‘Grieve over his death.’

b. *Ku-uy cwukum-i sulphu-ela.he-PS death-NM be.grieved-do-MP‘Be grieved at his death.’

(18)a. Ku-ka na-eykey Yumi-lul miw-e ha-tolok kangyohay-ss-ta.he-NM I-DT Yumi-AC be.hateful-do-CP force-PST-DC‘He forced me to hate Yumi.’

b. *Ku-ka na-eykey Yumi-ka mip-tolok kangyohay-ss-ta.he-NM I-DT Yumi-NM be.hateful-CP force-PST-DC‘He forced me to be hateful of Yumi.’

The fact that the transitive psych-verbs, but not the adjectival psych-verbs, mayoccur in imperatives and may be embedded under agentive control verbsindicates that only the former has the [+agentive] property. The transitivepsych-verbs, which are [)stative,+agentive], map their experience arguments asexternal arguments whereas the adjectival psych-verbs, which have the features[+stative, )agentive], map their experiencer arguments as internal arguments.

The crucial question is which of these features, [)stative] or [+agentive], isresponsible for the verb’s ability to have an external argument and to assignstructural case. Just by looking at psych-verbs, it is difficult to answer thisquestion because the adjectival psych-verbs, which do not have an externalargument and do not assign accusative case, always differ from transitive psych-verbs in both features. Kim (1990) provides a sentence pattern in Korean inwhich these two features can be distinguished. Certain classes of dyadic verbsdo not assign accusative case to their arguments, and they differ from transitiveverbs with respect to only one feature. Examples in (19) are from Kang (1987).

(19)a. Ku-ka/eykey ton-i sayngki-ess-ta.he-NM/DT money-NM come.to.exist-PST-DC

‘He came to have money.’b. Tomato-ka/ey pelley-ka manhi kki-n-ta.

tomato-NM/DT bug-NM a lot gather-PRS-DC‘Tomatoes gather a lot of bugs.’

c. Pucascip-i/ey totwuk-i cal tu-n-ta.rich house-NM/DT thief-NM well enter-PRS-DC‘(To) rich houses thieves often enter.’

d. I hengkeph-i/ey mwul-i an sumyetu-n-ta.this fabric-NM/DT water-NM NG penetrate-PRS-DC‘(Into) this fabric, water does not penetrate.’

128 E. H. Lee

123

The dyadic existential-possessive verbs and process (change-of-state) verbsexemplified in (19) do not assign accusative case to their arguments and insteadexhibit the same case-marking pattern as do the adjectival psych-verbs. It isimportant to note that verbs in (19), unlike adjectival psych-verbs, have thefeature [)stative]. These verbs share only the feature [)agentive] with theadjectival psych-verbs, as shown in (20).

(20)a. *Ton-i sayngki-ela.money-NM come to exist-MP‘Come to have money.’

b. *Ku-eykey ton-i sayngki-tolok kangyohay-ss-ta.he-DT money-NM come to exist-CP force-PST-DC‘I forced him to come to have money.’

Therefore, it is the [+agentive] feature rather than the [)stative] feature of averb that is responsible for its projection of an external argument and for itsaccusative case assigning ability.

Kim (1990, pp. 90–92) acknowledges the close matching of the verbs pickedout by -a iss and verbs with [)agentive] features, i.e., unaccusatives, but sheargues that unaccusativity is not a sufficient condition for the -a iss selection,pointing instead to telicity as the determining factor. However, as we havealready observed, -a iss can be used with verbs like sokha-ta ‘to belong to’,which is not telic.8 Kim (1990) raises two objections towards the analysis of-a iss as a reflex of unaccusativity in Korean. First, she points out that [+stative]adjectives, even if they are [)agentive] and therefore unaccusatives, are notcompatible with the -a iss form.

(21) *Ku-ka chakha-y iss-ta.he-NM be.nice-CN exist-DC[Intended] ‘He has been nice.’

This conclusion is based on her rather non-standard categorization of verbs inKorean. She claims that verbs and adjectives are the same category, differingonly in their [stativity], but contradictorily claims that some verbs are stative.Although adjectives conjugate like verbs, there is still a clear morphosyntacticdifference between verbs and adjectives in Korean. For example, verbs take thepresent tense marker -(nu)n while adjectives takes zero in the present.

