Axel Holvoet & Nicole Nau - Grammatical Relations and Their Non-Canonical Encoding in Baltic

Grammatical Relations
and their Non-Canonical Encoding in Baltic

Valency, Argument Realization and
Grammatical Relations in Baltic (VARGReB)
This series of five volumes is the outcome of a research project coordinated by Vilnius
University, aiming at bundling research efforts of scholars from different countries
in the domain of the Baltic languages (Lithuanian and Latvian). Baltic linguistics has
long been fraught with a strong diachronic bias, but there is now a growing body of
theoretically and typologically informed work on these languages. The series focuses
on a core domain of Baltic grammar, viz. that of clausal architecture, case marking
and grammatical relations, and the syntax-semantic interface. In addition to three
collections of articles the series will also comprise two monographs, dealing with
copular sentences and argument alternations respectively. The first volume to appear
in the series is concerned with non-canonical grammatical relations in Baltic. The
next volumes will deal with a variety of issues in the domain of valency and argu-
ment realization.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/vargreb
Volume 1
Grammatical Relations and their Non-Canonical Encoding in Baltic
Edited by Axel Holvoet and Nicole Nau
Grammatical Relations and their
Non-Canonical Encoding in Baltic
Edited by
Axel Holvoet
University of Warsaw & Vilnius University
Nicole Nau
Adam Mickiewicz University in Poznań & Vilnius University
John Benjamins Publishing Company

Amsterdam / Philadelphia
TM
The paper used in this publication meets the minimum requirements of
8
the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data
Grammatical relations and their non-canonical encoding in Baltic / Edited by Axel Holvoet
and Nicole Nau.
p. cm. (Valency, Argument Realization and Grammatical Relations in Baltic,
issn 2352-0159 ; v. 1)
Includes bibliographical references and index.
1. Baltic languages--Grammaticalization. 2. Baltic languages--Grammar. 3. Baltic
languages--Case. I. Holvoet, Axel. II. Nau, Nicole.
PG8022.G73 2014
491’.9--dc23 2014000518
isbn 978 90 272 5909 7 (Hb ; alk. paper)
isbn 978 90 272 7039 9 (Eb)
© 2014 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents
Preface vii
Argument marking and grammatical relations in Baltic: An overview 1

Axel Holvoet and Nicole Nau
Case and word order in Lithuanian infinitival clauses revisited 43

Peter Arkadiev
Non-canonical grammatical relations in a modal construction:

The Latvian debitive 97
Axel Holvoet and Marta Grzybowska
Alternations in argument realization and problematic cases

of subjecthood in Lithuanian 137
Kristina Lenartaitė-Gotaučienė
Subjecthood in specificational copular constructions in Lithuanian 181

Rolandas Mikulskas
Differential object marking in Latgalian 207

Nicole Nau
The independent partitive genitive in Lithuanian 257

Ilja A. Seržant
On the non-canonical marking of the highest-ranking argument

in Lithuanian and Icelandic: Steps toward a database 301
Björn Wiemer and Valgerður Bjarnadóttir
Language index 363

Name index 365
Subject index 367
Preface
The present volume is the first in a series of five, consisting of three collections
of articles and two monographs, dealing with issues of valency and argument
marking, clausal architecture and grammatical relations, and the syntax-seman-
tics interface in Baltic. This series is the outcome of a research project conducted
by Vilnius University and supervised by the Lithuanian Research Council. It is
financed from the European Social Fund under measure VP1-3.1-ŠMM-07-K
“Support to Research Activities of Scientists and Other Researchers” (also called
the Global Grant Measure), Priority 3 “Strengthening of Capacities of Research-
ers and other Scientists” of the Operational Programme for Human Resources
Development.
The aim of this international research project is to bundle the efforts of schol-
ars from Lithuania and other countries in advancing theoretically and typologi-
cally informed research into the Baltic languages. Baltic scholarship, traditionally
characterized by a strong diachronic bias, is now rapidly evolving, and there is
an increasing interest in the Baltic languages among linguists of the most various
persuasions. We hope that this series will mark a breakthrough in the fruitful ex-
change of ideas between scholars in the Baltics and other countries.
Our warmest thanks are due to the following external reviewers who, in spite
of numerous other commitments, have found time to assist us with their insightful
criticisms and comments: Mirjam Fried, Gerd Hentschel, Agata Kochańska, Ora
Matushansky, Heiko Narrog, Elena Paducheva, and Alena Witzlack-Makarevich.
Wayles Browne, of Cornell University, has kindly agreed to read every ar-
ticle in the volume, and to improve the language of the English manuscripts. In
addition to this, he has made invaluable comments on the contents of the ar-
ticles, which have allowed the authors to avoid many a grievous mistake. Need-
less to say, he cannot be blamed for any of the shortcomings that remain in the
contributions.
Our sincere thanks are also due to the administrative personnel of Vilnius
University for their inconspicuous but efficient technical support, and to the rep-
resentatives of the Lithuanian Research Council for their constructive supervi-
sion of the research project.
The Editors
Argument marking and grammatical
relations in Baltic
An overview
Axel Holvoet*, *** and Nicole Nau**, ***

* University of Warsaw / ** Adam Mickiewicz University in Poznań /
*** Vilnius University
1. The project1
The Baltic languages – especially Lithuanian but also Latvian – have figured
prominently in the historical-comparative literature because of their archaic fea-
tures, which make their testimony in many cases no less valuable than that of
Indo-European languages of much older attestation such as Sanskrit and Classi-
cal Greek. A diachronic bias is felt in Baltic scholarship to this very day, whereas
the equally valuable contribution which the Baltic data can make to synchronic
research – both theoretical and empirical – has not yet been fully appreciated.
The situation is changing, however. There is a growing body of theoretically
and typologically informed research into the Baltic languages, which can now no
longer be said to be a neglected province of linguistic scholarship. The series of
publications opened by the present volume has grown out of a research project
aiming at bundling research efforts in a domain in which much interesting work
is being done by scholars from various countries and representing various frame-
works – functional-typological linguistics, Cognitive Grammar, Minimalism, Role
and Reference Grammar and others. This domain can broadly be characterized as
that of clausal architecture, case marking and grammatical relations, and the syn-
tax-semantics interface. Apart from being a convenient common denominator for
1. We wish to thank Peter Arkadiev, Rolandas Mikulskas, Ilja Seržant and Björn Wiemer
for their comments on this introduction. The sole responsibility for its shortcomings rests
with us.
2 Axel Holvoet and Nicole Nau
much topical research, an area of investigation thus defined also commends itself
by its intrinsic interest, being as it is at the heart of linguistic structure.
In view of the manifold interconnections and entanglements between the con-
stituent parts of this research domain, the division of subject matter between the
three collections of articles to be published within this project is, to a certain ex-
tent, one of convenience rather than of sharply delineated thematic clusters. Still,
the contributions to this volume present a fundamental unity, being concerned,
from different angles, with issues in the domain of grammatical relations.
The languages at the centre of our research project constitute a separate
branch of the Indo-European family, with close genetic and areal links to the Sla-
vonic languages. The recorded history of the Baltic languages is young, with the
earliest continuous texts dating from the 16th century. The written attestation of
Lithuanian starts (but for a few brief prayer texts) in 1547, that of Latvian in 1585.
The 16th century also saw the publication of a few texts in Old Prussian, a Baltic
language spoken in East Prussia, completely extinct since the early 18th century.
The extant Old Prussian texts are of such poor quality as to be of limited use
even for historical-comparative purposes and completely useless in research on
linguistic structures in Baltic. On the other hand, the High Latvian dialects used
in former Polish Livonia form the basis of a separate written variety attested from
the 18th century onward. Under the name Latgalian it now has semi-official status
in Latvia, and recent research has shown that it possesses interesting typological
features partly different from those of Low Latvian (the basis of Standard Lat-
vian), which justifies its figuring as a distinct research object alongside Lithuanian
and Latvian (cf. Nau 2011a).
2. Alignment and case marking in Baltic
The Baltic languages use the nominative-accusative type of marking of gram-

matical relations, where agentive arguments of transitive clauses and intransitive
clause subjects are marked with the nominative and transitive patientive argu-
ments with the accusative. This canonical case-marking pattern is also used with
many predicates for arguments that are not prototypical agents and patients, for
example with verbs of perception or knowledge. But many predicates in the zone
of low semantic transitivity (in the sense of Hopper & Thompson 1980) diverge in
interesting ways from the canonical pattern, marking arguments with one of the
other morphological cases. When a marking pattern diverges from the canoni-
cal pattern (which tends to gloss over the subtler distinctions in semantic role),
types of case marking emerge that appear to be semantically driven, overriding,
as it were, syntactic defaults. Such phenomena in various languages have received
Argument marking and grammatical relations in Baltic 3
much attention lately, and the notion of ‘semantic alignment’ has been proposed
in a volume edited by Donohue and Wichmann (2008). This notion has certain
drawbacks, however. It seems problematic to apply it to languages that do not
consistently show split marking for intransitive subjects. With reference to the do-
main lying between canonical transitivity and intransitive one-place predication,
the notion of alignment can best be avoided because of the non-canonicity of the
structures involved. Both Lithuanian and Latvian are clearly nominative-accusa-
tive marking languages, but there is a considerable body of two-place predicates
in the zone of low semantic transitivity that, rather than being canonicized, opt
for basically intransitive patterns. One of the two arguments rather than both
is then marked with a structural case, usually the nominative, whereas the case
marking of the other will reflect its semantic role. That Lithuanian and Latvian,
and many Slavonic and some Germanic languages for that matter, have many
more of these structures than English is probably not due to a different pattern
of alignment, but at least partly to the fact that they have rich case systems with
one or several semantic cases. On the danger of basing alignment distinctions on
non-canonical structures cf. also Haspelmath (2011).
2.1 Case marking in Baltic: The repertoires
In the Baltic languages, case-forms of nouns, adjectives, pronouns and other de-
clinable lexemes are formally well differentiated. Each case has a formal marker,
and there is very little syncretism. From a purely morphological point of view,
there is thus no ‘unmarked case’, unlike what we observe in the Slavonic languag-
es, where nominative and accusative do not always have an inflectional ending.2
The dative case in the Baltic languages shows all the properties of a prototypi-
cal dative as described by Næss (2011). Dative-marked arguments and adjuncts
designate recipients, benefactives, experiencers, goals, and purposes. In Latvian,
where no lexical verb meaning ‘have’ is used, the dative also marks the possessor
in predicative possession, and in western dialects and some varieties of colloquial
speech also in attributive possession, ousting the genitive.3 As Næss and others
have noted, the dative defies classification as either a ‘structural’ or a ‘semantic’
case: it has clearly structural functions, as the marking of the indirect object in
ditransitive clauses, but on the other hand “it has uses which can only be ac-
2. For case morphology in the Baltic languages, see Ambrazas, ed. (1997); Andronov (2001);
Nau (2011b).
3. For a cognitive grammar account of the functions of the Latvian dative see Berg-Olsen
(2005).
counted for in semantic terms, such as marking experiencers or beneficiaries”

(Næss 2011: 573). Case alternations involving the dative will be described in both
of the following subsections. Furthermore, Lithuanian and Latvian have a range
of free datives, which are notoriously difficult to classify and will be the subject of
contributions to one of the following volumes.
The use of the genitive for marking verbal arguments is well established in
Lithuanian, where it is associated with low transitivity, quantification, or negated
existence. It occurs as an alternate to the accusative and as the lexical object of
certain verbs. Modern Latvian, on the other hand, has more or less given up the
genitive as a case for objects without preposition (for an empirical study of the
genitive in contemporary Latvian see Berg-Olson 1999, 2000). A dramatic decline
in the use of the genitive in all its functions is witnessed in Latvian dialects with
a geographical spread from east to west. In the easternmost dialects, the use of
the genitive is frequent and very similar to Lithuanian, while in some of the west-
ernmost dialects the genitive has been lost totally (even as the case of adnominal
possession).
Lithuanian further has an instrumental case, which is used for marking argu-
ments and adjuncts. In Latvian the instrumental has formally merged with the
accusative in the singular and the dative in the plural. It has left some traces which
show up in some of the functions of these cases (albeit in constructions with re-
stricted productivity), and some scholars still argue for distinguishing an instru-
mental case in Latvian (for an overview see Andronov 2001: 205–207). However,
from a purely synchronic point of view there is little ground for this claim, and
the constructions in question can better be explained as uses of the accusative and
the dative (see Holvoet 2010). Finally, both Lithuanian and Latvian have a locative
that is always used without prepositions. Its main function, as may be expected,
is in marking adjuncts, but it also plays a minor role in argument marking. This
role is more pronounced in Latvian than in Lithuanian, which is partly connected
with the locative taking over part of the functions of the defunct instrumental (as
noted by Mühlenbach, cf. Endzelīns & Mühlenbachs 1907: 187, an idea further
developed in Holvoet 1992).
It is interesting to compare the use of the instrumental and the locative across
the Baltic languages. The group of verbs that have an argument in the instrumen-
tal case in Lithuanian is semantically diverse, and corresponding verbs in Latvian
mark the respective argument in different ways: most often with a preposition (ei-
ther ar ‘with’ or par ‘about; for’), sometimes with the accusative, the dative, or the
locative. A locative argument in Latvian, on the other hand, may correspond to an
accusative object in Lithuanian. The following three examples illustrate some of
these correspondences. The examples were chosen from the parallel corpus of Lat-
vian and Lithuanian fiction (LiLa); the roots of the selected verbs are cognates.
(1) ‘turn (into)’ Latvian original: locative
Vēj-š, kas jau bija pārvērties īst-ā
wind-nom.sg what.nom ptc be.pst.3 turn.ppa.m.sg real-loc.sg
vētr-ā,
storm-loc.sg
Lithuanian translation: instrumental
Vėj-as, jau spėjęs virsti tikr-a
wind-nom.sg ptc be_in_time.ppa.m.sg turn.inf real-ins.sg.f
vėtr-a,
storm-ins.sg
‘The wind, which had already turned into a real storm’
(2) ‘fall in love’, Lithuanian original: accusative

ji į-si-mylėjo tok-į akiniuot-ą student-ą
she pfx-refl-love.pst.3 such-acc.sg bespectacled.acc.sg student-acc.sg
Latvian translation: locative
viņa ie-mīlējās tād-ā briļļain-ā student-ā.
she pfx-love.refl.pst.3 such-loc.sg bespectacled.loc.sg student-loc.sg
‘she fell in love with a certain bespectacled student’
(3) ‘believe’; Lithuanian original: instrumental

Dangir-a tikėjo ne juo, o motin-a!
Dangira-nom believe.pst.3 neg 3sg.ins but mother-ins.sg
Latvian translation: dative
Dangira ticēja nevis viņ-am, bet māt-ei!
Dangira-nom believe.pst.3 neg 3sg-dat but mother-dat.sg
‘Dangira didn’t believe him, she believed mother!’
In the last example, the use of the instrumental in Lithuanian may be a some-
what artificial innovation of the standard language (it is nowhere attested in Old
Lithuanian). The use of the instrumental and the locative as marker for coding
arguments in Lithuanian and Latvian is certainly worth further study, from a lan-
guage-specific as well as comparative, a synchronic as well as a diachronic point
of view. In the following, however, we will concentrate on the uses of the genitive
and the dative in differential or non-canonical argument marking.
2.2 Differential object marking in transitive clauses: Alternatives

to the accusative
2.2.1 Transitivity-related differential marking

In the Baltic languages, as in many other languages of the world, there are differ-
ent strategies for marking reduced semantic transitivity in the coding of two-place
predicates. Low transitivity in two-place predicates may be reflected by different
mechanisms (cf. Malchukov 2006; Næss 2007). So, for instance, the first argument
may receive nominative marking, as in the transitive structure, while the second
argument gets a type of marking other than the canonical accusative. Another
strategy, to be discussed further on (Section 3) has an alternative kind of marking
(specifically the dative) for the first argument, whereas the second argument may
have the nominative marking characteristic of intransitive predicates but may also
show other types of marking. We will here mention a number of instances where
case marking conforms to the transitive model at one end, i.e., the first argument
looks like a commonplace transitive subject, whereas the marking of the second
argument is affected by transitivity-reducing factors such as negation, partial af-
fectedness or mood (irrealis).
Alternations with the genitive are the most common and they are most gram-
maticalized in Lithuanian, whereas in modern Latvian there are only some trac-
es of a formerly more widespread use. An areal survey comparing the scope of
genitive marking of objects and subjects in Baltic, Slavonic and Fennic is given
in Koptjevskaja-Tamm and Wälchli (2001: 650–660). The first rule to mention is
the genitive of negation. In Lithuanian, the object of a transitive verb is regularly
marked with the genitive under negation, both in finite clauses (example (5)) and
in infinitive clauses depending on a negated complement-taking verb (7).
Lithuanian
(4) Kit-ą dien-ą jis pamatė med-į su
other-acc.sg day-acc.sg 3.nom.sg.m see.pst.3 tree-acc.sg with
graži-ais dideli-ais žied-ais.
beautiful-ins.pl.m big-ins.pl.m blossom-ins.pl
‘The other day he noticed a tree with beautiful large blossoms.’
http://lt.wikipedia.org/wiki/Tikrasis_baobabas
(5) Eism-o įvyk-is: vairuotoj-as ne-pamatė medži-o.
traffic-gen.sg accident-nom.sg driver-nom.sg neg-see.pst.3 tree-gen.sg
‘Traffic accident: driver fails to notice tree.’
http://www.balsas.lt/naujiena/572827/eismo-ivykis-vairuotojas-nepamate-
medzio
(6) Tiesa, ne vis-i žmon-ės gali ger-ti kav-ą […]

true neg all-nom.pl.m people-nom.pl can.prs.3 drink-inf coffee-acc
‘True, not all people can drink coffee.’
http://sveika.lt/kava-ir-jos-teigiamas-poveikis/
(7) Kodėl kai kur-ie žmonės ne-gali ger-ti
why some-nom.sg.m people-nom.pl neg-can.prs.3 drink-inf
kav-os?
coffee-gen
‘Why can some people not drink coffee?’
http://www.sveikatos-namai.lt/wordpress/?p=65
The genitive of negation is an old inherited feature in Lithuanian; it is also well

known in the Slavonic languages. In Latvian, in contrast, objects of negated transi-
tive verbs regularly appear in the accusative. This is true for the standard language
as well as central and western dialects. In the 19th century, genitive marking was
still well attested as an alternative to accusative marking in central dialects and
marginally found in western dialects (see Mīlenbahs 2009 [1890/1891]). The
alternation between genitive or accusative was not connected to a difference in
meaning (Mīlenbahs 2009 [1890/1891]: 42). In Modern Standard Latvian, there
are only some traces of a genitive of negation left. It appears especially with the
negative pronouns nekas ‘nothing’ and neviens ‘nobody’, ‘no, not any’ (as a modi-
fier), or after the negative focus particle ne. Compare the following two render-
ings of an utterance reported in a news text. With a simple negation the object is
in the accusative, while with the additional particle ne it is in the genitive:
Latvian4
(8) Voika: Prasīb-ai par KNAB vadītāj-a atkāpšan-os
Voika request-dat.sg for KNAB director-gen.sg resignation-acc.sg
no amat-a ne-redzu pamatojum-u.
of post-gen.sg neg-see.prs.1sg reason-acc.sg
‘Voika: I don’t see a reason for requesting that the director of KNAB
[Corruption Prevention and Combating Bureau] should resign from his post.’
(headline)
4. Published March 1, 2013 in the electronic version of the newspaper Diena at: http://www.
diena.lv/latvija/viedokli/voika-prasibai-par-knab-vaditaja-atkapsanos-no-amata-neredzu-
pamatojumu-13996105.
(9) Lidz ar to prasīb-ai par KNAB vadītāj-a atkāpšan-os

Therefore request-dat.sg for KNAB director-gen.sg resignation-acc.sg
no amat-a ne-redzu ne maz-āk-ā
of post-gen.sg neg-see.prs.1sg neg little-comp-gen.sg.m.def
pamatojum-a.
reason-gen.sg
‘[…“] Therefore I don’t see the least reason for requesting that the director of
KNAB should resign from his post [,” said Voika.]’ (later in the same text)
In what way the particle ne and the negative pronouns affect the choice of object
case-marking deserves to be investigated in detail. It seems tempting to assume
that they have a sequentializing effect, i.e., induce incremental conceptualization
of the object, a factor associated by Huumo (2009) with the use of the partitive in
Finnish. In written texts, accusative and genitive marking are used alternatively
with no apparent difference in meaning. An interesting recent example is the fol-
lowing. In a telephone interview in a morning TV show on September 9, 2013,
the Latvian Minister of Culture denied the possibility of her resignation.5 She was
quoted in various media in one of the following forms:
(10) Prasīb-ai par man-u demisij-u
request-dat.sg for my-acc.sg resignation-acc.sg
ne-redzu nevien-u pamatot-u argument-u.
neg-see.prs.1sg no.one-acc.sg reasonable-acc.sg.m argument-acc.sg
‘I don’t see any reasonable argument for the request of my resignation.’6
(11) Prasīb-ai par demisij-u
request-dat.sg for resignation-acc.sg
ne-redzu nevien-a racionāl-a argument-a.
neg-see.prs.1sg no.one-gen.sg reasonable-gen.sg.m argument-gen.sg
‘I don’t see any reasonable argument for the request of [my] resignation.’7
The form the Minister actually used in the interview was the genitive, and the
pronoun neviena was emphasised. She also used the adjective racionāls and not
5. The interview was online at http://www.ltv.lv/lv/raksts/09.09.2013-grende-noliedz-ka-

noslepenojusi-auditoru-zinjojumu-par-operu.id17688/ (accessed 01.10.2013). The minister
actually lost her job shortly afterwards.
6. http://www.tvnet.lv/zinas/viedokli/477986-jaunzemegrende_neredzu_nevienu_pamatotu_
argumentu_demisijai
7. http://www.apollo.lv/zinas/ministre-prasibai-par-demisiju-neredzu-neviena-racionala-
argumenta/609018
pamatots for ‘reasonable’, thus, the second version is a more faithful rendering of
the second part of the sentence.
In Latgalian, the situation regarding the genitive of negation is in between
that of Lithuanian and Latvian: in more conservative variants of Latgalian, the
genitive is regularly used, while in many recent written texts the accusative is
the rule, except for negative pronouns (see Nau, this volume, for details). An-
other use of the genitive in Latgalian that is related to the genitive of negation is
in utterances that contain the speaker’s judgement of the situation as doubtful,
impossible, or undesired. Consider the following example from a short story by a
contemporary writer:
Latgalian (example from Nau, this volume)
(12) Ej tī saprūti tūs sīvīš-u.
go.prs.2sg ptc understand.prs.2sg dem.gen.pl woman-gen.pl
‘Try and understand women.’
The verb saprast ‘understand’ usually governs an accusative object. By the use of
the genitive the speaker indicates that from his subjective point of view it is im-
possible to understand women. This subjective use of the genitive is reminiscent
of some uses of the partitive case in Finnish. For example, in Finnish the use of
the partitive instead of the nominative-accusative in the question ‘Do you have
a pencil?’ implies that the speaker doubts a positive answer, or, for reason of po-
liteness, suggests that a negative answer is acceptable (cf. Kiparsky 1998).
The second situation where the genitive marks the object of a transitive verb
in Latgalian and Lithuanian is what is traditionally called the partitive genitive.
A more accurate term is ‘genitive of indefinite quantity’, as it is used with mass
nouns in the singular and count nouns in the plural to express an indefinite quan-
tity or an unspecified number.
(13) Jī i maiz-is pierka i trauk-u.
3pl.nom and bread-gen.sg buy.pst.3 and bowl-gen.pl
‘They bought bread and dishes’
Lithuanian
(14) Jie sugrįžo, valgė duon-os ir
3.nom.pl.m return.pst.3 eat.pst.3 bread-gen.sg and
gėrė vanden-s jo nam-uose.
drink.pst.3 water-gen.sg 3.gen.sg.m house-loc.sg
‘They went back, ate bread and drank water in his house.’ (The Bible, 1 Kings
13 : 19; New King James Version: So he went back with him, and ate bread in
his house, and drank water.)
(15) Gavau laišk-ų.

receive.pst.1sg letter-gen.pl
‘I received some letters.’ (Ambrazas, ed. 1997: 486)
In contrast, the use of the accusative may signal a definite, known quantity:
Lithuanian
(16) Gavau laišk-us.
receive.pst.1sg letter-acc.pl
‘I received the letters.’ (Ambrazas, ed. 1997: 486)
The alternation is not present in Latvian, where only accusative marking is pos-
sible if there is no overt quantifier:
Latvian
(17) Saņēmu vēstul-es.
receive.pst.1sg letter-acc.pl
‘I received letters / some letters / the letters.’
Definiteness is not the only parameter associated with this case alternation. More
important, especially in Lithuanian, is the association of case marking with aspec-
tuality.. In Lithuanian (unlike what we observe in Finnic) the notion of indefinite
quantity introduced by the partitive genitive is incompatible with the incremental
quantification implied by (progressive) imperfectives:
(18) Šeima kaip tik sėdėjo prie stal-o
family-nom.sg just sit.pst.3 at table-gen.sg
ir gėrė arbat-ą / *arbatos.
and drink.pst.3 tea-acc / *tea.gen
‘The family was just sitting at the table and having tea.’
The association of the partitive genitive with the description of the event (rather
than with properties of the referent of the object noun phrase) is also reflected
in its co-occurrence with verbal prefixes that mark the event as bounded, for ex-
ample the delimitative prefix pa-; see Seržant, this volume, for a thorough analysis
of the partitive genitive in Lithuanian and the various factors that determine its
use. Another instance of the interplay of verbal prefixes and case marking is ac-
cumulation: the meaning ‘having a large amount of something as the result of
an action’ is jointly produced by the genitive marking of the object and the ac-
cumulative verbal prefix Lithuanian pri- (and similar Latgalian pī‑, see Nau, this
volume), for example:
Lithuanian
(19) Ji pri-kepė pyrag-ų.
3nom.sg.f pfx-bake.pst.3 pie-gen.pl
‘She (has) baked a lot of pies.’ (Ambrazas, ed. 1997: 606)
Differential object marking of this type – the choice of the genitive instead of the
accusative in the context of negation, irrealis, or quantification – and the interplay
of object marking and verbal prefixes has so far been mainly investigated in Slavic
languages, for example in recent publications by Filip (2001, 2005), Borschev,
Partee and others (Borschev et al. 2008; Partee et al. 2011, 2012), or Kagan (2010,
2013); for a bibliography of earlier publications see Corbett (1986). The Lithu-
anian and Latgalian data will certainly enrich the discussion.
In Latvian, another type of alternation between the accusative and the dative
is found with some verbs of physical impact, most prominently the verb sist ‘beat,
strike, hit’. This verb is commonly used with the nominative-accusative frame, as
in example (20):
(20) es ne-spēju saprast cilvēk-us,
1sg.nom neg-be.able.prs.1sg understand.inf human-acc.pl
kas sit sun-i
who.nom hit.prs.3 dog-acc.sg
‘I am unable to understand people who beat a dog’
(http://www.sapforums.lv/index.php?t=46747)
The use of the dative instead of the accusative is associated with certain meanings,
which however are not always easy to formulate. Consider the following example,
which is very close in meaning to the previous one:
(21) Man šokēja tas, ka
1sg.dat shock dem.nom.sg.m that
sit sun-im, kur-š rej…
hit.prs.3 dog-dat.sg which-nom.sg.m bark.prs.3
‘I was shocked [to read] that [people] beat a dog that barks…’
(http://www.calis.lv/forums/tema/18153059-ka-palidzet-bailigam-
tramigam-sunim/1/)
Other verbs that show this alternation are spert ‘kick’, durt ‘stab; prick’ or kost ‘bite’
and a few others. The choice of the dative or the accusative is not associated with
properties of the agent or the patient, which in both situations are clearly distin-
guished by their prototypical features (volitional agent, affected patient). What
might have triggered the dative in example (21) is that the beating takes place in
a specific situation (which then is reiterated): each time a dog barks it receives a
blow. Mühlenbach (1902/1903: 222) characterized the difference as one of aktion-

sart: an opposition between semelfactive and durative. According to him, the da-
tive marks a single act and he translates sist sunim as “dem Hunde einen oder auch
einige Hiebe versetzen” (‘to strike the dog a blow or several blows’) while in sist
suni the accusative has an intensive or durative meaning translated into German
by “den Hund schlagen, hauen” (‘to beat the dog’). Indeed, in many instances the
construction with the dative is best translated as ‘strike someone a blow’, as in the
following example, where there is only one blow which even misses its intended
target:
(22) Strādāju naktsklubā un man ceļā maisījās kāds pārītis, kas skaidrojās ar
vārdiem.
Pēkšņi meiten-e sita puis-im
suddenly girl-nom.sg hit.pst.3 boy-dat.sg
un trāpīja man pa sej-u.
and reach.pst.3 1sg.dat on face-acc.sg
‘I was working at a night club and there was a couple who came in my way.
They were arguing. Suddenly the girl struck the boy a blow and hit my face.’
(http://cosmo.lv/forums/topic/114941-/?sort=desc&pnr=2#postid-487591)
In this example, and in several other instances found with sist + dative, the focus
is on the action carried out by the speaker, while the affectedness of the patient
is not important. One may thus speculate that the use of the dative implies that
the patient is less affected. However, in other instances the opposite seems to be
true: the dative can be used to foreground the affectedness of the patient, treating
it as an experiencer or recipient – someone receiving a blow, as in the following
newspaper headline where the verb is used metaphorically:
(23) Īr-es tirg-us sit rīdziniek-iem.
rent-gen.sg market-nom.sg hit-prs.3 inhabitant.of.Riga-dat.pl
‘The market for (apartment) rents hits the inhabitants of Riga.’
(http://news.lv/Latvijas_Vestnesis/2006-11-01/Ires_tirgus_sit_ridziniekiem)
The degree of affectedness of the dative marked objects of verbs such as sist ‘hit’
thus may vary. The dative is however never used for objects of transitive verbs
that imply a change of state, such as nosist ‘kill’, nodūrt ‘stab to death’, sakost ‘cause
severe harm by biting’.
It is possible to treat sist with dative and sist with accusative as separate verbs
with different case frames. The transitive verb sist₁ ‘beat, hit’ would then display
canonical case marking, while sist₂ ‘hit, strike, deal a blow’ would show non-ca-
nonical case marking, just as the verb trāpīt ‘hit (a goal), reach’, which always
combines with the dative. The meaning ‘strike someone a blow’ is also expressed
by the prefixed verb iesist that in this meaning governs the dative. The same would
then apply to the other verbs in this group. However, the semantic difference be-
tween uses of sist, spert, durt etc. with the accusative and with the dative are prob-
ably not consistent enough to justify the distinction of two lexical units.
2.2.2 Clause-type-specific differential object marking

Special types of marking are used in subordinate and embedded clausal structures
with non-finite forms (for Lithuanian cf. Arkadiev 2013).
The genitive marks the object of a transitive verb in purpose clauses with a
non-finite verb-form depending from a verb of motion. In Latgalian there is a
special form for these constructions, the supine. It is an inherited form that is
also found (or was found in the 20th century) in Central dialects of Latvian and
in eastern dialects of Lithuanian. In most dialects and in the standard languages,
the supine has been replaced by the infinitive. Lithuanian has, however, carried
over the genitive marking from the supine to the infinitive, whereas in Latvian the
accusative is used as in other infinitival clauses.
(24) jei nu-skrēja vylk-a systu
3sg.f.nom pfx-run.pst.3 wolf-gen.sg beat.sup
‘she ran out to beat the wolf ’
Central Latvian dialect with supine (Raģe 1964: 117)

(25) Valis àizgãja gùos ķęŕ-tu
Valis.nom go.out.pst.3 cow.gen.sg catch-sup
‘Valis went out to catch the cow.’
Standard Lithuanian (Ambrazas, ed. 1997: 557)

(26) atėjo draug-o aplankyti
come.pst.3 friend-gen.sg visit.inf
‘[he] came to visit a friend’
Standard Latvian
(27) Ērik-s devās apciemot tuv-us draug-us
Ēriks-nom leave.for.pst.3 visit.inf close-acc.pl.m friend-acc.pl
‘Ēriks went to visit (some/his) close friends’ (Mio-2)
The Lithuanian construction has been studied together with another purpose
construction, where an infinitive combines with the dative, for example:
Lithuanian (Ambrazas, ed. 1997: 557)

(28) iššove žmon-ėms pagąsdinti
shoot.pst.3 man-dat.pl scare.inf
‘[he] fired to scare people’
In this volume, Peter Arkadiev analyses Lithuanian purpose constructions with

a genitive or a dative, starting with a critical discussion of a recent account by
Franks and Lavine (2006).
The nominative is sometimes used for the object of an infinitive in embed-
ded clauses whose implicit subject is not controlled by a main clause nominative
subject, and in impersonal independent infinitival clauses. The use of nominative
objects with an infinitive is mainly observed in dialects and has caught the atten-
tion of linguists as an areal phenomenon shared by Finnic, Slavic and East Baltic
languages (see Larin 1963; Timberlake 1974; Ambrazas 2001). The main type is
impersonal necessitive constructions like the following:
East High Lithuanian dialect (Ambrazas 2001: 391)
(29) Reikia šien-as grėbti
need.prs.3 hay-nom.sg rake.inf
‘It is necessary to rake the hay’
Central Latvian dialect (Endzelin 1922: 409)

(30) vaiga issukât galv-a
need.prs.3 pfx.comb.inf head-nom.sg
‘One has to comb out the head’
This type of marking, however, differs from all the preceding ones in that it oc-
curs only in a specific syntactic configuration: the infinitive has an implicit subject
(agent), but it cannot be realized syntactically, and the fact of the first argument
position being blocked is probably a precondition for the use of the nominative,
as argued by Comrie (1975).
Another issue that has been mentioned in connection with the nominative
object with the infinitive is the specific type of marking occurring with the debi-
tive, a form of the Latvian verb expressing necessity. When a verb takes the debi-
tive form, its nominative subject goes into the dative and its accusative object into
the nominative (except for first and second person pronouns and the reflexive
pronoun):
Latvian
(31) Bērni lasa grāmat-u.
child-nom.pl read.prs.3 book-acc.sg
‘The children are reading a book.’
(32) Bērniem jālasa grāmata.

child-dat.pl deb-read book-nom.sg
‘The children have to read a book.’
To call this nominative an object would, however, be problematic. The debitive

construction (though originally bi-clausal) is now monoclausal. Within this
monoclausal debitive construction the configuration dat–nom is too far from the
canonical configuration nom–acc for such notions as object, which are inherent
to the transitive clause model, to be applicable. Neither can the second argument
be regarded as a subject; Holvoet and Grzybowska (this volume) argue that gram-
matical relations with the debitive are diffuse.
2.2.3 Differential subject marking

Alternative ways of marking intransitive subjects occur in Baltic as well. Though
with considerably lesser regularity than is seen with objects, the genitive is used
under negation instead of the nominative with verbs that have existential implica-
tions in Lithuanian. The following examples are taken from the recent study by
Aleksandravičiūtė (2013).
Lithuanian (Aleksandravičiūtė 2013)
(33) Kuršėn-uose ne-liko darb-o –
Kuršėnai-loc neg-remain.pst.3 job-gen.sg
uždaryta plyt-ų gamykl-a.
close.ppp.sg.f brick-gen.pl factory-nom.sg
‘There was no job left in Kuršėnai – the brick factory was closed.’
(34) Mokykl-oje ne-buvo fizik-os mokytoj-o.
school-loc.sg neg-be.pst.3 physics-gen teacher-gen.sg
‘There was no physics teacher at the school.’
This use can also be found in Latgalian, as in the following example from a fairy
tale:
Latgalian
(35) a zalt-a naud-ys na-pa-za-ruodēja vys
but gold-gen money-gen neg-pfx-refl-show.pst.3 ptc
‘but the golden money didn’t show up at all!’
In Latvian this genitive has been retained only with the negated verb būt ‘be’.
Very rarely it occurs with other verbs, as the use of the genitive has been severely
restricted when compared with Lithuanian.
Such alternations, now usually described as differential subject and object
marking, stand alongside other alternations opposing different assignments of
objecthood (with verbs of the ‘spray/load’ type) and subjecthood (with verbs
of the ‘swarm’ type). Baltic is in line with many other languages here, but, as in
Slavic, the number of alternating patterns with the ‘swarm’ type is greater than,
say, in English. It also includes a pattern without apparent subject assignment.
We will here illustrate the phenomenon with examples from Latvian (Lithuanian
examples are discussed by Kristina Lenartaitė-Gotaučienė, this volume). In the
first of the following sentences the verb mudžēt ‘teem’ combines with a nomina-
tive and a prepositional argument, where the nominative is used for the place (lit.
‘the yard was teeming of sparrows’). In the second example, the place is marked
with the locative and the actor with the nominative (lit. ‘in Latvia studios were
teeming’), while the third example shows locative and prepositional marking (lit.
‘in the square it was teeming of people’), thus there is no nominative subject. Note
that in all three examples the word-order is Place–Actor.
Latvian
(36) Cit-us rudeņ-us mūsu sēt-a mudžēja no
other-acc.pl autumn-acc.pl our courtyard-nom.sg teem.pst.3 of
zvirbuļ-iem un cit-iem sīk-iem putniņ-iem.
sparrow-dat.pl and other-dat.pl.m small-dat.pl.m bird-dat.pl
‘In other autumns our courtyard used to be teeming with sparrows and other
small birds.’
http://meeting.oho.lv/meeting.php?cmd=dienasgr_pub&did=677138&kid=
388163
(37) Vēl pirms pus gadsimt-a Latvij-ā mudžēt mudžēja
still before half century-gen.sg Latvia-loc teem.inf teem.pst.3
dažād-as kinoamatier-u studij-as.
various-nom.pl.f film.amateur-gen.pl studio-nom.pl
‘As late as fifty years ago various amateur film studios were still teeming in
Latvia.’
http://www.ltv.lv/lv/raksts/06.01.2013-province.-laika-glabataji-.id5138/
(38) Vakar-ā pilsēt-as centrāl-ajā laukum-ā
evening-loc.sg town-gen.sg central-loc.sg.def square-loc.sg
mudžēja no cilvēk-iem …, saldum-u un mantiņ-u
teem.pst.3 of people-dat.pl sweet-gen.pl and small.item-gen.pl
pārdevēj-iem.
vendor-dat.pl
‘In the evening the central town square was teeming with people […], sweets-
and bric-à-brac vendors.’
http://unenanne.tumblr.com/Makedonija
Such alternations are usually dealt with in the context of the syntax-semantics
interface, and they appear to be lexically anchored, though, on the other hand,
the existence of recurrent patterns allowing these alternations to be subsumed
under a few major types (the ‘swarm’ type, the ‘spray paint’ type etc.) seems to call
for a constructional treatment. An overview of ‘projectionist’ approaches (those
projecting such alternations from the lexicon, e.g., through lexical rules) and
those invoking the notion of construction is given in Levin and Rappaport Hovav
(2005: 186–236). Especially on a constructional approach it seems tempting to
establish a connection with certain grammatical alternations discussed under the
heading of differential marking. Is there, for instance, a difference of principle
between the subjectless marking strategy with mudžēt in (38) and the subjectless
structures with ‘be’? Compare, in Lithuanian, the similarity between the subject-
less construction with knibždėti (the counterpart of Latvian mudžēt) and those
with ‘be’ and a partitive genitive:
(39) Filharmonij-oje knibždėjo įvairiausi-ų žmoni-ų.
philharmonic-loc.sg swarm.pst.3 various-gen.pl people-gen.pl
‘The concert hall was swarming with all kinds of people.’
(from Lenartaitė-Gotaučienė, this volume)
(40) Mūsų Lietuv-oje yra įvairiausių žmonių […]
our Lithuania-loc be.prs.3 various-gen.pl people-gen.pl
‘There are all kinds of people in this Lithuania of ours…’
http://www.asirpsichologija.lt/index.php?id=470&category=4
An interesting question is whether these sentences contain instances of the same

genitive, or whether they are really different cases, one lexical and the other
structural. Whatever the answer, both constructions pose similar problems from
the point of view of grammatical relations. In her article, Kristina Lenartaitė-
Gotaučienė (this volume) examines the problem of subjecthood in Lithuanian
constructions with predicates of the swarm type, posing the question whether the
notion is, in any useful sense, applicable here.
The phenomena mentioned until now will already have sufficed to make clear
that there is no sharp line of division between lexically restricted grammatical
marking alternations (grammatically conditioned alternations applying to some
verb classes) and variety between lexically assigned marking patterns (different
marking patterns for different lexemes within the broadly defined class of two-
place predicates). As Holvoet and Grzybowska point out, the marking pattern
grammatically assigned by the Latvian debitive coincides with a pattern lexically
assigned by a smallish class of verbs. We can conveniently distinguish different
types of factors associated with deviant encoding – individual lexical require-

ments, grammatically determined splits, clause-type related splits etc., but all
types of conditioning may reflect common underlying mechanisms. Probably the
same factor underlies the use of a subjectless construction in the Latvian example
with mudžēt (38) and the indeterminacy of (otherwise canonical) subject assign-
ment in the English counterparts with swarm.
3. Baltic in a typological context: Alternatives to the transitive pattern
With two-place predicates, the transitive pattern (nominative subject and accu-
sative object) is used by the largest group of verbs. In a pilot study carried out
with the material of a typological project of the Institute of Linguistic Studies at
St Petersburg, Natalia Perkova investigated non-canonical marking of two-place
predicates in Latvian and Lithuanian in comparison with other European lan-
guages. In her sample of 129 verbs, 63 of the Latvian and 55 of the Lithuanian
equivalents are used with nominative-accusative marking (Perkova, MS). These
figures may not be representative for the languages at large, but they show two
tendencies that certainly hold: first, although it is a dominant pattern, nomina-
tive-accusative marking is less prevalent in the Baltic languages than in most SAE
languages, and second, the dominance of the pattern is greater in Latvian than in
Lithuanian. One reason is that Latvian has more or less given up the genitive as
a case for objects without preposition,8 and the translation equivalents of several
Lithuanian verbs governing a genitive are transitive verbs in modern Latvian, for
example verbs meaning ‘wait’, ‘search’, ‘loose’, ‘want’, ‘hate’, ‘listen to’ and ‘learn’.
Examples from the parallel corpus of translated fiction (LiLa) – all examples
are from texts originally written in Lithuanian and translated into Latvian.
(41) a. Na, o aš mokausi magij-os!
ptc but 1sg.nom learn.prs.3 magic-gen.sg Lithuanian
b. Jā, bet es mācos maģij-u!
ptc but 1sg.nom learn.prs.3 magic-acc.sgLatvian
‘Well, but I am studying magic!’
(42) a. Kaip bit-ė ieško med-aus vis-ose gėl-ėse,
like bee-nom.sg search.prs.3 honey-gen all-loc.pl flower-loc.pl
ieškok Mokym-o visur
search.imp.2sg lesson-gen.sg everywhere Lithuanian
8. Cf. Berg-Olsen (1999, 2000).

b. Tāpat kā bit-e meklē med-u vis-os

such as bee-nom.sg search.prs.3 honey-acc all-loc.pl
zied-os, meklējiet Mācīb-u visur
flower-loc.pl search.imp.2pl lesson-acc.sg everywhere Latvian
‘As the bee is looking for honey in all flowers, look for a Lesson
everywhere’
(43) a. Placeb-as laukė tinkam-o moment-o
Placebo-nom wait.pst.3 favorable-gen.sg.m moment-gen.sg
Lithuanian
b. Placebo gaidīja izdevīg-u brīd-i
Placebo wait.pst.3 favorable-acc.sg.m moment-acc.sgLatvian
‘Placebo was waiting for a favorable moment’
Note that we are speaking here about translation equivalents – Lithuanian and
Latvian verbs with the same meaning, independent of whether they are cognates
(such as Lithuanian mokytis, Latvian mācīties ‘learn’) or not (such as Lithuanian
ieškoti, Latvian meklēt ‘look for’, or Lithuanian laukti, Latvian gaidīt ‘wait’). It
seems that in most instances the two languages use different lexemes, but it is
dubious (and a question for further research) whether the choice of root, rather
than its meaning, is responsible for the case marking. Evidence against such a
thesis is provided by the fact that in the past several of these verbs were governing
the genitive in Latvian, too. The genitive as a lexical case was still found in sev-
eral Latvian varieties (central dialects and the developing standard language) in
the 19th century and sporadically in the 20th century, and is still used in eastern
Latvian dialects and other varieties of Latgalian. Here we find the same verbs as in
standard Latvian, but the case marking differs, for example:
Latgalian
(44) Bit-e meklej mad-s.
bee-nom.sg search.prs.3 honey-gen.sg
‘The bee is looking for honey.’
For a discussion of verbs governing the genitive in Latgalian, see Nau, this
volume.
Verbs that select a genitive object in Lithuanian and Latgalian belong to sev-
eral semantic groups. They have in common that the second argument is less
affected than in the prototypical transitive situation. These data thus confirm cur-
rent theories of transitivity splits which predict that verbs selecting another case
frame than the transitive model are lower on the hierarchy of transitivity, a no-
tion going back to Hopper and Thompson (1980) and Tsunoda (1981). The data
do not fully correspond to the hierarchy originally proposed by Tsunoda (1985).
Table 1.
Tsunoda’s classes and examples Case-marking of the patient argument
Class Predicates in Lithuanian in Latvian
1 Direct effect on patient kill, break, bend accusative accusative
1A Resultative
1 Direct effect on patient hit, shoot9 accusative accusative
1B Non-resultative
2 Perception see, hear, find accusative accusative
2A Patient more attained
2 Perception listen to, look at genitive accusative
2B Patient less attained (‘listen to’)10 (‘listen to’)
3 Pursuit search, wait, await genitive accusative
Tsunoda placed verbs of pursuit (‘search’, ‘wait’) in the middle of the hierarchy,
higher than verbs of knowledge, feeling, or relationship. Now in Lithuanian and
Latgalian several verbs of these lower groups (especially verbs of knowledge) do
combine with the accusative. In Latgalian, where we witness a tendency to ex-
tend the transitive model (see Nau, this volume), the genitive with the verbs for
‘wait’ and ‘search’ is more stable than with the verb ‘love’. However, in its first half,
Tsunoda’s hierarchy holds (if we regard only the alternative accusative vs. geni-
tive): the split occurs between class 2A and 2B, as shown in Table 1.
Tsunoda’s hierarchy has been critically discussed by several researchers.
Malchukov (2005) gives a short summary of the critique and develops an alterna-
tive approach by decomposing the hierarchy and starting with a two-dimensional
semantic map instead of the one-dimensional scale. An important insight of his
discussion is that verbs of pursuit and verbs of knowledge or feeling deviate from
prototypical transitive predicates in different ways. With pursuit verbs, there is
only a less affected patient, while the predicates further to the end of Tsunoda’s
hierarchy involve also a decrease of agentivity of the first participant (Malchukov
2005: 90). The genitive objects of the verbs discussed here can be seen as a local
strategy reflecting non-prototypical transitivity from the side of the patient only.
9. The verbs that show the accusative–dative alternation discussed above belong to this
group.
10. With the predicate ‘look at’, the second argument is marked in both Lithuanian and Latvian
with a preposition or with the locative.
A more pervasive alternative marking strategy occurs when an experiencer

receives dative marking, a phenomenon highly characteristic of the Baltic lan-
guages and especially Latvian. These instances raise more fundamental problems
than those mentioned above because here subject properties become diffuse, even
if the construction contains a nominative argument that could raise reasonable
claims to subjecthood.
To start with, there are a number of constructions with one overt argument
marked with a dative rather than a nominative:
Lithuanian
(45) Man šalta.
1sg.dat be.cold.prs.3
‘I am cold.’ (Ambrazas, ed. 1997: 602)
(46) Jam visada sekasi.
3.dat.sg.m always be.lucky.prs.3
‘He is always lucky.’ (Ambrazas, ed. 1997: 600)
Latvian
(47) Man sāp.
1sg.dat hurt.prs.3
‘I am in pain.’
(48) Viņ-am vienmēr veicas.
3-dat.sg.m always be.lucky.prs.3
‘He is always lucky.’
Some of these verbs are alternatively used with a pattern where the experiencer is
in the nominative:
Lithuanian
(49) Ligoni-ui pagerėjo / Ligon-is pagerėjo.
patient-dat.sg get.better.pst.3 patient-nom.sg get.better.pst.3
‘The patient improved.’ (Ambrazas, ed. 1997: 630)
Latvian
(50) Man skauž / Es skaužu.
1sg.dat envy.prs.3 1sg.nom envy.prs.1sg
‘I am envious.’
Some non-verbal predicates compete with a middle voice (formally reflexive)

verb with nominative coding:
Latvian
(51) Man bail / Es baidos.
1sg.dat afraid 1sg.nom fear.prs.1sg
‘I am afraid.’
(52) Man prieks / Es priecājos.
1sg.dat joy 1sg.nom rejoice.prs.1sg
‘I am happy.’
There is no obvious semantic difference between the alternatives. For the last two
examples one may state that the first variant (with a non-verbal predicate and a
dative) is more colloquial. For the Latvian verb skaust ‘be envious’ the pattern
with the dative is much more frequent.
While this is a pattern the Baltic languages have in common, there are also
some differences between Lithuanian and Latvian regarding its range of applica-
tion. For example, only in Lithuanian do we also find monovalent verbs with an
experiencer in the accusative (verbs denoting unpleasant physiological experi-
ence, see Wiemer & Bjarnadóttir, this volume, Section 3.2.2.3). In Latvian, ac-
cusative-marked experiencers are only found together with a nominative subject.
According to Bickel and Nichols (2011), this latter situation (though also found
in Latin) is less common. Languages that code experiencers in clauses with two
arguments as O, G, or T,11 usually extend this marking to experiencers that are
the single argument of an intransitive clause (Bickel & Nichols 2011: 315). Instead
of extending the accusative coding to arguments of monovalent verbs, Latvian
rather seems to abolish accusative marking of experiencers as arguments of biva-
lent verbs (see below our discussion of “dative drift”).
Constructions like those illustrated here might, at first sight, seem to be remi-
niscent of split intransitivity (see Nichols 2008 for such an approach). However,
this type of marking is basically restricted to certain types of predicators, those
of adjectival or nominal origin. Verbs occurring in clauses of this type, such as
Lithuanian sektis in (46), Latvian veikties in (48), usually occur also in a two-place
construction in which the second argument is in the nominative, cf.
Latvian
(53) Kā Jums veicas darb-i tād-ā karstum-ā?
how 2pl.dat succeed.prs.3 work-nom.sg such-loc.sg heat-loc.sg
‘How are you getting on with your work in such heat?’
https://twitter.com/Bitelv/status/94014355779235840
11. O[bject] = (direct) object of a two-place argument, G[oal] = indirect object of a three-place
argument, T[theme] = (direct) object of a three-place argument.
Sentences like (46), (48) could therefore be said to contain implicit nominative
subjects. In this sense these constructions constitute a transition to another type,
in which a dative experiencer regularly co-occurs with a nominative stimulus
argument:
Lithuanian
(54) Man patinka kat-ės.
1sg.dat like.prs.3 cat-nom.pl
‘I like cats.’ (Ambrazas, ed. 1997: 608)
Latvian
(55) Man patīk kaķ-i.
1sg.dat like.prs.3 cat-nom.pl
‘I like cats.’
Dative experiencers have been much discussed in the linguistic literature, es-
pecially with regard to their status as subjects or objects. Languages that figure
prominently in this discussion are South Asian languages as well as Icelandic and
other Germanic languages, to a lesser degree Russian (see, for example, Verma &
Monahan, eds., 1990; Bayer 2004; Bickel 2004; Moore & Perlmutter 2000). The
Icelandic data have become a kind of benchmark for ‘oblique subjects’ as they
pass a series of syntactic subjecthood tests formulated by Zaenen, Maling and
Thráinsson (1985) on the basis of Keenan’s well-known list of prototypical sub-
ject properties (Keenan 1976). Attempts to apply these tests to other Germanic
languages, notably German, have yielded poor results, cf. Andrews (2001) and
Sigurðsson (2004). The Baltic data have also come under scrutiny lately, see
Holvoet (2013) for a discussion of non-canonical subjects in Baltic and elsewhere,
and Holvoet (forthcoming) specifically on Latvian; see also Seržant (2013b) on
the canonicization of such structures, and Seržant (forthcoming) for a diachronic
and contact linguistic account of dative experiencers in Baltic and areally related
languages. In the following we will focus on Latvian.
There is of course no straightforward answer to the question whether the da-
tive experiencers in Latvian and Lithuanian are subjects or objects (or neither).
First, the answer will depend on the theoretical framework or approach towards
grammatical relations, for example, whether they are seen as global or as con-
struction-specific (Barðdal 2003; Bickel 2011), or whether they are defined as
universal or as language specific – for example, Verma & Monahan (1990: 4) ar-
gue for language internal criteria to establish subjecthood in Hindi. As in this
introduction we endeavor to be as neutral as possible, we will only point out some
properties of Latvian dative experiencers that will have to be taken into account
in most frameworks.
The properties that we want to discuss here have been associated with gram-
matical subjects, but it is likely that they rather arise from the fact that experienc-
ers often are the topic of a sentence, and/or that they denote a sentient being, most
often a human and frequently the first person (cf. Bickel 2004; Næss 2011). This
becomes evident when examining word-order and control properties, which in
turn are related to each other.12
A dative Experiencer frequently appears at the beginning of a clause. The or-
der Experiencer–verb–Stimulus is probably the least marked, for example:
(56) Taču man skauž šis viņ-as perfektum-s.
but 1sg.dat envy.prs.3 dem.nom.sg.m 3-gen.sg.f perfection-nom.sg
‘But I am envious of this perfection of hers.’
(http://attiecibas.espati.lv/index/view/id/2158/page/2/Skaudiba-starp-
draudzenem-ka-rikoties/)
The Stimulus may also appear before the verb, which puts the verb into the rhe-
matic position but does not change the relation between the arguments. If how-
ever the Stimulus is in topic position at the beginning of the clause, the effect is
greater and often equal to a difference expressed in English by different lexical
items. A good example is the verb patikt, which, depending on which element is
topicalized will be translated as ‘like’ or ‘appeal to’.
(57) Viņ-am patīk šī idej-a.
3-dat.sg.m like.prs.3 dem.nom.sg.f idea-nom.sg
or:
Viņ-am šī idej-a patīk.
3-dat.sg.m dem.nom.sg.f idea-nom.sg like.prs.3
‘He likes this idea.’
(58) Šī idej-a viņ-am patīk.
dem.nom.sg.f idea-nom.sg 3-dat.sg.m like.prs.3
‘This idea appeals to him.’
The person hierarchy interferes with these rules: If the Stimulus is a first or second
person and overtly expressed by a pronoun, the unmarked word-order is Stimu-
lus–Experiencer. A nominative pronoun in the position behind the verb at the
end of the clause is always emphasised. Compare:
12. Compare also Primus (2011: 200): “… a common structural property of experiencers with
non-causative verbs that can be explained by their classification as proto-agents. Despite their
variable case coding, they are always (or preferably) structural subjects, i.e. they occur sen-
tence-initially in terms of basic order […]. This structural function explains other syntactic
subject properties of oblique experiencers in many languages”.
(59) Es viņ-am patīku.

1sg.nom 3-dat.sg.m like.prs.1sg
‘He likes me.’, ‘I appeal to him.’
(60) Viņ-am patīku es.
3-dat.sg.m like.prs.1sg 1sg.nom
‘It’s me he likes.’, ‘It’s me who appeals to him.’
As can be seen in the last two examples, agreement is independent of word-or-

der. Unlike word-order, agreement is strictly grammatical. Only a nominative ar-
gument can trigger verb agreement. In the third person, agreement can be seen
when the verb takes the form of a participle, for example in the perfect tense:
(61) Viņ-am šī idej-a vienmēr ir patikusi.
3-dat.sg.m dem.nom.sg.f idea[f]-nom.sg always be.prs.3 like.ppa.f.sg
‘He has always liked this idea.’
Word-order is thus not a subject parameter in Latvian, unless subject is equated

with topic (as often happens in traditional grammar). The situation is similar with
regard to reflexive pronouns. Both a dative Experiencer and a nominative Stimu-
lus may control a reflexive pronoun. The controller does not have to be overtly
expressed.
(62) Kas tev ne-patīk sev-ī?
what.nom 2sg.dat neg-like.prs.3 refl-loc
‘What don’t you like in yourself?’
(http://amigos.lv/lv/qna?id=74485)
(63) ne-patīk paš-ai sev-ī tas.
neg-like.prs.3 emph-dat.sg.dat refl-loc dem.nom.sg.m
‘That’s something I don’t like in myself.’
(http://www.atceries.lv/lv/diskusijos.zinutes/4520)
(64) Dažreiz es pat-s sev apnīku.
sometimes 1sg.nom emph-nom.sg.m refl.dat tire.of.prs.1sg
‘Sometimes I am tired of myself.’
(http://vejaskrejejs.blogspot.com/2011/07/dazreiz-apnik-dazreiz-nezinu-ko-
darit.html)
The reflexive possessive pronoun may be controlled by the Experiencer as well,

but its choice is not obligatory: quite often a non-reflexive pronoun is found in
these constructions.
(65) [man] apnīk sav-s skat-s spogul-ī

1sg.dat tire.of.prs.3 reflposs-nom.sg.m sight-nom.sg mirror-loc.sg
‘[I am] tired of my sight in the mirror’
(http://www.maminuklubs.lv/forum/20120829-2011-gada-junija-briniskigas-
mamminas-ar-berniniem/?page=128)
(66) Visvairāk [man] kremt man-a bezspecīb-a.
most.of.all 1sg.dat annoy.prs.3 poss1sg-nom.sg.f weakness-nom.sg
‘I am especially annoyed by my weakness.’ /
‘My weakness is what annoys me most.’
(http://www.calis.lv/forums/tema/18159165-jaukaa-gruutnieciiba/1/)
Which option is more frequent and which factors influence the choice is a ques-
tion for further empirical research. In any case the Latvian data confirm that con-
trol of reflexive pronouns is associated with information structure and semantic
role rather than with syntactic subjecthood (cf. Haspelmath 2001; Bickel 2004;
2011; for Latvian cf. Holvoet & Grzybowska, this volume).
While both dative and accusative can mark experiencers in Latvian, they are
not equally prominent: the dative dominates. Several of the verbs that accord-
ing to normative grammar combine with an accusative are also found with a da-
tive in independent constructions,13 for example the verb interesēt ‘interest, be of
interest’:
(67) Mani ļoti interesē šī tēm-a.
1sg.acc very.much interest.prs.3 dem.nom.sg.f topic-nom.sg
‘I am very interested in this topic.’ (Mio-2)
(68) Tēt-im ļoti interesē politik-a.
daddy-dat.sg very.much interest.prs.3 politics-nom.sg
‘Daddy is very interested in politics.’ (Mio-2)
No difference in meaning has been found between the variants. Individual verbs
differ in how often the dative is used and for how long it has been attested. Our
general impression is that there is a kind of “dative drift” that has been going on
at least since the early 20th century,14 as witnessed by examples like (69), (70).
The pattern with a dative argument expressing an experiencer and a nominative
13. On this variation see also Berg-Olsen (2005: 114–118), to whose conclusions and call for
further research we subscribe.
14. We prefer to speak of a dative drift and not of ‘dative sickness’, a term used for the descrip-
tion of similar phenomena in Icelandic and evidently stemming from a prescriptive approach
to grammar.
expressing a stimulus or theme is extended to predicates that were originally used

with another case frame.
(69) Pēdējā jau arī par to nemaz nesajuta skaudību, jo
viņ-ai tā liet-a lieliski amizēja.
3-dat.sg.f that.nom.sg.f affair-nom.sg greatly amuse.pst.3
‘The latter didn’t feel any jealousy on that account, for the whole matter
amused her greatly.’ (Augusts Deglavs, 1910–1911)
(70) Protams, baltu patriot-iem jau ļoti
of_course Baltic patriot-dat.pl ptc very
kairināja dabūt viņ-u rok-ā.
tempt.pst.3 get.inf 3-acc.sg hand-loc.sg
‘Of course the Baltic-German patriots were very keen on laying their hands
on him.’ (Augusts Deglavs, 1910–1911)
The verb amizēt, from German amüsieren, governs an accusative in the original
language; kairināt ‘stimulate, entice, tempt’ has causative morphology and would
therefore be expected to govern an accusative. But while the tendency to replace
accusative marking of the experiencer with a dative is unmistakable, it has not
yet become quite predominant and the two models still compete. For example, in
the 3.5 million word corpus of modern standard Latvian (Mio-2), the 3rd person
present tense form interesē ‘interests’ appears 98 times with an accusative and 18
times with a dative argument. For the verb-form kaitina ‘annoys’ the respective
figures are 13 and 1, while the form uztrauc ‘worries’ is used 33 times with an
accusative and never with a dative (the use with a dative is attested in Internet
sources, but it seems to be marginal).
The dative drift manifests itself in several ways – not only in the substitution
of a dative for an original accusative in the marking of the second argument, but
sometimes also in a complete reorganization of the case marking pattern. An in-
teresting example is simpatizēt ‘be sympathetic’ (cf. Berg-Olsen 2005: 113–114).
The original pattern with this verb corresponds to the English construction X is
sympathetic to Y, where X is the person experiencing the emotion (nom in Eng-
lish and Latvian) and Y the stimulus or theme (dat in Latvian). But marking the
stimulus with the dative is unusual in Latvian. It may reflect a construal in which
the second argument is a goal towards which the emotion is directed, or a recipi-
ent. While this is the only pattern documented in dictionaries, another pattern
is actually more frequent in contemporary Latvian, viz. one with reversed case
marking: the experiencer is marked with the dative and the stimulus/theme with
the nominative. Compare:
(71) Lai gan visi viņi ir zagļi un krāpnieki, viņi gūst panākumus,
un Defo viņ-iem simpatizē.
and Defoe 3-dat.pl.m be.sympathetic.prs.3
‘While all of them [= the characters in Defoe’s novels] are thieves and swin-
dlers, they meet with success, and Defoe is sympathetic to them.’
(http://www.kurbijkurne.lv/forums/lofiversion/index.php?t25317.html)
(72) Visi mani paziņas grib balsot par sociāldemokrātiem,
jo viņ-iem simpatizē Jur-is Bojār-s.
because 3-dat.pl.m be.sympathetic.prs.3 Juris-nom Bojārs-nom
‘All my friends want to vote for the social democrats because they are sympa-
thetic to Juris Bojārs.’
(http://aluksniesiem.diena.lv/laikraksta-arhivs/kadas-ir-jusu-prieksvelesanu-
prognozes-66546)
Note that the word-order in both examples is Experiencer–Stimulus, which cer-

tainly helps in construing the intended meaning. We suspect that the pattern
conforming to the general dative drift will ultimately oust the original one, as a
pattern where the dative marks the Stimulus is unusual for Latvian. In the cur-
rent situation, where both constructions are used, there is a subtle difference in
meaning and use: the receding construction with the dative stimulus is still often
used when the object of the affinity is an abstract concept (simpatizē reformām
‘is sympathetic to reforms’), an ideology, or a group of people representing an
ideology (simpatizē komunistiem ‘is sympathetic to the communists’), while the
construction with a dative experiencer is often synonymous with ‘like’, especially
when the object of the emotion is a single person. However, this difference is not
always observable and it seems that most speakers either use both constructions
synonymously, or use only one of them.
4. Differential and non-canonical marking
In this section we will briefly dwell on the notions that are usually invoked in
describing the structures overviewed in Section 3. They include the notions of
‘differential marking’ and ‘non-canonical marking’. Both are useful but fraught
with difficulties. The notion of differential marking is unproblematic as long as
we remain agnostic about grammatical relations; as soon as we start speaking of
‘differential marking of subjects’ or ‘differential marking of objects’, meaning the
first or second argument of a two-place predication respectively, we are begging
the question of whether something that is a subject in one language should also
be a subject in another. The notion of non-canonical marking is problematic for
similar reasons. While we are justified in speaking of non-canonical marking of

arguments, referring to patterns that do not conform to the canonical patterns for
transitive and intransitive clauses in the given language, speaking of ‘non-canoni-
cal marking of subjects’ is, again, begging the question.
As mentioned in Section 3, Baltic has numerous constructions with expe-
riencer arguments marked with the dative, in which subject properties seem to
be spread over two arguments. The question is not straightforward: if, alongside
a dative experiencer, we find a nominative-marked argument, one of the pos-
sibilities is that this argument is the subject. This would be the model described
by Onishi (2001) as S + E (intransitive subject plus extension), or rather (as the E
argument is clearly less oblique), E + S. But in fact we find not only nominatives,
but also accusatives and genitives marking the second argument:
(73) Bērn-iem patīk pasak-as un piedzīvojum-i –
child-dat.pl please.prs.3 fairytale.nom.pl and adventure-nom.pl
jo vairāk pārvērtīb-u un burvestīb-u, jo labāk.
the more metamorphosis-gen.pl and magic- gen.pl the better
‘Children like fairy-tales and adventures – the more metamorphoses and
magic the better.’
http://epadomi.lv/atputas_briziem/05042012-davanu_horoskops_ko_
davinat_katrai_zodiak
(74) Ir skaidr-s, ka katr-am sun-im
be.prs.3 clear-nom.sg.m that every-dat.sg.m dog-dat.sg
vajag māj-u.
be.needed.prs.3 home-acc.sg
‘It is obvious that every dog needs a home.’
http://manamsunim.lv/index.php?lang=lv
(75) Bet Matild-ei palika žēl sav-as
but Mathilda-dat become.pst.3 sorry refl.poss-gen.sg.f
mās-as, kaut gan viņa ir slikta,
sister-gen.sg though 3-nom.sg.f be.prs.3 bad-nom.sg.f
tomēr viņas māsa.
‘But Mathilda began to feel sorry for her sister: even though she’s wicked, it’s
her sister.’
http://www.stastulaiks.lv/lv/viesistaba/konkursa-darbi-balso!/930/
The case of iekāroties ‘feel a craving for’, which can show all three cases for the
marking of the second argument, is discussed in Holvoet (forthcoming). One
might be tempted, considering the accusative marking in (74), to assume that
this argument is an object. But while it is true that this argument corresponds to
the direct objects of some languages (cf. John needs money), anybody who would
want to call the accusative in (74) an object because in research on differential

marking it tends to be, as a second argument, compared to the objects of other
languages, would be making a methodological mistake, as a distinction has to be
made, as Haspelmath (2010) points out, between comparative concepts (such as
O – the second argument in two-place predications, potentially though not al-
ways identified with the object of canonical transitive structures) and descriptive
categories (such as the direct object in a specific language). The notion of object
presupposes a transitive structure, and the pattern dat–acc is not a canonical
transitive pattern. By default we must assume that the structure in (74) is intran-
sitive and that the accusative is not an object – unless tests should yield strong
evidence to the contrary. In fact, all three structures illustrated in (73)–(75) are
instances of a similar pattern, which is intransitive but does not conform to the
pattern E + S. Even if the second argument is marked with the nominative, is does
not show consistent subject properties; if it is oblique-marked, it will have even
fewer subject properties (for details cf. Holvoet 2013). This situation can be char-
acterized as one of diffuseness of grammatical relations.
A similar situation can be observed with verbs of pain. In Latvian, the expe-
riencer with a verb of pain is in the dative, while the name of the body part is in
the nominative:
(76) Bija tāda dziesmiņa par Jānīti, kurš svētdien bija vesels,
bet pirmdien viņ-am sāpēja galv-a.
but on_Monday 3-dat.sg.m ache.pst.3 head-nom.sg
‘There was a song about one Johnny, who was OK on Sunday but had a head-
ache on Monday.’
http://news.lv/Rigas_Balss/1995/05/04/Pirmdien-Janitim-sapeja-galva
In Lithuanian, on the other hand, there is some variation: in some regional and
idiolectal varieties the name of the body part is in the nominative, in others (and
increasingly so in the standard language) it is in the accusative:
(77) Dar iki dabar man skauda koj-os,
still until now 1sg.dat hurt.prs.3 leg-nom.pl
gydausi pūsles ant pėdų.
‘Even now my legs hurt and I am treating the blisters on my feet.’
http://milinoff.wordpress.com/page/2/
(78) Po įvyki-o jam skaudėjo rank-as,
after accident-gen.sg 3.dat.sg.m hurt.pst.3 hand-acc.pl
bet šiuo metu viskas praėjo.
‘After the accident his hands hurt, but now everything is over.’
https://www.google.com/#q=%22jam+skaud%C4%97jo+rankas%22
These differences between and within Baltic languages pose the question of which
construction was the original one, and the answer to this question will partly de-
termine the account one could give for the alternation of case assignments and,
in particular, the marking of the second argument. Recent publications yield a
considerable diversity of approaches to these constructions, invoking alignment
differences (Piccini 2008), obliqueness clines and adjustments (Holvoet 2013)
and competing conceptualization frames (Berg-Olsen 2005: 68–70, 113–114 for
Latvian, Seržant 2013 for Lithuanian).
The accusative marking in the Lithuanian construction, rather atypically oc-
curring without concurrent nominative marking on another argument, is reminis-
cent of another type of Lithuanian constructions involving one-place predications
with accusative marking on the only (overt) argument:
(79) Pamenu, kaip neseniai skridau:
mane pykino, aš baiminausi, kad
1sg.acc sicken.pst.3 1sg.nom fear.pst.1sg that
vėluoju į kitą lėktuv-ą.
be.late.prs.1sg in other-acc.sg aeroplane-acc.sg
‘I remember a recent flight: I felt sick and was afraid to be late for the connect-
ing plane.’
This construction is, however, different in that the marking on the verb is overtly
causative, and it may take a nominative subject:
(80) Net ir mažiausi-as kiek-is alkoholi-o
even smallest-nom.sg.m quantity-nom.sg alcohol-gen.sg
ar tabak-o kvap-as mane pykina.
or tobacco-gen.sg smell-nom.sg 1sg.acc sicken.prs.3
http://www.londoniete.lt/namai/utopija-atgal-i-gamta
‘Even the smallest quantity of alcohol, or the smell of tobacco, make me
sick.’
Verbs of pain are, on the other hand, basically intransitive. A nominative causer
subject may appear with Latvian sāpēt:
(81) Man tik ļoti sāpēja viņ-a rīcīb-a,
1sg.dat so much hurt.pst.3 3-gen.sg.m behavior-nom.sg
tik ļoti, ka es pat nespēju uz viņu paskatīties.
‘His behaviour hurt me so much, so much that I couldn’t even look at him.’
http://spoki.tvnet.lv/literatura/Gerda-34/672514
In the Lithuanian construction with the accusative of the body part, however, no
stimulus subject in the nominative can surface:
(82) Nuo ryški-os švies-os jam skaudėjo ak-is.

from intense-gen.sg.f light-gen.sg 3.dat.sg.m hurt.pst.3 eye-acc.pl
‘His eyes hurt from the intense light.’
( 83) *Ryšk-i švies-a jam skaudėjo ak-is.
intense-nom.sg.f light-nom.sg 3.dat.sg.m hurt.pst.3 eye-acc.pl
Intended meaning: ‘The intense light hurt his eyes.’
This shows, on the one hand, that the accusative in (78) does not align with the
clearly object-like accusative in (79), but the nominative in the alternative con-
struction (77) cannot be construed as a subject with the semantic role of stimulus
either. The alternation, then, opposes a nominative without clear subject proper-
ties to an accusative without clear object properties.15
The problems posed by the constructions dealt with in this section are more
fundamental than those posed by the classical instances of differential argument
marking. If the alternation occurs at the lower end of the two-place predication,
we are still able to identify the first argument as a subject and are in doubt only as
to the object status of the second argument. Problems arise, of course, when we
are to draw the borderline between two-place predicates that are low in transitiv-
ity (look at the pictures) and one-place predicates with a second argument that is
not object-like (sit on a bench). But what if we are not able to identify either of the
two arguments of a two-place predicate as either a subject or an object? For situ-
ations like this it seems preferable to use the notion of non-canonical marking, as
it is clear that cases like these will occur in the zone of low transitivity, where the
canonical notions of subject and object are not strictly applicable. Applying the
notions of differential marking or splits to such situations is problematic because
we do not know whether, investigating morphosyntactic marking across lexemes,
we are dealing with comparable things, that is, e.g., whether the first arguments
with two different two-place predicates will in both instances be subjects.
The situation is, at any rate, more complicated than could appear at first sight.
We have, in different languages, canonical two-place structures with nominative
subjects and accusative objects. We can also imagine structures where the nomi-
native and/or accusative are not assigned but language-internal evidence allows
us to call the first argument a subject and the second argument an object. But
there is also a third type of situation in which the type of case marking is non-ca-
nonical and grammatical relations are diffuse. In this case it makes sense to speak
15. Seržant (2013a) assumes the accusative to be an object because of its regularly being re-
placed with the genitive under negation, begging the question of whether this is a sufficient
criterion for objecthood.
of non-canonical marking of (first and second) arguments, but not of subjects and
objects, which cannot be identified.
5. Canonical marking
The notion of canonical marking of grammatical relations has until now be used
rather informally, and there have been no serious attempts at defining it. In the
framework of Canonical Typology (Brown, Chumakina & Corbett 2013), a Ca-
nonical definition of subjecthood would have to involve a cluster of features asso-
ciated with the very notion of subject and object, allowing us to create a conceptual
space in which the actual instances of subjects and objects could be situated. It is
not quite clear whether the canonical approach, which until now has been applied
mainly to morphosyntactic features, is able to deal with a phenomenon as com-
plex as grammatical relations. In any case, the constituent features of subjects and
objects as language-specific categories might well be so divergent that universal
canonical definitions would have to involve the comparative concepts used for
alignment typology.
In the light of such a notion of canonicity every configuration diverging from
nominative for the intransitive subject and nominative and accusative for transi-
tive predicates would be non-canonical.
The consequences of such a view would be somewhat unexpected. The nomi-
native subject in (44), for example, would be non-canonical because it does not
co-occur with an accusative. And actually it could be said to be non-canonical
in the sense that it is hard to say whether it is a transitive or an intransitive sub-
ject. Of course, in a nominative-accusative marking language transitive and in-
transitive subjects have enough in common to make this question irrelevant for
practical purposes. On the other hand, when a verb assigns accusative without
concomitant nominative marking on another argument, the accusative cannot be
recognised as a canonical object, because that cannot be defined without a nomi-
native. Our default assumption, on meeting a structure like (74), should therefore
be that the accusative is not an object, save evidence to the contrary.
6. Concluding remarks and chapter summaries
The issues dealt with in this volume appear in the literature under various head-
ings. In this introduction we have mentioned differential marking, non-canonical
marking, clause-type-specific marking and argument alternations. Each prob-
lem can be discussed separately, and for each notion we will find a number of
examples that match it. Still, we hope to have shown how closely all these prob-
lems are interrelated. The manifold interconnections lend a certain underlying
unity to the articles brought together in the present volume. We will briefly sum-
marize them here.
Peter Arkadiev examines atypical case marking on objects in two types of
infinitival clauses in Lithuanian. One is traditionally called the dativus cum in-
finitivo; it expresses purpose and consists of an original infinitive of purpose with
an object or subject appearing in the dative. The other expresses purpose with
verbs of motion and consists of an infinitive with a genitival object. Historically, at
least the dative was once assigned directly, across what was eventually to become
a clause boundary. In fact, it is hard to escape the impression that genitive and
dative case are still being assigned directly across clause boundaries, though this
would be difficult to capture in any theoretical framework. Arkadiev proposes an
interesting account that, though basically couched in Minimalist terms, diverges
from standard Minimalist assumptions about case. He operates with the notion
of complementizing case, proposed for Australian languages by Dench and Evans
(1988), i.e., a case that can be assigned to a clause and can spread to several of its
subconstituents. The same empty functional head Cdat assigns case to NPs (pini-
gai namui ‘money for a house’) and clauses (pinigai namui statyti ‘money to build
a house’). Though the discussion is, to some extent, theory-internal, concerning
as it does the mechanisms of case assignment in Minimalism, the issues raised are
of broader interest.
Axel Holvoet and Marta Grzybowska deal with grammatical relations in
constructions with the debitive, a necessitive form of the Latvian verb. Though
the debitive is an inflectional verb form, which would lead us to examine the char-
acteristic case marking pattern it induces in the context of splits and differential
case marking, the authors discuss it as an instance of non-canonical grammati-
cal relations, arguing that the debitive construction shows diffuse grammatical
relations without clear subject or object properties in any of the arguments. They
show that similar patterns can also be lexically assigned. They argue for diffuse
grammatical relations as an essential element of non-canonicity, and point out
that in discussing splits or differential argument marking it is often impossible to
circumvent the issue of grammatical relations.
Kristina Lenartaitė-Gotaučienė searches for a satisfactory account of gram-
matical relations with Lithuanian predicates displaying alternations in argument
realization usually referred to as the swarm-alternations. Many of the construc-
tions participating in these alternations have no subjects marked as such by mor-
phosyntactic means. The author analyses the semantic properties of the predicates
with which these alternations are associated, and the discourse-pragmatic prop-
erties of the sentence types involved in the alternations, and attempts to single out
the factors that block the default association of subjecthood with agentivity. She
then introduces a notion of subjecthood involving features at the levels of gram-
matical structure, thematic relations and discourse structure. All these features
coincide in the case of the prototypical subject but various dissociations occur in
the non-prototypical configurations instantiated by the swarm-alternations.
Rolandas Mikulskas’ article deals with the problem of singling out the subject
in a subtype of copular constructions, the so-called specificational constructions,
in Lithuanian. While in other copular constructions, both predicative (predica-
tional) and equative, the assignment of grammatical relations follows the trajec-
tory–landmark configuration (the discourse entity that has to be identified being
selected as trajector, grammatically marked as the subject), agreement patterns in
the specificational constructions show that in some languages the pre-copular NP
is treated as subject whereas in others this function is assigned to the post-copu-
lar NP. This problem constitutes, of course, a challenge for any universal concept
of subjecthood, and also for the content requirement principle, one of the tenets
of Cognitive Grammar. Having examined first Minimalist, then Cognitivist ap-
proaches to this problem, the author proposes a solution in terms of agnation,
a notion borrowed from Systemic Functional Grammar. Though specificational
constructions (which in themselves are also equative structures), compared with
their counterparts in the ‘predicative’ domain, i.e., descriptional-identifying equa-
tive constructions, represent the opposite direction of referent identification, they
may, in some languages, borrow their grammatical architecture from their closely
related agnates. Finally, agreement patterns in specificational constructions show
a great deal of language-specific syntactic inertia and are not so relevant in the
conceptualization of specificational relation. In contrast, referential properties of
two nominals of the construction and its fixed information structure are decisive
in this respect. This is another instance of a two-argument construction in which
neither of the two arguments can be unequivocally assigned subject status.
Nicole Nau investigates the use of the genitive as an object marker in con-
servative varieties of Latgalian. The main part of the paper is concerned with dif-
ferential object marking – the alternation between the accusative and the genitive
with transitive verbs. The genitive is most regularly used in negated finite clauses
and infinitive clauses depending on a negated predicate, as well as in constructions
with the supine. Furthermore, it is used as a stylistic device to mark a situation as
unreal or impossible. With mass nouns and plurals a partitive genitive is used, but
it is subject to several restrictions. In the second, smaller part of the paper the au-
thor gives an overview of lexical predicates governing a genitive object and takes
a look into derivational processes that turn transitive verbs into verbs govern-
ing the genitive. As this is the first description of differential and non-canonical
object marking in Latgalian, its main goal is an accurate systematic description
of the data, leaving interpretation and comparison with Lithuanian – where the
situation is similar but not identical – for future studies.
Ilja Seržant examines the semantic contrasts associated with an instance of
differential marking of objects, viz. that between the accusative and the indepen-
dent partitive genitive (IPG) in Lithuanian. He discusses various meaning facets
of the IPG, and claims that it always transforms the NP into one with indefinite
reference, even with inherently definite NPs such as anaphoric or demonstrative
pronouns. He operates with the notion of boundedness, arguing that the Lithu-
anian independent partitive allows both for bounded-set and for unbounded-set
readings. While the first reading is indefinite but existential (like English ‘some’)
and its referent may be picked up by the following discourse, the unbounded
reading is weaker and refers to a fully arbitrary and undelimited set (but not
universal).
As regards the interaction between aspectuality and the IPG, Seržant, assum-
ing the bidimensional approach, claims that the IPG (object) always marks a non-
culminating VP which furthermore is typically incompatible with progressive
contexts and less compatible with other contexts typical of imperfective viewpoint
aspect. At the same time, it is always compatible with delimitatives (non-culmi-
nating perfectives). Interestingly, following similar approaches found in the lit-
erature, Seržant views boundedness as a notion cutting through the boundary
between the nominal and the verbal domains: the aspectual boundedness of a
predication interacts with the boundedness of the noun phrase and may affect
nominal case marking. One could compare this to the purposive case marking
applying both to the nominal and to the clausal domains in Arkadiev’s account
(this volume).
The paper by Björn Wiemer and Valgerður Bjarnadóttir aims at establish-
ing and comparing inventories of verbs with non-canonical argument marking
in two languages: Lithuanian and Icelandic. It is meant to be a first contribution
within a larger project that will show Lithuanian case-marking in contrast to oth-
er languages. It sets out the theoretical background, methodology and aims of the
project. Arguments are defined by their rank according to the Actor-Undergoer
hierarchy established in Role and Reference Grammar. Non-canonical marking is
any marking other than the nominative for the highest-ranking (most agentive)
argument of one-place or two-place predicates and/or accusative for the lower
ranking (most patientive) argument of two-place predicates. A classification of
verbs into ten lexical classes – including emotive, cognitive, modal etc. – allows
us to investigate the coding patterns in relation to verbal semantics. In their paper
the authors concentrate on highest-ranking arguments, which in both languages
may be marked by the genitive, the dative, and the accusative. Lithuanian makes
more use of the accusative, Icelandic of the dative. The genitive plays a minor role
in both languages. Lithuanian has further patterns involving the instrumental
(typically associated with the function of inanimate causer) and marginally the
locative.
Abbreviations
acc accusative nom nominative

dat dative pfx prefix
deb debitive pl plural
dem demonstrative ppa past active participle
f feminine ppp past passive participle
gen genitive prs present
inf infinitive pst past
ins instrumental ptc particle
ptc particle refl reflexive
loc locative reflposs reflexive possessive
m masculine sg singular
neg negative sup supine
Sources
LiLa = Lietuviešu-latviešu-lietuviešu paralēlo tekstu korpuss. Availabe online at: http://www.

korpuss.lv/lila/
LLVV = Latviešu literārās valodas vārdnīca. 7.2 S-T, 1991. Rāga: Zinātne.
Mio-2 = miljons-2.0, corpus of contemporary standard Latvian. Available online at: www.
korpuss.lv.
References
Aikhenvald, Alexandra Y., Robert M. W. Dixon, & Masayuki Onishi (eds). 2001. Non-Canoni-
cal Marking of Subjects and Objects [Typological Studies in Language 46]. Amsterdam/
Philadelphia: John Benjamins. DOI: 10.1075/tsl.46
Aleksandravičiūtė, Skaistė. 2013. The semantic effects of the Subject Genitive of Negation in
Lithuanian. Baltic Linguistics 4: 9–38.
Ambrazas, Vytautas (ed). 1997. Lithuanian Grammar. Vilnius: Baltos lankos.
Ambrazas V. 2001. On the development of the nominative object in East Baltic. In Circum-
Baltic Languages, vol. 2: Grammar and Typology [Studies in Language Companion Series
55], Östen Dahl & Maria Koptjevskaja-Tamm (eds), 391–412. Amsterdam/Philadelphia:
John Benjamins.
Andrews, Avery D. 2001: Non-canonical A/S marking in Icelandic. In: Non-Canonical Marking
of Subjects and Objects [Typological Studies in Language 46], Alexandra Y. Aikhenvald,
Robert M. W. Dixon, & Masayuki Onishi (eds), 85–111. Amsterdam/Philadelphia: John
Benjamins.
Andronov, Aleksey V. 2001. A survey of the case paradigm in Latvian. Sprachtypologie und
Universalienforschung 54(3): 197–208.
Arkadiev, Peter M. 2013. Marking of subjects and objects in Lithuanian non-finite claus-
es: A typological and diachronic perspective. Linguistic Typology 17(3): 397–437. DOI:
10.1515/lity-2013-0020
Barðdal, Jóhanna. 2003. ‘Oblique Subjects’ in Icelandic and German. Working Papers in Scandi-
navian Syntax 70: 61–99. Available at: http://org.uib.no/iecastp/barddal/JB-wpss.pdf
Bayer, Josef. 2004. Non-nominative subjects in comparison. In Non-nominative subjects, vol.
1 [Typological Studies in Language 60], Peri Bhaskararao & Karumuri Venkata Subbarao
(eds), 49–76. Amsterdam/Philadelphia: John Benjamins.
Berg-Olsen, Sturla. 1999. A syntactic change in progress: The decline in the use of the non-prepo-
sitional genitive in Latvian, with a comparative view on Lithuanian. MA thesis, University
of Oslo.
Berg-Olsen, Sturla. 2000. The Latvian non-prepositional genitive – a case losing ground. Res
Balticae 6, 95–146.
Berg-Olsen, Sturla. 2001. Subjects and valency-changing mechanisms in Latvian. Sprachtypolo-
gie und Universalienforschung 54(3): 200–225.
Berg-Olsen, Sturla. 2005. The Latvian dative and genitive: A Cognitive Grammar account. PhD
thesis, University of Oslo.
Bickel, Balthasar. 2004. The syntax of experiencers in the Himalayas. In Non-nominative sub-
jects, vol. 1 [Typological Studies in Language 60], Peri Bhaskararao & Karumuri Venkata
Subbarao (eds), 77–112. Amsterdam/Philadelphia: Benjamins.
Bickel, Balthasar. 2011. Grammatical relations typology. In The Oxford Handbook of Language
Typology, Jae Jung Song (ed), 399–444. Oxford: Oxford University Press.
Bickel, Balthasar & Johanna Nichols. 2011. Case marking and alignment. In The Oxford hand-
book of case, Andrej Malchukov & Andrew Spencer (eds), 304–321. Oxford: Oxford Uni-
versity Press.
Borschev, Vladimir, Paducheva, Elena V., Partee, Barbara H., Testelets, Yakov, & Yanovich, Igor.
2008. Russian genitives, non-referentiality, and the property-type hypothesis. In Formal
Approaches to Slavic Linguistics: The Stony Brook Meeting 2007 [Formal Approaches to
Slavic Linguistcs 16], A. Antonenko, J. F. Bailyn, & C. Bethin (eds), 48–67. Ann Arbor:
Michigan Slavic Publications.
Brown, Dunstan, Marina Chumakina and Greville G. Corbett (eds) 2013. Canonical Morphol-
ogy and Syntax. Oxford: Oxford University Press.
Comrie, Bernard. 1975. The antiergative: Finland’s answer to Basque. Chicago Linguistic Society
11: 112–121.
Corbett, Greville G. 1986. The use of the genitive or accusative for the direct object of negated
verbs in Russian: A bibliography. In Case in Slavic, Richard D. Brecht & James S. Levine
(eds), 361–372. Columbus, Ohio: Slavica Publishers Inc.
Donohue, Mark. 2008. Semantic alignment systems: what’s what, and what’s not. In The ty-
pology of semantic alignment, Mark Donohue & Søren Wichmann (eds), 24–75. Oxford:
Oxford University Press. DOI: 10.1093/acprof:oso/9780199238385.003.0002
Donohue, Mark & Wichmann, Søren (eds). 2008. The Typology of Semantic Alignment. Oxford:
Oxford University Press. DOI: 10.1093/acprof:oso/9780199238385.001.0001
Endzelin, J[an]. 1922. Lettische Grammatik. Riga: Kommisionsverlag A. Gulbis.
Endzelīns, Jānis & Kārlis Mīlenbachs. 1907. Latviešu gramatika. Rīga: K. J. Zichmanis.
Filip, Hana. 2001. Nominal and verbal semantic structure: analogies and interactions. Language
Sciences 23: 453–501. DOI: 10.1016/S0388-0001(00)00033-4
Filip, Hana. 2005. On accumulating and having it all. Perfectivity, prefixes and bare argu-
ments. In Perspectives on aspect, H. J. Verkuyl, H. de Swart, & A. van Hout (eds), 125–148.
Dordrecht: Springer. DOI: 10.1007/1-4020-3232-3_7
Franks, Stephen and Lavine, James E. 2006. Case and word order in Lithuanian. Journal of Lin-
guistics 42(1): 239–288. DOI: 10.1017/S0022226706003896
Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages.
In Non-canonical marking of subjects and objects [Typological Studies in Language 46],
Alexandra Y. Aikhenvald, Richard M.W. Dixon, & Masayuki Onishi (eds), 53–83. Amster-
dam/Philadelphia: Benjamins.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic
studies. Language 86(3): 663–687. DOI: 10.1353/lan.2010.0021
Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typol-
ogy. Linguistic Typology 15: 535–567. DOI: 10.1515/LITY.2011.035
Holvoet, Axel. 1992. Bemerkungen über die Entwicklung des lettischen Kasussystems: Der In-
strumental. In Colloquium Pruthenicum Primum. Papers from the First International Con-
ference on Old Priussian, Wojciech Smoczyński & Axel Holvoet (eds), 143–149. Warsaw:
Wydawnictwa Uniwersytetu Warszawskiego.
Holvoet, Axel. 2010. Between morphosyntax and the paradigm: Some puzzling patterns of case
distribution in Baltic and their implications. Acta Linguistica Hafniensia 42(2): 175–198.
DOI: 10.1080/03740463.2010.521445
Holvoet, Axel. 2013. Obliqueness, quasi-subjects and transitivity in Baltic and Slavonic. In The
Diachronic Typology of Non-Canonical Subjects [Studies in Language Companion Series
140], Ilja A. Seržant & Leonid Kulikov (eds), 259–284. Amsterdam/Philadelphia: John
Benjamins.
Holvoet, Axel. Forthcoming. Non-canonical subjects in Latvian: an obliqueness-based ap-
proach. In Contemporary Approaches to Baltic Linguistics, Peter M. Arkadiev, Björn
Wiemer, & Axel Holvoet (eds). Berlin: De Gruyter.
Hopper, Paul & Thompson, Sandra. 1980. Transitivity in grammar and discourse. Language
56(2): 251–299.
Huumo, Tuomas. 2009. Fictive dynamicity, nominal aspect, and the Finnish copulative con-
struction. Cognitive Linguistics 20(1), 43–70. DOI: 10.1515/COGL.2009.003
Kagan, Olga. 2010. Genitive objects, existence and individuation. Russian Linguistics 34: 17–39.
DOI: 10.1007/s11185-009-9051-x
Kagan, Olga. 2013. Semantics of genitive objects in Russian. Dordrecht: Springer. DOI:
10.1007/978-94-007-5225-2
Keenan, Edward L. 1976. Towards a universal definition of ‘subject’. Subject and Topic, Charles
N. Li (ed), 303–333. New York: Academic Press.
Kiparsky, Paul. 1998. Partitive case and aspect. In The Projection of Arguments: Lexical and
Compositional Factors, Miriam Butt & Wilhelm Geuder (eds), 265–307. Stanford: CSLI
Publications.
Koptjevskaja-Tamm, Maria & Wälchli, Bernhard. 2001. The Circum-Baltic languages. An ar-
eal-typological approach. In Circum-Baltic Languages, vol. 2: Grammar and Typology
[Studies in Language Companion Series 55], Östen Dahl & Maria Koptjevskaja-Tamm
Lavrin, B. A. 1963. Ob odnoj slavjano-balto-finskoj izoglosse. Lietuvių kabotyros klausimai 6:
87–107.
Levin, Beth & Rappaport Hovav, Malka. 2005. Argument Realization. Cambridge: Cambridge
Univerity Press.
Malchukov, Andrej L. 2005. Case pattern splits, verb types and construction competition. In
Competition and Variation in Natural Languages: The Case for Case, Mengistu Amberber &
Helen de Hoop (eds), 73–117. London: Elsevier. DOI: 10.1016/B978-008044651-6/
50006-9
Malchukov, Andrej. 2006. Transitivity parameters and transitivity alternations: Constraining
co-variation. In Case, Valency and Transitivity [Studies in Language Companion Series
77], Leonid. I. Kulikov, Andrej L. Malchukov, & Helen de Hoop (eds), 329–357. Amster-
dam/Philadelphia: John Benjamins.
Mīlenbahs, Kārlis. 2009 [1890/1891]. Par akuzatīvu un ģenitīvu pie noliegtiem tranzitīviem
verbiem. In Kārlis Mīlenbahs. Darbu izlase divos sējumos. 1. sējums, 42–49. Rīga: LU
Latviešu valodas institūts.
Moore, John & Perlmutter, David M. 2000. What does it take to be a Dative Subject? Natural
Language and Linguistic Theory 18: 373–416. DOI: 10.1023/A:1006451714195
Mühlenbach, Karl. 1902/1903. Über die vermeintlichen Genitive oder Ablative auf -ů oder u im
Lettischen. Indogermanische Forschungen 13: 220–260.
Næss, Åshild. 2007. Prototypical Transitivity [Typological Studies in Language 72]. Amster-
dam/Philadelphia: John Benjamins. DOI: 10.1075/tsl.72
Næss, Åshild. 2011. Varieties of dative. In The Oxford Handbook of Case, Andrej Malchukov &
Andrew Spencer (eds), 572–580. Oxford: Oxford University Press.
Nau, Nicole. 1998. Latvian [Languages of the World/Materials 217]. München: Lincom Europa.
Nau, Nicole. 2011a. A Short Grammar of Latgalian. München: Lincom Europa.
Nau, Nicole. 2011b. Declension classes in Latvian and Latgalian: Morphomics vs. Morphopho-
nology. Baltic Linguistics 2: 141–177.
Nichols, Johanna. 2008. Why are stative-active languages rare in Eurasia? A typological per-
spective on split-subject marking. In The typology of semantic alignment, Mark Donohue &
Søren Wichmann (eds), 121–139. Oxford: Oxford University Press. DOI: 10.1093/acprof:
oso/9780199238385.003.0005
Onishi, Masayuki. 2001. Introduction: Non-canonically marked subjects and objects. Param-
eters and properties. In Non-Canonical Marking of Subjects and Objects [Typological Stud-
ies in Language 46], Alexandra Y. Aikhenvald, Robert M. W. Dixon, & Masayuki Onishi
Partee, Barbara H., Borschev, Vladimir, Paducheva, Elena V., Testelets, Yakov, & Yanovich, Igor.
2011. Russian Genitive of Negation Alternations: The Role of Verb Semantics. Scando-
Slavica 57(2): 135–159. DOI: 10.1080/00806765.2011.631775
Partee, Barbara H., Borschev, Vladimir, Paducheva, Elena V., Testelets, Yakov, & Yanovich, Igor.
2012. The role of verb semantics in genitive alternations: genitive of negation and genitive
of intensionality. In The Russian Verb [Oslo Studies in Language 4(1)], Atle Grønn & Anna
Pazelskaya (eds), 1–29.
Perkova, Natalia. MS. Non-canonical argument marking in two-place predication: the case of
the Baltic languages. Unpublished paper, Stockholm 2013.
Piccini, Silvia. 2008. Traces of non-nominative alignment in Lithuanian: The impersonal con-
structions in Indo-European perspective. Baltistica 43(3): 437–461.
Primus, Beatrice. 2011. Case, grammatical relations, and semantic roles. In The Oxford Hand-
book of Case, Andrej Malchukov & Andrew Spencer (eds), 261–275. Oxford: Oxford Uni-
versity Press.
Raģe, S. 1964. Erģemes, Lugazu un Valkas izloksnes fonētika un morfoloģija. Valodas un
literatūras institūta raksti 8: 5–142.
Seržant, Ilja. 2013a. Rise of canonical objecthood with the Lithuanian verbs of pain. Baltic Lin-
guistics 4: 187–211.
Seržant, Ilja. 2013b. Rise of Canonical Subjecthood. In The Diachrony of Non-canonical Sub-
jects [Studies in Language Companion Series 140], Ilja A. Seržant & Leonid Kulikov (eds),
309–336. Amsterdam/Philadelphia: John Benjamins.
Seržant, Ilja. Forthcoming. Dative experiencer constructions as a Circum-Baltic isogloss. In
Contemporary Approaches to Baltic Linguistics, Peter M. Arkadiev, Björn Wiemer, & Axel
Holvoet (eds). Berlin: De Gruyter.
Seržant, Ilja A., Fedriani, Chiara, & Kulikov, Leonid. 2013. Introduction. In The Diachrony of
Non-canonical Subjects [Studies in Language Companion Series 140], Ilja A. Seržant, &
Leonid Kulikov (eds), ix–xxvi. Amsterdam/Philadelphia: John Benjamins. DOI: 10.1075/
slcs.140
Sigurðsson, Halldór Á. 2004. Icelandic non-nominative subjects: Facts and implications. In
Non-Nominative Subjects [Typological Studies in Languages 61], Peri Bhaskararao &
Karumuri V. Subbarao (eds), vol. 2, 137–159. Amsterdam/Philadelphia: Benjamins.
Timberlake, Alan. 1974. The Nominative Object in Slavic, Baltic, and West Finnic. München:
Verlag Otto Sagner.
Tsunoda, Tasaku. 1981. Split case-marking in verb types and tense/aspect/mood. Linguistics
19: 389–438.
Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21: 385–396. DOI:
10.1017/S0022226700010318
Verma, Manindra & K.[aravannur] P.[uthanvettil] Mohanan. 1990. Introduction to the expe-
riencer subject construction. In Experiencer Subjects in South Asian Languages, Manindra
Verma & K. P. Mohanan (eds), 1–11. Stanford: CSLI Publications.
Zaenen, Annie, Maling, Joan, & Thráinsson, Höskuldur. 1985. Case and grammatical func-
tions: the Icelandic passive. Natural Language and Linguistic Theory 3: 441–483. DOI:
10.1007/BF00133285
Case and word order in Lithuanian
infinitival clauses revisited
Peter Arkadiev
Institute of Slavic Studies, Russian Academy of Sciences / Russian State
University for the Humanities / Sholokhov Moscow State University
for the Humanities / Vilnius University
This paper considers the Lithuanian constructions with the Dative and Genitive
marking of direct objects of transitive verbs in purpose infinitival clauses, stud-
ied in Franks and Lavine (2006). I adduce empirical evidence and conceptual
arguments both speaking against the analysis proposed by Franks and Lavine
(2006), and argue for a different account based on recent “non-orthodox” pro-
posals in case theory. My analysis of the Lithuanian constructions is inspired by
a typological comparison with Australian languages possessing “complementiz-
ing” and “associating” case marking and morphological case-stacking. I propose
that the mechanism of multiple case assignment in syntax is operative in Lithu-
anian and show how it can naturally account for the Dative-plus-Infinitive and
Genitive-plus-Infinitive constructions.
1. Introduction1
Lithuanian, being a nominative-accusative language, encodes the direct object of

transitive verbs with the Accusative2 case, cf. (1a). In ordinary finite clauses this
Accusative encoding is overridden by the Genitive when the verb is negated, cf.
(1b), and frequently also when the object is partitive (see Seržant, this volume).
1. I thank my Lithuanian consultants for their patience and help, and David Erschler, James
Lavine, Ora Matushansky, Rolandas Mikulskas and Nicole Nau for their useful comments and
criticism on the earlier version of this paper. All faults and shortcomings remain mine.
2. Following the tradition of works in linguistic typology, I capitalize labels of language-par-
ticular grammatical categories and features.
44 Peter Arkadiev
(1) a. Jon-as perskait-ė laišk-ą.

Jonas-nom.sg read.through-pst(3) letter-acc.sg
‘Jonas read the letter.’
b. Jon-as ne-perskait-ė laišk-o.
Jonas-nom.sg neg-read.through-pst(3) letter-gen.sg
‘Jonas didn’t read the letter.’
The situation in infinitival clauses based on transitive verbs is, however, more
complex. In addition to (i) the “canonical” Accusative encoding of the object in
clauses selected by most verbs taking infinitival complements (e.g., verbs denot-
ing modality, volition, manipulation), cf. (2), and (ii) the Genitive encoding in
the presence of negation, either on the Infinitive, cf. (3a), or on the matrix verb if
the latter belongs to group (i), cf. (3b), there exist three other constructions with
“non-canonical” marking of the direct object of a transitive Infinitive.
(2) Jon-as nor-i [perskaity-ti laišk-ą].
Jonas-nom.sg want-prs(3) read.through-inf letter-acc.sg
‘Jonas wants to read the letter.’
(3) a. Dėking-a Onut-ė pažadėj-o [ne-palik-ti
grateful-nom.sg Onute-nom.sg promise-pst(3) neg-leave-inf
mūs-ų]…
we-gen
‘Grateful Onutė promised not to leave us.’ (LKT3)
b. Jon-as ne-nor-i [perskaity-ti laišk-o].
Jonas-nom.sg neg-want-prs(3) read.through-inf letter-gen.sg
‘Jonas does not want to read the letter.’
(iii) In impersonal matrix constructions (usually headed by emotional or modal

predicates or by the copula būti ‘be’) the object of the Infinitive can be in the
Nominative, cf. (4):
(4) J-am ne-patik-o [laukel-is ar-ti].
3-dat.sg.m neg-like-pst(3) field-nom.sg plough-inf
‘He did not like to plough the field.’ (Ambrazas (ed.) 1997: 638)
(iv) The object of the purposive infinitival clause occurring with verbs of motion
is in the Genitive, cf. (5):
(5) išvažiav-o [keli-o taisy-ti].
drive.out-pst(3) road-gen.sg repair-inf
‘(they) went to repair the road.’ (Ambrazas (ed.) 1997: 638)
3. Corpus of Lithuanian Language (LKT, http://tekstynas.vdu.lt/).

