'Telic Entity' as a Proto-Property of Lexical Predicates
Farrell Ackerman and John Moore
UC San Diego
Proceedings of the LFG99 Conference
The University of Manchester
Miriam Butt and Tracy Holloway King (Editors)
1999
CSLI Publications
http://www-csli.stanford.edu/publications/

1. Telicity and Transitivity(1)

    Linking theories, such as the Lexical Mapping Theory, account for recurring patterns of grammatical encodings associated with the arguments of predicates cross-linguistically. These proposals are designed to account for the grammatical function status of arguments associated with simple predicates or the alternative grammatical functional encodings associated with related predicates. Alternative function encodings are often divided into two types (cf. Simpson 1982, Ackerman 1990, 1992, and Joshi 1992, Dubinsky and Simango 1996, and Sadler and Spencer 1998, among others):

(i)    Morphosyntactic: Lexical semantics preserved; grammatical function assignments altered (e.g. passive,
                                    locative inversion).
            (ii) Morphosemantic:    Grammatical function alternation corresponds to a semantic contrast (e.g. causative, locative
                                                  alternation).

    This paper will examine a class of morphosemantic alternations, where the semantic contrast is in terms of telicity, and the encoding alternation is realized on the object argument in affirmative clauses containing personal verb forms.(2) This is exemplified in the following Finnish and Estonian examples:(3)

(1)    Finnish:         a.    Matti            osti         maito-a     (tunni-n).
                                    Matti-sg/n    bought   milk-sg.(hour-acc)
                                    'Matti bought milk (for an hour).'
                                b. Matti         osti    maido-n     (tunni-ssa).
                                    Matti-sg/n bought milk-sg.a   (hour-iness)
                                    'Matti bought the milk (in an hour).
                                    (Kiparsky 1998:279, cf also Heinämäki 1984)

(2) Estonian:             a. Ma     ehitasin         endale             suvilat         (kaks nädalat).
                                      I-n     built-1sg/ipf myself-sg-all cottage-p (two-part week-part)
                                     'I was building the cottage for myself (for two weeks).'
                                  b. Ma     ehitasin         endale             suvila         (kahe nädalaga).
                                      I-n     built-1sg/ipf myself-sg-all  cottage-a (two-geN week-gen-com)
                                     'I built the cottage for myself (in two weeks).'
                                     (adapted from Erelt M. et. al., 1997:SY36)

The durative adverbials in the (a) examples co-occur with atelic predicates, while the time span adverbials in the (b) examples co-occur with telic predicates. We see that this semantic contrast corresponds to a partitive/accusative encoding alternation within personal affirmative clauses in both languages.

    The verbal lexicon of Estonian is commonly divided into two large government classes:

(3)     Partitive verbs - verbs that govern partitive case only (cf. Tuldava 1994:187 and Saagpakk 1982:lxvi):

        Feelings                             Senses                                 Unresolved Actions
        armastama   'love'                kuulama       'listen'               aitama        'help'
        rõõmustama 'gladden'          kuulma          'hear'               juhtima       'direct'
        imetlama      'admire'            maistma        'taste'               jätkama     'continue'
        kartma         'fear'                mäletama      'remember'       lööma        'strike'
        kiitma           'praise'            nautima         'enjoy'              ootama      'await'
        põlgama       'despise'          nuusutama     'smell'               otsima       'seek'
        tundma         'feel, know'     nägema         'see'                  segama     'disturb'

(4)     Aspectual verbs - verbs that govern either the partitive or accusative case (cf Erelt et. al. 1993:50):

        avastama     'discover'         looma          'construct'        parandama     'improve'
        saavutama   'attain'              kujundama   'shape'             koostama       'put together'
        keetma        'cook'              voltima         'fold'                moodustama  'form'

These classes, as well as the usage patterns of the relevant verbs form the basis of the following descriptive generalization:

(5) The partitive/accusative alternation in Finnish and Estonian correlates with an atelic/telic contrast.(4)