8 Kim (1990) attributes these exceptions to idiosyncratic lexical meaning. For example, nam-ta‘to remain’ connotes that the expectation was that all is gone, and the state of affairs described byremain is set against such an expectation; sokha-ta ‘to belong to’ connotes that the unmarkedexpectation is no membership, i.e., ‘‘free floating of things in the universe’’; and concayha-ta‘to exist’ suggests that the unmarked state of affairs is nonexistence. However, it is difficult tounderstand why such expectations are unmarked or how such expectations about an opposite stateprovide a boundary for the verbs in question. Even if we concede that these verbs have suchpragmatic connotations, that does not seem to make them express the results of some action.


123

(22)a. Kolay-nun phoyulyu-ey sokha-n-ta.whale-TP mammal-LC belong to-PRS-DC

(stative verb sokha- ‘belong to’)‘Whales belong to mammals.’

b. Ku-ka chakha-ta.he-NM be nice-DC (adjective chakha- ‘be nice’)‘He is nice.’

Given that unergative versus unaccusative is a categorization among verbclasses, and auxiliary selection as a reflex of unaccusativity in other languages isrelevant to verbs only, the fact that -a iss does not occur with adjectives cannotinvalidate the claim that it reflects unaccusativity in Korean. Notice further thatadjectival predicates are not compatible with -a iss because an aspect markerdoes not normally occur with adjectives in languages in general.

Second, Kim (1990) points out that -a iss occurs with some potentiallyvolitional action verbs such as sit, stand, lie, arrive, come, and go, and therefore-a iss cannot encode unaccusativity. Note that these verbs are ambiguous withrespect to the agentivity feature: If these verbs denote a voluntary action, theyare [+agentive], e.g., please arrive on time; if they denote an involuntary event,they are [–agentive], e.g., the mail has arrived. We may see it as a true lexicalambiguity or as a logical polysemy derived from the event structure. Forexample, accomplishments such as cip-ey ka- ‘to go home’ refer to protractedevents describing an on-going process and thus compatible with -ko iss, as in(23a), but they can also be perceived as instantaneous transitions triggering achange of location and occur with -a iss, as in (23b).

(23)a. Ku-ka cip-ey ka-ko iss-ta.he-NM home-LC go-CN exist-DC‘He is going home.’

b. Ku-ka cip-ey ka-a iss-ta.he-NM home-LC go-CN exist-DC‘He is home.’ (He has come home.)

In (23a), the subject seems to be the agent of the locomotion while in case of(23b) the subject simply undergoes a change of location. I believe that thisexplains why the locomotion verbs are compatible with -a iss, which requires[)agentive] verbs.

This apparently peculiar morphosyntactic pattern in which the potentiallyagentive and volitional posture and locomotion verbs go along withnon-agentive morphosyntax is cross-linguistically common (Lee 1991). Loco-motion verbs such as go, sit down, stand up, etc. pattern together with stativeverbs rather than with action verbs in many languages. For example, loco-motion verbs in French, e.g., aller ‘to go’ and venir ‘to come’, use the auxiliaryverb etre ‘to be’ in the passe compose instead of using avoir ‘to have’. TheItalian verb go and come also take the unaccusative auxiliary essere ‘be’.

130 E. H. Lee

123

English also formerly allowed the perfect to be used with the copula be for theverbs go and come in limited cases. Also, posture verbs are allowed in the there-construction, which is not agentive as in (24b). (24b) is compared to (24a) withanother action verb, which cannot be put in there-construction.

(24)a. *There laughed children watching TV.b. There sat some students eating pizza in the lounge.

Although these locomotion verbs are potentially agentive, there is a clear dis-tinction among the verbs of motion between those that specify a manner ofmotion like roll, walk, swim, etc. and those that specify a direction but not amanner like arrive, come, go, rise, fall, etc. (Hoekstra 1994; Levin andRappaport 1992; Rosen 1984). Levin and Rappaport (1995) also observe adifference among certain spatial configuration verbs. Certain verbs of spatialconfiguration, including hang and stand, allow a transitive causative use whileother verbs in this class, including loom and slouch, do not.