(v) Finally, in other kinds of purpose infinitives adjoined to verbs or nouns, the
object is in the Dative, cf. (6).
(6) iššov-ė [žmon-ėms pagąsdin-ti].
shoot-pst(3) people-dat.pl frighten-inf
‘(he) fired to scare the people.’ (Ambrazas (ed.) 1997: 557)
The main foci of this paper are constructions with the Dative and the Genitive
marking illustrated in (5) and (6). These constructions, which I will call, respec-
tively, Dative-plus-Infinitive and Genitive-plus-Infinitive, have been extensively
studied by Steven Franks and James Lavine (2006) (further FL06), and the goal
of this contribution is to review and supplement their analysis in the light of new
and more comprehensive empirical data and to propose a different treatment, fol-
lowing certain recent proposals in the Minimalist case theory.
The structure of the paper is as follows. In §2 I briefly outline the main con-
ceptual tenets of the Minimalist case theory necessary for the understanding of
the article. In §3 I give a summary of FL06’s analysis and claims, and in §4 offer a
revision of FL06’s empirical claims based on my own research. In §5 I go beyond
Lithuanian and show how comparable data from other languages, including both
those related to Lithuanian (i.e., Latgalian) and those completely unrelated (i.e.,
some Australian languages), can help us better understand the Lithuanian pat-
terns of “non-canonical” case marking in infinitival clauses. In particular, I claim
that such notions as “associating” and “complementizing” case, first introduced by
Dench and Evans (1988) for the languages of Australia, and the generalized mech-
anism of multiple case assignment in syntax (cf. Matushansky 2008, 2010; Erschler
2009) can account for the Lithuanian data in a conceptually satisfactory way. In §6
I give a detailed outline of my own analysis of the Lithuanian constructions with
the Dative and Genitive case marking of the object of infinitival clauses.
2. Basics of the Minimalist case theory
Since FL06 address the Lithuanian constructions with Dative and Genitive ob-
jects of the Infinitive from the point of view of recent generative case theory
and argue that the Lithuanian data are problematic for the general assumptions
of this theory (cf. also Anderson, to appear, for a similar argument based on
a broader range of data from Lithuanian), I will start by briefly presenting the
main tenets of the current “mainstream” generative (Minimalist) case theory, of
course in a simplified fashion (see also Hornstein et al. 2005: Ch. 4; Bobaljik &
Wurmbrand 2009). However, further on in this paper I assume the reader to be
familiar with the most general architecture of the Minimalist theory (cf. Chomsky
2000, 2001; Hornstein et al. 2005) and its basic notions, such as binary branching
46 Peter Arkadiev
X-bar-syntactic structure, features, operations like Merge, Move and Agree, and
the division between the so-called “narrow syntax” and Phonological Form (PF).
Initially, case in generative grammar was assumed to be an abstract feature
of noun phrases required for the well-formedness of a syntactic derivation. The
so-called “case filter” (Chomsky 1981: 49) required every phonologically overt NP
to have case (even in languages with only vestigial case distinctions, like English,
or no morphological case at all, like Chinese). Most discussion of case in Govern-
ment and Binding theory was concerned with the relation of case with NP licens-
ing and had little to do with morphological case proper.4 In most work on case
theory it was assumed that NPs receive case only once and that once a case value
is assigned it can no longer be substituted by another case value. Such a view has
been largely retained in the “mainstream” Minimalist work, see, e.g., Chomsky
(2001: 6ff; Hornstein et al. 2005: Ch. 4), where case is viewed as an “uninterpre-
table” feature on nominals which has to be “checked” or “valuated” via an Agree
operation for the derivation to be licit, and where “once the Case value is deter-
mined, N no longer enters into agreement relations” (Chomsky 2001: 6). Such
a view of case theory has been recently challenged in such work as McFadden
(2004), Landau (2006), Keine (2010), Preminger (2011) and some others, who all
consider case to be primarily a morphological phenomenon, determined in syn-
tax but independent of the issue of NP licensing and not reducible to the Agree
operation, and by Béjar and Massam (1999), Merchant (2006) and Matushansky
(2008, 2010), who argue that case can sometimes be assigned more than once to
the same NP.5
Regardless of the licensing issue, which won’t concern us any more, at least
since Chomsky (1981: 170) discussions of case in generative theories have in-
volved a distinction between the so-called “structural” and “inherent” types of
case. Basically, structural case is determined by the general properties of the syn-
tactic configuration the NP occurs in (hence the synonymous term “configura-
tional” case), while inherent case is determined by the lexical properties of the
NP’s governor, e.g. by a particular verb or preposition. The distinction between
structural and inherent case proves to be especially useful for languages with rich
case systems such as, e.g. German, Icelandic, Slavic and Baltic. Here, structural
cases are those associated with the core grammatical relations of subject and di-
rect object, i.e. nominative and accusative, as well as the adnominal genitive, while
other cases or rather uses of cases, e.g. adverbal genitive, dative, instrumental
etc., are regarded as inherent cases. Under the current Minimalist architecture of
4. Cf. the starting sentence of Bobaljik & Wurmbrand (2009: 44).

5. In fact, proposals along these lines go back at least as early as McCreight Young (1988).
syntax, inherent case can be viewed as assigned or checked by lexical heads (V or

P), while structural case is assigned/checked by functional heads. In particular,
accusative can be viewed as assigned by the “little v” head present with transitive
verbs, and nominative as assigned by the (finite) T head (Chomsky 2001: 6), see
the tree diagram in (7). In (7) case assignment is indicated by dotted arrows sub-
scribed with the relevant case value.
(7) TP
3
T′
3
T[fin] vP
3
nom
NPi v′
3
v[tr] VP
3
acc
V NPk
Further developments in case theory, e.g., Babby (1986, 1994), Woolford (2006),
proposed to subdivide non-structural case into “semantic” and “lexical”, draw-
ing a distinction between non-structural case which is determined solely by id-
iosyncratic specifications of particular lexical items and case that is transparently
related to some kind of semantic role, e.g., recipient or beneficiary dative. For in-
teresting proposals along these lines concerning Lithuanian, see Anderson (2011,
forthcoming).
The distinction between different kinds of case reveals itself in the phenom-
enon of case competition, where, depending on the morphosyntactic context,
different cases may appear on an NP (arguably) bearing the same grammatical
relation and occupying the same structural position. A paradigm example of case
competition is the Genitive of negation rule in Russian and Lithuanian, see ex-
ample (1b) and (3) above. Here the Genitive is arguably a semantic case triggered
by the Neg head; the crucial property of this rule consists in the empirical gener-
alization that the Genitive of negation can replace only the structural Accusative
case but not any other case, lexical or semantic, cf. (8).
(8) a. Aldon-a dav-ė broli-ui obuol-į.
Aldona-nom.sg give-pst(3) brother-dat.sg apple-acc.sg
‘Aldona gave her brother an apple.’
48 Peter Arkadiev
b. Aldon-a ne-dav-ė broli-ui / *broli-o

Aldona-nom.sg neg-give-pst(3) brother-dat.sg/*gen.sg
obuoli-o / *obuol-į.
apple-gen.sg/*acc.sg
‘Aldona didn’t give her brother an apple.’
For a more detailed discussion of lexical vs. inherent case in Lithuanian, see FL06
(247–249) and Anderson (forthcoming). What is important here is that under
the Minimalist architecture of grammar, case competition and the distinction be-
tween lexical, semantic and structural case can be regarded as stemming from the
fact that lexical case is assigned to the NP “upon first merge, immediately when it
is introduced into the derivation” (Preminger 2011: 151), which, together with the
view that case values are determined once and for all, automatically implies that
NPs assigned lexical case cannot further receive semantic or structural case. The
possibility for semantic case to substitute for structural case, or for one structural
case to substitute for another under specific conditions such as non-finiteness,
can therefore be captured by establishing local relations between an NP not as-
signed lexical case and some functional head, e.g. Neg in the case of Genitive of
negation.
The last crucial property of the “mainstream” Minimalist conception of case
is the view of case assignment/checking as a local operation. The most general
constraint on locality of all operations (Move and Agree, the latter comprising
case assignment) is the so-called Phase Impenetrability Condition (Chomsky
2000: 108; 2001: 13) in (9).
(9) In phase α with head H, the domain of H (its complement) is not accessible
to operations outside of α; only H and its edge (specifier) are.
Phases are thus syntactic objects (parts of a derivation) which are inaccessible to
further operations once constructed; in general it is assumed that after a phase is
built, it is transferred to the semantic and phonological interfaces for interpreta-
tion and spell-out. Phasehood is determined by the nature of particular functional
heads; thus, (transitive) vPs and CPs are considered phases, while (at least non-
finite) TPs are not, which, for instance, makes it possible to capture the familiar
distinction between control and raising structures (see e.g. Chomsky 2001: 7–9).
For case theory, the notion of phase has obvious consequences in that case
assignment is constrained by the Phase Impenetrability Condition in (9) (see e.g.
McFadden 2010). In particular, this means that a head outside of the vP cannot
assign case to the direct object NP unless it raises to the edge of the vP – an as-
sumption crucial for FL06’s analysis of the Lithuanian constructions discussed in
the paper. Incidentally, this implies that with the Genitive of negation the direct
object has to raise to Spec,vP in order to be accessible to case assignment by the

Neg head (which is obviously located higher than v in syntactic structure), cf. the
diagram in (10), where phase boundary is indicated by the square.
(10) NegP
3
Neg vP
gen 3
NPdo vP
3
gen
NPsbj v′
3
x
v VP
3
V NPdo
To summarize this necessarily short and simplistic discussion of the “mainstream”

Minimalist case theory, let me recapitulate its main theoretical assumptions:
(11) i. each NP has to bear a unique case value;
ii. case values are assigned/checked by lexical and functional heads;
iii. case assignment/checking can only occur in local configurations con-
strained by the Phase Impenetrability Condition;
iv. case assignment/checking can only occur once for each NP;
v. in instances of case competition, structural, but not lexical/inherent case
can be overridden;
vi. NPs must move from their base positions in order to receive structural
case from heads located outside of their phases.
These assumptions constitute the conceptual basis of FL06’s analysis of the Lithu-
anian Dative-plus-Infinitive and Genitive-plus-Infinitive constructions summa-
rized in the next section. My own analysis of these constructions will challenge
most of the points in (11) and will be based on recent revisions to the Minimalist
case theory which propose an entirely different conception of case.
3. Franks and Lavine (2006)
FL06 is to date the only publication dealing with the non-canonical marking
of objects in Lithuanian Infinitive clauses, including the Dative-plus-Infinitive,
Genitive-plus-Infinitive and Nominative-plus-Infinitive constructions, from a
50 Peter Arkadiev
synchronic and theoretical point of view. Before that, these constructions had
received only a diachronic and historical-comparative treatment in Ambrazas
(1981, 1987), cf. also Schmalstieg (1987: 145–152, 174–179, 214–220), and had
been noticed in Sawicki (1992), dedicated to the more general issue of the expres-
sion of goal and purpose in Lithuanian.
Since in my paper I deal only with Dative-plus-Infinitive and Genitive-plus-
Infinitive constructions, leaving aside the Nominative-plus-Infinitive construc-
tions, I will not go into FL06’s treatment of the latter. The main empirical claims of
FL06 with respect to the Dative-plus-Infinitive and Genitive-plus-Infinitive con-
structions are outlined in the following subsections. Most examples and gram-
maticality judgments come from FL06 and are not necessarily endorsed by the
author of the present article.
3.1 Structural vs. inherent case
According to FL06, case alternation in infinitival clauses is possible only with Ac-
cusative direct objects, which receive case via general mechanisms of structural
case assignment, and not with indirect objects assigned inherent case by the verb,
cf. (12) showing a verb assigning Instrumental case vs. (13) with a regular transi-
tive verb with an object in the Accusative. The reason for word order variation in
(12) and (13) will be discussed in §3.2.
(12) a. Mes rūpin-a-mė-s vaik-ais.
we:nom take.care.of-prs-1pl-rfl child-ins.pl
‘We take care of children.’
b. Mes pastat-ė-me ligonin-ę [rūpin-ti-s
we:nom build-pst-1pl hospital-acc.sg take.care.of-inf-rfl
vaik-ais].
child-ins.pl
c. *Mes pastat-ė-me ligonin-ę [vaik-ams
we:nom build-pst-1pl hospital-acc.sg child-dat.pl
rūpin-ti-s].
take.care.of-inf-rfl
‘We built a hospital to take care of children.’ (FL06: 250)
d. Atėj-o [rūpin-ti-s draug-u].
come-pst(3) take.care.of-inf-rfl friend-ins.sg
e. *Atėj-o [draug-o rūpin-ti-s].
come-pst(3) friend-gen.sg take.care.of-inf-rfl
‘He came to take care of a friend.’ (FL06: 255)
(13) a. Mes gyd-o-me vaik-us.

we:nom treat-prs-1pl child-acc.pl
‘We treat children.’
b. Mes pastat-ė-me ligonin-ę [vaik-ams gydy-ti].
we:nom build-pst-1pl hospital-acc.sg child-dat.pl treat-inf
‘We built a hospital to treat children.’ (FL06: 252)
c. Daktar-as atėj-o [vaik-o gydy-ti].
doctor-nom.sg come-pst(3) child-gen.sg treat-inf
‘The doctor came to treat the child.’
The contrast between the behaviour of inherent case in (12) and structural case in
(13) naturally falls out of the assumption already stated above that inherent case,
being an idiosyncratic feature of individual lexical items (in this case verbs) and
assigned in the most local configuration by V to its complement directly upon
Merge, cannot be overridden by any other mechanisms of case assignment, all of
which necessarily apply at later stages of the derivation.
In addition to that, FL06 show that a case alternation is obligatory in purpose
infinitival clauses, but not in goal infinitival clauses with verbs of motion. Thus,
the replacement of the Accusative by the Dative is obligatory, cf. (14), while the
change from the Accusative to the Genitive is not, cf. (15).
( 14) *Pastat-ė daržin-ę [sukrau-ti šien-ą].
build-pst(3) hayloft-acc.sg keep-inf hay-acc.sg
intended: ‘They built a hayloft to keep hay.’ (FL06: 254)
(15) Parvažiav-o [pasiim-ti suknel-ę].
come.back-pst(3) take.rfl-inf dress-acc.sg
‘She came back to take the dress.’ (ibid.)
3.2 Case and word order
Probably the most crucial observation of FL06, from which they draw very im-
portant conclusions for their analysis, consists in the link between case alterna-
tion and “object shift” whereby the Dative and Genitive direct object normally
occupies a position before the Infinitive, in contrast to both the Accusative di-
rect objects of finite and “canonical” Infinitive clauses and the (indirect) objects
bearing inherent case. In other words, though the pragmatically neutral word
order in Lithuanian is Verb-Object (VO), in the infinitival constructions with
“non-canonically” marked direct objects the neutral order is rather OV, while VO
is accepted only under specific discourse conditions such as narrow focus on the
object, cf. (16) and (17).
52 Peter Arkadiev
(16) a. Mes pastat-ė-me ligonin-ę [vaik-ams gydy-ti].

we:nom build-pst-1pl hospital-acc.sg child-dat.pl treat-inf
‘We built a hospital to treat children.’ (FL06: 252)
#
b. Mes pastat-ė-me ligonin-ę [gydy-ti vaik-ams].
we:nom build-pst-1pl hospital-acc.sg treat-inf child-dat.pl
acceptable only under the narrow focus interpretation ‘We built a hospital
to treat precisely children (and not somebody else)’ (FL06: 251–252)
(17) a. J-ie dėj-o pastang-as [ilg-am kar-ui
3-nom.pl.m put-pst(3) effort-acc.pl long-dat.sg.m war-dat.sg
užbaig-ti].
finish-inf
‘They made efforts to end the long war.’
b. #J-ie dėj-o pastang-as [užbaig-ti ilg-am
3-nom.pl.m put-pst(3) effort-acc.pl finish-inf long-dat.sg.m
kar-ui].
war-dat.sg
acceptable only under the narrow focus interpretation (FL06: 251–252)
However, FL06 (256–257) note that the OV preference is attested only with the
Dative-plus-Infinitive constructions, and is not observed with the Genitive-plus-
Infinitive, where both orders are claimed to be neutral in terms of information
structure, cf. (18).
(18) a. Siunt-ė mergait-ę [parvež-ti daktar-o].
send-pst(3) girl-acc.sg bring-inf doctor-gen.sg
b. Siunt-ė mergait-ę [daktar-o parvež-ti].
send-pst(3) girl-acc.sg doctor-gen.sg bring-inf
‘He sent the girl to fetch a doctor.’ (FL06: 256)
Examples like (18a) are accounted for by FL06 (256–257) as an intermediate dia-
chronic stage in the development from a construction with the neutral OV order
and obligatory Genitive marking of the object to the default Infinitive construc-
tion with Accusative object and VO order, like the one shown in (15). Syntacti-
cally, FL06 propose that in examples like (18a) “[t]he matrix and embedded verbs
appear to be functioning as a single syntactic unit, which together assign lexical
genitive to their shared object argument” (FL06: 257).
Further, FL06 claim that in the Infinitive constructions with OV order the
Dative or Genitive object not just occurs before the verb but is moved out of the
VP, which is indicated by the position of manner adverbials demarcating the left
edge of the VP, cf. (19a). The final position of the displaced object is arguably the
left edge of the Infinitive phrase (InfP), which is indicated by the position of epis-
temic adverbials residing outside the VP, cf. (19b), and the OSV order in examples
where both the Dative object and the Dative subject co-occur, cf. (20).
(19) a. Pastat-ė daržin-ę [InfPšien-ui [VPsaugiai sukrau-ti]].
build-pst(3) hayloft-acc.sg hay-dat.sg safely keep-inf
‘They built a hayloft to keep the hay safely.’
b. Išvažiav-o [InfPkeli-o tikriausiai [VPtaisy-ti]]
drive.out-pst(3) road-gen.sg probably repair-inf
‘They went probably to repair the road.’ (FL06: 260)
(20) Pastat-ė daržin-ę [InfPšien-ui [vPmums sukrau-ti]].
build-pst(3) hayloft-acc.sg hay-dat.sg we-dat keep-inf
‘They built a hayloft for us to keep hay.’ (FL06: 266)
The correlation between purpose/goal semantics, non-canonical case marking

and displacement of the object naturally raises the question whether the Dative/
Genitive NP does not in fact occupy a position in the main clause where it re-
ceives its Dative or Genitive case from the main verb. And indeed, both Dative
and Genitive NPs can appear as purpose/goal adjuncts without an Infinitive (see
Sawicki 1992), cf. (21) and (22).
(21) a. Čia bu-s lentyn-a knyg-oms.
here be-fut(3) shelf-nom.sg book-dat.pl
‘Here will be a shelf for books.’ (Kerevičienė 2008: 182)
b. Žem-ė keli-a-s darb-ui ir kūryb-ai.
earth-nom.sg get.up-prs(3)-rfl work-dat.sg and creation-dat.sg
‘Earth is getting up for work and creation.’ (Kerevičienė 2008: 182)
(22) a. Išėj-o pien-o.
go.out-pst(3) milk-gen.sg
‘(He/she) went for milk’ (Ambrazas (ed.) 1997: 557)
b. Išsiunt-ė sūn-ų daktar-o.
send-pst(3) son-acc.sg doctor-gen.sg
‘(He/she) sent the son to get the doctor.’ (Ambrazas (ed.) 1997: 557)
However, FL06 show by syntactic tests that the displaced dative or genitive object
in fact forms a constituent with the Infinitive. This is evidenced by the following
diagnostics. First, in many cases it is impossible to omit the Infinitive, since the
object is not always (semantically) licensed in the matrix clause, cf. (23) vs. (21),
(24) vs. (22):
54 Peter Arkadiev
(23) Iššov-ė žmon-ėms *(pagąsdin-ti).

shoot-pst(3) people-dat.pl frighten-inf
‘He fired to frighten the people / *for people.’ (FL06: 271)
(24) Išvažiav-o keli-o *(taisy-ti).
drive.out-pst road-gen.sg repair-inf
‘They went to repair the road / *for the road.’ (FL06: 271)
Second, constituency tests such as coordination (25), tai-“clefting” (26) or frag-

menting (27) also indicate that the Dative (and Genitive) NPs belong to the In-
finitive clause rather than to the matrix clause.
(25) Pastat-ė daržin-ę [šien-ui sukrau-ti] ir [grūd-ams
build-pst(3) hayloft-acc.sg hay-dat.sg keep-inf and grain-dat.pl
apsaugo-ti].
protect-inf
‘They built a hayloft to keep hay and protect grain.’ (FL06: 272)
(26) Tai [šien-ui sukrau-ti] pastat-ė daržin-ę.
it hay-dat.sg keep-inf build-pst(3) hayloft-acc.sg
‘It is (for them) to keep hay that they built a hayloft.’ (FL06: 273)
(27) a. K-am pastat-ė daržin-ę?
what-dat build-pst(3) hayloft-acc.sg
‘For what purpose did they build a hayloft?’
b. Šien-ui sukrau-ti.
hay-dat.sg keep-inf
‘To keep hay.’ (FL06: 272–273)
All this indicates that the relation between the “independent” Dative and Geni-
tive expressions of goal and purpose shown in (21) and (22) and the Dative-plus-
Infinitive and Genitive-plus-Infinitive constructions is probably only a historical
one (cf. Ambrazas 1981, 1987 on the origins of the Dative-plus-Infinitive con-
struction), while synchronically both the Dative and Genitive direct object belong
to the embedded infinitival clause.
3.3 Argument vs. adjunct
Finally, FL06 claim that Dative-plus-Infinitive and Genitive-plus-Infinitive claus-

es differ in their syntactic status: the former are purpose adjuncts, whereas the
latter are rather (optional) goal arguments of motion verbs. The evidence for this
distinction comes from the difference in wh-extraction possibilities from the two
types of construction: the Dative-plus-Infinitive construction does not allow ex-

traction of wh-words, cf. (28), while the Genitive-plus-Infinitive construction
freely allows such question formation, cf. (29). (In Generative grammar, it is as-
sumed that wh-extraction is prohibited from adjuncts but permitted from argu-
ments. The exact details of this generalization, supported by data from English
and some other languages, are not relevant here. See, e.g., Szabolcsi 2006.)
(28) *K-ami atneš-ė vanden-s [ti palaisty-ti]?
what-dat bring-pst(3) water-gen.sg pour-inf
intended: ‘What did he bring some water in order to pour on?’ (FL06: 277)
(29) K-oi atėj-o [ti aplanky-ti]?
who-gen come-pst(3) visit-inf
‘Whom did he come to visit?’ (FL06: 278)
FL06 propose to capture this putative difference between the Genitive-plus-In-

finitive and the Dative-plus-Infinitive constructions in terms of different syntac-
tic structures: while the purpose Dative-plus-Infinitive clause is adjoined to the
matrix VP or NP, the Genitive-plus-Infinitive construction is selected by a special
Asp(ect) head associated with verbs of motion (FL06 remain somewhat vague as
to what the independent motivations for postulating such a functional head are,
see below). In the next section their analysis is spelled out in more detail.
3.4 FL06’s analysis
The Lithuanian constructions with the Dative and Genitive marking of the direct
object of the Infinitive raise two main problems for the “mainstream” Minimalist
case theory. First, they clearly pose problems for the usual assumption that NPs re-
ceive case only once and locally. Indeed, if the structure of a transitive verb phrase
in Lithuanian is as in (30), and the Accusative case on the direct object is valued
by the transitive v, as in finite and ordinary infinitival clauses, then what causes
the replacement of this Accusative by Dative or Genitive in purpose clauses?
(30) vP
3
NPsbj v′
3
v[+acc] VP
3
acc
V NPdo
56 Peter Arkadiev
Since the “non-canonical” case marking in the Lithuanian constructions in ques-

tion is associated with purpose semantics and, moreover, with a particular type
of verb, i.e. motion verbs in the case of the Genitive-plus-Infinitive construction,
the source of the non-Accusative case on the direct object of the Infinitive can
only be located outside of the vP, e.g. in the CP domain of the purpose infinitival
clause in case of the Dative-plus-Infinitive or in the matrix clause in case of the
Genitive-plus-Infinitive. However, in both cases the direct object must have al-
ready received its Accusative case from its own v, and further case assignment is
impossible, cf. (31), showing a putative structure of the purpose infinitival clause
with the null purpose C head being the possible source of the Dative case (the
illicit case assignment of the Dative case by the C to the object NP already marked
Accusative by the v is shown by the crossed arrow).
(31) CP
3
C[purp+dat] TP
3
dat
T[inf] vP
3
pro v′

x 3
v[+acc] VP
3
acc
V NPdo
The only way to ensure that some higher head is able to successfully assign the
Dative case to the direct object is to stipulate that “v has two variants, one with
features valuing Case and the other without” (FL06: 248), and that is what FL06
actually do in their analysis (FL06: 275). Another domain where such a “defec-
tive” or “inactive” v is needed in Lithuanian is the Genitive of negation shown in
example (3) above. Once the VP containing an object NP is merged with such a
“defective” v, Accusative is not assigned and the object thus remains available for
case assignment from some other head.
However, suspending the case assignment ability of v does not suffice to de-
rive the actual case marking, since the relation between C and the object in the
position of the complement of V cannot be established yet for other, independent,
reasons: vP is a phase, hence its constituents, including the object, cannot be ac-
cessed by heads outside of the vP due to the Phase Impenetrability Condition (9),
unless they move to some position at the edge of the vP. This is precisely what
happens in the Dative-plus-Infinitive construction, where, according to FL06, the
object raises and adjoins to the left of the vP as in (32). The diagram shows that
unless the direct object NP raises outside of the phase delineated by the square,
assignment of the Dative to it by the C head is ruled out.

(32) TP (FL06: 274)
3
T vP
3
vP CP
4 3
NPi C[purp+dat] TP
3
T[inf] vP
dat
3
NPdo vP
3
x
PROi v′
3
v[–acc] VP
3
V NPdo
In a similar vein, FL06 account for the Genitive-plus-Infinitive construction: here,

the infinitival TP with a (now optionally) “defective” v is selected by the matrix
VP itself embedded under the already mentioned aspectual projection respon-
sible for the Genitive case assigned to the object of the Infinitive, which again has
to move and adjoin to the left of the lower vP, see (33).
58 Peter Arkadiev

(33) TP (FL06: 277)
3
T vP
3
NPi v′
3
v AspP
3
Asp[+gen] VP
3
V′ TP
4 3
T[inf] vP
gen
3
NPdo vP
3
x
PROi v′
3
v[–acc] VP
3
V NPdo
Discussing the nature of the Lithuanian “object shift” and its relation to case as-
signment, FL06 (244–249, 285–286) call this kind of movement “agnostic”, be-
cause the object NP, unable to have its case features valued by the closest head (v),
moves to a higher position, where its case features can be valued by some – yet
unmerged – head whose precise nature is not clear at the stage of derivation where
movement applies (in a fashion similar to other successive-cyclic movement op-
erations, e.g. long-distance wh-fronting). The only reason for such a movement,
in FL06’s words, is “the possibility of salvation coupled with the certainty of death”,
i.e., the crash of the derivation if the object remains in its base position where nei-
ther its closest available head assigns case to it, nor any higher head can reach it
due to the Phase Impenetrability Condition. This kind of movement operation is
peculiar in that it is not “driven by the need to check features” (FL06: 243), and in
this respect FL06’s proposal is unorthodox, since such movement had not been
proposed before as a possible mechanism of case-assignment.
3.5 Summary
To summarize, the main empirical arguments (E) and analytical conclusions (A)
of FL06 are as follows:
E1. Only structural (Accusative), and not inherent, case can be replaced by the
Dative or Genitive in the infinitival purpose constructions.
E2. Replacement of the Accusative by the Dative is obligatory, while replacement
of the Accusative by the Genitive is optional.
E3. The constructions display a strong (with the Dative-plus-Infinitive) and weak
(with the Genitive-plus-Infinitive) preference for the preverbal position of
the object, and this preverbal position is at the left edge of the embedded vP.
E4. In both constructions, the object forms a constituent with the Infinitive.
E5. Dative-plus-Infinitive clauses are adjuncts whereas Genitive-plus-Infinitive
clauses are arguments.
A1. Lithuanian transitive verbs can have both an “active” v assigning Accusa-
tive to the direct object and an “inactive” v unable to assign case; this makes
it possible to formulate the account not violating the general assumption
of case theory that case of an NP can be assigned/valued only once in the
derivation.
A2. The object needs to move “agnostically” to the left edge of the vP in order to
enable some higher head to assign case to it; the Dative case is assigned by
the null purposive C, and the Genitive is assigned by the Asp(ectual) head
associated with verbs of motion; in both cases, Phase Impenetrability Condi-
tion is not violated.
In the next section I will critically review both theoretical and empirical parts of
the FL06’s argument, showing that their account of the non-canonical case mark-
ing of the direct object of infinitival clauses in Lithuanian is not entirely satisfac-
tory for conceptual reasons and rests on incomplete and not fully accurate data,
and therefore should be abandoned.
4. Franks and Lavine (2006) revised
There are several empirical as well as conceptual problems with FL06’s analysis,
which will be discussed here together with additional data, coming both from
native speakers and the Internet. I will start by briefly listing problems with FL06’s
60 Peter Arkadiev
theoretical assumptions and then will proceed with a more lengthy discussion
of empirical data. Note that if in the previous section grammaticality judgments
were given according to FL06 and their informants and sources, the responsibility
for the examples presented in this section lies on me and my interpretation of the
sources and my informants’ judgments.
4.1 Conceptual problems of FL06
Though most of the details of the analytical proposal advanced by FL06 seem to
inevitably follow from the general assumptions about the nature of case assign-
ment in Minimalism outlined in §2 (i.e., unique and local assignment/valuation
of case features by the closest available case-assigning head), they neither look
really explanatory nor seem to fully account for the data.
First of all, FL06’s proposal (A1 above) that v comes in two variants, one with
a [+acc] feature and the other with a [–acc] feature, does not offer an explanation
of case variation (not only in the Infinitive clauses, but also with the Genitive of
negation) but rather looks like a mere restatement of the facts. There is no reason
independent of case marking to postulate a non-case-assigning v bearing no obvi-
ous semantic difference from its case-assigning partner, in particular in that both
varieties of v are associated with an agentive theta-role of the external argument
merged in their specifier.6 In my view, stipulations like this should be avoided in
a truly explanatory account.
Second, the same can be said about the Asp projection postulated for verbs
of motion and assigning Genitive to the object of the infinitival clause or to the
“independent” Genitive NP denoting purpose of motion. The motivation FL06
(276) give for such a move is at best insufficient, and since FL06 do not “provide
a theory of aspect or event structure” (ibid.) which would independently sup-
port the introduction of such an aspectual head, nor explore its repercussions
elsewhere in the grammar or semantics, I can only conclude that this Asp head is
again a stipulation in order to have a suitable case assigner for the Genitive.
Finally, the other problem of FL06’s account of case and word order in Lithua-
nian infinitival clauses, now not only conceptual but also empirical, is their failure
to account for the availability of the VO order in these constructions, especially
in the Genitive-plus-Infinitive construction with verbs of motion. Indeed, since
6. FL06’s remark (p. 248) that the “defective” v “would also be needed for intransitive clauses”
does not seem warranted, since unaccusative intransitive clauses without agentive subjects ar-
guably lack a vP projection altogether while unergative intransitive clauses differ from transi-
tive ones not in the nature of their functional head v, but in the fact that their lexical V either
does not select a complement at all or assigns to it some inherent case.
FL06 (256) themselves note that in the Genitive-plus-Infinitive construction both

word orders are considered neutral, cf. example (18) above, this implies that the
object of the Infinitive does not necessarily need to move to the edge of the vP in
order to be assigned Genitive from the matrix clause.7 This fact seems to under-
mine the whole argumentation of FL06, since it means that (i) the object can be
somehow accessed by a non-local case-assigning head even inside the VP and,
consequently, (ii) that “agnostic” movement is not necessary in order to rescue
the derivation, which can happily converge even if the object remains in situ. This,
in my view, casts doubt on the close link between case marking and word order
(A2 above) assumed by FL06.
In the next subsection I will present further empirical evidence showing,
among other things, that case marking and word order in infinitival clauses of
Lithuanian should preferably not be lumped together and most probably involve
synchronically independent phenomena, and that, therefore, FL06’s analysis is
untenable.
4.2 Testing FL06’s empirical claims
According to the data I collected, of the empirical points (E1)–(E5) listed in §3.5,
only (E4) concerning the constituency of the infinitival constructions holds with-
out any further qualification. Below I will re-evaluate each of the FL06 empirical
claims on the basis of new data.
(E1) Structural vs. inherent case. Surprisingly, it turns out that not only struc-
tural Accusative, but also inherent Genitive can sometimes be replaced by the
Dative in the infinitival constructions with OV order, at least for some speakers
(partial acceptability is indicated by the % sign), cf. example (34).
(34) a. J-ie nor-i [išveng-ti kar-o].
3-nom.pl.m want-prs(3) avoid-inf war-gen.sg
‘They want to avoid war.’
b. %J-ie dėj-o pastang-as [kar-ui išveng-ti].
3-nom.pl.m put-pst(3) effort-acc.pl war-dat.sg avoid-inf
7. As has been already mentioned in §3.2, the account of the VO order in the Genitive-plus-
Infinitive construction hinted at by FL06 (257) considers such cases as involving an entirely dif-
ferent syntactic structure, i.e. a complex predicate assigning the Genitive to the object, but since
this analysis is not outlined in any detail and is not supported by any independent evidence, I
see no reason to seriously discuss it here.
62 Peter Arkadiev
c. %J-ie dėj-o pastang-as [išveng-ti kar-ui].

3-nom.pl.m put-pst(3) effort-acc.pl avoid-inf war-dat.sg
‘They made efforts to avoid war.’
In particular, the quantificational Genitive assigned by verbs with the cumulative

preverb pri- is always obligatorily replaced by the Dative, cf. (35)–(36):
(35) a. Vaik-ai skin-a gėl-es.
child-nom.pl pick-prs(3) flower-acc.pl
‘The children are picking up flowers.’
b. Vaik-ai pri-si-skyn-ė gėli-ų.
child-nom.pl prv-rfl-pick-pst(3) flower-gen.pl
‘The children have picked up (lots of) flowers.’
(36) a. *Vaik-ai atneš-ė krepš-į [prisiskin-ti gėl-ių].
child-nom.pl bring-pst(3) basket-acc.sg pick.up-inf flower-gen.pl
b. Vaik-ai atneš-ė krepš-į [gėl-ėms prisiskin-ti].
child-nom.pl bring-pst(3) basket-acc.sg flower-dat.pl pick.up-inf
‘The children brought a basket to pick up flowers.’
Examples parallel to (34) with the Genitive replaced by the Dative in purpose
infinitival clauses are also found on the Internet, cf. (37):
(37) a. Veiksm-ai [ši-oms problem-oms išveng-ti].
action-nom.pl this-dat.pl.f problem-dat.pl avoid-inf
‘Actions in order to avoid these problems.’8
b. …dokument-ai [finansin-ei param-ai prašy-ti].
document-nom.pl financial-dat.sg.f support-dat.sg ask-inf
‘documents in order to ask for financial support.’9
It is important to note that this unexpected phenomenon does not seem to be

correlated with the non-standard use of the Accusative case on the object of verbs
such as išvengti ‘avoid’ or prašyti ‘ask’, since the latter use is attested much less
frequently than the Dative marking with the Infinitive, cf. (38).10
8. http://support.google.com/webmasters/bin/answer.py?hl=lt&answer=76401
9. http://kekstas.darbastalis.lt/istorija/
10. To be sure, the standard variants with the Genitive marking of the object are many times
more frequent.
(38) Google search 04.01.2013:11

problemoms (Dat) išvengti ‘to avoid problems’ > 40 results
išvengti problemas (Acc) 2 results
išvengė problemas (Acc) ‘avoided problems’ 0 results
paramai (Dat) prašyti ‘to ask for support’ 7 results
prašyti paramą (Acc) 0 results
prašo paramą (Acc) ‘asks for help’ 4 results
In contrast to the replacement of the inherent Genitive by the Dative in purpose

infinitival clauses, the replacement of cases other than Accusative (e.g. Instru-
mental) by the Genitive in goal infinitives with verbs of motion seems to be con-
sistently banned.
(E2) At least for some native speakers, Accusative to Dative conversion is not
obligatory in purpose infinitival constructions. Accusative is also possible, at least
as a marginal option, especially in the VO order, cf. (39).
(39) a. %Jon-as nupirk-o medžiag-as [pastaty-ti nam-ą].
Jonas-nom.sg buy-pst(3) material-acc.pl build-inf house-acc.sg
‘Jonas bought materials to built a house.’
b. %Mes pastat-ė-me ligonin-ę [gydy-ti vaik-us].
we.nom build-pst-1pl hospital-acc.sg treat-inf child-acc.pl
‘We built a hospital to treat children.’
Though examples like (39) seem to be on the verge of grammaticality, Accusa-

tive tends to become the primary option when another Dative NP is present in
the infinitival clause. This will be discussed below in relation to FL06’s obser-
vation (E4).
(E3) The correlations between case marking and “object shift” postulated by
FL06 turn out to have a much subtler nature.
First, the native speakers I consulted (five Vilnius University students) all tend
to freely allow VO in both types of Infinitive constructions, cf. (40) and (41). Ac-
cording to their judgments, the VO order, consistent with the general pattern of
the language, is often regarded as neutral, whereas the OV order is associated with
focus or emphasis. Contra FL06, Dative-plus-Infinitive and Genitive-plus-Infini-
tive constructions do not seem to differ in this respect either from each other or
from Accusative-plus-Infinitive constructions.
11. Since LKT does not provide morphological annotation, using it for the study of infinitival
constructions is virtually impossible. I obtained my “corpus” results by simple Google searches
performed in January and February 2013. The number of occurrences has been established
manually by filtering out all irrelevant data (e.g. examples showing a different type of construc-
tion) and multiple occurrences of identical examples.
64 Peter Arkadiev
(40) Jon-as atėj-o [aplanky-ti draug-o].

Jonas-nom.sg come-pst(3) visit-inf friend-gen.sg
‘Jonas came to visit his friend.’
(41) Mes pastat-ė-me ligonin-ę [gydy-ti vaik-ams].
we:nom build-pst-1pl hospital-acc.sg treat-inf child-dat.pl
‘We built a hospital in order to treat children.’
Second, the analysis of elicited material is partly supported by corpus data. Thus,
for the Genitive-plus-Infinitive construction, the VO order is clearly preferred to
the OV order, as shown in (42).
(42) Google search 03.01.2013:
aplankyti draugo ‘to visit a friend’ ~ 80 results
draugo aplankyti ~ 35 results
pasiimti vaiko ‘to take the child’ ~ 200 results
vaiko pasiimti ~ 160 results
nusipirkti laikraščio ‘to buy a paper’ 45 results
laikraščio nusipirkti 5 results
The OV order seems to be preferred only when the matrix verb is itself a prenom-
inal attributive participle, cf. (43):
(43) a. [[draug-o aplanky-ti] atėj-ęs] berniuk-as.
friend-gen.sg visit-inf come-pst.pa.nom.sg.m boy-nom.sg
‘the boy who came to visit his friend’12
b. [[vaik-o pasiim-ti] atvyk-us-i] mam-a
child-gen.sg take.rfl-inf arrive-pst.pa-nom.sg.f mother-nom.sg
‘the mother who came to take along her child’13
Thus, the correct generalization about the word order in the Genitive-plus-Infini-
tive construction is that the latter does not significantly differ in its word order
possibilities from the regular Infinitive clauses with the Accusative marked object,
cf. the comparable statistics for the latter in (44).14
12. http://www.delfi.lt/news/daily/crime/article.php?id=14784007
13. http://www.15min.lt/komentaras/2492729
14. The two-tailed version of Fisher’s exact test applied to the data on OV vs. VO order in
Genitive-plus-Infinitive and Accusative-plus-Infinitive structures yielded p > 0.2 for “visit a/the
friend” and p > 0.089 for “buy a/the newspaper”, neither of which indicates a statistuically sig-
nificant relation between case and word order.

aplankyti draugą ‘to visit a friend’ ~ 70 results
draugą aplankyti 20 results
nusipirkti laikraštį ‘to buy a paper’ ~ 110 results
laikraštį nusipirkti 30 results
However, for the Dative-plus-Infinitive construction, the Internet data show a

clear preference for OV order, though VO order is evidently a well-established,
though yet minor, pattern, cf. (45).
durims uždaryti ‘to close doors’ 30 results
uždaryti durims 7 results
namui pastatyti ‘to build a house’ ~ 300 results
pastatyti namui 10 results
mašinai nusipirkti ‘to buy a car’ 60 results
nusipirkti mašinai 11 results
nuomai sumokėti ‘to repay the loan’ ~ 55 results
sumokėti nuomai ~ 10 results
knygai skaityti ‘to read a book’ ~ 70 results
skaityti knygai 10 results
The VO order occurs e.g. when the object NP is heavy, cf. (46) and (47).
(46) neelektrini-ai įtais-ai [atidary-ti ir uždary-ti

non.electric-nom.pl.m device-nom.pl open-inf and close-inf
dur-ims, lang-ams ir langin-ėms].
door-dat.pl window-dat.pl and shutter-dat.pl
‘non-electric devices for opening and closing of doors, windows and
shutters’15
(47) Vis dėlto ne pat-s geriausi-as laik-as
however neg very-nom.sg.m best-nom.sg.m time-nom.sg
[rašy-ti laišk-ui apie sav-e]…
write-inf letter-dat.sg about self-acc
‘However, it’s not the best time to write about oneself…’16
However, the Google data are inconclusive, cf. (48) showing heavy postverbal Da-
tive objects vs. (49) with heavy preverbal Dative objects.
15. http://isdv.upv.cz/portal/pls/portal/portlets.ozs.det?pozk=729339&plan=en
16. http://www.rasyk.lt/dienorastis/195020/195020.html
66 Peter Arkadiev

pastatyti namui ar butui ‘to build a house or a flat’17 vs.
no: namui ar butui pastatyti
malti mėsai ir žuviai ‘to chop meat and fish’18 vs.
no: mėsai ir žuviai malti

butui ar mašinai nusipirkti ‘to buy a flat or a car’ 3 results vs.
no: nusipirkti butui ar mašinai
buto nuomai sumokėti ‘to pay the rent for the flat’ 6 results vs.
sumokėti buto nuomai 2 results
In fact, it looks like a considerable part of the instances of the Dative-plus-In-

finitive construction with the OV order is constituted by set phrases which are
not created anew but memorized and repeated and often involve non-referential
objects,19 e.g., mėsai malti mašinėlė ‘mincing machine’, lit. “meat-dat chop-inf
machine”, indas kavai virti ‘coffee maker’, lit. “vessel coffee-dat boil-inf”, pini-
gai nuomai sumokėti ‘money to repay the loan’, lit. “money loan-dat pay-inf”,
paskola namui pirkti ‘loan to buy a house’, lit. “loan house-dat buy-inf” etc. This
hypothesis, however, cannot be substantiated without a statistical analysis, which
is very hard to conduct given the current state of the Lithuanian corpus.
The data at hand, in my view, warrant the conclusion that at least for the
Genitive-plus-Infinitive construction the “object shift” is not connected with case
marking at all. As to the Dative-plus-Infinitive construction, it seems that it is
undergoing a gradual shift towards the neutral VO word order, and thus the two
parameters – case marking and word order – should better be kept apart and not
lumped together in the account of these constructions.
(E4) Though FL06’s observations on the position of adverbials in Dative-plus-
Infinitive and Genitive-plus-Infinitive OV structures, indicative of the object be-
ing shifted to the left periphery of the infinitival clause, is largely confirmed by my
consultants, examples like (20) above where both the Dative subject and the Da-
tive object are present are considered marginal or even sometimes ungrammati-
cal. In such cases, marking of the object with the Accusative is preferred, cf. (50).
17. http://verslas.delfi.lt/nekilnojamas-turtas/article.php?id=19144292&com=1&s=1&no=140
18. http://www.mokslai.lt/referatai/referatas/smulkinimo-irengimai-prekybinese-ir-maitini-
mo-imonese-puslapis5.html
19. I thank Rolandas Mikulskas for pointing out to me the possible relevance of the referential
status of the object in these constructions.
(50) a. %Atidėj-o-me pinig-ų [nam-ui sūn-ui

put.by-pst-1pl money-gen.pl house-dat.sg son-dat.sg
pasistaty-ti].
build:rfl-inf
b. okAtidėj-o-me pinig-ų [sūn-ui pasistaty-ti
put.by-pst-1pl money-gen.pl son-dat.sg build:rfl-inf
nam-ą].
house-acc.sg
‘We put by some money for our son to build a house.’20
Moreover, this tendency to avoid two Dative NPs in an infinitive clause is not
limited to the co-occurrence of the object and the subject in the Dative case, but is
operative in prohibiting the Dative of the direct object in the presence of a Dative
indirect object regardless of word order, cf. (51):
(51) a. *pinig-ai vaz-ai motin-ai nupirk-ti
money-nom.pl vase-dat.sg mother-dat.sg buy-inf
b. *pinig-ai vaz-ai nupirk-ti motin-ai
money-nom.pl vase-dat.sg buy-inf mother-dat.sg
c. pinig-ai nupirk-ti motin-ai vaz-ą
money-nom.pl buy-inf mother-dat.sg vase-acc.sg
‘money in order to buy a vase for the mother’
In the Genitive-plus-Infinitive construction, the Genitive on the direct object is

compatible with the Dative indirect object, but not with the Dative subject (overt
subjects of infinitival clauses appear in the Dative in Lithuanian), cf. (52).
(52) a. Jon-as atėj-o [vaz-os motin-ai padovano-ti].
Jonas-nom.sg come-pst(3) vase-gen.sg mother-dat.sg give-inf
‘Jonas came in order to give the vase to his mother.’
b. *Jon-as atėj-o pas mane [laišk-o mums
Jonas-nom.sg come-pst(3) at me.acc letter-gen.sg we.dat
parašy-ti].
write-inf
intended: ‘Jonas came to me in order for us to write a letter.’
Thus it seems that Genitive-plus-Infinitive constructions selected by verbs of mo-

tion do not have a subject position at all. This is, however, not true of the other
purpose Infinitive constructions, which can have their own Dative subjects, as
20. I thank Rolandas Mikulskas for an illuminating discussion of such examples. Note that the
interpretation of the Dative sūnui in (50) as the beneficiary (‘to build a house for our son’) is
ruled out by the reflexive form of the verb pasistatyti ‘to build for oneself ’.
68 Peter Arkadiev
(50b) shows. Such examples are paralleled by other kinds of infinitival clauses
with Dative subjects, not expressing purpose, cf. (53).
(53) a. … pakeis-ti būd-ą [vis-iems skaity-ti knyg-as]
change-inf way-acc.sg all-dat.pl.m read-inf book-acc.pl
‘to change the way everyone reads books’21
b. tikimyb-ė [vaik-ams susirg-ti alergij-a]
probability-nom.sg child-dat.pl.m fall.ill-inf allergy-ins.sg
‘the probability that children become allergic’22
It is clear from these data that whatever the source of the Dative case on the object
of purposive Infinitive is, some other case-assigner is responsible for the Dative
case on the subject of infinitival clauses in Lithuanian.
(E5) The adjunct vs. complement status of Dative-plus-Infinitive resp. Geni-
tive-plus-Infinitive clauses diagnosed by wh-extraction is not uncontroversial,
since the contrast between (28) and (29) and similar examples in (54) and (55)
can actually be explained as stemming from a Complex NP Constraint violation,
without recourse to the adjunct/complement distinction.
(54) *K-ami atidėj-ai [np pinig-us [InfP ti nusipirk-ti]]?
what-dat put.by-pst.2sg money-acc.pl buy.rfl-inf
intended: ‘What did you put by money to buy?’
(55) K-oi darbinink-ai nuvažiav-o [InfP ti taisy-ti]?
what-gen worker-nom.pl drive.out-pst(3) repair-inf
‘What did the workers go to repair?’
The grammatical variant of (54) involves pied-piping of the whole Infinitive

clause, cf. (56):
(56) [InfP K-am nusipirk-ti]i atidėj-ai pinig-us ti?
what-dat buy-inf put.by-pst.2sg money-acc.pl
lit. ‘What to buy did you put by money?’
Moreover, passivization of the matrix clause, which dissociates the infinitival

clause from the noun, improves extraction, cf. (57):
21. http://www.johns-company.com/index.php?lang=lt&cat=400&month=2009-08&id=54486
22. http://www.alergija.info/view.php?page=104&rpid=2
(57) a. Nauj-oji ligonin-ė buv-o

new-nom.sg.f.def hospital-nom.sg aux-pst(3)
pastaty-t-a [vaik-ams gydy-ti].
build-pst.pp-nom.sg.f child-dat.pl treat-inf
‘The new hospital was built to treat children.’
b. K-ami nauj-oji ligonin-ė buv-o
who-dat new-nom.sg.f.def hospital-nom.sg aux-pst(3)
pastaty-t-a [ti gydy-ti]?
build-pst.pp-nom.sg.f treat-inf
lit. ‘Whom was the new hospital built to treat?’
This suggests (if we take the extraction diagnostic seriously at all, which is not
compelling pending an independent investigation of this phenomenon in Lith-
uanian) that there is no syntactic difference between Genitive-plus-Infinitive
and Dative-plus-Infinitive constructions in terms of the argument vs. adjunct
distinction.
The observed properties of Dative-plus-Infinitive and Genitive-plus-Infini-
tive constructions are summarized in Table 1.
Among the features listed, (d-ii) and (e) are most probably a reflection of a pro-
cessing-related surface ban on two Dative argument NPs in the same clause, while
others call for a deeper structural explanation. My account of case marking in the
constructions in question, in contrast to that of FL06, will rest on the assumption
that word order does not play any important role in this phenomenon, which, I be-
lieve, is warranted by the facts discussed above. (Of course, the analysis proposed
by FL06, or some other analysis taking word order into account, might be inevita-
ble for an earlier stage of Lithuanian where the OV order in the Dative-plus-Infini-
tive and Genitive-plus-Infinitive constructions was obligatory – if such a stage has
ever existed.) This means that, first, the conceptually problematic mechanism of
Table 1. Properties of Dative-plus-Infinitive and Genitive-plus-Infinitive constructions

Dative-plus-Infinitive Genitive-plus-Infinitive
(a) replacement of inherent case marginally possible impossible
(b) accusative retention marginally possible possible
(c) VO order possible, though less frequent preferred
(d) overt subject i. possible with Acc object impossible
ii. marginal with Dat object
(e) Dative indirect object i. possible with Acc object possible with Gen object
ii. impossible with Dat object
(f) wh-extraction possible possible
70 Peter Arkadiev
case assignment via “agnostic movement” proposed by FL06 is unnecessary, and,

second and most importantly, that the case assignment in Genitive-plus-Infinitive
and Dative-plus-Infinitive constructions in Lithuanian is not subject to the famil-
iar constraints on long-distance dependencies, such as the Phase Impenetrability
Condition (contra McFadden 2010). In the next section I will present typological
data which suggests an alternative analysis in terms of multiple case assignment
in syntax, which will be applied to Lithuanian data in §6.
5. Typological parallels
“Non-canonical” marking of objects of infinitival or purpose clauses seems to be

an infrequent phenomenon cross-linguistically; e.g., in a recent monograph on
the typology of purpose clauses (Schmidtke-Bode 2009) such patterns are not
mentioned at all. In this section I will first look for phenomena similar to the
Lithuanian Genitive-plus-Infinitive and Dative-plus-Infinitive constructions in
the related languages and then will proceed to a more remote but instructive par-
allel with the languages of Australia.
5.1 The Baltic and Slavic Supine
In the neighbouring languages (Latvian, Latgalian, Estonian, Slavic), no direct

parallels to the Lithuanian Dative-plus-Infinitive and Genitive-plus-Infinitive
constructions are found, with the exception of the Latgalian Genitive-plus-Su-
pine construction occurring with verbs of motion (for more details on object case
marking in Latgalian and on the Supine construction, see Nau, this volume).
In Latgalian, the Supine is a verbal form morphologically distinct from the
Infinitive (the formal difference between the two is often realized by root vocal-
ism only) and cognate to a very similar form attested in older Lithuanian (see
below). Clauses headed by the Supine denote purpose of motion; with the Supine
of transitive verbs, the direct object is normally encoded in the Genitive case, as
in (58).
Latgalian
(58) Rogon-a izsyutej-a bōrineit-i [drēb-u valāt]
witch-nom.sg send-pst(3) orphan-acc.sg cloth-gen.pl beat:sup
‘The witch sent out the orphan to beat clothes.’ (Nau 2011: 61)
By contrast, the Infinitive in Latgalian, which normally occurs with a different

range of matrix verbs, requires the Accusative marking of the direct object, cf.
(59). However, examples are attested where the Infinitive is used instead of the
Supine with verbs of motion, retaining the Genitive case of the object, as in (60).
Latgalian
(59) Bōrineit-ia sōk-a [viaļāt driāb-is].
orphan-nom.sg begin-pst(3) beat:inf clothes-acc.pl
‘The orphan began to beat clothes.’ (Nau 2011: 61)
(60) Jei aizguoj-a iz klāv-u [dacierp-t pādej-ūs
3:nom.sg.f go.out-pst:3 to barn-acc.sg shear-inf last-gen.pl
vušk-u].
sheep-gen.pl
‘She went out to the barn in order to shear the last sheep.’ (Nau 2011: 79)
In Lithuanian, the Supine construction with the Genitive object was well-attested
in older language up to the beginning of the 20th century, and is reported to
have been used in the North-Eastern Aukštaitian dialects (Zinkevičius 1966: 390),
which border on Latgale, cf. (61)–(63).
Old Lithuanian
(61) Atei-s [sudi-tu giw-u ir nůmirusi-u].
come-fut(3) judge-sup living-gen.pl and dead-gen.pl
‘He will come to judge the living and the dead.’
(VE 1579,23 18:8, Schmalstieg 1987: 174)
19th century Lithuanian

(62) [Svetim-uos mišk-uos malk-ų pirk-tų] važinėj-o.
alien-loc.pl.m forest-loc.pl wood-gen.pl buy-sup ride-pst(3)
‘They rode to buy wood in other people’s forests.’ (BA 1859,24 II)
North-Eastern Aukštaitian
(63) ažusuk [al-aus atsiger-tų]
drop.in:imp.2sg beer-gen.sg have.a.drink-sup
‘Drop in to drink some beer.’ (Zinkevičius 1966: 390)
Beyond Baltic, the Supine with the Genitive direct object was also (vestigially) at-
tested in the older Slavic languages: Old Church Slavonic (Vaillant 1966: 127–129;
1977: 171–172; Lunt 2001: 159–160), Old Russian (Ivanov 1990: 356–357), Old
Czech (Vaillant 1966: 129), see example (64).
23. Baltramiejus Vilentas. Enchiridion. Königsberg, 1579.

24. Antanas Baranauskas. Anykščių šilelis. 1858–1859. http://antologija.lt/text/antanas-
baranauskas-anyksciu-silelis
72 Peter Arkadiev
Old Church Slavonic

(64) id-ǫ [ugotova-tъ měst-a vamъ].
go-prs.1sg prepare-sup place-gen.sg you:dat.pl
‘I am going in order to prepare a place for you.’
(Sav 103025 J 14:2, Lunt 2001: 160)
Among modern Slavic languages, the Supine as a verbal form distinct from the
Infinitive is still attested in Slovene (Brezar et al. 2005: 114), example (65), and
Lower Sorbian (Steenwijk 2003), example (66), but in both these languages the di-
rect object of the Supine is marked by the canonical Accusative and not Genitive.
Slovene
(65) Še-l je [gleda-t nov-i film].
go-pst.m aux.prs.3sg watch-sup new-acc.sg.m film(acc.sg)
‘He went to watch the new film.’ (Brezar et al. 2005: 114)
Lower Sorbian
(66) Witśe pojěd-u [Lenk-u pyta-t].
tomorrow go-fut.1sg Lenka-acc.sg look.for-sup
‘Tomorrow I shall go and look for Lenka.’ (Steenwijk 2003: 333)
Thus, Slovene and Lower Sorbian have chosen a path of development of the origi-
nal Genitive-plus-Supine construction opposite to that of Lithuanian: the former
retained the form but shifted the object case marking to the “canonical” pattern,
while the latter (partly) retained the “non-canonical” case marking but got rid of
the distinction between the Supine and the Infinitive, cf. (67).

(67) NP-Gen V-Inf Lithuanian
NP-Gen V-Sup
NP-Acc V-Sup Slovene, Lower Sorbian
In fact, FL06 (252–257), following Schmalstieg (1987), recognize the Supine origin
of the Lithuanian Genitive-plus-Infinitive construction, but do not synchronically
analyze the latter along the lines of the former, which would involve, in particular,
the assumption that the source of the Genitive on the object is located inside the
non-finite construction and not in the matrix clause. I will explore this hypothesis
further in §6, after bringing forward a very different typological parallel.
25. Sava’s Book, ca. 1030.

5.2 The Australian “complementizing” and “associating” case
Constructions with “non-canonical” case marking of the object of purpose claus-

es are attested in some Australian languages, where the so-called “complemen-
tizing” and “associating” functions of case are singled out by Dench and Evans
(1988) (further DE88). In this subsection I will explore these notions and the
data behind them in some detail. My intention is to show that the Australian
case marking strategies, “exotic” though they may seem at first glance, can be
very instructive for the understanding of the case marking mechanisms in the
Lithuanian constructions discussed in this paper (for further parallels between
Lithuanian and Australian case marking see Arkadiev 2013b). From a theoretical
perspective, I largely follow the spirit of recent proposals in Richards (2007, 2013),
Matushansky (2008, 2010), Erschler (2009), though I diverge from these authors
in the more specific details of my analysis.
Complementizing case is a feature of a subordinate (usually, though not nec-
essarily non-finite) clause and can spread to some or all of its subconstituents
(DE88: 18–23), cf. (68) from Warumungu with the Dative complementizing case
appearing both on the nominalized predicate and on its object.
Warumungu (Pama-Nyungan > Warumungic, Northern Australia)
(68) api-jirra warnapartt=arna [ngapa-ka pari-nji-kki].
walk-towards tomorrow=1sg.fut water-dat get-nml-dat
‘I will go tomorrow to get water.’ (DE88: 19)
Associating case appears on arguments of nominalized verbs instead of the or-

dinary “canonical” case-marking (DE88: 31–32). The crucial difference between
complementizing and associating case is that while the former is assigned from
outside to the whole subordinate clause and can in principle appear on its predi-
cate head, the latter is assigned inside the nominalized clause and never appears
on its head. Consider example (69) from Dhalandji, where the Accusative on the
nominalized head of the relative clause ‘cutting’ is arguably a complementizing
case occurring in agreement with the nominal head ‘man’, while the Dative on the
object ‘meat’ of the relative clause is an instance of associating case.
Dhalandji (Pama-Nyungan > South-Western branch, Western Australia)
(69) ngatha nhaku-nha [kanyara-nha [murla-ku warni-lkitha-nha]].
1sg.nom see-pst man-acc meat-dat cut-rel.ds-acc
‘I saw the man cutting meat.’ (DE88: 31)
Complementizing and associating case functions (or, to put it otherwise, mech-

anisms of case assignment) are postulated by DE88 in addition to such more fa-
miliar types of case functions as relational (appearing on arguments and adjuncts
74 Peter Arkadiev
and indicating their semantic or syntactic relation to the predicate) and adnomi-
nal (appearing on NPs embedded into other NPs). A further, apparently still more
“exotic” type of case function is the so called “modal” case, attested in the Tangkic
family, which involves NPs and other constituents of the clause showing partic-
ular case marking depending on the tense or mood of the predicate, see below.
Different functions of case have different sources and domains of application,
i.e. are associated with different lexical or functional heads:
(70) – relational case is assigned in the VP/vP domain;
– adnominal case is assigned in the NP/DP domain;
– “modal” case (DE88: 23–28) is assigned in the TP domain by (varieties) of
finite T;
– associating case is assigned in the TP domain by (varieties) of nonfinite T;
– complementizing case is assigned in the CP domain by varieties of C.26
The crucial question with respect to (70) is what the syntactic objects to which
different kinds of case are assigned are. I will assume the simplest answer pro-
posed by Matushansky (2008, 2010) following Stowell (1981), i.e., that (by de-
fault) case is assigned by a head to its complement and subsequently percolates to
all subconstituents of the latter. Thus, complementizing case is assigned by the C
to the TP, modal and associating cases are assigned by the T to the vP/VP and so
on; from this, in particular, follows, that the relational Accusative is assigned by
the v to the VP and not directly to the object NP/DP. Crucial evidence for such a
view of case assignment is presented by Matushansky (2008, 2010) on the basis of
predicate nominal case marking in Russian (Matushansky’s analysis, incidentally,
can be extended to similar constructions in Lithuanian as well) and even more so
by the Australian examples discussed below (cf. also the already presented (68)
and (69)).
In a number of Australian languages cases assigned at different levels of struc-
ture do not exclude each other but are expressed by stacked case suffixes whose
order normally reflects the scope of case-assigning domains. Thus in Kayardild,
belonging to the Tangkic family, a single noun may bear up to four consecutive
case markers, e.g. adnominal, relational, modal and complementizing, as in ex-
ample (71); what is most important, in this language modal, associating and com-
plementizing cases appear on all subconstituents of the relevant domain, strongly
supporting the hypothesis of case assignment by the head to its complement with
26. Of course, different kinds of complementizing case can in principle be associated with dif-
ferent layers of Rizzi (1997)’s “extended left periphery”, but I am not going to pursue this ques-
tion here.
subsequent percolation outlined above, cf. the putative structure of (71c) in (72).27
For a detailed illuminating discussion of case marking in Kayardild, see Evans
(1995a, 1995b28).
Kayardild (Tangkic, Northern Australia; Evans 1995a: 102–103, 115–116)
(71) a. dangka-karra-nguni mijil-nguni
man-gen-ins net-ins
‘with the man’s net’
b. maku yalawu-jarra yakuri-na dangka-karra-nguni-na
woman catch-pst fish-m:abl man-gen-ins-m:abl
mijil-nguni-na.
net-ins-m:abl
‘The woman caught fish with the man’s net.’
c. maku-ntha yalawu-jarra-ntha yakuri-naa-ntha
woman-c:obl catch-pst-c:obl fish-m:abl-c:obl
dangka-karra-nguni-naa-ntha mijil-nguni-naa-nth.
man-gen-ins-m:abl-c:obl net-ins-m:abl-c:obl
‘The woman must have caught fish with the man’s net.’
(72) CP
C-Oblique domain
ei
C[must] → TP-Obl
ei
NP T′
M-Ablative domain
womanobl ei
T[pst] → VP-Abl-Obl
catchi ep
VP-Abl-Obl PP-Abl-Obl
relational Instrumental
3 3 domain
NP Vi P[ins] → NP-Ins-Abl-Obl
adnominal Genitive
fishabl-obl 3 domain
N → NP-Gen-Ins-Abl-Obl
netins-abl-obl mangen-ins-abl-obl
27. In (72) I assume for the sake of consistency that the relational Instrumental case is assigned
by a null P(repositional) head responsible for the interpretation of the adjunct instrument
phrase. The case itself is thus void of semantics and only reflects the presence of a semantically
interpretable but in this instance phonologically null case assigner. A different account, assum-
ing that semantic case on NPs does not necessarily need any external case-assigner, is certainly
also possible.
28. For an alternative analysis of Kayardild data see Round (2013).
76 Peter Arkadiev
Turning to non-finite clauses, it is worth noting that in Kayardild there is a spe-

cial verbal form similar to the Indo-European Supine, which appears with matrix
verbs of motion and assigns the Allative to the object. In (73a) this Allative, which
Evans classifies as modal rather than associating for reasons irrelevant to the cur-
rent discussion, appears to the left of the “outer” “modal” case assigned by the ma-
trix T; the structure of (73a) is given in (73b). Note that I assume that the Supine
is a special kind of non-finite T head and that there is no CP layer intervening
between the Supine construction and the matrix clause; this seems to be empiri-
cally supported for Kayardild, but I won’t go into the details of argumentation,
see Evans (1995: 486–487), where it is argued that Kayardild Supine clauses share
features of both finite and non-finite structures.
Kayardild
(73) a. balmb-u ngada warra-ju [bijarrba-ring-ku
tomorrow-m:prop 1sg:nom go-pot dugong-m:all-m:prop
raa-jiring-ku].
spear-sup-m:prop
‘Tomorrow I will go to spear dugong.’ (Evans 1995a: 487)
b. TP
3
NPi T′
M-Proprietive domain
Inom 3
T[pot] → VP-Prop
gom ei
VP-Prop TP-Prop
eh ei
Adv Vm NP T′
M-Allative domain
tomorrowprop proi ep
T[sup] → VP-All-Prop
to-spearprop-k 3
NP Vk
dugongall-prop
Looking further into Australian languages, we find the most striking paral-
lel to the Lithuanian Dative-plus-Infinitive construction in the Pama-Nyungan
languages Nyamal and Jiwarli. These languages have non-finite purpose claus-
es whose object appears in the Dative, cf. (74) and (75). In particular, examples
(74b) and (75b) can be translated into Lithuanian literally, with the use of the
Dative-plus-Infinitive construction, while example (74a) could be rendered by
the Genitive-plus-Infinitive construction (and in Latgalian, by the Genitive-plus-
Supine construction). The diagram in (76) shows the proposed structure of (74b);
I assume that the Dative is a complementizing case assigned by the purposive

C head.
Nyamal (Pama-Nyungan > South-Western branch, Western Australia)
(74) a. Ngunti-rna-rna jilya [kurti-larta yurta-yu].
send-pst-1sg child get-purp fish-dat
‘I sent the child to get fish.’ (Dench 2009: 761)
b. Ngunya-ngku mangkurla-lu warnta kurti-la [punga-lartara-lu
that-erg woman-erg stick get-prs hit-purp-erg
yukurru-ku].
dog-dat
‘That woman is getting a stick to hit the dog.’ (ibid.: 767)
Jiwarli (Pama-Nyungan > South-Western branch, Western Australia)

(75) a. Ngatha kamurri-a-rru [pirru-wu thika-lkarringu].
I(nom) get.hungry-prs-now meat-dat eat-purp
‘I am becoming hungry to eat meat.’ (Austin 2009: 4)
b. Kuwarti kurriya purra-rninyja [patha-rrkarringu-ru jiriparri-yi].
now boomerang toss-pst pelt-purp-erg echidna-dat
‘Next (he) threw a boomerang to hit echidna.’ (ibid.)
(76) TP
3
T[prs] vP
Ergative domain
ep
DPi-Erg v′
6 3
thaterg womanerg v → VP-Erg
getk ei
VP CP-Erg
Dative domain
3 fp
NP Vk C[purp] → TP-Dat(-Erg)
stick(abs) to-hitk-erg fy
Tk vP-Dat(-Erg)
3
NP v′
proi 3
vk VP-Dat(-Erg)
3
NP Vk
dog(abs)dat
78 Peter Arkadiev
(74b) and (76) raise the obvious question why the object of the purpose clause
bears only the Dative case and not the double Dative-Ergative, especially since
otherwise Nyamal allows case-stacking, cf. (77), where the Dative complementiz-
ing case (occurring in agreement with the corresponding relational case) follows
the Elative associating case in a nominalized relative clause.
Nyamal
(77) a. Wurtama-la nyumpalanga-mu [mayi-kapu-ku kama-njanu-ku].
wait.for-ant you.du.dat-ant food-elat-dat cook-rel-dat
‘He’ll wait for you two who are cooking food.’ (Dench 2009: 766)
b. VP Dative domain
3
V → NP-Dat
wait ep
NPi CP-Dat
you.twodat 3
C TP-Dat
3
NP T′
Elative domain
proi ro
T[rel]k → vP-Elat-Dat
cookingdat 3
vk VP-Elat-Dat
3
NP Vk
food(abs)elat-dat
The only empirically tenable answer to the question regarding the presence of
case-stacking in (77) and its absence in (74) (cf. Dench 2009: 766–768) is that
there exist language-specific morphological restrictions on the co-occurrence or
co-expression of several cases (see also DE88: 35–43). Thus, in Djapu (DE88: 40–
41) relational case markers must be omitted before the complementizing case
markers, cf. (78), and Locative case markers are deleted after the (adnominal)
Oblique, cf. (79).
Djapu (Pama-Nyungan >Yuulnguan, Northern Australia):
(78) ngayi rongiyi-n [nha-nhara-ngur malu-‘mirringu-wal].
he return-prf see-nml-abl father-kin-(*rel.case)-obl
‘He came back from seeing his father.’ (DE88: 40)
(79) waanga-ngur [yapa-‘mirringu-wal(*-ngur) ngarra-kalangu-wal(*-ngur)].
camp-loc sister-kin-obl(*-loc) I-obl-obl(*-loc)
‘at my sister’s camp’ (ibid.: 41)
Further, in Lardil (a Tangkic language related to Kayardild), in a fashion reminis-

cent of the behaviour of structural vs. inherent cases in Lithuanian or Slavic, the
modal “future” case replaces the structural Accusative case but attaches outside of
the inherent Instrumental case (Richards 2007, 2013), cf. (80).
Lardil (Tangkic, Northern Australia):
(80) a. ngada nguthungu warnawu dulnhuka-n beerr-u nyith-u.
I:nom slowly cook month.fish-acc ti-tree-ins fire-ins
‘I slowly cooked the month-fish on a fire of ti-tree wood.’
(Richards 2013: 48)
b. ngada nguthungu-thur warnawu-thur dulnhuka-r
I:nom slowly-fut cook-fut month.fish-fut
beerr-uru-r nyith-uru-r.
ti-tree-ins-fut fire-ins-fut
‘I will slowly cook the month-fish on a fire of ti-tree wood.’ (ibid.)
In fact, Richards (2007, 2013) draws a direct parallel between the Lardil “future”
case shown in (80) and the Genitive of negation rule in Russian, which applies
only to the structural Accusative case, and this parallel can certainly be extended
to Lithuanian, cf. (8) above.
Anyway, it remains an open question where the co-occurrence restrictions
and rules governing case-stacking and case-resolution belong (syntax vs. mor-
phology, case values vs. case markers etc.), and most probably this is a domain of
intra- and interlinguistic variation (cf. Richards 2007, 2013 and Erschler 2009 for
very different proposals). In the discussion of Lithuanian in the next section I will
assume that phenomena similar to those shown in (76), (78)–(80) belong to the
domain of morphology rather than syntax, and – admittedly, for aesthetic rea-
sons – I believe that to assign cross-linguistic variation and sometimes quite idio-
syncratic language specific phenomena to morphology, where many irregularities
belong anyway, is conceptually more attractive than trying to capture them in
syntax by postulating ad hoc solutions and unnecessary complications.
To conclude this section, I would like to say that since the mechanism of
multiple case assignment is anyway necessary not only to account for the phe-
nomena in the Australian languages, but also elsewhere (cf. Plank (ed.) 1995 on
Suffixaufnahme in the world’s languages), and even not only for the instances of
overt multiple case marking (cf. McCreight Young 1988; Béjar & Massam 1999;
Matushansky 2008, 2010; Erschler 2009), I see no conceptual obstacles to extend-
ing this mechanism beyond Australia and, specifically, to accounting for the Lith-
uanian constructions along the lines of the schematic analysis of the Australian
data presented above.
80 Peter Arkadiev
6. Back to Lithuanian: A new analysis
I will start by drawing upon an arguably simpler case hinted at in the previous
section, viz. the Genitive of negation. As has been already mentioned, this rule
applies to the Accusative direct objects but not to objects marked by inherent cas-
es (cf. (8) above), and targets the objects of the infinitival complement clauses as
well (cf. (3b) above). The Lithuanian Genitive of negation can be easily accounted
for with the mechanism of multiple case assignment (cf. a similar proposal for
the much more complicated Russian data in Erschler 2009) as a kind of “mod-
al” case assigned by the Neg head to the vP. The analysis has two components:
the relatively straightforward syntactic one shown in the tree diagram in (81a)
corresponding to (8b), and the morphological case resolution rules outlined in a
simplified form in (81b).29 From now on the case values realized morphologically
will be underlined in the diagrams.
29. I assume the simplest model of case, whereby the case values assigned in syntax directly
match the language specific values of morphological case, and are not decomposed into some
more abstract features (cf. Matushansky 2008, 2010 and Keine 2010 for the latter view). Also,
my model of case resolution is cast in terms of simple rules, though a more sophisticated ap-
proach, e.g., in the spirit of Optimality Theory (see Erschler 2009) is probably preferable. I leave
the choice of a more adequate formal representation of case resolution for future research.
Note also that I do not propose any mechanism assigning the Nominative case to the sub-
ject of the finite clause, because this issue is largely irrelevant for the purposes of the present
paper. The most natural and empirically plausible solution under the current theory would be
to assume that the Nominative is assigned by the finite T to the vP/VP, and percolates to the
NP occupying its specifier, which then can, but need not, move to Spec,TP. Integrating the
Nominative into the system of case resolution rules for Lithuanian presumably would not be
difficult, but this can be left for further research. Finally, I do not assume that Nominative is just
the “default” or “unmarked” case appearing where no other case is assigned; such a view does
not seem to be warranted for Lithuanian or at least for the constructions I am discussing; it is
also worth noting that in their analysis of the Lithuanian Nominative-plus-Infinitive construc-
tions, FL06 (278–284) explicitly argue against treating Nominative as default case. I thank Ora
Matushansky for suggesting that I clarify this issue.

(81) a. NegP
Genitive domain
3
Neg → vP-Gen
eo
NP v′ Accusative domain
Aldonanom;gen 3
v → VP-Acc-Gen
givei eo
NP V′
Dative domain
appleacc;gen 3
Vi → NP-Dat-Acc-Gen
brotherdat;acc;gen
b. case resolution rules (simplified):
[acc][gen] → [gen]
[α-case30][gen] → [α-case]
The long-distance Genitive of negation like in (3b) repeated here as (82a) falls
out naturally as well (for the reasons of pure simplicity of exposition I treat the
infinitival clause as a bare TP), see (82b).
(82) a. Jon-as ne-nor-i [perskaity-ti laišk-o].
Jonas-nom.sg neg-want-prs(3) read.through-inf letter-gen.sg
‘Jonas does not want to read the letter.’
b. NegP
3 Genitive domain
Neg → vP-Gen
eo
NPi VP-Gen
Jonasnom;gen 3
V TP-Gen
want eo
T[inf] vP-Gen
to-readk 3
NP v′
Accusative domain
proi 3
vk → VP-Acc-Gen
3
Vk NP
letteracc;gen
30. “α-case” means “any case value”.

82 Peter Arkadiev
The analysis I propose for the Dative-plus-Infinitive and Genitive-plus-Infinitive

constructions is basically similar to that of the Genitive of negation, though more
complex. The syntactic part of the analysis is fairly simple: I assume that the Da-
tive and Genitive cases are assigned by some higher heads at least to the whole vP
containing the object of the Infinitive and then percolate to its subconstituents.
The question of what these case-assigning heads are will be addressed separately
for each construction. The morphological side is more intricate, since it has to
capture most of the concrete phenomena discussed in §§3 and 4. Generally, just
like with the Genitive of negation, the “outer” Dative and Genitive cases are real-
ized when they combine with the “inner” structural Accusative and deleted oth-
erwise, but this is certainly not sufficient because, as has been shown in §4, first,
the replacement of the Accusative by the “outer” cases is optional in the Infinitive
constructions (in contrast to the obligatory Genitive of negation rule), and, sec-
ond, Dative in purposive infinitival clauses can replace not only the structural
Accusative but the putatively inherent Genitive as well.31 Together with the soft
constraint against the Dative case marking of the direct object in the presence of
a Dative subject or indirect object this yields the resolution rules listed in (83).32

(83) i. [acc][dat] → [dat] or [acc]
ii. [acc][gen] → [gen] or [acc]
iii. [gen][dat] → [gen] or marginally [dat]
iv. [α-case][dat] → [α-case]
v. [α-case][gen] → [α-case]
vi. Surface constraint: *[tp NPdat … NPdat]
Let us now turn to the analysis of individual constructions. For the Dative-plus-
Infinitive I propose that the Dative is a complementizing case assigned by C[purp]
to the infinitival TP, which actually follows the proposal in FL06 (274), with the
only difference that my analysis does not assume any locality restriction and hence
does not require the case marking to be in any way linked to word order (compare
also the Nyamal purposive construction in (74b) and (76)). Cf. the diagrams in
31. As pointed out by David Erschler, perhaps the best material to demonstrate this kind of
case interaction would be negated purpose infinitival clauses with a competition between the
Dative of purpose and the Genitive of negation. However, my consultants almost unanimously
rejected examples of negated purpose Infinitives, so this question remains unresolved.
32. From the comparison of (81b) and (83) it is clear that the case resolution rules have to
include information about the head assigning the “outer” case – otherwise there would be no
way to account for the different outputs of the [acc][gen] input for the Genitive of negation and
for the Genitive-plus-Infinitive construction. This can be implemented in different ways, see
Erschler (2009) and Matushansky (2010) for fairly different proposals.
(84a) and (84b) differing in word order only; in (84b) the displaced object is left-
adjoined to the infinitival TP rather than vP, which seems to better capture the
adverbial position facts established by FL06.33
(84) a. NP daržinė sukrauti šienui ‘hayloft to keep hay’

3
NP CP
Dative domain
hayloft 3
C[purp] → TP-Dat
3
T[inf] vP-Dat
to-keepk 3
NP v′
pro 3 Accusative domain
vk → VP-Acc-Dat
3
Vk NP
hayacc;dat
b. NP daržinė šienui sukrauti ‘id’.
3
NP CP Dative domain
hayloft 3
C[purp] → TP-Dat
ei
NPi TP-Dat
hayacc;dat ei
T[inf] vP-Dat
to-keepm 3
NP v′
vm → VP-Acc-Dat
3
Vm NPi
33. Note that I remain agnostic as to what in particular triggers the preferred movement of the
object in the Dative-plus-Infinitive construction to its left edge position. This could be e.g. con-
struction-specific or information structure properties of the C head. What is crucial is that this
movement is not related to case assignment in any way and occurs after case is assigned – just
as most other instances of overt NP movement attested in Lithuanian are associated with infor-
mation structure or weight rather than with case.
84 Peter Arkadiev
Infinitive constructions with overt Dative subjects, being not limited to pur-
pose Infinitives, suggest a different analysis, i.e., the one where the subject NP in
Spec,vP gets associating Dative assigned to the vP by the T[inf] head. If an Infini-
tive clause with an overt subject gets embedded under C[purp], its direct object can
potentially also receive the complementizing Dative from the latter, but such a
surface structure is often ruled out by the constraint against two Datives in one
TP (83vi), so the “inner” Accusative” is realized instead of the “outer” Dative, cf.
the diagram (85).
(85) NP pinigai sūnui pastatyti namą / %namui

3 ‘money for the son to buy a house’
NP CP
C-Dative domain
money 3
C[purp] → TP-Dat
3
NPi T′
A-Dative domain
sondat 3
T[inf] → vP-Dat-Dat
to-buildk 3
NPi v′
3 Accusative domain

*double dative → vk → VP-Acc-Dat-Dat
3
Vk NP
houseacc;dat
The ban on the Accusative-to-Dative conversion in the presence of a Dative indi-

rect object is accounted for in the same way, cf. (86).
(86) NP pinigai nupirkti motinai vazą / *vazai

3 ‘money to buy a vase for mother’
NP CP
C-Dative domain
money 3
C[purp] → TP-Dat
3
T[inf] vP-Dat
to-buyi 3
NP v′
vi → VP-Acc-Dat
ei
NP V′
vaseacc;dat 3 relational Dative domain
Vi → NP-Dat-Acc-Dat
*double dative →
mother dat;acc;dat
It is important to note that the “double-dative” constraint is violable, since, first,

not all speakers reject examples like (50a) (=85), and, second, not all structures
with more than one Dative NP can be thus ruled out, but only those where there is
an alternative variant of case-marking (i.e., Accusative). Cf. a perfectly felicitous
modal Infinitive clause (87) with a Dative subject and a Dative indirect object.
Such violability can in principle be handled by OT-style constraint interaction,
cf. Erschler (2009).
(87) a. Kaip [mums padė-ti j-am pripras-ti prie nauj-ų
how we.dat help-inf 3-dat.sg.m get.used-inf at new-gen.pl
nam-ų]?
house-gen.pl
‘How can we help him to get accustomed to the new home?’34
b. TP
associating Dative domain
ei
T[inf] → vP-Dat
to-helpi 3
NP v′
wedat 3
vi VP-Dat
3 relational Dative domain
Vi → NP-Dat-Dat
*double dative
hedat
34. http://www.paukstis.lt/forumas/viewtopic.php?f=23&t=6006&start=810
86 Peter Arkadiev
Turning to the Genitive-plus-Infinitive construction, I propose to analyze it sim-

ilarly to the Kayardild Supine shown in (73), i.e. as an associating case assigned
by the special variety of non-finite T.35 This functional head T[sup] is constrained
to co-occur with verbs expressing motion and is semantically interpreted as de-
noting an event serving as the goal or purpose of the matrix motion event. I as-
sume that such a functional head is available in all languages where non-finite
clauses denoting goal or purpose of motion display specific morphology or syntax
(or both). Thus, of the languages discussed in this paper, the following possess
a distinct T[sup] with differing morphological and/or case assigning properties,
cf. Table 2.
The only difference between the Supine constructions in Latgalian and ear-
lier/dialectal Lithuanian, on the one hand, and in standard Lithuanian, on the
other, is that in the latter the morphological realizations of T[sup] and T[inf] are
identical, though the functional heads differ in their interpretation, co-occur-
rence possibilities and case assigning properties. The structure of the Lithuanian
Genitive-plus-Infinitive construction is thus as shown in (88).
Table 2. Varieties of the Supine

Language Morphology Case assignment
Latgalian ≠ Inf Genitive
Lithuanian = Inf Genitive
Slovene ≠ Inf Accusative
Kayardild -jiring Allative
35. An alternative solution would be to follow the lines of the Dative-plus-Infinitive construc-
tion and postulate a special C head constrained to co-occur with verbs of motion and select-
ing the Infinitive T in Lithuanian and a Supine T in Latgalian; this C head would then assign
the complementizing Genitive. In my view, such an analysis is unnecesarily complicated and
should be rejected for reasons of economy, unless it turns out that it is supported by empirical
data. I thank David Erschler for drawing my attention to this possibility.
(88) TP Jonas atėjo aplankyti draugo

3 ‘Jonas came to visit a friend’
NPi T′
Jonasnom ru
T[pst] VP
camek ro
VP TP
ty ei Genitive domain
NPi Vk T[sup] → vP-Gen

to-visitm 3
NPi v′
Accusative domain
pro 3
vm → VP-Acc-Gen
3
Vm NP
friendacc;gen
The only question which still remains unresolved concerns the relationship be-
tween the Dative-plus-Infinitive and Genitive-plus-Infinitive constructions and
the corresponding “independent” cases appearing on the NPs denoting purpose
and goal, as in examples (21) and (22), repeated here as (89) and (90). Of course,
it would be desirable to have a common account of these clearly related uses of
both Dative and Genitive.
(89) a. Čia bu-s lentyn-a knyg-oms.
here be-fut(3) shelf-nom.sg book-dat.pl
‘Here will be a shelf for books.’ (Kerevičienė 2008: 182)
b. Žem-ė keli-a-s darb-ui ir kūryb-ai.
earth-nom.sg get.up-prs(3)-rfl work-dat.sg and creation-dat.sg
‘Earth is getting up for work and creation.’ (Kerevičienė 2008: 182)
(90) a. Išėj-o pien-o.
go.out-pst(3) milk-gen.sg
‘(He/she) went for milk.’ (Ambrazas (ed.) 1997: 557)
b. Išsiunt-ė sūn-ų daktar-o.
send-pst(3) son-acc.sg doctor-gen.sg
‘(He/she) sent the son to get the doctor.’ (Ambrazas (ed.) 1997: 557)
In the analysis of the purposive Dative in (89) I will follow FL06 (p. 274), assum-
ing that the Dative case on both independent purposive Dative NPs and direct
objects of purpose Infinitives has the same source, viz. the null purposive C. That
a C head can combine both with clauses and NPs is no surprise, cf. English for
88 Peter Arkadiev
and numerous instances in many other languages. Thus the structure of (89b) is
as shown in (91).

(91) TP
ei
NPi T′
earthnom ei
T[prs] VP
get.upk tp
VP CP
2 3
NPi Vk C[purp] → NP-Dat
work and creationdat
My proposal concerning the purposive Genitive with verbs of motion, however,

diverges from that of FL06. The latter postulate a common source for both the
independent Genitive and the Genitive on the object of the Infinitive, viz. the
head Asp associated with verbs of motion. Since I dispense with such an aspectual
head, I see two alternative ways of analyzing the independent Genitive, both of
which find some cross-linguistic support. The first possible analysis rests on the
default interpretation of the independent Genitive with verbs of motion, i.e. that
the motion event occurs in order to get or obtain the referent of the Genitive NP.36
This kind of semantics can be naturally captured by assuming that the indepen-
dent Genitive has the underlying structure of the Supine (= Genitive-plus-Infini-
tive) construction with a silent verbal head interpreted as ‘get’, ‘obtain’ or ‘bring
back’, cf. the structure of (90b) in (92).
36. Cf. the analysis of intensional transitive verbs by den Dikken et al. (1996) arguing for a
structure with a “concealed” complement clause. I thank Ora Matushansky for drawing my at-
tention to the relevance of this parallel.
(92) TP
ei
T[pst] vP
sentk 3
vk → VP-Acc
tp
VP-Acc TP
2 3
NP Vk T[sup] → vP-Gen
sonacc 3
v → VP-Acc-Gen
3
V NP
∅get doctoracc;gen
This kind of analysis can be indirectly supported by the fact that there are lan-
guages where case markers encoding the object serving as the purpose of motion
historically go back to verbal stems meaning ‘get’ or ‘bring’. One such language
is Adyghe (North-West Caucasian), cf. (93a) with the lexical fully inflected verb
‘bring’ and (93b) with its stem (with a vowel alternation) attached to the purpo-
sive adjunct and serving as a case marker.
Adyghe (North West Caucasian, own fieldwork)
(93) a. pŝaŝe-m psә q-ә-hә-ʁ.
girl-obl water dir-3sg.a-bring-pst
‘The girl brought (some) water.’
b. pŝaŝe-m gʷegʷenә-r ә-št-jә psә-he ḳʷa-ʁe.
girl-obl pitcher-abs 3sg.a-take-and water-bring go-pst
‘The girl took the pitcher and went to fetch water.’
However, the analysis of the independent purposive Genitive as a Supine con-

struction with an elided or null verb is not the only possibility. It might be the
case that the analogy between the Genitive-plus-Infinitive construction and the
independent Genitive of purpose with verbs of motion is just accidental, and that
the independent Genitive is just a semantic case appearing on a particular kind of
adjuncts, possibly assigned by some kind of null P(reposition). Note that it is not a
problem for such an analysis that purpose NPs with verbs of motion receive a dif-
ferent encoding from other types of purpose NPs, since this is what often happens
in languages, cf. Russian in (94) where two different overt prepositions assigning
different cases are employed.
90 Peter Arkadiev
Russian (personal knowledge)

(94) a. Zdes’ bud-et polka dlja knig.
here be-fut.3sg shelf for1 book(gen.pl)
‘Here will be a shelf for books.’
b. Devušk-a poš-l-a za vod-oj.
girl-nom.sg go-pst-sg.f for2 water-ins.sg
‘The girl went to fetch water.’
Though for reasons of symmetry the analysis with the reduced Supine construc-
tion seems to me to be more attractive for the independent Genitive denoting
purpose of motion, I prefer to leave this issue unresolved.
Thus, from the analysis of the Lithuanian constructions in this section and
its comparison to the Australian data presented in §5 it can be inferred that case
marking in individual languages boils down to the interaction of the universal
mechanism of case assignment by a head to its complement and such language-
specific issues as (i) the case inventory, (ii) the case-assignment specifications of
particular lexical and functional heads (note that (i) and (ii) probably are just two
facets of the same phenomenon), and (iii) the (morphological) rules of case-reso-
lution. Note that the issue of locality crucially invoked in much of the current the-
orizing about case and in particular in FL06’s analysis of Lithuanian plays no role
in the proposed conception of case assignment – though it might turn out that it
is still relevant for some (but definitely not all) phenomena associated with case.
7. Conclusions and implications
The Lithuanian Dative-plus-Infinitive and Genitive-plus-Infinitive constructions

pose very peculiar problems for a formal analysis, and, indeed, for an analysis
in any theoretical framework. The reason for this is a complex and cross-lin-
guistically not widespread interplay between case-marking and non-finite sub-
ordination, which is an outstanding characteristic of Lithuanian in general (see
Arkadiev 2013a, 2013b).
In this paper I presented empirical evidence arguing for a revision of the only
analysis of these constructions so far proposed in the literature (Franks & Lavine
2006), and have drawn my own analysis upon “very exotic” typological parallels
from Australian languages with “complementizing” and “associating” case and
overt multiple case marking, which, in my view, help us better understand the
nature of the actually no less “exotic” Lithuanian constructions (cf. Erschler 2009
and Matushansky 2008, 2010 for similar proposals concerning completely differ-
ent data).
The analysis presented above has some immediate consequences for the for-
mal theory of case (cf. also Merchant 2006). Metaphorically speaking, instead of
assuming that “all languages are like English”, the belief which has guided much of
the Government and Binding theory of “abstract case”, I propose to assume that in
fact “all languages are like Kayardild”. In particular, this means that:
1. NPs may receive case from many (potentially all) functional as well as lexical
heads which c-command them; this is implemented by a very general concep-
tion of case assignment, viz. by default a head assigns case to its complement
as a whole, and this case feature subsequently percolates down to (potentially
all) subconstituents of that complement.
2. Morphological realization of these multiple cases assigned in syntax is subject
to language-particular rules and constraints, which do not belong to “narrow
syntax”; to put it more accurately, languages may vary as to whether they have
any syntactic constraints on case percolation (e.g., it is obvious that in Lithu-
anian and many other languages finite T as well as overt C and (some) overt
Ps block case percolation) and in the nature of morphological or morphosyn-
tactic case resolution rules:
2a. some languages, like Kayardild or Nyamal, allow simultaneous morpho-
logical realization of several layers of case on a nominal; this is the stron-
gest empirical evidence for the syntactic mechanism of multiple case
assignment;
2b. other languages (arguably the majority) do not allow overt multiple case
marking in morphology, but in some (and probably many) of them the
mechanism of syntactic multiple case assignment reveals itself in alterna-
tions of case marking like the ones discussed in this paper.
The outlined conception of case assignment, largely inspired by Matushansky

(2008, 2010) but not following the latter in all details, has some implications for
the architecture of grammar in Minimalist case theory. While in the “classic”
case theory (e.g., Chomsky 1981: 162–176; Stowell 1981: 110–125; see Bobaljik
& Wurmbrand 2009 for a review) case assignment is a local operation, some-
times assumed to be just a reflection of Agree ultimately constrained by the Phase
theory of the Minimalist program (Chomsky 2001: 6ff.), the current analysis and
the data supporting it, by contrast, imply a non-local view of case. Since case is
assigned by a head to its complement and percolates down to all subconstituents
of the latter, case assignment per se is still a strictly local (head-complement)
operation, but case percolation is unbounded and in particular pays no attention
to (at least some) phase boundaries. A possible way to reconcile the novel view
of case and the independently motivated Phase theory is to exclude case percola-
tion from “narrow syntax” and to transfer it to PF, where case realization belongs,
92 Peter Arkadiev
anyway. This move, however, necessarily requires that PF-spellout occur not as
soon as each phase is constructed, but only after the whole derivation in “narrow
syntax” is completed (cf. Richards 2013).
Setting aside potential far-reaching implications of the analysis presented in
this paper, I would like to conclude by saying that I hope to have shown that, first,
typological comparison between languages apparently having as little to do with
each other as Lithuanian and the Australian languages can elucidate the phenom-
ena attested in both of them, and, second, that an adequate (formal or informal)
theory of case and its relations to such phenomena as non-finiteness and subor-
dination has to take into account a broad range of empirical data from all kinds
of languages, including such more or less “exotic” ones as Baltic or Australian.
The last point may seem trivial, but for the fact that Lithuanian data have so far
only rarely figured in any kind of theorizing about case, and almost no attempts
have been made to find a really adequate place for these data in the typology and
theory of case. I hope that this paper has served to partly fill this gap.
Abbreviations
a agent loc locative

abl ablative m masculine
abs absolutive m:abl modal ablative
acc accusative m:all modal allative
all allative m:prop modal proprietive
ant anticipatory mood neg negation
aux auxiliary nml nominalization
c:obl complementizing oblique case nom nominative
dat dative obl oblique case
def definiteness pa active participle
dir directional pl plural
do direct object pot potential
ds different subject pp passive participle
du dual prf perfect
elat elative prs present
erg ergative pst past
f feminine ptcl particle
fin finite purp purposive
fut future rel relativization
gen genitive rfl reflexive
imp imperative sbj subject
inf infinitive sg singular
ins instrumental sup supine
kin kinship possessive tr transitive
References
Ambrazas, Vytautas. 1981. Zur Geschichte einer indogermanischen Konstruktion (Dativus

cum infinitivo im Baltischen). Kalbotyra 32(3): 12–24.
Ambrazas, Vytautas. 1987. Die indogermanische Grundlage des Dativus und Nominativus
cum infinitivo im Baltischen. Indogermanische Forschungen 92: 203–219.
Ambrazas, Vytautas (ed). 1997. Lithuanian Grammar. Vilnius: Baltos Lankos.
Anderson, Cori. 2011. Case theory and case alternations: Evidence from Lithuanian. Baltic
Linguistics 2: 9–35.
Anderson, Cori. Forthcoming. Non-canonical case patterns in Lithuanian. In Contemporary
Approaches to Baltic Linguistics, Peter Arkadiev, Axel Holvoet, & Björn Wiemer (eds).
Berlin & Boston: De Gruyter Mouton.
Arkadiev, Peter [Arkad’ev P. M.] 2013a. O tipologičeskom svoeobrazii litovskoj glagol’noj siste-
my [On the typological peculiarity of the Lithuanian verbal system]. In Issledovanija po
tipologii slavjanskix, baltijskix i balkanskix jazykov [Studies in the typology of Slavic, Baltic
and Balkan Languages], Vjačeslav V. Ivanov & Petr M. Arkad’ev (eds), 320–361. St Peters-
burg: Aleteia.
Arkadiev, Peter. 2013b. Marking of subjects and objects in Lithuanian non-finite clauses: A
typological and diachronic perspective. Linguistic Typology 17(3): 397–437.
Austin, Peter. 2009. Issues in the synchrony and diachrony of case marking in the Manthar-
ta and Kanyara languages, Australia. Handout of talk at Case in and Across Languages,
Helsinki.
Babby, Leonard H. 1986. The locus of case assignment and the direction of percolation: Case
theory and Russian. In Case in Slavic, Richard Brecht & James Levine (eds), 170–219.
Columbus, OH: Slavica.
Babby, Leonard H. 1994. Case theory. In Noam Chomsky: Critical assessments. Vol. I: Linguis-
tics: Tome II, Carlos P. Otero (ed), 630–652. London, New York: Routledge.
Béjar, Susanne & Massam, Diane. 1999. Multiple case checking. Syntax 2(2): 65–79. DOI:
10.1111/1467-9612.00015
Bobaljik, Jonathan D. & Wurmbrand, Susi. 2009. Case in GB/Minimalism. In The Oxford Hand-
book of Case, Andrej Malchukov & Andrew Spencer (eds), 44–58. Oxford: Oxford Univer-
sity Press.
Brezar, Mojca Schlamberger, Pogacnik, Vladimir, & Perko, Gregor. 2005. Parlons slovène. Paris:
L’Harmattan.
Chomsky, Noam. 1981. Lectures on Government and Binding: The Pisa Lectures. Dordrecht:
Foris.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on
Minimalist Syntax in Honor of Howard Lasnik, Roger Martin, David Michaels, & Juan
Uriagereka (eds), 89–115. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael
Kenstowicz (ed), 1–73. Cambridge, MA: MIT Press.
den Dikken, Marcel, Larson, Richard K., & Ludlow, Peter. 1996. Intensional ‘transitive’ verbs
and concealed complement clauses. Revista di linguistica 8: 29–46.
Dench, Alan. 2009. Case in an Australian language. Distribution of case and multiple case
marking in Nyamal. In The Oxford Handbook of Case, Andrej Malchukov & Andrew
Spencer (eds), 756–769. Oxford: Oxford University Press.
94 Peter Arkadiev
Dench, Alan & Evans, Nicholas. 1988. Multiple case-marking in Australian languages. Austra-
lian Journal of Linguistics 8(1): 1–47. DOI: 10.1080/07268608808599390
Erschler, David. 2009. On case conflicts in Russian: An Optimality-Theoretic approach. In
Studies in Formal Slavic Phonology, Morphology, Syntax, Semantics and Information Struc-
ture. Proceedings of FDSL 7, Leipzig 2007, Gerhild Zybatow, Uwe Junghanns, Denisa Len-
ertova, & Petr Biskup (eds), 119–131. Frankfurt am Main: Peter Lang.
Evans, Nicholas. 1995a. A Grammar of Kayardild. Berlin & New York: Mouton de Gruyter.
DOI: 10.1515/9783110873733
Evans, Nicholas. 1995b. Multiple case in Kayardild: Anti-iconic suffix ordering and the dia-
chronic filter. In Double Case. Agreement by Suffixaufnahme, Frans Plank (ed), 396–428.
Oxford: Oxford University Press.
Franks, Steven & Lavine, James E. 2006. Case and word order in Lithuanian. Journal of Linguis-
tics 42(1): 239–288. DOI: 10.1017/S0022226706003896
Hornstein, Norbert, Nuñes, Jairo & Grohmann, Kleanthes K. 2005. Understanding Minimalism.
Cambridge: Cambridge University Press. DOI: 10.1017/CBO9780511840678
Ivanov, V. V. 1990. Istoričeskaja grammatika russkogo jazyka. [A Historical Grammar of Rus-
sian] 3rd ed. Moscow: Prosveščenie.
Keine, Stefan. 2010. Case and Agreement from Fringe to Core. A Minimalist Approach. Berlin &
New York: Mouton de Gruyter. DOI: 10.1515/9783110234404
Kerevičienė, Jurgita. 2008. The Lithuanian Dative and its English Counterparts (A Case Study
in Cognitive Grammar). Doctoral Dissertation, Vilnius University.
Landau, Idan. 2006. Severing the distribution of pro from case. Syntax 9(2): 153–170. DOI:
10.1111/j.1467-9612.2006.00087.x
Lunt, Horace G. 2001. Old Church Slavonic Grammar. 7th edn. rev. Berlin & New York: Mouton
de Gruyter. DOI: 10.1515/9783110876888
Matushansky, Ora. 2008. A case study of predication. In Studies in Formal Slavic Linguistics.
Contributions from Formal Description of Slavic Languages 6.5, Frank Marušič & Rok Žau-
cer (eds), 213–239. Frankfurt am Main: Peter Lang.
Matushansky, Ora. 2010. Russian predicate case, encore. In Formal Studies in Slavic Linguis-
tics, Proceedings of FDSL 7.5, Gerhild Zybatow, Philip Dudchuk, Serge Minor, & Ekaterina
Pshehotskaya (eds), 117–135. Frankfurt am Main: Peter Lang.
McCreight Young, Katherine. 1988. Multiple Case Assignments. PhD Dissertation, MIT.
McFadden, Thomas. 2004. The Position of Morphological Case in the Derivation: A Study on
the Syntax-Morphology Interface. PhD Dissertation, University of Pennsylvania. ‹http://
www.ling.upenn.edu/~tmcfadde/diss.html›
McFadden, Thomas. 2010. Structural case, locality and cyclicity. In Explorations of Phase The-
ory: Features and Arguments, Kleanthes K. Grohmann (ed), 107–130. Berlin: Mouton de
Gruyter.
Merchant, Jason. 2006. Polyvalent case, geometric hierarchies, and split ergativity. In Proceed-
ings of the 42nd Annual Meeting of the Chicago Linguistic Society, Jacqueline Bunting (ed),
47–67. Chicago, IL: Chicago Linguistic Society.
Nau, Nicole. 2011. A Short Grammar of Latgalian. München & Newcastle: LINCOM.
Preminger, Omer. 2011. Agreement as a Fallible Operation. Doctoral dissertation, MIT. <http://
dspace.mit.edu/handle/1721.1/68196>
Richards, Norvin. 2007. Lardil “case stacking” and the structural/inherent case distinction. Ms.
<http://ling.auf.net/lingbuzz/000405>
Richards, Norvin. 2013. Lardil “case stacking” and the timing of case assignment. Syntax 61(1):
42–76. DOI: 10.1111/j.1467-9612.2012.00169.x
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar, Liliane
Haegeman (ed), 281–337. Dordrecht: Kluwer. DOI: 10.1007/978-94-011-5420-8_7
Round, Erich R. 2013. Kayardild Morphology and Syntax. Oxford: Oxford University Press.
Sawicki, Lea. 1992. Genitive and dative in goal and purpose expressions in contemporary Lith-
uanian. Linguistica Baltica 1: 93–101.
Schmalstieg, William R. 1987. A Lithuanian Historical Syntax. Columbus, OH: Slavica.
Schmidtke-Bode, Karsten. 2009. A Typology of Purpose Clauses. Amsterdam/Philadelphia: John
Benjamins. DOI: 10.1075/tsl.88
Steenwijk, Han. 2003. The use of the supine in Lower Sorbian. In Dutch Contributions to the
Thirteenth International Congress of Slavists. Linguistics [Studies in Slavic and General Lin-
guistics 30], Jos Schaeken, Peter Houtzagers, & Janneke Kalsbeek (eds), 331–342. Amster-
dam: Rodopi.
Stowell, Tim. 1981. Origins of Phrase Structure. PhD Dissertation, MIT.
Szabolcsi, Anna. 2006. Strong vs. weak islands. In The Blackwell Companion to Syntax Vol. 4,
Martin Everaert & Henk van Riemsdijk (eds), 479–532. Oxford: Blackwell.
Vaillant, André. 1966. Grammaire comparée des langues slaves. Tome III. Le verbe. Paris:
Klincksieck.
Vaillant, André. 1977. Grammaire comparée des langues slaves. Tome V. La syntaxe. Paris:
Klincksieck.
Woolford, Ellen. 2006. Lexical case, inherent case, and argument structure. Linguistic Inquiry
37(1): 111–130. DOI: 10.1162/002438906775321175
Zinkevičius, Zigmas. 1966. Lietuvių dialektologija. Lyginamoji tarmių fonetika ir morfologija
[Lithuanian Dialectology: Comparative Phonetics and Morphology of the Dialects].
Vilnius: Mintis.
Non-canonical grammatical
relations in a modal construction
The Latvian debitive
Axel Holvoet and Marta Grzybowska

