Sie sind auf Seite 1von 342

Typological variation in grammatical relations

Dissertation zur Erlangung des akademischen Grades doctor philosophiae (Dr. phil.) Eingereicht an der Philologischen Fakultät der Universität Leipzig

von Alena Witzlack-Makarevich

Leipzig, im Juni 2010

Contents

Acknowledgements

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

viii

Abbreviations

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

ix

1 Introduction

1

1.1 Grammatical relations

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

1

1.2 Goals and overview

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

7

2 Major developments in the study of grammatical rela- tions

11

2.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

11

2.2 Anderson (1976)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

11

2.3 Keenan (1976)

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

13

2.4 Dixon (1994, 2009)

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

14

2.5 Role and Reference Grammar

 

19

2.6 Radical Construction Grammar

 

26

2.7 Bickel (in press) and the present approach

 

28

2.8 Conclusion

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

33

3 Methodological background

 

35

3.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

35

3.2 The goals of modern linguistic typology

 

35

3.3 Multivariate approach

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

36

3.4 Summary and outlook

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

39

4 Argument structure and semantic roles

 

41

4.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

41

4.2 Valence and argument structure

 

41

4.3 Semantic roles

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

48

4.3.1 Role-list approaches to semantic roles

 

48

4.3.2 Decompositional approaches to semantic roles

 

49

4.3.2.1 Dowty (1991) .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

50

4.3.2.2 Primus (1999)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

55

iv

CONTENTS

4.3.3 Argument roles: wrapping up and looking forward 56

4.4

Conclusion

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

59

5 Argument selection

 

61

5.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

61

5.2 Argument selection and alignment

 

61

5.3 Argument selection and alignment splits

 

65

5.4 Conclusion

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

70

6 Dependent marking and referential properties

 

73

6.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

73

6.2 Role and reference aspects of argument

 

74

6.3 Referential properties

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

75

 

6.3.1 Animacy and humanness

 

78

6.3.2 Definiteness and specificity

 

81

6.3.3 Lexical class

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

82

6.3.4

Person

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

85

6.3.5

Other referential features

 

87

6.4 Referential hierarchy

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

88

6.5 Interaction of features

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

90

6.6 Inflectional classes

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

93

6.7 Conclusion

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

94

7 Dependent marking and predicates

 

99

7.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

99

7.2 Generalized predicate classes and the default class

 

100

7.3 A note on argument realization

 

114

7.4 Split and fluid intransitivity

 

118

 

7.4.1 Split intransitivity

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

119

7.4.2 Fluid intransitivity

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

131

7.5 Conclusion

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

136

8 Dependent marking and clausal conditions

 

139

8.1 Introduction

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

139

8.2 Tense, aspect and mood

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

140

8.3 Morphological form of the predicate

 

144

8.4 Clause type (main vs. subordinate clause)

 

147

8.5 Polarity

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

149

8.6 Scenario

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

150

8.7 Conclusion

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

157

CONTENTS

v

9

Head marking

159

9.1 Introduction

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

159

9.2 On the nature of agreement

160

9.2.1 Introduction

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

160

9.2.2 Feature matching vs. feature unification

 

162

9.2.3 On marking and its locus

165

9.2.4 Grammatical versus pronominal agreement

 

169

9.3 Argument selection and head marking

172

9.4 Head marking and argument referential properties

175

9.5 Head marking and predicate class

177

9.6 Head marking and conditions

179

9.7 Head marking and hierarchical alignment

 

181

9.8 Conclusion

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

194

10Diathesis alternations

195

. 10.2Voice and diathesis

10.1Introduction

. 10.3Passive and antipassive as argument selectors

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

195

195

198

10.3.1Argument promotion and argument demotion

 

201

10.3.2Splits of argument promotion and demotion

 

205

10.4Diathesis as a condition on argument selection

206

10.5On syntactic transitivity of derived clauses

214

10.6Conclusion

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

216

11Variables of argument selectors

.

