'A part of the young (people) climbed onto the roof of the house.'

Arto Anttila (Boston University) &

Vivienne Fong (National University of Singapore)

Email: A. Anttila: anttila@bu.edu

V. Fong: ellfongf@nus.edu.sg

===================================================================

An archetypal part-whole expression (PWE) in Finnish is a quantifier phrase comprising a quantifier and a nominal phrase. Finnish quantifiers can be broadly divided into two semantically motivated classes: (i) count quantifiers that are syntactically modifiers and agree with their heads (e.g., "harva", "joku", "kaikki"); (ii) mass quantifiers that are syntactically heads and take partitive nominal complements (e.g., "hiukan", "kolmasosa", "30%").

Both types of quantifiers may additionally take elative nominals. This leads to some puzzling alternations between nominative and elative with count quantifiers, and between partitive and elative with mass quantifiers. The meaning distinction is sometimes clear ((1)), but sometimes elusive ((2)).

(1) a. jokainen mies `every man'

b. jokainen miehi-stä `every one of the men'

(2) a. osa tä-tä kaupunki-a `part of this city'

b. osa tä-stä kaupungi-sta `part of this city'

Earlier work on PWEs has offered semantic explanations for the choice of case. Chesterman (1991) and Alho (1992) claim that the partitive is indefinite, and the elative definite. Leino (1993) proposes that the partitive denotes unlimited substance, and the elative a specific set/mass. However, neither proposal seems entirely satisfactory. For example, `this city' in (2a) is both specific and definite. Fong and Anttila (1998) propose that the elative denotes specific quantity. Two predictions follow. First, quantifiers whose evaluation requires reference to the size of the set denoted by the nominal (proportional quantifiers like "kolmasosa", "30%"; superlatives like "paras") should require elatives; other quantifiers ("kilo", "runsaasti") should allow partitives freely. Evidence from the Suomen Kuvalehti 1987 corpus shows that these predictions hold, but only statistically. Second, phrases denoting unspecified quantities should reject elative (*"kaksi muutama-sta kissa-sta" vs. "kaksi kolme-sta kissa-sta"). This holds.

Unfortunately, the semantic explanation has some interesting counterexamples. When the PWE functions as an adverbial with elative case ((3)), we get partitive nominals with specific quantity readings. Elative nominals are ruled out ((4)), although in other contexts these quantifiers characteristically take elative nominals ((5)).

(3) a. viide-stä prosenti-sta miehi-ä 'from five per cent of

the/all men'

b. pääosa-sta miehi-ä 'from most of

the/all men'

(4) a. *viide-stä prosenti-sta miehi-stä 'from five per cent of

the/all men'

b. *pääosa-sta miehi-stä 'from most of

the/all men'

(5) a. viisi prosentti-a miehi-stä 'five per cent of

the/all men'

b. pääosa miehi-stä 'the majority of

the/all men'

Thus, while the elative nominal always denotes specific quantity, the partitive nominal is compatible with both non-specific quantity/indefinite readings as well as specific quantity/definite readings. We propose that the syntax of the PWEs in (3-5), where the quantifiers are heads of the construction, rules out the occurrence of double elatives, an instance of Case OCP (Mohanan 1994). Since syntax can override semantics, we take it that the semantic interpretation of partitives must involve genuine optionality and context-dependence.

We conclude by presenting an account of the case selection facts in terms of Optimality Theory. The analysis predicts that the choice of case depends on semantics, but only as a tendency, which lends partial support to non-semantic theories of partitive constructions (e.g. Abbott 1996).

Jóhanna Barddal

Lund University

The basic tenet of Construction Grammar (Goldberg 1995) is that there exists a correspondence between meaning and form. Every linguistic structure has a syntactic form and a semantic meaning. This linguistic structure is the construction. A Construction Grammar approach to morphological case assumes that morphological case is constructional, in the sense that morphological case is a property of arguments, and as such it participates in argument structure constructions (Barddal 1999a). It further predicts that all morphological cases should be productive if the constructions they appear in reach a sufficiently high level of type frequency.

Braine et al. (1990) experiment with action verbs and find that they can occur with two argument structures: the intransitive constructions with a Theme subject and the transitive construction with an Agent subject and a Theme object (for instance the verb roll). Braine et al. therefore argue that verbs are assigned argument structure on the basis of their meaning.

A study of the behaviour of novel verbs in Icelandic (Barddal 1999a, 1999b) has revealed that new verbs acquire their argument structure in three different ways: via Cluster Attraction, Isolate Attraction and Argument Structure Borrowing. Further, there are great differences in frequencies of different case patterns. The pattern with a nominative subject and an accusative object is by far the most common, while the pattern with an oblique subject and a nominative object occurred only once in a corpus of 912 verbs. Consider this example:

(1) Okkur anal‡sera›ist ?etta ?annig a› ...

we (dat.) analysed this (nom.) such that ...

'We came to the analysis that ...'

In Icelandic the Oblique Subject Construction is commonplace for many psychological predicates; considering that, it is strange that the construction only appears once in our corpus of novel verbs.

The interesting question arises whether this construction is less frequent because it is less productive, whether it is less frequent because the verbs which associate with it are perhaps not highly represented amongst novel verbs, or whether the strong Icelandic language purism has next to eliminated all innovative uses of these case patterns.

Since a corpus of novel verbs can be skewed in many ways, we can experiment with nonsense verbs with certain meanings, in order to answer our question. By doing that, we can determine the case patterns which are really productive in the language system. This paper will report on such an experiment, with nonce verbs, testing both adults and children!

References:

Barddal, Jóhanna. 1999a. Case in Icelandic - A Construction Grammar Approach. To appear in Tijdschrift voor Skandinavistiek 1999.

Barddal, Jóhanna. 1999b. Case and Argument Structure of some Loan Verbs in 15th Century Icelandic. Ms. Department of Scandinavian Languages, Lund University.

Braine, Martin D. S., Ruth E. Brody, Shalom M. Fish, Mara J. Weisberger and Monica Blum. 1990. Can Children Use a Verb without Exposure to its Argument Structure? Journal of Child Language 17:313-342.

Goldberg, Adele E. 1995. Constructions. A Construction Grammar Approach to Argument Structure. The University of Chicago Press, Chicago and London.

Hans C. Boas

Research Associate

University of California at Berkeley/University of North Carolina at

Chapel Hill

2619A Benvenue Ave.

Berkeley, CA 94704

U.S.A.

email: hcb@email.unc.edu

phone: 510-841-7329

This paper is a contribution to the discussion of the syntactic and semantic status of resultative constructions which involve a matrix verb (e.g., sweep) denoting an event and a secondary predicate (e.g., clean) which describes the resultant state that is caused by that particular event. Consider the following sentences:

1. a. Rachel swept the floor. 2. a. Dave washed the jeans/himself/ *the soap.

b. Rachel swept the floor clean. b. Dave washed the soap out of his eyes.

c. *Rachel swept herself clean. c. *Dave washed himself/the jeans out of his eyes.

3. a. *Carol ran her Nikes. 4. a. *Sue coughed her cat.

b. Carol ran her Nikes threadbare. b. *Sue coughed her cat sick

c. *Carol ran herself threadbare. c. Sue coughed herself sick.

5. a. *The dog barked the postman. 6. a. *Jenn danced Pat.

b. *The dog barked the postman hoarse. b. Jenn danced Pat tired.

c. The dog barked itself hoarse. c. Jenn danced herself tired.

Goldberg (1995: 180) argues that "the resultative can only apply to arguments that potentially (although not necessarily) undergo a change of state as a result of the verb." While this proposal correctly suggests that the postverbal NP always has to be a patient, it does not have a mechanism to determine whether the patient has to be realized as a "regular" internal argument (cf. (1a, b)) or a nonsubcategorized internal argument (cf. (2b): lexically transitive, (3b): lexically intransitive). In addition to that, Goldberg's account does not predict whether the postverbal NP can be realized as a regular object (cf. (1b), (2b), (3b)), a so-called "fake object" (cf. (4c), (5c), and (6c)), or both (cf. (6b), (6c)). Thus, substituting the resultative predicate 'hoarse' with 'off the property' in (5c) renders the sentence grammatical:

7. The dog barked the postman off the property.

In my paper I will first briefly summarize the main points against an analysis of resultatives as Small Clauses (cf. Hoekstra (1988)) or predication-type structures (cf. Carrier & Randall (1992)). Then, I will show that Goldberg's constructional approach to resultatives lacks predictive power as to what kinds of postverbal NPs are allowed in resultative constructions.

The main part of my paper is concerned with the analysis of resultatives in terms of complex predicates. In particular, I will argue that there are different verb classes that select for different kinds of postverbal NPs and resultative predicates. I will present evidence that the semantic information associated with the matrix verb contains the crucial information that determines the formation of complex resultative predicates. Thus, differences in acceptability between, e.g., (5b) and (7) are due to the fact that 'bark' is only able to form a complex predicate with a resultative phrase when the proper semantic criteria are met. Finally, I will show that the distribution of complex resultative predicates is linked to the formation of causative predicates.

Carrier, J. and J. Randall (1992): "The Argument Structure and Syntactic Structure of Resultatives," in: Linguistic Inquiry 23, 173-234.

Hoekstra, T. (1988): "Small clause result," in: Belgian Journal of Linguistics 7, 125-151.

Goldberg, A.E. (1995): Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press.

The Link between Structure and Meaning in Verbs Denoting Transfer of a Message

Esther Borochovsky - Tel Aviv University, Israel

Verbs which denote the transfer of a message, describe an action with three participants: the message addresser (Agent), the message addressee (Dative) and the message itself (Object). The full expected syntactical frame is therefore: NP (noun phrase)1 + V (verb) + NP2 + S (sentence/clause). Indeed, this is, for example, the syntactic frame of the verb "tell" in the sentence: "David (NP1) told his friend (NP2) that he is getting married (S/clause)".

And yet, many verbs denoting transfer of a message appear in unexpected syntactic frames: sometimes, for pragmatic reasons, a noun-phrase, other than one of the three expected elements, is added to a syntactic frame, thus creating a larger than usual syntactic frame. On the other hand, sometimes, due to semantic or pragmatic reasons, we find a reduced syntactic frame, where one or two of the semantic elements are not realized. Examples: "David reported to the teacher about Dan, that he copies his homework." (In Hebrew this sentence is acceptable: "David hilshin lamora al Dan shehu ma'atik shiurim"). The additional element is "about Dan", and the verb's syntactic valence is four instead of three.

In the following sentences, however, the verb valence is reduced from the expected valence of three, to two, and even one:

1. David complains that his neighbor steals his newspaper.

- There is no realization of the addressee due to a pragmatic reason: the message is more important (to the speaker) than the person who receives it.

2. David shouted at the teacher.

3. David talked to Dan.

In these two examples, the "message" element is not realized, either for a semantic reason - it is implied or incorporated in the meaning of the verb (example 2), or due to a pragmatic reason - the wish to stress the fact that a conversation took place (example 3).

4. He talks all the time.

In this sentence both the addressee and the content are not realized, and the verb, focusing on the action of talking, allows a syntactical frame which realizes only the addresser.

As shown above, the basic meaning of verbs dictates its syntactical structure, and particular idiosyncrasies of many verbs cause changes in this expected basic frame.