A generalization such as (5) is consistent with claims by Tsunoda (1981) and Hopper and Thompson (1981) that telicity contributes to the transitivity of a clause. In this paper, we derive (5) from a proto-role theory of argument selection, following Dowty (1991) and a tradition within LFG that dispenses with both atomic thematic roles and thematic hierarchies and recognizes a role for telicity in argument selection. (cf. Ackerman 1990, 1992, Zaenen 1993, and Joshi 1993, Markantonatou 1995). In addition, we propose a new proto-patient property: telic entity. (cf. also Tsunoda 1981, Hopper and Thompson 1981, 1982, Grimshaw 1990, Smith 1991, Krifka 1992, 1998, Tenny 1994, Ramchand 1997, Butt 1998, Kiparsky 1998, Filip 1999, among others.)

2. Telic Entity as a Thematic Proto-Property

2.1. Proto-Properties and Syntagmatic vs. Paradigmatic Selection

Dowty (1991) proposes that the functional encoding of arguments is best formulated in terms of proto-roles, where atomic role labels such as agent and patient are interpreted as proto-type cluster categories based on proto-agent and proto-patient properties:

(6)    Proto-Roles and Proto-Properties:

        Proto-Agent properties                                               Proto-Patient properties

        - volitional involvement in event or state                         - undergoes change of state
        - sentience                                                                    - incremental theme
        - causing an event or change of state                              - causally affected
        - movement (relative to position of other participant)      - stationary (relative to another participant)
        - exists independently of the event                                  - does not exist independently of the event, or not at all

    Under this approach, grammatical function encoding of arguments is regulated by the argument selection principle:

(7)   (Syntagmatic) Argument Selection Principle:
        In predicates with grammatical subject and object, the argument for which the predicate entails the greatest number
        of Proto-Agent properties will be lexicalized as the subject of the predicate; the argument having the greatest number of
        Proto-Patient entailments will be lexicalized as the direct object (Dowty 1991:576).

Because this selection principle applies to co-arguments of a single predicate, we call it the Syntagmatic Argument Selection Principle. Its operation is illustrated in (8):

(8)     The builders                               built                        the house.
          volitional                                                                     undergoes change of state
          sentient                                                                       incremental theme
          causing change of state               <-  syntagmatic ->     causally affected
          movement relative to OBJ                selection              stationary relative to SUBJ
          exists independently                                                     lack of independent existence
          most Proto-Agentive: SUBJ                                   most Proto-Patientive: OBJ

    In Ackerman and Moore (1993, 1995, and 1999), we propose an extension to the Syntagmatic Argument Selection Principle to handle morphosemantic, paradigmatic contrasts. For example the direct/indirect object encoding alternation associated with certain Spanish psych-verbs corresponds to a semantic contrast in terms of the proto-patient property change of state:

(9) Example of Paradigmatic Selection:
      a.    Los perros lo molestan siempre que llega ebrio.
            'The dogs harass him (DO) every time he comes home drunk.'
      b.   Los perros le molestan (* siempre que llega ebrio).
            'Dogs bother him (IO) (* every time he comes home drunk).'
            (Treviño 1990, 50b & 51b)

This morphosemantic alternation yields a paradigmatic contrast in the degree of proto-patientivity with respect to the experiencer argument:

(10) Change of State more proto-patientive: DO (lo)

                            paradigmatic selection

        no Change of State less proto-patientive: IO (le)

Based on paradigmatic alternations like (10), as well as similar alternations in causative and other constructions, we propose the following selection principle:

(11) Paradigmatic Argument Selection Principle:
        In predicates where a 'single' argument exhibits alternative encodings, the most proto-typical alternant with respect to a
        particular proto-role will be realized with a less oblique encoding than an argument that is less proto-typical with respect
        to the same proto-role.

The point of the paradigmatic strategy is to provide a theory of morphosemantic encoding alternations, complementing the syntagmatic strategy. The two encoding strategies are well-formedness conditions over the lexicon: the syntagmatic strategy is a condition on well-formed lexical items, while the paradigmatic strategy is a condition on related lexical items. Thus, the lexical entries in (12) are faithful to both selection principles.