11.1Introduction

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

217

217

. 11.2Coding vs. behavior

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

218

. 11.4Mono-clausal vs. cross-clausal

11.3Locus of marking

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

220

221

11.5Clause-linkage type

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

223

11.6Controller vs. controllee

231

11.7Matrix predicate

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

233

11.8Argument treatment

237

11.9Conclusion

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

239

12Discussion and conclusion

241

. 12.2Overview of the system of variables and their values

12.1Introduction

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

241

242

12.3Deriving traditional alignment

 

248

12.4On the interaction of argument selectors

256

12.5Database realization

263

vi

CONTENTS

12.6Conclusions and prospects for further research

 

271

A Language sample

 

275

B Examples of database coding

 

281

B.1

Ritharngu

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

281

B.2

Itzaj Maya

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

283

B.3

Puma

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

284

B.4

B.5

. Imbabura Quechua

Udihe

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

284

286

References

 

289

Acknowledgements

The research reported here was supported by grants BOBO 2471/3-1 and BI 799/3-1 (“Typological variation in the processing of grammatical rela- tions”) from the German Research Foundation as part of the research group “Grammar and processing of verbal arguments” (FOR 742). First and foremost, I wish to express my gratitude to Balthasar Bickel, who is responsible for arousing my interest in linguistic typology. Through- out the entire process of writing this thesis, I have benefited immensely from

his ideas, professional enthusiasm, and inspiration. And thanks a lot for in- volving me in all the exciting research projects before, during, and after writing this thesis.

I would like to express my deepest gratitude to Johanna Nichols, who

spent a lot of time and energy in carefully reading the draft of this thesis and providing precious advice and encouragement.

I would further like to thank my “third” supervisor Bernard Comrie for

the discussion and prompt reading of the pre-final draft of this thesis. Many other people have accompanied me during writing this thesis and provided me with their support. Martin Haspelmath always made time for me when I needed an advice and it was a great pleasure to discuss with him issues related and unrelated to this thesis. For valuable theoretical dis- cussions thanks to Fabian Heck, Jeff Good, and Andrej Malchukov. I also profited from comments by audiences of several conferences and seminars, such as ALT7 (Paris), ALT8 (Berkeley), the Work in Progress series at the MPI-EVA, and the Typological Colloquium series at the University of Leipzig.

This thesis also benefitted from the input I received at all the formal and in- formal meetings of the RHIM-EuroBABEL research project and the Syntactic Typology Reading Group at the MPI-EVA. Language-specific thanks to Peter Cole, Spike Gildea, Antoine Guillaume, Alice Harris, Iren Hartmann, Gabriella Hermon, Silvia Kutscher, Zarina Molochieva, Irina Nikolaeva, Simon Overall, Donald Stilo, and Thomas Wier. Needless to say, none of them is to blame for any errors in the interpretation of the data.

viii

CONTENTS

Many thanks to my fellow PhD students Falko Berthold, Birgit Jänen, Robert Schikowski, and Matthias Urban, who read the drafts of this thesis at various stages. Another round of thanks to Robert Schikowski for exten- sive discussions and argumentations at an early stage of writing-up. Thanks to Diana Forker and Diana Schackow for valuable comments on individual chapters. Thanks to Martina Ernszt for translating the theses into German and to Kai and again Robert for polishing them. Database and LaTeX thanks to Lennart Bierkandt, Sven Siegmund, and particularly to Taras Zakharko. Social thanks to my friends and former and present University of Leipzig and MPI EVA colleagues: Birgit, Christfried, David, Dejan, Diana, Eva, Jenny, Falko, Fabian, Lennart, Matthias, Martina, Robert, Dr. Schiering, Taras, Zaira, Zarina, and many others. Even more special thanks to Tanya for all the practical support, Kai for his emotional support, and C. for her intellectual support.

Abbreviations

1,2,3

person

A

(unless specified otherwise) most agent-like argument

ABL

of transitive (and ditransitive) predicate ablative

ACT.PTCP

active participle

ADD

additive focus

ADJ

adjective

ADV

adverb(ial)

AGR

agreement

ALL

allative

anim

anim

ANTIP

antipassive

ASP

aspect marker

ATT

attenuative Aktionsart

AUX

auxiliary

BEN

benefactive

BM

boundary marker

CLF

classifier

CLM

clause-linkage marker

COM

comitative

COMP

complementizer

COMPL

completive

COND

conditional

CONT

continuous

COMPL

completive status

COP

copula

CVB

converb

d

dual

DAT

dative

DECL

declarative

DEF

definite

DEM

demonstrative

DIR

direct

x

CONTENTS

DIST

discourse particle

DIST

distal

DS

different subject

DUR

durative

DYN

dynamic

EMPH

emphatic

EPEN

epenthetic

ERG

ergative

e

exclusive

EXCL

exclusive

EXCLA

exclamation

EXPL

expletive

EXP

experiential aspect

exper.