Patrick Brandt

Utrecht Institute of Linguistics OTS

Trans10, 3512 JK Utrecht - The Netherlands tel. +31-30-2536057

Several puzzles arise in the analysis of ditransitive constructions and especially in the context of the phenomenon labelled 'dative shift'. One is that while in the prepositional realization, the prepositional indirect object may scope over the (indefinite) direct object, the direct object never scopes over the (indefinite) indirect object (cf. Larson (90)):

1a) The teacher assigns an exercise to every student (several exercises)

b) The teacher assigns a student every exercise (one student)

Elaborate analyses involving considerable amounts of stipulations about possible QR- adjunction sites and Small-Clause structure have been offered for this subtle though robust phenomenon (e.g. Aoun and Li (89)).

In this talk I present evidence that the reason for the asymmetry is actually much more straightforward and leads us directly to what is really behind 'dative shift': In the double object construction, the indirect object receives a definite (existence and uniqueness presupposing) interpretation. Scopally inert due to its semantics, even if the indirect object found itself in the scope of the direct object at some relevant level of representation, it wouldn't be distributed over.

Most tellingly, we encounter the following fact in Russian: In the double object construction (but not in what corresponds to the prepositional realization), the indirect object has to have a 'definiteness' marker, otherwise the construction is ungrammatical:

2a) ... on dal kakomu-to studentu knigu

... he gave some -def student book

2b) *... on dal kakomu-nibud studentu knigu

... he gave some -indef student book

Noticeably, everything else stays the same in Russian: No prepostion disappears, no structure is added. What changes is the word order and the c-command relations between the two objects (Barss and Lasnik (86)). To large degrees, the same holds for other 'scrambling' languages as diverse as Sakha (Yakut) and German.

The semantic effects and structural properties of 'dative shift' are highly reminiscent of what has been found to hold of (short distance) scrambling, which has been argued convincingly not to derive from movement but from different base-generated structures (Neeleman and Weerman (98)).

In this vein, I argue against a technically complicated transformational analysis of dative shift but propose that the two constructions correspond to different base-generated 'scrambled' orders. On independent grounds, the indirect object is argued to correspond to an adjunct (modifier/ secondary predicate) rather than an argument (pace Marantz (84)), its structural position determined by selectional restrictions and locality of predication rather than by devices such as thematic hierarchies and associated mapping principles.

The analysis is not only reasonably simple, but also strongly supported by recent findings about topic-focus articulation in its relation to word order and stress patterns (Cinque (93), Reinhart (95)): In the double object construction, the indirect object escapes the sentence's default stress corresponding to its focus. Being definite ('given' in Heim's (82) terminology), the indirect object belongs to the sentence's topic which is at odds with carrying stress.

Following Partee, we may represent the (event-)semantics of the two constructions in terms of tripartite structures roughly as following (cf. Haijkova, Partee, Sgall (98)):

3a) Pam gives a beggar a dollar

OP e, x [Pam give x (e) & to a beggar (e)] [Pam give a dollar (e) & to a beggar (e)]

3b) Pam gives a dollar to a beggar

OP e,x [Pam give a dollar (e) & to x (e)] [Pam give a dollar (e) & to a beggar (e)]

Aoun, J. and Y.-H.A. Li (1989): Scope and Constituency. Linguistic Inquiry 20

Barss, A. and H. Lasnik (1986): A note on Anaphora and Double Objects. Linguistic Inquiry 17

Cinque, G. (1993): A Null Theory of Phrase and Compound Stress. Linguistic Inquiry 24

Hajicova, E., Partee, B.H. and Sgall, P. (1998): Topic-Focus Articulation, Tripartite Structures, and Semantic Content. Dordrecht: Kluwer (Studies in Linguistics and Philosophy 71)

Heim, I. (1982): The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation, University of Massachusetts, Amherst.

Larson, R. (1990): Double Objects Revisited: Reply to Jackendoff. Linguistic Inquiry 21

Marantz, A. (1984): On the Nature of Grammatical Relations. Cambridge, MA: MIT Press (Linguistic Inquiry Monographs 10)

Neeleman, A. and F. Weerman (1998): Flexible Syntax. A Theory of Case and Arguments. Manuscript Utrecht Institute of Linguistics OTS

Reinhart, T. (1995): Interface Strategies. Manuscript Utrecht Institute of Linguistics OTS

Beaumont Brush

beaumont.brush@sil.org

SUNY-Buffalo Linguistics

It is widely assumed that the lexical entry for a verb must stipulate the types of complements that the verb takes. This assumption misses the generalization that semantically related verbs take the same pattern of complementation, i.e. allow some and disallow others. Clearly we do not want to say that complementation is arbitrary, but listing complement types in the lexical entry without further motivation does just this. Another possibility is that with a sufficiently rich semantic representation it is possible to derive the complements a particular verb can take from the representation itself instead of simply listing that information in the lexical entry.

This paper provides semantic representations in a Role and Reference Grammar framework for the basic mental predicates 'think', 'know', 'want', 'feel', 'see', and 'hear' that are motivated according to the semantics of the individual verb and that also link up naturally to the semantics of complement types, e.g. psych-action, direct perception, cognition, and propositional attitude. We represent mental predicates as relations between the mind with itself, the body, and the world (i.e. what is external to the mind and body), and use those elements in the representation of the predicate semantics as well as the semantics of the complement type. Complement types are then defined according to what sentential elements are related (juncture, e.g. core, nucleus, clause) and how (nexus, e.g. subordination, coordination, cosubordination), with the morphosyntax of each construction varying from language to language. Thus we have a natural and non-arbitrary way to link the semantics of predicates with the semantics of syntactic complement types.

The stability of syntactico-semantic links diachronically in English

Ruth Carroll

Pembroke College, Oxford

ruthcarroll@yahoo.com

Levin 1993 provides an extensive survey of the syntactic behaviour of English verbs that can be linked to the verbs' semantics. In the book's introduction, Levin calls for further research to account "in a principled and systematic way" for the "attested patterns of behavior exhibited by verbs in English and other languages" (1993:14). This paper presents the results of research into the verbs of fourteenth-century English.

The paper demonstrates that there has been little change in the syntactico-semantic links since the Middle English period. This is of interest both for the study of the syntax/semantics interface and for language change and the history of English.

The search for components of meaning relevant to syntax begins with the study of verb classes. One such class is the Cooking Verbs. In Modern English such verbs accept the Causative/Inchoative Alternation, the Instrument Subject Alternation, Resultative Phrases, and the Adjectival Passive Participle (Levin 1993:243-244). Middle English Cooking Verbs have these syntactic characteristics as well:

Causative/Inchoative:

(1) a. Take pork and rost it (FC 120/1)

Take pork and roast it

b. take a peece of salt beof ... & let hit roste (MB 121/7-9)

take a piece of salt beef ... and let it roast

Instrument Subject:

(2) a. let hem seþe ouer þe fuyre (MB 75/10)

let them simmer over the fire

b. hete defieþ and seþiþ rawe þingis (Trev.Barth. 132/35)

heat digests and stews raw things

Resultative Phrase:

(3) fry hem broun (FC 116/9)

fry them brown

Adjectival Passive Participle:

(4) & plaunt it with smale briddes istyued (FC 176/5)

and plant it with small birds stewed

That the syntax is linked with semantics and not with lexemes is seen clearly when relevant lexical changes are taken into consideration. The verb seethe is still used in English, but no longer as a Cooking Verb. The paper shows that while the word served semantically as a Cooking Verb, it could be used in the syntactic constructions characteristic of the Cooking Verbs. However, now that the verb has undergone semantic change, it no longer has these syntactic characteristics:

Causative/Inchoative:

(5) a1. * I seethed him with rage.

a2. * His anger seethed him with rage.

a3. * The events of the evening seethed him with rage.

b. He seethed with rage.

Instrument Subject:

(6) a. He seethed with rage.

b. * Rage seethed him.

Resultative Phrase:

(7) * He seethed purple with rage.

Adjectival Passive Participle:

(8) * a seethed man

Through a systematic treatment of a variety of Middle English verbs, this paper demonstrates that the syntactico-semantic links in English have proved more resistant to change than other aspects of the language.

REFERENCES:

Levin, B. 1993. English Verb Classes and Alternations: A Preliminary Investigation.

The University of Chicago Press.

Barth = Seymour, M.C., ed. 1975. On the properties of things: John Trevisa's translation of Bartholomæus Anglicus De Propreitatibus Rerum. The Clarendon Press.

MB = Heinrich, F, ed. 1896. Ein mittelenglisches Medizinbuch. Max Niemeyer.

FC = Hieatt, C.B and S. Butler, eds. 1985. The Forme of Cury. Curye on Inglysch, 98-145. EETS ss 8.

Kristin M. Eide & Tor A. Åfarli, Helsinki, Sept. 2-4, 1999

NTNU, Trondheim.

1. Bowers (1993) proposes that the subject-predicate relation is formed by a predication operator (a function from a property to a propositional function) that heads an independent functional predication projection, as shown e.g. in (1a). He also suggests that this operator may be lexicalised in certain cases, for instance as in (1b).

(1) a. ...make [PrP John [Pr' [Pr OP] [AP crazy]]]

b. ...regard [PrP John [Pr' [Pr OP/as] [AP crazy]]]

In our talk we will first show that many different types of element may lexicalise the predication operator. We then go on to argue that a given visible element is often multifunctional in that it may potentially lexicalise different types of functional operator, and we propose that syntactic representations are structured objects essentially composed by functional operators that are made visible by various types of element by insertion or movement.

2. Using Mainland Scandinavian data, we first show that the particle som in structures like (2) is the lexicalisation of the predication operator.

(2) ...anse [Jon som gal]. (...consider Jon as crazy.)

We then argue that til in resultative small clauses like (3a) and for in small clauses like (3b) lexicalise the predication operator, as well.

(3) a. ...gjøre [Jon til forbryter]. (...make Jon to a criminal.)

b. ...ta [Jon for kelner]. (...take Jon for (being) a waiter.)

We will also argue that copular verbs lexicalise the predication operator. Generalising that idea, we furthermore argue that ordinary main verbs lexicalise the predication operator.

3. Not only do different elements lexicalise the predication operator; we also show that any given element that may lexicalise that operator has potentially various other functions besides. For instance, som may be used as a complementizer or comparative particle in addition to its use as a predicational particle. Likewise, være 'be' may be used e.g. as an auxiliary verb or an existential verb. That is, the very same element appears to be capable of being "recycled" in different syntactico-semantic roles.

We explain this phenomenon in terms of a notion of multifunctionality, arguing that polysemy is a derivative notion. Assuming that syntactic elements make functional operators visible by "supporting" them (cf. the notion of do-support), we thus argue that the content of any syntactico-semantic projection is at least partly constituted by the content of the operator and partly by the content of the supporting element or marker. Therefore, the supporting element always underdetermines the content of the projection it heads. Specifically, we shall argue that any visible element is a potentially versatile structural position marker, and we propose that a syntactic string is the derivative visible expression of a rudimentary semantic representation constituted by structurally ordered operator tokens, the latter in effect constituting the basic underlying logical form of the sentence. We furthermore claim that the syntactic string reflects this rudimentary operator structure in a homomorphic fashion (cf. also Bouchard 1995).