(25)a. The laundry hangs on the clothesline.b. They hang the laundry on the clothesline.

(26)a. The surly youth slouched against the wall.b. *I slouched the surly youth against the wall.

They attribute this contrast to the fact that looming and slouching are posturesthat are necessarily internally caused, unlike hanging, sitting, or standing, whichare postures that can be brought about by an external cause. Verbs of inher-ently directed motion and externally caused spatial verbs are all unaccusatives.

When these posture verbs express the assumption or maintenance of aposition and are agentive, they are used in the -ko iss form. When they arenon-agentive, that is, when describing the position of their subject with respectto some location, i.e., simple position similar to existential sentences, they canbe used in the -a iss form. Given that the two classes of exceptions, namelyadjectives and locomotion verbs, can be given a reasonable explanation, we canmaintain the perfect correlation between the -a iss selection and unaccusativity.

So far I have argued, following Kim (1990), that [+agentivity] is the relevantfeature distinguishing between unergatives and transitives, on the one hand,and unaccusatives and passives, on the other, in Korean. I also argued that-ko iss takes [+agentive] verbs, and -a iss takes [)agentive] verbs. However, thenotion of agentivity in Kim (1990) that has guided the discussion thus far isproblematic since it requires volition and animacy. In Dowty (1979) andJackendoff (1976, 1990), agent involves the notions of doer, volition, andcausers, and it is defined as the first argument of the abstract operator DO(originally proposed in Ross 1972). As shown in (27), however, an inanimatesubject—and therefore non-agentive—can appear as an external argumentin a -ko iss sentence. Since [+agentive] necessarily entails [+animate] or


123

[+volitional], we cannot use agentivity in a typical sense for characterizing-ko iss.

(27)a. Hay-ka ttu-ko iss-ta.sun-NM rise-CN exist-DC‘The sun is rising.’

b. Ku pyeng-i manhun salam-ul cwuk-i-ko iss-ta.the disease-NM a lot people-AC die-CS-CN exist-DC‘That disease is killing many people.’

In fact, objections have been raised to such a notion of agency that it entailsan animate or volitional subject. Van Valin and Wilkins (1996) point out severalproblems with this approach to agents. First, agency cannot be entirelyattributed to the lexical meaning of the verb alone, but it is related to the NPargument. For example, whether break takes DO or not depends on the ani-macy of the argument, e.g., John broke the vase versus the rock broke the vase.Even with an animate subject, the same verb can have different interpretationswith respect to the issue of intention, volition, and control. Each of thesereadings would require a different lexical representation for break, leading to aproliferation of logical structures for break in the lexicon, which is not desir-able. Moreover, it is not at all clear that these differences in interpretationshould be a direct function of the lexical representation of a verb. Furthermore,while there are arguments which are ‘pure’ themes and experiencers, there areno ‘pure’ agent arguments because an agent is always also an experiencer,effector, theme, etc. Hence, unlike the other roles, agent is always a secondaryinterpretation added to other, lexically determined roles. Holisky (1987) pro-poses a radically different account of agent. According to Holisky, most verbsare simply unmarked for agency, and the interpretation of an argument as anagent is a pragmatic inference or implicature and not an inherent property ofthe verb’s semantic representation. His pragmatic principle states, ‘‘You mayinterpret effectors and effector-themes which are human as agent (in theabsence of any information on the contrary)’’ (pp. 118–119). In the vastmajority of these cases the actual lexical requirement of the verb is that itssubject be a doer (effector) and not necessarily an agent. The role ‘‘effector’’ or‘‘doer’’ is the first argument of the activity predicate (do), and it includes notonly agents but also forces and instruments.