experiencer

F

feminine

FIN

final

FV

final vowel

FOC

focus

G

most goal-like argument of ditransitive clause

GEN

genitive

GEN.LOC

generic locative

h

honorific

h.anim

higher animate

HIAF

high affectedness Aktionsart

IDEOPH

ideophonic

i

inclusive

IMP

imperative

inanim

inanimate

INCOMPL

incompletive status

IND

indicative

INDEF

indefinite

INDIV

individuation particle

INF

infinitive

INS

instrumental

INTERJ

interjection

INTR

intransitive

INTS

intensive Aktionsart

IPFV

imperfective

IRR

irrealis

l.anim

lower animate

LAT

lative

LOC

locative

M

masculine

CONTENTS

xi

MIR

mirative

MOM

momentaneous

N

noun

N-

non- (e.g. NPST nonpast)

n.a.

not applicable

NATIV

nativizer

NEARPST

near past

NEG

negation, negative

NEP

Nepali

NEUT

neuter

NMLZ

nominalizer/nominalization

NOM

nominative

ns

nonsingular

OBL

oblique

OBLIG

obligative

ON

superessive

OPT

optative

ORIG

originative

p / pl.

plural

P

patient-like argument of canonical transitive verb

PFV

perfective

POLINT

polar interrogative

POSS

possessive

PRED

predicative

PREVERB

preverb

PRF

perfect

PROB

probability

PROG

progressive

PROH

prohibitive

PROX

proximal/proximate

PRS

present

PST

past

PTCL

particle

PTCP

participle

PUNCT

punctual

PURP

purposive

PVB

preverb

Q

question particle/marker

REAL

realis

RECIP

reciprocal

RECPST

recent past

REFL

reflexive

REM

remote

xii

CONTENTS

REP

reportative

RES

resultative

RESTR

restrictive

s / sg.

singular

S

single argument of canonical intransitive verb

SEQ

sequential

SG

singular

SIM

simultaneous

SREL

subject relativizer

SS

same subject

T

most theme-like argument of ditransitive clause

UT

unspecified time

TEL

telic

TNS

default tense

TOP

topic

TR

transitive

V

verb, verbal derivational affix

VALID

validator (= first hand witness knowledge)

VOC

vocative

Chapter 1

Introduction

1.1 Grammatical relations

Grammatical relations are a broad and ill-defined notion and, in principle, this term can be used to refer to any grammatical dependency relations. In practice, however, it is used to denote the relations between a clause or a predicate and its arguments. 1 The major relational categories subject, di- rect object, and indirect object are among the most basic concepts of many models of grammar and are often regarded, either explicitly or implicitly, as universal (cf. Chomsky 1981; Dixon 1994, 2009a). Moreover, they belong to the fundamental concepts in descriptions of most languages. Chung, for in- stance, claims that ‘all languages have rules referring to subject and direct object, which are central to the syntactic organization’ (Chung 1978: 99f.). If, by contrast, a linguist finds that the categories of subject and object are not useful or applicable for the description of a language, this decision re- quires explicit justification (cf. Durie 1985, 1987 on subject in Acehenese; Nakayama 2001 on object in Nuuchahnulth).

Until the early 1970’s, surface morphological criteria, such as case mark- ing, agreement, and constituent order, played a pivotal role in identifying in- dividual grammatical relations. To a certain extent the tradition of regarding

1 Synonymously to ‘grammatical relation’ or ‘grammatical function’ some lin- guists use such terms as ‘syntactic function’ (Falk 2006) or ‘syntactic role’ (Croft

2001).

2

1.1 Grammatical relations

these morphosyntactic properties as major tests of subjecthood and object-

hood has survived until today and can be vividly illustrated by the following

statement:

[I]t should be possible to identify the linguistic clues ensuring

communication; e.g. the clues helping to distinguish the subject

and the object. [

I want to consider the following three fac-

tors, belonging to different grammatical levels, which may help

to identify the elements of the sentence: the organising power of

verbal valency; the nominal and verbal inflection and the word

order. It will be shown that these factors cooperate in order to

facilitate the identification of the subject and the direct object.