Raffaella Folli

Italian, like English and many other languages, has the possibility of using the same verb, when it belongs to the class of 'change of state verbs', to produce both a causative/transitive and an inchoative/intransitive sentence:

(1)a. il vento ha rotto la finestra

(1)b. the wind broke the window

(2)a. la finestra si h rotta

(2)b. the window broke.

As we can see, the transitive form is the same in the two languages, but in (2)a. Italian forms the inchoative variant by inserting si. A comparative analysis supports the view that the intransitive member of the alternation is derived form the transitive one by means of a process of deletion or neutralisation of the causer argument. The derivation then goes from the prior transitive structure to the derived intransitive (Burzio 1986, Manzini 1986, Cinque 1988).

This line of thought is adopted by Levin and Rappaport-Hovav (1995). They believe that the priority of the transitive member of the alternation in analysis can be supported by the consideration of those languages in which the intransitive variant shows a certain degree of increase in the morphological complexity of the verbal projection. Italian would be in this sense an example of such an increase. Higginbotham (1997), however, suggests that the intransitive form is both conceptually and, at least in English also syntactically, prior. The conceptual argument is that there are cases, such as English sink, where detransitivization would imply, wrongly, that to say the boat sank would invariably implicate that there was an event that was the cause of its sinking.

The literature cited above, and other work such as Chierchia (1989), has considered only cases where si is obligatory in the inchoative in Italian. But a closer examination reveals that there are three distinct classes of verbs showing the transitive-inchoative alternation: (a) a class for which si is obligatory in the inchoative; (b) a class in which si is impossible; and (c) a class in which si is allowed but not necessary. I will argue that, in (a), the inchoative is indeed derived from the causative; that in (b) the causative is derived from the inchoative; and that in (c) likewise the inchoative is prior, but the form with si carries a distinctive meaning, the elaboration of which carries implications for the interpretation of si itself. Moreover, the inchoative verbs in classes (a)-(c) in general have distinctive types of meaning: (a) comprises (unaccusative) V of "independently determinable change of state" (e.g., when a thing is broken (rotto), it is in pieces); (b) comprises V expressing change of position or degree, without an implicated final state (e.g., invecchiare, "age"); and (c) comprises V whose final state is given as the maximum change that may be obtained (e.g., bruciare ,"burn", reaching up to bruciato "burnt", but no farther). Accordingly, I will propose a semantic formalization for the possible derivations in the three classes defined above.

I propose various syntactic and semantic tests that distinguish among classes (a)-(c), as follows:

1. For class (a), the conjunction of the inchoative with the denial of the past participial form is a contradiction; not so with class (c), without si:

(1) La finestra si e' rotta, ma non si e' rotta (contradictory)

The window is broken, but is not broken

(2) La casa e' bruciata (in parte), ma non si e' bruciata

(completamente) (non-contradictory)

The house burned, but it is not burnt

But the addition of si gives again a contradiction:

(3) La casa si e' bruciata, ma non si e' bruciata

The same can be seen in English by comparing (4) and (5):

(4) I burned the toast but it was not burned (contradiction)

(5) I burned the toast but it was not burnt

These facts indicate that the effect of the participle with si is to yield a final state; hence also that in (1) the participle gives the final state of having been broken (hence, in pieces), not the event of breaking. The causative (6) in Italian expresses a relation between a cause (something Gianni did) and the broken state of the window; but the English (7) expresses a relation between the cause and the breaking:

(6) Gianni ha rotto la finestra (=John caused the window to be broken)

(7) John broke the window (=caused the window to break)

2. All inchoatives can be non-finite complements of explicitly causative constructions with fare ("cause") (in which case si does not surface). Extending some observations of Centineo (1995), we may observe that the interpretations strictly follow the classification (a)-(c); specifically, V in (a) are always interpreted as transitive, those in (b) and (c) as intransitive, and in (c) secondarily as transitive:

(8) Gianni fa rompere la finestra (=Gianni made somebody cause the window to be broken)

(9) Gianni fa invecchiare il vino (=Gianni cause the wine to age)

(10) Gianni fa bruciare il libro (=Gianni caused the book to burn; Gianni made somebody burn the book)

3. Standard tests for telicity reveal that forms with si are telic, and so involve two events (Higginbotham 1998), whereas forms without si are atelic, involving only one:

(11) Il legno si e' bruciato in/*per un'ora

The wood burned in/for an hour

(10) Il legno e' bruciato in/per un'ora

The wood burned in/for an hour

The ungrammaticality of class (b) V with si, illustrated in (11), now follows from the fact that no final state can be associated with the process in question:

(11) *Il vino si e invecchiato

(Compare the English "The wine is (properly) aged," which does not mean that is has grown older, but rather that it has reached the state of being sufficiently old.)

In conclusion, the causative-inchoative alternation cannot be decided in favor of one or another form as primitive in the general case. Language-particular facts, including for instance the fact that the Italian participle with essere, unlike the English forms, always give a final state, must be taken into consideration.

'I saw the sunset yesterday.'

Timo Järvinen

Timo.Jarvinen@conexor.fi

The crucial issue for a syntactic model is how to obtain a correspondence between syntax and semantics. My framework is based on the dependency. Generally, the dependency models are inherently more "word oriented'' than constituent-structure models which use abstract phrase categories. A parallelism between syntax and semantics can not be obtained in word-based descriptions. Therefore, another syntactic primitive than a word is needed. The concept of nucleus, proposed by a French linguist Lucien Tesnière (1959), is such a primitive.

The nucleus is a minimal unit in syntactic description, and therefore it corresponds to a node in a dependency graph. A nucleus always contains both the structural and the semantic centre. Though in many cases these centres coincide in a same orthographic word, it is not always the case. A very familiar instance of a dissociative nucleus is a preposition and its complement.

Nucleus-based descriptions are advantageous, because the functional structure consisting of the relations between nuclei corresponds closely to the semantic organisation of the sentence. Tesnière maintained that there is never a structural connexion (i.e. syntactic relation) without a semantic connexion (1959, Ch. 22).

The notion of nucleus is not only valid for describing the structure of the sentence, but because it is the minimal semantic unit, it also provides a more adequate basis for lexicographic descriptions than a word-based lemma, as Tapanainen and Järvinen (1998) shows.

The linguistic phenomenon targeted in my presentation is discontiguity. Elements that appear both in continuous form as a single word and in discontiguous forms as separate words provide prima facie evidence in favour of the abstract notion of nucleus. For example, certain verbal complexes, such as the German verb untergehen, 'set', in specific environments appear obligatorily as discontiguous: Die Sonne geht im Westen unter. However, both contiguous and discontiguous manifestations are semantically equivalent.

Though nucleus-based descriptions are more abstract than distributional models, the level of abstractness is motivated by the observable linguistic phenomena rather than a hypothetical cognitive-conceptual organisation. The nucleus-based model does not describe underlying invariances or syntactic universals, but the surface-syntactic phenomena. Every language contains specific structures that are prone to a variation and change.

For example, in German the determiners are the main locus of morphosyntactic information, which mean that they are very tightly integrated into syntactic structure, and therefore intranuclear. There are also the fusional forms, such as im (in + dem), which are indicative of the degree of integratedness of the determiner to the prepositional nucleus. In English, the determiners do not exhibit such properties, and therefore the solution adopted in Järvinen and Tapanainen (1997) to treat determiners as extranuclear is adequate.

I argue that monostratal syntactic descriptions that use an abstract concept of nucleus as a primitive are motivated both theoretically and practically (see also Järvinen and Tapanainen, 1998). My presentation will focus on the basic types of nuclei, verbal complexes and prepositional nuclei. I illustrate how word order variations and complex long-distance phenomena are analysed in a semantically motivated way using nucleus-based structural descriptions.

Timo Järvinen and Pasi Tapanainen, 1998.

Towards an implementable dependency grammar. In Proceedings of the Workshop "Processing of Dependency-Based Grammars", (eds.) Sylvain Kahane and Alain Polguère, Université de Montréal, Quebec, Canada, 15th August 1998, pp. 1-10.

Timo Järvinen and Pasi Tapanainen, 1997.

Dependency Parser for English. Technical Reports of the Department of General Linguistics, University of Helsinki. 43 pp.

Pasi Tapanainen and Timo Järvinen, 1998.

Dependency Concordances. International Journal of Lexicography, (11)3 pp. 187-203.

Lucien Tesnière, 1957.

Éleménts de Syntaxe Structurale. Paris, Klincksieck.

Seppo Kittilämailto:(sepkit@utu.fi)

The prototypical transitive event has often been given the following definition: in a prototypical transitive event a volitionally acting agent acts on a patient (which is not coreferential with the agent) that undergoes the activity described by the sentence. The sentence pattern languages use to describe this kind of activity (e.g. she killed him) can be regarded as the prototypical transitive sentence of the language in question. This sentence pattern can in most languages also be used to describe other kinds of activity, the semantic transitivity of which is considerably lower than that of prototypical transitive actions (e.g. he has a book, he sees him). The structure of the prototypical transitive sentence can be manipulated or altered in many ways, which often change the meaning of the sentence in some way. These alternations can be divided into two groups according to their nature. Firstly, there are obligatory alternations. This means that there are some unprototypical actions or events and states that can never, because of their inherent (less transitive) semantics, take the morphosyntactic form of the prototypical transitive sentence. Second, there are prototypical transitive sentences that can, and under certain circumstances must, be marked differently from the prototypical transitive sentence (e.g., if there is a considerable change in the affectedness of the patient). These differences are not caused by some inherent semantic aspects of a particular action (type), but their use can only be explained in reference to a concrete situation. Many of the transitivity alternations have a semantic cause (e.g. reflexive, reciprocal, mediopassive, certain antipassives), but there are also alternations that change the morphosyntactic marking of the sentence (e.g. most passives, some antipassives, object-deletion) without affecting the inherent semantics of the action itself in any considerable way. Semantic (or pragmatic) transitivity features that in some languages require a specific morphosyntactic marking include volitionality or intentionality of the agent, identity of the agent (animate vs. inanimate instigators/causes of the action), affectedness of the patient, indviduation of agent or patient, tense, mood or aspect of the action, overall agentivity of the action (die vs. come), dynamicity or stativity of the action and coreference between agent and patient (=reflexive). In addition to these, sentences that describe habitual actions (e.g. birds eat worms) are often marked differently from those that describe concrete events or actions (the birds are eating worms in our garden). The morphosyntactic means that languages use to mark these differences include case-marking, morphology and agreement of the verb, addition/deletion of some clitic or particle, and word-order. Sometimes similar kinds of actions can/must be described by different lexemes or verb roots. In my presentation I am going to give an overview of the typology of some of the most important transitivity alternations and analyse briefly the underlying semantic reasons. I hope to be able to show that the traditional, purely morphosyntactic, approach to transitivity is insufficient and that the relationship between syntactic and semantic transitivity is not so direct as often assumed.

Marcus Kracht

II. Mathematisches Institut

Arnimallee 3

D -- 14195 Berlin

email: kracht@math.fu-berlin.de

For most theories of semantics cases are irrelevant at best. For these theories the ideal language has fully regimented word order but no morphology. This means that cases have to disappear at the interface from syntax to semantics, and likewise have to be inserted when passing from the semantics to the syntax. Syntax must provide a structural substitute for cases. This line of thinking is exemplified in the Minimalist Program, where cases are treated as uninterpretable features. In the MP, case is instrumentalized to get the word order facts of languages right, by assuming that they can trigger sometimes overt and sometimes covert movement of elements.

