mai ellin tungseth - verbal prepositions and arguments

8/11/2019 Mai Ellin Tungseth - Verbal Prepositions and Arguments

1/201


2/201


3/201

Volume 121

Verbal Prepositions and Argument Structure. Path, place and possession

in Norwegian

by Mai Ellin Tungseth

General Editors

Werner AbrahamUniversity of Vienna / Rijksuniversiteit

Groningen

Elly van GelderenArizona State University

Linguistik Aktuell/Linguistics Today (LA)provides a platform for original monograph

studies into synchronic and diachronic linguistics. Studies in LA confront empirical

and theoretical problems as these are currently discussed in syntax, semantics,

morphology, phonology, and systematic pragmatics with the aim to establish robust

empirical generalizations within a universalistic perspective.

Linguistik Aktuell/Linguistics Today (LA)

Advisory Editorial Board

Cedric BoeckxHarvard University

Guglielmo CinqueUniversity of Venice

Gnther GrewendorfJ.W. Goethe-University, Frankfurt

Liliane HaegemanUniversity of Lille, France

Hubert HaiderUniversity of Salzburg

Christer PlatzackUniversity of Lund

Ian RobertsCambridge University

Lisa deMena TravisMcGill University

Sten ViknerUniversity of Aarhus

C. Jan-Wouter ZwartUniversity of Groningen


4/201

Verbal Prepositionsand Argument Structure

Path, place and possession in Norwegian

Mai Ellin TungsethUniversity of Troms

John Benjamins Publishing Company

Amsterdam / Philadelphia


5/201

Library of Congress Cataloging-in-Publication Data

ungseth, Mai Ellin.

Verbal prepositions and argument structure : path, place and possession in Norwegian /

by Mai Ellin ungseth. p. cm. (Linguistik Aktuell/Linguistics oday, - ; v. )

Includes bibliographical references and index.

. Norwegian language--Grammar. . Linguistics--Norway. . Space and time in

language. I. itle.

PD.

.'--dc

(Hb; alk. paper)

John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or anyother means, without written permission from the publisher.

John Benjamins Publishing Co. P.O. Box Amsterdam Te Netherlands

John Benjamins North America P.O. Box Philadelphia -

Te paper used in this publication meets the minimum requirements of

American National Standard for Information Sciences Permanence of

Paper for Printed Library Materials, z39.48-1984.

8TM


6/201

Table of contents

Acknowledgements

chapter 1. Setting the scene: Events, participants, paths, and places 1

1.1 Introduction 1

1.2 Te P chameleon 2

1.3 Te interace between argument structure and syntax 7

1.3.1 Lexicalism vs. constructionism 8

1.3.2 Ramchands First phase syntax 14

1.3.3 Te argument structure o adpositional phrases 20

1.4 Main findings o the book 22

chapter 2. Prepositions: Paths and places 252.1 Introduction 25

2.2 Locative prepositional phrases and verbs o motion 29

2.2.1 ypes o verbs and interpretations o the PP 29

2.2.2 Goal o motion interpretations: Tomas (2001),

Folli & Ramchand (2002) 31

2.3 Diagnostics 34

2.3.1 emporal adverbials 34

2.3.2VP constituency tests

36

2.3.3 Ordering o adverbials 37

2.3.4 PP-topicalization and clefing 39

2.3.5 Distribution o anaphora 40

2.3.6 Accent placement/prosody 41

2.3.7 Summary o the findings 42

2.4 Expressions o location 43

2.4.1 A note on external vs. internal location 46

2.5 Directional expressions 46

2.5.1 Simplex and complex directional PPs 47

2.5.2 Til 48

2.5.3 Internally complex directional PPs 52


7/201

vi Verbal prepositions and argument structure

2.6 Analysis 542.6.1 Prepositions, argument structure and interpretations 562.6.2 Ambiguous PPs and locative and directional interpretations 572.6.3 Te directional reading 572.6.4 Te locative reading 60

2.7 Unambiguously directional PPs 612.7.1 Simplex directional PPs 622.7.2 Internally complex directional PPs 652.7.3 Directional PPs as adjuncts? 66

2.8 Summary 68

chapter 3. Possession and the notion of affected participants 693.1 Overview and aims 693.2 Previous treatments of double object constructions 70

3.2.1 Oehrle (1976) 713.2.2 Marantz (1984) 723.2.3 Kayne (1984) 733.2.4 Baker (1988) 743.2.5 Larson (1988, 1990) and Jackendoff (1990) 75

3.2.6 Hellan (1991) 783.2.7 farli (1992) 793.2.8 Den Dikken (1995) 803.2.9 Pesetsky (1995) 833.2.10 Harley (2002) 83

3.3 Norwegian: Verb types and interpretations 843.3.1 What types of verbs? 853.3.2 Te interpretation of the added participant 90

3.3.3 Analysis 953.3.4 Benefactive double object constructions 963.3.5 Deriving the structure 1003.3.6 Summary of the findings for Norwegian 103

3.4 German 1043.4.1 Distribution of beneficiaries and maleficiaries 1053.4.2 Analysis 1113.4.3 Creation verbs 1133.4.4 ransitive verbs which do not refer to creation events 1163.4.5 Unaccusative verbs and dative participants 120

3.5 Summary and remaining puzzles 122


8/201

Table of contents vii

chapter 4. Abstract places and results 125

4.1 Introduction 125

4.2 Til and lexically specified endpoints 130

4.3 Intransitive predicates and til 132

4.3.1 What are the properties of semelfactives? 133

4.3.2 What are the properties of degree achievements? 135

4.4 Data: Semelfactives and degree achievements 136

4.4.1 Semelfactives with and without til 137

4.4.2 Degree achievements with and without til 142

4.4.3 Interim summary 148

4.5 Analysis 149

4.5.1 Semelfactives 1494.5.2 Degree achievements 152

4.5.3 Summary 154

4.6 Transitive semelfactives 155

4.6.1 Interpretations with and without til 155

4.6.2 DP vs. PP, and properties of event participants 159

4.6.3 Analysis and structures: Transitive semelfactives 162

4.7 Summary 165

chapter 5. Denouement: Summary and conclusions 167

Bibliography 175

Index 183


9/201


10/201

Acknowledgements

Tis book is a revised version of my doctoral dissertation in linguistics, whichI defended at the University of roms in May 2006. I want to thank my supervi-sor Peter Svenonius, who has shown deep interest in the topics at the heart of thisdissertation, demonstrated patience almost beyond limits, and his door has always

been open for discussion of trivial as well as more difficult problems. I am alsoindebted to Gillian Ramchand, for discussion of theoretical as well as more banalissues. In addition, I want to express my thanks to the external members of mycommittee, or farli and Heidi Harley, for useful discussions and suggestions asto how to make this work more consistent.

I also want to thank Christer Platzack and Halldr Sigursson, who both en-couraged me to publish this dissertation, and also to Werner Abraham and Elly

van Gelderen for showing enthusiasm and interest in my book. I am also greatly

indebted to Gunnar Hrafn Hrafnbjargarson for his magician-like knowledge ofLAEX; thanks a lot for helping me with the layout of this book.

Last, but not least: thanks to all my friends and family, and especially toTomas, for showing great patience with me throughout the process of writingthis work, and for making me understand that det ordner seg (everything willbe fine).


11/201


12/201

chapter 1

Setting the scene

Events, participants, paths, and places

1.1 Introduction

An important assumption which orms the basis or this book, is that the nature

o argument structure is essentially syntactic, and as such conditioned by the same

principles which govern syntactic structures in general (c. e.g., Baker 1988; Hale &

Keyser 1993). Tis places this book well within a constructionist tradition which seeks

to void the lexicon o most o its argument structure inormation (c. Borer 2005,

Ramchand to appear). Instead, the inormation traditionally thought to be properties o

individual lexical items is treated as properties o syntactic structures themselves.