University of Warsaw / Vilnius University
The aim of the paper is to account for the pattern of grammatical relations
with the debitive, an inflectional form of the Latvian verb expressing necessity.
The authors argue that the debitive construction displays what they call diffuse
grammatical relations. They show that neither of the arguments in the debitive
construction shows clear subject or object properties, and introduce the notion
of a demoted intransitive subject occurring alongside a less-oblique datival ar-
gument, in a configuration reminiscent of ditransitive structures with demoted
direct objects. Such patterns with diffuse grammatical relations are also assigned
lexically, not only in Latvian but also in Lithuanian, many Slavonic languages
and some Germanic languages such as Icelandic; in the case of the debitive, the
diffuseness of grammatical relations is associated with the modal character of
the construction, a factor known to induce non-canonical argument marking.
The authors also dwell on the diachronic implications of the issues raised in the
article, introducing the notion of obliqueness adjustments, i.e., changes in the
case-marking of arguments bringing it in line with syntactic obliqueness.
1. Introductory remarks1
In this article we will be dealing with what has been referred to as non-canoni-
cally marked grammatical relations (cf. Aikhenvald, Dixon & Onishi (eds), 2001).
We will be focusing on Latvian constructions containing the debitive, an inflec-
tional form of the verb expressing necessity. The non-canonical marking pattern
with this form seems to be connected with its modal function. After explaining
the formal properties of the debitive, we will examine grammatical relations in
1. Our thanks are due to Wayles Browne, Nicole Nau, Natalia Perkova, and an external re-
viewer for many useful remarks. For all remaining mistakes we remain solely responsible.
98 Axel Holvoet and Marta Grzybowska
constructions with this form, arguing that this is an instance of diffuse gram-
matical relations, without clear concentration of subject properties in any NP. We
will then place the debitive in the context of other, lexically assigned valency pat-
terns in Latvian and elsewhere and we will argue for the notion of a demoted
intransitive subject occurring alongside a less-oblique dative marked experiencer
argument in a configuration reminiscent of the oscillations in the assignment of
grammatical relations in ditransitive structures. We will discuss both synchronic
and diachronic implications of such structures with diffuse grammatical relations.
Basing ourselves on the comparison with lexically assigned patterns having dif-
fuse grammatical relations, we will attempt to reassess the debitive in its capacity
as a grammatical construction. Finally, we will briefly assess the evidence of the
debitive in the context of valency splits, and we will conclude with a few remarks
on the notion of non-canonicity as applied to grammatical relations.
2. The debitive: A modal form with atypical syntactic properties
The debitive is a form of the Latvian verb expressing necessity. In typology, the
more common term for debitive is necessitive. Necessitive meaning tends to be
expressed by affixes occupying a separate slot distinct from that of mood, so that
a necessitive form can itself be conjugated for mood. For this reason De Haan
(2006) proposes the term ‘modal affix’ in order to set formations expressing ne-
cessity and possibility apart from mood, even though in the traditional grammars
of many languages they are described as moods. This is shown by the Hungarian
potential (possibility) form in (1), also marked for conditional (irrealis) mood
(Rounds 2001: 59):
(1) Mond-hat-nék valami-t?Hungarian
say-pot-irr.1sg something-acc
‘Might I say something?’
The Latvian debitive contains a modal prefix jā-; it is prefixed to a stem form coin-
ciding with the 3rd person present indicative, except in the case of būt ‘be’, where
it is added to the infinitive, yielding jābūt. The resulting form is uninflected, and
tense and mood are expressed by different forms of the auxiliary būt ‘be’, which is
usually omitted in the present indicative but surfaces in other forms. Compare:
(2) Man (ir) jā-gaida šeit.Latvian
1sg.dat be.prs.3 deb-wait here
‘I must wait here.’
Non-canonical grammatical relations in a modal construction 99
(3) Tev bū-tu ilgi jā-gaida.Latvian

2sg.dat be-irr long deb-wait
‘You would have to wait a long time.’
The debitive does not inflect for person, but the ‘target’ of the modal predicate, i.e.,
the person on whom a necessity is imposed (we borrow this term from Narrog
2010), is expressed by the dative. The debitive can thus be compared to a series of
‘impersonal’ modal verbs and predicators characteristic of both Slavonic and Bal-
tic, such as Latvian vajag ‘one must, should’, Lithuanian reikia ‘one must, should’,
Russian možno ‘one can, may’, Polish wolno ‘one may, is allowed’ etc., which also
combine with modal targets in the dative. When this dative NP is lacking, the
clause refers to an unspecified human modal target. Compare:
(4) Man tur jā-iet. Latvian
1sg.dat there deb-go
‘I must go there.’
(5) Man vajag tur iet. Latvian
1sg.dat be_needed.prs.3 there go.inf
(same meaning)
In the standard language, the original object (the noun marked as object in the
verb’s basic, lexically assigned valency pattern) appears in the nominative, cf. (6),
unless it is a 1st or 2nd person pronoun or a reflexive pronoun; these are (with rare
exceptions for 1st and 2nd person pronouns mentioned by Endzelin 1923: 753) in
the accusative, cf. (7).
(6) Man jā-dzer ūden-s.
1sg.dat deb-drink water-nom
‘I must drink water.’
(7) Man tevi jā-brīdina.
1sg.dat 2sg.acc deb-warn
‘I must warn you.’
In many Latvian dialects, however, and in colloquial and substandard Latvian,

the accusative is also used for nouns and 3rd person pronouns. Apart from the
fact that prescriptive grammarians warn against this use (starting with Endzelīns
1932: 43), no data on the actual distribution of case forms or the conditions fa-
vouring the use of nominative or accusative are available. We will venture a few
suggestions on this but corpus data are needed if more reliable data are to be
gained.
We will return to the morphosyntactic properties of the debitive construc-

tion in greater detail below, as this is closely connected with the main topic of
this article; the general remarks given at this stage are meant only to facilitate the
interpretation of the example sentences.
There are no restrictions as to the type of necessity expressed by the debitive:
it can be used both in the domain of event (root) and in that of propositional
(epistemic) modality (these terms are taken from Palmer 2001). We will illustrate
this with two examples: (8) is deontic, whereas (9) is epistemic:
(8) Jā-kaunas gan bū-tu jums, ka tik vieglprātīgi
deb-be_ashamed prtc be-irr 2pl.dat that so thoughtlessly
esat ielīd-is lamat-ās.
be.prs.2pl run_into-ppa.nom.sg.m trap-loc.pl
‘It’s you who should be ashamed of having run into the trap so thoughtlessly.’
(M. Zīverts)
(9) Nikolaj-s izgāja, bet saimniek-s nodomāja, ka
Nikolaj-nom go_out.pst.3 but farmer-nom reflect.pst.3 comp
Nikolaj-am jā-būt kād-am sektant-am.
Nikolaj-dat deb-be some-dat.sg.m sectarian-dat.sg
‘After Nikolaj had left, the farmer reflected that Nikolaj must be some kind of
sectarian.’ (Ā. Erss)
Historically, the debitive derives (as first pointed out by Prellwitz 1904) from a
bi-clausal structure comprising a possessive construction of the mihi est type and
an infinitival relative purpose clause whose relative pronoun (based on the Indo-
European relative pronoun *ye/o-) accreted to the verb. The original meaning is
partly retained (alongside the new modal meaning) in Old Latvian:
(10) Man ir Barriba jaehd, ko juhs
1sg.dat be.prs.3 food.nom.sg deb.eat rel.acc 2pl.nom
ne sinnat.
neg know.prs.2pl
‘I have meat (i.e., food) to eat that ye know not of.’
(Glück’s New Testament, 1685, John 4.32)
Subsequently a shift from, say, ‘I have some bread to eat’ to ‘I have to eat bread’ oc-
curred. On the details of the grammaticalization of the debitive as a modal form
cf. Holvoet (1998).
Though, as mentioned above, the debitive cannot properly be called a mood,
presumably it is an inflectional rather than a derivational form of the verb.
There is no change in category (the debitive is a verbal form), the debitive is de-
rived with absolute regularity from just any verb, and its meaning is completely
predictable. Moreover, derivation doesn’t seem to be an option here because mo-

dality is not very close to lexical meaning: whereas valency-changing formations
(causatives and anticausatives) have a fair chance of being described as deriva-
tional, one would hesitate to extend such an interpretation to a formation ex-
pressing a modal predication. True, as can be seen from examples (4)–(7), case
marking with the debitive differs from what it is with the basic forms of the verb,
and possibly (though this remains to be investigated), this different case marking
reflects a different assignment of grammatical relations. So perhaps the debitive
is a valency-changing device after all? Bielenstein (1864: 211) had certainly de-
scribed the debitive as a kind of passive, and Nau (1998: 39) also describes the
debitive as a voice, albeit distinct from the passive. Voice is not typically deri-
vational, of course, and if one describes the debitive as inflectional but distinct
from mood, voice is obviously an option. But we know from typology that not
every inflectional form changing case-marking and/or grammatical relations is
voice: many cases have been reported of what is now called ‘splits’, the changes
in case marking and grammatical relations associated with tense and aspect dif-
ferences in Georgian (Boeder 1979) being a classical example. We will return to
this further on; for the time being, suffice it to say that we prefer not to describe
the debitive as a value of the category of voice because its function is modal and
there are no grounds to assume that it has ever been anything else but modal.
While Narrog (2010) convincingly argues for a typologically well-documented
connection between event modality and non-canonical argument marking, we
are not committed to classifying all these instances of non-canonical marking
under voice. Within a traditional (Jakobsonian) view of the verbal system as a
system of correlations we are faced with the dilemma of classifying the debitive
with either mood or voice, as it cannnot be outside the system of correlations; but
the Latvian debitive evidently does not fit into such a view of verbal categories,
and here we would be better served with a notion of ‘gram’ (as used in Bybee et al.
1994) rather than with traditional grammatical categories.
3. Monoclausal and biclausal debitives
It follows from the above that the construction with the debitive is used where the
Standard Average European languages would use modal verbs. Though modal
verbs are sometimes thought of as auxiliaries (especially in the English-grammar
tradition, cf. Lightfoot 1974), most authors describe them as complement-taking
predicates (cf. Noonan 2007); in the latter case, we must describe the sentence as
bi-clausal (a raising or control construction). But as the Latvian debitive marker
fuses with the verb form, the result is apparently a monoclausal structure; at least
evidence for a biclausal structure is lacking. There are, however, exceptions.
There is a subtype of the debitive in which this form is derived from a com-
pound verb stem consisting of the auxiliary būt and the past active participle – the
stem of the compound (perfect) tenses of the Latvian verb. The meaning is then
‘(one) must have seen’ etc.
(11) Vispārējs princips saka, ka
tulk-am ir savlaicīgi jā-būt redzēj-uš-am
interpreter-dat.sg be.prs.3 previously deb-be see-ppa-dat.sg.m
katr-u rakstī-t-o run-u,
every-acc.sg write-ppp-acc.sg.def speech-acc.sg
kura tiks nolasīta pasākumā.
‘The general principle says that an interpreter must have previously seen every
written speech that is to be delivered at an event.’
http://www.tulkojumi.de/par_tulkosanu.php
(12) Darba nav daudz, man liekas, superīga haltūra, tur
paš-us seriāl-us nemaz nav obligāti
self-acc.pl.m serial-acc.pl not_at_all be.prs.3.neg obligatorily
jā-būt redzēj-uš-am.
deb-be see-ppa-dat.sg.m
‘There is not much work, seems to me it’s a super job, you’re not at all obliged
to have seen all these serials yourself.’
http://klab.lv/community/darbs/33080.html?mode=reply
In such constructions the nominative as a case for marking the original object is
rejected by native speakers. This is evidently connected with the fact that the verb
governing this complement is not fused with the debitive marker but remains
intact, thereby retaining its usual accusative marking. It is only when the markers
of the modal and the embedded predicate are fused into one predicative form that
this form is able to assign nominative.
Debitive constructions of this type are structurally quite similar to structures
with impersonal modal verbs, like vajag ‘it is necessary’. Compare:
(13) Tev to film-u jā-būt redzēj-uš-am.
2sg.dat this.acc.sg film-acc.sg deb-be see-ppa-dat.sg.m
(14) Tev to film-u vajag būt
2.sg.dat this.acc.sg film-acc.sg be_needed.prs.3 be.inf
redzēj-uš-am.
see-ppa-dat.sg.m
‘You must have seen this movie.’
But for the fusion of the modal predicator and the infinitive, the two structures
are exactly parallel, and as we would describe (14) as a bi-clausal control struc-
ture, we will describe the construction with the debitive in (13) in a similar fash-
ion. We can here differentiate two clausal structures: the dative argument belongs
to the matrix clause (and controls the PRO-argument of the embedded clause),
whereas the accusative object belongs to the embedded clause. The specific fea-
ture of this variety of the debitive construction consists in that the auxiliary of the
compound verb form, let’s call it the Anteriority (Ant) head, fuses with the Modal
head (Mod), as shown in diagram (15):
(15) S
ei
NP AuxP
ei
Aux ModP
ei
Mod S
ei
NP AntP
ei
Ant VP
ep
V NP
5
tev (ir) jā-būt PRO redzējušam šo filmu
2sg.datbe.prs.3 deb-be see.ppa.dat.sg.m dem film.acc.sg
‘You must have seen this film’
We could correspondingly refer to this variety of the debitive as the ‘bi-clausal

debitive construction’. In what follows it will not further interest us, because the
main emphasis will be on those cases where matrix clause and embedded clause
(that is, what would be matrix and embedded clause with an independent modal
verb like vajag) are fused into one. We cannot therefore distinguish the heads
Mod and V, and there is not much sense in positing an embedded clause in this
case. A schematic representation would be (16):

(16) S
wp
NP AuxP
qi
Aux AntP
qi
Ant ModP
ei
Mod VP
ei
V NP
5
Tev ir bijis jālasa šī grāmata
2sg.dat be.prs.3 be.ppa.nom.sg.m deb-read dem book.nom.sg
One aspect of the bi-clausal debitive construction, however, must be briefly dis-
cussed, viz. the case form of the participle. In a finite compound perfect tense
form it would be in the nominative, agreeing with the subject:
(17) Es Jān-i esmu redzēj-is.
1sg.nom John-acc be.prs.1sg see-ppa.nom.sg.m
‘I have seen John.’
With what does the dative in (13) agree? At first sight it appears to agree with the
dative NP controlling the PRO-subject of the embedded clause. But this guess
would probably be amiss. The predicate nominal in an embedded infinitival clause
is in the nominative when it is controlled by a nominative subject in the matrix
clause, but in the dative in all other instances, even when, say, the controller in the
main clause is in the accusative (cf. Holvoet 2005 for details):
(18) Esi cenšos [PROi būt godīg-s.]
1sg.nom try.prs.1sg be.inf honest-nom.sg.m
‘I try to be honest.’
(19) Es jūsi lūdzu [PROi būt man-iem
1sg.nom 2pl.acc ask.prs.1sg be.inf 1sg.poss-dat.pl.m
vies-iem.]
guest-dat.pl
‘I request you to be my guests.’
As we have agreement with the nominative in (17), it is a priori conceivable that

there could also be agreement with a datival quasi-subject. But as the dative also
occurs as a default agreement case (alternatively, we could say that the PRO-sub-
ject is dative-marked if there is no nominative-marked agreement controller in
the matrix clause), we have no means of distinguishing between agreement and

non-agreement (or default agreement). The case form of the participle cannot
therefore be used as evidence in a discussion on subjecthood in the bi-clausal
debitive construction. This kind of evidence must also be used with caution in
a discussion of grammatical properties in the basic (monoclausal) debitive con-
struction, where the predicate nominal with jābūt ‘one must be’ is also in the
dative:
(20) Jums jā-būt man-iem vies-iem.
2pl.dat deb-be 1sg.poss-dat.pl.m guest-dat.pl
‘You must be my guests.’
As this construction has become monoclausal, one could, perhaps, make a case
for the predicate nominal agreeing in case with the datival quasi-subject. Proof of
this is, however, impossible to provide, and agreement of the predicate nominal
in the debitive construction will not further be used as evidence. Only agreement
of participles occurring as parts of compound forms of the auxiliary may be sig-
nificant, as we will see further on.
4. The argument structure of modals
Whereas epistemic modals have no argument structure of their own (their com-
plement is a complete proposition, without any argument being singled out),
event modals often single out an animate subject that is felt to experience neces-
sity or possibility and also doubles as an agent with respect to the action which is
to be performed:
(21) You should drink less coffee.
The syntactic position reserved for such a modal target may, it is true, be occupied
by an inanimate argument that is not an agent:
(22) That vase should be standing on the chimney-piece.
This lack of sensitivity to animacy and semantic role is reminiscent of the epis-
temic modals, and could be viewed as evidence for an interpretation of modals as
raising verbs. But several Slavonic and Baltic languages have impersonal modals
that can be used only if there is an animate ‘experiencer’ of event modality. In Ei-
de’s (2005: 47–49) terminology, they are always directed, two-place or ‘ought-to-
do’ deontics and cannot be construed as non-directed, one-place or ‘ought-to-be’
deontics.
(23) Mne nado schodit’ v magazin. Russian

1sg.dat be_needed run.inf into shop.acc.sg
‘I must run to the shop.’
( 24) *Vaz-e nado stojat’ na kamin-e.
vase-dat.sg be_needed stand.inf on chimney-piece-loc
Intended meaning: ‘The vase should stand on the chimney-piece.’
In a similar way, such modals cannot be used to express epistemic modality:
(25) *Maš-e sejčas uže nado byt’ doma. Russian

Mary-dat.sg now already be_needed be.inf home
Intended meaning: ‘Mary should already be at home by now.’
On this class of modals in Slavonic, cf. Hansen (2001: 121–122, passim). As shown

in Holvoet (2007: 146–149), Latvian has done away with all these restrictions,
e.g., the impersonal modal vajadzēt is syntactically exactly parallel to Lithuanian
reikėti, but unlike its Lithuanian counterpart it extends to all types of modality
including epistemic modality, and it shows no restrictions with regard to the ani-
macy of the modal target. This is shown by (26):
(26) Tiem vajaga būt bij-uš-iem
these.dat.pl.m be_needed.prs.3 be.inf be-ppa-dat.pl.m
varen-iem ozol-iem.
mighty-dat.pl.m oaks-dat.pl
‘These must have been mighty oaks.’ (G. Janovskis)
The Latvian ‘quirky’ modals also function as ‘ought-to-be’ or one-place (non-di-

rected) deontic modals. This goes both for vajag and for the debitive, but will be
illustrated only for the debitive here:
(27) Šajā gadījumā gald-am jā-stāv pēc iespējas tālāk
in this case table-dat.sg deb-stand as far as possible
no ieejas, lai būtu laba iespēja pārredzamīb-ai.
from entrance so_that be.irr good possibility overview-dat.sg
‘In this case the table should stand as far as possible from the entrance so that
there is a good possibility of taking a general view.’
http://www.apollo.lv/zinas/majas-kabineta-fen-sui/389420
This is an exception confirming the rule: in languages having a personal necessi-

tive modal alongside impersonal ones, it is the personal one that is used in epis-
temic function and ought-to-be deontic function, in the lack of animate modal
targets. But Latvian never developed a personal necessitive modal, overriding the
usual restrictions on impersonal modals instead. Let us note that several Slavonic
languages that now have personal necessitive modals have borrowed them from
German (cf. Polish musieć and Czech muset from German müssen, a widespread
borrowing discussed by Hansen 2001).
We can now see case assignment with the Latvian debitive in the context of
the impersonal and voice-sensitive modals discussed by Narrog: despite the lack
of any restrictions on the use of the debitive, the debitive aligns with a type of
modals typically associated with event modality and with animate modal sub-
jects, and its extension into other types of modal expressions is a language-spe-
cific feature associated with the failure of Latvian to develop or borrow a personal
necessitive verb.
In view of the above we need not expect grammatical relations in the deb-
itive construction to correspond to the standard, lexically specified assignment
of grammatical relations with any given verb. Rather we must view them in the
context of other verbs taking datival quasi-subjects alongside other nominative or
oblique marked arguments. These will include modal verbs like vajadzēt, but also
other verbs in the zone of low transitivity (in the sense of Hopper & Thompson
1980), with experiencer arguments. In the discussion that follows, we will simply
refer to the datival NP in the debitive construction as the A argument, and to the
nominatival or accusatival argument as the B argument.
5. Grammatical relations and case marking with the debitive:

Earlier discussion
The debitive construction has been mentioned in the literature as a problem for
grammatical relations and for case marking. The discussion of grammatical rela-
tions with the debitive starts with Fennell (1975), who already suggests describ-
ing the dative NP as a subject, citing such evidence as reflexivization. Of course
it pops up in traditional Latvian grammars as well, where the question is often
posed which noun phrase in the debitive construction is to be considered the
subject (e.g., Bergmane et al. 1962: 233).
Another way of looking at the debitive construction is concentrating on case
marking while taking grammatical relations more or less for granted. The mark-
ing of the second argument was touched upon by Timberlake (1974), who men-
tioned the Latvian debitive in connection with the Fennic nominative object. The
parallelism is indeed striking and deserves to be mentioned here. In Fennic the
object of an embedded infinitive is in the nominative if the higher verb has no
nominative subject:
(28) Sinu-n täytyy luke-a tämä kirja.Finnish

2sg-gen be_needed.3sg read-inf dem book.nom
‘You must read this book.’
The Fennic nominative object rule is subject to the same animacy constraint as
the B argument with the Latvian debitive – 1st and 2nd person pronouns with
embedded infinitives are not in the nominative, but in the accusative – a special
case available only for pronouns (for nouns the genitive is said to function as an
accusative):
(29) Sinu-n täytyy kutsu-a minu-t.Finnish
2sg-gen be_needed.3sg invite-inf 1sg-acc
‘You must invite me.’
This comparison with the Fennic nominative object rule draws the debitive into
another domain of investigation, that of case marking strategies. Authors working
in this domain attempt to account for the ways in which, grammatical relations
being given, languages tend to encode them. Comrie (1975) introduces the mutu-
ally complementary mechanisms of discriminatory case (case used to distinguish
semantic roles) and language economy (case differentiation is avoided where no
discriminatory function is to be performed). In more recent publications, inspired
by Optimality Theory, we find a triad of competing motivations – distinguishing
and identifying use of case, and economy (cf. De Hoop & Malchukov 2010).
It might seem superficially tempting to view the Latvian debitive in the context
of such discussions. Even conceding that the debitive has become a monoclausal
structure, one might claim, say, that the nominative is a strategy of encoding the
object in the lack of a nominative subject, etc. But such speculations presuppose,
of course, that one is able to identify an object as such, and to go on investigating
why it is coded the way it is. We will be arguing here that this may not be taken
for granted.
Given that the debitive is apparently an inflectional form of the verb, the
change of case marking, which might also be one of grammatical relations (this is
to be investigated in the present paper) gives the impression of a split, similar to,
say, the tense-aspect triggered split known from Georgian, mood-triggered split
as with the object of the Finnish imperative etc. But whereas a tense-triggered
split is something unexpected (we do not conceive of a tense feature as a distinct
predicate with its own argument structure), a split associated with a modal affix is
less unexpected. Modality is a category that has been shown to interact with argu-
ment marking (Narrog 2010), and this also applies to instances where the modal
marker is affixal.
6. Grammatical relations with the debitive
6.1 Morphosyntactic tests
Morphosyntactic properties such as agreement and case marking are the first that
have to be tested before we pass on to behavioural properties. The Latvian con-
struction with the debitive contains, in its literary variety, a dative marked A argu-
ment and a nominative marked B argument.
Agreement is the first feature to be examined. The impossibility of establish-
ing agreement with the dative marked A argument has already been mentioned
above, so we will concentrate on the B argument. The auxiliary can be in a com-
pound tense and will then contain a past active participle capable of agreeing with
a nominative B argument, and which we would expect to agree if B is a subject. It
usually doesn’t, however:
(30) Šajos gad-os ir bij-is jā-lasa
those.loc.pl.m year-loc.pl be.prs.3 be-ppa.nom.sg.m deb-read
dažād-as ziņ-as,
various-nom.pl.f news_item-nom.pl
tai skaitā arī šokējošas, kad pašai diktorei trīcējušas kājas no uztraukuma.
‘In those years one has had to read various news items, among them shocking
ones, which made the newsreader’s legs shake with excitement.’
http://www.diena.lv/sabiedriba/pasi-mazakie-648149
By way of comparison, we will cite an example with patikt ‘please’, for which dat-
nom is the lexically assigned case pattern, as illustrated in (31):
(31) Man patīk šī film-a.
1sg.dat please.prs.3 this.nom.sg.f film-nom.sg
‘I like this film.’
In the compound tenses of patikt, the past active participle consistently agrees
with the B argument:
(32) Maz-ajām meiten-ēm visvairāk bija
little-dat.pl.f.def girl-dat.pl most be.pst.3
patik-usi indieš-u princes-e.
please-ppa.nom.sg.f Indian-gen.pl princess-nom.sg
‘The little girls had liked the Indian princess most of all.’
http://www.jrt.lv/nijaramas-pasakas?page=1
The Latvian Academy Grammar (Bergmane et al. 1959: 618) states that agreement
with the nominative NP in the debitive construction is possible, but rare. While
this may be true, low relative frequency does not entitle us to treat it as marginal.
We find constructions with agreement even in contemporary informal internet
texts. Not unexpectedly, we find it in contexts where the nominative NP is pre-
verbal, as in (33):
(33) Es gan spriežu no malas, jo man par laimi
1sg.nom ptcl judge.prs.1sg from aside because 1sg.dat fortunately
šād-a izvēl-e nav bij-usi jā-izdara.
such-nom.sg.f choice-nom.sg be.prs.3.neg be-ppa.nom.sg.f deb-make
‘True, I can judge only from the sidelines, as I have never been compelled to
make such a choice.’
http://www.calis.lv/forums/tema/18191705-piespiedu-karta/
But topicalization and sentence-initial position do not seem to be a necessary

condition for agreement:
(34) Kad 1940. gad-ā Soikan-am bij-usi jā-izdara
when 1940 year-loc.sg PN-dat be-ppa.nom.sg.f deb-make
izvēl-e starp konservatorij-u un Māksl-as akadēmij-u,
choice-nom between conservatoire-acc and art-gen.sg academy-acc
viņ-š izšķīr-ies par labu pēdēj-ai.
3-nom.sg.m decide-ppa.nom.sg.m in_favour latter-dat.sg.f
‘When, in 1940, Soikans had to make a choice between the conservatoire and
the Academy of Fine Arts, he decided in favour of the latter.’
http://www.diena.lv/arhivs/magiska-realisma-stasts-12559774
What, then, can be concluded from the agreement pattern? Agreement with the
nominative marked NP is considerably less regular than with the lexically as-
signed nominative with patikt, but it is not replaced with agreement with the A
argument – instead, we have (partial) loss of agreement, which probably simply
reflects the grammaticalization of the debitive and a concomitant decategorial-
ization (for this notion cf. Hopper 1991). It could be objected that the pluperfect
in clauses like (32) is also a grammatical form, so that the loss of agreement of
the participle would not have been unexpected. While this is certainly true, an
explanation is available for this: the perfect forms of the verb have remained much
closer to their source, the copular construction, and indeed have not yet become
completely dissociated from it. The source construction of the debitive has, on the
other hand, become completely obscured.
Other morphosyntactic properties yield very weak evidence. Replacement of
the nominative or accusative with other cases, e.g., the genitive of negation, is
a type of evidence that could be used. Replacement with a genitive of negation

is exceedingly rare, but nevertheless attested. It should be added that, like other
types of adverbal genitive, the genitive of negation has been in decline in Latvian
at least since the 19th century, and in the course of the decades following the
Second World War it has completely fallen into disuse (cf. Berg-Olsen 1999). One
of the positions in which it has been retained longer than in others is with the
emphatic negation ne ‘not even, not as much as’. In the presence of this particle,
we sometimes find a genitive instead of a nominative:
(35) Māter-u Jur-is iztaisīj-is, ka nav
Māters-gen.pl Juris-nom arrange-ppa.nom.sg.m comp be.prs.3.neg
jā-maksā ne kapeik-as.
deb-pay not_even kopeck-gen.sg
‘Juris Materis has arranged things so that not a kopeck is to be paid.’
The genitive of negation is characteristic of both intransitive subjects and transi-

tive objects, as shown by the following examples:
(36) Mūsu valst-s ne-dod ne kapeik-as,
1pl.gen state-nom.sg neg-give.prs.3 not_even kopeck-gen.sg
tāpēc esam aizkavēj-uš-ies.
therefore be.prs.1pl get.behind-ppa-nom.pl.m
‘The state doesn’t give a kopeck, that’s why we have got behind.’
http://politiskirepresetie.lv/?id=22
(37) Tas nozīmē, ka no šīs firm-as Latvij-ā
that mean.prs.3 that from this.gen.sg.f company-gen.sg Latvia-loc
ne-paliks ne kapeik-as.
neg-remain.fut.3 not_even kopeck-gen.sg
‘That means that not as much as a kopeck will be left over from this company
in Latvia.’
http://www.diena.lv/arhivs/labi-sakts-darbs-ir-japabeidz-10592013
The use of the genitive in (35) does not, therefore, yield any evidence as to wheth-
er the B argument is an object or an intransitive subject.
We might finally mention the fact that the nominative with the debitive con-
struction may also replace an adverbial accusative. For instance, an accusative of
duration may either be replaced with the nominative (38) or be retained (39):
(38) Vēl vien-s gad-s tev jā-pabarojas.
still one-nom.sg.m year-nom.sg 2sg.dat deb-get_fat
‘You have to be fattened for one more year.’
(from a folk tale, cited Endzelīns-Mühlenbachs 1907: 197)
(39) Vis-u vakar-u tām bija vien-ām

all-acc.sg evening-acc.sg 3.dat.pl.f be.pst.3 one-dat.pl.f
paš-ām jānosēž kakt-ā.
self-dat.pl.f deb-sit_out corner-loc.sg
‘For the whole evening they had to sit in a corner by themselves.’
(Jānis Veselis)
This tendency to treat adverbials of duration in the same way as objects probably
points to some kind of similarity between the two (evidently because patients and
themes typically ‘measure out’ the action), but it is doubtful whether the replace-
ment of the accusative with the nominative in the debitive construction can tell us
anything about grammatical relations with the debitive. The nominative rule for
adverbials of duration is similar to the genitive of negation rule, which optionally
applied to these adverbials at the time when this rule was still operative for direct
objects:
(40) Ja arī kāds redaktors Krievijā būtu sociāldemokrāts,
Viņ-š ne-vien-as dien-as ne-drīkstētu
3-nom.sg.m neg-one-gen.sg.f day-gen.sg neg-be_allowed.irr
sav-ā viet-ā palikt.
poss.refl-loc.sg place-loc.sg stay.inf
‘Even if there were any social-democrat editor in Russia, he wouldn’t be
allowed to stay in his place for one single day.’ (Augusts Deglavs)
As in the case of the genitive of negation rule, the case-marking rule applied to
direct objects optionally applies to accusative-marked temporal adverbials. What
this means in terms of grammatical relations is not quite clear. The application of
the genitive of negation rule to adverbials does not make them into objects, and
the application of the rule ‘change accusative to nominative with the debitive’ is
probably also just a rule of case marking, without clear implications for the pat-
tern of grammatical relations.
To sum up the evidence of morphosyntax: the nominative marked B argu-
ment with the debitive shows certain morphosyntactic subject features, though
not consistently; unambiguous object properties in the B argument cannot be
found.
6.2 Syntactic tests
Syntactic tests should, in principle, be decisive in establishing subjecthood. A

set of behavioural tests developed by Zaenen, Maling and Thráinsson (1985),
basically going back to Keenan (1975), has yielded good results for Icelandic.
But the results are rather poor for other Germanic languages like German (cf.
Andrews 2012, though cf. also Barðdal & Eythórsson 2006, who attempt to rela-
tivize the differences), and also for Slavonic and Baltic. An important test is that
of syntactic embedding, e.g. one can check whether a datival subject of an embed-
ded infinitival clause can be controlled by a higher clause complement, yielding
something like
(41) *Johni hopes [dati to please the book]
‘John hopes he will like the book.’
This test works for Icelandic but not for any Baltic or Slavonic language (cf.
Holvoet 2013). Unfortunately it cannot be applied to the debitive, which, prob-
ably partly for semantic and partly for syntactic reasons, is not amenable to em-
bedding. The debitive has no non-finite forms at all. Its nearest counterpart, the
modal verb vajadzēt, does not allow any embedding based on subject properties
of the dative A argument:
( 42) *Jān-is cer [PRO ievajadzēties naud-u]
John-nom hope.prs.3 become_needed.inf money-acc
Intended meaning: ‘John hopes to find himself in need of money.’
One test that does seem to work to some extent is the reflexivization test. Let us
see how it works for the Latvian debitive. Only the dative argument can control
reflexive pronouns proper:
(43) Viņam ir dāvanas, zēni, lielas dāvanas, un tādēļ
viņ-š mums jā-uzlūko par sev līdzīg-u
3-nom.sg.m 1pl.dat deb-regard as refl.dat equal-acc.sg
‘He’s got gifts, boys, great gifts, and therefore we should regard him as our
equal.’ (Augusts Deglavs)
Control of reflexive possessive pronouns, however, yields different results. Re-

flexivity tests involving reflexive possessives are not often used in the literature,
because many languages do not distinguish reflexive from non-reflexive 3rd per-
son pronouns (e.g., English his, her are ambiguous with regard to reflexivity).
When they are distinct, they can yield valuable evidence because their application
is somewhat wider that that of tests with reflexives proper. When, for instance,
we take a Latvian verb like patikt ‘please’ (a verb of a type that has often been
discussed in the literature on non-canonical grammatical relations, cf. Icelandic
líka), we will be unable to detect any ability to control reflexivity in the dative ar-
gument, because the reflexive pronoun sevis has no nominative. When, however,
we allow evidence from reflexive possessives, we see that the dative can control at
least a certain type of reflexives:2
(44) Tev patīk sav-i klasesbiedr-i?
2sg.dat please.prs.3 refl.poss-nom.pl.m classmate-nom.pl
‘Do you like your classmates?’ http://www.formspring.
me/r/tev-pat-k-savi-klasesbiedri/253602291771249079
Applying this test to the debitive, we see that the dative-marked A argument con-
trols reflexive possessive pronouns as well:
(45) Tā ir pasaul-es lielāk-ā
dem-nom.sg.f be.prs.3 world-gen.sg greatest-nom.sg.f.def
nelaim-e, ka vecāk-iem jā-redz sav-i
misfortune-nom.sg comp parent-dat.pl deb-see poss.refl-nom.pl.m
bērn-i aizej-am.
child-nom.pl go_away-cvb
‘The greatest misfortune in the world is that parents have to watch their chil-
dren go away.’
http://www.calis.lv/forums/tema/15206129-engelisu-maminas/51/
Though regular, control of reflexive possessives by the dative NP is not without

exceptions, as shown by the following:
(46) Galu galā, valsts brīvības cīnītāju pirmais un vispārēji svarīgais iemesls lai
atgūtu brīvību, ir
lai mūsu vecāk-iem un vecvecāk-iem ne-būtu
so_that 1pl.gen parent-dat.pl and grandparent-dat.pl neg-be.irr
jā-redz viņ-u mocītāj-u simbol-us un
deb-see 3-gen.pl tormenter-gen.pl symbol-acc.pl and
2. Certain differences have been noted with reference to the antecedents of the reflexive prop-
er and the reflexive possessive. Rappaport (1985) says that the Russian reflexive possessive svoj
may have arbitrary reference. This can also be said of Latvian savs, cf.
Ja esmu piedzim-is šai pasaul-ē, tad arī
if be.prs.1sg be.born-ppa.nom.sg.m this.loc.sg world-loc.sg then also
man sav-s zem-es kaktiņ-š pienākas.
1sg.dat refl.poss-nom.sg.m land-gen.sg corner-nom.sg be.due.prs.3
‘If I was born in this world, then a corner of land of my own is my due as well.’
http://www.tvnet.lv/zinas/viedokli/
325226-latvija_joprojam_var_nopelnit_naudu/comments/page/10
It seems, however, that these uses are emphatic (‘of one’s own’) rather than reflexive, and that
properly reflexive uses must be bound.
slavināšan-u mūsu paš-u māj-ās.

glorification-acc 1pl.gen self-gen.pl home-loc.pl
‘Ultimately the first and foremost motive for the champions of national free-
dom to regain liberty was that our parents and grand-parents would not be
forced to look at the symbols and glorification of their tormentors in our own
home country.’ http://www.peticijas.com/forum/post/721929
A Google search reveals that such instances, though not very frequent, are too
frequent to be written down to performance errors. Control of reflexive posses-
sives by nominative subjects is presumably absolutely regular, but in the case of
dative marked NPs this regularity seems to be less absolute. There could perhaps
be additional factors, perhaps connected with semantic role (agency) or with cat-
egorical features like animacy, that may cause deviation from it. This would have
to be separately investigated. Redzēt, it might be noted, is low in transitivity. But
with darīt savu darbu ‘do one’s job’, where the subject is agentive, we also find,
alongside an enormous preponderance of jādara savs darbs, occasional uses with
lack of reflexivization:
(47) Manuprāt, katr-am kultūr-as darbiniek-am un arī
to_my_mind every-dat.sg.m culture-gen worker-dat.sg and also
viņ-u vadoš-ajiem darbiniek-iem ir jā-dara
3-gen.pl senior-dat.pl.m.def worker-dat.pl aux.prs.3 deb-do
viņ-u darb-s.
3-gen.pl work-nom.sg
– un tas ir nevis priekš ķeksīšiem, bet cilvēkiem.
‘To my mind every culture worker, and also their senior workers, should do
their job.’ http://www.estars.lv/raksti/2/15421
With a similar order of nominals (dat – nom) the nominative does not com-
monly control reflexivization, as seen in the following:
(48) Bet siev-ām vīr-i aizvien jā-redz ar vis-iem
but wife-dat.pl husband-nom.pl always deb-see with all-dat.pl.m
viņ-u veikal-iem
3-gen.pl business-dat.pl
‘But wives should always view their husbands in the context of their (sc. their
husbands’) business occupations.’ (Pāvils Rozītis)
However, one does find examples where the B argument is topicalized and occurs
sentence-initally, and then it may control reflexivization:
(49) Atgriežoties pie biskvīta, kad tas izcepies,

tas jāņem ārā, jā-atbrīvo no sav-a
it.nom.sg.m deb-take out deb-loosen of poss.refl-gen.sg.m
folij-a žodziņ-a un cepam-ā papīr-a,
foil-gen.sg enclosure-gen.sg and baking-gen.sg.m.def paper-gen.sg
lai atdziest.
in_order_that get_cold.prs.3
‘To get back to the sponge cake, when it’s ready, it should be taken out, loos-
ened from its foil enclosure and from the baking paper to let it get cold.’
The tests, then, yield no spectacular results. The dative NP controls reflexivization
rather regularly (though not with absolute consistency), but the nominative can
also control reflexivization when topicalized and moved to sentence-initial posi-
tion. It seems that such oscillations as we find in texts reflect genuine oscillations
in usage: prescriptive grammar pays a lot of attention to the use of the nomina-
tive with the debitive, but grammarians do not go so far as to regulate patterns of
reflexivization.
Outside the mainstream generative tradition, where control of reflexivization
is explained configurationally, it has been suggested that it might be determined
by other factors, e.g., Jackendoff (1973) argues for thematic relations, Pollard and
Sag (1994) suggest obliqueness (o-command), Culicover and Jackendoff (2005)
suggest a mix comprising also linear order. The examples discussed above strong-
ly point to topicality as a determining factor. If one identifies obliqueness with the
hierarchy of grammatical relations, then reflexivization is a test for subjecthood.
But it seems more convenient to characterize the obliqueness hierarchy as a hier-
archy of noun phrases based on inherent discourse saliency (inherent topicwor-
thiness). Obliqueness so defined will rest on such features as semantic role and
animateness. There is a default correspondence between grammatical relations
and degrees of obliqueness, but we must allow for the possibility of the status of
subject or direct object not being conferred; in this way, we can capture the fact
of oblique marked NPs being intuitively felt to be subjects because of their status
of least oblique NP, while at the same time not satisfying the subjecthood criteria
associated with fully-fledged subjects.
The reflexivization test is thus not a very strong one and we do certainly not
regard it as sufficient proof of subjecthood. But in the absence of better syntactic
tests it might perhaps serve as an indication of which of two arguments can raise
stronger claims to subjecthood. This, however, seems to be dependent on such
factors as information structure and word order, and that is a question to which
we will turn now.
6.3 What do the subjecthood tests show?
We saw above that, in constructions with the debitive, reflexivization may be con-
trolled both by the A argument and the B argument, but not in exactly identical
conditions: in (49) the nominative NP is topicalized and fronted, and this is evi-
dently a condition for the test to work. Constructions with the debitive are not ex-
ceptional in this respect. Constructions with patikt show the same indeterminacy:
both the dative marked A argument and the nominative marked B argument may
control reflexivization, but in the case of the B argument a condition is that it must
be topicalized and occur sentence-initially. Compare the following examples:
(50) Tev patīk sav-i klasesbiedr-i?
2sg.dat please.prs.3 reflposs-nom.pl.m classmate-nom.pl
‘Do you like your classmates?’ http://www.formspring.
me/r/tev-pat-k-savi-klasesbiedri/253602291771249079
(51) Pirmdzimtais var būt krietns lielais brālis vai laba lielā māsa.
Tād-s viņ-š patīk sav-iem vecāk-iem.
such-nom.sg.m 3-nom.sg.m please.prs.3 reflposs-dat.pl parent-dat.pl
‘The firstborn can be a decent elder brother or a good elder sister. That’s how
his parents like him.’
http://wow7.blogs.lv/2010/11/10/cela-cirtejs-diplomats-vai-dumpinieks-1/
We will not say that by changing information structure and word order in typical
transitive structures like John ate a sandwich we are changing the assignment of
grammatical relations, because this assignment reflects only the default assign-
ment of topichood, secondary topichood etc.: it is only relative topicworthiness
that can be grammaticalized in the assignment of grammatical relations specified
for a specific verb in the lexicon. But when we consider constructions with dif-
fuse grammatical relations, as the ones discussed here, the situation is different.
Barðdal (2001) notes that with Icelandic verbs like líkar the dative has many sub-
ject properties in the configuration dat-nom, but when the pattern of topicaliza-
tion and word order are changed, the nominative becomes a fully-fledged subject.
Barðdal explains this by suggesting that subjecthood is construction-specific, and
that there are simply two distinct constructions, dat-nom and nom-dat. It may
be justified from the viewpoint of Construction Grammar to postulate two dis-
tinct constructions here, but let us note that the difference between the two con-
structions reduces essentially to information structure. Usually a reversal in the
pattern of information structure does not lead to the rise of a new construction.
Two distinct constructions may differ, among other things, in topic-comment
structure, e.g., existential constructions will typically have place as topic, where-
as locational constructions will have theme as topic. But these constructions also
have morphosyntactic peculiarities of their own that allow us to set them apart as
distinct constructions irrespective of topic-comment structure. Cf. the following
examples from Lithuanian:
(52) Pieštuk-ai yra stalči-uje.
pencil-nom.pl be.prs.3 drawer-loc.sg
‘The pencils are in the drawer.’
(53) Stalči-uje yra pieštuk-ų.
drawer-loc.sg be.prs.3 pencil-gen.pl
‘There are pencils in the drawer.’
(54) Pieštuk-ų yra stalči-uje.
pencil-gen.pl be.prs.3 drawer-loc.sg
‘Where you can find pencils is in the drawer.’
(52) is locational with topicalized theme, (53) is existential with topicalized place,
and (54) is existential with topicalized theme. This shows that the existential con-
struction may depart from its typical topic-comment structure without ceasing
to be existential. In the case of the two varieties of clauses with Latvian patikt,
however, the difference seems to reduce to topic-comment structure.
One might object here that in Icelandic the differences are more marked, as
Icelandic quasi-subjects pass more subjecthood tests than in Baltic, and that this
justifies positing two distinct constructions. Within a different, say, generative
framework, one might capture this by saying that there are two distinct underly-
ing structures. In both cases this claim would have a very specific meaning, that
is, in a constructional framework one would have to say that both constructions
(dat-nom and nom-dat) inherit the subject construction, associating this role
either with the dative or with the nominative; and in a Minimalist framework one
would have to say that either the dative or the nominative is generated, in under-
lying structure, in a position such as to enable control of reflexivization through
the mechanism of c-command.
The logic of the explanation would be, in both cases, inescapable provided
one accepts that the subjecthood tests are indeed indicative of subjecthood. But
what if they are connected with topichood? After all, subjecthood is a prototypi-
cal category – it rests on a cluster of properties partly semantic (agent and topic),
partly syntactic (behavioural properties), and partly morphosyntactic (coding
properties). A fully-fledged subject will therefore consist in a certain pairing of
meaning and form. Where the formal properties are absent, relying on the seman-
tic properties – agenthood, topichood – leads, of course, to circularity, and there-
fore researchers are right in insisting on syntactic properties, that is, behavioural
properties that can be brought to light by subjecthood tests. But if the behavioural
tests reflect topichood rather than subjecthood, the reasoning becomes, again, cir-
cular. We find that subjecthood encodes, among other things, topichood; where
the morphosyntactic features (nominative, agreement) are present, we take them
as a basis for identifying the subject; when they are lacking, we identify the subject
on the basis of features actually relying on topichood. In fact, the correct solution
would be to say that in those cases, there is no subjecthood.
Of course, if we find tests sensitive also to case marking, they will be more
conclusive. Reflexivization tests may be assumed not to be highly sensitive to case
marking, but when agreement is concerned we will expect a connection, because
agreement involves, among other things, case agreement. But there are other
tests, basically syntactic in nature, that might also be more sensitive to case, viz.
coordination tests.
6.4 Coordination tests
The coordination tests are meant to show what can function as a pivot-control-
ler in coordination. The datival quasi-subject in the construction with patikt
cannot:
(55) Bērn-iem patīk un *(viņ-i) prasa labāk
child-dat.pl please.prs.3 and 3-nom.pl.m demand.prs.3 rather
tās pasak-as, nekā grāmat-u.
that.acc.pl.f fairy.tales-acc.pl than book-acc.sg
‘The children like them and call for those fairy tales rather than for a book.’
http://www.apollo.lv/komentari/zinas/507005/1
The nominative argument can, provided it is topicalized and fronted:
(56) Bet nu tā pa lielam jau, protams, galvenais,

ka vārd-s piestāv un Ø patīk vecāk-iem.
that name-nom.sg fit.prs.3 and Ø please.prs.3 parent-dat.pl
‘But, broadly speaking, the most important thing is that the name fits and that
the parents like it.’
http://www.calis.lv/forums/tema/18110672-meitenu-vardi/72/
In a way, this confirms what was found above for the reflexivization test when
applied to patikt. The tests that work for the nominative require an additional
factor, viz. topicalization and initial position. For the dative argument only the
reflexivization test works, not the coordination test, perhaps because here only
one of two criteria is met and nominative case is lacking.
With the debitive the possibilities of equi-deletion in coordination are ex-

tremely limited. A nominative B argument does not seem to be able to control a
conjoined clause pivot:
(57) Salas bija pārvērstas par mīnu laukiem,
ciemat-i daudz-iem bija jā-atstāj un *(tie)
village-nom.pl many-dat.pl aux.pst.3 deb-leave and 3.nom.pl.m
kļuva gandrīz ne-apdzīvot-i
become.pst.3 almost neg-uninhabited-nom.pl.m
‘The islands were turned into minefields, and the villages had to be abandoned
by many and they became almost uninhabited.’
http://www.balticgreenbelt.uni-kiel.de/fileadmin/fileexchange/Outcomes
Documentation/Project_Newsletter/No._7/BGBPanoramaNo7_LV.pdf
On the other hand a nominative B argument can, on condition of being topical-

ized and fronted, be a pivot controlled by a nominative subject itself, cf. (56).
Note, however, that this example contains an overt present tense auxiliary ir,
which might perhaps strengthen the subject properties of the nominative, mak-
ing the whole structure more similar to a copular construction:
(58) Vid-es, sociāl-ie un ekonomisk-ie
environment-gen social-nom.pl.m.def and economic-nom.pl.m.def
jautājum-i ir savstarpēji cieši saistī-t-i un Ø
issue-nom.pl be.prs.3 mutually closely connect-ppp-nom.pl.m and
ir jā-risina vienoti.
be.prs.3 deb-solve conjointly
‘Environmental, social and economic issues are closely interrelated and must
be dealt with conjointly.’
http://archive.politika.lv/temas/vide_un_ilgtspeja/17419/
By contrast, the dative quasi-subject does not seem to have this property:
(59) Bīni satiku, kura tik ļoti skumst, ka

viņ-ai jā-strādā un *(viņ-a) ne-var
3-dat.sg.f deb-work and 3-nom.sg.f neg-be_able-prs.3
ballīt-ē piedalīties.
party-loc.sg take_part.inf
‘I met Bīne, who is ever so sad she has to work and can’t go to the party.’
http://klab.lv/users/kripatina/95616.html
We saw above that the dative marked A argument passes the reflexivization test,
which is presumably less dependent on case marking. Pivot-controller proper-
ties seem to be more restrictive: they may require the controller to be a topic but
also to have the same case marking as an overt subject in the place of the pivot
would have. It is, in this sense, a better subjecthood test but, for both the lexically
determined nominative with patikt and the nominative NP with the debitive, it
is indicative of a kind of conditional subjecthood, on which we will comment in
the following section.
7. Some theoretical considerations
7.1 Recoverable subjecthood
Patikt and the debitive show, basically, a nominative subject whose subject prop-
erties are partly suspended in the normal configuration usually discussed in the
literature on non-canonical subjects, that is, dat–nom, but recovered when we
diverge form this pattern. We call this recoverable subjecthood (correspondingly,
we could speak of suspended subjecthood in the construction with a less-oblique
dative marked noun phrase).
The notion that subjecthood features can be suspended leads us to ask wheth-
er such a situation could not be accounted for by using the notion of demoted
subject. There is an analogue to this in the linguistic literature: direct objects are
known to be deprived of some of their object properties in the vicinity of indirect
objects within the ditransitive construction; in Relational Grammar, such demot-
ed direct objects are characterized as chômeurs (Perlmutter & Postal 1983: 95–99).
This demotion is shown by the possibilities of passivization: from John gave Mary
the flowers we can derive Mary was given the flowers, whereas The flowers were
given Mary is acceptable only to some speakers of English. Below, we will argue
for an analogous notion of demoted intransitive subject.
7.2 Demoted intransitive subjects
As argued in Holvoet (2013), the configuration dat–acc (or dat–gen) in in-

transitive structures has an analogue in ditransitive structures. The latter show
alignment differences which Blansitt (1984) had already characterized as a
choice between dative and dechticaetiative patterns; Malchukov, Haspelmath &
Comrie (2010) reformulate this as indirective and secundative alignment. Di-
transitive structures show a certain indeterminacy as to which argument in an
‘agent–recipient–theme’ or ‘agent–experiencer–theme’ configuration
should be selected as the second-ranking argument immediately below the sub-
ject. Languages will either select the theme or the recipient/experiencer according
to whether they base the choice on grammatical function (identity of grammati-

cal function between two-place and three-place predications), animacy or seman-
tic role – these competing strategies are studied in Kittilä (2006). Alongside the
unambigous models in which either the theme or the recipient aligns with the
direct object of monotransitive constructions, there are also the so-called neutral
constructions characteristic, say, of English, in which the so-called intransitive
object outranks, i.e., is less oblique than, the direct object, which is sometimes
referred to as a demoted direct object. While this situation is, perhaps, to a certain
extent due to the loss of case distinctions in English, it reflects a more general
regularity, viz. that recipients/experiencers in ditransitive constructions compete
with patients/themes in terms of syntactic obliqueness, which is actually also the
basic reason for the alignment differences.
What we want to suggest here is that there is a similar competition in intransi-
tive clauses, and that experiencer arguments with their typical experiencer mark-
ing (dat) may occur in intransitive structures alongside arguments that could
be characterized as demoted intransitive subjects on the analogy of the demoted
objects of ditransitive structures.
The crucial point to be made here is that even when a language chooses the
theme as the direct object, the dative recipient/experiencer will outrank it on the
nonobliqueness-obliqueness hierarchy, which can be seen in languages with case
and free word order, where the dative will precede the accusative in ditransitive
constructions (60) while following it in other dat–acc constructions (61):
(60) On-a da-l-a rebenk-u podarok.Russian
3-nom.sg.f give-pst-f.sg child-dat.sg present.nom.sg
‘She gave the child a present.’
(61) On pripisa-l nesčastnyj slučaj ošibk-e.
3.nom.sg.m attribute-pst accident.nom.sg mistake-dat.sg
‘He attributed the accident to a mistake.’
Normally direct objects outrank all other objects, including certain dative objects,
but they are outranked by recipient and experiencer datives. For such datives the
traditional term ‘indirect object’ is available, but there is no traditional name for
the dative analogously outranking an intransitive subject as in (6) or (31); though
authors attempting to characterize it in terms of grammatical relations sometimes
call it a quasi-subject or an oblique subject. One will scarcely want to call it an
indirect object, as this term presupposes the presence of a direct object and is
therefore not applicable to intransitive structures. The best way to characterize
this configuration is therefore to say that we have a demoted intransitive subject
alongside a quasi-subject.
7.3 The debitive among other structures with demoted intransitive subjects
The point we want to make here will of course be that the debitive construction is
also among those exhibiting the two features we have introduced in Sections 7.1
and 7.2. In its basic variety it displays a nominative B argument that has no fewer
subject properties than the A argument at the light of the tests usually applied, but
is clearly less oblique.
Of course, our explanation will have to accommodate two facts already men-
tioned at the outset, viz. (i) the accusative used to mark the B argument when it is
a 1st or 2nd person pronoun or a reflexive pronoun, and (ii) the accusative con-
sistently marking the B argument in part of the Latvian dialects and apparently
to an increasing extent in substandard Latvian. We will therefore argue that the
accusative can be a means of encoding not only objects but also demoted intransi-
tive subjects.
8. Shifts in the case marking of the B argument
The tendency to replace the nominative with the accusative could appear, at first
glance, to reflect the introduction of object marking for what was actually a kind
of object. As mentioned above, the argument structures of modal and embedded
predicates are fused in the debitive construction, and as dat–nom debitive struc-
tures reflect a basic nom–acc valency of the embedded predicate, it would not be
unexpected if this object marking were transferred to the debitive construction.
What results from this shift is the case-marking pattern dat–acc, a rare con-
figuration. Usually the accusative is (lexically) assigned only if the nominative is
also assigned, a regularity reflected in Burzio’s generalization (Burzio 1986: 178),
which says that accusative is assigned only if there is an external argument marked
as agent. This will usually mean that acc is assigned if nom is assigned. Even if
this generalization is correct, it must probably allow for some motivated excep-
tions, and the debitive construction would be one of them. The A argument can be
claimed to be, at some level, an agent, even if the dative marking it receives also re-
flects another thematic role, that of experiencer with the modal predicate. Assum-
ing covert case stacking, as is fashionable in Minimalist theory (cf. Sigurðsson’s
2003 account of the dat argument in the Faroese dat–acc configuration as bear-
ing simultaneously inherent dat and structural nom to explain the acc marking
on the B argument), one could say the A argument is marked both dat and nom,
which would account for acc marking on the B argument.
On the other hand, looking at the matter from a diachronic point of view, we
could assume the B argument, which originated as the nominative subject of an
existential construction, to have undergone a syntactic reanalysis as object of the

new modal construction, with concomitant shift in case marking. If Haspelmath’s
‘behaviour-before-coding principle’ (Haspelmath 2010) is correct, we would ex-
pect the nominative to become an object syntactically first, and only then to ac-
quire the corresponding marking. It would be difficult, however, to document this
by showing that the B argument, even when marked with a nominative, actually
behaves like an object. The tests would have to be behavioural, i.e., they would
consist in passivizing, debitivizing, nominalizing etc. the verb and seeing what
will happen to the NP under investigation. But the debitive is itself a derived form
from which no further derivation is possible: it cannot be passivized, for instance,
that is, the debitive can be derived from a passive (though this is exceedingly
rare), but debitive morphology renders it impossible to derive a passive from a
debitive. The putative objecthood of the nominative marked B argument cannot,
therefore, be put to the test. One thing is striking, however, from a diachronic per-
spective. As mentioned above, the debitive was in place, as a modal construction,
in the 17th century, even though the source meaning is still attested in this period
alongside the new one. As long as the source construction existed alongside the
new modal one, the syntactic structure of the latter may have been different from
what it is now; the nominative argument was perhaps still a subject, as in the
source construction illustrated in (10). But at the same time there was already a
subjectless construction derived from intransitive verbs:
(62) wiņśch irr Eliass kam ja nahk
3.nom.sg.m be.prs.3 Elias.nom who.dat deb come
‘This is Elias, which was for to come.’
(Glück’s New Testament, 1685, Mt 11.14)
Such an impersonal debitive construction was a sufficient condition for an ex-

panded variety with a direct object to arise in the case of transitive verbs. If the
B argument is an object, it must therefore have been more or less since the 17th
century, as the preconditions for its rise were already present. As the shift in case
marking is still an ongoing process in the early 21st century, the time lag between
the putative syntactic change and the implementation of its morphosyntactic con-
sequences would be considerable.
It would probably be safer to assume that, during this long period of retention
of the nominative, the B argument did not become an object. We will argue that
in many languages we find constructions with dative A arguments outranking ac-
cusative B arguments we have no reason to regard as objects.
Evidence may be drawn from a number of constructions similar to the
above-mentioned constructions with patikt but for the fact that the B argument is
encoded by the accusative. The construction with vajadzēt (already mentioned

above as a modal verb) is the most characteristic example:
(63) Man vajag šo naud-u.
1sg.dat be_needed.prs.3 this.acc.sg money-acc.sg
‘I need this money.’
With this verb the accusative has supplanted the genitive, which is still regularly
used in texts from the late 19th century:
(64) … lūkodami tik uz to, kurš par visiem nelaimīgāks un
kur-am vispirms palīdzīb-as vajaga
rel-dat.sg.m first_of_all help-gen be_needed.prs.3
‘… taking into account only who is the unhappiest of all and needs help above
all others.’ (Kaudzītis brothers, Mērnieku laiki, 1879)
The verb vajadzēt represents a small and heterogenous class of verbs also com-
prising the non-verbal predicator žēl ‘sorry’, used with būt ‘be’ as an auxiliary
and occurring traditionally with a dat–gen, more recently also with a dat–acc
pattern; and iekāroties ‘feel a craving for’, with which the B argument may be
in the genitive, the accusative but also in the nominative (it is therefore tran-
sitional between the vajadzēt type and the patikt type). For details cf. Holvoet
(forthcoming).
In the case of vajadzēt one could, as with the debitive, venture that the accusa-
tive is introduced because the B argument is an object. But there is little to com-
mend such a judgement: whereas the configuration nom–acc is a canonically
transitive pattern in a nominative-accusative language like Lithuanian, the con-
figuration dat–acc certainly is not, and we would probably be justified in saying
that such a configuration has at least as many chances of being intransitive as of
being transitive. Is there any additional evidence? Passivization is impossible:
( 65) *Tava palīdzīb-a tiek vajadzē-t-a.
Your help-nom.sg aux.prs.3 need-ppp-nom.sg.f
Intended meaning: ‘Your help is (being) needed.’
We should also look at similar oscillations outside Baltic. The Germanic ‘like’ con-
structions come to mind here. Compare the Faroese ‘like’ constructions with ac-
cusatives alongside the Icelandic nominatives:
(66) Jón-i lík-ar þessi bók.Icelandic
John-dat like-prs.3 this book.nom
‘John likes this book.’
(67) Mær líkar henda film-in.Faroese

1sg.dat please.prs.3 this film-acc.def
‘I like this film.’ (cited from Woolford 2003)
One could view this shift as the first step in a chain of changes leading to a mod-
ern English construction like I like this film. It has been noted that constructions
of this type (called ‘blended constructions’) existed in Middle English (Trousdale
2012: 63), e.g.
(68) The more that a man beheld hym the bettre hym schuld like hym.
‘The more a man beheld him the better he should like him.’
However, this construction coexisted with the present-day construction in which

the experiencer is a subject and the stimulus an object:
(69) He made me loþen þat þat i most liked.
‘He made me hate that which I liked most.’
(68) does not seem to be an intermediary stage between the old and the new
construction here: the new constructions arose when a new categorization frame
was introduced alongside the old one and gradually ousted it, a point made by
Allen (1986). The blended construction was a process separate from the ultimate
change of categorization frame, and it consisted simply in a shift from nomina-
tive to accusative. We do not need to view this as a first step toward the rise of a
transitive structure – a structure that would be transitive only formally anyway, as
semantically it remains low in transitivity in the sense of Hopper and Thompson
(1980). Still less can the construction with the dative and the accusative be con-
sidered transitive, or on its way to becoming transitive.
Such a change in case marking, we propose, is possible in certain specific pat-
terns of morphosyntactic marking characterized by diffuse grammatical relations.
As there is no evidence for transitivity even if B is marked with the accusative,
we must characterize the syntactic structure not as A–O, but as S–E (intransitive
subject plus extension) in Onishi’s (2001) notation. To be more precise, we could
write E–S, as the datival argument is clearly less oblique than the nominative or
accusative marked argument. S is what we call a demoted intransitive subject.
The shift from nominative to accusative seems to be a not so uncommon
strategy for the marking of demoted intransitive subjects in constructions with
oblique marked quasi-subjects. This can be explained on the assumption that the
case marking of noun phrases in a clause also tends to be based on hierarchical
principles. The hierarchical ordering of cases at the morphological level (accord-
ing to the case hierarchy, cf. Blake (2001: 89–90) tends to reflect the obliqueness
cline at the syntactic level (cf. the Generalized Hierarchy Principle in Primus
1999)).
Morphosyntactic and syntactic change in constructions with diffuse gram-
matical relations often reflect obliqueness adjustments: their purpose is to bring
the hierarchical ordering of cases in accordance with syntactic obliqueness when
an obliqueness mismatch occurs. An obliqueness mismatch is what we observe
in clauses like (18), where a more-oblique argument bears a case higher in the
case hierarchy than a less-oblique argument (in this case, nominative and dative
respectively). The introduction of the accusative in instances like (34), (36) is a
way of lessening the obliqueness mismatch.
This claim calls, of course, for further explanation. If the cline of morphosyn-
tactic obliqueness is adjusted to that of syntactic obliqueness, we would expect
still further adjustment, to the point where we get, say, a shift from dat–nom to
nom–acc, as instantiated by English like. What we witness here is a partial adjust-
ment. It may, of course, go further and lead to the rise of a canonical transitive
nom–acc pattern. What we want to stress, however, is that we also need some-
how to characterize the dat–acc structure. It is clearly intransitive, and this claim
is not in contradiction with Hopper and Thompson’s (1980) Transitivity Hypoth-
esis, to which we subscribe. There are various degrees of semantic transitivity,
but formal transitivity is only canonical or it is absent: low semantic transitivity
manifests itself in the use of intransitive structures alongside canonical transitive
structures (say, ‘A likes B’ alongside ‘B is pleasing to A’). Structures like (61) are
therefore not half-transitive but intransitive. Secondly, there is no reason to view
the configuration dat–acc as an unstable, transitional stage between the more
stable configurations dat–nom and nom–acc; nor is there any reason to view
the configurations dat–nom, dat–gen and dat–acc, viewed as one group, as
transitional stages on the way to the canonical nom–acc confuguration. In sev-
eral Germanic, Baltic and Slavonic languages all these configurations are stable
enough, being, in many cases, attested over a period of nearly a thousand years.
9. The debitive and voice
There are, as we have seen, no clear subject properties in the debitive construc-
tion, and it would be more precise to speak of diffuse grammatical relations. In
the context of discussions on valency splits we can thus qualify the debitive as an
inflectional form that changes not only the morphosyntactic encoding of noun
phrases but also the pattern of grammatical relations. Nevertheless the debitive
differs in this respect from voice operations, which typically carry over noun
phrases from one grammatical relation to another (object to subject, subject to

non-core argument) but do not serve to render grammatical relations diffuse.
Of course there is a connection between debitive and voice. The Academy
Grammar (1959: 620) states that passive debitive forms exist in principle, but are
exceedingly rare. No authentic examples from the literature are given. Indeed pas-
sive debitive forms seem to be a feature of Biblical and archaizing language, where
they are used to refer to fated or divinely ordained events:
(70) Katr-ai ielej-ai jā-top pildī-t-ai un
every-dat.sg.f valley-dat.sg deb-become fill-ppp-dat.sg.f and
katr-am kaln-am un pakaln-am nolīdzinā-t-am.
every-dat.sg.m mountain-dat.sg and hill-dat.sg level-ppp-dat.sg
‘Every valley shall be filled, and every mountain and hill shall be made low.’
(Revised 1965 Latvian Bible, Luke 3.5)
Such constructions are possible in the elaborated code because the morphology
allows them, but they do not occur in spontaneous speech. In fact the very dif-
fuseness of grammatical relations in the debitive construction makes passiviza-
tion superfluous. The passive enables suppression of the agent and usually, though
not always, also promotes the original object to subject position. When a debi-
tive construction is derived from a construction with an active verb, the original
subject becomes a datival quasi-subject that can be suppressed without further
morphosyntactic changes. The original object becomes a demoted intransitive
subject, i.e., a nominative-marked noun phrase with suspended subject proper-
ties. We have pointed out above that demoted intransitive nominative subjects
are characterized by recoverable subjecthood, i.e., the ability of functioning as
a fully-fledged subject when topicalized and occurring clause-initially. What is
normally achieved by the passive construction is achieved, in the case of the debi-
tive, by simply moving the demoted nominative subject to clause-initial position
and treating it as main topic, and suppressing the dative A argument reflecting
the agent. The demoted intransitive subject thereby becomes a full intransitive
subject and no passive morphology is needed.
With this versatility of the debitive construction in mind we may now attempt
to tackle, in a diachronic perspective, the question of the relationship between the
accusative marking of the B argument and its grammatical status. If the B argu-
ment were an object, it would have been amenable to a change of case-marking
from nominative to accusative. Such a process is quite conceivable, and it would
have been an analogue to Old Latin constructions with the gerundive. Alongside
a construction in which the gerundive appears in a copular construction with the
original object as grammatical subject, one also finds in pre-Classical Latin an
impersonal variety with an accusative (Ernout & Thomas 1959: 286).
(71) Agita-nd-ae sunt vigili-ae.Latin

keep-grdv-nom.pl.f be.prs.3pl watch-nom.pl
(72) Agita-nd-um est vigili-as.
keep-grdv-nom.sg.n be.prs.3sg watch-acc.pl
‘One must keep watch.’ (Plautus, Trinummus 4.2.27)
We could imagine this reanalysis having led to a reorganization of the whole con-
struction, which would have become impersonal (‘one has to…’), with accusative
objects in the case of transitive verbs. This is not what actually happened: the
construction with the accusative was abandoned, probably on the authority of the
writers of the classical period. But had it imposed itself, we could imagine, as the
standard construction, an expanded variety of (70) in the form
(73) Nobis agita-nd-um est vigili-as.
1pl.dat keep-grdv-nom.sg.n be.prs.3sg watch-acc.pl
‘We have to keep watch.’
This would provide us with a plausible scenario for a debitive variety with an ac-
cusative object. It is not excluded that in the Latvian dialects with consistent use
of the accusative for the B argument this is actually the correct interpretation.
This pattern is not typologically implausible because it is not, in this case, lexically
assigned but occurs as a result of a grammatical operation.
In many Latvian dialects, however, we observe the considerably more com-
plex situation described in the opening part of this article. As long as we consider
the debitive in isolation, as the grammatical construction it undoubtedly is, we
will be tempted to interpret it as involving a clear-cut assignment or reassignment
of grammatical relations, with the B argument as, say, a nominative object as in
Fennic infinitival constructions. But our view of grammatical relations with the
debitive must change from the moment we realize that the pattern occurring with
the debitive is also assigned lexically, or, to be more precise, both patterns – that
with the nominative and that with the accusative for the B argument – may be
assigned lexically, that the patterns are related, that they may even co-occur with
the same verb (the example iekāroties above), and that both are associated with
a specific configuration involving diffuseness of grammatical relations, with a
demoted intransitive subject outranked by a less-oblique datival quasi-subject.
These instances of lexical assignment entitle us to take a similar view of the debi-
tive construction. Pursuing the Latin analogy we would normally expect either
full subjecthood for B (vigiliae sunt agitandae) or full objecthood for B (agitan-
dum est vigilias). But the existence of the lexically assigned construction with a
datival quasi-subject and a demoted intransitive subject provided a third path
of development, and the debitive has evidently been drawn into the orbit of this
lexically assigned construction in part of the Latvian dialects and in the standard
language.
How the debitive will develop further is difficult to predict. We have described
the debitive construction as an instance of diffuse grammatical relations with a da-
tival quasi-subject and a nominative- or accusative-marked demoted intransitive
subject. This pattern need not remain stable: it might well shift to a pattern with
generalized accusative for the B argument, which we would then perhaps have to
recognize as a direct object. This would be an analogue to the Latin construction
Nobis agitandum est vigilias. But we could also imagine the debitive construction
splitting into two distinct constructions, of the kind we may probably assume to
have coexisted at one stage in pre-Classical Latin, viz. (67) and (68). It will cer-
tainly be interesting to follow the development of the debitive construction in the
course of the years to come.
10. The debitive and splits
In the literature on splits, the mood that has received most attention is the im-
perative, which shows deviant case marking in several languages: in Finnish, the
object of an imperative is in the nominative, presumably because the agent is as-
sumed to coincide with the addressee and the need for explicit marking of the
object is weakened. For the same reason in Hindi, ergative marking on the agent
is optionally dispensed with in the imperatival construction. These instances are
straightforwardly dealt with in terms of case marking, grammatical relations be-
ing assumed to be stable. Or rather, authors dealing with such splits and with the
typology of case marking in general tend to be agnostic about grammatical rela-
tions, evidently holding these to be language-specific notions, instead operating
with notions like A and P, which are partly defined in terms of thematic roles.
The discussion on case marking with the Latvian debitive could therefore also,
in principle, be integrated in a domain of research concentrating on case mark-
ing as seen as a means of coding thematic relations. But the notions of A and P,
as defined with reference to the transitive structure and distinguished from the
intransitive subject S, are actually not purely semantic but refer to the canoni-
cal syntactic realizations of the core thematic roles (cf. the discussion on this in
Haspelmath 2011). It is therefore not always possible to circumvent the issue of
grammatical relations, especially when these are non-canonical, as in the case of
the Latvian debitive. And another reason why the Latvian debitive is special is, of
course, that it collapses two argument structures, that of the modal predicate and
that of its complement. These two elements taken together set the Latvian debitive
apart from such classical cases of (broadly speaking) modal splits as the Finnish
imperatival construction or embedded infinitival clauses with impersonal modals
as illustrated in (26), (27). It would be wrong not to take into account the fact that
the datival argument in the debitive is not only an agent, but also a modality target
and, as such, an experiencer; as it would be wrong not to see that, even when ac-
cusative marking is used for B, the configuration dat–acc too strongly diverges
from the canonical transitive pattern for the labels A and P to be immediately
applicable.
Splits, then, are of different kinds. The split with the Latvian debitive is special
in that it is modality-driven, not mood-driven, with the consequences already
pointed out above. But its main interest probably lies in non-canonicity, which
should here be understood as diffuseness. As Sands and Campbell (2001: 279)
point out, there is nothing non-canonical about the nominative used to mark
objects of imperatives and infinitives in Finnish, because that’s simply the way to
mark such objects in the given constructions in Finnish. We could, in a similar
way, regard a nominative object with a Latvian debitive as canonical if it were not
for the diffuseness of grammatical relations which prevents us from identifying it
as an object. Diffuseness is therefore probably an important element of non-can-
onicity. It certainly is in the case of the debitive, where non-canonicity cannot be
claimed to be associated with lexical assignment.
Abbreviations
acc accusative nec necessitive

aux auxiliary neg negative
comp complementizer nom nominative
cvb converb pl plural
dat dative pot potential
deb debitive ppa past participle active
def definite ppp past participle passive
f feminine poss possessive
fut future ppp past passive participle
gen genitive prs present
grdv gerundive prtc particle
inf infinitive pst past
irr irrealis refl reflexive
loc locative rel relative pronoun
m masculine sg singular
References
Aikhenvald, Alexandra Y., Robert M. W. Dixon, & Masayuki Onishi (eds). 2001. Non-Canon-
ical Marking of Subjects and Objects [Typological Studies in Language 46]. Amsterdam/
Philadelphia: John Benjamins. DOI: 10.1075/tsl.46
Allen, Cynthia L. 1986. Reconsidering the history of like. Journal of Linguistics 22(2): 375–409.
Andrews, Avery D. 2001. Non-canonical A/S marking in Icelandic. In Non-Canonical Marking
of Subjects and Objects [Typological Studies in Language 46], Alexandra Y. Aikhenvald,
Robert M. W. Dixon and Masayuki Onishi (eds), 85–111. Amsterdam/Philadelphia: John
Benjamins.
Barðdal, Jóhanna. 2001. The perplexity of Dat-Nom verbs in Icelandic. Nordic Journal of Lin-
guistics 24: 47–70. DOI: 10.1080/033258601750266187
Barðdal, Jóhanna & Eythórsson, Thórhallur. 2006. Control infinitives and case in Germanic:
‘Performance error’ or marginally acceptable constructions? In Case, Valency and Transi-
tivity [Studies in Language Companion Series 77], Leonid Kulikov, Andrej Malchukov, &
Peter de Swart (eds), 147–177. Amsterdam/Philadelphia: John Benjamins.
Berg-Olsen, Sturla. 1999. A syntactic change in progress: The decline in the use of the non-
prepositional genitive in Latvian, with a comparative view on Lithuanian. MA thesis, Oslo
University.
Bergmane, Anna. 1959, 1962. Mūsdienu latviešu literārās valodas gramatika. I. Fonētika un
morfoloģija. II. Sintakse. Rīga: Latvijas PSR Zinātņu akadēmijas izdevniecība.
Bielenstein, August. 1863/1864. Die lettische Sprache nach ihren Lauten und Formen erklärend
und vergleichend dargestellt. I–II. Berlin: Ferdinand Dümmler.
Blake, Barry J. 2001. Case. 2nd edition. Cambridge: Cambridge University Press. DOI:
10.1017/CBO9781139164894
Blansitt, Edward L. 1984. Dechticaetiative and dative. In Objects:Towards a Theory of Gram-
matical Relations, Lund: Department of Scandinavian Languages, 2001.] Frans Plank (ed),
127–150. London & New York: Academic Press.
Boeder, Winfried. 1979. Ergative Syntax and Morphology in Language Change: The South Cau-
casian Languages. In Ergativity: Towards a Theory of Grammatical Relations, Frans Planck
(ed), 435–480. London: Academic Press.
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. DOI: 10.1007/978-94-009-4522-7
Bybee, Joan, Perkins, Revere & Pagliuca, William. 1994. The Evolution of Grammar. Tense, As-
pect and Modality in the Languages of the World. Chicago & London: The University of
Chicago Press.
Comrie, Bernard. 1975. The antiergative: Finland’s answer to Basque. Papers from the eleventh
regional meeting, Chicago Linguistic Society 11, 112–121.
Culicover, Peter W. & Jackendoff, Ray. 2005. Simpler Syntax. Oxford: University Press. DOI:
10.1093/acprof:oso/9780199271092.001.0001
De Haan, Ferdinand. 2006. Typological approaches to modality. In The Expression of Modality,
William Frawley (ed), 27–69. Berlin/New York: Mouton de Gruyter.
Eide, Kristin Melum. 2005. Norwegian Modals [Studies in Generative Grammar 74]. Berlin-
New York: Mouton de Gruyter.
Endzelīns, Jānis. 1901/1975. Ursprung und Gebrauch des lettischen Debitivs. Bezzenbergers
Beiträge 26: 66–74 = id., Darbu izlase I, Rīga: Zinātne, 143–150.
Endzelīns, Jānis [J. Endzelin]. 1923. Lettische Grammatik. Heidelberg: Carl Winter.
Endzelīns, Jānis. 1932/1980. Dažādas valodas kļūdas [Various language mistakes]. 3rd edn.
Rīga: A. Gulbis. Reprinted Darbu izlase 3.2, Rīga: Zinātne, 9–45.
Endzelīns, Jānis & Mühlenbachs, Kārlis. 1907. Latviešu gramatika. Rīga: K. J. Zichmanis.
Ernout, Alfred & Thomas, François. 1959. Syntaxe latine. 2e éd. Paris: Klincksieck.
Fennell, Trevor G. 1973. The subject of Latvian verbs in the debitive mood. In Baltic Literature
and Linguistics, Arvids Ziedonis, Jaan Puhvel, Rimvydas Šilbajoris, & Mardi Valgemäe
(eds), 213–221. Columbus, Ohio: Association for the Advancement of Baltic Studies.
Hansen, Björn. 2000. The German modal verb müssen and the Slavonic languages–the recon-
struction of a success story. Scando-Slavica 46: 78–92. DOI: 10.1034/j.1600-082X.2000.
d01-7.x
Hansen, Björn. 2001. Das slavische Modalauxiliar. Semantik und Grammatikalisierung im Rus-
sischen, Polnischen, Serbischen, Kroatischen und Altkirchenslavischen. München: Verlag
Otto Sagner.
Harris, Alice C. 1981. Georgian Syntax. A Study in Relational Grammar [Cambridge Studies in
Linguistics 33]. Cambridge: University Press.
In Non-Canonical Marking of Subjects and Objects [Typological Studies in Language 46],
Alexandra Y. Aikhenvald, Robert M. W. Dixon, & Masayuki Onishi (eds), 53–83. Amster-
Haspelmath, Martin. 2010. The Behaviour-before-Coding Principle in syntactic change. In Es-
sais de Typologie et de Linguistique Générale: Mélanges Offerts à Denis Creissels, Franck
Floricic (ed), 541–554. Lyon: Presses Universitaires de l’École Normale Supérieure.
Holvoet, Axel. 1998. Notes on the rise and grammaticalization of the Latvian debitive. Linguis-
tica Baltica 7: 101–118.
Holvoet, Axel. 2005. Agreement strategies in infinitival clauses in Baltic. In Prace Bałtystyczne 2.
Język, literatura, kultura [Baltic Studies 2. Language, Literature, Culture], Ona Vaičiulytė-
Romančuk & Norbert Ostrowski (eds), 31–41. Warszawa: Uniwersytet Warszawski.
Holvoet, Axel. 2007. Mood and Modality in Baltic. Kraków: Wydawnictwo Uniwersytetu
Jagiellońskiego.
Holvoet, Axel. 2013. Obliqueness, quasi-subjects and transitivity in Baltic and Slavonic. In The
Diachronic Typology of Non-prototypical Subjects, Ilja Seržants & Leonid Kulikov (eds),
257–282. Amsterdam & Philadelphia: John Benjamins.
Holvoet, Axel. Forthcoming. Non-canonical subjects in Latvian: An obliqueness-based ap-
proach. In Contemporary Approaches to Baltic Linguistics, Peter M. Arkadiev, Axel
Holvoet, & Björn Wiemer (eds). Berlin: Mouton de Gruyter.
Hopper, Paul. 1991. On some principles of grammaticalization. Approaches to Grammaticaliza-
tion vol. 1 [Typological Studies in Language 19.1], Elizabeth Closs Traugott & Bernd Heine
Hopper, Paul & Thompson, Sandra. 1980. Transitivity in grammar and discourse. Language 56:
251–299.
Keenan, Edward L. 1976. Towards a universal definition of ‘subject’. In Subject and Topic,
Charles N. Li (ed), 303–333. New York: Academic Press.
Kittilä, Seppo. 2006. Object-, animacy- and role-based strategies. A typology of object marking.
Studies in Language 30(1): 1–32. DOI: 10.1075/sl.30.1.02kit
Lightfoot, David. 1974. The Diachronic Analysis of English Modals. In Historical Linguistics.
Proceedings of the First International Conference on Historical Linguistics, Edinburgh, 2nd-
7th Sept., 1973. Vol. I: Syntax, Morphology, Internal and Comparative Reconstruction,
John Anderson & Charles Jones (eds), 219–249. Amsterdam: North Holland Publishing
Company.
Malchukov, Andrej, Haspelmath, Martin, & Comrie, Bernard. 2010. Ditransitive constructions:
A typological overview. In Studies in Ditransitive Constructions. A Comparative Handbook,
Andrej Malchukov, Martin Haspelmath, & Bernard Comrie (eds), 1–65. Berlin: Mouton
de Gruyter.
Malchukov, Andrej L. & de Hoop, Helen. 2011. Tense, aspect, and mood based differential case
marking. Lingua 121: 35–47. DOI: 10.1016/j.lingua.2010.07.003
Narrog, Heiko. 2010. Voice and non-canonical case marking in the expression of event-ori-
ented modality. Linguistic Typology 14: 71–126. DOI: 10.1515/lity.2010.003
Nau, Nicole. 1998. Latvian [Languages of the World, Materials 217]. München & Lancaster:
Lincom Europa.
Noonan, Michael. 2007. Complementation. In Language Typology and Syntactic Description.
Vol. II. Complex Constructions. 2nd edition, Timothy Shopen (ed), 52–150. Cambridge:
Cambridge University Press.
Onishi, Masayuki. 2001. Introduction. In Non-Canonical Marking of Subjects and Objects
[Typological Studies in Language 46], Alexandra Y. Aikhenvald, Robert M. W. Dixon, &
Masayuki Onishi (eds), (eds) 2001, 1–51. Amsterdam/Philadelphia: John Benjamins.
Palmer, Frank R. 2001. Mood and Modality. 2nd edn. Cambridge: University Press. DOI:
10.1017/CBO9781139167178
Perlmutter, David M. & Postal, Paul P. 1983. Some proposed laws of basic clause structure. In
Studies in Relational Grammar, David M. Perlmutter (ed), 81–128. Chicago & London:
University of Chicago Press.
Pollard, Carl J. & Sag, Ivan A. 1994. Head Driven Phrase Structure Grammar. Chicago/London:
University of Chicago Press.
Prellwitz, Walther. 1904. Zur Entstehung des lettischen Debitivs. Bezzenbergers Beiträge 28:
319.
Primus, Beatrice. 1999. Cases and Thematic Roles. Ergative, Accusative and Active. Tübingen:
Niemeyer. DOI: 10.1515/9783110912463
Rappaport, Gilbert C. 1986. On anaphor binding in Russian. Natural Language and Linguistic
Theory 4(1): 97–120. DOI: 10.1007/BF00136266
Rounds, Carol. 2001. Hungarian. An Essential Grammar. London/New York: Routledge.
Sands, Kristina & Campbell, Lyle. 2001. Non-canonical subjects and objects in Finnish. In
Non-Canonical Marking of Subjects and Objects [Typological Studies in Language 46],
Alexandra Y. Aikhenvald, Robert M. W. Dixon, & Masayuki Onishi (eds), 251–305.
Amsterdam/Philadelphia: John Benjamins.
Sigurðsson, Halldór Ármann. 2003. Case: abstract vs. morphological. In New Perspectives on
Case Theory [CSLI Lecture Notes 156], Ellen Brandner & Heike Zinsmeister (eds), 223–
267. Stanford, CA: Center for the Study of Language and Information.
Sigurðsson, Halldór Ármann. 2004. Icelandic non-nominative subjects. Facts and impli-
cations. In Non-Nominative Subjects. Vol. 2 [Typological Studies in Language 61], Peri
Bhaskararao & Karumuri Venkata Subbarao (eds), 137–159. Amsterdam/Philadelphia:
John Benjamins.
Timberlake, Alan. 1974. The Nominative Object in Slavic, Baltic, and West Finnic. München:
Verlag Otto Sagner.
Woolford, Ellen. 2003. Burzio’s generalization, markedness and locality constraints on nomina-
tive objects. In New Perspectives on Case Theory [CSLI Lecture Notes 156], Ellen Brandner
& Heike Zinsmeister (eds), 299–327. Stanford, CA: Center for the Study of Language and
Information.
tions: the Icelandic passive. Natural Language and Linguistic Theory 3(4): 441–483. DOI:
10.1007/BF00133285
Alternations in argument realization
and problematic cases of subjecthood
in Lithuanian
Kristina Lenartaitė-Gotaučienė
Klaipėda University / Vilnius University
The paper focuses on subject properties in Lithuanian sentences containing

verbs participating in the swarm alternation. In neither variant of the Lithu-
anian swarm alternation can either of the two arguments (i.e., the locational and
the non-locational argument) be considered a prototypical subject, i.e., be said
to display properties on the lexical (semantic), grammatical, and discourse lev-
els of representation that are typical of a prototypical subject. The discrepancies
between subject properties on distinct levels of representation are triggered by a
set of discourse-pragmatic and semantic features characteristic of the swarm al-
ternation. One of the basic discourse-pragmatic features characteristic of some
of its variants is that they function as presentational constructions and pragmat-
ically their subjects are not construed as topics but as sentence foci. Subjects of
presentationals are obligatorily marked by a focus pitch accent. This property of
presentationals conditions a spread of subject properties over the (pro)nominal
constituents of a clause. Other semantic-pragmatic properties contributing both
to alternation in subject realization and discrepancies between subject proper-
ties on distinct levels of representation are: (i) the obligatory requirement of a
location participant on the lexical-semantic and discourse-pragmatic levels of
representation; (ii) the pragmatic requirement for verbs displaying presenta-
tional constructions to be informationally light in context; (iii) a decrease in
agentivity properties of agent arguments triggered by the discourse-pragmatic
function of presentationals and realized on the lexical-semantic and, in some
instances, the (morpho)syntactic level of representation; (iv) the semantic-prag-
matic requirement for non-locational arguments of the swarm alternation to be
indefinite plurals or mass terms rather than singular NPs and to denote uniden-
tifiable referents.
138 Kristina Lenartaitė-Gotaučienė
1. Introduction1
The phenomena of so-called argument alternation, involving a single set of ar-

guments but allowing variations in their syntactic realization (for example, the
instances of dative alternation, locative alternation, etc. widely discussed in the
linguistic literature) are a challenge for all theoretical frameworks. According
to Levin and Rappaport Hovav (2005: 194f.), argument alternations raise three
broad and interrelated theoretical questions: (i) do the variants, despite their ob-
vious relatedness and near-paraphrase meaning relation, really differ in meaning,
(ii) what mechanisms give rise to variants, i.e., should the variants be treated as
derivationally linked syntactically and/or semantically, or should they be consid-
ered to be associated with a single verb root but with two distinct event types/ver-
bal ‘profiles’ or with two independent constructions, (iii) what determines which
variant is chosen in a certain context.
The paper focuses on a particular type of argument alternations in Lithu-
anian (1) similar to the so-called ‘swarm alternation’ in English (2). We inves-
tigate questions (i) and (iii), and are not directly concerned with the issue of
whether the variants involved in the swarm alternation in Lithuanian should be
treated as derivationally linked or whether they should be viewed as independent
constructions.
(1) a. Filharmonij-oje knibždėjo įvairiausi-ų žmoni-ų. (CCLL)
philharmonic-loc.sg swarm.pst.3 various-gen.pl people-gen.pl
‘All kinds of people were swarming in the concert hall.’
b. Filharmonij-a knibždėjo (nuo) įvairiausi-ų
philharmonic-nom.sg swarm.pst.3 (with) various-gen.pl
žmoni-ų.
people-gen.pl
‘The concert hall was swarming with all kinds of people.’
(2) a. Bees2 swarm in the garden.
b. The garden swarms with bees. (Dowty 2000: 112)
1. At various stages of the preparation of this paper, I have benefited from insightful com-
ments and stimulating discussion by Axel Holvoet, Rolandas Mikulskas, Peter Arkadiev,
Natalia Zaika, and Valgerður Bjarnadóttir. I am also greatly obliged to an anonymous reviewer
for very helpful comments, and I would like to thank Wayles Browne for his careful reading of
the manuscript, his useful observations and the many improvements he has made to my Eng-
lish text. The shortcomings that still remain are all my own.
2. Here and throughout the paper, boldfacing in the examples is mine – KLG.
Alternations in argument realization in Lithuanian 139
The paper looks at the discourse-pragmatic and semantic properties that con-
tribute to the fact that in neither of the variants of Lithuanian swarm alternations
could either of the two constituents be considered a prototypical subject. Actually,
it is a characteristic property of verbs displaying argument alternations that those
of their arguments that may alternatively be realized as a subject or object of a
clause do not have the properties on distinct levels of representation that are typ-
ical of prototypical subjects or objects. The realization of subjecthood in Lithua-
nian swarm alternations is, then, the main topic of the paper. Taking into account
Keenan’s (1976) influential study offering a multifactorial definition of the con-
cept of ‘subject’, it is obvious that the ‘subject’ notion necessarily involves certain
properties of different domains, viz. functional properties, semantic properties,
coding properties, behavioural properties, and structural properties. In Lithua-
nian, all the variants of the swarm alternation demonstrate discrepancies between
properties of distinct domains that typically coincide when we have instances of
the so-called ‘prototypical subject’.3 For example, in sentence (1a) the event par-
ticipant (people) is animate, agentive, volitionally causing the event and, from the
point of view of thematic relations of the NP constituents, the most appropriate
candidate for subjecthood in the given predicate-argument structure. However,
this constituent lacks the other properties characteristic of subjects. It has neither
the nominative case-marking nor the topical status typical of a subject. In (1b)
the location (philharmonic) may be considered the subject of the clause due to its
nominative case and topical status. However, taking into account the occurrence
of an agentive participant in the event structure, the locational subject reveals a
departure from the principle of assigning subjecthood to the highest theta-role.
In this paper, for the purpose of identifying the ‘prototypical subject’ and
explaining various instances of departure from it in realizing subjecthood in
Lithuanian swarm alternations, we adopt Livitz’s (2006) approach. According to
Livitz, ‘subjecthood’ can be understood as prominence on three linguistic lev-
els: the lexical (semantic), the syntactic, and the pragmatic. Consequently, there
are three primitives of subjecthood: thematic subjecthood, grammatical subject-
hood, and discourse subjecthood. A prototypical subject is the most prominent
at each level, thus combining all three primitives. Non-prototypical instances of
subjecthood can be explained as a dissociation between some or all of these three
types of subjects. Livitz proposes a primitives-based subjecthood analysis mainly
for the purpose of explaining non-nominative subjects in Russian. As mentioned
3. Actually, the notion ‘prototypical subject’ and its characteristic properties are universal
and ‘prototypical’ only to a certain degree. Establishing a universal definition of subjecthood is
made impossible by the variety of behavioural and other properties associated with the subject
cross-linguistically.
above, Lithuanian swarm alternations comprise various instances of dissociation

between the three primitives of subjecthood. For example, in (1a) the constituent
people satisfies the requirements of thematic subjecthood, but the NP is neither a
grammatical subject (grammatical/syntactic subjecthood is determined by sub-
ject-verb agreement, and, consequently, in languages like Lithuanian, by nomi-
native-case marking), nor a discourse subject (i.e., a topic). In (1b), the location
subject satisfies the requirements for both grammatical and discourse subject-
hood, but not for thematic subjecthood. The idea of splitting up the primitives of
subjecthood provides a basis for explaining various departures from the charac-
teristic properties of subjects in the Lithuanian swarm alternation.
Furthermore, the discourse-pragmatic properties characteristic of Lithuanian
swarm alternations can also give rise to discrepancies between the primitives of
subjecthood. First of all, it is important to account for the discourse function of
some variants of the Lithuanian swarm alternation: some alternating variants (as
in (1a), (3a), and (3c)) serve as ‘sentence-focus constructions’. Alternative terms
for ‘sentence-focus constructions’ used in the literature are: ‘news sentence’, ‘neu-
tral description’, ‘all-new utterance’, ‘thetic sentence’, ‘event-reporting sentence’
(cf. Lambrecht 2000: 623; 1994: 138 for the references). According to Lambrecht
(2000: 623), the overriding function of the sentence-focus (SF) constructions is
presentational, i.e., “SF constructions serve either to introduce a discourse-new
referent or to introduce an event which involves a referent which is discourse-
new or contextually construed as such”. The two subtypes of SF are instances of
so-called ‘thetic’ sentences; these only recognize or reject some judgment material
without predicating this judgment of some independently recognized subject, i.e.,
the basic logical structure of thetic judgments is “A is” or “A is not”. In contrast,
the ‘categorical’ (or topic-comment) judgment, which is typically expressed in
the subject-predicate sentence type, involves both the act of recognizing a sub-
ject and the act of affirming or denying what is expressed about the subject; the
basic logical structure of categorical judgments is “A is B” or “A is not B” (cf.
Lambrecht 1994: 139). Thetic sentences are considered to be marked with respect
to categorical (or topic-comment) sentences. The subtype of event-reporting
thetic sentences (e.g., My dog got lost4 as an answer to the question “What hap-
pened?”) is sometimes also referred to by the term ‘event-central thetic sentences’.
Consequently, sentences introducing a discourse-new referent (e.g., Once upon a
time there was a king) are sometimes referred to as ‘entity-central’, though they
are mostly known as ‘presentational’ sentences. Since the variants of the Lithu-
anian swarm alternation, as in (1a), (3a), and (3c), are entity-central, i.e., they
4. Here and throughout the paper, small capitals indicate the focus accent of a sentence.
refer to the presence of a new or hitherto unnoticed entity, they will henceforth
be called ‘presentational’ constructions in this paper (almost all the instances of
presentational constructions involved in Lithuanian swarm alternations are en-
tity-central except for the two attested examples with the verb knibždėti (‘swarm’)
that are event-central presentationals).
Turning again to the question of dissociation between the primitive proper-
ties of subjects in the Lithuanian swarm alternation, two important consequences
of the functioning of the given sentences as discourse-pragmatic presentational
constructions should be accounted for: (i) presentationals are characterized by
“the absence of a topic relation between the proposition and that argument which
functions as the topic in the categorical counterpart” (ibid.: 144); that is, in pre-
sentational constructions, the grammatical subject is not construed as a topic, but
as a focus of the clause, (ii) presentationals “are often characterized by a non-ca-
nonical distribution of subject properties over the (pro)nominal constituents of a
sentence” (Gast & Haas 2011: 128). The two consequences are closely interrelated.
First of all, for the sentence to be interpreted as a presentational construction, the
subject needs to be marked in such a way that it will not be construed as topic.
Therefore Lambrecht (2000: 624f.) proposes that the marking of the presentation-
al constructions “crucially involves ‘detopicalization’ of the subject constituent,
i.e., cancellation of those grammatical features that are conventionally associated
with the subject as the topic of a sentence”. One of the ways of achieving non-top-
ical construal is associating the subject constituent with grammatical properties
that are conventionally characteristic of focus arguments. Since in predicate-fo-
cus constructions (i.e., pragmatically topic-comment structured propositions, in
which the subject is the topic and the predicate expresses information about this
topic) the unmarked focus argument is the object, the coding of the non-topical
subject of presentationals may be achieved by grammatical and pragmatic fea-
tures normally associated with the object of a predicate-focus sentence. In many
languages the NP introducing a new discourse referent exhibits properties of both
subjects and objects (cf. Lambrecht 2000 for references). Thus, in presentational
constructions the distribution of subject properties over the (pro)nominal con-
stituents of a sentence is triggered by the need to ‘detopicalize’ the subject for the
given discourse-pragmatic purposes. All the variants of the Lithuanian swarm
alternation are treated as pragmatically unmarked only when their non-locational
constituent (i.e., people in (1a–b) and lilac in (3a–c)) bears a focus accent.
(3) a. Motin-a siuva, lang-as praver-t-as,

mother-nom.sg sew.prs.3 window-nom.sg open-ppp-nom.sg
darž-e alyv-os kvepia <…>. (CCLL)
garden-loc.sg lilac-nom.pl smell.prs.3
‘Mother is sewing, the window is open, in the garden there is a smell of
lilac.’
b. Darž-as alyv-omis kvepia.
garden-nom.sg lilac-ins.pl smell.prs.3
‘The garden smells of lilac.’
c. Darž-e alyv-omis kvepia.
garden-loc.sg lilac-ins.pl smell.prs.3
‘In the garden, there is a smell of lilac.’
However, not all variants of the Lithuanian swarm alternation are presentation-
als on the level of discourse-pragmatic representation. Sentences with a topical
location-subject, as in (1b), (3b), function as predicate-focus constructions and
express a pragmatically topic-comment structured proposition, while sentences
(1a), (3a), and (3c) are presentationals. Sentences (1b) and (3b) demonstrate the
coincidence of the grammatical and discourse properties typically characteristic
of subjects but a departure from typical thematic subject properties. The prag-
matic structure of (3a) is construed in such a way that the grammatical (i.e.,
nominative-marked) subject of the clause is not a topic. The variant illustrated in
(3c), in which neither participant has nominative marking, is represented only by
verbs denoting smell emission. The given variant explicitly realizes a non-canoni-
cal spread of subject properties over two nominal constituents, where neither of
the two constituents may be assigned grammatical subjecthood (however, both of
them have other characteristic properties of subjecthood in distinct domains of
representation).
These discourse-pragmatic properties are interrelated with certain seman-
tic properties of predicates and their arguments. First of all, verbs displaying the
swarm alternation denote not merely an activity/process performed by a certain
event participant, but rather an activity/process taking place in a certain location.
For example, in (1a–b), (2a–b) and (3a–c), it is not the ability of the given event
participants to swarm or emit a smell but the presence of this activity/process in a
certain location that is predicated by the verbs. Thus, location participants are ob-
ligatorily required in the predicate-argument structure of these verbs in order to
realize the swarm alternation. The semantics of swarm-alternation verbs therefore
correlates with the fundamental property of discourse-pragmatic presentationals
“to introduce the NP referent into the discourse world of the interlocutors by as-
serting its presence in a given location” (Lambrecht 1994: 179). Another semantic
property of NPs designating non-locational participants5 in the swarm alterna-

tion (i.e., those whose presence is introduced by discourse-pragmatic presenta-
tionals) is the requirement for their referents to be expressed as indefinite plurals
or mass terms rather than singular NPs. This semantic property is also in ac-
cordance with the fact that “in many languages certain constructions expressing
thetic propositions are restricted to, or at least strongly preferred for, “indefinite”
NPs, i.e. NPs with unidentifiable referents” (ibid.: 143). And finally, the discourse-
pragmatic functions of presentationals, i.e., the introduction of a discourse-new
referent correlating with the meaning of the verb denoting the presence of an
activity/process in a given location, in fact leads to a situation in which agent
arguments of the swarm alternation can be said to be agentive or “pseudo-agen-
tive” to a certain degree. In other words, the semantic properties of the agent
participant become irrelevant when on the discourse-pragmatic level of repre-
sentation a sentence functions as a presentational construction. For example, in
(1a) and (2a) the semantic properties of the agent participants, i.e., the volitional
involvement and/or responsibility for carrying out the swarming event etc., are ir-
relevant for the discourse-pragmatic function of the clause. Thus, in order to give
an appropriate explanation of non-prototypical subject realization in Lithuanian
swarm alternations, it is important to take into account all the above-mentioned
discourse-pragmatic and semantic factors.
The problem and the assumptions of the research having been introduced,
the structure of the paper remains to be presented. Section 2 provides an overview
of various instances of the swarm alternation in Lithuanian. Section 3 contains
an analysis of the semantic factors of the swarm alternation which both correlate
with the discourse-pragmatic function of presentationals and are relevant for the
realization of subjecthood. Section 3.1 gives an analysis of the semantic properties
of predicates and their arguments characteristic of all instances of the Lithuanian
swarm alternation, while Section 3.2 explores one particular group, smell-emis-
sion verbs, and points out their semantic properties relevant for realizing the
construction in which neither NP has the nominative marking typical for gram-
matical subjects. Section 4 gives an analysis of the discourse-pragmatic function
of the presentational construction, characteristic of some variants of the Lithu-
anian swarm alternation, and highlights the relevant consequences of this dis-
course-pragmatic function for realizing subjecthood. Section 5, dealing with the
evidence for discourse-pragmatic and semantic levels of representation, explores
non-prototypical subjects in the Lithuanian swarm alternation. The analysis of
non-prototypical subjects is based on the idea of splitting up ‘subjecthood’ into
5. Since such arguments of the swarm alternations may display several thematic roles, for
convenience they will be referred to as non-locational participants/arguments.
primitives on three linguistic levels, i.e., lexical (semantic), syntactic and prag-
matic, and explains the non-prototypical instances of subjecthood in the Lithu-
anian swarm alternation as various instances of dissociation between three types
of primitives of subjecthood.
2. Overview of Lithuanian swarm alternations
The phenomenon most widely known as the swarm alternation has been much
commented on in linguistic literature (cf. Fillmore 1968: 48–49; Anderson 1971;
Jackendoff 1990: 178–179; Levin 1993: 53–53, 233–238 among others), yet only
a few comprehensive accounts of swarm alternations in different languages have
been given (for English cf. Salkoff 1983; Dowty 1999, 2000, 2001; Rowlands 2002;
Muñoz 2011; for Czech cf. Fried 2005; for Dutch cf. Hoeksema 2008; for Lithu-
anian cf. Lenartaitė 2010: 132–146). Most of these studies take for granted that a
description of the phenomenon in a given language should start with exploring
the semantics of the verbs and establishing the semantic classes of verbs involved
in the alternation. As for the present research, verb semantics is taken into ac-
count at least in identifying the specific semantic properties of smell-emission
verbs which are different from other verb classes, and it seems reasonable to di-
vide swarm-alternation verbs into semantic classes. Lithuanian swarm-alterna-
tion verbs fall into the following semantic classes:
i. Verbs denoting (multidirectional) movement of entities or substances:6

knibždėti (‘teem, swarm’), srūti (‘stream’), plūsti (‘flow, pour’) etc., cf. sentenc-
es (4a–b) and also (1a–b).
ii. Verbs denoting sound emission: aidėti (‘echo’), skambėti (‘sound, resound’),
ūžti (‘buzz’) etc., cf. (5a–b).
iii. Verbs denoting a kind of light emission or presence of a color in a given loca-
tion: spindėti (‘shine, glow’), blizgėti (‘glister, glitter’), raudonuoti (‘redden, to
be red’), geltonuoti (‘yellow, to be yellow’) etc., cf. (6a–b).
iv. Verbs denoting smell emission: kvepėti (‘smell, scent’), smirdėti (‘stink’), dvok-
ti (‘stink, reek’) etc., cf. (7a–c) and also (3a–c).
v. Verbs with the prefix pri-, denoting massive (usually directed) movement to
some location, occupying it completely or spreading over it to a relatively
large extent: privažiuoti (‘arrive massively’), pribėgti (‘come running massive-
ly’), pribėgti (‘flow in, run (about liquid)’), privarvėti (‘drop in, dribble in’),
prisirinkti (‘gather massively’) etc., cf. (8a–b).
6. After Levin and Rappaport Hovav (1995: 91), the given semantic class (like the correspond-
ing class in English) may be identified as “verbs of substance emission”.
(4) a. Prakait-as plūdo kakt-omis, nugar-omis.7

sweat-nom.sg flow.pst.3 forehead-ins.pl back-ins.pl
‘Sweat was pouring down foreheads and backs.’
b. Kakt-os, nugar-os plūdo prakait-u.
forehead-nom.pl back-nom.pl flow.pst.3 sweat-ins.pl
‘Foreheads and backs were flowing with sweat.’
(5) a. Šeštadienį gatv-ėse skambėjo muzik-a.8
Saturday street-loc.pl resound.pst.3 music-nom.sg
‘On Saturday music resounded in the streets.’
b. Šeštadienį gatv-ės skambėjo nuo muzik-os.
Saturday street-nom.pl resound.pst.3 with music-gen.sg
‘On Saturday the streets resounded with music.’
(6) a. Obelys jau pražydusios, ir kregžd-ės
apple-tree.nom.pl already blossom.ppa.nom.pl and swallow-nom.pl
sugrįž-usi-os, piev-ose geltonuoja pien-ės9 <…>
return-ppa-nom.pl meadow-loc.pl yellow.prs.3 dandelion-nom.pl
‘The apple-trees are already in blossom, and the swallows are back, [and]
dandelions shine yellow in the meadows.’
b. Piev-os geltonuoja nuo pieni-ų.
meadow-nom.pl yellow.prs.3 with dandelion-gen.pl
‘The meadows are yellow with dandelions.’
(7) a. Ventiliacij-os ang-oje dvokė
ventilation-gen channel-loc.sg stink.pst.3
pelėsi-ai (buvo bjauru kvėpuo-ti).10
mould-nom.pl be.pst.3 disgusting breathe-inf
‘In the ventilation channel there was a stench of mould, the air was foul
to breathe.’
b. Ventiliacij-os ang-a dvokė pelėsi-ais.
ventilation-gen channel-nom.sg stink.pst.3 mould-ins.pl
‘The ventilation channel stank of mould.’
c. Ventiliacij-os ang-oje dvokė pelėsi-ais.
ventilation-gen channel-loc.sg stink.pst.3 mould-ins.pl
‘In the ventilation channel, there was a stench of mould.’
7. http://eia.libis.lt:8080/archyvas/viesas/20110131124538/http://www.culture.lt/satenai/
?leid_id=679&kas=straipsnis&st_id=2275
8. http://www.jurbarkiskis.lt/index.php/kultura/6774-etadien-gatvse-skambjo-muzika
9. http://mano.zebra.lt/gyvenimai/kwakwa/dienorastis/205265/
10. http://mano.zebra.lt/gyvenimai/~/klubai/th_ydomios_istorijos/79604/
(8) a. Į sal-ę prisirinko žmoni-ų.