We argue here in line with Nordlinger (1998) that case is not only independent from syntactic structure, but that it can be a substitute for syntactic structure. Hence, there are languages, notably languages with rich case morphology, in which the syntactic structure is more flat than is currently assumed. Such languages include many of the Australian languages but also Finnish, German, Hungarian, Latin and Tagalog. (That eg Finnish word order is discourse constrained rather than syntactically constrained is argued for by Vilkuna (1989).)

In this presentation we shall provide a semantics in which cases are instrumental in providing the string-to-meaning translation. This will give additional evidence that cases survive at the interface. This semantics is based on the work of Vermeulen (1995), who proposed a new theory for the merge in DRT by introducing so-called referent systems. As we shall argue, the referent systems (in suitably modified form) will play the role of the argument structure in language. We assume that a functor (eg a verb) may select its arguments by means of a feature (eg case). This is stated in its argument structure. This feature must be overtly realized. If the functor selects several arguments, its argument structure consists of a sequence associating different variables with (possibly different) cases. For example, the meaning of the Latin verb {\sf dare} (to give) is an open formula containing three variables, the subject variable x, the object variable y, and the indirect object variable z. x is identified by nominative, y by accusative and z by dative case. These variables are in a sequence. Languages are parametrized as for whether they can associate the variables only following that sequence (English) or whether they can associate them by first match (Latin). This allows to generate free constituent order without scrambling. Order sensitivity is accounted for by a separate feature.

This model has several distinctive traits: while giving preference to the basic word order it allows certain languages to deviate from it rather drastically. In some sentences, the verb and its arguments may come in any order whatsoever, but the semantics still asigns the right meaning. Second, it allows languages to rely alternatively on word order, on constituency and on case to identify the meanings, or even on a mixture of these. Third, it can explain case agreement (eg within an NP) without any further mechanisms.

Nordlinger, Rachel: Constructive Case. Evidence from Australian Languages. CSLI Dissertations in Linguistics, CSLI, Stanford, 1998.

Vermeulen, Kees F. M.: Merging without Mystery or: Variables in Dynamic Semantics, Journal of Philosophical Logic 24(1995), 405 -- 450.

Vilkuna, Maria: Free Word Order in Finnish. Its Syntax and Discourse Function. SKS, Helsinki, 1989.

Pekka Lahdenmäki

pekka.lahdenmaki@helsinki.fi

The idea of binding theories is to capture the differences among referentially dependent elements and to restrict the class of possible antecedents of anaphors. In Finnish, third person possessive suffixes (Px) are clearly among the elements whose behaviour binding theories should explain: they must be bound inside a local domain.

Binding theories usually assume a hierarchical structure on which a control relation can be defined between a binder and a bindee. C-command relation in a phrase structure tree and o-command in an obliqueness hierarchy are well known examples. To my knowledge, no workable solution based on these hierarchies is available that correctly captures the behaviour of Finnish possessive suffixes. One reason for this is certainly the fact that there is no widely accepted way to derive phrase structure or even grammatical functions in Finnish.

Informally, a subject is always a possible binder of a Px, if semantic constraints do not intervene. However, sometimes an object may bind the same Px as well, resulting in ambiguity, as in example (1a). Example (1b) shows that this is not always the case:

(1a) Opettaja_i vei lapsen_j kotiinsa_i/j.

teacher.NOM bring.PAST.3SG child.GEN home.ILLATIVE.Px3SG

'The teacher took the child home.' (either the teacher's or the child's home)

(1b) Opettaja_i löi lasta_j laukullaan_i/*j.

teacher.NOM hit.PAST.3SG child.PAR bag.ADESSIVE.Px3SG

'The teacher hit the child with his/her bag.'

Even more unfortunate for an obliqueness-based account is that sometimes even an oblique may be the antecedent of a Px that is attached to an NP higher in the obliqueness hierarchy (i.e. non-oblique) as in (2):

(2) Opettajalla_i oli laukkunsa-i/*j

teacher.ADESSIVE be.PAST.3SG bag.NOM.Px3SG

'The teacher had his/her bag'

In this presentation I will argue that the correct place to account for these and other binding facts is a level of argument structure. There is independent motivation for a syntactic argument structure, which stems e.g. from studies of complex predicates in several languages. Argument structure is also becoming the locus of binding theory in constraint-based lexicalist theories, most clearly in Head-Driven Phrase Structure Grammar. The idea is roughly that argument structure is a list and the first member of such a list is a potential binder. In the analysis of the above examples (1a) and (2) nested lists are assumed, and nested lists yield additional first members of a list, thus additional potential binders.

Binding phenomena are widely used as diagnostic for phrase structure configurations or as part of the definition of the grammatical function 'subject'.One can question the validity of such tests if binding is better defined elsewhere. I think this is a welcome result. However, this does not mean that binding as a useful diagnostic tool in grammar development is lost. If one thinks linking in terms of sublinkings between a semantic structure, syntactic argument structure, grammatical function structure and surface string, then the way arguments are ordered at the level of syntactic argument structure (e.g. for the purposes to get the binding right) affects all the other levels as well.

Jaakko Leino

University of Helsinki

Department of general linguistics

jaakko.leino@helsinki.fi

Finnish is like most European languages in having a complement clause category which is used, among other contexts, as a complement of verbs of saying and verbs of cognition. The Finnish complement clause, että clause, sometimes occurs with a "dummy" pronoun se:

1. Ajattelen vain sitä, että ulkona paistaa aurinko.

think-1SG only it-PART that outside shine-3SG sun

'I only think [about the fact] that the sun is shining outside'

Also, the complement clause (että clause) may occur with a "dummy" adverb niin in contexts that are often superficially similar to those where the pronoun se is used:

2. Ajattelen niin, että huomenna varmaan paistaa aurinko.

think-1SG so that tomorrow probably shine-3SG sun

'I think that the sun will probably shine tomorrow'

As these examples suggest, the difference between se, että and niin, että is in part one of factivity. This is further supported by the fact that niin, että cannot occur as the complement of a factive predicate, whereas se, että can:

3. a. *Tiedän niin, että ulkona paistaa aurinko.

know-1SG so that outside shine-3SG sun

b. Tiedän sen, että ulkona paistaa aurinko.

know-1SG it-ACC that outside shine-3SG sun

'I know [the fact] that the sun is shining outside'

There is, however, more to this than just factivity. In my paper, I will argue that the external syntax of the että clause is ambiguos. It has a "nominal" and an "adverbial" reading. The former may express, for example, something that has been said, though of etc. The latter, in contrast, may express the way something was said, thought of etc. The pronoun se corresponds to the "nominal" reading, and the adverb niin corresponds to the "adverbial" reading. Thus, the words se and niin may be thought of as disambiguating elements.

The main point of my paper is that the pronoun se makes the että clause referential (perhaps even definite, but this is arguable). Therefore, the referent of the se, että construction is existentially presupposed. The existential presupposition has different semantic consequences in different contexts:

-- When the referent of the että clause is a state of affairs, the existential presupposition of the referent is tantamount to a factive presupposition: a complement that refers to an existing state of affairs is necessarily true.

-- When the referent of the että clause is a linguistic expression or a proposition, in other words a representation of a state of affairs, the reference is interpreted in terms of specificity. The existential presupposition of a linguistic expression, for example, is that the expression has been uttered (or written).

In sum, I will show that the external syntax of a sentential complement has radical consequences on the semantic interpretation of the complement. Also, I will explain why this is so in the case of the Finnish että clause complement.

AUTHOR'S NAME: Dianyu Li

AFFILIATION: THE AUSTRALIAN NATIONAL UNIVERSITY

TELEPHONE NO.: 61-02-62513549 or 61-02-62798219

E-MAIL: Dianyu.Li@anu.edu.au

POSTAL ADDRESS: DEPARTMENT OF LINGUISTICS

FACULTY OF ARTS

THE AUSTRALIAN NATIONAL UNIVERSITY

CANBERRA 0200

AUSTRALIA

Causative constructions in Mandarin are observed as unique in the present study in four respects: 1)Two of the four causative verbs, i.e. 'rang' and 'jiao' are speech act verbs which usually appear in permissive and instructive constructions but not causative constructions. When semantic bleaching occurs and aspectuality functions, they, instead, form causative constructions. 2) the other two causatives, i.e. 'shi' and 'ba' show similar semantic bleaching in forming causative constructions and, 'ba' in particular, demonstrates its semantic maintainance or bleaching in two distinct constructions: completive vs causative constructions. 3) The causative constructions formed by the two types of causatives are also observed as distinct in both tense domain patterns and clause structures, though their constituent structures look similar. 4) The type of the entity of the sentence subject, coupled with the type of the syntactic control, are considered as the two important factors for the causativity of the constructions.

The study has been conducted under the complex predicate approach.

Key Words: causative, instructive, completive, semantic bleaching, syntactic control, aspectuality

Anikó Lipták

HIL/University of Leiden

P.O. Box 9515

2300 RA Leiden

The Netherlands

***************************************************************************

Distributivity is usually defined relationally, after Choe (1987). This requires two terms: a set (distributive key), whose members are paired in a one-to-one fashion to instances of another term (distributive share). This definition is used in the recent Beghelli & Stowell 1997, too, which locates the distributive key and the share in two functional projections: in (the specifier of) DistP and ShareP respectively, DistP immediately dominating ShareP.

My talk will achieve two goals. 1. I will show with empirical data that the relational definition of distributivity is insufficient for describing distributivity in natural language. Besides the pairing odify the B & S (1997) framework based on Hungarian data. Since it eliminates uniform quantifier raising of any quantifier to any position, it is advantageous for Hungarian, where quantifiers move to specific positions, among others to DistP (Szabolcsi 1997). I will argue that while share-distributive items move to a DistP that dominates ShareP, predicate-distributive items raise to a DistP not dominating a ShareP. In the first case the distributivity feature is checked off by Shareo, in the latter by the head of the predicate. This structural difference explains the different behavior of different quantifiers with respect to negation, focusing of other constituents and scope interactions in Hungarian and other languages.

Anikó Lipták, University of Leiden/HIL

Malte Zimmermann, University of Amsterdam/HIL

We present a unified analysis of "binominal each"-constructions in a variety of languages (English, German, French, Finnish). Our analysis deviates from Safir & Stowell's (1988) standard analysis of English binominal 'each' in some important respects. It also accounts for typological differences between languages, thus furthering our understanding of the construction as a whole.

Safir & Stowell (1988) analyze the "binominal each"-construction in (1) as in (1'):

(1) The boys bought one book each.

(1')The boys<i> bought [DP one book<j> [QP PRO<j> [Q' each [null-object<i>]]]].

'Each' in (1) is treated as a quantifier. It takes the 'distributive' DP 'the boys' and a PRO controlled by the 'range'-DP 'one book' as arguments, and distributes elements of the latter over elements of the former. The complement of 'each' is a null-object anaphoric to 'the boys'.

S & S's analysis rests heavily on the assumptions that (a.) binominal 'each' is a quantifier; (b.) it follows the range-DP; (c.) it has to undergo LF-quantifier raising, so that the null-anaphor is locally A-bar-bound.