(12) a. molestara     <agr1 ,                             arg2>
            'harrass'         Causer (Proto-Agent)     Sentient (Proto-Agent)
                                                                        Change of state (Proto-Patient)
                                  SUBJ                            DO

        b. molestarb     <arg1 ,                             arg2>
            'bother'            Causer (Proto-Agent)    Sentient (Proto-Agent)
                                    SUBJ                           IO

Given that the selection principles refer to predicate entailments and lexically specified encoding options, molestara and molestarb must have different entailments which correspond to principled differences in function selection. Therefore, they must represent different, but related predicates.

2.2. The Role of Telicity

    As discussed above, Estonian accusative/partitive alternation exhibits a semantic contrast in telicity:

(13) a. Ma     ehitasin         endale                suvilat         (kaks nädalat).
            I-n     built-1sg/ipf myself-sg-all   cottage-p    (two-part week-part)
            'I was building the cottage for myself (for two weeks).'
        b. Ma     ehitasin         endale                suvila         (kahe nädalaga).
            I-n      built-1sg/ipf myself-sg-all   cottage-a (two-gen week-gen-com)
            'I built the cottage for myself (in two weeks).'

This is exactly the type of alternation that the Paradigmatic Argument Selection Principle should account for. However, in order for the alternation in (13) to be accounted for in this way, we need to show that there is (i) an alternation in obliqueness, and (ii) a contrast in proto-patientivity.

2.2.1 Obliqueness

There is evidence that this accusative/partitive alternation corresponds to a case alternation and not a contrast in grammatical function (Uuspõld 1969, Kiparsky 1998).(5) That is, in both cases, the noun phrase is a direct object; the alternation is one of surface case only. Thus, the obliqueness contrast is in terms of case; we express this via the case hierarchy in (14) (cf. Blake 1994:157).

(14) Case Hierarchy: nom > acc/erg > gen > part > dat > loc > abl/inst > other obl

Under this view, then, partitive objects are more oblique than accusative objects. This correctly captures the intuition that the accusative object in (13b) is less oblique than the partitive in (13a).

2.2.2 Contrast in Proto-patientivity

Under the standard set of proto-properties, there is no contrast in degree of proto-patientivity between the objects in (13); both objects bear the same set of proto-patient entailments:

(15) suvila     'cottage-a'     undergoes change of state         (proto-patientive)
        suvilat   'cottage-p'     incremental theme                     (proto-patientive)
                                           causally affected                       (proto-patientive)
                                           stationary relative to SUBJ        (proto-patientive)
                                           lack of independent existence    (proto-patientive)

The basic problem is that the examples appear to contrast with respect to telicity, but that telicity is not represented in the original set of predicate entailments interpreted as proto-properties. Since some notion of telicity appears to be the only relevant semantic difference, it must be represented as a proto-patient property, if we are to explain the morphosemantic contrast with the Paradigmatic Selection Principle. We call this telicity-related proto-patient property telic entity.   Based on the definition of telic predicate in (16), we define telic entity as in (17)

(16)    A lexical predicate P is telic iff for every event e and e', such that P(a1, , an, e) and P(a1, , an, e'), and where e'
           is a subevent of e, e and e' have the same boundaries (end-points). (cf. Krifka 1998).

(17)   An argument ai of predicate P is a telic entity iff P is a telic predicate and entails that a subpart of the denotation of the
          entity that corresponds to ai(under any use of P), expresses the end-point of any telic event denoted by P and its
          arguments.

