Since the publication of F. R.

Palmer's Mood and Modality in 1986,

when the topic of 'modality' was fairly unfamiliar, there has been con-
siderable interest in the subject as well as in grammatical typology in
general. Modality is concerned with mood (subjunctive, etc) and with
modal markers such as English modal verbs (can, may, must, etc.) and is
treated as a single grammatical category found in most of the languages
of the world. In this new edition, Palmer draws on a wealth of examples
from a wide variety of languages to investigate the category of modality
in detail. He discusses familiar features in a number of mainly European
languages, and also looks at less familiar features including 'evidential'
systems and the contrast of realis/irrealis, both to be found in unrelated
languages.

F. R. PALMER is Emeritus Professor of Linguistic Science at the Univer-

sity of Reading and is a Fellow of the British Academy. He is the author
of innumerable articles and many books, including Semantics (2nd
edition, 1981), Mood and Modality (1986) and Grammatical Roles and
Relations (1994), also published by Cambridge University Press. He has
lectured all over the world including in the USA, Brazil, India, China,
Japan, Australia and most of Europe.
CAMBRIDGE TEXTBOOKS IN LINGUISTICS

General editors: s. R. A N D E R S O N , J . B R E S N A N , B . C O M R I E ,
W. DRESSLER, C. J. EWEN, R. HUDDLESTON, R. LASS,
D. LIGHTFOOT, J. LYONS, P . H. MATTHEWS, S. ROMAINE,
N. V. SMITH, N. VINCENT

MOOD AND MODALITY

Second Edition
In this series
p. H. M A T T H E W S Morphology Second edition
B. C O M R I E Aspect
R. M. K E M P S O N Semantic Theory
T. B Y N O N Historical Linguistics
j . A L L W O O D , L . - G . A N D E R S O N a n d . D A H L Logic in Linguistics
D . B . F R Y The Physics of Speech
R . A . H U D S O N Sociolinguistics Second edition
A . J . ELLIOTT Child Language
P . H . M A T T H E W S Syntax .
A . R A D F O R D Transformational Syntax <-
L. B A U E R English Word-Formation
s. c . L E V I N S O N Pragmatics
G. B R O W N a n d o . Y U L E Discourse Analysis
R . H U D D L E S T O N Introduction to the Grammar of English
R. LASS Phonology
B . C O M R I E Tense
w . KLEIN Second Language Acquisition
A . J . W O O D S , P . F L E T C H E R and A . H U G H E S Statistics in Language
Studies
D. A. CRUSE Lexical Semantics
F. R . P A L M E R Mood and Modality
A . R A D F O R D Transformational Grammar
M. G A R M A N Psycholinguistics
w . C R O F T Typology and Universals
G. G . C O R B E T T Gender
H . j . G I E G E R I C H English Phonology
R. C A N N Formal Semantics
p. J . H O P P E R a n d E . c . T R A U G O T T Grammaticalization
j . L A V E R Principles of Phonetics
F . R . P A L M E R Grammatical Roles and Relations
B . B L A K E Case
M. A . J O N E S Foundations of French Syntax
A . R A D F O R D Syntactic Theory and the Structure of English: a Minimalist
Approach
R. D . VAN VALIN, J R , and R . J . L A P O L L A Syntax: Structure,
Meaning and Function
A . D U R A N T I Linguistic Anthropology
A . C R U T T E N D E N Intonation Second edition
J. K . C H A M B E R S and P . T R U D G I L L Dialectology Second edition
C. LYONS Definiteness
R. KAGER Optimality Theory
J . A . H O L M An Introduction to Pidgins and Creoles
c . G . C O R B E T T Number
c . J . EWEN andH. VAN DER H U L S T The Phonological Structure of Words
F . R . P A L M E R Mood and Modality Second edition
MOOD A N D MODALITY
SECOND E D I T I O N

F. R. PALMER
Professor Emeritus, Department of Linguistic Science,
University of Reading

CAMBRIDGE
UNIVERSITY PRESS
P U B L I S H E D BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge, United Kingdom

CAMBRIDGE UNIVERSITY PRESS

The Edinburgh Building, Cambridge CB2 2 RU UK www.cup.cam.ac.uk
40 West 20th Street, New York, NY IOOI 1 421 I USA www.cup.org
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
Ruiz de Alarcn 13, 28014 Madrid, Spain

Cambridge University Press 2001

This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements, ' "
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2001

Printed in the United Kingdom at the University Press, Cambridge

Typeface 9.5pt/13pt Times System 3b2 [CE]

A catalogue for this book is available from the British Library , ,..

ISBN 0 521 80035 8 hardback

ISBN 0 521 80479 5 paperback
To
ANNA, ELIZABETH and RACHEL
(not forgetting RICHARD and CLAIRE)

c
 CONTENTS

Preface
Note on the text
List of abbreviations

1 Introduction
1.1 Basic concepts
1.1.1 Realis and irrealis
1.1.2 The notion of assertion
1.2 Two basic distinctions
1.2.1 Mood and modal systems
1.2.2 Propositional and Event modality
1.3 The classification of modality in modal systems
1.3.1 Epistemic and Evidential
1.3.2 Deontic and Dynamic
1.4 Other types of modality
1.4.1 Presupposed
1.4.2 Negative and Interrogative
1.4.3 Wishes, fears, etc.
1.4.4 Past tense as modal '
1.4.5 Complex systems
1.5 Typology
1.6 Grammatical markers
1.7 Terminology
1.8 Summary of basic categories
1.9 Organization

2 Modal systems: Propositional modality

2.1 Epistemic modality
2.1.1 The three types v I,
2.1.2 Speculative and deductive
Contents
2.1.3 Deductive and Assumptive
2.1.4 MAY as concessive
2.1.5 Modifications
2.1.6 Past time reference
2.1.7 Inference and confidence
2.2 Evidential modality
2.2.1 Evidential systems
2.2.2 Reported
2.2.3 Sensory
2.2.4 Direct and indirect evidence
2.2.5 Other evidential possibilities
2.2.6 Hierarchy
2.2.7 Evidentials and certainty
2.3 Interrogative and Negative
2.4 Two other possibilities
2.5 Structure of propositional systems
2.6 Discourse and participants
2.6.1 Discourse systems
2.6.2 Participant systems
2.7 Declaratives
2.7.1 Declaratives and modal systems
2.7.2 Assertion and 'strong assertion'

3 Modal systems: Event modality

3.1 Formal systems
3.2 Deontic
3.2.1 Directives
3.2.2 Commissive
3.2.3 Modifications
3.2.4 Subjectivity
3.2.5 Past time reference
3.3 Dynamic
3.3.1 Ability and willingness
3.3.2 Past tense and time
3.4 Imperative and jussive
3.5 'Purposive' in Australian languages
3.6 An alternative analysis

4 Modal systems and modal verbs

4.1 Modal systems
x
 Contents

4.1.1 Formal identity of different systems 86

;
4.1.2 Possibility and necessity * 89
4.1.3 Possibility, necessity and negation in English 90
4.1.4 Possibility, necessity and negation in other languages 92
4.1.5 The notional connection 98
4.2 Modal verbs 100
4.2.1 Modal verbs in English '' 100
4.2.2 Modal verbs in other languages ' 101
4.2.3 Formal differences .' ' 103
4.3 Modal systems and other categories 104
4.3.1 Mood -V. 104
4.3.2 Future ' . 104
4.3.3 Negation 106

Indicative and subjunctive 107

5.1 Main and subordinate clauses ' 108
5.1.1 Main clauses ' 108
5.1.2 Subordinate clauses 111
5.2 Propositional modality 112
5.2.1 Speculative 112
5.2.2 Reported 113
5.2.3 Negative 116
5.2.4 Interrogative 120
5.2.5 Presupposed 121
5.2.6 Future 124
5.2.7 Conditional 124
5.2.8 Other types 125
5.3 Event modality 126
5.3.1 Directives 127
5.3.2 Purposive 128
5.3.3 Wishes, fears, etc. 131
5.3.4 Resultative 135
5.4 Imperative and Jussive 136
5.4.1 Morphology and syntax 137
5.4.2 Subjunctive 138
5.5 Subjunctive as subordinator 142
5.6 Other possibilities 144

Realis and Irrealis 145

6.1 Joint and non-joint marking 145
XI
Contents
6.2 Terminology
6.3 Morphology and syntax
6.3.1 Morphology
6.3.2 Linked clauses
6.3.3 Complement clauses
6.4 Joint systems
6.5 Non-joint systems
6.5.1 Binary systems
6.5.2 Realis, irrealis and unmarked
6.5.3 Wider systems
6.5.4 Complex systems
6.6 Propositional modality
6.6.1 Past, present and future
6.6.2 Speculative and Deductive
6.6.3 Interrogative
6.6.4 Negative
6.6.5 Negative-like categories
6.6.6 Reported
6.6.7 Presupposed
6.6.8 Conditional
6.6.9 Habitual Past
6.7 Event modality
6.7.1 Imperative and jussive
6.7.2 'Strong' and 'polite' commands
6.7.3 Obligative
6.7.4 Abilitive

7 Subjunctive and irrealis

7.1 Similarities and differences
7.2 A binary system
7.3 The typological status of Realis/Irrealis
7.4 Alternative subordinate constructions
7.4.1 Non-finite clauses
7.4.2 Direct speech
7.5 Speaker commitment
7.6 The subjunctive in English

8 Past tense as modal

8.1 'Real'and'unreal'
8.2 Modal-tense
Xll
 Contents
8.2.1 Modal verbs ' 203
8.2.2 Greek optative 204
8.3 Real and unreal conditions 207
8.3.1 Marking by tense 207
8.3.2 Marking by tense and mood 209
8.3.3 Interaction of modality and unreality 211
8.3.4 Further possibilities 214
8.4 Wishes 216
8.5 Russian 'subjunctive' 217
8.6 Explanation 218

References 222
Language index 231
General index 234
 PREFACE

Since the publication of the first edition (Palmer 1986) there has been
considerable interest in modality (as well as in grammatical typology in
general). A symposium on Mood and Modality held in the University of
New Mexico in 1992 was successful in bringing together over forty re-
searchers, and resulted in the publication of eighteen papers (Bybee and
Fleischmann 1995a). Yet, in contrast, a workshop on modality at the
International Congress of Linguists only ten years before had attracted only
four scholars. This symposium was followed in 1993 by a symposium on
Modality in Germanic languages, which resulted in a further eight papers
(Swan and Westvik 1997). Indeed, Bybee and Fleischmann (1995b: 1) suggest
that the first symposium succeeded in establishing for modality the kind of
status that had been established for tense and aspect by a symposium on
those subjects ten years before (Hopper 1982). It should, however, should be
pointed out that the first edition of the present volume was published six
years before the symposium in New Mexico and that a volume on modality
in English appeared thirteen years before that (Palmer 1979).
It was recognized in the first volume that the most appropriate name for
the relevant category is simply 'modality', and that 'mood' is more appro-
priate as the traditional name for indicative, subjunctive, etc., in both
classical and modern European languages. Nevertheless, it was thought that,
since the topic was fairly unfamiliar, Mood and modality might be the most
informative title, and that was the title used in the 1993 symposium.
I have been assisted by a whole host of scholars who have provided
examples or answered queries about their work. I fear that a list of names
would inevitably contain omissions, and hope they will accept my sincere
thanks and forgive me for not mentioning them individually.

xv
 NOTE ON THE TEXT

In addition to the usual conventions for the use of italics, quotation marks
and asterisks, the following notation is used:

In the text:
Initial capitals major typological categories
SMALL CAPITALS verbs, including modals
'single quotation marks' terms by the authors quoted

In the inset examples

SMALL CAPITALS grammatical categories and language forms where
the grammatical category is not established
- hyphen morphemic boundary in the language material
and corresponding division in the gloss
+ plus sign combined categories in the gloss represented by a
single element in the language material

To a large extent, the transcriptions, glosses and translations used in the

glosses are those of the original authors, except that
(i) no distinction is made between affixes and clitics;
(ii) where both a phonemic or phonetic and a morphophonemic
representation (showing morpheme boundaries) are given by the
author, either the morphophonemic version will be given or the
morphophonemic form will be given and followed by the
phonemic representation;
(iii) the abbreviations used for the names of the categories are
standardized for consistency - see list on next page.

xvi
 ABBREVIATIONS

1 first person
2 second person
3 third person ^
ABS absolute
ABL ablative
ABSOL absolutive
ACC accusative
ADM admirative
AFF affirmative
AG agent
AOR aorist
APP apparent
ASS assertion/assertive
ASSUM assumed/assumptive
AUD auditory
AUX auxiliary
BEN benefactive
CARIT caritative
CATEG.ASS categorical assertion
CAUS causative
CERT certain
CIRC circumstantive
CL classifier
CM conjugation marker
CNTF counterfactual
CNTR.ASS counterassertion
CO.AG co-agency
COMPL complementizer
COMPLET completive
COMPUL compulsional
XVll
List of abbreviations

CONCESS concessive
COND conditional
CONJ conjecture .;
COOP cooperative
COP copulative
COREF coreferential
CUST customary
DAT dative
DEB debitive
DEC declarative
DED deduced
DEF definite
DEFOC defocus
DEM demonstrative
DES desiderative
DIFF different event
DIR direct case
DS different subject
DUB dubitative
DUR durative
EF.INF. emphatic first-hand informatio
EMPH emphatic
ERG ergative
EV evidential
EXCL exclamative
EXPECT expectational
FACT factual
F.INF first-hand information
FOC focus
FREQ frequentative
F.SG feminine singular
FUT future
FUT.INT future intention
GEN generic
GEN. COND generic conditional
GEN. KNOW general knowledge
HAB habitual
HAB.P habitual past
HORT hortative
HR.EV heard evidence
XVlll
 List of abbreviations

HSY hearsay
H YPOTH hypothetica;
IGNOR ignorative
IMM.PAST immediate past
IMP imperative
IMPERS impersonal
IMPF imperfect
1MPFV imperfective
INC incompletive
INCH inchoatived
INCL inclusive
IND indicative
INDEF indefinite
INDIR indirect
INF/INFER inference/inferential
INFIN infinitive
INFREQ infrequentative
INT interrogative
INTEN intensifier
INTR intransitive
IRR irrealis
IRRELEV irrelevance
JUSS jussive
LING. E VID linguistic evidence
LOC locative
M.E. multiple event
M.SG masculine singular
MOD modal
NEG negative
NOM nominative
NONFEM non-feminine
NONFUT non-future
NONVIS non-visual
OBJ object
OBL obligative
OPT optative
PART participle
PAT patient
PER period
PERF perfect
List of abbreviations
PERFV perfective
PERM permission/permissive
PERS.AG personal agency
PERS.EXP personal experience
PL plural
PL.EXC exclusive plural -,
PLUP pluperfect
POSS possessive
POT potential
PRED predicate
PREDICT prediction
PREP preposition
PRES present
PROG progressive
PROH prohibitive
PUNCT punctual
PURP purpose, purposive
QUES, Q question
QUOT quotative
REAL realis
RECIP reciprocal
RED reduplication
REFL reflexive
RELEV relevance
REL.FUT relative future
REM remote
REP report, reported, reportive
REP. DEP report in dependent clause
REP.IND report in independent clause
RESP responsive
RESP.SCEP sceptical response
SAME same event
SCEP sceptical
SEC second-hand
SEE.EV seen evidence
SEMEL semelfactive
SENS.EVID sensory evidence
SEQ sequential
SG singular
SIM simultaneous
XX
 List of abbreviations

SIMULAT simulative
SM subject marker
SPEC speculated
ss same subject
STAT stative
SUB subject
SUBJ subjunctive
TOD.PAST today's past
TOP topic
UNCERT uncertain
VIS visual
YEST.PART yesterday's past

XXI
 I
Introduction

It has come to be recognized in recent years that modality is a valid cross-

language grammatical category that can be the subject of a typological study.
It is a category that is closely associated with tense and aspect in that all three
categories are categories of the clause and are generally, but not always,
marked within the verbal complex.
In notional terms all three are, in some way, concerned with the event or
situation that is reported by the utterance (though for simplicity the term
'event' will, throughout this volume, be used to cover events, actions,
situations, states, etc.). Tense, rather obviously, is concerned with the time of
the event, while aspect is concerned with the nature of the event, particularly
in terms of its 'internal temporal constituency' (Comrie 1976: 3). Modality is
concerned with the status of the proposition that describes the event.

1.1 Basic concepts

1.1.1 Realis and irrealis

Modality differs from tense and aspect in that it does not refer directly to any
characteristic of the event, but simply to the status of the proposition. One
possible approach to its analysis is to make a binary distinction between
'non-modal' and 'modal' or 'declarative' and 'non-declarative', and to
associate this distinction with the notional contrast of 'factual' and 'non-
factual, or 'real' and 'unreal'.
However, these terms are not really satisfactory (see 1.1.2), and in recent
years 'realis' and 'irrealis' have been used for this distinction. They have the
advantage that they are obviously technical, so that their use can avoid any
possible connotations of the more familiar terms. For Mithun (1999: 173) the
distinction is that 'The realis portrays situations as actualized, as having
occurred or actually occurring, knowable through direct perception. The
irrealis portrays situations as purely within the realm of thought, knowable
only through imagination.'
1
Introduction

Doubts have been raised about the validity of the realis/irrealis distinction
in a typological study (see 7.3), and it must be allowed that it is not necessary
for the analysis of everything that is subsumed under modality. Nevertheless,
the idea that notional features of realis and irrealis are grammaticalized as
the typological categories of Realis and Irrealis is a useful one and will be
followed in this book. (The terms are used initially as terms for the notional
features that determine the grammatical status of propositions, but are also
used, by extension, to refer to the typological categories 'Realis' and 'Irrealis'
with initial capitals - see 1.7. They are also used, in some languages, for the
related language-specific grammatical categories see 1.7 and 6.2.)
In all typological studies there is considerable variation in the ways in
which languages deal with grammatical categories, and there is probably
more variation with modality than with other categories. Indeed, there is
variation, at a quite fundamental level, in the categories that are treated as
Realis and Irrealis in different languages. One language may mark com-
mands as Irrealis, another may mark them as Realis, while yet another may
not treat them as part of a system of modality at all. There are similar
variations in the treatment of futures, questions, denials, reports, etc.
A few examples will show both what kinds of feature are involved and the
very different ways in which they are treated in different languages. First,
English uses a modal verb to distinguish a judgment about a proposition
from a categorical statement. Compare the first example, with no modal
verb, with the other two:
Mary is at home
Mary may be at home
Mary must be at home
Secondly, Spanish distinguishes what is believed to be true from what is
doubted, by the use of the indicative and subjunctive mood, as in Klein
(1975:356,353):
Creo que aprende
I believe that learn+3sG+PRES+iND
'I believe that he is learning'
Dudo que aprenda
I doubt that Iearn+3SG+PRES+SUBJ
'I doubt that he's learning'
Thirdly, forms that are actually given the grammatical labels of 'realis' and
'irrealis' in the Papuan language Amele (Roberts 1990: 371-5) are used to
distinguish 'remote past', 'today's past', 'habitual past' and 'present' from
'future', 'imperative', 'hortative', 'counterfactual' and 'negative'. Examples of
2
 1.1 Basic concepts

realis 'remote past' and irrealis 'future', with the realis marker -en and the
irrealis marker -eb on the verb in the subordinate clause, are:
ho bu-busal-en age qo-in
pig SIM-run OUt-3SG+DS+REAL 3PL hit-3PL+REM.PAST
'They killed the pig as it ran out'
ho bu-busal-eb age qo-qag-an
p i g SIM-run OUt-3SG+DS+IRR 3PL hit-3PL-FUT
'They will kill the pig as it runs out'

1.1.2 The notion of assertion

The contrast between 'factual' and 'non-factual' or 'real' and 'unreal' is not
clear enough to explain fully the distinction between Realis and Irrealis.
Explanations in terms of 'true' and 'untrue' are even less satisfactory, and, in
any case, it may be wise to avoid these terms in view of their logical
connotations (see Bybee, Perkins, and Pagliuca 1994: 239).
There is, however, a body of literature that has proved very helpful. It
has been argued that the use of the 'indicative' and the 'subjunctive', which
are the traditional terms used in many European languages for the
distinction between Realis and Irrealis, can be accounted for in terms of
'assertion' and 'non-assertion' (Bolinger 1968; Terrell and Hooper 1974;
Hooper 1975; Klein 1975). The most succinct account is that of Lunn (1995:
430), who directly links the choice of the indicative to assertion and the
choice of the subjunctive to non-assertion. Using examples of subordinate
clauses in Spanish, she suggests that a proposition may be unworthy of
assertion for three reasons:

(i) the speaker has doubts about its veracity:

Dudo que sea buena idea
I doubt that be+3so+PRES+suBJ good idea
T doubt that's a good idea'
(ii) the proposition is unrealized:
Necesito que me devuelvas ese libro
I need that me return+2so+PRES+supj that book
'I need you to return that book to me'
(iii) the proposition is presupposed:
Me alegra que sepas la verdad
me it pleases that know+2so+PRES+suBj the truth
'I'm glad that you know the truth'
The important thing about this analysis is that it shows quite clearly that
the choice of the Irrealis marker, the subjunctive, does not depend on the
distinction between what is factual and what is not (and still less on what is

3
Introduction
true and what is not true). It depends on the distinction between what is
asserted and what is not asserted. This is particularly clear from the third
example, which relates to what is presupposed, for here the proposition ('that
you know the truth') is clearly factual. It is known to, presupposed by, both
the speaker and the addressee. Indeed, Givn (1994: 304) comments that it
would appear to be 'super realis' since it is known to be true. It is not,
therefore, factuality, certainty or truth that is at issue here. What is at issue is
that nothing is being asserted, that there is no information value, because
both speaker and hearer accept the proposition. It is for that reason that the
proposition is treated as Irrealis, for propositions that are presupposed are
not asserted.
There are other types of non-asserted proposition. These are discussed in
1.4.

1.2 Two basic distinctions

1.2.1 Mood and modal systems

Basically, there are two ways in which languages deal grammatically with an
overall category of modality. These are to be distinguished in terms of (i)
modal system and (ii) mood. Both may occur within a single language, e.g.,
in German, which has a modal system of modal verbs and mood (indicative
and subjunctive), and in Central Porno (see below, this section, and 6.3.2). In
most languages, however, only one of these devices seems to occur or, at
least, one is much more salient than the other. In some of the European
languages the subjunctive appears to be falling into disuse, while in English it
has virtually disappeared and, at the same time, a modal system of modal
verbs has been created (see 4.3.1).
Typically with mood, all or most clauses are either Realis or Irrealis: the
system is basically ('prototypically') binary. An obvious example is the
contrast made in European languages between the indicative and the
subjunctive, the indicative marking clauses as Realis and the subjunctive
marking them as Irrealis. A similar distinction (though with rather a different
set of notional-grammatical functions) is one that is mostly found in the
description of some Native American languages and some languages of
Papua New Guinea. In the study of these languages the distinction has been
made in terms of (the grammatical categories) realis and irrealis, rather than
in terms of indicative and subjunctive.
Examples of the indicative/subjunctive contrast from Spanish and of the
realis/irrealis contrast from Amele were given in 1.1.1, but are repeated
here:
4
 1.2 Two basic distinctions

Creo que aprende

I believe that learn+3so+PRES+iND
'I believe that he is learning'
Dudo que aprenda
I doubt that learn+3sG+PRES+suBj
'1 doubt that he's learning'
ho bu-busal-en age qo-in ' . :.
pig siM-run out-3sc+DS+real 3PL hit-3PL+REM.PAST
'They killed the pig as it ran out'
ho bu-busal-eb age qo-qag-an
pig SIM-run OUt-3SG+DS+IRR 3PL hit-3PL-FUT
'They will kill the pig as it runs out'
Although the contrast of Realis/irrealis can be said to be basically or
prototypically binary, there are exceptions, especially that (i) Imperative and
Jussive are outside the indicative/subjunctive system of mood (see 5.4), and
(ii) where there are realis and irrealis markers, some clauses may be
unmarked for this distinction (see 6.5.2). However, even where there are
these exceptions, there is usually still a pair of forms that essentially mark a
binary contrast of notional realis and irrealis (see 7.2 for discussion).
Strictly, there is no typological difference between indicative/subjunctive
and realis/irrealis. Both make express the distinction between the notional
features of realis and irrealis and can be seen as instances of the typological
categories of Realis and Irrealis. However, there are some differences in their
distribution and syntactic functions, some of which are illustrated by the
examples given above:

(i) The subjunctive is found mainly in subordinate clauses (5.1, 5.5).

(ii) Irrealis and realis often co-occur with other grammatical
markers (6.1).
(iii) Both are often notionally redundant, but in different ways -
irrealis because of its co-occurrence with other markers,
subjunctive because, in subordinate clauses, its occurrence is
determined by the type of complementizer (7.1).
(iv) Unlike the indicative/subjunctive system, realis/irrealis systems
do not usually occur together with tense systems. In general past
and present are marked as realis, future as irealis.

For these reasons they will be dealt with in separate chapters, indicative/
subjunctive in Chapter 5 and realis/irrealis in Chapter 6. It must be
emphasized, however, that this is to some degree a decision based more on
Practical considerations than on a clear set of criteria for the distinction (see
Introduction

7.1). (For the problem of terminology and the decision to use the terms
'realis' and 'irrealis' for the notional features, the grammatical markers and
the typological categories, see 1.7.)
Within modal systems different kinds of modality are distinguished within
a single system of commuting terms. A clear example of such a system is that
of the English modal verbs when they are used to make judgments about the
factual status of a proposition as in:
Kate may be at home now
Kate must be at home now
Kate will be at home now
These can be interpreted and labelled as:
a possible conclusion speculative
the only possible conclusion deductive
a reasonable conclusion assumptive
A rather different system is the evidential system of Central Pomo (Mithun
1999: 191), where, as well as a (rarely used) unmarked form, there are forms
with markers for general knowledge, first-hand personal experience (usually
visual), auditory evidence, hearsay and inference:
hemul
rain fell
'It rained'
hemul-?ma
rain fell-GEN.KNOW
'It rained' (that's an established fact)
hemul-ya
rain fell-vis
'It rained' (I saw it)
hemul-nme'
rain fell-AUD
'It rained' (I heard it)
hemul-?do
rain fell-HSY
'It rained' (I was told)
hemul-?ka
rain fell-iNF
'It rained' (everything is wet)
(There are also two markers for personal experience of actions performed by
the speaker and for actions that affected the speaker. These are discussed in
2.2.5.) For a similar system in Tuyuca see 2.2.1.
The distinction between Realis and Irrealis is an integral part of systems of
6
 1.2 Two basic distinctions

mood. It can generally usually be applied to modal systems as well, in that

notionally the terms in a modal system are irrealis, and there is usually an
unmarked realis form. In English, for example, there are the three types of
(irrealis) modality that are marked by the modal verbs, but there is also a
realis form, the 'declarative' that contains no modal verb:
Kate is at home now
Similarly, in Central Pomo it would seem that there is an unmarked realis
form and five irrealis forms indicating different kinds of evidence for the
proposition.
However there are two types of situation in which it is less easy to apply a
simple Realis/Irrealis distinction. First, there are some languages with a
modal system but no unmarked (realis) declarative, so that it would seem
that either the system contains realis and irrealis forms, or there is no realis
form in the language - see 2.7.1. Secondly, although Central Pomo has a
modal system in which it is reasonable to see the Realis/Irrealis distinction, it
also has a completely independent system of mood with realis and irrealis
markers (see 6.3.2).
There is one final point. It is not always possible to draw a clear distinction
between mood and modal system, since, in some languages, the overall
system of modality has characteristics of both - see 6.5.3 and 6.5.4.

1.2.2 Propositional and Event modality

There is a clear contrast in the notional features involved in the following
pairs of examples (the first two already quoted in 1.2.1 and the second two
indicating permission and obligation):
Kate may be at home now
Kate must be at home now
Kate may come in now
Kate must come in now
The distinction is usually made in terms of 'epistemic' and 'deontic' modality,
and illustrated by the use of paraphrases using 'possible' and 'necessary':
It is possible (possibly the case) that Kate is at home now
It is necessarily the case that Kate is at home now
It is possible for Kate to come in now
It is necessary for Kate to come in now
The important distinction between the two pairs is indicated by the words
that' and 'for'. This clearly suggests that the first pair are concerned with the
speaker's judgment of the proposition that Kate is at home, whereas the

7
Introduction
second are concerned with the speaker's attitude towards a potential future
event, that of Kate coming in. For that reason a basic distinction may be
made between 'prepositional modality' and 'event modality'. (This contrast
is essentially the same as that of Jespersen (1924: 329-31), between 'con-
taining an element of will' and 'containing no element of will'.)
This distinction does not apply only to epistemic and deontic modality, as
illustrated here (where it is largely associated with the notion of possibility
and necessity - see 4.1). Other types of modality are to be included, especially
dynamic modality (Palmer 1979: 36-7; 1986: 102-3) within event modality
and evidential modality within prepositional modality. (Chung and Timber-
lake (1985: 244) use the term 'epistemological' for 'evidential', but 'epistemo-
logical' is too similar to 'epistemic' to be a useful term.) The terms
'epistemic', 'deontic' and 'dynamic' are taken from Von Wright (1951: 1-2).
In summary, epistemic modality and evidential modality are concerned
with the speaker's attitude to the truth-value or factual status of the
proposition (Prepositional modality). By contrast, deontic and dynamic
modality refer to events that are not actualized, events that have not taken
place but are merely potential (Event modality).

1.3 The classification of modality in modal systems

1.3.1 Epistemic and Evidential

Epistemic and Evidential systems are the two main types of Propositional
modality and were exemplified in 1.2.1 from English and Central Pomo. The
essential difference between these two types is (as is implicit in the discussion)
that with epistemic modality speakers express their judgments about the
factual status of the proposition, whereas with evidential modality they
indicate the evidence they have for its factual status. The English examples
were of three kinds of (epistemic) judgment - speculative, deductive and
assumptive. They may be regarded as examples of the typological categories
Speculative, Deductive and Assumptive. The Central Pomo examples in-
cluded five types of evidence: 'general knowledge', 'first-hand personal
experience (usually visual)', 'auditory evidence', 'hearsay' and 'inference'.
The first four of these exemplify the evidential categories of Reported(gen),
Visual, Auditory and Reported(l), while the last ('inference') is the epistemic
category Deductive.
One immediate observation is that the typological category Deductive
appears in both systems. There is an obvious reason for this - deduction, as
expressed by English MUST usually suggests that the judgment was based on
evidence. Coates (1983: 41) comments, 'In its most normal usage, Epistemic
8
 1.3 The classification of modality in modal systems

MUST conveys the speaker's confidence in the truth of what he is saying, based
a deduction from facts known to him (which may or may not be
necified).' Thus Kate must be at home may be a judgment based on the
observation that she is not in her office. In a similar way, the 'inference' form
in Central Pomo indicates that there is indirect, usually visual, evidence for
the proposition. Notionally these are very close and can both be treated as
Deductive. Because both a deduction and evidence are involved, it is reason-
able to treat the English form in a judgment system and the Central Pomo
form in an evidential system. However, purely for the purpose of discussion
and exemplification, Deductive will be discussed in the section on Epistemic
(2.1).
A second observation is that evidentials may occur within what is
primarily an epistemic (judgment) system. The modal verbs in German, for
instance, which are similar to those of English, express the judgments
Deductive and Speculative, but SOLLEN and WOLLEN are used as evidentials
(Hammer 1983: 231, 232):
Er soil steinreich sein
He SOLLEN+ 3SG+PRES+IND very rich be
'He is said to be extremely rich'
Er will eine Mosquito abgeschossen haben
he WOLLEN+3SG+PRES+IND a Mosquito shot down have
'He claims to have shot down a Mosquito' (plane)
The first indicates what is reported, the second what is claimed by the person
designated by the subject of the sentence. In both cases the speaker merely
provides the evidence for the proposition.
Within Evidential modality there are two main types, which may be
identified as Sensory and Reported, the former subsuming Visual, non-Visual
and Auditory (at least) and the latter, three sub-types of report. It could be
argued that these form distinct types of propositional modality, so that there
are three types - Epistemic, Reported and Sensory. In practice, however, the
binary distinction between Epistemic and Evidential modality will be main-
tained in this book.

1.3.2 Deontic and Dynamic

Deontic and Dynamic are the two main types of Event modality. In the
simplest terms the difference between them is that with deontic modality the
conditioning factors are external to the relevant individual, whereas with
dynamic modality they are internal. Thus deontic modality relates to
obligation or permission, emanating from an external source, whereas

9
Introduction

dynamic modality relates to ability or willingness, which comes from the

individual concerned. The distinction can be seen in:
John may/can come in now (permission)
John must come in now (obligation)
John can speak French (ability)
John will do it for you (willingness)
The first two exemplify the typological categories of (Deontic) Permissive
and Obligative. The second pair exemplify the categories of (Dynamic)
Abilitive and Volitive. This is, however, an oversimplification in three
respects.
First, although Deontic modality stems from some kind of external
authority such as rules or the law, typically and frequently the authority is
the actual speaker, who gives permission to, or lays an obligation on, the
addressee. It is for this reason that Searle (1983: 166) speaks of 'directives',
which he defines as 'Where we try to get [others] to do things'.
Secondly, Ability, one of the categories of Dynamic modality, has to be
interpreted rather more widely than in terms of the subjects' physical and
mental powers, to include circumstances that immediately affect them (but
not, of course, deontic permission). This can be seen in the contrast between
deontic and dynamic CAN, where deontic CAN indicates permission, while
dynamic CAN may indicate not merely ability, but also the possibility in a
more general sense, as shown by:

He can go now (Deontic: I give permission)

He can run a mile in five minutes (Dynamic: he has the ability)
He can escape (Dynamic: the door's not locked)
Thirdly, there is another category, that of Commissive, which is defined by
Searle as 'where we commit ourselves to do things', and exemplified by
English SHALL:

You shall have it tomorrow

Since the conditioning factor for the action lies with the speaker, not the
subject, this is best seen as another type of Deontic modality.
It must be admitted, however, that Commissive and the two forms of
Dynamic modality are much more clearly marked in English than in many
other languages.

1.4 Other types of modality

The sub-types of modality discussed in the last section are mainly relevant to
the analysis of modal systems. There are, however other grammatical
10
 1.4 Other types of modality

ories t hat a r e associated with modality, especially (but not solely) with
respect to mood.

1.4.1 Presupposed
The treatment of presupposed propositions as irrealis was discussed in 1.1.2,
with the example:
Me alegra que sepas la verdad
me it pleases that know+2sG+PRES+suBj the truth
'I'm glad that you know the truth'
The use of the subjunctive in subordinate clauses, as in this example, is,
perhaps, the clearest indication of its use for presupposition, since it is clear
that nothing is being asserted by the proposition in the subordinate clause. Yet
examples of presupposed propositions being marked as irrealis can be found
in main clauses too, e.g., the use of the subjunctive in Italian in a concessive
clause, where the speaker admits to the truth of the proposition (see 5.2.5):
sia pure come dici tu ma io non vengo
be+3sG+PRES+suBj perhaps as you say but I not come
'It may be as you say, but I'm not coming'
It is also relevant that, as the translation shows, English uses a modal verb
(may) here. More striking is the marking of the category 'admirative' as
irrealis in Caddo (Oklahoma - Chafe 1995: 357):
hus-ba?a-sayi-k'awih-sa?
ADM-1 +BEN+iRR-name-know-PROG
'My goodness he knows my name'
This Chafe finds 'surprising', yet it is clear that the speaker is not asserting
the proposition, but is indicating his surprise at what is obvious to the
addressee (see 6.6.7).
'Presupposed' is very similar to what Kiparsky and Kiparsky (1971) call
factive', with the argument that REGRET, RESENT and many other verbs and
constructions are 'factive predicates.' However, the term can be misleading,
as can be seen in the remark by Coates (1983: 235) that 'KNOW is the classic
example of a factive predicator'. For although know clearly suggests that the
speaker accepts the information as factual, presupposition is not involved - it
does not suggest that the addressee equally accepts the factuality. The
complements of KNOW are not Factive in the presuppositional sense.

1.4.2 Negative and Interrogative

Negatives and Interrogatives are classed together as being 'non-assertive' by
Quirk, Greenbaum, Leech and Svartvik (1985: 83) for grammatical reasons
11
 1

Introduction

in English - their association with certain 'non-assertive' forms such as any

and yet. It is, therefore, not surprising that they sometimes occur within
modal systems or are marked as irrealis where there is marking of mood.
A striking example is to be found in what is clearly an epistemic system in
Imbabura, a variety of Quechua (Cole 1982: 164), where the same marker is
used for both question and negation. Cole lists the markers as:

ma(ri) emphatic first-hand information

mi first-hand information
shi conjecture ,,
cha(ri) doubt
chu Yes-No question
chu negation

The first two are probably to be seen as indicating evidence from the senses,
while the second two are markers of judgments. The third pair shows that the
same marker is used for Interrogative and Negative. Examples are:
kan-paj ushi-wan Agatu-pi-mi
you-of daughter-with Ageto-in-F.iNF
'I met your daughter in Ageto'
kaya-shi kan-paj churi shamu-nga
tomorrow-coNj you-of son come-3+FUT
'I suppose your son will come tomorrow'
Juzi-ka kitu-man chaya-shka-ch
Jos-TOP Quito-to arrive-PERF-DUB
'Perhaps Jos has arrived in Quito'
mayistru-chu ka-ngui
teacher-QUES be-2sg
'Are you a teacher?'
fluka-ka mana chay llama-ta shuwa-shka-ni-chu
I-TOP not that sheep-Ace steal-PERF-2sG-NEG
'I didn't steal that sheep'
Similarly, but in a system of mood (realis and irrealis), Interrogative and
Negative are both marked as irrealis in Caddo (Chafe 1995: 334-5; see 6.6.3,
6.6.4):
kuy-t'ayi-bahw
neg-1+AG+iRR-see
'I don't see him'
sah?yi-bahw-nah
2+AG-+IRR-See-PERF
'Have you seen him?'
12
 1.4 Other types of modality

It may be thought that interrogatives and negatives are never marked as

Trrealis in European languages. It is true that there is no indication of them
being treated as Irrealis in main clauses, but there is evidence of such
treatment in subordinate clauses. Thus reported (indirect) questions are
expressed with the subjunctive in Latin (see 5.2.4), while a negated verb of
saying in Spanish and Italian is also usually followed by the subjunctive (see
5.2.3).

1.43 Wishes, fears, etc.

The status within modality of wishes and fears is a little more obscure,
although it is clear that notionally they express attitudes towards proposi
tions whose factual status is not known or propositions that relate to
unrealized events. They are thus partly deontic, partly epistemic. There are
plenty of examples of wishes and fears being expressed by the subjunctive in
subordinate clauses in Latin and other languages (see 5.3.3), but the
subjunctive is also used in main clauses, as in Latin and Classical Greek:
modo valeres (Cic. Att. 9.22)
only be well+2sG+iMPERF+suBj
'If only you were well'
m: sous diaphthire:i gmous (Eur. Ale 315)
not your ruin+3sG+PRES+suBJ marriage
'I'm afraid she may ruin your marriage'
Similarly in the West African language Fula (Arnott 1970: 299ff.), wish is
one of the features signalled by what is identified as subjunctive:
njuutaa bale
be.long+2sG+suBJ in days
'May you live long!'
Expressions of wanting, desire and preference relate similarly to unrealized
events and may be marked as irrealis, but generally only in subordinate
clauses, as in Spanish:
quiero que estudias ms
I want that study+3sG+PRES+suBJ more
'I want you to study more'

1.4.4 Past tense as modal

There is one further feature that affects modal systems - the use of past tense
forms to express 'unreality', 'tentativeness', 'potentiality', etc. This often
'unctions independently of mood, but may interact with it. This is discussed
ln
Chapter 8, but two illustrations may be given.

13
 1

Introduction

First, three of the English modal verbs have what are formally present and
past forms:
WILL will would
CAN can could
MAY may might
All three past tense forms can be used with past time reference, but past time
reference is not the most common function of the past tense forms. Rather
they are used to express greater tentativeness, as in:
He'll be there now
He'd be there now
He may be there now
He might be there now
He can't be there now
He couldn't be there now
Secondly, in a large number of languages, a distinction is made between
'real' and 'unreal' conditionals, as can be illustrated by English:
If the children are here, John will be happy
If the children were here, John would be happy
If Mary comes, John will stay
If Mary came, John would stay
The second example of each pair is unreal. In the first of these second
examples (which refers to a present situation, where the facts are known),
there is a clear implication that the children are not here so that John will not
be happy. In the second, which refers to the future, where the facts are not
knowable, the implication is that it is unlikely that Mary will come. This
distinction, as can be seen from the examples, is marked by tense, with
present tense forms in both clauses of the real conditionals and past tense
forms in those of the unreal ones.
This use of the past tense will be referred to as the 'modal-past'.