We propose a competing analysis based on the following observations/theoretical considerations: (a.) In German, French, and Finnish, the elements corresponding to binominal 'each' are morphologically more complex than the ordinary adnominal quantifier. This indicates that these elements might not be simple heads, but X'-projections. (b.) Binominal 'each'-elements can also be found in prenominal positions (German, Finnish), or in both positions (French). (c.) Our analysis does not involve LF-quantifier raising of the binominal phrase in order to license the anaphoric element.

We analyze binominal 'each' cross-linguistically as the head of a predicate small clause, whose subject is a PRO controlled by the distributive DP. The complement of the binominal element is anaphoric to this PRO. This explains the basic anaphoric behaviour of binominal expressions (Burzio 1986). The base-generated structure in all languages is:

(2) The boys<i> bought [DP one book [PredP PRO<i> [Pred' each [anaph]]]]

The binominal PredP predicates over the range-DP as a whole. It can be A-bar-fronted to a specifier position in the extended projection of the range-DP. This gives the prenominal word-order, e.g. in German:

(3)Die Jungen<i> haben [XP[PredP PRO<i> jeweils]1 X[DP[zwei Bücher]t1]]gekauft. the boys have each-anaph two books bought

'The boys have bought two books each.'

Fronting is an A-bar-process. This explains why the prenominal binominals block A-bar extraction (Rizzi 1990) from within the range-DP (Tellier & Valois 1993). To illustrate with a German was-für-split-example:

(4) Was<i> haben die Jungen (*jeweils) t<i> für zwei Bücher gekauft?

what have the boys each for two books bought

*'What kind of two books have the boys (each) bought.'

This way, word-order variation of binominals among languages and within one language is reduced to A-bar predicate fronting. Furthermore, our analysis enables us to compute the meaning of the relevant sentences from their S-structure in a strictly compositional fashion. LF-movement is not required for interpretive reasons. This makes our analysis more economical,

and it lessens the importance of a syntactic level LF in interpreting natural language expressions.

Malte Zimmermann

Universiteit van Amsterdam/ Holland Institute of Generative Linguistics (HIL)

c/o Romaanse Taalkunde, kr. 320

Spuistraat 134

1012 VB Amsterdam

The Netherlands

ph.: + (0)20-5254635

Lennart Loenngren

e-mail: lennart.lonngren@hum.uit.no

Semantic description could gain from a redefined and enlarged concept of synonymy. Traditionally, synonymy can only hold between two words belonging to the same part of speech, for example Sw. enorm - ofantlig, glo - stirra. Sometimes also "syntactic derivation", i.e. formation of abstract nouns, for instance ankomma - ankomst, god - godhet, is regarded as a case of synonymy.

The scope can be widened on the word level to include pairs like i - innehaalla, syfte - foer att, which are not normally considered synonymous. This is a way to account for paraphrases like Boken innehaaller sakfel - Det finns sakfel i boken; Syftet med hans resa var att knyta kontakter - Han reste dit foer att knyta kontakter. The same kind of equivalence relation holds between med and anvaenda in paraphrases like oeppna doerren med en mejsel and anvaenda en mejsel till att oeppna doerren. Contrary to this, med in Han diskuterade problemet med sin vaen has no synonym, the PP here being an argument of the verb. Med can be characterized as semantic in the first case, and syntactic in the second.

One further step is to accept synonymy between free and bound units, i.e. between words and morphemes. This is transparent in languages with morphological cases. For instance, the semantic instrumental case ending in Russian can be synonymous with use, as in Ivan rezhet khleb nozhom. 'Ivan cuts the bread with a knife', or with like as in the sentence On voet volkom. 'He howls like a wolf', but the same case ending in On komanduet polkom 'He commands the regiment' is syntactic. The genitive ending in Sw. Pelles klocka is synonymous with tillhoera, in mammas koettbullar with laga, but the same ending in pappas ankomst is syntactic.

When there is no case morphology synonymy can nevertheless be established between words and non-phonemic or totally implicit "case markers". For instance, in Sw. dansa en sommar and minnas en sommar there is synonymy in the first phrase with the preposition under, in the second phrase the connection is syntactic.

We have talked about synonymy between semantic units. But as a matter of fact there can also be "synonymy" between syntactic units. The relation holding between a subject/complement and its predicate, which is normally expressed by means of a case ending or non-phonemically, can be expressed by so-called delexicalized words, and cross-level synonymy may occur. For instance, verbs like utfoera, aatnjuta, undergaa, underkasta function as empty connectors between arguments and predicates; cf. Vera aer populaer - Vera aatnjuter popularitet. Even nouns can be empty connectors, as in foeremaalet foer hans laengtan, offret foer bedraegeriet; cf.: Han reste till Moskva - Maalet foer hans resa var Moskva.

This approach to semantic description has been outlined in my book "Valency structures in Russian ", Oslo 1998.

Tamara Loenngren

University of Tromsoe

Faculty of Humanities

Department of Russian

N-9037 Tromsoe

Norway

e-mail: tamara.lonngren@hum.uit.no

Causal constructions, the kernel of which is the logical proposition X is the cause of Y, have been investigated in modern Russian, for example by Klobukov (1984), but no attention has been payed to the syntax and semantics of such constructions in Old Russian. In this paper I will have

a closer look at the behaviour of one particular word occurring in causal constructions, namely the Old Russian adverb abie.

This word, which can be traced back to Sanscrit ahnaya and which has equivalents in the Greek texts underlying most of Old Russian religious literature, is not known in modern Russian; it disappeared already in the 17th century. But in Medieval texts, especially hagiographic, it plays en important role. The reason is that its meaning, approximately 'at once', 'suddenly', sometimes 'soon', was well apt to render "the reality of unreality", miracles. It serves as a transition signal between the spheres of modus and dictum. For instance, the saint blew at the eyes of the blind, and "abie" the latter regained his vision. Or a fettered man uttered the name of the saint, and "abie" the fetters fell off.

Syntactically we are talking here about paratactic constructions. In my material half of them are asyndethic, the rest contain the conjunction i "and". The word abie has quite a fixed position: it occurs in the second clause, initially or after the conjunction i. This clause is the modus part of the construction. There is a certain amount of variation in the syntax and semantics of these parts, which effects also the meaning and function of the word abie.

I have investigated the behaviour of abie using a material which comprises 139 examples . These are taken from Old Russian texts which have been only recently discovered and available, namely the three-volume "Sobornik", an anthology of vitae of Old Greek saints, compiled and edited by Nil Sorsky (ab. 1433-1508). The publication of this unique autograph is now being prepared at the University of Tromsoe.

Satu Manninen

University of Edinburgh

Recent work puts forward the hypothesis that adverbials are merged into the unique speficier positions of functional and of light v heads (cf. Alexiadou 1997, Laenzlinger 1998, Cinque 1999, Manninen 1999). These hypotheses are largely embedded within the framework of antisymmetry and of the Linear Correspondence Axiom (LCA) of Kayne (1994), building on the idea that adverbials contain features which allow them to be licensed in a specifier-head (checking) relation with a matching functional or light v head. Thus, even though the licensing of adverbials is a syntactic phenomenon, it is connected to and reflects the semantic properties of functional and of light v heads.

In this paper, I investigate the licensing and distribution of VP adverbials (eg manner, place, and time adverbials), concentrating on the relation between hierarchical structure and linear word order. Two claims which are of importance to the theory of adverbials are made: firstly, I argue that VP adverbials are licensed via semantic feature checking against an appropriate light v head, that is, as specifiers of an appropriate light v head. I further argue that light v heads are hierarchically ordered with regard to each other so that a stricthierarchical structure is formed; the hierarchical structure (ie the hierarchical ordering of light vPs) is determined by the properties of UG and is always the same across languages. Secondly, the requirement that hierarchical structure directly reflects linear word order and vice versa is relaxed. This claim is based on the observation that, although in some languages hierarchical structure always directly corresponds to linear word order and vice versa, in others there is no constraint on the mutual linear ordering of sentence final VP adverbials. The fact that VP adverbials are allowed to permute with regard to each other is shown by the Finnish examples in (1) and (2) below:

(1) a. Sirkku juoksi ontumalla rannalla

ran with a limp on the beach

(1) b. Sirkku juoksi rannalla ontumalla

ran on the beach with a limp

(2) a. Sirkku juoksi ontumalla keskiyöllä

ran with a limp at midnight

(2) b. Sirkku juoksi keskiyöllä ontumalla

ran at midnight with a limp

Within in the LCA, hierarchical structure must always directly correspond to linear word order: crucially, if A is higher up than B in hierarchical order, then A precedes B in linear word order. Thus, assuming that VP adverbials are merged into the specifier positions of light v heads and hierarchical structure, ie the hierarchical ordering of light vPs, is fixed, then either (1a) or (1b), or (2a) or (2b), must reflect the original "base generated" structure while the other one reflects a derived structure and is a result of movement. However, it is difficult to find theoretical justification for such an analysis; nor is it supported by empirical evidence from individual languages. In particular, if sentences such as (1a-b) and (2a-b) differed from each other in the sense that one of them reflects the original "base generated" structure while the other one is a result of movement, then one would expect to find differences in their respective interpretation (eg one would expect one of the sentences to have syntactically marked word order and to involve focussing).

In this paper, I will show, by using examples from various languages, that hierarchical structure is always fixed ; however, evidence from individual languages seems to contradict the view that hierarchical structure directly corresponds to linear word order. I will argue, contra Kayne (1994), that the LCA cannot establish a linear ordering between elements appearing in the specifier of vP positions. This explain why VP adverbials are able to permute with regard to each other. I will further argue that the availability of this mechanism is a property that UG allows individual languages to have, so that some languages allow variation in the mutual linear ordering is their VP adverbials while others do not.

References:

Alexiadou, Artemis (1997) Adverb Placement: a Case Study in Antisymmeric Syntax. John Benjamins.

Cinque, Guglielmo (1999) Adverbs and Functional Heads: a Cross-Linguistics Perspective. Oxford University Press.

Kayne, Richard (1994) The Antisymmetry of Syntax. The MIT Press.

Laenzlinger, Christopher (1998) Comparative Studies in Word Order Variation: Adverbs, Pronouns and Caluse Structure in Romance and Germanic. John Benjamins.

Manninen, Satu (1999) Manner Adverbials and the Structure of Finnish Sentences: a Minimalist Approach. Ph.D Thesis. University of Edinburgh.

Sergio Meira

meira@ruf.rice.edu

------------------------------------------------------------------------------------------------------------------

Very little attention has been given to Amazonian languages in the literature on 'split-S' or 'active-stative' languages; yet a growing number of Amazonian cases of such languages is being reported. The aim of the present paper is to show two case studies of Amazonian languages that

do seem to belong to the active type at first sight, but which reveal unsuspected complexities as one probes deeper into their grammar.

In the Tupi-Guaranian language family, there are several languages described as active-stative; in fact, one of the first languages given as an example of such systems was precisely Guarani (Klimov 1974). However, Rodrigues' work on Tupinamba' reveals that the two conjugations of intransitive verbs are actually a broader grammatical phenomenon which interacts with the verb-noun itself, making it less clear whether or not Tupi-Guaranian languages are actually comparable to the better known North American active-stative languages (Muskogean, Siouan, Panoan, etc.).