At the present, we have arrived at a stage in the development o linguistic theory

which has provided us with advanced tools enabling us to investigate languages inmuch greater detail than what has hitherto been possible. In addition, the ocus

has been shifed; earlier, the main ocus was a more macrocomparative one, which

examines different languages in order to determine the minimal set o parameters

necessary to account or variation and ultimately also acquisition. However, more

recently, the importance o microcomparative research has also been recognized,

where one instead o taking a birds-eye view o language, makes detailed analyses

o closely related languages or dialects, in order to discover and describe the range o

variation possible.Tis book is a contribution to that discussion; by investigating in detail a limited

number o instances in one particular language, and in which the argument structure

o a prepositional phrase uses with that o a verbal predicate, I hope not only to come

up with generalizations and analyses or the language under investigation, but ulti-

mately also to shed light on the nature o argument structure in general. Here, I have

chosen to ocus mainly on data rom Norwegian, and mainly on the properties o the

notoriously flexible preposition til to, but in chapter 3, which ocusses on beneac-

tive double object constructions, I also make a comparison to German, which is a

closely related language, but which behaves consistently different with respect to the

property under investigation. Te amount o literature written on the topic o prepo-

sitional phrases is vast, and here, I have chosen to ollow the line o research according


13/201

Verbal prepositions and argument structure

to which the argument structure o adpositional phrases can be seen to decompose

into a hierarchy o unctional projections where the interpretation is a

Te present chapter is organized as ollows: I start by discussing the motiva-

tion which lies behind my choice o topic. Ten I move on to laying out the theo-

retical ramework which will be important in the subsequent chapters. In the last

section, I give a short summary o the book, where I present the main findings and

conclusions reached.

1. Te P chameleon

Te purpose o this section is to give a very brie overview o the immense flexibilityo adpositions both in terms o classification and interpretation, and as such, it is

not an attempt to survey the enormous amount o literature written on the topic o

adpositions, both rom a semantic and a more syntactic perspective. For instance,

the whole discussion concerning prepositions and case will not be considered here,

or two reasons: (i) I have chosen to look at data rom Norwegian, which does not

mark nouns or morphological case and (ii) my main concern in this book is argu-

ment structure, and while case sometimes is seen to interact with argument struc-

ture and interpretation, I still think that case is basically a syntactic phenomenon.1

Some questions relating to the inventory o linguistic categories have provednotoriously hard to answer, and many o those questions centre around the proper

classification o the category P. Chomsky (1970) proposed the now well-known

binary classification system o +/V, +/N, which divide the lexical categories verb,

noun and adjective into classes with respect to distributional criteria. In this sys-

tem, verbs are classified as +V, N, nouns are V, +N, and adjectives are +V, +N.

However, Chomsky in his original proposal did not include the category P in this

classification system, this was done later by Jackendoff (1977), who proposed that

adpositions are classified as V, N, and hence fill the gap in this system.However, the properties o adpositions make them particularly hard to classiy.

In some respects, they seem to belong in the verbal domain, in others, they seem to

be more noun-like. Adpositions also seem to stand with one leg in each camp with

respect to the distinction between lexical and unctional categories. In contrast to the

typical lexical categories N, V, and A, which are considered open classes where new

elements can be added, P is ofen treated as a closed class consisting o a small inven-

tory o items, which makes them look like unctional elements. However, in English

1. c. e.g., Svenonius (2002) or a view where the distribution o dative and accusative case in

Icelandic is dependent on whether the subevents introduced by a transitive verb are seen as over-

lapping (accusative) or distinct (dative). For a view o case as a PF phenomenon, which acts as an

instruction to PF in how to translate syntactic structure into morphology, see Platzack (2006).


14/201

Chapter 1. Setting the scene

P appears to be more open, in that new elements can be added to the inventory.

Examples here include words like regarding and concerning, which syntactically are

used in the same way as adpositions, but which originally were verbs. For more on

the grammaticalization from N to P, see, for instance. Knig & Kortmann (1991),

Heine & Kuteva (2002), van Gelderen (2004), or Svenonius (2007).

Moreover, adpositions are also meaningful items which can be used to denote

spatial relations between participants in an event, which makes them look more

like lexical elements. Authors like e.g., den Dikken (2003) argue that the category

P is lexical, but projects a full-fledged structure of extended projections, but others

again deny the existence of P as a lexical category, like e.g., Baker (2003). Accord-

ing to Grimshaw (1991), P is a functional head in the extended projection of the

noun, and it plays a role similar to that played by complementizers in the extendedprojection of V. Tis analogy between adpositions and complementizers is nothing

new; in English, for instance, the complementizer for is homophonous with the

prepositionfor, and the inifinitival marker is homophonous with the preposition

to, which made Emonds (1985) propose that adpositions and complementizers

belong to the same category (cf. also the proposal by Kayne 2004).

While the fact that the inventory of adpositions in a language is rather small

might indicate that P is best treated as a functional category, the lexical vs. func-

tional distinction depends very much on theory-internal criteria for what countsas lexical and what counts as functional, as argued e.g., by van Riemsdijk (1978).

But since this distinction is not crucial within the constructionist framework as-

sumed here, where the meaning of lexical items is treated as stemming from a

fine-grained de-compositional structure, I will not dwell more on that topic here.2

Since this book focusses almost exclusively on data from Norwegian, which does

not mark noun phrases with morphological case, I have chosen to stay agnostic

with respect to the discussion of adpositions and case, for instance, the very inter-

esting alternation in German between accusative and dative with a restricted set of

adpositions in combinations with predicates referring to motion events, which has

received a lot of attention within a generative perspective (cf. e.g., Abraham 2003;

den Dikken 2003; Noonan 2006 & Zwarts 2005a).

However, the topic of the proper classification of adpositions deserves more

attention. As already mentioned, adpositions share important properties both with

verbs and nouns, and attempts have been made to make them fit into one of these

categories. In this book, I hope to show that adpositions (specifically, prepositions)

are chameleon-like in their behaviour and interact with the verbal predicational

structure in different ways. In the next three chapters I will show three different

. But see, for instance Rauh (1993), Koopman (2000), den Dikken (2003), Noonan (2006) for

more on this topic.


15/201


ways in which adpositions and adpositional phrases do this, and I will argue that

their flexibility with respect to the positions they can appear in and the types of

interpretations that arise are best treated within a constructionist approach where

interpretation is a direct consequence of the syntactic structure projected.

In some respects, adpositions have noun-like qualities. For instance, some ad-

positions are developed from a nominal source. Svenonius (2007) mentions the

example instead of, which stems from the old noun stead, which means place. Tis

is also true for its Norwegian equivalent i stedet for lit. in the place of , where the

analogy is even clearer, since the noun sted, which means place is still in use in

the language. Another example comes from Northern Smi, where saji place, can

either be used as a noun, in which case it has its literal meaning, as in (1a), or it can

be used as a postposition meaning instead, as in (1b) (examples from Svenonius2007= his (26ab), p. 12):

(1) a. Don ohkkt mu sajis.

you sit me place.loc/at.place

Youre sitting in my place or Youre sitting instead of me.

b. Don ohkkt mu dbla.sajis.

you sit me usual place.loc

Youre sitting in my usual place.

P is also similar to nouns in not combining with tense/aspect morphology, which

is a typical property of verbs. In languages such as Hungarian, there is also a close

relation between adpositional elements and case suffixes, where the dividing line

between what counts as an adposition and what counts as a case suffix on a noun

is very hard to draw (cf. Asbury 2005).

On the other hand, it is also well known that adpositions have many things in

common with verbs. Both verbs and adpositions can assign case to their arguments,

and in chapter 3, I will argue that this is the case with the Ppreposition which

is essential in the licensing of an added Beneficiary or Recipient participant in

benefactive double object constructions. Based on the original idea by Benveniste

(1966), different authors have argued that the auxiliary HAVE is decomposable

into a main verb BE plus an abstract prepositional component (cf. e.g., Freeze

1992; Kayne 1993; or den Dikken 1995). In my analysis of double object construc-

tions, I will adopt a decompositional analysis of the possession relation which is

essentially similar to the one proposed by den Dikken (1995). Specifically, I pro-

pose that the Goal/Recipient argument is generated in the complement of a null

preposition P , which also case-marks the Goal. Assuming a split PP model, theTeme is then introduced as the external argument of thep head and case-marked

by the verb in the usual fashion. Te empty preposition must be licensed, which

can be achieved in one out of two ways: (i) either via incorporation into a verb


16/201


(here into the abstract verbal predicate Pred, which can be thought of as BE), or

(ii) it can be licensed by dative case morphology. As we will see, Norwegian, lack-

ing morphological dative, obligatorily employs strategy (i), while German has a

choice between the two strategies (contrary to den Dikkens original proposal).