Levin and Rappaport (1995) also point out that, although the notions ofchange of state and agentivity have figured in previous accounts of unaccus-ativity, these notions are too coarse. They propose, instead, a semantic dis-tinction between ‘‘internally’’ and ‘‘externally’’ caused eventualities, which ispartly based on Smith’s (1970) notion of control. Verbs like break and open,Smith proposes, describe eventualities that are under the control of someexternal cause that brings the eventuality about. Such intransitive verbs havetransitive uses in which the external cause is expressed as subject. Verbs likelaugh, play, and speak do not have this property: the eventualities that eachdescribes ‘‘cannot be externally controlled’’ but ‘‘can be controlled only by the

132 E. H. Lee

123

person engaging in it’’; that is, control ‘‘cannot be relinquished’’ to an externalcontroller (p. 170). Levin and Rappaport (1995, p. 91) use a similar notion ofinternally and externally caused eventualities. With an intransitive verbdescribing an internally caused eventuality, some property inherent to theargument of the verb is ‘‘responsible’’ for bringing about the eventuality. Foragentive verbs such as play and speak, this property is the will or volition of theagent who performs the activity. However, an internally caused verb need notbe agentive. For example, the verbs blush and tremble, which take animate—though non-agentive—arguments can nevertheless be considered as describinginternally caused eventualities because these eventualities arise from internalproperties of the arguments, typically an emotional reaction. Verbs with aninanimate—and thus clearly non-agentive—single argument, e.g., verbs ofemission, may also describe internally caused eventualities in the sense thatthese eventualities are conceptualized as arising from inherent properties oftheir arguments.

Although these researchers use different terms, ‘‘doer’’, ‘‘effector’’, ‘‘internalcauser’’, and ‘‘controller’’, they all seem to agree that the notion of agent is nota basic or fundamental semantic role and that an inanimate and nonvolitionalargument can also serve as the external argument of unergative and transitiveverbs. Following this line, I assume that agentivity in Korean does not requirevolition or animacy either. I will henceforth use the term ‘‘actor’’ instead of‘‘agent’’ to avoid confusion. Actor role includes inanimate forces as well asvolitional and animate agents.

4 Semantic analysis

In this section, I will provide a semantic analysis of the two markers using eventsemantics (Davidson 1967, Bach 1986, Parsons 1989, among others). The keyingredients of event-based approaches are as follows: (a) to include events(incomplete as well as atomic events), (b) moments and intervals of time in ourontology, and (c) using part-whole relations between events as a basic tool toexplain the progressive.

Let us define our model first. A model M is a structure hE, T, I, F i where E isan event structure, T is a set of times linearly ordered, I is the set of individuals,and F is the interpretation function. The event structure consists of a set ofeventualities EM = SM ¨ EM, where SM is the set of states, and EM is the set ofevents. A temporal precedence relation <, which is a strict partial order, and atemporal inclusion relation ˝, which is a partial order, are both operative onEM X EM.

First, I will characterize the verbs that take the -ko iss form and then definethe meaning of -ko iss. I assume that verbs which take -ko iss have an ACTORrole as their external argument. I assume that verbs which take -ko iss can have‘do’ as part of their lexical meaning, which is defined as follows.

(28) Actor Event (P, x, (y)) iff for every M, w, t, g,sh"x[P(x) fi �$y[doing(x,y)]]tM,w,t,g = 1


123

The -ko iss form is translated as follows.

(29) sVP-ko isst M,w,t,g, where VP denotes an event whose first argumentis an actor, i.e., [KO ISS] [–VP[+actor]]= kPkekx[P(x)(e) � doing (e, x)]

For illustrative purpose, (30) contains sample translations. The set of thematicroles are partial functions from the set of events E to the set of individuals I. ACstands for actor, and TH stands for theme.

(30)a. John-i talli-ko iss-ta. ‘John is running.’1) talli- ‘run’ � kxke[running’(e) � AC (e, x)]2) talli-ko iss- ‘is running’ � kxke[running’(e) � doing (e, x) � AC (e, x)]3) John-i talli-ko iss- � ke[running’(e) � doing (e, j) � AC (e, j)]4) John-i talli-ko iss-ta. ‘John is running’ � $e[running’(e) � doing (e, j)

� AC (e, j) � now ˝ s(e)] (s is a temporal measure function of anevent, i.e., an event’s running time

b. John-i cip-ul cis-ko iss-ta. ‘John is building a house.�1) cip-ul cis- ‘build a house’ � kxke[building’(e) � AC (e, x) � CAUSE

[BECOME [a house exists]]]2) cip-ul cis-ko iss ‘building a house’ � kxke[building’(e) � doing (e, x)