Many languages in the Cariban family show morphological active-stative properties in their verbal conjugations. However, the system lacks the semantic consistency of its North American counterparts. Upon closer inspection, the active category is seen to be almost entirely composed of reflexivized verbs, all derived, synchronically or diachronically, from transitive verbs by means of a process that, for some historical reason, did not entail the loss of the transitive A-marking conjugation. Thus, the morphological active properties of the language are actually properties of the reflexivization (detransitivization) process, not having necessarily much to do with the semantic motivations that underlie other active systems.

Klimov 1974. On the character of languages of active typology. Linguistics no. 131, 11-25.

P'urhepecha, or Tarascan, is an isolated language spoken in the state of Michoacan, Mexico. This suffixed language has 29 morphemes that make reference to space and may combine with a large number of voice morphemes. The descriptive analysis undertaken assigns an abstract meaning to each space morpheme and to identify the specific meaning of a space morpheme it is necessary to identify the phrase with which it is linked, in other words if the space morphemes refers, for example, to a 'frontal space', this frontal space will be interpreted as a person's face, a TV screen if a person or a TV is mentioned in the phrase with which the morpheme is linked. The function of this phrase is determined by the class to which the space morpheme belongs to, or by the given voice morpheme present in the verb.

Cristina Monzón

Colegio de Michoacán

Martínez de Navarrete 505

Zamora, Michoacán, México

e-mail cmonzon@colmich.edu.mx

Dr. Diane C. Nelson email: d.c.nelson@leeds.ac.uk

Dept. of Linguistics & Phonetics phone: +44 (113) 233-3563

University of Leeds fax: +44 (113) 233-3566

Leeds, England LS2 9JT

http://www.leeds.ac.uk/linguistics/staff/diane/Welcome.html

The problem of linking has yielded several key hypotheses within the generative literature (Universal Alignment Hypothesis, Perlmutter & Postal 1984; Uniformity of Theta Assignment Hypothesis, Baker 1988). These proposals map particular semantic roles to particular structural positions; for example, patients are taken to occupy underlying object positions, regardless of surface syntax. In this way, thematic relations are taken to directly reflect deep structural relations, which are in turn analysable from thematic relations in a given structure. In this paper it is argued that a single morphosyntactic process in Finnish yields two distinct surface outputs which share a single underlying syntactic representation predictable from the semantics, thus validating these linking hypotheses.

Pesetsky (1995) notes that English pairs like the following pose problems for the above proposals, since the same role (Experiencer) appears to be arbitrarily linked with a syntactic subject (a) and a syntactic object (b):

(a) John worried about the television set.

(b) The television set worried John.

Pesetsky argues that two distinct roles are represented here, which differ slightly in their interpretations. Specifically, he explains their distinct syntax and semantics by associating the Experiencer-object in (b) with an implicit causative marker.

Finnish appears to provide strong evidence in support of Pesetsky's analysis; a single affix is associated with Experiencer alternations as in the above English examples, and simultaneously functions as a causative marker. Both of these processes in Finnish involve the affix /-tt/. In the

first case, this affix productively derives causatives from transitive verbs, which has the effect of introducing a nominative agent/causer and demoting the underlying nominative subject to an optional oblique:

Hän kirjoitti omaelämänkerta-nsa.

she wrote autobiography-acc-Px3

'She wrote her autobiography'

Hän kirjoitu-tti omaelämänkerta-nsa (kirjailija-lla)

she wrote-caus autobiography-acc-Px3 (writer-adessive)

'She had (a writer) write her autobiography'

A second, morphologically identical process applies productively to a class of psych predicates ('FLIP-verbit'), yielding a partitive Experiencer 'subject' and an optional nominative causer:

Minä suri-n.

I(nom) grieved-1s

'I grieved'

Minu-a sure-tta-a (koira-ni kuolema).

I-part grieve-tt-3s (dog-Px1s death(nom))

'(My dog's death) grieved me'

These constructions raise some important questions relevant to the linking hypotheses mentioned above. Do they represent distinct morphological processes (Hakulinen & Karlsson 1979)? If they share the same morphological process, can the underlying structure of the input to this rule be predicted from the semantics? If the UAH, UTAH and Pesetsky's proposals are correct, the thematic relations in these examples should reflect underlying syntax in a systematic way.

This appears to be true: in both constructions, the /-tt/ affix introduces an additional role of causer/agent into the derivation. This is shown by the fact that both causatives and (at least most) intransitive Experiencer verbs with /-tt/ may undergo impersonal passivisation, a process which targets the agent role and excludes intransitives taking inanimate patient/themes as subject. Moreover, the partitive subjects are expected to be underlying experiencer objects. The data show that these 'subjects' fail most tests for subjecthood in Finnish: (a) they cannot bind anaphoric possessive affixes in adjoined clauses; (b) under negation, they remain in the partitive case while the nominative theme remains in the nominative; and (c), they are subject to word order constraints distinct from genuine nominative subjects.

Fabrice Nicol, maitre de conferences, Université de Paris-III, Sorbonne nouvelle, France

address : UFR du monde anglophone, rue de Santeuil, 75005 Paris

e-mail: FABRNI@aol.com ; fnicol@univ-paris3.fr

**********************************

Guéron (1985), Vergnaud and Zubizarreta (1992) propose a set of syntactic constraints to account for the distribution of the French inalienable possession construction (henceforth IPC) exemplified in (1) :

(1) Il lui a cassé le bras.

He her-clitic-dative broke the arm.

'he broke her arm'

In (1), the definite article is used in French, whereas it would be out in English. The locality property of this construction is derived from the Binding Theory in Guéron's framework, and from the theories of small clauses and syntactic chains by V & Z.

It appears that neither theory covers the full range of empirical data. In almost all the examples given by the authors, the only construction considered is the clitic construction, with a dative pronominal clitic as in (1). There are cases, though, in which a full dative noun phrase is legitimate. Casser (to break) is one of the verbs that accept the full-NP IPC, illustrated in (2) :

(2) Il a cassé le bras à Marie.

He broke the arm prep-dative Mary (possessor)

'He broke Mary's arm'.

There are remarkably stringent and so far unexplained conceptual constraints on (2). First, the possessor must be affected by a material medium (hand, weapon, etc.), in a bounded process. This is necessary but not sufficient : the verb must be a verb of contact, surface alteration, or partial entraining (following Jackendoff's 1990 typology). Furthermore, verbs acceptable in (1) but not in (2), satisfy the conjoined conditions (i) and (ii) :

(i) the limb is affected in a ±bounded process;

(ii) constraint (i) is relaxed for verbs of non-affecting contact (either static contact or inchoation of contact, and verbs of brushing/skimming (past)).

Finally, verbs that license neither the clitic construction nor the full-NP construction are:

1 all psychological and perception verbs (imagine, invent ; see, watch)

2 strikingly, verbs of impact (the strike class in Jackendoff's terms).

The above-mentioned constraints are accounted for in Jackendoff's (1991) framework. They point to an interesting correlation between semantic structure and cliticization. Assuming that (2) is the base structure for (1), it so appears that cliticization, a syntactic process, voids/relaxes a number of semantic constraints on French IPCs, though not all of them. This may be attributed to dative Case realization: morphological marking on the clitic or lexical marking through the preposition à.

Constraints on French IPCs therefore clearly undermine reductionist theories of theta-role assignement in (bare) phrase structure, as developed by Hale and Keyser (1993). These theories are bound to advocate strict autonomy of movement theory from semantic constraints; however clitic IPCs support Hornstein's (1999) recent proposal that semantic role assignment and movement theory are to be correlated, and Jackendoff's (1990, 1997) 'parallel model' for syntactic and semantic derivations.

Esa Penttilä

DO A/AN X as an Idiomatic Construction in English

Since idioms have usually been regarded as unsystematic anomalies of language, grammarians have rarely felt a need to deal with them in any detail. Idioms are simply viewed as idiosyncratic non-complex wholes with no systematically analyzable structure. Construction Grammar (CG), however, provides an opposite view of portraying idioms. According to CG, the basic meaning units of language are constructions, i.e., complex syntactic structures combining meaning and form in such a way that a large part of the meaning of each construction is regarded as independent of the particular items involved and is instead ascribed to the structural pattern of the locution (see, e.g., Goldberg 1995, Kay 1995). An approach like this leads to taking idioms seriously when describing a language. After all, idioms are by definition complex linguistic meaning units and in that sense prototypical constructions (see, e.g., Jackendoff 1997). Therefore, instead of treating them as marginal, CG brings idioms to the fore and views them as more or less central to the linguistic description; by analyzing them it is possible to gain a clearer insight into the structure of the whole language (see, e.g., Fillmore et al 1988, Jackendoff 1997).

In my paper, I will discuss the nature of idiomatic constructions in English within the framework of CG. More specifically, I will deal with the DO A/AN X construction in which X refers to a proper name; the most expected proper name in this context is the name of a person, but there are various other possibilities available as well. In my paper, I will analyze the systematic structure of this construction. The construction is discussed with the help of examples taken from the British National Corpus (BNC). Other constructions briefly touched upon in this paper include discontinuous idioms, such as bring X to light and lead X a merry chase (see Fraser 1970).

Fillmore, Charles J., Paul Kay, and Mary Catherine O'Connor (1988) Regularity and Idiomaticity in Grammatical Constructions: The Case of LET ALONE. Language 64: 501-538.

Fraser, Bruce (1970) Idioms Within a Transformational Grammar. Foundations of Language 6: 22-42.

Goldberg, Adele E. (1995) Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press.

Jackendoff, Ray (1997) The Architecture of the Language Faculty. Cambridge, MA: MIT Press.

Kay, Paul (1995) Construction Grammar. In Jef Verschueren, Jan-Ola Östman and Jan Blommaert (eds.), Handbook of Pragmatics: Manual. Amsterdam: John Benjamins. 171-177.

Marja Pälsi

University of Helsinki

Marja.Palsi@helsinki.fi

Most of the work in Construction Grammar has been done on English. Finnish exemplifies a very different kind of language, agglutinating, with its 14 nominal cases, four of which (nominative, genitive, accusative and partitive) participate in expressing the aspectual boundedness opposition on the object NP.

In my paper I will present a grammar fragment of the Finnish object, using a formalism based on Fillmore and Kay 1996. I want to show that the role of the object and that of the resultative phrase are much more independent in Finnish than in the corresponding English sentences, and therefore there is no strictly independent Resultative Construction in Finnish.

Goldberg 1995 posits a Resultative (1a,b) Construction for English:

1.a He wiped the tools clean.

1.b He ate himself sick.

Finnish has similar sentences, too (2a and 2b corresponding to 1a and 1b, respectively):

2.a Nuiji pihvi pehmeäksi!

Pound - the steak (object: nominative; bounded) - tender (translative) !

2.b Hän söi itsensä kipeäksi.

S/he - ate - her/himself (object: genitive; bounded) - sick (translative).

They follow the same pattern: the syntax "(S) V O X(directional phrase)" has the meaning 'S does V (in (2a), to the normal valency object of the verb; in (2b), to the prototypical or contextually

inferrable object), which causes O to move to (state) X'.

For (2a), we do not need a separate construction to derive that meaning. Verbs - being constructions, too - together with general Finite Clause Object Constructions take care of the semantic relation between verb and object. A bounded object gives the sentence a resultative meaning. A Directional Case Construction associates the directional phrase with the object.

For (2b), a separate Constructional Object Construction is needed, to make sense of the object, which, indeed, cannot be the valency object of the verb, but is the object of the whole construction instead.