Given these definitions, we can formalize a contrast in telicity as a contrast in proto-patientivity, by positing two lexically-related predicates that contrast in their proto-patient entailments and have different case-government patterns. (6)
 

(18) a. ehitamaa     <arg1 ,                                         arg2>
            'build'             various Proto-Agent props        various Proto-Patient props
                                                                                  telic entity
                                  SUBJ                                       DO
                                                                                  accusative

        b. ehitama b    <arg1 ,                                         arg2>
            'build'             various Proto-Agent props         various Proto-Patient props
                                                                                   (no telic entity)
                                 SUBJ                                          DO
                                                                                    partitive

The related lexical entries in (18) conform with the Paradigmatic Argument Selection Principle and represent the distinction exhibited by Estonian aspectual verbs.

2.3. Telic Entity vs. Incremental Theme

    Before exploring some consequences of our analysis of telicty-induced morphosemantic alternations, we digress to contrast the new proto-patient property telic entity with Dowty's incremental theme property. In (19) we see an example of an incremental theme argument:

(19) Max mowed the lawn.

The lawn is an incremental theme because it measures out the event (the status of the lawn reflects the degree to which the event is completed). Based on Dowty's discussion, we define incremental theme as in (20):

(20)     Incremental Theme: The argument xi of a lexical predicate P(x1, , xn) is an incremental theme of P iff the
            denotation of P entails that there is a homomorphic mapping (preserving the part-of relation ) from the structured
            denotation of xi into the event structure structured denoted by P and its arguments (cf. Dowty 1991).

Thus, the incremental theme property establishes a homomorphic mapping between the structure of an argument and the corresponding event structure. This mapping preserves the part-of relation:

(21) a.     Kim drank water.
        b.     Kim drank the water.

In both examples, the object is an incremental theme argument. They differ in the structure of the objects. In (21a) the mass-term object (water) is cumulative; that is each subpart of 'water' itself has the property of being water. In contrast, the definite object (the water) in (21b) is quantized; that is, each subpart of 'the water' does not qualify as an instance of the water. These terms follow Krifka (1998), where he shows that the presence of an incremental theme-like homomorphism guarantees a parallel mereology between incremental theme arguments and the structure of the events. In particular, the incremental theme mapping establishes the following relationships:(7)

(22)     Structure of argument           Event structure
            cumulative            ->                    atelic
                                        incremental theme
            quantized             ->                     telic
                                        incremental theme

Thus, while there is a close relationship between incremental themes and telicity, they are distinct. In (21), there are incremental themes, but the events are either atelic or telic. Therefore, a predicate can entail the incremental theme property for one of its argument without entailing the telic entity proto-property.

    Conversely, a predicate can entail telic entity, without entailing the property incremental theme:

(23) a. Tunnen                 hästi     oma     sõpra.
            know-1sg.pres    well     own-p friend-p
            'I know my own friend well.'        (Vääri 1975:103)
        b. Tundsin             selles         noormehes         ära         olümpiavõitja.
            know-1sg.ipf    this-in       youngman-in     preverb olympic-champ-a
            'I recognized an olympic champion in this young man.'  (Kippasto, Nurk, and Seilenthal. 1997)

In (23) there is a stative/achievement contrast. This contrast comes from an alternation in telic entity, not incremental theme:

(24)  a. stative            entails                teli object
        b. achievement does not entail     telic entity

            neither entails incremental theme

We conclude that telic entity and incremental theme are independent proto-patient properties (cf. Ramchand 1997).

3. Telic Entity as a Predicate Entailment

    For Estonian aspectual verbs, we posit separate lexical entries on the basis of (i) contrasting case government, and (ii) contrasting lexical entailment of telic entity. This allows us to use the Paradigmatic Selection Principle to account for the direction of the case alternation. In English verbs like drink, there is no case alternation and the alternation in telic entity seems to depend on the semantics of the object argument. Therefore, there is no reason to posit separate lexical entries. This leads to a more general question: given that telicity is often determined by the nature of the object argument, what evidence is there for treating telicity as a predicate entailment, rather than computing it compositionally? We answer this question by surveying some of the cross-linguistic typology with respect to the exponence of telicity.