� AC (e, x)]3) John-i cip-ul cis-ko iss- � ke[building’(e) � doing (e, j) � AC (e, j)]4) John-i cip-ul cis-ko iss-ta. ‘John is building a house.’ � $e[building’(e)

� doing (e, j) � AC (e, j) � now ˝ s(e)]

Transitive verbs like build a house are a complex expression containing the processpart ‘‘building’’ and the result part ‘‘a house coming into existence’’. When -ko issoperates on such a predicate, it picks out only the process part and ignores theresult part due to its semantics of ‘‘doing’’. Therefore, no imperfective paradoxarises, and (30b) does not entail that there is or will be a house built. I assume thatless obvious cases, such as verbs of wearing and contact, e.g., ip- ‘wear’ andtha- ‘ride’ and transitive psych-verbs al- ‘(come to) know’ and cohaha- ‘like’, alsohave an actor role as their external arguments (they canalso be experiencer), and inthe -ko iss construction, they express that the actor is ‘‘doing’’ something, e.g.,assuming and maintaining the appearance, position, knowledge, and emotions,etc. This is supported by the fact that these verbs pass agentivity tests.

(31)a. Ku os-ul ip-ko iss-ela.that clothes-AC wear-CN exist-IMP‘Keep wearing that clothes.’

b. I ket-ul al-ala.this fact-AC know-IMP‘Know this.’

c. Ney iwus-ul salanghay-la.your neighbor-AC love-IMP‘Love your neighbor.’

134 E. H. Lee

123

Now let us define the semantics of -a iss. Verbs that take -a iss are unac-cusatives, which do not have ‘do’ in their logical structure. I define them as‘‘non-actor events’’ as follows.

(32) Non-actor Event (P) iff for every M, w, t, g,sh"x[P(x) fi :�$y[doing(x,y)]]tM,w,t,g = 1

When the -a iss form is complemented by stative verbs such as sal-ta ‘to live’, itindicates that the state of living continues. On the other hand, when it iscomplemented by punctual unaccusatives such as cwuk-ta ‘to die’, one mustfind some duration to which the marker can apply. I assume that the state ofbeing dead provides such a duration for the verb in question. Note that x diesentails x is (becomes) dead. Therefore, I assume that cwuk- ‘die’ is a complexpredicate consisting of an instantaneous action of dying and a state of beingdead, as in (34a). The -a iss form is translated as follows.

(33) sVP-a isstM,w,t,g, where VP denotes an event whose first argument is notan actor, but its only argument is a theme, that is, [A ISS] [–VP[-actor]]= kPkekx[P(x)(e) � IN (e, x)] (IN (e, x) reads, ‘x is in the state e’).

Below are some sample translations.

(34)a. John-i cwuk-e iss-ta. ‘John is dead.’1) cwuk- ‘die’ � kxke[e = e1 ¯ e2 � dying’(e1) � BECOME[dead’(e2, x)]

� TH (e, x)]2) cwuk-e iss ‘dead’� kxke[BECOME[dead’(e, x) � IN (e, x) � TH (e, x)]]3) John-i cwuk-e iss � ke[BECOME[dead’(e, j) � IN (e, j) � TH (e, j)]]4) John-i cwuk-e iss-ta. ‘John is dead’ � $e[BECOME[dead’(e) � IN (e, j)

� TH (e, j) � now ˝ s(e)]]b. John-i sal-a iss-ta. ‘John is alive.’

1) sal- ‘live’ � kxke[living’(e) � TH (e, x)]2) sal-a iss ‘alive’ � kxke[living’(e) � IN (e, x) � TH (e, x)]3) John-i sal-a iss � ke[living’(e) � IN (e, j) � TH (e, j)]4) John-i sal-a iss-ta. ‘John is alive’ � $e[living’(e) � IN (e, j) � TH (e, j)

� now ˝ s(e)]

(35) is an example of a passive sentence. The passive suffix hi- suppresses theACTOR role, which was originally in the representation of a transitive verbcap- ‘catch’.