Even though superfluously, the Constructional Object Construction can be used as an alternative way to derive the meaning of (2a), too. Two mechanisms overlap here, and produce the same sentence.

There are restrictions on the use of the Constructional Object Construction. Transitive verbs cannot have their valency object overtly expressed, and the verb must be used in its literal sense.

The object tends to be bounded.

The Constructional Object Construction is also instantiated by various idiom constructions, one of which is exemplified by (2b).

The Constructional Object Construction and verb lexeme constructions unify with general object and subject constructions, in which boundedness plays an important role.

Thus, my contention is that we need to keep apart the objects of different constructions (here the verb lexeme construction and the Constructional Object Construction). Moreover, this methodological necessity is not limited to this particular case.

Another point to note is that the constructions I will be discussing are best related to each other in a gradient, or overlapping fashion.

References:

Fillmore, Charles and Paul Kay, ms. Construction Grammar. CLSI Lecture Notes, Stanford. University of California, Berkeley. Latest version: 1996.

Goldberg, Adele E. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press.

Joseph Quer

Background. It has been sometimes observed (cf. LeGrand 1975, Davison 1980, Carlson 1981, Dayal 1995, 1998) that despite the well-established incompatibility of free choice (FC) readings with episodicity, English FC any occasionally appears in non-modal, episodic contexts. Characteristically, those cases involve modification of the any DP by a relative clause (RC), as in (1) (from Dayal 1995). LeGrand (1975) discussed this sort of examples under the label subtrigging.

(1) a. John talked to any woman *(who came up to him)

b. Any man *(who saw the fly in the food) didn't eat dinner

Outline. On the basis of disambiguating parallel evidence in Catalan and Spanish I argue that the cases such as (1) where any is subtrigged do constitute modal contexts in that they involve the conditional-like interpretation tied to characterizing past sentences (see Krifka et al. 1995). In English, past morphology is ambiguous between perfective and imperfective readings, which blurs the distinction between episodic and generic/characterizing sentences, respectively. I show that aspectual distinctions play a decisive role, as is expected in conditional readings. Catalan and Spanish display such aspectual distinctions overtly with past tenses. The basic claim is that cases like (1) are examples of conditional-like sentences headed by an operator like GEN(eric) or HAB(itual) where the FC modified by the RC contributes the antecedent/restrictor, therefore they are modal contexts. Most of the instances of subtrigged FC discussed in the literature are identified here as characterizing sentences in the past, thus being amenable to the conditional interpretation associated with generic/characterizing statements. The contrast in Catalan (2) corresponding to (1a) clearly indicates that subtrigged FC readings are only viable with imperfective past morphology, which lends support to the hypothesis that such occurrences of FC do not constitute exceptions to the anti-episodicity property of FC.

(2) a. *Va parlar amb qualsevol dona que se li va apropar

b. Parlava amb qualsevol dona que se li apropava

Under the term subtrigging, though, another set of examples such as (3) has been included that cannot be reduced to the same account just sketched.

(3) At the end of his speech, the president thanked any volunteer who had taken part in the rescue operation

I take such cases to be genuine instances of subtrigged FC, for the main predication can be in a past episodic tense in Catalan (cf. (4)). I show that they are also related to modality in the sense that the domain of the individuals denoted by the subtrigged any DP is defined in a different model than the default one where the main clause is evaluated. Catalan marks this overtly with the FC determiner and subjunctive morphology on the verb in the RC (mood shift indicates model shift, cf. Quer 1998), English solely with any. In a nutshell, the FC description is interpreted de dicto not in the epistemic model of the speaker, but rather in the implicit model anchored to the subject of the main predication, hence its reported-speech flavour.

(4) Va felicitar qualsevol voluntari que {hagués/*havia} participat en l'operació de rescat

Languages with overt aspectual and mood oppositions are thus argued to provide us with empirical arguments in favour of the inherent modal nature of FC, also in subtrigging contexts.

Jarno Raukko

University of Helsinki

If we examine, say, a polysemous verb and try to classify and describe its various meanings, we might be tempted to think that a semantically plausible classification would match a classification of the verb's valency patterns (or the like). We might want to accept this view because we would like to see syntax and semantics interact and agree with one another - also when polysemy is concerned. I am not aiming at throwing this view out; it is a reoccurring pattern and always useful to keep in mind; but I wish to challenge its absolute primacy and point out a few important counterprinciples, with illustrative examples.

A related macro-level phenomenon has to do with zero-derivation: at least in languages like English we can have polysemy across word-class boundaries, if we allow the definition of polysemy to go across them. It will be interesting to look for evidence that language users might feel a verbal use and a nominal use to be semantically closer to one another than two verbal uses or two nominal uses of the same word. A workable example is _date_: informants more easily relate "I'm dating her" and "She's my date" than either of the two to "This item dates back to the Middle Ages" or "What is today's date?". (My material comes from experiments using questionnaires on one polysemous word at a time.)

Coming back to polysemous verbs and their valency patterns vs. polysemy patterns, I am mainly looking at the English _get_ 'receive, become, arrive, start, may, must, understand, answer, etc.' and the Finnish _pitää_ 'like, must, hold, keep, have, organize, consider, resist, etc.'. It is clear that some relevant semantic distinctions - such that e.g. informants saliently come up with in sorting tasks - correlate with syntactic divisions. But I wish to show that in the case of both verbs,

some semantic units (meaning types) go across syntactic patterns. And if certain meaning types allow for flexible syntax, meaning differences cannot be merely detected by looking at differences in sentence structure.

Ruth H. Sanders and Alton F. Sanders

We (one linguist, one computer scientist) have designed a comprehensive syntactic parser, 'Syncheck,' able to identify virtually all correct syntactic structures of written German. It is intended as the basis for a writing aid for intermediate and advanced learners of German. But a writing aid should be able to give information about errors, as well as identifying whether or not a given sentence is acceptable. Therefore we have designed an 'error grammar' to work on top of the syntactic parser. Where the parser itself identifies correct structures, the error grammar identifies incorrect structures so that, having found them, it can give a specific error message to the student. Since students' errors are semantic as well as syntactic, we have attempted to categorize patterns of the most common semantic errors as well, but by using syntactic methods. We will provide examples of these semantic errors. The 'language' used by the linguist to specify the patterns assumes a comprehensive German lexicon (made available to us for research purposes from Lingsoft Inc. of Helsinki through its designers at the University of Helsinki). (These patterns are treated as regular expressions although the set of languages described by the patterns is not truly regular because of the potentially infinite alphabet.) We use finite state methods to screen the input German text for these patterns, and produce an appropriate error message to the user if they are found.

......................................

Ruth H. Sanders

Professor of German, Miami University

Oxford, Ohio 45056 U.S.A.

phone: 513-529-2526, fax: 513-529-1807

Alton F. Sanders

Professor of Systems Analysis, Miami University

Oxford, Ohio 45056 U.S.A.

Mari Siiroinen

Department of Finnish

University of Helsinki

In my paper I shall discuss some of the ways to code the experiencer in Finnish. I shall also say something about the discourse functions associated with these different ways of coding.

As is well known, the transitive clause is particularly suitable for giving a linguistic expression to events that feature two participants: an agent and a patient. The agent acts consciously and volitionally, causing the patient to undergo a change. The agent is coded as the subject, the patient as the object. The transitive clause is, however, also used to code a variety of less prototypically transitive events and states. These can be spatial relations, possessive relations etc. (Taylor 1989:206-217, Croft 1991: 198-212)

In terms of less prototypically transitive events, mental events are a case in point. They have two participants with clearly distinct roles, namely the experiencer and the stimulus. But which of the two is to become the subject? This is in fact not a simple issue to resolve, and there are a variety of structures to code mental events. Finnish, for instance, has these:

1. a. Minä pelkää-n si-tä.

I fear-1SG it-PTV

b. Se pelotta-a minu-a.

It frighten-3SG I-PTV

The experiencer is the subject and the stimulus is the object with 1a.The stimulus is the subject and the experiencer is the object with 1b. These two kinds of clauses express two different perspectives.

There is still a third alternative. Most of the emotive verbs with an experiencer object (1b) can be used like this:

2. a. Minua pelottaa tulevaisuus.

I-PTV frighten-3SG future

b. Minua pelottaa.

I-PTV frighten-3SG

Here, the experiencer occupies the standard position of the subject (before the verb). It is not the subject, however: it has the object case (partitive) and the verb does not agree with it. If present, the stimulus is the subject: it is in the normal subject case (nominative) and the verb agrees with it. But this construction is often subjectless (2b). I shall refer to this special construction as the experiencer construction.

In my paper I shall discuss the syntactic and semantic properties of the experiencer construction. Based on corpus studies, I shall also outline the discourse functions of the experiencer construction and other constructions typically used for coding emotive expressions.

References:

Croft, William. 1991. Syntactic Categories and Grammatical Relations. The Cognitive Organization of Information. Chicago & London: The University of Chicago Press.

Taylor, John R. 1989 Linguistic Categorization. Prototypes in Linguistics Theory. Oxford: Clarendon Press

Ida Toivonen

Department of Linguistics

Stanford University

Particles such as in, up, away in the Germanic languages are relatively well studied in the linguistic literature (Den Dikken (1995), Svenonius (1994), Vinka (1998)). It has been noted that in some languages the particles must follow the direct object (Danish), in some languages they must precede the direct object (Swedish) and in yet other languages both positions are allowed. Several purely syntactic accounts have been offered (Svenonius (1994), Den Dikken (1995)), but recent studies have shown that the semantics of the relevant constructions is relevant as well (Vinka (1998), Toivonen (to appear); see also Åfarli (1985)). This paper explores further how semantics interacts with syntax in the verb-particle constructions. The focus will be on Swedish.

It has been claimed that Swedish particles must precede the direct object. Although there is certainly some truth to this generalization, there are counterexamples.⁽¹⁾

• Peter lämnade (kvar) boken (kvar).

P. left ('in.place') book.the ('in.place')

'Peter left the book in place.'

• Johan förföljde (*ner) mannen (ner) till stranden.

J. followed (*down) man.the (down) to beach.the

'Johan followed the man down to the beach.'

When both orderings are possible, there is often a clear difference in meaning between the two, as discussed in Teleman et al. (to appear) (see also examples (3-4)). It has previously been argued that particles that follow the direct object denote directionality, whereas particles preceding the direct objects are resultative (Toivonen (to appear)). If a particle cannot denote directionality, it never appears after the direct object. This generalization, however, cannot alone account for all the syntactic facts, such as the fronting of particles and placement of reflexive pronouns.⁽²⁾

• Många skriver in sig i Viktväktarna.

many write into self in Weight.Watchers

'Many people join/sign up for Weight Watchers.'

• Hon skriver sig in i akademien.

she writes self into in academy.the

'She writes her way into the academy.'

Example (4) implies that 'she' will be 'in the academy' as a result of her writing, but 'in' still follows the object. In a more complete account of the syntax of the verb-particle constructions, finer distinctions in theta-roles must be made and idiomaticity must also be considered.

Although several of the semantic facts affect the word order, it is not necessary to place the whole burden of explanation on the phrase-structure. A model where the lexicon, the argument structure, and the syntax can be analyzed separately will help to analyze differences and similarities between different dialects and languages.