1.4.5 Complex systems

Ideally, there would be discrete systems of modal markers, such that each of
the typologically relevant modal systems was in a one-to-one correspondence
with a formal system in each language. In practice, the situation, as might be
expected in a typological study, is far less simple and neat. There are three
ways in which the systems are far from being orderly and simple.
First, two apparently different systems may use the same (or most of the
same) set of markers. This is so for the epistemic and deontic/dynamic
systems of English, which use the same modal verbs, especially MAY, CAN and
14
 1.4 Other types of modality

MUST, as was illustrated in 1.2.1. (Indeed many languages have the same
forms for epistemic and deontic modality - see 4.2.1.) CAN, for instance, may
be used as epistemic, deontic or dynamic:
He can't be in his office now (epistemic)
He can come in now (deontic, permission)
He can run a mile in four minutes (dynamic, ability)
However, there are some formal distinctions between the uses, e.g., that for
positive epistemic modality MAY is used rather than CAN, that MAY is an
alternative to CAN only for deontic, and that the past tense form could is used
to refer to past time only when it is dynamic. These features are discussed in
much more detail in 4.2.3.
Secondly, in a few cases, notionally identical modal categories may be
treated as members of one system in one language, and of another system in
another language. This is so with Deductive which, because it involves a
judgment based upon evidence, may be a term in a judgment system or in
an evidential system, as in English and Central Porno respectively (see
1.3.1).
Thirdly, a single formal system often contains forms that belong to two or
more modal systems. Thus in the German system of modal verbs not only do
most of the modal verbs function as both epistemic and deontic, but also
there are forms (SOLLEN and WOLLEN) that are clearly evidential (see 1.3.1).
The system in Imbabura (Cole 1982: 164, described in 1.4.2) has an even
more varied set of terms.
Fourthly, there may not be a simple one-to-one relation between the terms
of a formal system and some of the typlogically relevant notional features.
Thus in Lega (Bantu, Eastern Zaire - Botne 1997: 511-22, and personal
communication), there are three particles, whose primary functions seem to
be those of Reported, Speculative and Sensory. These are familiar categories,
but there are additional features associated with two of them. The term for
Reported often expresses doubt or disbelief:

nkumgwg (bono), mbo bazongo b kulyg merende

I hear that whites EV eat frog
'I hear that Westerners eat frogs (though I find that unlikely)'
The term used for Sensory (Visual and Auditory) may also express both
confidence of the kind associated with Deductive and reports that can be
trusted:
amp mbul zsabka
EV rain rain
'It is already raining' (I can see it)

15
Introduction

amp kukr momponge

EV she.is.pounding rice
'She's assuredly pounding rice' (I can hear it)
amp Kisang ndilc kw isoko; kikpu kyg tak gno
EV Kisanga went to market; basket is not here
'Kisanga surely went to the market; her basket is not here'
amp Mok korwa
EV Moke tired
'Moke is tired' (Moke told me)
See 2.2.4 for more details.
Fifthly, the same formal system may contain not only modal categories,
but others such as those of tense, e.g., in Ngiyambaa (Donaldson 1980:
159-62):

(i) Imperative
nindu bawunn-ga yuwa-dha
you+NOM middle-Loc lie-iMP
'You lie in the middle!'
(ii) Past
yurun-gu rjidjiyi
rain-ERG rain+PAST
'It rained'
(iii) Present
yururjgu nidja-ta
rain-ERG rain-PRES
'It is raining'
(iv) Irrealis
yurun-gu nidja-1-aga
rain-ERG rain-CM-lRR
'It might/will rain'
(v) Purposive
nadhu bawurj-ga yuwa-giri
I+NOM middle-Loc lie-PURp
'I must lie in the middle'
yururj-gu nidja-1-i
rain-ERG rain-CM-PURP
'It is bound to rain'
In fact, the situation in Ngiyambaa is even more complex than suggested
above (Donaldson 1980: 251-65). There is a set of clitics that relate to
features that are clearly modal and to others that are clearly not. Donaldson
divides them into three types, 'Belief clitics, 'Knowledge' clitics and 'Evi-
dence' clitics. Examples are:
16
 1.4 Other types of modality

'Belief' clitics:

(i) Assertion
wanaiy-ba:-na yana-nhi
NEG-ASS-3ABS Walk-PAST
'He didn't walk' (again)
(ii) Categorical assertion
guni:m-ba];a-nu: balu-y-aga
m0ther+ABS-CATEG.ASS-2OBL die-CM-IRR
'Your mother is bound to die'
(iii) Counter-assertion
guyan-baga:-dhu gaja
Shy+ABS-CNTR.ASS-INOM be+PRES
'But I'm shy!'
(iv) Hypothesis
gali:-ninda-gila niyanu balu-y-aga
water-CARIT-HYPOTH We+PL+NOM die-CM-IRR
'We'll probably die for lack of water'

'Knowledge' clitics:
(i) Exclamative
minja-wa:-ndu dha-yi
what+ASS-EXCL-2NOM eat-PAST
'What did you eat?' / 'You ate what?'

guya-wa:-ndu dha-yi
fish+ABS-EXCL-2NOM eat-PAST
'So you ate a fish!' / 'What? You ate a fish'
(ii) Ignorative
minjarj-ga:-dhu dha-yi
what+ABS-IGNOR-lNOM eat-PAST
'You ate something, I don't know what' / 'I don't know what you ate'
guya-ga:-ndu dha-yi
fish+ABS-IGNOR-2NOM eat-PAST
'Did you eat a fish' / 'You ate a fish, I don't know'

'Evidence' clitics:

(i) Sensory evidence

rjindu-gara girambiyi
yOU+NOM-SENS.EVID sick+PAST
'One can see you were sick'

gabuga:-gara-lu rjamumiyi
egg+ABS-SENS.EVID-3ERG lay+PAST
'It's laid an egg by the sound of it'

17
Introduction

(ii) Linguistic evidence >;! ;??

rjindu-dhan girambiyi
yOU+NOM-LING.EVID sick+PAST
'You are said to have been sick'

The evidence clitics are clearly markers of evidentials, but most of the others
are, at best, marginally modal.

1.5 Typology
It is now well understood that grammatical typology cannot be undertaken
on the basis of purely formal, grammatical marking. As Croft (1995: 88)
says, 'In general, cross-linguistic identification cannot be accomplished on
purely formal (structural) grounds for two reasons. First, variation across
languages is too great . . . Second, formal definitions are internal to the
structural system of a single language . . . For these reasons, typologists
generally use definitions . . . that are "external" to the linguistic system, that
is semantic, pragmatic or discourse-based definitions'. (The features involved
in the definitions are referred to in this book as 'notional', thereby avoiding
any debate about semantic vs pragmatic, etc.)
It may be useful to consider briefly the ways in which notional and formal
criteria may be related.
First, a single category may be identified across languages, e.g. Speculative
in English, Tamil, Ngiyimbaa and Quechua (2.1.2).
Secondly, a category may be distinguished formally in one language, but
may be only one of several categories indicated by a single marker in
another. Thus, Future is uniquely marked in Amele, but is only one of five
categories indicated by an irrealis marker in Manam (both of them Papuan
languages - see 6.1).
Thirdly, a single marker may indicate a number of different notional
categories that are not formally distinguished in any language, but are
nevertheless of typological interest. Thus Presupposed is indicated by sub-
junctive in Spanish (5.2.5). It is not uniquely distinguished elsewhere, but is
of typological interest in that it accounts for irrealis marking in Caddo
(6.6.7).
Finally, it is sometimes useful to compare a set of forms that is formally
marked with a set whose formal status is doubtful in another language. Thus
the modal verbs in English are a formally defined set (4.2.1), but it is useful to
compare them with similar sets not only in German, where there are some
formal markers, but also in the Romance languages where their formal status
is very much in doubt (see 4.2.2).
18
 1.6 Grammatical markers

1.6 Grammatical markers

Tue grammatical markers of modality are very varied, and it would be
nhelpful to exemplify them in detail here. There is, however, a discussion of
the grammatical status of modal verbs in 4.2 and of the markers of realis and
rrealis in 6.3.1. However, a few comments may be made.
Basically there are three types of marker: (i) individual suffixes, clitics and
particles, (ii) inflection and (iii) modal verb.
Some examples of individual items have already been given, e.g. those of
Imbabura in 1.4.2. They are found as markers both of terms in modal
systems and of mood.
Wholly inflectional forms that involve the form of the lexical verb itself as
well as tense, mood and person-number are characteristic of indicative and
subjunctive. There are, however, simultaneous markers of both mood and
person-number in some languages that are described in terms of realis/
irrealis, and there is at least one such language (Takelma) where the forms
involve the lexical verb also (see 6.3.1).
The use of modal verbs seems to be confined to modal systems.
There are also some syntactic conditions on the occurrence of mood
markers. In 6.3.2 there is one example of a Native American language and
several examples of Papuan languages where realis and irrealis are marked
only in 'linked' clauses.
Finally, in some languages there are large sets of particles or clitics. Thus
Mithun (1986: 99) says that there are over fifty different evidential particles
in Cayuga (Northern Iroquioian). There is no easy way of deciding to what
extent such sets are relevant to a grammatical study of modality though it is
likely that most of the members of the set will not be modal in any of the
senses discussed here. Such large sets are usually best regarded as lexical
rather than grammatical. But some of the items may be of typological
interest.

1.7 Terminology
Some remarks about terminology and notation have already been made, but
it may be useful to consider the issues briefly here.
In a typological study there are three distinct but related sets of concepts
(three levels of discourse) - the typological categories that are seen as
common to various different languages, the grammatical markers associated
w
ith them in individual languages and the notional features that justify the
ypological identification. It might seem wise to use different terminology
an
d different notation for each of the three levels, but this is virtually
lrn
Possible for several reasons, and would probably bring confusion rather
19
Introduction
than clarity; it is actually both normal and helpful to use the same terms at
all levels. For instance, grammar books talk about the 'future tense' and say
that it indicates the (notional) future. Indeed, the names for grammatical
terms are generally derived from the notional features associated with them.
There are, admittedly, alternative names for some of the categories and
Lyons (1977: 30), following Bar-Hillel (1970: 370), suggests that a distinction
should be made between 'declarative', 'interrogative' and 'imperative' on the
one hand and 'statement', 'question' and 'command' on the other. This
advice is not consistently followed here for two reasons. First, there are no
obvious pairs of terms for some of the categories, and, moreover, there is
little consistency among writers. Secondly, provided ambiguity can be
avoided, it is sometimes actually helpful to use the same terminology. It is
more illuminating, for instance, to say that the Latin subjunctive with a
negative is used for a negative imperative (and so compare it with the
morphologically defined imperative) than to say that it is used for a
prohibition.
One convention that will be used is the use of capitals to distinguish the
typological categories, e.g., Realis, Irrealis (1.1.1), Speculative, Deductive,
Assumptive, etc. (1.3.1). However, this convention will not always be used,
for two reasons. First, the same terms are often used for the language-specific
grammatical categories, e.g., imperative and future, which may be instances
of Imperative and Future. In such cases it is simpler to use the form without
initial capitals, unless it is important to stress the typological status. (One
case where the distinction is important is with (grammatical) realis and
irrealis, since indicative and subjunctive are also instances of (typological)
Realis and Irrealis.). Secondly, it seems hardly worthwhile using capitals
throughout for the terms to describe types of modality - event, propositional,
epistemic, deontic and dynamic, which refer to the main typological classes,
and never to language-specific grammatical categories. Initial capitals will be
used only when they are first mentioned.
There is no simple way of distinguishing the notional and grammatical
categories, except by using terms such as 'question' and 'command', where
these are readily available. One possibility might have been the use of
quotation marks for the notional categories, but quotation marks are needed
to indicate those terms used by individual authors that differ from those used
in this book. Instead, where there is a danger of confusion the glosses
'notionally' or 'notional' can be added to indicate notional categories.
There is also a problem with the facts that there is wide variation in the
terminology used by different authors and that their terminology is often
different from that used in this book. An extreme case is the use of the terms
20
 /. 7 Terminology
'report', 'quotative' and 'hearsay', which are used by different authors in
various ways (see 2.2.2). It is for that reason that quotes will be used for all
uch authors' terms, if they are different from those used in this book. The
authors' terms will also be used in the glosses to the examples (where they are
shown in abbreviated small capitals), but the abbreviations used for these
categories will be standardized (a list is given in the preliminary pages).
In summary:

(i) The terms for the grammatical categories will generally be the
same as those for the typological categories, though an exception
is made in the use of 'indicative/subjunctive' in some languages
and 'realis/irrealis' in others to relate to the typological Realis/
Irrealis.
(ii) Typological categories may be identified by the use of initial
capitals. There will be no variation in the terms used for them.
(iii) Where necessary notional categories will be glossed with
'notional(ly)'.
(iv) Terms used by individual authors will be indicated by quotation
marks. These terms will also be shown (in small capitals) in the
glosses to the examples but in a standardized abbreviated form.

Something needs to be said too about the choice of terms used for the
typological categories. The guiding principles here are to avoid confusion
(such as that created by 'report', 'quotative' and 'hearsay'), and to use terms
that are reasonably familiar and not too off-putting to the reader. The issues
that were considered (and the decisions that were made) are as follows:

(i) There are some familiar grammatical terms, such as Imperative

and Jussive, for which there are no sensible alternatives.
(ii) There are other terms with both adjectival and nominal forms
such as Interrogative/Interrogation and Negative/Negation. For
consistency with (i) the adjectival form will be used.
(iii) Most of the terms have the ending -ive but Visual is preferred to
the etymologically correct, but wholly invented, Visive. Two
other adjectival forms are Future and Conditional.
(iv) There is no obvious term for what is presupposed -
'Presuppositive' is too unfamiliar and clumsy. The participial
(-ed form) 'Presupposed' is preferred.
(v) For the confusing 'report','hearsay' and 'quotative' the term
'Reported' (with -ed) is preferred to 'Quotative' (with -ive).
There are, however, three sub-types of Reported and to avoid
21
Introduction

the confusion that would be created by using the other terms,

these will be referred to as 'Reported (2)', 'Reported (3)' and
'Reported (Gen)', to indicate second-hand information, third-
hand information and what is generally said to be true (see
2.2.2).
(vi) No adjectival forms are suggested for wishes and fears or for
habitual past, though 'Desiderative' and possibly Timitive'
might have been used for the first two.

1.8 Summary of basic categories

It may be useful to summarize the main typological categories and their
classification, already mentioned.
Within modal systems these categories can, not surprisingly, be arranged
neatly, but there are other categories that are not easily classified, most of
them found mostly, but not entirely, with Realis/Irrealis mood.
Propositional modality
Epistemic
Speculative
Deductive
Assumptive
Evidential
Reported: Reported (2), Reported (3), Reported (Gen)
Sensory: Visual, non-Visual, Auditory
Event modality
Deontic
Permissive
Obligative
Commissive
Dynamic
Abilitive
Volitive
Other important categories, that may be marked as Irrealis but are mostly
found with mood, are: Future, Negative, Interrogative, Imperative-Jussive,
Presupposed, Conditional, Purposive and Resultative, Wishes ('Desidera-
tive') and Fears ('Timitive'?) and, less commonly, Habitual-Past.

1.9 Organization
This book is basically in two parts.
The first part, Chapters 2, 3 and 4, is concerned with modal systems.
Chapter 2 deals with propositional modality, with main sections on epistemic
and evidential modality, and Chapter 3 with event modality, with main

22
 1.9 Organization
ections on deontic and dynamic modality, although in both chapters there
r e some sections dealing with other, but related, matters. Issues affecting
both types of modality are considered in Chapter 4.
The second part, Chapters 5, 6 and 7, is concerned with mood. Chapter 5
deals with indicative and subjunctive and Chapter 6 with realis and irrealis,
while issues affecting them both are considered in Chapter 7.
Chapter 8 deals with the issue of past tense being used as modal.

23
 1

2
Modal systems: Propositional
modality

As was suggested in 1.2.2, epistemic modality and evidential modality are

concerned with the speaker's attitude to the truth-value or factual status of
the proposition and may thus be described as 'propositional modality'. The
basic difference between epistemic modality and evidential modality is that
with epistemic modality speakers make judgments about the factual status of
the proposition, whereas with evidential modality they indicate what is the
evidence that they have for it.
However, as was noted in 1.3.1, there are two respects in which epistemic
and evidential systems are, in practice, not always wholly distinct. First, the
category Deductive, which involves both a judgment and evidence, may
occur in both types of systems, and this is true also, though less commonly,
of other categories (e.g., Assumptive in Tuyuca - see 2.1.3). It would not be
helpful to deal with these twice, once in a section on epistemic modality and
once in a section on evidential modality and, for that reason, they will be
considered in the section on epistemic modality (2.1). The section on
evidential modality (2.2) will deal with purely evidential features, most
importantly with Reported and Sensory. Secondly, evidential categories
sometimes occur within what are primarily epistemic (judgment) systems
(e.g., the uses of SOLLEN and WOLLEN in German that were exemplified in
1.3.1), but these can still be considered in the section on evidentials (2.2.2).
Later sections deal with what seem to be other types of propositional
modality.

2.1 Epistemic modality

2.1.1 The three types

There are three types of judgment that are common in languages, one that
expresses uncertainty, one that indicates an inference from observable
evidence, and one that indicates inference from what is generally known.
These may be identified typologically as Speculative, Deductive and Assump-
24
 2.1 Epistemic modality
ve respectively. An alternative name for the first is 'Dubitative', but, since
t u e forms do not generally indicate positive doubt (Bybee 1985: 179),
'Speculative' is to be preferred.
There seem to be few languages that have a system with all three markers,
but English is an exception, using the three modal verbs MAY, MUST and WILL:

John may be in his office

John must be in his office
John'll be in his office

The first indicates that the speaker is uncertain whether John is in his office.
With the second, the speaker makes a firm judgment, on the basis of
evidence, e.g. that the office lights are on, that he is not at home, etc. With
the third, the judgment is based on what is generally known about John, e.g.
that he always starts at eight, that he's a workaholic, etc. (a reasonable
conclusion).
MUST seems to draw a firmer conclusion than will, so that the three might
be explained in terms of:
a possible conclusion
the only possible conclusion
a reasonable conclusion *-
There are, it may be argued, two not entirely compatible contrasts in the
English system. The first involves the strength of the conclusion, and
distinguishes between what 'may' be and what 'must' be, i.e., between what is
epistemically possible and what is epistemically necessary (see 4.1.2). This
distinguishes Speculative (MAY) and Deductive (MUST). The second distin
guishes between an inference from observation and an inference from
experience or general knowledge, i.e., between Deductive (MUST) and As
sumptive (WILL). In many languages it seems that one of these contrasts is
either the only contrast in an epistemic system or, at least, the more
important contrast. Not surprisingly, the single contrast between inference
from observation and inference from experience is mostly found in languages
with evidential systems.
The common element in the two contrasting pairs is Deductive, but the
notional features associated with it differ somewhat depending on which
contrast is the more salient. For these reasons the two contrasting pairs will
be considered in separate sections, first dealing with Speculative and Deduc
tive, the second with Deductive and Assumptive.
However, although these three are the most common categories of
Prepositional modality, there are different systems in some languages.
25
Modal systems: Propositional modality

Ladakhi (Tibeto-Burman), for instance, has a marker for 'inference'

(= Deductive) in a basic set of modal markers, but also a second set of
markers that distinguish four types of inference. However, these involve
evidential features and are discussed in 2.2.1.

2.1.2 Speculative and deductive

English then, has the contrast of Speculative and Deductive, as in:
John may be in his office
John must be in his office
There is one further point. These examples illustrate judgments about
propositions relating to the present, but with MAY (and, less commonly,
MUST) the proposition can refer to the future:
John may come tomorrow
They may also be used with HAVE plus the past participle for propositions
relating to the past:
John may/must have been in his office
Many of the European languages have the same contrast, e.g., Danish:
Det kan vsere sandt
that may+3sG+PRES be true
'That may be true'
Det m vaere sandt
that must+3so+PRES be true
'That must be true'
Danish is a Germanic language like English, but the same contrast is found
in the Romance language, Italian:
Pu essere nell ufficio
can/may+3sc+pRES be in the office
'He may be in the office'
Deve essere nell ufficio
must+3sG+PRES be in the office
'He must be in the office'
It is found also in Modern Greek, but through the use of impersonal verbs:
bori na ine sto yrafio tus
can+iMPERS that they are in the office their
'They may be in their office'
prepi na ine sto y ratio tus
must+iMPERS that they are in the office their
'They must be in their office'
26
 2.1 Epistemic modality

This is not a feature of modern European languages alone. In Tamil

Oravidian, S. India and Sri Lanka - Thiagarayan 1981; Asher 1982) there
r e a number of potentially modal forms, but only two are in the form of a
bal su ffix, -laam and -um. These Asher (1982: 16772) calls 'debitive' and
'nermission', though they are used epistemically as Speculative and Deduc-
tive as well as deontically (see 3.2.1). In the example below the term
'concessive' is used to indicate that 'permission' is used in the epistemic
sense:
Kantacaami vantaalum vara-laam
Kandaswami come+coNCESS come-PERM
'Kandaswami may perhaps come'
Gaixeecan ippa Mannaarkutiyile irukka-n.um
Ganesan now Mannargudi+Loc be-DEB
'Ganesan must be in Mannargudi now'
The two types of judgment are often found in complex systems of the type
that were considered in 1.4.5, e.g., in Ngiyambaa (Donaldson 1980: 159-62):
yurun-gu rjidja-1-aga
rain-ERG rain-CM-iRR
'It might/will rain'
yururj-gu rjidja-1-i
rain-ERG rain-CM-PURP
'It is bound to rain'
It was also noted in 1.4.2 that Imbabura has markers for 'conjecture' and
'doubt', but a clearer illustration is to be found in Inga, which, like
Imbabura, is a variety of Quechua (Levinsohn 1975: 14-15, 22). The
distinction is made between 'action deduced' and 'action speculated', the full
system being:

mi action witnessed - affirmative

chu action witnessed - negative
si action reported to the speaker
cha action deduced by the speaker as having occurred
(char- probability reinforced)
sica action speculated as possible by the speaker

Levinsohn calls these markers 'deictic aspect enclitics', because, when not
attached to the main verb, they indicate the 'rheme' - what is normally new
f o r m a t i o n , but this is not relevant to the issue of modality. He gives
samples of them, but with no indication of the modal status in the
translations and without glosses for the enclitics:

27
Modal systems: Propositional modality

nis pufluncuna-mi ..-', (15)

then they.slept-Mi
'Then they slept' (witnessed affirmative) ,.
chica nuca mana yachani-chu (jl)
that I not I.know-CHU
T don't know that myself (witnessed negative)
chacapi-si yallinac (22)
on.bridge-i; they .were.crossing ;.
'They were crossing on the bridge' (reported)
chihoraca mal-cha cado carca fiide del bautismo (23)
at.that.time bad-CHA it was baptismal certificate
'At that time my baptismal certificate must have been incomplete'
(deduced)
chipica diablo-char ca (19)
there devil-CHAR it.was
'A devil was presumably there' (deduced, probability reinforced)
yuyanacurca-sica chica cajta huamra (27)
they.were.thinking-siCA that be child
T suppose they were thinking he was a child' (speculated)
It will be noted that there is a further distinction here for Deductive in terms
of probability - 'deduced' and 'deduced, probability reinforced'. This is
discussed in 2.1.5.
One final point may be made. It is no coincidence that Tamil 'debitive' and
'permissive' forms are used for both epistemic and deontic modality, for, as
noted in 1.2.2, the same is true of English MAY and MUST. This can be
explained in terms of epistemic and deontic possibility and necessity, which is
discussed in 4.1.

2.1.3 Deductive and Assumptive

English has the contrast between Deductive and Assumptive, as shown in
2.1.1:
John must be in his office
John'11 be in his office
The difference between WILL and MUST is illustrated in (see Palmer 1990: 58):
It's nine o'clock - John will be in his office now
Yes, the lights are on, so he must be there
As with MAY and MUST, WILL can be used to refer to the future (but is then
difficult to distinguish from the 'pure' future use - see 4.3.2) and with HAVE to
refer to the past.
28
 2.1 Epistemic modality

The contrast is also found in-a number of other languages, but in most of
them Deductive and Assumptive occur together in systems that include
evidential markers of report and sensation. One such language is Tuyuca
(Colombia - Barnes 1984: 257), where the markers are labelled 'apparent'
a n d 'assumed'. The full paradigm is:

diga ap-wi
soccer play+3so+PAST-vis , i.:
'He played soccer'
(I saw him play)

diga ap-ti
soccer play+3sc+PAST-NONVis ->.
'He played soccer' /
(I heard the game and him, but I didn't see it or him) ,

diga ap-yi
soccer play+3sc+PAST-APP
'He played soccer'
(I have seen evidence that he played: his distinctive shoe print on the
playing fields. But I did not see him play)

diga ap-yigi
soccer play+3sG+PAST-SEC
'He played soccer'
(1 obtained the information from someone else) '
diga ap-hyi
soccer play+3so+PAST-ASSUM
'He played soccer'
(It is reasonable to assume that he did)

These can reasonably be identified as instances of the typological categories

Visual, Auditory, Deductive, Reported and Assumptive respectively.
As has been already argued in 1.3.1, it is probably wrong to treat this as a
mixed system', one that contains markers of both judgments and evidentials.
't would be more reasonable to say that Deductive and Assumptive can be
se
en as both judgments and evidentials in that the relevant judgments are
cased upon evidence - in the case of Tuyuca from visual evidence and from
w
nat is more generally known. As such these fit into systems that are
Primarily either epistemic or evidential.
A pair of markers that can equally be identified as Deductive and
Assumptive is also to be found in Wintu (N. California, now almost extinct -
Schlichter 1986: 51-3): the 'inferential' -re- and what Schlichter calls the
x
Pectational' -?e. Examples of the 'inferential' (Deductive) are:

29
Modal systems: Propositional modality

heke man haraki-re'-m

somewhere EXCL go+coMPLET-lNF-DUB
'He must have gone somewhere' (I don't see him)
piya mayto'n dekna'sto'n piya ma'n biyaki-re'-m
those feet steps that EXCL be+coMPLET-iNF-DUB
'Those tracks of steps! That must have been him'
Examples of the 'expectational' (Assumptive) are:
tima min-el? pira'-?el
cold die-EXPECT starve-EXPEd
'He might freeze to death, he might starve' (it's cold and he's alone,
helpless, sick)
?imto'n nuqa'?-l
berries ripe-EXPECT
'The berries must be ripe' (it's that time of year)
Schlichter translates the assumptive as 'might' and 'must', but he also
comments that the speaker believes the statement to be true. In the case of
the Deductive it is 'because of circumstantial sensory evidence', most often
visual, and in the case of the Assumptive 'because of his experience with
similar situations, regular patterns, or repeated circumstances common in
human life'. These remarks together with the bracketed comments suggest
that these are examples of Assumptive, which would have been more clearly
indicated by a translation with WILL:
'He'll freeze to death, he'll starve'
'The berries will be ripe'
From the data given, it seems that the Assumptive in evidential systems
differs little from WILL in English, but that there is a slight difference between
the Deductive in such systems and English MUST. It appears that MUST is used
only when there is emphasis upon the deduction, whereas the Deductive in
evidential systems is used much more widely - whenever any degree of
deduction is involved. Oswalt (1986: 38-9) says of the 'inferential' in
Kashaya that it 'appears in many situations in which English would not
normally use "must have" '. Examples are:
kalikakh dima' si-qa-c-qh
book holding make-cause-self-iNF
'He has a picture taken of himself holding a book'
he?n in-iwa ma mace'-thi-qa-m
how doing-QUES you guard-NEG-iNF-RESP
'Why did you not guard her?'
He comments that 'the existence of the picture (which is seen) is the evidence
30
 2.1 Epistemic modality

hat one was taken (which act was not seen)' and that 'the lack of guarding
n ot seen but is inferred . . . because the prisoner is gone'.
It may also be noted that for one example in Makah, Jacobsen (1986: 12)
Hoes not use 'must' in the translation for an example of the marker for
'inference from physical evidence':
ha?ukaX. pi'dic
'I see you ate'
Similarly, Mithun (1986: 101-2), in the discussion of an inferential particle in
Wyandot, Seneca, Cayuga and Mohawk, uses 'must' in the translation for
only two out of eight examples.
There is some emphasis in some of the discussion on the fact that the
speaker believes the proposition to be true. Thus Schlichter (1986: 51) says of
the 'inferential' that 'the speaker believes his statement to be true' (see
above), while Oswalt (1986: 34) says 'The Kashaya Inferential suffix implies
no lack of certainty, merely lack of higher ranking evidence.' Whether
Deductive in these languages is significantly different from English MUST is
discussed in 2.7.1.

2.1.4 MAY as concessive

MAY is used as a concessive, i.e., in the sense of'although', in e.g.,

He may be rich, but he's not very lucky

He may have been rich, but he wasn't very lucky

There is a similar use of MOGEN in German (Hammer 1983: 261):

Er mag noch so gescheit sein, . . . aber . . .

he MOGEN +3SG+PRES+IND ever so intelligent be, . . . b u t . . .
'He may be intelligent, but . . .'

With a concessive clause the speaker does not indicate doubt about the
proposition, but rather accepts it as true, in order to contrast one state of
affairs with another. The use of MAY is to be explained, therefore, not in
terms of Speculative, but of Presupposed. In other languages, such as Latin,
Italian or Spanish, the subjunctive would be used (see 1.4.1 and 5.2.5).

2.1.5 Modifications
me languages have ways of indicating that the judgments are either weaker
or stronger. English, for instance, uses the past tense forms of the modals, the
m
dal-pas (see 1.4.4 and 8.2.1). These express a lower degree of commit-
me
n t , judgments that are more 'tentative'.
h
o r MAY and WILL the past tense forms might and would are used:

31
Modal systems: Propositional modality

Mary might be at the school by now ' ' ' ': '-."
Mary would be at the school by now

MUST has no past tense forms, but the modal-past equivalent is ought to
or, more commonly, should (even though formally this is the past tense form
of SHALL):

Mary ought to / should be at school by now

But this carries with it some notion of conditionality - 'provided things are as
I expect them to be', etc., and is not just a weaker or more tentative form of
MUST (Palmer 1979: 49, cf. 100-2).
There is another past tense form - could, which is also used to imply some
kind of inference, although the present tense form can is not used in a similar
sense, as in:

Mary could be at the school by now

This might suggest a possible conclusion, but one based on knowing facts
such as what time it now is or how long it takes to get there.
There is a similar use of past tense forms in German, but together with the
subjunctive. The present form of MGEN is most typically used to express a
weak epistemic judgment, and its past tense form in the subjunctive indicates
greater tentativeness (Hammer 1983: 227ff.):

Man mochte meinen, dass . . . (229)

one MGEN+3SG+IMPF+SUBJ think that. . .
'One might think that . . .'

However, the same forms of KNNEN and DRFEN are all also used in similar
senses:
Er konnte krank sein (227)
h e K.NNEN+3SG+IMP+SUBJ ill be
'He might be ill'
Er diirfte krank gewesen sein (228)
he DURFEN+3SG+IMPF+SUBJ ill been be
'He might well have been ill'

According to Hammer, 'mochte often denotes possibility or probability or

conveys a hesitant or polite doubt', while 'diirfte implies greater probability
than konnte, but at the same time more tentative and polite'. (The use of
similar forms in conditionals is discussed in detail in 8.2.)
A distinction can be made, then, in English and German between weaker
and stronger judgments. There appears to be a similar contrast in Inga
(Levinsohn 1975, see 2.1.2 above):
32
 )
' 2.1 Epistemic modality

chihoraca mal-cha cado carca fiide del bautismo (23)

at.that.time bad-CHA it was baptismal certificate
'At that time my baptismal certificate must have been incomplete'
(deduced)
chipica diablo-char ca (19)
there devil-CHAR it was
'A devil was presumably there' (deduced, probability reinforced)
However, there is a difference. In Inga, it seems, the contrast is between a
normal conclusion and a more positive one, though the translation in terms
of 'presumably' is a poor indication of reinforced probability. In English and
German the contrast is between a normal conclusion and a less positive one.

2.1.6 Past time reference

The present tense forms of MAY and WILL can be used to make a judgment in
the present about events in the past by the addition of have plus the past
participle:
Mary may/will/must have arrived by now
But it is not possible to use the past tense forms of these modals to indicate a
past judgment by the speaker. It is not possible to say:
*Yesterday Mary might/would arrive
In other words the proposition can be in the past, but the modality (the
judgment) cannot. (Admittedly, in literary language Mary might/would be
there by now could occur as an independent sentence, but only to represent
the thoughts of one of the characters, as in John relaxed - Mary would be
there by now. This is essentially reported speech (see below) with the verb of
reporting 'understood'.)
There is a reason for this. Inferences or conclusions are essentially
subjective and performative. They are actually made by the speaker, at the
time of speaking. They cannot be made in the past, although it is perfectly
possible to report them in the past with lexical verbs, e.g., Mary thought that
, Mary concluded that . . ., etc. Lyons (1977: 798) offers a theoretically
Possible example of objective epistemic modality, but it is contrived and he
concedes that the distinction between subjective and objective epistemic
modality 'is, to say the least, uncertain'.
MUST has no past tense form, so that its morphology alone rules out
Mary must be there yesterday. There is, however, an unexpected use of the
Past tense of HAVE TO to indicate that there was only one possible conclusion,
a
nd, usually, that it proved to be right.
I found the book at last - in the bookcase - it had to be there

33
Modal systems: Propositions modality

This, however, can be explained - HAVE TO is not a modal verb, and so not
subjective as MUST is. It can therefore be paraphrased as 'It was epistemically
necessary t h a t . . .', i.e., 'It couldn't have been otherwise than t h a t . . . '
However, the past tense forms of the modals are used in reported speech
with past tense verbs of reporting (and in that sense only may indicate past
time - see Palmer 1990: 43).
He may/will be there
He said he might/would be there
Since MUST has no past tense form, it may be used in similar constructions, or
replaced by had to:
He must be there
He said he must be there

2.1.7 Inference and confidence

Coates (1983: 41, 131, 177) says of the English epistemic modals:
In its most normal usage, Epistemic MUST conveys the speaker's
confidence in the truth of what he is saying, based on a deduction from
facts known to him (which may or may not be specified).
MAY and MIGHT are the modals of Epistemic Possibility, expressing the
speaker's lack of confidence in the proposition expressed.
WILL expresses the speaker's confidence in the truth of the proposition;
unlike epistemic MUST, the speaker's confidence is not based on a process
of logical inference. Instead, it is based on common sense or repeated
experience.
However, the repeated reference to 'confidence' may be a little misleading
since there are two issues here, one a matter of 'confidence' or degree of
commitment by the speaker, the other a matter of the strength of the
conclusion. For degrees of confidence can be expressed by adverbs:
Perhaps she's there
She's probably there
She's certainly there
These are not, or not wholly, equivalent to sentences with the modal verbs.
With MUST there is often some indication of the facts (often observable) on
which the inference is based. An example from Coates (1983: 41) is:
His teeth were still chattering but his forehead, when I felt it, was hot
and clammy. He said 'I must have a temperature.'
A similar example is found in Palmer (1979: 44):
All the X-rays showed absolutely negative. There was nothing wrong, so
it must just be tension, I suppose.
34
 2.2 Evidential modality
Tt is clear that it is the notion of deduction or inference from known facts
that is the essential feature of MUST, not just the confidence of the speaker,
which is expressed by the adverbs certainly, definitely, etc.
Similarly, will is not paraphrased by probably, but draws an inference, not
clearly from observable facts, but from generally known facts. Thus Mary
will be at the school by now indicates a conclusion based on knowing what the
time now is and what are Mary's movements.
By contrast, there seems to be little difference between Mary may be at
school and Perhaps Mary is at school, but this may be because both are
negative in the sense that the first suggests there is absence of good grounds
for a conclusion, the second that there is absence of real confidence. There is,
then, little or no difference, just as there is no difference between coffee
without milk and coffee without cream.
There is, then, a difference between inference and confidence. With the
former, speakers indicate that they are inferring from available information,
while in the latter they indicate the degree of confidence they have in what
they are saying. There is a clear difference between MUST and adverbs such as
certainly, the modal expressing inference and the adverb confidence. The
same is true of WILL, which is different from probably. There is less obvious
difference between MAY and perhaps, but they can still be seen, in a negative
sense, as instances of inference and confidence respectively.
However, markers of confidence may be added to the markers of inference.
With the modal verbs, it is possible for the speaker's commitment to be
modified by the addition of what Halliday (1970: 331) calls 'harmonic
combinations', which reinforce the speaker's commitment, or by 'hedges',
which reduce it. Coates (1983: 46, 138) notes for epistemic MUST the
occurrence of harmonic I'm sure, surely and certainly, and of the hedges, /
think (the most common), / mean, I suppose, I fancy, I take it and I would
guess. For MAY, she suggests harmonic perhaps, possibly, and hedges I
suppose, I think, I don't know, I wouldn't know, I'm not sure, I mean and It
seems to me. These do not belong to a grammatical system and an analysis of
them is beyond the scope of this work. (For discussion of similar issues see
2
-2.7 and 2.7.2.)

2.2 Evidential modality

2.2.7 Evidential systems

hough there are formal systems with multiple terms that are primarily
ld
ential, there are basically only two types of purely evidential categories,
Ported and Sensory (evidence of the senses). Indeed, in Ngiyambaa there
35
Modal systems: Propositioned modality

are, according to Donaldson (1980: 275-6), just two 'Evidence clitics', those
for 'sensory evidence' and 'linguistic evidence' as in:
nindu-gara girambiyi
yOU+NOM-SENS.EVID sick+PAST

'One can see you were sick'

nindu-dhan girambiyi
yOU+NOM-LING.EVID sick + PAST
'You are said to have been sick'
('Sensory evidence' in Ngiyambaa may involve any of the five senses, see
2.2.3.)
However, many formal systems include other categories, including those
discussed as judgments in 2.1. One clear example of such an extended
evidential system (from Central Pomo) was given in 1.2.1. Another is that of
Tuyuca (Brazil and Colombia - Barnes 1984: 257, cf. also Malone 1988 for a
reanalysis of the data), where there are evidential categories labelled 'visual',
'non-visual', 'apparent', 'second-hand' and 'assumed':

diga ap-wi
soccer play+3sG+PAST-vis
'He played soccer'
(I saw him play)

diga ap-ti
soccer play+3sc+PAST-NONVis
'He played soccer'
(I heard the game and him, but I didn't see it or him)

diga ap-yi
soccer play+3sG+PAST-APp
'He played soccer'
(1 have seen evidence that he played: his distinctive shoe print on the
playing fields. But I did not see him play)

diga ap-yigt
soccer play+3sc+PAST-SEC
'He played soccer'
(I obtained the information from someone else)

diga ap-hiyi
soccer play+3sG+PAST-ASSUM
'He played soccer'
(It is reasonable to assume that he did)
The first two indicate evidence of the senses, the fourth evidence from what

36
 2.2 Evidential modality
speaker has been told, while the third and fifth indicate judgments based
the evidence from what has been observed or is known.
A rather different system is suggested for Hidatsa (Siouan, USA) by G. H.
Matthews (1965: 99-100). This is interesting because of the glosses provided,
though there are serious problems about the interpretation (see 2.7.1).
Matthews suggests that the final clause of each sentence ends in a mood
morpheme; six such morphemes are glossed as follows:

Emphatic: 'indicates that the speaker knows the sentence to be

true; if a sentence that ends with the Emphatic is false, the
speaker is considered a liar'
Period: 'indicates that the speaker believes the sentence to be
true; if it should turn out otherwise, it would mean that he was
mistaken, but by no means a liar'
Quotative: 'indicates that the speaker regards what he has said to
be something that everyone knows'
Report: 'indicates that the speaker was told the information given
in the sentence by someone else, but has no evidence of its truth
value'
Indefinite/Question: 'both indicate that the speaker does not
know whether or not the sentence is true. The Indefinite also
means that the speaker thinks the listener does not know;
whereas the Question indicates that the speaker thinks that the
listener does know.'

The first three are illustrated by:

waco ikipi kure ho ski
man pipe carried EMPH
'The man (sure) did carry the pipe'
waco ikipi kure ho c
man pipe carried PER
'I suppose the man carried the pipe'
waco ikipi kure ho wareac
man pipe carried QUOT
'The man carried the pipe, they say'
'The markers for 'Report' and 'Indefinite' are rahe (see 2.2.2 for an example)
ar
>d took respectively, while 'Question' 'has no actual phonemic shape'.)
Most of the extensive evidential systems that have been described are to be
und in Native American languages, but Foley (1986: 165) reports that there
r e ey
identials in some of the Papuan languages. He quotes Loeweke and
37
 1
Modal systems: Propositional modality

M a y (1980) for examples from F a s u , w h e r e t h e list includes 'seen', 'heard',

'deduced from evidence', ' h e a r s a y from a k n o w n source', ' h e a r s a y from an
unknown source' and 'supposition':

a-pe-re
SE[EN]-COme-[SE]EN
'[1 see] it coming'
pe-ra-rakae
COme-CUST-HEARD
'[I hear] it coming'
pe-sa-reapo
COme-PAST-DED
'[I've concluded] it's coming'
pe-sa-pakae
COme-PAST-HSY.UNKNOWN
'[I've heard] it's coming'
pe-sa-ripo
COme-PAST-HSY.KNOWN
'[I've heard] it's coming'
pe-sa-pi
come-PAST-supposE
'[I think] it's coming'

E v e n m o r e striking is t h e situation in L a d a k h i ( T i b e t o - B u r m a n - K o s h a l
1979), w h e r e there are n o less t h a n t h r e e systems relating t o epistemic
m o d a l i t y (as well as a set of d e o n t i c suffixes).
First, t h e r e is a system of six suffixes, of which four are evidential -
' r e p o r t i v e ' ( R e p o r t e d ) , ' o b s e r v e d ' (Visual), 'experiential' (not fully explained)
a n d 'inference' (Deductive). Of the o t h e r t w o , o n e ('possible') indicates
Speculative a n d the o t h e r is labelled 'generic':

palldan-ni spe-ha sill-at

Paldan-ERG book-DlR read-REP
'Paldan reads a book' (a report)
kho-e lcsrj-me csd-duk
he-ERG tree-die CIU-OBSERVED
'He cuts the tree' (direct observation)
kho che-arak
he gO-EXPERIENTIAL
'He goes' (speaker's feelings)
kho-a zur-mo sante duk i-ok
he-DAT pain-DiR very be die-iNF
'He will die, because he is very sick'

38
 2.2 Evidential modality

kho-e thore na ao thuk-cen <

he-DAT tomorrow my brother-DiR meet-POSSiBLE
'He is likely to meet my brother tomorrow'
i-ma zaktan sar-ne sarranok
sun-DiR daily east-ABL rise-GEN
'The sun rises daily in the east'

Secondly, four different types of inference that involve some evidential

features are indicated by suffixes together with -thig-. The four types are
'from sounds or habitual occurrences', 'observation not correctly remem
bered', 'unobserved, partial or vague knowledge' and 'guessed, as, for
instance, about events that occurred at a distance and hence cannot be seen
clearly':
dolma yon-thig-rek
Dolma come-iNF-souND
'Dolma is coming' (hearing footsteps, voice, etc.)
khoe kane pene rje khyer-thig-yot ^
he me from money take-iNF-OBSERVED
'He might have taken money from me'
kho i-khanpe nanrja duk-thig-son
he this-house in live-iNF-UNOBSERVED
'He might have lived in this house'
apumo rdemo yot-thig-duk
that-girl beautiful be-iNF-GUESSED
'That girl might be beautiful'

Thirdly, there are four copulas used for simple statements, definite
knowledge, experience or feeling, and what is seen or heard:

na mag-mi yin
I soldier-DiR be
'I am a soldier' (simple statement)
khon-rja pe-ne yot
he-DAT money-DiR be
'He has money' (definite knowledge)
rja go-a zur-mo rak
I head-DAT pain-DiR be
'I have a head-ache' (experience)
pu-mo rdemo duk
girl-DiR beautiful be
'That girl is beautiful' (seen)

39
Modal systems: Propositions modality

2.2.2 Reported
Examples of Reported (with labels such as 'second-hand', 'linguistic evi-
dence', 'hearsay') from Tuyuca and Ngiyambaa and Fasu were given in the
last section.
Reported is also found in languages that have a system of mood, but only
where the system is 'joint' (see 6.1), i.e., where the grammatical markers of
Realis and Irrealis co-occur with grammatical markers of other categories.
This co-occurrence suggests that these categories are notionally 'realis' or
'irrealis' (and themselves typologically Realis or Irrealis). Not surprisingly,
Reported is generally so marked as Irrealis, as can be seen from the system in
Hixkaryana (Carib, N. Brazil - Derbyshire 1979: 1435), where 'uncertain' is
the mark of Irrealis and 'hearsay' the term used for Reported (see 6.5.4):
nomokyan ha-ti
he.COme+NONPAST+UNCERT INTEN-HSY
'He's coming (they say)'
However, similar forms are to be found in languages that have what are
predominantly judgment systems, e.g., German (Hammer 1983: 231, 232)
and Danish (Davidsen-Nielsen 1990: 96):
Er soil steinreich sein
He SOLLEN+3SG+PRES+IND very.rich be
'He is said to be extremely rich'
Han skal vasre enormt rig
He SKULLE+PRES be enormously rich
'He is said to be enormously rich'
For some languages, it is necessary to recognize sub-categories of Re-
ported. Willett (1988: 57, 96) suggests that there are three kinds of reported
evidence:

1) Second-hand evidence: the speaker claims to have heard of the

situation described from someone who was a direct witness.
2) Third-hand evidence: the speaker claims to have heard of the
situation described, but not from a direct witness.
3) Evidence from folklore: the speaker claims that the situation
described is part of established oral history.