(2) a. Jens strikket Marit en genser.

Jens knitted Marit a sweater

Jens knitted Marit a sweater.

b. Marit bakte Jens en eplekake til bursdagen hans.

Mari baked Jens an apple.cake to birthday.the his

Marit baked Jens an apple cake for his birthday.

Te function of the incorporating preposition P is close to that of the overt

preposition til to, which can be incorporated into the verb with certain ditransitive

verbs (cf. (3)), analogous to double object constructions with verbs of transfer of

possession:

(3) a. Komiteen tildelte kandidaten prisen for beste hovedoppgave.

committee.the to.share candidate.the prize.the for best masters.thesis

Te committee awarded the candidate the prize for the best masters

thesis.

b. De tilbydde han jobben som trener for fotballaget.

they offered him job.the as coach for football.team.theTey offered him the job as coach for the football team.

Moreover, the fact that the entailment of possession cannot be cancelled in

sentences like the ones in (4), supports the claim that the notion of possession is

structurally present here.

(4) a. #Jens strikket Marit en genser, men ga den til rine.

Jens knitted Marit a sweater, but gave it to rine

Jens knitted Marit a sweater, but gave it to rine.

b. #Marit bakte Jens en eplekake, men ga den til hunden.

Marit baked Jens an apple.cake, but gave it to dog.the

Marit baked Jens an apple cake, but gave it to the dog.

Furthermore, many languages have verbal prefixes which are closely related to

ad-positions, and which affect the aspectual properties of the verbs they combine

with in various ways. Slavic has many examples of this type, the following Russian

examples are from Svenonius (2004) (= his (1), p. 1):

(5) a. Helder za-brosil mja v vorota anglian.

Helder into-threw ball in goal English

Helder kicked the ball into the English goal.

b. David sovsem za-brosil futbol.

David completely into-threw soccer

David completely gave up soccer.


17/201


c. Ricardo nervsno za-brosal mja.

Ricardo nervously incp-threw ball

Ricardo began to nervously throw the ball.

In (5a), za into is used in its spatial sense, while in (5b) the combination of verb

plus za has the idiomatic meaning to give up. In (5c), za is used as an inceptive

prefix which focusses on the initiation of the event.

Germanic particles and adpositions can have functions similar to those of

Slavic prefixes, as shown by the examples in (6). Tey can for instance mark that

the action is completed, as in the English example in (6a), which contains the

completive marker up. Adpositions can also signal ongoing action, which English

marks by means of on ((6b)), while German employs a PP with an on plus dative

case on the complement to mark that the action continues, as in (6c):

(6) a. Tey drank all the wine up in twenty minutes.

b. Tey drank on all night.

c. Sie haben an einem Haus gebaut.

they have on a.dathouse built

Tey built on a house.

In chapter 4, I will show instances from Norwegian where what looks like a par-

ticle til to (which can sometimes also appear without an overt ground argument)

specifies a non-spatial endpoint to an event where the endpoint is otherwise under-

specified, so that the interpretation of the event is vague between denoting a single

change or a process. Tus while a sentence like (7a) is vague between denoting a

single jump or a jumping process, (7b) can only have the first type of meaning:

(7) a. Jens hoppet.

Jens jumped

i. Jens jumped (once).

ii. Jens was jumping.

b. Jens hoppet til. Jens jumped to

Jens jumped (once).

Tat the addition of a prepositional phrase can make the event telic by adding a

spatial endpoint is nothing new. It is also well-known that one and the same verb,

which when it appears without a spatial PP is interpreted as unbounded, can com-

bine with different types of spatial PPs, resulting in different types of interpreta-

tions. Consider the examples in (8), with the verb sykle bike. Te example in (8a)

shows that this verb can combine with a directional PP with til to, which makesthe event telic by providing a spatial endpoint. In addition, the same verb can also

combine with a locative PP with a locative preposition like i in, as in (8b), where

the interpretation of the event for some speakers is vague between a locative and a


18/201


directional reading for the PP. Te verb sykle can also combine with a route PP like

mot towards, as in (8c), where a bounded reading for the event is never available.

(8) a. Jens syklet til butikken. Jens biked to store.the

Jens biked to the store.

b. Jens syklet i grfa.

Jens biked in ditch.the

Jens biked in the ditch (ambiguous).

c. Jens syklet mot byen.

Jensbikedtowards town.the

Jens biked towards the town.

In the second chapter, I argue that properties of cases like the ones in (8) can be

handled properly in a framework where the interpretation of the V-PP complex

follows on the one hand from the position which the spatial PP is merged into in

the verbal predicational structure, and on the other from the fine-grained decom-

positional structure of spatial prepositional phrases. I thus hope to show that at

least for the cases at hand, the flexible nature of adpositions can be properly han-

dled within a constructionist framework where the interpretation of elements is a

direct consequence of the fine-grained syntactic structures in which they appear.

1. Te interface between argument structure and syntax

In traditional generative grammar, verbs are assumed to be listed in the lexicon to-

gether with argument structure information which specifies the number and what

types of arguments (in terms of -roles) it combines with in the syntax. Tis argu-

ment structure information then forms the basis for the projection of the syntactic

structure, as stated in the Projection Principle (cf. Chomsky 1981).

A longstanding problem within generative grammar concerns the division

of labour between the lexicon and the syntactic component. In this connection, a

number of important questions arise: How much information should be associated

with the lexical entries themselves, and how much can be treated as the result of

syntactic operations? What is the correct level of representation for stating argu-

ment structure generalizations? Why is it that specific semantic roles get linked to

specific syntactic positions? And how can argument structure alternations such as

the causative-inchoative alternation, but also the limitations on such operations,

best be treated without losing explanatory power?Tis debate dates back as early as Chomsky (1970), who was the first to shed

light on the topic of the division of labour between the lexical component and

the transformational component. In a time when people lightheartedly proposed


19/201


transformations all over the place, Chomsky argued that the properties of different

types of nominalizations in English are best treated as stemming from different

base-generated structures, and not from the application of various complicated

transformations.

Roughly, the types of approaches to these questions fall into two main camps.

On the one hand, we have researchers working within the lexicalist tradition, who

assume that the correct level for stating argument structure information is the

lexicon (cf. Levin & Rappaport Hovav 1995, inter alia; Reinhart 2006 & Reinhart

2002). On the other, we have authors who seek to derive the argument structure

behaviour of lexical items from the syntactic structure in which they appear (cf.

e.g., Hale & Keyser 1993, inter alia; van Hout 1998; van Hout 2000; ravis 1992;

Ritter & Rosen 1998; Borer 2005, among others).

1..1 Lexicalism vs. constructionism

According to lexicalist approaches, the lexicon is an autonomous module with

its own rules and combinatorial mechanics. Te argument structure associated

with particular lexical heads is treated as deriving from the lexical semantics of

that head. In that way, the lexical meaning of a predicate is assumed to determine

its syntactic properties. Te information contained in the lexical entries is highly

specific, and argument structure flexibilities are captured in terms of lexicaloperations directly on argument structure frames. Te lexicon is then linked to

the syntactic component via a system of internally ordered linking rules which are

responsible for mapping event participants into specific structural positions.

One problem for a lexicalist approach concerns the nature of argument struc-

ture alternations. Many verbs are notoriously flexible with respect to the syntactic

structures they can appear in. For instance, a verb like sink or break can either

appear in an intransitive version where it combines with a single internal argu-

ment, or it can appear in a transitive frame where it combines with an agent and

a patient. Tis alternation is quite common, and has come to be known as the

causative-inchoative alternation.

(9) a. Te boat sank.

b. John sank the boat.

c. Te window broke.

d. John broke the window.

Other instances where one verb may be associated with different types of syntac-

tic frames, include, among other things, resultative constructions, as in (10), anddouble object constructions, as in (11):

(10) a. Te dog barked the entire neighbourhood awake.

b. John danced Mary across the room.


20/201


c. I cooked the chicken black.

d. She painted the house red.

(11) a. John sent a letter to Mary. b. John sent Mary a letter.

c. Mary knitted a sweater for John.

d. Mary knitted John a sweater.