3.1. Finnish

Filip (1999) proposes a compositional account of telicity for partitive/accusative alternations in Finnish:

(25) a. Kalle     lämmittää             saunaa.
            Kalle-n warm-3sg.pres   sauna-p
            'Kalle is warming up the sauna.'
        b. Kalle     lämmittää          saunan.
            Kalle-n warm-3sg.pres sauna-a
            'Kalle will warm up the sauna.'
            (Karlsson 1983:80)

Filip assumes that accusative objects, when incremental themes, are quantized; partitive objects are underspecified for quantization. Working in a unificational framework, Filip uses the equivalent of an incremental theme function to ensure that the quantized accusative object maps to a telic event structure. Hence, under her analysis, the telicity of the event is determined by the quantization of the object, which, in turn is keyed to case marking.

3.2. Czech

    In contrast, Filip (1999) proposes a lexical treatment of telicity for the following Czech contrast:

(26) a. Psal             dopis.
            write.past   letter.sg.acc
            'He wrote (a/the letter)/He was writing (a/the letter).'
        b. Pre-psal             dopis. (8)
            over-write.past letter.sg.acc
            'He rewrote a/the letter.'

The unprefixed predicate is underspecified for telicity, while the prefixed predicate derives a telic, perfective predicate (as well as changing other aspects of the predicate meaning). Following Filip, we can represent the telicity of the prefixed predicate as an aspect of the predicate's lexical semantics. This yields the paradigmatic contrast in (27).

(27) a. Pre-psal         <arg1 ,                                     arg2>
            'rewrite'             various Proto-Agent props    various Proto-Patient props
                                                                                  telic entity
                                     SUBJ                                    DO
                                                                                  accusative

        b. Psal                <arg1 ,                                     arg2>
            'write'                 various Proto-Agent props    various Proto-Patient props
                                                                                   (no telic entity)
                                       SUBJ                                   DO
                                                                                   accusative

In Czech, morphological means yield lexical verbal predicates which differ with respect to their entailment sets.

3.3. Estonian

In Estonian, there are a number of partitive verbs that only govern the partitive case; these were given in (3). However, some of these verbs may combine with prefix-like adverbs or preverbs, yielding complex verbs that govern only the accusative case (cf. Remes 1982 and Sulkala 1996). As in Czech, this morphological operation yields a contrast in telicity:

(28) a.     Madis     joob                 teed.
                Madis     drink-3sg/ind  tea-p
                'Madis is drinking tea.'
        b.     Madis     joob                 oma     tee       ära.
                Madis     drink-3sg/ind   own    tea-a    preverb
                'Madis will drink up his tea.'

Thus, Estonian has a preverbal system which functions semantically much like the prefixal system of Czech. In addition, a contrast in telicity yields an encoding alternation with respect to the object case (cf. Hassellblatt 1990 and Pusztay 1994). This suggests that in Estonian, as in Czech, telicity can be a lexical property:

(29) a.     ärajooma         <arg1 ,                                         arg2>
                'drink up'             various Proto-Agent props        various Proto-Patient props
                                                                                            telic entity
                                            SUBJ                                       DO
                                                                                            accusative

        b.     jooma              <arg1 ,                                           arg2>
               'drink'               various Proto-Agent props           various Proto-Patient props
                                                                                          (no telic entity)
                                       SUBJ                                           DO
                                                                                           partitive

Given this independent motivation for treating telicity as a lexical property in Estonian, we propose to extend it to aspectual verbs, where there is no contrast in verb morphology, but only a contrast in case government:

(30) a. Laps         voltis             kahte     paberlaevukest.
            child-n     fold-3sg-ipf  two-p    paper-boat-p
            'The child was folding two paper boats.'
        b. Laps         voltis             kaks     paberlaevukest.
            child-n     fold-3sg-ipf   two-a   paper-boat-a
            'The child folded two paper boats.'