References:

Åfarli, Tor A. 1985. Norwegian verb particle constructions as causative constructions. Nordic Journal of Linguistics 8: 75-98.

Den Dikken, Marcel. 1995. Particles. On the Syntax of Verb-Particle, Triadic, and Causative Construction. New York: Oxford University Press.

Telemann, Ulf, Staffan Hellberg and Erik Andersson. To appear. Svenska Akademiens Grammatik.

Toivonen, Ida. To appear. Swedish place expressions. Proceedings of NELS 29.

Svenonius, Peter. 1994. Dependent Nexus. Subordinate predication structures in English and the Scandinavian languages. PhD diss., University of California at Santa Cruz

Vinka, Mikael. 1998. Two distinct verb particle constructions in Swedish. Proceedings of ConSOLE 6.

ÅT = Tidningen Åland

Ildiko Toth

Third person plural agreement marking has been shown to be able to license the so called "quasi-universal" and "quasi-existential" arbitrary readings for null subjects in pro-drop languages. (cf. Cinque (1988) for introduction of these two terms, Suñer (1983) and Jaeggli (1986) and Belletti & Rizzi (1988) for syntactic analyses of these readings). Quasi-existential reading is exemplified in (1).

(1) Pro cantan en la casa vecina.

Arb sing-3PL in the house neighbouring

'Arb are singing in the house next door.'

In (1) the interpretation of arb is synonymous with an existentially quantified NP subject. The third person plural agreement is purely formal. The sentence is also true if it is only one single person who is singing next door. The sentence has a specific time reference.

The quasi-universal reading arises when 3PL agreement appears on the verb and the tense specification of the sentence is not specific.

(2) a. Li, odiano gli stranieri.

there hate-3PL the foreigners

'There, they hate foreigners.'

b. In questo paese pro temuto il terremoto.

in this country fear the earthquake

'In this country people fear the earthquake.'

In my paper I will first give a short summary of the data concerning inflected infinitivals in Hungarian. I will group constructions with inflected infinitives into two groups according to the thematic properties of the matrix predicate. In the first group the main predicate has an experiencer argument of its own in addition to the subject infinitival clause. This is exemplified in (3).

(3) János-nak_i kellemetlen [ec_i az igazságot bevalla-ni-a]

John-DAT unpleasant the truth-ACC admit-INF-3SG

'It is unpleasant for John to admit the truth.'

In the second group the matrix predicate is monadic. It takes a sole argument, an infinitival clause containing an inflected infinitive.

(4) Ilyen késõn már muszáj [valaki-nek otthon len-ni-e].

so late already must someone-DAT home be-INF-3SG

Someone must be at home at such a late hour.

Given that infinitives can have 3PL agreement marking (to license their subject and the experiencer argument of the matrix predicate if it has one), the question can be raised whether quasi-existential and/or quasi-universal readings are available with 3PL marking on the infinitive. This is the main topic of my paper.

Given the two classes differentiated above, there are four possibilities:

¶ Quasi-existential reading for null experiencer arguments with third person plural marking on the infinitive.

· Quasi-universal reading for null experiencer arguments with third person plural marking on the infinitive.

¸ Quasi-existential reading for the null subject of the infinitival subject clause with third person plural marking on the infinitive.

¹ Quasi-universal reading for the null subject of the infinitival subject clause with third person plural marking on the infinitive.

I will take these possibilities one by one. Semantic/thematic properties can be invoked to exclude possibility ¶, since it can be independently argued that quasi-existential reading is not available to experiencer arguments. In case of ·, ¸, and ¹, however, I will argue that neither thematic/semantic nor syntactic considerations can account for the fact that neither the quasi-universal nor the quasi-existential reading is available in case of inflected infinitives with 3PL agreement marking. In other words, if previous analyses of quasi-existential and quasi-universal readings are on the right track, we expect (5b), (6b) and (7b) to be grammatical on the relevant readings, contrary to facts.

(5) a. Ebben az országban pro fél-nek [a vulkánok közelében lak-ni].

this-in the country-in fear-3PL the vulcanos near live-INF

'In this country people fear to live near vulcanos.'

b. #Ebben az országban félelmetes pro_i [ec_i vulkánok közelében lak-ni-uk]

this-in the country-in frightening vulcano near live-INF-3PL

'In this country it is frightening for the people to live near vulcanos.'

(6) a. Pro kopog-tak. Bizonyára áll-nak az ajtó eltt.

knock-3PL Surely stand-3PL the door in-front-of

'Someone(s) knocked. Someone(s) must be standing in front of the door.

b. #Pro kopog-tak. Áll-ni-uk kell az ajtó eltt.

knock-3PL stand-INF-3PL must the door in-front-of

'Someone(s) knocked. Someone(s) must be standing in front of the door.

(7) a. Itt bizonyára sokat dolgoz-nak.

here surely much work-3PL

'Here, people must work a lot.'

b. #Itt sokat kell dolgoz-ni-uk.

here much must work-INF-3PL

'Here, people must work a lot.'

I will propose an explanation of these contrasts based on the lexical specifications of each agreement morpheme. In particular, the lexical feature composition of 3PL agreement suffix (in null subject languages) appearing on finite verbs is compatible with unspecified (generic) person while the 3PL agreement suffix of infinitivals is only compatible with referential subjects. Such a conclusion is not surprising given that in Hungarian two different agreement paradigms are used on finite verbs and on infinitives.

Finally, the idiosyncratic nature of agreement suffixes and the semantics compatible with them will be supported by some crosslinguistic data, comparing Italian to Spanish and Hungarian to Portuguese.

Maija Vilkkumaa

1. Kirjat jäivät keittiöön.

book-PL stay-PL-PAST kitchen-ILL

The books were left at the kitchen.

(lit. 'The books stayed to the kitchen)

2. Jätin kirjat keittiöön.

leave-SG1-PAST book-PL kitchen-ILL

I left the books at the kitchen

3. Jätin takin junaan.

leave-1SG-PAST coat-GEN train- ILL

I left my coat in the train.

4. a) Jäin kotiin.

stay-1SG-PAST home-ILL

I stayed at home [although I had intended to leave].

Etsuyo Yuasa

Department of East Asian Languages and Literatures

The Ohio State University

Prototypically, there is certain correspondence between syntactic categories and semantic categories: (i) syntactic verbs, adjectives, and prepositions correspond to predicates in semantics, which express properties or relations, and (ii) syntactic NPs correspond to semantic arguments (Wierzbicka 1986, Croft 1991). As McCawley (1982, 1988) notes, however, such correspondence is not absolute, and a syntactic NP can be a predicate in semantics expressing properties (1).

(1) John is a lawyer.

In this paper, we investigate subordination markers in Japanese, and show another type of categorial mismatches in syntax and semantics.

The first part of this paper focuses on the synchronic analysis of Japanese subordination, and three different groups (2)-(4) are identified in accordance with their properties.

(2) [Tamago-ga nai kara] suupaa-made it-te ku-ru (Sub-1)

egg-Nom not have since supermarket-to go-Pres

"Since we do not have eggs, I will go to a supermarket."

(3) [Boku-ga kodomo-no toki-ni] sono-uta-ga hayat-ta (Sub-2)

I-Nom child-Gen time-Dat the-song-Nom become.popular-Past

"When I was a child, the song became popular."

(4) [Iraku-ga iji-o har-u kagiri] jookyoo-wa sinkoku-da (Sub-3)

Iraq-Nom be perverse-Pres as far as situation-Top serious-be

"As far as Iraq is perverse, the situation is very serious."

The first group of subordination markers (Sub-1) exemplified in (2) is prototypical subordination markers or postpositions which connect a subordinate clause with a main clause. The second group of subordination markers (Sub-2) as in (3) consists of an abstract noun and a postposition, which show some parallelism to the Noun Complement Construction or relative clauses. The third group of subordination markers (Sub-3)(4) shares syntactic properties with Sub-1 (the occurrence of determiners or the prenominal form of adjectival nouns before Sub-3, ga-no conversion), but they behave like Sub-2 in semantics (predicate function, no attributive adjectives or a relative clause modifying Sub-3).

In the second part of this paper, I will show that the phenomena observed with Japanese subordination markers are not isolated cases, and similar patterns are also found in instances of decategorization of bleaching (as discussed in Grammaticalizaiton literatures Givon 1979; Traugott and Heine1991; Hopper and Traugott 1993). In decategorization or bleaching, the categorial stauts of the affected item also becomes less clear, because the shift of the meaning and the shift of grammatical properties of the original form take place at different speeds. Observing the history of different groups of Japanese subordination markers, I will claim that Sub-3 subordination markers are on the process of grammaticalization which Sub-1 subordination markers have already completed (5):

(5) Sub-2 (3) Sub-3 (4) Sub-1 (2)

semantics NP---------------------------------------> Postposition--------------->

syntax argument-->Predicate----------------------------------------------------->

In conclusion, I will propose that there is different sensitivity of syntax and semantics to diachronic change (e.g., semantics systematically changes much faster than syntax), and that causes discrepancies between syntax and semantics.

Croft, William. 1991. Syntactic categories and grammatical relations. Chicago: Universith of Chicago Press.

Culicover and McNally, 1998. Syntax and Semantics 29: The Limits of Syntax. San Diego: Academic Press.

Bresnan, Joan. 1995. Categorial mismatch. Theoretical Approaches to African Linguistics, ed. by Akinbiyi Akinlabi, 19-46. Trenton, New Jersey: Africa World Press, Inc.

Harada, S. 1971. Ga-no conversion and idiolectal variations in Japanese. Gengo Kenkyuu 60, 25-38.

Ide, Itaru. 1967. Keishiki meishi toha nani ka. In Nihon no Bunpoo 3, ed. by Tokieda. Tokyo: Meijishoin.

Hopper, Paul J. and Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press.

Jackendoff, Ray. 1997. The Architecture of the Language Faculty. Cambridge: The MIT Press.

McCawley, James D. 1982. The nonexistence of syntactic categories. Thirty Million Theories of Grammar by James D. McCawley, 176-203. Chicago: The University of Chicago Press.

McCawley, James D. 1988. Syntactic Phenomena of English. Chicago: University of Chicago Press.

Newmeyer, Frederick J. 1996. The discrete nature of syntactic categories: Against a prototype-based account. Ms.

Okutsu, K. 1986. Iwayuru Nihongo-joshi-no Kenkyuu. Tokyo. Bonjinsya.

Rosch, E. 1978. Principles of categorization. Cognition and Categorization, 27-48. Hillsdale, New Jersey: Lawrence Erlbaum Associates, Publishers.

Ross, John Robert. 1973. Nouniness. Three Dimensions of Linguistics Research, ed. by Fujimura, 137-157. Tokyo: TEC.

Rubba, Jo. 1994. Grammaticization as semantic change. In Perspectives on grammaticization, ed. by W. Pagliuca, 81-101. Amsterdam: John Benjamins Publishing Company.

Sadock, Jerrold. 1991. Autolexical Syntax. Chicago: University of Chicago Press.

Traugott, Elizabeth Closs and Bernd Heine. 1991. Approaches to Grammaticalization 1. Amsterdam: John Benjamins Publishing Company

Wiersbicka, Anna. 1986. What's in a noun? Studies in Language 10-2, 353-389.

Notes:

1. Example (1) is adapted from Telemann et al. (to appear).

2. Example (4) is from ÅT, Feb 24 1999.