In lexicalist frameworks, argument structure alternations like these have either

been captured in terms of both being derived from a common base, or in terms

of treating one alternant as underlying and the other as derived. Reinhart (2002)

proposes different types of operations that operate directly on theta grids (lexi-

cal entries) to change their argument-taking properties. In her system, unaccusa-

tives are derived from causatives via a reduction operation which takes away the

agent. Tus, she assumes that all unaccusatives are uniformly derived from verbs

which have the feature [+c] (cause), and whose second argument is a theme. Te

lexicalist position is nicely summed up in the following quote from Reinhart

(2002), p. 284:

[L]inguistic practice is guided by the principle of Lexicon Uniformity, which

states that each verb-concept corresponds to one lexical entry with one thematic

structure, and entails that the various thematic forms of a given verb are derived

by lexicon-operations from one thematic structure.

However, some verbs are notoriously flexible with respect to the different grammati-

cal environment they can appear in, they seem virtually unconstrained. Te examples

in (12) are from Ramchand (to appear), the ones in (13) are from Borer (2003):

(12) a. John ate the apple.

b. John ate at the apple.

c. Te sea ate into the coastline.

d. John ate me out of house and home.e. John ate.

f. John ate his way into history.

(13) a. Te factory horns sirened throughout the raid.

b. Te factory horns sirened midday and everyone broke for lunch.

c. Te police car sirened the Porche to a stop.

d. Te police car sirened up to the accident site.

e. Te police car sirened the daylight out of me.

Still, verbal flexibilities are not entirely general, as the pairs of examples in (14) show:(14) a. Mary danced.

b. * John danced Mary.

c. Mary yawned.


21/201

1 Verbal prepositions and argument structure

d. * Te lecture yawned Mary.

e. I painted John a picture.

f. * I painted John a wall.

Lidz & Gleitman (2004) have argued, based on the findings from language ac-

quisition experiments both with English-speaking children and children speaking

Kannada, that the children seem to have access to a universal principle of argument

linking, which they state as follows ( cf. Lidz & Gleitman 2004: 157):

(15) Every participant in an event as it is mentally represented shows up as a syntactic

phrase in a sentence describing that event.

Kannada, a Dravidian language spoken in Southern India, gives compelling

evidence that some elements of argument structure must be innate. Unlike Eng-lish, Kannada has an overt causative verbal affix, which can never occur unless

in causative contexts. Tus, the presence of this morpheme is a reliable sign of

a causative context. Still, children acquiring Kannada treated argument number

as an indication of causativity, in exactly the same way as children acquiring

English did, instead of treating the causative morpheme as a clue, despite the

stability of the latter cue. Lidz & Gleitman (2004) take this to indicate that some-

thing like the principle in (15) must be available.3Also, as was pointed out to me

by Heidi Harley (p.c.) the distribution of clausal complements has hitherto beenignored in the literature on argument structure flexibilities, but the ungram-

maticality of examples such as the one in (16) (from Harley & Noyer 2000) need

to be accounted for. As far as I can see, such examples form counterevidence

to the hypothesis that syntactic structure building is totally unconstrained, and

that pragmatic and conceptual knowledge alone are responsible for ruling out

impossible structures.

(16) *Te giraffe falls that the zebra jumps.

In the third chapter, I look at benefactive double object constructions in Norwe -

gian, and I argue that while there is much variation among speakers with respect to

which predicates allow the addition of an extra beneficiary participant, the predi-

cates which permit this must all be compatible with a creation interpretation. In

addition, pragmatic factors concerning use are seen to be at play. Tis conclusion

supports a purely constructionist treatment of benefactive double object construc-

tions, where the impossible structures are not ruled out by grammar per se, but

instead by pragmatic considerations.

. For more on experiments which show that childrens ability for interpretive coercion is

restricted by available argument structure frames, see the discussion in Lidz (1997).


22/201

Chapter 1. Setting the scene 11

Levin & Rappaport Hovav (1995) discusses pairs like (17) (from Levin &Rappaport Hovav (1995), (their (1213) p. 8586), where there seems to be agap in the pattern which is unexpected on the assumption that the intransitiveform is underlying and the transitive form derived via the addition of an externalargument. On the basis of such gaps in the paradigm, they conclude that (17a) isderived from (17b) by decausativization, which removes the external argument.

(17) a. Te wind cleared the sky. b. Te sky cleared.

c. Te waiter cleared the table. d. * Te table cleared.

Authors working within constructionist-based approaches to argument structurego in the opposite direction. On the basis of the extreme freedom of some verbs toappear in different frames as demonstrated by such examples as (12) and (13), theyreject the special status of the lexicon. Rather, the interpretation of lexical itemsis driven solely by the properties of syntactic structures. Te theory of Hale &Keyser (1993), inter alia, is similar to constructionist approaches in may ways. Intheir theory, argument structure is nothing more than the syntactic configurationprojected by lexical items. Te following quote from Hale & Keyser (2002) (p. 1)

sums up their syntax-based approach to argument structure nicely:

Argument structure is determined by properties of lexical items, in particular, bythe syntactic configurations in which they must appear. Tere are just two rela-tions, complement and specifier, defined so as to preclude iteration and to permitonly binary branching.

In this system, theta role assignment reduces to predication; there are no thematicroles apart from the relations between event participants and the lexical relational

structure. Te fact that while an intransitive sentence like (18a) has a transitivecounterpart ((18c)), while a sentence like (18b) does not ((18d)), can be deriveddirectly from structural differences:

(18) a. Te pot broke. b. Te engine coughed.

c. I broke the pot. d. * I coughed the engine.

Te lexical entry for the verb break consists of a root () and a verbal host (V)which may be empty. Te root appears as the complement of the verbal compo-nent, and has the property that it requires a specfier, which they assume is animportant property of the root. Te structure for (18a) is shown in (19):


23/201


(19) V

DP

the pot

V

V

break

Likewise, the lexical entry for cough consists of a root plus a verbal element. How-

ever, unlike break, the root for cough does not require a specifier. Te structure for

(18b) is shown in (20):

(20)

V

V

cough

Te causative variant in (18c) is derived when the structure in (19) is embedded

under a verbal head as in (21), where the argument introduced by the root is in-

terpreted as a subject of change. Te external argument is introduced outside of

the verbal projection.(21) V1

V1 V2

DP

the pot

V2

V2

break

Because the structure in (20) does not introduce a specifier (the subject of unerga-

tives is introduced outside the verb phrase), there is no internal argument which

can be licensed by V1, and the derivation in (22) crashes:

(22) V1

V1 V2

V2

cough


24/201


Tus, the main differences between unaccusative and unergative predicates is as-

sumed to follow from one essential property, which they have to stipulate: unac-

cusative roots introduce a specifier argument (which is interpreted as the subject of

change), while unergatives never do this; their subject is introduced higher up in the

structure, as true external arguments. An unergative predicate cannot be embedded

under a higher verbal head because this would violate restrictions on predication.

However in Hale and Keysers system, although the nature of argument

structure is syntactic, these processes are still assumed take place in the lexicon

(therefore the name L-syntax), although obeying the same principles as other

syntactic derivations.

Borer (2005) argues that a modular theory of grammar where the lexicon forms

a separate component which is linked to the syntactic component via specific in-terface or linking rules has the consequence that the same type of information is

represented twice; once in the lexicon, and once in the syntax. From the point of

view of economy, this is an undesirable result. According to her system, the best

solution is to completely abandon the lexicon as an independent module. Instead

she assumes that lexical items (in her terms listemes) are devoid of any syntactic

properties which regulate their syntactic distribution. Interpretation is simply a

matter of the functional event structure in which the arguments are merged as

specifiers. Any structure can, in principle, be generated, and impossible structuresare ruled out by pragmatic considerations and real world knowledge.

Te existence of flexibilities such as the ones demonstrated in (12) and (13)

above may tempt us into assuming a radical constructionist approach like the one

assumed by Borer (2005), where lexical items do not contain any information rel-

evant to argument structure at all. Still regularities such as the ones in (14) are

quite widespread, and it is hard to see how the non-existence of certain patterns

can be made follow simply from pragmatic considerations.

In this book, I will follow a moderate constructionist approach to the nature of

argument structure like the one assumed by Ramchand (to appear). Tis approach

is moderate in that it does not completely deny the presence in the lexicon of selec-

tional information of some sort, which constrains the way in which lexical items are

associated with structural positions. I will adopt a Ramchandian position according

to which the flexibility of the generative system is constrained by syntactically rel-

evant categorial features which are attached to listemes in the lexicon. In this way,

the lexicon is not totally devoid of argument structure information of the type that

radical constructionists want to get rid of. Still, the approach is constructionist at

heart, because it rejects a view of the lexicon as an independent module with its own

. For an approach which is similar in spirit, although proposed within a more radical

Distributed Morphology framework, see Harley & Noyer (2000).