It is not easy to find suitable names for these sub-categories. It might have
been possible to use the terms 'Quotative', 'Hearsay' and 'Folklore', but
there is great inconsistency in the literature in the use of these terms, and this
might lead to confusion. Moreover 'folklore' is misleading in that the third
sub-category usually refers to what is generally believed to be true, not
40
 2.2 Evidential modality

sarily a s P a r t f folklore or tradition. For these reasons the sub-

eories are to be labelled 'Reported (2)', 'Reported (3)' and 'Reported
r n)' ('Reported' may be used to refer either to the undifferentiated
n r v 0 r to the overall category that includes the sub-categories.)
In many languages the same form is used for all three types of Reported.
-rhis seems to be the case with the examples provided by Mithun (1986:
102-3) for 'quotative particles' in the Northern Iraquoian languages
Tavuga, Oneida, Mohawk and Seneca. Two of the examples are clearly from
story, while there is no indication in the others whether the evidence is
'second-hand' or 'third-hand'. The same appears to be true of Tuyuca, where
Barnes (1984: 261) talks of a speaker 'reporting information that was relayed
to him', and gives an example of Reported (Gen), in this case part of a
legend:
ana kITre baka-yigi
'A snake bit him'
Similarly the example from Hixkaryana given above is glossed as 'they say',
but Derbyshire goes on to say that the same 'hearsay' particle has to be
repeated continually through a story that is being related, and Givn (1982:
34-5) reports that in Sherpa (Tibeto-Burman) the 'indirect/hearsay' markers
is used in passages from religious texts (see also 2.7.1).
Yet some languages have markers to distinguish the subtypes. Thus Fasu
(2.2.1) has different markers for 'hearsay from a known source' and 'hearsay
from an unknown source', i.e., Reported (2) and Reported (3):
pe-sa-pakae
COme-PAST-HSY.UNKNOWN

'[I've heard] it's coming'

pe-sa-ripo
COme-PAST-HSY.KNOWN
'[I've heard] it's coming'
Similarly for Hidatsa (Siouan, USA), G. H. Matthews (1965: 99-100)
distinguishes between what he calls 'quotative' and 'report' (i.e., again,
reported (2) and Reported (3), respectively). These are glossed (see 2.2.1):
Quotative: 'indicates that the speaker regards what he has said to
be something that everyone knows'
Report: 'indicates that the speaker was told the information given
in the sentence by someone else, but has no evidence of its truth
value'
samples are:
41
Modal systems: Propositions modality

waco ikipi kure ho wareac

man pipe carried QUOT
'The man carried the pipe, they say'
waco wira rakc ho rahe
man pipe carried REP
'The man carried the pipe, they say'
In Central Pomo too (see 2.2.1 for examples), there is a distinction between
'general knowledge' and 'hearsay'. 'General knowledge' is to be identified as
Reported (Gen), while 'hearsay' seems to be either Reported (2), or Reported
(2) + Reported (3) (the example is glossed T was told').
A rather different distinction is to be found in German, which has yet
another type of Reported (see 1.3.1). For while the modal verb SOLLEN
indicates what others say, WOLLEN indicates what the person represented by
the subject says (Hammer 1983: 231, 232):
Der Geschftsfiihrer solte schon nach Hause gegangen sein
the manager SOLLEN+3SG+PAST already to house gone to.be
'The manager was said to have gone home already'
Er will eine Mosquito abgeschossen haben
he WOLLEN+3SG+PRES+IND a Mosquito shot.down have
'He claims to have shot down a Mosquito' (plane)
WOLLEN can also be used for non-verbal communication by the subject, as in:
Er wolle mich nicht erkennen
he WOLLEN+3SG+PAST me not to recognize
'He pretended not to recognize me'
There is one further point to note concerning German. As briefly men-
tioned above, the subjunctive is used to indicate what was said, i.e., Reported
(2), as in (see 5.2.5):
Bei seiner Vernehmung berief sich H. auf Notwehr. Er
in his examination appealed H to self-defence, he
sei mit S. in Streit geraten und
be+3sG+PRES+suBj with S. in quarrel fallen and
habe sich von diesem bedroht gefuhlt
have+3sc+PRES+suBj self by him threatened felt
'In the course of his cross-examination, H. pleaded self-defence. He had
become involved in a quarrel with S. and had felt himself to be
threatened by him'
The second sentence is not asserted by the speaker (or writer), but is shown
to be what was said (by H.).
A further distinction between a report that is reliable and a report that >s
unreliable is to be found in Lega, but is discussed in 2.2.4.
42
 2.2 Evidential modality

2.2J Sensory
ituation with Sensory is like that of Reported in that, although some
aees have a single category of Sensory, others have sub-categories, of
. visual (evidence from seeing) and Auditory (evidence from hearing)
the most obvious. In practice, however, there are generally three possi-
... s . (i) a single marker (Sensory), (ii) a marker for seeing and another for
. r- t n e other senses (which may be identified as Visual and non-Visual) and
r") markers for seeing and hearing (Visual and Auditory). Only rarely is
there a marker for the senses other than seeing and hearing (but see below).
Ngiyambaa (Donaldson 1980: 275-6) is of the first type - there is a clitic
which is used for sensory evidence that may involve all five senses:
nindu-gara girambiyi
yOU+NOM-SENS.EVID Sck+PAST
'One can see you were sick'
gabuga:-gara-lu namumiyi
egg+ABS-SENS.EVFD-3ERG lay+PAST
'It's laid an egg by the sound of it'
(The chicken concerned was out of sight.)
yura:bad-gara rjidji guuga-nha
rabbit+ABS-SENs.EviD here+cmc be inside-PREs
rjama-ra-bara-dhu-na " '"'''"
feel-PRES-CATEG. ASS-1NOM-3 ABS
'I can tell there's a rabbit in here. I (can) feel it for sure'
(The speaker had her hand in a burrow.) E

dhagun-gir-gara nina dhirjga: ga-ta

earth-nasty-with-SENS.EViD this+ABS meat+ABS.be+PRES
'This meat tastes nasty with earth'
(Said while attempting to eat it.)
wara:y-gara-dhu-na bungiyamiyi dhinga:-dhi:
bad+ABS-SENS.EViD-lN0M-3ABS change.with.fire+PAST meat+ABS-loBL
T have burnt my meat so it's no good, to judge by the smell'
(Said outside the house where the meat was cooking.)
TT
e sen
ses involved here are, quite obviously, those of sight, hearing, touch,
ta
ste and smell respectively.
he second type, in which there is one marker for sight (Visual) and
er (non-Visual) for all the other senses is represented by Tuyuca
'Barnes 1984: 260). Examples repeated from 2.2.1 are:
diga ap-wi
soccer play+3so+PAST-vis
'He played soccer'
(I saw him play)
43
Modal systems: Propositional modality
diga ap-ti
soccer play+3so+PAST-NONVis
'He played soccer'
(I heard the game and him, but 1 didn't see it or him)

In spite of the fact that the second example refers to hearing, Barnes states
that this 'non-visual evidential may be used to report how someone, some
thing or some event smelled, sounded, tasted or felt (smells, sounds, tastes or
feels)'. Examples are:

yoro sush-ta (ta - 3PL+PAST)

'They smelled (of liquor) a long way off'
mtru btski (ti = 'other' (1, 2 or inanimate) +PAST)
'The motor roared'

It is also used to refer to the speaker's own emotion, pain or knowledge:

paga pn-ga (ga = 'other'+PRES)

'My stomach hurts'
t s-ga
T like it'

The third type with separate markers for both Visual and Auditory are less
common. Indeed Oswalt (1986: 43) says 'Kashaya, Southern Porno and
Central Porno are apparently rare among the languages of the world in
having an evidential for a specific non-visual sense, the Auditory.'
For Central Porno, Mithun (1999: 181) offers a very neat paradigm that
clearly marks Auditory as well as Visual (plus Reported and Deductive, as
well as a simple Declarative, for which see 2.7.1). The full paradigm was
given in 1.2.1. Examples (repeated) of Auditory and Visual are:

hemul-ya
rain.fell-vis
'It rained' (I saw it)
hemul-nme'
rain.fell-AUD
'It rained' (I heard it)

For Kashaya, Oswalt notes a pair of markers for Visual and a marker for
Auditory. The Visual markers are what Oswalt (1986: 36) calls 'factual' and
'visual', although the formal difference between them is that of imperfective
and perfective respectively. These form a pair, and while both refer to what
the speaker sees or has seen, the 'factual' may also be used, more widely, to
refer to classes of actions that have been generally observed or are common
knowledge: , ,.

44
 2.2 Evidential modality
h
qowaq-w (qow'q )
pack-FACT
'(I see) he is packing'
qowaq-ya (qowahy)
pack-vis
'(I just saw) he packed, I just saw him pack'

The Auditory markers occurs with marker for both perfective and
imperfective:

mo-Vd-Vnna (mo'dun)
run-IMPFV-AUD
'I hear/heard someone running along'
mo-mac-Vnna (moma'cin)
run-PERFV-AUD
'I just heard someone run in'

However, Auditory marker are not limited to Native American languages.

As noted in 2.2.1, both a Visual marker and an Auditory marker are to be
found in the Papuan language Fasu (Foley 1986: 165; Loeweke and May
1980):

a-pe-re
SE[EN]-COme-[SE]EN
'[I see] it coming'
pe-ra-rakae
COme-CUST-HEARD
'[I hear] it coming'

In some languages there is a marker that is used mainly, but not

exclusively, for hearing. Gordon (1986b: 76-7) provides examples in Mar-
icopa (Yuman) for sight and hearing:

m-iima-'yuu
2-dance-SEE.EV
'You danced' (I know because I saw)
m-ashvar-a
2-Sng-HR.EV
'You sang' (I know because I heard it)

owever, she comments that the 'hearing evidential' is used for 'first-hand
n
owledge . . . gained not by having seen the event, but having otherwise
e
nsed (usually heard) the state or event'. Moreover, it is most strongly
s
ciated with verbs strongly associated with sound, e.g. 'sing', and with
Ver
bs0fsayingasin:

45
Modal systems: Propositional modality

Pam-sh 'im nyip ny-mhank ii-'a

Pam-suBj say me 3/1 -like say-HR.EV
'Pam told me that she likes me'
Similarly, for Makah (Nootka, Washington State), Jacobsen (1986: 9-10)
has examples of an unmarked declarative and one with the marker for
hearing {qad?i)\
wiki'caxaw
'It's bad weather'
wiki'cax akqad?i
'It sounds like bad weather'

Jacobsen, however, notes that the same marker can be used for personal
feelings, and has, in recent times, been extended to refer to direct observation.
There are no attested examples of a specific marker for the senses other
than seeing and hearing. However, in Kashaya, which unusually (see above)
distinguishes Visual and Auditory, there is a 'catch-all class', which 'happens
to include inference on evidence found apart from the act' (Oswalt 1986: 43).
This is what Oswalt calls 'Inferential I', which expresses 'lack of high ranking
evidence'. Examples are (Oswalt 1986: 38-9):

mu cohtoc-q
'He must have left, he has left'
cuhni' mu?ta-qa
'Bread has been cooked' (I can smell it)

In Maricopa, as noted above, the other senses can be expressed by the

marker usually used for hearing.
The situation in Ladakhi (discussed with examples in 2.2.1) is complex. It
would appear that there is an evidential for Visual in the basic set of
evidentials, one for Auditory (or Auditory-Assumptive) in a set of suffixes
used for inference and a copula related to Visual-Auditory. Examples,
repeated from 2.2.1, are:

kho-e lcan-me csd-duk

he-ERG tree-die cut-OBSERVED
'He cuts the tree' (direct observation)
dolma yorj-thig-rek
Dolma come-iNF-souND
'Dolma is coming' (hearing footsteps, voice, etc.)
pu-mo rdemo duk
girl-DiR beautiful be
'That girl is beautiful' (seen)

46
 2.2 Evidential modality

ddition there is an 'experiential' (not fully explained) in the basis set plus
copula for experience and feelings:
kho che-srsk
he gO-EXPERIENTIAL
'He goes' (speaker's feelings)
na go-3 zur-mo rak
I head-DAT pain-DiR be
'I have a head-ache' (experience)
Finally it may be noted that even in English there is some connection
between sensation and modality. The most common way of expressing what
one sees, hears, smells, tastes or feels is with the modal verb CAN:
I can see the moon
I can hear a funny noise
I can smell something burning
I can taste salt in this
I can feel something hard here
All of these indicate that the speaker has the sensation, not that he has the
ability to have it. The relevant point is that English does not normally
present information about sensation with simple declarative statements, but
chooses instead to use a modal form, just as Ngiyambaa indicates sensory
evidence with an evidential clitic.

2.2.4 Direct and indirect evidence

There are some languages in which there is a single marker of modality that
may indicate either what is said (Report) or what may be inferred to be so
(Deductive). For Turkish, Aksu-Koc Ayhan and Slobin (1986: 159) exem
plify a contrast between what they call 'direct experience' with the suffix -dl
and 'indirect experience' with the suffix ml, giving as an example of the
latter:

Ahmet gel-mis,
Ahmet come-Mij
'Ahmet came / must have come'
This may be interpreted as:

(a) inference: the speaker sees Ahmet's coat hanging in the front
hall, but has not yet seen Ahmet.
(b) hearsay: the speaker has been told that Ahmet has arrived, but
has not yet seen Ahmet.
(Cf
- also Lewis (1967: 101).)

47
Modal systems: Propositional modality

However, the inferential interpretation is possible only where the inferenr

is drawn from a state resulting from a previous process, such as the coat
hanging on the wall resulting from Ahmet's arrival. As a result only th
'hearsay' interpretation, not the 'inferential' interpretation is, therefore
possible for:
yag mur yag-acak-mis,
rain rain-FUT-m;'.
'It is reported that it will rain'
*'It will probably rain'
In Sherpa too, there is a contrast between what Givn (1982: 32) calls
'direct' evidence and 'indirect-hearsay' evidence, illustrated by:
ti lag a ki-yin-no
he work do-Aux-be+DiR
'He is working (and I have direct evidence to support this)'
ti lag a ki-yin-way
he work do-Aux-be+iNDiR
'He is working (I have indirect/hearsay evidence)'
There is something similar in Abkhaz (N. W. Caucasus - Hewitt 1979:
196), where there is a suffix on the verb in certain tense forms to 'indicate
that the assertion is made as a result of inference or hear-say', e.g.:
ds-r-s-x-zaap'
him-they-kill-PERF-lNFER/HSY
'Apparently they have already killed him'
However, Abkhaz has a number of constructions that express various modal
notions.
Another possible example is an Iranian language of the USSR, Tajik.
Rastorgueva (1963: 64) distinguishes between 'evident' and 'non-evident' (or
'narrative') forms, the latter being glossed as referring to 'an occurrence or
action known to the speaker, not on the basis of personal experience, but
from collateral sources, i.e. from someone else's words or from logical
inference'. The non-evident verb forms are compounds formed from the
perfect, including a form of the copula. An example given is:
aka fodi: meomadaast, rost-mi
'(they say that) Mr Shodi is coming. Is that true?'
There is no example of inference, but another example indicates non-
immediate awareness, which can be seen as a (correct) inference in the past:
man avval u:ro nafinoxtam, diktat karpa binam, A^mad budaast
'At first I didn't recognize him; then when I looked carefully, I saw it
was Ahmed' . .

48
 2.2 Evidential modality

levant to add that English uses apparently in both senses, as in the

U
elation of the Abkhaz example.
tra
ther more striking is the situation in Lega (Bantu, Zaire - Botne 1997:
9? and personal communication), where, it seems, there is a clear
st between indirect, unreliable evidence and direct, reliable evidence.
C
are three particles in this language with grammatical status in that they
nvariant and occur only in clause-initial position.
One of these, labelled 'possibility', indicates 'weak inference, often based
sensory input', and so can be seen as a marker of the judgment Speculative
(and is of no further interest here):
mbe mbula zkok ka
SPEC rain fall
'Maybe it will rain'
The other two are evidential. One (umbo) is used for Reported, but may
also be used to express doubt after verbs of saying and in questions, as in:
nkumgwg (bono), mbo bazongo b kulyg merende
I hear that EV whites eat frog
'I hear that Westerners eat frogs' (though I find that unlikely)
mbo Amsi ndileko Misisi?
QUOT Amisi went to Misisi
'Is it really the case that Amisi went to Misisi?'
The other evidential (amp) is used where the evidence is believed to be
'particularly robust or cogent'. The evidence is of three main types:

1) Direct sensory evidence, particularly from sight or hearing;

2) Strong inferential evidence (Deductive);
3) Report that can be trusted.

Examples are (the first and last from personal communication):

amp mbul zsabka
EV rain rain
'It is already raining' (I can see it)
amp kukr momponge
EV she is pounding rice
'She's assuredly pounding rice' (I can hear it)
amp Kisang ndile kw isoko; kikpu kyg tak gno
EV Kisanga went to market; basket is not here
'Kisanga surely went to the market; her basket is not here'
amp Mok korwa
EV Moke tired
'Moke is tired'(Moke told me) ,..

49
Modal systems: Propositions modality

(The last example could also be said if the speaker can see that Moke is
tired.)
The essential difference between the two evidentials is that the first
indicates evidence that is less than reliable, the second evidence that is wholly
reliable. This contrast between two types of Reported is particularly striking

2.2.5 Other evidential possibilities

There are other possibilities. In Makah (discussed above in relation to
Auditory), there is an unmarked form that indicates visual or direct
experience, but also a marker (caqi) for 'uncertain visual'. There are, thus
two types of Visual:
apac
'It's a canoe'
apaccaqil
'It looks like a canoe'
A different feature is found in Central Pomo (Mithun 1999: 181). In
addition to the five evidentials illustrated in 1.2.1 and 2.2.3, it has markers
for personal experience of actions performed by the speaker and for actions
that affect the speaker (Mithun does not suggest labels for these):
da--w-la
pulling-seize-PERF-PERS.AG
'I caught it' (I know because I did it)
da--w-wiya
pulling-seize-PERF-PERS.AG
'1 got caught' (I know because it happened to me)
The related language Kashaya has forms that Oswalt (1986: 34-6) calls
'performative', which are used to refer to actions performed by the speaker as
in:
qowaq-wela
'I am packing (a suitcase)' (imperfective)
qowaq-mela
'I just packed' (perfective)
In addition, there is another marker, (-yowa), which Oswalt (39-40) calls
'personal experience', that is used in the initial stages of narrative in place of
all the other markers except 'quotative' (= Reported):
mul--yow-e' hayu chno-w
then-ASS-PERS.EXP-NONFINAL d o g SOUnd-ABSOL

'Then (I saw, heard, judged) the dog barked'

This is more than a general marker of Sensory, since it includes inference
50
 2.2 Evidential modality

t is equally not a general marker of prepositional modality, since it

excludes reports (Reported).

2.2.6 Hierarchy
U (1986: 43) says of the evidentials in Kashaya that they lie on a
hierarchy:
performative > factual > visual > auditory > inferential > quotative
(i.e. 'Performative' > Declarative > Visual > Auditory > Deductive >
Reported)

Here those that precede have priority over those that follow. Oswalt suggests
that the order seems to be universal, even for English where evidential
concepts are expressed by verbs. 'Thus, someone speaking of an act he
himself is performing or has performed, would not normally attribute
knowledge of that event to a lower type of evidence . . . ' (than the
'Performative'). Similarly, 'a speaker of any language would not normally
say "I heard someone approaching the door" when he can see the person
approaching'.
Oswalt gives no specific examples from Kashaya, but there is much more
exemplification and discussion from Tuyuca (Barnes 1984: 262-4). The
hierarchy here is:

visual > non-visual > apparent > second-hand > assumed

(i.e. Visual > Non-visual > Deductive > Reported > Assumptive)

The lowest in the hierarchy is 'assumed' (Assumptive), which is used when

the speaker has prior knowledge about the state of things or about habitual
or general 'behaviour patterns', but 'only when no information about the
state or event is being or has been received'. In that sense it is the negative
member of the evidential system, used when no ready evidence is available.
In contrast, Visual is the preferred evidential. It is used whenever a speaker
has seen, or is looking at, a state or event. Even if he has or has had another
type of evidence, he will still use the Visual evidential rather than that
appropriate to the other types. Thus, if a man hears his dog yelp when killed
V a jaguar, he will still use the Visual evidential, not the non-Visual, if he
at
er sees evidence of the kill. The visual evidence takes precedence over the
a
uditory.
B
arnes stresses the importance of giving visual information whenever
ssible. Thus if a person is asked, on leaving his mother's house, whether
e ls
in, he will reply 'She was' with the Visual evidential, to indicate the
st
visual information, rather than 'She is', which would require the
51
Modal systems: Prepositional modality
Assumptive, since there is no other information available concerning her
present whereabouts.

2.2.7 Evidentials and certainty

There was a discussion in 2.1.7 of the differences between markers of
inference and markers of certainty, with only the former strictly belonging to
a modal system. There are, however, formal systems that contain markers of
both types, e.g., in Hanis Coos (Oregon - Frachtenberg 1922: 385-8; cf
Mithun 1999: 182):
kwa 'it seems, as if, like, kind of
yiku, k "maybe, perhaps, I guess'
hakwat, kwai 'it seems, as if
qn suspicion
qaiku supposition
qaini' sudden recollection 'Oh, I recollect'
natsi' 'I doubt'
hefin hearsay, *I was told, it is said'
il 'surely, certainly'
cku knowledge by evidence, 'it must have been that'
Most of these denote 'degrees of certainty and knowledge' and six of them
look rather like types of Speculative, though these, together with the marker
for 'surely, certainly', are more like indicators of degrees of confidence than
of inference or deduction (see 2.1.7). Yet the system also contains what are
clearly epistemic-evidential markers - Reported ('hearsay') and Deductive
('knowledge by evidence'). The whole system is, therefore, of more interest to
the issue of modality than, say, the particles in German (see 2.6.1).

2.3 Interrogative and Negative

Interrogative and Negative are sometimes, but not commonly, found as
members of an epistemic modal system. Interrogative, for example is a term
in a modal system in Menomini (Algonquian, USA), where the system
according to Hockett (1968: 2378) is illustrated by:
pi'w he comes, is coming, came
pi'wen he is said to be coming, it is said that he came
pi'? is he coming, did he come?
piasah so he is coming after all! (despite our expectations to
the contrary!)
piapah but he was going to come! (and now it turns out that he
is not!)
It should not, perhaps, be surprising that some languages use the saioe
marker for both Negative and Interrogative, since these can be seen as 'non"
52

k
 2.3 Interrogative and Negative

' j n > e.g., English (see 1.4.2). A clear example is Imbabura (Cole
3
? 164)- The full system is given in 1.4.2, but the relevant examples (with

elm) are:
mayistru-chu ka-ngui
teacher-QUES be-2sc
'Are you a teacher?'
uka-ka mana chay llama-ta shuwa-shka-ni-chu
I-TOP not that sheep-ACC steal-PERF-2sc-NEG
'I didn't steal that sheep'

Another example of the same form being used for both negation and
interrogation is to be found in Tiwi (Australia - Osborne 1974: 43). In this
language there is an unmarked 'indicative', but four marked moods -
'imperative', 'subjunctive', 'compulsional' and 'incompletive'. The 'subjunc
tive' has three functions - for negation, for non-PAST interrogation and for
what is possible but unlikely. The first two functions are illustrated by:

a-inunkwa ji-ma-ksoimi
NEG he-suBJ-do
'He doesn't do it'
pu-ma-ta-wari
they-suBJ-FUT-fight
'Are they going to fight?'

A different possibility is that there is no specific marker for Interrogative,

but that questions are indicated by a marker that indicates lack of knowledge
on the part of the speaker. This is illustrated in the system of 'knowledge'
clitics of Ngiyambaa (Donaldson 1980). There are two such clitics, identified
as 'exclamative' and 'ignorative'. The 'ignorative' is used together with
counterfactuaP and 'hypothesis' to express what might have been and
'perhaps' respectively:

minjan-ga:-ma-ndu dha-yi (253)

What-IGNOR-CNTF-2NOM eat-PAST
'You might have eaten I don't know what' (but you didn't)
guya-gila-ga:-lu dha-yi (257)
fish+A8S-HYPOTH-lGNOR-3ABS eat-PAST

'Perhaps he ate a fish'

owever, it indicates a question when no other clitic occurs:

minjan-ga:-ndu dha-yi (260,262)

What+ABS-IGNOR-2NOM eat-PAST
'You ate something, I don't know what'/'I don't know what you ate'
53

tisM.-
Modal systems: Propositional modality

guya-ga:-ndu dha-yi (260,263)

fish+ABS-IGNOR-2NOM eat-PAST
'Did you eat a fish?7'You ate a fish, I don't know'
The 'exclamative' is used in a similar way to express both ignorance and
surprise:
minja-wa:-ndu dha-yi (260,262)
what+ABS-EXCL-2NOM eat-PAST
"What did you eat?7'You ate what?
guya-wa:-ndu dha-yi (260,262)
fish+ABS-EXCL-2NOM eat-PAST
'So you ate a fishP/'What? You ate a fish!'
The indication of a question by a form that marks ignorance on the part of
the speaker is shown, even more clearly, in the 'participant system' of Kogi,
to be discussed in 2.6.2.
Rather differently, in Huichol (Mexico - Grimes 1964: 27) it appears that
unmarked forms are often taken to be questions, whereas statements are
specifically marked as 'assertive', the reverse of what is normal in most
languages. Grimes says 'Any assertion can be agreed with or disagreed with,
so that a plain assertion is the semantic equivalent of an English "Yes or No"
question. In a conversational context, however, modeless assertions are most
likely to be interpreted as questions.' Examples are:
pe-t a
ASS-direction go
'He left'
mz tikuucu
deer asleep
'Is the deer asleep?'
However, it should be pointed out that a tentative question is commonly
signalled by an expression of ignorance in most languages. Thus in English
the verb WONDER may be used to suggest a question, or possibly a request:
I wonder if he's arrived
I wonder if you can help me
(But see Bolinger 1978 for the potential distinction of question and rumina-
tion marked by if and whether.) Conversely, an interrogative may be used to
express doubt:
We're expecting him, but will he come?
Two other languages that exhibit Interrogative or both Interrogative and
Negative within a modal system are Hixkaryana (Carib, N. Brazil) and
Serrano (Uto-Aztecan, California), but in both languages there is also a

54
 2.4 Two other possibilities

*i.m of mood (realis and irrealis) and, for that reason, they are discussed in
5ystei u
6 5.4.
There are also languages that have more that one interrogative marker.
Tuyuca has a 'first-hand' and a 'nonfirst-hand' interrogative (discussed
h ' flv in 2.5), while Khezha (Tibeto-Burman) has five types of marker for
V s-No questions and six types for W^-questions (discussed in 2.6.1).

2.4 Two other possibilities

There is a 'remote' suffix in Kashaya (now rarely used), that indicates an
activity in the remote past, often the personal experience of the speaker, or if
less remote, a past that is irretrievable 'because the participants are dead, or
because the world has changed so much' (Oswalt 1986: 40):
men si-yi? ci?-thi-miy
this do-PL+HAB-NEG-REM
'They never used to do that in the old days'
It may be significant, however, that the example itself relates to a habitual
activity, for a similar example (labelled 'habitual') is found with an irrealis
marker (see 6.6.9) in Bargam (Papua - Roberts 1990: 384, quoting an
unpublished manuscript by Hepner):
miles-eq leh-id teq anamren aholwaq-ad in
return-ss+iRR go-DS+iRR then owner see-ss+siM 3SG
didaq tu-ugiaq
food PERF-give+HAB.P+3SG
'When (the pig) would return and then the owner would go and, on
seeing it, used to give it food'
Roberts (1990: 283, fn. 13) notes that English also uses an (irrealis?) modal
for expressing habitual past - When we were children, we would often play at
being grown ups. It could be added that the use of a form of WILL suggests
tendencies rather than just habitual action - that we tended to play rather
than that we actually did play, and that it may be this that permits habitual
past actions to be seen as irrealis. The relevance of this is, however questioned
b
y Bybee et al. 1994: 239 (see 7.3).
Kashaya (Oswalt 1986: 42) also has a means of expressing 'what proved to
true', where what was believed in the past is in contrast with the present
Known situation. Here there are two suffixes, -bi-, which is labelled Tnfer-
e
ntial II' and -wa-, which is a marker of the verbal feature 'absolutive':

khe hi?baya-?-bi-w
v
my man-ASS-iNF-ABSOL
'It t u r n e d o u t t o b e m y h u s b a n d ' ...';;.

55
Modal systems: Propositional modality

Slightly different are the forms noted by Hockett (1958: 237-8) for
Menomini (Algonquian) - see 2.3 for the full paradigm:

piasah so he is coming after all! (despite our expectations to the contrary!)

piapah but he was going to come! (and now it turns out that he is not!)

In both languages there is a contrast between what was expected and what
proved to be true. With the Menomini examples, moreover, it would seem
that nothing is being asserted - possible translations might be 'It is surprising
that he is coming' and i t is surprising that he isn't coming.' On this
interpretation these could be handled in terms of presupposition (as in 1.4.1
and 6.6.7). The Kashaya example does not lend itself to a similar interpret-
ation, but there is still the element of unexpectedness and it may be that this
is more important than the assertion. These examples also have something in
common with the use of English had to (2.1.6). All of them are concerned
with the recognition of a present time fact, but relating it to what was known
or assumed in the past.

2.5 Structure of propositional systems

The organization of this chapter is in terms of two types of propositional
modality, Epistemic modality and Evidential modality, and of two main
types of Evidential modality, Reported and Sensory. This is essentially the
same as that of Willett (1988: 96) who arranges the 'meanings' of the
'evidentials' as:

I. Direct Evidence
A. Visual
B. Auditory
C. Sensory
II. Indirect Evidence
A. Reported
1. Second-hand
2. Third-hand
3. From Folklore
B. Inferring
1. From results
2. From reasoning

Although the individual categories are usually clearly marked, the

grouping of them into a structure such as that suggested by Willett appears
to be largely based on notional criteria and this may seem to be a weakness
of them (see, e.g., Croft 1955: 88). However, sometimes there is formal
justification. An example of this can be seen in the analysis of the Tuyuca

56
 2.5 Structure of propositions systems

Hentials, which have been discussed several times (see, e.g., 2.2.1). The five
t e 2 ories are Visual, non-Visual, Deductive, Reported and Assumptive and

r n es (1984: 267) suggests that the first three are 'first-hand' and the last
' n onfirst-hand\ However, Malone (1988: 123) argues that only the first
are first-hand, so that Deductive is nonfirst-hand. Malone's argument
based on the fact that Tuyuca has two types of Interrogative markers
that are to be distinguished as 'firsthand' and 'nonfirsthan, the choice
being determined by the type of response expected. By this test, Deductive
firsthand like Visual and non-Visual, because it is a possible response to a
first-hand Interrogative. (Another analysis in terms of first-hand and
nonfirst-hand is to be found in Botne (1997: 523-5), following Chafe (1986:
263)andFrawley(1992:412-15).)
Similarly, but more ingeniously, Steele (1997: 289-90) offers an analysis of
Luiseio (Uto-Aztecan, S. California) that provides a formal basis for a
structural analysis that makes the contrasts quotative/non-quotative, asser
tion/non-assertion and modal/non-modal. There is, in this language a
'particle complex', which, she argues, has four parts. The third part marks
person and number and is of no relevance here. The first, second and fourth
parts are marked as zero or with a particle (shown as X if other than kun),
the possibilities being:
Non-quotative
Non-modal assertion 0-0-3-0 OR 0-0-3-X
Non-assertion X-0-3-0
Modal assertion X-0-3-X
Quotative
Assertion 0-kun-3-0 OR 0-kun-3-X
Non-assertion X-kun-3-0 OR X-kun-3-X
Examples are (with the particle complex analysed into its four parts):
noo n takwayak (0-0-n-O)
I PART.COMPL Sick
'I am sick'
heyiqus sum (su-0-m-0)
dig+PAST PART.COMPL
'Were they digging?'
heyi xumpo (xu-0-m-po)
dig PART.COMPL
'They should dig'
nookunun takwayak (0-kun-n-0)
PART.COMPL sick
'I'm sick, so I'm told' . . . .

57
 p
Modal systems: Propositional modality
heyiqus sukunum (su-kun-m-O)
dig+PAST PART.COMPL
'They were digging - is that what you said?'
It can be seen that the quotative/non-quotative distinction is made bv
Olkun in second place and assertion/non-assertion by OIX in first place. On
this analysis it would seem that 'modal assertion' is a form of non-assertion
(as would be expected - see 1.1.2) and that for non-quotative non-assertion
only it is marked by X in fourth position. What is important, of course, is
that all three distinctions are marked formally in the language and are not
purely notional.

2.6 Discourse and participants

2.6.1 Discourse systems

Modals have an important part to play in discourse, as the participants
express their opinions and attitudes and, in general, interact with one
another. It is not wholly surprising, therefore, that there are systems which
are more directly concerned with discourse relations.
A good example of a system that is more concerned with discourse than
modality is that of Cashibo (Shell 1975: 178-93). The forms used for third
person singular are:

ka declarative
kaisa report in independent clause
isa report in dependent clause
kara interrogative
ria response
riapa response where question implies scepticism
karaisna interrogative report

Examples are:
Jorgeni'n ka an lpiz ^aasiaa
George DEC+3 his pencil break+PAST+3
'George broke his pencil'
JorgenY kasa an lpiz tfaasiaa
George REP.IND+3 his pencil break+PAST+3
'George broke his pencil, it is said'
an lpiz isa ^asiaa kion ka
his pencil REP.DEP+3 break+PAST+3 CLOSE OF QUOTE DEC+3
Jorgenin ?i kaaa
George me say+PAST+3
'George told me that he broke his pencil'

58
 2.6 Discourse and participants

Jorge kara k w an
George I N T + 3 go+PRES+iNT
'Is George going?'
(a) ria k w ann
(he) RESP+3 go+PRES+3
'Yes, he is going'
(a) riapa k w ann
(he) RESP.SCEP+3 go+PRES+3
'Yes, he certainly is going!'
an ka oka ?i karaisna k w an kion
he DEC+3 ask+pAST+3 me INT.REP+3 go+PAST+1/2 CLOSE OF QUOTE
'He asked me if I went'
Cashibo, then, h a s a very full discourse system, a n d , in p a r t i c u l a r , forms for
making statements, a s k i n g q u e s t i o n s , giving replies a n d giving e m p h a t i c
replies (the first t w o being i n t e r p r e t a b l e as Declaratives a n d Interrogatives).
There a r e s o m e discourse features in t h e ' b e l i e f clitics of N g i y a m b a a
(Donaldson 1980: 2 5 2 - 5 ) , which were exemplified in 1.4.5 a n d a r e further
discussed in 2.7.2. D o n a l d s o n defines t h e m as:
Assertion 'used to draw the addressee's attention to a
statement'
Categorical assertion 'the speaker presents the statement . . . as
significant for its absolute truth'
Counter-assertion 'either contradicts a previous statement or is
intended to counter some presupposition the
speaker suspects his addressee of entertaining'
Hypothesis 'marks a statement as an unconfirmed hypothesis
on the part of the speaker'
Examples, r e p e a t e d from 1.4.5, a r e :
waija:y-ba:-na yana-nhi
neg-ASS-3ABS walk-PAST
'He didn't walk (again)'
gunhm-bata-nu: balu-y-aga
mOther+ABS-CATEG.ASS-20BL die-CM-IRR
'Your mother is bound to die'
guyan-baga:-dhu ga[a
Shy+ABS-CNTR.ASS-INOM be+PRES
'But I'm shy!'
gali:-ninda-gila rjiyanu balu-y-aga
Water-CARlT-HYPOTH We+PL+NOM die-CM-IRR
'We'll probably die for lack of water'

c o u n t e r - a s s e r t i o n ' is obviously a m a t t e r of discourse. 'Categorical asser-

59
 p

Modal systems: Propositional modality

tion' also seems to be related to discourse in that it can be triggered by what

someone else has said. In one example a place is being discussed and
someone intervenes with:
nadhi-la:-bara-dhu badhiyi
there+CIRC-EST-CATEG.ASS-lNOM COme+PAST
(EST=ESTablished reference)
'That's exactly where I've (just) come from!'
Another example is translated 'I've seen it already.'
It has not always been noted that, in an indirect way, English has a formal
indication of discourse features, found in constructions associated with the
NICE properties of the English auxiliary verbs. (This is an acronym
suggested by Huddleston (1976: 333) and based on Palmer (1974: 15).) These
features are, with examples:
Negative I can't go
Inversion Must 1 come?
'Code' He can swim and so can she
Emphatic Affirmation He will be there
Obviously, negation may be used to deny what someone has said, 'inversion'
to form a question and 'emphatic affirmation' to reaffirm what may have
been doubted. 'Code', in particular, is very relevant to discourse. It allows
the use of modal verbs without a lexical verb, the lexical verb being 'under-
stood' (recoverable) anaphorically from a previous lexical verb as in:
A. Can he do it?
B. Yes he can.
There could be a series of such modal verbs - Well he may, But he must, etc.,
so that anyone joining the conversation would not know what the topic was.
Moreover, as was seen in 2.3, Interrogative and Negative are sometimes
members of modal systems, and they too play an important role in discourse.
Many languages have sets of particles that may be used in discourse, e.g.
the 'modal particles' of German (Curme (1905: 368 (1960: 350-2))). The
most important ones (with Curme's glosses) are:
ja 'truly, why, don't you see, you know'
doch 'after all, though, just, truly, surely . . .'
denn 'evidently, as is well known, as I learn . . .'
schon 'never fear, no doubt, surely, as a matter of course'
wohl 'indeed, certainly'
However, these are not directly relevant to discourse and probably cannot be
defined as members of a grammatical system.
More precisely definable is the set of sentence-final particles in Chinese
which are glossed as follows by Li and Thompson (1981: 238ff.):

60

i
 2.6 Discourse and participants

le currently relevant state

ne response to question
ba solicit agreement
ou friendly warning
a/ya reduce forcefulness
ma question

Some examples are

yyng de le (264)
same it LE
'It's the same (you're wrong in thinking that what you have is different)'
tmen y au san tio ni ne (301)
they exist three CL cattle NE
'(Listen) they have three cows'
wo he bn bei ba (308)
I drink half glass BA
'I'll drink half a glass, OK?'
wo yo d n! ou (308)
I will hit you ou
'Let me tell you, if you do this, I will hit you'
n li a/ya cf. ni li (315)
you come A/YA you come
'You come here' 'You come here!'
ni hao ma (305)
you well MA
'Are you well? (= How are you?)'
Even more striking is the situation in Khezha (Tibeto-Burman - Kapfo,
quoted by Bhat 1999: 80-1), where there are eight types of marker for
Yes-No questions and another eight for Mi-questions, the choice being
largely related to features of discourse.
For Yes-No questions the relevant issues are:

(i) The speaker knows and expects confirmation;

(ii) The speaker has reliable information and expects confirmation;
(iii) The speaker indicates uncertainty (a negative marker);
0v) The speaker presumes the proposition to be correct, but is
uncertain;
(v) The speaker takes it for granted that the addressee will agree;
( v 0 The speaker is amazed at the proposition, asking for
reaffirmation;
(vii) T h e S p e a ker has heard an unusual rumour and is asking for
verification;

61
 w
Modal systems: Propositional modality

(viii)
The speaker is unconvinced, as the event is unusual, and wants
reaffirmation.
Examples are:
n meri-e ni-a-dire
you Mary-ACC love-REAL-iNT
'You are in love with Mary, aren't you?'
no mhechs-a-ya/ni
you work-REAL-lNT
'Are you working?'
pfs w-d-mo
father come-PERF-iNT
'I presume my father has come?'
pfa w-d-ne
father come-PERF-iNT
'Perhaps my father has (probably) come, hasn't he?'
n-ns mri-e ni--yo
you-NOM Mary-ACC love-REAL-iNT
'Oh! You are in love with Mary? (I can't believe it)'
i-z lss--mom
your-mother sick-REAL-iNT
'Is your mother ill?'
no we- ni--l
you me-ACC love-REAL-iNT
'Do you love me?'
(The suffix ~a is glossed as 'imperfective' by Kapfo, but Bhat is almost
certainly correct in treating it as realis, contrasting with irrealis future.)