into hall-acc.sg gather.pst.3 people-gen.pl
‘People have gathered into the hall.’
b. Sal-ė (pilnutėl-ė) prisirinko (žmoni-ų).11
hall-nom.sg full-nom.sg gather.pst.3 people-gen.pl
‘The hall filled with people.’
The first four semantic classes (i)–(iv) of the Lithuanian swarm alternation are
nearly the same as those distinguished by Dowty (1999, 2001) for English in-
stances of the swarm alternation, cf. examples (9–12).
(9) a. Snails are crawling in the garden.
b. The garden is crawling with snails. (Dowty 2001)
(10) a. Music resounded in the hall.
b. The hall resounded with music. (Dowty 2001)
(11) a. Fireflies glowed in the field.
b. The field glowed with fireflies. (Dowty 2001)
(12) a. Garlic reeked on his breath.
b. His breath reeked with garlic. (Dowty 2001)
According to Hoeksema (2008), the set of swarm-alternation predicates in Dutch

is also roughly similar to that of English. However, Lithuanian and English also
exhibit differences in this respect. In English, Dowty (1999, 2001) distinguishes
one more class, i.e., verbs indicating degree of occupancy/abundance: abound, be
rich, rife, rampant (with). In Lithuanian, these verbs do not allow the abundantly
occurring and location-occupying event participant to surface as a subject. On
the other hand, Lithuanian has a group of verbs not represented among the se-
mantic classes of English swarm verbs, viz. class (v). Verbs with the prefix pri-,
denoting massive movement of entities or a substance to a given location, thus
occupying it completely or spreading over it in relatively large quantities, are the
largest class of swarm-alternation verbs in Lithuanian. Semantically the group is
very similar to class (i), as the verbs of both classes (i) and (v) denote movement
of entities or a substance. However, group (v) should be set apart because of the
semantic and coding properties associated with the verbal prefix pri-. Almost all
intransitive verbs denoting movement and semantically compatible with the pre-
fix pri- display the swarm alternation. It is precisely the prefix pri- that provides
the semantic feature of complete occupation of a given location or dispersion over
it. Pri- is also responsible for the obligatoriness of a location participant in the
11. http://www.zodynas.lt/terminu-zodynas/p/pilnutelis
argument structure of the verbs. Moreover, predicates with pri-

predicate-
do not allow entities/substance event participants to surface as nominative-
marked subjects – these are obligatorily marked with the partitive genitive, as in
(13a)–(14a).
(13) a. Siauči-ant audr-ai buvo nuplėš-t-a
rage-cvb.prs storm-dat.sg be.pst.3 rip.off-ppp-nom.sg.f
stog-o dang-a ir pro stog-ą į
roof-gen.sg covering-nom.sg.f and through roof-acc.sg into
kambar-į pribėgo vanden-s12 [/*vand-uo].
room-acc.sg flood.pst.3 water-gen.sg [water-nom.sg]
‘During the storm the roofing was ripped off and through the roof water
has flooded into the room.’
b. Kambar-ys pribėgo vanden-s.
room-nom.sg flood.pst.3 water-gen.sg
‘The room flooded with water.’
(14) a. Į kapin-es priėjo žmoni-ų / *žmon-ės.
to cemetery-acc.pl come.pst.3 people-gen.pl / people-nom.pl
‘A lot of people came to the cemetery.’
b. Nors kapin-ės piln-os priėjo žmoni-ų,
although cemetery-nom.pl full-nom.pl come.pst.3 people-gen.pl
bet dar daugiau jų liko stovėti gatv-ėse.13
but even more they.gen.pl leave.pst.3 stand.inf street-loc.pl
‘Although the cemetery was crowded with people, even more remained
standing in the streets.’
Some class (i) verbs denoting movement, e.g., knibždėti (‘swarm’), kušėti (‘teem’),
more frequently display partitive-genitive agent subjects, as in (15b), (16), but
nominative-marked agent subjects are also possible, as in (15a). The meaning of
the partitive genitive is compatible with the requirement for non-locational argu-
ments in the swarm alternation to be expressed as indefinite plurals or mass terms
rather than as singular NPs (see Section 3.1 for more details).
(15) a. Ar Jums malon-u maty-ti šunis,
if you.dat.sg pleasant-n see-inf dog.acc.pl
kuri-ų kail-yje knibžda blus-os?!(CCLL)
which-gen.pl fur-loc.sg teem.prs.3 flea-nom.pl
‘Are you pleased to see dogs whose fur is teeming with fleas?!’
12. http://nepriklausomasekspertas.lt/NT/?q=node/16
13. http://www.plienosparnai.lt/page.php?622
b. Šuns kail-yje knibžda blus-ų.

dog.gen.sg fur-loc.sg swarm.prs.3 flea-gen.pl
‘Fleas are swarming in the dog’s fur.’
(16) Ten tarybini-ais laik-ais paprastai knibždėte
there Soviet-ins.pl times-ins.pl usually swarm.cvb
knibždėjo šnip-ų bei KGB agent-ų <…>.14
swarm.pst.3 spy-gen.pl and KGB agent-gen.pl
‘In Soviet times it [the restaurant] used to be swarming with spies and KGB
agents.’
In Lithuanian, smell-emission verbs display three variants of the swarm alterna-

tion, i.e., the construction with source/smell participant subject, as in (17a), the
construction with locative subject, as in (17b), and the construction in which nei-
ther NP has nominative marking, as in (17c).
(17) a. Lauk-e kvepėjo alyv-os.
outdoors-loc.sg smell.pst.3 lilac-nom.pl
‘Outdoors lilacs smelled.’
b. Lauk-as kvepėjo alyv-omis.
outdoors-nom.sg smell.pst.3 lilac-ins.pl
‘[The air] outdoors smelled of lilac.’
c. Lauk-e kvepėjo alyv-omis. (CCLL)
outdoors-loc.sg smell.pst.3 lilac-ins.pl
‘Outdoors there was a smell of lilac.’
It is no coincidence that in some other languages it is precisely the smell-emission

verbs that display the variant of the swarm alternation with two oblique-marked
NPs, cf. examples (18) from Russian and (19) from Czech. The semantic proper-
ties of the smell-emission verbs contributing to realizing the construction with
two oblique-marked NPs are commented on in 3.2.
(18) V komnat-e paxlo cvet-ami, i
in room-loc.sg smell.pst.3.n flower-ins.pl and
ona soedinjalas’ dlja menja s ix zapaxom. (RNC)
she was.associated for me with their smell.
‘In the room there was a smell of flowers and for me she was associated with
their [sc. the flowers’] smell.’
(19) V kuchyn-i vonělo skořic-í. (Fried 2005: 481)
in kitchen-loc.sg smell.pst.3.n cinnamon-ins.sg
‘In the kitchen there was a smell of cinnamon.’
14. http://www.efoto.lt/zurnalas_foto/uz_kadro_projektas_satria
Two examples with the verb knibždėti (‘swarm’) in which neither NP has nomina-
tive case-marking have also been found, cf. sentences (20)–(21). In the present
examples non-locational participants are realized as PPs and also otherwise differ
from examples such as in (15b) and (16), where non-locational participants are
partitive genitive-marked. This partitive genitive is assigned lexically rather than
structurally; like the PP-marking it is directly assigned by the predicate. But parti-
tive genitive NPs, as in (15b) and (16), cannot be clearly identified as grammatical
subjects, and their syntactic subject function may be viewed as being obscured by
partitive genitive-marking, whereas the PP-marking of the relevant participants
in (20)–(21) precludes any association with subject status.
(20) Kiekvien-oje šal-yje tiesiog knibžda nuo barakud-ų.15
every-loc.sg country-loc.sg just teem.prs.3 with barracuda-gen.pl
‘Every country just teems with barracudas.’
(21) Kit-ose cikl-o viet-ose tiesiog
other-loc.pl cycle-gen.sg place-loc.pl just
knibžda nuo veik-ėjų.16
swarm.prs.3 with character-gen.pl
‘The other places of the cycle just teem with characters.’
Although sentences (20)–(21) are the only attested instances involving the verb
knibždėti and are considered by the vast majority of native speakers to be only
marginally acceptable, they are included in this research as particular instances
of event-central presentationals. Impersonal constructions involved in the swarm
alternation in Dutch and German (cf. Hoeksema 2008) may also be treated as
event-central presentationals, cf. sentences (22a–c) from German. The imperson-
al constructions appearing in the swarm alternation in German as in (22c) (and
also in Dutch) are also referred to as ‘formulaic presentational’ constructions;
they are introduced by a kind of existential formula (cf. there is/are in English).
Event-central presentationals are discussed in more detail in Section 4.
(22) a. Ameisen wimmeln in der Küche.
ants swarm in the kitchen
‘Ants are swarming in the kitchen.’
b. Die Küche wimmelt von Ameisen.
the kitchen swarms with ants
‘The kitchen is swarming with ants.’
15. http://www.lrytas.lt/?id=12653832491263876911&view=6
16. http://www.elibrary.lt/resursai/Ziniasklaida/LLKC/liaudies_kultura/Lk06_3.pdf
c. Es wimmelt von Ameisen in der Küche. (Hoeksema 2008: 3)

It swarms with ants in the kitchen
In Lithuanian, the swarm-alternation verbs partly correspond to those of English

(and, according to Hoeksema 2008, of Dutch). Thus, it is obvious that the ability
to realize the given type of alternations and the semantic properties of predicates
involved are interrelated. The semantic properties of the predicates and their ar-
guments that are relevant for assigning subject status to different arguments in
various instances of the swarm alternation are discussed in the next section.
3. Semantic properties of the verbs and their arguments

in Lithuanian swarm alternation
The section is organized as follows. Section 3.1 analyzes the semantic properties
of predicates and their arguments characteristic of all instances of the Lithuanian
swarm alternation. The research concerns only those properties that are relevant
for NP realization in argument alternations and, consequently, for assigning sub-
ject status to different arguments. Thus, the section deals with the following ques-
tions: (i) the semantics of swarm-alternation verbs and their argument structure
(with an obligatory position for a locative participant), (ii) the semantic require-
ment for non-locational participants to be expressed as indefinite plural or mass
terms rather than singular NPs, and the partitive-genitive marking of non-lo-
cational arguments as indicative of a decrease in agentivity properties, (iii) the
semantic requirement for locational participants in the location-subject variant
to be conceived as being completely or mostly ‘filled’ with the non-locational par-
ticipants, (iv) the meaning differences displayed by variants/constructions of the
swarm alternation. Then, in Section 3.2, a particular group of verbs of smell emis-
sion are explored, with particular reference to the semantic properties relevant to
realizing the construction in which neither participant has the nominative mark-
ing characteristic of the grammatical subject.
3.1 General instances
All Lithuanian swarm-alternation verbs have an obligatory meaning component

specifying that the activity/process predicated by the verb takes place/exists in a
given location, therefore, a position for a locative participant is required in argu-
ment structure. For instance, in (23)–(25), the existence of a location participant
is absolutely irrelevant for the activities/processes predicated by the verbs, while
in all the variants of the swarm alternation, as in (26)–(28), the location partici-
pant is obligatorily included in the meaning of the verbs.
(23) Baland-į žyd-inči-os gėl-ės
April-acc.sg bloom-ppra-nom.pl flower-nom.pl
kvepia, bet ne tokios ryški-os.17
smell.prs.3 but not so bright-nom.pl
‘In April blooming flowers smell, but [they are not] so bright.’
(24) Muzik-a skambėjo grakščiai, plastiškai.18
music-nom.sg resound.pst.3 gracefully plastically
‘Music resounded gracefully and softly.’
(25) a. Skruzdėl-ės knibžda. (MLD)
ants-nom.pl swarm.prs.3
‘Ants are swarming.’
b. #Skruzdėl-ių knibžda.
ants-gen.pl swarm.prs.3
‘Ants are swarming.’
(26) a. Šoki-ų aikštel-ėje smirdėjo prakait-as ir
dance-gen.pl square-loc.sg stink.pst.3 sweat-nom.sg and
alkohol-is.
alcohol-nom.sg
‘On the dance floor there was a reek of sweat and alcohol.’
b. Šoki-ų aikštel-ė smirdėjo prakait-u ir
dance-gen.pl square-nom.sg stink.pst.3 sweat-ins.sg and
alkoholi-u.19
alcohol-ins.sg
‘The dance floor reeked of sweat and alcohol.’
(27) a. Miškel-yje skambėjo paukšči-ų bals-ai.
grove-loc.sg resound.pst.3 bird-gen.pl voice-nom.pl
‘Voices of birds resounded in the grove.’
b. Netoliese es-ant-is miškel-is skambėjo
nearby be-ppra-nom.sg grove-nom.sg resound.pst.3
nuo įvairiausi-ų paukšči-ų bals-ų.(CCLL)
from various-gen.pl bird-gen.pl voice-gen.pl
‘The nearby grove resounded with voices of various birds.’
17. http://www.gamtosturtai.lt/docs/AROMATU%20OAZE.doc
18. http://www.fortvio.lt/lt/spauda.html
19. http://kaituisejaiatsiradokitas.blogas.lt/kai-tu-isejai4-34.html
(28) a. Valgykl-oje knibždėjo tarakon-ų /tarakon-ai.

canteen-loc.sg swarm.pst.3 cockroach-gen.pl cockroach-nom.pl
‘In the canteen cockroaches were teeming.’
b. <…> valgykl-a knibždėjo tarakon-ų <…>. (CCLL)
canteen-nom.sg swarm.pst.3 cockroach-gen.pl
‘The canteen was teeming with cockroaches.’
As for (25b), partitive-genitive marking of the single argument carries a presup-

position of a location in which the activity of swarming takes place. That is, sen-
tence (25b) is only possible if the location is presupposed and the meaning of the
predicate is not just one of predicating the given property, but of predicating the
presence of the property in the presupposed location. Moreover, swarm-alterna-
tion verbs permit variations in case marking, as in (28a). Although the partitive
genitive marking of non-locational participants is preferred, nominative case is
also possible.
Verbs with the prefix pri-, which constitute the largest semantic class of
swarm-alternation verbs, do not allow entity/substance participants to surface as
nominative case-marked subjects: these are obligatorily marked with the partitive
genitive, as in (29)–(30). Pri- is also responsible for the obligatory location par-
ticipant in the argument structure of the verb.
(29) a. Jo trob-oje prisigrūdo žmoni-ų <…>20
he.gen cottage-loc.sg crowd.pst.3 people-gen.pl
[/*žmon-ės]
[/people-nom.pl]
‘People were crowding in his cottage.’
b. Jo trob-a prisigrūdo žmoni-ų.
he.gen cottage-nom.sg crowd.pst.3 people-gen.pl
‘His cottage was crowded with people.’
(30) a. Mišk-e priželia krūmokšni-ų / *krūmokšni-ai.
forest-loc.sg spring.up.prs.3 scrub-gen.pl / scrub-nom.pl
‘Brushwood proliferates in the forest.’
b. Užteršt-as mišk-as priželia krūmokšni-ų,
polluted-nom.sg forest-nom.sg spring.up.prs.3 scrub-gen.pl
samanoja medži-ai.21
be.moss-covered.prs.3 tree-nom.pl
‘The polluted forest is overgrown with brushwood, [and] trees are covered
with moss.’
20. http://ebiblioteka.mkp.emokykla.lt/kuriniai/portugalijos_karalius/,format.pdf
21. http://referataiplius.lt/zmogus-ir-gamta-13952.html
The corresponding verbs without pri- do not allow either alternations in argu-
ment realization or variation in case-marking, cf. sentence (31a–b). Marking an
NP with the partitive genitive is not compatible with the thematic role of agent,
which entails volitionally and purposefully acting entities. Agentive participants
of an event marked with the partitive genitive are usually conceived of as deper-
sonalized groups in which the responsibility of each entity for carrying out the
activity is irrelevant. Therefore, the obligatory (as in verbs with pri-) or optional
(as in (28a)) marking with the partitive genitive may be considered to code a de-
crease in agentivity properties of the respective participants which correlates with
the discourse-pragmatic functions of presentationals displayed by some variants
of the Lithuanian swarm alternation, i.e., the introduction of a discourse-new ref-
erent, causing the agent arguments of the swarm alternation to be ‘agentive to a
certain degree’; see Section 4 for more details.
(31) a. Žmon-ės grūdosi į bažnyči-as
people-nom.pl crowd.pst.3 into church-acc.pl
dėko-ti Diev-ui22 <…>
thank-inf god-dat.sg
‘People were crowding into the churches to thank God.’
b. *Žmoni-ų grūdosi į bažnyči-as
people-gen.pl crowd.pst.3 into church-acc.pl
dėko-ti Diev-ui.
thank-inf god-dat.sg
The partitive genitive denoting a decrease in agentive properties also correlates

with another semantic property characteristic of non-locational NPs in the
swarm alternation, viz. the requirement for them to be expressed as indefinite
plural or mass term and not singular NPs. In English, according to Dowty (1999,
2000, 2001), this requirement is valid only for the locative-subject variant: “Lo-
cation-subject forms sound natural only when the with-phrase is an indefinite
plural or mass term, but not a singular NP”, cf. (32). In Lithuanian, though the
requirement holds first of all for the locative-subject variant, in the variant with
a non-locational participant-subject individualization of entities performing an
activity is also dispreferred, cf. examples (32), (33):
(32) a. The table crawled with a hundred cockroaches.
b. ?The table crawled with seventy-three cockroaches. (Dowty 2001: 5)
22. http://printfu.org/atkuriama
(33) a. Aukštesni-uose šlait-uose knibžda šimt-ai

higher-loc.pl incline-loc.pl swarm.prs.3 hundred-nom.pl
alpinist-ų / ?40 alpinist-ų / ?101 alpinist-as.
climber-gen.pl / 40 climber-gen.pl / 101 climber-nom.sg
‘Hundreds of climbers/ ?40 climbers/ ?101 climbers swarm on the higher
inclines.’
b. <…> aukštesn-ieji šlait-ai knibžda nuo
higher-nom.pl incline-nom.pl swarm.prs.3 with
šimt-ų alpinist-ų.23 [*nuo 40 alpinist-ų / *nuo
hundred-gen.pl alpinist-gen.pl [with 40 alpinist-gen.pl / with
101 alpinist-o]
101 alpinist-nom.sg]
‘The higher inclines swarm with hundreds of climbers/ *with 40 climbers/
*with 101 climbers.’
c. Filharmonij-oje knibždėjo šimt-ai žmoni-ų /
philharmonic-loc.sg swarm.pst.3 hundred-nom.pl people-gen.pl /
?130 žmoni-ų / ??125 žmon-ės
130 people-gen.pl / 125 people-nom.pl
‘Hundreds of people/ ?130 people/ ??125 people swarmed in the concert
hall.’
d. Filharmonij-a knibždėjo nuo tūkstanči-ų
philharmonic-NOM.SG swarm.PST.3 with thousand-gen.pl
žmoni-ų / *nuo 130 žmoni-ų / *nuo 125 žmon-ių
people-gen.pl / with 130 people-gen.pl / with 125 people-gen.pl
‘The concert hall was swarming with thousands of people/ *130 people/
*125 people.’
The individualization of the non-locational participants is at least irrelevant in the

swarm alternation. The given semantic properties of non-locational participants
also correlate with the partitive genitive marking denoting a decrease in agentive
properties. Moreover, considering semantic factors that contribute to alternations
in argument realization (which actually involve the assignment of subject status
to different arguments), it is important to account for the semantic properties
of both candidates for the subject position, i.e., locational and non-locational
participants. In the literature exploring the swarm alternation the holistic read-
ing entailing the ‘filling’ (or at least ‘filling’ to a relatively large extent) of a loca-
tion participant has been noted. In order to be realized as a subject, the location
participant has to be perceived as completely (or mostly) ‘filled’ by the non-lo-
cational participants. The earliest and most familiar observation about the location-
23. http://www.technologijos.lt/n/mokslas/idomusis_mokslas/S-34093/
subject variant of the swarm alternation (cf. Anderson 1971; Dowty 2001) is that
the location-subject form seems to entail that the activity ‘fills’ the entire location
to a greater extent than the form with non-locational subject (or according to
Dowty’s terminology, the agent-subject form) does. Dowty cites example (34a–b)
of Anderson as illustration of the given observation, cf. also examples (35a–b) for
Lithuanian.
(34) a. Ants are crawling on the table, but they’re confined to this one corner of it.
b. # The table is crawling with ants, but they’re confined this one corner of it.
(Dowty 2001)
(35) a. Virtuv-ėje knibžda skruzdėli-ų / skruzdėl-ės, bet jų
kitchen-loc.sg swarm.prs.3 ant-gen.pl / ant-nom.pl but they.gen
daugiausia šit-ame (virtuv-ės) kamp-e.
most this-loc.sg kitchen-gen.sg corner-loc.sg
‘Ants are crawling in the kitchen, but they’re mostly in this corner of the
kitchen.’
b. ??Virtuv-ė knibžda skruzdėli-ų, bet jų
kitchen-loc.sg swarm.prs.3 ant-gen.pl but they.gen
daugiausia yra tik šit-ame kamp-e.
most be.prs.3 just this-loc.sg corner-loc.sg
‘??The kitchen is crawling with ants, but they’re mostly only in this one
corner.’
Thus, in Lithuanian as well as in English, the semantic properties of the non-lo-

cational participants as a group of non-discrete and non-individualized entities
(and a decrease in agentive properties characteristic of some instances of the Lith-
uanian swarm alternation) and the fact that the locational participant is perceived
as completely (or mostly) ‘filled’ by the non-locational participants, both con-
tribute to the realization of the location-subject variant. That is, non-locational
participants, being conceived as non-discrete, non-individualized and less agen-
tive, are less appropriate candidates for subjecthood than locational participants,
which are perceived as ‘filled’ by the non-locational participants and thus fully or
somewhat affected. The variant with location-subject indicates that the location
participant is ascribed the property of complete ‘filling’ (i.e., the property result-
ing from an activity/process of certain entities/substance and affecting a given
location by means of ‘filling’ it completely or to some extent). The variant with
non-locational participant subject denotes presence of entities/of a substance per-
forming a given activity or conditioning a given process in a given location.
All the semantic properties of predicates and their arguments discussed above
are interrelated in one way or another with the properties of the participants re-
quired by a discourse-pragmatic function of presentationals. Moreover, as shown
in Section 5, the semantic properties of swarm-alternation predicates and, conse-

quently, the semantic features of their arguments, viz. lack of certain features that
are typical of subjects, contribute to alternations in argument realization, i.e., the
assignment of subject status to different arguments. In the following subsection, a
particular group of smell-emission verbs are explored with particular reference to
the semantic properties relevant for realizing one particular variant of the swarm
alternation, the one in which neither participant has nominative marking and
both are oblique-marked.
3.2 Semantic properties of smell-emission verbs and the variant

of the swarm alternation with two oblique-marked NPs
Section 2, presenting an overview of instances of the Lithuanian swarm alterna-

tion, mentioned that in Lithuanian, as in other languages such as Russian and
Czech, it is precisely the smell-emission verbs that display the variant of the
swarm alternation with both NPs oblique-marked. It is obvious that some seman-
tic properties of smell-emission verbs contribute to realizing the variant with two
oblique NPs, cf. the three-way alternation pattern for Lithuanian (36a–c) (=17)
and for Czech (37a–c), and also the construction with two oblique NPs for Rus-
sian (38) (=18).
(36) a. Lauk-e kvepėjo alyv-os.
outdoors-loc.sg smell.pst.3 lilac-nom.pl
‘Outdoors lilacs smelled.’
b. Lauk-as kvepėjo alyv-omis.
outdoors-nom.sg smell.pst.3 lilac-ins.pl
‘[The air] outdoors smelled of lilac.’
c. Lauk-e kvepėjo alyv-omis. (CCLL)
outdoors-loc.sg smell.pst.3 lilac-ins.pl
‘Outdoors there was a smell of lilac.’
(37) a. V kuchyn-i voně-l-a skořic-e.
in kitchen-loc.sg.f give.off.fragrance-pst-sg.f cinnamon-nom.sg.f
‘Cinnamon24 smelled in the kitchen.’
b. Kuchyň voně-l-a skořic-í.
kitchen:nom.sg.f smell-pst-sg.f cinnamon-ins.sg.f
‘The kitchen smelled of cinnamon.’
24. Small caps indicate that the NP is not the topic. The nominative is unambiguously a focal
constituent (Fried 2005: 481).
c. V kuchyn-i voně-l-o skořic-í.

in kitchen-loc.sg.f smell-pst-sg.n cinnamon-ins.sg.f
‘In the kitchen, there was the smell of cinnamon.’ (Fried 2005: 481)
(38) V komnat-e paxlo cvet-ami, i
in room-loc.sg smell.pst.3.n flower-ins.pl and
ona soedinjalas’ dlja menja s ix zapaxom. (RNC)
she was.associated for me with their smell.
‘In the room there was a smell of flowers and for me she was associated with
their [sc. the flowers’] smell.’
However, as the corresponding examples in Russian (39) and in Czech (40) show,
verbs of smell emission are not the only ones that can realize the variant with two
oblique NPs. In Lithuanian, two examples with the verb knibždėti (‘swarm’) were
also found in which neither of the NPs has nominative marking, cf. (20)–(21) in
Section 2. Though rare, they are also taken into account here. Unlike other in-
stances of the swarm alternation, which function as entity-central presentationals,
(20)–(21) and (39)–(40) are event-central presentationals, i.e., they introduce a
new event into discourse, see also Section 4 for other instances of event-central
presentationals involved in the swarm alternation.
(39) V Kievsk-oj oblast-i kiše-l-o fal’sifikacij-ami …25
in Kiev-loc.sg district-loc.sg be.rife-pst.3-n falsification-ins.pl
‘In Kiev district, falsifications were rife.’
(40) (v nadcházejících volbách se v boji o hlasy voličů … střetnou také sdružení …)
Kandidát-y se to hemží především na
candidate-ins.pl.m refl expl swarm.prs.3sg first.of.all on
pravé straně (komunálního politického spektra).
right.loc.sg.f side.loc.sg
‘(in the up-coming elections, the competition for voters will include also civic
organizations …) Swarming of candidates occurs especially on the right side
(of the local political spectrum).’ (Fried 2005: 482)
A very important observation concerning verbs of olfactory perception and their

ability to realize the variant with two oblique NPs was made by Paducheva (2007).
The fact that a smell may exist/be presented independently of its source contrib-
utes to this realization, cf. (41)–(42) for Russian and Lithuanian respectively.
25. http://news.meta.ua/single/2010-11-18903268/
(41) V syr-om vozdux-e pax-l-o

in moist-loc.sg.m air-loc.sg smell-pst-n
kvaš-en-oj kapust-oj.
pickle-ppp-ins.sg.f cabbage-ins.sg
‘In the moist air, there was a smell of sauerkraut.’ (Padučeva 2007)
(42) Sal-ėse dvokia prakait-u, ne-plaun-am-os
hall-loc.pl stink.prs.3 sweat-ins.sg neg-wash-ppra-nom.pl.f
grind-ys!26
floor-nom.pl
‘There is a stench of sweat in the halls [and] the floors are not cleaned.’
The constructions with two oblique NPs, as in (41)–(42) also cf. (36c), (37c), and
(38)), denote that there is/exists a given smell in the given location, while the
source of the smell may have been removed from the given location some time
ago. Thus, according to Fried (2005: 492), constructions like (37c) “simply reports
the presence of a sensory effect in a place”. Moreover, in constructions with two
oblique NPs it is precisely a smell, not a (visible) source of a smell, that is a refer-
ent of the instrumental NPs. For example, as in (43), the construction with two
oblique NPs is unacceptable when the referent of an instrumental NP is visible
and, consequentially, definite and referential.
( 43) ??Kambar-yje kvepėjo į vaz-ą pamerk-t-omis
room-loc.sg smell.pst.3 into vase-acc.sg put-ppp-ins.pl.f
dvidešimt penki-omis rož-ėmis.
twenty five-ins.pl.f rose-ins.pl
‘In the room, there was a smell of twenty-five roses put into the vase.’
The fact that in double-oblique constructions the instrumental NP refers to a

smell rather than the source of a smell becomes evident when we examine the
constructions in a broader context. Nominative NPs, as in (44a–b), denote the
source of a smell while in the constructions with two oblique NPs the instrumen-
tal NP denotes the smell itself, cf. (44c–d). Furthermore, taking into account the
discourse function of presentationals like the initial sentences in (44a–d), it is
obvious that the sentences with nominative NPs denoting the sources of a smell,
as in (44a–b), serve, on the discourse-pragmatic level of representation, to present
precisely a source of a smell in a given location, while sentences with two oblique
NPs (44c–d) are predisposed to present a smell in a given location.
26. http://m.lrytas.lt/?id=12285682701226373344&view=9&p=6
(44) a. Kiem-e kvepėjo rož-ės. Jos šiemet

courtyard-loc.sg smell.pst.3 rose-nom.pl 3.nom.pl.f this.year
buvo labai didel-ės ir graži-os.
be.pst.3 very large-nom.pl.f and beautiful-nom.pl.f
‘The roses smelled in the courtyard. This year they were very large and
beautiful.’
b. Kiem-e kvepėjo rož-ės. ?T-as
courtyard-loc.sg smell.pst.3 rose-nom.pl that-nom.sg.m
kvap-as buvo sald-us ir malon-us.
smell-nom.sg be.pst.3 sweet-nom.sg.m and pleasant-nom.sg.m
‘The roses smelled in the courtyard. That smell was sweet and pleasant.’
c. Kiem-e kvepėjo rož-ėmis. ??Rož-ės
courtyard-loc.sg smell.pst.3 rose-ins.pl rose-nom.pl
buvo labai didel-ės ir graži-os.
be.pst.3 very large-nom.pl.f and beautiful-nom.pl.f
‘In the courtyard, there was a smell of roses. The roses were very large and
beautiful.’
d. Kiem-e kvepėjo rož-ėmis. T-as
courtyard-loc.sg smell.pst.3 rose-ins.pl that-nom.sg.m
kvap-as buvo sald-us ir malon-us.
smell-nom.pl be.pst.3 sweet-nom.sg.m and plesant-nom.sg.m
‘In the courtyard, there was a smell of roses. That smell was sweet and
pleasant.’
On the other hand, event participants denoting a sound or visual properties can-
not be separated from their sources. For instance, verbs of auditory perception
allow a sound (e.g., music etc.) or a source of a sound (e.g., a guitar) to surface as
the subject of the clause, as in (45)–(46). Even though a sound can be realized as
a clause subject, it cannot exist independently of an instrument producing it. An
instrument may be not explicitly realized, but its existence is always presupposed
and present in a given event of sound emission. Therefore auditory-perception
verbs cannot use the construction with two oblique NPs to denote ‘existence/
presence of a sound in a given location’.
(45) <…> centrin-ėje aikšt-ėje visą tą laiką
central-loc.sg square-loc.sg all.this.time
skambėjo varp-ų muzik-a.27
resound.pst.3 carillon-gen.pl music-nom.sg
‘Carillon music resounded in the central square all this time.’
27. http://www.mytrips.lt/Pasakojimas/Olandija-ir-geliu-paradas/404
(46) Scen-oje skambėjo gitar-a,

stage-loc.sg resound.pst.3 guitar-nom.sg
akordeon-as, birbyn-ė <…>28
accordion-nom.sg hornpipe-nom.sg
‘Guitar, accordion and hornpipe resounded on the stage.’
Verbs of visual perception usually do not allow visible entities/substances to be

split into something like a source of a visible property and a visible property, and
constructions with two oblique participants are impossible here. It is visible enti-
ties/substances present in a given location, as in (47a), or a location ‘filled’ with
the given entities/substances, as in (47b), that may be realized as clause subject.
(47) a. Dang-uje mirgėjo aitvar-ai.
sky-loc.sg flicker.pst.3 kite-nom.pl
‘Kites flickered in the sky.’
b. Prie Nemun-o dang-us mirgėjo nuo aitvar-ų.29
near Neman-gen.sg sky-nom.sg flicker.pst.3 with kite-gen.pl
‘Near the Neman, the sky was flickering with kites.’
Thus, in Lithuanian, as well as in Russian and Czech, the constructions with two
oblique NPs containing smell-emission verbs denote the presence of a smell in a
given location. The main factor contributing to the realization of the construc-
tions with two oblique NPs is the semantic properties of arguments denoting a
smell/the source of a smell, viz. a smell may exist/be present independently of
its source. In other words, constructions with two oblique NPs serve to denote
the presence/existence of a smell without presupposing the source of that smell.
On the discourse-pragmatic level of representation the constructions with two
oblique NPs also exclusively present a smell.
The next section explores the discourse-pragmatic function of presentationals
displayed by some variants of the Lithuanian swarm alternation, with particular
reference to the pragmatic properties that contribute to alternations in assigning
subject status to different arguments.
28. http://ausra.pl/2009/0919/muz-seimos.htm
29. http://www.lrytas.lt/-12406674341239176629-zapy%C5%A1kio-danguje-tuntai-
v%C4%97jo-g%C5%ABsius-gaudan%C4%8Di%C5%B3-aitvar%C5%B3-nuotraukos-video.
htm
4. Discourse-pragmatic functions of the Lithuanian swarm alternation
As mentioned in the Introduction, the discourse function of presentationals dis-

played by some variants of the the Lithuanian swarm alternation has these main
consequences: (i) a grammatical subject of a sentence functioning as a presen-
tational construction is most characteristically not construed as a topic, but as a
clausal focus; (ii) presentationals are often characterized by non-canonical dis-
tribution of subject properties over the (pro)nominal constituents of a sentence.
Furthermore, particular properties of presentationals are interrelated with certain
semantic properties of arguments of swarm-alternation verbs: (i) the semantics of
Lithuanian swarm verbs and also their argument structure obligatorily require a
position for a locative participant, hence the requirement for a location participant
correlates with the fundamental property of discourse-pragmatic presentationals,
which is not simply to introduce a new NP referent, but rather to introduce the
NP referent by asserting its presence in a given location; (ii) the semantic require-
ment to express non-locational participants as indefinite plurals or mass terms,
not as singular NPs, correlates with the preference of thetic propositions for ‘in-
definite’ NPs (NPs with unidentifiable referents); (iii) the semantic properties of
the non-locational participants as a group of non-discrete and non-individual-
ized entities, and the marking of non-locational participants with the partitive
genitive to indicate a decrease in agentive properties, correlate with the pragmatic
function of presentationals: presentational function renders the semantic proper-
ties of the agent participant irrelevant and the agent argument of such sentences
can be said to be agentive to a certain degree, or ‘pseudo-agentive’. All these dis-
course pragmatic and semantic properties are to be discussed in more detail in
the present section. The analysis of presentationals is mainly based on Lambrecht
(1994, 2000). The main notions concerning the issue, i.e., ‘presentational con-
structions’, ‘sentence-focus constructions’, ‘predicate-focus constructions’ ‘thetic’
vs. ‘categorical’ judgments etc., were presented in the Introduction, therefore they
are not repeated here, unless it is necessary additionally to comment on new or
already known characteristics of these phenomena.
First of all, the variants of the Lithuanian swarm alternation functioning as
presentationals and the means to realize discourse-pragmatic presentationals are
discussed in more detail. As mentioned above, not all variants of the Lithuanian
swarm alternation function as presentationals. The location-subject variants, as in
(48b) and (49b), are so-called predicate-focus (PF) constructions and they are prag-
matically structured as propositions in which the subject is topic and the predicate
expresses information about the topic, while in non-locational subject variants,
as in (48a) and (49a), which pragmatically belong to the so-called sentence-focus
(SF) constructions, the proposition is so structured that both subject and predicate
are in focus and the sentences serve to introduce a new discourse referent, or, to be
more precise, they refer to the presence of a new or hitherto unnoticed entity. The
constructions with two oblique- NPs containing smell-emission verbs, as in (49c),
are also pragmatically presentationals and serve to refer to the presence of a smell
in a given location (see Section 3.2 for more details).
(48) a. <…> vyresn-iesiems broli-ams bažnyči-oje suteik-t-as
elder-dat.pl brother-dat.pl church-loc.sg give-ppp-nom.sg
Sut vir
ti
ni
m-o sak ra
ment-as, po ku ri-o
Confirmation-gen.sg Sacrament-nom.sg after which-gen.sg
į na m-us pri va vo sveči-ų <…>30
žia
into house-acc.pl come.pst.3 guest-gen.pl
‘<…> in the church the elder brothers were given the Sacrament of
Confirmation, after that [a lot of] guests came into the house.’
b. Pas mus kiem-e šiandien ryt-e
at we.acc courtyard-loc.sg today morning-loc.sg
dram-a buvo. Policij-os piln-as
drama-nom.sg be.pst.3 police-gen.sg full-nom.sg
kiem-as privažiavo.31
courtyard-nom.sg come.pst.3
‘In our courtyard, there was a drama this morning. The courtyard got full
of police.’
(49) a. <…> nam-as ir aplink-a traukė
house-nom.sg and surroundings-nom.sg draw.pst.3
ak-į tvark-a ir jaukum-u.
eye-acc.sg tidiness-ins.sg and comfort-ins.sg
Daržel-yje kvepėjo gėl-ės <…>32
flower.garden-loc.sg smell.pst.3 flower-nom.pl
‘The house and surroundings drew the eye with their tidiness and cosi-
ness. In the flower garden, flowers smelled.’
b. Vasari-o mėnes-į Gvadelup-os or-o
February-gen.sg month-acc.sg Guadeloupe-gen.sg air-gen.sg
uost-as kvepėjo gėl-ėmis.33
port-nom.sg smell.pst.3 flower-ins.pl
‘In February the airport of Guadeloupe smelled of flowers.’
30. http://www.pzinios.lt/Rubrikos/Aktualijos/Laukia-Kaledu-ir-tecio!
31. http://www.supermama.lt/forumas/index.php?showtopic=894444&st=12
32. http://ausra.pl/0606/negrizti.html
33. http://www.girnos.lt/content/karib%C5%B3-j%C5%ABroje-pirat%C5%B3-nesutikome-
bet-radome-lietuvos-%C5%BEem%C4%97lap%C4%AF
c. Vakar Vilni-aus bank-e kvepėjo

yesterday Vilnius-gen.sg bank-loc.sg smell.pst.3
valerijon-u, mėt-ų laš-ais ir validoliu. (CCLL)
valerian-ins.sg mint-gen.pl drop-ins.pl and Validol-ins.sg
‘Yesterday in the Bank of Vilnius branch there was a smell of valerian,
mint drops and Validol.34’
One of the fundamental properties of presentationals is that pragmatically the

subject is not construed as a topic but as a sentential focus. All the variants of the
Lithuanian swarm alternation are treated as pragmatically unmarked only when
their non-locational participant has a focus accent (the focus accent is indicated
by small caps), cf. (48)–(49). The swarm alternation is characteristic of intransitive
verbs. Intransitive PF sentences are pragmatically unmarked when the predicate
has a focus accent (here the notion ‘predicate’ is used in the sense of ‘what is said
of the subject’). In PF sentences with transitive predicates, typically an argument
that serves to form an intransitive predicate, i.e., an object, is an unmarked focus
argument, cf. Lambrecht (2000: 625f.). In the locative-subject variant of the Lithu-
anian swarm alternation, as in (48b) and (49b), the grammatical subject is the
topic of the sentence and the predicate (i.e., predicate portion) has focus accent,
thus, a topic-comment relation, as characteristic of predicate-focus sentences,
holds between the subject and the predicate. In instances when a location-subject
is the focus argument, as in (50), it functions pragmatically as a presentational
construction introducing a new event.
(50) Meninink-ų rūm-uose (dabar Prezidentūr-a)
artist-gen.pl house-loc.pl now Presidential.Palace-nom.sg
įvyko klub-o steigiam-asis susirinkim-as.
take.place.pst.3 club-gen.sg constituent-nom.sg assembly-nom.sg
Žmoni-ų prisirinko piln-a balt-oji sal-ė. (CCLL)
people-gen.pl gather.pst.3 full-nom.sg White-nom.sg Hall-nom.sg
‘In the House of Artists (now the Presidential Palace) the founding meeting
of the club was held. The White Hall was brimming with people.’
In (49a) the grammatical subject is not a topic but a focus argument, hence “the
proposition expressed by the sentence is not pragmatically construed as being about
the referent of this NP. It is not the absence of any topical element that defines a SF
construction but the absence of a topic-comment relation between the proposition
and that argument which in the unmarked case [i.e., in a PF sentence – KLG] func-
tions as the topic of a sentence, i.e. typically the subject” (Lambrecht 2000: 627–
628). Thus, presentationals, as a particular instance of the SF, serve not to express
34. A heart medicine.

a topic-comment relation between the proposition and the subject, but to state the
existence/presence of the referent of the subject NP in a given location.
Concerning the characteristic properties of SF constructions, and the presen-
tational constructions in particular, in Lithuanian swarm alternations we note a
strong preference35 for the focus argument (i.e., non-locational participant) to be
realized in clause-final position. Nevertheless, the clause-final positioning of non-
locational participants is probably not directly material to the issue. It seems that
in Lithuanian there is a tendency to place any focus argument (not only the gram-
matical subject of presentationals) in clause-final position. A well-established
fact about SF constructions is that “[i]n English, and in other languages relying
on prosodic focus marking, a SF construction is minimally characterized by the
presence of a pitch accent on the subject and by the absence of prosodic promi-
nence on the predicate portion of the sentence” (Lambrecht 2000: 617). Thus, it is
precisely the focus pitch accent on the subject, leading to the absence of a topic-
comment relation between the proposition and the subject, that characterizes SF
constructions, and presentationals among them.
It is not a coincidence that the focus pitch accent on the subject serves to
mark presentationals. Since, for the sentence to be interpreted as a presentational
construction, the grammatical subject of a sentence needs to be marked in such
a way that it will not be construed as sentence topic, it is obvious that the focus
pitch accent on the subject serves as the most appropriate means to ‘detopicalize’
the subject. Therefore Lambrecht (2000: 624f.) proposes that the marking of SF
constructions “crucially involves a ‘detopicalization’ of the subject. The process of
‘detopicalizing’ the subject is defined by the Principle of Detopicalization:
SF marking involves cancellation of those prosodic and/or morphosyntactic sub-
ject properties which are associated with the role of subjects as topic expressions
in PF sentences. (Lambrecht 2000: 624)
Lambrecht (ibid.: 624–625) also emphasizes that the terms ‘detopicalization’ and

‘cancellation’ in the Principle of Detopicalization do not suggest a synchronic der-
ivational relationship between a PF and a corresponding SF construction. What
the Principle of Detopicalization tries to capture is natural paths of grammatical-
ization rather than rules of sentence formation. The process of ‘detopicalization’,
realized by marking the subject as the focus argument, leads to further prosodic
35. For instance, on the basis of CCLL data, in only 13 instances of 196 containing the verb
form kvepia (PRS.3) ‘smell’ is the non-locational participant not in clause-final position. Simi-
larly with the verb form dvokia (PRS.3) ‘stink’ – 3 of 26, dvokė (PST.3) ‘stank’ – 3 of 32, smirda
(PRS.3) ‘stink’ – 0 of 4, smirdėjo (PST.3) ‘stank’ – 1 of 8, knibžda (PRS.3) ‘swarm’ – 49 of 253,
knibždėjo (PST.3) ‘swarmed’ – 27 of 179 etc.
and morphosyntactic consequences, captured by the Principle of Subject–Object

Neutralization:
In a SF construction, the subject tends to be grammatically coded with some
or all of the prosodic and/or morphosyntactic features associated with the focal
object in the corresponding PF construction. (Lambrecht 2000: 626)
Since in PF constructions the unmarked focus argument is the object, the coding
of non-topical subjects of presentationals may be achieved by using grammatical
and pragmatic features normally associated with the object of a PF sentence. That
is, in presentationals a distribution of subject properties over the (pro)nominal
constituents of a sentence is the natural consequence of the subject’s need to
‘detopicalize’, cf. Lambrecht (2000); Gast and Haas (2011) for more details. The
swarm alternation is characteristic of intransitive verbs. For this reason, the pre-
sentational constructions realized by different variants of the Lithuanian swarm
alternation show a spread of subject properties over both nominal constituents
obligatorily required in the predicate-argument structure of the sentence. Con-
structions with two oblique NPs, as in (49c) etc., explicitly spread the subject
properties over the (pro)nominal constituents of a sentence. However, in fact, all
the variants of the swarm alternation testify to a spread of subject properties over
the (pro)nominal constituents of a sentence.
SF constructions cover several subtypes pragmatically similar and yet consti-
tuting different phenomena. One subtype of the SF construction is the so-called
‘existential sentence’, which involves verbs of existence, especially the verb be.
Another subtype is ‘Locative Inversion’ (cf. Levin 1993; Bresnan 1994; Levin &
Rappaport Hovav 1995). Some languages, for instance English, use several dis-
tinct constructions to code different subtypes of SF constructions. According to
Gast and Haas (2011), constructions that serve to code presentational utterances
can be divided into ‘formulaic’ and ‘non-formulaic’ ones. In English at least two
constructions may be used to encode presentational utterances: (i) constructions
introduced by the ‘existential formula’ there is/are, there’s, as in (51), and (ii) the
so-called ‘(locative) inversion’ constructions, as in (52).
(51) There is [a man in your kingdom who has a spirit of the holy gods in him …]36
(Gast & Haas 2011: 129)
36. The NP in the brackets introduces a new participant.

(52) On one hand flashes37 a 14-carat round diamond; on the other hand
sparkles an 8-carat stone flanked by the diamond-studded initials W.N.
[Philadelphia Inquirer, “To the Top the Hard Way” 1–D]
(Levin & Rappaport Hovav 1995: 255)
In Lithuanian, typically both existential constructions, as in (53), and presenta-

tionals, as in (54), are marked by a focal accent on the subject (also cf. the presen-
tationals involved in the swarm alternation, as in (48a), (49a), (49c) etc.), for more
details on Lithuanian existentials see Kalėdaitė (2002) and Holvoet (2005).
(53) Stalči-uje yra sąsiuvinis. (Holvoet 2005: 140)
drawer-loc.sg be.prs.3 notebook:nom.sg
‘There is a notebook in the drawer.’
(54) Vien-ą kart-ą gyveno paukštis.38
one-acc.sg time-acc.sg live.pst.3 bird-nom.sg
‘Once upon a time there lived a bird.’
Hence, it is important to identify what existential and other subtypes of presenta-

tionals have in common that causes them all to function as SF constructions. First
of all, both existential and other presentationals have similar semantic proper-
ties. According to Lambrecht, “the most common and grammatically most clearly
marked presentational clause type is characterized across languages by the pres-
ence of a limited set of predicates whose arguments have a highly non-agentive
and often locative case-role, such as “be”, “be at”, “live”, “arrive”, “have”, “see”,
etc.” (Lambrecht 1998: 180). That is, verbs occurring in presentationals usually
denote the presence/existence/appearance of an argument in a particular location
and are treated as being “informationally light in context”. The verb provides a
possibility for introducing a new participant and does not contribute any addi-
tional information beyond the fact that the entity exists, cf. Levin and Rappaport
Hovav (1995: 220–277). Existentials typically involve verbs such as ‘be’. However,
Lithuanian swarm alternation verbs occurring in the presentational construc-
tions, as well as verbs displaying locative inversion constructions in English, may
be treated as predicating additional information beyond the mere existence of
the entity. According to Levin and Rappaport Hovav (1995: 253), for the verb to
be treated as informationally light in context, the verb and the entity the verb is
predicated of need to be mutually predictable. Swarm alternation verbs and their
37. Here, to allow easier identification of locative inversion, small capital letters mark any rel-
evant verbs and accompanying auxiliaries, cf. Levin and Rappaport Hovav (1995: 218).
38. http://banga.balsas.lt/lt/2content.content_view_diary_other/1878367.506310-=
%283545140906
non-locational arguments are usually mutually predictable, i.e., the activity/pro-

cess the verb denotes is characteristic of, or closely associated with, the semantic
specifications of the argument the verb is predicated of. Thus, both existentials and
other subtypes of presentationals (hence also presentationals entering the Lithu-
anian swarm alternation and English locative inversion constructions) serve, by
virtue of the “informational lightness in context” of their verbs, to present a new
participant by asserting its presence/existence in a given location.
As mentioned above (cf. Section 3.1), some properties of presentationals are
interrelated with certain semantic properties of arguments of swarm-alternation
verbs. The fact that location participants are obligatorily required in the predi-
cate-argument structure of the Lithuanian swarm verbs correlates with the fun-
damental property of discourse-pragmatic presentationals, which is to present or
introduce NP referents to the ‘space’ or ‘scene’ of discourse rather than asserting
their mere existence (Lambrecht 1994: 179). Therefore, it is not surprising that, in
many languages, at least some subtypes of presentationals (e.g., Lithuanian swarm
alternation or English locative inversion constructions) the locative expression
serves as a topical reference point for the entity being introduced via the presen-
tational construction (Lambrecht 2000: 637). Thus, a presentational construction
needs at least two semantic-pragmatic properties: (i) verbs must be information-
ally light in context, and (ii) a location participant must be obligatorily included
in the argument structure.
Other semantic properties of non-locational participants of the Lithuanian
swarm alternation as a group of non-discrete and non-individualized entities/
substances account for their being indefinite rather than definite. They satisfy
the preference of thetic propositions (of which presentationals are a subtype, cf.
the Introduction) for NPs with unidentifiable referents (Lambrecht 1994: 143).
Thus, we have here one more instance of conformity between semantic and prag-
matic properties contributing to the realization of presentationals. Furthermore,
the above-mentioned semantic properties of non-locational participants, includ-
ing agent participants, correlates with the fact that agent participants introduced
by presentationals can be said to be ‘pseudo-agentive’. According to Lambrecht
(ibid.: 183), “[i]n such cases, the agentivity of the predicate is subordinated to the
presentational function of proposition and the predicate is in fact pragmatically
construed as non-agentive”. Thus, discourse-pragmatic presentationals serve to
present/introduce an agent participant, but do not convey information about its
agentivity, and therefore the semantic properties typical of agent participants, viz.
volitional involvement and/or responsibility for the execution of an event etc., are
irrelevant for the discourse pragmatic function of presentationals.
The given discourse-pragmatic requirement to ‘deagentivize’ predicates, to-
gether with the relevant semantic properties of agent participants (non-discrete
and non-individualized entities/substance) explain why in some languages, such

as Dutch and German, typically agentive predicates may be pragmatically con-
strued as non-agentive in formulaic presentational impersonal constructions,
cf. (55a–c) (=22) for German and (56a–c) for Dutch. Formulaic presentational
impersonal constructions in German and Dutch swarm alternations are event-
central rather than entity-central. The given formulaic presentational impersonal
constructions from German and Dutch, as well as event-central instances of the
swarm alternations from Russian (39), Czech (40), and Lithuanian (20)–(21), are
displayed by verbs of visual perception, i.e., by verbs denoting visually perceptible
movement of entities/substance. As discussed in Section 3.2, visual-perception
verbs are unlike smell-emission verbs in not allowing visible entities/substances
to be split into something like a source of a visible property and a visible prop-
erty, therefore presentationals involving visual-perception verbs may introduce a
visible moving entity/substance in a given location, as in (47a) or introduce the
event, as in (20)–(21), (39)–(40), (55c), and (56c).
(55) a. Ameisen wimmeln in der Küche.
ants swarm in the kitchen
‘Ants are swarming in the kitchen.’
b. Die Küche wimmelt von Ameisen.
the kitchen swarms with ants
c. Es wimmelt von Ameisen in der Küche. (Hoeksema 2008: 3)
It swarms with ants in the kitchen
‘In the kitchen it is swarming with ants.’
(56) a. Mieren krioelen in de keuken.
Ants crawl in the kitchen
‘Ants are crawling (in large numbers) in the kitchen.’
b. De keuken krioelt van de mieren.
The kitchen crawls with the ants
‘The kitchen is crawling with ants.’
c. Het krioelt van de mieren in de keuken.
It crawls with the ants in the kitchen
‘In the kitchen it is crawling with ants.’ (Hoeksema 2008: 3–4)
The discourse-pragmatic functions of presentationals discussed in this section

are exploited in the following final section of the paper, which explores non-
prototypical subjects in the Lithuanian swarm alternation. The main purpose of
Section 5 is to argue for splitting ‘subjecthood’ into primitives on three linguis-
tic levels, viz., lexical (semantic), syntactic, and pragmatic, in order to explain
divergences between the three types of primitives of subjecthood displayed by the

Lithuanian swarm alternation.
5. Non-prototypical subjects in the Lithuanian swarm alternation
The present section will start with attempting an answer to the question: Why is
the procedure of splitting up ‘subjecthood’ into primitives revealing when trying
to explain subject properties in the Lithuanian swarm alternation? First of all,
properties conditioning alternations in argument realization should be discussed,
with particular reference to the swarm alternation. As we saw in the previous
sections, the alternations in subject realization illustrated by the swarm alterna-
tion usually take place when NP candidates for subjecthood display a decrease in
certain properties on semantic, discourse-pragmatic or (morpho)syntactic levels
of representation (or to be more precise, the semantics of the swarm-alternation
verbs condition the lack of certain subject properties in their arguments). For
instance, the discourse-pragmatic functions of presentationals typically require
agentive predicates to be construed as pragmatically non-agentive, therefore agent
arguments which the corresponding ‘deagentivized’ predicates are predicated of
are also treated pragmatically as non-agentive (cf. Section 4). Secondly, the prop-
erties of ‘deagentivized’ predicates and their agent arguments correlate with se-
mantic and (morpho)syntactic properties of the same agent arguments, i.e., these
may be or are obligatorily marked with the partitive genitive, which indicates a
decrease in agentive properties, cf. (57) (also see Section 3.1 for more detail).
(57) a. Akropol-yje knibždėjo žmoni-ų, keliauj-anči-ų
Acropolis-loc.sg swarm.pst.3 people-gen.pl travel-ppra-gen.pl
su įvairi-omis turistin-ėmis firm-omis.39
with various-ins.pl travel-ins.pl agency-ins.pl
‘People traveling with various travel agencies were swarming on the
Acropolis.’
b. Akropol-yje knibždėjo žmon-ės, keliauj-ant-ys
Acropolis-loc.sg swarm.pst.3 people-nom.pl travel-ppra-nom.pl
su įvairi-omis turistin-ėmis firm-omis.
with various-ins.pl travel-ins.pl agency-ins.pl
‘People traveling with various travel agencies were swarming on the
Acropolis.’
39. http://perkunas.byethost33.com/filosofija/sofija/sofija20.html
c. Akropol-is knibždėjo (nuo) žmon-ių.

Acropolis-nom.sg swarm. pst.3 (with) people-nom.pl
‘The Acropolis was swarming with people.’
This decrease in agentivity properties, based on semantic and pragmatic grounds

and correspondingly marked on the (morpho)syntactic level of representation,
enables the other event participant, viz. the location participant − when it is in-
terpreted as ‘filled’ by non-locational participants and thus fully or somewhat
affected – to be an appropriate candidate for subjecthood. Thus, alternations in
argument realization, i.e., assignment of the basic syntactic positions (subject or
object) to different arguments, are possible only when the relevant arguments
display certain discrepancies between properties of semantic, pragmatic or
(morpho)syntactic levels of representation that are typical of the prototypical sub-
ject of a certain language. That is, alternations in argument realization take place
when a certain argument does not display all or a sufficient number of properties
on distinct levels of representation that would guarantee its being assigned subject
or object position. It seems that arguments displaying all or a sufficient num-
ber of properties on the relevant levels of representation that are characteristic
of a prototypical subject40 do not allow alternations in argument realization (i.e.,
alternations involving a single set of arguments, as the swarm alternations, loca-
tive alternations etc.). Semantic properties of the verbs displaying alternations in
argument realization account for their arguments not having a sufficient number
of properties to let us conceive them as the only and undoubted candidates for
subjecthood.41 Therefore, in order to explain the different patterns of subject as-
signment in Lithuanian swarm alternations we must look for the properties/ab-
sence of the properties of distinct levels of representation that are characteristic of
prototypical subjects. Though swarm-alternation constructions obligatorily lack
certain properties that prototypical subjects would have, at least in some vari-
ants of these alternations (cf. 57b–c), the respective arguments are still treated as
40. For the properties typically associated with prototypical subject/objects cf. Dowty (1991),
Van Valin and LaPolla (1997) etc. Cf. also Hopper and Thompson’s (1980) famous study on
transitivity, where properties typical of both subjects and objects are discussed in detail, since
transitivity is traditionally treated as a property of the entire clause, involving an activity be-
ing ‘transferred’ from agent to patient, so that transitivity obligatorily entails at least two event
participants that are candidates for subjecthood and objecthood respectively.
41. For instance, according to Levin (2006), change of state verbs, such as break, crack, widen
etc., do not allow alternations in object realization. Also it is important to note that it is precise-
ly the patient arguments of change of state verbs that display the greatest number of properties
characteristic of objects. Cf. also numerous other studies by Levin and Levin and Rappaport
Hovav on the alternations in English.
subject-like enough, i.e., as displaying some of the properties of prototypical

subjects. That is, even though subjects of all instances of the swarm alternation
fall short of prototypical subjecthood, they still may be considered subjects on
some level/levels of representation even if not on all. As stated in the Introduc-
tion, we adopt Livitz’s (2006) three levels and three corresponding primitives of
subjecthood.
A prototypical subject is most prominent at each level, i.e., a prototypical
subject combines all three primitives. The three primitives can, however, operate
independently of each other, producing subjects that are ‘prototypical’ to a cer-
tain degree (ibid.: 78–79). Hence, instances of non-prototypical subjecthood can
be explained as a dissociation between properties of the three types of subjects.
Livitz proposes the primitives-based subjecthood analysis mainly for the purpose
of explaining non-nominative subjects in Russian and similar instances in other
languages (Korean, Kinande); her account is couched in Minimalist terms. Livitz’s
analysis of Russian non-nominative subjects is based on the fact that in Russian
there is a strict requirement for an overt subject (including non-nominative sub-
jects) to be realized in directly preverbal position and a strict constraint on verb-
initial sentences (ibid.: 4–5, 9–11). An exception to the general restriction against
verb-initial sentences is presentationals, or in Livitz’s terminology “constructions
in a story telling context”, cf. (58a–b) and also verb-initial sentences with the into-
national emphasis of narrow focus on the sentence-final element (which usually
occurs in preverbal position in intonationally neutral instances). Thus, according
to Livitz, the obligatory requirement for the subject to be placed in preverbal posi-
tion serves as the basis for Russian non-nominative subjects (since they obligato-
rily take preverbal position) still to be considered subjects at some level/levels of
representation.
(58) a. Sobralis’ kak-to tri kolleg-i…
got.together.pst.pl.3 once three collegues-nom.pl
‘Once three colleagues got together.’
b. Žili-byli ded i bab-a. I
lived-was.pst.pl.3 old.man.nom.sg and old.women-nom.sg and
byla u nix kuročk-a rjab-a.
be.pst.sg.f to them hen-nom.sg speckled-nom.sg.f
‘There once lived an old man and an old woman. And they had a speckled
hen.’ (Livitz 2006: 10)
Lithuanian also demonstrates a strong tendency for subject realization in prever-

bal position and the same exceptions to the general tendency. Nevertheless, in
both languages examples contradicting the general tendency may be found, cf.
(59) and (60) for Russian and Lithuanian respectively. Since presentationals are
minimally characterized by the presence of a pitch accent on the subject and by

the absence of prosodic prominence on the predicate portion of a sentence (cf.
Section 4), and the given examples display prosodic prominence on the intransi-
tive predicate, as in (59), and on the direct object of ditransitive predicates, as
in (60), it is obvious that the given verb-initial sentences are PF constructions
rather than presentationals. Therefore, in Lithuanian (and probably in Russian)
the correlation between subjecthood and verb-initial position may be treated as a
strong tendency rather than a strict requirement, and it cannot serve as a basis for
an analysis of non-prototypical subjects in the Lithuanian swarm alternation. In
the present paper the alternations in argument realization, being a consequence
of the absence of certain properties that are typically associated with subjects, is
treated as a sufficient basis for proposing non-prototypical subjects and a justifi-
cation for splitting up subjecthood into primitives on three linguistic levels. This
paper does not, however, aspire to provide a complete description of the mecha-
nisms of realization of the three types of subjecthood primitives.
(59) Rovno polgod-a nazad ja possorilsja s
just half.year-gen.sg ago 1.sg get.angry.pst.sg.1.m with
otc-om i ušel iz dom-a. <…> Ušel
father-ins.sg and leave.pst.sg.m from home-gen.sg leave.pst.sg.m
ja iz-za to-go, čto otec zapretil
1.sg because.of this-gen.sg that father.nom.sg prohibit.pst.sg.m
mne stroit’ otnošenij-a s devušk-oj.42
me.dat build.inf relationship-nom.pl with lass-gen.sg
‘Just half a year ago I got angry with my father and left home. I left home for
the reason my father prohibited me to build a relationship with my lass.’
(60) <…> paskolinau klasiok-ui fleš-ą, šiandien atidavė
loan.pst.sg.1 classmate-dat.sg flash-acc.sg today return.pst.3
jis man t-ą fleš-ą <…>.43
3.sg me.dat that-acc.sg flash-acc.sg
‘I loaned a USB flash drive to a classmate. Today he returned the USB flash
drive.’
The primitives of subjecthood should now be introduced in more detail. According

to Livitz, lexical (semantic) subjecthood is defined by assignment of the thematic
subject: “Thematic Subject is the argument that receives the highest Thematic role
42. http://www.podslyshano.ru/doma-ne-pojmut/17250-rovno-polgoda-nazad-ia-possorilsia-
s-otcom-i-yshel-i-doma.html
43. http://www.nea.lt/wap;forum;zinutes;it;a4cdd3ba;1;ASC;9ruadtu06tr608utu7b906mev7
from the predicate, according to the traditional Theta-role hierarchy” (ibid.: 76).

Livitz adopts the theta-role hierarchy proposed by Belleti and Rizzi (1988):
Agent > Experiencer > Theme > Goal/Source/Location > Manner/Time
In the case of the prototypical subject, the thematic subject must also become
both grammatical and discourse subject. In the Lithuanian swarm alternation,
non-locational arguments may have the thematic roles of agent ((57) and (61)),
theme (62), and source (63a); location participants are obligatorily included in the
argument structure of the verbs (cf. Section 3.1). With respect to the above-pre-
sented theta-role hierarchy, agent or theme arguments, as compared with location
arguments, are the ones that undoubtedly receive the highest thematic role from
the predicate and, consequently, become thematic subjects. However, source and
location thematic roles share a level on the hierarchy, therefore neither of them
is a better candidate for thematic subjecthood. Hence, their relatively low posi-
tion in the theta-role hierarchy and the fact that both arguments have theta-roles
of the same rank may also be considered as a factor conditioning alternation in
subject realization. In Lithuanian swarm alternations the thematic role of source
is assigned to the relevant arguments only by smell-emission verbs, as in (63a).
In principle, anything may be perceived as having a certain smell and it does not
matter whether it is an inherent property of a certain thing or not.
In Lithuanian, grammatical subjecthood is determined by subject-verb
agreement and nominative case marking. It is precisely the (morpho)syntactic
properties that are indicative of grammatical relations, therefore the nominative
case-marked NPs of various instances of the Lithuanian swarm alternation (cf.
(57b), (61b), (63a–b)) could arguably be considered subjects, though they lack
other prototypical subject properties on other levels of representation. Moreover,
these (morpho)syntactic properties are the most constant indicator of subject sta-
tus, since both thematic subjecthood and the topical status of discourse subject-
hood are fairly unstable. Nevertheless, as mentioned above, in order to explain
alternations in subject realization we need to account for the presence/absence of
characteristic subject properties on all levels of representation.
The last primitive of subjecthood is discourse subjecthood. According to
Lambrecht, there exists a strong correlation between subject and topic in the
topic-comment type of sentences, and it is a universal feature of natural language
that the subject of a sentence will be interpreted as topic and the predicate as a
comment about this topic unless the sentence contains morphosyntactic, pro-
sodic, or semantic clues to the contrary. “The subject can therefore be charac-
terized as the unmarked topic expression and the topic-comment structure
as the unmarked presuppositional structure of a sentence” (1994: 136). In
exploring alternations in subject realization it is important to note that some vari-

ants of the Lithuanian swarm alternation display a departure from the typical
subject and topic correlation caused by the discourse-pragmatic properties of
presentationals.
To conclude the section, the interaction between the three primitives of sub-
jecthood in the Lithuanian swarm alternation should be discussed. Such sentenc-
es as (57) and (61)–(63) demonstrate the main types of the Lithuanian swarm
alternation with respect to the thematic roles of non-locational participants. In
(57) and (61) the thematic roles assigned to arguments by the predicates are those
of agent and location respectively, in (62) and (63b–c) those of theme and loca-
tion, in (63a) those of source and location.
(61) a. Dan-ėje besidarbuoj-ant ugniagesi-ams
Dane-loc.sg work-cvb.prs fireman-dat.pl
krantin-ėse prisirinko smalsuoli-ų <…>44
embankment-loc.pl gather.pst.3 inquisitive-gen.pl
‘While firemen were working in the Dane, inquisitive people gathered on
the embankments [of the river Dane].’
b. Up-ės krantin-ės prisirinko smalsuoli-ų.
river-gen.sg embankment-nom.pl gather.pst.3 inquisitive-gen.pl
‘The embankments of the river were packed with inquisitive people.’
(62) a. Kiem-e priaugo krūm-ų,
courtyard-loc.sg grow.pst.3 bush-gen.pl
sod-as – neįžengiam-as…45
garden-nom.sg impassable-nom.sg
‘Bushes were growing in the courtyard [and] the garden was
impenetrable.’
b. Kiem-as priaugo krūm-ų.
courtyard-nom.sg grow.pst.3 bush-gen.pl
‘The courtyard was overgrown with bushes.’
(63) a. Kiem-e kvepėjo balt-ų roži-ų
courtyard-loc.sg smell.pst.3 white-gen.pl rose-gen.pl
krūmai.
bush-nom.pl
‘In the courtyard white rose bushes smelled.’
44. http://www.15min.lt/naujiena/aktualu/lietuva/danes-upeje-pasirode-naftos-tersalai-56-
202604
45. http://burgis.lt/mirusiuju-miestai-medziai-kelias
b. Kiem-as kvepėjo rož-ėmis.

courtyard-nom.sg smell.pst.3 rose-ins.pl
‘The courtyard smelled of roses.’
c. Kiem-e kvepėjo rož-ėmis. (CCLL)
courtyard-loc.sg smell.pst.3 rose-ins.pl
‘In the courtyard there was a smell of roses.’
In the instances with nominative agent arguments, as in (57b), the given argu-
ment is both thematic and grammatical subject, but it displays a departure from
the properties of discourse subjecthood, i.e., for the discourse function of pre-
sentational construction it is ‘deagentivized’ and construed as the focus of the
sentence (see Section 4 for more detail). In other instances with agent argu-
ments, as in (57a) and (61a), the given argument is obligatorily (61a) (partitive
genitive is required by the verbal prefix pri-) or optionally (57a) marked with a
partitive genitive. Hence, the given agent arguments are thematic, but they are
neither grammatical nor discourse subjects. In the instances with theme argu-
ments, as in (62a), the given argument is the thematic subject, since it receives
the highest theta-role from the predicate, but it is neither a grammatical subject
(due to the obligatory partitive genitive marking) nor a discourse subject (because
of the pragmatic functions of presentational sentences). In (63c) the predicate
also assigns the thematic roles of theme and location to the respective arguments,
however the pattern shows a spread of subject properties over the two nominal
constituents of the sentence, triggered by the discourse function of the sentence
(cf. Section 4). Hence, in instances like (63c), the theme argument may be consid-
ered thematic subject, while the location argument may be considered discourse
subject (due to its topical status). However, neither of them is grammatical sub-
ject. The source thematic role in Lithuanian swarm alternations is displayed only
by smell-emission verbs, cf. (63a). In this sentence the source argument can be
considered grammatical subject, but not discourse or thematic subject. On the
thematic hierarchy presented above, the thematic roles of source and location
are at the same level, so that neither is eligible for thematic subjecthood. As for
the location-subject variant, as in (57c), (61b), (62b), and (63b), here the location
arguments are both grammatical and discourse subjects, while the non-location
argument is the thematic subject.
Thus, in neither variant of Lithuanian swarm alternation can either of the
two arguments be considered a prototypical subject displaying subjecthood on all
levels of representation. Moreover, all the variants display spread of subject prop-
erties of distinct levels over both arguments of predication. Location arguments
are always discourse subjects. In some instances, as in (57a–b), (61a), (62a), (63a),
and (63c), the discourse subjecthood of location arguments is triggered by the
discourse function of presentationals represented by the sentence. However, in

the location-subject variant, as in (57c), (61b), (62b), and (63b), the location ar-
guments receive discourse subject status by virtue of their specific interpretation
(i.e., ‘filling’ semantics) receiving nominative marking (and becoming grammati-
cal subjects) and discourse topic-comment structure.
Furthermore, all nominative arguments are considered grammatical subjects
despite lacking properties of other types of subjecthood. As for the instances in
which neither argument has nominative marking, as in (57a), (61a), (62a), and
(63c), neither of them may be considered grammatical subject, in other words
these sentences may be considered impersonal. In Malchukov and Ogawa’s (2011)
terms, they are T-impersonals, i.e., impersonals sensitive to topicality. In instances
such as (57a), (61a), the agent arguments have partitive genitive marking denot-
ing a decrease in agentivity properties (see Section 3.1 for more detail). The given
semantic and (morpho)syntactic properties correlate with the requirement of the
discourse-pragmatic presentationals for agentive predicates, and, consequently,
for their agent arguments, to be construed as non-agentive. Hence, the agentive
status of the corresponding arguments is neutralized on all levels of representa-
tion. As mentioned above, in such instances as (63c), displayed by smell-emission
verbs, the theme participants may be considered thematic subjects, but due to the
discourse-pragmatic function of presentationals they do not receive either the
status of topic of the proposition or nominative case-marking, cf. Section 4 on the
spread of subject properties over (pro)nominal constituents of the sentence as a
consequence of the discourse function of presentationals.
6. Conclusion
Alternations in argument realization, i.e., variations in assignment of the basic

syntactic positions of subject or object to different arguments of a predicate, take
place when the semantics of the predicates involved is such that their arguments
lack all or a sufficient number of the properties on different levels of representa-
tion that are typical of subjects. Not surprisingly, therefore, in neither variant of
the Lithuanian swarm alternation can either of the two arguments be considered
a prototypical subject, i.e., displaying subjecthood on all levels of representation –
lexical (semantic), grammatical and discourse. Furthermore, all instances of the
Lithuanian swarm alternation display a spread of subject properties of different
levels of representation over both nominal arguments of the clause. The spread
of subject properties over both (pro)nominal arguments is conditioned by a set
of discourse-pragmatic and semantic features characteristic of the Lithuanian
swarm alternation.
First of all, on the discourse-pragmatic level of representation, some variants

of the Lithuanian swarm alternation function as presentationals. One of the fun-
damental properties of presentationals is that pragmatically their subjects are not
construed as topics, but as sentence foci, and, in Lithuanian as in other languages,
presentationals are obligatorily marked by a focus accent on the subject. Focus
accent-marking is typically an object feature. Therefore, in presentationals, the
marking of the subject is, in a way, characteristic of the object, partially neutral-
izing its subject properties, and the other argument (i.e., the location argument)
may take over part of the subject’s properties and even be realized as grammatical
subject. Thus, the spread of subject properties over nominal NPs of a clause dis-
played by presentationals is a consequence of the focal status of the subject.
Another fundamental property of discourse-pragmatic presentationals illus-
trated by the Lithuanian swarm alternation is that they introduce a new referent
into discourse, but do not convey information about the newly-introduced refer-
ent. Therefore presentationals do not realize the topic-comment type of prop-
osition; rather, they are a kind of thetic judgments. For similar reasons, some
instances of the Lithuanian swarm alternation that function as presentationals
and involve agentive predicates are pragmatically construed in such a way that
the agentive properties of the predicates are subordinated to the presentational
function of the proposition, and the predicates and their agent arguments are in
fact construed as reduced in agentivity. This irrelevance of agentivity properties
for the discourse-pragmatic functions of presentationals is supported by the fact
that in some instances of the Lithuanian swarm alternation that involve agentive
predicates, agent arguments are preferably /obligatorily marked with the partitive
genitive, which is indicative of a decrease in agentivity properties.
Other factors contributing to discrepancies between typical subject proper-
ties on different levels of representation are semantic-pragmatic in nature. The
crucial semantic property for the Lithuanian swarm alternation is that a position
for a locative argument is obligatorily required in predicate-argument structure.
The given semantic property correlates with the property of discourse-pragmatic
presentationals consisting in presenting or introducing an NP referent into the
‘space’ or ‘scene’ of discourse rather than of asserting its mere existence. Another
property of discourse presentationals is that their verbs must be information-
ally light in context. The informational lightness of verbs provides a possibility
of introducing a discourse-new participant and contributes to realizing the main
function of the presentational constructions. The presentational constructions
involved in the Lithuanian swarm alternation may be treated as predicating ad-
ditional information beyond the mere existence of the entity and thus not being
informationally light in context. However, in Lithuanian swarm-alternation pre-
sentational constructions, both verbs and the referents of their NPs are typically
mutually predictable, which causes the verbs to be treated as informationally light

in context. Furthermore, the semantic properties of non-locational participants
as a group of non-discrete and non-individualized entities/substances require
them to be realized as indefinite plurals or mass terms, not as singular NPs. These
semantic properties also conform to the pragmatic preference of thetic proposi-
tions, of which presentational constructions are a subtype, for NPs with uniden-
tifiable referents.
All the above-mentioned discourse-pragmatic and semantic properties of
predicates/arguments/alternating constructions contribute to the discrepancies
between properties characteristic of subjects on different levels of representa-
tion; thus, in neither variant of Lithuanian swarm alternations can either of the
two arguments be considered a prototypical subject. Nevertheless, the nomi-
native-marked arguments of the Lithuanian swarm alternation are considered
grammatical subjects despite lacking subject properties on the other levels of rep-
resentation, i.e., lexical (semantic) and discourse (though different instances of
nominative arguments may have distinct functional motivations). Instances of
the Lithuanian swarm alternation in which neither argument is nominative can
be regarded as a kind of impersonal constructions, viz., topicality-sensitive im-
personal constructions.
Constructions with two oblique-marked NPs displayed by smell-emission
verbs serve to denote the presence/existence of a smell without presupposing
presence of the source of that smell in a given location. Thus, the main factor con-
tributing to realizing constructions with two oblique-marked NPs is the semantic
properties of smell-emission verbs, viz. the ability of these predicates to realize a
participant denoting a smell separately from its source.
Abbreviations
acc accusative loc locative

cvb converb m masculine
dat dative n neuter
def definite nom nominative
expl expletive pl plural
f feminine ppa past participle active
gen genitive ppl past participle
imp imperative ppp past participle passive
inf infinitive pppra present active participle
ins instrumental sg singular
Sources
CCLL Corpus of the Contemporary Lithuanian Language. http://tekstynas.vdu.lt/tekstynas/

MLD Modern Lithuanian Dictionary. http://dz.lki.lt/static/english.html
RNC Russian National Corpus. http://www.ruscorpora.ru
References
Anderson, Stephen. 1971. The role of deep structure in semantic interpretation. Foundations of
Language 6: 387–396.
Bresnan, Joan. 1994. Locative Inversion and Architecture of Universal Grammar. Language 70:
72–131. DOI: 10.2307/416741
Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67: 547–619.
Dowty, David. 1999. Argument Alternations, Lexical Polysemy and Thematic Roles: A Case
Study. http://www.ling.ohio-state.edu/~dowty/swarm-bees.hdt.norway
Dowty, David. 2001. The Semantic Asymmetry of ‘Argument Alternations’ (and Why it Mat-
ters). In Groninger Arbeiten zur germanistischen Linguistik 44. Groningen: Center for Lan-
guage and Cognition. http://ling.ohio-state.edu/~dowty/papers/groningen-00.pdf
Dowty, David. 2000. ‘The Garden Swarms with Bees’ and the Fallacy of ‘Argument Alternation’.
In Polysemy. Theoretical and Computational Approaches, Yael Ravin & Claudia Leacock
(eds), 111–128. Oxford: Oxford University Press.
Fried, Mirjam. 2005. A frame-based approach to case alternation: the swarm-class in Czech.
Cognitive Linguistics 16(3): 475–512. DOI: 10.1515/cogl.2005.16.3.475
Gast, Volker & Haas, Florian. 2011. On the distribution of subject properties in formulaic pre-
sentationals of Germanic and Romance: A diachronic-typological approach. In Imper-
sonal Constructions, Anna Siewierska & Andrej Malchukov (eds), 127–167. Amsterdam/
Philadelphia: John Benjamins.
Hoeksema, Jack. 2008. The swarm alternation revisited. In Theory and Evidence in Semantics,
Erhard Hinrichs & John Nerbonne (eds), 53–80. Stanford: CSLI Publications.
Holvoet, Axel. 2005. Intranzityvinių sakinių tipai: egzistenciniai, lokatyviniai ir posesyvin-
iai sakiniai [Types of intransitive clauses: existential, locative and possessive clauses].
In Gramatinių funkcijų tyrimai [Studies in Grammatical Functions], Axel Holvoet &
Rolandas Mikulskas (eds), 139–160. Vilnius: Lietuvių kalbos institutas.
Jackendoff, Ray S. 1990. Semantic structures. Cambridge, MA: MIT Press.
Kalėdaitė, Violeta. 2002. Existential Sentences in English and Lithuanian. A Contrastive Study
[European University Studies, Linguistics 248]. Frankfurt am Main: Peter Lang.
Keenan, Edward. 1976. Towards a Universal Definition of ‘Subject’. In Subject and Topic: A
New Typology of Language, Charles N. Li & Sandra Thompson (eds.), 303–333. New York:
Academic Press.
Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus and the Men-
tal Representation of Discourse Referents. Cambridge: Cambridge University Press. DOI:
10.1017/CBO9780511620607
Lambrecht, Knud. 2000. When subjects behave like objects: An analysis of merging of S and
O in Sentence-Focus constructions across languages. Studies in Language 24(3): 611–682.
DOI: 10.1075/sl.24.3.06lam
Lenartaitė, Kristina. 2010. Argumentų raiškos alternavimas lietuvių kalboje [Alternations in

Argument Realization in Lithuanian]. PhD thesis, Vilnius: Institute for the Lithuanian
Language.
Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. Chicago/
London: University of Chicago Press.
Levin, Beth. 2006. English Object Alternations: a Unified Account. Unpublished draft. Stanford
University. http://www.stanford.edu/~bclevin/alt06.pdf
Levin, Beth & Rappaport Hovav, Malka. 1995. Unaccusativity: At the Syntax–Lexical Semantics
Interface. Cambridge, MA/London: MIT Press.
Levin, Beth & Rappaport Hovav, Malka. 2005. Argument Realization. Cambridge: Cambridge
University Press.
Livitz, Inna. 2006. What’s in a Nominative? Implications of Russian Non-Nominative Subjects for a
Crosslinguistic Approach to Subjecthood. BA thesis, Cambridge, MA.: Harvard University.
Malchukov, Andrej & Akio Ogawa. 2011. Towards a typology of impersonal constructions.
A semantic map approach. In Impersonal Constructions, Anna Siewierska & Andrej
Malchukov (eds), 19–57. Amsterdam/Philadelphia: John Benjamins.
Padučeva, Elena V. 2007. Genitiv otricanija i Nabljudatel’ v glagolax tipa zvenet’ i paxnut’ [Gen-
itive of negation and Observer with verbs of the type zvenet’ and paxnut’]. http://lexico-
graph.ruslang.ru/TextPdf1/paxnut_2007.pdf
Portero Muñoz, Carmen. 2011. A Functional Discourse Grammar approach to the Swarm-al-
ternation as a case of conversion. In Morphosyntactic alternation in English: functional and
cognitive perspectives, Pilar Guerrero Medina (ed), 316–338. Sheffield: Equinox.
Rowlands, Rachel. 2002. Swarming with bees: Property predication and the swarm alterna-
tion. MA thesis, University of Canterbury. www.lacl.canterbury.ac.nz/ling/documents/
Rowlands_Thesis.pdf
Salkoff, Maurice. 1983. Bees are swarming in the garden: A systematic synchronic study of
productivity. Language 59(2): 288–346. DOI: 10.2307/413576
Van Valin, Robert D. & LaPolla, Randy L. 1997. Syntax: Structure, Meaning and Function.
Cambridge: Cambridge University Press. DOI: 10.1017/CBO9781139166799
Subjecthood in specificational copular
constructions in Lithuanian
Rolandas Mikulskas
Institute for the Lithuanian Language / Vilnius University
Superficially, copular sentences serving specificational function, such as The

winner of the race is John or The tallest girl in the class is Molly, look like inverted
structures as compared with their more usual predicative counterparts, cf. John
is the winner of the race and Molly is the tallest girl in the class. Though in terms
of word order the specificational copular construction can be thought of as
derived from its predicative counterpart by means of inversion, this inversion
is strongly motivated by the communicative demands of specification and adds
new structural properties to the construction: the former predicative nominal
‘moved’ into precopular position inevitably acquires an existential presuppo-
sition with regard to its unique referent, thus exhibiting role definiteness, and
the information structure of the construction becomes bound to the different
referential functions of its two nominals (the role defining NP is always a topic
and the role specifying NP a comment). There are thus good functional and
structural reasons to treat this specificational construction as a distinctive sub-
type of copular predication. Syntactically, however, the specificational copular
construction, as compared to its predicative counterpart, remains, in many re-
spects, a non-canonical predication. For instance, pseudoclefts (which represent
a special case of the specificational copular construction), pose serious challeng-
es to Binding Theory because of certain well-known connectivity effects. An-
other problem with this kind of copular predication − and this will be the topic
of this article − is the inconsistency of subject assignment in such constructions
across languages (and sometimes even within the same language). For example,
in English, Danish, Swedish and French, the morphosyntactic marking of the
subject is conferred on the first nominal of the specificational copular construc-
tion, while in Lithuanian, Russian, as well as in Italian and German, the second
nominal of the construction acquires this marking. The fact that semantically
equivalent structures acquire opposite patterns of morphosyntactic coding of
the main grammatical relation, suggesting opposite directions of conceptual-
ization of essentially the same specificational relation, poses a challenge to the
main principle of Cognitive Grammar, the so-called content requirement. If one
182 Rolandas Mikulskas
sticks rigorously to the morphosyntactic coding used in specificational copular

sentences, one encounters difficulties with defining grammatical relations in
terms of conceptual relevance, i.e., as a trajector / landmark configuration.
1. Introduction1
1.1 Preliminary remarks
In the literature, problems of specificational predication are mainly discussed in

the context of (inverted) copular constructions (as copulas we can consider not
only be (Lith. būti), but its aspectual (lexical or morphological) counterparts be-
come, remain (Lith. tapti, likti, tebe-būti) as well). This is not a frequent topic,
though the discussion of theoretical and descriptive issues of predication in copu-
lar clauses has a long history in linguistics (cf. Higgins 1973; Heggie 1998; Moro
1997; Rothstein 2001: 205−338 inter alia). Examples of prototypical specification-
al copular clauses are:
(1) The winner of the race is John.
(2) The president of the club is Peter.
Their exact counterparts in Lithuanian and Russian, two languages structurally

close to each other, would be, correspondingly:
(1) a. Varžyb-ų nugalėtoj-as yra Jon-as.Lithuanian
race-gen winner-nom be.prs.3 John-nom
(2) a. Klub-o prezident-as yra Petr-as.
club-gen president-nom be.prs.3 Peter-nom
(1) b. Pobeditel’ sorevnovanij – Ivan.Russian
winner.nom race.gen.pl John.nom
(2) b. Prezident klub-a − Pëtr.
president.nom club-gen Peter.nom
The same sort of specificational predication can be found in Lithuanian and

Russian constructions with causative predicates (of the type to appoint, to elect),
which contain, as a constituent part, a copular predication (of state) − their gen-
eral meaning is ‘to cause sb. to become / be sth.’:
1. The pre-final draft of this paper has benefited greatly from comments by Peter Arkadiev,
Ilja Seržant and Nicole Nau. I am also grateful for the comments of the anonymous reviewers.
All faults and shortcomings are mine.
(3) a. Klub-o prezident-u buvo Lithuanian

club-gen president-ins.sg be.pst.3
paskir-t-as / išrink-t-as Jon-as.
appoint-ppp-nom.sg.m elect-ppp-nom.sg.m John-nom
b. Prezident-om klub-a byl naznačen / Russian
president-ins club-gen be.pst.m appoint.ppp.nom.sg.m
izbran Ivan.
elect.ppp-nom.sg.m John.nom
But cf.
c. *The president of the club was appointed / elected John.English
However, we find specification not only in the domain of copular constructions

proper, but also:
a. in loose appositional constructions:

(4) The greatest Victorian poet, Tennyson, died early.
(example from Van Langendonck 2007: 129)
b. in Lithuanian in reinterpreted comparative constructions of the type toks

NP1 kaip NP2 (‘such NP1 like NP2’), which represents a special case of close
apposition (especially those with case attraction):
(5) Toki-o žmog-aus kaip Jon-o į sveči-us
such-gen person-gen.sg like John-gen to guest-acc.pl
niek-as nekvies
nobody-nom neg-invite.fut.3
‘Nobody will invite (to his home) a person like John’
c. in Lithuanian mono-clausal counterparts of English pseudoclefts (where case

attraction is also observed):
(6) Ką mėgstu, tai raudon-ą vyn-ą.
what.acc like.prs.1sg that red-acc wine-acc
‘What I like is red wine’
In this paper, while analyzing some alignment2 problems in specificational claus-

es, I restrict myself to copular structures only. The choice is not accidental because
2. I use alignment here as a cover term encompassing morphosyntactic encoding (case mark-
ing and agreement) of core arguments, which in copular predication amount to subject NP and
predicative complement (NP, ADJ or PP). In the cognitive perspective adopted in this paper,
this term naturally also encompasses the linking of grammatical relations in specificational
copular constructions to relevant semantic functions of their NPs, viz. role definition and its
specificational copular constructions are closely related to other types of copular

constructions and largely inherit the descriptive and theoretical problems associ-
ated with these (especially those concerning the classification of subtypes, align-
ment, semantics of the copula, and the referentiality of the two NPs in the copular
construction).
1.2 Formal approaches to specificational copular constructions:

Challenges for Binding Theory
In recent decades a set of specific problems pertaining to specificational copular

constructions have been a source of concern for formal semanticists and genera-
tive grammarians. The former are mainly concerned with establishing the type
of copular constructions represented by specificational constructions, a problem
closely related to the semantics of the copula (i.e., whether we should posit two
copulas, be1, and be2, or just one be) and the referentiality status of the precopular
and postcopular NPs in the construction (definite, indefinite, specific). Are speci-
ficational constructions an inverted subtype (with discourse-motivated fronting
of the second NP, cf. Heggie 1988) of a predicative copular construction of the
type NP1 be NP2, which is assumed to be the basic type of copular predication
(Williams 1983; Partee 1986), or do they represent a special (‘asymmetric’) in-
stance of the equative predication (Heycock & Kroch 1998, 1999; Heycock 2012;
Heller 2005)?
Proponents of the generative approach derive specificational as well as pred-
icative copular constructions from underlying structures with small clauses, the
difference between the two types depending on which NP – the ‘logical subject’
(in the predicative construction) or the ‘predicative’ (in the specificational con-
struction) – is raised into the sentence-initial SpecIP position (Moro 1991, 1997;
Mikkelsen 2004). In some languages, like English, Danish, Swedish, French, the
‘raised’ predicative NP becomes ‘syntactic subject’ of the specificational copular
clause. Thus, for generative grammarians, the copula be3 is a raising verb of the
specification. As such, this term is comparable to the terms (distribution of) grammatical rela-
tions and argument linking.
3. Caroline Heycock and Anthony Kroch (1999: 381−382) extend the small clause analysis to
specificational constructions with the aspectual verbs remain and become, cf.:
(i) The best solution remains instant retreat.
(ii) At this point our real problem becomes John.
same sort as consider, believe, expect etc.4 Movement of the inherently predica-
tive NP into SpecIP position is motivated by contextual presuppositions: the
first NP of a specificational construction is universally topical (‘discourse-old’)
(Mikkelsen 2004).
One of the major concerns for generativists, though, is the challenge posed
to syntactic binding theory by the so-called connectivity effects in pseudoclefts,
which (on one reading) represent a special case of specificational copular con-
structions. ‘Connectivity effects’ is a general term for a cluster of syntactic and
semantic phenomena, such as (the principles of) distribution of anaphoric el-
ements, Negative Polarity Items (NPIs) licensing, de dicto readings of NPs in
opaque contexts (i.e., under the scope of an intensional operator) etc., that are
usually assumed to hold only under a c−command configuration, but show up
in environments where this configuration is absent. It is generally assumed in
generative theory that the c−command configuration, as a structural principle
of syntactic representation, holds in a local domain (i.e., under the TP node).
Thus the prediction is that in English pseudoclefts, which are bi-clausal struc-
tures, a nominal of the postcopular phrase cannot be c−commanded by the sub-
ject nominal, which is ‘buried’ inside the precopular (free relative) clause. In fact
we find postcopular nominals that behave as if they were c−commanded, i.e., as
if a c−command configuration in pseudoclefts held across their copula. That the
syntactic locality condition is violated in pseudoclefts is clearly shown by the fact
that with respect to the well known Principles A, B, and C (illustrated respectively
in the following examples) of the standard Binding Theory (Chomsky 1981), both
NPs contained in them behave in the same way as they do in the corresponding
mono-clausal paraphrases, cf.:
(7) a. What Maryi was was proud of herselfi/*j.
b. Maryi was proud of herselfi/*j.
(8) a. What Maryi was was proud of her*i/j
b. Maryi was proud of her*i/j.
(9) a. What shei was was proud of Mary*i/j.
b. Shei was proud of Mary*i/j.
We can see, for example, that in sentence (7a) principle A of standard Binding
Theory is violated: the anaphor herself is licensed in the postcopular phrase even
4. To be more exact, the copula verb is structurally comparable to the raising-to-subject coun-
terparts of these verbs, i.e., to their passive forms, cf. Mary considers John (to be) the real mur-
derer ~ John is considered (to be) the real murderer and John is the real murderer, or it can be
compared to such original raising-to-subject verbs as seem, appear, turn out, happen etc., cf.
John seems (to be) happy and John is happy.
though it is not c−commanded by the antecedent Mary, which is embedded in-

side the precopular free relative. Despite this violation of the locality condition,
the bound reading of the sentence is grammatical.
One of the recent solutions to this problem is reached by postulating the rele-
vant c−command configuration at a more abstract level of the syntactic represen-
tation of pseudoclefts. Caroline Heycock and Anthony Kroch (1999: 365) suggest
that this level should be “more abstract than LF under most current conceptions”.
According to them, the relevant representation of these sentences “is built up in
the process of discourse interpretation and may constitute the interface with the
conceptual-intentional system of mind” (ibid.). These authors suggest that only
on the surface are pseudoclefts bi-clausal (equative) copular constructions, serv-
ing as a pragmatic ‘information-packaging’ device in discourse for focusing new
information. The relevant ‘post-LF’ level of syntactic representation, needed for
semantic interpretation (including binding relations), is reached by the pragmati-
cally driven ‘unidirectional’ derivational process in the course of which mono-
clausal structure (similar to the corresponding simple sentence paraphrases) is
arrived at by applying to the surface structure a logical ‘iota-reduction’ operation.
The mono-clausal Lithuanian pseudoclefts, which supposedly have evolved from
bi-clausal copular structures, empirically support this suggestion.
The cognitive approach: Challenges for the content requirement

1.3
principle
None of these problems bother those working in the framework of Cognitive

Grammar, as all kinds of underlying structures or raising procedures are imper-
missible there. But it seems that the varying alignment in specificational con-
structions, depending on the language type, poses no insignificant challenge to
the universality of the main principle of Cognitive Grammar, viz. the content
requirement,5 (Langacker 1987: 52−55): 488) also known as the principle of
form−meaning pairing, or the symbolic thesis (Taylor 2002: 592).
The fact that in different languages different nominals (a Cognitive Gram-
mar term roughly equivalent to the traditional notion of noun phrase; Langacker
5. The content requirement is the requirement that “the only structures permitted in the
grammar of a language (or among the substantive specifications of universal grammar) are
(1) phonological, semantic, or symbolic structures that actually occur in linguistic expressions;
(2) schemas for such structures; and (3) categorizing relationships involving the elements in (1)
and (2). Hence no descriptive constructs are permitted that lack both phonological and seman-
tic content. Furthermore, overt structures cannot be derived from hypothetical “underlying”
structures having a substantially different character” (Langacker 1987: 53−54).
1987: 126), bearing different semantic roles in specificational copular construc-

tions, show the morphosyntactic coding and syntactic behaviour of a subject,
challenges the aspiration of Cognitive Grammar to define the subject and the
object (or the predicative complement in our case) of the clause in conceptu-
al terms, as trajector and landmark (or as figures of the first and second focal
prominence) of the predicated relation. In Cognitive Grammar this definition of
the main grammatical relations is supposed to hold universally across languages
(Langacker 2008: 378−382). The aim of this paper is to tackle this problem in some
depth, and to provide an attempt at reconciling the principles of cognitive theory
with the atypical (non-canonical) syntax of specificational copular constructions.
In what follows I will argue that the main grammatical relation, i.e. the
subject, plays a minimal, if any, role in the conceptualization of specificational
copular constructions even in such subject-prominent languages as English. My
argument is based, first of all, on the empirical fact that in different types of lan-
guages different nominals, exhibiting different referential properties, receive the
morphosyntactic coding of the subject in the specificational construction under
discussion. Secondly, it is based on the observation that the asymmetry needed
for conceptualization (and processing) of the specificational relation in the con-
struction is ensured by the different referential status of its two nominals and its
fixed topic−comment structure. This asymmetry, which can, in cognitive terms,
be formulated as a trajector/landmark configuration, is, in this construction, es-
tablished independently of clausal grammar, i.e., externally, by the relevant con-
textual presuppositions.
Meanwhile, in Section 2, I will present the treatment of copular constructions
in Cognitive Grammar, which is mainly based on instantial semantics, a notion
elaborated by Ronald Langacker. In Section 3 I will provide a definition for the
specificational copular construction, which turns out to be a contextually moti-
vated subtype of the equative construction. In Section 4 I will briefly review the
definition of the grammatical relations in Cognitive Grammar and envisage diffi-
culties in implementing it in some cases of non-canonical marking. In Section 5 I
will discuss in detail the issue of subjecthood in specificational copular construc-
tions of two different types of languages. Additional arguments for my claim will
be adduced from pragmatically marked cases of the English-type specificational
constructions, which are erroneously excluded from the domain of specificational
predication by proponents of the formal approaches. In Section 6, I will present
my concluding remarks.
2. The treatment of copular constructions in Cognitive Grammar
Copular constructions of the type NP be ADJ (Jane is tall) are unproblematic in

Cognitive Grammar, because adjectives, being themselves relational predications,
naturally serve, along with the copula, the role of clausal predicates, the copula
merely providing finiteness features for the complex predicate of the clause. Such
copular sentences predicate some property of, or ascribe some feature to, the
subject nominal, and can be classified as the ascriptional subtype of the copular
construction.
The nature of the predication in copular constructions of the type NP1 be
NP2 is not so straightforward. In Cognitive Grammar it is explained through the
type–instance relation established by the predicative nominal (NP2): the referent
of the subject nominal in the copular construction is identified with an instance of
the type represented by the postcopular (predicative) nominal (on ‘instantial se-
mantics’ see Langacker 1991: 55–71). Two semantic effects can be created through
the identificational relation between two nominal referents: that of class inclusion
and that of equation. Class inclusion entails that the referent of the subject nomi-
nal, through its identification with a type instantiation, is included into a broader
class of same-type referents: typically the postcopular NP is indefinite, cf. Cicero
is an orator and (a) philosopher. Equation, in turn, is found when two specific
referents (coded by definite NPs) are equated in an identificational relation as one
and the same individual, cf. Cicero is Tully. Accordingly, the two cases represent
the class-inclusional and equative subtypes of the copular construction.
The advantage of the cognitive approach to the classification of (sub)types
of copular construction consists, firstly, in that in all cases (even in those with a
postcopular adjective) we can get by with one be, that of identity, and, secondly,
it consists in the naturalness of the classification, differences between (sub)types
depending mainly on the grammatical class or referential properties of the pred-
icatives (more on this type-classification of copular constructions see Mikulskas
2009: 141−151).
For the purpose of the present paper only copular constructions of the type
NP be NP are relevant, especially those with two definite nominals, representing
the equative subtype. Notably, this subtype of the copular construction is not ho-
mogeneous. Functionally, we can discern two species of equative structures: iden-
tificational proper and descriptional-identifying. The following sentences represent
the first kind of equatives:
(10) a. My cousin Harvey is the guy who got drunk at our wedding.
b. The guy who got drunk at our wedding is my cousin Harvey.
(example from Langacker 2008: 523)
In both instances − a and b – the equation of two definite nominal descriptions

(the referents of which, each in its own way, are known to the speaker and the
hearer) serves the purpose of identifying their designates as one and the same
person. We observe that this construction shows some sort of symmetry by vir-
tue of the exchangeability of (10a) and (10b). Though word order, representing
topic−comment structure in these sentences, depends on the presuppositional
question, by inverting the word order we do not create a new type of copular
construction.6
Sentences (11)−(12) represent the descriptional-identifying kind of the equa-
tive subtype:
(11) John Smith is our English teacher.
(12) Jane is the tallest girl in the class.
A distinguishing feature of these sentences is that the referents of their predicative

nominals are not necessarily unique: their uniqueness depends on support from
the relevant frame (or cognitive model) information, cf. Jane is the tallest girl in
the class, but last year it was Mary; John Smith is our English teacher besides two
others teaching English at our school. Naturally only in the latter case can we speak
of equative predication. In English, we are luckily able in this case to distinguish
a postcopular nominal designating a unique referent from its non-unique (‘pred-
icative’) use formally, by using different question words, who and what respec-
tively, in presuppositional questions: Who is John Smith? as against What is John
Smith? Only the former refers to the unique position occupied by John Smith.7
This kind of equation, especially from the perspective of the identificational con-
struction proper, can be seen as a limiting case of class-inclusion.
6. Possibly objections might be raised against this statement: one could find grounds, say,
to treat (10b) as representing a limiting case of the specificational construction. For example,
Caroline Heycock (2012: 209) lists a similar sentence One gang member turned out to be some-
one I knew among specificational constructions. Indeed, superficially such sentences (if we
leave aside functional differences) are reminiscent of the specificational constructions under
discussion, but descriptions like The guy who got drunk at our wedding or One gang member
(whom I somehow got to know about) in the above examples hardly satisfy the specifications
of role definiteness. Besides, these sentences lack one important feature of the specificational
construction (as I define it), viz. their postcopular ‘specifiers’ (though they are known at least
to the speaker) are not chosen ‘from the finite list of known candidates’ (for the definition see
section below).
7. In Lithuanian this test is not available: in both cases we must use the same question word
kas. So for the Lithuanian counterpart of sentence (11) Jonas Kalvaitis yra mūsų anglų kalbos
mokytojas, irrespective of the presupposed referentiality of its postcopular nominal, only one
question can be posed: Kas yra Jonas Kalvaitis?
3. Defining specificational copular constructions
Equative constructions of the descriptional-identifying kind show no such sym-

metry as can be observed in equatives proper: by inverting word order, and cor-
respondingly topic−comment structure, in these constructions we create a new
subtype of the copular constructions, viz. the specificational copular construc-
tion. By the inverting procedure a new structural property is added, establishing
the asymmetry of the specificational predication: the nominal in the precopular
position necessarily acquires the property of role definiteness8 – it defines some
role, the existence of the unique referent of which is presupposed and which is yet
to be identified, or specified, by the postcopular definite nominal (prototypically a
proper name, but not necessarily, cf. The culprit is me; The real problem is your par-
ents; Heycock 2012: 213). The specifier of the definite role is usually chosen from
a finite list of known candidates.9 It is important to note that the role definite-
ness of the first NP of the specificational construction is established by the very
presuppositional question. For example, the sentence Our English teacher is John
Smith possibly presupposes the question Who is your English teacher? – the poser
of which has no doubt about the existence of some individual teaching us English
and asks only to identify, or specify, this referent. It should also be clear from this
observation that the role-defining nominal is always the topic of the sentence, and
the specifier always its comment, providing new, rhematic, information.
Now we are in a position to define the specificational copular construction
in more rigorous terms. The unidirectionality of the specificational relation be-
tween the two nominal descriptions in this subtype of the copular construction
is iconically reflected in the asymmetry of two kinds of nominal definiteness, the
precopular NP displaying role definiteness whereas the postcopular NP is charac-
terized by individual definiteness. Another structural property defining the spec-
ificational constructions under consideration is that, independently of discourse
pragmatics, they have a fixed topic−comment (or theme−rheme) structure in that
the role description is always the topic, and the specifier is always its comment
(needless to say this known−new information structure naturally conforms to the
asymmetry of the role–specifier disposition).
By now it should be clear that the specificational constructions are an equa-
tive subtype of the copular predication and that they have their counterparts in
the ‘predicative’ domain of this kind of predication, viz. descriptional-identifying
8. Cf. the role / value distinction in Langacker (1991: 71−73) and the notions of attributive /
referential usage of a definite description in Donnellan (1966).
9. An example of the specificational construction from Heycock (2012: 209) is highly indica-
tive of this principle, cf. The murderer is one of those men over there.
equatives. There are good conceptual and structural reasons to consider the lat-
ter to be primary structures with respect to the former. For example, taking the
matter from a truth-conditional perspective, we can produce an utterance like
The winner of the race is Jane if and only if we really know that Jane won the race,
i.e., that the utterance Jane is the winner of the race is true. In terms of word order
it seems reasonable to view specificational constructions as inverted structures,
their source structure being descriptional-identifying equatives.
One should not forget that this inversion is not of the complement fronting
sort, that is, pragmatically motivated by considerations of contrasting new infor-
mation, but is a device for producing a new predicational type, driven by the cor-
responding contextual presuppositions (and, as we have seen, it carries structural
consequences, such as role definiteness of the precopular NP and fixed informa-
tion structure of the sentence). Alternatively, wishing to avoid undesirable deri-
vational overtones in this approach, we can formulate the undeniably systematic
relations between specificational constructions and their predicative counterparts
in terms of agnation, a notion borrowed from systemic-functional grammar.10 By
the same token, locative and existential constructions will be agnates in Lithu-
anian, standing in the same kind of inverted word order (resp. topic−comment)
relation:
(13) a. Pieštuk-ai yra stalči-uje.
pencil-nom.pl be.prs.3 drawer-loc.sg
‘The pencils are in the drawer’
b. Stalčiu-je yra pieštuk-ų.
drawer-loc.sg be.prs.3 pencil-gen.pl
‘There are (some) pencils in the drawer’
4. Short review of the subject definition in Cognitive Grammar
Before getting into the midst of the problems concerning subjecthood in specifi-
cational constructions, we will briefly recall some relevant points of the definition
10. The term agnation designates a ‘regular and systemic’ relationship between two grammati-
cal constructions which have the same major lexical items, but are different in structure, as in
(i) and (ii):
(i) The dog bit the man.
(ii) The man was bitten by the dog.
More about relationships of ‘enation’ and ‘agnation’ between constructions see Heyvaert
(2003: 35–40).
of grammatical relations in Cognitive Grammar. At the very beginning we should

note that in the theoretical framework of Cognitive Grammar the main gram-
matical relations – subject and (direct) object – are defined as syntactic figures of
conceptual relevance. ‘On the surface’, subject and object are recognizable from
their morphosyntactic encoding (prototypically nominative vs. accusative or
fixed positions in the syntactic sequence, respectively) as well as from well-known
behavioural characteristics (Keenan 1976). Importantly, in Cognitive Grammar,
the subject / object alignment also encodes viewing or perspective, i.e., some at-
titude (construal) of the conceptualizer toward the situation profiled. A situation
viewed from a certain stance is in its very essence asymmetrical. In terms of ar-
gument structure, verbal lexemes are also usually asymmetrical: this asymmetry
is prototypically supported by the corresponding semantic roles of force-dynam-
ically interacting participants of the verbal process.11 For example, in the canon-
ical transitive event, an agent affects a patient. So we can say that verbal lexemes
provide, for encoding purposes, structural templates which are explicitly encoded
in the corresponding morphosyntactic marking. Of course, for discourse purpos-
es these standard structures may be overridden using the constructional devices
that are at hand in languages, cf. subject / object inversion in passives.
It is also important that the conceptualization of the content of the verbal
lexeme (i.e., the process) in Cognitive Grammar is a dynamic cognitive procedure
formulated in attentional terms as sequential scanning. This unidirectional pro-
cessing of the verbal relation therefore has its time dimension (processing time),
which also adds to the asymmetry of the conceptualized content.
Finally, two salient participants of the verbal relation, even if only abstractly
figuring in it, are chosen as reference points for processing purposes. They set the
endpoints of the verbal relation and are usually formulated in spatial-dynamic
terms, as its trajector and landmark. But they are also understood in attentional
terms, as foci of primary and secondary salience. Prototypically, in the process of
conceptualization of the verbal relation, the salience of the trajector and the land-
mark is supported by the putative semantic roles, thus forming a natural sequence
of reference points for a better understanding of the verbal content (often imply-
ing complexity of the event, cf. Peter smashed the window and (consequently) The
window smashed to pieces). In the finite transitive clause both the trajector and the
landmark of the verb are elaborated by the relevant nominal descriptions and are
morphosyntactically encoded as the subject and the object of the sentence.
11. Cf. “The profiled relationship is rendered asymmetrical by the very fact of one participant
being elevated to the status of relational figure − what varies is whether this status is inspired
by objective factors (i.e., the content of conceived relationship), or whether its origin is purely
subjective, in the sense of being imposed extrinsically as part the conceptualization process
itself ” (Langacker 2002: 223).
Of course, the morphosyntactic encoding of the trajector / landmark con-

figuration as a grammatically relevant asymmetry facilitates the task of concep-
tualizing the clausal predication, but its role in this matter may be thought of
as secondary in the case of rich verbal semantics, there being other sources for
maintaining the relevant asymmetry. But, for example, in semantically interpret-
ing such well-known symmetrical structures as Marsha resembles Hilda / Hilda
resembles Marsha, the role of the morphosyntax is much more prominent, being
in this case indicative of the construal rather than of the semantics of the predi-
cate (more see Langacker 1991: 305−321, 2001, 2008: 363−382, 512−524).
4.1 Some problems envisaged in the Cognitive Grammar approach
It is generally suggested in Cognitive Grammar (at least in its Langackerian ver-

sion) that the morphosyntactic encoding of the main grammatical relations (viz.
subject and object) in conceptualizing matter directly reflects the trajector/land-
mark configuration, the latter constituting the conceptual content of the former.
This view neatly conforms to the content requirement principle, fundamental in
Cognitive Grammar. But in some cases the ‘form-meaning’ pairing is, in this re-
spect, not so straightforward.
Take, for example, Lithuanian inferential evidentials with non-canonical
marking of grammatical relations, recently discussed in Lavine (2010):
(14) Ing-os nuramin-t-a vaik-as.
Inga-gen.sg calm.down-ppp-nonagr child-nom.sg
‘Inga must have calmed the child down’
In this construction the subject is in the genitive and the object is in the nomina-
tive, non-canonical case marking (according to the Minimalist approach adopted
by Lavine) being determined by the morphology of the (non-agreeing) neuter
past passive participle.12 So, when identifying grammatical relations in this con-
struction, we cannot rely on the standard nominative/accusative marking, but
must resort to our knowledge of the transitive verb’s semantics (or to the standard
construal of the transitive event).
The situation is not so clear, in this respect, in the copular or copular-like
constructions, where one cannot rely on verb semantics at all. Take, for example,
another Lithuanian construction, recently intensively discussed in the literature
(Holvoet 2005, 2006; Mikulskas 2006):
12. In Lithuanian passive participle constructions the indirectly introduced agentive argument
is always in the genitive.
(15) Onut-ės (yra) žali-os ak-ys.