25/201


specific lexicon-internal processes which are opaque to the syntactic computational

system. Instead, the only type o argument structure inormation present in the lexi-

con is categorial eatures which places restrictions on the syntactic computation.4

1.. Ramchands Firstphasesyntax

Ramchand (to appear) presents a view o argument structure which is essentially con-

structionist, in that it seeks to abandon the need or a separate lexical module where

individual lexical items are listed together with their argument structure inormation,

and where alternations are treated in terms o specific lexicon-internal operations.

At first this might seem similar to the L-syntactic structures assumed in the

work by Hale and Keyser, but there is one important difference between the twoapproaches. While Hale and Keyser assume that the nature o argument structure

is syntactic, and as such as obeying the same restrictions as syntax proper, they

still make a distinction between L-syntax, which takes place in the lexicon, and

S-syntax, which takes place in the syntactic component. However, Ramchand as-

sumes an essentially constructionist approach where there are no lexicon-specific

derivations; everything happens in the derivational component.

In the Ramchandian model, the inormation which is traditionally associated

with lexical items can be decomposed into a combination o maximally three sub-events, each represented by a separate unctional projection. Tere is a causing or

initiation subevent InitP (or initiation), a process subevent ProcP which denotes a

transition or change, and a result subevent ResP which gives the endpoint or final

state o the event.

Each subevent introduces and licenses different types o event participants,

which appear in the respective specifier positions. InitP introduces different types

o causers or external arguments, ProcP licenses the Undergoer, the participant

which is interpreted as undergoing change or transition with respect to the pro-

cess, and ResP introduces the Resultee, which is the participant that is interpreted

as the holder o the result state. Tis splitting up o the verb phrase into separate

layers is similar to other types o decompositional structures or the verb phrase,

where the VP is split up into two separate layers, v and V (e.g., Chomsky 1995), but

differ rom these in assuming a specific projection Result which introduces a final

state which is part o the verbal predication.

In this model there is no one-to-one correspondence between telicity and the

presence o ResP in the decompositional structure o the predicate, a view which

I share. Tus, I adopt the assumption that ResP is only present in case the predi-cate explicitly expresses a result state. In addition, there are also various other ways

in which entailments o telicity can arise, or instance with verbs o creation/con-

sumption, where the telicity o the event is directly correlated with the bounded-

ness o the Undergoer participant (c. e.g., Verkuyl 1972, inter alia& Krifa 1998).


26/201


With these verbs, the internal argument gives rise to a Path or scale which is

homomorphic with respect to the process denoted by the event, and a bounded

Path gives rise to a bounded event. Te term path originally stems from work on

the semantics of spatial prepositions, but it has been generalized to cover a broad

range of cases (cf. e.g., Hay et al. 1999, Beavers to appear, etc.).

Consequently, direct objects whose inherent properties give rise to a scale

which is homomorphic with the process, are not Undergoers, but Paths. Paths are

homomorphic with respect to the process, and appear as complement to ProcP.

Rhemes, on the other hand, appear in the complement to ResP, where they serve

to further define a final point/result state entailed by the lexical predicate. As we

will see, this is exactly parallel to the Paths in the denotation of directional prepo-

sitional phrases, where Path-denoting directional PPs appear as complements toProcP, while location-denoting PPs which are interpreted as goals of motion can

only get this type of interpretation in the complement to ResP.

Something similar has been argued for the prepositional domain, where Zwarts

(2005b) argues the Paths denoted by spatial prepositional phrases can either be

bounded or unbounded, and where the (un)boundedness of the PP contributes to the

(un)boundedness of the event as a whole. According to this view, when a bounded PP

combines with a verb which does not provide an endpoint, the PP is responsible for

the boundedness of the event as a whole, as I will argue in the second chapter.When no endpoint is entailed by the lexical predicate (hence, ResP is absent

from the structure), as with verbs of manner of motion, a directional prepositional

phrase combines directly with the process head, and entailments of telicity arise if

the PP-denotation is bounded. Zwarts argues that while unbounded spatial Paths

can be concatenated to form a new Path (a Path which is towards the house can be

concatenated with another Path which is also towards the house, and the result is a

new Path, which is also towards the house), this is not so for bounded Paths.

As we will see in Chapter 4, this notion of concatenation can also be transferred

to the eventual domain, where I will argue that the extended activity readings for

semelfactive and degree achievement predicates are derived via an operation of

S-summingwhich is essentially similar to concatenation in the event domain.

Te view of participant roles as relating to specific structural positions is simi-

lar to proposals by e.g., Hale & Keyser (1993), ravis (1992), Ritter & Rosen (1998),

Borer (2005), who all assume that event participants are interpreted in relation to

functional heads in the event structure that is built up. In this way, theta role as-

signment reduces to relations between specifiers and heads in the decompositional

structure of the verb phrase. Participant roles can be simple or complex; complexroles arise when one and the same participant is linked to different specifier posi-

tions simultaneously. In Chapter 3, I will argue that when an additional benefactive

participant appears with verbs of creation/consumption, as in a sentence likeJohn

baked Mary a cake, the extra participant gets a composite role; it is simultaneously


27/201


a Resultee, but also a Recipient, where the Recipient interpretation stems from

an empty preposition with possessional content, similar to proposals by e.g., den

Dikken (1995), Pesetsky (1995) or Harley (2002).

raditional theta role labels like Agent, Teme, or Experiencer have no autono-

mous status in this system. Instead, they are simply the results of entailments aris-

ing from different structural relations between a participant in a specifier and a

functional head. Tis is a desired effect, given the notorious difficulties on getting a

coherent classification of the properties of different thematic roles, how many there

should be, what their appropriate labels are, etc (cf. e.g., the discussion in Dowty

1989 or Baker 1996 & Baker 1997, who try to reduce the number of thematic roles to

just a few, related to either verbal entailments (Dowty) or to specific structural posi-

tions (Baker)). However, while I will occasionally employ labels such as Beneficiary,Recipient, or Goal, I will simply use them as handy descriptive labels, although such

labels have no independent status in the model that I am working in.

In effect, a constructionist approach also effectively eliminates the need for

specific linking rules which posit rules for how the participants in a verbs argu-

ment structure are mapped into syntactic positions. Linking has, for instance, been

assumed to take place according to a thematic hierarchy (cf. e.g., Larson 1988 &

Grimshaw 1990), which stipulates the internal order of participants. However, in

a constructionist model, the interpretation of arguments is simply the result of thesyntactic positions they appear in. (23) shows the maximal decompositional structurefor a verb:

(23) InitP

DP

DPInitiator

Init'

Init ProcP

DP

DPUndergoer

Proc'

Proc ResP

DP

DPResultee

Res'

Res XP

...

In this system, ProcP forms the core of the verbal predication, which Ramchand

assumes is obligatorily present for all dynamic verbs, even with punctual verbs.


28/201


At first, this might sound counterintuitive because of the label Proc(ess)P and the

way in which we conceive of processes as durational. However, ProcP might be

treated as representing a pure transition where nothing is said with respect to the

duration of the transition. Tis assumption might, however, turn out problem-

atic for theory-internal reasons, namely the stipulation that ProcP is always pres-

ent with dynamic predications. If ProcP represents a transition, we predict that it

should not be necessary with activity predicates, where there is no change going

on. So the least costly solution is probably to abandon the transition idea, and as-

sume that ProcP represents dynamicity or duration, which might be short.5

In Chapter 4 I discuss the properties of semelfactive predicates in Norwegian,

and there, I simply adopt the assumption that ProcP is obligatorily present. How-

ever, nothing in my proposal really hinges on that assumption, and ProcP couldprobably also be absent without affecting the argument. All possible verb meanings

can be built out of this minimal inventory of subevents. In the subsequent chapters,

I will employ examples from Norwegian to show how this works for different types

of verbs. As already mentioned, I share the view that in order to account for both

argument structure flexibilities and their restrictions, insertion must somehow be

constrained so as to avoid the generation of ungrammatical structures. Te way in

which Ramchand represents this is by tagging a lexical item with categorial fea-

tures (in terms of Init, Proc, Res) which regulate what heads in the decompositionalstructure which the lexeme are associated with. An argument can be identified with

more than one subevent, which is represented by coindexation. Tus, the lexical

entry for a basic unergative verb like dance can be represented like this:

(24) dance: [Initi, Proc

i]

On this basis, we generate a structure like (25), for a sentence likeJohn danced:

(25) InitP

DP

John

Init'

Init

dance

ProcP

DP

John

Proc'

Proc

danced

. In Ramchands system, ProcP is crucially not present with stative predications, which are

just instances of InitP embedding ResP.