(Schøsler 2001: 273)

]

If the German sentences in (1) are considered from this perspective, ein

Mann ‘INDEFsM.NOM man’ and der Hund ‘DEFsM.NOM dog’ can be iden-

tified as the subject, as they trigger agreement on the verb (suffix -t for

‘3sPRS’) and the definite articles are in the nominative case, whereas einen

Mann ‘INDEFsM.ACC man’ in (1b) can be identified as the object of the

clause, as it does not trigger agreement on the verb and bears the accusative

case marking: 2

(1) German

a. Ein

Mann

starb.

INDEFsM.NOM man ‘One man died.’

die.1sPST

b.

Der

DEFsM.NOM dog

ge-biß-en. PTCP-bite-PTCP ‘The dog has bitten a man.’

Hund

ha-t

einen

Mann

have-3sPRS INDEFsM.ACC man

2 Glossing follows the Leipzig Glossing Rules (http://www.eva.mpg.de/lingua/ files/morpheme.html) with some additions (see Abbreviations). I took the liberty to standardize the original abbreviations in examples reproduced from other sources and, if necessary, to modify the glosses in accordance with the Leipzig Glossing Rules. Examples without a reference to a source were elicited from native speakers.

1.1 Grammatical relations

3

The 1970s saw a fundamental change in the discussion of grammatical

relations. It was triggered by an increasing interest in languages exhibit-

ing ergative traits and appearance of a number of important descriptive

accounts. Thus, Dixon’s (1972) grammar of Dyirbal was followed by a dis-

cussion of a range of previously largely ignored languages with ergative

characteristics (e.g. Comrie 1973, 1979c on Chukchi; Blake 1976 on some

Australian languages; Woodbury 1977 on West Greenlandic Eskimo). As in

these languages morphological criteria do not identify subjects familiar from

most European languages, some of the frequently asked questions at that

time were whether such languages (as well the so-called Philippine-type and

active languages, cf. Schachter 1976) have a subject at all and which of the

arguments it is.

As an illustration of the problems arising when trying to identify gram-

matical relations in languages with ergative characteristics, consider the

following examples from Kâte, one of the languages discussed in Anderson

(1976) when examining possible criteria of subjecthood in such languages

(cf. Section 2.2). In Kâte, the only argument of an intransitive verb is in the

unmarked absolutive, such as ŋic ‘man.ABS’ in (2a). In transitive clauses,

one of the arguments is also in the absolutive case, such as ŋic ‘man.ABS’

in (2b), whereas the other argument is marked with the ergative marker,

such as qâto-zi ‘dog-ERG’ in the same example: 3

(2) Kâte (Trans New Guinea; Papua New Guinea; Pilhofer 1933: 44)

a.

b.

hâmo-jec.

man.ABS one die-NEARPST.3sS/A ‘One man died.’

qâto-zi

dog-ERG man.ABS bite-NEARPST.3sS/A ‘The dog has bitten the man.’

ŋic

moc

ŋic

ki-jec.

If we compare Kâte examples in (2) with the corresponding German clauses

in (1), it turns out that the two categories established by case marking in

3 Not every A argument in Kâte is marked with the ergative marker, its distribu- tion is conditioned by a range of factors including rhematicity and part of speech distinction (for a recent account, see Suter 2010+). These details are, however, irrelevant for the present discussion.

4

1.1 Grammatical relations

Kâte (ergative-marked qâto-zi ‘dog-ERG’ on the one hand and absolutive-

marked ŋic ‘man.ABS’ on the other) do not correspond to the two cate-

gories of subject and direct object established by case marking in German

(nominative-marked einen Mann ‘INDEFsM.ACC man’ and der Hund ‘DEFs-

MASC.NOM’ versus accusative-marked den Mann ‘DEFsMASC.ACC man’).

As Anderson summarizes this discrepancy, in ergative languages like Kâte

“the morphology appears to establish the existence of a category which in-

cludes subjects of some verbs, and objects, but not subjects of other verbs”

(Anderson 1976: 3).