Ann-gen.sg be.prs.3 green-nom.pl.f eye-nom.pl
‘Ann has green eyes.’
The precopular nominal (presumably the subject) is in the genitive and its post-
copular attribute – in the nominative. When identifying grammatical relations
we can, in this case, rely only on structural analogy with other constructions
(namely, on the analogy with the ascriptive copular construction Onut-ėNOM yra
žali-ųGEN.PL aki-ųGEN.PL (mergin-aNOM) ‘Ann is a girl with green eyes’ (lit. Ann is
a girl of green eyes’) or, alternatively, with the possessive Dat + esse construction
Onut-eiDAT yra žali-osNOM.PL ak-ysNOM.PL ‘Ann has green eyes’ (lit. ‘to Ann are
green eyes’); more see Mikulskas 2006: 41−50).
In the equatives of the type NP1nom be NP2nom, either nominal can, in prin-
ciple, be chosen for the grammatical function of subject, the choice depending
mainly on the communicative perspective, which is reflected in the syntax of the
sentence:
(16) a. Mark Twain is Samuel Clemens.
b. Samuel Clemens is Mark Twain.
The situation is somehow more complicated in specificational equatives, as we

will see later.
We can thus conclude that there are constructions in language with regard to
which, in identifying grammatical relations, we cannot rely on the morphosyn-
tactic marking of their nominals. If, in these cases, one sticks rigorously to the
content requirement principle, suggesting that morphosyntax is always indica-
tive of the conceptual relevance of the clausal elements marked and vice versa,
one risks falling into the trap of circularity. It is clear that, especially in the case
of copular predication, some other grounds, independent of morphosyntactic
marking, must be found for judging the conceptual relevance of both nominals
in the construction, i.e., for establishing the trajector/landmark configuration re-
flected in it.
5. Subjecthood in specificational copular constructions
Now we shall take a look at the alignment of grammatical relations in specifica-

tional constructions in a few languages of different types. In English (as well as in
Danish, Swedish, French and some other European languages) the grammatical
role of subject is undoubtedly assigned to the precopular nominal characterized by
role definiteness. In terms of constituency the postcopular specifier can be thought
of, then, as verbal complement. In English13 this alignment clearly manifests itself
in verb number agreement and in pronominal reference in tag-questions:
(17) a. The cause of the riot was / * were the pictures on the wall.
(18) The tallest girl in the class is Molly, isn’t it / *she?
(example from Mikkelsen 2004: 93)
But in the Italian translation of (17a) the picture of the alignment is the reverse:
(17) b. La causa della rivolta *fu / furono le foto del muro.

(examples from Moro 1997: 28)
Not surprisingly, Portuguese, Spanish, and Catalan pattern with Italian showing
the NP2 agreement in their specificational constructions (Heycock 2012: 211).
But, surprisingly, French, another Romance language, shows in its specificational
sentences the NP1 agreement, an agreement pattern as in English,14 cf.:
(19) L’état c’est / *ce suis *je / moi.
the state that be.prs.sg.3 / that be.prs.sg.1 me
‘The state is me’ (example from Heycock 2012: 213)
Likewise German, in contrast to its Germanic relatives,15 consistently shows the

NP2 agreement in its specificational sentences (consequently, in this language the
second nominal of this copular construction is selected for the subject role), cf.:
13. Danish, Norwegian and Swedish (at least their standard varieties) have lost all traces of
agreement on finite verbs, but the first nominal in their specificational constructions is un-
doubtedly in the subject position. For instance, in Danish, the subjecthood of the NP1 in speci-
ficational copular construction can be detected by various tests based on verb second word
order (for example, position of negation in ‘inverted’ copular construction), or on pronominal-
ization (for example, tag-questions) (Mikkelsen 2004: 22−28, 120−129).
14. Nicole Nau, though, draws my attention to the fact of possible variation of verbal agree-
ment in French specificational constructions. According to her, some people (citing grammars
as authorities) maintain that ce sont is correct and that c’est should be avoided in such examples
as L’enfer c’est les autres (Sartre; literally: ‘hell is the others’), i.e., they would rather say L’enfer ce
sont les autres. I am not in a position to judge to what extent this variation in French occurs and
if it is statistically relevant.
15. However, an experimental study of agreement patterns in German and Dutch specifica-
tional constructions (Fischer 2003; referred to in Heycock 2012: 223−224) shows that in such
sentences as Singular NP1 be Plural NP2 (The real problem is your parents), Dutch speakers are
much less confident than Germans in choosing the second NP for agreement. There is wide-
spread speculation that inconsistency in NP2 agreement in Dutch specificational sentences
is due to restricted scrambling possibilities in this language when compared with German.
However, Faroese does not allow scrambling at all and hence might be expected to show NP1
(20) Das eigentliche Problem sind / *ist deine Eltern.

the real problem be.prs.3pl / be.prs.3sg your parents
‘The real problem is your parents’
(21) Die Königin von England bin / *ist ich.
the queen of England be.prs.1sg / be:prs.sg.3 I
‘The queen of England is me’ (examples from Heycock 2012: 211−212)
In specificational constructions of the languages with rich morphology, such as

Russian and Lithuanian etc., it is also the postcopular specifier that receives the
morphosyntactic marking of the subject. In Russian, in sentences referring to the
past, this alignment reveals itself not only in verbal number agreement, but also
in gender agreement:
(22) Edinstvenn-yj, kto stal na našu storonu, *by-l / by-l-a
only one-nom.sg.m who stood on our side be-pst.m / be-pst-f
Varvar-a.
Barbara-nom.sg
‘The only person who defended us was Barbara’
(example from Geist 2008: 95)
In Lithuanian, in such cases, gender agreement is manifested in the particip-

ial forms of the copular verb (serving, in the following example, an evidential
function):
(23) Šios mokykl-os direktori-us
this.gen.sg.f school-gen.sg headmaster-nom.sg
*esąs / esanti toki-a Janina Petraitienė.
be.ppa.nom.sg.m/ be.ppa.nom.sg.f such-nom.sg.f pn-nom.sg
‘The head-master of this school reportedly is (a certain) Jane Petraitiene’
agreement in specificational construction like English, another language without scrambling.

But as Caroline Heycock shows in her study (2009), agreement in Faroese specificational sen-
tences is variable and depends on the syntactic context: for example, sentences of the type NP1
be NP2 tend to show NP2 agreement, while in those of type NP1 Modal be NP2 the picture is
the reverse. These agreement facts complicate the theoretical problems connected with sub-
ject assignment in specificational constructions even more than the (overstudied) dichotomy
English-type vs. Italian-type specificational constructions would suggest (if one is a follower
of Generative Grammar, one inevitably has to postulate, in the case of variable agreement, two
different underlying ‘deep’ structures (or ‘logical forms’) for essentially the same specificational
proposition). For the sake of simplicity, though, my further discussion bears on this traditional
dichotomy.
In Lithuanian specificational constructions, in addition, the copular verb agrees

with the first and the second person pronouns occurring in the post-copular
specifier position:
(24) Klub-o prezident-as esu / *yra aš.
club-gen.sg president-nom be.prs.1 / be.prs.3 I.nom
‘The president of the club is me’
An additional argument in favour of postcopular subjects in the specificational

constructions in this type of languages can be adduced: in Lithuanian (as well as
in Russian) the precopular NPs can receive the instrumental marking (mirroring
the same case marking variability in their predicative counterparts). Semantically,
copular clauses with instrumental marking of their predicative NPs differ mini-
mally from those with nominative marking: the former bearing in their meaning
some overtones of temporality (i.e., change over time) of the situation (mainly
in references to the occupation of the subject referent) described.16 So there are
good reasons to treat such copular clauses as representing one and the same con-
struction with variable case marking on the same (postcopular) nominal (though
there are attempts, in the literature, to assign them different underlying syntactic
structures, cf. Pereltsvaig 2008: 10–14, 66–94). Such an attitude toward different
case marking on the same nominal is even more justified in the case of specifica-
tional constructions, where the role-defining nominal with instrumental marking
has the same presupposition of existence of its role-unique instantiator as that
with the nominative marking (the uniqueness of the instantiator being the object
of the same presuppositional question):
(25) Mokykl-os direktori-us / -umi (pernai) buvo
school-gen head-master-nom / ins (last year) be.pst.3
Jon-as Petrait-is.
PN-nom
‘(Last year) the headmaster of the school was John Petraitis’
The variability of case marking on the first nominal of the construction cannot, of
course, be compatible with the subjecthood of this nominal.
16. Such a semantic distinction between two morphosyntactic codings of the predicative NPs
in copular constructions is sufficient for our purposes, but, of course, it is an oversimplification.
On a more elaborated view, adopted in Timberlake (1990), the temporality of predicative situa-
tions signalled by the instrumental is seen as depending on the lexical semantics of predicative
nouns (descriptive vs. restrictive nouns), which in its turn correlates (or interacts) with aspec-
tuality and tense of the copula.
5.1 Encountering problems
Now we can see the contours of the problem. On the one hand, the alignment
of grammatical relations in specificational copular constructions in the two dif-
ferent types of languages which we have considered shows reverse pictures, and
this suggests different construals of the same equative predication. On the other
hand, the specificational constructions of all these languages show a very simple
unidirectional predicational scheme, which is clearly motivated by the same kind
of presuppositional questions concerning the identity of the unique instantiator
of the role defined: the definite role description is specified (or its unique referent
is identified) by equating it with some individual, not vice versa. Technically, the
informativeness of the predication in the specificational copular construction is
ensured by the different scope of descriptions, or by the different referential po-
tential of its two nominals. So there hardly is any space for alternative construals
in the case of specificational constructions.17 This discrepancy in conceptualiza-
tion / coding must somehow be accounted for.
17. The anonymous reviewer suggested to me that “before the strongest conclusion is drawn
from the data under consideration, the possibility should at least be entertained that the con-
ceptual configuration evoked by specificational constructions is ambivalent in the sense that
it offers competing motivations for the choice of the primary and the secondary figure within
the scene”. Indeed, some basis for conceptual ambivalence in this respect can be found in the
fact that the specifier nominal in the specificational construction designates an ‘actual’ entity
and thus is conceptually more salient than the role nominal, which designates a ‘virtual’ entity.
Undoubtedly, salience of this sort, along with topicality, becomes crucial for the assignment of
the subject role to the first nominal in the predicative counterpart of the specificational con-
struction. The canonical alignment of grammatical relations goes in line, in this case, with the
common topic−comment information structure, in which the establishment of the trajector /
landmark configuration is unproblematic: the topical nominal in the predicative construction,
which designates the definite (‘actual’) entity, naturally becomes the trajector of the profiled
equational relation and, consequently, the subject of the clause. In the comment part of the
structure this entity is matched against, or identified with, the relevant type instance, which
is essentially a ‘virtual’ entity “conjured up by the speaker and hearer solely for purposes of
making a type attribution” (Langacker 1991: 68), and which naturally serves the function of
the landmark in the equation of two entities. In the Lithuanian-type specificational construc-
tions, however, conceptual motivation for the assignment of the subject role to the same nom-
inal, occupying postcopular position, is admittedly different if only because there we are not
dealing with a topic–comment information structure but with a topic-referent identifying one.
We may hypothesize that in this case the specificational relation is processed in the opposite
direction, starting from the (postcopular) specifier nominal. This nominal (which has the mor-
phosyntactic coding of the subject) thus becomes a reference point for the conceptualization
of the specificational construction. Then the unique referent implied by the role nominal (and
which yet has to be specified) serves the function of the landmark in this construal. We may
Needless to say, postulating two different underlying syntactic structures in

Lithuanian-type and English-type specificational constructions, in line with the
generativist tradition, will make the problem even worse. Surely there is no alter-
native for one who attaches a paramount importance to the constituency struc-
ture of the clause. But constituency is not fundamental in the cognitive analysis of
the clause: one can arrive at the same conceptualization of its content by alterna-
tive compositional ways (Langacker 2000: 147−170). Grammatical relations, be-
ing figures of conceptual relevance in Cognitive Grammar, encode, by definition,
some construal of the clausal content. However, in the specificational construc-
tions of different languages we encounter the same construal, clearly reflecting
the same communicative perspective by distinct referential functions of their two
nominals, but with different alignments of the grammatical relations in the sen-
tence. Maybe other structural characteristics of the nominals than their morpho-
syntactic coding could be more important in the matter of conceptualization of
the specificational propositions?
5.2 Some preliminary speculations
The defining structural properties of the specificational construction provide good

grounds for maintaining the asymmetry of the clausal predication. Indeed, the
asymmetry of the specificational relation is supported by the different semantic
roles of the two nominals in the construction − definite role / (the identification
of its unique incumbent) specifier – and, correspondingly, by their different refer-
ential potential; this asymmetry is even more enhanced by those roles being tied
to topic−comment structure, which is achieved by posing corresponding presup-
positional questions. So, in a specificational copular construction, the role defin-
ing nominal is always the topic and its specifier the comment. The semantics of
the copular verb being of no relevance, the different semantic and informational
speculate further that particular languages (like Lithuanian, Russian, Italian, German and so
on) focus on this particular motivation and conventionalize the corresponding alignment of
grammatical relations in their specificational constructions. But in that case we are compelled
to draw one undesirable conclusion, namely, that there may exist more than one ‘convention-
alized’ construal of the specificational constructions in one and the same language! This is
because, as we will see later, in the English specificational constructions with unmarked and
marked word order different nominals are selected for the grammatical role of subject: the role
nominal in the first case, and the specifier nominal in the second. It seems therefore to me that
while the relative conceptual salience of the ‘actual’ participant in the specificational situation
is undoubtedly important for subject selection, at least in the Lithuanian-type specificational
constructions it is not sufficient to trigger alternative construal of this highly externally moti-
vated structure.
functions of the two nominals in the construction provide good grounds for laying
down a trajector/landmark configuration. The role-defining nominal, also provid-
ing known information, naturally becomes the starting point in the conceptualiza-
tion of the specificational relation, i.e., its trajector, and its postcopular specifier,
providing new information, naturally establishes the endpoint of this relation, i.e.,
becomes its landmark.
However, this conceptually relevant trajector / landmark configuration is sup-
ported by subject marking only in the English-type specificational constructions,
thus facilitating the task of conceptualization of the content predicated. In the
Lithuanian-type specificational constructions, by contrast, subject marking falls
on the postcopular nominal, and this alignment entails the opposite directionality
of the conceptualization of essentially the same relation. Intuitively such ‘back-
ward’ processing of the specificational relation is less natural if not hardly plau-
sible. Though, admittedly, ‘backtracking’ and ‘reconceptualization’18 plays some
subsidiary role in the matter of processing the specificational relation. This role
may be more prominent in the case of Lithuanian-type specificational construc-
tions than in the English-type. Indeed, the morphosyntactic coding of the subject
in the Lithuanian-type specificational constructions is essentially the same as that
in their predicative counterparts, which are, as I have stated earlier, conceptually
primary with respect to the former. The same alignment of grammatical relations
may favour, at least to some extent, reconceptualization of the specificational
construction as an ‘inverted’ predicative construction. Thus, Lithuanian speak-
ers may feel a closer relation between the two agnate constructions than English
speakers.
We can try to account for these alignment differences in the specificational
constructions of the different types of languages in terms of grammaticalization,
the English-type constructions being more grammaticalized in that subject mark-
ing is conferred on their topics. Consequently, the Lithuanian-type specificational
constructions, whose alignment is the same as that of their predicative counter-
parts, have somehow not yet evolved into a new (sub)type of copular construction.
But such speculation is not correct and, as we shall see later, is even in contradic-
tion with the facts. It would be more correct to say that in the English-type specifi-
cational constructions the topical nominal, occupying the clause-initial position,
18. To think of conceptualization as strictly linear processing is an oversimplification. Cf. what

Ronald Langacker says in this respect: “Processing occurs simultaneously in various dimen-
sions and at multiple levels of organization. There is not invariably any sequence of access for
the elements of a complex conception, nor is one fully adhered to in actual practice. And given
the pressures of online processing, any actual rendition is likely to be discontinuous and com-
plicated by factors like backtracking and reconceptualization” (Langacker 2008: 501).
becomes its subject by default. Also worth noting is that in languages with rela-
tively free clausal word order it is not unusual for a rhematic complement to be
coded as the subject of the sentence. Take, for example, Lithuanian existentials:
(26) Kiem-e auga žilvit-is
courtyard-loc grow.prs.3 willow-tree-nom
‘There is a willow-tree growing in the courtyard’
We could suggest that, the referential properties of the two nominals and their
fixed roles in the information structure of the Lithuanian-type specificational
constructions being the same as in the English-type constructions, different sub-
ject marking in the former plays no direct role in the processing of the specifica-
tional relation. Subject marking is relevant in this case insofar as it is indicative
of the systemic relation of the specificational copular construction to its agnate
counterpart in the predicative domain.
5.3 Yet more puzzles
Though, structurally, specificational copular constructions are clearly motivated

by corresponding contextual presuppositions and have a fixed topic−comment
structure, to some minimal degree one can pragmatically manipulate informa-
tional accents in the construction, changing its intonational contour and/or word
order for the expression of contrast without essentially affecting the type of the
predication. For our purposes it is important that the specifier can be fronted
and moved into initial position in both types of languages. This pragmatically
motivated movement of the constructional complement carries no grammatical
consequences for the Lithuanian-type specificational construction, but it inverts
the selection of the subject nominal in those of the English-type (so we can add
the feature of subject/complement reversibility to the structural characteristics of
the latter). The fronting of the specifying complement in specificational construc-
tions can be induced by the appropriate presuppositional question, cf.:
(27) Q: Who is the mayor?
A: The mayor is John.
(28) Q: Is the mayor Sam?
A: No, John is the mayor (Sam is the fire chief).
Thus, in (28) A, the specifier John, fronted into initial position, by default becomes
the subject of the clause (cf. John is the mayor, isn’t he / *it?). Maintaining that
(28) A remains a specificational predication creates a problem of constituency for
the generative analysis, because in a raising structure the same nominal cannot
occupy SpecIP position while being at the same time in CompIP position. To
avoid this problem, generative grammarians (cf. Mikkelsen 2006; Heycock
2012: 216−21719 inter alia) claim that the (28) A-type answers are predicative
constructions, not specificational ones. But the structural type of copular con-
structions in all the languages considered so far is established for communicative
purposes of nominal specification externally, by posing corresponding questions
in presupposition, and it minimally depends on the alignment pattern occurring
in them, which is mainly an internal matter of clausal grammar, depending, if not
on verbal semantics (as is the case in copular predication in general), then on the
very syntax of the language type. So, while escaping problems inherent to their
theory, proponents of the generative approach fail to do justice to the adequate
type description of specificational constructions, including relevant aspects of
their grammar.
Taking into account instances of the English-type specificational construc-
tions with pragmatically marked word order, it becomes even clearer that if one
wants to come to terms with the content requirement principle, fundamental
in Cognitive Grammar, in dealing with the conceptual organization of specifi-
cational constructions, one should not stick to the morphosyntactic coding of
their nominals. The conceptual organization (and presumably also the trajector/
landmark disposition) in marked cases of the English-type specificational con-
struction remains the same as it was in their unmarked counterparts: the speci-
ficational relation is, in both cases, processed unidirectionally, starting from the
19. Arguing that in both (i) and (ii) the answers are predicative copular sentences, Caroline
Heycock refers to comparable pitch accent variability in examples (iii) and (iv), which is ‘the
norm for English’, but specificational sentences, because of the fixed pitch accent on their focal
NP, are out of this ‘default rule’ (thence the infelicity of (ii) A’), only predicative sentences con-
form to it:
(i) Q: Who was the culprit? (John or Bill?)
A: John was the culprit?
(ii) Q: What was John? (Was John the culprit or the victim?)
A: John / he was the culprit.
A’: *The culprit was John.
(iii) Q: Who checked the cockpit?
A: John checked the cockpit.
(iv) Q: What did John check?
A: John checked the cockpit.
But this argumentation, based on ‘parallel accentuation patterns’, is superficial and clearly in-
correct: taking into account the communicative perspective established by question (i), the
answer to it can hardly be called predicative, and for the same reason the ‘infelicitous’ (ii) A’
does not represent specificational predication.
definite role description, while the new alignment of grammatical relations in

them goes the opposite way. In the Lithuanian-type specificational constructions,
both in unmarked and marked cases, alignment is the same, its asymmetry being
opposite to the conceptually motivated direction of the processing of the specifi-
cational relation.
6. Concluding remarks
The fact that in different types of languages (and even in the same language type,
if we take into account pragmatically marked cases, not to mention the com-
plex situation in Dutch and Faroese specificational constructions mentioned in
Footnote 15) different nominals, with different referential properties and serv-
ing different semantic functions in specificational predication, are chosen for the
grammatical function of subject, renders it extremely difficult, if not impossible,
to define grammatical relations in the specificational copular constructions in
conceptual terms, i.e., as a trajector/landmark configuration, which is the usual
practice in Cognitive Grammar.
The conceptual organization of the specificational copular construction is
based on the referential properties of its nominals and on their fixed roles in the
informational structure of the construction. The best candidates for the roles of
trajector and landmark are the role-defining nominal and its specifier, respectively
serving as the starting point and the end point of the specificational predication.
This configuration is supported by the alignment of grammatical relations only
in the unmarked cases of the English-type specificational construction, while the
conceptualization (processing) of the Lithuanian-type specificational construc-
tions (as well as of the pragmatically marked cases of English type) dispenses with
the direct support of the grammar.
It is clear that, generally, the logic of morphosyntactic coding in specifica-
tional constructions does not reflect the logic of their conceptual organization: in
most cases the latter is not recoverable from the former. In this sense we can speak
of a violation of the content requirement principle in the syntax of specificational
constructions, which is an undesirable conclusion for Cognitive Grammar.
Deprived of conceptual relevance, the subjecthood of the first nominal in the
English-type specificational constructions is mainly indicative of the syntactic
properties associated with initial position in this type of languages. Inasmuch as
the subjecthood of the complement is not an exclusive property of the Lithu-
anian-type specificational constructions, it is also a syntactic property of this type
of languages. But also worth noting is the fact that in the Lithuanian-type speci-
ficational constructions, subject marking falls on the same nominal as it does in
their predicative counterparts, thus emphasizing the agnate (or paradigmatic) re-
lation between them: this may have some relevance for the conceptualization of
the specificational predication.
Taking a general view of the matter, it is not unexpected that the syntax of the
contextually driven constructions considered in this article does not exactly con-
form to the specific communicative purposes, showing some kind of language-
type specific inertia.
Abbreviations

dat dative NONAGR nonagreeing
f feminine pl plural
fut future PN personal name
gen genitive ppa past participle active
ins instrumental ppp past passive participle
loc locative prs present tense
m masculine pst past
neg negation sg singular
References
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.

Donnellan, Keith S. 1966. Reference and Definite Descriptions. Philosophical Review 75: 281–
304. DOI: 10.2307/2183143
Fischer, Golda. 2003. The problem is / are your parents: Resolving number conflicts in equative
sentences in Dutch and German. Honours dissertation, University of Edinburgh.
Geist, Ljudmila. 2008. Predication and Equation in Copular Sentences. In Existence: Semantics
and Syntax [Studies in Linguistics and Philosophy 84], Ileana Comorovski & Klaus von
Heusinger (eds), 79–105. Dordrecht: Springer.
Heggie, Lorie. 1988. The Syntax of Copular Structures. PhD dissertation, University of South-
ern California.
Heller, Daphna. 2005. Identity and Information: Semantic and Pragmatic Aspects of Specifica-
tional Sentences. PhD dissertation, The State University of New Jersey, Rutgers.
Heycock, Caroline. 2009. Agreement in specificational sentences in Faroese. Nordlyd: Tromsø
Working Papers in Language and Linguistics 36: 56–77.
Heycock, Caroline. 2012. Specification, equation, and agreement in copular sentences. Canadi-
an Journal of Linguistics 57(2): 209–240. DOI: 10.1353/cjl.2012.0033
Heycock, Caroline & Kroch, Anthony. 1998. Inversion and equation in copular sentences. In
Papers in Linguistics 10, Artemis Alexiadou, Nanna Fuhrhop, Ursula Kleinhenz, & Paul
Law (eds), 71–87. Berlin: Zentrum für allgemeine Sprachwissenschaft.
Heycock, Caroline & Kroch, Anthony. 1999. Pseudocleft Connectedness: Implications for the
LF Interface Level. Linguistic Inquiry 30(3): 365–397. DOI: 10.1162/002438999554110
Heyvaert, Liesbet. 2003. A Cognitive−Functional Approach to Nominalization in English [Cog-
nitive Linguistics Research 26]. Berlin & New York: Mouton de Gruyter.
Higgins, Francis R. 1973. The Pseudo-cleft Constructions in English. PhD dissertation, MIT,
Cambridge, MA [Published by Garland Press, New York, 1979].
Holvoet, Axel. 2005: Intranzityvinių sakinių tipai: egzistenciniai, lokatyviniai ir posesyvin-
iai sakiniai [Intransitive clause types: existential, locational and possessive clauses]. In
Gramatinių funkcijų tyrimai [Studies in Grammatical Functions], Axel Holvoet & Rolandas
Mikulskas (eds), 139–160. Vilnius: Institute for the Lithuanian Language.
Holvoet, Axel. 2006. Dėl sintaksinio dviprasmiškumo ir teminės-reminės struktūros [On syn-
tactic ambiguity and theme-rheme structure]. Acta Linguistica Lithuanica 55: 116–124.
Keenan, Edward L. 1976. Towards a universal definition of ‘subject’. In Subject and Topic,
Charles N. Li (ed), 303–333. New York: Academic Press.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, vol. 1: Theoretical Prerequisites.
Stanford, CA: Stanford University Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar, vol. 2: Descriptive Application.
Stanford, CA: Stanford University Press.
Langacker, Ronald W. 2000. Grammar and Conceptualization, 2nd edn. Berlin & New York:
Mouton de Gruyter.
Langacker, Ronald W. 2001. Topic, Subject, and Possessor. In A Cognitive Approach to the Verb:
Morphological and Constructional Perspectives [Cognitive Linguistics Research 16], Hanne
Gram Simonsen & Rolf Theil Endresen (eds), 11–48. Berlin: Mouton de Gruyter.
Langacker, Ronald W. 2002. Concept, Image, and Symbol. Berlin & New York: Mouton de
Gruyter.
Langacker, Ronald W. 2008. Cognitive Grammar: A Basic Introduction. Oxford: Oxford Univer-
sity Press. DOI: 10.1093/acprof:oso/9780195331967.001.0001
Lavine, James. 2010. Mood and transitivity in Lithuanian: The case of the inferential evidential.
Baltic Linguistics 1: 115–142.
Mikkelsen, Line. 2004. Specifying Who: On the Structure, Meaning, and Use of Specificational
Copular Clauses. PhD dissertation, University of California, Santa Cruz.
Mikkelsen, Line. 2006. Specificational copular clauses. Talk given at the Kobe Area Circle of
Linguistics, Kobe Shoin Women’s University, November 22, 2006.
Mikulskas, Rolandas. 2006. Pastabos dėl sintaksinio dviprasmiškumo sąvokos ir kiti susiję
dalykai [Notes on the notion of syntactic ambiguity and other related matters]. Acta Lin-
guistica Lithuanica 55: 1–53.
Mikulskas, Rolandas. 2009. Jungties konstrukcijos ir jų gramatinis kontekstas [Copular con-
structions and their grammatical context]. Acta Linguistica Lithuanica 61: 113–156.
Moro, Andrea. 1991. The raising of predicates: copula, expletives and existence. In MIT Work-
ing Papers in Linguistics Volume 15: More Papers on Wh−Movement, Lisa Cheng & Hamida
Demirdash (eds), 119–181. Cambridge, MA: MIT.
Moro, Andrea. 1997. The Raising of Predicates. Cambridge: Cambridge University Press. DOI:
10.1017/CBO9780511519956
Partee, Barbara. 1986. Ambiguous pseudoclefts with unambiguous be. In Proceedings of NELS
16, Stephen Berman, Jae-Woong Choe, & Joyce McDonough (eds), 354–366. Amherst,
MA: GLSA.
Pereltsvaig, Asya. 2008. Copular Sentences in Russian: A Theory of Intra-Clausal Relations [Stud-
ies in Natural Language and Linguistic Theory 70]. Dordrecht: Springer.
Rothstein, Susan. 2001. Predicates and Their Subjects [Studies in Linguistics and Philosophy
74]. Dordrecht, Boston & London: Kluwer.
Taylor, John R. 2002. Cognitive Grammar. Oxford: Oxford University Press.
Timberlake, Alan. 1990. The aspectual case of predicative nouns in Lithuanian texts. In Verbal
Aspect in Discourse. Contributions to the Semantics of Time and Temporal Perspective in
Slavic and Non-Slavic Languages [Pragmatics and Beyond New Series 5], Nils B. Thelin
(ed), 325–347. Amsterdam /Philadelphia: John Benjamins.
Van Langendonck, Willy. 2007. Theory and Typology of Proper Names. Berlin: Mouton de
Gruyter. DOI: 10.1515/9783110197853
Williams, Edwin. 1983. Semantic vs. Syntactic Categories. Linguistics and Philosophy 6: 423–
446. DOI: 10.1007/BF00627484
Differential object marking in Latgalian
Nicole Nau
Adam Mickiewicz University in Poznań / Vilnius University
This article discusses variation in object marking in Latgalian, based on sources

where this phenomenon is most pronounced. Differential object marking in
Latgalian consists in the choice of the genitive instead of the accusative for ob-
jects of transitive verbs. Genitive marking regularly appears in negated clauses
and in constructions with the supine. It is optionally used in clauses that por-
tray a situation as unreal, unlikely, or undesired from the point of view of the
speaker. Apart from these clause-related factors, genitive marking is common
though less regular with mass nouns and count nouns in the plural in descrip-
tions of single events, while for habitually occurring events the accusative is
used. The use of the partitive genitive is increasingly coupled with prefixing of
the verb. A further type is the use of the genitive with cumulative verbs. There
is also a small group of non-derived verbs that select the genitive. Those that
have a nominative subject tend to become transitive verbs in modern written
Latgalian.
1. Introduction1
Latgalian differs from Latvian, its closest relative, in showing symmetrical dif-
ferential object marking similar to that found in Lithuanian: with certain types of
objects, or with certain verbs, or in certain constructions the accusative competes
with the genitive as the case for object marking. In most of these cases there is
reason to hold that genitive marking is an inherited feature that Latgalian and
Lithuanian have retained, while Latvian has expanded the use of the accusative.
Differential object marking (DOM) has attracted the interest of linguists of
various persuasions. The first case-studies of differential object marking (starting
with work by Bossong, see especially Bossong 1991) considered languages where
only certain noun phrases are overtly marked as objects, while other objects are
1. I am grateful to Axel Holvoet, Björn Wiemer, Ilja Seržant, Peter Arkadiev, and an external
reviewer for their valuable comments on an earlier version of this paper.
208 Nicole Nau
unmarked. In such a case DOM is part of restricted marking (cf. Sinnemäki

2012ms.). Restricted marking appears to be widespread in the languages of the
world. Languages that have been studied more often include Spanish, Romanian,
Turkish, and Semitic languages. The contrast between marking vs. non-mark-
ing of objects or subjects is called asymmetrical differential marking in de
Hoop & Malchukov (2008; see also other works by these authors), while symmet-
rical differential marking concerns the choice between two overt marking
strategies. The language most often cited as an example is Finnish. De Hoop &
Malchukov (2008) argue that the two kinds of differential marking should be dis-
tinguished, as different principles play a role in triggering case-marking in each
of them. Research on differential object marking of both kinds is tightly connect-
ed to research on transitivity, starting with Hopper & Thompson (1980). Differ-
ential marking is not the same as (non)canonical marking, though both terms
are used in research on more or less the same linguistic phenomena. The latter
seems to depend more on the framework used and basic assumptions underlying
the research than the former (which is one reason why “differential marking” is
preferred and “non-canonical marking” is avoided in this paper), and the range
of phenomena treated as “non-canonical” varies accordingly.2 “Non-canonical”
marking presupposes “canonical” marking, which is often only vaguely defined,
using a concept of prototypical transitivity, which in turn depends on one or an-
other understanding of the syntactic functions A, P, and S (see Haspelmath 2011
for a helpful analysis of seemingly subtle yet crucial differences among the “Dixo-
nian”, “Comrian”, and “Bickelian” approach to these concepts).
The aim of this paper is to present a systematic overview of the competi-
tion between accusative and genitive marking in Latgalian texts, and to formulate
questions for further research. Since the Latgalian data have been known only to a
small group of scholars of Baltic languages and most linguists are unfamiliar with
the language, this paper will include a large number of examples. All examples are
taken from authentic sources. For background information on the grammar of
Latgalian the reader is referred to Nau (2011a).
2. For example, Onishi (2001: 43) in his introduction to an edited volume on non-canonical

argument marking states: “We don’t regard variations in case-marking exclusively due to (2)
[referential status of NPs] and/or (3) [overall clause types with respect to aspect, mood, polar-
ity, genericity, etc.] as ‘non-canonical’”, while in the following sentences he acknowledges that
other authors in the same volume will deal with just these variations.
1.1 Latgalian varieties and the data for this research
Not all linguists acknowledge Latgalian as a separate language, but instead treat
Latgalian varieties as part of the Latvian language. In this paper, if not otherwise
indicated the label “Latvian” will refer to Modern Standard Latvian and thus be op-
posed to “Latgalian”. This is not to deny that Latgalian dialects (in the narrow sense
of “dialect”, referring to traditional rural dialects) can be appropriately grouped
together with other dialects of the Latvian language and that for several phenom-
ena dialect continua exist. Still, Latgalian in my understanding comprises various
varieties that are fruitfully treated together and set apart from those varieties of
Latvian that resemble Modern Standard Latvian to a much higher degree. They
can be roughly divided into the following groups:
a. the Latgalian (Non-Selonian) dialects of the High Latvian dialect group

(documented in dialect descriptions, texts recorded and transcribed by Lat-
vian dialectologists, folklore texts collected and transcribed by ethnographers
since the late 19th century);
b. semi-standardized written Latgalian of the 18th and 19th century (texts
mostly written by non-native speakers, for the greatest part religious texts);
c. standardized written Latgalian of the 20th century (documented in a con-
siderable amount of published texts, fiction and non-fiction, normative dic-
tionaries and a normative grammar (Bukšs & Placinskis 1973); 1945–1990
publications appeared mainly in exile communities in Western Europe and
North America);
d. contemporary spoken and written Latgalian with several varieties. These va-
rieties show ongoing merging of traditional dialects, a growing influence of
standard Latvian, and the emergence of a new standard Latgalian (document-
ed in audio recordings and written texts, often published on the Internet3).
This research is based on texts of groups (a) and (d). Latgalian publications of the
20th century mentioned under (c) show considerably fewer peculiarly Latgalian
features and are often very close to standard Latvian in vocabulary, morphology
and syntax. The question of how far this variety displays differential object mark-
ing is left for further research.
My main source of a variety of type (a) is fairy tales collected and transcribed
at the end of the 19th century by the Polish ethnographer Stefania Ulanowska in
the parish of Viļāni in Central Latgalia (further referred to as UP). The digitalisa-
tion, editing and translation of these texts is currently being undertaken within
3. A first corpus of contemporary written Latgalian (1 million words) has been compiled and
is now available online (currently in a test version): MuLa-1.0.
210 Nicole Nau
the project Dziedzictwo Językowe Rzeczypospolitej (The Linguistic Heritage of the

Polish-Lithuanian Commonwealth).4 The database built in this project will contain
the Latgalian texts in two forms: in the original writing of the publication and in
a transliteration according to current orthography, which facilitates reading and
searching. Misprints, inconsistent spellings and certain morphological irregulari-
ties are corrected in a form given in brackets. For convenience, in this paper only
the edited form will be used in examples from UP. The following example shows
the three forms:
original form in UP: łopsa pi-dziera asznia i ap-jedia gobułu galiś
transliterated form: lopsa pīdzēra [pīdzēre] ašņa i apjēde [apēde] gobulu galis
form used here: lopsa pīdzēre ašņa i apēde gobulu galis
‘the fox drank (some) blood and ate a piece of meat’
Examples from other sources will always be given in the original form without
editing.
My main source for contemporary Latgalian are texts written by Ilze Sperga
(IS), a prolific contemporary author who also comes from Viļāni, which has the
advantage that her variety is very close to that of UP in terms of phonology and
morphology. For this paper I investigated one of her short stories completely (Kur
sauleitei sāta (Where the Sun has his home), further referred to as KS), considering
all instances of accusative and genitive object marking. This allows for a modest
statistical analysis which may give a first impression of the amount of differential
object marking in Latgalian. The text comprises 8150 running words. Ignoring
some instances where the case of the object could not be determined (for reasons
explained in Section 2), I counted 209 instances of accusative objects and 179 in-
stances of objects with genitive marking. Further information is given in Table 1.
By choosing these sources, I consciously chose somewhat conservative vari-
ants of Latgalian. The language of my sources is more different from Modern
Standard Latvian than most varieties of contemporary Latgalian are (especially
concerning written Latgalian). The fairytales in UP were collected at a time when
most speakers of Latgalian had little if any contact with Latvians living in the
western parts of today’s Latvia. Today, presumably all speakers of Latgalian in
Latvia are in daily contact with Standard Latvian, at least through radio, tele-
vision and written media; furthermore, most of them received their education
in Latvian (except for those who attended a Russian-medium school). This has
led to a heavy influence of Latvian on the syntax of Latgalian. Ilze Sperga’s prose
4. The project is funded by the Polish Programme for the Development of the Humanities
(Narodowy Program Rozwoju Humanistyki) and carried out at Adam Mickiewicz University in
Poznan 2012–2013. See www.inne-jezyki.amu.edu.pl.
Table 1. Accusative and genitive marking of objects in KS (8150 words)

Singular Plural Proper nikas Other Total
noun noun name pronoun
accusative 103 64 8 0 34 209 = 54%
genitive 43 60 9 34 33 179 = 46%
Total 146 124 17 34 67 389 = 100%
Note: The negative pronoun nikas ‘nothing, nobody’ has been counted separately for reasons explained in
Section 3. The category “other pronoun” contains several traditionally distinguished kinds of pronouns,
such as interrogative and indefinite, besides personal pronouns. See Section 4.2 for further discussion of
pronouns.
stands out among contemporary Latgalian written texts in showing significantly

more peculiarly Latgalian constructions, such as can be found in UP and other
traditional texts. This becomes evident when searching the newly available corpus
of contemporary Latgalian (MuLa-1.0), where several of Ilze Sperga’s texts are
included. In some of the subsections below I will discuss data from more “main-
stream” contemporary sources that show the decline of differential object marking
in modern Latgalian. My focus however will be on the conservative variants that
have retained differential object marking – or maybe even developed it further.
1.2 Towards a systematic description
There have been only a few very brief descriptions of object marking in Latgalian
or one of its varieties, all of them confined to remarks within grammars or dialect
descriptions. Information can be found in Endzelin’s comprehensive grammar of
Latvian (Endzelin 1922), where Latgalian dialects are treated as part of the Lat-
vian language, in Gāters’ overview of Latvian dialect groups (Gāters 1977), and
in a few descriptions of individual local dialects. Unfortunately most dialect de-
scriptions are restricted to phonology and morphology, and the rare descriptions
which consider syntactic phenomena (for example Ūsele 1998) do so in an un-
systematic, rather impressionistic way. Following a tradition inherited from Latin
grammar, the uses of a case-form are presented as instances of various meanings
of that case, such as genitivus partitivus, genitivus qualitatis, genitivus materiae,
etc. The construction in which a case-marked noun appears is of secondary im-
portance in this approach. Bukšs, who devotes six pages of his grammar to the
uses of the genitive in standard Latgalian (Bukšs & Placinskis 1973: 292–298), is
more systematic; still his focus is on semantic rather than syntactic distinctions.
The competition between genitive and accusative marking, which is the top-
ic of this paper, has many aspects and may be approached and classed in vari-
ous ways. My research has first been inspired by work on transitivity, especially
212 Nicole Nau
Hopper and Thompson’s (1980) classic paper and the concept of prototypical
transitivity presented in Næss (2007). It is evident that genitive marking in Lat-
galian is associated with a lower degree of transitivity than accusative marking.
A Latgalian clause that corresponds to the semantic prototype of transitivity (a
two-place predicate with a volitional agent and an affected patient) contains a
finite verb, a nominative subject triggering agreement, and an accusative object.
This syntactic model is used also with many two-place predicates that are not
prototypically transitive, for example, verbs of perception (‘see’, ‘hear’), knowledge
(‘know’), or consumption (‘eat’, ‘drink’). Furthermore, many three-place predi-
cates (typical examples are verbs of physical transfer such as ‘give’ or ‘sell’) have a
nominative subject and an accusative object that have the same formal properties
as the arguments in a prototypically transitive clause. All and only these verbs
will be called “transitive” in this description of Latgalian. This use of the term
“transitive” is in line with traditional grammar, as well as with the “Comrian”
approach to grammatical relations5 (see Haspelmath 2011; Tsunoda 1985: 387).
Consequently, verbs that select another type of argument marking (for example,
nominative and genitive, or dative and genitive) are intransitive (not transitive) –
and not “less transitive”. Only on a semantic level may we speak of “less transitive”,
“prototypical transitive” and the like, as I did above.
The first point of my investigation, then, is to explore under which circum-
stances the accusative marking of the object of a transitive verb – a verb that has
an accusative object in its frame – is replaced by genitive marking. The reasons
may be properties of the clause or construction (Section 3), or properties of the
object (Section 4). Clause-related properties that play a role in Latgalian are po-
larity (negation triggers genitive marking) and mode (irrealis triggers genitive
marking). Constructions that change the case-marking of objects are construc-
tions with the supine and the infinitive. The cases discussed in Section 3 have in
common that they concern all transitive verbs and all kinds of objects. Section 4,
on the other hand, concerns the marking of certain types of objects, and some-
times genitive marking occurs only with a subclass of transitive verbs.
A different question is the object marking of (potentially) intransitive verbs,
discussed in Section 5. Here, the competition of genitive and accusative marking
means a competition between the transitive and an intransitive pattern. In Sec-
tion 5.1 I present two-place intransitive predicates, many of which are increasingly
5. A difference is that the traditional approach does not distinguish between the P of a transi-
tive clause and the T of a ditransitive clause.
used as transitive verbs. In Section 5.2 I briefly discuss derivational means that
turn a transitive verb into an intransitive verb.6
By presenting the data in this way I hope to prepare the ground for more gen-
eral questions (or questions of more general interest). One of these is the connec-
tions between the different groups of differential marking. Can they all be united
by a common “idea”, and can this idea be formulated more precisely than by a
general reference to low transitivity (after all, not all instances of low transitivity
are associated with differential object marking)? Or, on the contrary, are there
language-internal arguments for setting apart one type of genitive-marking from
another type? For example, Kagan (2010, 2013) argues for a closer association (in
Russian) of the genitive of negation and the intensional genitive, as opposed to the
partitive genitive. I am still far from drawing a general map of genitive marking in
Latgalian, but will discuss some of the findings of the present research under this
perspective in the concluding Section 6.
Before starting the syntactic analysis, a closer look at the formal expression
of accusative and genitive of nouns and pronouns is in order, which will be the
topic of Section 2.
2. Morphology
With nouns, genitive and accusative are morphologically clearly distinct. Table 2
shows the case endings of nominative, accusative and genitive in different declen-
sion classes (for more information see Nau 2011a and 2011b).
As Table 2 shows, accusative and genitive are always clearly distinguished
within one number. Syncretism is found only across numbers: in class I and IV
the accusative singular has the same form as the genitive plural, while the genitive
singular has the same form as the accusative plural in classes IV–VI as well as for
the couple of class I nouns that historically belonged to an u-declension (class III)
and have retained a genitive in -s, for example, gods ‘year’, mads ‘honey’. In order
to determine which case is used for object marking in a given text, one thus has
6. Note: While I thus keep up the terminological difference between transitive and intransi-
tive verbs, my division of topics into three classes treated in Chapters 3, 4, and 5, respectively,
is essentially the same as that by Haspelmath (2001: 56), who groups conditions under which
argument marking in European languages deviates from the canonical scheme of nominative
subjects and accusative objects into the classes (i) reference-related conditions, (ii) clauses-
related conditions, and (iii) predicate-related conditions, or the one used by Næss (2007), who
in her Chapter 5 treats factors that correlate with a deviation from prototypical transitivity in
separate sections with the headings ‘Semantic features in verb subcategorization’, ‘Properties of
argument NPs’, and ‘Clause-level properties’.
214 Nicole Nau
Table 2. Case-endings of nouns

Singular Plural
I I < III II IV V VI I–III IV V–VI
nom -s -s -sj -a -e -sj -i -ys -(i)sj
acc -u -u -i -u -i -i -(u)s -ys -(i)sj
gen -a -s -a -ys -isj -sj -u -u -u
Notes:
– Stem alternation is ignored here, as it appears only in addition to endings and does not contribute
to the distinction of case-forms within one declension.
– In UP the accusative plural of nouns of declensions I–III is almost always signaled by -s
(undergoing various forms of assimilation), but this ending has been edited as -us.
– In 20th century’s written Latgalian usually -as instead of -ys and -es instead of -is are used
for the genitive singular and the nominative and accusative plural of classes IV and V.
to know whether the object is in the singular or the plural. In most cases this can
be decided from the context and/or from encyclopedic knowledge. In example (1)
meadow (singular), shepherd (singular) and sheep (plural) had been mentioned
before; besides, it is usual that one shepherd guards several sheep on one mead-
ow. The ending -u can thus be interpreted accordingly as accusative singular for
‘meadow’ and ‘shepherd’, but as genitive plural in the case of ‘sheep’.
(1) atroda tū pļov-u i vušk-u i
find.pst.3 dem.acc.sg meadow-acc.sg and sheep-gen.pl and
gon-u
shepherd-acc.sg
‘(she) found the meadow and (the) sheep and the shepherd’ (UP)
Some instances remain open to interpretation, for example:
(2) dzed-s […] sviļpoj tuos na-šmuk-uos dzīsm-is

grandpa-nom whistle.prs.3 dem.?.? neg-pretty-?.? song-?.?
‘grandpa is whistling those indecent songs (acc) / that indecent song (gen)’
(KS)
With personal pronouns the situation becomes more difficult. First and second
person pronouns have several dialectal variants of case-forms, and even one and
the same speaker may use more than one form for a given case. Table 3 lists the ac-
cusative and genitive variants given by Cibuļs and Leikuma (2003; dialectal vari-
ants of the stem vowel are disregarded here), plus the forms manis and tevis given
by Bukšs and Placinskis (1973) for standard Latgalian and which are often used in
Table 3. Case-forms of pronouns (cf. Cibuļs & Leikuma 2003: 48; Bukšs & Placinskis
1973: 142)
1sg 2sg 1pl 2PL 3rd: m.sg f.sg m.pl f.pl
nom es tu mes jius jis jei jī juos
acc mani tevi myus jius jū jū jūs juos
myusus jiyusus
gen mane teve myusu jiusu juo juos jūs jūs
mani tevi myus jius jū jū
maņa teva
maņs tevs
mans tevis
manis
written texts of the 20th century.7 As can be seen, most variation is found with the
genitive. Potentially ambiguous forms within one column are highlighted.
Syncretism of genitive and accusative is regular for the masculine plural of
third person pronouns, but it also appears with first and second person pronouns.
In all these instances, it is the form of the accusative that takes the place of the
genitive rather than the other way around (unlike Slavic).8 Going by the above
table, the form jū in example (3) (with a single referent) can be determined as
accusative, while the form mani in example (4) would be ambiguous.
(3) a jei jau iz-līn dasmat-ā stund-ē
ptc 3sg.f.nom already pfx-climb.prs.3 tenth-loc.sg hour-loc.sg
i meklej jū pa baznīc-u
and search.prs.3 3sg.acc about church-acc.sg
‘yet she climbs out [of her tomb] already at ten o’clock and looks for him about
the church’ (UP)
(4) Maņ tevi tagad juo-pamat, a tu mani
1sg.dat 2sg.acc now deb-leave.prs.3 and 2sg.nom 1sg.acc/gen
meklēsi
search.fut.2sg
‘I have to leave you now, and you will look for me […]’ (UP)
However, the interpretation of the forms mani, tevi in UP as genitive forms is not
straightforward. The use of these forms for both accusative and genitive is typical
7. For the historical background see Endzelin (1922: 372–374).

8. Endzelin (1922: 420–421) reports the use of genitive forms instead of the accusative in va-
rieties of Latvian; his examples are from central and western varieties of Latvian, not from
Latgalian.
216 Nicole Nau
Table 4. Accusative and genitive forms of personal pronouns in UP

4a First person singular
Function mani (acc) mane9 (gen) acc for gen gen for acc
Object (104) 92 12 11/19 4/85
acc expected (85) 81 4
gen expected (19) 11 8
Compl. of Prep. (18) 3 15 3/18 0/0
4b Second person singular
Function tevi (acc) teva (gen) acc for gen gen for acc
Object (56) 46 10 9/19 0/37
Compl. of Prep. (19) 5 14 3/17 0/2
4c Third person singular (jū both genders, juo masculine, juos feminine)
Function jū (acc) juo/juos (gen) acc for gen gen for acc
Object (477) 414 63 38/96 5/381
acc expected (381) 376 5*
Compl. of Prep. (83) 64 29 6/35 0/58
* all juo (masculine)
for southern Latgalian dialects, but not for central dialects (Leikuma 2010: 63), to
which the language of Viļāni belongs. In the writings of Ilze Sperga no syncretism
is found: mani and tevi are unambiguous accusative forms, while for the genitive
of the first person singular mane and (rarely) manis are used, for the second per-
son singular respectively teve or (rarely) tevis. The suspected “syncretism” in UP
may also be explained in another way: as the use of accusative forms in functions
where the genitive is expected. To decide this question, more and broader inves-
tigations are needed. For this paper, I investigated the use of singular personal
pronouns in all fairytales of UP. The results are shown in Tables 4a–c.
9. Including 1 instance of manis as complement of a genitive-governing preposition.

As “object” I counted all appearances as complement of predicates.10 “Geni-

tive expected” refers to cases that will be discussed in detail in the next sections:
(i) the object of a negated verb, or of an infinitive in the scope of negation, (ii) the
object in a construction with the supine, or an infinitive in the function of a su-
pine, (iii) the object of verbs that as a rule govern the genitive in UP, such as me-
klēt ‘look for’ in example (4) (for a list of these verbs see Section 5). “Accusative
expected” refers to objects of transitive verbs in affirmative clauses. There are a
few doubtful instances, where the texts do not contain enough data for us to be
sure which case to expect: the verb nūrauduot ‘to mourn’ appears only a single
time, with juo as object (counted as genitive where accusative was expected); the
verb syuteit ‘to send’ (with various locational prefixes) was classed as “accusative
expected”, although it appears a few times with an object in the genitive; however,
accusative is by far more common. For complements of prepositions, accusative
is expected with the prepositions ar ‘with’, ap ‘about’, par ‘about, over’, pret ‘against’,
caur ‘across’, while genitive is expected with pi ‘at’, nu ‘from’, deļ ‘for’, piec ‘after’,
zam ‘under’, bez ‘without’, da ‘up to’. The preposition iz ‘to; onto; on’ is used with
both genitive and accusative (without a clear difference in meaning11) and was
therefore excluded from the count.
The tables show that the majority of uses allow a distinction of accusative
and genitive, but there is a slight tendency to use the accusative forms mani, tevi
and jū in positions where the genitive is expected, especially as objects. Some
examples will be discussed in later sections. In contrast, the number of instances
where a genitive form is used instead of an accusative is very low.
I conclude that the forms mani, tevi, and jū (singular) in UP can (and should)
be classified in all of their occurrences as accusative forms and not as variants of
the genitive. This in turn allows the thesis that accusative marking supersedes
genitive marking with personal pronouns more often than with nouns.
3. Genitive marking with transitive verbs: Clause-level properties
In this section I investigate the change of case-marking from accusative to geni-

tive under conditions that are associated with the whole clause, while semantic
properties of the object or the predicate are of no or only secondary importance.
Here, it is justified to say that the genitive is used “instead of the accusative”: with
10. In addition to verbal predicates, complements of the non-verbal predicatives bais ‘fear,
afraid’ and žāl ‘sorry’ were also taken into account.
11. That is, the distribution known from Latvian, where uz + genitive = ‘on’, while uz + accusa-
tive = ‘to’, is not observed in the texts under investigation.
218 Nicole Nau
the same lexical material, the genitive is used when a grammatical condition ap-
plies while the accusative is the unmarked choice for affirmative clauses with a
finite verb and realis modality.
3.1 Genitive marking in negated clauses
All varieties of Latgalian use the genitive for marking objects in negated clauses,
but they differ with respect to the extent of this use. In Bukšs’ grammar of standard
Latgalian a short and strict rule is given: „The negated object is in the genitive”
(“Nūlīdzamais objekts stōv genitivā”, Bukšs & Placinskis 1973: 296). The author
provides some short examples but no further comment. For UP and Ilze Sperga’s
prose the rule holds with only a few exceptions. In most modern texts, however,
accusative marking is found quite often. Examples from such texts will be given at
the end of this section. First, I will discuss object marking under negation in my
main sources, distinguishing several formal types.
As in Latvian and Lithuanian, clause negation (“it is not the case that P”) is
expressed in Latgalian by adding the negative prefix to the finite verb. In such
clauses, the direct object is regularly used in the genitive in Ulanowska’s fairy tales
as well as in the modern texts by Ilze Sperga. The rule affects all kinds of objects
(mass and count nouns, proper names, pronouns) and holds for all tense and
mood forms of the verb.
(5) tī vac-ī brauc otkon, Juoņ-a
dem:nom.pl.m old-nom.pl.m.def ride:prs.3 again Juoņs-gen
na-jem
neg-take:prs.3
‘the elder brothers ride out again, they don’t take John [with them]’ (UP)
(6) može jei man-e na-ap-jess
maybe 3sg.f.nom 1sg-gen neg-pfx-eat.fut.3
‘maybe she won’t eat me’ (UP)
(7) jis taid-u slykt-u drēb-u nikod na-beja
3sg.m such-gen.pl bad-gen.pl cloth-gen.pl never neg-be:pst.3
nosuojis
wear:ppa:m.sg
‘he had never worn such bad clothes’ (UP)
(8) tov-ys laim-is nikas tev na-at-jims!
2sg.poss-gen.sg.f luck-gen.sg nobody.nom 2sg.dat neg-pfx-take.fut.3
‘nobody will take your luck away’ (UP)
(9) Dodz-s to lob-s, Pīter-a na-kūstu.

Dodzs-nom ptc good-nom.sg.m Pīters-gen neg-bite.cnd
‘Dodz (the dog) is good, he wouldn’t bite Pīters’ (KS)
(10) Na-saun zvierbuļ-a, kas vēļ nav
neg-shoot.prs.2sg sparrow-gen.sg wh.nom yet neg.be.prs.3
izparā-t-s.
pfx-hatch-ppp-nom.sg.m
‘Don’t shoot a sparrow before it is hatched.’ (KS)
In Ilze Sperga’s short story (KS) investigated here, there is one finite negated clause
with an object in the accusative. It contains an idiomatic expression (maisīt kādam
golvu, literally “stir someone’s head”), which may prevent the change of case:
(11) Tu na-maisi puik-am golv-u ar sov-im
2sg.nom neg-stir.prs.2sg boy-dat.sg head-acc.sg with rpo-dat.pl.m
zmej-im!
snake-dat.pl
‘Don’t you baffle the boy (“stir the boy’s head”) with your [talking about]
snakes!’ (KS)
In UP, several instances of accusative marking in negated clauses are found that
are not in idioms and must be accepted as exceptions to the widely observed rule
of genitive marking. One regularity within the exceptions was already mentioned
by Endzelin (1922: 420; among others, he cites the example from UP below),
namely the use of accusative in a negated clause that is contrasted to an affirma-
tive clause:
(12) tu na-jem naud-u, pa-jem ūgl-is
2sg.nom neg-take.prs.2sg money-acc pfx-take.prs.2sg coal-acc.pl
‘Don’t take the money, take the coals!’ (UP)
Similar examples were also found in texts by Ilze Sperga (see also (20) below)
and by other authors. Note that in this construction, the activity denoted by the
predicate itself (the act of taking) is not negated.
Of the remaining instances of accusative marking in negated clauses in UP,
it appears that many contain a personal pronoun (examples (14)–(16)), while
nouns (example (13)) are less often accusative-marked (more research is needed
to prove this thesis):
(13) a tēteit-i na-redz i na-dzierd
but daddy-acc neg-see.prs.3 and neg-hear.prs.3
‘but she doesn’t see nor hear her daddy’ (UP)
220 Nicole Nau
(14) nikas jū na-redzēja

nobody 3sg.acc neg-see.pst.3
‘nobody saw him’ (UP)
(15) kab man-i sun-i na-sa-rau-tu
ptc 1sg-acc dog-nom.pl neg-pfx-tear-cond
‘if only the dogs won’t tear me off ’ (UP)
(16) es tev-i na-sysšu
1sg.nom 2sg-acc neg-kill.fut.1sg
‘I won’t kill you’ (UP)
With the personal pronouns that I have checked systematically (see Section 2
above), accusative marking is frequent. For the first person singular the accusa-
tive mani (10 tokens) is even more common than the genitive mane (5 tokens) as
object in a negated clause. For the third person singular I found 14 tokens of the
accusative jū and 21 tokens of a genitive form (8 masculine juo, 13 feminine juos)
in finite negated clauses.
Negation is further expressed with the negative focus particles ni and na
along a verb in the affirmative form. The first particle is used in coordinations
(ni …, ni … ‘neither … nor …’), while the second negates single elements of a
clause (see also Nau 2011a: 91). In Ilze Sperga’s writing, the object is in the geni-
tive if the clause entails the meaning ‘not P’, otherwise it is in the accusative. The
following examples illustrate this:
(17) Ni jei dorb-a zyna, ni dorb-s juos.
neg 3sg.f.nom work-gen know.prs.3 neg work-nom 3sg.f.gen
‘Neither does she know work nor does work know her.’ (KS)
(entails ‘she doesn’t know work, work doesn’t know her’)
(18) Na cylvāk-am tuo sprīs-t, kū Dīvs
neg man-dat.sg dem.gen.sg.m decide-inf wh.acc God-nom
struopēs voi aizmoksuos.
punish.fut.3 or reward.fut.3
‘It’s not for man to decide (“that”) whom God will punish or reward.’ (KS)
(implies ‘man doesn’t decide that’)
(19) Jis tū na Pīter-am soka.
3sg.m.nom dem.acc.sg neg Pīters-dat say.prs.3
‘It’s not to Pīters that he says that’, ‘He doesn’t say it to Pīters’ (KS)
(does not entail ‘he doesn’t say it’)
As with verbal negation in example (12) above, the accusative is also used when a
negated object is contrasted to a non-negated object of the same verb:
(20) Lai lobuok mozgoj greid-u, na mun-us kaul-us
ptc better wash.prs.3 floor-acc.sg neg my-acc.pl.m bone-acc.pl
‘She’d better wash the floor, not my bones.’ (IS)
The rule that negation of the object with non-negation of the event triggers ac-
cusative marking (or is at least compatible with it) may also explain the accusative
in the following example with negation expressed by the verbal prefix:
(21) A jis at-soka: Nā, šys vīn-u
ptc 3sg.m.nom pfx-say.prs.3 neg log.nom.sg.m one-acc.sg
na-puordūškys, – pierc divi!
neg-sell.pfa.m.sg buy.prs.2sg two
‘But he replied: No, I won’t sell [just] one – buy two!’ (UP)
The negated clause in (21) is the swineherd’s answer to the princess’ request to
sell her one of his dancing pigs. He does not deny the possibility of selling (any of
his pigs). While we have different verbs in this sentence, the meaning is similar
to that of example (12) or (20). We may rephrase it as ‘I won’t sell one, but I will
sell two’.
Objects of an infinitive that depends on a negated predicate may be in the
accusative or the genitive. Genitive marking is usually found with negated pos-
sibility, as in the following examples:
(22) oc-u na-var at-rau-t nu tuos meitin-is
eye-gen.pl neg-can.prs.3 pfx-tear-inf from dem.gen.sg.f girl-gen.sg
‘he can’t tear his eyes away from this girl’ (UP)
(23) Es na-muoku durov-u attaisē-t!
1sg.nom neg-can.prs.1sg door[pl]-gen open-inf
‘I can’t open the door!’ (UP)
(24) Tik Pišk-ys na-var jimt leidza
ptc Piška-gen neg-can.prs.3 take.inf along
‘He just can’t take along Piška’ (KS)
With other negated matrix predicates, there is more variation. Again, in many ex-
amples with accusative marking the object is expressed by a personal pronoun. It
is possible that intensionality plays a role with matrix predicates other than those
of negated possibility. Compare the following examples:
222 Nicole Nau
genitive marking
(25) Jau vieli beja, sīv-a na-gribēja guņ-s
already late.adv be.pst.3 wife-nom.sg neg-want.pst.3 fire-gen.sg
kūr-t.
light-inf
‘It was already late, the wife didn’t want to light a fire.’ (UP)
(26) Ka jis na-ļūbej tuo stuosteit, a vajag.
that 3sg.m.nom neg-like.prs.3 dem.m.sg.gen tell.inf but need.prs.3
‘That he doesn’t like to tell this, but has to.’ (KS)
accusative marking
(27) A jī na-grib ni par kaid-u naud-u
ptc 3pl.nom.m neg-want.prs.3 neg for some-acc.sg money-acc
jū puordūt
3sg.acc sell.inf
‘But they don’t want to sell him [= the little boy] for any price’ (UP)
(28) valn-am navā val-is iz-dzeit jū
devil-dat.sg neg.be.prs.3 freedom-gen.sg pfx-drive.inf 3sg.acc
‘the devil may not (is not allowed to) chase him [= the servant] away’ (UP)
(29) At-skrēja sulain-i, pa-za-vēra vysur,
pfx-run.pst.3 servant-nom.pl pfx-rfx-look.pst.3 everywhere
na-dreikstēja miš-u puormaisēt
neg-dare.pst.3 mass-acc.sg disturb.inf
‘The servants came running, they looked everywhere, they didn’t dare disturb
the mass’ (UP)
In the sentences with an accusative object, the existence of the referent of this
object is independent of the predicate, while the existence of the fire in example
(25) and the thing told in (26) is relative to the possible world evoked by it (if
she doesn’t want to light a fire, there probably won’t be a fire; if he doesn’t tell it,
there will be no story). This would explain why personal pronouns, which usually
refer to persons and objects already in existence, often are in the accusative in
such constructions. However, my data are too scarce to allow a definite conclu-
sion about the importance of this feature.
The following example is interesting in that the negation has floated from the
infinitive (where it would be expected according to the meaning of the clause) to
the matrix verb.12 The meaning is ‘they told us not to scold and not to beat her’,
12. As Axel Holvoet pointed out, there is a parallel in Polish, where nie kazać + infinitive may
mean either ‘not command to do’ or ‘forbid to do’. An example found on the Internet: Nie pow-
iem, Teść nie kazał mówić ‘I won’t say, my father-in-law told me not to say [where I caught the
but the construction literally translates ‘they didn’t tell us to scold and beat her’.
The direct object is in the genitive.
(30) Buor-t juos na-lyka i sis-t juos na-lyka
scold-inf 3sg.f.gen neg-tell.pst.3 and beat-inf 3sg.f.gen neg-tell.pst.3
i kai jū iz-vuicei-t da dorb-a, ka na-var
and how 3sg.acc pfx-teach-inf to work-gen.sg if neg-can.prs.3
ni sis-t, ni lomuo-t?
neg beat-inf neg abuse-inf
‘They told us not to scold her and not to beat her, but how can we teach her
to work if we may neither beat nor berate?’ (UP)
Note that this construction formally meets the definition of negative raising given
by Noonan (2007: 100),13 but deviates from the typical instances such as I don’t
believe he’ll come = I believe he won’t come. Other than with predicates of propo-
sitional attitude, it is not evident that the meaning of ‘not order to do’ should be
‘order not to do’, and indeed in Latgalian not all instances of nalikt ‘not order’ have
this meaning (the same holds for Polish, where, as I noted perusing examples
found on the Internet, nie kazać only in a minority of instances has the “raising”
meaning). For example, at the beginning of one fairy-tale the malevolent step-
mother orders her husband to kill the sons of his first marriage, but at the end
these sons turn out to be strong and useful for the household, so she ‘didn’t order
to kill (get rid of) them anymore’ (Latgalian: nalyka gubēt jūs vairuok, Polish: nie
kazała gubić ich więcej), which is not the same as ‘ordered not to kill’.
The combination of an infinitive with the negative prefix is not frequent.
The few examples with a direct object found so far show accusative marking, for
example:
(31) Tys nav tai kai ar azar-a bols-u,
dem.nom.sg.m neg.be.prs.3 so as with lake-gen.sg voice-acc.sg
kū var na-dzierdē-t, ka klausuos viej-a.
wh.acc can.prs.3 neg-hear-inf if listen.prs.3 wind-gen.sg
‘It’s not as with the voice of the lake, which one might not hear (which one
may fail to hear) if one listens to the wind.’ (KS)
fish]’ [http://www.wedkuje.pl/wedkarski-rekord,szczupak,90391]. It is interesting to note that

the Polish translation of the sentence in (30) given in UP does not use this construction, but nie
pozwolili ‘didn’t allow’. There is only one example in UP where nalyka ‘didn’t tell to’ = ‘told not
to’ in UP has the translation equivalent nie kazał.
13. “Negative raising is the name applied to the situation where a negative marker appears to
be removed from the complement clause with which it is logically associated and raised to the
ordinary position for negatives within the matrix clause” (Noonan 2007: 100).
224 Nicole Nau
Note that the clause “one might not hear the voice” does not entail “one doesn’t
hear the voice”. The modal meaning of possibility is foregrounded, the negation is
backgrounded, and I suppose that accusative marking was favored for this reason.
As stated at the beginning of this section, in modern texts we often find accu-
sative marking within negated clauses. It would be interesting to investigate when
this trend set in and how it proceeded. While I suppose influence from Standard
Latvian to be the main factor in the decline of genitive marking, it may well be
that such influence met with a tendency already at work in Latgalian. Another
question for further investigation is possible parallels with Russian, which now-
adays uses the genitive of negation only under certain circumstances (see Kagan
2013 for a recent monographic treatment). Within the scope of this paper I am
not able to determine the degree to which genitive marking of objects in negated
clauses is still common in contemporary written Latgalian, or whether it appears
in situations which trigger genitive marking in Russian. From a cursory perus-
al of the corpus MuLa-1.0 I gained the impression that genitive and accusative
marking are roughly the same in frequency (recall that this corpus includes many
texts by Ilze Sperga, who uses the genitive consistently). I took a closer look at
texts written by Valentīns Lukaševičs (another writer from central Latgalia whose
prose I highly value, though his syntax is less conservative) and found that only
the negative pronoun nikas ‘nothing, nobody’ was used as a genitive object in ne-
gated clauses (this pronoun does not appear in the accusative in these texts at all),
while nouns and other pronouns are marked by the accusative as in affirmative
clauses.
(32) Tod mes Zig-u ilg-uok-u laik-u na-redziejom.
then 1pl.nom Ziga-acc long-comp-acc.sg time-acc.sg neg-see.pst.1pl
‘After that we didn’t see Ziga [= a man’s name] for quite a while.’ (VL)
(33) ni-vīn-s tai eisti tū vēļ navā pamaniejs
neg-one-nom so really dem.acc.sg yet neg.be.prs.3 notice.pst.3
‘nobody has really noticed this yet’ (VL)
(34) A sābr-ūs dzeivoj Dyņuk-s, nikuo na-dora.
ptc neighbour-loc.pl live.prs.3 Dyņuks-nom nothing.gen neg-do.prs.3
‘And next door lives Dyņuks, who doesn’t do anything.’ (VL)
(35) Bet ar cegl-u ni-vīn-s na-var nikuo
but with brick-acc.sg neg-one-nom neg-can.prs.3 nothing.gen
iz-darei-t.
pfx-do-inf
‘But nobody can do anything with a brick.’ (VL)
This situation seems to be just one step behind that of Modern Standard Latvian,
where objects in negated clauses are always in the accusative, only the negative
pronouns nekas ‘nothing’ and neviens ‘nobody’ sometimes appear in the genitive,
though not as often as in Latgalian. This can be seen when comparing the one-
million-word corpora of both languages. In the Latgalian corpus MuLa-1.0, the
accusative nikū has only 13 tokens while the genitive nikuo is attested with 388
tokens. In the Latvian corpus miljons-1.0 we find 331 instances of the accusative
neko ‘nothing, (not) anything’.14
3.2 From negation to irrealis
In Ilze Sperga’s short story (KS) there are several instances of genitive objects in
formally affirmative clauses with a pragmatically implied negation. These sen-
tences express a situation that the speaker holds to be unreal, impossible. There
is a strong factor of subjectivity, subjective judgment – meanings such as ‘I don’t
believe this is true’, ‘I doubt this will happen’, ‘I wish this would happen/had
happened’.
The following example may be considered a borderline instance of overt
negation:
(36) Nā, tai na-var byut, ka Dīv-s radzātu tuo
neg so neg-can.prs.3 be.inf that God-nom see.cond dem.gen.sg.m
vys-a i vēļ ī-sa-vārtu, kū kur-s dora na
all-gen.sg.m and ptc pfx-rfx-look wh.acc each-nom do.prs.3 neg
pa jam. Da i vēļ atrostu laik-a nū-struopēt
prep 3sg.m.dat ptc ptc ptc find.cond time-gen.sg pfx-punish.inf
vys-us nakryškuon-us i slaist-us.
all-acc.pl.m condemned[noun]-acc.pl and slacker-acc.pl
‘No, it can’t be that God sees all this and what is more notices what each
one does wrong. And in addition would find the time to punish all the con-
demned and the slackers.’ (KS)
(implies: ‘God does not see all this’; ‘God does not find the time’)
Note the use of irrealis mood in Latgalian. Another subgroup of sentences with
implied negation comprises counterfactual conditionals, where the predicate is in
the compound (past) form of the irrealis mood (byutu + past active participle).
14. The number of tokens of the genitive form Latvian nekā cannot be determined automat-
ically, as this form is homographic with an adverb meaning ‘in no way’, which is much more
frequent.
226 Nicole Nau
(37) Byutu koč kačalān-a at-ness, i to vaira

be.cond ptc kitten-gen.sg pfx-carry.ppa.m.sg ptc ptc more
poļz-is byutu.
use-gen be.cond
‘Had he brought home a kitten / I wish he had brought home a kitten, there
would have been more use after all.’ (context: Granny is reasoning after her
grandson had given his brother a pet rat) (KS)
(entails ‘he didn’t bring home a kitten’)
(38) Byutu sov-u bagadzeļņ-u sa-kuortuojs, a to
be.cond rpo-gen.pl mess-gen.pl pfx-put.in.order.ppa.m.sg ptc ptc
bais i vyd-ā īt.
afraid ptc centre-loc go.inf
‘He should have cleaned up all his mess15 (I wish he had cleaned up his mess),
now one fears even to enter.’ (context: Granny complains about her messy
husband) (KS)
(entails ‘he didn’t clean up the mess’)
A slightly different instance is the use of the genitive in the following example:
(39) Kab na tuo vādar-a, voi ta tys

if neg dem.m.sg.gen belly-gen.sg qu ptc dem.m.sg.nom
zmej-s byutu juos preciejs.
snake-nom.sg be.cond 3sg.f.gen marry.ppa.m.sg
‘If there hadn’t been this belly, would the pig ever have married her?’ (Mar)
(implies: ‘he wouldn’t have married her’)
Here, we are not dealing with counterfactuality in the real world (after all, he
did marry her), but with incompatibility with a possible world imagined by the
speaker. The same holds in the following example, where the verb is in the simple
form of the conditional:
(40) kurs ta tī āstu tūs zuoļ-u
who.nom ptc ptc eat.cond dem.gen.pl medicine[pl]-gen
‘who would want to eat this medicine’ (context: if the calf that was given anti-
biotics were slaughtered, its meat would not be suitable for humans to eat)
(Mar)
(implies: ‘nobody would eat it’)
15. I have translated bagadzeļne as ‘mess’, the meaning it has in this context. The primary mean-
ing of bagadzeļne is ‘home, shelter for the poor and homeless’.
Irrealis meaning may also be expressed with other forms of the verb. In the fol-
lowing examples these are simple future (41) and imperative16 (42):
(41) Da i kas zač-am slid-u dūs.
ptc ptc wh.nom hare-dat skate-gen.pl give.fut.3
‘Who will/would give a hare skates.’ (KS)
(implies: ‘nobody gives skates to a hare’)
(42) Ej tī saprūti tūs sīvīš-u.
go.prs.2sg ptc understand.prs.2sg dem.gen.pl woman-gen.pl
‘Try and understand women.’ (KS)
(implies: ‘you can’t understand women’)
The use of the genitive for objects of transitive verbs in irrealis clauses is fairly
regular in Ilze Sperga’s writing. While I haven’t so far found similar examples in
other texts, speakers of Latgalian attest that they are perfectly natural. Research
on (spontaneous) spoken Latgalian may lead to more insights about this usage.
3.3 Constructions with the infinitive or the supine
Latgalian has two inherited non-finite forms of verbs: the infinitive and the su-
pine. Both are formed from the same stem but take different endings, which in
turn trigger different morphophonological shapes of the word-form (see Nau
2011a: 60–61 for details). The ending of the infinitive is a palatalized -tj (the pala-
talization is not indicated in contemporary and 20th-century orthography), while
the ending of the supine is -tu. The supine is (or was in the mid-20th century)
also found in some Central Latvian dialects in North Latvia as well as in Selonian
dialects bordering on Latgalian dialects (for lists of these dialects see Endzelin
1922: 713; Gāters 1977: 139). In western dialects, standard Latvian, and standard
Lithuanian the supine has been lost.
The supine is used in constructions with verbs of motion and expresses an ac-
tion that is the purpose of the motion. The object of a transitive verb in the supine
is in the genitive. It more often precedes the verb, but word order is not strict, as
can be seen in the following examples.
(43) jei pameta maiz-i i nu-skrēja vylk-a systu
3sg.f.nom leave.pst.3 bread-acc and pfx-run.pst.3 wolf-gen.sg beat.sup
‘she left the bread and ran outside to beat the wolf ’ (UP)
16. The imperative of the second person singular is formally identical to the indicative present.
I do not indicate the imperative function in the morphological glosses.
228 Nicole Nau
(44) a jei na-zyna, ka jī nū-brauce

but 3sg.f.nom neg-know.prs.3 that 3pl.m.sg pfx-ride.pst.3
kautu otkait ļauž-u
kill.sup again people-gen
‘but she didn’t know that they had left to kill people once more’ (UP)
Word order may be rather a matter of information structure. Constructions with

a supine and an object that precedes it are often found at the beginning of a para-
graph. Compare the following fragment in Latgalian and its Polish translation (as
given by Ulanowska). While Polish does not have a supine and uses the infinitive
here, the word order is the same as in Latgalian:
(45) Jis guoja, tys tāvs jau, cik dīnys viežu giutu i ni vairuok, ni mozuok, vyss pa
treis viņ vieži giva, ar tū viņ i dzeivuoja. Vot tys dāls saslyma i nūmyra, – nūīt
tys tāvs giutu viežu, – sagiun div viņ, – a piec tuo jau nivīnys nadabuoja.
Chodził ten ojciec codzień raki łapać i ani więcej, ani mniej, tylko wciąż po
trzy raki łapał i tem tylko żyli. Syn zasłabł i umarł, – poszedł ojciec łapać
raki, – złapał tylko dwa, a potem już ani jednego nie dostał.
‘The father went each day to catch crayfish [object – supine/infinitive] and
always caught three, no more, no less, and that was all they had for living. One
day the son got ill and died. Again the father went to catch crayfish [supine/
infinitive – object] and caught only two, but afterwards he didn’t get a single
one anymore.’ (UP)
The supine may be replaced by an infinitive while genitive marking is preserved.

This happens probably more often in modern texts, but is attested in UP, too.
Compare:
supine
(46) syuta jau durak-u iz mīst-u telerdž-u pierktu
send.pst.3 ptc simpleton-acc.sg to town-acc.sg plate-gen.pl buy.sup
‘they sent the simpleton to town to buy plates’ (UP) – supine
infinitive instead of supine

(47) jis brauce iz dzeraun-i ūgļ-u pierkt’
3sg.f.nom ride.pst.3 to village-acc.sg coal-gen buy.inf
‘he went to the village to buy coal’ (UP) – infinitive (ending -tj)
The use of the genitive in constructions with an infinitive that replaces a supine
has a parallel in Lithuanian (cf. Arkadiev, this volume). In Central Latvian di-
alects, on the other hand, a supine with a genitive object has been replaced by
an infinitive with the object in the accusative. This can be seen in dialects of
Northern Livland. The following examples are taken from the autobiography of a
farmer, written at the beginning of the 20th century and published in 2008, pre-
serving dialectal features.
Central Latvian
(48) Kad tēvs mani pirmoreiz baznīcā veda, piebraucām pie Mellāķīša kroga un
tēv-s iet tūlin iekšā no rīt-a šņab-a
father-nom go.prs.3 at.once inside of morning-gen schnapps-gen
no-dzer-tu.
pfx-drink-sup
‘When my father drove me to church for the first time, we passed Mellaķītis’
pub and father went in at once in the morning to drink schnapps.’ (Ādama
stāsts)
(49) Un nu man-s tēv-s ie-iet Šimk-a
and now my-nom.sg.m father-nom.sg pfx-go Šimkus-gen
muiž-as krog-ā dūš-u uzprāvī-t, šņab-i
manor-gen.sg pub-loc.sg courage-acc.sg fix-inf schnapps-acc.sg
ie-dzer-t
pfx-drink-inf
‘And now my father enters the pub at Šimkus Manor in order to improve his
courage, to drink schnapps.’ (Ādama stāsts)
In Latgalian, accusative marking is more usual in genuine infinitive construc-

tions. For example, with phasal verbs such as ‘start’ only accusative marking is
possible.
(50) jam jei vāz-u i suok vec-i
take.prs.3 3sg.f.nom stick-acc.sg and start.prs.3 old.man-acc.sg
sist
beat.inf
‘she takes a stick and starts to beat the old man’ (UP)
With an infinitive depending on the verb gribēt ‘want’ an object is sometimes in

the genitive, though accusative marking is more common. Possibly the choice
of the genitive depends on additional factors (this needs further investigation).
Compare the following two examples:
accusative marking
(51) Dzed-s gribēja iz-lomuot tū dzāruoj-u,
Grandpa-nom.sg want.pst.3 pfx-scold.inf dem.acc.sg drunkard-acc.sg
a na-sacēja nikuo.
ptc neg-say.pst.3 nothing.gen
‘Grandpa wanted to tell off the drunkard, but didn’t say anything.’ (KS)
230 Nicole Nau
genitive marking
(52) Ka kotr-ys vec-s gribēs Dīv-a školē-t,
if each-nom.sg.m old.man-nom.sg want.fut.3 God-gen educate-inf
kas tī beig-uos vaira parād-a pastateis
wh:nom here end[pl]-loc still order-gen.sg establish.fut.3
‘If any old man wanted to educate God, who would maintain order in the end.’
(KS)17
The sentence in (51) describes a real situation, that in (52) a hypothetical situ-
ation. The second clause of (52) (‘who would maintain order’) is an example of
pragmatically implied (subjective) negation (‘nobody would maintain order’),
thus one of the types discussed in the previous section (compare with examples
(40) and (41)). Such pragmatic implications are typically contained in rhetorical
questions with infinitive construction, where again genitive objects can be found.
The following example is similar to (39), which contained a rhetorical question
with a finite verb as the second part of a conditional sentence with the same type
of protasis (kab na NP ‘if there weren’t/hadn’t been NP’)
(53) Kab na tuos, i dūmoj, kai mieneš-a
if neg dem.gen.sg.f ptc think.prs.2sg/3 how month-gen.sg
beig-u sa-gaideit, kai bārn-a iz-audzēt.
end-gen pfx-wait.inf how child-gen.sg pfx-raise.inf
‘If she hadn’t this [money], imagine, how would she make ends meet (lit. “how
to meet the end of the month”), how could she raise the child.’ (KS)
(implied: ‘she wouldn’t make ends meet, she couldn’t raise the child’)
As in the instances discussed in Section 3.2, genitive marking in such construc-

tions is characteristic of the prose of Ilze Sperga. The wider distribution of this
usage is a matter for further research.
Infinitives depending on a predicate expressing deontic modality (possibil-
ity and necessity) as a rule combine with accusative objects, except for negated
possibility as discussed above. The main device for expressing deontic necessity
in Latgalian is the verb vajadzēt ‘must, be needed’ (for the marking of nominal
arguments of this verb see Section 5.1 below), for example:
(54) vajaga maiz-i pelnīt
need.prs.3 bread-acc.sg earn.inf
‘one has to earn one’s bread’ (UP)
17. Note that vaira, which in other contexts has the meaning ‘more’, is not a quantifier in this
sentence.
In UP, there are no genitive objects of infinitives depending on vajadzēt; in KS I

found one, but in other texts by Ilze Sperga there were more examples with ac-
cusative marking. There may be varieties of Latgalian where genitive marking is
more usual. Cibuļs and Leikuma (2003: 83) write that the object of an infinitive
depending on a form of vajadzēt is ‘usually’ (parasti) in the genitive. However,
their own examples given earlier on the same page show accusative marking.
The authors further mention that “in dialects sometimes the nominative is used”
(izloksnēs reizēm lieto arī nominatīvu). One such example was found in UP:
(55) cikpat tei maiz-e at-dūt vajaga
anyhow dem.nom.sg.f corn-nom.sg pfx-give.inf need.prs.3
cyuk-om niu!
pig-dat.pl now
‘anyhow now we have to give this corn to the pigs!’ (UP)
4. Different marking for different kinds of NPs
4.1 Mass nouns and indefinite quantity
In this section I will examine instances that are traditionally treated under the
name genitivus partitivus (for Latvian including Latgalian cf. Endzelin 1922: 411–
412), in Latgalian dolomais genitivs (cf. Bukšs & Placinskis 1973: 297). An early
insightful description of the use of the genitive for marking partial objects is given
by Strods, who writes in his Latgalian dictionary:
If a noun is used as an object of an affirmative verb, then it is in the accusative;
if it is the object of a negated verb, it is in the genitive […]. But if a mass noun is
used as an object then it is in the genitive with affirmative verbs, too, except for
the case in which we refer with this noun to a known, definite part of the object.
For example, when one buys milk, butter, meat [pīnu, svīstu, gaļu – acc] in the
shop one says: dūd maņ pīna ‘give me milk.gen’, dūd maņ svīsta ‘give me butter.
gen’, dūd maņ gaļas ‘give me meat.gen’, dūd mads ‘give honey.gen’, dūd cukra
‘give sugar.gen’. On the other hand, if we talk about all the known amount we say:
dūd bārnim pīnu ‘give the children the milk.acc’; padoļ olu ‘distribute the beer.
acc’; pōrdūd madu ‘sell the honey.acc’; pōrdūd svīstu ‘sell the butter.acc’ etc. By
this form we understand a known amount of milk, beer, honey, butter, of which
all can be bought, sold or distributed, but in the shop the client doesn’t buy the
whole object, only a part of it.18 (Strods 1933: 15)
18. Free translation by the author of this article. The original text goes: “Jo lītas vōrds ir kai pa-
piļdeitōjs pēc pastypryunojūšō darbeibas vōrda, tad jys stōv akuzativā, – jo pēc nūlīdzūšō, tad
232 Nicole Nau
It is interesting that Strods here presents the genitive as the unmarked case for
mass nouns as objects, while accusative marking requires a motivation. In the
texts under investigation mass nouns do indeed frequently appear with genitive
marking; however, the accusative is found often as well. For example, in UP the
noun gaļa (or gale) ‘meat’ as an object of a transitive verb was found 9 times in
the genitive and 8 times in the accusative. Furthermore, the explanation given by
Strods is insufficient, although it provides a useful point of departure. There are
several open questions which await further investigation, such as: Do all mass
nouns potentially appear with genitive marking, or is there a restriction to a se-
mantic subclass (such as substances)? Which transitive verbs allow genitive mark-
ing of such objects? (In my texts, there is a relatively small class of basic verbs
that often appear with genitive-marked mass nouns: ‘give’, ‘buy’, ‘carry’, ‘take’, ‘eat’,
‘drink’, but I don’t know whether there is any systematic restriction.) Are there
restrictions (or preferences) with regard to the aspectuality of the verb? I suspect
that the answers to these questions will not be identical for texts from different
periods (such as UP and KS), different genres, or different dialects. In this paper,
however, I restrict myself to the presentation of some examples and rather general
comments.
A mass noun in the genitive most often refers to an indefinite quantity of the
referent. The following are typical examples from UP:
(56) at-skrīn iz jū pel-eit-e, soka tai:
pfx-run.prs.3 to 3sg.acc mouse-dim-nom.sg say.prs.3 so
“Ī-dūd tu maņ maiz-is!”. Jis
pfx-give.prs.2sg 2sg.nom 1sg.dat bread-gen.sg 3sg.m.nom
ī-deve maiz-is.
pfx-give.pst.3 bread-gen.sg
‘a little mouse came running to him and said: “Give me bread!” He gave [the
mouse] bread.’ (UP)
(57) a valn-s jau nas naud-ys i ber capur-ē
and devil-nom ptc carry.prs.3 money-gen and pour.prs.3 hat-loc
‘and the devil carries (= fetches) money and pours (it) into the hat’ (UP)
genitivā […]. Bet, jo vīlas lītas vōrds ir kai papiļdeitōjs, tad ari pēc pastyprynojūšim darbeibas
vōrdim stōv genitivā, izjamūt tū gadijumu, kad mes ar šū lītas vōrdu nūsaucam zynamu, nūteik-
tu tō prīksmeta daļu, pīm.: jo būdē pērk pīnu, svīstu, gaļu, tad soka: dūd maņ pīna, dūd maņ
svīsta, dūd maņ gaļas, dūd mads, dūd cukra un otkon, jo runojam par vysu zynomu daudzumu,
tad sokom: dūd bārnim pīnu; padoļ olu; pōrdūd madu; pōrdūd svīstu u. t. t. Ar šū formu mes
saprūtam zynomu pīna, ols, mads, svīsta daudzumu, kurū vysu var jimt un nūpērkt, pōrdūt vai
padaleit, bet būdē pērcēja pērk na vysu tū prīkšmetu, bet tikai jō dalu” (Strods 1933: 15). I am
grateful to Lidija Leikuma who brought this passage to my knowledge.
Count nouns in the plural can be treated in the same way as mass nouns:
(58) Jī i roza-dzeivuoja, i maiz-is pierka i