(35) John-i cap-hi-e iss-ta. ‘John is being caught.’1) cap- ‘catch’ � kxkyke[catching’(e) � AC (e, x) � TH (e, y)]2) cap-hi- ‘be caught’ � kyke[BECOME[caught’(e, y) � TH (e, y)]]3) cap-hi-e iss � kyke[BECOME[caught’(e, y) � IN (e, y) � TH (e, y)]]4) John-i cap-hi-e iss � ke[BECOME[caught’(e, j) � IN (e, j) � TH (e, j)]]5) John-i cap-hi-e iss-ta. � $e[BECOME[caught’(e) � IN (e, j)

� TH (e, j) � now ˝ s(e)]


123

Finally, let me provide two distinct lexical representations of verbs likeanc- ‘to sit’, as follows. AT (x, y) means x is at (in the location of) y.

(36)a. anc-1 � kxke[sitting’(e) � AC (e, x)]b. anc-2 � kxkyke[sitting’(e) � BECOME[AT (x, y) � TH (e, x)]]

The agentive verb in (36a) can co-occur with -ko iss while the unaccusative verbin (36b) can appear with -a iss.

5 Conclusion

In this paper, I have argued that an analysis of Korean imperfective aspectmarkers reveals a strong correlation between the syntactic argument structureand the semantics of aspect. Korean provides a good test case for unaccus-ativity and its semantic basis due to its overt case marking and interestingpsych-predicate minimal pairs. I argued that the difference between -ko issand -a iss derives partly from their different argument structures and that apurely aspectual dichotomy of progressive versus resultative is not adequate.The -ko iss form takes predicates that have an external argument to indicate anon-goingness of some active action while the -a iss form takes predicates whosesole argument is a theme subject to express a state in which the theme argumentis located. I argued that the distinction between -ko iss and -a iss is a reflex ofunaccusativity in Korean and proposed agentivity as the semantic basis for it(Kim 1990), with a revised notion of agents (actors), including inanimate forces.I also provided a compositional semantics of the two markers, using event-based semantics.

This paper has broader implications for the syntax–semantics interface,manifested by the semantic correlation between argument structure and internaltemporal structure of events. Dowty (1991) points out that the plausibility of theunaccusativity hypothesis rests on the fact that the intransitive predicates arguedto be unaccusative on syntactic grounds usually turned out to entail relativelypatient-like meanings for their arguments while those argued to be syntacticallyunergative were usually agentive in meaning. That is, given the assumption thatdirect objects are somehow inherently patient-like, it is plausible that the formerbut not the latter are ‘underlying’ objects promoted to subjects. In Dowty’s sys-tem, thematic roles are not discrete, and there are only twoproto-roles, i.e., Proto-Agent and Proto-Patient. The membership of each category is determined by anumber of entailments. From the list of Proto-Agent entailments, the mostimportant for the unergative/unaccusative contrast is volition, and verbs withvolition are always unergative. Among the Proto-Patient entailments, incre-mental themehood, i.e., whether or not it is telic, is highly significant for thedistinction between unaccusative and unergative. The appeal to two entailments,each from a different proto-role, gives us two possible loci for a semanticboundary, namely ‘Agency’ and ‘Telicity’. Agentive atelic verbs are definitelyunergative while non-agentive telic verbs are definitely unaccusative; other

136 E. H. Lee

123

combinations can vary in a language as well as cross-linguistically. If the mostimportant distinction is between agentivity and lack of it, then agentive verbs,regardless of telicity, will be unergative; if the distinction between telicity and lackof it is primary, then atelic verbs, regardless of agentivity, will be unergative, andtelic verbs will be unaccusative. Dowty further speculates that ‘active’ languagessuch as Lakhota are closer to exemplifying the former while the unaccusativitymanifested in Italian ismore like the latter. Similarly,VanValin (1990) argues thatinherent lexical aspect and agentiveness are the primary semantic parametersgoverning the distinction between unergative and unaccusative verbs and thatlanguages vary with respect to which parameter governs the split. For instance, inItalian and Georgian the relevant contrast is between classes of verbs with dif-ferent lexical aspect properties, i.e., telicity, whereas in Acehnese and Tsova-Tushthe contrast turns crucially on agentiveness.