2.6.2 Participant systems

In some languages there are systems indicating more direct reference to the
participants in the discourse.
In Kogi (Chibchan, N. Colombia) there is, Hensarling (1982) tentatively
suggests, an 'evidential' system indicating 'who knows what about the
situation being discussed'. She suggests a matrix for five particles, as shown
in:
Speaker Hearer Gloss
ni + + remind
na + - inform
shi + - ask
skaN - - doubt
ne - ? speculate

62
 2.6 Discourse and participants

Examples are:
ni-gu-ku-
REMIND-do-I-NEAR.PAST

'I did it just a while ago, as you know'

na-gu-jg
INFORM-do-INTERMEDIATE. PAST
'I tell you he did it some time ago'
shi-n
ASK-be+PROXIMATE
'Is that the way it is?'
skarj-gu
DOUBT-d0+PROXIMATE
"Who knows if it did just now?'
nbbi nqgutse n haqgna
lion little SPECULATE think+PUNCTILIAR
' "I wonder if it is a small lion", he thought'
(n appears in sentence-final position here, but it can occur in the same place
as the others.)
'Remind' relates what both speaker and addressee know, 'Inform' what the
speaker knows but the addressee does not, 'Ask' what the addressee knows
but the speaker does not, 'Doubt' what neither know and 'Speculate' what
the speaker does not know (the addressee's knowledge not being considered).
A more complex system is proposed by Lowe (1972) for Nambiquara
(Brazil). He suggests that there is a two-dimensional system involving:

event verification: individual, collective

speaker orientation: observation, deduction, narration

The 'speaker-orientation' system is clearly an evidential system with the three

familiar terms Visual, Deductive and Report:

observation I report what I saw

deduction I tell my deduction
narration I was told
ne
event verification' system involves speaker and addressee - whether the
P e aker alone or both speaker and addressee saw the event, deduced that it
cc
urred or were told about it. Lowe provides the following glosses:
individual, observation: 'I report to you what I saw the actor doing'
individual, deduction: 'I tell you my deduction of an action that
must have occurred because of something I
';' ' see or saw'

63
Modal systems: Propositional modality
individual, narration: 'I was told by someone that a certain action
occurred'
collective, observation: 'I report what both I and the addressee saw
the actor doing'
collective, deduction: 'From what the speaker and the addressee
saw, they deduce that a certain action must
have taken place'
collective, narration: 'Both speaker and addressee were told that a
certain event took place'

The individual verification third-person forms for the verb wa3kon3 (with
he glossed as 'past' and ra2 as 'definite aspect' and two of the person +
3

verification-orientation markers underlined) are given as:

wa3kon3a'h3ra2 'He worked'

wa3kon33h3ra2 'He must have worked'
wa" kon3ta'h3ra2 'I was told that he worked (past)'

The corresponding collective variation forms (with only final wa2 glossed as
'indefinite mood') are given as:

wa3kon3tait'ti2tu3wa2 'Both you and 1 saw that he worked'

wa3kon3nait'ti2tu3wa2 'He worked, as deduced from what we saw'
wa3kon3ta1t'?ti2tu3wa2 'It was told us that he worked'

2.7 Declaratives
It is generally the case that, in languages with epistemic modal systems, there
is a form that is unmarked for modality, and which, notionally, simply makes
an unqualified assertion. It can thus also be seen as Realis, while the modal
forms are Irrealis. This form is identified as the Declarative. Thus, for
English, the Declarative is marked by the absence of a modal verb, although
it will be marked for the other verbal categories, tense and aspect. (There is
usually another unmarked form, the Imperative, but this is notionally not
assertive but directive, and is best discussed together with deontic modality -
see 3.4.)
Since it is unmarked for modality, the Declarative is not to be seen as in
any sense 'stronger' than a modal form. It simply asserts without indicating
the reasons for that assertion or the speaker's commitment to it. It is perfectly
possible for a speaker to say John is in his office without being wholly sure or
when there is strong evidence for it, i.e., in circumstances where John may be
in his office or John must be in his office would also be appropriate.
It is perfectly possible to express modal notions without the use of modal
systems. This can be achieved through the use of lexical verbs as in:

64
 2.7 Declaratives

I think that Mary is in her office

I saw John in his office this morning
mmatically, the first of these is not an example of the modal judgment
lative, a n c j equally the second is not an example of the Visual evidential.
v are both Declaratives, and the relevant notional interpretation is in
rfns of asserting that 'I think' and that 'I saw'. (The situation is different
A generally simpler in most of the mood systems discussed in Chapters
i_7 where there is generally a system of two, Realis and Irrealis. Grammati-
ally neither is unmarked, both being of equal status within the system.)

2.7.1 Declaratives and modal systems

In many languages there is a Declarative form that is independent of any
modal system. This is so of English where there is a contrast between a
Declarative and epistemic modals:
Mary is at home
Mary may/must/will be at home
In terms of the Realis/Irrealis distinction, the situation seems to be clear - the
Declarative with no modal verb is Realis and the modal forms are Irrealis.
However, this simple situation - unmarked Realis and a system of marked
realis forms - is not typical of all languages with modal systems. There are
several different situations.
First, in a few languages, the Declarative is a term in the same formal
system as the modal categories. This is so in Ngiyambaa (New South Wales,
Australia - Donaldson 1980: 159-62), as fully explained and exemplified in
1.4.5. Here past and present (Declaratives) occur in the same system as
markers of Speculative ('irrealis') and Deductive ('purposive') as well as
Imperative (discussed in 3.4).
Secondly, and much more problematic, is the situation in which there
appears to be no formally and notionally unmarked Declarative because all
the categories not only belong to a single formal system, but also are all
notionally evidential. This is so in Tuyuca (see 2.2.1), where the 'strongest'
evidential is Visual, and so the closest to Declarative, but it does not simply
assert, it also indicates that there is visual evidence for the proposition,
samples, now repeated and glossed with the evidential labels used in this
book. are:

Tuyuca
diga ap-wi
soccer play+3sG+PAST-vis
'He played soccer'

65
Modal systems: Propositional modality
diga ap-ti
soccer play+3sG+PAST-AUD
'He played soccer'
diga ap-yi
soccer play+3sG+PAST-DED
'He played soccer'
diga ap-yigi
soccer play+3sG+PAST-QUOT
'He played soccer'
diga ap-hlyi
soccer play+3sc+PAST-ASS
'He played soccer'
It is instructive to compare an apparently very similar system in Central
Porno (repeated from 1.2.1) where there is both an unmarked form (though it
is rarely used) and a set of marked evidential categories (though 'general
knowledge' seems very close to being an instance of Declarative):
hemul
rain fell
'It rained'
hemul-?ma
rain fell-GEN.KNOw
'It rained' (that's an established fact)
hemul-ya
rain fell-vis
'It rained' (I saw it)
hemul-nme'
rain fell-AUD
'It rained' (I heard it)
hemul-?do
rain fell-HSY
'It rained' (I was told)
hemul-?ka
rain fell-iNF
'It rained' (Everything is wet)
Central Porno, then, has a Declarative and an evidential system, whereas
Tuyuca has no Declarative but simply an evidential system.
There is, in fact, another language for which it has been claimed that there
is no 'unmodalized declarative' (Lyons 1982: 110), this claim being based on
the assertion by G. H. Matthews (1965: 98) that all main sentences in Hidatsa
end in a 'Mood' (for details see 2.2.1). However, this is not strictly true, since

66
 2.7 Declaratives
hown that the 'Period' marker does not occur with sentences that end
'' p a s t tense marker (G. H. Matthews 1965: 110, where it is stated that
W
+ Period is 'reduced' by a rule to Past). In fact, then, there is an
marked Declarative with past tense.
Thirdly, even when the Declarative is formally distinct from the evidential
tom sometimes it can be used only where there is direct or visual evidence,
system*
ru s in Serrano (Hill 1967: 18), which is discussed in 6.5.4 because it has
ood as well as a modal system, the simple past is used only to report first-
hand knowledge. The speaker, that is to say, identifies himself as a witness,
fjjll gives as an example:
'i:p bi' wahi' pinq
here he+PAST coyote pass
'The coyote passed here' (I saw him)

The importance of direct evidence is also shown in the account of

Hixkaryana (see 2.2.2 and 6.5.4). Derbyshire (1979: 145) says that the
distinction between 'hearsay' and 'eyewitness' is crucial and that 'in any
connected discourse concerning events that have not been witnessed by the
speaker', the 'hearsay' evidential occurs in almost every clause.
Perhaps even more striking is the contrast in Sherpa between direct
evidence and 'indirect-hearsay' evidence as illustrated in 2.2.4. Givn (1982:
34-5). reports that a Lama narrating the Life of Buddha told the bulk of the
story in the perfective/past using the hearsay/indirect evidence suffix, and
that in the entire narrative only two direct evidence suffixes occurred, and
that these were both in direct quoted speech. Givn comments that the story
'is undoubtedly considered the "truest" of all stories for a devout Tibetan
Buddhist such as the story-teller himself, yet is told in the "hearsay/indirect
evidence mode" because it was not witnessed by the speaker'.
Fourthly, as was noted in 2.1.3, in some languages, e.g. Kashaya and
Makah, the Deductive is used much more widely than English MUST. It seems
that, while MUST lays some emphasis on the fact that a judgment (based on
evidence) is being made, Deductive in these languages is used for any
statement for which there is circumstantial evidence. This relates to the
question whether, in terms of the speaker's belief, the Deductive is stronger
or
weaker than the unmarked Declarative. On Kashaya, Oswalt (1986: 34)
avs
the Inferential suffix implies no lack of certainty, merely lack of higher
rar
>kmg evidence'. For Wintu, Schlichter (1986: 51) says that for Assumptive
Well as Deductive the speaker believes the statement to be true 'because of
Clr
cumstantial sensory evidence'. For Deductive this evidence is most often
Ui
, and for Assumptive it is 'because of his experience with similar
67
Modal systems: Propositions modality

situations, regular patterns, or repeated circumstances common in human

life'.
However, it may be that the same is true of epistemic MUST and WILL in
English. Someone who says John must be in his office or John 11 be in his offiCe
must, surely, believe that John is in his office. The issue of Declarative vs
modal categories is not simply a matter of belief, but of the contrast between
merely asserting and indicating either that a judgment has been made or that
there is evidence for the proposition. The difference between English and
languages such as Kashaya is that there is a greater tendency to use modal
forms in Kashaya. Yet the fact that this language uses Deductive where
English would use Declarative, does not necessarily imply that this Deductive
is more indicative of the speaker's belief than English MUST.
There is one final point. The Declarative may not be the unmarked form.
Thus in Huichol (Mexico - Grimes 1964), an unmarked form is usually taken
as a question, while the Declarative has the 'assertive marker' (see 2.3).

2.7.2 Assertion and 'strong assertion'

In some languages there are, it seems, both a 'weaker' and a 'stronger'
Declarative. Thus Imbabura (Cole 1982: 164) has forms for both 'emphatic
first-hand information' and 'first-hand information' as in (the second
repeated from 1.4.2):
fiuka-ta miku-naya-n-mari
I-ACC eat-DES-3-EF.INF
'1 want to eat'
kan-paj ushi-wan Agatu-pi-mi
you-of daughter-with Ageto-in-F.iNF
'I met your daughter in Ageto'
It was also noted in 1.4.5 and 2.6.1 that there are two 'belief clitics in
Ngiyambaa (Donaldson 1980: 252-5). These are defined as:

Assertion 'used to draw the addressee's attention to a statement.

It is literally equivalent to a statement with "I assert t h a t . . . " '
Categorical assertion 'the speaker presents the s t a t e m e n t . . . as
significant for its absolute truth. Literally, [it] is equivalent to
prefacing the statement with "I categorically assert t h a t . . . " '

Examples (repeated) are:

waga:y-ba:-na yana-nhi (254)

NEG-ASS-3ABS walk-PAST
'He didn't walk' (again)
68
 2.7 Declaratives
wana:y-ba:t-na yana-nhi (254)
NEG-CATEG-ASS-3ABS walk-PAST
'He absolutely didn't walk (again)V'He never walked again'
Similarly, for Hidatsa (see 2.2.1), G. H. Matthews (1965: 99-100) distin-
ishes between 'Emphatic' and 'Period', with the comments:

Emphatic: 'indicates that the speaker knows the sentence to be

true; if a sentence that ends with the Emphatic is false, the
speaker is considered a liar'
Period: 'indicates that the speaker believes the sentence to be
true; if it should turn out otherwise, it would mean that he was
mistaken, but by no means a liar'

Examples (repeated) are:

waco ikipi kure ho ski
man pipe carried EMPH
'The man (sure) did carry the pipe'
waco ikipi kure ho c
man pipe carried PER
T suppose the man carried the pipe'
The distinction, it is claimed, is one of knowledge vs belief, but the situation
in all three languages is very similar.
If modal systems are treated in terms of Realis and Irrealis, it seems that in
these three languages there is not only an Irrealis system, but also a two-term
Realis system, in which two degrees of strength of assertion are distinguished.

69
 3
Modal systems: Event modality

Deontic and dynamic modality refer to events that are not actualized, events
that have not taken place but are merely potential, and may, therefore, be
described as 'event modality'. The basic difference between deontic and
dynamic modality is that with dynamic modality the conditioning factors are
external to the person indicated as the subject (that he is permitted, ordered,
etc., to act), whereas with deontic modality they are internal (that he is able,
willing, etc., to act). There are, however, some other points. First deontic
modality is generally dependent on some kind of authority, often the
speaker. Secondly, Commissive (where the speaker guarantees that the action
will take place) may also be included under deontic modality. Thirdly,
dynamic ability may sometimes be interpreted in terms of the general
circumstances that make action possible or impossible (see 1.3.2) rather than
the actual ability of the subject.
However, it must be admitted that a great deal of material for this chapter
comes from English, for English appears to have a more extensive system of
deontic modality than many other languages.

3.1 Formal systems

As already noted several times (see especially 1.4.5), the same forms are often
used for epistemic and for deontic/dynamic systems. Equally the same form
(e.g. CAN) may be either deontic or dynamic. Moreover, as also noted in
1.4.5, modal markers may occur in the same formal system as other
grammatical markers, e.g., the 'purposive' marker in Ngiyambaa, which not
only may be either epistemic or deontic, but also occurs in a formal system
containing tense markers.

3.2 Deontic
The most common types of Deontic modality are the 'directives', 'where we
try to get others to do things' (Searle (1983: 166)).
70
 3.2 Deontic

3.2-1 Directives
kinds of directive are expressed in English by MAY and MUST, the two
rial verbs that also express epistemic Speculative and Deductive, though in
,. o q u ial speech CAN is preferred to MAY, but MAY still survives - see Palmer
(l990:71)-asin:

You may/can go now

You must go now
These can be identified as Permissive and Obligative.
There are similar pairs of verbs in other European languages, e.g. German:
Du magst herein kommen
You can/may+2sG+PRES in.here come
'You may come in'
Du musst herein kommen
You must+2so+PRES in.here come
'You must come in'
So also in Italian, where, however, the third person form is used instead of
the second for politeness:
Pu entrare
can/may+3sG+PRES come.in
'You may come in'
Deve entrare
must+3sG+PRES come.in
'You must come in'
Modern Greek also uses the same form as for the epistemic counterparts, but
with one difference - that the Permissive form is not impersonal like
epistemic Speculative, but agrees with the subject:
boris na fiyis
can+2sc+PRES that you.leave
'You may leave'
prepi na fiyis
must+iMPERS that you.leave
'You must leave'
milarly, although French has two verbs, POUVOIR and DEVOIR, the form
normally used for obligation is the impersonal ilfaut.
father more surprising, in Danish and North Frisian, the modal verb
se
Q for epistemic deduction is also used for both deontic permission
ncl
obligation, although other forms are available to avoid ambiguity.
he
relevant verb is matte in Danish (Davidsen-Nielsen 1990: 84, 187,
19
4): . ,,

71
Modal systems: Event modality

Der m vasre flere andre grunde

There MTTE+PRES be several other reasons
'There must be several other reasons' (epistemic necessity)
Du m danse en dans til
you MTTE+PRES dance a dance more
'You may dance another dance' (deontic possibility)
Vi m bare tage chancen
we MTTE+PRES just take the chance
'We'll just have to take the chance' (deontic necessity)
There are pairs of forms in many other languages. Thus in Tamil
(Dravidian - Asher 1982: 167-70) the suffixes -taam and -um are used, as
with epistemic modality - see 2.1.2 - for permission and obligation (Debitive)
respectively:
ven_um-n_i\aakkaa, naalekki avan peeca-laam
want-coND tomorrow he speak-PERM
'If he wants, he can speak tomorrow'
avan arjke pooka-n_um
he there go-DEB
'He must go there'
Similarly in Lisu (Lolo-Burmese - Hope 1974: 122, 126), there is a set of
intransitive verb stems that include forms for obligation and permission,
though these belong to a wider system that includes dynamic modals - see
3.3.1:
sa nya ami khwa wa-a
Asa TOP field hoe obligatory-DEC
'It is obligatory for Asa to hoe fields'
sa nya ami khwa da-a
Asa TOP field hoe acceptable-DEC
'It is acceptable for Asa to hoe fields'
The modals of permission and obligation (Permissive and Obligative) can
be interpreted (like epistemic Speculative and Deductive) in terms of possi-
bility and necessity. This is discussed in detail in 4.1.2.

3.2.2 Commissive
Commissives are defined by Searle (1983: 166) as 'where we commit ourselves
to do things'. They are signalled in English by the modal verb SHALL. They
can usually be seen either as promises or as threats; the only different
between these seeming to be whether what the speaker undertakes to do is r
is not welcome to the addressee:

72
-Tpr

3.2 Deontic

John shall have the book tomorrow

You shall do as you are told
the speaker commits himself to ensuring that the event takes place, by
anteeing to arrange that John will receive the book and that the
8
Hressee w m ^0 what is demanded, (SHALL is also used together with WILL,
m e formal dialects of English, for future time reference, but in this
ire use SHALL occurs only with first person subjects, SHALL used with 2nd
3rd person subjects specifically signals a Commissive.)
Most languages do not have a specific grammatical form for Commissives,
hut it may also be noted that in Ngiyambaa (Donaldson 1980: 160, 161) the
'irrealis' marker is used epistemically for 'might' and 'likely', but deontically
for authoritative 'shall' (see 1.4.5):
yururju rjidja-1-aga
rain-ERG rain-CM-iRR
'It might/will rain'

warja:y-ndu-gal dhagurma-gu yana-y-aga

NEG-2NOM-PL cemetery-DAT go-CM-iRR
'You shall none of you go to the cemetery'

It may be significant that the last example and other similar examples quoted
by Donaldson are negative. This appears to be related to, but not identical
with, the issues of negation with possibility and necessity discussed in 2.1.6
and 3.3.2. (Ngiyambaa has a different form for obligation, the 'purposive',
which will be discussed in 3.5.)

3.2.3 Modifications
Just as the past tense forms of epistemic modals can be use to indicate more
tentative, weaker, judgments, so past tense forms of some of the deontic
modals are used to weaken the force of the modality, MUST has ought to and
should (for any difference between these see Coates 1983: 7783), and MAY and
CAN have might and could). Morphologically, should is formally the past tense
rm of SHALL, but notionally it functions as the modified form of MUST.
!he position is clearest with ought to (should). There are two points. The
rst
is that these verbs differ from MUST in that the speaker admits the
Possibility that the event may not take place. This is seen in:
He ought to /should come, but he won't
*He must come, but he won't
Th
e
second example here is most unlikely, if not anomalous; if the speaker
ks
that the obligation may not be fulfilled, ought to/should would be used.
73
Modal systems: Event modality

Secondly, ought to/should (plus have) can refer to past events, whereas MUST
cannot:
You ought to/should have come

Must have can be interpreted epistemically (2.1.6), but never deontically.

The explanation for these two points is that ought to and should are
essentially conditional - referring to what would occur or would have
occurred. The conditionality does not refer strictly to the modal, but to the
event expressed in the proposition (Palmer 1979: 102; 1990: 125). Thus the
following glosses are appropriate:
You ought to come
'You have an obligation to come, and you would come if you fulfilled it'
You ought to have come
'You had an obligation to come and you would have come if you had
fulfilled it'
The situation with might is less simple. It is often used in questions as a
more tentative, more polite, form for asking permission (Palmer 1990: 80):
Might I come in at the moment on this, Chairman?
But it is also used to make a quite positive suggestion, as in (Palmer 1990:
187):
You might try nagging the Abbey National
You might have told me
Here again, as with ought to/should, there is conditionality, referring to events
that would take place or would have taken place under certain conditions.
With ought to/should the conditions were the carrying out of an obligation;
with might they seem to be those of behaving in a sensible or proper way. But
this is a much stronger notion than mere permission, and it is clear that might
have does not mean 'You would have acted thus if you had had permission.
Apparently then, might expresses a stronger kind of deontic modality than
MAY - a positive suggestion rather than mere permission.
Degrees of obligation are, however, signalled in other languages in different
ways. Thus in Albanian (Newmark, Hubbard and Prifti 1982: 102-3; Frawley
1992: 423) the 'must'/'should' contrast is marked by different modal verbs:
kjo kmish duhet lar
this shirt must wash
'This shirt must be washed'
kjo kmish do lar
this shirt want wash
'This shirt needs washing' .

74
 3.2 Deontic

$2.4 Subjectivity
deontic modals are often used to indicate permission and obligation
anating from the speaker, but it cannot be claimed that they are always
biective in this sense. The speaker may not be involved in, e.g.,
You can smoke in here
You must take your shoes off when you enter the temple
However, generally there is an implication that the speaker agrees with the
ermission or obligation. For obligation it is relevant that English has a form
that is an alternative to MUST, and which generally indicates that the speaker
takes no responsibility for the obligation. This is HAVE TO. There is a potential
contrast between:
You must come and see me tomorrow
You have to come and see me tomorrow
The first could be no more than a suggestion or an invitation. The second
suggests that there is some compelling reason independent of the speaker. If
there is not, the addressee might take offence, regarding it as presumptuous
of the speaker to say what he or she has to do.
Similarly, BE SUPPOSED TO is used instead of should or ought to where the
speaker is not responsible:
You should go to London tomorrow
You are supposed to go to London tomorrow
It is difficult to judge to what extent directives, especially Obligative, are
subjective in this sense. There is, in Tiwi (Australia - Osborne 1974: 44), a
marker that is labelled 'compulsional' (see 2.3):
a-u-kaaimi
he-coMPUL-do
'He has to do it'
a-u-ra-kajimi
he-COMPUL-FUT-do
'He will have to do it'
"is, Bybee (1985: 167) suggests, on the basis of the translations, does not
cate obligation imposed by the speaker, but merely asserts that the agent
s an
, Wigation. However, Osborne says that this 'compulsional' form is
T lv alent to "must" or "have t o " ' , which does not support Bybee's
sgestion. He also states that it occurs in the non-PAST or the present, which
P le s that it does not occur in the past. This might well suggest that it is, in
' objective (that the speaker imposes the obligation), as will be argued in
he n
ext section.

75
Modal systems: Event modality

3.2.5 Past time reference

It was suggested in 2.1.7 that, because the epistemic modals are essentials
subjective, they are not used in past tense forms to indicate a judgment in the
past.
It is equally true, for English at least, that the past tense forms of the
deontic modals are not used for past time reference. The past tense forms of
MAY and CAN, might and could, are not used to indicate permission in the
past, while MUST has no past tense forms. It is, therefore, not possible to sav
with a deontic sense:
*I might/could/must(ed) come yesterday
{Could is possible in a dynamic sense - 3.3.2.) Moreover, for an obligation in
the past, which by definition cannot be imposed by the speaker at the present
time, the past tense form of HAVE TO is available:
He had to be in London yesterday
As with the epistemic modals (2.1.7), past tense forms may be used in
reported speech (with MUST unchanged or replaced by had to):
He may(can)/ must come on Tuesday
He said he might(could)/must/had to come on Tuesday
There is a slightly different situation with the past tense forms should and
ought to, in that they do not strictly lay an obligation, but merely indicate
what the speaker thinks is right. It is, therefore, possible to indicate such
deontic judgments in relation to past events. English does this with should
have and ought to have:
He should have/ought to have gone to London the next day
The restrictions on past tense are, however, essentially a feature of English.
There is no similar restriction on, for instance, the comparable German modal
MSSEN, which has past tense forms that may be used for past time reference:
Ich musste fleissig arbeiten
I mussen+3sG+PAST+iND hard work
'I had to work hard'
The same is true of the related verbs in the Romance languages.

3.3 Dynamic

3.3.1 Ability and willingness

There appear to be two types of dynamic modality, expressing ability a n
willingness (Abilitive and Volitive), which are expressed in English by CA^
and WILL: : ;

76
 3.3 Dynamic

My destiny's in my control. I can make or break my life myself

Why don't you go and see if Martin will let you stay?
Aaain it is to be noted that the verbs here are verbs also used for other
of modality. CAN is used both for epistemic modality (but only when
typ
ated) and for deontic modality. It can thus be seen as expressing a third
e of modality, one that is internal to the subject of the sentence. Similarly
. permissive forms of other European languages that were discussed in
-! 7 1 are used to indicate ability. In many languages there is no formal
distinction between permission and ability, but in English, however, the
distinction is clear, in that MAY is not used to indicate ability (see 4.2.3). As
mentioned in 1.3.2, however, dynamic CAN is used not only to refer to
physical and mental powers, but also to include circumstances that might
affect the person involved, as (repeated) in:
He can run a mile in under four minutes (ability)
He can escape (there is nothing to stop him)
In Lisu (Lolo-Burmese - Hope 1974: 122-6) there are distinct forms to
indicate two types of ability, one in the sense of 'knowing how', the other in
the sense of physical ability:
sa nya ami khwa kwu-a
Asa TOP field hoe mentally.able-DEC
Asa is able (knows how) to hoe fields'
sa nya ami khwa kwhu-a
Asa TOP field hoe able-DEC
'Asa is (physically) able to hoe fields'
sa nya ami khwa da-a
Asa TOP field hoe acceptable-DEC
'It is acceptable for Asa to hoe fields'
There are three other types of dynamic possibility, indicating freedom from
taboo, no hindrance and having sufficient courage:
sa nya ami khwa ty-a
Asa TOP field hoe freedom.taboo-DEC
'It is not taboo for Asa to hoe fields'
sa nya ami khwa bala-a
Asa TOP field hoe no.hindrance-DEC
Asa is free to hoe fields'
sa nya ami khwa pa-a
Asa TOP field hoe able.couragewise-DEC
tt Asa dares to hoe fields'
We
ver, the Lisu verbs are members of a fairly large group, which includes

77
Modal systems: Event modality

not only the deontic and dynamic forms, but also others with translations 'Jt
is normal for Asa to hoe the field', 'Asa's hoeing the field has happened
before', 'Asa is tired of hoeing the field.'
It is worth adding that there are European languages that distinguish th.es
two kinds of ability, by using the verb 'know' for mental ability, e.g. French
savoir:
II sait nager
he know+3sG+PRES+iNDic to swim
'He can swim'
However, there are no formal grounds for treating SAVOIR as a modal verb
(see 4.2.2). More strikingly, in Chinese the modal verb hui has the general
meaning of 'know how'. Hockett (1968: 62) comments that 'If a person
knows how to speak Swahili, he hwi speak Swahili; if he smokes he hwi
smoke.' He adds that it is used even with inanimate objects - 'one can say
that a high wind hwi. . . blow down a tent or that an electron hwi behave
in accordance with the equations of wave mechanics'.
One further point may be made in connection with Lisu. There is a form to
express 'able couragewise', which can be translated by dare in English. DARE
is a modal verb, though restricted to non-assertive environments (see 1.4.2).
It can, then, be seen as a marker of another type of dynamic modality
(Palmer 1990: 111-12).
WILL is used in English not only as an Assumptive (2.1.3) and to indicate
future (4.3.2), but as a Volitive, to express willingness. Attested examples that
clearly indicate such willingness rather than mere futurity are (Palmer 1990:
134):
Why don't you go and see if Martin will let you stay?
She loves him and she won't leave him.
Will you stand by the anchor?
There is one formal difference between future WILL and volitive WILL - only
the latter is used in the protasis {If- clause) of a conditional sentence:
It'll rain tomorrow -> If it rains tomorrow
John'll help you - If John'll help you
Corresponding verbs in other European languages usually have much
wider meanings, expressing wishing and wanting, and, in the case of German.
WOLLEN also has the evidential meaning of what is said, claimed or pretende
by the subject of the sentence (see 2.2.2).
There are examples in some languages of a form that indicates intentio*1-
e.g., the 'intentives' of Tonkawa (Texas - Hoijer 1931: 289-90) and Maid1'
(California - Shipley 1964: 46-52):

78
 3.3 Dynamic
heul-aha'a 'I shall catch him'
?jRj?ts 'I'm going to go'
Ability c a n ^ e interpreted in terms of dynamic possibility, as permission
v be interpreted in terms of deontic possibility (see 3.2.1). It might be
ected that there would be a similar use of MUST to express what is
essarily so in a similar way. Such a use of MUST is, it seems, fairly rare, but
"/perhaps, found in (Palmer 1990: 130):
He's a man who must have money
Another example is (Perkins 1982: 260):
You must go poking your nose into everything
However, these two are slightly different. The first suggests an overwhelming
desire, the second an overwhelming propensity. The second sense can also be
expressed by can't help as in:
You can't help poking your nose into everything

3.3.2 Past tense and time

One respect in which these dynamic modals differ in English from the
epistemic and deontic modals is that they can be used in the past tense to
refer to past time:
When I was younger 1 could run much faster
All he would accept was our thanks
However, neither of them can be used to refer to an event that was actually
achieved as a result of the ability or willingness. Thus it is not possible to say:
*I ran fast and could catch the bus
*I asked him and he would come
et there is no problem with the corresponding negative form or habitual
forms:

I ran fast, but couldn't catch the bus

I asked him but he wouldn't come
I always ran fast and so could catch the bus
I always asked him and he would come
Th
ere
is, however, no similar restriction on the almost equivalent forms BE
ABLE T O
and BE WILLING TO
I ran fast and was able to catch the bus
I asked him and he was willing to come
Th
e
is a very interesting explanation for this - that a modal, being irrealis,
'"appropriate to refer to what is a realis situation - a completed or a past

79
Modal systems: Event modality

action. But negation and habitual past can be regarded as irrealis, as wi]i L
illustrated in 6.6.4 and 6.6.9, and it may be for that reason that the past ten
forms can be used when there is negation or reference to habitual action f
the situation is then irrealis. For more discussion see Palmer 1990: 92
154-7.
Past tense forms, as with other modals, can be used as 'modifications'
express ability and willingness more tentatively, especially when makine a
offer as in:
1 can/could do that for you
I will/would do that for you
However, there is often an implication of conditionality - 'if you wanted me
to', 'if you asked me'.
Could is also used, rather like might, for a suggestion:
You could try nagging the Abbey National
You could have told me

3.4 Imperative and jussive

Most languages have a specific form that can be identified as Imperative. In
languages such as English where the modal systems are marked by modal
verbs the imperative is quite independent of the modal systems, in English
being indicated by the simple form of the verb, e.g. Come here.
Notionally, Imperative is closely associated with deontic modals. It is
clearly directive and usually portrayed as indicating a command. In fact it is
often thought to be the strongest of the directives, one that emanates from
someone in authority, which, therefore, does not expect non-compliance.
However, there are two points to be noted about its relationship to the modal
verbs.
First, it can be used not only to give commands, but also simply to give
permission or advice as in:
Come in!
Don't worry about it
Thus, Come in! may be interpretable as either You may come in or You must
come in. Secondly, Imperative is performative and subjective in that the
speaker actually gives the 'command' in the act of speaking. For this reason.
unlike the directives, it does not normally occur in a subordinate clause-
Compare:
'You must come' I said that she must come
'Come in!' *I said that come in
(But see 5.4.1.)
80
 3.4 Imperative andjussive

T yens /iv977: 747) argues that the imperative can only be, strictly, second
and never third person (or first person). This may, however, be no
Pe . a n a terminological issue, since first and third person 'imperatives'
00
ften simply called 'jussives'. Bybee (1985: 171) suggests that where there
iX
{ 11 set of person-number forms the term 'optative' is used, but this is not
'S lv suitable in view of the fact that the term is used traditionally for the

* tive' mood in Classical Greek (8.2.2). (But see also 5.4.2 for data for
"opt**
u- h 'optative' might be more appropriate.) The term 'Jussive' (plus
Imperative) is preferred here.
Imperative and Jussive may belong to a modal system. Thus in Afar
rrushitic, Ethiopia - Bliese 1981: 139-46), there is, in the same formal
vstem, 'imperative', 'jussive', a 'subjunctive' and a 'consultative', as exempli-
fied by:
imperative: 'ab 'do'
jussive: 'nakay 'let me drink'
subjunctive: 'rabu 'may I die'
consultative: a'boo 'shall I do it?'

All of these could be regarded as deontic - as examples of event modality,

relating to possible events in the future.
It was noted in 1.4.5 that in Ngiyambaa (Donaldson 1980: 159),
Imperative occurs in the same system (of verbal inflection) as the Permissive
and the Commissive (whose forms also have epistemic functions) - see
3.2.1, 3.2.2:

nindu bawun-ga yuwa-dha

you+NOM middle-Loc lie-iMP
'You lie in the middle!'

However, past and present Declaratives (see 2.7.1) also belong to the same
formal system in Ngiyambaa. Imperative is not, then, formally, a member of
a de
ntic system, or even, more strictly, of a purely modal system, but a
member of a mixed system.
similarly, in many of the Native American Indian languages, suffixes that
Imperative and Jussive are often members of a larger system. There are
Sets o f s
uch suffixes in Tonkawa (Texas - Hoijer 1931: 83-94) and Maidu
Shipley 1964; 46-51), both quoted by Mithun (1999: 171), in which both
perative and Jussive ('hortative', 'exhortative') appear. For Maidu,
P eY has the following tense-'mode' paradigms:
present-past indicative, future indicative, habitual past indicative,
past punctual indicative;
81
Modal systems: Event modality

subjunctive;
monitive optative ('might'), intentive optative ('going to'),
hortatory optative ('let');
interrogative;
imperative I and imperative II.

Examples from the system for Tonkawa are:

Declarative naxadj ganaw-o-'o 'I married'
Assertive do.na-na'a 'He lies'
Exclamatory 'awac'a'la hedoxa-giw 'The meat is all gone'
Interrogative yaxa-ga? 'Did you eat?'
Intentive henl-a'ha'a 'I shall catch him'
Imperative 'andjo-w 'Wake up'
Potential ya.dj-V-n'ec 'I might see him'
Exhortative hama'am-dox-a'dew-e'l 'Let him be burned up'

(The terms 'exclamatory' and 'exhortative' are offered by Mithun 1999: 171.)
There are two kinds of imperative in a number of North American
languages (Mithun 1999: 171). In Maidu (North California - Shipley 1964:
51-2), 'Imperative I' 'is used when the action of the order is to be carried out
in the presence of the speaker or when there is no interest in the place of the
ordered action', while 'Imperative IF is used 'when the ordered action is to be
carried out in the absence of the speaker':

Imperative I slpi
'Sing!'

Imperative II mymk plkydi dkpajtipad

'Stick it in his door!'

In Cheyenne (Algonquian, Montana, Leman 1980: 41) the difference is

between commands to be carried out immediately and those to be carried out
later:

meseestse 'Eat!'
msheo?o 'Eat!' (later)

This is true also of Takelma (Sapir 1922: 94), and for a number of other
languages (Bybee 1985: 171). Also, there are often two kinds of command in
languages with mood systems, but differing in terms of politeness - see 5.
and 6.7.2.
Maidu and Cheyenne also have jussives:
'Hortative' mseheha 'Let him eat!'
'Hortative' ?yk'jts 'Let me go!'

82
 3.5 'Purposive' in Australian languages

3 5 'Purposive' in Australian languages

category identified as 'purposive' in Australian languages expresses
ation (and epistemic necessity) in main clauses (see 1.4.5). An example
m Ngiyambaa (Donaldson 1980: 162) is:
rjadhu bawun-ga yuwa-giri
1+NOM middle-LOC lie-PURP
'I must lie in the middle'
H wever, a 'purposive' may also be used in a main clause to suggest a result
f offl an unknown cause, as in Dyirbal (Dixon 1972: 69):
balan ugumbil bangul yatangu balgali
CL+NOM woman+NOM CL+ERG man+ERG hit+puRP
'Something happened to enable or force the man to hit the woman'
In subordinate clauses the same marker is used for both purpose and
result. Thus in Yidiny (Dixon 1977: 345-6), the purposive in a subordinate
clause usually indicates purpose:
d_ada 4ud.u:mbu gar.bagarba:alrju nud_u wawa:lna
child+ABS aunt+ERG hide+PAST not see+puRP
'Auntie hid the child so that it should not be seen'
But it can also be used to express 'a natural result':
nayu burawurjal duga:l djnbid_inbi:lna
1 +SUB Burawugal+ABs grab+PRES struggle+RED+puRP
I grabbed the water sprite woman and as a result she kicked and
struggled
The same suffix is also used in some Australian languages to express
indirect commands, as in in Ngiyambaa (Donaldson 1980: 280):
nadhu-na rjiyiyi girma-1 rjinu:
I+NOM-3ABS say+PAST wake-PURP you+OBL
'I told her to wake you'
Typologically this is of interest because these functions of the 'purposive'
are very like those of the subjunctive in Latin, which can also be used for
Purpose, result and for indirect commands (see 5.3.2, 5.3.4, 5.4.2). Notionally
se
different uses can, perhaps, be seen as related in terms of showing the
ec
t of some cause - an intended effect with purpose and with indirect
or
nmands, an actual effect with results. The similarity between the obli-
1Q
n and result interpretations in main clauses (not paralleled in Latin) can,
whaps, be shown by different possible translations - 'The man had to hit
e
woman' and 'It was inevitable that the man hit the woman.' There is a
se
for the action, either the obligation imposed on the man or some
Un
known or unstated factor.

83
Modal systems: Event modality

Similarly for Dyirbal, Dixon (1972: 689) speaks of 'implicated verb

complexes' which are again marked with the purposive inflection and indicate
either an intended action or a natural consequence (both in relation to a
previous event):
naila cUngalijiu biligu
1+SUB run+PRES/pAST climb+PURP
Tm running (to a tree) to climb it'
bayi yar.a wayjiklm yalu
CL+NOM man+NOM come.uphill+PRES/PAST towards.here
barjgun dundungu man^ali
CL+ERG bird+ERG point.out+PURp
'The man came uphill towards here, resulting in the bird's pointing out
his presence'
Dixon (1977: 346) points out (for Yidiny) that the semantic distinction
between purpose and result may be neutralized in the negative:
nayu gungagunga:r. gali:na garu nanan
1+SUBJ north+RED go+puRP by and by 1+OBJ
namu:ray nudu jiumadna
smell+ABS not smell+PURP
'I must go by the north so that she will not smell me'
Moreover, there are examples in which it is not possible to decide whether
the purpose or result sense is intended, or whether there is simply indetermi-
nacy between the two possible meanings, as in Dyirbal (Dixon 1972: 68):
balan 4ugumbil bangul yajangu balgan bacjjgu
CL+NOM woman+NOM CL+ERG man+ERG hit+PREs/past fall+PURp
'Man hits woman causing her to fall down'

3.6 An alternative analysis

Bybee (1995: 6; cf. 1985: 166) argues that 'event modality should be divided
into speaker-oriented and agent-oriented modality'. 'Agent-oriented modal-
ity' she says 'encompasses all modal meanings that predicate conditions on
an agent with regard to the completion of an action referred to in the main
predicate, e.g., obligation, desire, ability, permission and root possibility-
She also says, 'markers of directives, such as imperatives, optatives o r
permissives, by which a speaker attempts to move an addressee to action, ar
called speaker-oriented'. In particular, it should be noticed that age 11 '
oriented modals include deontic statements that describe obligations an
permission.
There is a slightly strange dichotomy, for there is considerable differeI1
between (i) permission and obligation, for which the conditions factors

84
 3.6 An alternative analysis
ternal to the agent and (ii) desire and ability for which the factors are
nerally internal to the agent. This is the basis of the deontic/dynamic
,. tjnCtion. Moreover, permission and obligation (especially permission)
Aen stem from the speaker and are thus more speaker-oriented than agent-
riented. Indeed, it is rather difficult to understand what is the difference in
Bvbee's analysis between (agent-oriented) 'permission' and (speaker-or-
iented) 'permissive'. In view of this, the deontic/dynamic dichotomy seems
more useful than that of agent-oriented and speaker-oriented.

85
 4
Modal systems and modal verbs

The modal systems described in the last two chapters share a number of
features, not only in the systems themselves, but also, for many languages in
the use of modal verbs and the association with possibility and necessity. A
detailed discussion of these has been left to this chapter in order to avoid too
much repetition and cross-referencing. These issues hardly affect evidentials
so that the discussion is almost entirely concerned with epistemic, deontic
and dynamic modality.