(31) a. voltimaa         <arg1 ,                                             arg2>
            'fold'                 various Proto-Agent props             various Proto-Patient props
                                                                                          telic entity
                                    SUBJ                                             DO
                                                                                          accusative

        b. voltima b        <arg1 ,                                            arg2>
            'fold'                  various Proto-Agent props            various Proto-Patient props
                                                                                         (no telic entity)
                                     SUBJ                                           DO
                                                                                         partitive

This move allows us to use the Paradigmatic Selection Principle to account for the direction of the case alternation in Estonian, both in partitive and aspectual verb classes.

    Given the parallelisms between Estonian aspectual verbs and Finnish concerning semantic contrasts and case-marking alternations, we can extend the analysis in (31) to Finnish. That is, rather than deriving the telicty compositionally, we propose that it be a lexical property, as we have argued is the case with cognate aspectual verbs in Estonian. In other words, we argue that in Czech, Estonian, and Finnish, telicity is treated as an entailment of lexical predicates; that is, telicity has been grammaticized in these languages. In English, on the other hand, there is no evidence for lexical telicity - the contrast has no morphological or case government reflexes. Thus, in English, telicity has not been grammaticized in the same way.

    To summarize, the encoding factors that motivate lexical telicity are: (i) verbal morphology (prefixes, preverbs), and (ii) case government. This translates into the well-known typological distinction between head- and dependent-marking:

(32) The expression of telicity:

        Czech:             head marking                                               telicity is
        Estonian:         head marking and dependent marking          grammaticized
        Finnish:           dependent marking
---------------------------------------------------------------------------------------------------------------
        English:           determined compositionally                          no grammaticization

When telic entity is a lexical entailment and it is expressed by an encoding alternations, then the Paradigmatic Selection Principle correctly predicts the direction of the alternation.

4. Conclusions

    In this paper we have argued for the need to posit telic entity as a proto-property entailment of lexical predicates. In expanding the class of grammatically relevant predicate entailments to include telic entity we have explicitly incorporated into Dowty's framework one of the properties hypothesized to be crucial for distinguishing between degrees of semantic transitivity by Tsunoda (1981) and Hopper & Thompson (1982).

    The need for a telic entity proto-property only became apparent when faced with a class of telicity-induced encoding alternations. The fact that paradigmatic selection reveals the need for a telic entity proto-property raises a larger point: should additional transitivity factors that are proposed by Hopper and Thompson, but not reflected in Dowty's proto-properties be incorporated into a proto-type theory of thematic roles?

    Under our account, it is important that the construct predicate be recognized as an information unit determinative for argument selection (Perlmutter 1979, Mohanan 1995, and Ackerman and Webelhuth 1998, among others). Predicates can be interpreted lexically (Ackerman and Webelhuth 1998 and Frank 1996), or composed of elements in phrasal syntax (see Alsina 1996 and Butt 1998).  By appealing to grammatically relevant predicate entailments and an independently motivated notion of obliqueness, one can develop a general theory of morphosemantic alternations. In addition, we have done this without reifying proto-roles, thereby permitting arguments to be compared with one another with respect to their sets of proto-properties.  This assures a role for Dowty's original counting procedures as relevant for linking.



References

Ackerman, F. 1990. Locative Alternation vs. Locative Inversion. WCCFL 9, CSLI Publications, 1-4.

Ackerman, F. 1992. Morpholexical Relatedness. in I. Sag and A. Szabolcsi eds., Lexical Matters. CSLI Publications, Stanford, 55-83.

Ackerman, F. and J. Moore. 1993. Grammatical Relations and Affected Arguments. talk delivered at the Sixth Biennial Conference on Grammatical Relations, Vancouver, September 1993.

Ackerman, F. and J. Moore. 1995. Proto-Properties, Grammatical Relations, and Case Marking. talk delivered at the Colloque de Syntaxe et Semantique de Paris, October 1995.

Ackerman, F. and J. Moore. 1999. Syntagmatic and Paradigmatic Dimensions of Causee Encodings. Linguistics and Philosophy 22, 1-44.

Ackerman, F. and G. Webelhuth. 1998. A Theory of Predicates. CSLI, Stanford.

Alsina, Alex. 1996. The Role of Argument Structure in Grammar: Evidence from Romance. CSLI, Stanford.