29/201


In (25) the lexical encyclopaedic content of the verb is linked to two heads, viz.

Init and Proc, and the participantJohn is simultaneously the Initiator and the Un-

dergoer. Te verb dance does not entail any endpoint or result state, so ResP is not

present. Te entry for a basic unaccusative verb like break can be represented in

the following way:

(26) break: [Proci, Res

i]

Break obligatorily entails a Result state, which is the state of being broken. Te

same participant is simultaneously Undergoer and Resultee. Te structure for an

example like the stick broke is shown in (27):

(27) ProcP

DP

e stick

Proc'

Proc

break

ResP

DP

the stick

Res'

Res

break

Break also allows the addition of a causer. Tis is done by the addition of InitP,

which licenses an extra participant. Te structure resulting from causativization

is shown in (28). Tis is essentially the same type of structure assumed for all telic

transitive verbs.

(28) InitP

DP

John

Init'

Init

break

ProcP

DP

the stick

Proc'

Proc

break

ResP

DP

the stick

Res'

Res

break


30/201


Some verbs which do not themselves entail entail the presence of an endpoint

or result state can nevertheless give rise to telic readings when they combine with

different types of material which identifies a final result state, for instance particles

or resultative secondary predicates.

In Chapter 2, I discuss a class of examples where some verbs of manner of

motion which themselves refer to atelic activities can combine with a locative PP

which defines the endpoint of the event. An example of the right kind is a sentence

likeJens syklet i grfa Jens biked in the ditch, which is ambiguous between a loca-

tive and a directional reading for the PP. Te relevant structure for the directional

reading is shown in (29):

(29)

i grfa

InitP

DP

Jens

Init'

Init

syklet

ProcP

DP

Jens

Proc'

Proc

Syklet

ResP

DP

Jens

Res'

Res PlaceP

On the directional reading, I assume that the endpoint is provided by a null Res

head, the content is not identified by the lexical verb, which does not entail an end-

point with pure manner of motion verbs. Te locative PP, which I assume is an in-

stance of the category Place, then appears in the complement to the null Res head.

Different languages differ with respect to the availability of a null ResP in theirinventory, but Norwegian seems to be quite free in this respect, as a null ResP can

be employed both with verb-particle constructions, locative PPs which get a direc-

tional interpretation, and also, for instance, with resultative constructions.


31/201


1.. Te argument structure o adpositional phrases

Since the main topic o this book is essentially various ways in which adpositional

phrases (specifically:prepositional phrases) combine with different types o verbs,and what interpretations that arise, I will not in any great detail consider the inter-

nal unctional structure projected by different types o adpositional phrases, the

main reason or which being that it is a broad topic which deserves more attention

than what is possible within the limits o this book.

Spatial adpositions can ofen be characterized as denoting asymmetric relations

between a Figure and a Ground, where the Figure is the entity which is located or in

motion, and the Ground is the reerence object which it is being located in relation

to. Tese terms originally stem rom work within gestalt psychology, but have beenemployed by work by almy (almy 1978 & almy 2000). Te ollowing sentences

show that the Figure-Ground relation is asymmetrical and cannot be reversed:

(30) a. Te bicycle was next to the house.

b. # Te house was next to the bicycle.

c. Te cat is under the table.

d. # Te table is over the cat.

Although I will not actively employ the Figure-Ground distinction in this book, I still

assume that patterns like the ones observed by e.g., almy and Svenonius are robustacts about how language expresses spatial relations. Probably, the distinction can also

be transerred to some o the other uses which adpositions have been put to.

Because o the indirect relationship between the external argument and the

verb as opposed to that o the verb and its complement, Kratzer (1996) and others

have argued that the external argument is introduced by a unctional head v which

dominates the verb phrase. Analogous to this, and because o the asymmetrical

relation between Figure and Ground in adpositional phrases, people have pro-

posed that the Figure is introduced as the external argument o a unctional head

p (c. e.g., van Riemsdijk 1990; Rooryck 1996; or Svenonius 2003).

In chapter 4, I assume that the preposition til to, which combines with semel-

actives and degree achievements is different rom the homophonous spatial prepo-

sition til in having the property that it does not introduce an external argument. In

a split PP model o the prepositional phrase, these Ps could be assumed to be bare

Ps lacking the p layer which is necessary or introducing the external argument.

However, nothing in my proposal hinges on this assumption, so I have chosen to

simply use the label PP also or the external-argument-lacking version o til.

For spatial adpositional phrases, I will adopt the insight o much recent workon the syntax o spatial PPs according to which the main distinction between

location and goal is represented in terms o differences in the internal unctional

structures projected by the respective PPs (c. e.g., van Riemsdijk & Huybregts 2002;


32/201


Koopman 2000; den Dikken 2003; Svenonius to appear, inter alia & Noonan

2006). For locative PPs, I will assume that they are PlacePs, even on their direc-

tional reading. Unambiguously directional PPs, on the other hand, are always

PathPs. Accordingly, a locative PP like in the house will be assigned a (schematic)

structure like the one in (31), while a directional PP like into the house will be

assigned a structure like the one in (32), where a Path head to takes a PlaceP

complement.6

(31) PlaceP

Place

in

DP

the house

(32) PathP

Path

to

PlaceP

Place

in

DP

the house

While I do not deny the fact that the extended projections of adpositions may and

probably also do contain more functional structure than what I will be assuming in

this work, and which contribute in specific ways to interpretation, I have chosen to

remain largely agnostic with respect to these issues, because my main concern is with

the external distribution of adpositional phrases, and how they interact with the ver-

bal predicational structure in specific ways. In addition to the ordinary inventory of

prepositions, I assume that the inventory of adpositional phrases also contains null

elements. As I argue in Chapter 3, the possession relation decomposes syntactically

into an abstract verbal predicate plus a null prepositional component P, where the

prepositional component incorporates into Pred. (cf. 1.2).

Te distinction between Place and Path also has its parallel in the event domain.

In exactly the same way as Path embeds PlaceP in directional adpositional phrases

like into, ProcP embeds ResP in the verbal domain. Tis suggests that they might

probably be treated as isomorphic; the verbal version arises when the complex is

. Svenonius (to appear) argues that the complement of Place is a KP (for case), but I abstract

away from that in the following representations.


33/201


incorporated under a verbal projection (InitP), while the prepositional version arises

when no such incorporation takes place. While this might seem a tempting idea,

exploring its full limitations lies beyond the scope of the present work.

For instance, in a resultative like we painted the barn red, ProcP embeds the

resultative AP red, which defines a final state, parallel to the role of PlaceP has in

PathP. A simplified structure for we painted the barn red is given in (33):

(33)

red

InitP

DP

We

Init'

Init

painted

ProcP

DP

the barn

Proc'

Proc

painted

ResP

DP

the barn

Res'

Res AP

In that way PathP can be thought of as the parallel in the adpositional domain tothe verbal ProcP and PlaceP as the parallel to ResP. Tis is also reflected in the way

in which these projections combine with each other and the interpretations that

result from the combination. PathP can only combine with ProcP, where it gives

rise to a Path which is homomorphic with respect to the process denoted by the

verb. In the same way, PlaceP only combines with ResP, and it is interpreted as a

location or state.

1. Main findings of the book

Tis is a book about prepositional phrases, argument structure, and how the argu-

ment structure of prepositional phrases is integrated into the argument structure


34/201


of the verb to give rise to different types of interpretations. Basing my discussion

mainly on data from Norwegian, I develop a model of the relation between argu-

ment structure and syntactic structure where the interpretation of event partici-

pants is a consequence of the structural positions they appear in in a fine-grained

decompositional model of the verb phrase. Te analysis is then applied to three

different types of verb-PP relations. With this, I hope to show that there are good

reasons to assume that the nature of argument structure is essentially syntactic,

and that there is no need for specific argument structure-changing operations tak-

ing place in the lexicon.