4.1 Modal systems

4.1.1 Formal identity of different systems

Notionally, epistemic modality and deontic/dynamic modality might seem to
have little in common. As suggested in 1.2.2, epistemic modality is concerned
solely with the speaker's attitude to the truth value or factual status of the
proposition (propositional modality), whereas deontic and dynamic modality
refer to events that are not actualized, events that have not taken place but
are merely potential (event modality). Yet in English (and many other
languages) the same forms are used for both types. The following, for
instance, can all be interpreted either epistemically or deontically:
He may come tomorrow
The book should be on the shelf
He must be in his office
Where the same verbs are used for the different types of modality, there
are often slight differences in the forms, which will be summarized in 4.2.3-
Such differences suggest that the types are grammatically as well as notion-
ally different, but do not explain why the forms are basically the same. One
example of a difference is to be found in Modern Greek, where an impersonal
form is used for epistemic possibility, but a personal form of the verO
(agreeing with the subject) is used for (deontic) permission or (dynamlC'
ability, as in:
86
 4.1 Modal systems

ta pej bo na fiyun vrio

the children BORO+3SG+PRES+IMPFV that Ieave+3PL+PRES+PERFV tomorrow
'The children may leave tomorrow'
ta pej born na fiyun vrio
the children BORO+3PL+PRES+IMPFV that leave+3PL+PRES+PERFV tomorrow
'The children may/can leave tomorrow'

the firSL' t n e m d a l verb (bori) is singular and does not agree with the
lural subject, and so is to be interpreted epistemically. In the second the
erb (born) is plural and agrees with the subject and so is to be interpreted
deontically (permission) or dynamically (ability).
The use of the same form for both propositional and event modality is not
a purely European phenomenon. The same is true of the Dravidian language
Tamil (Asher 1982: 171), where two suffixes that are identified as 'permission'
and 'debitive' are used for both epistemic and deontic modality:

Kantacaami vantaalum vara-laam

Kandaswami come+coNCESS come-PERM
'Kandaswami may perhaps come'
veen.-um-n.n.aakkaa naalekki avan peeca-laam
want-coND tomorrow he speak-PERM
'If he wants, he can speak tomorrow'
Gan_eecan ippa Mannaarku(;iyile irukka-n.um
Ganesan now Mannargudi+Loc be-DEB
'Ganesan must be in Mannargudi now'
avan anke pooka-n_um
he there go-DEB
'He must go there'

There are other non-Indo-European languages in which a single form may

be used either deontically or epistemically, as has already been partially
noted (1.4.5, 3.2.2 and 3.5) for Ngiyambaa (Donaldson 1980: 160-2):

yururj-gu rjidja-1-i
rain-ERG rain-CM-puRP
'It is bound to rain'
rjadhu bawurj-ga yuwa-giri
I+NOM middle-Loc lie-PURP
T must lie in the middle'
yururju nidja-1-aga
rain-ERG rain-CM-iRR
'It might/will rain'

87
Modal systems and modal verbs
waga:y-ndu-gal dhagurma-gu yana-y-aga
NEG-2NOM-PL cemetery-DAT go-CM-iRR
'You shall none of you go to the cemetery'
(But see 3.2.2 for comments on the second pair of examples.)
In some languages, there are pairs for 'may' only. Thus in Tutatulabal
(Steele 1975: 207, quoting Voegelin) the 'permissive' suffix is used in the two
senses, e.g.:
hatdaiw-aha-bi
cross-PERM-SM
'You may cross it'
wi:-aha-dza
run-PERM-SM
'It might run'
In colloquial Cairene Arabic (Gary and Gamal-Eldin 1982: 98-9) the
same form is used for both kinds of 'must', but there are different forms for
'may':
laazim tiXallas bukra
must you+SG-finish tomorrow
'You must finish tomorrow'
laazim jikuun hinaak
must he.be there
'He must be there'
ti?dari tifuuti min hina
you+SG.can/may you+SG.pass from here
'You can/may pass through here'
jimkin jikuun hinaak
probable/possible he.be there
'He may be there'
Bybee et al. (1994: 195) also provide an example from Abkhaz for 'must'
and from Lao for 'may/can'. They also note that there is a similar epistemic-
deontic pairing for a form expressing both weaker obligation and a weaker
judgment comparable to English should (see 2.1.5 and 3.2.3) in Baluchi. The
form (marked 'subjunctive') may be a matter of mood rather than of modal
system, but still illustrates the relation between deontic and epistemic:
a ea bbart
3SG 3PL take.away +3SG+SUBJ
'He ought to take them away'
ma bskly adda kassa byzana
IPL perhaps there someone know+lPL+suBjl
'Perhaps we know someone there'

88
 4.1 Modal systems

The examples here show the same forms being used for epistemic and
, o n t j c modality not only for both propositional and event modality, but
l o for both deontic and dynamic. Identical forms are also used for deontic
nd dynamic modality as in English:

He can come in now

He can run a mile in under four minutes

This is true of similar forms in other languages, e.g., of POTERE in Italian. It is

a lso true of Lisu (duscussed in 3.3.1):

sa nya ami khwa da-a

Asa TOP field hoe acceptable-DEC
'It is acceptable for Asa to hoe the field'
'Asa is (physically) able to hoe fields'

4.1.2 Possibility and necessity

There is, in fact, a simple explanation for why the same forms are used in
the different types of modality. The explanation is in terms of possibility and
necessity, which are, Lyons (1977: 787) says, 'the central notions of
traditional modal logic'. Indeed, the terms 'epistemic', 'deontic' and
'dynamic' are taken from a pioneering work on modal logic by Von Wright
(1951: 1-2,28).
Thus epistemic Speculative and Deductive can be interpreted in terms of
what is epistemically possible and what is epistemically necessary:

John may be in his office

= It is epistemically possible that John is in his office
John must be in his office
= It is epistemically necessary that John is in his office

Similarly, deontic Permissive and Obligative can be interpreted in terms of

what is deontically possible and deontically necessary:
You may/can go now
= It is deontically possible for you to go now
You must go now
= It is deontically necessary for you to go now

There is one reservation. The word necessary itself is not used in an

Pyemic sense in ordinary language (as opposed to logical terminology). It
uld not be normal to say *It is necessary that John is in his office, although
ls
possible to say, in semi-logical language, It is necessarily the case that
oli
n is in his office. There is no problem with the possible - It is possible that
olr
n is in his office is perfectly normal.

89
Modal systems and modal verbs I

Moreover, the epistemic and deontic uses can be distinguished in terms

'possible/necessary that' and 'possible/necessary':
It is possible that John is in his office
It is 'necessary' that John is in his office
It is possible for John to be in his office
It is necessary for John to be in his office
The first two are epistemic, the last two deontic (although the third might b
interpreted as epistemic also).
The importance of possibility and necessity in the modal systems is also
shown by the distribution of the verbs in terms of negation. However, the
situation with the forms used for negation of the modal verbs differs
considerably in the various languages of the world and will be examined in
greater detail in the next two sections.

4.1.3 Possibility, necessity and negation in English

There are two different ways in which a modal expression may be negated, as
can be seen in the modal forms for epistemic possibility in English:
Mary may be at school
Mary may not be at school
Mary can't be at school
With the use of the notions of possibility and necessity, it is easy to explain
the difference between the two types of negation. One is to be interpreted in
terms of 'possible not', the other in terms of 'not possible'. With 'possible
not' it may be said that there is negation of the proposition, whereas with
'not possible' there is negation of the modality.
It is possible that Mary is not at school ('may not', proposition negated)
It is not possible that Mary is at school ('can't', modality negated)
Formally, the distinction is made, it can be seen, by changing the verb for
'not possible' - using CAN in place of MAY.
A similar distinction can be made between 'necessary not' and 'not
necessary'. Yet there are no comparable forms of MUST that are used for the
negation of epistemic necessity. The only forms available are:
Necessary John must be in his office
Necessary not John can't be in his office
Not necessary John may not be in his office
What is immediately noticeable about these negative forms is that they are
the same as those used for epistemic possibility, but in the reverse order.
There is quite a simple explanation for this. It involves the logical relations
between possibility and necessity, in that 'not possible' is logically equivalent

90
r 4.1 Modal systems

essary not' and, conversely, 'not necessary' is equivalent to 'possible

t0
0 t \ in logical terms:
Not possible = Necessary not
Not necessary = Possible not
r r expressing 'necessary not', English uses the form for 'not possible'
't) and for 'not necessary' it uses the form for 'possible not' (may not).
Thus the forms of MAY and CAN provide all the negatives for epistemic
ibility and necessity. There is suppletion - the negative forms for
stemic necessity are provided, suppletively, by the forms for epistemic
ssibility. Lyons (1977: 802) remarks that 'in English at least, possibility
ther than necessity should be taken as primitive in the analysis of epistemic
modality'.
There are, then, two characteristics of these negative forms, the use of a
different verb (CAN for MAY) and suppletion. These two features are to be
found with the deontic modals too, but in a rather different way. The
possibility forms are illustrated by:
Possible Mary may/can come tomorrow
Possible not Mary needn't come tomorrow
Not possible Mary may not/can't come tomorrow
Examples of the necessity forms are:
Necessary Mary must come tomorrow
Necessary not Mary mustn't come tomorrow
Not necessary Mary needn't come tomorrow
It can be seen here that with deontic modality it is the necessity that is
'primitive' in Lyons's sense. For it is with the necessity forms that a different
verb is used (NEED in place of MUST) and for 'Possible not' the 'Not necessary'
form needn't is required.
There is no problem with dynamic modality. Normally only the modality
is negated:
He can't run a mile in four minutes (is unable)
He won't answer (is unwilling)
Bei
ng able or willing not to do something is more likely to be expressed by
abl
e not to and willing not to.
Ihere is, then, an important question: why is it that may not and can't have
1
terent interpretations? And the same question may be asked of mustn't and
e
"n t. Formally, they are clearly negations of modals, and, if the modality
re
formally associated with the modal, the expected interpretation of the
two would be 'not possible' and of the second two 'not necessary', with
a tion of the modality. The interpretation of may not as 'not possible' but

91
Moaal systems and modal verbs

of mustn't as 'necessary not' would suggest that the grammar and

notional interpretation are not in a one-to-one relationship, and that ^
sense, one of the negative markers is in the 'wrong' place. (If it is argu e c j., ''
the negation belongs with the lexical verb, the problem remains, but with
e
negative in the 'wrong' place for different forms.)
It should be said, that, although there does not appear to be a consist
over-all pattern for negation in English, some scholars have looked W
an
explanation. Cormack and Smith (forthcoming) argue that the 'split' betw
negation of the modality and negation of the proposition mainly (but n
entirely) corresponds to possibility/necessity, while Coates (1983: 237-Q\
argues that it depends on the deontic/epistemic distinction, with only deonti
mustn't being exceptional. In fact, the contrasts between epistemic may not
and can't and between deontic mustn't and needn't provide exceptions to
both views. The first view fails to account for may not (possibility with
negation of the proposition) and needn't (necessity with negation of the
modality). The second is contradicted not only by the 'exceptional' mustn't
(deontic with negation of the proposition - and the same is true of shouldn't),
but also by can't (epistemic with negation of the modality).
It will be helpful to look at other languages.

4.1.4 Possibility, necessity and negation in other languages

It might be expected that there are languages in which the negation of all the
relevant modals is regular in the sense that there is no suppletion in terms of
the logical equivalences and no use of alternative verbs. In fact, in an
investigation into over thirty languages and dialects (Palmer 1995, 1997),
none was discovered, the closest being Modern Greek in which only the
deontic 'not necessary' form was irregular. The forms were:
bori na ine sto y ratio tus
BORO+3SG+IMPFV that they .are in.the office theirs
'They may be in their office' (epistemic possibility)
den bori na ine sto yrafio tus
NEG BORO+3SG+IMPFV that they.are in.the office theirs
'They can't be in their office' (not possible)
bori na min ine sto yrafio tus
BORO+3SG+IMPFV that NEG they.are in.the office theirs
'They may not be in their office' (possible not)
prepi na ine sto yrafio tus
PREPI+3SG+IMPFV that they.are in.the office theirs
'They must be in their office' (epistemic necessity)

92
 T
4.1 Modal systems

den prepi na ine sto yrafio tus

NEG PREPI+3SG+IMPFV that they.are in.the office theirs
'They may not be in their office' (not necessary)
p r e pj na min ine sto yrafio tus
PREPI+3SG+IMPFV that NEG they.are in.the office theirs
'They can't be in their office' (necessary not)
boris na fiyis
BORO+2SG+IMPFV that you.leave
'You may leave' (deontic possibility)
en boris na fiyis
NEG BORO+2SG+IMPFV that you.leave
'You can't/may not leave' (not possible)
boris na mi fiyis
BORO+2SG+IMPFV that NEG you.leave
'You needn't leave' (possible not)
prepi na fiyis
PREPI+2SG+IMPFV that you.leave
'You must leave' (deontic necessity)
prepi na mi fiyis
PREPI+3SG+IMPFV that NEG you.leave
'You mustn't leave' (necessary not)
Only the deontic 'not necessary' form is missing from this paradigm. To
express that, there are two possible forms, both using different verbs: '
:
en ine anangi na fiyis
NEG is necessary that you.leave
'You needn't leave' (not necessary)
en xriaxete na fiyis
NEG it.is.needed that you.leave
'You needn't leave' (not necessary)
Where there is irregularity in the sense being discussed here, there are
several possibilities, of which three are the most important:

(i) the use of a different verb

(ii) logical suppletion
(iii) the positioning ('misplacement') of the negative marker.

these will be considered in turn.

The use of a different verb has already been illustrated for English and
Modern Greek. It is also found in German (as well as other languages):
Du magst herein kommen
you MGEN+2SG+PRES+IND here come
'You may/can come in'

93
 r
Modal systems and modal verbs
Du darfst nicht herein kommen
you DURFEN+2sG+PRES+iND not here come
'You can't/may not come in'
Du kannst nicht herein kommen
you K6NNEN+2SG+PRES+IND not here come
'You can't/may not come in'
Where there is logical suppletion, it is of two types. In the first type, the
possibility forms are used suppletively for the necessity forms. This has
already been illustrated for epistemic modals in English, but an unrelated
language for which there is a similar situation is Assamese (personal
communication from Jiyoti Tamuli). The distinction between 'possible not'
and 'not possible' is made by the position of the negative marker:
ofisot thak-ibo pare
office be-iNF he.can
'He may be in his office'
ofisot na t^ak-ibo pare
office not be-iNF he.can
'He may not be in his office'
ofisot ttiak-ibo ncoare
office be-iNF NEG+he.can
'He can't be in his office'
For the negatives of necessity, however, the possibility forms >iire used
suppletively, as in English:
ofisot t^ak-ibo lage
office be-iNF he.must
'He must be in his office'
ofisot thak-ibo ncoare
office be-iNF NEG+he.can
'He can't be in his office'
ofisot na thak-ibo pare
office not be-iNF he.can
'He may not be in his office'
With the second type of suppletion, it is the forms that negate the modal
('not possible' and 'not necessary') that are used suppletively for the forms
that negate the proposition, 'possible not' and 'necessary not'. This can be
illustrated from Danish epistemic modals, where there are two forms only
(Davidsen-Nielsen 1990: 78-87):
Det kan ikke vaere sandt
that can not be true
'That can't be true'
94
 4.1 Modal systems

Det beh0ver ikke vaere sandt

that must not be true
'That may not be true'
need a little explanation. In Danish (and most Germanic languages)
negative normally occurs after the verb, so that it is the modal that is
Hy negated in the examples above, and the expected interpretation
uld be 'not possible' and 'not necessary', with negation of the modality.
p r negation of the proposition ('possible not' and 'necessary not') there is
inpletion, the first example expressing 'necessary not' and the second
' ssible n o t ' Biit; as can be seen, suppletion is not in terms of possibility
forms being used for necessity forms, as in English, but of forms negating the
modal being used for forms negating the proposition.
This feature is also to be found in the deontic modals of a formal variety
of Welsh, where the two forms are:
(Ni) gewch chi ddim ddod yfory
(not) get you not come tomorrow
'You can't come tomorrow'
('d oes) ddim rhaid i chi ddod yfory
(It is not) not necessary to you come tomorrow
'You needn't come tomorrow'
These are literally 'not possible' and 'not necessary', but are also used to
express 'necessary not' and 'possible not'.
The third possibility is that the negation is, or appears to be, misplaced in
the sense the grammatical marking and the notional interpretation do not
correspond. To make this point quite clear, the Modern Greek examples
above may be considered again. Because they involve subordination with the
conjunction na, it can be seen that the negative may occur either in the main
clause or in the subordinate clause. Moreover, when the negative occurs in
Ae main clause (which contains the modal) the interpretation is 'not possible'
or
not necessary', and, when it occurs in the subordinate clause (which
contains the lexical verb) the interpretation is 'possible not' or 'necessary
n
t This is what might be expected - negation of the modal verb indicates
negation of the modality.
This is so also in Catalan, according to Picallo (1990: 287):
En Jordi pot no haver sortit
the Jordi can not have+iNHN leave+PART
'It is possible that Jordi hasn't left'
En Jordi no ha pogut sortir
the Jordi not has can+PART leave+iNFiN
'Jordi hasn't been able to leave'

95
Modal systems and modal verbs

However, here, as the translations show, there is the further point th-
negation of the main verb is always associated with propositional (episterni
modality and negation of the modal verb with event (deontic/dynamioi
modality.
To avoid any possibility of ambiguity here, two distinctions are needed
the grammatical distinction between the modal and lexical verb and th
notional distinction between the modality and the proposition. In a regular
system such as that of Modern Greek (for the most part), (grammatical*
negation of the modal indicates (notional) negation of the modality and
(grammatical) negation of the lexical verb indicates (notional) negation of
the proposition.
There are many languages, however, in which there is irregularity in this
sense. Most commonly it occurs where there is grammatical negation of the
deontic necessity modal. Consider the following examples from Kinyarwanda
and Arabic:
agomba kwinjira
he.must to.come.in
'He must come in'
ntagomba kwinjira
NEG+he must to.come.in
'He mustn't come in
la:zim jizi
must you.come.in
'You must come in'
ma:/mu laizim jizi
NEG must you come in
'You mustn't come in'
In both sets of examples, the modal verb is formally negated, in Kinyar-
wanda by its morphology, in Arabic by the fact that the negative precedes it
Yet, notionally, it is the proposition, not the modality, that is negated. The
interpretation is 'necessary not', not 'not necessary'. There is, it may be said,
'misplacement' or, better, 'reversal' of the marking of the negative.
The same is true of French and Italian:
il faut partir
it is.necessary to.go
'We must leave'
il ne faut pas partir
it NEG is NEG necessary to go
'We mustn't go'

96
 4.1 Modal systems

Deve venire
dovere+3sG+PRES+iND to.come
'You (he/she) must come'
Non deve venire
not dovere+3sG+PRES+iND to.come
'You (he/she) mustn't come'
/The Italian form is, in fact, ambiguous in that it may also have the regular
nterpretation, with no displacement, of 'You needn't come', but it is still the
only normal way of expressing 'necessary not'.)
The situation in the Germanic languages is not completely certain since the
negative can usually occur in only one position - between the modal and the
lexical verb. Nevertheless, it is usually the case that the negative follows the
verb that it (notionally) negates. The regular interpretation of a Germanic
form with a negative would, therefore, be that it is the modality that is
negated.
On this interpretation the German deontic necessity form is regular and
that of Norwegian is irregular. Within German negated MSSEN expresses
'not necessary' and DRFEN is required suppletively for 'necessary not' (see
above):
Du musst nicht herein kommen
you MSSEN+2SG+PRES+IND not here come
'You needn't come in'
Du darfst nicht herein kommen
you DRFEN+2SG+PRES+IND not here come
'You may not come in'
In Norwegian, by contrast the negation of the necessity verb indicates
'necessary not' and a non-modal form is used for 'not necessary':
Han m ikke forlate rommet
he must not leave the.room
'He mustn't leave the room'
Han trenger ikke forlate rommet
he must not leave the.room
'He needn't leave the room'

Grammatically the negative belongs with the modal. Notionally, it is the

dality that is negated in German, but the proposition in Norwegian.
English is also irregular in this sense with:
You mustn't come
TV '
ls
!s, perhaps, even more striking, because the negative is cliticised on the
97
Modal systems and modal verbs

modal, and might thus seem to suggest even more strongly (but wrongly) t ^
it is the modality that is negated.
However, this irregular displacement of the negative is not confined
deontic possibility. Consider the two possible interpretations of English-
Mary may not come tomorrow
In the epistemic sense it is regular ('possible not'), but in the deontic sens
(now very formal and rather unusual), it is irregular ('not possible' refusin
permission).
This reversal of negative marking with deontic necessity is widespread in
the languages of the world. It occurred in ten of the twenty languages
investigated in Palmer 1995. (The languages investigated in Palmer 1996 were
all Germanic, and reversal was much rarer.) It is a puzzle why this should be
There may, perhaps, be two reasons - that there is comparatively little need
of the 'not necessary' form and that placing the negation early in the sentence
emphasizes the prohibition. Thus 'necessary not' usurps the form more
naturally associated with 'not necessary'.
For a detailed study of the relationship between modality and negation,
see D e H a a n (1997).

4.1.5 The notional connection

The close relationship between epistemic and deontic modals has, of
course, been noted before. Joos (1964: 195) comments 'within the modal
system English does not distinguish between duty and logic'. But it has
seldom been noted that there is no immediately obvious reason why the
same forms should be used for expressing the speaker's degree of commit-
ment to truth and for getting other people to do things. It is by no means
obvious that permission is a 'related notion' to possibility, or requirement
to certainty, as Steele, Alemajias and Wasow (1981: 21) comment without
justification or explanation.
The most detailed and plausible explanation is found in Sweetser (1982
and 1990) who argues (1982: 492) that the 'epistemic world is understood in
terms of the sociophysical world', an idea that is virtually identical with that
of locative case theory (Anderson 1971). Sweetser suggests that while MAY IS
'an absent potential barrier in the sociophysical', epistemic MAY is the
'parallel case in the world of reasoning' and that its meaning would be that
'there is no barrier to the speaker's process of reasoning from the available
premises to the conclusion expressed'. This idea of 'no barrier' had also been
anticipated by Ehrman (1966) who interprets it as nihil obstt. Simile
arguments hold for MUST.

98
 4.1 Modal systems

. w o u l d be no more than plausible speculation if it were not the case

Sweetser adduces evidence from other forms in English. She finds a
' lar ambiguity in verbs such as INSIST, SUGGEST, EXPECT:

I insist that you go to London

I insist that you did go to London
I suggest that you leave the room now
I suggest that you left the room to avoid being seen
The following sentence is ambiguous between the two senses:
I expect him to be there
To these observations can be added the fact that verbs such as ask, promise
and swear can be used epistemically or deontically:
I asked if he had come
I asked him to come
I promise you he's here
I promise you he'll come
I swear he's here
I swear I'll give it to you

Sweetser even notes a somewhat similar ambiguity in causal conjunction:

He came because he heard me screaming
He heard me screaming, so he came
(You say he's deaf, but) he came, so he heard me screaming , ,
(You say he's deaf, but) he heard me screaming, because he came
The first two are concerned with the reasons for the action, the third and
fourth with the reasons for the judgment. Similar arguments can be found for
therefore, since, although, despite and anyway.
The same is true of other languages. For example, INSISTIR in Spanish is
used in both senses, though the indicative in the subordinate clause indicates
the epistemic use and the subjunctive indicates the deontic sense (Klein 1975:
356, who does not specifically note the epistemic/deontic distinction):
Insisto que aprende
I.insist that learn+3sG+PRES+iND
'I insist that he's learning' /
/
Insisto que aprenda
I.insist that learn+3so+pRES+suBj
'I insist that he learn'
It may be added that it has been argued, quite convincingly, that epistemic
Ses
have developed, diachronically, from dynamic/deontic uses (see Trau-
gott 1989).

99
Modal systems and modal verbs

4.2 Modal verbs

Modal verbs are used in all four of the main types of modality that are founri
in modal systems -judgments, evidentials, deontic and dynamic (though to
very limited degree with evidentials).

4.2.1 Modal verbs in English

It is, perhaps, fortunate for a study of modal verbs in which there k
considerable discussion of English that there is no doubt that English has a
set of modal verbs that can be formally defined. They are MAY, CAN, MUST
OUGHT (TO), WILL and SHALL, and marginally, NEED and DARE (including
might, could, would and should). The facts have been stated many times (e.s
Palmer 1979, 1990) and will only be briefly summarized here.
First, they are members of a larger set of auxiliary verbs, which exhibit
what Huddleston (1976: 333) has called the NICE properties - their
occurrence with negation, inversion, 'code' and emphatic affirmation (see
Palmer 1987: 14-21), as in (see 2.6.1):
I can't go
Must I come?
He can swim and so can she
He will be there
These properties the modal verbs share with the other auxiliary verbs BE and
HAVE but, in addition, they have formal features of their own:

(i) They do not co-occur: there is no *will can come, *may shall be,
etc. (though in a few dialects there are some very restricted
possibilities of co-occurrence such as might could- see Brown
1991 and comments by Cormack and Smith forthcoming).
(ii) They have no -s forms for their third person singular: e.g. *He
oughts to come. The form wills exists, but as a form of the lexical
verb, as does cans, in entirely different senses. (DARE has forms
both with and without -s: He dares to come, Dare he cornel This
suggests that it functions both as a lexical verb and as a modal.)
(iii) They have no non-finite forms: e.g. no *to can or *canning; there
is no */ hope to can come tomorrow.
(iv) They have no imperatives: *Can be here! *Must come now!
(v) MUST has no morphologically past tense form, although the
others do (could, should, might, etc.); of those forms, only could
is used to refer to past time (though all may occur in reported
speech).
(vi) There are suppletive negative forms (4.1.3).
100
 4.2 Modal verbs

(vii) There are formal differences between the modal verbs, in their
epistemic and deontic senses, in terms of negation and tense
(4.2.3).

4.2.2 Modal verbs in other languages

j ns tructive to compare the English verbs with similar verbs in German
A French. German certainly has six potential candidates for the title of
odal verb: KONNEN, DRFEN, MOGEN, MSSEN, SOLLEN and WOLLEN, obvious
nates 0 f the English ones (though, unlike the English modals, they have
nfinitive forms and fewer restrictions on their use). These are used both
epistemically and deontically, epistemic MAY being translated by either
KONNEN or MOGEN, and deontic MAY by either KNNEN or DRFEN, while both
epistemic and deontic MUST are translated by MUSSEN (see Hammer 1983:
223ff- for examples):
Er kann/mag krank sein
he KONNEN/MOGEN+3SG+PRES+IND ill be
'He may be ill'
Du kannst/darfst den Bleistift behalten
thou KONNEN/DURFEN+2SG+PRES+IND the pencil keep
'You can keep the pencil'
These are less clearly identified as members of a grammatical system than the
English modals, but they have some idiosyncrasies, notably:

(i) There is no final -/ in the third person singular of the present

indicative: e.g. kann, mag (cf. gibt 'gives').
(ii) In subordinate clauses the order of the elements of the verbal
complex is changed if one of them is a modal verb in the
infinitive occurring with the finite form of another auxiliary verb
(as it cannot occur in English). The usual rule is that the finite
form occurs last, but in these circumstances it occurs at the
beginning of the verbal complex (Hammer 1983: 224):
Es war klar, dass er sich wiirde anstrengen mussen
It was clear, that he himself be+3sG+iMPF+suBj exert must+iNF
'It was clear that he will have to exert himself
(Note that the word order is not *anstrengen mussen wiirde.) There are
similar modal auxiliaries in the Scandinavian languages, but they 'differ in
complex ways from language to language, making translation difficult'
(Haugen 1976: 80).
There are similar verbs in the Romance languages - French POUVOIR and
EV
OIR, Italian POTERE and DOVERE, Spanish PODER and DEBER. The situation

101
Modal systems and modal verbs

in spoken French and Spanish is complicated by the use of impersonal //

peut que and puede que for epistemic possibility and // faut for deon
necessity in French, but the Italian verbs are used in both epistemic ann
deontic senses:
pu essere nell' ufficio
POTERE+3SG+PRES+IND be in.the office
'He may be in the office'
deve essere nell' ufficio
DOVERE+3SG+PRES+IND be in.the office
'He must be in the office'
pu venire domani
PODERE+SSG+PRES+IND come tomorrow
'He may/can come tomorrow'
deve venire domani
DOVERE+3SG+PRES+IND come tomorrow
'He must come tomorrow'
There appear to be no clear grammatical indications that the modal verbs in
the Romance languages are a clearly defined set. (For French see Ruwet
(1967: 185ff.), though there is one marginal feature noted by Kayne 1975:
1-27.) However, the verbs of Italian and Spanish are of typological interest
in terms of their functions with negation.
The use of auxiliary verbs to express modality is very characteristic of
European languages, but not exclusive to them. Thus Mandarin Chinese (Li
and Thompson 1981: 173-88) has a set of auxiliary verbs that can be
formally defined, in that:

(i) They must co-occur with a lexical verb (or, at least, one must be
understood from the context),
(ii) They have no aspect markers or intensifiers.
(iii) They cannot precede the subject,
(iv) They cannot have objects.

All of these are semantically modal; the list is:

ylnggai, ylngdang, gai 'ought to, should'

nng, nnggu, hui, kyi 'be able to'
nng, kyi 'have permission'
gn 'dare'
kn 'be willing to'
di, bixii, biyo, bdi 'must, ought to'
hui 'will, know how'
102
 4.2 Modal verbs

Although the verbs of Chinese have no inflections, these auxiliaries appear

he essentially verbal. There is also a set of auxiliary verbs in Cashibo (Peru
shell 1975: 178-91) that are inflectional and marked for person-number
h the stems may combine), but which may, perhaps, belong more to a
Hscourse system than a modal system (see 2.6.1).

4.2.3 Formal differences

Although the same modal verbs may be used in English for both epistemic
nd deontic modality, generally the distinction is quite clear, and it can be
een from the previous discussion that there are also some formal distinctions
between the modals in their two uses:

(i) Deontic MUST has negative mustn't and a suppletive needn't, but
epistemic MUST has no morphologically related negative,
(ii) May not negates the modality when deontic (no permission), but
the proposition when epistemic ('It may be that it is not so'),
(iii) MAY and MUST followed by have are always epistemic, never
deontic.
(iv) MAY is replaceable by CAN only in the deontic use, though can't
may be epistemic.
Furthermore, when it refers to the future, MUST is almost always deontic; the
epistemic sense is provided by BE BOUND TO (see Palmer 1979: 45-6, and, for
discussion, Coates 1983: 4 2 - 3 , and Palmer 1983: 291). Moreover, might is
closely related in its meaning to present tense may only in its epistemic sense.
In Modern Greek, similarly, although the same forms are used for deontic
and epistemic modality, there are differences in the syntax. First, as was seen
n 4.1.1 and 4.1.4, forms of the verb BORO are used for possibility, but in the
epistemic sense an impersonal (third person singular) form is found, while in
the deontic (and dynamic) there is a fully inflected form that agrees with the
subject.
Moreover, in many languages, if the lexical verb is marked as past, e.g., as
m English with may have and must have above, only an epistemic interpret-
a
tion is possible. Examples from Modern Greek are:

o janis prepi na fiyi avrio

the John PREPI that leave+3sc+PRES+PERFV tomorrow
'John must leave tomorrow'

o janis prepi na efiye

the John PREPI that leave+3sG+PAST
'John must have left'

103
Modal systems and modal verbs
4.3 Modal systems and other categories

4.3.1 Mood
The decision to treat modal systems and mood in separate parts of this boot
rests largely on the fact that most languages can be characterized as havin
one or the other. However, some languages have both, in two possible wavs
First, in a few languages that have a mood system with realis and irrealj
markers (Chapter 6), the categories associated with irrealis may form what
clearly a modal system. Examples are Hixkaryana and Serrano, in which the
categories marked as irrealis clearly belong to a familiar evidential system
This is discussed in some detail in 6.5.4. Rather differently, Central Pomo has
a system of evidentials that is independent of its mood system (see 2.2.1 and
6.3.2).
Secondly, the Romance languages have a system of mood, marked by
indicative and subjunctive, but also have a set of modal verbs. However, as
noted in 4.2.2, the modal verbs have not been fully grammaticalized, and in
French and Italian, at least, the subjunctive appears to be losing ground - it
is used much less in colloquial French and has largely disappeared from the
dialects of Southern Italy. Moreover, in English, the appearance of the
modal system (of modal verbs) has been accompanied by the disappearance
of Anglo-Saxon mood (see Lightfoot 1979, Plank 1984, who differ in their
views about the way in which this development took place). There are, then,
reasons for suggesting that, in general, the two are not likely to co-exist, or
that, if they do, one will, in time, replace the other.

4.3.2 Future
WILL and SHALL are formally modal verbs by the criteria proposed in 4.2.1,
yet they are often used to refer to future time (though SHALL is now much less
common, especially in American English, in this sense). Indeed, traditional
grammar books treated them as markers of a 'future tense' in English.
However, as first argued by Fries (1927), it is reasonable to argue that
English does not have a future tense, this being an idea carried over from
Latin grammar. There are several arguments for this, but one important one
is that, quite simply, they do not often indicate pure futurity, but are usually
associated with conditional futures. Indeed BE GOING TO is a better candidate
for the 'future tense' in English. (See Palmer 1990: 160-1, and the very
persuasive arguments of Huddleston 1995.)
It is not surprising that modal verbs should have future time reference-
The future is not fully known and it is always no more than a reasonable
104
 4.3 Modal systems and other categories

motion that a future event will ensue. Indeed, Lyons (1977: 677, 816)

says:
Futurity is never a purely temporal concept; it necessarily includes an
element of prediction or some related notion.
What is conventionally used as a future tense . . . is rarely, if ever,
used solely for making statements or predictions, or posing or asking
factual questions, about the future. It is also used in a wider or narrower
range of non-factive utterances, involving supposition, inference, wish,
intention and desire.

It may be added that even languages that have future tenses that are not
formally modal, but belong within the inflectional system, often use these
tenses for similar purposes. They are used in an assumptive sense, like
English WILL, in, e.g., French and Italian (Lyons (1968: 310; Lepschy and
Lepschy 1977: 139):

Ca sera le facteur
that be+3sG+FUT+iND the postman ' '
'That'll be the postman' (epistemic)

Suonano, sar Ugo

they.are.ringing be+3sc+FUT+iND Ugo
'The bell's gone; it'll be Ugo'

In colloquial Spanish, moreover, the future tense is normally used not to

express future time, but in this epistemic sense. Future time is signalled by the
verb IR ('to go') plus the infinitive.
Similarly, although English uses SHALL with second and third person forms
as a Commissive (and WILL can be interpreted in the same way, though not
as unambiguously as SHALL), this is true also of the French future tense:

You shall have it tomorrow

You'll have it tomorrow
Vous l'aurez demain

The connection between future and modality can also be shown histori-
cally. There are also plenty of examples of future tenses that are historically
erived from subjunctives; this is true of some Latin forms (see Handford
y
47: 15). Other languages have future tenses that have their origins in a
m
dal-type auxiliary. Thus in Modern Greek, tha is a reflex of THELO: T
lsn
, and the marker of future in Swahili is derived from (ku)-taka '(to)
ls
h , while the futures of modern Romance languages were derived from the
mfinitive plus HABEO T have',
finally, it will be seen in the next two chapters that future time is signalled

105
 "!

Modal systems and modal verbs

by mood, both by the subjunctive and by irrealis. This too illustrates th

potentially modal characteristics of future time reference.

4.3.3 Negation
The most important relationship between modality and negation is the on
that involves possibility and necessity. But there are a few other points t
note.
First, negation (usually together with interrogation) is involved in both
modal systems (2.3) and mood (5.2.3 and 6.6.4).
Secondly, some languages have what might be thought to be negative
modal verbs. Thus Latin may use the verb NOLO T refuse' in negative
commands (as well as ne plus the subjunctive - see 5.4.2), as in:
Nolite facere
refuse+2PL+iMP do
'Don't do it'
Welsh has a verb PEIDIO T cease' that is used in a similar way:
peidiwch gadael iddo fynd
cease+2PL+iMP from let him go
'Don't let him go'

106
 5
Indicative and subjunctive

it was noted in 1.2.1 that many languages have the grammatical category of
mood, and that in European languages particularly, but not exclusively, this
is dealt with in terms of the distinction between the indicative and the
subjunctive. An example from Spanish is (Klein 1975: 356):
Insisto que aprende
I.insist that learn+3so+PRES+iND
'I insist that he is learning'
Insisto que aprenda
I.insist that Iearn+3SG+PRES+SUBJ
'I insist that he learn'
It was also noted that other languages, particularly the languages of the
Americas and of Papua New Guinea, have markers that are labelled 'realis'
and 'irrealis'.
It was suggested that basically there is no typological difference between
indicative/subjunctive and realis/irrealis, and that both are instances of mood
(Realis/Irrealis). There are, however, considerable differences between the
functions of what have been labelled 'subjunctive' and 'irrealis', and,
therefore, for practical reasons they will be dealt with in separate chapters (5
an
d 6). A further chapter (7) will discuss the similarities and differences
between them.
Most of the examples in this chapter will be taken from the classical
an
guages, Latin and Greek and the Romance languages, especially Spanish,
lnce
there is a large body of literature concerning the subjunctive in that
n
guage. Some mention is made of German, which has considerable uses of
e
subjunctive in its literary form, although in other Germanic languages,
8- Danish, the use of the subjunctive is almost dead. However, there are
er
languages that have been usefully and plausibly described in terms of
ICat
. ive and subjunctive. In some of the Bantu languages of Africa, for
an
ce, the basic structure of verbs consists of a sequence of markers for
107
Indicative and subjunctive

subject, tense, verb stem and mood, indicative being marked by final -a a n H
subjunctive by final -e (though no tense distinctions occur with subjunctive
Examples from Swahili (Steere 1943: 43, 57) are:
n-me-pend-a
I-PERF-love-IND
'I have loved'
ni-pend-e
I-love-suBJ
'Let me love'

Examples from another Bantu language, Luvale (Zimbabwe and Zaire -

Horton 1949) and from the non-Bantu, West African, language Fula (Arnott
1970: 299ff.) are to be found later in this chapter (5.1.1 and 5.4.2). There will
also be discussion of some Semitic languages, particular the Ethiopian
languages, but with these there is a problem concerning the status of the
subjunctive (see 5.4.2).

5.1 Main and subordinate clauses

Jespersen (1924: 314) noted that one of the functions of the subjunctive is
simply that of being subordinate, in that it is typically the mood used in
subordinate clauses. It is, in fact, no coincidence that the term 'subjunctive' is
a translation of the Classical Greek hypotaktik which literally means
subordinate. Indeed, in Latin, the subjunctive was increasingly used in
subordinate sentences even where there seemed to be no notion of irrealis
(see below, 5.5).
However, the subjunctive is also used in main clauses, and its uses there
are rather more simply explained than its uses in subordinate clauses. For
that reason, main clauses will be considered first.

5.1.1 Main clauses

Lakoff (1968: 172ff.) recognizes six different uses of the subjunctive in main
clauses in Latin. Her terminology is a little idiosyncratic (or at least very
different from that used in this book), and, for that reason they are
characterized here in terms of the modal categories used in the presen
volume, with Lakoff's terms in brackets :
Jussive ('Imperative'):
Naviget! haec summa est, hie nostri nuntius esto
sail+3so+PRES+suBJ this point is, this of.us message let.it.be
Virg. (Aen. 4- 2 3 7 )
'Let him sail, this is the point, let this be our message'

108
 5.1 Main and subordinate clauses

Volitive ('Optative'):
Ut ilium di. . . perduint (PI. Aul. 785)
that him gods destroy+3PL+PRES+suBj
'May the gods destroy him!'
Obligative ('Jussive'):
Sed maneam etiam, opinor (PI. Trin. 1136)
But remain+lSG+PRES+suBJ still, 1.think
'But I should still stay, I think'
Obligative ('Deliberative'):
Quid agam iudices? (Cic. Verr. 5.2)
what do+lsG+PRES+suBJ jurymen
'What am I to do, gentlemen of the jury?'
Speculative ('Potential'):
lam apsolutos censeas quom incedunt infectores
(PI. Aul. 520)
now paid.off think+2so+PRES+suBj when come.in dyers
'You may think they are already paid off, when in come the dyers'
Presupposed ('Concessive'):
Sit fur, sit sacrilegus . . .
be+3so+PRES+suBj thief be+3sG+PRES+suBJ temple.robber
at est bonus imperator (Cic. Verr. 5.4)
yet he.is good general
'Though he is a thief, though he is a temple-robber . . . he is a good
general'
These can be related to the three types of non-assertion discussed first in 1.1.2
and the distinction of epistemic versus deontic/dynamic modality (1.1.2 and
1.3.2). The first four are deontic (and the subjunctive is used because the
proposition is unrealized). The last but one is epistemic (the speaker has
doubts about the veracity of the proposition). The last one is an example of
presupposition (nothing is being asserted).
Similar examples can be found in Italian (Lepschy and Lepschy 1977:
223-4):
entri pure
enter+3so+PRES+suBJ if.you.please
'Please come in'
9 che venga anchelui
that come+3sG+PRES+suBJ also him
'Let him come too'
potessi venire anch' io
can+lso+PRES+suBJ come also I '''
'If only 1 could come too'

109
Indicative and subjunctive

sapessi che lusso

know+2so+PRES+suBj that grand
'You should see how grand'
che sia finite
that be+3sG+PRES+suBj finished
'I wonder if it's finished'
sia pure come dici tu ma io non vengo
be+3sG+PRES+suBj perhaps as say you but I not come
'It may be as you say, but I'm not coming'
The first is usually regarded as the polite command - Imperative (see 3 \
but, unlike the Imperative proper, in which the verb is in the second person
it has a verb in the third person. In form it is like the second, which is Jussive
The third is a wish (Volitive) and the fourth another example of obligation
(Obligative). All are deontic (and irrealis). The last but one is epistemic
expressing possibility or a degree of doubt (Speculative), while the last is
probably best seen as concessive indicating something presupposed, though it
too could also be interpreted as Speculative, expressing doubt, as in the
previous example. Again they exemplify the three types of non-assertion.
There is something very similar in the West African language Fula (Arnott
1970: 299ff.). Not surprisingly, the term 'subjunctive' is used. There is a
paradigm of the verb that has functions very similar to those of the Latin
subjunctive (though it lacks tense distinctions).
Four types of directives are illustrated:
Injunction
rjgaraa
come+2sG+suBj
'Come on!'
Report for instruction or permission
minasta-na?
come in+lsG+suBJ-iNT
'May I come in?'
Offer or request for permission or invitations with HAA
haa njahen
HAA gO+lPL+SUBJ
'Let's go'
Obligation with SEY
sey rjgurtodaa
SEY come.out+2sG+suBj
'You ought to come out'
These are examples of Imperative, Permissive, Jussive and Obligative respe0
tively. There is also an example of a wish (Volitive):

110
 5.7 Main and subordinate clauses
Wish or prayer
njuutaa bale
be.long+2sc+suBJ in.days
'May you live long!'
il 0 f these the proposition is unrealized. No example of epistemic usage
^aker has doubts about the veracity of the proposition) was given, but
(the syZ'*^
k an example that is perhaps to be explained in terms of presupposition
there is * . . . . . ,
that nothing is being asserted:
Expostulation or rhetorical question
njoooaa
Sit+2SG+SUBJ
'What? You sit down!'
An example of the subjunctive in Swahili used as jussive (see 5.4.2) was
given above. Examples of the same use in another Bantu language, Luvale
(Horton 1949: 302-4) are:
tu-y-e
we-go-suBJ
'Let's go!'
va-iz-e (veze) waxi
they-come-suBJ quickly
'Let them come quickly'

5.1.2 Subordinate clauses

Examples of the use of the subjunctive in subordinate clauses are even more
varied. They include not only those uses that were discussed in Chapters 2 and
3, but others such as future reference and wishes and fears. For that reason the
contents of the next two sections will be distinguished in terms of propositional
modality and event modality rather than in terms of epistemic, deontic and
dynamic. There is then a section on the imperative, which often involves the
subjunctive, and a section showing the use of the subjunctive for no other
re
ason, apparently, than that the verb occurs in a subordinate clause.
As argued in 1.1.2, the distinction between the indicative and the sub
junctive is associated with assertion and non-assertion, and one of the
re
asons for non-assertion is that the speaker has doubts about the veracity of
e
Proposition. But there is some ambiguity about the word 'speaker'. In a
am clause it clearly refers to the actual speaker, the originator of the
terance. It is this speaker who doubts the veracity of the proposition, as in:
Quiz viene
maybe come+3sG+PRES+suBJ
'Maybe he's coming'

111
Indicative and subjunctive

In a subordinate clause, however, the relevant 'speaker' is the reno

speaker, indicated by the subject of the main verb, as can be seen in:
Maria duda que sea buena idea
Maria doubts that be+3sG+PRES+suBJ good idea
'Mary doubts that's a good idea'
(This is a modification of the example in 1.1.2, which was unhelpful becai
the actual speaker and the reported speaker were the same, both T ) i
general, the choice of the subjunctive in main clauses depends on the attitude
beliefs, etc., of the actual speaker. In subordinate clauses, however, it depend
on the attitudes, beliefs, etc. of the person represented by the subject of th
main clause (with those attitudes, beliefs, etc., indicated by the main verb)
This is even true of the use of the subjunctive for what is presupposed, as in-
Le alegra que sepas la verdad
him it.pleases that know+2sc+PRES+suBj the truth
'He's glad that you know the truth'
Generally, however, the actual speaker also accepts that the proposition is
true. (This is debatable, but will not be discussed here.)
No examples of the uses of subjunctive in subordinate clauses are given
here - they are to be found in the following sections, which exemplify both
main and subordinate clauses.