Blake, W. 1994. Case. Cambridge University Press, Cambridge.

Butt, M. 1998. Constraining Argument Merger. in E. Hinrichs. et. al. eds., Complex Predicates in Nonderivational Syntax. Syntax and Semantics Vol. 30. Academic Press, San Diego. 73-114.

de Hoop, H. 1996. Case Configuration and NP Interpretation. Garland, New York.

Dowty, D. 1991. Thematic Proto-Roles and Argument Selection. Language 67, 547-619.

Dubinsky, S. and S. Simango. 1996. Passive and Stative in Chichewa. Language 72, 749-781.

Erelt, M. et. al. 1993. Eesti Keele Grammatika. Vol. II Süntaks. Eesti Teaduste Akadeemia Keele ja Kirjanduse Institut, Tallinn, Estonia.

Erelt, M. et. al. 1997. Eeesti Keele Käsiraamat. Tallinn, Estonia.

Filip, H. 1999. Aspect, Eventuality Types, and Nominal Reference, Garland, New York.

Frank, Anette. 1996. A Note on Complex Predicate Formation: Evidence from Auxiliary Selection, Reflexivization, and Past Participle Agreement in French and Italian. in M. Butt and T. Holloway King, eds., On-line Proceedings of the First LFG Conference, http://www-csli.stanford.edu/publications/LFG/lfg1.html.

Hassellblatt, C. 1990. Das estnische Partikelverb als Lehnübersetzung aus dem Deutschen. Veröffentlichungen der Societas Uralo-Altaica, Wiesbaden, Germany.

Heinämäki, O. 1984. Aspect in Finnish. in C. de Groot and H. Tommola, eds., Aspect Bound. Foris, Dordrecht, 153-177.

Hopper, P. and S. Thompson. 1981. Transitivity in grammar and discourse. Language 56, 251-299.

Hopper, P. and S. Thompson. 1982. eds., Studies in Transitivity. Syntax and Semantics Vol. 15. Academic Press, San Diego.

Joshi, S. 1993. Selection of Grammatical and Logical Functions in Marathi. Doctoral Dissertation, Stanford University.

Karlsson, F. 1983. Finnish Grammar. Werner Söderström Osakeyhtiö, Helsinki.

Kiparsky, P. 1998. Partitive Case and Aspect. in M. Butt and W. Geuder eds., The Projection of Arguments: Lexical and Compositional Factors. CSLI Publications, Stanford. 265-308.

Kippasto, A., A. Nurk, and T. Seilenthal. 1997. Magyar-Észt Vonzatszotar. Bibliotheca Studiorum Hungaricorum in Estoniae, Turku.

Krifka, M. 1992. Thematic Relations as Links between Nominal Reference and Temporal Constitution. in I. Sag and A. Szabolcsi eds., Lexical Matters. CSLI Publications, Stanford, 29-53.

Krifka, M. 1998. The Origins of Telicity. in S. Rothstein ed., Events and Grammar. Kluwer, Dordrecht. 197-235.

Maling, J. 1993. Of Nominative and Accusative: The Hierarchical Assignment of Grammatical Case in Finnish. in A. Holmberg and U. Nikanne, eds., Case and Other Functional Categories in Finnish Syntax. Mouton de Gruyter, Berlin, 49-74.

Markantonatou, S. 1995. Modern Greek Deverbal Nominals: an LMT Approach. Journal of Linguistics 31, 267-299.

Mohanan, T. 1995. Argument structure in Hindi. CSLI Publications, Stanford.

Perlmutter, D. 1979. Predicate: A grammatical relation. In P. Hubbard and P. Tiersma eds., Linguistic Notes from La Jolla, 127-150.

Pusztay, J. 1994. Könyv az Észt Nyelvröl. Folia Estonica. Savariae, Szombathely, Hungary.

Ramchand, G. 1997. Aspect and Predication: the Semantics of Argument Structure. Oxford University Press, Oxford.