3pl.nom ptc pfx:rfx-live.pst.3 and bread-gen.sg buy.pst.3 and
trauk-u
bowl-gen.pl
‘They started keeping house, bought bread and dishes’ (UP)
(59) jei pa-jēme ļauž-u, aiz-moksuoja jim, […]
3sg.f.nom pfx-take.pst.3 people-gen pfx-pay.pst.3 3pl.m.dat
‘she took (some) people, paid them’ (UP)
In the above examples, reference was made to an indefinite quantity of a substance

or an indefinite number of objects/people. A slightly different, though closely re-
lated meaning is the designation of a part of a given whole (a partitive in a narrow
sense), as exemplified in the following two sentences:
(60) Soka tī saiminīk-i: “Pa-jem tu
say.prs.3 dem.nom.pl.m farmer-nom.pl pfx-take.prs.2 2sg.nom
sev naud-ys!”
rfx.dat money-gen
‘The farmers say: “Take (some of the) money for yourself!”’ (UP; context: the
hero had rescued a large amount of money in a haunted barn and given it to
the farmers)
(61) Ej tu, vec-s at-nes jiudiņ-a
go.prs.2sg 2sg.nom old.man-nom pfx-carry.prs.2sg water-gen
dzeiv-oa
alive-gen.sg.m.def
‘You go, old man, fetch (some of the) water of life’ (UP; context: the addressee
possessed a well of water of life)
As stated by Strods in the quotation above, the accusative is used with mass nouns
with reference to a known amount, usually the whole amount available in the
context, as well as for individuation in cases where the noun may refer to a mass
and to an element of the mass. Both cases are illustrated in the following example
with the word maize in its second meaning ‘grain, wheat and the like’ (in examples
above it occurred in its first meaning ‘bread’).
234 Nicole Nau
(62) ituos skudr-is losa jau tū maiz-i

dem.nom.pl.f ant-nom.pl gather.prs.3 ptc dem.acc.sg grain-acc.sg
sevkur-a skudr-a i nas sov-u maiz-i iz
each-nom.f ant-nom.sg ptc carry.prs.3 rpo-acc.sg grain-acc.sg to
orūd-a […] Da gaism-ys, juos iz-laseja vys-u
box-gen.sg Until light-gen.sg 3pl.f.nom pfx-gather.pst.3 all-acc.sg
maiz-i
grain-acc.sg
‘the ants now gather the wheat, each ant carries its grain into the box […] by
morning they have sorted out all the wheat’ (UP; context: the lazy hero had
thrown all sorts of grain together and then had ants sort it out)
The above example also shows the typical combination of a mass noun in the ac-
cusative with a determiner: a demonstrative pronoun, a possessive pronoun, or
the word vyss ‘all’. The determiner is not obligatory, as definiteness doesn’t have to
be marked in Latgalian:
(63) Nu, kū ta tu jimsi? voi naud-u, voi ūgl-is?
ptc wh:acc ptc 2sg.nom take.fut.2sg q money-acc q coal-acc.pl
‘Well, what will you take? The money or the coals?’ (UP; context: the heroine
was shown a box of money and a box of coals and had to choose between
them for her reward)
Besides total amounts and definite quantities, also indefinite specific amounts are
expressed by a noun phrase in the accusative. In this situation a modifying adjec-
tive is often found, as in the following example.
(64) taisa lyl-u gūd-u, dzekovaj jim i
make.prs.3 big-acc.sg feast-acc.sg thank.prs.3 3pl.dat and
moksoj lyl-u naud-u
pay.prs.3 big-acc.sg money-acc.sg
‘[the king] organizes a big feast, thanks them and pays (them) big money’
(UP)
Not all instances of genitive vs. accusative marking of mass nouns as objects in
finite affirmative clauses can be explained by the opposition of indefinite/part vs.
definite (or specific)/whole. The accusative is also used for indefinite quantities
of a mass or group in two related situations: (i) when reference is made to a type
rather than to a concrete event, (ii) when talking about habitual situations. The
first can be illustrated by a clause from Strods’ (1933: 15) own text quoted above:
(65) jo būd-ē pērk pīn-u, svīst-u, gaļu,

if shop-loc.sg buy.prs.3 milk-acc.sg butter-acc.sg meat-acc.sg
tad soka: […],
then say.prs.3
‘when one buys milk, butter, or meat in a shop, one says: […]’
A habitual situation is depicted in the following example from UP:
(66) I beja tī kēv-eit-e, kur jiudin-i vada

and be.pst.3 there mare-dim-nom rel water-acc transport.prs.3
‘And there was a mare that carried water’ (UP)
In Ilze Sperga’s writing the rules for genitive and accusative assignment of objects
containing a mass noun or a plural noun are similar to those found in UP. In de-
scriptions of habitual events the accusative is used:
(67) gastiņč-am nas peirāg-us i tod kūpā dzer
present-dat.sg bring.prs.3 cake-acc.pl and then together drink.prs.3
čaj-u
tea-acc
‘on her visits she brings along cakes and then they have tea together’ (KS)
However, the description of a habitual event may also contain a genitive object.
In such cases the verb appears with a prefix, as happens in Lithuanian and in
Slavic languages (see Seržant, this volume, for Lithuanian and further referenc-
es; of course, it cannot be taken for granted – or even expected – that Latgalian
behaves just as Lithuanian). Further research is needed regarding the range and
meaning of possible prefixes, as well as their obligatoriness. In example (68) the
prefix ī- has a clear local meaning (ī-līk ‘puts in’), while in (69) (ī-dūd ‘gives’) and
(71) (ī-dzer ‘drinks’) it is a mere marker of boundedness.19
(68) tok nu reit-a, cikom Pīters vēļ guļ,
but of morning-gen.sg while Pīters-nom still sleep.prs.3
Pišk-ys kletk-ā ī-līk iedīņ-a
Piška-gen cage-loc.sg pfx-put.prs.3 food-gen.sg
‘but in the morning, while Pīters is still asleep, (she) puts food into Piška’s
cage’ (KS)
19. Because of the different functions a prefix may have, and because the function often arises
from the construction rather than the morpheme, I always gloss prefixes with a label referring
to the form, “pfx”, without indicating the function it has in the particular example (for example,
local, completive, cumulative).
236 Nicole Nau
(69) Ka Pīter-s nav iz-dariejs nikuo švak-a,

if Pīter-nom neg.be.prs.3 pfx-do.ppa.m.sg nothing.gen bad-gen.sg.m
bab-a jam ī-dūd kompet-u.
granny-nom 3sg.m.dat pfx-give.prs.3 candy-gen.pl
‘If Peter hasn’t commited anything wrong, granny gives him sweets.’ (KS)
also possible reading: ‘… granny gives him a candy (accusative singular)’.
In descriptions of single events, genitive marking also occurs together with prefix-
ing, while accusative marking occurs with non-prefixed verbs (again, as in Lithu-
anian and Slavic, but more investigation on the Latgalian specifics is needed):
(70) Tik gord-ūs blīņ-u iz-cap-us-e.
so delicious-gen.pl.def pancake-gen.pl pfx-bake-ppa-f.sg
‘She has made such delicious pancakes.’ (KS)
(71) Reit-a dorb-i ap-darei-t-i, ī-dzers
morning-gen.sg work-nom.pl pfx-do-ppp-pl.m pfx-drink.fut.3
kopej-is i liksīs kaid-u strēč-eit-i
coffee-gen and put.fut.3.rfx some-acc.sg moment-dim-acc.sg
at-stīpt kuoj-u.
pfx-stretch.inf leg-gen.pl
‘When the morning jobs are done, (she) will drink coffee and lie down for a
little while to stretch her legs.’ (KS)
(72) Dzed-s verās iz juos i dzer sov-u
Grandpa-nom look.prs.3 to 3sg.gen.f and drink.prs.3 rpo-acc.sg
kopej-u.
coffee-acc
‘Grandpa looks at her and drinks his coffee.’ (KS)
4.2 Other cases: Pronouns, proper names, and count nouns
While genitive marking of mass nouns and count nouns in the plural is fairly
regular, the motivation for genitive marking of other classes of nominals in finite
affirmative clauses is not clear. Such instances occur in all texts under investiga-
tion, but they differ from each other and do not show a common tendency.
As already mentioned, the negative pronoun nikas ‘nothing, no one’ is over-
whelmingly used in the genitive when it appears as the object of a transitive verb.
The use of the genitive with this pronoun is much more widespread across vari-
eties of Latgalian and diachronically more stable than the use of the genitive with
other kinds of noun phrases in negated clauses.
Another pronoun containing the same root (the interrogative/indefinite pro-

noun kas ‘who, what, something, someone’) is vysskas ‘all kinds of things’. It be-
haves like a mass noun and is therefore often used in the genitive. Compare the
following two examples from Ilze Sperga’s short story, where vysskas ‘all kinds of
thing’ contrasts with vyss ‘everything’:
(73) Bab-a vysakuo muok.
Granny-nom.sg all.kinds.gen know.prs.3
‘Granny knows (how to do) all kinds of things.’ (KS)
(74) Pīter-s muok vys-u.
Pīters-nom know.prs.3 everything-acc
‘Pīters knows (how to do) everything.’ (KS)
The genitive of nikas ‘nothing, no one’ and vysskas ‘all kinds of things’ is clearly
connected to broader rules of object marking which have been discussed in the
previous sections. Quite different is the behavior of personal pronouns and proper
names. Variation in object case marking with these two types occurs in different
texts and, curiously, with opposing tendencies.
As shown above, in UP personal pronouns to some extent seem to attract
accusative marking in contexts that usually trigger genitive marking – as objects
of negated transitive clauses, as mentioned in 3.1, or as arguments of intransitive
verbs that will be discussed in Section 5. This tendency has not been observed in
the writings of Ilze Sperga.
In Ilze Sperga’s writing there may be a tendency to use a proper name in the
genitive instead of the accusative. Table 1 above showed that the percentage of
genitive-marked objects in KS is higher with proper names (all in the singular in
this story) than with common nouns in the singular. Here are two examples with
names from different declension classes (both masculine, referring to a man or
boy) where no motivation for the choice of genitive marking could be found:
(75) Nikuo nadūšu! Ni čarkys, ni pusčarkys. Ej iz sātu! – jei bolsā gamanej
i dzan Pedz-is prūm.
and drive.prs.3 Pedze-gen away
‘I won’t give you anything! Not one glass of liquor, not even half a glass. Go
home! – she yells and drives Pedze away.’ (KS)
(76) Īraudziejs Pīter-a, dzed-s ap-sa-ruove i suoce
see.ppa.m.sg Pīters-gen grandpa-nom pfx-rfx-tear.pst.3 and start.pst.3
sviļpuot.
whistle.inf
‘Having spotted Pīters, grandpa stopped and started whistling.’ (KS)
238 Nicole Nau
Both verbs (dzeit pruom ‘to drive away’ and īraudzeit ‘to spot, notice’) usually
govern the accusative and there are no other factors (to my knowledge) that could
have triggered genitive marking here. Note that, unlike in Slavic languages, the
accusative differs from the nominative in all declension classes (see Table 2), so
the genitive cannot be motivated by a need to distinguish subjects from objects.
Nothing of this sort has been found in UP, where proper names are rare anyhow.
There are also individual instances of genitive marking of common nouns in
the singular, especially in UP, for which it is hard to find a motivation. I will dis-
cuss two of them and suggest possible explanations, but these may turn out to be
wrong in the course of further investigation of a larger amount of data.
First, it is possible that a definiteness/indefiniteness distinction is signaled by
case marking. This would be a feasible explanation for the contrast between the
following two sentences. My free translation shows this interpretation:
(77) Jei cīši nūzaprīcuojās, ka pazagrīžkūte iz tāvu atpakaļ i
deve jam sov-u gradzyn-u nu pierst-a.
give.pst.3 3sg.m.dat rpo-acc.sg ring-acc.sg from finger-gen.sg
‘She was very happy that she should return to her parents and gave him the
ring from her finger.’ (UP)
(78) Vot jis i nūvad jū iz ceļa, nūzarunoj jī, ka Juoņc [Juoņs] praseis kieniņa, lai jam
atdūd jū par sīvu, a jei pajem,
paškiņkoj jam gradzyn-a, ka jei jau pi cyt-a
give.prs.3 3sg.m.dat ring-gen that 3sg.f ptc to other-gen
na-īs!
neg-go.fut.3
‘Now he leads her on the way, they agree that John will ask the king to give
her to him for a wife, and she goes and presents him with a ring [and says]
that she won’t marry another man.’ (UP)
There are no more instances of the verb paškiņkuot ‘to give as a present’ in UP. In
modern texts it is found with accusative marking.
The second example involves the noun zyrgs ‘horse’, which appears four times
in the genitive instead of an expected accusative in UP. However, these four in-
stances do not seem to be related and it is most doubtful that animacy could play a
role here (other animate referents are not marked with the genitive as objects, and
in one of the examples the word refers to the meat of a slaughtered horse). One of
these instances is the following:
(79) Pa-jem tu ituo zyrg-a, a

pfx-take.prs.2sg 2sg.nom dem.gen.sg.m horse-gen.sg and
kai tev vaidzēs jū,
if 2sg.dat need.fut.3 3sg.acc
izej uorā, apzagrīz treis reizis iz pādai, aizsviļp, – jis atskrīs!
‘Take this horse, and when you’ll need it, go out, turn around on your heel
three times, whistle and it will come running!’ (UP)
This is evidently not a case of indefiniteness. It is possible (though maybe a bit

far-fetched) to motivate the genitive in this example by the temporary/incomplete
nature of the transaction: though the hero is told to “take” the horse, it obviously
will not stay with him afterwards, for he has to call it each time. However, as this
is the only instance of the verb (pa)jimt ‘take’ with a genitive object in UP, the
explanation is mere speculation.
5. Intransitive verbs with genitive arguments
The constructions discussed in the two previous sections involved a transitive

verb – by definition, a verb which selects a nominative subject and an accusative
object (and possibly another argument) – where the accusative marking of the
object under certain conditions is changed into genitive marking. The focus of the
present section is on verbs that select genitive objects,20 either on a basic lexical
level or as a result of prefixation.
5.1 Lexical semantics of verbs selecting genitive arguments
The first intransitive model to be presented here is NOM–GEN, that is, two-place
predicates that select a nominative subject and a genitive object. Most verbs with
this case frame are also found (as an alternative) with accusative objects. With
some verbs (for example meklēt ‘look for’) the accusative is only an occasional
variant, while with others (for example žāluot ‘love’) genitive and accusative are
equally often found in object marking. Table 5 presents a basic list containing the
most important of these verbs (i.e. those most frequently found, undisputedly
used with genitive objects, non-prefixed).
20. I use the term “object” in the traditional way for non-subject core arguments.
240 Nicole Nau
Table 5. Verbs with nominative subject and genitive object (nom–gen)
Verb Meaning Comment
meklēt ‘look for’
gaideit ‘wait for’ also with prefix: sagaideit ‘await, expect, meet’
gribēt ‘want’
praseit ‘ask (for something)’, in UP also ‘ask someone’ (other texts prefer dative or the
‘demand’ preposition nu ‘from’)
in UP also puorpraseit ‘beg pardon’ (may be a calque
from Polish przepraszać) and aizpraseit ‘invite’ (may be a
calque from Polish zapraszać)
saukt ‘call (for someone)’ not in the meaning ‘call by name’
klauseit ‘obey’
klauseitīs ‘listen to’
vērtīs ‘look at’ also with prefix: apsavērt, pasavērt
žāluot ‘love’
ļūbeit ‘love’, ‘like’
beitīs ‘fear’
dabuot ‘get’
raudzeit ‘try’
There are some other candidates for this list, but my data are too scarce for
me to be sure about their case frame.21 The verbs in this list belong to differ-
ent semantic groups: pursuit (‘search’, ‘wait’), verbal interaction (‘demand’, ‘call’,
‘invite’, ‘obey’), perception (‘listen’, ‘look’), emotion (‘love’, ‘fear’), and other (‘get’,
‘try’). They have in common a medium to low degree of semantic transitivity:
the situations they denote differ from prototypical transitivity in some but not all
respects. On Tsunoda’s scale (Tsunoda1985; see also Malchukov 2005 for a fur-
ther development) they occupy a middle position, from 2.b (perception, patient
less attained) to 5 (feeling).22 Næss (2007) provides a simple but effective way of
describing prototypical and non-prototypical transitivity, using the three features
volitionality (vol), instigation (inst) and affectedness (aff). A prototypical agent
acts volitionally, is the instigator of the event and not affected by it. A prototypical
patient is affected by the event and shows neither volitionality nor instigation.
The two core participants of an event such as KILL can thus be modeled as: A =
21. For example, in UP there is just one instance of puorspēlēt ‘lose’, where it takes a genitive ob-
ject (‘he lost the war’). This verb was found a few times in modern sources, where it combined
with the accusative, but it is rare anyhow (the meaning ‘lose’ is more often expressed by zaudēt,
with accusative). The verb puor-spēlēt ‘lose’ may be a calque from Polish prze-grać.
22. Verbs of Tsunoda’s class 4 (‘know’, ‘remember’, ‘forget’) are used with the transitive frame in
Latgalian.
[+vol, +inst, –aff], P = [–vol, –inst, +aff]. Applying this framework to the
verbs in Table 5 one finds that most of them have a volitional agent [+vol], a
feature that may account for their nominative subjects, which they share with
transitive verbs. The patient, on the other hand, is [–aff], or only weakly affected,
which accounts for the non-accusative marking of the object. Most of the Stan-
dard Latvian equivalents of the Latgalian verbs in Table 5 are transitive verbs. The
Latvian verb klausīt ‘obey’ governs either the accusative or the dative (accusative
is slightly more frequent in the corpus miljons-2.0) and is sometimes even found
with a locative (klausīt vecākos ‘obey/listen to one’s parents’ found in the corpus
miljons-2.0).
The verb beitīs ‘fear’ differs from the other verbs. The genitive of the object
historically goes back to an ablative and the verb shows no tendency to become
transitive. The equivalents in Modern Standard Latvian, bīties and its more fre-
quent synonym baidīties ‘fear’, are never used with an accusative. They select either
a genitive or, more commonly today, a prepositional object with the preposition
no ‘of, from’. In modern Latgalian texts such as those of the corpus MuLa-1.0 a
combination with the preposition nu is frequent as well. In UP the verb beitīs
regularly combines with a genitive object, for example:
(80) “Vylk-a mait-a, kuo tu beistīs?”
wolf-gen.sg carrion-nom.sg wh.gen 2sg.nom fear.prs.2sg
“Beistūs Juoņ-a Palnurušk-ys, jis te iraida!”
fear.prs.1sg Juoņs-gen Palnuruška-gen 3sg.m.nom here be.prs.3
‘Wolf ’s carrion, what are you afraid of? – I am afraid of John-from-the-Ashes,
he is here!’ (UP)
However, I found one example with the second person singular form tevi which I
interpret as accusative (see Section 2):
(81) Nu, jis tevi beisīs tagad, na tu juo!
ptc 3sg.m.nom 2sg.acc fear.fut.3 now neg 2sg.nom 3sg.m.gen
‘Well, now he will fear you, not you him!” (UP)
This example may just show the general tendency to use the accusative form tevi
instead of the genitive. Note that the third person pronoun in ‘you (will) not (fear)
him’ is in the genitive.
The pioneering role of pronouns is more evident with other verbs. With me-
klēt ‘search’, there are 18 instances of genitive marking and 3 instances of accusa-
tive marking in UP, the latter all with a personal pronoun (one time first person
singular and two times third person singular; see examples (3) and (4) above).
With the verb gaidēt ‘wait’, the proportion is 16:5; again, in all 5 instances of
242 Nicole Nau
accusative marking the object is a personal pronoun (first, second, and third per-
son singular).
Three of the verbs are borrowings from an East Slavic language. The verb
dabuot ‘get’ is an old loan that is also found in Latvian (dabūt), where it mostly
governs the accusative (in Modern Standard Latvian always). There has been a
gradual shift from genitive to accusative marking which is somewhat slower in
Latgalian than in other dialects of the Latvian language. In Latgalian, genitive
objects are still more common with this verb, but accusative marking appears
already in UP and is rather frequent in contemporary texts. Examples with gen-
itive marking are (82) from UP and (84) from KS, wile (83) shows accusative
marking.23
(82) kienin-š ruod-s, ka dabuoja taid-a
king-nom.sg glad-nom.sg.m that get.pst.3 such-gen.sg.m
boguot-a znūt-a
rich-gen.sg.m son.in.law-gen.sg
‘The king was glad that he got such a rich son-in-law’ (UP)
(83) vidišk-ys nū-guoja iz dzeraun-i i dabuoja
middle[adj]-nom.sg.m pfx-go.pst.3 to village-acc.sg and get.pst.3
tāva-meit-u
father.gen-daugther-acc.sg
‘the middle (brother) went to the village and got a farmer’s daughter’ (UP)
(84) Kod bruoļ-s pa-beigs vuiceitīs i dabuos
when brother-nom.sg pfx-finish.fut.3 learn.inf get.fut.3
lob-a dorb-a, jis Pīter-i jims da sev-e.
good-gen.sg.m job-gen.sg he Pīters-acc take.fut.3 to refl-gen
‘When his brother has finished his education and gets a good job, he will take
Pīters to live with him’ (KS)
The other two Slavic loans, žāluot and ļūbeit ‘love, like’ are used in UP with both
the genitive and the accusative (examples (85) and (86)). In modern texts they
appear mainly with the accusative; ļūbeit appears a few times in MuLa-1.0 with an
object in the genitive, while žāluot doesn’t appear at all in the meanng ‘love’, only
in the meaning ‘regret’.
23. Regarding the contrast between (82) and (83), Ilja Seržant suggested that gender may play
a role for the case selection, but no such correlation was found.
(85) jam beja meit-a, kūm-am, i

3sg.m.dat be.pst.3 daughter-nom.sg godfather-dat.sg and
tei meit-a tuo puiškin-a žāluoja
dem.nom.sg.f daughter-nom.sg dem.gen.sg.m boy-gen.sg love.pst.3
‘the godfather had a daughter, and this daughter loved that boy’ (UP)
(86) Kienineit-e žāloj cīši sov-u veir-u
princess-nom.sg love:prs.3 very rpo-acc.sg husband-acc.sg
‘The princess loved her husband very much’ (UP)
So far, no semantic or stylistic difference has been observed between clauses with
accusative marking and clauses with genitive marking. Possibly the investigation
of a larger corpus and discussions with native speakers may bring to light some
factors that favor one or the other case. For the time being, I can only state that
the verbs of Table 5 (except for beitīs ‘fear’) show a tendency to change from an
intransitive model to the transitive model of case-marking.
A property connected to this change is the possibility of undergoing valency
change in two constructions that affect the marking of core arguments. These are,
in Latgalian as in Latvian, the passive and the debitive, a form expressing neces-
sity. In the passive, accusative objects are promoted to nominative subjects, in the
debitive accusative objects (except for the pronouns for first and second person
singular and the reflexive pronoun) become nominative marked arguments with
partial subject properties (see Holvoet & Grzybowska, this volume, for Latvian).
In the texts that were examined closely for this paper, none of the verbs of Table 5
was used in a passive form – there are no constructions like ?tyka/beja maklāts
‘was searched for’, ?tyka/beja dabuots ‘was gotten’. In a construction with the debi-
tive, a genitive object selected by the verb remains in the genitive, while an accu-
sative object usually changes to nominative (though not always, accusative is also
found). Compare škola (nominative) juoveic ‘has to finish school’ versus dorba
(genitive) juodaboj ‘has to get a job’ in the following example:
(87) Jurs to, skait, jau doktur-s, tik ka škol-a
Jurs-nom ptc ptc already doctor-nom.sg ptc that school-nom.sg
juo-veic i lob-uok-a dorb-a juo-daboj.
deb-do and good-comp-gen.sg.m work-gen.sg deb-get
‘Look, Jurs is a doctor already, he just has to finish school and to get a better
job.’ (KS)
An example for the debitive with a genitive object from UP is the following:
244 Nicole Nau
(88) maņ jau beja saiminīk-a juo-meklej

1sg.dat ptc be.pst.3 master-gen deb-search
‘after all I had to look for a man for the household’ (UP; context: a wife
explains to her husband why she was getting married after several years of his
absence)
In other modern texts, however, where the verbs of Table 5 ever more often are
used as transitive verbs, we find the debitive construction with a nominative ar-
gument, as in the following examples from the corpus of contemporary Latgalian
texts:
(89) juo-daboj naud-a jaun-im dorb-im
deb-get money-nom new-dat.pl work-dat.pl
‘one has to get money for new tasks’ (MuLa-1.0)
(90) poš-om jō-meklej material-i
self-dat.sg deb-search material-nom.pl
‘one has to look for materials oneself ’ (MuLa-1.0)
(91) i tod juo-gaida viļcīn-s
ptc then deb-wait train-nom.sg
‘and then one has to wait for the train’ (MuLa-1.0)
Passive constructions are also found in the corpus, as well as in Latgalian texts on
the Internet. They appear most often in modern texts that show a considerable
influence from Latvian, for example:
(92) tod tyka lyk-t-ys golv-ys kūpā i tyka
then aux.pst.3 put-ppp-nom.pl.f head-nom.pl together and aux.pst.3
maklā-t-s īspiejami lob-uok-ais variant-s.
search-ppp-sg.m possible-adv good-comp-nom.sg.m.def variant.nom.sg
literally: ‘then heads were put together and the best variant possible was searched
for’ (http://lakuga.lv/2011/01/19/latgolys-radeja-i-latgaliskuo-informaceja/)
(93) tyka pīveik-t-i vys-i konkurent-i i
aux.pst.3 beat-ppp-nom.pl.m all-nom.pl.m competitor-nom.pl and
dabuo-t-a gūd-pyln-uo 1. vīt-a
get-ppp-nom.sg.f honor-full-nom.sg.f.def 1st place-nom.sg
‘all competitors were outperformed and the honorable first place was gotten’
(MuLa-1.0)
An example for dabuots ‘gotten’ was found also in a traditional fairy tale in a
southern Latgalian subdialect (from the collection by Šmits, where Latgalian texts
unfortunately have been edited in an unsystematic way):
(94) Vīn-am ķēniņ-am beja dabuo-t-s nu cyt-as

one-dat.sg.m king-dat.sg be.pst.3 get-ppp-sg.m from other-gen.sg.f
kēnests gudr-s un muocī-t-s kač-s
kingdom.gen.sg clever-nom.sg.m and teach-ppp-nom.sg.m cat-nom.sg
‘A king had got a clever and well trained cat from another kingdom.’ (liter-
ally: “to a king had been gotten a cat”) (fairy tale Velni mānās, from Dagda,
http://valoda.ailab.lv/folklora/pasakas/gr14/14A0517.htm)
In my eyes these examples do not count as evidence for the ability of genitive ob-
jects to become nominative subjects in a passive construction. Rather they prove
that in varieties of Latgalian where they occur the respective verbs are used with
the transitive pattern. However, further research is needed to find out what comes
first: an accusative in the active voice or the ability to undergo valency change in
the passive construction.
Another intransitive pattern of two-place predicates with a genitive is dat–
gen. The dative argument expresses a non-volitional affected experiencer. The
frame is found with a couple of verbs and two predicative adverbs. In Tsunoda’s
(1985) system, they are further down the scale: they belong to group 5 (feeling:
‘desire’, ‘need’, ‘sorry’, ‘afraid’) and 6 (relationship: ‘lack’, ‘have enough’). In contrast
to the verbs of Table 5, there is no tendency to use these predicates according to
the transitive model, neither in Latgalian nor in Latvian.
Examples:
(95) Dīv-am tože lob-u cylvāk-u vajag.
God-dat also good-gen.pl person-gen.pl need.prs.3
‘God also is in need of good people’ (KS)
(96) At-soka zyrg-s: “Maņ bais Juoņ-a Palnurušk-ys!”
pfx-say.prs.3 horse-nom.sg 1sg.dat afraid Juoņs-gen Palnuruška-gen
‘The horse replied: “I am afraid of John-from-the-Ashes!”’
(97) Pīter-am žāl dzed-a ar bab-u.
Pīters-dat sorry grandpa-gen with grandma-acc
‘Pīters feels sorry for his grandparents’ (KS)
Table 6. Predicates with a dative and a genitive argument (dat–gen)

Predicate Meaning Comment
gribētīs ‘desire’ also: aizagribēt
vajadzēt ‘need’ also: aizvajadzēt
tryukt ‘lack’
pītikt ‘be/have enough’
žāl ‘sorry for’
bais ‘afraid of ’
246 Nicole Nau
This small group of predicates is semantically and syntactically quite heteroge-

neous, but they also have several features in common. The dative argument is
often not expressed and it is possible to treat these predicates as having only one
core argument (with the dative as an optional peripheral extension). Except for
tryukt ‘lack’, they often combine with an infinitive clause instead of an NP. An
object of an infinitive depending on žāl/žāļ ‘regret, be sorry (to do)’ may also be
in the genitive:
(98) Jam žāļ, tam tāv-am, puordūt dāl-a
3sg.m.dat sorry dem.dat.sg.m father-dat.sg sell.inf son-gen
‘He is sorry, the father, to sell his son’ (UP)
The predicates bais ‘afraid’ and pīteikt ‘(have) enough’ also combine with a prepo-
sitional phrase (bais nu ‘afraid of ’, pīteik ar ‘enough with’). Nominal objects have
thus a lower token frequency with these predicates than with those of Table 5.
Nevertheless, genitive marking is more stable even in contemporary texts. Only
with vajadzēt several instances of accusative marking were found; for an example
see the second line of (79) above with the third person pronoun jū in the ac-
cusative. Note that accusative marking is the rule in Standard Latvian (see also
Holvoet & Grzybowska, this volume). This does not make vajadzēt a transitive
verb, however, as it lacks a nominative subject.
5.2 Intransitivation by derivation
The interplay of prefixation, or prefixation together with reflexivisation, and ob-

ject marking needs further investigation. In the literature on Latgalian, there are
only vague hints as the following by Bukšs:
Nasalyktais verbs dzert prasa objektu akuzativā, pīm. dzert yudini, bet salyktais –
objektu genitivā: pasadzert yudiņa. (Bukšs & Plačinskis 1973: 296)24
The simple verb dzert ‘drink’ governs an object in the accusative, for example
dzert yudini ‘to drink water’, but the derived (literally “compound”) verb governs
the genitive: pasadzert yudiņa. [pa-sa-dzert: pfx-rfx-drink ‘drink one’s fill’]
In UP the verb pasadzert appears 15 times, but it is never combined with an

object.
24. Note that in this quote, in contrast to Strods (1933) cited above, the accusative is assumed
to be the unmarked case for the mass noun iudiņs ‘water’.
One of the productive devices that clearly trigger genitive marking of objects
is the cumulative prefix pī-.25 It indicates that a large amount of something has
been affected or come into existence as the result of the action denoted by the
verb (cf. Lindāne 1986 for a Selonian (non-Latgalian) subdialect of High Latvian;
a Lithuanian equivalent is pri-; Filip 2008 analyses the interplay of a cumulative
prefix with aspect and eventuality type in Slavic languages):
(99) piec razboinīk-im palyka daudz šmuk-u drēb-u,
after robber-dat.pl remain.pst.3 a.lot pretty-gen.pl cloth-gen.pl
kur-u jī beja pī-zog-uš-i
rel-gen.pl 3pl.nom be.pst.3 pfx-steal-ppa-pl.m
‘The robbers had left behind a lot of pretty clothes which [genitive] they had
amassed by stealing’ (UP)
(100) Pavasar-ūs jis pī-taisēja vysaid-u
spring-loc.pl 3sg.m.nom pfx-make.pst.3 all.kind.of-gen.pl
sviļp-u
whistle-gen.pl
‘In spring time he made all kinds of whistles’ (KS)
Naturally, nouns in the plural or mass nouns are most often found as objects of a
verb with the cumulative prefix pī-. The following example however shows that it
may also combine with a singular count noun. Note that pīdzemdēt is not neces-
sarily cumulative and may also be used as a transitive verb. The use of the genitive
(in my interpretation) implies that the new child will be an addition to an already
existing large group of children:
(101) Pī-dzemdēs bārn-a, puor-dzers naud-u.
pfx-give.birth.fut.3 child-gen.sg pfx-drink.fut.3 money-acc.sg
‘She’ll give birth to yet another child, she’ll waste the money on drink’ (KS)
While verbs with the prefix pī- often have a genitive object, verbs with the prefixes
sa- or puor- usually have an accusative object, even if it is a mass noun. Compare
the two following examples from a fairy tale from UP, which denote the same situ-
ation in two variants (as in fairy tales many things have to happen three times):
25. The prefix pī- has also some other meanings, and not all verbs with pī- are intransitive.
248 Nicole Nau
(102) Nūskrēja iz teiruma, nūzavēra kur ļauds struoduoja,

sa-jēme nu jūs maiz-i, bīz-u-pīn-u,
pfx-take.pst.3 from 3pl.gen bread-acc thick-acc.sg-milk-acc.sg
nū-nese tam puišk-in-am
pfx-carry.pst.3 dem.m.sg.dat boy-dim-dat.sg
‘[The fox] ran to the field, observed where people were at work, took from
them (their) bread, cottage cheese (acc), brought it to the boy.’ (UP)
(103) a lops-a nū-skrēja iz teirum-a, pī-zoga
ptc fox-nom.sg pfx-run.pst.3 to field-gen.sg pfx-steal.pst.3
gal-is, maiz-is, bīz-a pīn-a deļ puišk-in-a
meat-gen bread-gen thick-gen.sg.m milk-gen for boy-dim-gen.sg
‘but the fox ran to the field, stole meat, bread, cottage cheese (gen) for the
boy’ (UP)
In (103) the cumulative prefix and genitive marking of the object highlight the
fact that several items were stolen (gathered by stealing), while the description in
(102) just reports the act of taking the items into possession. The accusative mark-
ing in (102) may also imply that all the available bread and cottage cheese were
taken (indicated in my translation by ‘their’).
The genitive object of a verb with the cumulative prefix does not change in the
debitive construction:
(104) iz zīm-ys juo-pī-pļaun sīn-a
to winter-gen.sg deb-pfx-mow hay-gen.sg
‘(a large amount of) hay has to be mowed for winter’ (KS)
In contrast to intransitive verbs such as meklēt ‘search’ or dabuot ‘get’, cumulative

verbs also appear in the passive in traditional texts from (Central) Latgalian as
well as in the writing of Ilze Sperga. The genitive of the object is retained, and the
verb is marked by the default features masculine singular:
(105) a beja jam maiz-is i gal-is pī-gatavē-t-s
and be.pst.3 3sg.m.dat bread-gen and meat-gen pfx-cook-ppp-m.sg
deļ septiņ-u god-u
for seven-gen year-gen.pl
‘and they had bread and meat prepared (“amassed by cooking”) for seven
years’ (UP)
6. Conclusions
The main aim of this paper was to present the regularities of differential object
marking in Latgalian. The main sources were texts by one contemporary writer
and a collection of orally transmitted fairy tales from the late 19th century. The
chosen sources represent syntactically conservative varieties where on the whole
genitive objects are more frequent then in most contemporary written texts. In
the analysis, a principled distinction was made between transitive verbs, where
accusative marking is the unmarked situation that under certain circumstances is
replaced by genitive marking, and intransitive verbs governing a genitive object.
The term “differential object marking” should probably be used only in the first
situation, while “variation in object marking” may be a suitable cover term for
both. Verbal prefixation, which is at the border of grammar and lexicon, provides
a connection between the two groups. Table 7 gives a summary of types of differ-
ential object marking discussed in this paper, including prefixation. All examples,
except for the ones marked “UP”, are taken from Ilze Sperga’s short story (KS).
For examples that have not been presented before morphological glossing is given
below.
(106) Jei pat moz-ūs sīr-eņ-us taisa
she even small-acc.pl.m.def cheese-dim-acc.pl make.prs.3
‘she even makes small cheese’ (sweet curd) (KS)
(107) Dzed-s to pat-s lizeik-u na-taisa
grandpa-nom ptc self-nom.sg.m spoon-gen.pl neg-make.prs.3
‘Grandpa himself doesn’t make spoons’ (KS)
(108) Dīv-eņ-š redz vys-u
god-dim-nom.sg see.prs.3 all-acc.sg
‘Our Lord sees everything’ (KS)
(109) Vys-a jis na-redz vys
all-gen.sg he neg-see.prs.3 ptc
‘He doesn’t see everything, not at all’ (KS)
(110) dūs čark-u
give.fut.3 glass.of.liquor-acc.sg
‘(she) will give (him) a glass of liquor’ (KS)
(111) Jis muok pļau-t zuol-i
he know.prs.3 mow-inf grass-acc.sg
‘He knows how to mow (the) grass’ (KS)
250 Nicole Nau
Table 7. Differential object marking in Latgalian (summary)

Accusative Genitive
same verb, affirmative negative
same kind Jei pat mozūs sīreņus taisa (106) Dzeds to pats lizeiku nataisa (107)
of object ‘she even makes small cheeses’ ‘Grandpa doesn’t make spoons’
Dīveņš redz vysu (108) Vysa jis naredz vys (109)
‘God sees everything’ ‘He doesn’t see everything’
realis irrealis
dūs čarku (110) kas začam slidu dūs (41)
‘will give [him] a glass (of liquor)’ ‘who will give skates to a hare’
infinitive supine
Jis muok pļaut zuoli (111) guojuši sīna pļautu (112)
‘he knows how to mow grass’ ‘(if) they had gone to mow hay’
same verb, count noun mass noun
different kind (UP) īdeve jam gruomotu (113) (UP) īdeve maizis (114)
of object ‘gave him a book’ ‘gave (him) bread’
Pīters muok vysu (74) baba vysakuo muok (73)
‘Pīters knows everything’ ‘Granny knows all kinds of things’
lexically related mass noun limited mass noun unlimited
verb, same kind dzers sovu kopeju (72) īdzers kopejis (71)
of object ‘will drink his coffee’ ‘will drink (some) coffee’
unmarked cumulative
Jei pat mozūs sīreņus taisa (106) jis pītaisēja vysaidu sviļpu (100)
‘she even makes small cheese’ ‘he made (lots of) different whistles’
(112) guoj-uš-i sīn-a pļau-tu

go-ppa-pl.m hay-gen.sg mow-sup
‘(if) they had gone to mow (the) hay’ (KS)
(113) ī-deve jam gruomot-u
pfx-give he.dat book-acc.sg
‘gave him a book’ (UP)
(114) ī-deve jam maiz-is
pfx-give he.dat bread-gen.sg
‘gave him bread’ (UP)
In the first group distinguished in Table 7 (“same verb, same kind of object”),
grammatical or pragmatic (stylistic) factors trigger the use of the genitive for ob-
jects of transitive verbs regardless of the lexical material. Semantic properties of
the verb or the object seem to play no role here. The first and the third case are the
most regular and allow us to formulate general rules:
i. In negated clauses the object of a transitive verb is in the genitive.

ii. In constructions with the supine the object of a transitive verb is in the
genitive.
The use of the genitive in clauses presenting a situation as unreal, on the other
hand, is optional. It was found most regularly in the prose of Ilze Sperga, who
uses the genitive as a stylistic device especially in inner monologues of a hero or
the narrator. Further research is needed to determine how widespread this use is
among speakers and writers of Latgalian. For the time being, I propose the fol-
lowing rule:
In clauses that present a situation as unreal, the genitive may be used for
the object of a transitive verb to emphasize the speaker’s subjective point of
view.
The next group comprises instances where genitive marking is connected to prop-
erties of the object noun phrase. In Table 7 I included only the (seemingly) un-
disputed instance of mass nouns and the semantically related pronoun vysskas ‘all
kinds of things’. In Section 4.1 I quoted Strods (1933) who presented the genitive
marking of mass nouns as objects of transitive verbs as the unmarked type. How-
ever, in the texts under investigation the situation is less clear. In the fairy tales
(UP), Strods’ rule holds in many, but not all instances: mass nouns as objects do
not always appear with genitive marking where expected. An important, but still
not sufficient, modification of the rule is that in clauses that present a habitual
situation the accusative is used with mass nouns. In the writings of Ilze Sperga,
genitive marking of mass nouns seems to require verbal prefixes, a situation simi-
lar to that in Russian (though one has to keep in mind that the effects of prefixing
in Latgalian and Russian are not the same). More detailed research is needed here,
involving a much larger body of sources and discussion of (real and constructed)
examples with native speakers. For the time being I can only assume that there is
no simple rule such as “mass nouns are marked genitive” in contemporary Lat-
galian. I suppose that there is a tendency to couple the genitive marking of noun
phrases designating masses or collections (the latter expressed by count nouns in
the plural) with prefixing of the verb, which in turn may have different functions.
Thus, it is possible that the group in the middle of the table (“same verb, different
kind of object”) will merge with the third group (“lexically related verb, same or
different kind of object”). As a next step, the genitive marking will become redun-
dant – as the relevant semantic features are expressed by the verb – and may be
given up. Further investigation of contemporary texts may show whether this is
what happens in modern Latgalian.
252 Nicole Nau
There is only a relatively small group of non-prefixed verbs in Latgalian that

govern a genitive (and are thus by my definition intransitive verbs). In Section 5.1
I listed twelve simple verbs that appear in my sources with a nominative subject
and a genitive object. With the exception of beistīs ‘fear’, these verbs are alterna-
tively used as transitive verbs even in UP, the most conservative of my sources,
and many contemporary writers regularly use them with an accusative object
(with Ilze Sperga’s texts standing out by retaining the genitive more often). I have
not been able to find characteristics that distinguish the verbs of this group from
other verbs that have only the transitive frame – the lower degree of semantic
transitivity may be a necessary condition, but it is not a sufficient one. It is possi-
ble that these verbs are the remnants of a formerly more numerous group of verbs
with genitive objects. Some other verbs occur occasionally with this case frame
in UP but not in other sources I have seen so far, for example sorguot ‘guard’ or
syuteit ‘send’. From a synchronic point of view the genitive object is just an idio-
syncratic behavior of certain lexemes. This may be why it is so easily given up.
With verbs and non-verbal predicates that govern a genitive and a dative, but do
not have a nominative subject, the genitive seems to be more stable. This group is
still smaller, it comprises only half a dozen lexemes (see Section 5.1).
While some of the verbs whose genitive object is increasingly replaced by an
accusative object are intensional, others are not, and not all intensional verbs are
intransitive. Latgalian therefore does not provide material to prove or disprove
Kagan’s thesis that genitive marking with intensional verbs correlates with the use
of the genitive in negated clauses, while neither of these uses correlates with the
use of the partitive genitive (Kagan 2010; 2013). Nevertheless, the Latgalian data
may give some support: It seems to be true (further research is needed) that vari-
eties (such that those of UP and KS) that use the genitive of negation as a general
rule also more often use the genitive governed by verbs such as meklēt ‘search’,
gaideit ‘wait’ and dabuot ‘get’, while the decline of the partitive genitive with mass
nouns as objects of simple (non-prefixed) verbs proceeds independently. Another
observation is that the consistent use of the genitive of negation in Ilze Sperga’s
writings correlates with, and is probably related to, her use of the genitive in sty-
listically marked clauses with irrealis mode.
Differential object marking in Latgalian is itself subject to variation: different
tendencies are found in texts from different times, by different authors, or of dif-
ferent genres. The multifaceted nature makes its description even more difficult
and calls for further studies. For example, as mentioned several times in this pa-
per, rule (i) formulated above is clearly present in the writings of Ilze Sperga, but
not observed in many other contemporary texts where the accusative is used with
transitive verbs under negation. On the other hand, some tendencies in genitive-
marking are found only in certain sources, for example those relating to prop-
erties of the NP discussed in Section 4.2. The use of the accusative of personal
pronouns instead of an expected genitive in UP may be due to an influence of
southern Latgalian dialects that is absent in the language of Ilze Sperga. In con-
trast, Ilze Sperga’s use of the genitive instead of the accusative with (male) proper
names may be an idiosyncratic phenomenon.
With respect to object marking, as in other areas of grammar, Latgalian
proves to be somewhere in between Lithuanian, where genitive marking of ob-
jects is more common and stable, and Latvian, where it has been given up almost
completely. Latgalian syntax has barely been subjected to standardization, and a
great amount of variation is found in texts from the past 150 years. This variation
poses a challenge for a coherent synchronic description, which was the aim of
the current investigation. On the other hand, it opens possibilities for diachronic
research on the dynamics of change in object marking, for example regarding
the decline of the genitive of negation, or new rules for the use of the partitive
genitive.
Abbreviations

comp comparative pfa future active participle
cond conditional pfx prefix
dat dative pl plural
deb debitive ppa past active participle
def definite (ending) ppp past passive participle
dem demonstrative pronoun prs present tense
dim diminutive pst past
f feminine ptc particle
fut future q question particle
gen genitive refl reflexive
inf infinitive REFL.POSS reflexive possessive pronoun
loc locative sg singular
log logophoric pronoun sup supine
m masculine wh who/what pronoun
neg negation
254 Nicole Nau
Sources
Ādama stāsts. Mazsalaciešu dzīve, ieradumi un tikumi Ā. Purmaļa autobiogrāfijā 19. un 20. gs.
mijā. 2008. Ed. Sanita Reinsone. Rīga: Zinātne.
IS = texts from Ilze Sperga’s blog at www.naktineica.lv, also available at the portal www.lakuga.
lv.
KS = Kur sauleitei sāta, short story by Ilze Sperga published 2008 on her former web-site at
http://saprge.wordpress.com
Mar = Marabeju kolns, short story by Ilze Sperga, published 2011 on her web-site at http://www.
naktineica.lv/?p=5490.
Miljons-1.0 = corpus of contemporary Latvian, 1 million words. Access through www.korpuss.
lv. [accessed in August 2013]
Miljons-2.0 = corpus of contemporary Latvian, 3.5 million words. Access through www.
korpuss.lv. [accessed April-August 2013]]
MuLa-1.0 = corpus of contemporary written Latgalian, 1 million words. Access through www.
korpuss.lv. [accessed April-August 2013]
UP = Ulanowska, Stefania. 1895. Łotysze Inflant polskich, a w szczególności gminy Wielońskiej,
powiatu Rzeżyckiego. Obraz etnograficzny. Część III. Zbiór wiadomości do antropologii
krajowej, t. XVIII, 232–406.
VL = texts written by Valentins Lukaševičs, published in a Latvian newspaper (Latvijas Avīze)
and on the Internet, available at the portal www.lakuga.lv.
References
Bossong, Georg. 1991. Differential object marking in Romance and beyond. In New analyses
in Romance linguistics, Dieter Wanner & Douglas A. Kibbee (eds), 143–170. Amsterdam/
Bukšs, Miķelis & Placinskis, Jurs. 1973. Latgaļu volūdas gramatika un pareizraksteibas vōrdneica
[Latgalian Grammar and orthographic dictionary]. 2nd edn. München: Latgalischer
Verlag.
Cibuļs, Jurs & Leikuma, Lidija. 2003. Vasals! Latgaliešu valodas mācība. Rīga: n.i.m.s.
de Hoop, Helen & Malchukov, Andrej L. 2008. Case marking strategies. Linguistic Inquiry 39:
565–587. DOI: 10.1162/ling.2008.39.4.565
Endzelin, Jan. 1922. Lettische Grammatik. Riga: Kommisionsverlag A. Gulbis.
Filip, Hana. 2005. On accumulating and having it all. Perfectivity, prefixes and bare arguments.
In Perspectives on aspect, Henk J. Verkuyl, Henriëtte de Swart, & Angeliek van Hout (eds),
125–148. Dordrecht: Springer. DOI: 10.1007/1-4020-3232-3_7
Gāters, Alfrēds. 1977. Die lettische Sprache und ihre Dialekte. The Hague: Mouton.
In Non-canonical marking of subjects and objects [Typological Studies in Languages 46],
Alexandra Aikhenvald, R. M. W. Dixon, & Masayuki Onishi (eds), 56–83. Amsterdam/
Hopper, Paul J. & Thompson, Sandra A. 1980. Transitivity in grammar and discourse. Language
56: 251–299.
Kagan, Olga. 2010. Genitive objects, existence and individuation. Russian Linguistics 34: 17–39.
DOI: 10.1007/s11185-009-9051-x
Kagan, Olga. 2013. Semantics of genitive objects in Russian. Dordrecht: Springer. DOI:
10.1007/978-94-007-5225-2
Leikuma, Lidija. 2010. Par dažiem Latgales latviešu folkloras pieraksta jautājumiem: Lettische
Volkslieder (1869). Baltu Filoloģija 19: 53–70.
Lindāne, Irma. 1986. Dignājas izloksne. Rīga: Zinātne.
Malchukov, Andrej L. 2005. Case pattern splits, verb types and construction competition. In
Competition and variation in natural languages: The case for case, Mengistu Amber &
Helen de Hoop (eds), 73–117. London: Elsevier. DOI: 10.1016/B978-008044651-6/
50006-9
Nau, Nicole. 2011a. A short grammar of Latgalian. München: Lincom Europa.
Nau, Nicole. 2011b. Declension classes in Latvian and Latgalian: Morphomics vs. Morphopho-
nology. Baltic Linguistics 2: 141–177.
Næss, Åshild. 2007. Prototypical transitivity [Typological Studies in Languages 72]. Amster-
dam/Philadelphia: John Benjamins. DOI: 10.1075/tsl.72
Noonan, Michael. 2007. Complementation. In Language typology and syntactic description.
Volume II: Complex constructions, Timothy Shopen (ed), 52–150. Cambridge: Cambridge
University Press.
Onishi, Masayuki. 2001. Introduction: Non-canonically marked subjects and objects: Param-
eters and Properties. In Non-canonical marking of subjects and objects [Typological Studies
in Languages 46], Alexandra Aikhenvald, R. M. W. Dixon, & Masayuki Onishi (eds), 1–51.
Sinnemäki, Kaius. 2012ms. A typological perspective to differential object marking. Manu-
script, version 24 Aug. 2012. <http://people.uta.fi/~kaius.sinnemaki/dom_web.pdf> (02
March 2013).
Strods, Pīters. 1933. Pareizraksteibas vōrdneica. Rēzekne [published by the author].
Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21: 385–396. DOI:
10.1017/S0022226700010318
Ūsele, Veneranda. 1998. Tilžas izloksnes apraksts. Rīga: Latviešu valodas institūts.
The independent partitive
genitive in Lithuanian
Ilja A. Seržant
University of Konstanz / Vilnius University
The aim of the paper is to give a semantic description of the independent or

bare partitive genitive (IPG) in Lithuanian in rather neutral, functional terms.
The IPG is a multi-faceted category that bears on the domains of quantifica-
tion and (in)definiteness. On its quantificational reading, the IPG encodes an
implicit quantifier, arbitrary in its value. I have used the notion of (un)bound-
edness (re-)introduced in Paul Kiparsky’s (1998) seminal paper on the partitive
case in Finnish. NP-internally, the IPG has two main readings: unbounded and
bounded reading. The first reading provides the concept of the participant rather
than ‘zooming in’ on particular instantiations. It is extremely weak referentially,
probably the weakest option available in Lithuanian. This reading is restricted to
those verbs in Lithuanian that allow their arguments to be kind-referring NPs
(e.g., the subject of the existential to be, or object of to know). On the bounded
reading, in turn, the IPG encodes an undetermined but delimited set, the read-
ing is existential and resembles indefinite plurals. The individuals introduced
by this reading are stored in the discourse model and may be picked up by
anaphoric pronouns in the following discourse. They never constitute prima-
ry or foregrounded information of the message, though. Furthermore, I have
claimed that the incremental-theme verbs and verbs of transfer in East Lithua-
nian interact with the IPG-marked object with respect to their aspectual prop-
erties. Here only the bounded reading of the IPG is available. This explains the
ban on the occurrence of IPG in imperfective contexts in Lithuanian (such as
progressive, generic and iterated atelic contexts) with incremental-theme verbs,
because the imperfective interpretation induces an inherently unbounded event
which is not compatible with the bounded reading of the IPG. Both bounded
and unbounded values are assumed to be originally two different readings of
the same implicit quantifier that have, however, acquired different distributions
in the course of time.
258 Ilja A. Seržant
1. Introduction1
The present paper is devoted to the independent partitive genitive in Lithuanian,

cf. the genitive argument in (1) below:
(1) Nusipirkau pien-o
buy.pst.1sg milk-gen.sg
‘I bought (some) milk.’
The occurrence of this kind of genitive is not directly licensed by some overt head,
hence it will be referred to as independent partitive genitive here (abbreviated:
IPG). Instead, different kinds of semantic considerations are decisive for an NP to
be marked with the IPG, when the latter overrides the NP’s structural case mark-
ing; these range from referential and quantificational properties of the respective
NP’s referent to the referential and quantificational properties of the predicate.
Generally, Lithuanian does not have morphological or other grammaticalized
means to mark (in)definiteness of an NP. Bare NPs freely allow definite inter-
pretation in this language. This is, however, different with the bare NPs that are
case-marked by the IPG – these NPs can only have indefinite interpretation with
no presupposition.2 The fact that proper names are excluded from occurring in
the IPG in Lithuanian is telling in this respect. Equally telling is the observation
that abstract nouns are especially likely to have the IPG marking if occurring in
the object position. In many instances such abstract-noun objects remind one
of objects of light verbs in other languages, cf. such frequent expressions as, e.g.,
turėti kantrybės [have patience.gen] ‘to be patient’. The verb turėti is a sort of light
verb here lacking its original lexical meaning of possession, while the abstract
noun kantrybės ‘of patience’ is not individualized, providing only the concept.
The inherent indefiniteness of the bare IPG case-marked NPs in contrast to other
1. My thanks go, first of all, to Kristina Lenartaitė-Gotaučienė, whose contribution was es-
sential. Without her this paper would never have been possible. The paper profited much from
the comments of my colleagues in the project Valency, Argument Realization and Grammatical
Relations in Baltic, Vilnius University: (alphabetically) Peter Arkadiev, Axel Holvoet, Rolandas
Mikulskas, Nicole Nau and Björn Wiemer. Special thanks go to the external reviewer. All dis-
claimers apply.
2. Seeming exceptions are anaphoric pronouns marked with the IPG. I will discuss examples
with anaphoric pronouns marked with the IPG, cf. examples (16), (17) below, in more detail. At
this point, it suffices to say that these IPG-marked pronouns semantically embed presupposed
kinds or supersets introduced in the preceding discourse as input but in their overall meaning
(in the output) refer to the actual participant extracted from this superset/kind just as an in-
stantiation of such a kind or as a member of such a superset. This participant is indefinite and
not previously mentioned.
case-marked NPs is thus striking in Lithuanian. The IPG must be regarded as

a particular sort of morphologically-marked indefinite description in this lan-
guage. In general, one would expect a case to express some thematic relationship
that the respective NP bears to its predicate (cf. Blake 1994: 1–2) and not to alter
the referentiality of the respective NP. Here, however, the IPG is triggered by the
quantificational and/or determination properties of its host NP and, in some in-
stances, of its VP. Differently from the ‘regular’ cases, it cannot encode any of the
semantic or syntactic roles.
The IPG vs. NOM/ACC alternation represents a subtype of the widely-dis-
cussed phenomena of Differential Object- and Differential Subject Marking (in-
ter alia, Bossong 1998; Aissen 2003; Kittilä et al. 2011). Cross-linguistically, this
phenomenon may be governed not only by the NP-internal properties but also
by the tense or aspectual properties of the verb phrase (cf. Lestrade & de Hoop
2011). This is exactly what is found with the IPG of Lithuanian: the alternation
with the structural case is governed not only by indefiniteness of reference and
quantity but may also depend on the aspectual interpretation of the verb. Thus,
one may distinguish between NP-internal and VP-internal functions of the IPG.
The latter – differently from other similar instances such as the partitive case in
Finnic languages – are mostly restricted to incremental-theme accomplishments
and are only possible with mass nouns and plurals in Lithuanian (exceptions are
found with temporary transfer, cf. example (51) below).
As to its syntactic properties, the IPG takes part in a meaningful alternation
with structural cases only: it overrides the nominative case of unaccusative in-
transitive subjects and accusative direct objects of a particular group of verbs.
Although verbs do not subcategorize for the IPG, there are lexical restrictions.
Only a restricted number of verb classes allow the IPG alternation in Lithuanian.
Somewhat strikingly, the IPG of Lithuanian can (restrictedly) be coordinated
with accusatives and nominatives:
(2) Nusipirkau pien-o ir bandel-ę
buy.pst.1sg milk-gen.sg and roll-acc.sg
‘I bought milk and a roll.’
(3) Atvažiavo man-o brol-is, tėv-ai, kel-i
arrive.pst.3 1sg-gen brother-nom.sg parents-nom.pl some-nom.pl
vaikyst-ės draugai ir visoki-ų kit-ų
childhood-gen.sg friend:nom.pl and various-gen.pl other-gen.pl
sveči-ų
guest-gen.pl
‘My brother, parents, some friends from childhood, and various other guests
have arrived.’
Otherwise, only those constituents may be coordinated that “have an identical

dependence relation to their head” (Ambrazas et al. 2006: 460), cf. ungrammatical
(4) with an attempt to coordinate the beneficiary and the theme:
(4) Nusipirkau bandel-ę *ir tėv-ui

buy.pst.1sg roll-acc.sg *and father-dat
[Intended meaning] ‘I bought a roll for my father.’
This example would have been fully grammatical if the two NPs were not coor-
dinated. Moreover, it makes a certain type of participants indistinguishable with
regard to their encoding. Thus, a subset of S and O – to use Comrie (1978)’s well-
known labels –morphosyntactically pattern alike (S-split), which is by no means
typical for an accusative language such as Lithuanian.
The paper is structured as follows. First, I will discuss the quantificational and
determination properties of the IPG (Section 2). Section 3 deals with its interaction
with the aspectual and actional properties of the verb. Finally, Section 4 discusses
the reasons for the generalization and lexicalization of the IPG in the construc-
tions with an intensional predicate. Finally, Section 5 draws some conclusions.
2. Implicit quantifier
2.1 Evidence for an implicit quantifier
In what follows I will assume that the IPG evokes a quantifier/determiner (hence-
forth: Q) following the analogous analysis of the Russian IPG in Pesetsky (1982),
Neidle (1988), to mention but a few. Consider example in (2) (repeated here for
convenience as (5)):
(5) Nusipirkau pien-o ir bandel-ę
Buy.pst.1sg milk-gen.sg and roll-acc.sg
‘I bought milk and a roll.’
What is the reason for choosing the IPG marking of the first object NP and accu-
sative marking with the second object NP? As regards the second NP the answer
is simple: singular discrete entities cannot be marked by the IPG in Lithuanian
by definition; if they are, alternatively and exceptionally, they will be coerced into
a mass noun reading.3 Thus, there is no meaningful alternation available with
3. Cf. O pyragėlio pageidausit? [But pie.gen.sg wish.fut.2sg] ‘Would you like some pie?’
(http://www.lastfm.de/user/SkyBlueEyed4eve). Here, the NP pyragėlis ‘pie’ is treated as a mass
noun and not as a discrete count noun.
singular discrete referents (with some very few exceptions, see temporary transfer
in examples (51)–(52) below).4
The situation is, however, different with the first object NP, which allows both
IPG and the functionally (but not morphologically) unmarked accusative. One
of the functions of the IPG is to cause its NP to be marked as quantificational,
with an inherent indefinite quantity and, consequently, indefinite reference. That
is to say, there is an implicit quantifier/determiner that implies an indefinite quan-
tity of individuals, the references of which are also indefinite (cf. English some).
For the sake of clarity, I will assume that the IPG evokes an implicit quantifier Q
which has an indeterminate value by default, that is to say, in the lack of some
other overt quantifier present in the clause. Otherwise, it assumes the value of that
overt quantifier and (in many instances) becomes also syntactically dependent on
that quantifier.
The assumption of an implicit quantifier evoked by the IPG finds support
from those verbs that require their arguments to be quantified. Here, the IPG
patterns with NPs having an overt quantifier or measure phrase. Thus, the IPG
is triggered by such verbal quantifiers as: pri-, at-, per- meaning ‘a lot’ as well as
už- ‘a little bit’ (in the relevant meaning of these prefixes) – often only in combi-
nation with the reflexive suffix/prefix ‑si-, cf. už-si-kąsti duonos ‘to eat bread (ipg)
a little bit’, per-si-valgyti obuolių ‘overeat apples (ipg)’ (Ambrazas et al. 2006: 503).
In addition, there is a number of verbs that typically take a measure phrase such
as stokoti ‘to lack’, trūkti ‘to lack’, užtekti ‘to have enough of ’, mažėti ‘to decrease’,
daugėti ‘to increase’, etc. The semantics of these verbs leads to a straightforward
syntactic dependency such that verbs like pri-važiuoti ‘to arrive massively’ can
occur with partitive-genitive-marked subjects only (cf. Wiemer & Bjarnadóttir,
this volume):
(6) Pri-važiavo žmoni-ų / *žmon-ės
quant-drive.pst.3 people-gen.pl / *people-nom.pl
‘There have arrived a lot of people.’
Contrast (6) with the same verb važiuoti but without the quantifier:
(7) Važiavo *žmoni-ų / žmon-ės

drive.pst.3 *people-gen.pl / people-nom.pl
‘People drove/rode.’
4. This is different in Russian, where there is a meaningful alternation available in the context
of negation and a number of intensional verbs. Lithuanian, by contrast, has generalized the IPG
here (see Kuryłowicz 1971).
Such verbs as pri-važiavo do not allow non-quantified arguments: the nominative

žmonės makes the sentence in (6) ungrammatical. At the same time, an NP with
an overt quantifier such as daug ‘many’ or a measure phrase (in the nominative
case) such as (pilnas) kiemas ‘full yard’ are fully regular:
(8) Piln-as kiem-as žmoni-ų pri-važiavo
full-nom yard-nom people-gen.pl quant-drive.pst.3
‘There have arrived a full yard of people.’
(9) Daug žmoni-ų pri-važiavo
many people-gen.pl quant-drive.pst.3
‘There have arrived a lot of people.’
The restriction is not conditioned morphologically, since the measure phrase Pil-
nas kiemas in (8) is nominative subject of the verb.
There is, however, one restriction on the use of partitive genitive with overt
quantifiers: the reference of the NP – whether with or without an overt quantifier
or measure phrase – must be indefinite as in (6), (8) and (9). Formally speaking,
the IPG introduces here a variable that is bound by an existential quantifier by
default, if there are no explicit quantifiers or measure phrases in the clause. If
there is an overt quantifier, this variable is bound by the latter, e.g., by the one
incorporated in the verb as in (6), by an overt quantifier as in (9), or by a measure
phrase as in (8).
This being said, I will not dwell any longer on overt quantifiers triggering the
partitive genitive, since these instances do not represent an independent partitive
genitive taking part in a meaningful alternation with one of the structural cases.
2.2 Values of the implicit quantifier/determiner
The value of the implicit quantifier is vague – which is expected given that it does
not have an overt realization:
(10) Aš nusipirkau bandeli-ų
1sg.nom bought roll-gen.pl
‘I bought some rolls.’
The participant encoded by the IPG has inherently narrower scope with other
operators, cf. the universal quantifier visur ‘everywhere’ in the following exam-
ple. Here, the reading under which there would be everywhere one and the same
group of mistakes (i.e. specific mistakes) is excluded:
(11) Paklaustas, kokios klaidos lėmė jo pralaimėjimą, Mindaugas sakė:

asked which mistakes decided his losing Mindaugas said
“Visur buvo klaid-ų. Atsiprašau, vėliau pakalbėsim.”
everywhere be.pst.3 mistake-gen.pl excuse_me later talk.fut.1pl
‘As he was asked about which mistakes decided his failure, Mindaugas said:
“Mistakes were everywhere. Excuse me, let us discuss (this) later.”’5
Even though the value of Q remains vague, it has certain restrictions:
a. it is not compatible with contexts where it would have the value of ‘one’ item
from a set, even though logically and historically (cf. Seržant 2012a) this
might have been possible, cf. (12) below;
b. in some instances, it has become incompatible with contexts where it would
have an unbounded quantity value (to be explained below).
As to (a), contrast the following example from Ancient Greek, where the IPG al-
lows the value one (Seržant 2012a):
(12) Adrḗst-oio égēme thygatr-õn (Ancient Greek6)
Adrastos-gen.sg marry.aor.3sg daughter-gen.pl
‘He married a daughter of Adrastos.’ (Hom. Il. 14.121)
Now I turn to (b), which is a more complicated issue. I assume that the implicit
quantifier evoked by the IPG generally may have two readings:
i. bounded indeterminate reading

ii. unbounded indeterminate reading
The unbounded indeterminate reading is found when, literally, no boundaries for

the quantity implied are set, neither on the upper nor on the bottom edge. Here,
the IPG denotes only the concept or sort of things (a non-technical term used in
McNally 1998) that are involved in the situation referred to by the verb. On this
reading, the characteristic properties of the respective kind are crucial for the dis-
course, rather than a particular set of instantiations thereof. Naturally, denoting
just a concept and providing just for the properties of the participant, it cannot
have a specific quantity value. The unbounded value is a non-specific indefinite
reading with no existential commitments.
The bounded reading, in turn, implies certain boundaries on the quantity
of the individuals, even though these boundaries are not stated explicitly and
5. http://www.mamosdienorastis.lt/?p=6&sub=4137&pr=80.
6. The source language is indicated only for the non-Lithuanian examples.
remain vague. On this reading, there is a particular, finite set of individuals/en-

tities that cannot be extended (non-cumulative7). This reading is an existential,
indefinite plural reading. In contrast to the unbounded reading, the participant
introduced by the IPG on this reading can be anaphorically picked up in the fol-
lowing discourse.
Essentially, these readings are not just two different interpretations that can be
disambiguated only pragmatically or contextually. There are certain grammatical
contexts – as will be shown below – that display only one of these two readings.
In what follows I will elaborate on this distinction between two different read-
ings with examples and further details.
2.2.1 Bounded indeterminate reading

Consider example (13) (about the airplane crash with the former Polish presi-
dent) featuring the bounded indefinite reading:
(13) Nustatyt-a, kad kabinoje buvo asmenų, kurie
ascertained-ppp-n that in_cabin be.pst.3 person.gen.pl who
ne-buvo įgul-os nari-ai. Vien-o bals-as
neg-be.pst.3 crew-gen.sg member-nom.sg one-gen.sg voice-nom.sg
identifikuo-t-as, kit-ą ar kit-us turi
identify-ppp-nom.sg.m other-acc.sg or others-acc.pl must.prs.3
nustatyti Lenkij-o specialist-ai.
identify.inf Poland-gen.sg expert-nom.sg
‘It has been ascertained that there were persons in the cabin who weren’t crew
members. One (person)’s voice has been identified, the other or the others
will be identified by the Polish experts.’8
The quantity of the persons that must have been in the cabin before the crash is
delimited (bounded) and individuated as the following clauses imply. The IPG
invokes here a bounded or delimited set from which individuals can be discur-
sively accessed. This is evinced by the anaphoric reference vieno (balsas) ‘(the
voice) of one [person]’ as well as by kitą ‘other’ and kitus ‘others’ that directly
refer to voices but indirectly to the respective persons with these voices, that is, to
particular members of the set. Notably, the semantics of the IPG is ambiguous in
the first sentence of this example between the bounded and unbounded readings.
7. x is cumulative if: x plus x is also x (e.g., apples plus apples are also apples) (Quine 1960: 19;
Krifka 1989: 39; Kiparsky 1998).
8. http://www.15min.lt/naujiena/aktualu/pasaulis/kaczynskio-lektuvo-kabinoje-pries-pat-
katastrofa-buvo-zmoniu-kurie-nebuvo-igulos-nariai-nustate-tyrejai-57-99670
The bounded reading can only be assumed on the basis of the context provided
by the second sentence.
The situation is somewhat different in examples (14) and (15). Here, it is the
telic predicate in the past tense that requires or triggers the bounded reading of
the object seeds, because the action of buying has been accomplished and another
action is being carried out (scatter and cook, respectively), which can only be in-
terpreted in such a way that only a particular set/number of seeds has been affect-
ed. Here, the IPG creates a set of entities that are picked up in the next clause by
the anaphoric pronoun although the very referents remain inaccessible through-
out the discourse:
(14) Nu-pirkau gėli-ų sėkl-ų ir jas palei
tel-buy.1sg flower-gen.pl seed-gen.pl and 3.acc.pl.f along
keli-ą išbėriau
road-acc.sg scatter.pst.1sg
‘I bought [some] flower seeds and scattered them at the edge of the road.’9
(15) Mam-a davė pači-os užaugint-ų
Mum-nom.sg give.pst.3 self-gen.sg.f cultivate.ppp-gen.pl
pomidor-ų, iš-spaudžiau juos ir netyčia
tomato-gen.pl tel-press.pst.1sg 3.acc.pl.m and accidentally
išviriau skani-ą sriub-ą.
cook.pst.1sg tasty-acc.sg.f soup-acc.sg
‘My mother gave me [some] tomatoes which she herself had grown, I pressed
them and accidentally cooked a tasty soup.’10
The quantity of the entities implied is consequently particular/delimited, non-cu-

mulative: in (14), the speaker has bought a particular amount of flower seeds and
no more than this, while the very amount – being irrelevant for the discourse –
remains backgrounded, implicit and indeterminate. Example (15) is fully anal-
ogous. Differently from (14) and (15), the predicate in (13) does not have such
an entailment on its subject and may readily take an unbounded subject as well.
Indeed, it does so in the examples immediately below in (18).
Before I turn to the discussion of the unbounded reading in the next subsec-
tion, I briefly touch upon the examples in which the anaphora itself is embedded
under the IPG:
9. http://www.inspiration.lt/2010/03/netobulas-asotis/
10. http://m.lrytas.lt/-11746027321172481245-p2-laim%C4%97s-akimirkos-%C5%A1uolyje-
%C4%AF-bedugn%C4%99.htm
(16) … burokėli-ų sriub-a, apie kuri-ą rašei

beet-gen.pl soup-nom.sgi about which-acc.sg.f write.pst.2sg
po Naujamečio iš-viriau jos ir aš dabar
after New-Year tel-cook.pst.1sg 3sg.gen.fi’ also I now
‘… the beet soup, about which you wrote after New Year. I have also cooked
it now.’11
It is obvious that the person did not cook exactly the instance that is referred to
in the preceding clause. The pronoun jos ‘of it’ anaphorically refers back not to the
instance but to the kind of soup introduced by the first clause; it can be rendered
literally by ‘I cooked some of that sort of soup’. The predicate iš-viriau ‘I have
cooked’ due to its telicity requires boundedness on its object – as in (14) and (15).
The IPG extracts a delimited set of the kind introduced by the NP burokėlių sriuba
‘beet soup’ in the preceding clause. Crucially, the IPG jos is not a presupposition
here, since the specific portion of the soup that has been cooked is not introduced
previously in the discourse and cannot be accessed by general world knowledge.
The following example is analogous:
(17) Tauragės tinklinio komanda … į Kroatiją vyko už savo asmenines lėšas.
“Iš kur jie gavo tų pinig-ų,
From where 3.nom.sg.m get.pst.3 this.gen.pl money-gen.pl
mes visai nesidomėjome.”
‘The basketball team of Tauragė went to Croatia on its private funds. From
where did they get that money, we didn’t ask at all.’12
Here again, the IPG embeds the definite DP tie pinigai ‘these (nom) money (nom)’
that in itself is presuppositional and referring back to the kind asmenines lėšas
‘private funds’ introduced in the preceding clause. In the output, in turn, tų pinigų
‘of that money’ evokes a delimited set out of this presupposed kind being itself not
presupposed or definite. Again, it can be rendered by lit. ‘where did they get some
of that sort of money’.
Crucially, the IPG-marked object never constitutes primary information in
the discourse, even in these instances, where its referent is stored in the model
and picked up anaphorically in the following clauses. Thus, the exact money that
has been used to cover the travel expenses to Croatia in (17), the very portion
of the soup that has been cooked in (16) as well as those seeds that have been
11. http://gyvenimas.delfi.lt/receptai/article.php?id=8417053&com=1&s=2&no=120#ixzz2Ie
y3ZcHJ
12. http://www.aspauda.lt/repository/silokarcema/pagegiu%20archyvas%20PDF/2009-06-
30%207-8.pdf
scattered along the road in (14) themselves do not constitute central information
of the respective discourses. In (17), for example, the stress is on the information
that it was private and not public money, and, in (16), the sort of the soup is rel-
evant and not its particular instantiation that has been cooked.13
To conclude, the IPG itself cannot anaphorically pick up a discursively pre-
supposed participant in its output, but it can take presupposed referents (typically
kinds) as its input.
2.2.2 Unbounded indeterminate reading

Now I turn to the unbounded indeterminate reading. Consider examples (18)
and (19), where the IPG introduces a particular kind of people described by the
relative clause:
(18) Yra toki-ų žmoni-ų, kur-iems gali
be.prs.3 such-gen.pl people-gen.pl whom-dat.pl.m can.prs.2sg
ne-rašyti ne-skambinti, ne-sikalbėti
neg-write.inf neg-phone.inf neg-speak.inf
ir jie vis tiek bus patys artimiausi. Taip, ne?
‘There are such people that you don’t have to write to, to call, to speak with,
and they will nevertheless be the closest to you. Isn’t it so?’14
(19) Kiekvien-as mūsų pažįsta žmoni-ų, kurie
each-nom.sg.m 1pl.gen know.prs.3 people-gen.pl who-nom.pl.m
yra liekn-i, nors nuolat kemša.
be.prs.3 slim-nom.pl.m though constantly gobble.prs.3
šokolad-ą
chocolate-acc.sg
Dažnai slapta net pykstame ant jų ar likimo, kad ne visi gali valgyti tai, ką nori,
ir nestorėti.15
‘Each of us knows people who are slim, although they constantly consume
chocolate. We are often even secretly annoyed by them or by the fate that not
everyone can eat whatever (s)he would want to and not get fat.’
These examples do not imply any restriction as to quantity and cumulativity holds.
There might be just one individual or two hundred of them who would all satisfy
13. This has also been claimed in Seržant (2012b) for the IPG in Ancient Greek, where it seems
to have primarily backgrounding function.
14. http://www.formspring.me/r/yra-toki-moni-kuriems-gali-nera-yti-neskambinti-
nesikalb-ti-ir-jie-vis-tiek-bus-patys-artimiausi-taip-ne/350581344012617113?switch=hidden
15. http://www.lrytas.lt/gyvenimo-budas/tarp-musu/?p=3
the description given in (18) or (19). Existential sentences as in (18) or (19) assert
existence of particular kinds/subkinds, i.e. good people or slim people. This can be
tested by modifying the sentence as ‘There is such a kind/sort of people whom …’
for (18) and ‘Each of us knows the kind/sort of people that are …’. This inser-
tion does not change the meaning of the sentences or, at most, slightly. The IPG-
marked NPs here evoke characteristic properties related to those kinds while they
do not zoom in on particular instantiations. This reading of the IPG somewhat
resembles NPs used predicatively, which also do not introduce a set of individu-
als but rather refer to the characteristics that a particular referent is equated with.
The NP just provides the concept (cf. the property <e, t> type in Partee 1986, sug-
gested for the genitive of negation in Russian in Partee & Borschev 2004; Kagan
2005; Borschev et al. 2008).
Observe the scopal interpretation of the IPG in example (19) with regard to
the universal quantifier kiekvienas ‘every/each’: if taken as a quantity/set – since
there are most probably different sets of slim people that each of the persons ad-
dressed might know (even if somewhat intersecting) – it has narrow scope. How-
ever, if taken as a kind, it has wide scope, since everyone knows one and the same
kind of slim people (cf. McNally 1998: 359). The latter interpretation is even more
likely in (19) and, hence, the kind analysis of the IPG. The reason is that it is
precisely kind interpretation that is anaphorically picked up in the following dis-
course by the third person pronoun jie ‘they’ in (18) and jų ‘them’ in (19). If their
antecedent, the IPG would have introduced sets of instantiations of the kind that
each of us knows, one would expect these instantiations to be picked up by the
personal pronouns, as in the case of the bounded reading in (13) or (14) above.
However, this is not the case in (19) where the pronouns obviously refer to the
kind slim people.
Typical for the unbounded reading is the use of the abstract nouns with lexi-
cally empty verbs, quite parallel to light-verb constructions in other languages.
Here it is only the noun that provides lexical information about the action. I men-
tioned above the following predicate:
(20) Svarbiausia turėti kantryb-ės ir treniruotis.
Important.super.n have.inf patience-gen.sg and train.inf
‘The most important thing is to be patient and to train.’16
Thus, Ambrazas et al. (2006: 511) state that the construction turėti (lit.) ‘to have’
plus abstract noun in genitive constitute the general pattern, cf. turėti reikšmės (lit.
16. http://www.valstietis.lt/Pradzia/Laisvalaikis/Pramogos/Ugnies-sou-su-balandziais-ir-
Isjunk-sviesa
‘to have significance.gen’) ‘to be of significance. If these nouns get modified by an

adjective, however, then the accusative case-marking is preferred:
(21) Tai turi reikšm-ės
this have.prs.3 significance-gen.sg
‘This is of importance.’
(22) Tai turi didel-ę reikšm-ę
this have.prs.3 big-acc.sg significance-acc.sg
‘This is of great importance.’
This is probably because the modification endows the referent of the NP with
more individuation and, hence, the accusative is more preferable here. Note that
semantically examples with the IPG can be interpreted as patterning somewhat
with object incorporation. The object is disindividualized, it cannot be focused on
its own (only the whole VP may be in focus), it constitutes one lexical meaning
and one predicate with the lexically empty verb turėti. The insertion of a modify-
ing adjective as in example (22) splits up the incorporation and, hence, non-in-
corporating accusative marking is preferable.
The kind or concept interpretation of the IPG NP on the unbounded reading
is supported by its compatibility with adjectives taking kinds as their comple-
ments such as visoks ‘any/different kinds of ’ or įvairius ‘various’:
(23) Visoki-ų laik-ų užėjo
any-gen.pl time-gen.pl come_up.pst.3
‘There were all sorts of times.’ (LKŽ, vide sub užeiti)
Note that these adjectives predominantly modify kinds due to their lexical se-
mantics. The speaker of (23) does not speak about an indefinite number of peri-
ods in a purely temporal sense as non-diverse entities of the same sort but rather
about different kinds of periods. In a sense, these adjectives modify the concept of
the NP ‘time’ and not its particular instantiations.
2.2.3 Bounded vs. unbounded indeterminate readings, summing up

To summarize, the bounded indeterminate reading (henceforth: the bounded
reading) is an indefinite plural or existential reading that denotes a set of indi-
viduals (analogously for mass nouns: a limited quantity of parts of a mass), while
the unbounded indeterminate reading (henceforth: the unbounded reading) does
not evoke any particular set. Instead, I assume that it just provides the concept or
a description of a particular kind of the participant. On this kind-reading, it can
be treated as a kind-individual, as can be observed from examples like (19), in
which it is anaphorically picked up in the following discourse, or examples like
(11), in which as a kind-individual it has wide scope with regard to a quantifier.

While this reading allows referring to the kind of the respective NP, it cannot
have generic interpretation (as kinds typically have), because the situations de-
scribed in the examples above are not about all or nearly all individuals fitting the
description of that kind. In the latter conditions, nominative/accusative case has
to be used. Seržant (2012b) argues that the IPG of Ancient Greek functions quite
like a restrictive relative clause – it restricts the kind of the respective participant.
Generics, in turn, provide kind reference by virtue of the totality of the members
of the kind. Thus, a typical test for kinds of this type is its compatibility with such
predicates as ‘to die out’. Here, the judgement about the kind is logically derived
from the judgements about all its members (a kind dies out only if all its members
die out).
However, similar to generics, this reading is also unbounded, since there is
no requirement for a particular limit of potential individuals. To use a metaphor
for the unbounded reading, the description of the participant remains on a high
zoom-out level “not seeing” the particular individuals but only their concept.
I have noted above that, for many occurrences, distinctions between the two
readings are typically provided only by context. The optionality crucially hinges
upon whether or not the predicate allows both readings. Thus, telic predicates
denoting accomplished actions are compatible with the bounded reading only,
while atelic predicates such as pažinti ‘to know’ or būti ‘to be, to exist’ allow both
readings and further context is needed to disambiguate them. This situation is
somewhat reminiscent of the Carlsonian unified analysis of English bare plurals
(Carlson 1977) that also may have two different readings very much dependent
on whether the predicate may have the individual-level (generic) interpretation or
not, contrast the generic reading in (24) vs. the indefinite plural reading in (25):
(24) Horses are smart.
(25) Horses ran into the lake.
On Carlson’s (1977) unified account English bare plurals encode only kinds, while
there is a realization relation that derives the existential indefinite plural reading
such as (25) from the unified kind reading, taking into account the individual vs.
stage level value of the predicate (I skip the formalism). The existential quantifier
is provided by the verb in instances like (25) and does not stem from the NP itself,
which is thus the same in both (25) and (24) as to its referential properties. This
is furthermore an elegant explanation for why the bare NPs always take narrow
scope with regard to other quantifiers in the sentence: since the existential quan-
tifier comes from the predicate, it is naturally within the same scope as the verb
with respect to other quantifiers such as, e.g., negation.
Thus, a question would be legitimate about whether there might be a uni-

fied analysis for both the unbounded, kind/concept reading and the indefinite
plural, bounded reading of the Lithuanian IPG in the same vein. Note, however,
that not only the bounded reading but also the unbounded reading – contrary
to the English bare plurals – cannot have the property of inclusiveness (totality)
and, hence, does not have the generic reading, unlike the English bare plurals
(the respective structural case is mandatory here in Lithuanian). Furthermore,
as I will show below (Section 3.2) in more detail, the two readings of the IPG are
not in complementary distribution across the Lithuanian predicates, unlike the
two readings of the English bare plurals. Crucially, while the bounded reading
is found to interact with the boundedness value of the incremental-theme verbs,
this is not true for the unbounded reading, which is found only with NP-internal
functions. The VP-related functions of the IPG are mainly found with incremen-
tal-theme verbs. Here, one might expect to find the unbounded reading of the
IPG in the progressive reading of the verb, because progressive itself does not
impose any boundary on the event, nor – with incremental-theme verbs – on the
object NP (cf. English I am eating sandwiches). However, the unbounded reading
is blocked, the accusative case-marking being the only option here in Lithuanian
(differently from Finnish).
3. Implicit quantifier and aspect
In the examples above, the implicit quantifier Q is a typical D(eterminer)-quan-

tifier (term introduced in Löbner 1985, Partee 1995) with regard to its domain
of application, i.e., a quantifier that applies internally to its host XP. However,
with some incremental-theme verbs and, dialectally, with some verbs of transfer
(see example (51) below), this quantifier interacts with the actional properties of
the verb in a uniform way, thus patterning rather as an A(dverb)-quantifier here
(term introduced in Löbner 1985; Partee 1995). The implicit quantifier Q here
undergoes a development otherwise well-established with overt quantifiers that
start out as D-quantifiers but develop into A-quantifiers and not vice versa (cf.
Keenan & Paperno 2012), cf. D in I ate a little bit of the cake vs. A in I ate the cake
a little bit.
Before I turn to the discussion of the data (Section 3.2), I introduce the no-
tional inventory on aspect and actionality – just as much as will be necessary for
the present paper – in Section 3.1.
3.1 Extending (un)boundedness onto the verbal domain
3.1.1 Actionality vs. aspect in narrow sense

I adhere to the bidimensional approach to aspectuality (Bertinetto 1997; Smith
1997; Sasse 2002; just to mention a few). That is, I distinguish between the viewpoint
aspect or aspect1 (as per Sasse 2002) and actionality (also termed, e.g., Aktionsart
or aspect2 in Sasse 2002). Crucially, both dimensions operate with boundaries of
an event: while actionality is about inherent (cf., inter alia, Depraetere 1995) or in-
trinsic (Sasse 2002) boundaries, viewpoint aspect is about temporal or established
boundaries (Sasse 2002: 205–206). Actionality refers to the inherent organization of
an event such as referred to by Vendler’s classes (activity, achievement, accomplish-
ment or state, Vendler 1957[1967]) and their various subsequent modifications as
well as by such compositional properties as telicity (telic vs. atelic) or dynamicity.
The viewpoint aspect (aspect1 in Sasse 2002), in turn, refers to the proper-
ties that the speaker establishes in a particular utterance and that pertain to such
domains as discourse organization or pragmatics and not to the semantics of the
event itself. Traditionally, one distinguishes the perfective and imperfective aspect
here. The former implies that a particular event is represented as included into
the reference point of the narration, while the latter entails that it is the reference
point that is included into the duration of the event, or, more precisely, the im-
perfective is often simply non-committal as to its temporal delimitation. In other
words, the event is either represented as not delimited or, minimally, as non-com-
mittal with regard to its boundaries.
Lithuanian does not have morphological means to encode viewpoint aspect,
which therefore can only be disambiguated contextually (cf. Arkadiev 2011 and
the literature therein). In what follows I will refer to viewpoint aspect in the sense
of a contextually inferred interpretation.
The imperfective viewpoint can invoke different interpretations of the simple,
non-derived verb in Lithuanian. Thus, such a form as valg-o (eat-pres.3) ‘eats/is
eating’ can have an episodic (actual) progressive interpretation ‘is eating’ if so re-
quired by the context, e.g., by the subjunction kol ‘while’. It can, furthermore, have
non-episodic interpretations such as generic (‘John eats sandwiches’) or habitual
(‘John eats sandwiches every morning’). To this extent, the Lithuanian verb is am-
biguous. Notably, there are morphological means to derive dedicated habitual or
iterative verbal stem formations, for instance, valg-y-dav-o (eat‑#‑iter‑pst.3) ‘ate
regularly’. These belong primarily to the domain of actionality (aspect2) and are
not dependent on the choice of the viewpoint, although there is some systematic
affinity between such actional classes as, e.g., habituals or states and the imperfec-
tive viewpoint, as is well known (cf. Sasse 2002: 210). I will discuss these deriva-
tions in more detail in a separate Section 3.3 below.
In what follows I will briefly describe the framework that allows establishing
a link between the nominal quantificational properties and such properties of the
verb (Section 3.1.2). Then, I will turn to the discussion of the data (Section 3.2).
Even though Lithuanian does not have dedicated markers for the viewpoint as-
pect (cf. Arkadiev 2011), as it will turn out from the discussion of the data, the
viewpoint interpretation of the verb is relevant for the assignment rules of the IPG
with incremental-theme verbs in Lithuanian as well as with transfer verbs in East
Lithuanian.
3.1.2 Aspect, actionality and the interaction with the object

In general, it is a well-known fact that certain verbs allow their objects to determine
their actional class (aspect2 property), being themselves ambiguous between ac-
complishments or activities (Verkuyl’s generalization in 1972), cf. English:
(26) He ate apples
(27) He ate the apples
The lack of a boundary in the object NP apples in (26) makes the whole VP to eat
apples pattern with activities such as to work, e.g., with respect to the commonly
used tests like in an hour/for an hour. The opposite is true of (27). Here, the de-
limited NP the apples supplies an inherent endpoint and the whole VP to eat the
apples becomes an accomplishment. In other words, the delimitation of the object
NP is projected on the verb, or, alternatively, the delimitation of the object NP
disambiguates the accomplishment reading of the verb to eat, while lack of such
a delimitation features the activity reading of this verb (Tenny 1992: 5). In any
event, we observe that the quantificational value of the object NP is mirrored by
the verb and, hence, becoming the value of the whole VP.
Now, the case with the IPG in Lithuanian is not fundamentally different from
what we observe in the English examples above: the quantificational value of the
IPG is rendered by the verb and, consequently, by the whole VP here. However,
before I come to the presentation of the data illustrating this point, I first intro-
duce a framework that will allow interaction between the quantificational proper-
ties of the object NP and the aspectual and actional interpretation of its verb and,
consequently, VP.
Since the majority of the verbs to be discussed in Section 3.2 are incremen-
tal-theme verbs (cf. Krifka 1989, term coined in Dowty 1991), some remarks on
this type of event encoding are in order. Incremental-theme verbs are typically
accomplishment verbs that establish a so-called homomorphic relationship with
their object NPs; this entails that every subpart of the event the verb denotes is
unique, and is coupled with a particular unique subpart of the object NP (Krifka
1992: 39). Thus, in a sentence like John ate the roll every specific subpart of the
roll corresponds to a specific subpart of the event of eating. The object is said to
‘measure out’ the event (Tenny 1994). From this it naturally follows that, if the in-
cremental theme is bounded, then the event in itself is also bounded. The typical
incremental-theme verbs that I will be dealing with below are consumption verbs
valgyti ‘to eat’, gerti ‘to drink’ or verbs of creation like virti ‘to cook’.
I believe that this relation holding between the object and the predicate with
the incremental-theme verbs has been transferred to other predicates and their
subjects or objects that are not typical incremental-theme verbs to begin with. To
give an example, consider example (6) repeated here as (28) for convenience:
(28) Pri-važiavo žmoni-ų / *žmon-ės
quant-drive.pst.3 people-gen.pl / *people-nom.pl
‘There have arrived a lot of people / too many people.’
In this example the relationship between the verb and the object is read into a
homomorphic one – the subparts/subsets of the subject žmonių ‘people’ can be
mapped onto the subevents of the arrivals. The verb requires its IPG subject NP
to have distributive reading while the collective reading is blocked. That is to say,
different sets of people correspond to particular arrivals: e.g., first came John and
Mary, then came a neighbor and some other people, etc., finally, too many people
arrived and the inherent endpoint massive arrival has been achieved. The very
event of arrival is not typically homomorphic not only with singular subjects and
collective plurals but also with distributive plurals, since the regular verb to arrive
is accomplished with every subpart of such a subject, and there is no inherent
endpoint of a higher level comprising different arrivals like in the Lithuanian ex-
ample above, cf. Several well-known linguists have arrived.
I suggest, following Kiparsky (1998)’s argumentation on Finnish, that these
are the incremental-theme verbs that gave rise to instances with an interaction
between the object quantifier and the actionality of the VP found with tempo-
ral transfer in Eastern Lithuanian (to be discussed below, cf. example (51)) or in
Finnish examples like (34) (below). There is no homomorphic relationship be-
tween the verb and its object in these instances, but the object quantifier never-
theless quantifies over the predicate in the way the quantifier of an incremental
theme object measures the event.
In order to be able to account for the Lithuanian data presented in the next
subsection, I have to adopt an approach that would be able to model the interac-
tion found between some verbs and their objects. I am not going to discuss pos-
sible options of the syntactic implementation of such an interaction. In formal
approaches, such an interaction will assume a node where the quantificational
properties of the verb and those of the object will interact, generating the aspec-
tual and actional interpretation of the whole clause. This, in turn, presupposes
that the aspectual and actional values of the predicate, on the one hand, and the
quantificational properties of the object NP (or, where necessary, also of the sub-
ject NP), on the other, must be compatible with each other in order to yield a
common interpretation of the whole sentence. Such a parallelism between the
quantification of the event structure and of the nominal has been argued for in
the literature (cf., inter alia, Verkuyl 1972; Krifka 1989, 1992; Filip 1989; Kiparsky
1998; Borer 2005 and most recently Champollion 2010).
One of the most influential approaches which integrates event structure
quantification and NP quantification is Kiparsky (1998). Kiparsky’s concern is
primarily to account for the assignment of the partitive case – as opposed to the
accusative case (traditionally, genitive) – in Finnish. His main claim is that the
unboundedness of the VP is the discriminating factor that requires the assignment
of the partitive case to the object NP, whereas boundedness motivates accusative.
According to Kiparsky, a VP is unbounded if the predicate and/or the object is
unbounded. There are verbs that are inherently unbounded (such as psych verbs)
and, hence, inherently take the partitive case, while there are accomplishment
verbs that alternate between partitive and accusative giving rise to unbounded
and bounded interpretations, respectively. Kiparsky (1998) defines unbounded-
ness as follows: a predicate or NP is unbounded “iff it is cumulative, divisive and
not diverse” (I skip the formulaic definitions):
(29) Unboundedness: cumulative, divisive and not diverse
x is cumulative if: x plus x is also x (e.g., apples plus apples are also apples);
x is divisive if: any subpart of x is also x (e.g., a subpart of apples is also
apples);
x is diverse if: x is not atomic and its members are not related by a subpart
relation (e.g., one apple is not diverse).
Indeed, this approach has a number of advantages: it coherently accounts for

most of the instances of the partitive case assignments in Finnish and, crucially,
it provides a unified model for the verbal and nominal quantification as well as
for the interaction between both. Furthermore, it makes an important distinction
between (un)boundedness and (a)telicity that are notoriously used synonymously
in the literature (e.g., in Doetjes 1997).
There are, however, some problems with this account which are even more
relevant for Lithuanian or Russian than for Finnish. The main problem lies in how
it deals with what one might call “restrictedly unbounded” or “weakly bounded”
quantities. These are indefinite portions of something: either (i) NPs headed by
such weak quantifiers/determiners as some, or (ii) verbs that denote only some
portion of an action (by means of their morphology, lexically or contextually),
e.g., Lithuanian pa-valgyti ‘to eat a little bit/somewhat’. Lack of clear-cut, definite
boundaries might point to the unboundedness of these portions but, on the other
hand, the presence of some – even if vague – cut-off point unequivocally suggests
bounded interpretation. I illustrate this in more detail.
Borer (2005) has pointed out that a problem for Kiparsky’s approach emerges
with such NPs as apples and some apples which are not explicitly discriminated
here, even though both yield contradictory results with regard to the bounded-
ness and telicity value of their VPs, contrast (30) vs. (31):17
(30) He ate apples (for an hour/*in an hour)
(31) He ate some apples (in an hour/*for an hour)
On Kiparsky’s approach both NPs are cumulative and not diverse, while their divi-
siveness value is dependent upon whether or not the singular atoms are included
in the set. To give an example, if one assumes a value of two apples for apples and
some apples, then both would emerge as non-divisive (since one apple fits neither
the description apples nor the description some apples), alternatively, if singular
atoms are excluded with both then both become divisive (Borer 2005: 42ff.). Fur-
thermore, a boundary at the upper edge is explicitly claimed with some apples,
because one cannot infinitely add some apples to some apples and get a set that
would still fit the description of some apples. Even though, I concede, this bound-
ary is vague and is subject to an individual interpretation, it nevertheless exists
and is explicitly signaled by the quantifier some. Thus, if one has two apples for
some apples and add another some apples, e.g., also two apples then four apples
might still fit the description of some apples. Once, however, one has reached, say,
1000 apples the description of some apples will no longer do, at least, not for most
speakers.18 Note that this is not the case with the description apples. There is, thus,
a difference between apples and some apples in terms of boundedness.
Having no distinction between apples and some apples would predict that
both VPs in (30) and (31) must pattern alike. Crucially, only some apples yield
a telic interpretation of the event, as pointed out in Borer (2005). Any account
17. As Eleanor Coghill pointed out to me (p. c.), the telic reading is somewhat odd, because one
usually would not time how long it takes to finish a vague amount of food. Nevertheless, (31) is
better with “in an hour” than with “for an hour”.
18. Thus, one of the reviewers has pointed out that boundedness of some is a matter or prag-
matics. I disagree, I believe the exact amount that no longer fits the description of some is in-
deed a matter of personal interpretation. However, the very existence of a boundary with some
is not a matter of pragmatics or context, but is an inherent semantic property of some.
must therefore discriminate between apples and some apples. Thus, Borer (2005)
suggests that bare plural NPs (i.e. the plural marker itself) induce plurality, which
excludes atomic members and encompasses only various combinations of several
members, while the quantifier some includes also the atomic singulars in its set.
This makes some apples a non-divisive entity (because at least one apple cannot
have a subpart that would fit the description some apples) and apples a divisive
one. This allows Borer to get the value bounded with some apples (because it is
non-divisive and, hence, violating (29)) and the value unbounded with apples
(because it conforms to (29)). Borer’s approach is, however, less compatible with
mass nouns quantified by some (some water), which also yield a telic event inter-
pretation but are unbounded. Additionally, as noted above, I assume that some
apples is also not cumulative sensu stricto.
Be that as it may, on the descriptive level I state that the reading in (30) is to
some extent parallel to (but not identical with) the unbounded reading of the Lith-
uanian IPG discussed above. By contrast, (31) is an existential reading evoking a
non-empty set of entities (or a set of portions for mass nouns) that have naturally
a quantificational value which is expressed by some. I suggest that (31) is parallel
to the bounded reading found with the Lithuanian IPG. Note, however, that the
functional overlap between the unbounded reading of the Lithuanian IPG and the
English bare plurals, on the one hand, and the bounded reading with English some,
on the other hand, is only partial. Thus, English some, being an overt expression,
can be focused (at least, on its quantifier reading), while the implicit quantifier of
the Lithuanian IPG cannot. The implicit quantifier is inherently backgrounded
in the discourse – it is referentially so ‘weak’ that it cannot be a focus on its own.
Thus, the question ‘How much water did you drink?’ cannot be answered by (32)
but only by (33):
(32) Vanden-s išgėriau
Water-gen drink.pst.1sg
‘I drank [some] water.’
(33) Šiek tiek / truputį vanden-s išgėriau
some / some water-gen drank.pst.1sg
‘I drank [some] water.’
The utterance in (32) can only be used when the whole VP is in focus, e.g., an-
swering the question ‘What did you do?’.
In the same way, the analogy between the bare plurals in English and the
unbounded reading of the IPG in Lithuanian is not absolute. Recall that a typical
function of English bare NPs is to have a generic reading (cf., inter alia, Carlsson
1977). This reading is not available with the IPG; instead the default, structural
cases (i.e. the nominative in the subject and accusative in the object position)
have to be used in Lithuanian to denote genericity. The value of a sort of universal

quantifier such as all is not available with the IPG on any of its readings. The un-
bounded reading of the IPG is even a weaker reference than that of bare plurals in
English. It denotes only the concept and is, in a sense, non-extensional.
I turn back to the notion of (un)boundedness. I have noted above that – as
far as Kiparsky’s approach is concerned – it is problematic with indefinite plu-
rals/existentials like some X, because the latter are bounded portions/sets not
only semantically but also with respect to their behavior with other categories
such as their contribution to the actionality type. Now, it seems that the same
problem also exists with those predicates that morphologically or lexically de-
note a particular portion/part of an action. In the case of accomplishment verbs,
Kiparsky’s model predicts correct results only when this portion exhausts the ac-
tion completely, including the culmination (P-property in Dahl 1981). However,
if just a particular portion/part of an accomplishment event or of an activity event
is referred to by the predicate, then this portion must intuitively be bounded, but
Kiparsky’s model treats it as unbounded. Consider the following example from
Kiparsky (1998):
(34) Hän avasi ikkunaa (Finnish)
3sg open.pst.3sg window.part.sg
(i) ‘He was opening the window.’
(ii) ‘He opened the window (partly).’19
Both readings encode an indeterminately quantified action. Crucially, while (i)

does not make any commitments as to the boundaries of the event featuring the
progressive reading, (ii) does imply such a boundary. Reading (ii) highlights
here the delimitative reading (= cessative reading in the Finnish tradition, cf.
Huumo 2010: 90). This reading entails that the action has taken place for a while
and then has stopped for whatsoever reason without reaching its natural bound-
ary (culmination/telos/inherent endpoint), if such a boundary is implied by the
actionality type of the respective verb at all.20 Now, while reading (i) is perfectly
19. Kiparsky (1998), who cites this example, adduces two other readings available: (iii) ‘He
opened the window for a while’ (cf. temporary transfer below) and (iv) ‘He opened the window
again and again’. Those are, however, not relevant at this point.
20. I will use here the notion delimitative adopted in, inter alia, Sasse (2002: 206), first intro-
duced in Maslov (1959). Maslov refers to delimitatives as to aktionsart (Russ. ‘sposob dejstvija’),
i.e. as pertaining to the domain of actionality and not to aspect sensu stricto. Delimitatives are
typically derived from homogenous non-culminating predicates such as to walk or to sleep (cf.
Mehlig 2006 for this argument on Russian). However, as Mehlig (2006: 253ff.) notes, there is
a number of accomplishment verbs like pisat’ ‘to write’, pit’ ‘to drink’ or even otkryt’ ‘to open’
consistent with Kiparsky’s model, reading (ii) suffers from problems similar to the
ones mentioned above in relation to the boundedness value of some apples in the
nominal domain. This reading cannot be regarded as unbounded in any non-the-
ory-dependent sense – it obviously implies certain boundaries, although again, I
concede, these boundaries are not definite.21 It is, therefore, not cumulative, since
not every partly opening the window plus partly opening the window will also yield
partly opening the window. Consider the situation when the two processes partly
opening the window have the value of half-opening the window. Here the result will
no longer fit the description partly opening the window but rather the description
completely opening the window. This is not the place to further discuss the Finnish
data but I remark that the discriminating semantic factor in Finnish thus does not
seem to be boundedness in Kiparsky’s terms but rather totality as has been claimed
in the literature, see Huumo (2010) for the most comprehensive recent account:
those VPs that have a positive value of the feature totality (culminating VPs) mark
their objects with accusative, while those that have a negative value thereof use
partitive. The feature totality (conditionally) implies, of course, the positive value
of the feature boundedness but not necessarily vice versa as in the case of the de-
limitatives (no biconditionality).
Similarly, the partitivity account adhered to in Filip (1989) or Krifka (1998)
is generally coherent, though it faces similar problems. It crucially relies on the
notion of partitivity that is assumed to be the common feature both of verbal
aspects such as progressive and of mass nouns/bare plurals, given that the incre-
mental-theme relation between the object and the verb holds. In Krifka (1998),
the imperfective aspect is obtained by extracting some parts from the whole de-
notational base of a telic verb. Hence the parallelism with nominal partitivity: the
imperfective aspect is interpreted here as referring to a part of a telic event in the
same way as nominal partitives refer to a part of the NP they embed (cf. Krifka
1998). Again, Finnish data seem to match this account, because it is the partitive
case here that yields the progressive interpretation of the event. However, this
approach does not take into account the perfective non-accomplished, delimi-
tative readings of the Finnish partitive, as in reading (ii) ‘to open the window
partly/somewhat’ in example (34) above. That is, a part of an event may be either
as well that can be conceptualized as homogenous, if the focus is on the activity taking place
before the inherent culmination/endpoint.
21. The lack of definite boundaries might have been the reason for treating it as unbounded.
Indeed, to some extent this reading behaves as unbounded: somewhat opening the window plus
somewhat opening the window may also yield somewhat opening the window, tests on divisive-
ness will give analogical results. However, this reading is only restrictedly unbounded and
above/below some level it stops being cumulative or divisive, respectively.
bounded (i.e., be temporally delimited) or not. This distinction is less relevant for
Finnish but is essential for Lithuanian (and Russian, see Seržant, forthcoming a).
Thus – as I will show below (Section 3.2) – only if the part of the event encod-
ed by an incremental-theme verb is bounded, may the IPG be used, while if it
is unbounded then only accusative can be used in Lithuanian. While essentially
correct, the partitivity account in Filip (1989), Krifka (1989) or Kiparsky (1998) is
thus not fine-grained enough with respect to aspectuality.
Tatevosov and Ivanov (2009) fix this problem by distinguishing two func-
tions (operators in their terms): (a) non-culmination (actionality/aspect2) and
(b) perfectivity/imperfectivity (viewpoint aspect/aspect1). This decompositional
approach is coherent with the data discussed here. At this juncture, there is a way
to capture the similarities and differences between the partitive case in Finnish
and the IPG in Lithuanian (and Russian). The (a) function always creates non-cul-
minating accomplishments in both Lithuanian and Finnish (and also in Russian),
but not necessarily imperfective aspect as assumed in Krifka (1998) or progressive
as claimed in Filip (1989). This function belongs rather to the domain of actional-
ity and not to the viewpoint aspect, as Tatevosov and Ivanov (2009: 93–95) cor-
rectly maintain. This is also intuitively more likely, since the lack of culmination
is inherent here to such VPs exactly as the presence of the latter is inherent for the
accomplishment VPs. The authors argue that the output of the non-culmination
function is the input to the viewpoint aspect (Tatevosov & Ivanov 2009: 94). Thus,
the perfective interpretation is indeed found morphologically marked in Russian
as well as contextually disambiguated in Bagwalal (Nakh-Daghestanian, Andic)
(Tatevosov & Ivanov 2009: 93–94) and in Finnish with the delimitative readings
such as (ii) in (34) above. Furthermore, this account is compatible with the imper-
fective reading (i) in (34) that for some reason has become more prominent in the
literature.22 The imperfective reading must be analysed as follows: the partitive
case is responsible for the actional class of non-culmination of the VP, yielding a
non-culminating activity VP to open a window. Now, this activity VP may occur
in an imperfective (e.g. progressive ‘I am opening the window’) or perfective con-
text (e.g., ‘I somewhat opened the window’).
For Lithuanian, I argue that not only unboundedness2 (cf. Sasse 2002) on the
level of actionality, i.e. the non-culmination of the VP, is an essential condition but
that boundedness1 (cf. Sasse 2002: 205–206) induced by the viewpoint perfectivity
is equally important, in contrast to Finnish. That is to say, viewpoint perfectivity is
22. Cf. also Metslang (2001) on Estonian, who also adheres to a straightforward relationship
between the partitive case and imperfectivity. However, what she means here is non-culmina-
tion, which belongs rather to the domain of actionality (inherent to a VP) and not to the view-
point aspect.
an equally important parameter for assigning the IPG to the objects of the incre-
mental-theme verbs and temporal transfer besides non-culmination (cf. Holvoet
1991 for a similar argument on Polish). Notably, these factors do not play any role
for those verbs that do not allow such an interaction between the nominal and
verbal partitivity, e.g. the verb pirkti ‘to buy’. With these verbs the IPG operates
NP-internally only, e.g., with būti ‘to be’, pažinti ‘to know’. Notably, the Finnish
synonyms of latter verbs equally do not show any interaction on the VP level.
3.2 Interaction of the predicate and nominal (un)boundedness

in Lithuanian
Dahl (1984: 10) argues that the perfective aspect of Russian is incompatible with
unbounded objects:
(35) On na-pisal pis’m-a (Russian)
3.nom.sg.m tel/perf-write.pst.sg.m letter-acc.pl
‘He wrote some letters.’
*‘He wrote letters.’
In (35), only the bounded reading of the object – ‘some letters’ – will make a
grammatical sentence, while the unbounded reading ‘letters’ is not possible here.
This is because the event of “writing is an unbounded activity as long as the object
is not delimited in some way” (Dahl 1984: 10). In other words, the incremental-
theme verbs require their objects to be bounded in perfective contexts. As I argue
below, the situation with the IPG of Lithuanian is fully parallel.
Koptjevskaja-Tamm and Wälchli (2001: 652) argue that aspectual considera-
tions – unlike some other languages of the linguistic area such as Russian or Finn-
ish – are not relevant for the IPG in Lithuanian. Indeed, at first glance, unlike
Russian, Lithuanian allows the IPG to occur with both telic (36) and atelic predi-
cates (37) and seems to be independent of the choice of the actionality type, cf. the
verb ger-ti ‘to drink’ with telicity overtly marked in (36) by means of the prefix iš-
as opposed to the morphologically unmarked gėrė with no such marking in (37):
(36) Jis iš-gėrė vanden-s/vanden-į
3.nom.sg.m tel-drink.pst.3 water-gen.sg/water-acc.sg
‘He drank up (some) water / water.’
(37) Jis gėrė vanden-s/vanden-į
3.nom.sg.m drink.pst.3 water-gen.sg/water-acc.sg
‘He drank (some) water / water.’
However, this is not the whole story. Combinations of the IPG with different as-
pectual readings of the incremental-theme verbs reveal the interaction of the IPG
with different viewpoint contexts in Lithuanian. In what follows I will argue that
Lithuanian has generalized only the bounded reading of the IPG for the interac-
tion with the aspectual properties of the predicate, while the unbounded reading is
excluded from this interaction function in Lithuanian; the latter remains available
only for NP-internal function.
The first evidence comes from delimitatives. Delimitatives are formed by
means of the prefix pain Lithuanian. Delimitatives from incremental-theme
verbs typically take IPG-marked objects. Contrast pa-gėrė in (38) marked explic-
itly as a delimitative allowing perfective viewpoint contexts only, with its etymo-
logical counterpart in (41) that is ambiguous as to its aspectual interpretation:
(38) Jis pa-gėrė vanden-s/??vanden-į ir
3.nom.sg.m delim-drink.pst.3 water-gen.sg/??water-acc.sg and
išėjo iš virtuvės
walk_out.pst.3 from kitchen
‘He drank water (for a while) and then left the kitchen.’
*‘He was drinking (some) water.’
In the context of delimitatives, accusative is restricted to a very specific use, name-

ly, if the ontological class of the NP has to be emphasized (cf. Paykin, forthcom-
ing, on a similar phenomenon in Russian):
(39) Jon-as tris dien-as pa-gėrė vien tik
John-nom three.acc days-acc.pl delim-drink-pst.3 only
vanden-s/vanden-į ir pasveiko
water-gen.sg/water-acc.sg and recovered
‘John drank only water for three days and recovered.’
(40) Kol ne-atvežė ger-o vyn-o, jis
until neg-deliver.pst.3 good-gen.sg.m wine-gen.sg 3.nom.sg.m
kurį laiką pa-gėrė Kagor-ą
some_time delim-drink.pst.3 Cahors-acc
‘Until a good wine was delivered, he drank Cahors wine for some time.’
Elsewhere, the IPG is the default case-marking with incremental-theme delimita-

tives, to the extent that no examples with accusative can be obtained from Google
search. This is different from the unmarked stem ger-ti ‘to drink’, where accusa-
tive is not only used for emphasizing class but also and mainly in imperfective
contexts.
(41) Jis gėrė vanden-į