5.2 Propositional modality

5.2.7 Speculative
An example of the subjunctive being used to express epistemic possibility
(Speculative) in Latin was given in 5.1.1. Further examples from Italian and
Spanish are:
che sia finito
that be+3so+PRES+suBJ finished
'I wonder if it's finished'
Quiz viene
maybe come+3sG+PRES+suBJ
'Maybe he's coming'
These are examples of main clauses, although the Italian example 1S
introduced by che which is usually a subordinating conjunction ('that').
An example from a very different language, Baluchi (Bybee et al- 1""
195), was given in 4.1.1:
ma bakly adda kassa byzans
IPL perhaps there someone know+lso+suBjl
'Perhaps we know someone there'

112
 5.2 Propositional modality

subjunctive may also be used in subordinate clauses after verbs of

{ but here there is usually a choice between indicative and subjunctive,
hoice being determined by the speaker's judgment about the truth of the
the
a n t proposition. Thus Butt and Benjamin (1988: 228, 227) state that
r
'believe' and similar verbs such as PARECER 'appear', SUPONER 'suppose'
, S0S PECHAR 'suspect' may also be followed by the subjunctive if there is
tancy or what is stated is more hypothetical. Similarly Lepschy and
oschy (1977: 227) state that either mood may be used with Italian CREDERE
i;eve', PARERE 'appear' and SEMBRARE 'seem'. An example with CREDERE is:
Credo che tu abbia/hai ragione
I.think that you have+3sG+PRES+suBj/iND right
'I think you are right'
The subjunctive is also used in Italian after impersonal expressions that
express possibility and probability such as probabilelimprobabile che 'it is
likely, unlikely that' and pud darsi che, pud essere che 'it may be that'
(Lepschy and Lepschy 1975: 225), and after similar expressions in Spanish.
Slightly differently, in literary German the subjunctive is used with verbs
of belief, but only with the past tense (and also with verbs of saying - see
5.2.2):
Ich glaubte er ware krank
I thought he be+3sc+iMPF+suBj ill
'I thought he was ill'
A different issue is the use of the subjunctive with negated verbs of belief.
This is discussed in 5.2.3.

5.2.2 Reported
The subjunctive is often used to indicate what is reported. In German, it may
be used in main clauses to indicate that it is what is said or was said and not
part of the writer's or speaker's own statement:
Bei seiner Vernehmung berief sich H. auf Notwehr. Er ;
in his examination appealed H to self-defence, he
sei mit S. in Streit geraten und
be+3sG+PRES+suBj with S. in quarrel fallen and
habe sich von diesem bedroht gefuhlt
have+3sG+PRES+suBj self by him threatened felt
'In the course of his cross examination, H. pleaded self-defence. He had
become involved in a quarrel with S. and had felt himself to be
threatened by him'
Th
e
second sentence here is what H. claimed, not part of the writer's own
re
Port.
113
Indicative and subjunctive

In subordinate clauses, the subjunctive is usually to be found in Qe

n

(examples are from Hammer (1983: 265-71), and are mostly from a lit
style): ary

Ich glaubte er ware krank

I thought he be+3so+iMPF+suBJ ill
'1 thought he was ill'
Er sagte er ware krank
he said he be+3so+lMPF+suBj ill
'He said he was ill'
However, in colloquial German, the indicative may be used:
Er glaubte, ich war krank
he thought I be+3sG+iMPF+iND ill
'He thought I was ill'
In a literary style, the subjunctive is often used even with a present tense
verb of reporting, especially when, as in all of the above examples, the
conjunction dass is not used. With dass the indicative is more likely:
Er sagt, er msse nach Hause
he says he must+3so+PRES+suBj to house
'He says he must go home'
Er sagt, dass er nach Hause muss
he says that he to house must+3sG+PRES+iND
'He says he must go home'
The indicative without dass may suggest that the speaker accepts the reported
proposition as true, as in German:
Er sagte, er schwimmt gern
he said he swim+3sG+PRES+iND with.pleasure
'He says he likes swimming'
Italian does not generally use the subjunctive after a verb of reporting that
is neither negated or interrogative (5.2.3, 5.2.4), but it does so after si dice
'one says', which indicates 'hearsay' (Reported (3)). Compare (Lepschy and
Lepschy 1977: 226):
Ada dice che i soldti sono partiti
Ada says that the soldiers be+3PL+PRES+lND left
'Ada says that the soldiers have left'
Si dice che i soldti siano partiti
One says that the soldiers be+3PL+PRES+SUBJ left
'They say that the soldiers have left'
This is also true of German:

114
 5.2 Propositional modality

Man sagt, er sei gestorben

One says he be+2sG+PRES+suBJ died
'They say he's died'
does not use the subjunctive for reported statements (see 7.4.1).
p-r a relative clause within a subordinate clause of reported speech in
U0vvevc '
1 u5 nlaced in the subjunctive if the subordinate clause is part of what is
Latw 1 v
reported:
Dicit se de Gallis . . . postulare triumphum quos
he.says self from Gauls to.demand triumph whom
acie vicerit (Liv. 36.40.3)
by.battle defeat+3so+PERF+suBj
'He says he claims a triumph from the Gauls, whom he has defeated in
battle'
If it is part of the information being provided by the speaker (the original
speaker) the indicative is used:
Diogenes . . . dicere solebat Harpalum, qui temporibus illis
Diogenes to.say used Harpalus who in.times those
praedo felix habebatur, contra deos
brigand happy be.held+3sG+iMP+iND against gods
testimonium dicere (Cic. D.N. 3.34)
witness to.speak
'Diogenes used to say that Harpalus (who at that time was generally
thought to be a fortunate brigand) was a witness against the gods'
There is an exactly parallel situation in German (Hammer 1983: 268):
Er sagte, er werde das Buch kaufenda sein
he said he be+3sG+iMPF+suBj the book to.buy as his
Onkel. . . es ihm empfohlen htte
Uncle . . . it to.him recommended have+3sG+iMPF+suBj
'He said he would buy a book as his uncle . . . had recommended it to
him'
Er sagte, er bewerbe sich um diese Stelle,
he said he apply+3sG+PRES+suBJ self to this job
fiir die er gar nicht geeignet ist
for which he at all not suited be+3so+pRES+iND
'He said he was applying for this job, for which he is not at all suitable'
*ne subjunctive is also used in Latin for a reported cause:
Aristides . . . nonne ob earn causm expulsus est patria
Aristides NEG INT for that cause expelled is from.country
quod praeter modum iustus esset (Cic. T.D. 5.36.105)
because beyond mean just be+3sG+iMpF+suBJ
'Was not Aristides exiled because (it was said) he was excessively just?'

115
Indicative and subjunctive

(For an example from Classical Greek, but with the optative, see 8.2.2.)
The subjunctive is similarly used in German, e.g. (Hammer 1983: 269)-
Papa m5chte auch gern selbst lenken, Mama will es
Papa would too with.pleasure self to drive, Mama wishes it
aber nicht weil es die Nerven angreife
however not, because it the nerves strain+3so+PREs+suBj
'Daddy would like to drive, but Mummy doesn't want him to because
(she says) it is a strain on the nerves'
Only slightly different is the Spanish example offered by Lavandera (197.
19):
Mientras que a vos no te falte nada, como
as.long as to you not you lack+3sG+PRES+suBj nothing as
vos decis . . .
you say
'As long as you don't need anything as you say . . .'
Here the comment como vos decis clearly indicates that the reason for
inaction (presumably) was given by the addressee.

5.2.3 Negative
The most common association of Negative with subjunctive is in subordinate
clauses where the superordinate clause is negated. This association is rare in
main clauses, in spite of the fact that within event modality the subjunctive is
often used with negative imperatives (5.4.2) and that there are instances of
the association of negative with irrealis in main clauses (6.6.4). However, one
example of subjunctive with negative in a main clause is to be found in
Luvale (Bantu - Horton 1949: 302-4), where it is used with a particle
translated as 'not yet':

kanda tu-mum-on-e
not.yet we-him-see-sUBJ
'We haven't seen him yet'
In subordinate clauses the subjunctive is regularly used after negated verbs
of belief and report, but rather different issues seem to be involved and. f r
that reason, the two types will be dealt with separately.
With verbs of belief, if the verb is negated, the subjunctive is generally used
in the subordinate clause in the Romance languages. Compare from Spanis
(Klein 1975: 353):
Creo que aprende
I.believe that Iearn+3SG+PRES+IND
'I believe that he is learning'
116
 5.2 Propositional modality
No creo que aprenda
not I.think that learn+3sG+PRES+suBj
'I don't think that he is learning'
arnpl es of similar negated form in Italian (Hall 1964: 222) and French
(Bloomfield 1933: 273) are:
Non credo che sia Corelli
not I.think that be+3so+PRES+suBj Corelli
'I don't think that it's Corelli'
Je ne pense pas qu'il vienne
I not think that he come+3so+PRES+suBj
'I don't think he'll come'
The subjunctive is also used with verbs that express doubt even though
there is no formal negation, e.g., in Spanish, (Klein 1975: 356, 353):
Dudo que aprenda
I.doubt that learn+3sG+PRE$+suBJ
'I doubt that he's learning'
From this it might be suggested that the use of the subjunctive is not directly
determined by the negative, but by the expression of doubt. On this view the
subjunctive is used because 'not-think' is an expression of doubt, and verbs
of doubting take the subjunctive, as shown by the previous Spanish example
and its Italian counterpart:
Dubito che impari
I.doubt that learn+3sc+pRES+suBJ
'I doubt that he's learning'
This view is supported by the fact that with a verb of belief that is not
negated, the use of the subjunctive indicates the speaker has some doubt
about the truth of the proposition, as was noted in 5.2.1.
There is, however, another way of looking at this. I don't think he is stupid
can be interpreted as 'I think he's not stupid.' This has been treated in
theoretical studies as 'negative raising' or 'negative transportation', which
suggests that the negative originates in the subordinate clause, but is raised
or
transported to the main clause. An even better analysis is in terms of the
scope of the negation. With the most obvious interpretation there is narrow
scope in / don't think he is stupid, since only the subordinate clause is
(notionally) negated. In contrast, on the less likely interpretation 'It's not the
as
e that I think he's stupid', there would be wide scope, negating the whole
Se
ntence.
"i this respect it is relevant that, with a negative imperative, the indicative,
n
t the subjunctive, is used:

117
 ^

Indicative and subjunctive

No crea usted que es tonto
not think+3sG+PRF.s+suBj you that be+3sG+PRES+iND stupid
'Don't think that he is stupid'
Here, the interpretation is not in terms of 'not-think', i.e., 'doubt', but '
terms of a command not to think - 'Not (think(he is stupid))' rather th
'Think (not(he is stupid))'. In terms of scope, it appears that the contrast'
between narrow scope with the subjunctive (as in the previous examples) and
wide scope (as here) with the indicative.
The indicative is also to be found with negative questions:
i,No es verdad que ha dicho eso?
not is truth that have+3so+PRES+iND said that
'Isn't it true that he said that?'
Butt and Benjamin (1988: 227) comment that negative questions are not
semantically negative, and so the indicative is used, but in terms of scope it is
clear that this does not mean 'Is it true that he didn't say that?' (with narrow
scope).
Now, what is important for comparison of verbs of belief with verbs of
saying is that the subjunctive is used in the former where there is narrow
scope. This, as will now be seen, does not apply to verbs of saying.
With verbs of saying in Romance the verb 'say' is generally followed by
the indicative unless it is negated. If it is negated either the indicative or the
subjunctive may be used, as in Italian:
Dico che ha torto
I.say that have+3sc+PRES+iND wrong
'I say that he is wrong'
Non dico che lui abbia/ha torto
not I.say that he have+3sG+PRES+suBj/rND wrong
'I do not say that he is wrong'
The crucial point here is that there is now wide scope - the interpretation of
the negated example is 'It is not the case that I say he is wrong' not T say that
he is not wrong.' With verbs of saying, unlike the verbs of belief, the
subjunctive occurs where there is wide scope of the negation.
The same situation (with the same argument) applies to 'know', as in
Spanish:
Yo sabia que l estaba ah
I knew that he be+3sG+PAST+iND here
'I knew that he was here'
Yo no sabia que l estaba/estuviera ah
I not knew that he be+3sG+PAST+iND/suBj here
'I did not know that he was here

118
 5.2 Propositional modality

second does not mean T knew that he was not here', but 'It is not the
that I knew that he was here'.
Since the explanation in terms of narrow scope (and 'not-think equals
b) is not available for negated verbs of saying, a different reason for the
0 f the subjunctive with negated verbs of saying must be sought. In fact,
here is a fairly obvious answer - that what is not said is not asserted. It will
seen in 6.6.4 that, in languages where the mood distinction is made in
rms of Realis/Irrealis, Negative is similarly associated with Irrealis.
It may be noted, however, that 'deny', also requires the subjunctive in
Spanish:
Niego que haya venido
I.deny that have+3sG+PRES+suBj come
'I deny that he has come'
For with 'deny' there is (notionally) narrow scope. The example above means
'I say that he has not come.' So both wide scope and narrow scope of the
negative are associated with the subjunctive.
The situation in German is similar to that in the Romance languages. The
subjunctive is found even with present tense verbs if they are negated:
Er sagt, er ist miide
He says he be+3sG+PRES+iND tired
'He says he is tired'
Er sagt nicht, er ware miide
he says not he be+3so+iMPF+suBJ tired
'He does not say he is tired'
A rather different example of negative and subjunctive is to be found in
causal clauses in Latin. These are introduced by a causal conjunction and are
generally followed by the indicative:
Torquatus . . . filium suum quo is contra imprium in hostem
Torquatus son his because he against rule in enemy
pugnaverat, necari iussit (Sail. C. 52)
fight+3sG+PLUP+lND be.killed he.ordered
'Torquatus ordered his son to be executed, because he had fought
against the enemy contrary to orders'
However, the subjunctive is used when the clause is negated, i.e., it is not the
tr
ue reason:

Pugiles . . . ingemescunt non quod doleant. . . sed

boxers groan not because be.in.pain+3PL+PRES+suBJ but
quia profundenda voce omne corpus intenditur (Cic. TD. 2.23.56)
because with.bursting voice whole body be.stretched+3sG+PRES+iND

119
Indicative and subjunctive
'Prize-fighters groan not because they are in pain, but because their
whole body is made more tense by the burst of sound'
(This may be compared with the subjunctive for a reported cause in 5 i ->
The same kind of feature is found in Spanish (Lavandera 1978: 21):
Yo no lo digo porque a mi me moleste
I not it say because to me me bother+3sG+PRES+suBj
'I don't say it because it bothers me' {i.e. 'not because . . .')
There is something similar in the more stereotyped expression:
No es que no me guste
not is that not me pleases+3sG+PRES+suBJ
'It isn't that I don't like it'
Here a possible cause is being rejected.
Slightly differently, Spanish uses the subjunctive after sin que 'without', to
indicate that the relevant action did not take place:
Desech el cigarrillo sin que el profesor
he.threw.away the cigarette without that the master
lo viese
it see+3sG+iMPF+suBj
'He threw away the cigarette without the master seeing it'
Similarly in Italian (Lepschy and Lepschy 1977: 225):
E entrato senza che noi lo sentissimo
is entered without that we him hear+lpL+iMPF+suBj
'He came in without our hearing him'

5.2.4 Interrogative
There are two points concerning Interrogative.
First, interrogation and negation often function in very similar ways, so
that together they can be characterized as 'non-assertive' (Quirk et al. 1985:
83). There are, therefore, examples of verbs of belief and verbs of saying
where interrogation, like negation, induces the subjunctive as in the Spanish
and Italian:
^Cree Vd que venga?
believe you that come+3sG+PREs+suBJ
'Do you think he will come?'
sai se sia vero?
you.know if be+3SG+PRES+suBJ true
'Do you know if it's true?'
This is also true of German even with present tense verbs of saying ( se
5.1.2):
120
 5.2 Propositional modality

Sagt er er ware miide

Says he he be+3sG+iMPF+suBJ tired
'Does he say he is tired?'
There is an interesting opposite situation with negative interrogatives in
Cnanish, which use the indicative. This was discussed in 5.2.3 with the
examp' e:
^No es verdad que ha dicho eso?
not is truth that have+3so+PRES+iND said that
'Isn't it true that he said that?'
Secondly, the subjunctive is used in reported (indirect) questions, as in
Spanish and Italian, where the indicative is more informal:
mi chiese se fosse/era possibile
me he.asked if be+3so+iMPF+suBj/iND possible
'He asked me if it would be possible'
The subjunctive is always required for indirect (reported) questions in Latin,
though the indicative is used in direct questions:
Quid agis? -> Rogo quid agas
What do+2sc+PRES+iND -> I.ask what do+2sG+PRES+suBj
'What are you doing?' -> 'I ask what you are doing'
Rogavi pervenisset-ne Agrigentum (Cic. Ven. 2.4.12.27)
I.asked arrived+3sG+PLUP+suBi-iNT Agrigentum
'I asked if he had arrived at Agrigentum'

5.2.5 Presupposed
It was noted in 1.1.2 that the subjunctive is used in Spanish where the
proposition (in a subordinate clause) is presupposed, with the example:
Me alegra que sepas la verdad
me it.pleases that know+2sG+PRES+suBj the truth
'I'm glad that you know the truth'
There is a similar situation in Italian (as well as other languages), as can be
seen from the Spanish and Italian examples:
Sp. Lamento que aprenda
It. Mi displace che impari
I.regret that learn+3sG+PRES+suBJ
'I regret that he learns/is learning'
Verbs like REGRET have what Kiparsky and Kiparsky (1971: 245-8) call
active complements' (see 1.4.1), in that the proposition in the complement is
P r esupposed, i.e., accepted as true by the speaker (and, the speaker assumes,
y the addressee). The essential point, as explained in 1.1.2, is that, since the

121
Indicative and subjunctive

proposition is presupposed, nothing is being asserted and, when there is non

assertion, the subjunctive is used.
Other examples of the subjunctive in a subordinate clause from Spanish
and Italian are:
me molesta que te quejes tanto
me it.bothers that VOU+REFL complain+2sG+PRES+suBj so
'It bothers me that you complain so much'
mi sorprende che tu dica questo
me it.surprises that you say+2so+PRES+sUBJ that
'It surprises me that you say that'
According to Given (1994: 304), the subjunctive is used in Spanish with
verbs such as SENTIR in the sense of 'sympathize', SORPRENDERSE, 'be
surprised', MOLESTARSE, and ALEGRARSE 'be glad'. There are similar verbs in
Italian.
The subjunctive is similarly used with impersonal constructions that
express emotion (Spanish and Italian):
fu una lstima que no me lo dijeras
it.was a pity that not me it tell+2sG+PAST+suBJ
'It was a pity that you didn't tell me'
peccato che sia gi partit
it.is pity that be+3so+PRES+suBJ already left
'It is a pity that he has already left'
With some expressions, Givn states, there is a choice of indicative or
subjunctive depending on the 'degree of emotional reaction', the subjunctive
expressing greater emotion, e.g. HORRORIZARSE 'be shocked' and SER INCRE-
BLE 'be incredible'. In Italian there is a stylistic choice according to Lepschy
and Lepschy (1977: 225) with verbs and expressions such as RINCRESCERE, 'be
sorry', STUPIRSI 'be surprised', ESSERE CONTENTO/SCONTENTO 'be pleased/
displeased', ESSERE ARRABIATO 'be cross' etc.
The contrast between the use of the indicative for what is asserted and the
subjunctive for what is presupposed is well illustrated by verbs that can be
used in two senses. Compare the Spanish and Italian pairs (Klein 1975,
Lepschy and Lepschy 1977: 228):
Siento que aprende
I feel that learn+3so+PRES+iND
'I feel that he is learning'
Siento que aprenda
I.feel that learn+3so+PRES+suBJ
'I regret that he should learn' ,; -..
122
T

5.2 Propositional modality

Si capisce che sono arrabiati
One understands that be+3PL+PRES+iND cross
'It's clear that they are cross'
Si capisce che siano arrabiati
One understands that be+3PL+PRES+suBj cross
'It's understandable that they should be cross'
In the first of each pair the speaker presents the proposition as new
infonnation and the indicative is used, while in the second the speaker
expresses his attitude towards shared knowledge, and the subjunctive is used.
Powerful evidence that what is relevant is not the emotional attitude itself,
but the fact that there is no new information, is shown by the use of the
subjunctive after el hecho que 'the fact that' (Butt and Benjamin 1988: 221):
el hecho que Espafla no tenga petrolio
the fact that Spain not have+3sG+PRES+suBJ oil
explica las dificultades econmicas
explains the difficulties economic
'The fact that Spain doesn't have any oil explains the economic
difficulties'
Although there is specific reference to 'fact' here, it is clear that the
proposition is not being asserted, but is presupposed. By contrast where the
'fact' {el hecho) is new information, the subjunctive is not used (Butt and
Benjamin 1988: 229):
No lo hace por de hecho que no le gusta
not it does for of fact that no her please+3sG+PRES+iND
'She doesn't do it, because she doesn't like it'
Finally, but quite importantly, the subjunctive may be used in concessive
clauses to grant or admit that something is true. These may be in a main
clause in apposition to a following clause (see 5.1.1), or an (oblique)
subordinate clause as in Latin:
Sit fur, sit acrilegus . . .
be+3so+pRES+suBJ thief be+3sG+PRES+suBJ temple robber
at est bonus imperator (Cic. Ven. 5.4)
yet he.is good general
'Though he is a thief, though he is a temple-robber . . . he is a good
general'
Quamvis sis molestus, numquam te esse
although be+2sG+PRES+sUBJ troublesome never you to.be
confitebor malum (Cic. T.D. 2.25.61)
I.will.confess evil
'Although you are troublesome, I will never admit you to be evil'

123

I
Indicative and subjunctive

Similar examples from Italian and Spanish are:

sia pure come dici tu ma io non vengo
be+3sc+PREs+suBJ perhaps as say you but I not come
'It may be as you say, but I'm not coming'
Aunque sea difcil, lo har
although be+3so+PRES+SUBJ difficult it I.will.do
'Although it is difficult, I will do it'
What is conceded is not asserted, but is presupposed, and the subjunctive i
used.

5.2.6 Future
Futurity is often marked as irrealis in languages in which mood is described
in terms of realis/irrealis (6.6.1). This is much less common with indicative/
subjunctive.
In main clauses, the only instance is the use of the subjunctive to refer to
the future in Homeric Greek:
ou gr po: toous don anras, oud
not for ever such see+lso+AOR+iND men nor
do:mai (Hom. //. 1.262)
see+lsa+AOR+sUBJ
'I have never seen such men, nor shall I see'
However, in temporal clauses, there is a widespread use of the subjunctive
to refer to hypothetical future events, with the indicative used for past events.
Examples from Classical Greek and Spanish are (the first of each pair has the
indicative, the second the subjunctive):
epe d etelete:se Dareos. . . , Tissaphrne:s
when but die+3sG+AOR+iND Darius Tissaphernes
diabllei tn Kron (Xen. An. 1.1.3)
slanders the Cyrus
'When Darius died, Tissaphernes slanders Cyrus'
epeidn d diaprkso:mai h domai, h:kso: (Xen. An. 2.3.29)
when but finish+ ISG . AOR+SUBJ what I want I.will.come
'When I have finished what I want to do, I will come'
Cuando lleg en Inglaterra, vino a ver-me
when arrive+3sG+PAST+iND in England he.came to see.me
'When he arrived in England, he.came to see.me'
Cuando se termine la guerra, volver a
when REFLfinish+3sG+PRES+suBJthe war I.will.return to
Inglaterra
England
'When the war ends, I will return to England'

124
 5.2 Propositional modality

There are similar examples in locative clauses:

Dondequiera que era, me escribia
wherever that be+3sG+iMPF+iND to.me he.wrote
'Wherever he was, he would write to me'
Dondequiera que yo est, te escribir
wherever that I be+lsG+PRES+suBJ to.you I.will.write
'Wherever I am, I will write to you'
The subjunctive may also be used in Classical Greek in the apodosis (the.
//"-clause) of a real conditional clause (see 8.3) that refers to the future
(though the future indicative is also possible):
en touto poi:is, hamart:seis
if this do+2SG+PRES+SUBJ, err+2SG+FUT+IND
'If you do this (should do this), you will be wrong'

5.2.7 Conditional
The subjunctive is used in conditional clauses in a number of languages. In
Latin, for instance, the present subjunctive is used in:
Si hoc facias, erres
if this do+2SG+PRES+SUBJ err+2SG+PRES+SUBJ
'If you did/were to do this you would be wrong'
However, 'modal tense' (1.4.4, 8.2) is usually involved as well, and, for that
reason, a detailed discussion of subjunctive in conditional sentences is left to
8.3.2.

5.2.8 Other types

There is a distinction in Spanish between:
El que asasin a Smith est loco
he that 1C11+3SG+PAST+IND to Smith is mad
'The man who killed Smith is mad' >
and
El que asasinara a Smith est loco
he that kill+3sG+PAST+suBj to Smith is mad
'The man who killed Smith is mad'
e precise interpretation of these and the status of the form asasinara is a
m
atter of some debate (see Rivero 1975, 1977; Rojas 1977), but it seems that
ne first refers to a particular person, while the second has a 'whoever' sense
"at links the killing with the insanity. Paraphrases would be 'I conclude that
certain person (the one who, in fact, killed Smith) is insane' and 'I conclude
"at whoever killed Smith is, thereby, shown to be insane.'

125
Indicative and subjunctive

The subjunctive is being used, it would appear, for indefiniteness. Su

indefiniteness is treated as irrealis because it is not asserted that a cert-'
person was the killer.
Slightly different is the use of the subjunctive in:
Busco un empleado que hable ingls
I.look.for an employee who speak+3sG+PRES+suBJ English
T'm looking for an employee who speaks English'
Busco a un empleado que habla ingls
I.look.for to an employee who speak+3sc+PRES+iND English
'I'm looking for an employee who speaks English'

The first means that I am looking for an employee (anyone) who can speak
English; the second means that I am looking for a certain employee who can
in fact, speak English. (However, as can be seen, the distinction is also
marked by the use of the preposition a before specific objects.) There are
similar constructions in Italian (Lepschy and Lepschy 1977: 229):

Cerco una ragazza che sappia cinese

I.look.for a girl who know+3sG+PRES+SUBJ Chinese
T am looking for a girl who speaks Chinese'

Cerco una ragazza che sa cinese

I.look.for a girl who know+3sc+pREs+iND Chinese
'I'm looking for a girl who speaks Chinese'

This too could be seen in terms of indefiniteness, but may also be

compared with the use of the subjunctive for 'relative purpose' (5.3.2), and
also with its use in complement clauses of wanting (5.3.3). Comparison with
the latter suggests that the subjunctive is determined or governed by the main
verb.
To make the point in a better way, it may be said that the whole relative
clause is within the semantic scope of the verb, in that speaking English, etc.,
is part of what the speaker is looking for. The indicative, by contrast, shows
that the relative clause is not within the scope of the verb. It may be recalled
that, similarly, relative clauses in Latin that are within complement clauses of
indirect speech are in the indicative if they are not part of what was said (see
5.1.2). This too can be seen as a matter of scope - the indicative shows that
the relative clause is not part of what is being reported.

5.3 Event modality

Most of the types of event modality for which the subjunctive is used are very
different from those expressed by modal verbs, apart from Directives.

126
 5.3 Event modality

5JJ Directives
. sU bjunctive is also used to express the deontic notions of weak obligation
nressed in English by should, the modal-past equivalent of must (2.1.5)):
Italian
Sapessi che lusso
know+2sG+PRES+suBJ that grand
'You should see how grand'
Latin
Iniurias fortunae . . . defugiendo relinquas (Cic. T.D. 41)
wrongs of.fortune by.fleeing leave+2so+PRES+suBJ
'Flee from and leave behind you the blows of fortune'
Latin
Sed maneam etiam, opinor (Pi. 7.1136)
but remain+lSG+PRES+suBJ still, I.think
'But I should still stay, I think'
An example from a very different language, Baluchi (Bybee et al. 1994:
195), was given in 4.1.1:
a e bbart
3SG 3PL take away+3sc+suBJ
'He ought to take them away'
There is, however, a problem with some of these. Those with second
person subjects are not always easily distinguished from what are treated as
'polite imperatives' (to be discussed in 5.3.2), and those with first and third
person subjects are not easily distinguished from jussives ('first and third
person imperatives'), which are discussed in 5.4.1. Examples from Spanish
are:

Tome su libro
take+3sG+pRES+suBJ your book
'Take your book'
Hablemos de otra cosa
talk+lPL+PRES+suBj about other thing
'Let's talk about something else'
Que entre
that enter+3sc+PRES+suBJ
'Let him come in'
The imperfect and pluperfect subjunctive are used in Latin to refer to what
s
hould have been done in the past. Here they are clearly distinct from
operatives and jussives, which are essentially performative, and so relate to
tlle
present:

127
 Indicative and subjunctive

sed tu dictis, Albane, maneres (Virg. Aen. % *,

but thou to.words Alban remain+3sG+iMPF+suBJ
'But thou, Alban, shouldst have kept thy word'
Adservasses hominem (Cic. Verr. 65)
keep+2sG+PLUP+suBJ man
'You should have kept the man'
' One other use that seems to belong here is the 'deliberative', with which
speakers question what they should (or perhaps must) do:
Latin
Quid agam iudices? (Cic. verr. 5 21
what do+lso+PRES+suBj jurymen
'What am I to do, gentlemen of the jury?'
Greek
:moi eg p:i b:? (Eur. Hec. 1056)
alas I where go+lso+PRES+suBi
'Ah me! Where shall I go?'
These too can refer to the past:
haec cum viderem, quid agerem . . . ? (Cic. Sest. 19)
these when I.saw what do+lSG+iMPF+suBJ
'When I saw this, what was I to do?'
There is not a great deal to be said about directives and subordinate clauses.
In most European languages the infinitive is used after expressions of
permission and obligation (see 7.4.1), e.g. after Latin licet 'it is allowed' and
oportet 'it is necessary', though Modern Greek uses a conjunction with a finite
verb (see 4.1.4 for examples). However, some languages use subjunctive with
such constructions. This is so with the Western Nilotic languages (Bavin 1995:
112-13). Thus in Acholi the subjunctive is used after wek for permission,
myero for obligation and gwok for prohibition. An example of myero is:
myero a-ngwec-i
must 1 +SG-run-suBJ
'I must run'

5.3.2 Purposive
Clauses that express purpose are marked with a conjunction plus the
subjunctive in Latin and Classical Greek:
Haec acta res est ut ii nobiles
this done thing is that those nobles
restituerentur in civitatem (Cic. Rose. Am- 51-1
be.restored+3PL+iMPF+suBJ in state
'This was done so that those nobles should be restored to the state

128
 5.3 Event modality
tn gr kkon ae di kolzein, hin'
the for bad always it is necessary to.punish in.order.that
ameino:n e:i (Plat. Leg. 944d)
better be+3sG+PRES+suBJ
'For we must punish the bad man that he may be better'
The same kind of construction is used in Spanish and Italian:
Le prest el dinero para que pudiese
him l.lent the money in order that can+3so+iMPF+suBJ
comprar su billete
buy his ticket
'I lent him the money in order that he could buy his ticket'
ti scrivo affinch tu capisca la situazione
you I.write in order that you understand+2so+PRES+suBJ the situation
'I am writing to you so that you understand the situation'
The usual way of expressing negative purpose is simply to use a negative in
the subordinate clause, with the same conjunction. However, Latin generally
uses tie in place of ut (and only rarely ut ne):
Dolabella, ne collum tonsori committeret, tondere
Dolabella lest neck to.the.barber entrust+3sG+iMPF+suBJ to.shave
filias suas docuit (Cic. Tusc. 5.20.58
daughters his taught
'So that he should not entrust his neck to a barber, Dolabella taught his
daughters how to shave'
Similarly, Classical Greek usually has the negative m: in the subordinate
clause after a conjunction, but Homeric Greek may use m: alone:
apstiche m: ti no:se:i (Horn. //. 1.522)
depart lest something notice+3sc+AOR+suBJ
'Depart lest she notices anything'
Latin ne and Classical Greek m: are here associated with the subjunctive,
whereas their other negatives, non and ouk, are generally found with the
indicative. It is tempting, for that reason, to treat these negative forms as
nrealis and realis respectively. But the conditions for their use are complex,
a
nd, in both languages, there are exceptions to the general rule (see especially
the use of non with the subjunctive for 'result' in 5.2.3). It is beyond the scope
oi
this book to deal with them in detail, but attention is drawn to their use
or
fears (5.3.3) and negative imperatives and jussives (5.4.2).
Purpose may be expressed with conjunctions other than those that
translate 'in order that', e.g. with Latin dum 'until':
Expecta . . . dum Atticum conveniam (Cic. An. 1. 1.4)
wait until Atticus meet+lSG+PRES+suBJ
'Wait until I meet Atticus' . . .

129
Indicative and subjunctive

Without the notion of purpose the indicative is used:

Dum anima est, spes esse dicitur (Cic. An. 9 in
while life be+3sc+PRES+iND hope to.be is.said
'It is said that while there is life there is hope'
Purpose can also be expressed with a relative clause plus the subjunctiv
(traditionally 'relative purpose'):
Scribebat tamen orationes, quas alii dicerent (Cic. Brut. 6 20M
he.wrote however speeches which others say+3pL+iMPF+suBj
'However, he wrote speeches for others to give'
Here quas has the function of ut. . . eas 'so that . . . them'. Idiomatic Latin
requires this relative construction when there is a pronoun in the subordinate
clause.
Slightly different, but interesting in this context, is a construction in
Tigrinya (Palmer 1962: 38), where the relative is followed by the imperfective
(which is the usual form in subordinate clauses):
qwal'a zihaqqf, zissnkl bahuq may
child REL+it.be.nursed, REL+it.be.cooked dough, water
zimasa' zahaqqwansaba
REL+it.be.drawn REL+it.be.churned milk
'A child to be nursed, dough to be cooked, water to be drawn, milk to
be churned'
Some of the examples treated in terms of 'indefiniteness' in 5.2.8 (those
after 'look for') might also be treated in terms of purpose.
It is worth noting that purpose clauses have much in common with the
complements of directives. Thus in Latin, ut and the negative ne with the
subjunctive are used both to express 'in order that' and after verbs of
ordering, requesting, etc. Compare:
Caesar singulis legionibus singulos legatos et quaestorem
Caesar to.single to.legions single legates and quaestor
praefecit ut eos testis quisque suae virtutis
put.in.charge that them witnesses each of.his virtue
haberet (Caes. B.C. 1.52)
have+3sG+iMPF+suBj
'Caesar put the legates and the quaestor each in charge of a legion so
that everyone might have them as witnesses of his valour'
Rogat et orat Dolabellam ut de sua provincia
he.asks and he.begs Dolabella that from his province
decedat (Cic. Ven. 1-29-72'
withdraw+3SG+PRES+sUBj
'He asks and implores Dolabella to leave his province'

130
 5.3 Event modality

So too, in Russian, the infinitive is used both with 'want' and for purpose
hen the subject of the two clauses is the same, while chtby ('that' plus
fticle -by) with the past tense is used for both types when the subjects are
afferent (and the past tense with the particle -by is often called the
'subjunctive'-see 8.5):
on khochet plavat'
he wants swim+iNFiN
'He wants to swim'
my prishli poetit' bol'novo
we have.come visit+iNFiN patient
'We've come to visit the patient'
h
ja khochu, chto-by vy bol'she eli '-'
I want that-by you more eat+PAST+M.SG
'I want you to eat more'
chto-by nikto ne znal ob etom, nado molch
that-by no-one not know+PAST+M.SG about this necessary be.silent
'So that no-one should know about this, we must be silent'
There is a comparable feature in Ngiyambaa (Donaldson 1980: 2801), in
that the marker for purpose clauses (the 'purposive', which is also used to
mark Obligative - see 1.4.5) is also used in subordinate clauses with verbs
expressing 'knowing how'/'remembering' and 'wanting' (with the 'caritative'
case marker also for the latter):
rjadhu dhi:rba-nha gururja-giri
I+NOM know-PRES swim-PURP
'I know how to swim'
bura:y wagayma-giri-ninda gar.a
child+ABs play-PURP-CARiT be-PRES
'The child wants to play'
This is hardly surprising, since purpose clauses express what the subject
wants or intends, and they are quite close in meaning to complement clauses
with the intention, etc., expressed by lexical items, as in:
He did it hoping/intending that they would come
He did it so that they would come

5.3.3 Wishes, fears, etc.

he subjunctive is used for wishes and fears. The term 'Desiderative' is
v
ailable for the former (a possible alternative, 'Volitive', has been used in a
Hferent sense - see 3.3.1), but there is no recognized term for the latter,
though the formation from Latin would be 'Timitive' (1.7).

131
Indicative and subjunctive

In main clauses, for wishes for the future, the present subjunctive is i,
sed
in, e.g., Latin, Italian and Portuguese:
Ut ilium di . . . perduint (PI A .
that him gods destroy+3PL+PRES+suBJ
'May the gods destroy him!'
Dio vi benedica
God you bless+2pL+PRES+suBj
'May God bless you'
Venha a dia
come+3sG+PRES+suBj the day
'May the day come!'
It is also used in Fula, as noted in 5.1.1:
njuutaa bale
be long+2sG+suBJ in.days
'May you live long!'
Impossible wishes for the present and past are expressed in Latin and
Italian by the imperfect subjunctive:
modo valeres (Cic. An. 9.22)
only be well+2sG+iMPF+suBj
'If only you were well'
potessi venire anch' io
can+lsG+PRES+suBJ come also I
'If only I could come'
Similarly, impossible wishes for the past are expressed by the pluperfect in,
e.g., Latin:
utinam ne. . . tetigissent litora puppes (Catull. 64.171)
that not touch+3PL+PLUP+suBJ shores ships
'Would that their ships had not touched the shore'
However, generally, in European languages, wishes have the same form as
conditionals (this is true even of the languages exemplified above), and are
therefore considered in detail in 8.4.
In subordinate clauses after verbs of wanting and hoping, Spanish normally
requires the subjunctive if there is a different subject (Givn 1994: 286):
Quiero que venga
I.wish that come+3sG+PRES+suBJ
'I wish that he would come'
Espero que venga pronto
I.hope that come+3sc+PRES+suBJ soon
'I hope he comes soon'

132
 5.3 Event modality

So does Italian:
Voglio che venga
I.wish that come+3sG+PRES+sUBJ
'1 wish that he would come'
Spero che Ugo ritorni sabato
I.hope that Ugo return+3sG+suBJ Saturday
'I hope that Ugo comes back on Saturday'
Latin may have either the accusative and infinitive construction (see 7.4.1)
or the subjunctive for wanting:
Ut mihi aedis aliquas conducat volo (PI. Merc. 560)
that to.me house some buy+3sG+PRES+suBJ I.wish
'I want him to rent a house for me'
Nolo me in tempore hoc videat senex (Ter. And. 819)
I don't wish me in time this see+3sG+PRES+suBj old.man
'I don't want the old man to see me now'
But for 'hope' the accusative and infinitive is the normal construction in
Latin (see 7.4.1).
Fears are usually expressed with a verb of fearing plus the subjunctive in a
subordinate clause, as in Spanish:
Temo que haya muerto
I.fear that have+3so+PRES+suBj died
'I fear that he has died'
Latin and Classical Greek require simply ne and m: with subjunctive
(these are the arguably 'irrealis' forms used for negative purpose - see 5.3.2):
Timeo ne laborem augeam (Cic. Leg. 1.4)
I.fear work increase+lso+PRES+suBj
'I am afraid that I shall increase my work'
Ddoika gr m: ou hsion :i . . . apagoreein
I.fear for lest not righteous be+3sc+PRES+suBJ . . . to refuse
(Plat. Rep. 368B)
'For I am afraid it will not be righteous to refuse'
There is, however, an interesting usage in Classical Greek where an
e
xpression of fear can be indicated without a verb of fearing, simply by the
subjunctive preceded by the negative me:. Often, however, this expresses
"Me more than an unwelcome possibility, like English 'I'm afraid that' as
in:

m: d: n:as hlo:si (Horn.//. 16.128)

lest indeed ships take+3PL+A0R+suBj
'I'm afraid they'll take the ships'

133
Indicative and subjunctive

all m: ou tu :i khalepn . . . thnaton

but lest not this be+3sG+PRES+suBJ difficult death
ekphugein (P ,
38a)
to.avoid
'I suspect it is not the avoidance of death that is the difficulty'
It may even express no more than 'perhaps'

all m: touto ou kal.s ho:molog:samen (Plat \je

but lest this not well agree+lPL+AOR+suBj
'Perhaps we have not fairly conceded this'

m: sous diaphthire:i gmous (Eur. Ale i\<\

not your ruin+3sc+PRES+sOBJ marriage
'She may ruin your marriage'

There is something similar in To'aba'ita (Austronesian, Solomon Islands -

Lichtenberk 1995: 294-8), although there is no mood marking. There is a
form that expresses epistemic uncertainty, but also signals apprehension
(Lichtenberk calls it 'apprehensional' and translates it as 'lest'). This too is
used in main clauses, in subordinate clauses after a verb of fearing and also
for negative purpose:

ada keka fanga sui na'a

lest they+SEQ eat COMPLET PERF
'They may have finished eating' (the speaker is worried about getting
food)
nau ku ma'u 'asia na'a ada laalae to'a baa ki keka
I I+FACT fear very lest later people that PL they+SEQ
lae mai keka thaungi kulu
go hither they+SEQ kill US+INCL
'I am scared the people might come and kill us'
nau ku agwa 'i buira fau ada wane 'eri ka riki nau
I I+FACT hide at behind rock lest man that he+SEQ see me
'I hid behind a rock so that the man might not see me'

There are similar examples in Diyari (Australia - P. Austin 1981: 225).