Korean belongs to the latter type of language: it is the presence or absence ofthe actor role that determines whether a verb is unergative or unaccusative. If itis indeed the case that the distinction between unergative and unaccusativeverbs cuts across two features, namely agentivity and telicity, and languagesmay differ as to which feature is given more weight, we may come up with aninteresting typology of languages not only for the unergative/unaccusativedistinction but also for aspect and event structure in general. Active or agentivelanguages (such as Korean, Tsova Tush) will be initiator (external argument)-oriented languages while telicity languages (such as Georgian and Italian) willbe delimitation (incremental theme)-oriented languages.

Ritter and Rosen (2000) suggest that predicates form different naturalclasses depending on a language’s classification as an I-language (initiatorlanguage) or D-language (delimitation language). In D-languages, accom-plishments and achievements form a natural class (both have terminal bounds)while in I-languages, accomplishments and activities form a natural class (bothhave initial bounds). It is interesting to note that Korean also falls into theI-language group, which is expected given its emphasis on the agentivity side ofcriteria for the uanccusative/unergative distinction. For example, accom-plishments in the past do not necessarily entail that the event has been com-pleted in Korean because common nouns are not marked for singular/plural,and both count and mass nouns employ an optional numeral classifierconstruction to specify their number/quantity. In (35), sakwa can denote‘an apple’, ‘apples’, or even a piece of an apple.

(37) John-I sakwa-lul mek-ess-nuntey sakwa-ka acikJohn-NM apple-AC eat-PST-but apple-NM stillnam-a iss-ta.remain-CN exist-DC‘John ate *an apple/apples, but there are still some left.’

Moreover, accomplishments with a bare object can be modified by an ‘in’adverbial as well as a ‘for’ adverbial.


123

(38)a. John-i il nyen-tongan cip-ul ci-ess-ta.John-NM one year-for house-AC build-PST-DC‘John built a house for a year.’

b. John-i il nyen-maney cip-ul ci-ess-ta.John-NM one year-within house-AC build-PST-DC‘John built a house in a year.’

The next step in this line of inquiry, then, will be to investigate whethergeneralizations drawn from a cross-linguistic comparison of I-languages versusD-languages are valid by examining more languages and devising a method ofuniversal representation of the correlation between syntactic argument struc-ture and aspectual event structure.

References

Ahn, Yoonja. 1995. The aspectual and temporal system of Korean: From the perspective of the twocomponent theory of aspect. PhD dissertation, University of Texas, Austin.

Bach, Emmon. 1986. The algebra of events. Linguistics and Philosophy 9: 5–15.Burzio, Luigi. 1986. Italian syntax: A government and binding approach. Dordrecht: Reidel.Choe, Hyun Bae. 1971. Wurimalpon ‘Our Grammar�. Seoul: Cengumsa [first published in 1929].Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.Davidson, Donald. 1967. The logical form of action sentences. In The logical decision and action,

ed. N. Rescher, 81–95. Pittsburgh: University of Pittsburgh Press.Dowty, David. 1979. Word meaning and montague grammar. Dordrecht: Reidel.Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67: 547–619.Holisky, Dee Ann. 1987. The case of the intransitive subject in Tsova-Tush (Batsbi). Lingua 71:

103–132.Hopper, Paul, and Sandra Thompson. 1980. Transitivity in grammar and discourse. Language 56:

251–299.Jackendoff, Ray. 1976. Toward an explanatory semantic repretation. Linguistic Inquiry 7: 89–150.Jackendoff, Ray. 1990. Semantic structures. Cambridge: MIT Press.Kang, Young Seo. 1987. Korean syntax and universal grammar. PhD dissertation, Harvard

University.Kim, Namkil. 1986. The progressive in Korean. In Studies in korean language and linguistics, ed.

N. Kim, 98)117. East Asian Studies Center, USC.Kim, Young Joo. 1990. The syntax and semantics of Korean case: The interaction between lexical

and syntactic level of representation. PhD dissertation, Harvard University.Kim, Yookyung. 1993. The resultative progressive in Korean. CLS 29, 251–252. Chicago:

University of Chicago.Kindaichi, Haruhiko. 1950. Kokugo doosi no itibunrui [A classification of Japanese verbs]. Gengo

Kenkyu 15: 48–63.Kinsui, Satoshi. 1995. Iwayuru ‘sinkoo-tai’ ni tuite [On the so-called progressive]. In Tukisima

Hirosi hakase koki kinen kokugogaku ronsyuu, ed. Tukisima Hiroshi, 169–197. Tokyo: KyukoShobin.