Sadler, L. and A. Spencer. 1998. Morphology and Argument Structure. in A. Spencer and A. Zwicky eds., The Handbook of Morphology. Blackwells, Malden, MA, 206-235.

Saagpakk, P. 1982. Estonian-English Dictionary. Yale University Press, New Haven.

Simpson, J. 1983. Aspects of Warlpiri Morphology and Syntax. Doctoral Dissertation, MIT.

Smith, C. 1991. The Parameter of Aspect. Kluwer Academic Publishers, Dordrecht.

Sulkala, H. 1996. Expression of Aspectual Meanings in Finnish and Estonian. in M. Erelt. ed., Estonian: Typological Studies 1. Publications of the Department of Estonian of the University of Tartu. 165-217.

Tenny, C. 1994. Apectual Roles and the Syntax-Semantics Interface. Kluwer, Dordrecht.

Treviño, E. 1990. Non-Canonical Subjects in Spanish: Evidence from Causative and Psych Verbs,' ms. University of Ottawa.

Tsunoda, T. 1981.Split Case-Marking Patterns in Verb-Types and Tense/Aspect/Mood. Linguistics 19:389-438.

Tuldava, J. 1994. Estonian Textbook. Research Institute for Inner Asian Studies. Bloomington, Indiana University

Uuspõld, E. 1969. On the vat-Construction in Modern Estonian. Annual Meeting of the Research Group for Generative Grammar. 32-36.

Vääri, E. 1975. Viron Kielen Oppikirja. Suomen Kirjallisuuden Seura, Helsinki, Finland.

Zaenen, A. 1993. Unaccusativity in Dutch: Integrating Syntax and Lexical Semantics. in J. Pustejovsky ed., Semantics and the Lexicon. Kluwer, Dordrecht. 129-161.

Farrell Ackermanfackerman@ucsd.edu
John Moore moorej@ucsd.edu

1. We wish to thank the following people for helpful discussion; all shortcomings are our responsibility: Chris Barker, Hana Filip, Mark Gawron, C. Hasselblatt, Anu Nurk, Kazuto Matsumura, Maria Polinsky, Eric Potsdam, Gert Webelhuth, and participants of LFG99.

2. We define telicity more precisely below.  Since aspect interacts with alternative case marking for objects in the Finnic languages, we restrict our focus here to transitive clauses.

3. We gloss partitive and accusative -p and -a respectively. The accusative label represents a certain abstraction over the data, as there is a degree of case syncretism that obscures the distinction between accusative and genitive case in some instances. See Maling (1993) for discussion of this issue with respect to Finnish data; similar arguments hold for Estonian.

4. An alternative account links the partitive/accusative alternation to a contrast in specificity (cf. de Hoop 1996). Kiparsky (1998) argues against this approach.

5. For example both partitive and accusative arguments participate in Raising to Object.

6. We should emphasize that not all telic eventualities entail telic entity for an argument; in fact, as Filip (1999) points out, telicity can be determined by non-arguments.  Furthermore, it is possible that a non-object argument may bear the telic entity entailment, as in:

 (i) We put the ball in the bucket.

Unless there is a contrast in the telic entity entailment with respect to a particular non-object, we make no predictions about how these should be encoded.

7. Krifka's formulation differs from Dowty's in the following respect: Dowty (1991, p. 567) defines incremental theme as a homomorphic mapping from object structures to event structures (preserving the part-of relation). Krifka (1998) defines a series of principles that effectively define an isomorphism between event structures and object structures (again, preserving the part-of relation). However, Krifka's proof (p. 214) that a telic event corresponds to a quantized object (and that an atelic event corresponds to a cumulative object) only makes use of the mapping from event structures to object structures - the converse of Dowty's incremental theme. Nevertheless, it is possible, using Krifka's system, to achieve the same results by only referring to the mapping from object structures to event structures. Hence, it appears that a homomorphism in one direction or the other is all that is needed.

8. There should be a hachek over the r in Pre-psal; the same is true in (27a).