3.nom.sg.m drink.pst.3 water-acc.sg
‘He drank/drank up/was drinking water / the water / water (and not beer).’
As has been mentioned, delimitatives entail that the process had run for a while
and stopped for whatsoever reasons without reaching its natural end (if such an
end is presupposed at all). The utterance in (38) is thus bounded, but the bound-
ary is arbitrary and does not coincide with any intrinsic boundary (culmination).
It is a non-culminating bounded or perfective event. There is thus a sort of ‘agree-
ment’ between the indefinite bounded reading of the object ‘some water’ and the
bounded reading of the predicate that equally does not provide a definite bound-
ary. Additionally, both the predicate and the object agree in their non-commit-
ment to the full exhaustiveness (culmination for the verb and inclusiveness for the
object). To further test the hypothesis about the necessity of a semantic concord
between the value of the implicit quantifier of the IPG and that of the verb, let us
embed the same verb into the context of the imperfective viewpoint (42)–(45)
where, e.g., English would require a progressive verb:
(42) Kol jis valgė bandel-es/?bandeli-ų,.
while 3.nom.sg.m eat.pst.3 roll-acc.pl/?roll-gen.pl
kitijau pradėjo dirbti
‘While he was eating rolls, others already started working.’
(43) Jon-as valgė ramiai ??bandeli-ų/bandel-es,
John-nom eat.pst.3 quietly ??roll-gen.pl/roll-acc.pl
bet staiga pašoko ir išbėgo į kiemą
‘John was quietly eating rolls, but suddenly jumped up and ran into the
yard.’
(44) Kol viriau makaron-ų sriub-ą/*sriub-os,
while cook.pst.1sg noodles-gen.pl soup-acc/*soup-gen,
man sudegė bulvės
1sg.dat burn.pst.3 potato-nom.pl
‘While I was cooking noodle soup, my potatoes got burned.’
(45) Kai Jonas paklausė Eglę dėl ryto planų, ji negalėjo iškart atsakyti,
nes virė *sriub-os/sriub-ą.
Because cook.pst.3 *soup-gen/soup-acc
‘When John asked Eglė about tomorrow’s plans, she could not answer right
away, because she was cooking soup.’
(46) Kol Jon-as virėsi kiaušini-us/?kiaušini-ų mišrainei,

while John-nom cook.pst.3.refl egg-acc.pl/?egg-gen.pl salad-dat
kaip tik rodė žinias.
just as show:pst.3 news
‘While John was cooking himself eggs for a salad, the news was on (on TV).’
The results achieved with mass noun objects are clear-cut: the IPG case marking
of the object is blocked here and only the accusative case marking is allowed.
The results with plural count nouns are similar.23 Indeed, the progressive context
in (42)–(46) activates the unbounded (progressive) reading of the verb, which
would have been in conflict with the bounded reading of the IPG.24 One would
expect the unbounded reading of the IPG here, but it is unavailable for the inter-
action with the actional properties of the incremental-theme verbs in Lithuanian
(unlike Finnish) and only accusative is an option here.
I conclude that an incremental-theme predicate must not be unbounded (be
in an imperfective context) in order to be able to combine with the IPG-marked
objects in Lithuanian. The IPG is most natural with delimitative verbs marked
with the prefix pa-, but it can also co-occur with culminating predicates such as
in (47):
(47) Jis iš-gėrė al-aus
3.nom.sg.m tel-drink.pst.3 beer-gen.sg
‘He drank (some) beer.’
Note that the interpretation of the whole VP in (47) is similar to (42)–(46) – in

both instances, at a minimum, there is no commitment as to whether the culmi-
nation has been achieved, and the conversational implicature of (47) would be
that the event has not been fully exhausted. Needless to say, an implicature can
be defeated. That is to say, (47) may in fact be uttered in the situation where the
action has been exhausted de facto, but the speaker just did not want to commit
her-/himself to the totality (something like English there has been beer drink-
ing would neither entail nor exclude the fact that the whole amount of the beer
23. Both native speakers whom I have consulted unequivocally prefer accusative. However, in
example (46), both speakers reported that they could imagine someone using the IPG, while
not preferring it themselves. Speaker 2 rejected the IPG with (42) and (43), while speaker 1
reported that she accepts them with plural count nouns but not with mass nouns.
24. I haven’t found any example on Google of the verbs virti ‘to cook’, valgyti ‘to eat’ or gerti ‘to
drink’ in different forms occurring in the context with a progressive interpretation (evoked by
kol ‘while’) and the IPG . Furthermore, such constellations seem to be unattested in the Lithu-
anian Corpus at http://tekstynas.vdu.lt/tekstynas/ (it is not tagged, which is why only around
300 hits of these verbs could be checked).
appropriate for the situation has been consumed). In order to commit oneself to
exhaustiveness, accusative has to be used:
(48) Jis iš-gėrė al-ų
3.nom.sg.m tel-drink.prt.3 beer-acc.sg
‘He drank up his beer.’
Another type of unbounded event (imperfective viewpoint) is the habitual read-

ing of the verb. As expected, we find it ungrammatical with the IPG, even though
the verb valgyti ‘to eat’ combines with the IPG elsewhere:
(49) Jis visada valgė tik ??bandeli-ų/bandel-es
3.nom.sg.m always eat.pst.3 only ??roll-gen.pl/roll-acc.pl
‘He always ate only rolls.’
I summarize: The IPG has the bounded reading only when interacting with the as-
pectual and actional properties of incremental-theme predicates. It is, therefore,
compatible with the bounded (readings of) predicates only and occurs only in the
context of the perfective viewpoint. If the verb is marked as unbounded or con-
textually disambiguated as such, the IPG case marking of the object is typically
blocked, and the structural case marking has to be used.
The data presented so far may give the impression that it is the predicate (or,
more precisely, some predicate quantifier) that binds the quantificational value
of the IPG, which per se should be just regarded as vague in such a case. This
explanation, however, would not be able to account for the fact that the same
predicates in an imperfective context (progressive, habitual, generic) do not have
this ability. One would have expected that the imperfective context would display
the unbounded reading of the IPG. Recall that the latter is attested NP-internally
(see Section 2.2.2 above).
The following examples demonstrate that the bounded reading of the IPG can
provide viewpoint disambiguation of an incremental-theme predicate. While ac-
cusative can in principle occur in both perfective and imperfective contexts, the
IPG highlights only the perfective reading of an incremental-theme verb. Con-
sider the following example, where valgė ‘ate/was eating’ is in itself ambiguous as
to its viewpoint interpretation:
(50) Jis valgė bandeli-ų
3.nom.sg.m eat.pst.3 roll-gen.pl
‘He ate rolls (a while and stopped).’ (perfective/bounded)
*‘He was eating rolls.’ (*imperfective/unbounded)
While the verb valgė ‘ate/was eating’ is ambiguous between the bounded and the
unbounded reading as long as the object is marked accusative, the IPG in (50) dis-
ambiguates the bounded reading, while blocking the unbounded one, cf. progres-
sive in (42) and habitual in (49) above which are both incompatible with the IPG.
Thus, we have observed that the IPG disambiguates the morphologically am-
biguous verb valgė ‘ate/was eating’, which cannot be interpreted as unbounded
in the presence of the IPG object marking. This can additionally be illustrated
with the IPG objects yielding temporary transfer interpretation – a phenomenon
attested also elsewhere in the Eastern Circum-Baltic Area (cf. Seržant, forthcom-
ing-a). In eastern Lithuanian dialects, the IPG may override the accusative object
of the verbs of transfer in order to encode that the result of transfer is to be tempo-
rally delimited (Ambrazas et al. 1976: 25), i.e., bounded in our terms. Contrast the
IPG marked object in (51) with the implication of temporariness of the results of
the transfer and the canonical accusative object in (52) with no such implication:
(51) Duok man peiliuk-o (Eastern Lithuanian)
give.imp 1sg.dat knife-gen.sg
‘Give me a/the knife for a moment!’ (from Jablonskis 1957: 578)
(52) Duok man peiliuk-ą
give.imp 1sg.dat knife-acc.sg
‘Give me a/the knife!’
The transfer verbs are typically achievements that do not presuppose a process
phase (like activities or like accomplishments with their pre-phase) that the im-
plicit quantifier of the IPG could quantify over. The resultant state, i.e. the after-
effects, is the only phase with these verbs that is lasting and hence has potential to
be measured. This is why the implicit quantifier applies exactly at this stage of the
action. Crucially, here as in the examples above, we find only the bounded reading
of the IPG projected on the whole event.
Notably, the temporary transfer construction displays a typologically quite
rare constellation whereby a quantifier that is formally a D(eterminer)-quanti-
fier being hosted in the NP or even more precisely in N acts functionally as an
A(dverb)-quantifier quantifying over the whole event while in itself being affected
holistically. Such a quantifier seems to be unattested, cf. the overview in Corbett
(1994: 202; 2000: 251).25
Table 1 summarizes the results for instances of interaction between the IPG
and the aspectual and actional properties of the verb. A somewhat simplified
25. As Wayles Browne (p.c.) pointed out the English adjective occasional may be somewhat par-
allel. Thus a pregnant woman may ask Is it OK to have an occasional drink during pregnancy?
Table 1. The Lithuanian IPG and its interaction with aspect and actionality of the verb
with a comparison to the Finnish partitive
Non-culminating Culminating
Imperfective Perfective
Lithuanian Acc Part Acc
Finnish Part Part Acc
Table 2. The denotation of the Lithuanian IPG and the Finnish partitive in aspectually
relevant contexts
Totality/culmination Bounded/perfective
Lithuanian IPG no yes
Finnish partitive no yes & no
picture of the Finnish partitive case (both the Finnish partitive case and the Lith-
uanian IPG are labeled as Part in Table 1) is added for comparison (see Seržant,
forthcoming-b, for an areal comparison).
As can be observed from the Table 1 above, in Lithuanian the aspectually rel-
evant IPG is sensitive to viewpoint aspect boundedness. The Finnish partitive is
more sensitive to totality/P-property (see Huumo 2010), or boundedness2 of Sasse
(2002), while both the IPG and the partitive case entail no commitments as to
totality and are most naturally and frequently interpreted as non-total (recall that
in Lithuanian this is valid only for incremental-theme verbs and (dialectally) for
temporary transfer) (see Table 2).
3.3 Iterated events and the IPG
Finally, remarks on the object marking of iterated events are in order. I mentioned
above that a habitual event, being in itself unbounded, is compatible with accu-
sative only and blocks the IPG, because the latter consistently has the bounded
reading when interacting with aspect/actionality of the verb. This is, however, not
the whole story. There are iterative and habitual predicates that tend to be used
with the IPG rather than with the accusative. While more research on this point is
needed, the alternation between the IPG and accusative is anchored in the seman-
tic composition of the iterative and habitual events. These events can naturally be
decomposed into two components: the underlying simple event and an operator
that multiplies the simple event into a habitual or iterative event. While the over-
all iterative or habitual event is unbounded, the simple events in the scope of the
operator need not be so. Thus, the habitual event in (54) contains a multiplication
of the simple event in (53) which is itself bounded:
(53) Jon-as iš-gėrė kav-os
John-nom tel-drink.pst.3 coffee-gen
‘John drank some coffee.’
(54) Ryt-ais Jon-as visada
morning-ins.pl John-nom always
(iš-gėrė)/iš-ger-davo kav-os
(tel-drink.pst.3)/tel-drink-pst.hab.3 coffee-gen
‘John used to drink coffee in the morning.’
Different from (54), the habitual event in (55) is a weaker statement than in (54):
while (54) entails (55), the reverse is not true. The event in (55) states only the
habit of drinking:
(55) Ryt-ais Jon-as visada gėrė/gerdavo kav-os
morning-ins.pl John-nom always drank/drink.pst.hab.3 coffee-gen
Recall that telic predicates – if the temporal boundary coincides with the telos –
may use accusative to encode inclusiveness, the definiteness of the object affected.
Thus, a sentence as in (56) multiplies the event in (57) which is, in turn, total
(+ P-property in Dahl’s 1981 terms), and not the one in (53):
(56) Ryt-ais Jon-as visada iš-ger-davo kav-ą
morning-ins.pl John-nom always tel-drink-pst.hab.3 coffee-acc
‘John used to drink up (a certain amount of) coffee in the morning.’
(57) Jon-as iš-gėrė kav-ą
John-nom tel-drink.pst.3 coffee-acc
‘John drank up the coffee.’
In other words, when the simple event that is in the scope of the multiplier is
bounded, then the speaker uses the IPG when implying no commitments on ex-
haustiveness and the accusative case in order to communicate exhaustiveness.
However, if the reiterated simple event is not marked as telic or bounded,
then the accusative is more natural (according to the native speakers) and also the
most common option:26
26. I have checked 50 occurrences of the iterative/habitual gerdavo (drink.pst.hab.3) with an

overtly expressed object on Google: 49 were accusative objects against 1 with the IPG marked
object.
(58) Ryt-ais Jon-as visada gėrė/ger-davo

morning-ins.pl John-nom always drink.pst.3/drink-pst.hab.3
kav-ą.
coffee-acc
(59) Jon-as gėrė kav-ą
John-nom drink.pst.3 coffee-acc
‘John drank coffee.’
4. Intensional contexts and negation
In this section I briefly touch upon the determiner facet of the partitive genitive.
The discussion here is unavoidably suggestive rather than conclusive. This is so
because the partitive genitive has been generalized in the contexts of intension-
ality and negation, thus turning into a syntactically rather than a semantically
driven phenomenon. The partitive genitive entered the syntactic valence of the
intensional verbs and became mandatory in the context of negation and in some
other constructions. In spite of this, the original motivation for the partitive geni-
tive here is primarily semantic as is also true, e.g., with some Russian intensional
predicates, cf. ždat’ ‘to wait’, xotet’ ‘to want’ with Acc/IPG alternation as opposed
to the Lithuanian laukti ‘to wait’, norėti ‘to want’ both governing genitive only as
well as with the Russian genitive-under-negation.
It seems that what semantically links the contexts discussed in Section 2 with
the intensional contexts and the context of negation is the unbounded reading of
the IPG. There is a consensus that intensional contexts and negation, i.e., the two
main subgroups here, are, as Partee (2008: 307) states, “conducive to decreased
referentiality”. This has to do with the fact that both the negated and intensional
predicates allow their arguments to have no existential presupposition. In other
words, the referents of the arguments of these verbs need not exist in the par-
ticular discourse model to yield a pragmatically and grammatically coherent
expression.
Intensional contexts are contexts that evoke concepts, not referents, and are
not to be confused with intenTional contexts that represent a specific subset of
intenSional contexts. They are opposed to the extensional contexts which refer to
aspects of the real world (cf. Cruse 2000: 21). Typically, a verb requires its argu-
ments to be extensional or referring expressions. The situation is different with
the so-called intensional verbs. Intensional verbs typically have two readings:
(i) the specific (in terms of scopal specificity, cf. von Heusinger 2002) or trans-
parent reading, and (ii) the opaque, non-specific reading, i.e., with no existential
presupposition (Quine 1960: §32; Zimmermann 1993). The latter has been argued
to be, more precisely, a non-referential, existentially non-committal property-de-
noting reading (Borschev et al. 2007). By way of example consider the following
two English sentences with the intensional verb to plan (from Lyons 1999: 170):
(60) Tom plans to bring up three children on his own
a. they’re horrible brats and I wish him luck.
b. but first he needs to find a woman to bear them for him.
The verb to plan induces (scope) ambiguity in sentences such as (60), which can
be highlighted by the additional contexts in a. or b. The context in (60a) displays
the specific (transparent) reading of the cardinal NP in (60), because it makes the
existence of the object’s referent (i.e., three children) mandatory. The context of
(60b), by contrast, implies a non-existential opaque/narrow-scope reading: the
existence of the object NP is assumed only for the imagined world evoked by the
verb to plan. Or, to be more precise, on the non-existential reading, the proper-
ties of the participant of the imagined (planned) situation are given, but not a real
referent.
It has been claimed in the literature that the intensional verbs allow their object
NPs to have property-denoting reading <e, t> (Zimmerman 1993; van Geenhoven
& McNally 2005; Kagan 2012). The narrow scope found with intensional verbs fol-
lows naturally: only quantified expressions (i.e., instantiations) have the potential
to have wider scope. The partitive genitive on its unbounded reading is yet another
option to encode reference to the concept of a particular kind of the respective NP
while not committing oneself to a particular reference.
The intensional context can be invoked either by the lexical semantics of the
respective verb (e.g., intentional verbs or verbs of request) or by the intentional
component of verbs of movement.27 The following verbs require their objects to
be marked with genitive in Lithuanian: ieškoti ‘to look for’, norėti ‘to want’, siekti
‘to strive for’, trokšti, geisti ‘to desire, wish’, ilgėtis ‘to long for’, etc., cf. the full list in
Ambrazas et al. (2006: 486), cf. also Endzelīns (1951: 558) on Latvian.
(61) Noriu stal-o/stal-ą su keturi-omis kėd-ėmis
want.prs.1sg table-gen.sg/table-acc.sg with four-ins.pl.f chair-ins.pl
‘I want (to have) a table with four chairs.’
27. I will not discuss the experiencer verbs (such as bijoti(s) ‘be afraid of ’, išsigąsti ‘be frightened
of ’) that encode their stimulus with the genitive. I believe (on the basis of comparative evidence
from the ancient Indo-European languages) that this use of the genitive stems from the origi-
nally ablative function. The latter is not directly related to the partitive (family of) readings. In a
similar vein, Genitive objects of such verbs as šalintis ‘to avoid’, saugotis ‘to beware of ’ are rather
related to the originally ablative use of the genitive.
The verbs of request (such as prašyti ‘to ask for’, klausti ‘idem’) may also be inten-
sional, i.e., they do not necessarily require their object to exist:
(62) Aukcione bus parduodamas Hitlerio laišk-as,
kuri-ame jis prašė nuolaid-os automobile-ui
which-loc 3.nom.sg.m ask.pst.3 discount-gen.sg car-dat
‘A letter by Hitler will be auctioned off in which he asked for a discount on a
car.’28
The utterer of (62) does not make any commitments about whether or not there
existed a referent for the IPG NP.
Lithuanian, as well as Latvian folklore and dialectal texts (Endzelīns 1951: 559,
562–563), have a construction with the genitive of purpose, also widely attested in
a number of Russian and Belarusian dialects (Markova 1988: 99, 1989; Seržant,
forthcoming-a). This construction consists of a motion verb and a complement
clause comprising an infinitive and its object in genitive. The matrix verb must
be a verb of motion, while the infinitive is optional and may be omitted yield-
ing one clause (Franks & Lavine 2006; Arkadiev, this volume), contrast (63) and
(64) vs. (65):
(63) Važiavom į mišk-ą egl-ės
go.pst.1pl to forest-acc.sg Christmas_tree-gen.sg
‘We went to the forest to get a Christmas tree.’29
(64) Es iešu sein-a telīt-ēm (Latvian folklore)
1sg.nom go.fut.1sg hay-gen calf-dat.pl
‘I will go to get hay for the calves.’ (Endzelīns 1951: 559)
(65) Einu ruoštis, važiuosiu egl-ės pirkti
go.prs.1sg prepare.inf drive.fut.1sg Christmas_tree-gen.sg buy.inf
‘I will go and prepare myself. I will go to buy a Christmas tree.’30
The genitive-of-purpose is semantically closely related to the genitive of intention

discussed above. In the genitive-of-purpose construction, it is the verb of motion
28. http://www.15min.lt/naujiena/laisvalaikis/ivairenybes/aukcione-bus-parduodamas-
hitlerio-laiskas-kuriame-jis-prase-nuolaidos-automobiliuiberlynas-61-104790#ixzz2Im2-
nVxa8
29. http://eia.libis.lt:8080/archyvas/viesas/20110131122940/http://www.culture.lt/satenai/
?leid_id=729&kas=straipsnis&st_id=3127
30. http://www.supermama.lt/forumas/lofiversion/index.php/t890132-300.html
that adds the intentional component.31 The latter projects the embedded event
and the object’s referent into the domain of modality. Given the modal nature of
the purpose clauses, we expect to find grammatical markers signalling non-refer-
entiality of the purpose subclauses. Thus, e.g., in Russian, the non-referentiality of
the event encoded by the purpose clause is marked by the subjunctive particle -by:
(66) Ja poedu v gorod, čtoby (Russian)
1sg.nom drive.fut.1sg to town in_order_to
kupit’ elk-u
buy.inf Christmas_tree-acc
‘I will drive to town in order to buy a Christmas tree.’
It is likely that the function of the partitive genitive in the genitive-of-purpose

construction is parallel to the function of the subjunctive particle -by in Russian
čto-by ‘in order to’. Recall that one of the functions of the partitive genitive within
the domain of determination is to allow the speaker to make no existential com-
mitments with regard to the referent. Analogically to the original function of the
partitive genitive in this Lithuanian construction, the subjunctive particle -by in
the Russian purpose construction is semantically triggered by the non-reality/
non-referentiality of the whole event encoded by the purpose clause.32
Notably, the canonical, accusative object marking can replace the genitive-
of-purpose assumedly due to the analogy with other matrix predicates that are
not motions and therefore regularly have accusative here. Thus, secondarily, both
options with verbs of motion are available in Lithuanian: accusative and genitive.
As a result, there is no difference in terms of referentiality of the respective NPs,
as the following examples illustrate:
(67) Vien-oje pasak-oje mergait-ė Žaliakepuraitė ėjo
one-loc.sg.f story-loc.sg girl-nom Žaliakepuraitė go.pst.3
aplankyti senel-ę Mild-ą
visit.inf grandmother-acc Milda-acc
‘In one story, Žaliakepuraitė went to visit her grandmother Milda.’33
31. Cf. the periphrastic future formations based historically on the verb of movement ‘to go’ in
English or French.
32. For a formal syntactic account of the genitive-of-purpose clause in Lithuanian see Franks
and Lavine (2006) and Arkadiev (this volume).
33. http://www.jurbarkas.rvb.lt/lt/vaikams/128-literaturinis-rytmetis-qskaityti-balsu-smaguq
(68) Marij-a ėjo aplankyti sav-o giminait-ės Elzbiet-os

Mary-nom go.pst.3 visit.inf refl-gen relative-gen Elisabeth-gen
‘Mary went to visit her relative Elisabeth.’34
Evidently, both examples contain definite NPs in the relevant position, showing
that there is no semantic contrast between the accusative and genitive marking
anymore.
As to the genitive-of-negation, Lithuanian – differently from Russian and
similar to Polish – has grammaticalized the former IPG marking with any predi-
cate that is negated. By grammaticalization I mean here the development of a
grammatical rule that requires genitive case-marking of the object in the con-
text of negation. Originally, the genitive marking of the object had probably been
driven semantically (depending on the referential properties of the object NP,
etc.) and emphasis, but it turned into a semantically ‘empty’ rule in Lithuanian.
We observe thus increase in internal dependency (Haspelmath 2004; cf. also Givón
1979: 208) and semantic bleaching (Traugott 2003).
5. Conclusions
In this paper I have discussed only instances of a syntactically (but not necessarily
semantically) independent partitive genitive in Lithuanian.
I have claimed that the IPG when applied NP-internally attests two readings:
(i) the indeterminate but bounded reading, and (ii) the indeterminate unbounded
reading. The first reading is synonymous to existential indefinite plurals in that it
evokes a delimited set of individuals. These individuals typically have wide scope
and can be anaphorically back-referred. In contrast, the second reading denotes
just a concept, a kind of the respective participant without referring to any of its
instantiations. This reading is found only with atelic verbs, e.g., as subject of būti
‘to be’ or object of pažinti ‘to know’. The kind or concept introduced by this read-
ing may also be anaphorically picked up and have wide scope over a quantifier. In
both cases, the participant encoded by the IPG is always discursively background-
ed and never constitutes the primary information in the discourse.
One could assume that the two readings may be derived from the kind read-
ing, very much in parallel to the Carlsonian generalized account of bare plurals in
English (Carlson 1977). The type of the predicate would provide for disambigu-
ation, since telic predicates never allow for the unbounded reading. On this ac-
count, telic predicates would display the bounded reading due to their entailments
34. http://dievogarba.wordpress.com/2011/07/02/
on the object, while the atelic, stative predicates would allow both. Such an account
seems, however, inappropriate synchronically. The reason is that the two readings
are not complementarily distributed. There is a class of verbs, namely the incre-
mental-theme verbs and the transfer verbs (the latter only in Eastern Lithuanian),
which on their non-culminating reading induced by the IPG could theoretically
be unbounded, e.g., in an imperfective context. Crucially different from, e.g., Finn-
ish, though, these contexts are not grammatical with the IPG in Lithuanian. The
bounded reading of the IPG is the only reading available for the interaction with
the aspectual properties of the verb. The analysis of the IPG and the comparison
to the Finnish partitive highlights the double nature of the intermediate stage be-
tween the telos/culmination and the beginning of an action. This stage crucially
may be either bounded, as displayed, e.g., by delimitatives, or unbounded, as dis-
played by, e.g., the progressive context. While in Finnish this difference is not rel-
evant for case assignment rules, in both cases the object being marked with the
partitive case, it is important for Lithuanian, since Lithuanian can only mark those
non-culminating stages that are perfective.
As regards the intensional contexts, I have only briefly illustrated the main
groups that assign the partitive genitive. Note that the partitive genitive is not
syntactically independent here. These are mainly the verbs of intention, the geni-
tive-of-purpose construction, and the clausal negation: they require the direct
object and unaccusative subject to be case-marked with genitive in terms of a
syntactic dependency with no synchronic traces of the original distribution of
the IPG here.
I have not discussed the genitive-under-negation rule on this occasion. In
fact, the genitive is only mandatory for the object position, while in the subject
position (applies to unaccusative subjects only) an alternation with the nomina-
tive case marking is still possible. This is an intriguing topic and needs a separate
investigation. For the time being, however, the interested reader is referred to the
analogous alternation in Russian, widely discussed in the literature (inter alios,
Babby 2001; Borschev et al. 2007; Krasovitsky et al. 2011; Padučeva 1997, 2005;
Partee & Borschev 2002; Partee 2008; Rakhilina, ed., 2008).
Abbreviations
acc accusative m masculine

aor aorist n neuter
dat dative nom nominative
delim delimitative prefix ppp past passive participle
f feminine perf perfectivizing prefix
fut future pl plural
gen genitive pst past
hab habitual quant quantifying prefix
imp imperative refl reflexive
ins instrumental super superlative
loc locative tel telicizing prefix
References
Aissen, Judith. 2003. Differential Object marking: Iconicity vs. Economy. Natural Language and
Linguistic Theory 21: 435–448. DOI: 10.1023/A:1024109008573
Ambrazas, Vytautas (ed). 2006. Dabartinės lietuvių kalbos gramatika. A Grammar of Modern
Lithuanian. Vilnius: Lietuvių kalbos institutas.
Arkadiev, Peter. 2011. Aspect and Actionality in Lithuanian on a typological background. In
Langues baltiques, langues slaves, Daniel Petit, Claire Le Feuvre, & Henri Menantaud (eds),
61–92. Paris: CNRS Editions.
Babby, Leonard H. 2001. The genitive of negation: a unified analysis. In Annual Workshop on
Formal Approaches to Slavic Linguistics: The Bloomington Meeting 2000 [Formal Approach-
es to Slavic Linguistics 9], Steven Franks, Tracy Holloway King, & Michael Yadroff (eds),
39–55. Ann Arbor: Michigan Slavic Publications.
Bertinetto, Pier Marco. 1997. Il dominio tempo-aspettuale: Demarcazioni, intersezioni, contrasti.
Torino: Rosenberg & Sellier.
Blake, Barry J. 1994. Case. Second Edn. Cambridge: Cambridge University Press.
Borer, Hagit. 2005. Some notes on the syntax of quantity. In Aspectual Inquiries [Studies in
Natural Language and Linguistic Theory 62], Paula Kempchinsky & Roumyana Slabakova
(eds), 41–68. Dordrecht: Springer.
Borschev, Vladimir, Elena V. Paducheva, Barbara H. Partee, Yakov G. Testelets, & Igor
Yanovich. 2008. Russian genitives, non-referentiality, and the property-type hypothesis.
In Formal Approaches to Slavic Linguistics: The Stony Brook meeting [Formal Approaches
to Slavic Linguistics 16), Andrei Antonenko, John F. Bailyn, & Christina Y. Bethin (eds),
48–67. Ann Arbor: Michigan Slavic Publishers.
Bossong, Georg. 1998. Le marquage différentiel de l’objet dans les langues d’Europe. In Actance
et Valence dans les Language de l’Europe, Jack Feuillet (ed), 193–258. Berlin/New York:
Mouton de Gryuter.
Carlson, Greg N. 1977. A unified analysis of the English bare plural. Linguistics and Philosophy
1: 413–457. DOI: 10.1007/BF00353456
Champollion, Lucas. 2010. Parts of a whole: Distributivity as a bridge between aspect and mea-
surement. Ph.D. dissertation, University of Pennsylvania.
Comrie, Bernard. 1978. Ergativity. In Syntactic Typology: Studies in the Phenomenology of Lan-
guage, Winfred Lehmann (ed), 329–394. Austin: University of Texas Press.
Corbett, Greville. 1994. Systems of grammatical number in Slavonic. Slavonic and East Eu-
ropean Review 72(2): 201–217. A revised version of: Systems of grammatical number
in Slavonic. In Studies in Number and Quantification, David Gil (ed), European Science
Foundation Programme in Language Typology: Theme 7, Noun Phrase Structure: Work-
ing Paper no. VII/19.
Corbett, Greville. 2000. Number [Cambridge Textbooks in Linguistics]. Cambridge/New York:
Cambridge University Press. DOI: 10.1017/CBO9781139164344
Cruse, Alan. 2000. Meaning in Language. An Introduction to Semantics and Pragmatics. Oxford:
Oxford University Press.
Dahl, Östen. 1981. On the definition of the telic-atelic (bounded-unbounded) distinction. In
Syntax and Semantics, vol. 14: Tense and Aspect, Philip J. Tedeschi & Annie Zaenen (eds),
79–90. New York: Academic Press.
Dahl, Östen. 1984. Perfectivity in Slavonic and other languages. In Aspect Bound. A Voyage Into
the Realm of Germanic, Slavonic and Finno-Ugrian Aspectology, Casper de Groot & Hannu
Tommola (eds), 3–22. Dordrecht/Cinnaminson: Foris Publications.
Declerck, Renaat. 1989. Boundedness and the Structure of Situations. Leuvense Bijdragen 78:
275–308.
Declerck, Renaat. 1991. A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha.
Depraetere, Ilse. 1995. On the necessity of distinguishing between (un)boundedness and
(a)telicity. Linguistics and Philosophy 18(1): 1–19. DOI: 10.1007/BF00984959
Doetjes, Jenny S. 1997. Quantifiers and Selection. On the distribution of quantifying expres-
sions in French, Dutch and English. The Hague: Holland Academic Graphics. (https://
openaccess.leidenuniv.nl/handle/1887/19731)
Dowty, David. 1991. Thematic Proto-Roles and Argument Selection, Language 67: 547–619.
Endzelīns, Jānis. 1951. Latviešu valodas gramatika [Grammar of the Latvian Language]. Riga:
Latvijas valsts izdevniecība.
Filip, Hana. 1989. Aspectual Properties of the AN-Construction in German. In Tempus – As-
pekt – Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen
[Tense – Aspect – Mood. Lexical and Grammatical Forms in the Germanic Languages]
[Linguistische Arbeiten 237], Werner Abraham & Theo Janssen (eds), 259–292. Tübingen:
Max Niemeyer Verlag.
Franks, Steven & James E. Lavine. 2006. Case and word order in Lithuanian. Journal of Linguis-
tics 42(1): 239–288. DOI: 10.1017/S0022226706003896
Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press.
Haspelmath, Martin. 2004. On directionality in language change with particular reference to
grammaticalization. In Up and down the cline: The nature of grammaticalization [Typologi-
cal Studies in Language 59], Olga Fischer, Muriel Norde, & Harry Perridon (eds), 17–44.
Amsterdam: John Benjamins.
Holvoet, Axel. 1991. Transitivity and Clause Structure in Polish. Warsaw: Slawistyczny Ośrodek
Wydawniczy.
de Hoop, Helen. 2003. Partitivity. In The Second Glot International State-of-the-Article Book,
Lisa Cheng & Rint Sybesma (eds). Berlin/New York: Mouton de Gruyter.
Huumo, Tuomas. 2010. Nominal aspect, quantity, and time: The case of the Finnish object.
Journal of Linguistics 46: 83–125. DOI: 10.1017/S0022226709990223
Jablonskis, Jonas. 1957. Rinktiniai raštai. I tomas. Vilnius: Valstybinė grožinės literatūros
leidykla.
Kagan, Olga. 2005. Genitive case: A modal account. In Proceedings of Israel Association for
Theoretical Linguistics 2, Yehuda Falk (ed).
Kagan, Olga. 2012. Semantics of Genitive Objects in Russian. A Study of Genitive of Negation and
Intensional Genitive Case. Dordrecht/Heidelberg/New York/London: Springer.
Keenan, Edward L. & Denis Paperno. 2012. Overview. In Handbook of Quantifiers in Natural
Language, Edward L. Keenan & Denis Paperno (eds), 941–950. Dordrecht: Springer. DOI:
10.1007/978-94-007-2681-9_18
Kiparsky, Paul. 1998. Partitive case and aspect. In The projection of arguments. Lexical and
compositional factors, Miriam Butt & Wilhelm Geuder (eds), 265–307. Stanford: CSLI
Publications.
Kittilä, Seppo, Jussi Ylikoski, & Katja Västi (eds). 2011. Case Animacy and Semantic Roles
[Typological Studies in Language 99]. Amsterdam/Philadelphia: John Benjamins. DOI:
10.1075/tsl.99
Krifka, Manfred. 1989. Nominalreferenz und Zeitkonstitution. Zur Semantik von Massentermen,
Pluraltermen und Aspektklassen. München: Wilhelm Fink Verlag.
Krasovitsky, Alexander, Matthew Baerman, Dunstan Brown, & Greville G. Corbett. 2011.
Changing semantic factors in case selection: Russian evidence from the last two centuries.
Morphology 21: 573–592. DOI: 10.1007/s11525-010-9178-x
Koptjevskaja-Tamm, Maria & Bernhard Wälchli. 2001. The Circum-Baltic languages: An areal-
typological approach. In Circum-Baltic Languages: Typology and Contact vol. 2, Östen Dahl
& Maria Koptjevskaja-Tamm (eds), 615–750. Amsterdam/Philadelphia: John Benjamins.
Kuryłowicz, Jerzy. 1971. Słowiański genetivus po negacij. In Sesja naukowa międzynarodowej
komisji budowy gramatycznej języków słowiańskich, Stanisław Urbańczyk (ed), 11–14.
Wrocław: Ossolineum.
Lestrade, Sander & Helena de Hoop. 2011. On case and tense: the role of grounding in differen-
tial case marking. Unpublished manuscript, Radboud University Nijmegen.
Löbner, Sebastian. 1985. Natürlichsprachliche Quantoren – Zur Verallgemeinerung des Be-
griffs der Quantifikation. Studium Linguistik 17/18: 79–113.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. DOI:
10.1017/CBO9780511605789
Markova, Nina. 1988. Roditel’nyj padež prjamogo objekta v russkom fol’klore Karelii [Direct
object genitive in the Russian folklore of Karelia]. Jazyk russkogo fol’klora [The Language
of Russian Folklore], E. B. Artemenko (ed), 96–104. Petrozavodsk: Petrozavodskij gosu-
darstvennyj universitet.
McNally, Louise. 1998. Existential sentences without existential quantification. Linguistics and
Philosophy 21: 353–392. DOI: 10.1023/A:1005389330615
Mehlig, Hans Robert. 2006. Glagol’nyj vid i vtoričnaja gomogenizacija oboznačajemoj situacii:
K upotrebleniju delimitativnogo sposoba dejstvija v russkom jazyke [Verbal aspect and
secondary homogenization of the situation: On the use of the delimitative aktionsart in
Russian]. In Semantika i Struktura slavjanskogo vida 4 [Semantics and Structure of Slavic
Aspect 4], Volkmar Lehmann (ed), 235–276. München: Sagner.
Metslang, Helle. 2001. On the developments of the Estonian aspect. In Circum-Baltic Languages
vol. 2: Grammar and Typology, Östen Dahl & Maria Koptjevskaja-Tamm (eds), 443–479.
Neidle, Carol. 1988. The Role of Case in Russian Syntax. Dordrecht: Kluwer Academic Publish-
ers. DOI: 10.1007/978-94-009-2703-2
Padučeva, Elena V. 1997. Roditel’nyj sub”ekta v otricatel’nom predloženii: sintaksis ili seman-
tika? [Subject Genitive of Subject in a negated sentence: syntax or semantics?]. Voprosy
Jazykoznanija 2: 101–116.
Paducheva, Elena V. 1998. On non-compatibility of Partitive and Imperfective in Russian. The-
oretical Linguistics 24(1): 73–82. DOI: 10.1515/thli.1998.24.1.73
Padučeva, Elena V. 2005. Ešče raz o genitive sub”ekta pri otricanii [Once again about the subject
genitive with negation]. Voprosy Jazykoznanija 5: 84–99.
Partee, Barbara H. 1986. Noun phrase interpretation and type shifting principles. In Stud-
ies in discourse representation theory and the theory of generalised quantifiers, Jeroen
Groenendijk, Dick de Jongh, & Martin Stokhof (eds), 115–143. Dordrecht: Foris.
Partee, Barbara H. 1995. Quantificational Structures and Compositionality. In Quantification
in Natural Languages, Emmon Bach, Eloise Jelinek, Angelika Kratzer, & Barbara H. Partee
(eds), 541–601. Dordrecht: Kluwer Academic Pulishers.
Partee, Barbara H. 2008. Negation, intensionality, and aspect: Interaction with NP semantics.
In: Theoretical and Crosslinguistic Approaches to the Semantics of Aspect [Linguistik Aktuell
110], Susan Rothstein (ed), 291–317. Amsterdam: John Benjamins.
Partee, Barbara H. & Vladimir Borschev. 2002. Genitive of negation and scope of negation in
Russian existential sentences. In Annual Workshop on Formal Approaches to Slavic Lin-
guistics: The Second Ann Arbor Meeting 2001 [Formal Approaches to Slavic Linguistics 10],
Jindrich Toman (ed), 181–200. Ann Arbor: Michigan Slavic Publications.
Partee, Barbara H. & Vladimir Borschev. 2004. The semantics of Russian genitive of negation:
The nature and role of perspectival structure. In Proceedings of Semantics and Linguistic
Theory 14, Robert B. Young (ed), 212–234. Ithaca NY: CLC Publications.
Paykin, Katia. Forthcoming. Russian Partitives and Verbal Aspect. In Partitivity, Silvia Luraghi
& Tuomas Huumo (eds). Berlin: De Gruyter.
Pesetsky, David M. 1982. Paths and categories. Ph.D. dissertation, MIT, Cambridge.
Quine, Willard van Orman. 1960. Word and Object. Cambridge MA: MIT Press.
Rachilina, Ekaterina V. (ed). 2008. Ob”ektnyj genitiv pri otricanii v russkom jazyke. Moscow:
Probel.
Sasse, Hans-Jürgen. 2002. Recent activity in the theory of aspect: Accomplishments,
achievements, or just non-progressive state? Linguistic Typology 6: 199–271. DOI:
10.1515/lity.2002.007
Seržant, Ilja A. 2012a. Morphosyntactic properties of the partitive genitive in the sub-
ject position in Ancient Greek. Indogermanische Forschungen 117: 187–204. DOI:
10.1515/indo.2012.117.2012.187
Seržant, Ilja A. 2012b. Pragmatics and Semantics of the bare Partitive Genitive in Ancient
Greek. Sprachtypologie und Universalienforschung 65(2): 113–136.
Seržant, Ilja A. Forthcoming a. Independent partitive genitive in North Russian. To appear in
Contemporary approaches to dialectology: The area of North, Northwest Russian and Be-
larusian vernaculars [Slavica Bergensia 13], Ilja A. Seržant & Björn Wiemer (eds), 2014.
Bergen: University of Bergen.
Seržant, Ilja A. Forthcoming b. Independent partitive as a Circum-Baltic isogloss. Forthcoming

paper.
Smith, Carlotta. 1997. The Parameter of Aspect. 2nd edition. Dordrecht: Kluwer. DOI:
10.1007/978-94-011-5606-6
Tatevosov, Sergei & Mikhail Ivanov. 2009. Event structure of non-culminating accomplish-
ments. In Cross-linguistic Semantics of Tense, Aspect and Modality, Lotte Hogeweg, Helen
de Hoop & Andrej Malchukov (eds), 83–130. Amsterdam/Philadelphia: John Benjamins.
Tenny, Carol. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer.
DOI: 10.1007/978-94-011-1150-8
Timberlake, Alan. 2004. A Reference Grammar of Russian. Cambridge: Cambridge University
Press. (http://www.dilbilim.info/doktora/Belgeler/Actionality.pdf)
Traugott, Elisabeth 2003. Constructions in Grammaticalization. In The Handbook of Historical
Linguistics, Brian Joseph & Richard Janda (eds), 624–646. Malden, MA: Blackwell. DOI:
10.1002/9780470756393.ch20
Vendler, Zeno. 1957[1967]. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press.
Verkuyl, Henk J. 1972. On the Compositional Nature of the Aspects. Dordrecht: Reidel. DOI:
10.1007/978-94-017-2478-4
van Geenhoven, Veerle & Louise McNally. 2005. On the property analysis of opaque comple-
ments, Lingua 115(6): 885–914. DOI: 10.1016/j.lingua.2004.01.012
Zimmermann, Thomas Ede. 1993. On the proper treatment of opacity in certain verbs. Natural
Language Semantics 1(2): 149–179. DOI: 10.1007/BF00372561
On the non-canonical marking
of the highest-ranking argument
in Lithuanian and Icelandic
Steps toward a database
Björn Wiemer*, *** and Valgerður Bjarnadóttir**, ***

* Johannes Gutenberg University of Mainz / ** Stockholm University /
*** Vilnius University
This paper is a first report on an ongoing project aiming at building up a da-

tabase of non-canonical argument marking in Lithuanian in contrast to other
languages with relatively rich systems of morphological cases. The language
with which we begin the comparison is Icelandic. The overarching aim consists
not only in a unified inventorisation of relevant units, but in disclosing (i) regu-
larities in the alternation of coding patterns and (ii) the factors underlying such
variation. We will concentrate on case marking; this however implies agreement
patterns as well, insofar as in these two languages only nominatival NPs trigger
agreement on the finite verb (= predicate of the clause). We further concen-
trate on verbs with non-canonical marking of their highest-ranking argument
(HRA); the notion of HRA is defined on the basis of the Actor-Undergoer
Hierarchy as developed in Role & Reference Grammar (see 2.1). What we pres-
ent here is not yet a comprehensive description of the whole body of data, but
rather a problem paper which details some conceptual issues and highlights
some prominent types of argument coding and their relation to lexical groups.
1. Sketch of the project1
The main purpose of our investigation consists in a comprehensive picture of

the case assignment of core arguments in Lithuanian and Icelandic, which in
1. We appreciate the helpful comments by Nicole Nau and an anonymous reviewer. We are
furthermore obliged to Kristina Lenartaitė for her contentful comments on an earlier version
of this paper and to Wayles Browne for his comments and thorough language-editing. We fur-
thermore want to thank Jurgis Pakerys, Birutė Rivitytė, Aurelija Usonienė, Benita Riaubienė
302 Björn Wiemer and Valgerður Bjarnadóttir
some way or another deviate from the predominant coding pattern both within
these languages and known in general for languages with a nominative-accusa-
tive alignment. The features that are somehow associated with (or which condi-
tion) the specific flagging of arguments will be gathered in a database, for which
the relevant predicative units will be given entries according to unified criteria.
Let us stress immediately that we want to capture a maximally comprehensive
inventory of such patterns and of their alternations for each language. On this
basis we want to establish which patterns are productive, which are recessive (in
diachronic terms of type frequency); for the less productive patterns we want to
provide exhaustive inventories of predicative items. This implies that we will reg-
ister not only verbs, but also adjectives and other stems with nominal inflections
if they are used predicatively. However, in this first account the predicative use of
words other than verbs will remain out of consideration.
This descriptive task, in turn, will give a more convenient possibility to sur-
vey and test assumptions related to the nature of, and motives for, non-canonical
marking of arguments. Since we have begun doing this for only two, albeit not
closely related, languages, we want to provide more subtle criteria, test them and
give some in-depth case studies, once we have grasped the principal network of
factors influencing (or conditioning) specific marking patterns. By the same to-
ken, such an investigation can serve the purpose of a pilot study for analogous
research extended to other languages (in parallel to other projects with similar
objectives, on which see 1.2). Last not least, after we will have understood enough
about the contemporary patterns and alternations of non-canonical argument
coding in Lithuanian and Icelandic we may attempt to trace back diachronic
changes in these patterns, as far as corpus data for earlier stages of these languages
and already existing investigations allow us to do so.2
The article is structured in the following way. In this section we sketch the
project; we also indicate in which way it differs from ongoing similar projects.
Section 2 gives information on the theoretical background pertaining to the syn-
tax-semantics interface. In Section 3 we present a scaffolding for the analysis of
the relevant data which we have gathered so far for Lithuanian and Icelandic. First
of all, we will be concerned with identifying the main coding patterns (including
and especially Auksė Razanovaitė and Rolandas Mikulskas for their unfailing replies. Needless
to say, all remaining mistakes and inconsistencies are exclusively our own responsibility.
2. In this respect the situation is much better for Icelandic than for Lithuanian. Cf. such work
as Andrews (1976), Thráinsson (1979), Zaenen, Maling and Thráinsson (1985), Sigurðsson
(1989), Jónsson (1996), Barðdal (2001, 2006), Eythórsson and Barðdal (2005), but also the fact
that for Icelandic from the start we had a list of verbs with ‘quirky subjects’ at our disposal
which accounted for earlier stages of the language. See further 3.1.
On the non-canonical marking of the highest-ranking argument 303
their alternations) insofar as they deviate from a canonical nominative-accusative

pattern, with tendencies toward lexical groupings and with the main differences
between both languages which come to the fore already in a more superficial anal-
ysis. In addition, we point at some phenomena which, to our knowledge, have not
received due attention in studies on variation in the flagging of arguments within
and across languages. Section 4 sums up these observations and gives an outlook.
1.1 On the choice of Lithuanian and Icelandic and the framework
A comparison of non-canonical argument marking in Lithuanian and Icelandic

is profitable, among other things, for the following reasons. First, although the
two languages are not closely related in genealogical terms, they have inherited
and retained several morphological cases; it is therefore as a rule possible to dis-
cern core arguments not only on the basis of their behavioural properties, but
also because of their morphological coding (according to the well-known subject
properties after Keenan 1976). Second, Lithuanian and Icelandic have never been
in contact with each other. They belong to different parts of Europe and are, as
it were, separated from each other by languages with more typical SAE features.
Differences and coincidences in core argument marking can thus be considered
as not being induced by contact, and the chance of areal diffusion that might
have affected both languages in this domain of alignment appears to be extremely
low. Third, the IE heritage of morphological cases has been retained to a differ-
ent extent: Lithuanian has six cases,3 Icelandic only four. Thus, the possibilities of
variation for non-canonical marking based on oblique cases are more restricted
in Icelandic than in Lithuanian. However, we may remark at once that despite the
more reduced number of cases in Icelandic, adpositional marking of arguments
does not turn out to be more prominent than in Lithuanian (see 3.2.2). Thus, ad-
positional marking of core arguments turns out to be rare in both languages.
1.2 Differences in comparison to similar projects
Our investigation shows parallels with at least two other ongoing projects con-
cerned with non-canonical argument marking and the question to which extent
non-canonical marking is motivated lexically. However, both of them differ from
ours in the way data are accessed, the sort of data and the typological and areal
range of comparison.
3. We disregard the vocative, since its functions are unrelated to the coding of arguments or
adverbials.
First, clear parallels exist with the St. Petersburg project headed by Sergej Saj
and entitled “Verb argument structure variation and verb classification in lan-
guages of various structural types” (Russ. “Variativnost’ argumentnoj struktury
glagolov i klassifikacija glagol’noj leksiki v raznostrukturnyx jazykax”; cf. Saj 2011;
Saj, forthcoming). Its particular research issues are the following ones: The object
of investigation is only two-place predicates, with clausal arguments having been
excluded from the start. Only verbs are considered. Among them verbs have also
been included whose argument ranking highest in terms of agentivity (our HRA,
see below) is coded with the nominative. The distribution of x- and y-argument
of two-place predicates was established on the basis of an animacy-agentivity hi-
erarchy (Saj 2011: 427f.), not on the basis of logical decomposition. However, the
results of these different criteria of building argument hierarchies in many cases
coincide with ours (e.g., for verbs like ‘like, please’, Lith. patikti, Russ. nravit’sja:
x=dat, [+human] vs. y=nom/clause, [±human]). Further research will show how
far such coincidences go. In any case, almost all lexical concepts which according
to Saj (2011: 428) are commonplace for predicates with a non-canonically marked
human x-argument, occur in Lithuanian (‘please, like’, ‘ache, hurt’, ‘lack’, ‘suffice, be
enough’, ‘need’, ‘remain’); see the relevant subsections of Section 3. Furthermore,
in the Petersburg project the basis of the selection of predicative units was a ques-
tionnaire with 130 lexical concepts,4 for which translational equivalents in other
languages were requested from informants. The crosslinguistic basis of compari-
son is rather broad: at a first stage, responses concerning 16 languages were col-
lected. Among them we find Lithuanian, but not Icelandic.
The second project with which our investigation shows some affinity is the
“Leipzig Valency Classes Project” (Malchukov et al., cf. http://www.eva.mpg.de/
lingua/valency/files/project.php). This is a large-scale project on the typology of
verb classes defined on the basis of their valency properties (inspired by Apresjan
1967 on Russian and Levin 1993 on English). It is much broader in terms of cod-
ing properties, since – apart from its typological range – it does not concentrate
specifically on non-canonical marking, but aims at a broad typology of valen-
cy-changing alternations (ranging from unmarked dative shift, as in English, to
voice-changing operations).5 Data was collected via questionnaires “for a select-
ed sample of 70 verbs. These verbs are conceived of as representative of the ver-
bal lexicon and have been reported in the literature to show distinctive syntactic
4. Russ. “predikatnye smysly” (lit. ‘predicative senses’), supplied in the questionnaire as core
clauses with the respective predicative and its arguments in Russian.
5. Most interesting (to the Leipzig Project as well as to us) are “alternations which contribute
the most to verb classification in the sense that they are neither restricted to few verbs, nor ap-
ply across the board” (from the website).
behaviour both within and across languages” (from the website). Here, converse-
ly, Icelandic is in the sample, but Lithuanian is not.
By comparison: Our project – apart from, on its initial stage, being restricted
to Lithuanian and Icelandic – does not start from an onomasiological perspective,
and the items have not been collected with the aid of questionnaires. Instead, we
have collected items from dictionaries and already published material (see 3.1) on
the basis of their coding properties for the Highest Ranking Argument (HRA), a
notion which we derive from the Actor-Undergoer Hierarchy (AUH) of Role &
Reference Grammar (see 2.1). The criterion was that the HRA was not marked
with the nominative; the coding of other arguments was used only as a criterion
of subclassification (see 3.2). Lexical (and other) groups were established only
after predicative units had been extracted. Furthermore, from the start we have
divided the items to be investigated by parts of speech, although in this contribu-
tion we will deal only with verbs. Moreover, at later stages we also want to account
for clausal arguments, which have been excluded from the Petersburg project.
1.3 Delimitation of units assembled in the database
In this first account we limit our interest only to units for which the HRA is
marked other than with the nominative. In general, non-canonical core argument
marking in languages with predominant nom–acc alignment6 applies when the
most agentive argument is not marked with the nominative and/or the most pa-
tientive argument is not marked with the accusative. Thus, a subcase of non-ca-
nonical marking obtains when the single argument of a one-place predicate (S in
Dixon’s 1994 terms) or the most agentive argument of a two- or three-place pred-
icate (Dixon’s A) is not marked with the nominative. In Lithuanian and Icelandic
this subcase comprises the following patterns:
one-place
Lithuanian
(1) Mokykloje padaugėjo mokinių. (single argument in gen)
school.loc increase.pst.3 pupil.gen.pl
‘In school, the number of pupils has increased.’ (lit. ‘… there have increased
(the) pupils.’)
6. Probably one hardly finds a language with only one consistent alignment pattern, even if
only coding patterns in unmarked voice are considered (cf. Haspelmath 2011: 552–558). We
therefore speak of predominant nom-acc-alignment if these cases are used in the unmarked
voice of prototypical transitive predicates (‘kill’, ‘break’) and also otherwise apply prominently.
This prevalent pattern is characteristic of the ‘European standard’, to which both Lithuanian
and Icelandic conform despite the groups of non-canonical case patterns that are the topic of
this article.
(2) Joną purto (nuo šalčio). (single argument in acc)

pn.acc shaking.prs.3 from frost.gen
‘Jonas is shaken (from the cold.)’ (lit. ‘(it) is shaking Jonas …’)
(3) Jonui pagerėjo. (single argument in dat)
pn.dat become_better.pst.3
‘Jonas became better.’ (lit. ‘(it) became better to Jonas.’)
(4) Nuo ežero padvelkė vėsum-u. (single argument in ins)7
from lake.gen blow.pst3 freshness.ins
‘It (suddenly) blew with freshness from the lake.’
Icelandic
(5) Vinds gætti. (single argument in gen)
Wind.gen notice.pst.3sg
‘There was a little wind.’
(6) Mig syfjar. (single argument in acc)
I.acc get_sleepy.prs.3sg
‘I am getting sleepy.’
(7) Mér leiddist. (single argument in dat)
I.dat bore.pst.3sg
‘I was bored.’
two-place
Lithuanian
(8) Eglė šlykštisi aplinkos melagingumu.(nom–ins)
pn.nom loathe.prs.3 surrounding.gen mendacity.ins
‘Eglė loathes the mendacity of the/her environment.’
(9) Žmonės išsigando kraujo(nom–gen)
people.nom become_scared.pst.3 blood.gen
‘People became scared of the blood.’
(10) Jonas norėjo, kad Eglė ateitų.(nom–fin.compcl)
pn.nom want.pst.3 comp pn.nom come.sbjv.3
‘Jonas wanted Eglė to come [lit. … that Eglė come].’
(11) Jonas norėjo susitikti su Egle.(nom–infinitival compl.)
pn.nom want.pst.3 meet.inf with pn.ins
‘Jonas wanted to meet with Eglė.’
7. For a discussion of debatable instances (concerning the number and type of arguments) the
reader is referred to Section 3.2.
(12) Jonas padėjo Eglei. (nom–dat)

pn.nom help.pst.3 pn.dat
‘Jonas helped Eglė.’
(13) Jonas svajojo apie Eglę. (nom–PP)
pn.nom dream.pst.3 about pn.acc
‘Jonas dreamed of Eglė.’
(14) Jonui patiko Eglė.(dat–nom)
pn.dat like.pst.3 pn.nom
‘Jonas liked Eglė. / Eglė pleased Jonas.’
(15) Jonui pagailo Eglės.(dat–gen)
pn.dat take_pity.pst.3 pn.gen
‘Jonas took pity on Eglė.’
(16) Jonui patiko, kad Eglė gražiai apsirengė.
pn.dat like.pst.3 comp pn.nom beautifully dress.pst.3
‘Jonas was pleased that Eglė had dressed beautifully.’ (dat–fin.compcl)
(17) Jį erzino, kad niekas jo neklausė.
he.acc annoy.pst.3 comp nobody.nom he.gen neg.ask.pst.3
‘It annoyed him that nobody (had) asked him.’ (acc–fin.compcl)
Icelandic
(18) Sveinn vill að Sigga komi (nom–fin.compcl)
pn.nom want.prs.3SG comp pn.nom come.sbjv.3sg
‘Sveinn wants Sigga to come.’
(19) Sveinn vill hitta Siggu (nom–nonfin.compcl)
pn.nom want.prs.3SG to meet.inf pn.acc
‘Sveinn wants to meet Sigga.’
(20) Sveinn hjálpaði Siggu(nom–dat)
pn.nom help.pst.3sg pn.dat
‘Sveinn helped Sigga.’
(21) Sveinn þarfnast Siggu (nom–gen)
pn.nom need.prs.3sg pn.gen
‘Sveinn needs Sigga.’
(22) Sveinn hugsar um Siggu(nom–pp)
pn.nom think.prs.3sg of pn.acc
‘Sveinn thinks of Sigga.’
(23) Sveini líkar Sigga (dat–nom)
pn.dat like.prs.3sg pn.nom
‘Sveinn likes Sigga.’
(24) Sveini líkar við Siggu (dat–pp)

pn.dat like.prs.3sg with pn.dat
‘Sveinn likes Sigga.’
(25) Sveini skilst að Sigga komi á morgun
pn.dat understand.prs.3sg comp pn.nom comes.prs.3 tomorrow
‘It seems to Sveinn / Sveinn thinks that Sigga will come tomorrow.’
(dat–fin.compcl)
(26) Sveini leiðist að bíða(dat–nonfin.compcl)
pn.dat bore.prs.3sg to wait.inf
‘Sveinn is bored by waiting.’
(27) Svein dreymir Siggu(acc–acc)
pn.acc dream.prs.3sg Sigga pn.acc
‘Sveinn dreams about Sigga (dream while asleep).’
(28) Svein dreymir um Siggu(acc–pp)
pn.acc dream.prs.3sg um pn.acc
‘Sveinn is daydreaming of Sigga.’
(29) Svein langar að fara (acc–nonfin.compcl)
pn.acc want.prs.3sg to go.inf
‘Svein wants to go.’
(30) Svein langar að Sigga komi (acc–fin.compcl)
pn.acc want.prs.3SG comp pn.nom come.sbjv.3sg
‘Svein wants Sigga to come.’
At least for Lithuanian we may add dative-marked external possessors. The pos-
sessor counts as part of the core and, thus, is treated like an argument in syntactic
terms. However, it is not licensed by the predicate8 and, therefore, its argument
status is at least debatable on a semantic level. See, for instance:
(31) Jonui badė panages.(datposs–acc)
pn.dat pierce.pst.3 finger_tip.acc.pl
lit. ‘(it) prickled the finger tips at him.’ [idiomatic meaning ‘He was nervous /
unquiet.’]
(32) Jonui skauda skrandį / skrandis.(datposs–acc/nom)
pn.dat ache.prs.3 stomach.acc / .nom
lit. ‘(it) hurts the stomach at Jonas.’ = ‘Jonas’ stomach hurts.’
We will dwell upon this later (see 3.2.2.1).
8. Cf. Payne and Barshi (1999: 3), König (2001: 970f.).

Of course, a full account of ‘deviations’ from nom–acc-alignment must in-

clude predicates for which the more agent-like argument is coded with the nom-
inative, but the other, more patient-like is not marked with the accusative. Thus,
one should count as verbs with non-canonically marked arguments, for instance,
bjaurėtis, šlykštėtis (8), padėti (12) with an oblique case other than acc, verbs like
svajoti (13) with a PP, and verbs like norėti which can take a clausal complement
(10)–(11), but with the more agent-like argument in the nominative. Further-
more, Lithuanian has ditransitives demonstrating the well-known alternation
of ‘spray paint’-verbs, which however does not affect the HRA. Finally, certain
groups of Lithuanian two-place verbs show an acc–ins-alternation for the second
argument (lower ranking in terms of AUH),9 and there is a rarer alternation with
two-place verbs in which a nom–acc-pattern alternates with a nom–ins-pattern
with mutually exchanged arguments; cf. Lenartaitė (2011: 187ff.), from where we
cite the following pair of examples:
(33) a. Purvai aptiško drabužius.
dirt.nom.pl sprinkle.pst.3 clothes.acc
‘Dirt sprinkled on the clothes.’
b. Drabužiai aptiško purvais.
clothes.nom sprinkle.pst.3 dirt.ins.pl
‘The clothes were sprinkled with dirt.’
The types mentioned in the last paragraph will be excluded from our analysis,
and we will concentrate only on verbs (and other predicative units) for which
the more agent-like or the single argument lacks nominative coding, thus cases
like (1)–(7) or (14)–(17), (23)–(30). The other argument may be coded with the
nominative, but its way of coding will be relevant only for the subclassification
of the units included into our database, not for the question whether it will be
considered at all.
Issues which instead will concern us during the first stages of the project are:
(i) determination of lexical groups in which non-canonical marking of the HRA
can be found; (ii) their main coding patterns, together with patterns of their al-
ternation. Already at this preliminary stage we have to distinguish external pos-
sessors, particularly in Lithuanian, where they show a frequent overlap with
peripheral argument functions, such as ‘maleficient’ (see 3.2.2.1).
9. Cf. Lenartaitė (2011: 187–189), Anderson (forthc.: Ch. 1) for detailed analyses within dif-
ferent frameworks.
2. Theoretical background
For our purpose we needed to choose an apparatus which in general distinguishes

explicitly between the semantic and the (morpho)syntactic level, and, first of all,
in which the assignment of argument positions would not end in a vicious cir-
cle. Many theories that have worked with lists of case roles (semantic roles, theta
roles) at some point or other end up with arbitrary decisions as for the specific
semantic nature and number of these roles and, more importantly, as for their
hierarchy among each other (cf. Primus 2009: 264f. for a summary; cf. also Dahl
1985). A theory which largely avoids such pitfalls is Role & Reference Grammar
(RRG), and this is the reason why we have decided to use it in the description of
our material, although there are some reference- and clause-related conditions on
case choice10 which have not sufficiently been incorporated into this theory.
Therefore, we first give an introduction to basic notions used within RRG
(2.1) and relevant for our analysis in Section 3. In 2.2 we comment on the relation
of these notions to other, in many respects similar approaches within alignment
typologies.
2.1 Role & Reference Grammar: Some necessary notions
RRG is a functional approach at capturing the syntax-semantics (or even syntax-

pragmatics) interface. It is a lexicalist syntax, insofar as argument positions are
strictly tied to (or: conditioned by) the meaning of predicative elements (verbs
etc.). RRG is monostratal, i.e., it rejects any kind of underlying, or deep, structure.
Instead, a rather straightforward mapping procedure of predicates, arguments
and adjuncts into the Layered Structure of the Clause (LSC),11 and vice versa, is
assumed. As concerns relational structure, RRG, like a couple of other theories,
assumes generalized semantic roles. These are distinguished from grammatical
relations, though they are crucial for the formulation of many language-specific
linking rules between argument structure and the syntax. These generalized roles
are called ‘macroroles’, and there are two of them: Actor und Undergoer.
10. These were surveyed in Haspelmath (2001: 56–59). Since these factors are not easy to cap-
ture in logical decomposition, they yield rather cumbersome formal representations.
11. The LSC consists of Nucleus (occupied by the predicate) and its Core (housing the argu-
ments of the predicate), plus an optional Periphery and language-specific Pre- and Post-Core
slots (cf. Van Valin 2005; Van Valin & LaPolla 1997: 35–40 and passim), which are however of
no relevance for our present concern.
ACTOR UNDERGOER

Arg. of 1st arg. of 1st arg. of 2nd arg. of Arg. of
DO do’ (x, … pred’ (x, y) pred’ (x, y) pred’ (x)
[ ’ = increasing markedness of realization of argument as macrorole]
Figure 1. Actor-Undergoer hierarchy (cit. from Van Valin 2005: 61)
Macroroles are understood as “generalizations across classes of specific argu-

ment positions in logical structure” (Van Valin & LaPolla 1997: 142; cf. also Van
Valin 2005: 60–67). Therefore, arguments are defined via logical decomposition
(see below); they can be arranged along a scale between a most agent-like and a
most patient-like pole. This scale is called the Actor-Undergoer hierarchy (AUH);
see Figure 1.
The difference between DO and do’ is that between a prototypical agent (i.e., a
conscious, controlling instigator of a change affecting or effecting another object,
DO) and any sort of ‘causer’, including non-intended, accidental causation, also
causation by a fact (expressed by a clausal complement, e.g. That he came late /
His late arrival upset everyone). The causer corresponds to the x-argument of the
do’-predicate.
If an Actor is assigned it is always the most agent-like argument available on
that scale; correspondingly, for the Undergoer it is the most patient-like argu-
ment. In this way, RRG does justice to an acknowledged fundamental opposition
between cardinal arguments of a transitive predication (A + O in Dixon’s terms;
see 2.2).12 Most importantly, irrespective of their specific ontological nature (and
the names given to roles like, e.g., ‘agent’, ‘patient’, ‘theme’, ‘goal’, ‘experiencer’,
‘stimulus’) it is the internal hierarchy between arguments on such a scale which
influences the assignment of Actor and Undergoer. This hierarchy is a necessary,
although not always a sufficient condition, since the status of Actor and Under-
goer can also depend on coding and behaviour properties of arguments, which
surface in language-specific pivots and controllers, united under the heading of
Privileged Syntactic Arguments (PSA). In a sense, thus, macroroles themselves
serve a pivotal role in the bidirectional linking algorithm between semantics (sc.
logical structures) and syntax (see Figure 2), and the notion of macroroles is not
entirely free of a hybrid (semantico-syntactic) nature (see 2.2).
12. Cf. Van Valin (1991: 162), where Actor and Undergoer are characterized as “the two pri-
mary arguments of a transitive predication, either one of which may be the single argument of
an intransitive verb”. Cf. furthermore (1991: 162f.): “[Macroroles] correspond to what pretheo-
retically are called ‘logical subject’ and ‘logical object’, but these terms are not used, because
‘subject’ and ‘object’ are normally used to refer to syntactic, not semantic, relations.”
SYNTACTIC FUNCTIONS: PSA Direct core arguments Oblique core arguments
Language-
Privileged syntactic argument [PSA] selection:
specific
Highest ranking MR = default (e.g. English)
Lowest ranking MR = default (e.g. Dyirbal)
SEMANTIC MACROROLES:
ACTOR UNDERGOER
Arg. of 1st arg. of 1st arg. of 2nd arg. of Arg. of
DO do’ (x, … pred’ (x, y) pred’ (x, y) pred’ (x)
Transitivity = No. of macroroles [MRα]
Transitive – 2
Intransitive –
1
Atransitive = 0
Argument position in LOGICAL STRUCTURE
Universal
Verb class Logical structure

STATE predicate’ (x) or (x, y)

ACTIVITY do’ (x, [predicate’ (x) or (x, y)])
ACHIEVEMENT INGR predicate’ (x) or (x, y)
SEMELFACTIVE SEML predicate’ (x) or (x, y)
ACCOMPLISHMENT BECOME predicate’ (x) or (x, y)
ACTIVE ACCOMPLISHMENT
do’ (x, [predicate’1 (x, (y))]) & INGR predicate’2 (z, x) or (y)
CAUSATIVE α CAUSE β, where α, β are LSs of any type
Figure 2. Summary of RRG linking system (cit. from Van Valin 2005: 129)
In the following we will call the highest-ranking argument (henceforth: HRA)

that argument position which is highest in terms of the AUH. This corresponds
to the left-most position within the AUH in the logical structure of the respective
predicative unit: it must be the x-argument of DO, or otherwise the left-most
argument of M-transitive predicates (= x-argument of left-most do’-predicate),
but it can even be the Undergoer with M-intransitive state-predicates (see shortly
below).
The positions on the AUH are determined by logical decomposition based
on primitive semantic notions ‘borrowed’ from Dowty (1979) and dependent on
Vendler’s (1967 [1957]) so-called Aktionsarts: States, Activities, Achievements
and Accomplishments,13 together with a battery of syntactic tests showing (dis)
13. To these four classes Active accomplishments and Semelfactives have been added later; cf.
Van Valin (2005: 32–42). Vendlerian Aktionsarts have to be understood as aspectual classes of
predicates on clause level, as admitted by Van Valin himself, e.g., in Van Valin (1991: 159): “The
ability for progressive, telic, punctual use. The most primitive classes are states
and activities. States may be one- or two-place predicates, symbolized as pred’ (x)
or pred’ (x, y), respectively; Activities contain a do’-predicate. Other aspectual
classes as well as causative predicates are derived from States and Activities with
the help of operators (in capital letters; see lower part of Figure 2 and examples
below); the causative operator conjoins LSs of any kind.14
This way of presenting argument structure and its coding in the syntax has
the following advantages. First, argument positions are ascribed in a non-circular
way, since assignments and their hierarchy do not depend on (and are not in-
fluenced by) some preconceived inherent properties of arguments, first of all of
typical referents associated with animacy or other ontological hierarchies. Labels
like ‘effector’, ‘perceiver’, ‘cognizer’, ‘experiencer’, ‘theme’, ‘stimulus’ etc. (but also
‘agent’ and ‘patient’) remain convenient as a mnemotechnic device, but they do
not have any relevant status for the theory. On the other hand, in many instances
the coding properties of arguments turn out to depend on referential properties
or clause-related conditions like individuation, animacy, definiteness or affirma-
tive/negative predication (cf. Haspelmath 2001: 56–59), which since Hopper and
Thompson (1980) have frequently been shown to influence transitivity on clause
(or utterance) level. So far RRG has hardly accounted for these factors, and they
are not easy to include into linking mechanisms like those in Figure 2. However,
this problem appears to be rather of a technical nature.15 Anyway, for the time
being note that reference- and clause-related conditions like those analysed with
great care in Lenartaitė-Gotaučienė (this volume), Nau (this volume), Seržant
(this volume) have not been paid due attention to so far in RRG.
A second advantage of the approach taken by RRG consists in the following.
The AUH defined on the basis of logical decomposition allows for a partial ex-
planation of why certain groups of predicates show behaviour that deviates from
canonical alignment (nom–acc, as for Lithuanian and Icelandic). In short, non-
canonical marking shows up in two constellations. Either there are no arguments
placed at the ‘edges’ of the AUH, otherwise: the arguments occupy positions in
the AUH which are very close to each other. See marking patterns like those in
(8)–(10) and (14)–(15) above. Or there is only an Undergoer (but no Actor):
Aktionsart of a verb is the basis of its interpretation in a clause, but the interpretation of the
clause as a whole is not always the same as that of the verb alone.”
14. Ditransitives are decomposed into a hierarchy of more primitive one- or two-place predi-
cates, connected by the cause operator.
15. In principle, reference-related conditions can be accounted for in the decomposition of
argument NPs, which is done in the form of qualia structures.
either because there is no other argument at all (see examples (1)–(7)), or because
a second argument closer to the Actor-pole of the AUH does not show PSA-prop-
erties. The latter situation will be discussed in 3.2.2.8. In either instance, the fact
that with such predicates no argument is marked with the nominative, indicates
that for some semantic reason the nominative (as an unmarked structural case)
is avoided if the HRA is not most agentive; this often implies that the HRA does
not show a clear contrast to a second argument in terms of ‘distance’ on the AUH.
From this angle, there is reason to assume that the more a language has predicates
with non-canonically marked core arguments, the more the semantic nature of
the arguments (influenced by their ontological characteristics, the degree of af-
fectedness and other things) prevails over syntactic factors. This corresponds to
the relation between two competing functions of case marking, characterizing vs.
discriminatory functions: the former functions explain why certain semantic roles
tend to be coded in identical manner (thus, those which belong under the same
argument position in Figure 1), whereas the latter functions are useful to explain
why arguments of the same predicate are usually treated differently in the syntax
irrespective of their semantic nature (cf. Primus 2009: 262f.). One may assume
that, in principle, the proportion (type frequency) of verbs with non-canonically
marked arguments relative to the entire stock of verbs in a given language serves
as a good indicator of the degree to which the characterizing functions of oblique
cases prevail over their discriminatory functions (in that same language). English,
for instance, ‘lumps together’ (neutralizes) very many specific semantic roles for
syntactic purposes, whereas languages in the Eastern part of Europe (e.g., Lithu-
anian) in general prove more restrictive in this respect, i.e., semantics often pre-
vails over syntactic contrasts in the coding of arguments.
Regardless of this, argument hierarchies remain valid even for characterizing
functions. By the same token, one can capture the fact that, at least in Lithu-
anian, there are no predicates with co-occurrent dative-marked and instrumen-
tal-marked arguments. This restriction follows from the combination of the two
rules given in [1] and [2] below.
The basis of lexical entries contains their logical decomposition, and the
transitivity of predicative units is captured semantically (not in terms of syntac-
tic valency slots), as it depends on the number of macroroles. Predicates with
two macroroles are M-transitive, if predicates have only one macrorole they are
M-intransitive, and if they lack either macrorole they are called M-atransitive.
Thus, for a verb to count as M-transitive, its LS must contain minimally two core
arguments. However, this is only a necessary condition; in addition, both argu-
ments must behave as macroroles, as with Lith. mėgti ‘like’ in (34). If this condi-
tion is not met, e.g., because the x-argument is not coded with a nom-NP and
does not show any properties of a PSA (Equi-NP-deletion, control of reflexives,
etc.), the verb is classified as M-intransitive (see patikti ‘like, please’ in example
(35)). If, furthermore, neither argument of a two-place verb shows coding and/
or behavioural properties of privileged syntactic constituents, the verb becomes
M-atransitive, as occurs with Lith. reikėti ‘need’ in (36) and Icel. vanta ‘need’ in
(37). If the number of macroroles (and, thus, M-transitivity) cannot be deduced
from the predicate’s LS per se, its lexical entry is marked as [MR 1] (M-intransi-
tive) or [MR 0] (M-atransitive):
Lithuanian
(34) Jolytė mėgsta obuolius.
pn.nom like.prs.3 apple.acc.pl
‘Jolytė likes apples.’
like’ (Jolytė, obuoliai); generalized for mėgti: like’ (x, y)
(35) Jolytei patinka nauja suknelė.
pn.dat like.prs.3 new.nom.sg.f skirt.nom.sg.f
‘Jolytė likes her new skirt. / The new skirt pleases Jolytė.’
like’ (Jolytė, nauja suknelė) [MR 1]; generalized for patikti: like’ (x, y) [MR 1]
(36) Jolytei reikia paramos.
pn.dat need.prs.3 support.gen.sg
‘Jolytė needs support.’
need’ (Jolytė, parama) [MR 0], generalized for reikėti: need’ (x, y) [MR 0]
Analogically for Icelandic:

(37) Siggu vantar aðstoð
pn.acc need.prs.3SG support.acc.sg
‘Sigga needs support.’
As for M-intransitive predicates, the question arises of whether the single macro-
role has to be qualified as Actor or Undergoer. The answer depends on whether
the LS contains an Activity predicate or not: if it does, the macrorole is an Actor,
if not it is an Undergoer.16 For instance, with patikti in (35) we are dealing with
an Undergoer.
In languages with predominant nom–acc-alignment (like Lithuanian and
Icelandic) all predicative units coding their non-HRA with the nominative can
be considered as M-intransitive (see Lith. patikti above). M-atransitive are verbs
like Lith. pagailėti ‘take pity on/feel pity for’ (or pagailti; see example (15)) and,
trivially, one-place verbs for which the single argument is not coded with the
16. See the ‘Default Macrorole Assignment Principles’ in Van Valin and LaPolla (1997: 152f.)
and Van Valin (2005: 63). M-intransitive predicates with an Actor are usually identical with
unergatives, M-intransitive predicates with an Undergoer with unaccusatives.
nominative. As a rule, the single macrorole of M-intransitive predicates which

fall under our inquiry has to be qualified as an Undergoer, because in their logical
structure no primitive do’-predicate occurs. Problematic, however, are ‘swarm’-
and ‘smell’-verbs showing an alternation of loc–nom for the location argument
and of ins–nom for the source argument, respectively (see 3.2.2.6).
In RRG there are also rules for assigning the dative and the instrumental,
respectively. Since they are based on logical decomposition and the AUH, these
rules are claimed to hold irrespective of the alignment type (nom–acc or erg–
abs) and the specific case system of the language at hand. The rule concerning the
dative is very simple (Van Valin 2005: 110):
[1] Assign dative case to non-macrorole arguments (default).
The validity of this rule becomes immediately obvious with ditransitive verbs of
languages with inventories of morphological cases like Lithuanian, Russian, or
German. But it is also quite obvious for all two-place verbs which deviate from
the canonical nom–acc coding pattern included in the lists in 1.3: state-verbs for
which the y-argument is nom-coded (like Lith. patikti, Icel. líka, Germ. gefallen
‘to like, please’) and the x-argument appears in the dative, or, conversely, two-
place verbs for which the x-argument appears in the nominative, but the y-argu-
ment takes the dative (e.g., Lith. padėti, Icel. hjálpa, Germ. helfen ‘help’).
As concerns the instrumental, the rule goes as follows (Van Valin 2005: 110):
[2] Assign instrumental case to non-MR b argument if, given two arguments, a
and b, in a logical structure, with (α) both as possible candidates for a par-
ticular macrorole and (β) a is equal or higher (to the left of b) on the AUH, b
is not selected as that macrorole.
Complicated as this sounds on first sight, it is easy to relate with the structure of
causal event chains, such as described, among many others, by Croft (1998). If
causal event chains are broken down into representations of logical decomposi-
tion, they look as in (38c), which is a generalisation of the LS of sentence (38a) (cf.
Van Valin & LaPolla 1997: 121):
(38) a. Tom is cutting the bread with a knife.
b. LS: [do’ (Tom, [use’ (Tom, knife)])]
CAUSE [[do’ (knife, [cut’ (knife, bread)])]
CAUSE [BECOME cut’ (bread)]]
c. generalized:
[do’ (a, [use’ (a, b)])]
CAUSE [[do’ (b, [act’ (b, c)])] CAUSE [BECOME result’ (c)]]17
The ordering from left to right in (38b) and (38c) iconically reflects the compo-
nents of causal event chains. If we now ‘cut off ’ the part until the first CAUSE-
operator, we ‘lose’ variable a, which represents the highest-ranking argument in
terms of the AUH. We are left with b as the next highest in this hierarchy. Since
condition (β) of the assignment rule in [2] does not obtain any more, b now can
become Actor and be coded as a PSA (nominatival subject):
(38) d. The knife cuts the bread.
e. [do’ (Tom, [use’ (Tom, knife)])] CAUSE
[[do’ (knife, [cut’ (knife, bread)])] CAUSE [BECOME cut’ (bread)]]
Rule [2] thus formulates the internal hierarchy between animate agents (conscious
instigators) and instruments and explains under which condition instruments
can be coded as a PSA in the nominative.18 The question however is whether
they really do get coded in the nominative. Furthermore, this hierarchy-based
rule cannot capture the use of the Lithuanian instrumental for the description of
illnesses or weather phenomena. These denote states and do not involve any hu-
man effector. We will come to these issues in Section 3.2.2.8. In Section 3 we will
also draw attention to instances in which logical decomposition does not supply a
reliable key for a distinction between argument and adjunct status, nor for deter-
mining the number of core arguments.
2.2 Comparison to some other frameworks
RRG owes a lot to Dowty’s decompositional semantics (see 2.1), and its macro-
role concept resembles Dowty’s (1991) notion of agent and patient proto-roles.
A crucial difference exists in that Dowty accepts prototype features as defining
properties of specific semantics roles (and their typical syntactic behaviour). In
practically all other respects Dowty’s framework can be considered as possibly
the one that is closest to RRG. Among other things, the proto-roles are char-
acterized in sets of opposed features (causing vs. caused, physically active vs.
17. act’ is meant as joker for any kind of (goal-directed) activity, result’ as a joker represent-
ing any state resulting from that activity. Instead of BECOME, INGR is imaginable as well; for
the causal event structure the difference between both aspectual operators is irrelevant.
18. Similar observations on role hierarchies have been made in other (mostly non-decomposi-
tional) approaches, too (cf. Primus 2009: 264f. for a brief survey).
physically changed, etc.). For a concise overview of other approaches based on

semantic primitives, decomposition and/or relational structure of arguments cf.
Butt (2009: 34–38).
As has been pointed out recently by Haspelmath (2011), most prominent
theories in the field of alignment typology rest on assumptions about hierarchies
(or contrasts) of semantic roles (as arguments of predicates) which betray a mix-
ture of semantic and syntactic notions. Together with this, different alignment
frameworks mostly yield identical (or at least very similar) results in their classi-
fication of numerical valency and the treatment of coding patterns which deviate
from some accepted prototypical (or default) alignment of transitive, as well as of
ditransitive, predicates (as well as in attempts to give non-circular definitions of
transitivity in relation to the coding of core arguments). However, as a rule, most
frameworks implicitly apply their notions as both descriptive and comparative
concepts, and they are rather ‘sloppy’ in, or altogether tacit about, borderline in-
stances of argument vs. adjunct status. In these respects RRG is no real exception.
Since a comprehensive discussion of these issues clearly goes beyond the aims of
this contribution, the following short comparison with the Dixonian approach
may suffice.
Dixon’s (1994) core arguments A, O, S, E are conceived of as grammatical
categories. They do not show one-to-one relationships with semantic roles (Agent,
Patient, Experiencer, etc.). A and O exist only as correlative notions, whereas the
relation between S and E is asymmetric: E is in opposition (or: complementa-
ry) to S, but S does not need (or imply) E. All four categories are to be regard-
ed as prototypes, which are partially defined on semantic grounds (cf. Onishi
2001: 2f.) They are thus hybrid categories. While the notion of macroroles in RRG
is not deprived of hybrid nature as well (see 2.1), they do not rely on prototypes
and coincidences with Dixon’s notions are only partial. For instance, on the one
hand, an Undergoer coincides with a prototypical O, and S most probably shows
the properties of an Undergoer if the predicate cannot be decomposed with a
do‘-predicate. On the other hand, prototypical A is coextensive with an Actor, but
the latter represents S if the predicate is one-place and has to be decomposed with
a primitive do’-predicate. Anyway, these are only coextensions, but the notions
are intensionally different.
One striking thing in Onishi’s (2001) typological overview of non-canoni-
cally marked core arguments is that he does not mention the instrumental at all.19
19. In Lithuanian (and Slavic), the instrumental regularly shows up in causative ditransitives
other than transfer verbs (in the broadest sense), and also as the case coding the instrument,
the means (which may or may not be controlled by an agent/instigator), or some natural force
(see 3.2.2.8).
Dixon (1994) does not mention the instrumental as a case marking E (in the
extended intransitive pattern) either. The typical case to be used in a S–E core
argument relation seems to be the dative (which also shows up as E in ditransi-
tives in languages like German, Icelandic, Lithuanian and the Slavic ones). Does
this mean that, from a typological perspective, the instrumental is not involved
in non-canonical marking? As shown in 2.1, the instrumental can be treated as a
core argument if it represents a segment within a causal event chain. The general
neglect of the instrumental as a core argument in the typological literature may
have been caused by the crosslinguistically frequent interference (or polyfunc-
tionality) of the instrumental with functions other than the instrument (or the
means), first of all with the ergative or the passive agent, with manner or comita-
tive (cf. Malchukov & Narrog 2009: 529; Narrog 2009).
3. A first analysis of verb classes and case marking patterns

In this section we will describe our methods used for data collection, and pres-
ent the preliminary results of our analysis together with considerations on some
problems to be treated more systematically in subsequent research.
3.1 Provenance of the data
We have selected our database from dictionaries and from published linguistic
works. For Icelandic there already existed a list compiled by Jónsson (1998). As
for Lithuanian, we needed to compile and create lists ourselves. We have checked
systematically the three volumes of Lietuvių kalbos veiksmažodžių junglumo žody-
nas by Sližienė (2004; henceforth NS) and two dictionaries of modern Lithuanian:
Dabartinės Lietuvių Kalbos Žodynas (2000, DLKŽ) and the electronic online ver-
sion of Lietuvių Kalbos Žodynas (LKŽe). All items were entered into Excel files.
For claims made below many items have been checked in the electronic corpus of
modern Lithuanian (Lietuvių kalbos tekstynas, LKT) and with selective question-
naires presented to native speakers.
At present our lists still show some gaps, but they suffice for the purpose of
this report paper. Since we have already checked existing dictionaries more or less
systematically, we are quite confident that only exceptional coding patterns (and
their alternations) have evaded our attention and that the overall proportion of
groups given below will not change too much when the database assumes a more
complete stage. In the discussion below we have included some additional units
not yet entered in the database, if they helped illustrate some specific point.
3.2 The coding patterns and lexical groupings
The data have been classified according to the following criteria:
i. lexical groups;
ii. coding patterns and, if more than one pattern is attested for a predicative
item, their alternations.
Apart from that, we have checked whether the respective items have alternative
codings that conform to the canonical nom-acc-pattern and, if so, whether and
to what extent differences in case choice correlate with differences in meaning.
We have not yet systematically checked for whether the given verb permits a
clausal argument (= non-first order argument). This amounts to Dixon’s (1991)
distinction between two classes of predicates, Primary-A and Primary-B; cf.
Onishi (2001: 24): “Primary-A […] never take complement clauses (e.g. ‘run’, ‘hit’,
‘eat’ and ‘give’) and Primary-B which may allow a complement clause to fill a
core function slot (e.g., ‘think’, ‘see’, ‘like’ and ‘surprise’).” This analysis requires a
separate study and exceeds the scope of this exploratory paper (see however end
of Section 3.2.2.3).
3.2.1 The lexical groups

Let us begin with the lexical groups. We grouped the verbs into the following
classes:
(lex1) meteorological
Verbs referring to natural or meteorological (climatic etc.) phenomena. Exam-
ples: Lith. aušti ‘dawn’, apdrėbti ‘cover with snow’, Icel. daga ‘dawn’, fenna ‘cover
with snow’.
(lex2) quantifying
Verbs predicating some aspects of quantification. Mostly this concerns changes
in the amount or number of the respective object denoted, which in Lithuanian
is consistently coded with the genitive (see 3.2.2.5). For instance, Lith. apstėti,
daugėti ‘increase, grow’, pakakti, užtekti ‘suffice’, stokoti, stigti, trūkti ‘lack’; Icel.
fjölga ‘increase’, vaxa ‘increase, grow’, fækka ‘diminish’, duga, nægja ‘suffice’, bresta,
skorta, vanta ‘lack’.
Existential and presentational sentences also belong to the field of quantifica-
tion. In Lithuanian they are consistently marked with the genitive if an indefinite
quantity is indicated (cf. Seržant, this volume). However, we have not counted
verbs in constructions like (39) where the genitive is not licenced by the verb or
with explicit quantifiers, such as a numeral (40):
(39) Šiandien pas mus yra / atėjo svečių.

today at we.acc be.prs.3 / come.pst.3 guest.gen.pl
‘Today (some) guests stay at our place / with us.’
(40) Mergaitei yra 10 metų.
girl.dat be.pst.3 year.gen.pl
‘The girl is ten years old.’
The reason is that we are first and foremost interested in the properties of lexemes,
not of constructions.
(lex3) physiological
Verbs denoting some kind or other of bodily experience, usually connected with
an illness or unpleasant feeling. The difference with regard to emotive predicates
(see next group) is that the feeling shows direct physiological symptoms; for in-
stance, Lith. atleisti ‘release’, kratyti ‘shake’, purtyti ‘shake’ (non-literal ‘be disgust-
ed’) or berti (išberti) ‘scatter, cover (e.g., with pimples)’:
(41) Jam veidą beria spuogais:
he.dat face.acc cover.prs.3 pimple.ins.pl
vaikui petį votimis išbėrė.
child.dat shoulder.acc ulcer.ins cover_up.pst.3
‘His face becomes covered with pimples: the child’s shoulder has been covered
with ulcers.’ (NS, sub verbo)
Note that in Lithuanian such predicates often go with an external possessor, which
is coded in the dative. In Icelandic the use of external possessors is much more
restricted. On this issue see further 3.2.2.1.
Icel. verkja ‘ache’, klæja ‘itch’, svíða ‘hurt’, blæða ‘bleed’ and also less direct
physiological symptoms: svima ‘feel dizzy’, klígja, velgja ‘feel nauseated’, as well
as verbs denoting improvement in health (Icel. batna, Lith. gerėti ‘get better’) or
decline in health (Icel. deprast, lakra, versna ‘get worse’).
(lex4) emotive
Verbs referring to emotional states, either negative or positive; for instance Lith.
atsibosti ‘become bored’, prailgti ‘be tired of ’, rūpėti ‘worry, be concerned’, patikti
‘like’, vilioti ‘be attracted to’, Icel. leiðast ‘be bored’, lengjast ‘be tired of ’, óa við,
ugga ‘be afraid, fear’, langa ‘want’, líka ‘like’.
(lex5) epistemic
This term is to be taken broadly, since it refers to mental states based on some
kind or other of impression. All seem-verbs belong here (Lith. atrodyti, rodytis,
dėtis, nusiduoti etc., Icel. sýnast, virðast, þykja), but also some closely related ones
like Lith. matytis or šviestis ‘look like, appear’ as in
(42) J-oms švieči-a-si, kad taip bu-s ger-ai.
they.f.dat.pl appear.prs3.refl comp this_way cop.fut.3 good.adv
‘They understood [from some impression] that this way it will be alright.’
This class can sensibly be united with perceptual verbs (see lex6), but, jointly with
the latter ones, should be clearly distinguished from fructitive verbs (see lex8).
(lex6) perceptual
Here we mean verbs denoting impressions transmitted via some concrete sensory
experience. For Lithuanian we have found only verbs referring to olfactory expe-
rience (Lith. dvokti ‘stink’, dvelkti ‘smell’), as for Icelandic we find verbs referring
to auditory experience heyrast ‘hear,’ misheyrast ‘mishear’ or visual missýnast ‘fail
to see’ and of taste bragðast ‘taste’
(43) Líklega misheyrðist mér.
probably mishear.pst.3SG 1sg.dat
‘I probably misheard.’
(lex7) cognitive
This group consists of cognitive verbs such as Lith. sapnuotis ‘dream’, dingtelėti
‘come to mind’, užkristi ‘forget’ and Icel. dreyma ‘dream’, gruna ‘suspect’, furða ‘be
surprised’.
(lex8) fructitive
As fructitive we qualify verbs which denote successful or, if negated, unsuccessful
performances of some action (achievement of a goal). For instance, Lith. klotis,
sektis, eitis, nutikti ‘succeed’ (44), taikytis ‘manage, happen’ (45), Icel. auðnast,
lánast, takast ‘succeed, manage’ (48), farnast (49), ganga ‘do well/badly’. Some of
them denote happenings; for instance, Lith. tekti1 ‘fall into one’s lot’, sukakti ‘pass
(about anniversaries)’ (47), Icel. áskotnast ‘get by accident’ (50).
Lithuanian
(44) Jiems sekasi (su darbu).
they.dat.m.pl succeed.prs.3 with work.ins
‘They are doing well with (their) work.’
(45) Kartais ir aklai vištai taikosi

sometimes even blind.dat.sg.f hen.dat.sg.f happen.prs.3
grūdą rasti.
grain.acc.sg find.inf
‘Sometimes even a blind hen happens to find a grain.’
(46) Jam teko garbė pristatyti miesto merą.
he.dat get.pst.3 honor.nom introduce.inf town.gen mayor.acc
‘He got the honor of introducing the mayor of the city.’
(47) Dėdei sukako 75 metai.
grandpa.dat come_to_age.pst.3 year.nom
‘Grandpa has reached the age of 75.’
Icelandic
(48) Henni auðnaðist að ljúka prófi
she.dat manage.pst.3sg to finish.inf exam.dat
‘She managed to finish her exam.’
(49) Þeim farnaðist vel á nýja staðnum
they.dat do_well.pst.3sg at new place
‘They did well in the new place.’
(50) Mér áskotnaðist þessi bók um daginn
1sg.dat get.pst.3SG this book.nom about day.acc
‘I got this book the other day.’
The common denominator behind these verbs is that per se they do not denote
any particular event or process,20 but serve to assert (or ask for) something that
can be checked as true or false. Some of them belong to Dixon’s (1991) class of
‘Secondary verbs’; such predicates supply a “semantic modification of some other
verb, with which they are in syntactic or morphological construction” (quoted
after Onishi 2001: 24). Some of them behave like control verbs (e.g., examples
(45)–(46)). We have chosen the somewhat ‘exotic’ term ‘fructitive’,21 since we do
not want this class to be confused with factive predicates in the classical sense,
i.e., with predicates for which the complement is logically presupposed and thus
true even if the complement-taking predicate is negated (e.g., Lith. Visiems buvo /
Niekam nebuvo gaila, kad P ‘Everybody / Nobody regretted that P’ → P holds in
either case). In some cases this assertion is linked up with subjective evaluation
20. Some of the predicates subsumed under ‘happenings’ include also verbs with rather con-
crete descriptive content, but with a very limited range of collocations (e.g., Lith. sukakti in
example (47)).
21. The term was proposed by Wayles Browne.
(compare, for instance, examples (44) and (49)), but it differs from a purely emo-
tional reaction.
(lex9) modal
Most of these are auxiliaries, i.e., verb-like lexemes that have lost their argument
structure and combine with the infinitive of a ‘full’ verb contributing the lexical
meaning (and the argument structure) to a single, but complex predicate; com-
pare Lith. derėti2, tekti2, Icel. bera ‘be obliged to’. In both languages all modal com-
plex predicates which are of concern here code the HRA of the lexical verb in the
dative (see examples (51)–(52)). It would thus be most correct to say that this
dative argument “belongs” to the lexical verb, but that the entire modal complex
requires it to be coded with the dative; for instance:
Lithuanian
(51) Tau tenka nueiti ten pačiam.
2sg.dat have_to.prs.3 go(away).inf there self.dat.m.sg
‘You have to go there yourself / alone.’
Icelandic
(52) Mér ber að fara einni.
1sg.dat bear.prs.3sg leave.inf alone.dat.f.sg
‘I have to go alone.’
Other items behave like control verbs. A case in point is Lith. praversti ‘be worth,
advisable, useful’ (see example (53a)), which could also be included into the next
subgroup (lex10) especially if used with a first-order NP (see example (53b)):
Lithuanian
(53) a. Tau praverstų turėti guminius batus.
2sg.dat be_worth.sbjv.3 have.inf rubber boots.acc
‘It would be good for you to have (= put on / take with you) rubber boots.’
b. Tau praverstų guminiai batai.
2sg.dat be_worth.sbjv.3 rubber boots.nom
‘Rubber boots would come in handy for you.’
(lex10) suitability, appropriateness

In Icelandic, there are a couple of verbs denoting the notion of appropriateness:
hæfa, henta, passa, haga, láta, fara. In Lithuanian we have found tikti, derėti1 and
praversti mentioned in lex9.
Lithuanian
(54) Mėsos ir pūkų gamybai tinka
meat.gen and down.gen production.dat suit.prs.3
mūsų lietuviškos vištinės žąsys.(Žemės ūkis, 1997)
our Lithuanian.nom.pl.f goose.nom.pl
‘Our Lithuanian geese are well suited for the production of meat and down.’
Icelandic
(55) Honum hentar ekki að vera í skóla
He.dat suit.prs.3sg not be.inf in school
‘Being in school does not suit him.’
We want to stress that the lexical classes are not conceived of as mutually exclu-
sive, although we have put each lexeme only into one class. Fortunately, only a
few items raise problems concerning their classification into these lexical groups.
We have already mentioned Lith. praversti. Another problematic item is Lith.
knibždėti ‘swarm, teem’, which can be considered both a perceptual verb and a
verb marked for its feature of indefinite quantity. We have included it into the
perceptual group (lex6). Yet another item in point is Lith. reikėti1 ‘need’ (with a
first-order, not a clausal argument), which combines emotive and quantificational
semantics (and has been included into the latter group, lex2).
As concerns now the preliminary picture, let us have a look at Figure 3, which
shows the type frequencies and proportions of the lexical groups distinguished
above.
With regard to Lithuanian a caveat is necessary. This figure and the following
ones do not account for 38 items with prefixes that can be considered as phasal
(and thus purely temporal) modifications of activity verbs or as resultative (com-
pletive) specifications of mutative verbs (e.g., ap-aušti ‘dawn’) or of diffuse ac-
tive-accomplishment verbs (e.g., api-berti ‘cover (up)’).
Lithuanian is very productive in verb-internal stem derivation, both with
prefixes and suffixes. The derivational patterns are very heterogeneous as for the
possible contributions to the meaning of the derived stem (for a survey over pre-
fixation cf. Wiemer 2013: 225–227). Relevant for our present concern is the fact
that prefixes need not alter the lexical meaning of the deriving stem, they may
also just modify the temporal character of the situation denoted by the deriving
stem.22 There are basically two types. In one type the prefixes focus on one of the
22. Temporal modifications as well as lexical modifications which turn an atelic simplex into a
telic prefixed verb stem can, but need not, be accompanied by an intransitive-transitive change
(e.g., dribti ‘fall’ → ap-dribti ‘fall upon, cover up’, likti ‘remain’ → pa-likti ‘leave’, snigti ‘snow’ →
pri-snigti ‘cover with snow’).
Lexical groups
%
50
40
30
20
10
0
Meteoro- Quanti- Physio- Emotive Epistemic Cognitive Perceptual Fructitive Modal Suitability
logical fying logical
Lithuanian 13.3% 5.6% 40.0% 19.0% 5.6% 2.6% 2.6% 7.2% 3.1% 1.0%
verbs
Icelandic 19.3% 6.2% 22.1% 12.6% 0.9% 8.7% 1.4% 26.0% 0.5% 2.3%
verbs
Figure 3. Proportion of the lexical groups of verbs with non-canonical marking
of their highest-ranking argument (HRA) in Lithuanian and Icelandic
phases of an activity (i.e., atelic) verb, thus being simply ingressive, egressive or
semelfactive modifications of atelic stems (e.g., baisėti-s ‘be afraid’ → pa-si-baisė-
ti [the same], bosti ‘be weary of ’ → at-si-bosti, nu-si-bosti ‘get weary of ’, daigyti
‘stitch’ → su-daigyti ‘begin to stitch, stitch suddenly’, pa-daigyti ‘to stitch’, diegti
‘stitch’ → už-diegti ‘begin to stitch’), or they serve to set a temporal limit to the per-
formed action (e.g., pa-diegti ‘hurt for a while’).23 Another type of temporal mod-
ification applies to mutative verbs (like aušti ‘dawn’) and to verb stems that are
diffuse as for their telic or atelic character (= Active accomplishments in RRG). In
these cases the prefix interacts with the heterogeneous nature of phases and the
inherent boundary of the deriving stems. The ‘output’ of this interaction is a resul-
tative (or otherwise: completive) focus on the verbal action; this, in turn, can be
accompanied by a holistic effect (on which cf. Lenartaitė 2011: Ch. 4.3). Compare,
for instance, berti ‘cover’ → api-berti, iš-berti ‘cover up’, dribti ‘cover with snow’ →
ap-dribti ‘cover up totally with snow’; compare the following pair of verbs:
(56) Drimba lauke sniegas ubago kąsniais.
fall, cover.prs.3 outdoors snow.nom beggar.gen bit.ins.pl
‘Outdoors snow covers everything with huge flocks.’ (LKŽe, sub verbo)
23. Overlaps between both subkinds of this type of temporal modification are possible and
should be considered item by item. Temporal delimitation is associated to the prefix pa-, in-
gressive meaning with už-; however, pa- can also be used as a marker of ingression.
(57) Apdribo mane visą, visus mano drabužius.

cover_up.pst.3 1sg.acc all.acc.sg.m all.acc.pl.m my clothes.acc.pl
‘It has covered me all over, all my clothes.’ (NS, sub verbo)
Therefore, from all counts concerning Lithuanian we have excluded prefixed verbs
with either kind of modification. We furthermore excluded all morphologically
derived verbs whose affixes did not show any discernible difference of meaning
in comparison to the bases from which they derive (e.g., rodyti-s → pa-si-rodyti
‘seem’, dėti-s → su-si-dėti ‘seem’), or if they happen to be derived from the same ba-
sis with a variant of the same suffix (e.g., purt-in-ti and purt-y-ti ‘shake’, both being
morphological causatives with -in- and -y- in rather free distribution). Altogether
such cases amount to 38 (≈16%) out of 229 verb stems. If we did not exclude
these instances, the comparison with Icelandic would be skewed, because Icelan-
dic verbal prefixes do not cause the kind of temporal modification that we see in
Lithuanian and do not show derivational verbal morphology without changes of
lexical meaning.
Let us now go on with the basic counts of the Icelandic verbs (Figure 3).
There are also caveats necessary for Icelandic. We have excluded idiomatic con-
structions with dative external possessors. These constructions are included in
traditional accounts on oblique subjects such as the list of Bernódusson (1982),
Jónsson (1998). These datives are possessive, because their referent is the pos-
sessor of an inalienable entity, very often a body part. They are not justified by
the valency of the verb and therefore we exclude them from our list (see further
3.2.2.1 on Icelandic external possessors). We have also excluded other idiomatic
expressions such as
(58) Honum féll allur ketill í eld
he.dat fell.pst.3sg all.nom.sg.m kettle.nom in fire
‘He was stupefied.’
(59) Henni féllust hendur
she.dat fell.pst.3pl hand.nom.pl
‘She gave up.’
Having in mind that the figures are based on different sum totals of verb items
for both languages (see Figures 3–6) and that we have not yet conducted tests on
significance, some differences between Lithuanian and Icelandic become evident
already from a more cursory comparison. There are no meteorological verbs with
a dative argument in Lithuanian, while Icelandic has at least some. Icelandic also
shows a larger proportion of verbs denoting physiological phenomena with a da-
tive-HRA, whereas this lexical group is the absolutely dominating one for the
accusative-coded HRA in Lithuanian, for which also the proportion of emotive
Dative verbs, lexical groups

%
50
40
30
20
10
0
Lithuanian 0.0% 9.1% 7.8% 32.4% 14.3% 6.5% 1.3% 18.2% 7.9% 2.6%
verbs
Icelandic 10.4% 5.0% 18.7% 13.7% 1.7% 5.0% 2.5% 38.6% 0.8% 3.7%
verbs
Figure 4. Proportion of the lexical groups of verbs with DAT-coded (HRA)
Accusative verbs, lexical groups

%
70
60
50
40
30
20
10
0
Lithuanian 23.9% 0.0% 65.1% 11.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
verbs
Icelandic 32.2% 8.2% 27.9% 12.0% 0.0% 14.2% 0.0% 4.9% 0.0% 0.5%
verbs
Figure 5. Proportion of the lexical groups of verbs with ACC-coded (HRA)
and epistemic verbs with a dative-HRA is considerably larger. In turn, Icelan-

dic has many more fructitive verbs for both dative- and accusative-coded HRA,
while Lithuanian has none with accusative-coded HRA. There are no epistem-
ic, perceptual or modal verbs with accusative-marked HRA in either language;
Genitive verbs, lexical groups

%
100
90
80
70
60
50
40
30
20
10
0
Lithuanian 0.0% 100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
verbs
Icelandic 0.0% 42.8% 0.0% 0.0% 0.0% 0.0% 28.6% 28.6% 0.0% 0.0%
verbs
Figure 6. Proportion of the lexical groups of verbs with GEN-coded (HRA)
in this respect these lexical groups remarkably differ from emotive verbs. As for
genitive-coded HRA, Lithuanian shows a restriction to the ‘quantity’-group, but
anyway in both languages such verbs are in a clear minority against verbs with a
dative- or accusative-coded HRA.
Now, let us comment on Dative Sickness in Icelandic. Dative Sickness
(DS), or dative substitution, is a well-known phenomenon and has been in-
tensively studied (cf. Svavarsdóttir 1982; Zaenen et al. 1985; Halldórsson 1982;
Rögnvaldsson 1983; Svavarsdóttir et al. 1984; Smith 1994; Eythórsson 2000a, b,
2002; Smith 2001; Barðdal 2001a: 134–138, 2004, 2008; Minger 2002; Jónsson
2003; Jónsson & Eythórsson 2005; Friðriksson 2008). It refers to a change or, more
properly, a variation that occurs to a considerable degree at the level of the indi-
vidual speaker in the case marking of the HRA, more precisely: when accusative
is replaced by dative. For instance,
(60) Mig vantar hníf > Mér vantar hníf
1sg.acc need.prs.3sg knife.acc > 1sg.dat need.prs.3sg knife.acc
‘I need a knife.’
Table 1. Lexical groups affected by Dative Sickness in Icelandic

Number %
meteorol. 0 0.0
quantifying 15 14.2
physiological 49 46.2
emotive 20 18.9
epistemic 0 0.0
cognitive 22 20.8
perceptual 0 0.0
fructitive 0 0.0
modal 0 0.0
suit 0 0.0
Σ 106
This change has been documented for other Germanic languages, such as Ger-
man, Old English, Faroese.24 DS only affects acc-marked HRAs that can qualify
as experiencers (animate, usually human). Among the respective verbs are bresta
‘lack’, dreyma ‘dream’, gruna ‘suspect’, hrylla við ‘shudder at’, hungra ‘hunger’, iðra
‘repent, regret’, kitla ‘tickle’, klígja ‘feel nauseated, sick’, langa ‘want’, lengja eftir
‘long for’, skorta ‘lack’, sundla ‘feel dizzy’, svengja ‘feel, hungry’, svima ‘feel dizzy’,
svíða ‘smart, sting’, syfja ‘grow sleepy’, ugga ‘fear’, undra ‘wonder, surprise’, vanta
‘lack, need’, verkja ‘hurt, ache’, þrjóta ‘run short of, lack’, þverra ‘decrease’, þyrsta
‘feel thirsty’.
Table 1 shows to what extent the lexical groups can be affected by DS.
With meteorological verbs, instead, inanimate accusative (or dative) marked
HRAs are often replaced by the nominative (‘Nominative Sickness’); for instance:
(61) Storminn lægði > Stormurinn lægði
wind.acc subside.pst.3SG > wind.nom subside.pst.3SG
‘The wind subsided.’
There is a difference among the lexical groups for which DS can occur. It is more
prevalent with verbs of need and lack (e.g., vanta ‘lack’), emotive verbs (e.g., að
langa ‘want’) and physiological verbs (e.g., kitla ‘tickle’), but it is less common
among cognitive verbs as dreyma ‘dream’ and gruna ‘suspect’, which can even oc-
cur with nominative subject (Jónsson & Eythórsson 2005: 233).
DS has been attested in Latvian (Seržant 2013b: Ch. 3.2), but it does not oc-
cur in Lithuanian. We are however not that sure with respect to an equivalent of
24. In German it is already documented in mediaeval times (Seefranz-Montag 1983: 162–163).