It could be argued that fears and wishes are best treated as epistemic.
since they indicate attitudes to propositions rather than unrealized events-
Indeed, Givn (1994: 280) refers to hopes and fears as 'epistemic anxiety-
Wanting, however, seems to be different in that the emotion is more
directed to the event, and so should be treated as deontic, a kind
directive. A comparison of 'wish' and 'want' may bear out this point in two
ways. First, telling someone what is wanted is often a direction for actio
whereas an expression of wish is not so obviously so. Secondly, wishes, li

134
 5.3 Event modality

and hopes, can relate to the past as well as the present or future,
e
h e r eas
wanting cannot:
I wish John had come
I fear John came
I hope John came
??1 want John to have come
rThe j a s t example may not be impossible, but is very unnatural.) These two
a t u r e s suggest that wanting should be treated as deontic, the others as

epistemic
However, it must be admitted that it is not always easy to distinguish
wanting and wishing. Givn (1994: 283) offers two translations for the
Spanish verb QUERER in:
Quiero que estudies ms
1 want/wish that study+2so+PRES+suBj more
?'I want you to study more'
'I wish you would study more'
The point of the question mark is that the infinitive complement suggests a
'stronger manipulation' than the Spanish, but Spanish has no obvious
alternative.

5.3.4 Resultative
The construction used in Latin to express results (traditionally called
'consecutive' clauses) is almost identical with that used to express
purpose. If the clause is positive, ut plus the subjunctive is used in both
cases. If it is negative, a purpose clause has ne instead of ut (see 5.3.2),
but a consecutive clause retains ut with a following negative non.
Examples are:
Adeo turbati erant. . . ut quosdam consul manu ipse
so disturbed they.were that some consul by .hand self
reprehenderit (Liv. 34.14)
Seize+3SG+PERF+SUBJ
'They were so disturbed that the consul himself seized them by the hand'
Vulneribus confectus ut am se sustinere non posset
by .wounds weakened that still self hold.up not was.able (Caes. B. a 2.25)
'So weakened by his wounds that he could no longer hold himself up'
The use of the subjunctive here has puzzled scholars because results are
sser
>tially factual and the indicative might therefore be expected. Moreover,
this context the arguably 'realis' negative (5.3.2) occurs with the subjunc-
IVe
- Moore (1934: 108) says 'The subjunctive implies, not that the result is

135
Indicative and subjunctive

unreal, but that it is causally connected and logically subordinate to the ,

clause.' Alternatively, it can simply be argued that since the true functi
the subjunctive is to mark subordination, it is precisely that that is exem
fied here (see below). Hamp (1982: 118) comments 'The generalisation of
subjunctive would be a Latin phenomenon.' This may be strictly correct k
is a little misleading, for the similarity between purpose and result claus
found elsewhere. Even in English, so that introduces both types, though th
purpose clause may also contain should:
He worked hard so that he became rich
He worked hard so that he should become rich
The first sentence is ambiguous between the two senses.
Even more striking is similarity with the use of the purposive in the
Australian Ian guages. This was discussed in some detail in 3.5, but it may be
useful to repeat two examples from Yidiny (Dixon 1977: 3456), one expres-
sing purpose, the other a 'natural result':
c[ac|.a tj.uc|,u:mbu ga[bagar.banalnu nuc|.u wawa:lna
child+ABS aunt+ERG hide+PAST not see+PURP
'Auntie hid the child so that it should not be seen
rjayu burawunal duga:l 4.inbic)jnbi:lna
1+SUBJ Burawugal+ABS grab+PRES struggle+RED+PURP
'I grabbed the water sprite woman and as a result she kicked and
struggled'
It can be seen on reflection that purpose and result clauses are alike in that
one expresses an intended result, the other an actual result. It seems that
there is a good case for believing that purpose and result are, in a variety of
languages, closely related, or even indistinguishable, concepts. Even if Hamp
(1982) is correct in arguing that the ut of purpose and the ut of result have
different historical origins, it must be significant that they have now fallen
together, except in the negative.
Yet one might have expected that the contrast between intended and
actual results would be marked by the distinction of subjunctive and
indicative, but the main distinction between purpose and result clauses is
made by the two negatives ne and non. It is almost as if these, rather than
subjunctive/indicative, are the distinguishing markers of 'irrealis' intended
results and 'realis' actual results.

5.4 Imperative and Jussive

Imperative and Jussive were briefly discussed in 3.4, where they appear to
be terms in modal systems. They also occur in languages that hav

136
 5.4 Imperative and jussive

x ^ive/subjunctive mood, and, in this respect, the mood system is not

strictly binary.

5.4.1 Morphology and syntax

ditional grammars recognize second and third person 'imperatives' in
tin and Classical Greek (the latter being instances of Jussive - see 3.4). The
mperatives' for Classical Greek are, for the verb 'loose', given as:
singular dual plural
2 l-e l-eton l-ete
3 l-eto: l-eto:n lonto:n

Examples from Classical Greek are:

all' ei doki, plo:men, hormstho: tchus
but if it.seems sail+lPL+PRES+suBJ set.forth+3sG+PRES+JUSS swift
(Soph. Phil. 526)
'If thou wilt, let us sail and let him set forth with speed'
epmeinon, Ar:ia tekhea do: (Horn. //. 6.340)
wait+2so+AOR+iMP of.Ares equipment put on+lsc+PRES+suBJ
'Wait, let me put on my war harness'
(Both of these, however, also contain subjunctives with the function of first
person jussives - see 5.4.2.)
There are no first person imperatives/jussives in Latin and Classical
Greek, but they occur in other languages - an example from Amharic is
given below. Imperatives may be marked for gender as well as number, e.g.
in many Semitic languages that mark gender in their verbal paradigms.
They are also marked for aspect, in, e.g., Classical Greek and Russian, but
not for tense.
As noted in 3.4, imperatives and jussives do not normally occur in
subordinate clauses or in questions, for the obvious reason that they are
performative - the speaker is actually giving a command. There are,
however, a few exceptions to this. There is one example of a second person
imperative in a subordinate clause in Classical Greek, but, since it occurred
m
a comedy, it may have been a linguistic joke:
all' osth' h dr:son (Ar. Av. 54)
but you.know what do+2sc+PRES+iMP
'You know what do it!'
Wmlarly, there is one example of a jussive in a question:
keistho: nmos hum:n (Plato Legg 801d)
lay.down+3so+PRES+JUSS law for.us
'Shall the law be laid down for us?'

137
Indicative and subjunctive

What appear to be jussives are commonly found in questions in \rnh

aric
(Ethiopian Semitic - Cohen 1936: 179-80):
lawsada-w
take+3sG+JUSS-it
'Must he take it?'
man lawsad
what take+lso+juss
'What shall I take?'
Yet, as the translations show, this rather suggests that these jussives a
perhaps, not to be seen as Jussives, but as deontic Obligatives, more like th
subjunctive in European languages. There is more discussion of the relatio
between Jussive and the subjunctive in 5.4.2.
Similarly, it would not be expected that imperatives can refer to past time
in the way that deontic modals (must have, etc.) can. Yet there is in Syrian
Arabic (Cowell 1964: 36) a construction that looks extremely like a past tense
imperative - the use of the perfect form of KAN 'to be' with the imperative
but with the meaning 'should have':
kant kol lamma kant fal-be:t!
be+2sc+PERF eat+iMPF when be+2so+PERF in.the-house
'You should have eaten when you were at home'

5.4.2 Subjunctive
The subjunctive is used instead of the imperative as a polite command, but in
the third person, in both Italian and Spanish:
entri pure
enter+3so+PRES+suBJ if.you.please
'Please come in'
Tome su libro
take+3so+PRES+suBJ 3so+poss book
'Take your book'
Yet the subjunctive can also be used with second person subjects, to express
weak obligation, as noted in 5.3.1.
In Latin and Classical Greek the subjunctive is used in negative com-
mands, once again with the negatives ne and me: (see 5.3.2), though they may
also use the imperative:
Latin
Ne sis patruus mihi (Hor. Ser. 2.3.88)
not be+2sc+PRES+suBJ uncle to.me
'Don't come the uncle with me'

138
 5.4 Imperative and jussive

Greek
m: th:sthe nmon me:dena (Dem. 3. 10)
not place+2so+AOR+suBJ law none
'Do not pass any law'
., jy ^ e subjunctive is used in Spanish (but not Italian):
No hablas con l
not speak+2so+PRES+suBJ with him
'Don't talk to him'
c o r subordinate clauses, Latin uses the subjunctive with or (rarely)
ithout the conjunction ut or, for the negative, tie:
Rogat et orat Dolabellam ut de sua provincia
he.asks and he.begs Dolabella that from his province
decedat (Cic. Verr. 1.29.7)
withdraw+3sG+PRES+suBj
'He asks and implores Dolabella to leave his province'
Oppidanos hortatur, moenia defendant (Sail. 7 56)
townsmen he.urges walls defend+3PL+PRES+suBj
'He urges the townsmen to defend the walls'
suisque imperavit ne quod omnino telum in hostes
and.to.his.men ordered that.not any altogether weapon in enemies
reicerent (Caes. B.C. 1.46.2)
throw.back+3PL+iMPF+suBJ
'and ordered his men not to throw back any missile at all at the enemy'

However, for a few verbs, notably IUBEO 'I order', the accusative and
infinitive are used - see 7.4.1:
cum . . . eos . . . suum adventm expectare
since them his arrival to.wait.for+PRES+iNFiN
iussisset (Caes. B.C. 1.27)
he.had.ordered
'since he had ordered them to wait for his arrival'

There is a very similar use of the subjunctive in both Italian and

Spanish:

Gli hanno ordinato che tacesse

to.him they .have ordered that be.quiet+PAST+suBJ
'They ordered him to be quiet'
Le mandaron que les-siguiera
him they.ordered that them-follow+3sG+PRES+suBj
'They told him that he should follow them'
Wever, in both languages an infinitive may also be used:

139
Indicative and subjunctive

Gli hanno ordinato di tacere

to.him they.have ordered PREP be.quiet+iNFiN
'They ordered him to be quiet'
Le mandaron seguir-les
him they .ordered follow+iNFiN-them
'They ordered him to follow them'
This use of the infinitive construction as an alternative to the Subjunc
construction will be discussed in detail in 7.4.1.
The subjunctive is similarly used as a (notional) jussive. Examples of fi
person use in Classical Greek and Spanish are:
all' ei doki, plo:men, (Soph. />;,/;. 526
but if it.seems sail+lpL+PRES+suBj
'If thou wilt, let us sail'
Levantemo-nos
raise+ IPL+PRES+SUBJ-IPL+REFL
'Let's get up'
Examples of third person use from Italian and Spanish are:
Che venga anche lui
that come+3sG+PRES+suBj also him
'Let him come too'
Que entre
that enter+3sG+PREs+suBj
'Let him come in'
Use of the subjunctive as jussive in non-European languages has already
been illustrated for Hausa and the Bantu languages Swahili and Luvale.
Examples repeated from the introductory section of 5 and from 5.1.1 are:
haa njahen
HAAgO+lPL+SUBJ
'Let's go'
ni-pend-e
I-love-suBj
'Let me love'
va-iz-e (veze) waxi
they-come-SUBJ quickly
'Let them come quickly'
There was also some discussion of the jussive in Semitic languages in 5.4.1-
In fact, the forms that can be treated as Jussive for both formal and notiona
reasons are sometimes more like subjunctives, and are even treated by sorn
scholars as subjunctives. In particular, they occur in subordinate clauses-

140
 5.4 Imperative and jussive

for Syrian Arabic, Cowell (1964: 345ff.) uses the term 'subjunctive' to
to a form that is similar to the Amharic Jussive, which may function as
fS
'optative' (i.e., Jussive) as in:
n-rh as-sinama
lPL-go+suBJ to.the cinema
'Let's go to the cinema'
. ' su bjunctive' is used in subordinate clauses with 'overt expressions of
hortation, suggestion, wish, fear, intention, etc.'. In addition, it is found
th 'must', 'may', 'be able', 'know how to', 'forget to' and many others, not
11 of them describable as modal. Examples are
?ana baddi ?srza ial-be:t
I I.want I.go+SUBJ to.the-house
'I want to go home'
la.zam ?ufi bwaidi
must I.keep+SUBJ to.promise.my
'1 must keep my promise"
There is another respect in which this jussive (or 'subjunctive') is like the
European subjunctive - it is used as the negative of the imperative in (Cowell
1964: 345):
ruh
go+2so+iMP
'Go!'
la truh
not go+2sc+JUSS
'Don't go !'
However, there is no uniformity with the form that is most generally used
in subordinate clauses. In Tigre (Ethiopian Semitic) the jussive is used for
purpose (Leslau 1945:200):
?sttu ?3gsl tatrai
to.him in.order.that complain+3sG+JUSS
'in order to complain to him.
in
the closely related language, Tigrinya, it is the imperfect that is used
(Leslau 1941: 142):
ma?3nti kat-blla
in.order. that-eat+3F-sc+iMPF
'so that she could eat'
Amharic normally uses the imperfect in indirect commands (Cohen 1936:
304);

141
Indicative and subjunctive

snd-imatu azzaza
that-they come+iMPF he.ordered
'He ordered them to come'
However, if the superordinate verb itself is in the imperative the jussiv
6
be used (Cohen 1936: 357): "^
ygzazwaccaw yamtu
order.them they.come+juss
'Order them to come'
(This is certainly not direct speech, which would require the impera
'Come'.)
This is, perhaps, the best place to note the 'optative' mood in Nahuatl
(Aztec - Andrews 1975: 52, 384), which has marking for number and perso
and has both present and past tense. This is used for wishes, relating to both
present and past. However, Andrews adds, 'There is no special imperative
mood for the purpose of giving commands. Command sentences and
exhortation sentences are simply wish sentences.' It might, perhaps, be
suggested that this is more like a subjunctive and that it provides an example
of a language in which expressions that are treated elsewhere as jussives and
imperatives (as well as desideratives - 5.3.3) are indicated entirely by the
subjunctive.

5.5 Subjunctive as subordinator

It was suggested earlier (5.1) that the subjunctive is typically associated with
subordination. Usually there is some notionally irrealis feature involved as
well, but sometimes it seems simply to be a marker of subordination. This
became increasingly true of Latin, as shown by an example from Late Latin:
Pugnatum . . . incerto Marte, donee proelium nox
it.was.fought with, uncertain Mars until battle night
dirimeret (Tac. H. 4.35)
break off+3sc+iMPF+suBj
'The fight went on indecisively until night broke it off
There are other examples where the use of the subjunctive may seem to
lack any motivation in terms of irrealis. For instance, it is sometimes used in
Latin simply to distinguish between different kinds of subordinate clause.
The conjunction cum, for instance, can be used to mean 'when' or (causal)
'since', as indeed can English as:
As I was going to the shop, I saw John
As I was going to the shop, I couldn't wait
Latin uses the indicative and subjunctive to distinguish these two senses:

142
 5.5 Subjunctive as subordinator

Cum haec leges, consules habebimus (Cic. An. 5.12.2)

when these read+2sG+FUT+iND consuls we.shall.have
'When you read this, we shall have consuls'
Quae cum ita sint, Catilina, perge quo
which since thus be+3pL+PRES+suBJ Catiline proceed whither
coepisti (Cic Cat. 1.5)
you.began
'Since this is so, Catiline, pursue the course you have begun'

ver here it could be argued that 'since' indicates shared knowledge,

. mat this is another example of presupposition being marked by an
irrealis mood.
There is also a contrast between the moods in:
Ante quam pro L.Murena dicere instituo, pro
before that for L.Murena speak begin+lSG+PRES+iND for
me ipso pauca dicam (Cic. Mur. 1)
me self few let.me.say
'Before I begin to speak for L.Murena, let me say a little on my own
behalf
Ante quam veniat in Pontm, litteras ad
before that come+3so+PRES+suBj in Pontus letter to
Cn. Pompeium mittet (Cic. Agr. 2.20.53)
Cn. Pompeius he.will.send
'Before he arrives in Pontus, he will send a letter to Cn. Pompeius'

However, the problem is not the use of the subjunctive in the second, but the
use of the indicative in the first (both refer to the future - see 5.2.6). It may be
that it is the immediacy of the event that warrants the indicative.
Similarly for Fula, Arnott (1970: 3056) indicates that the subjunctive is
used not only with verbs of (i) enjoining, (ii) wishing, (iii) fearing, taking
care, (iv) requesting, (v) permitting or agreeing, (vi) causing or arranging,
but also for other types of clause. Thus (Arnott 1970: 310-11), it is required
w
ith haa 'until', ooke 'before', etc., even though the clauses are notionally
realis:

6e-ngac).ay ka remuki haa 6e-timmina

'They continue farming until they finish'
ooke 6e-njottoo, 'o-'yami gorko 'on . . .
'Before they arrived, he asked the man . . .'

^uriilarly, in a very different language, Mangarayi (Australia - Merlan 1982:

'8, 184), there is a prefix that is labelled 'irrealis', which is used for epistemic
Possibility, but which is also often the only marker of subordination:

143
Indicative and subjunctive

a-jiani-yug
IRR-talk-AUX
'He might talk'
gawa-j muyg jan? ya-ma-ji
bury-PAST-puNCT dog die IRR-AUX-PASTPUNCT
'He buried the dog when it died / that died'

5.6 Other possibilities

So far the contrast has been between one set of indicative forms and one s
of subjunctive forms. But it is perfectly possible to allow that there might h
a further system within either of the two moods.
In Swahili (and, indeed, other Bantu languages) the subjunctive is not
marked for tense-aspect as the indicative is. There is a single set of forms for
subjunctive, but other sets for present continuous, past, future, consecutive
conditional, etc. Swahili does not appear to have any modality systems, such
as those of English or the evidential systems. It seems that there is here the
converse of what is found with modal systems in that there is one marker of
irrealis, but a system of markers of realis.
There is what might seem to be the opposite situation in Maasai (East
Africa - Tucker and Mpaayei 1955: 61 ff., 96ff.), in that, for subordination at
least, there are several types of subjunctive, although the choice of marker is
not free but determined by the subordinating verb. These different forms are
labelled 'subjunctive (a)', 'subjunctive (b)', 'infinitive (a)', 'infinitive (b)' and
the 'N-tense' and there are (at least) five classes of predicator:

(i) 'Le takes either of the two subjunctives,

(ii) Verbs of wanting, liking, disliking, etc. (including 'ought'), take
the N-tense.
(iii) Infinitive (a) is used mostly after verbs of motion, but also after
verbs of liking, 'help', 'do again', 'do in the near future', 'start',
'repeat', 'do early in the morning',
(iv) Infinitive (b) is used with 'get to do', 'know how to', 'finish', "be
able to', 'dare',
(v) Either infinitive (b) or the present tense is used after 'refuse', 'b e
afraid', 'forget', 'lack'.

144
r
6
Realis and Irrealis

. w a s noted in 1.2.1 that, in many languages, particularly the Native

American languages and those of Papua New Guinea, mood is described in
terms of the grammatical markers of realis and irrealis. Although the
distinction is basically the same as that between indicative and subjunctive,
both being markers of the typological categories Realis and Irrealis, there are
sufficient differences to warrant discussion of them in a separate chapter. Yet
there is not always a clear distinction between them - see 7.1 for discussion.

6.1 Joint and non-joint marking

There are basically two ways in which realis and irrealis markers function. In
some languages their main function is to co-occur with other grammatical
categories. In others they mainly occur in isolation and are themselves the
only markers of specific notional categories.
For instance, in Amele (Papuan - Roberts 1994: 372) an irrealis marker is
required whenever a future marker is present in the sentence:
ho bu-basal-en age qo-qag-an
pig siM-run.out-3sG+DS+iRR 3PL hit-3PL-FUT
'They will kill the pig as it runs ou
By contrast in another Papuan language, Muyuw (Bugenhagen 1994: 18,
quoting a personal communication), the irrealis marker is itself the indication
ffuture:
yey b-a-n Lae nubweg
I iRR-lsG-go Lae tomorrow
'I will go to Lae tomorrow'
nctly, these might seem not to be comparable in that, in the Amele
a
mple, there is a syntactic relationship between two grammatical markers,
ea
'is and future, whereas in the Muyuw example there is simply a specific
ail
wnatical marker, irrealis, that indicates futurity. Yet from a typological
lr,
t of view, they are similar, in that the common notional feature of
145
Realis and Irrealis

futurity is associated in both languages with the grammatical cateenr

& l
,. y of
irrealis.
These different functions of realis and irrealis will be distinguished in te
of being 'joint' and 'non-joint' markers, the joint markers being those th
co-occur (obligatorily) with other grammatical markers.
However, there are two important reservations. First, the distinction is n
clear-cut, in that both features often appear in a single language, althou h
the mood system of any one language usually contains markers that ar
predominantly either joint or non-joint. It will, therefore, sometimes K
convenient to refer to the systems themselves as either 'joint' or 'non-joint'
Secondly, non-joint systems are often not strictly binary, in that construc
tions may not only be marked as realis or irrealis, but may also be unmarked
(see 6.5.2). More problematic for the realis/irrealis distinction is the fact that
in some languages, realis and irrealis markers occur together with other
markers in even wider systems, which, prima facie, look more like modal
systems than instances of mood (but see 6.5.3).
For Caddo, spoken in Oklahoma, Chafe (1995: 3519) reports that
pronominal prefixes of the verb may be distinguished as either realis or
irrealis and that the choice is determined by sets of grammatical markers that
occur before these pronominal prefixes. Irrealis prefixes are used in conjunc
tion with grammatical markers indicating negation, prohibition, obligation,
conditional, and also in conjunction with 'simulative', 'infrequentative' and
'admirative' prefixes. (In addition to the 'simple negative' prefix, there is also
a 'temporal negative' expressing 'not for a long time', and, in addition to the
conditional, a 'generic conditional' and a 'negative conditional'):
ky-a-yibahw
NEG-1 + AG+IRR-see

'I don't see him'

ka-sah?-yibahw
PROH-2+AG+IRR-See
'Don't look at it'
kas-sa-ny?aw
OBL-3+AG+IRR-Sng

'He should/is obliged to sing'

hi-t'a-yibahw
COND-1+AG+IRR-See
'If I see it'
dy-a-yibahw
SIMULAT-1 + AG+IRR-see
'As if I saw it'

146
 6.1 Joint and non-joint marking

ws-a-yibahw
INFREQ-1 + AG+IRR-See
'I seldom see it'
hs-ba-?asa-yik'awih-sa?
ADM-1 +BEN+iRR-name-know-PROG
'My goodness he knows my name'

Realis prefixes, it is implied, although Chafe gives no examples, occur with

oressions that refer to the present or past states and events. They also
ccur with two markers of future ('future' and 'future intention'):

ci-yibahw-?a?
1 +AG+REAL-See-FUT
'I'll look at it'
ci-yibahw-a?
1 + AG+REAL-See-FUT.INT
'I'm going to look at it'

However, both realis and irrealis each have one function in which they
occur alone, with no other grammatical marker, and are thus non-joint. A
non-joint irrealis prefix indicates a (yes-no) interrogative:

sah?-yibahw-nah
2+AG+IRR-See-PERF
'Have you seen him?'

A non-joint realis prefix indicates an imperative:

yah?-yibahw
2+AG+REAL-see
'Look at it'

The Caddo system is, then, predominantly, but not entirely, joint.
A non-joint system is found in Manam (Papuan - Lichtenberk 1983,
discussed in Bugenhagen 1994: 9-11). The system here is binary - Lichten
berk (1983: 181) states 'The realis-irrealis opposition is crucial to the Manam
v
erbal system in the sense that every finite verb must be specified (by means
r a subject/mood prefix) for one of the two moods.' Realis is used for (i) past
events, (h) present events and (iii) habitual events:

u-no?u
lSG+REAL-jump
'I jumped'
lira i-pura-pura
rain 3sG+REAL-come-RED
'It is raining'

147
jvcrtto Linu 11 i c u t u

?i-zen-zrj
lPL.EXC.REAL-cheW betel-RED
'We (habitually) chew betel-nuts'
Irrealis is used for (i) future events, (ii) commands, exhortations
warnings ('lest'), (iii) counterfactual events and (iv) sequences of custorna
or habitual activities:
si n-gu mi-sa?-i
loincloth POSS-ISG lsG+iRR-wash-3so+OBj
'I will wash my loincloth'
go-moan?o
2sG+iRR-eat
'Eat'
nu u-rre nora bo?an-be go-pra
ISG+IRR lsG-want yesterday SIM-FOC 2sG-come
'I wish you had come yesterday'
ma ga-?orin-i ?i ga-plo-0
garden lPL.EXC+lRR-clear-3sG+OBJ tree lPL.EXC+iRR-fell-3PL+OBj
'We clear (the bush for) a (new) garden, we fell the trees'
There is one important point to be noted here. When a marker is joint, i.e.,
occurs with another grammatical marker, it is redundant in that it adds
nothing to the notional features associated with that other marker, apart
from characterizing them as realis or irrealis. By contrast, non-joint markers
are the sole indicators of the notional features and so are not redundant.
However, it is seldom the case that all the terms in a system are joint (and
redundant) or non-joint (and non-redundant). Moreover, occasionally either
realis or irrealis may occur with the same grammatical marker (e.g., with
Future in Central Porno see - 6.5.1), with a difference of meaning, and so
non-redundantly, although joint. There is a debate concerning redundancy,
which will be discussed in 7.3.

6.2 Terminology
Although they are transparent, it is, perhaps, a little unfortunate that the
terms 'realis' and 'irrealis' have been adopted as grammatical terms in pla ce
of the traditional terms 'indicative' and 'subjunctive'. Bybee et al. (1994: 23oj
report that the earliest use of the terms in the relevant sense is not to
found in their corpus before 1970, the earliest instance being in the analyslS
of the Australian language Maung by Capell and Hinch (1970).
It might have seemed wise, in the interests of consistency, to use only n
set of terms, either 'indicative' and 'subjunctive' or 'realis' and 'irrealis
both sets of data, preferably the former. However, this is probably imp r a

148
 6.2 Terminology

al now that both traditions are firmly established, and, moreover, there is
ffle benefit in dealing with the two types of phenomena separately (with
heir different terminologies, and in separate chapters),
fhe result of this, however, is that there is, unfortunately, a problem with
h e terminology in that 'realis' and 'irrealis' can be used in at least three

senses.
First, the terms may be used to refer to the notional features, rather than
'factual' and 'non-factual', or 'real' and 'unreal' (see 1.1.1). It may be noted
that Givn (1994) uses these terms in his analysis of the Spanish subjunctive.
Secondly, the terms are used by linguists as labels for the formal grammatical
categories. Thirdly, if indicative/subjunctive and realis/irrealis are seen as the
grammatical markers of mood, it would be appropriate to recognize 'Realis'
and 'Irrealis' as the relevant typological (cross-language) categories.
As argued in 1.7, there is no problem about the terminology to be used to
distinguish a typologically valid category from the notional features associ-
ated with it, since typological categories can be indicated by the use of initial
capitals. Thus the typological grammatical categories of Realis and Irrealis
are associated with the notions of realis and irrealis. Confusion between the
terms as names of the language-specific categories and as labels for the
notional features is possible, but where necessary can be avoided by talking
of'notionally realis' and 'notionally irrealis'.
There is another point, which was briefly discussed in 1.5. Typologically
the categories in joint and non-joint systems discussed above in 6.1 are very
similar, and it is convenient to use the same terminology for them. Thus,
although 'future' is not distinguished grammatically from 'command' and
'counterfactual' in Manam (6.1), it is still reasonable to treat it as an instance
of the typological category Future. Commonly, however, writers express the
facts by saying that categories such as future 'are' realis or irrealis, either
because of their co-occurrence with realis or irrealis markers or because they
af
e signalled by the realis or irrealis markers. This clearly mixes the
grammatical and notional levels, but may be allowed because it is a fairly
obvious and instructive way of combining the facts of both joint and non-
joint systems.
There is one further point. The term 'irrealis' is used in the description of
s
orne languages as the name of one of the terms in a modal system, where it
Would be more appropriate to use a term such as 'dubitative'. An example
can
be found in the account of Ngiyambaa (Donaldson 1980: 160, 162)
here 'irrealis' occurs in the same system as 'purposive' which is used both
e
ntically and epistemically, and in its epistemic use is essentially Deductive
(see
1.4.5, 2.1.2 and 4.1):

149
Realis and Irrealis
Irrealis
yurur)-gu nidja-1-aga
rain-ERG rain-CM-iRR
'It might/will rain'

Purposive
rjadhu bawurj-ga yuwa-giri
I+NOM middle-LOC lie-PURP
'I must lie in the middle'
yururj-gu rjidja-1-i
rain-ERG rain-CM-PURP
'It is bound to rain'
This is not relevant for the treatment of irrealis as mood. The term is used
here to refer to a term in a modal system (rather than in a system of realis/
irrealis mood).

6.3 Morphology and syntax

6.3.1 Morphology
Often realis and irrealis markers are single isolated forms, whether words,
clitics or affixes. This is so in Kiowa (Oklahoma - Watkins 1984: 171) in
which there is a single suffix to:
hgy -b-to.
perhaps lsG-go-iRR
'Maybe I'll go'
Similarly in Maricopa (see 6.5.3) the marker is -ha:
ny-aay-ha
1/2-give-lRR
'I will give it to you'
There is a similar situation with the non-joint markers in the Papuan
languages. According to Bugenhagen (1994: 36) the markers in eight such
languages are verb prefix for subject and mood, pronoun for subject and
mood, verb prefix (for mood only), preverbal particle and sentence-initial
particle.
In some languages, however, there are markers that are cumulative
exponents of several grammatical categories. Thus in Caddo (Oklahoma -
Chafe 1995: 352-3), there are prefixes that are simultaneously markers of 0)
person, first, second, third and 'defocussing', (ii) the grammatical relations
(see Palmer 1994) agent, patient and beneficiary, as well as (iii) realis and
irrealis. Part of the whole system is:

150
 6.3 Morphology and syntax
Realis prefixes
agent patient beneficiary
First person ci- ku- ku-
Second person yah?- si- si-
Irrealis prefixes
First person ta-/ti- ba- ba-
Second person sah?- sa?a- sa?u-

There are, in fact, forty-eight possibilities. There is some syncretism,

however, such as that shown by the formal identity of the first three pairs of
atient and beneficiary above. This is, perhaps, striking in that cumulative
exponence of this kind is somewhat reminiscent of the subjunctive (see 7.1).
Even more like indicative/subjunctive is the system described for Takelma
(Southern Oregon) by Sapir (1922: 94), and discussed by Mithun (1999: 5).
Here there are six 'tense-modes' - 'aoris, 'future', 'potential', 'inferential',
'present imperative' and 'future imperative'. Each verb has two stem forms
yowo- and yu- for 'run', baxam- and baxma- for 'come', etc. The first stem, 'a
derived or amplified form of the base', is used for the aorist, which is the
most frequently used of the tense modes. The second stem, 'the generally
more fundamental verb-stem', is used for the 'less important tense-modes'.
Sapir's examples are, unfortunately, far from transparent, e.g.:
Aorist tlomom 'He killed him'
Imperative doum 'Kill him'
Now the aorist is used to refer to past events, the present and the immediate
future, and is clearly realis, while the others are all potentially irrealis. The
realis/irrealis distinction is marked, then, in the choice of verb stem. This
kind of redundant marking is not at all uncommon in inflected languages. In
Latin, for instance, there are different stems for the non-perfect and perfect
tenses and moods, e.g. am- for non-perfect and amav- for perfect with the
verb 'love', and, more strikingly, fer- and tul- non-perfect and perfect with
'carry', even though the person-number paradigms are also different. (For
further discussion, see 7.1.)
A further possibility is that there is double marking, as in Alsea, once
s
Poken in Oregon (Buckley 1988, quoting an unpublished manuscript,
Frachtenberg 1918). Here the distinction is marked both by complementizers
~ realis mis and irrealis sis - and, according to Mithun (1999: 175), by affixes
tn
at are cumulative exponents of Realis/Irrealis and Inchoative/Completive
(Page references to Buckley):
temhu mis-axa wi'J-x (12)
and then REAL+coMPL-back come-REAL+coMPLET
'And then, after she came back'

151
Realis and lrrealis
mis-i/bc mi^hu' X.q'-s-t-ax
(19)
REAL+COMPL-3PL+SUB at last CrOSS-?-STAT-REAL+COMPLET

'When they finally got across'

sips tqaild-i (21,

IRR+COMPL+2PL+SUB Want-IRR+COMPLET
'If you desire it'

i'mst tm-in tqailt-9X sis

t h u s and-lSG+SUB Want-REAL+COMPLET IRREAL+COMPL
kexk-ai'-m
C3)
assemble-iRR+iNCH-iNTR
'For that reason, I want [the people] to reassemble'

There are two other possibilities. First, there are languages in which there
are realis and irrealis markers and, in addition, constructions that are
unmarked (discussed in 6.5.2). Secondly, Realis may be unmarked (as it
usually is where there are modal systems of the type presented in Chapters 2
and 3). Possible examples of this are Tolkapaya Yavapai and Mojave, which
are discussed in 6.3.3 and 6.6.5.

6.3.2 Linked clauses

In the Caddo examples presented by Chafe (6.1), the joint markers occurred
not only within a single clause, but also within a single word. One repeated
example (for Negative and Irrealis) is:
kuy-t'a-yibahw
NEG-1 +AG+IRR-See
'I don't see him'
However, there are both Native American languages and Papuan languages
in which these joint markers are associated with separate clauses of a
complex sentence.
One such language is Central Porno (Pomoan, N. California - Mithun
1999: 176-7, cf. Mithun 1995: 368-73), which also has a modal system (see
1.2.1, 2.2.3). Here the realis/irrealis distinction is made in linked clauses (of a
single type, but where the links would be translated as 'and', 'while', 'then,
etc.). In the first clause the verb has markers that are cumulative exponents ot
(i) realis/irrealis, (ii) 'same'/'different', which shows whether the linked
clauses refer to the same or to different events, (iii) for realis only,
'simultaneous'/'sequential' to distinguish between simultaneous and sequen-
tial events. The co-occurring grammatical category that determines tne
choice of realis and irrealis is in the final clause. Thus, in the examples below,
the categories of imperative and conditional occur in the second clause an

152
 6.3 Morphology and syntax

e determine that the verb in the first clause is marked as irrealis

e+jRRealis and different+iRReahs respectively):
(sarnc
qh hni'-?el d--hi m'tya-1
water bread-the make-SEMEL-SAME+iRR 3PL-PAT
qa'-w-''-ka-m
bitng-gO-IMPF+PL-CAUS-IMP
'Make the water bread and invite them to eat it'
men mi-hla, mu'l ?a' ? h ? le ?a'
so say-DiFF+iRR that 1+AG sit-C0ND 1+AG
'//"she said that, I'd stay longer'
The next examples show markers of present and past in the second clause
and the marking of realis (same+simultaneous+realis and different+sequen-
tial+realis) in the first clause:
?a' Edna-to -'l y-'h-du- hntil-?el
1+AG Edna-PAT house-to go-PERF-iMPF-SAME+siM+REAL Indian-the
a'n-'d-an-ya mu'tu
talk-IMPF-IMPF-PERS.EXP 3 + PAT
'I go to Edna's house and talk Indian to her'
to' met ?-n'-ya-w-li mya' m
1+PAT such by.gravity-set-DEFOC-PERF-DiFF+SEQ+REAL 3+POSS father
d-'?-du-w h-w
like-REFL-IMPF-PERF nOt-PERF
'I was nominated and his father didn't like it' ('When I was nominated,
his father didn't like it')
There is a similar situation in the Papuan language Amele, as described by
Roberts (1990: 371-5). However, one important difference is that, although
realis/irrealis marking occurs with linked clauses ('clause chaining'), it does
so only where the clauses have different subjects. Again, the verb of the first
('medial' in Roberts's terminology) clause is marked as realis/irrealis (cumu
latively by a marker for realis/irrealis, number and person and 'different
subject'), while the verb of the final clause is marked for the co-occurring
grammatical category (cumulatively with number and person). The following
examples illustrate realis marking (together with person and number and
different subject) on the verb of the first clause and markers of past and
Present (plus person and number) on the verb of the final clause:
ho bu-busal-en age qo-in
pig SIM-run.OUt-3SG+DS+REAL 3PL hit-3PL+REM.PAST
'They killed the pig as it ran out'
ho bu-busal-en age qo-igi-na
pig SIM-run.OUt-3SG+DS+REAL 3PL hit-3PL-PRES
'They are killing the pig as it runs out'

153
Realis and Irrealis

Similarly, other examples show irrealis marking in the medial clause anri
future and imperative in the final clause:
ho bu-busal-eb age qo-qag-an
pig siM-run.out-3sG+DS+iRR 3PL hit-3PL-FUT
'They will kill the pig as it runs out'
ho bu-busal-eb age qo-ig-a
pig siM-run.out-3sG+DS+iRR 3PL hit-3PL-iMP
'Kill the pig as it runs out'
These examples show how mood is marked on the first of the linked clauses
(the 'medial' clause), but is dependent on the grammatical marker in the
second (the 'final' clause). (For a full list of categories involved, see 6.4.)
The translations of the linked ('chained') clauses suggest that they include
both oblique subordinate clauses (temporal, purpose, conditional, etc.) and
coordinate clauses with 'and', 'but', etc. In Amele, at least (but probably in
other languages), these two types of clause can be distinguished syntactically
in that a subordinate medial clause can be moved - either embedded in the
final clause or transposed after it (Roberts 1994: 13-15). This can be seen
from:
ho qo-qag-an nu dana age h-oiga-a
pig hit-3+PL-FUT PURP man 3PL come-3PL-TOD.PAST
'The men came to kill the pig'
The subordinate ho qo-qag-an nu 'to kill the pig' can be transposed:
dana age ho qo-qag-an nu h-oiga-a
man 3PL pig hit-3+PL-FUT PURP come-3PL-TOD.PAST
'The men came to kill the pig'
dana age h-oiga-a ho qo-qag-an nu
man 3PL come-3PL-TOD.PAST pig hit-3+PL-FUT PURP
'The men came to kill the pig'
By contrast no transposition is possible with the coordinate:
ho busale-i-a qa dana age qo-iga-a
pig run.out-3sG-TOD.PAST but man 3PL hit-3PL-TOD.PAST
'The pig ran out, but the men killed it'

6.3.3 Complement clauses

The last section dealt with a special kind of complex sentence, in which two
clauses are involved in the determination of mood marking, but realis/irrealis
marking is also to be found in more familiar types of subordinate clause,
both oblique clauses and complement clauses (in the same way as indicative/
subjunctive). Thus in Mojave (Yuman, Arizona - Munro 1976: 54-5) the

154
 6.3 Morphology and syntax

vrealis suffix appears on 'subordinate verbs to mark unreal or hypothetical

ituations: desires, conditionals, obligations, counterfactuals and occasional
f itures'. Examples of an oblique and a complement clause are:
huwalyapay ya-?-a?a;v-9 ?-?ahot-e
Walapai x-1-understand-iRR 1-good-FUT
'If I learn Walapai, I'll be all right'
humi'c ?atay iyu'-h-a'r-pc
children more have-iRR-want-TENSE
'She wants to have more children'
A similar pair from Alsea, once spoken in Oregon (Buckley 1988 and
frachtenberg 1920) was given in 6.3.1:
sips tqaild-i (24)
IRR+COMPL+2PL+SUBJ Want-IRR+COMPLET
'If you desire it'
.mst tm-in tqaialt-sx sis
thus and-lsG+suB WANT-REAL+COMPLET REAL+COMPL
kexk-ai'-m (42)
assemble-iRR+iNCH-iNTR
'For that reason, I want [the people] to reassemble'
There are plenty of examples of similar complement clauses in Papuan
languages. In a table showing the functions of irrealis in these languages,
Bugenhagen (1994: 36) indicates that, of the six languages with extensive
mood marking, all use irrealis for 'want' and 'lest', while five use it for
'purpose', 'ability' and 'obligation'. Examples of 'want' from Manam and
Sursurunga are:
tam-gu i-rre go-pra zma
father-ISG 3SG+REAL 2so+iRR-come tomorrow
'My father wants you to come tomorrow'
iau nem ngo i-na han ur Ukarumpa
ISG+REAL want that ISG-IRR go to Ukurumpa
'I want to go to Ukarumpa'
Here, clearly, 'want' is expressed with the use of an irrealis complement
clause. The same is true of ability and weak obligation, as well as wanting, in
Ma
ngseng (although with ability the irrealis marker is optional):
i pavurvu (aro) i ro vait
3SG able (IRR) 3SG go.up coconut
'He is able to climb coconut trees'
i sivna ako aro i oma i
3SG good that IRR 3SG do 3SG
'He should do it'

155
Reads and irrealis
tho n-ve aro te oriet pua ponange
ISG say-thus IRR we hunt crocodiles today
'I want us to hunt crocodiles today'
However, the status of the forms is less clear in Muyuw (Bugenhagen 1 QQ
19, quoting a personal communication), where examples of 'want' 'ahi-
i < < , , ' "ity'
and obligation are:
sivina-g b-a-n lae
need/desire-1 so iRR-lso-go Lae
'I want to go to Lae'
kadiloka b-i-vag
ABILITY IRR-3SG-do
'He can do it'
awoum b-i-vag
PROH IRR-3SG-do
'He must not do it'
There is no problem with 'want', but a possible analysis of the other two
examples is not in terms of complements of the markers of ability and
prohibition, but in terms of joint marking of irrealis and grammatical
markers for ability and prohibition.
Complement clauses can also be found in conjunction with mood-marked
linked ('chained') clauses. For Amele, Roberts (1990: 373-4) lists five types
of complement subordinate clause that also require irrealis marking -
intentive, desiderative, habitual desire, abilitative, purpose. The subordinate
clauses are marked in several different ways - with 'relative future', impera-
tive and infinitive:
ho bu-basal-eb age qo-qaq-a bili tawe-ig-abe
pig siM-run.out-3sG+DS+iRR 3PL hit-3pL-REL.FUT be stand-3PL-TOD.PAST
'They stood about to kill the pig as it ran out'
ho bu-basal-eb age qo-ig-a ade-o-na
pig sim-run.out-3sG+DS+iRR 3PL hit-3PL-iMP 3PL+OBJ-3SG-PRES
'They want to kill the pig as it runs out'
ho bu-basal-eb age qo-oc nu ihoc
pig siM-run.out-3so+DS+iRR 3PL hit-iNFiN for able
'They are able to kill the pig as it runs out'
It is important to understand that these subordinate complement clauses
themselves are not marked as irrealis in the way that in other languages
complement clauses are marked with subjunctive. Rather, it is the markers oi
subordination in these clauses (relative future, imperative and infinitive) that
determine that the previous ('medial') clause is marked as irrealis. (Notice
too that the markers in the main clause are irrelevant in this respect - in the

156
 6.4 Joint systems
two they are future and present, but the 'medial' clause is marked as
firs'
;rrealis)
renerally the use of irrealis in a subordinate clause is redundant in that it
determined by the construction and allows no contrasts of meaning. An
resting exception is to be found in Mojave (Yuman, Arizona - Munro
074- 54-5)- The distinction between (unmarked) realis and irrealis is used to
. tjnguish wanting, an attitude towards a notionally irrealis proposition,
j ]j|dng, an attitude towards a notionally realis proposition:
?-isay-0 ?-a:r-mot-m
1-fat-iRR I-want-NEG-TENSE
'I don't want to get fat'
?-isay-k ?-a:r-mot-
1-fat-ss l-want-NEG-STAT
'I don't like being fat'

6.4 Joint systems

It is, perhaps, rather surprising that the two Native American joint systems
that have been discussed in detail, Caddo (6.1) and Central Pomo (6.3.2) are
very different in two important ways. First, as shown in 6.3.2, the syntax is
very different. The co-occurring forms in Caddo are within a single clause
(and in all the examples given, within a single word), whereas in Central
Pomo there are linked clauses with the mood marker being in one single
clause, and the co-occurring grammatical marker in another. In this respect,
Central Pomo is rather like some of the Papuan languages, as described by
Roberts (1994: 36). Secondly, there is a difference in their treatment of some
of the grammatical categories. Those illustrated for Caddo are:

Realis: future, future intention, imperative

Irrealis: negation, prohibition, obligation, conditional,
simulative, infrequentative, admirative, temporal negative,
generic conditional and negative conditional.
The categories illustrated for Central Pomo were:
Realis: imperfective, perfective
Irrealis: conditional, imperative, future (basically).
(Negative and Interrogative do not affect mood in Central Pomo, while
uture is usually, but not always, marked as irrealis. Notice that future and
'imperative are realis in Caddo - see the relevant sections of 6.6.1 and 6.7.1.)
Takelma (6.3.1) is, perhaps, to be regarded as another Native American
ail
guage with a joint system. If so, it would seem that realis is unmarked
157
i^euiii una rreuiis
since the (realis) 'aoris has the basic stem of the verb. It has a joint system '
that a set of categories occurs with the derived (irrealis) stem. The categor'
would then be:

Realis: present/past ('aoris)

Irrealis: future, potential, inferential, present imperative and
future imperative

These three Native American Indian languages differ considerably from

each other, but in many of the Papuan languages the realis/irrealis system i
more homogeneous with fairly predictable sets of obvious realis and irrealis
categories. Examples from Amele were given in 6.3.2. For that language
Roberts (1990: 275) lists the grammatical categories that are associated with
realis and irrealis as:

Realis: habitual past, remote past, yesterday's past, today's past

present tense
Irrealis: future, imperative, hortative, prohibitive, counterfactual/
prescriptive, apprehensive.