While attempting to answer the question what is the subject and object

in languages with ergative traits like Kâte, it became a common practice to

extend the inventory of tests beyond morphological marking and word or-

der and to include syntactic processes, such as Equi-NP deletion, raising,

conjunction reduction, and the behavior of the reflexives, as diagnostics of

grammatical relations (for the discussion of some other tests see the con-

tributions in Li 1976 and Plank 1979). For instance, in the discussion of

Kâte, Anderson considers a type of clause-chaining which is characterized

by marking all but the last chained clause with special verb forms indicating

the relative temporal relation of each clause to the following one, such as

sequential and simultaneous verb forms in (3). Moreover, if the intransitive

subject or the agent argument of a transitive clause is identical to the intran-

sitive subject or the agent argument of the following clause, the verb of the

first clause is not marked for the person and number of this argument, such

as the verbs hone-lâ ‘see-SEQ.SS’, gasacke-lâ ‘run-SEQ.SS’, and lo-lâ ‘take-

SEQ.SS’ in (3a). If, however, the intransitive subject or the transitive agent

argument of the following clause has a difference reference, the verb of the

preceding clause is marked for the person and number of its intransitive

subject or the transitive agent argument, such as lo-ha-me ‘take-SIM-3sDS’

and hone-pe ‘see-SEQ.1sDS’ in (3b):

(3) Kâte (Trans New Guinea; Papua New Guinea; Pilhofer 1933: 35f.)

a.

bec

pig.ABS see-SEQ.SS run-SEQ.SS gun.ABS take-SEQ.SS mulu-nʒa-ŋ.

shoot-IRR-2s

‘If you see a pig, you would run, take a gun, and shoot it.’

hone-lâ

gasacke-lâ

tepe

lo-lâ

1.1 Grammatical relations

5

b. fiuc

lo-ha-me

hone-pe

wise-wec.

illegitimately take-SIM-3sDS see-SEQ.1sDS flee-REM.PST.3s ‘While he was stealing, I saw him, and he ran away.’

This construction is used by Anderson (1976) to argue that Kâte has a sub-

ject after all and it can be identified on the basis of switch-reference markers

distribution. In contrast to German, Kâte subject is not case-marked consis-

tently.

Though common in the investigation of grammatical relations, the prac-

tice of picking out a particular morphosyntactic feature of a range of possi-

bilities and then treating it as providing the only right diagnostics requires

a strong motivation. Anderson’s explanation of ranking switch-reference

marking above case marking in case of Kâte is based on the postulated pri-

macy of syntactic tests over morphological tests for determining grammat-

ical relations. But is it a valid explanation? Moreover, this approach runs

into a number of other problems. First, though Anderson does not discuss

Kâte any further, one could consider other syntactic constructions of Kâte

and it is quite possible that they might provide support for a different kind

of subject than the one identified by the switch-reference construction. In

cases of a mismatch, the uncomfortable question of which one to choose or

how to weight different criteria becomes unavoidable (for examples and dis-

cussion see Van Valin and LaPolla 1997; Croft 2001; for the discussion of this

problem as presented by ditransitives see Hudson 1992; Malchukov et al.

2007). Second, to identify the subject in Kâte Anderson proposes to rely

on the evidence from switch-reference marking. But other languages (e.g.

English and German) do not have any construction reminiscent of the Kâte

switch-reference marking. In such languages other constructions are tradi-

tionally used to determine the grammatical relations. But now the question

arises whether it is legitimate to equate the grammatical relations estab-

lished on the basis of different criteria. Why should the relation established

by switch-reference marking in Kâte be associated with the relation estab-

lished by conjunction reduction in German or English?

Such an approach to characterizing language structures has been

severely criticized for suffering from what Croft labels ‘methodological op-

portunism’, i.e. using “language-specific criteria when the general criteria

do not exist in the language, or when the general criteria give the “wrong”

6

1.1 Grammatical relations

results according to one’s theory” (Croft 2001: 30). Being logically inconsis- tent, ad hoc and often circular, cross-linguistic methodological opportunism is not a rigorous scientific method for discovering the properties of language structures and thus must be abandoned (Croft 2001).

One of the accepted alternatives to the methodological opportunism in the domain of grammatical relations research is to consider all morphosyn- tactic properties of grammatical relations without prioritizing among them. Once this equitable approach is accepted, one has to recognize that various morphosyntactic features of arguments (i.e. what used to serve as subject and object tests) do not necessarily identify a single set of grammatical re- lations (e.g. one subject and one object or one ergative and one absolutive) in a language. Instead, every single construction can, in principle, estab- lish a different grammatical relation. Thus, instead of viewing grammati- cal relations as uniform categories, it became common to regard them as construction-specific categories (cf. Comrie 1978b; LaPolla 2006; Moravc- sik 1978a; Van Valin 1981, 1983, 2005; Van Valin and LaPolla 1997; Dixon 1994; Palmer 1994; Croft 2001; Bickel 2004, 2010+b).

The increasing interest in s