The spreading in Icelandic is much later, probably in the second half of the 19th century, but the
onset is earlier, as can be seen by a few isolated examples from older texts (Halldórsson 1982).
Nominative Sickness in Lithuanian. One can come across a nom- vs. acc-alter-
nation of the single argument of meteorological verbs as in (62) (registered by the
LKŽe, sub verbo):
(62) a. Upė šąla.
river.nom freeze.prs.3
‘The river is freezing.’ (LKŽe, sub verbo)
b. Jau langą šąla.
already window.acc freeze.prs.3
lit. ‘(it) is already freezing the window.’ (LKŽe, sub verbo)
We cannot be sure which alternative is the diachronically earlier one, in particular

if the chronological order between nom- and acc-marking corresponds to the
nom>acc-order stated for Lithuanian ‘pain/ache’-verbs by Seržant (2013a).
3.2.2 Patterns of coding

Now let us turn to the coding patterns. As much as possible, we distinguish be-
tween one- and two-place verbs, although some problems arise with determining
the number of arguments. We must confess that we have been unable to sup-
ply reliable criteria applicable to all critical cases encountered. We therefore have
largely restricted ourselves to more or less elaborate discussions on the reasons
for this difficulty. For the same reason the tables below do not separate one- and
two-place predicates.
Two things can be said in advance. First, we have not encountered any con-
vincing cases of adpositional marking of core arguments for those groups of
predicates which satisfied our selection condition, namely: that the HRA would
be marked other than with the nominative. Second, we have not come across
non-canonical coding with three arguments; in fact, this seems hardly imagin-
able, because it would practically amount to a weird constellation without a pro-
totypical agent (argument of DO in the AUH), but with a causer (= x-argument
of do’) and two arguments closer to the Undergoer-pole of the AUH. Apart from
this, another caveat concerning the empirical picture is in place, which we dwell
upon in the next subsection.
3.2.2.1 External possessors

In Lithuanian, but not in Icelandic, inalienable possession is regularly indicat-
ed by a so-called external possessor, i.e., a dative-NP whose referent stands in a
natural part-whole relationship to another referent of the situation denoted by
the predicate, but which is given independent syntactic status outside of argu-
ment NPs. It is thus treated like an additional argument in the core, although it is
not (on any account) licensed by the verb; instead it must be interpreted via the
argument-NP to which it shows the part-whole (or ‘possessive’) relationship. Com-

pare two different uses of the dative as a possessor (63) and an experiencer (64):
(63) Ligoniui atėmė rankas ir kojas.
patient.dat take_off.pst.3 arms.acc and legs.acc
‘The patient was deprived of his/her arms and legs [because of a natural
force].’ (NS, sub verbo)
(64) Jonui skauda galvą / galva.
pn.dat ache.prs.3 head.acc / head.nom
‘Jonas has a headache.’ (lit. ‘To Jonas (it) hurts the head / the head hurts.’)
The dative-NP ligoniui in (63) denotes a referent who is to be thought of as the

natural and, thus, inalienable ‘whole’ with regard to the referent of the object-NPs
rankas ir kojas (here atimti ‘take off ’ denotes a physiological process resp. event
not controlled by a human agent). The dative-NP Jonui in (64) denotes the same
kind of relationship w.r.t. the NP galvą (galva). However, only in (64) can the NP
which denotes the part be omitted (64a), while it cannot in the former instance
without yielding a nonsensical, at best elliptical utterance (63a):
(63) a. *Ligoniui atėmė.
(64) a. Jonui skauda.
Obviously, for an external possessor to be interpretable the part with which it

is mutually related has to be made explicit. The predicate requires only the con-
stituent which denotes the most patient-like participant of the action, while the
inalienable possessor itself is ‘licensed’ only via its mutual association with this
patient-like participant. The situation for skaudėti is different, because no pa-
tient-like argument is required; the dative-NP denoting the affected person is li-
censed by the verb itself, and it therefore is not to be characterized as an external
possessor.25 We will come back to this point in connection with ‘ache’-verbs (see
3.2.2.9).
As concerns external possessors in Icelandic, there is a syntactic restriction
imposed on the distribution of possessive datives. They are limited within a prep-
ositional phrase which denotes the static position of an entity belonging to the
semantics of the sentence:
(65) Hún settist á bak hestinum
she.nom sit.pst.3SG on back.acc horse.dat
‘She sat on the back of the horse.’
25. We thank Axel Holvoet for having brought this to our attention.
Even provided this condition holds, the possessive dative is rare and consid-
ered very formal or poetic, and it can alternate with other constructions (cf.
Höskuldur Thráinsson 2005: 217–219). It occurs though quite often in idiomatic
expressions:
(66) Sigrúnu kom ekki dúr á auga
pn.dat come.pst.3SG not sleep.nom on eye.dat
‘Sigrún got no sleep at all.’ (idiom.) (Jónsson’s list)
(67) Sigrúnu lá mikið á hjarta
pn.dat lie.pst.3SG a lot.nom on heart.dat
‘Sigrún was very anxious.’ (Jónsson’s list)
(68) Sigrúnu vafðist tunga um tönn
pn.dat wind.pst.3SG tongue.nom around tooth.acc
‘Sigrún was at a loss for an answer.’ (idiom.)
In Icelandic, there is another pattern for marking an external possessor, which

occurs, for instance, with predicates denoting somebody’s state of health or a un-
pleasant physiological experience. In this pattern the possessor gets marked with
the accusative: accposs – í + PPbody_part. We will comment on that in 3.2.2.4.
3.2.2.2 Rough proportions of patterns

Table 2 supplies the type frequencies classified by the oblique cases used for mark-
ing the HRA.
3.2.2.3 Verbs taking accusative arguments

The most obvious fact is the high number of acc-coded HRA in Lithuanian (see
Figure 5). The absolute majority of them are verbs denoting unpleasant physio-
logical experience. Here a sort of mismatch between argument positions in log-
ical decomposition and syntactic coding appears. The LS of predicates with one
Table 2. Lithuanian and Icelandic one- and two-place verbs distinguishing
the case of the HRA
Lithuanian Icelandic
N % N %
ACC 109 54.5 183 42.2
DAT 77 38.5 244 56.2
GEN 9 4.5 7 1.6
INS 3 1.5 absent absent
LOC 2 1.0 absent absent
Σ 200 434
syntactic valency consists of the primitive predicate feel’ (x, [pred’]), with pred’
being a joker for any sort of appropriate state. Compare an example for deginti
‘burn (trans.)’:
(69) a. Šiąnakt mane labai degino.
this_night 1sg.acc very burn.pst.3
‘This night I felt hot.’ (lit. ‘(it) burned me’) (LKŽe, sub verbo)
b. feel’ (1sg, hot’) [MR1], MR=acc
In logical decomposition these predicates appear with two argument positions

of which the second one (y) is occupied by a one-place state predicate; see (69b).
The mismatch may probably be explained by lexical incorporation of this state
predicate (hot’) into the meaning of the verb root: deg-in-ti ‘burn (tr.), set on fire’
is the morphological causative of deg-ti ‘burn (intr.)’, and this lexical feature is
the basis of the metaphorical transfer underlying the meaning extensions of such
‘impersonal’ verbs (see below).
Regardless of this, the state can be modified by an aspectual operator, and
the reason for the (change of) state can be indicated in Lithuanian by a NP in the
nominative (70) or a PP (71), but not by an instrumental NP, and in Icelandic by
a PP (72):
(70) Sukrėtė mane šaltis.
jolt.pst.3 1sg.acc frost.nom
‘The frost shook me.’ (LKŽe, sub verbo)
(71) Joną purto nuo šalčio. (vs. *šalčiu)
pn.acc shake.prs.3 from frost.gen frost.ins
lit. ‘(it) is shaking Jonas from the frost.’
(72) Mig verkjaði af hungri.
1sg.acc hurt.pst.3SG of hunger.dat
‘I was aching from hunger.’
The LS for sentences (70)–(72) looks basically identical:
(73) [do’ (šaltis, Ø)] CAUSE [BECOME feel’ (Jonas/1sg, cold’)] .
The question is whether the part to the left of the CAUSE-operator can be con-
sidered as still incorporated into the verb meaning if the verb is used without a
nominatival subject indicating the cause (reason). Moreover, in Lithuanian, apart
from nom-NPs and PPs the reason can often be marked with the instrumental
(see 3.2.2.8), but not for all ‘physiological’ verbs (see example (71)). Among the
next tasks of the project we will have to establish (i) which factors influence the
choice of these three “rival” devices of marking the reason (force, source, cause) of
physiological states, (ii) their relative frequency, (iii) their argument status.
Other syntactically one-place verbs belonging to this lexical group are aižyti
‘sting’, atleisti ‘release’, badyti, daigyti, diegti, durti ‘stitch’, dusinti ‘suffocate, stifle’,
ėsti ‘eat (of animals)’, gelti ‘sting’, gniaužti ‘squeeze, knead’, griaužti ‘chew’, kratyti,
krėsti ‘shake’, laužyti, laužti ‘break’, plėšti ‘tear (apart)’, purtyti ‘shake’, raižyti, rėžti
‘cut’, spausti ‘press’, tampyti ‘pull’, troškinti ‘parch with thirst’. This group compris-
es 65% of acc-coded verbs with non-canonically marked HRA in Lithuanian.
Following Seržant (2013a), this group may be characterized by the conceptual-
ization of painful physiological experience through highlighting such states as
the endpoints of causal chains (in the sense of Croft 1998). From this cognitive
perspective, the accusative marks, as it were, the person affected as the target of
an unspecified (or lexically incorporated) stimulus. Physiological verbs in Ice-
landic comprise 27% of the verbs with acc-coded HRA. Like in Lithuanian they
also mostly consist of verbs denoting unpleasant physiological experience; for
instance, hungra ‘hunger’, svengja ‘feel, hungry’, þyrsta ‘feel thirsty’, verkja ‘hurt,
ache’, svíða ‘smart, sting’, kitla ‘tickle’, klígja ‘feel nauseated, sick’, sundla ‘feel dizzy’,
svima ‘feel dizzy’, syfja ‘grow sleepy’.
Without going into details, we can say that for almost all of the Lithuanian
verbs the usage referring to unpleasant physiological states can be captured as
a metaphoric extension from some more basic physical activity. The degree to
which this meaning transfer has become conventionalized may differ and should
be established in further research; and certainly this type of non-canonical mark-
ing does not constitute any closed class of verbal items, rather it is supplemented
continuously by further items. Roughly, the respective basic activity is mapped
onto the domain of physical (physiological) experience, and the specific dif-
ferences in the manner of physical activity are preserved in the “physiological”
meaning. Thus, for instance, kratyti, krėsti and purtyti ‘shake’ normally denote
involuntary, spasmic movements of the whole body, whereas verbs like bady-
ti, diegti, gelti ‘stitch, sting’ refer to feelings of itching experienced inside one’s
body, often only of specific parts or a restricted region. The verbs rėžti ‘cut’, spausti
‘press’ and tampyti ‘pull’ are used to express a person’s internal feeling of cutting,
pressing or stretching, respectively, either on the whole body or some significant
part of it. Other verbs are more specialized as for the affected body part, e.g.,
laužyti ‘break’ usually refers to one’s legs. From the list above only troškinti ‘parch
with thirst’ is not derived semantically from physical activity; morphologically it
is the causative of trokšti ‘be thirsty’. Finally, dusinti ‘stifle, suffocate (trans.)’ in its
use with non-canonically marked HRA is derived semantically from the goal-di-
rected activity initiated by a conscious agent (e.g., Nusikaltėlis savo auką dusino
pačiomis rankomis ‘The criminal suffocated his victim with his bare hands.’). In
the ‘physiological’ use of dusinti (and the other verbs) the first part of the causal
event chain (in Croft’s 1998 sense), i.e. this conscious initiator, is lacking. Note,
finally, that the verbs which underlie this pattern of meaning extension denote
not too specialized kinds of physical activity. Moreover, they are – with few ex-
ceptions – not causatives with inherent boundaries; they, thus, do not denote pro-
totypical situations of semantic transitivity (in the sense of Hopper & Thompson
1980), pace Seržant (2013a).
In Icelandic only very few of the acc-coded verbs have this metaphoric ex-
tension. The primary meaning of all of the above mentioned Icelandic verbs is
that of unpleasant physiological experience. The few exceptions from this are:
bora ‘drill’, bíta ‘bite’, blása ‘blow’. With the first two verbs the acc-NP denotes the
affected person (74), with the last one the affected body-part (75):
(74) Mig bítur í eyrun (af frosti)
1sg.acc bite.prs.3SG in ear.acc.pl from frost
‘My ears are cold/frostbitten.’
(75) Fótinn blés upp
leg.acc blow.pst.3sg up
‘The leg got swollen.’ (Jónsson’s list)
There is another important circumstance to be noted about this pattern in Lithu-

anian. Some verbs show an alternation: either the acc-NP denotes the person af-
fected (see (76a)) or, by way of synecdoche, it can alternatively denote the affected
body part of that person (see (76b)):
(76) a. Mane visą daigo.
1sg.acc all.acc.sg stitch.prs.3
‘I feel stitches everywhere.’
b. Jam daigo šoną.
he.dat stitch.prs.3 side.acc
‘It is itching his [lit. him] the side.’
In (76b), the person referent gets marked with the dative and, since it stands in
an inalienable part-whole relation to his/her body part, it represents an external
possessor. As mentioned in 3.2.2.1, we do not count external possessors as part
of the argument structure, although they add to the syntactic valence in the core.
This class shares some crucial semantic properties with a handful of ‘pain’-verbs
(skaudėti etc.); cf. Seržant (2013a) and see further in 3.2.2.9.
In Icelandic the accepted pattern is the one given in (77a):
(77) a. Mig klæjar alla.

1sg.acc itch.prs.3SG all.acc.f.sg
‘It is itching me all over.’
For some speakers however the pattern in (77b) is possible, too:
(77) b. Mér klæjar allri.

1sg.dat itch.prs.3SG all.dat.f.sg.
‘It is itching me all over.’
The reasons for these alternations in the languages are different, as the dative in
(76b) is due to the marking of external possession, while the dative in (77b) is due
to DS and has to be considered as belonging to the predicate’s argument structure
(see 3.2.2.1).
In Lithuanian the alternation shown in (76a–b) does not appear possible with
all verbs of the ‘physiological’ group. Verbs with meanings relating to more specif-
ic body parts prefer the pattern with the external possessor, while verbs denoting
physiological experience pertaining to the whole body (organism) choose the pat-
tern without the external possessor. Browsing through our data and informants’
replies gives the impression that the former group is more numerous.26 The latter
group is represented not only by dusinti ‘stifle’ and troškinti ‘parch with thirst’
(whose ‘derivational history’ anyway differs from the metaphorical transfer of the
other verbs, see above), but also by kratyti, krėsti, purtyti ‘shake’. At the moment,
we cannot present any figures, but to elaborate on this issue is clearly among the
next tasks to be done in research.
For the sake of completeness we have to at least mention here other, less fre-
quent coding options. Some Lithuanian verbs allow the affected body part to be
coded with the locative or a locational-directional PP (mostly į+acc ‘in(to)’ or
po+ins ‘under, at’). In such cases the experiencer is coded with the dative, e.g.
(78) Motinai sugniaužė po krūtine.
mother.dat squeeze.pst.3 under breast.ins
‘Mother felt a squeezing under her breast.’ (LKŽe)
This pattern is much more regular in Icelandic, where the body part is invariably
coded with a PP headed by í+dat ‘in(to), at’ (directed) or ‘under’ (undirected).
The experiencer is coded with the accusative:
26. Utterances like *Mane.acc diegia / gelia ‘(it) stitches me’, with or without an indication of
the affected body part (*Mane.acc diegia šone / į šoną ‘(it) stitches me into my side’) are hardly
accepted by informants.
(79) Mig verkjar í bakið

1sg.acc hurt.prs.3SG in back.acc
‘My back hurts.’
We have not attested instances in Lithuanian in which the experiencer would have
been coded with the accusative (see also f. 26). If an acc-NP naming the person
co-occurs with a directional PP, the latter is used in its primary (spatial) meaning,
the physiological experience refers to the whole person, and the figurative usage
of the verb can hardly be considered a conventionalized meaning. In (80) this
becomes obvious from the particle lyg used for comparison (≈ ‘as if ’):
(80) Mane lyg stūmė į ežerą.
1sg.acc as_if push.pst.3 in lake.acc
lit. ‘It somehow pushed me (in)to the lake.’ (NS, sub verbo)
The same applies to įsukti ‘screw into’ (for which see the next example).
Although we cannot say that stumti and įsukti have lexicalized ‘physiologi-
cal meanings’, the last remarks testify to the productivity of the pattern to which
the class of physiological verbs with non-canonically coded arguments belongs:
the most patient-like argument is marked with the accusative, since there is no
known (or communicatively important) reason (or force) which has evoked the
current state (or event), it is left unspecified, and there is no other participant
which could count as instigator in a causal chain. Briefly: no argument promotes
to a nominatival NP, because there is no participant qualifying semantically as
close to the Actor-pole of the AUH, while there is a participant located closely to
the Undergoer-pole. The same ‘logic’ works with external possessors if the most
patient-like participant is not the whole body, but only a part of it, as in (81) with
ranką ‘hand/arm.acc’:
(81) Jam ranką įsuko į kuliamąją.
he.dat arm.acc screw.pst.3 in threshing machine.acc
‘His hand was driven into the threshing machine.’ (NS, sub verbo)
(lit. ‘To him (it) screwed the hand into the threshing machine.’)
Probably, it was this more general pattern which supplied the ground for the fre-
quency of ‘physiological predicates’ with non-canonical coding.
Another salient, though considerably less frequent lexical class of one-place
verbs with an acc-coded argument denote meteorological phenomena (26% of
all verbs with acc-marked HRA in Lithuanian and 59% in Icelandic), such as
Lith. apdribti ‘cover up (with snow)’ and atlyti ‘soak off ’ and Icel. leggja ‘lay, put
(freeze)’, lægja ‘lower down’:
(82) Kol parvažiavau, apdribo mane visą.

until come_home.pst.1sg cover_up.pst.3 1sg.acc all.acc.sg
‘Until I arrived home, I got all over covered up with snow.’ (NS, sub verbo)
(83) Gerai arti – pūdymą atlijo, minkšta.
well plough.inf fallow_field.acc soak_off.pst.3 soft.n
‘It is (now) easy to plough: the fallow field has been soaked off, everything is
soft.’ (NS, sub verbo)
(84) Tjörnina lagði.
lake.acc lay.pst.3SG
‘The lake froze over.’ (Jónsson’s list)
(85) Storminn lægði.
Storm.acc abate.pst.3SG
‘The storm abated.’ (Jónsson’s list)
Other verbs belonging here are Lith. leisti (sniegą) ‘melt (snow)’, Icel. leysa taka
‘melt (snow)’, Lith. atleisti ‘release’ (about frost), Icel. leggja ‘lay (about frost,
freeze)’, hema ‘freeze’, héla, hríma ‘become covered with frost’ Lith. aplyti ‘make
wet’ (about rain), įmerkti ‘soak’ (e.g. about hay), Icel. flæða ‘flow’, fylla ‘be filled
(with water, rain)’, lægja ‘lower, abate (about wind, storm)’. Into this group we
have included also verbs like Lith. sunešti ‘bring, blow together’ (about snow),
because they refer to forces as natural phenomena, and Icel. reka ‘drift (because of
a current or storm)’, velkja ‘toss about (because of a current or storm)’. Compare:
(86) Kartais per naktį sunėša
sometimes through night.acc bring_together.prs.3
tokias pusnis, kad…
such.acc.pl.f snowbank.acc.pl.f that …
‘Sometimes overnight it blows together such snowbanks that …’
(NS, sub verbo)
Many of the meteorological predicates require a metaphoric reading, insofar as

the meteorological situations are described by non-meteorological verbs. This is
just as with the Lithuanian physiological verbs above, but with meteorological
predicates it concerns both languages. For instance, we have Lith. leisti and Icel.
leysa both with the original meaning ‘let loose, release’ and acquiring the meaning
‘melt’ (about snow), and Icel. leggja ‘lay’, which gets the meaning of ‘freeze’ (about
lakes and rivers).
Finally, an important group of Lithuanian verbs with an acc-coded argument
naming a person incudes verbs which denote the causation of emotional states,
such as erzinti ‘irritate’, nervinti ‘unnerve’, liūdinti ‘sadden’, graudinti ‘grieve’, pyk-
dyti ‘exasperate’, jaudinti ‘excite’, džiuginti ‘make glad’. This acc-coded argument
cannot however qualify as HRA. Since we are dealing here with causative verbs,
the argument structure is enhanced by an additional role which in terms of the
AUH ranks higher than the acc-coded person argument. This obtains irrespec-
tive of whether the causer is another person (87a) or some sort of fact (87b):
(87) a. Jis erzina mane
he.nom irritate.prs.3 1sg.acc
‘He irritates me.’
b. Mane erzina, kad geri tiek daug alaus.
1sg.acc irritate.prs.3 comp drink.prs.2sg that much beer.gen
‘It irritates me that you are drinking so much beer. / I am irritated by how
much beer you drink.’
The non-canonical pattern in (87b) results from productive semantic derivation

by which the (usually human) Actor of the original meaning (87a) is replaced
by a clausal complement (87b). The acc-marking of the Undergoer remains un-
changed, but, in contrast to predicates denoting physiological states (see above),
this argument does not become HRA, since it remains in opposition to another,
higher-ranking argument position, namely: the clausal complement. Compare
the LSs in (69b) and (73) with (87c):
(87) c. [α] CAUSE [BECOME feel’ (1sg, nervous’)],
with α representing any LS.
At least on an intuitive level, the difference between the clausal constituent cor-
responding to α in (87c) and the NP argument represented by šaltis ‘frost’ in (73)
lies in the fact that the latter can be omitted, while the former cannot: Joną purto
‘Jonas is shaken’ (see example (71)) does not sound elliptical even without con-
textual support, whereas Mane erzina (see example (87a–b)) does. Thus, the ar-
gument character of the clausal complement in (87b) should not raise doubts,
contrary to the PP nuo šalčio ‘from/because of the frost’ in (71).
The reason underlying this difference seems to be that the clausal complement
in (87b) (= α in (87c)) represents a proposition, while nuo šalčio corresponds to
an object of lower order (in terms of Lyons 1977). It is not clear to us whether this
difference brings about any more consequences for logical decomposition. But
a quite obvious conclusion following from our considerations is that emotional
predicates with acc-marked (human) Undergoer are two-place predicates which
need a [MR1] stipulation. Moreover, since the acc-coded argument denoting the
person in cases like (87b) has to be qualified as Undergoer, the proposition serv-
ing as cause and symbolized by α in (87c) must not equal a primitive do’-pred-
icate; for otherwise the single direct core argument in the accusative should be
qualified as Actor and the ‘Default Macrorole Assignment Principles’ (see 2.1)
would be violated.
As concerns Icelandic, 20% of its acc-coded verbs denote emotional states
such as langa ‘want’, lengja eftir ‘long for’, ugga ‘fear’. The Icelandic verbs with
acc-coded core argument denoting a person cover more lexical groups than the
Lithuanian ones. Rather many, or 22%, are cognitive verbs such as dreyma ‘dream’,
gruna ‘suspect’, undra ‘wonder, surprise’, minna ‘remember vaguely’. As for verbs
of the quantitative group, which for the most part have gen-coding on the argu-
ment in Lithuanian, 15% of Icelandic verbs of this lexical group code their argu-
ment with the accusative. These verbs mainly denote lack and need (e.g., vanta
‘lack, need’, þrjóta ‘run short of, lack’, bresta, skorta ‘lack’); they are two-place and
mark their other argument with the accusative, too.
3.2.2.4 Verbs taking dative arguments

In Lithuanian, emotive, epistemic, perceptual and fructitive verbs usually code
their HRA with the dative, while in Icelandic it is mainly fructitive verbs which
code their HRA with the dative, and there is some larger amount of meteorolog-
ical and physiological verbs using this case (see Figure 4). For Lithuanian see (in
square brackets we indicate the lexical group):
(88) Jam užbaiso – pradė[jo] šaukt. [emotive]
he.dat become_afraid.pst.3 begin.pst.3 shout.inf
‘He frightened (so that) he began to shout.’ (TDrIV202, Dglš)
(89) Jam truputį maišosi. [epistemic]
he.dat little_bit mix.prs.3
‘He is getting confused a little bit.’ (lit. ‘To him (it) is mixing up (in his head).’)
(NS, sub verbo)
Verbs denoting physiological states with a dat-coded argument are hard to find.
We can however adduce the following example:
(90) Ligoniui pagerėjo.27[physiological]
patient.dat get_better.pst.3
‘The patient has become better.’
This is coded in exactly the same way in Icelandic:
27. As concerns the antonym (pa)blogėti, we have not so far come across any instance with
non-canonically marked experiencer.
(91) Sjúklingnum batnaði

patient.dat get_better.pst.3SG
‘The patient has become better.’
Contrary to the ‘physiological’ verbs with acc-coded argument (3.2.2.3), the in-
choative Lith. (pa)gerėti and its antonym (pa)blogėti (see f. 28), as well as Icel.
batna, skána, versna do not describe any specific sort of pain or disease, but relate
to a person’s general medical condition. In Lithuanian they normally demonstrate
a canonical pattern, as in, e.g.,
(92) Ir mūsų ligonis jau po truputį gerėja.
and our patient.nom already bit_by_bit get_better.prs.3
‘And our patient is getting better step by step.’ (LKŽe, sub verbo)
In Icelandic the more general states of health are generally coded with the dative
as well; for instance, batna, skána ‘get better’, deprast, lakra, versna ‘get worse’,
while verbs denoting more specific sorts of diseases mark their HRA with the
accusative. There are however exceptions to that, as with the verb blæða ‘bleed’
which marks its argument with the dative. There is also an interesting acc–dat
alternation with the verb svíða ‘smart, hurt’; its argument is coded with acc when
it denotes physical pain (93a), and dat when it denotes an emotional state (93b):
(93) a. Mig svíður (í fingurinn)
1sg.acc hurt.prs.3 in finger.acc
‘I smart/My finger smarts.’
b. Henni svíður þetta sárt.
she.dat hurt.prs.3 this.nom badly
‘She feels bad because of this.’
The pattern in (93b) is always a two-place verb (dat–nom), but when the pain is
physical the experiencer is coded with the acc, and the verb has to be qualified
as one-place (93a).
Andrews (1982: 463) mentions this alternation: “Accusatives seem to appear
with essentially physiological states (including intense desires and dreams) while
the datives tend to be associated with psychological states.” He gives the following
examples:
(94) a. Mig kelur
1sg.acc freeze.prs.3
‘I am freezing / getting frost-bitten.’
b. Mér kólnar
1sg.dat get cold.prs.3
‘I am getting cold.’
The verb kala in (94a) suggests direct impact on the body or physical damage,
while the inchoative kólna in (94b) denotes simply the feeling of getting cold.
One obvious difference between the two languages is the fact that 26% of the
meteorological verbs in Icelandic have dative coding, while Lithuanian does not
have any dative coded meteorological verb. This group includes Icelandic verbs
like hvolfa ‘capsize’, kyngja niður ‘fall thick’ (snow), slá niður ‘strike’ (lightning):
(95) Eldingu sló niður
lightning.dat struck.pst.3sg down
‘Lightning struck.’
Andrews (1982: 461–463) explained the difference between the acc and dat-cod-
ing of the argument as follows: when the argument undergoes a sudden or mo-
mentary movement it is marked with dat (96a); on the other hand, when this
event affects the argument gradually it is marked with acc (96b):
(96) a. Bátnum hvolfdi
boat.dat capsize.pst.3sg
‘The boat capsized.’
b. Bátinn rak að landi
boat.acc drift.pst.3sg to land.dat
‘The boat drifted to the shore.’
Another explanation has been put forth by Svenonius, who claimed that for an
acc-coded HRA the cause of the event is constantly present throughout the pro-
cess, while the initiating force is not active during the process when the HRA is
coded with the dat (Svenonius 2001: 214).28 Both the acc- and the dat-coded
HRA construction in the meteorological group have a nom-coded HRA counter-
part (transitive/intransitive counterparts):29
(96) c. Straumurinn rak bátinn
current.nom drift.pst.3SG boat.acc
‘The current drifted the boat.’
28. This also explains the coding of verbs in other lexical classes such as the acc coding for
the verb klæja ‘tickle’ (constantly present) and dat coding for blæða ‘bleed’ (only an initiating
force).
29. This causative pattern where nominative agentive subject of the transitive construction is
‘suppressed’ and the original object shows up as a subject preserving its original object case (cf.
example (96a–b)) accusative, dative, or more rarely genitive has recently been dealt with by
scholars. Cf. Ottosson (2009), who refers to them as impersonal detransitives, and Cennamo et
al. (2011), who refer to them as oblique active intransitive.
d. Straumurinn hvolfdi bátnum

Current.nom capsize.pst.3SG boat.dat
‘The current capsized the boat.’
Among two-place verbs there are three lexical groups with a dat-coded HRA:
psychological (i.e., emotive plus epistemic) verbs, fructitive verbs and ‘pain’-verbs
(on which see 3.2.2.9).
3.2.2.5 Verbs taking genitive arguments

All verbs with a gen-coded HRA are associated to some aspect of quantification:
apstėti, daugėti, gausėti ‘increase’, mažėti ‘decrease’ and all pri-prefixed verbs (see
Figure 6). There are a few more verbs referring to quantification which take a dat-
NP denoting a person, as in
(97) Eglei trūksta / pakanka kantrybės.
pn.dat lack.prs.3 suffice.prs.3 patience.gen
‘Eglė lacks / has sufficient patience.’
The argument status of the dat-NP is, however, arguable (and, correspondingly,
the two-place status of the verb). We can at least conclude that verbs with gen-
marked arguments are predominantly one-place, i.e. the obliquely marked HRA
is their only argument.
As for Icelandic, genitive-coded HRAs are very rare (see Figure 6). Further-
more, there is a striking difference in comparison to the dative and accusative in
that the gen-marked HRA is never an experiencer (which corresponds to the
x-argument of pred’, thus closer to the Actor than to the Undergoer pole). They
could at best be labelled as theme (which corresponds to the y-argument of pred’,
thus a position very close to the Undergoer pole). As for the verb þurfa ‘need, be
needed’ the gen-marked argument is not the person needing but the thing need-
ed; compare:
(98) Þess þarf ekki
This.gen need.prs.3sg neg
‘It is unnecessary.’
We have to treat separately those verbs which are related to existence (often negat-
ed), i.e., with a meaning that someone/something is or is not there. For instance,
missa við ‘be absent’, as in
(99) Þegar Jóns missir við
when pn.gen miss.prs.3SG with
‘When Jon is no longer here.’ (Jónsson’s list)
The verbs njóta ‘be present’ and gæta ‘be perceptible’ are most frequently used
with negation denoting non-existence:
(100) Hans nýtur ekki lengur við.
he.gen enjoy.prs.3sg neg longer with
‘He is no longer here.’ (Jónsson’s list)
(101) Vinds gætti ekki.
wind.gen notice.pst.3sg neg
‘There was no wind.’ (Jónsson’s list)
This use is similar to the gen of negation for subjects in Lithuanian as in
(102) Tėvo nebuvo namie

dad.gen neg.be.pst.3 at home
‘Dad was not at home.’
The difference is, though, that in Icelandic the genitive is restricted to certain
verbs kenna, gæta, and it does not occur with the verb ‘to be’.
Finally, there are verbs with similar semantics as geta ‘be mentioned’ and ken-
na ‘be noticeable’, which are used equally with or without negation:
(103) Hans getur víða í fornum sögum
he.gen mention.prs.3sg widely in old sagas
‘He is mentioned in many sagas.’ (Jónsson’s list)
3.2.2.6 Locations
There are some few lexemes that are difficult to judge, because they show al-
ternations of coding patterns which involve a NP naming a locality (cf.
Lenartaitė-Gotaučienė, this volume). The verbs knibždėti and kvepėti are repre-
sentative of such cases. Knibždėti belongs to the famous class of “swarm”-verbs
(104), kvepėti ‘smell’ is the neutral verb denoting olfactory perception (105). Both
show characteristic alternations of coding patterns:30
(104) a. Oras knibždėjo tuntais muselių / nuo muselių.
air.nom swarm.pst.3 troops.ins fly.gen.pl from fly.gen.pl
‘The air was swarming with flies.’
b. Ore knibždėjo (tuntai) muselių.
air.loc swarm.pst.3 troops.nom fly.gen.pl
30. All examples but (104c) and (105b) are quoted from Lenartaitė (2011: 146f.). Beside kvepėti,
other olfactory verbs (dvelkti, dvokti, smirdėti, trenkti, tvoksti ‘stink’) show the same pattern
alternation as in (105a–b).
c. Ore knibždėjo muselės.

air.loc swarm.pst.3 fly.nom.pl
‘Flies were swarming in the air.’
(105) a. Virtuvė / virtuvėje kvepėjo česnaku.
kitchen.nom / kitchen.loc smell.pst.3 garlic.ins
b. Virtuvėje kvepėjo česnakas.
kitchen.loc smell.pst.3 garlic.nom
‘The kitchen smelt of garlic.’ / ‘In the kitchen it smelt of garlic.’
The problem consists in the following. These verbs could basically be considered
as one-place predicates (with the generalized LS: exist’ (x)), with the indication of
the locality as an adjunct: any state or activity takes place just in some spatial set-
ting, so that there would not be any reason why the locality should be considered
as an argument of knibždėti or kvepėti. However, the possible syntactic promotion
to a nominatival subject suggests that the locality has to be included into the ar-
gument structure of these verbs (so that they should be given LSs of the general
structure: be-loc’ (x, y)). Icelandic, on the contrary, does not permit this coding
pattern readily, at least it is rare and highly marked. Thus, the following remarks
will be restricted to Lithuanian.
When it comes to the other NP, at first sight no serious doubt arises as to its
argument status, regardless of whether in (105a–b) the stimulus (or whatever may
be the label for this semantic role) is coded with the nominative (česnakas) or the
instrumental (česnaku) or, as in (104a–b), with the genitive (muselių) vs. the PP
(nuo muselių). The genitive with knibždėti is motivated by the feature of indefinite
quantity, which we will not comment upon further,31 while the instrumental with
kvepėti can be explained as a special subcase of inanimate instigator (cause, force,
reason), which we will discuss shortly below.
And yet, on closer inspection, the argument status of this NP turns out to be
less obvious for olfactory verbs. As long as they refer just to the ‘mode’ of percep-
tion, and irrespective of an evaluative (negative or positive) component, they can
be used as zero-place verbs; for instance, Čia (darže, virtuvėje …) kvepia/dvelkia
lit. ‘Here (in the garden / in the kitchen) (it) smells/stinks’ or Kaip gardžiai kvepia!
lit. ‘How tasty (it) smells!’.
Furthermore, the neutral verb denoting acoustic perception skambėti ‘sound’
requires a complement (?Čia skamba __ ‘Here (it) sounds __’; Salėje skamba
*(muzika) ‘In the room *(music) sounds’). A further difference in comparison
to olfactory and ‘swarm’-verbs lies in the fact that skambėti does not show any
31. For a detailed analysis from a different point of view cf. Lenartaitė (2011: 129–146; this
volume).
nom–ins-alternation; instead, it can code its argument only in the nominative.

These observations appear to correlate with conceptually conditioned asymme-
tries between perception predicates arranged on hierarchies like those known
from Viberg (1984). Among other things, verbs denoting modes of perception
that are highest (and thus cognitively salient) on such a hierarchy, i.e., visual and
acoustic impression (Lith. matytis ‘look like’, atrodyti ‘seem’ and skambėti ‘sound’,
respectively), do not show coding alternations for a locality like verbs denoting
impressions of smell or like ‘swarm’-verbs. Moreover, it should be investigated
whether some among the ‘swarm’-verbs cannot be regarded as a subclass of verbs
denoting acoustic impressions.32 But, anyway, in contrast to ‘neutral’ auditory
verbs ‘swarm’-verbs include a semantic component of indefinite, but high quan-
tity of the entity that is conceived as the source (cause) of that impression. This
additional component marks them off from ‘ordinary’ perception verbs.
3.2.2.7 Internal perception

The previous considerations concerned only external perceptual stimuli. When it
comes to internal, usually unpleasant experience, skambėti (contrary to the olfac-
tory verbs) can be used, too. But in this meaning a locative argument denoting a
body part is mandatory:
(106) a. Man skamba ir skamba ausyse.
1sg.dat sound.prs.3 and sound.prs.3 ear.loc.pl
‘(It) is ringing heavily in my ears.’ (lit. ‘To me (it) is sounding and sound-
ing in the ears.’) (LKŽe, sub verbo)
The locative should be considered as an argument, since its omission would leave
us with an utterance that can at best be understood elliptically (?Man skamba).
The locative case or a locative PP (see example (107)) anyway indicates a person’s
body part. Then, the person is coded with the dative and has to be treated as an
external possessor (see 3.2.2.1). The LS of skambėti as a predicate of internal per-
ception, thus, should look like
(106) b. feel’ (ausys, [painful’]), ausys ‘ears’.
32. This would contrast with Engl. swarm, which does not by itself denote an acoustic impres-
sion, as Wayles Browne commented: “Indeed, when bees swarm, they buzz, and one can hear
them. But many other things swarm and create no sound. E.g. The bay was swarming with
sharks. The river is swarming with fish. My head is swarming with ideas. The city was swarming
with secret agents. The house is swarming with bacteria.”
There are other verbs related to acoustic perception which show the same pattern
(and can be used with reference to external stimuli, too): burgzti, dūgzti, gausti
‘hum’, spengti ‘tingle’, ūžti ‘ring (in one’s ears)’. Compare:
(107) Man nesveika – galvą diegia, per
1sg.dat unhealthy.n head.acc stitch.prs.3 through
ausis burzgia.
ear.acc.pl hum.prs.3
‘I am feeling bad: it is itching my head, it is humming through my ears.’
(LKŽe, sub verbo)
(108) Man spengia galvoje.
1sg.dat tingle.prs.3 head.loc
‘I feel a tingling in my head.’ (A. Usonienė, p.c.)
(109) Man dūzgia ausyse.
1sg.dat hum.prs.3 ear.loc.pl
‘It is humming in my ears.’ (A. Razanovaitė, p.c.)
Basically, the same pattern can be observed with verbs of light emission if used to
denote some sort of deviant physiological state:33
(110) Jam juoduoja / žaliuoja akyse.
he.dat be/turn_black.prs.3 / be/turn_green.prs.3 eye.loc.pl
‘It is shining black / green in his eyes.’
As for Icelandic, the same pattern exists for verbs denoting light emission (111).
But there are no equivalents of Lithuanian verbs denoting acoustic or olfactory
perception, or of ‘swarm’-verbs, with non-canonical marking:
(111) Mér sortnaði fyrir augum
1sg.dat become_black.pst.3SG before eyes.dat.pl
‘I almost fainted (lit. (It) became black before my eyes.).’ (Jónsson’s list)
Now we are left with the fact that, apart from a few verbs of olfactory perception
(see 3.2.2.6), there are no arguable one-place verbs whose argument would be
coded with the instrumental. However, in Lithuanian the instrumental plays an
important role as a device denoting inanimate causers (natural force, reason),
which, in turn, shows affinity with the notion of stimulus or source (as with olfac-
tory verbs) and the notion of means.34 We now turn to this issue.
33. We owe this observation to Kristina Lenartaitė-Gotaučienė (p.c.), who also gave the example.
34. Cf. Narrog (2009: 597–599) and Malchukov and Narrog (2009: 529) on crosslinguistically
frequent polysemy patterns of the instrumental.
3.2.2.8 Instrumental
We should be reminded of rule [2] for the assignment of the instrumental in RRG
which we elaborated on at the end of Section 2.1.
In Lithuanian the instrumental occurs quite regularly with non-canonical
coding (112a–b) if the pattern can be assumed to be derived from a transitive
verb with a canonical pattern (113)–(114); compare:
(112) a. Vaiką (visą) išbėrė spuogais.
child.acc all.acc.sg cover_up.pst.3 pimples.ins.pl
‘The child got covered all over with pimples.’
b. Vaikui veidą išbėre spuogais.
child.dat face.acc cover_up.prs.3 pimples.ins.pl
‘The child’s face was covered with pimples.’
(113) Turbūt negeras vėjas užpūtė,
probably bad.nom.sg wind.nom.sg blow_out.pst.3
kad burną išbėrė šašai.
comp mouth.acc cover_up.pst.3 pimples.nom.pl
‘Probably a bad wind made him/her catch a cold, so that pimples covered his/
her mouth.’ (LKŽe, sub verbo)
(114) Klojimo nebėra, kur javus išbersim?
barn-floor.gen neg.ptc.be.prs.3 where grain.acc.pl pour_out.fut.1pl
‘There is no more any barn-floor laying; where should we pour out the grain?’
(LKŽe, sub verbo)
In (112a–b) the instrumental-NP spuogais stands in the same semantic relation

to the affected person and his/her body part as does the nominative-NP šašai
in (113), although the hierarchy of syntactic relations differs: the sentences in
(112a–b) do not have a PSA, whereas in (113) the PSA is marked with the nomi-
native, and it is the name of the disease – thus, reason or cause of the state denot-
ed – which can be promoted to a PSA without a change of grammatical voice. All
sentences in (112)–(113) have two core arguments, and there is no prototypical
(human) agent, and this is why the name of the cause surfacing in (112a–b) as
an instrumental-NP, can be given PSA-status at all in (113). Contrary to this,
sentence (114) implies a human agent, and the affected other referent (= Under-
goer, Dixon’s O) is coded with the accusative, but there is no instrumental-NP
(neither could it be added) denoting a cause, and this correlates with a difference
in meaning. For the time being, let us consider išberti in (114) as the first meaning
(= išberti1 ‘pour out’), išberti in (112)–(113) as its second meaning, išberti2 ‘cover
(with pimples, spots’).
However the meaning shift between išberti1 and išberti2 may be captured, the
instrumental is typically used as a coding device for participants that can be char-
acterized as causes (forces, reasons, sources), and this transfers to clause types
without a PSA (thus, non-canonical patterns of case marking). Although rule [2]
in 2.1 explains why a non-human agent (natural force etc.) can become a nomina-
tival subject, the crucial point is that this ‘promotion’ to subject (PSA) need not
occur. Presumably, this, in turn, can be explained from the fact that natural forces
are not the most typical agents. In other words: there is nothing in the grammar
of Lithuanian which forces non-typical agentive arguments (= x-argument of do’)
eo ipso to be coded as a PSA (in active voice), even in the absence of prototypical
agents (= argument of DO’).
The same holds for the instrumental in another prominent use, namely: as
denoting the means35 of an action (normally with nouns denoting substances);
see example (115):36
(115) a. Darbininkai užpylė duobę smėliu.
workers.nom fill_by_pouring.pst.3 pit.acc sand.ins
‘The workers filled up the pit with sand.’
b. Duobę užpylė smėliu. / Mašiną užpylė sniegu.
pit.acc fill.pst.3 sand.ins / car.acc cover.pst.3 snow.ins
lit. ‘(It) filled up the pit with sand.’ / ‘(It) covered the car with snow.’
c. Smėlis užpylė duobę. / Sniegas užpylė mašiną.
sand.nom fill.pst.3 pit.acc / snow.nom fill.prs.3 car.acc
‘The sand filled up the pit.’ / ‘The snow filled up the car.’
Curiously, verbs denoting flooding demonstrate another pattern. For them the
reason (cause) can be left out (it is lexically implied), but marking the natural
force with the instrumental sounds awkward (116b), while they permit to code it
with the nominative (116c) (K. Lenartaitė-Gotaučienė, p. c.):
(116) a. Pavasarį tas pievas užlieja / apsemia / užpila / užtvindo.
lit. ‘In spring (it) floods these meadows.’
35. For a distinction between instrument, means and natural force cf. already Apresjan (21995
[1974]); Russ. orudie, sredstvo, stixija, respectively). Equivalent labels figure as well in other
approaches in which elaborate inventories of semantic roles are assumed. Apresjan (ibid.) para-
phrases the means (sredstvo) as the participant which is used up and, thus, ceases to exist as
such (contrary to what happens to the other two kinds of entities).
36. Other verbs showing this behaviour are išpilti, išversti, išmesti ‘scatter, cover (with spots)’
(originally two-place) and užpustyti, užnešti, padengti ‘cover (up)’ (originally three-place)
(K. Lenartaitė-Gotaučienė, p. c.).
b. ?Pavasarį tas pievas užlieja / apsemia / užpila / užtvindo vandeniu.ins.

to mean ‘In spring (it) floods these meadows with water.’
c. Pavasarį vanduo.nom / upė / polaidis užlieja / apsemia / užpila / užtvindo
tas pievas.
‘In spring water / the river / thaw flood floods these meadows.’
In all these classes (illustrated in (112)–(113), (115)–(116)) the instrumental re-

mains optional, i.e., the cause of the change of state need not be indicated; in
many cases (as with diseases in (112)–(113)) it is already implied by the verb.
An explanation for the different behaviour of verbs denoting processes (or
events) in which natural forces are involved as causes might go as follows. The
pattern with nom–ins-alternation requires that at least one of the following two
conditions be fulfilled: either (i) the natural force meaning is derived from an
originally three-place verb (see example (115a–c)), or (ii) an originally two-place
verb has undergone a meaning shift (and thus become polysemous), as in (114)
vs. (112)–(113). The verbs in (116) do not fulfill either condition.
3.2.2.9 ‘Ache’-verbs
Finally, let us dwell on the tiny group of verbs for which the meaning ‘ache, hurt’
is not derived from another meaning (as with the verbs in 3.2.2.3), but primary. In
Lithuanian it consists of more or less five items:37 skaudėti, sopėti, mausti, peršėti
(of skin or wounds) and niežėti ‘itch’. LKŽe also gives knietėti in the ‘physiological’
meaning of ‘itch’,38 but such usage seems to be exceptional (obsolete and/or dia-
lectal) and we could not find a single example in the corpus of standard Lithua-
nian (LKT). The same applies to svembti ‘ache, hurt’ (rare, see LKŽe for examples,
not attested in LKT).
In Icelandic only the following verbs exist: verkja ‘hurt’, svíða ‘smart’, klæja
‘itch’. All of them have the experiencer marked with the accusative. Some of the
Lithuanian verbs named above are peculiar insofar as they allow the person to be
coded with the dative without an indication of the body part. As a consequence
their dative-NP does not represent an external possessor, but a genuine argument
of the verb (see the discussion in 3.2.2.1).
Lith. sopėti and skaudėti are almost full synonyms, and both show an alterna-
tive pattern dat–acc/nom:
37. To this list one could add derivatives of some of these verbs, e.g. skausti (ingressive of
skaudėti), panižti (ingressive of niežėti). They virtually do not differ from the source verbs, or
differ only in actionality, but not in their coding pattern.
38. Compare: Jau senai man tą ausį.acc knieta ‘Already for a long time it has been itching this
(⊃ my) ear’ (Bgt.).
(117) Man skauda ranką / ranka.

(118) Man sopa ranką / ranka.
1sg.dat ache.prs.3 arm/hand.acc / nom
‘My hand hurts.’ (lit. ‘To me (it) hurts the hand. / To me the hand hurts.’)
In the corpus we have found examples of the same alternation for mausti, peršėti
and niežėti, too.
The dative indicating the person (experiencer) can easily be omitted. This
occurs almost always if the affected body part is coded with the nominative; for
instance:
(119) Sopa krūtinė ir pilvo raumenys.
ache.prs.3 breast.nom and belly.gen muscle.nom.pl
‘(My) breast and the muscles of (my) belly ache.’
(120) Vidurius spaudžia, šaltis krečia, drebulys
guts.acc press.prs.3 frost.nom shake.prs.3 chill.nom.pl
maudžia, valgymas nelenda
hurt.prs.3 eating.nom neg.climb.prs.3
‘(It) presses the guts, the frost shakes, chill hurts, the meal doesn‘t go in.’
(R. Kalytis: Verksmo vieta, 1995)
Moreover, all verbs allow the affected body part to be coded with the locative or a
locational PP. See an example with niežėti:
(121) (…) man prie ausies niežti.
1sg.dat at ear.gen hurt.prs.3
‘It hurts me near my ear.’ (Z. Mačionis: Profesorius Kazys Daukšas, 2000)
As for other ‘parameters’ of coding the five verbs seem to be less homogeneous.
Let us consider the following possibilities: (i) Can the affected body part be omit-
ted and only the person (Experiencer) be coded with the dative? (ii) Can the per-
son be coded as accusative (and the affected body part be omitted)? As for the
first possibility, informants would readily use skaudėti without an explicit affected
body part, less so sopėti. Thus, (122a) sounds natural – e.g. in a physician‘s con-
sulting room before medical treatment starts – but less so (122b):
(122) a. Man skauda.
1sg.dat ache.prs.3
b. ?Jai sopa.
she.dat ache.prs.3
‘It hurts me / her.’
We have not come across examples of type (122b) for sopėti in the corpus, either.
Furthermore, such usage is judged even worse with respect to the other three
verbs:
(123) Ar tau *(žandą) labai maudžia?
q you(sg).dat jaw.acc very hurt.prs.3
‘Does your jaw hurt you very much?’
(124) Jam peršti *(akis / odą).
he.dat itch.prs.3 eyes.acc / skin.acc
‘It is itching his eyes / skin a lot.’
On the other hand, consider the following corpus example, which shows that
mausti does not always require the affected body part to be expressed (by courtesy
of K. Lenartaitė-Gotaučienė):
(125) Frunzės karo akademijos ženkleliu draskoma praporščikė, kuri ilgus metus
ištarnavo griežtojo režimo kolonijoje kontroliere. − Afonia, man maudžia!
Draugas papulkininki, neskriauskite, vaikai, anūkai namuose laukia…
‘The ensign of the Frunze military academy was rent with a small sign and she,
who had served for many years in a strict regime colony as controller, said:
Afonia, something is aching me! Lieutenant-colonel, don’t do me harm, my
children and grandchildren are waiting for me at home.’ (CCLL)
We have been unable to find any equivalent example for peršėti and have regis-
tered only three hits for niežėti.
Let us finish by briefly examining whether any of the five verbs can be used
with an Experiencer-NP in the accusative. According to our informants, this op-
tion does not sound very felicitous with any of the five verbs, unless the acc-NP
denoting the Experiencer is modified by the quantifier visas ‘whole, entire’:
(126) Mane visą niežti.
1sg.acc all.acc.sg ache.prs.3
‘(It) aches me all over.’ (R. Mikulskas, p.c.)
It appears that emphasis is necessary for the Experiencer to be coded this way. In
the corpus, occasional examples paralleling this variant could be found only for
mausti, e.g.,
(127) […] kartais taip maudžia visą,
sometimes so ache.prs.3 all.acc.sg
kad tiesiog nežinai kur dėtis.
‘Sometimes it aches (me) all over that you don‘t know where to go.’
(Š. Navickis: Diagnozė, Kaunas 2001)
The Icelandic pain verbs can without problem be used without a body part:
(128) Mig klæjar.

1sg.acc itch.prs.3sg
‘I am itching / It is itching me.’
When a body part is indicated, it is marked with the PP í+dat:
(129) Mig verkjar í fótinn.

1sg.acc hurt.prs.3sg in leg:acc
‘My leg hurts.’
4. Conclusions and outlook
Let us now sum up our preliminary findings which resulted from a systematic
survey of non-canonical HRA-marking in Lithuanian and Icelandic after the
main body of data had been assembled. The main conclusions are:
1. A rough comparison of the distribution of oblique cases over lexical groups

shows that Lithuanian and Icelandic differ in the frequency of using the da-
tive vs. the accusative, first of all, with physiological, emotive, meteorological
and fructitive verbs. The overall preponderance for the dative in Icelandic can
probably be explained, first of all, from the high proportion of fructitive verbs
with this case; the preponderance for the accusative in Lithuanian is evidently
primarily due to the high proportion of physiological verbs with this case (see
Figures 3, 4 and 5).
2. As a lexically required (i.e., adverbal) case the genitive occupies an altogether
minor choice (against dative- and accusative marked HRAs) in both languag-
es. But while the genitive is being driven out in Icelandic, in Lithuanian it has
a firm standing at least for two reasons. First, the genitive occurs regularly
with one- and two-place verbs implying an indefinite quantity and predomi-
nantly prefixed with pri-; however, only with one-place verbs (e.g., privažiuoti
‘arrive (about a certain amount of people)’) can we speak of genitive-marked
HRA, with two-place verbs (e.g., pririnkti ‘collect, gather (a certain amount of
people/things)’) the genitive marks the more patient-like argument (poten-
tial Undergoer), while the HRA is marked with the nominative. We have not
taken either of these verbs into account in our survey, so that figures on type
frequency may change to some extent when we take into account one-place
verbs prefixed with pri-. Second, in Lithuanian the genitive alternates regu-
larly with the nominative or accusative for reasons on clause- and reference-
related levels, but not as such inherent to the lexical meaning of the verb.39
Here we have not considered reference- and clause-conditioned case alterna-
tion at all.
3. The large number of acc-coded verbs denoting unpleasant physiological
experience in Lithuanian (see 3.2.2.3) is due to frequent (productive) meta-
phoric extensions from more basic physical activities, while in Icelandic such
metaphorical extension occurs only with few verbs. However, even in Icelan-
dic we can observe a certain preference for accusative-marked HRAs if the
verb denotes physical pain, in contrast to experiencers of (unpleasant) emo-
tional states which are preferably marked with the dative in both languages
(see 3.2.2.4).
4. External possessors are prominent in Lithuanian (marked with the bare da-
tive), but a rare option in Icelandic. One pattern is subject to severe syntactic
restrictions, and it has been turning increasingly obsolete (showing up, for
instance, in idioms); see 3.2.2.1. Another pattern exists with the PP í + dat,
which is used in non-canonical marking of HRA with predicates denoting
unpleasant physiological experience (accexp – í + datbodypart). This pattern,
in turn, is rare in Lithuanian, though it is attested with the same lexical class
of predicates (see 3.2.2.2). Lithuanian ‘physiological’ verbs usually show the
pattern datexp–accbodypart. This is in line with the Lithuanian preference for
external possessors coded with the dative.
5. In both languages there are only a handful of verbs denoting physical pain as
their primary (non-derived) meaning. These verbs differ among themselves
as to whether they allow the person to be coded with the dative (Lith.) or the
accusative (Icel.) without an indication of the body part or not. If they do,
the oblique NP can be considered a genuine semantic argument of the verb
(3.2.2.9). If the body part is coded, the pattern is the same as with the marking
of external possessors (datposs–accpossessee in Lithuanian). Remarkably, the
preference for the dative in Lithuanian, which we observe in these instances,
runs counter to the otherwise prominent accusative-marking pattern with
‘physiological’ predicates which mostly derive from verbs with lexical mean-
ings from other domains (first of all, some sort of basic physical action; see
3.2.2.3). This observation confirms the impression that verbs specialising in
general denotations of physical discomfort or pain are not only a very infre-
quent kind of lexical units, but their peculiar status can show up also in devia-
tions from otherwise prominent coding patterns of the same language. In the
39. For some details on this issue cf. Lenartaitė (2011: Ch 4.3) and Seržant (this volume). Nau
(this volume) discusses clause- and reference-related factors in Latgalian. For an overview cf.
also Koptjevskaja-Tamm and Wälchli (2001: 647–660).
case of Lithuanian the preferred way of coding can be explained by analogy

with the frequent pattern of the external possessor.
6. There is no equivalent to Dative Sickness (as explained in 3.2.1) in
Lithuanian.
7. Adpositional marking of core arguments proves rare in both languages.
Therefore, the different number of morphological cases retained from the
(ultimately) common IE heritage does not exert any serious effect on adposi-
tional marking, which has remained marginal in Icelandic as well.
8. Names of locations pose problems: on what criteria can we ascribe argument
status to them, especially if they are not marked with the nominative? This
problem is prominent with ‘swarm’-verbs and verbs of olfactory perception.
The conclusion that nominative-marked names of locations are to be regard-
ed as semantically derived is corroborated by the fact that this pattern occurs
so rarely in Icelandic (see 3.2.2.6).
9. A similar problem arises with the instrumental, which in Lithuanian has sev-
eral functions beyond marking instruments; it generally denotes causes (rea-
sons, natural forces) and means. Although such case roles occupy a specific
argument position in the logical decomposition of causative predicates (see
rule [2] and example (38a–c)), such an argument can be incorporated lexi-
cally (as with verbs meaning ‘flood’) with the consequence that it cannot be
expressed as instrumental HRA if a proper Actor (= a-argument in [2] and
(38c)) has been suppressed. More remarkable is the fact that in case of Ac-
tor-suppression the instrumental argument (= b-argument in [2] and (38c))
is not required to be ‘promoted’ to a PSA (nominatival subject). Moreover,
the resulting non-canonical case marking of the HRA (provided the instru-
mental can be counted as a core argument) is impossible with an instrumen-
tal denoting an instrument. In other words: only instrumental NPs denoting
other functions (as parts of causal event chains) than that of an instrument
(e.g., force, cause, means) allow for non-canonical marking of HRA. Another
requirement for the HRA to be coded by the instrumental is that the verbal
predicate is semantically derived either from an originally three-place verb
(typically verbs belonging to the ‘spray-paint’ series) or from a two-place verb
via some sort of metaphor-based polysemization (see 3.2.2.8).
Among the tasks for the nearer future we want to mention here only those two
which directly evolve from the discussion above:
10. A systematic assembly of logical decompositions for verbs with identical (or
very similar) argument structure. Coping with this task however presupposes
that one has clarified whether marginal case roles (like those mentioned in 7
and 8 above) ‘count’ as argument positions in logical structure.
11. Establishing rules for coding patterns based on productive derivation from
other classes of verbs (‘other’ both in terms of argument structure and lexical
meaning). This concerns, first of all, the patterns with the instrumental (see
3.2.2.8), but might also be relevant for many predicates with a dative-NP (not
discussed in this article).
Of course, the database itself has to be completed and the factors influencing vari-
able patterns of coding have to be specified with more precision and on the basis
of token-frequency, too.
Abbreviations
acc accusative loc locative

adv adverb (suffix) nom nominative
dat dative nonfin nonfinite
cl clause pl plural
comp complement pn personal name
fin finite prs present tense
gen genitive pst past tense
ins instrumental sbjv subjunctive
List of nonstandard abbreviations
AUH Actor-Undergoer hierarchy PSA Privileged Syntactic Argument

DS Dative Sickness RRG Role and Reference Grammar
HRA Highest-ranking argument
Sources
Jónsson, Jóhannes G. 1998. A list of Predicates that Take a Quirky Subject in Icelandic, Ms. Uni-
versity of Iceland.
DLKŽ – Dabartinės Lietuvių Kalbos Žodynas (2000).
LKG – Lietuvių kalbos tekstynas, http://tekstynas.vdu.lt/
LKŽ – Lietuvių kalbos žodynas, 20 vol., Vilnius.
LKŽe – Lietuvių kalbos žodynas, electronic variant, http://www.lkz.lt/
Sližienė, Nijolė. 2004. Lietuvių kalbos veiksmažodžių junglumo žodynas vols. 1–3. Vilnius:
Mokslo ir enciklopedijų leidybos institutas. (Referred to as NS in examples.)
References
Anderson, Cori. Forthcoming. Non-canonical case patterns in Lithuanian. In Contemporary

Approaches to Baltic Linguistics, Peter M. Arkadiev, Axel Holvoet, & Björn Wiemer (eds).
Berlin: Mouton de Gruyter.
Andrews, Avery. 1976. The VP complement analysis in Modern Icelandic. North Eastern Lin-
guistic Society 6: 1–21.
Andrews, Avery. 1982. The Representation of Case in Modern Icelandic. In The Mental Rep-
resentation of Grammatical Relations, Joan Bresnan (ed), 427–503. Cambridge, MA: MIT
Press.
Apresjan, Jurij D. 1967. Ėksperimental’noe issledovanie semantiki russkogo glagola. Moskva:
Nauka.
Apresjan, Jurij D. 21995 [1974]. Leksičeskaja semantika (sinonimičeskie sredstva jazyka).
Moskva: Škola “Jazyki russkoj kul’tury”.
Barðdal, Jóhanna. 2001. Case in Icelandic – A synchronic, diachronic and comparative approach.
Lund University dissertation. [Published as Lundastudier i nordisk språkvetenskap A57,
Lund: Department of Scandinavian Languages, 2001.]
Barðdal, Jóhanna. 2004. The semantics of the impersonal construction in Icelandic, German
and Faroese: Beyond thematic roles. In Focus on Germanic Typology, Werner Abraham
(ed), 101–130. Berlin: Akademie Verlag.
Barðdal, Jóhanna. 2006. Construction-specific properties of syntactic subjects in Icelandic and
German. Cognitive Linguistics 17(1): 19–106. DOI: 10.1515/COG.2006.002
Barðdal, Jóhanna. 2008. Productivity: Evidence from Case and Argument Structure in Icelandic.
Amsterdam/Philadelphia: John Benjamins. DOI: 10.1075/cal.8
Butt, Miriam. 2009. Modern approaches to case: an overview. In The Oxford Handbook of Case,
Andrej Malchukov & Andrew Spencer (eds), 27–43. Oxford: Oxford University Press.
Cennamo, Michela, Eythórsson, Thórhallur, & Barðdal, Jóhanna. Forthcoming. The Rise and
Fall of Anticausative Constructions in Indo-European: The context of Latin and Germanic.
Submitted to a thematic volume of Linguistics.
Croft, W. 1998. Event Structure in Argument Linking. In The Projection of Arguments. Lexical
and Compositional Factors, Miriam Butt & Wilhelm Geuder (eds), 21–63. Stanford: CSLI
Publications.
Dahl, Östen. 1985. Case Grammar and Prototypes. Duisburg: Linguistic Agency, University of
Duisburg, Series A 146.
Dixon, Richard M. W. 1991. A New Approach to English Grammar, on Semantic Principles.
Oxford: Clarendon Press.
Dixon, Richard M. W. 1994. Ergativity. Cambridge: Cambridge University Press. DOI:
10.1017/CBO9780511611896
Dowty, Richard 1979. Word meaning and Montague Grammar. Dordrecht: Reidel. DOI:
10.1007/978-94-009-9473-71
Dowty, Richard. 1991. Thematic proto-roles and argument selection. Language 67: 547–619.
Eythórsson, Thórhallur. 2000a. Dative vs. nominative: changes in quirky subjects in Icelandic.
Leeds Working Papers in Linguistics 8: 27–44.
Eythórsson, Thórhallur. 2000b. Fall á fallanda fæti [Unstable case marking]. Íslenskt mál og
almenn málfræði 22: 185–204.
Eythórsson, Thórhallur. 2002. Changes in subject case marking in Icelandic. In Syntactic Effects
of Morphological Change, David Lightfoot (ed), 196–212. Oxford: Oxford University Press.
DOI: 10.1093/acprof:oso/9780199250691.003.0011
Eythórsson, Thórhallur & Barðdal, Jóhanna. 2005. Oblique subjects: A common Germanic in-
heritance! Language 81(4): 824–881. DOI: 10.1353/lan.2005.0173
Friðriksson, Finnur. 2008. Language Change vs. Stability in Conservative Language Communi-
ties: A Case Study of Icelandic. PhD Dissertation, Gothenburg University.
Halldórsson, Halldór. 1982. Um méranir: Drög að samtímalegri og sögulegri athugun [About
dativizings: preliminaries to a synchronic and diachronic investigation]. Íslenskt mál og
almenn málfræði 4: 159–189.
Haspelmath, Martin. 1999. External Possession in a European areal perspective. In External
possession Typological Studies in Language 39], Doris L. Payne & Immanuel Barshi (eds),
109–135. Amsterdam/Philadelphia: John Benjamins.
In Non-Canonical Marking of Subjects and Objects [Typological Studies in Language 46],
Alexandra Y. Aikhenvald, Richard M. W. Dixon, & Masayuki Onishi (eds), 53–83. Amster-
Haspelmath, Martin. 2011. On S, A, P, T and R as comparative concepts for alignment typology.
Linguistic Typology 15: 535–567. DOI: 10.1515/LITY.2011.035
Hopper, Paul J. & Thompson, Sandra A. 1980. Transitivity in Grammar and Discourse. Lan-
guage 56: 251–95.
Jónsson, Jóhannes G. 1996. Clausal Architecture and Case in Icelandic. Amherst: University of
Massachusetts dissertation.
Jónsson, Jóhannes G. 2003. Not so quirky: on subject case in Icelandic. In New Perspectives on
Case and Case Theory, Ellen Brandner & Heike Zinsmeister (eds), 127–164. Stanford: CSLI
Publications.
Jónsson, Jóhannes G. & Eythórsson, Thórhallur. 2005. Variation in subject case mark-
ing in Insular Scandinavian. Nordic Journal of Linguistics 28 (2): 223–245. DOI:
10.1017/S0332586505001435
Keenan, Edward. 1976. Towards a universal definition of ‘subject’. In Subject and Topic, Charles
N. Li (ed), 303–333. New York etc.: Academic Press.
König, Ekkehard. 2001. Internal and external possessors. In Language Typology and Language
Universals, Martin Haspelmath et al. (eds), vol. 2, 970–978. Berlin/New York: de Gruyter.
Lenartaitė, Kristina. 2011. Argumentų raiškos alternavimas lietuvių kalboje [Alternations in Ar-
gument Realization in Lithuanian]. Unpublished PhD thesis, Vilnius.
Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. Chicago/
London: University of Chicago Press.
Lyons, John. 1977. Semantics (2 vols). Cambridge: Cambridge University Press.
Malchukov, Andrej & Narrog, Heiko. 2009. Case polysemy. In The Oxford Handbook of Case,
Andrej Malchukov & Andrew Spencer (eds), 518–534. Oxford: Oxford University Press.
Maling, Joan. 2002. Það rignir þágufalli á Íslandi [Verbs with dative objects in Icelandic].
Íslenskt mál 24: 31–105.
Minger, David L. 2002. An Analysis of Grammatical Relations and Case Marking in Icelandic.
MA Thesis, University of California, Davis.
Narrog, Heiko. 2009. Varieties of instrumental. In The Oxford Handbook of Case, Andrej
Malchukov & Andrew Spencer (eds), 593–600. Oxford: Oxford University Press.
Onishi, Masayuki. 2001. Introduction: Non-canonically marked subjects and objects: Parame-
ters and properties. In Non-Canonical Marking of Subjects and Objects [Typological Stud-
ies in Language 46], Alexandra Y. Aikhenvald, Richard M. W. Dixon, & Masayuki Onishi
Ottosson, Kjartan. 2009. The Anticausative and Related Categories in the Old Germanic Lan-
guages. Ms. University of Oslo.
Payne, Doris & Barshi, Immanuel. 1999. External Possession [Typological Studies in Language
39]. Amsterdam/Philadelphia: John Benjamins. DOI: 10.1075/tsl.39
Primus, Beatrice. 2009. Case, grammatical relations, and semantic roles. In The Oxford Hand-
book of Case, Andrej Malchukov & Andrew Spencer (eds), 261–275. Oxford: Oxford Uni-
versity Press.
Rögnvaldsson, Eiríkur. 1983. Þágufallssýkin og fallakerﬁ í íslensku [Dative sickness and the
case system in Icelandic]. Skíma 16: 3–6.
Saj, Sergej S. 2011. Nekanoničeskoe markirovanie aktantov mnogomestnyx predikatov: opyt
kvantitativno-tipologičeskogo issledovanija. Acta Linguistica Petropolitana, vol. vii, part 3,
Nikolaj N. Kazanskij (ed), 424–430. Sankt-Peterburg: Nauka.
Saj, Sergej S. Forthcoming. Bivalent verb classes in the languages of Europe: A quantitative
typological study. Submitted to Language Dynamics and Change.
Seefranz-Montag, Ariane von. 1983. Syntaktische Funktionen und Wortstellungsveränderung.
Die Entwicklung “subjektloser” Konstruktionen in einigen Sprachen. Munich: Wilhelm Fink.
Seržant, Ilja. A. 2013a. Rise of canonical objecthood with the Lithuanian verbs of pain. Baltic
Linguistics 4: 187–211.
Seržant, Ilja A. 2013b. The diachronic typology of non-canonical subjects and subject-like
obliques. In The Diachronic Typology of Non-canonical Subjects [Studies in Language Com-
panion Series 140], Ilja A. Seržant & Leonid Kulikov (eds), 313–360. Amsterdam/Philadel-
phia. John Benjamins.
Sigurðsson, Halldór Á. 1989. Verbal Syntax and Case in Icelandic. Lund: Lund University dis-
sertation. [Published: Reykjavík: Institute of Linguistics, University of Iceland, 1992.]
Smith, Henry. 1994. Dative Sickness in Germanic. Natural Language and Linguistic Theory 12:
675–736. DOI: 10.1007/BF00992930
Smith, Michel B. 2001. Why quirky case really isn’t quirky (Or how to treat dative sickness
in Icelandic). In Polysemy in Cognitive Linguistics, Hubert Cuyckens & Britta E. Zawada,
Svavarsdóttir, Ásta. 1982. ‘Þágufallssýki [‘Dative Sickness’]’. Íslenskt mál og almenn málfræði 4:
19–62.
Svenonius, Peter. 2001. Case and event structure. ZAS Papers in Linguistics 26: 197–217.
Thráinsson, Höskuldur. 1979. On Complementation in Icelandic. Harvard University disserta-
tion, Cambridge, MA [Published: New York: Garland, 1979.]
Thráinsson, Höskuldur. 2007. The Syntax of Icelandic. Cambridge University Press. DOI:
10.1017/CBO9780511619441
Van Valin, Robert D., Jr. 1991. Another look at Icelandic case marking and grammatical rela-
tions. Natural Language and Linguistic Theory 9: 145–194. DOI: 10.1007/BF00133328
Van Valin, Robert D., Jr. 2005. Exploring the Syntax-Semantics Interface. Cambridge etc.:
Cambridge University Press. DOI: 10.1017/CBO9780511610578
Van Valin, Robert D., Jr. & LaPolla, Randy. 1997. Syntax. Structure, meaning and function. Cam-
bridge: Cambridge University Press. DOI: 10.1017/CBO9781139166799
Vendler, Zenon. 1967 [1957]. Linguistics in Philosophy. Ithaca: Cornell University Press.
Viberg, Åke. 1984. The verbs of perception: A typological study. Explanations for Language
Universals, Brian Butterworth, Bernard Comrie, & Östen Dahl (eds), 123–162. Berlin:
Mouton de Gruyter.
Wiemer, Björn. 2013. Značimost’ sposobov modifikacii glagol’nyx osnov dlja ocenki areal’noj
differenciacii baltijskix jazykov (po sravneniju s rjadom slavjanskix mikrojazykov).
In Tipologija slavjanskix, baltijskix i balkanskix jazykov (v svete jazykovyx kontaktov),
Vjačeslav Vs. Ivanov & Petr M. Arkad’ev (eds), 220–246. Sankt-Peterburg: Aletejja.
tions: The Icelandic passive. Natural Language and Linguistic Theory 3: 441–483. DOI:
10.1007/BF00133285
Language index
A H 313–329, 331–343, 345–346,

Adyghe 89 Hindi 23, 130 348–351, 352, 354–356
Ancient Greek 263, 267, 270 Hungarian 98 Lithuanian dialects (Aukštaitian)
Australian languages 34, 43, 71, 286
45, 73–74, 76, 79, 90, 92 I Lower Sorbian 72
Icelandic 23, 26, 36–37, 46, 97,
C 112–113, 117–118, 125, 301–307, M
Czech 71, 107, 144, 148, 156–157, 313, 315, 319, 321–330, 332–339, Middle English 126
160, 168 341–346, 348, 351, 354–356
Indo-European 1–2, 76, 100, N
D 290 North-West Caucasian 89
Danish 181, 184, 194–195 Nyamal 76–78, 82, 91
Dhalandji 73 J
Djapu 78 Jiwarli 76–77 O
Dutch 144, 146, 149–150, 168, Old Church Slavonic 71–72
195, 203 K Old Czech 71
Kayardild 74–76, 79, 86, 91 Old Lithuanian 5, 71
E Old Russian 71
English 3, 16, 18, 24, 27, L
36, 46, 55, 88, 91, 101, 113, Lardil 79 P
121–122, 126–127, 138, 144, 146, Latgalian 2, 9–11, 13, 15, 19–20, Pama-Nyungan 73, 76–78
149–150, 153, 155, 164–167, 35, 45, 70–71, 76, 86, 207–215,
218, 223–225, 227–231, 234– R
170, 181, 183–185, 187, 189–190,
236, 240–253, 355 Russian 23, 47, 71, 74, 79–80,
194–196, 199–203, 261, 270–
Latin 22, 128–130, 211 89–90, 99, 106, 114, 122, 139,
271, 273, 277–278, 283–284,
Latvian 1–11, 13–23, 25–28, 148, 156–157, 160, 168, 171–172,
286, 290, 292, 294, 304, 312,
30–31, 34, 70, 97–102, 106–111, 181–183, 196–197, 199, 211, 213,
314, 330
113–114, 118, 123, 129–131, 224, 251, 260–261, 268, 275,
Estonian 70, 280
207–211, 215, 217–218, 224– 278, 280–282, 289, 291–294,
F 225, 227–229, 231, 241–247, 304, 316
Faroese 123, 125–126, 195–196, 253, 291, 331
S
203, 330 Lithuanian 1–7, 9–11, 13–23,
Slavic languages 11, 71–72, 235,
Finnish 8–9, 108, 130–131, 208, 30–31, 34–37, 43–45, 47–51,
238, 247
257, 271, 274–275, 278–281, 55–56, 58–61, 66–74, 76,
Slovene 72, 86
284, 287, 294 79–80, 83, 86, 90–92, 97, 99,
South Asian languages 23
106, 118, 125, 137–144, 146,
G 148, 150, 153, 155–157, 160–161,
T
German 12, 23, 27, 46, 107, 113, 163–178, 181–183, 186, 189, 191,
Tangkic 74–75, 79
149, 168, 181, 195, 199, 316, 193, 196–201, 203, 207, 209,
319, 330 218, 227–228, 235–236, 247, W
Germanic languages 3, 23, 97, 253, 257–261, 263, 271–278, Warumungu 73
113, 330 280–282, 284, 286–287, 289,
291–294, 301–306, 308–309,
Name index
A F Krifka, M. 264, 273, 275,

Allen, C. L. 126 Fennell, T. G. 107 279–280
Andrews, A. 23, 113, 302, Filip, H. 11, 247, 275, 279–280 Kroch, A. 184, 186
342–343 Franks, St. 14, 43, 45, 49, 59,
90, 291–292 L
B Fried, M. 144, 148, 156–158 Lambrecht, K. 140–142, 161,
Babby, L. 47, 294 163–167, 173
Barðdal, J. 23, 113, 117, 302, 329 G Landau, I. 46
Béjar, S. 46, 79 Gast, V. 141, 165 Langacker, R. 186–188, 190,
Bickel, B. 22–24, 26 Gāters, A. 211, 227 192–193, 198–200
Bielenstein, A. 101 Lavine, J. 14, 43, 45, 49, 59, 90,
Borer, H. 275–277 H 193, 291–292
Bossong, G. 207, 259 Haas, F. 141, 165 Leikuma, L. 214–215, 231–232
Bukšs, M. 209, 211, 214–215, Haspelmath, M. 3, 26, 30, 121, Livitz, I. 139, 171–173
218, 231, 246 124, 130, 208, 212, 293, 305,
Burzio, L. 123 310, 313, 318 M
Heycock, C. 184, 186, 189–190, Malchukov, A. L. 6, 20, 93,
C 195–196, 202 108, 121, 176, 208, 240, 304,
Campbell, L. 131 Hoeksema, J. 144, 146, 149–150, 319, 349
Carlson, G. N. 270, 294 168 Maling, J. 23, 112, 302
Chomsky, N. A. 45–48, 91, 185 Hopper, P. 2, 19, 107, 110, Massam, D. 46, 79
Cibuļs, J. 214–215, 231 126–127, 170, 208, 211, 313, 336 Matushansky, O. 43, 45–46,
Comrie, B. 14, 108, 121, 260 Huumo, T. 8, 278–279, 287 73–74, 79–80, 82, 88, 90–91
Croft, W. 316, 335–336 McFadden, Th. 46, 48, 70
Culicover, P. W. 116 I Merchant, J. 46, 91
Ivanov, M. 71, 280 Mühlenbach, K. 4, 12, 111
D
Dahl, Ö. 278, 281, 288, 310 J N
de Haan, F. 98 Jackendoff, R. 116, 144 Narrog, H. 99, 101, 107–108,
de Hoop, H. 108, 208, 259 Jónsson, J. G. 302, 319, 327, 319, 349
Dench, A. 34, 45, 73, 77–78 329, 331 Nichols, J. 22
Dixon, R. M. W. 97, 305, 311, Noonan, M. 101, 223
K Næss, Å 3–4, 6, 24, 211–212,
318–320, 323, 350
Kagan, O. 11, 213, 224, 252, 240
Donohue, M. 3
268, 290
Dowty, D. 138, 144, 146, 153,
Keenan, E. L. 23, 112, 139, 192, O
155, 170, 273, 312, 317
271, 303 Ogawa, A. 176
E Keine, St. 46, 80 Onishi, M. 29, 97, 126, 208, 318,
Eide, K. M. 105 Kiparsky, P. 9, 257, 264, 274– 320, 323
Endzelīns, J. 4, 99, 111, 291 276, 278–280
P
Erschler, D. 43, 45, 73, 79–80, Kittilä, S. 122, 259
Paducheva, E. 157
82, 85–86, 90 Koptjevskaja-Tamm, M. 6,
Partee, B. 11, 184, 268, 271,
Evans, N. 34, 45, 73, 75–76 281, 355
289, 294
366 Name index
Perkova, N. 18, 97 Sližienė, N. 319 U

Pollard, C. J. 116 Stowell, T. 74, 91 Ulanowska, St. 209, 218, 228
Prellwitz, W. 100 Strods, P. 231–234, 246, 251
Preminger, O. 46, 48 Svenonius, P. 343 V
Van Valin, R. D. 170, 310–312,
R T 315–316
Rappaport, G. 17, 114, 138, 144, Tatevosov, S. 280 Vendler, Z. 272, 312
165–166, 170 Thompson, S. A. 2, 19, 107,
Richards, N. 73, 79, 92 126–127, 170, 208, 211, 313, 336 W
Thráinsson, H. 23, 112, 302, 333 Wälchli, B. 6, 281, 355
S Timberlake, A. 14, 107, 197 Wichmann, S. 3
Sag, I. A. 116 Tsunoda, T. 19–20, 212, 240,
Saj, S. 303–304 Z
245
Sands, K. 131 Zaenen, A. 23, 112, 302, 329
Subject index
A indirective vs. secundative modal 74, 76, 80

ablative 75, 92, 241, 290–291 121 morphological 46, 80, 301,
accomplishment 272–273, 275, allative 76, 86, 92 305, 316
278, 280, 312, 325 apposition 183 oblique 30, 78, 107, 116, 126,
accusative 2–14, 18, 20, 22, arguments 148, 156, 160, 165, 178, 303,
26–27, 29–33, 35–37, 43–44, adpositionally marked 303, 309, 314, 327, 344, 354–355
46–47, 50–52, 55–56, 59, 331, 356 relational 74, 78
61–64, 66, 69–70, 72–74, core 183, 239, 243, 301, 303, resolution 79–82, 90–91
79–87, 92, 99, 102–104, 108, 312, 314, 317–318, 331, 349, stacking 43, 78–79, 123
110–112, 122–131, 178, 192–193, 356 structural 3, 46–51, 258–259,
204, 207–208, 210–225, non-canonically marked 5, 271, 285, 314
228–243, 245–253, 259–261, 36, 97, 101, 208, 301, 303 theory 43, 45–49, 55, 59, 91
269–271, 275, 277, 279–280, oblique 29, 97–98, 107, 116, case assignment 34, 43, 45,
282, 284–289, 292–293, 295, 121–123, 126, 312 47–51, 56, 58, 60, 69, 70,
302, 305, 309, 328–330, 333, 73–74, 79–80, 83, 86,
335, 338, 341–344, 350–357 B 90–91, 107, 294, 301
achievement 272, 312, 322 backtracking 200 multiple 43, 45, 70, 79–80,
actionality 271–274, 278, boundedness 36, 235, 257, 266, 91
280–281, 287, 351 271–272, 275–276, 278–281, case marking 1–3, 6, 8, 10, 12,
activity 142–143, 150, 152–153, 287 19–20, 27, 32, 34, 36, 43,
155, 167, 170, 219, 272–273, 45, 53, 56, 59–61, 63, 66,
C 69–70, 72–75, 79, 82, 85,
278–281, 312, 315, 317, 325–326,
Canonical Typology 33 90–91, 97, 101, 107–109,
335–336, 346
canonical marking, canonicity 112, 119–121, 123–124, 126,
Actor 16, 36, 301, 305, 310–315,
3, 5, 33–34, 36, 97–98, 131, 101, 128, 130, 139, 140, 149,
317–318, 338, 340–341, 344,
208, 301–303 152–153, 173, 176, 183, 193,
356–357
case 197, 208, 212, 217, 237–238,
Actor-Undergoer Hierarchy
adnominal 74 243, 258, 269, 271, 282,
(AUH) 305, 309, 311–314,
associating 43, 73–74, 78, 284–285, 293–294, 301, 314,
316–317, 331, 338, 340, 357
86, 90 319, 330, 350, 356
adjunct 54, 68–69, 75, 89,
attraction 183 multiple 79, 90–91
317–318, 346
choice, clause-related c-command 91, 118, 185–186
affectedness 6, 12, 240, 314
conditions on 207, 212, class inclusion 188–189
agnation 35, 191
310, 313 Cognitive Grammar 1, 35,
agreement 25, 35, 46, 73, 78,
choice, reference-related 181, 186–188, 191–193, 199,
104–105, 109–110, 119, 140,
conditions on 212, 313 202–203
173, 183, 195–196, 211, 283, 301
competition 47–49 communicative perspective
Aktionsart 12, 272, 278, 312–313
complementizing 34, 45, 194, 199, 202
alignment 2–3, 31, 33, 121–122,
73–74, 77–78, 82 connectivity effects 181, 185
183–184, 186, 192, 194–196,
discriminatory 108 constituency 54, 61, 194, 199,
198–200, 202–203,
hierarchy 126–127 201
302–303, 305, 309–310, 313,
inherent 46, 48–51, 59–61, construal 27, 141, 192–193,
315–316, 318
69 198–199
semantic 3
368 Subject index
content requirement 35, 181, equative predication 184, 189, highest-ranking argument
186, 193–194, 202–203 198 (HRA) 301, 309, 312, 314–
control of reflexivization 116, ergative 77–78, 92, 130, 319 315, 317, 324, 326–329, 330–331,
118 experiencer 12, 21–30, 98, 105, 333–336, 339–344, 354–357
copular construction 107, 121–123, 126, 131, 173, 245,
ascriptional 188 290, 311, 313, 318, 332, 337–338, I
equative 186 341–342, 344, 351–354 idioms 219, 355
class-inclusional 188 external possessor 321, 332–333, indefinite quantity 9–10,
descriptional-identifying 35 336–337, 348, 352, 356 231–233, 261, 321, 325, 346, 354
identificational 189 individualization 153–154
specificational 181, 182, 187, F inferential evidentials 193
190, 201–204 functional heads 46, 48–49, infinitival clauses 13–14, 34,
counterfactual conditionals 74, 86, 90 43–45, 50–51, 55, 59–63,
225 67–68, 82, 131
G infinitive 6, 13–14, 34–35, 37,
cumulative 62, 207, 235, 247–
Generalized Hierarchy Principle 43–45, 49–57, 59–72, 76,
248, 250, 264–265, 275–277,
127 80, 82–90, 92, 98, 103,
279
genitive 3–11, 13–15, 17–20, 32, 107, 131, 178, 212, 216–217,
D 34–38, 43–57, 59–64, 66– 221–223, 227–231, 246, 250,
dative 3–6, 11–14, 20–30, 34, 72, 75–76, 79–82, 86–90, 253, 291, 295, 324, 357
36–37, 43–47, 49–57, 59, 92, 108, 110–112, 125, 131, of purpose/goal 44, 63, 82,
61–63, 65–70, 73, 76–78, 147, 149–150, 152–154, 161, 84, 88
81–87, 90, 92, 94–95, 169, 175–178, 193–194, 204, instantial semantics 187–188
98–99, 103–105, 107, 109, 207–208, 210–222, 224– instrumental 4–5, 37, 46, 50,
113–124, 126–128, 131, 138, 232, 234–243, 245–253, 63, 75, 79, 92, 158, 178, 197,
178, 204, 212, 240–241, 257–258, 261–262, 268, 204, 295, 314, 316–319, 334–
245–246, 252–253, 295, 275, 289–295, 320–321, 329, 335, 346, 348–351, 356–357
304, 308, 314, 316, 319, 321, 343–346, 354–355, 357 intonational contour 201
324, 327–330, 332–333, independent 88–90 irrealis 6, 11, 98, 131, 212, 225,
336–337, 341–344, 347, 352, intensional 213 227, 250, 252
354–357 negation 6–7, 9, 47–48, 56,
dative drift 22, 26–28 60, 79–82, 110–112, 213, J
of recipient/experiencer 122 224, 252–253, 268, 293 judgments
dative sickness 26, 329–330, partitive 9–10, 17, 35–36, categorical 140, 161
356–357 147, 149–150, 152–154, thetic 140, 177
dative-plus-infinitive 43, 161, 169, 175–177, 207, 213,
252–253, 257–258, 261–262, K
45, 49–50, 52, 54–57, 59, 63,
289–290, 292–294 kind reference 270
65–66, 68–70, 76, 82–83,
86–87, 90 of purpose 292, 294
L
debitive 14–15, 17, 34, 37, 97– quantificational 62
landmark 35, 182, 187, 192–194,
114, 116–117, 120–121, 123–125, genitive-plus-infinitive 43,
198, 200, 202–203
127–131, 243–244, 248, 253 45, 49–50, 52, 54–57, 59, 61,
landmark/trajector
Default Macrorole Assignment 63–64, 66–70, 72, 76, 82,
configuration 35, 182, 187,
Principles 315, 341 86–90
193–194, 198, 200, 203
definiteness 10, 92, 181, grammatical relations
Layered Structure of the Clause
189–191, 194, 234, 238, diffuse 34, 97–98, 117,
(LSC) 310
257–258, 313 126–127, 130
locality 48, 82, 90, 185–186,
role 181, 189–191, 194
H 345–347
habitual 234–235, 251, 272, locative 4–5, 16, 20, 37, 78, 92,
E
285–289, 295 131, 138, 148, 150, 153, 161, 163,
elative 78, 92
165–167, 170, 177–178, 191,
equation 188–189, 198
Subject index 369
204, 241, 253, 295, 337, 347, 142–143, 147–150, 152, predicate-focus constructions
352, 357 156–158, 171, 173, 175–176, 141–142, 161
locative inversion 165–167 178, 192–194, 197, 204, presentational sentences 137,
logical decomposition (logical 207, 211–213, 231, 238–241, 140, 141–143, 149, 153, 155,
structure, LS) 140, 304, 243–246, 252–253, 259, 157–158, 160–169, 171–172,
310–317, 333–334, 340, 346– 262, 270, 277, 294–295, 174–177, 321
347, 356–357 302, 304–305, 309, 314–317, entity-central 140
330–331, 334, 343, 346–347, event-central 141, 149, 157
M 349, 351, 352, 355, 356–357 formulaic 149, 168
macroroles 310–312, 314–315, nominative sickness 330– presupposition
318 331 contextual 185, 187, 191, 201
mass nouns 9, 35, 207, 231–234, existential 181, 289–290
236, 247, 251–252, 259, 269, O Primary-A, Primary-B
277, 279, 284 object predicates 320
Minimalism 1, 34, 60 differential marking of 28, Principle of Detopicalization
modal affix 98, 108 36 164
modal target 99, 105–106 direct 4, 6–7, 12, 14–15, 18, Principle of Subject–Object
modal verb 20, 23, 29–34, 36, 43–44, Neutralization 165
argument structure of 99, 46, 48, 50–51, 54–57, 59, Privileged Syntactic Argument
101–103, 107, 113, 125, 329 67, 70–72, 80, 82, 84, (PSA) 311–312, 314, 317,
impersonal 99, 102 88, 92, 97, 111–112, 116, 349–350, 356–357
modality 121–122, 124, 130, 141, processing 69, 187, 192, 200–
event 101, 105, 107 170–172, 207–208, 218, 201, 203
deontic 105 221–225, 230–235, 237–239, processing time 192
propositional 100 242–243, 245–247, 249, pronouns 3, 7–9, 14, 25–26,
mode 212, 252, 346 251–253, 258–259, 273–274, 36, 99, 108, 113–114, 197,
mood 6, 74, 92, 98, 100–101, 279, 281–282, 286, 289, 210, 213–218, 220, 222,
108, 130–131, 208, 218, 225 290, 294 224, 236–237, 241, 243, 253,
movement 58, 61, 69, 83, 144, indirect 3, 22, 50, 67, 69, 82, 257–258, 268
146–147, 168, 185, 201, 290, 84–85, 121–122 indefinite 237
292, 343 nominative 14, 107–108, personal 210, 214, 216–217,
M-transitivity (-intransitivity, 129, 131 220, 222, 237, 253, 268
-atransitivity) 315 shift 51, 58, 63, 66 proto-roles 317
obliqueness 31, 97, 116, 122, pseudoclefts 181, 183, 185–186
N 126–127
necessitive 14, 34, 98, 106–107, adjustments 97, 127 Q
131 mismatch 127 quantification 4, 10–11, 257,
negation 6–7, 9, 11, 15, 32, 44, o-command 116 275, 320–321, 344
47–48, 56, 60, 79–82, 92, Optimality Theory 80, 108
110–112, 195, 204, 212–213, R
216, 218, 220–225, 230, P reconceptualization 200
252–253, 261, 268, 270, 289, passive 37, 92, 101, 124, 128, 131, reference point 167, 198, 272
293–294, 345 178, 185, 193, 204, 243–245, rhetorical questions 230
pragmatically implied 225 248, 253, 295, 319 Role and Reference Grammar
negative focus particle 7 passivization 68, 121, 125, 128 1, 36, 357
negative pronouns 7–9, 224 percolation 75, 91
negative raising 223 Phase Impenetrability Condition S
nominative 2–3, 6, 9, 11, 14–16, 48–49, 56, 58–59, 70 sentence-focus constructions
18, 21–25, 27, 29–33, 36–37, phase theory 91 140, 161
43–44, 46–47, 49–50, 80, pivot/controller properties 120 sequential scanning 192
92, 99, 102, 104, 107–113, polarity 185, 208, 212 small clauses 184
115–121, 123–131, 139–140,
370 Subject index
specificational relation 35, 181, subjecthood V

187, 190, 198–203 discourse 139–140, 173, 175 verbs
specifier 48, 60, 80, 190, 194, grammatical 23, 26, control 323–324
196–201, 203 139–140, 142, 173 egressive 326
split intransitivity 22 recoverable 121, 128 fructitive 322, 329, 341,
starting point 200, 203 tests 23, 117–118 344, 354
state 12, 170, 182, 272, 286, 312, thematic 139–140, 173, 175 incremental-theme 257,
316–317, 333–334, 338, 342, subjectivity 225 271, 273–274, 281–282, 284,
346, 348–349, 351 subordination 90, 92 287, 294
stimulus 23–25, 27–28, 31–32, Suffixaufnahme 79 ingressive 326, 351
126, 290, 311, 313, 335, 346, 349 supine 13, 35, 37, 70–72, 76, of motion 34, 44, 51, 55,
subject 2, 6, 14–16, 18, 21–25, 86, 88–90, 92, 207, 212, 217, 59–61, 63, 67, 70–71, 76,
28–35, 46, 53, 66–69, 227–228, 250–251, 253 86, 88–89, 227, 292
80, 82, 84–85, 92, 97–98, syncretism 3, 214–215 mutative 325–326
104–105, 107–109, 111–113, prefixed 13, 207, 236, 251,
115–124, 126–130, 137, T 326, 327
139–143, 146, 148–150, temporary transfer 259, 261, raising 184
153–156, 159–161, 163–166, 274, 278, 286–287 secondary 323
169–178, 181, 183–185, theta-role hierarchy 173 semelfactive 12, 312, 326
187–188, 191–201, 203, 207, topicalization 110, 117, 119 swarm-type 146, 161, 167,
211–212, 239–240, 243, topic-comment structure 117– 316, 345, 347–348, 356
246, 252, 257, 259, 262, 265, 118, 173, 176, 187, 189–190, viewpoint aspect 36, 272–273,
274–275, 277, 293–294, 303, 199, 201 280, 287
311, 317, 331, 335, 343, 346, trajector 35, 182, 187, 192–194, voice 21, 101, 107, 127–128, 245,
350, 356 198, 200, 202–203 304–305, 349–350
dative 67–68, 84 transitivity
demoted intransitive 97–98, hierarchy of 19 W
121–123, 126, 128–130 semantic 2–3, 6, 127, 240, word order 16, 24–25, 28, 43,
differential marking of 28 252, 336 50–51, 60–61, 64, 66–67,
logical 184, 311 type-instance relation 188 69, 82–83, 116–117, 122, 181,
oblique 23, 122 189–191, 195, 199, 201–202,
U 227–228
prototypical 23, 35, 137, 139,
Undergoer 36, 301, 305,
143, 170–171, 173, 175–176,
310–313, 315–316, 318, 331, 338,
178
340–341, 344, 350, 355, 357
syntactic 24, 149, 184
unique instantiator 197–198
Compiled by Erika Jasionytė-Mikučionienė

Axel Holvoet & Nicole Nau - Grammatical Relations and Their Non-Canonical Encoding in Baltic

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Axel Holvoet & Nicole Nau - Grammatical Relations and Their Non-Canonical Encoding in Baltic

Uploaded by

Copyright:

Available Formats

Grammatical Relations

and their Non-Canonical Encoding in Baltic

John Benjamins Publishing Company

the American National Standard for Information Sciences – Permanence

Library of Congress Cataloging-in-Publication Data

© 2014 – John Benjamins B.V.

Argument marking and grammatical relations in Baltic: An overview 1

Case and word order in Lithuanian infinitival clauses revisited 43

Non-canonical grammatical relations in a modal construction:

Alternations in argument realization and problematic cases

Subjecthood in specificational copular constructions in Lithuanian 181

Differential object marking in Latgalian 207

The independent partitive genitive in Lithuanian 257

On the non-canonical marking of the highest-ranking argument

Language index 363

Axel Holvoet*, *** and Nicole Nau**, ***

2. Alignment and case marking in Baltic

The Baltic languages use the nominative-accusative type of marking of gram-

2.1 Case marking in Baltic: The repertoires

counted for in semantic terms, such as marking experiencers or beneficiaries”

(2) ‘fall in love’, Lithuanian original: accusative

(3) ‘believe’; Lithuanian original: instrumental

2.2 Differential object marking in transitive clauses: Alternatives

2.2.1 Transitivity-related differential marking

(6) Tiesa, ne vis-i žmon-ės gali ger-ti kav-ą […]

The genitive of negation is an old inherited feature in Lithuanian; it is also well

(9) Lidz ar to prasīb-ai par KNAB vadītāj-a atkāpšan-os

5. The interview was online at http://www.ltv.lv/lv/raksts/09.09.2013-grende-noliedz-ka-

(15) Gavau laišk-ų.

blow. Mühlenbach (1902/1903: 222) characterized the difference as one of aktion-

2.2.2 Clause-type-specific differential object marking

Central Latvian dialect with supine (Raģe 1964: 117)

Standard Lithuanian (Ambrazas, ed. 1997: 557)

Lithuanian (Ambrazas, ed. 1997: 557)

In this volume, Peter Arkadiev analyses Lithuanian purpose constructions with

Central Latvian dialect (Endzelin 1922: 409)

(32) Bērniem jālasa grāmata.

To call this nominative an object would, however, be problematic. The debitive

2.2.3 Differential subject marking

An interesting question is whether these sentences contain instances of the same

types of factors associated with deviant encoding – individual lexical require-

3. Baltic in a typological context: Alternatives to the transitive pattern

8. Cf. Berg-Olsen (1999, 2000).

b. Tāpat kā bit-e meklē med-u vis-os

A more pervasive alternative marking strategy occurs when an experiencer

Some non-verbal predicates compete with a middle voice (formally reflexive)

(59) Es viņ-am patīku.

As can be seen in the last two examples, agreement is independent of word-or-

Word-order is thus not a subject parameter in Latvian, unless subject is equated

The reflexive possessive pronoun may be controlled by the Experiencer as well,

(65) [man] apnīk sav-s skat-s spogul-ī

expressing a stimulus or theme is extended to predicates that were originally used

Note that the word-order in both examples is Experiencer–Stimulus, which cer-

4. Differential and non-canonical marking

similar reasons. While we are justified in speaking of non-canonical marking of

want to call the accusative in (74) an object because in research on differential

(82) Nuo ryški-os švies-os jam skaudėjo ak-is.

6. Concluding remarks and chapter summaries

acc accusative nom nominative

LiLa = Lietuviešu-latviešu-lietuviešu paralēlo tekstu korpuss. Availabe online at: http://www.

Lithuanian, being a nominative-accusative language, encodes the direct object of

(1) a. Jon-as perskait-ė laišk-ą.

(iii) In impersonal matrix constructions (usually headed by emotional or modal

the American National Standard for Information Sciences – Permanence

Argument marking and grammatical relations in Baltic: An overview 1

Case and word order in Lithuanian infinitival clauses revisited 43

Subjecthood in specificational copular constructions in Lithuanian 181

Differential object marking in Latgalian 207

The independent partitive genitive in Lithuanian 257

Language index 363

Axel Holvoet, and Nicole Nau, ***