In the second chapter, I examine instances where a verb of motion combines

with different types of spatial prepositional phrases, and I show that the verb and

the PP contribute to the interpretation in various ways. Te main purpose of thechapter is to get a clearer picture of how directional/goal of motion readings for

the PP arise. As we will see, in some cases, the notion of an endpoint is contributed

by the verb, and the PP simply serves to further specify this endpoint. In others,

the verb doesnt provide an endpoint at all; instead, it is the semantics of the prepo-

sitional phrase which determines whether the combination is telic or not.

Following a long line of research (cf. e.g., Koopman 2000; den Dikken 2003; or

Svenonius to appear, inter alia & Noonan 2006, to mention just a few) I assume that

spatial prepositional phrases contain a rich functional structure, where the distinc-tion between Path and Place will be the most important for the present analysis.

In combination with a limited set of verbs of motion a locative PP can get a

directional reading, in which case I assume that they are not PathPs, but PlacePs

which can only get a directional reading in a specific structural configuration. Te

distinction between Path and Place is also reflected in their different syntactic be-

haviour, which follows both from differences in the syntactic structures which the

PPs appear in as well as from the distinction between Path and Place. ruly direc-

tional PPs are relatively free with respect to the positions they appear in, which I

assume follows if they are PathPs, which have the property of being referentially

complete. However, locative PPs which only get a directional reading in com-

bination with a limited set of verbs, are not PathPs, but PlacePs, which can only

get a directional reading when they appear as the sister to a projection of a verb

which itself licenses an endpoint. Consequently, if this relation is disrupted, only

a locative reading emerges. Tus I assume that PlaceP is referentially incomplete

in that PlaceP itself is never able to give rise to a goal of motion interpretation.

However, when they refer to locations, PlacePs are even freer in their distribution

than PathPs. In that case, I assume that PlacePs are simply adjoined to a projectionof the verb, although nothing in principle rules out alternative analyses.

Te topic of the third chapter is benefactive and malefactive double object

constructions in Norwegian and German. I start out by looking at data from

Norwegian, where an added beneficiary participant is restricted to appearing with


35/201


verbs referring to events of creation or obtaining. For these cases, I argue that there

is a possession relation between the Recipient and the Teme, which decomposes

syntactically into a null preposition plus an abstract verbal predicate.

Te null preposition is licensed through movement into a position from which

it can incorporate into the verbal predicate. Alternatively, the null preposition can

be licensed by dative case morphology, in which case it does not have to move,

which is an option available in German. For instance, in a sentence likeJohn baked

Mary a cake, the RecipientMary is introduced as the internal argument of the null

preposition, while the theme is introduced by thep head in a split PP model.

German, in addition to permitting a beneficiary participant with verbs of cre-

ation or obtaining, also permits the addition of a dative-marked beneficiary or

maleficiary participant also with other types of verbs, specifically with transitiveand unaccusative verbs which refer to a final point. In these cases, I argue that the

added participant is best treated as an Experiencer which is introduced in a struc-

turally higher position than Recipient/Beneficiaries where it is interpreted as a

holder of the result state represented by the embedded predicate. Here, the notion

of possession is a purely pragmatic consequence.7

Te fourth chapter looks at (transitive and intransitive) semelfactives and de-

gree achievements, which on their base readings are vague between denoting single

changes or extended events. However, in Norwegian (and Swedish), a PP with til tocan combine with semelfactives and degree achievements, in which case only the

single change interpretation is available. Contrary to what has previously been as-

sumed (e.g., by Smith 1997 & oivonen 2003), I argue that the base readings of both

semelfactives and degree achievements are telic, but that the endpoint denoted by

the predicate is crucially underspecified or incomplete. As such, the endpoint can

be conceived of as non-distinct from the starting point of the event, and as a result,

singular events can be concatenated under an operation known as S-summing (cf.

Kamp 1979 & Rothstein 2004) which derives extended events from minimal events

in case certain restrictions are met. When til is present, S-summingcannot apply,

and only the singular reading is available. Tis follows, I argue, if the function of til

is to add content to the underspecified endpoint entailed by the predicate, which

blocks S-summing from applying. Te preposition til as it is employed here is actu-

ally static, and as such distinct from the Path-denoting preposition til, which might

seem surprising at first glance, but as we see, there are also other cases in which this

static til is employed, e.g., to mark possession and Experiencers.

. German also permits free datives with a wide range of other functions, among them the so-

called ethical datives, as in Er fhrt mir zu schnell lit. he drives me too fast, which are relatively

free in their distribution. However, in the present work, I have limited myself to only consider-

ing the two already mentioned classes of datives.


36/201

chapter 2

Prepositions

Paths and Places1

2.1 Introduction

In this chapter, I examine the syntactic and semantic properties of different types

of spatial prepositional phrases in Norwegian. Te purpose is to investigate the

different ways in which spatial PPs combine with verbs of motion to give rise to

locative and directional readings. Te terms locative and directional will here be

used as descriptive labels; while locative PPs serve to locate entities or events in

space, directional PPs specify a direction and an endpoint for the motion.2

In languages with morphological case, such as for instance German or Ice-

landic, there are a set of prepositions which alternate between combining with a

noun phrase marked either with dative or accusative case, and where the different

cases correlate with locative (dat) and directional (acc) interpretations of the PP

as a whole. However, this is an area which deserves more attention than what is

possible within the limits of the present discussion, and will therefore be put aside

for future research (but consider, for instance, the work done by den Dikken 2003;

Abraham 2003 & Zwarts 2005a).

Te term motion verb will be used as a descriptive label for a class of verbs

which refer to motion or displacement. Te properties of this class of verbs have

been described by Levin (1993), who makes further distinctions within the class ofverbs of motion. For the present purposes, the most important distinction will be

the one between verbs of manner of motion and verbs of inherently directed motion.

1. Some of the material included in this chapter has been published as ungseth (2003) and

ungseth (2005). However, the analysis developed here will deviate in certain respects from the

proposals put forth there.

2. In addition to these two categories of spatial PPs, we have instances of so-called route PPs,

which specify a path or route, but which in themselves never specify an endpoint. Examples in-

clude prepositions such as towards or along. I will not consider prepositional phrases with route

prepositions, for lack of space.


37/201


According to Levin (1993)

[verbs o manner o motion ] describe motion that typically, though not neces-

sarily, involves displacement, but none o them specifies an inherent directionas part o its meaning. All o these verbs have meanings that include a notion o

manner or means o motion. Tey differ rom each other in terms o the specific

manner or means (p. 264).

Verbs o inherently directed motion speciy the direction o motion, even i an

overt directional complement is not present.

Te question o how to properly characterize motion relations has been dis-

cussed extensively in the literature. Some authors treat motion events as reer-

ring to a series o snapshots, where each snapshot depicts the moving object at adifferent location. Tis line o thinking has been suggested e.g., by enny (1992)

or Verkuyl (1993). Jackendoff (1996) rejects the snapshot approach as imprecise;

since it only specifies a sequence o momentary states, and not as a sequence o

motion, it loses the act that the object is moving. Instead, he proposes that mo-

tion should be encoded as continuous change over time, where the position o the

theme is mapped onto a path in time. Tis is similar to the view held by e.g., Krifa

(1998), who assumes that motion verbs describe motion relations (MRs) between

an event e, a pathp and a figure x, where x is mapped ontop in e.An important aim o this chapter is to arrive at a deeper understanding o

the different ways in which directional readings can arise. Te meaning o the

verbal predicate and different types o spatial prepositional phrases contribute to

interpretation in different ways; sometimes, the directional interpretation comes

rom the verb, in other cases, the directional meaning derives rom the semantic

properties o unambiguously directional prepositional phrases.