Kudo, Mayumi. 1995. Asupekuto/tensu taikee to tekusuto: Gendai nihongo no zikan no hyoogen[Aspect-tense system and text: Temporal expressions in Modern Japanese]. Tokyo: Hitsuji.

Lee, Hyo Sang. 1991. Tense, aspect and modality: A discourse-pragmatic analysis of verbal suffixesin Korean from a typological perspective. PhD dissertation, UCLA.

Lee, Keedong. 1993. A Korean grammar on semantic-pragmatic principles. Seoul: HankwukMwunhwasa.

Lee, Hyo Sang. 1999. Dynamicity as an archi-concept in Korean. Studies in Language (to appear).Lee, EunHee. 2004. Different states described by aspectual markers -ko iss and -a/e iss in Korean.

Harvard Studies in Korean Linguistics 10: 547–557.Lee, EunHee. 2006. Stative progressives in Korean and English. Journal of Pragmatics 38: 695–717.

138 E. H. Lee

123

Levin, Beth, and Malka. Rappaport 1992. The lexical semantics of verbs of motion: The perspectivefrom unaccusativity. Roca 1992: 247–269.

Levin, Beth, and Malka. Rappaport 1995. Unaccusativity: At the syntax-lexical semantics interface.Cambridge: MIT Press.

Martin, Samuel. 1992. A reference grammar of Korean. Tokyo: Charles E. tuttle Company.Ogihara, Toshiyuki. 1998. The ambiguity of the -te iru form in Japanese. Journal of East Asian

Linguistics 7: 87–120.Parsons, Terence. 1989. The progressive in English: Events, states and processes. Lingustics and

Philosophy, 12: 213–291.Perlmutter, David. 1978. Impersonal passives and the unaccusative hypothesis. Berkeley Linguistics

Society 4: 157–189.Perlmutter, David, and Paul. Postal 1984. The I-advancement exclusiveness law. In Studies in

relational grammar 2, ed. D. Perlmutter and C. Rosen, 81–125. University of Chicago Press.Ritter, Elizabeth. and Sara. Rosen 2000. Event structure and ergativity. In Events as grammatical

objects, ed. C. Tenny and J. Pustejovsky, 187–238. Stanford: CSLI.Rosen, Carol. 1984. The interface between semantic roles and initial grammatical relations. In

Studies in relational grammar 2, ed. D. Perlmutter and C. Rosen, 38–80. University of ChicagoPress.

Ross, John. 1972. Act. In Semantics of natural language, ed. D. Davidson and G. Harmon, 70–126.Dordrecht: Riedel.

Shirai, Yasuhiro. 1998. Where the progressive and the resultative meet: A typology of imperfectivemorphology in Japanese, Korean, Chinese, and English. Studies in Language 22: 661–692.

Shirai, Yasuhiro. 2000. The semantics of the Japanese imperfective -teiru: An integrative approach.Journal of Pragmatics 32: 327–361.

Smith, Carlota. 1970. Jespersen’s ‘move and change’ class and causative verbs in English. InLinguistic and literary studies in honor of Archibald A. Hill, vol. 2: Descriptive Linguistics,ed. M. Jazayery, E. Polome, and W. Winter, 101–109. Berlin: Mouton.

Smith, Carlota. 1991. The parameter of aspect. Dordrecht: Kluwer.Sohn, Ho Min. 1999. The Korean language. Cambridge: The University of Cambridge Press.Van Valin, Robert. 1990. Semantic parameters of split intransitivity. Language 66: 221–260.Van Valin, Robert. 1993. A synopsis of role and reference grammar. In Advanced in role and

reference grammar, ed. R. Van Valin. Amsterdam: John Benjamins.Van Valin, Robert. and Wilkins. David 1996. The case for ‘effector’: Case roles, agents, and agency

revisited. In Grammatical constructions, ed. M. Shibatani and S. Thomson, 289–322. Oxford:Oxford University Press.


123

Argument structure and event structure: the case of Korean imperfective constructions

Documents