Some verbal predicates like e.g., fall imply an endpoint, and can give rise

to telic readings also when no goal PP is present. elicity can have different

sources, it can either be is part o the lexicalized meaning o the predicate or

it can derive rom our pragmatic/conceptual knowledge about the world. For

instance, while the verb fall is normally thought o as denoting a telic event, a

sentence like the moon has been falling for millennia is acceptable, which shows

the role o pragmatic/conceptual knowledge in determining telicity. In addition,

there are indeed examples o inherently telic predicates, like or instance leave,

collapse or arrive, where telicity is presumably part o the lexical meaning o

the predicate, which in the system adopted here amounts to saying that the Res

head is obligatorily present in their decompositional structure. Tus, telicity canbe seen to all into two types; telicity which is part o the inherent meaning o

the predicate, and associated with its syntactic decompositional structure, and

pragmatic/conventionalized telicity, which might derive rom extralinguistic


38/201

Chapter 2. Paths and places 2

factors. For more on this latter issue, consider, for instance Hay et al. (1999) or

Beavers (to appear).

elic predicates, whether the telicity is part of the core meaning of the predicate,

or derives from conventionalized use, can combine with a locative prepositional

phrase which further specifies that endpoint, as in (1a), where the interpretation of

the PP is directional.3Other verbs, like e.g., run, can optionally give rise to direc-

tional interpretations when they combine with a locative PP. Te example in (1b)

is for some speakers potentially ambiguous between a locative and a directional

reading of the PP, but when an unambiguously directional PP is present, only a di-

rectional reading arises, as in (1c). In yet other cases, the verbal predicate does not

specify a final point at all, as in (1d). Here, the locative PP can only be interpreted

as locating the event, and a directional reading can only arise if the prepositionalphrase itself is directional, as in (1e).

(1) a. John fell in the water (directional).

b. John ran in the room (ambiguous).

c. John ran into the room (directional).

d. John danced in the room (locative).

e. John danced into the room (directional).

Te difference between direction and location is reflected in the basic distinction

between Place and Path, where Place is associated with stative location and Path is

associated with motion and direction. I will follow recent proposals in the litera-

ture by assuming that Place and Path are functional heads in the extended projec-

tion of the preposition (cf. e.g., Koopman 2000; van Riemsdijk & Huybregts 2002;

or den Dikken 2003).4I start out by examining sentences like the ones in (1a) and

(1b) where a locative prepositional phrase in combination with a limited subclass

of verbs is ambiguous between a directional and a locative interpretation. I argue

that the directional and locative interpretations arise from differences in the syn-

tactic positions for the prepositional phrase in each case, which is consistent withtheir different behaviour with respect to a number of tests which are sensitive both

to syntactic, semantic and phonological distinctions.

3. However, the properties of the Ground argument of the preposition also seem to play a role

in determining whether a locative or directional interpretation is most appropriate. In fall in

the water,the wateris typically located on the ground, which is downward from the perspective

of the faller. An example likefall in the room, on the other hand, typically gets a locative read-

ing, because of the properties of the room, which is not typically something which it is easy toconceive of falling into.

4. Te term path was introduced by Jackendoff (1983) in order to describe PPs with motion

verbs.


39/201


Having done that, I move on to looking at the behaviour of unambiguously

directional (goal) prepositions like til to, which will be treated as instances of

the categoryPath. Although we should expect unambiguously directional PPs to

pattern essentially in the same way as locative PPs on their directional meanings,

we will see that the patterns are not entirely as expected. Tis fact which will have

to receive a principled explanation.

Spatial PPs vary with respect to their internal complexity. Both locative and

directional PPs can either be simplex (in a non-technical sense), by which I

mean that they consist of a single preposition, like e.g., at (locative) or to (di-

rectional). Tey can also be complex, like e.g., onto (directional) or in front of

(ambiguous), in which case they consist of a locative or directional particle plus

a simple locative preposition. In the following, I will occasionally use the termssimple and complex to describe PPs which contain a single preposition like

i in or til to.

In order to get a clearer picture of the patterns which emerge, I will first lay out

the data and diagnostics before moving on to the actual analysis where each type

of preposition-verb-combination will be treated separately. In combination with a

subset of verbs of manner of motion, PPs with locative prepositions are ambigu-

ous between referring to the endpoint of the event, or specifying the location for

the motion event. Such prepositional phrases are invariably instances of the cat-egory PlaceP; the differences in interpretation are the consequence of the different

structural positions they are merged into in the decompositional structure of the

verb phrase.

On their directional readings, these PPs appear low down in the verb phrase

as complements to the Res head in a Ramchandian decompositional structure like

the one outlined in the previous chapter. Here, they get a goal of motion interpre-

tation. On their locative reading, they appear higher up in the structure as adjuncts

to the verb phrase, where they specify the location for the event. PPs which are

unambiguously directional (i.e., they can never get locative readings) are PathPs,

and PathPs do not depend on an endpoint implied by the verb to be interpreted as

goals of motion. In contrast to PlacePs, they can also get a goal reading with a wide

range of atelic manner-of-motion verbs.

With atelic manner of motion verbs, I assume that PathPs appear in the com-

plement position to the Proc head where they give rise to a path or scale which

is homomorphic with respect to the verbal process. Tus, it is the properties of

the prepositional phrase which determine whether the event is telic or not, in

the fashion of Zwarts (2005b). If PathP is headed by a bounded preposition liketil to, which denotes a bounded path, the interpretation of the event as a whole

is bounded. If PathP is headed by an unbounded preposition like mot towards,

which denotes an unbounded path, the interpretation of the event is atelic.


40/201

Chapter 2. Paths and places 2

2.2 Locative prepositional phrases and verbs of motion

In Norwegian, and to a certain extent also in English, a sentence containing a verb

of motion and a prepositional phrase which consists of a locative preposition like i

in,p on, or under under, can be seen to be ambiguous between a telic reading of

directed motion and an atelic reading of located motion. A few examples with dif-

ferent prepositions are given in (2) (intransitive verbs) and (3) (transitive verbs):5

(2) a. Jens har syklet i grfa.

Jens has biked in ditch.the

Jens has biked in the ditch (ambiguous).

b. Per har hoppet i elva

Per has jumped in river.thePer has jumped in the river (ambiguous).

c. Flasken har rullet i keren.

bottle.the has rolled in field.the

Te bottle has rolled in the field (ambiguous).

(3) a. Hrek mtte trille vogna i garasjen.

Hrek must roll cart.the in garage.the

Hrek had to roll the cart in the garage.

b. Per burde bre skiene p lofet.

Per should carry skies.the on attic.the

Per should carry the skies in the attic.

c. Hans har skjvet pappeskene under senga.

Hans has pushed cardboard.boxes.the under bed.the

Hans has pushed the cardboard boxes under the bed (ambiguous).

A prepositional phrase like i grfa in the ditch, in (2a) can either be interpreted as

the endpoint/goal of the biking event, or it can simply specify its location.

2.2.1

ypes of verbs and interpretations of the PPWhile the sentences in (2) and (3) above are ambiguous between locative and

directional readings for the PP, the PPs in (4) can only be interpreted as the end-

point of the event. Te verbs used in (4) belong to Levins class ofput-verbs, which

are inherently directional, where a locative PP only serves to further specify the

endpoint already entailed by the predicate.

(4) a. Marit har satt brdet i ovnen.

Marit has put bread.the in oven.the

Marit has put the bread into the oven.

5. In the following, I will use complex verb forms in order to avoid issues regarding the posi-

tion of the direct object.


41/201


b. Hun har satt seg p stolen.

she has sat herself on chair.the

She got seated on the chair.

c. Dere burde putte bkene i esker. you should put books.the in boxes.the

You should put the books in boxes.

d. Hun har lagt boka under senga.

she has put book.the under bed.the

She has put the book under the bed.

However, not all verbs which refer motion or displacement can combine with

locative prepositions to give rise to directional interpretations. For instance,

the only available interpretation for the PPs in the sentences in (5) is locative.According to Levins classification, these verbs all belong to the class of verbs

of manner of motion. Although some manner of motion verbs do license

directional interpretations for locative PPs, this is never possible with the verbs

used in (5).

(5) a. Petter mtte svmmme i innsjen.

Petter had.to swim in lake.the

Petter had to swim in the lake.

b. Marit har jogget i skogen.Marit has jogged in forest.the

Marit has jogged in the forest.

c. Vi har spasert i parken.

we have strolled in park.the

We have strolled in the park.

d. Barna har danset i stua.

children.the have danced in living.room.the

Te children have danced in the living room.

Yet other verbs again do not refer to displacement, and can never combine withdirectional PPs. However, some

mai ellin tungseth - verbal prepositions and arguments

Documents