He provides a similar set of examples from Nobonob, Anjam and Bargam

(all from unpublished manuscripts). A pair of sentences from Nobonob
(Roberts 1990: 280) is:
ah ag e he-egeg danab lag lag qag-pig
woman 3PL food do-3pL+DS+siM+REAL man 3PL house tie-3PL+PAST
'As the women cooked the food, the men roofed the house'
ah ag e he-bepeg danab lag lag qag-kulag
woman 3PL food do-3pL+DS+siM+iRR man 3PL house tie-3pL+FUT
'As the women cook the food, the men will roof the house'
He notes that these three languages are unrelated, though all from the same
area (the Madang Province of Papua New Guinea), but he also provides
examples from languages in other, far removed, areas - Wojokeso (West
1983), Angaataha (Huisman 1973) and Gahuka (Deibler 1976). Roberts
(1990: 382-3, quoting an unpublished manuscript by D. Rucker) also gives
examples from Anjam, where present, immediate past and remote past are
marked as realis, while future, imperative and counterfactual are marked a
irrealis. Examples of remote past and future are:
e tabir yans-eqn-a-m Rut
lSG dishes WOrk-SIM+REAL-REM.PAST-lSG+REAL+DS R u t h
alarj-oqn-e-j
play-CNTF-REM.PAST-3SG
'While I washed the dishes, Ruth played'
158
 6.4 Joint systems

a wan-oqn-i-m naygi b-q-ab

3SG WOrk-SIM+IRR-FUT-3sG+IRR+DS 3PL COme-FUT-3PL
'He will be working when they come'
However, there is an obvious difference here - that the medial clauses as well
the final clauses contain the grammatical markers that determine the
|j s /i r realis distinction (remote past and future).
It is quite apparent that there is a close correspondence between the
otional and formal categories. Only propositions indicating past and
esent eV ents are treated as realis, for these propositions are asserted;
oropositions indicating future events and all unrealized, potential, events are
treated as irrealis, for these are not asserted. The unrealized, potential, events
are those indicated both by imperative, hortative, prohibitive, counterfactual/
prescriptive, apprehensive in main clauses and by intentive, desiderative,
habitual desire, abilitative, purpose in subordinate clauses.
There is one other curious point. Amele has a number of sentence-final
particles. Some of these occur only with irrealis final verbs (those that
determine irrealis in the medial verb), but others can occur with realis final
verbs, and, in spite of being notionally irrealis, do not affect the mood of the
medial verb. Thus with 'maybe' or the particle for question, the medial verb
still has realis marking if the final verb is realis (e.g., present tense):
ho bu-basal-en age qo-gi-na fa
pig SIM-run.OUt-3SG+DS+REAL 3PL hit-3PL-PRES DUB
'Maybe they are killing the pig as it runs out'
ho bu-basal-en age qo-gi-na fo
pig SIM-run.OUt-3SG+DS+REAL 3PL hit-3PL-PRES QUES
'Are they are killing the pig as it runs out?'
Other such particles are translated as 'nevertheless', 'let them', 'really' and
'always'. Roberts (1990: 379-88) reports that studies by other scholars,
mostly in unpublished manuscripts, show a similar situation in Nobonob,
Anjam, Bargam and Wojokeso (West 1983) and probably several other
languages.
In the Papuan languages, Future is always one of the categories marked by
'rrealis in joint systems (and in non-joint - see 6.5.1). There is some variation
w
'th the other potentially irrealis categories. A table presented by Roberts
' "94: 31) shows that where there is marking of irrealis on the medial verb
"h chained constructions, the marker indicates Future only in two
n
guages; Future and Imperative in five; Future, Imperative and Counter-
ct
ual in four; and Future, Imperative, Counterfactual and Habitual Past in
e
- (Where the only irrealis category is Future, there is, of course, no way of
0v
>ng that it is a marker of mood rather than tense.)

159
Realis and Irrealis

6.5 Non-joint systems

6.5.1 Binary systems

Although it is clear that realis and irrealis occur non-jointly to indicate
directly, contrasts such as present and past vs future, etc. (see 6.1 f. r e
there are few, if any, languages where there is a simple binary contra
realis and irrealis. This is hardly surprising, for such a binary contrast w
allow for a great deal of ambiguity. Most languages with non-joint mark'
have other devices to make further distinctions. In fact, most languages w' i,
extensive non-joint marking do not have binary systems of mood at all
will be clear from 6.5.2, 6.5.3 and 6.5.4.
However, one language that appears to have a purely binary contrast i
Manam (Papuan - Lichtenberk 1983: 182-91, discussed in Bugenhagen 1994.
9-11), which was exemplified in 6.1. Lichtenberk, it will be recalled, states that
'every finite verb must be specified . . . for one of the two moods'. He suggests
that realis is used for (i) past, (ii) present and (iii) habitual events, while irrealis
is used for (i) future events, (ii) commands, exhortations, warnings ('lest'), (iii)
counterfactual events and (iv) customary or habitual activities.
Yet irrealis frequently occurs in conjunction with particles to mark further
distinctions. Thus, while in isolation it will indicate a certain or near future
(see 6.1), together with the form masa 'perhaps' it indicates an uncertain
future:
tanpwa masa bul?a rja-ma?-i
chief perhaps 3sG+iRR-make-3sG+0BJ
'The chief will give a feast'

Together with the 'prospective' form ?na it is said to correspond to English

'going to':
?si n-gu mi-sa?-i ?na
loincloth POSS-ISG lsG+iRR-wash-3sG PROSP
'I am going to wash my loincloth'
Moreover, when the prospective form is preceded by abe 'already', ll
expresses imminence, and when preceded by a form meaning 'very (very)
near' it expresses 'the fact that an event almost took (has taken) place, bu
did (has) not'. There are other forms to indicate i e s and ability/permission,
and the irrealis form is also used in subordinate clauses after 'want'.
There is a potential problem here in that it could be argued that, sine
these forms regularly occur with irrealis, this is a joint system rather than
non-joint system (where the irrealis marking would be redundant). Howeve ,
it is clear that the basic distinction is made contrastively, with no otn

160
 6.5 Non-joint systems

ugj-s - between realis past and present and irrealis future. The system is,
efore, basically non-joint and it is more appropriate to see these co-
rring particles as having the function of 'fine-tuning' the basic distinc-
Motionally and typologically they are very different from the gramma-
1 categories that co-occur with irrealis in joint systems.

6.5-2 Realis, irrealis and unmarked

typical characteristic of mood is the binary system of Realis and Irrealis.
However, in some languages, although there is a contrast stated in terms of
a lis and irrealis, there are also forms that are unmarked for mood.
This is inevitably so in the Papuan language Amele, because mood is
marked only on linked constructions with different subjects, as illustrated in
6 3 2. Where there are the same subjects, the construction is unmarked for
m ood (Roberts 1994: 11):

age cabi na beli-me-ig ceta ceh-eig-an

3PL garden to.go-ss-3PL yams plant-3PL-YEST.PAST
'They went to the garden and planted yams (yesterday)'
There is no problem with this; it is enough to say that the realis/irrealis
contrast is restricted to certain constructions, and that in such constructions
there are still only the two possibilities.
In some languages there is a choice, in some constructions, between
unmarked and irrealis. Thus in Alamblak (Papua New Guinea - Roberts
1990: 390-1, quoting Bruce 1984), both the imperative and the hortative (i.e.,
the jussive) are used without a mood marker, but may also be used with
irrealis. Examples of the imperative are (see 6.7.1 for further details):
nuat wa-ya-n-t
sago.patty iMP-eat-2sG-3F.SG
'Eat the sago patty!'
(nik) wa-roh-twa-k
(2PL) IMP-St-FUT+IRR+IMP/HORT-2PL
'You all be seated'
Similarly, in Muyuw (Bugenhagen 1994: 18, quoting a personal communi
cation), present and past are either unmarked or marked as realis, while
uture is marked as irrealis:
nov i-weiy buluk
yesterday 3sc-kill pig
'Yesterday he killed the pig'
buluk nov bo n-ei-weiy
pig yesterday EMPH REAL-3SG-kill ... ..
'I know he really killed a pig yesterday' ; .

161
Realis and Irrealis
yey b-a-n Lae nubweig
I iRR-lsG-go Lae tomorrow
'I will go to Lae tomorrow'
(But it is not stated whether the presence of the emphatic marker determine
the realis marking, so that it is not entirely clear whether the choice of
unmarked or realis is determined or not.)
Slightly different is the situation in which there are three markers, but
where two of them might appear to be realis and irrealis. For Dani, another
Papuan language, Foley (1986: 163) suggests that there is a three-term
contrast - which he calls 'real', 'likely' and 'potential'. His examples are
taken from Bromley (1981: 28, 47), but with some reinterpretation:
wat-h-i
kill-REAL-lSG+AG
'I killed him'
was-o-ik
kill-likely-ISG+AG
'I will kill him'
wa?-l-e
kill-POT-lSG+AG
'I may kill him'
There is a similar situation in the Northern Iraquoian languages (New
York, Ontario, Quebec - Chafe 1995: 359-60). There is a three-way division
in Seneca:
wa?-ke-ke-?
FACT-1 AG-See-PUNCT
'I see/saw it' (direct perception or memory)
e-wa?-ke-ke -?
FUT-1 AG-see-PUNCT
'I'll see it' (prediction)
aa-ke-ke -?
OPT-1 AG-See-PUNCT
'I should/might see it' (obligation or possibility)
Chafe interprets these in terms of a reality continuum with 'factual' at one
end, 'optative' at the other and 'future' as intermediate.
However, it could be argued that the systems of Dani and Seneca are
modal systems rather than mood. In addition to the fact that there is a three-
term system, the use of 'optative' as both epistemic and deontic is typical oi
modal systems (see 4.1.1) and the gloss on 'factual' is reminiscent of <*n
evidential term. An argument against this is that modal systems do n
usually include Future.
162
 6,5 Non-joint systems

Rather differently, in Alamblak (Roberts 1990: 390) there are two negative
forms, each associated with a different irrealis marker (they are distinguished
a S 'certain' and 'uncertain' negatives and irrealis):

fifiji noh-r-fe-r
NEG.CERT die+IRR.CERT-IMM.PAST-3SG+MASC
'He did not die'
afe noh-rhwa-t-r
NEG.UNCERT die-FUT+IRR.UNCERT-3sG
'He will not die'
However, as can be seen, the choice of the irrealis marker is determined by
the negative. The 'certain' irrealis is also used alone for counterfactuals:
hik-r-fe-an-n
folloW+IRR.CERT-IMM.PAST-lSG-2SG
'I would have followed you'

6.5.3 Wider systems

It was noted in 6.2 that occasionally the term 'irrealis' is used for a term in a
modal system, where a term such as 'dubitative' would be more appropriate.
However, in some languages both 'irrealis' and 'realis' have been applied to
terms that belong to a wider system, which, on formal grounds, should be
analysed in terms of a modal system rather than (typically binary) mood.
There is, however, some justification for the terms and for considering them
here - that the notional features associated with them (and with irrealis in
particular) are similar to those associated with irrealis in the languages that
have been discussed.
Thus, in Kiowa (Oklahoma - Watkins 1984: 170-2) there is a marker of
'future' which is treated by Mithun (1999: 173) as irrealis (but Watkins's
'future' is used in the glosses in the examples below). It clearly indicates future:
em-can-t'o'-no hg y-mokhl-d-o'
2sG-arrive-iRR-and+DiFF now 2sG/lsG/PL-ready-be-FUT
'I'll have them ready when you come'
hgy -b-o'
perhaps lso-go-FUT ,. ,
'Maybe 1 will go'
"- 's also used in several (notionally) irrealis functions - what might have
occurred but didn't, conditional and obligation:
mgy -b'-o'
might.but.not lsc-go-FUT
'1 thought I might go, but didn't'

163
Realis and Irrealis

-th-?-t' g -tl
2sG+SG.OBJ-find-FUT and+SAME 2sG+sG.0Bj-send+iMP
'If you find him send him here'
k'yakomda m'-Q'-do kypd -thy-d'-t'
life indeed-good-because longtime lPL+SG.OBJ-with-hold-FuT
'Because life is so enjoyable, we ought to hang on as long as possible'
There is also an unmarked 'indicative' (= realis?), and markers for imperativ
and 'hearsay' (= Reported (3)?). The imperative is illustrated in the second
example above. An example of 'hearsay' is:
btho' m-k'd-olthQ -khop-md-hl
unknowing 2sG-very-head-hurt-become-HSY
'I didn't know you had a headache'
There is also a marker for negative imperative, though this occurs (jointly)
with 'future' (irrealis):
py t' mn-tl-t'
PROH all 2sG+DUAL-tell-FUT
'Don't tell them everything'
This 'future' (irrealis) in Kiowa has a number of uses that are associated with
irrealis in other languages, but clearly it does not belong to a binary system
or even a trinary system (of realis, irrealis and unmarked).
A more dubious example is that of Maricopa (Yuman, Arizona - Gordon
1986a: 27, 109). Here an 'irrealis' suffix is used (without any other gramma
tical marker) to indicate future, possible and contrary-to-fact:
ny-aay-ha
1/2-give-iRR
'I will give it to you'
haat nyi-ttpooy-nt-ha
dog+PLOBJ+PL-kill+PL.ACTION-tOO-IRR
'It might kill dogs too'
aanylyviim m-vaa-kis ?nym-yuu-ha
yesterday 2-come-coND 2/1 -see-iRR
'If you had come yesterday, you would have seen me'
Similarly, a realis marker indicates present or past (Gordon 1986a: 24, 25):
hot-m
good-REAL
'It is good'
aham-m
hit-REAL
'He hit him' . . . .

164
w 6.5 Non-joint systems

j-[owever, Maricopa also has markers for 'completed actions' (perfective),

jj(. c 0 i n pletive',
'desiderative' and two evidentials, Visual and Non-visual:
?-yuu-ksh
1-see-PERFV
'I saw it'
nyaa ?-imaa-iNC
I 1-dance-iNC
'I can dance'
m-we-lya
2-do-DES
'Please do it'
imaa-yuu ,
dance-vis
'He danced' (I saw it)
m-ashvar-?a
2-sing-NONVis
'You sang' (I heard you)
There is also an imperative, but this occurs jointly with realis:
k-truy-m
IMP-kill-REAL
'Kill it'
Another language with an 'irrealis' form with several functions is Mao
Naga (Tibeto-Burman - Giridhar 1994: 67-9). Here an irrealis marker is
used for reference to the future, but also for thwarted intentions and for
obligation:
ai izo ocii vuta le
I today home go IRR
'I will go home today'
Alemo-no oro hru le-Ti-e
Alemo-ERG pig buy IRR-RELEV-PRED
'Alemo wanted to buy a pig, but couldn't'
pfo-no idu ru-oTi-le
he-ERG yesterday write-iRRELEV-iRR
'He must have written yesterday'
However, 'irrealis' occurs, as can be seen, with markers of 'relevance' and
^relevance' and it can be replaced by 'inference':
pfo-no idu ru-oTi-ahi
he-ERG yesterday write-iRRELEV-iNF
'He must have written yesterday'

165
Realis and Irrealis

Strictly, all of these languages are languages with modal systems, but tt,
all have an 'irrealis' marker that indicates features that are associated /
the grammatical category of irrealis in other languages.

6.5.4 Complex systems

There is a complex situation in Hixkaryana (Carib, N. Brazil - Derbvsh'
1979: 143-5), in that there appears to be both an evidential modal system
the kind discussed in Chapter 2 and a system of realis/irrealis mood. MooH
appears to be indicated by a non-past and a non-past 'uncertain' form of th
verb, which can, perhaps be seen as realis and irrealis (though the system also
involves an 'intensifier' particle).
When occurring alone, the 'non-past uncertain' indicates a question:

nomokyaha
he.come+NONPAST
'He is coming'
nomokyano
he come+NONPAST.UNCERT
"Will he come?'

With the intensifier ha-, however, these express 'may' and 'must' respectively:
nomokyaha ha
he come+NONPAST INT
'He must certainly come'
nomokyan ha
he.come+NONPAST.UNCERT INTEN
'He may come'
With other particles, all preceded by the intensifier ha-, the non-past also
expresses:
'hearsay' with -ti
nomokyan ha-i
he.COnie+NONPAST.UNCERT INT-HSY
'He's coming' (they say)
'uncertainty' with -na
nomokyan ha-na
he.COme+NONPAST.UNCERT INT-UNCERT
'Maybe he'll come'
'deduction' with -mi
nomokyan ha-mt
he.COme+NONPAST.UNCERT INT-DED
'He is evidently coming' (on hearing the sound of an outboard motor)

166
 6.5 Non-joint systems
'positive doubt, scepticism' with -mpe
nomokyatxow ha-mpe
they.come+NONPAST.UNCERT INT-SCEP
'They are coming! I don't believe it'
The non-past may also be used with the intensifier plus another particle for
'certainty', 'prediction' or 'warning':
nomokyaha ha-mpi ni
he.come+NONPAST INT-PREDICT
'He's coming - be warned'
These are all categories found in modal systems.
Even more complex is the situation in Serrano (Uto-Aztecan, California -
Hill 1967: 21, and personal communication). There is a set of evidential
particles which 'specify the validity of the statement'. They are:

ha inferential (889)
kw3'3 potential 'can'
k w an 3 quotative
may 'may'
na'a volitative
pata intensive
qy 'not'
ta dubitative
The 'dubitative' seems to be a marker of irrealis. In isolation it too
indicates a question (although this is also signalled by intonation):
kwa'i ta-m kihwu:i
eat DUB-PL you fish+ACC
'Are you (pi.) eating fish?'
hai:rjkwa ta-bi mi
to.where DUB-he+PAST go
'Where did he go?'
It co-occurs with future and deductive ('inferential'):
'i:p t wahi' pinkiv
here DUB coyote pass+FUT
'The coyote will pass here'
'ama' t X ma:m
he DUB INF hear
'He must hear it'
However the 'potential' can also be an indicator of a question (Hill, personal
c
nvmunication), and the 'quotative' may also be used alone with reference
t0
the future:

167
Realis and Irrealis
kwi?- pi yi:?i
POT-you you+them dry
'Can you dry them?'
pimia' kwini- quib
with him QUOT-we dwell+FUT
'We would live with him (so we were told)'

Hixkaryana and Serrano could, then, also have been considered '
Chapter 2 rather than here, since they appear to have an evidential mo
system as well as mood. However, both have Interrogative as a term in th
system and Hixkaryana has Negative as well. These at least are more typical
of mood than of modal systems.

6.6 Prepositional modality

6.6.1 Past, present and future

There is very little to be said about reference to simple past and present.
These (perhaps to be identified as 'Non-future') are always marked as realis.
Some languages have more than one marker for past, e.g., Amele with both
'remote past', and 'today's past' both of which are realis.
It is important to realize, however, that it is not the case that reference to
the present or past time will always imply realis marking. There is nothing
strange about the fact that, for instance, habitual past is marked as irrealis in
Bargam (see 6.6.9), and other categories discussed in this section, such as
Speculative, Negative, Interrogative, etc., may imply irrealis marking, even
though there is reference to past or present time. It is only simple statements
(assertions) about present or past actions that are always realis.
By contrast, Future is the category most commonly associated with irrealis
in both joint and non-joint systems. Indeed, it was shown in 6.4 and 6.5.1
that in the Papuan languages of both types Future is always marked by
irrealis. Examples already quoted from (joint) Amele (Roberts 1990: 372)
and (non-joint) Muyuw (Bugenhagen 1994: 18) are:

ho bu-basal-en age qo-qag-an

pig siM-run.out-3sG+DS+iRR 3FL hit-3PL-FUT
'They will kill the pig as it runs out'
yey b-a-n Lae nubweig
I iRR-lsG-go Lae tomorrow
'I will go to Lae tomorrow'
The same is true of most of the Native American languages. In Kiowa (see
6.5.3), Future is one of the categories indicated by irrealis alone. In Takelm^

168
r 6.6 Propositional modality

, 1 o the future paradigm is in the irrealis set. In Serrano (see 6.5.4), too,
c iture is marked as irrealis ('dubitative'):
'i:p t wahi' pinkiv
here DUB coyote pass+FUT
'The coyote will pass here'
However, the situation is quite different in Caddo. In that language there
re two markers of future time, and both are marked as realis, not, as
expected, irrealis (repeated from 6.1):
ciyi-bahw-?a?
1 +AG+REAL-See-FUT
'I'll look at it'
ciyi-bahw-ah
1+AG+REAL-See-FUT INT
'I'm going to look at it'
There is also a different situation in Central Porno (Mithun 1995: 370).
Here the future occurs most commonly with irrealis:
te'nta-lil wa-'n-hi ?' q-be-w-?khe
town-to go-iMPF-SAMF.+iRR 1+AG toward-carry-PERF-FUT
'I'll go to town and bring it back'
However, the realis marker may be used instead to indicate greater expecta
tion of the event taking place (Mithun 1995: 379):
?' h-w-da m ?b'-n--w-?k>ie
1+AG nOt-PERF-DIFF+SIM+RF.AL 2+AG Suffer-DUR-REFL-PERF-FUT
'When I am no longer here, you will suffer'
It is natural to look for an explanation for the difference between Central
Porno and Caddo. There is a discussion of this in Chafe (1995: 358), but it
includes the imperative (see 6.7.1), which is also marked as realis. He
comments that since so many other categories in Caddo are 'functionally
motivated members of the Irrealis category', the occurrence of irrealis with
Imperative is 'an inconsistency'. He goes on to say that the use of realis with
Future is 'more problematic . . . especially in view of the fact that future is so
fundamental to the expression of irreality in the Papuan languages'. He
suggests two answers. The first is that 'reality is not a binary but a gradient
dimension in which imperatives and futures express ideas that are judged to
be more in accord with reality than, say, yes-no questions or negations,
speakers may have a relatively stronger expectation that commands will be
obeyed or that predicted events will take place.' He suggests that in the
northern Iroquoian languages Future is intermediate between realis and
lrr
ealis. The second is a diachronic explanation - that imperatives and

169
Realis and Irrealis

futures were established in Caddo before the realis/irrealis distinction w-

formed (from interrogative verbs). But such a diachronic explanation wo
be only half the story - it still needs to be explained why interrogatives a
futures were not absorbed into such a 'functionally motivated system' R
answers may be true.
There are other languages in which there is a close association of futn
with irrealis. In Burmese, the 'non-future' marker appears to be used to ref
to present and past events, while the 'future' marker is used for future event
(Okell 1969: 425, 426, 355, who, however, merely refers to them as 'verb
sentence markers'):
sneine-tain" mye? hpta?-te
Saturday-every grass CIU-NONFUT
'He cuts the grass every Saturday'
da-caflmo m-la-ta
that-because.of not-come-NONFUT
'So that's why they didn't come'
mne?hpafl s-me
tomorrow begin-FUT
'We shall begin tomorrow'
However, Comrie (1985: 51) points out that the contrast is not a temporal
one. The 'future' marker may be used to refer to judgments in the present
and past (Okell 1969:355):
hmafl-le-me
be.true-undoubtedly-FUT
'That may be true'
mcithi s-hp-me htifi-te
tamarind eat-ever-FUT think-NONFUT
'I think he must have eaten tamarinds before'
Similarly, Comrie (1985: 51) argues that the putative future tenses of
Dyirbal (Australian - Dixon 1972: 55) are essentially modal, but only because
the 'future' ('irrealis') is used for present habitual. Yet, by contrast, present
habitual is marked by the 'non-future' in Burmese. There are, moreover,
many examples of the use of what are generally regarded as future tenses to
refer to probability (Assumptive), e.g., the morphologically defined future
tenses of Italian, French or Spanish. English WILL, moreover, has the notional
functions of assumptive, future and habitual (see 2.1.3, 4.3.2 and 6.6.9).

6.6.2 Speculative and Deductive

Surprisingly perhaps, epistemic categories such as Speculative and Deductive
(and also Reported) that are regular members of the epistemic modal system

170
 6.6 Propositions modality

not listed as categories associated with irrealis in the tables for the
( notions of irrealis in Papuan languages presented by Bugenhagen (1994: 36)
a nd Roberts (1994: 31).

In the Native American languages, they are sometimes, but rarely,

ociated vvith irrealis. There are three marginal cases that were noted. First,
n Takelma (see 6.3.1), the 'inferential' is used for both Ability and
Speculative and the 'potential' is used for both Deductive and Reported.
Secondly, in Kiowa (Oklahoma - Watkins 1984: 171), Speculative is indi-
cated by irrealis plus 'perhaps' (though the example given refers to the future
a l s o, and Future is marked by irrealis):

hgy -b't
perhaps lSG+go+iRR
'Maybe I'll go'

Thirdly, Speculative is sometimes found in non-binary systems (6.5.2), where

an interpretation in terms of realis, irrealis and 'intermediate' is possible, the
irrealis being interpreted as Speculative. This is the interpretation given by
Chafe (1995: 339) for Seneca (examples repeated from 6.5.2):

wa?-ke-ke-?
FACT-1 +AG-See-PUNCT

T see/saw it'
es-ke-ke-?
FUT-1 +AG-See-PUNCT

'I'll see it'

aa-ke-ke-?
OPT-1 +AG-See-PUNCT

T should/might see it'

A similar trio from Dani was also noted in 6.5.2 (Foley 1986: 163):

wat-h-i
kill-REAL-lSG+AG
'I killed him'
was-0-ik
kill-likely-ISG+AG
T will kill him'
wa?-l-e
kill-POT-lSG+AG
T may kill him'

!n Burmese, however (Okell 1969: 355; Comrie 1985: 51), the irrealis
^ future') marker is also used for Speculative:

171
Realis and Irrealis

hma-lefi-me
be.true-undoubtedly-FUT
'That may well be true'
mcithi s-hp-me hti-te
tamarind eat-ever-iRR think-FUT
'I think he must have eaten tamarinds before'

In general, however, Speculative, Deductive and Reported do not appear

be marked by mood.
Mention should be made of Hixkaryana and Serrano (6.5.4) but thes
appear to have both a modal system and a system of mood. The modal
system in Hixkaryana clearly includes Speculative and Deductive, and that of
Serrano has Deductive (and both have Reported also - see 6.6.6).

6.6.3 Interrogative
In some languages the irrealis marker is used, without any other grammatical
marker, to indicate a question. (Some of the examples have already been
given but are repeated here.)
This is true of Caddo (6.1):
sah?yi-bahw-nah
2+AG+IRR-See-PERF
'Have you seen him?'
It is also true of Hixkaryana and Serrano (which have both a modal system
and a system of mood - see 6.5.4). An example from Hixkaryana is:
nomokyano
he .come+NONPAST .UNCERT
'Will he come?'
The situation is more complex in Serrano (Hill 1967: 21) in that the irrealis
('dubitative') is obligatory in questions - even though questions are also
marked by intonation (shown as '?' in the examples) - unless the potential is
present:
k w a'ita-m c ? kihwu:ci (20
i

eat DUB-PL you QUES fish+ACC

'Are you (pi.) eating fish?'
There is, however, one exception. The 'dubitative' marker is not required it
the 'potential' is already present (Hill, personal communication):
kwi-c pi yi:?i?
POT-you you+them dry
'Can you dry them?'

172
 6.6 Propositional modality

W usually, then, it is the potential alone that may occur without any other
' mark er for questions.
in Central Pomo, however, (Mithun 1995: 373, 381) although an interro-
f Hon marker is required it does not affect mood. Thus, interrogative will
c u r with irrealis when there is also a future marker, but with realis when
there is also a marker of past time, the situation being the same as if there
were no interrogative present, mood being determined solely by the (irrealis)
future and (realis) perfective:
?i-wa ma a--t-?khe ya-1 ca-'l
be-iNT 2+AG swinging-whip-M.E-FUT IPL+PAT house-to
d-m-ma-hi?
lead+PL-M.E.-COOP-SAME+IRR
'Are you going to whip us when you take us home?'
than da-s-'-ba-wa ma ?' ch?l-'i-w
hand pulling-wash-REFL-SAME+SEQ+REAL-iNT 2+AG hair comb-REFL-PERF
'Did you wash your hands and comb your hair?
Mithun (1995: 380-1) offers an explanation for the difference between the
marking of Interrogative in Central Pomo (no effect) and Caddo (irrealis).
She suggests that it is a matter of scope. In Central Pomo basic propositions
are categorized as realis or irrealis and then may be questioned:
QUESTION ( REALIS (PROPOSITION))

In Caddo, the scope relations are reversed, the marking of realis/irrealis

being sensitive to questions:
-REALIS (QUESTION(PROPOSITION))

6.6.4 Negative
Interrogative and Negative are often considered together because they
appear to function in similar ways (see 2.3), as they do in English where they
are both subsumed under 'non-assertion' - cf. Quirk et al. (1985: 83-4).
There are similarities between them in respect of realis and irrealis marking.
In Caddo (Chafe 1995: 354, 355) negatives, like interrogatives, co-occur
with irrealis:
kuy-t'ayi-bahw
NEG-1+AG+IRR-See
'I don't see him'
This is also true of Mesa Grande Diegueflo (Yuman, S. California -
Langdon 1970: 159):
?anya' puy ?axap-x-vu awa'p-x uma'w
I there I.go.in-IRR-SPEC they.want.it-lRR they+NEG
'They didn't want me to go there'

173
Realis and Irrealis

In Central Porno, however, the negative is like the interrogative in that

does not affect mood. Thus, where there are markers of both negative
conditional, the mood marker will be irrealis, and where there are marker
both negative and past, the mood marker will be realis, just as they would
no negative were present (Mithun 1995: 374-5):
ma me'n ?-w h-w-hla ma b-da
2+AG SUCh do-PERF NEG-PERF-DIFF+1RR 2+AG this-at
ma' bast thab'?i-w p^-w-w^ke
things bad lie-iNCH-PERF vis-perceive-PERF-FUT
'If you don't do that, you're going go see bad things happening round
here'
ranch-?el qd yh-t-a h-w
ranch-the good do-M.E-iMPF+PL NEG-PERF
?-n ya-1 q-'l ma-w
be-SAME+SIM+REAL 1+PL-PAT OUt-tO thrOW-PL+PERF
d-'?-c'i-w
wailt-REFL-IMPF+PL-PERF
'Because they didn't keep up the rancheria, they wanted to throw us
out'
There is a similar variation in the Papuan languages. Roberts (1990: 378)
specifically remarks that, although negation could have modal status (no
doubt thinking of the situation in the Native American languages), 'it does
not interact with the medial simultaneous realis-irrealis verb' in Amele.
However, he also notes that in Alamblak irrealis is associated with negation,
so that negative past (immediate past and remote past) as well as future is
marked as irrealis:

fiji noh-r-f-r
NEG.CERT die-lRR-lMM.PAST-3M.SG
'He did not die'
af noh-rhwa-t-r
NEG.UNCERT die-FUT-IRR.UNCERT-3SG
'He will not die'
(As noted in 6.5.2, there are two negative markers in Alamblak, each
associated with a different marker of irrealis.)
Similarly, Bugenhagen (1994: 19) reports that for just one of the eight
languages that he presents, Muyuw, negation is either unmarked for mood or
marked as irrealis (but never as realis) with both past and present as in:
nag i-n wa-ven
NEG 3so-go to-village
'He is not going/did not go to the village'

174
 6.6 Propositional modality
nag b-ei-n wa-ven
NEG iRR-3sG-go to-village
'He is not going/did not go to the village'
IVlithun (1995 381-2) provides the same explanation for the variation in
he realis/irrealis status of Negative as she does for Interrogative. It is a
m atter of scope:

Central Pomo NEGATIVE ( REALTS (PROPOSITION))

Caddo -REALIS (NEGATIVE(PROPOSITION))

6.6.5 Negative-like categories

There are several instances of the use of irrealis with categories or expressions
that are notionally to some degree negative.
Thus the infrequentative co-occurs with irrealis in Caddo (Chafe 1995:
357):
ws-a-yibahw
INFREQ-1 + AG+IRR-See
'I seldom see it'
This is best seen as a type of negation, since this sentence implies that
generally I do not see it. (Seldom is a 'semi-negative' in English - see Palmer
1987: 22.)
There are also several examples of the use of irrealis to refer to events that,
in some way or other, were not fulfilled or to states that no longer hold,
rather like the 'contrary to expectation' forms discussed in 2.4. Thus in
Kiowa (Oklahoma - Watkins 1994: 171) irrealis may indicate that an action
might have occurred, but did not (repeated from 6.5.3):
mgy -b'-o''
might but not lso-go-iRR
'I thought I might go, but didn't'
Somewhat similar is the situation in Tolkapaya Yavapai (Yuman, Cali
fornia - Hardy and Gordon 1980: 189-92). There is an infix, th, which is
labelled 'modal', but which appears to be more like an irrealis marker in spite
f the fact that there is another particle that is labelled 'irrealis'. This 'modal'
mfix has a number of functions, which may be grouped into three types:

(i) It is used for unreal conditions and (unreal) wishes:

m-vaa-th-m ny-'u-h yi-tha
2-came-MOD-DS 1/2-see-iRRAUx-MOD
'If you had come I would have seen you'
ma-ch'-yu-th-k wal'-yii-k '-yum
1-suB-be-MOD-ss I-wish-ss I-be
'I wish I were you'

175
Realis and Irrealis
(ii) It is used after 'try':
'wi '-yoov-a-k '-wi-th-k '-yum
money 1/3-make-iRR-ss 1-do-MOD-ss 1-be
'I'm trying to make money'
(Hi) It is used for 'what used to be (but is no longer)':
ma-ch m-se-ch m-yu-th-k m-yum
you-suB 2-fat-sUB 2-be-MOD-ss 2-be
'You used to be fat'
'-ima-th-k '-tlahv-k '-yum
1-dance-MOD-ss 1-tired-ss 1-be
'I was dancing, but now I'm tired'
The significant point is that in (iii) the form expresses what was previously so
but is no longer. (This may be relevant to the discussion of habitual past in
6.6.9. For habitual past also suggests that the actions of the past are no
longer being continued, so that this may be part of the reason why habitual
past can be marked as irrealis.)
Slightly differently, there are two irrealis markers in Nakanai (Papuan -
Johnston 1980: 64, discussed by Bugenhagen 1994: 24-5). One, 'non-
imminent irrealis' ge, is used to express future, doubt and possibility, the
typical categories associated with irrealis. The other, 'imminent irrealis' ga,
'encodes the notion of imminent or frustrated action' according to Johnston,
though Bugenhagen suggests that 'almost' would be a better translation, as
in:
eau ga tuga so-io, ouka
lsg IRR walk to-there, not
'I was about to proceed, but didn't'
eau ga la-lea
lSG IRR RED-Sick
'I'm getting sick'
(For examples of'non-imminent irrealis', see 6.7.2.)
A similar example of frustrated action was given for Mao Naga in 6.5.3
(repeated here):
Alemo-no oro hru le-Ti-e
Alemo-ERG pig buy IRR-RELEV-PRED
'Alemo wanted to buy a pig, but couldn't'

6.6.6 Reported
Reported occurs as one of the categories in many evidential systems (se
2.2.2). It might, therefore, be expected to be one of the categories marked as
irrealis with mood, but the only example noted was from one of the W
176
 6.6 Propositional modality

i neuages discussed in 6.5.4, which appear to have both mood and a modal
ystem. This is Hixkaryana (Carib, N. Brazil - Derbyshire 1979: 143-5),
uhere 'hearsay' occurs with 'uncertain', which is the marker of irrealis:
nomokyan ha-tt
he.COme+NONPAST+UNCERT INT-HSY
'He's coming (they say)'
The other language discussed in 6.5.4 was Serrano (Uto-Aztecan, California
Hill 1967: 89), but there 'quotative' (= Reported (3)?) can occur alone,
without any marker of irrealis:
ptmia' kwini- qucib
with.him QUOT-we dwell+FUT
'We would live with him (so we were told)'

6.6.7 Presupposed
Chafe (1995: 357) offers from Caddo (Oklahoma) an example of the irrealis
marker being used with the Admirative prefix Ms:
hs-ba?a-sayi-k'awih-sa?
ADM-1 +BEN+iRR-name-know-PROG
'My goodness, he knows my name'
Chafe says 'This last usage is perhaps surprising', since 'the event or state is
real enough', and suggests that the fact that it is contrary to expectation is
responsible for the use of the Irrealis, 'as if the speaker were saying "It's
unreal that he knows my n a m e " ' . But it is no more surprising than the use of
the subjunctive in Spanish and Italian for what is presupposed (5.2.5), as
would be clear if the translation were 'I'm amazed that he knows my name.'
The explanation that there is no new information is equally valid for this use
of the irrealis in Caddo.

6.6.8 Conditional
Conditionals are marked as irrealis in Caddo (Chafe 1995: 356):
hi-t'a-yibahw
COND-1+AG+IRR-See
'If I see it'
^o too are other markers that are notionally similar - generic conditioaal
an
d simulative:
nas-t'a-yi-bahw
gen coND-1-AG+iRR-see
' W h e n e v e r 1 see it'

177
Realis and Irrealis

dy-ayi-bahw
SIMULAT-1 +AG+IRR-see
'As if I saw it'
Similarly in Central Porno (Mithun 1995: 370), both real and unr
conditionals (see 8.1.) are marked as irrealis, though only the latter has th
conditional marker:
wa-q-hi ?e lo'-h-du-w?khe
gO-level-SAME+IRR COP help-IMPFV-IMPFV-PERFV-FUT
'If I go, I'll be helping out'
me'n mi-hla mu'l ?a' ?h'?le ?a'
so say-DiFF+iRR that 1+AG sit+C0ND 1+AG
'If she said that, I'd stay longer'
(This 'conditional' is also used to express obligation - see 6.7.3.)
There is an interesting situation concerning conditionals in the languages
considered by Bugenhagen (1994). He distinguishes what he calls 'counter
factuals' and 'hypotheticals'. These are what have been called 'unreal' and
'real' conditionals (see 8.1), although it is obvious that Bugenhagen's choice
of terms avoids the danger of confusion with realis and irrealis. Yet there
seems to be some genuine interplay between these two different types of
'unreality' in this sense and 'irreality', in that Papuan languages deal with
mood in conditionals in different ways. Generally both clauses of conditions
of both types are marked as irrealis, as in Manam:
?ti t?e-o rja-pra-ra Bogi n-l?a
boat one-3sG iRR-arrive-assume Bogia ISG+IRR
'If a boat should come, I would go to Bogia'
'If a boat had come, I should have gone to Bogia'
Yet in Sursurunga, quite paradoxically, the protasis ((/-clause) of counter-
factual (unreal) conditionals is marked as realis:
ngo -k-te han balbal us rin na
if 3SG+REAL-DEF-EMPH go again blow SUBJ rain 3SG+IRR
han kopkom kuluk namnam
go grow good SUBJ food
'If it had kept on raining regularly, the crops would have grown well'
Only in one of the languages quoted by Bugenhagen (1994: 33), Sinangoro, is
the distinction between hypotheticals and counterfactuals marked q m,:e
simply by realis and irrealis in both clauses. Only the hypothetical, with realis
markers, is exemplified:
yema yasi be iayoma-ni nai, ba iayo-m
if canoe 3SG+REM+REAL come-iMPF time ISG+REM+REAL go-iMPF
'If the canoe comes, I will go'
178
 6.7 Event modality

6.6.9 Habitual Past

little surprising, perhaps, is the co-occurrence of habitual past with irrealis
gargam (Papua - Roberts 1990: 384, quoting an unpublished manuscript),
w as mentioned in 2.4:

miles-eq leh-id teq anamren aholwaq-ad in

return-ss+iRR go-DS+iRR then owner see-ss+siM 3SG
didaq tu-ugiaq
food PERF-give+HAB.P+3SG
'When (the pig) would return and then the owner would go and, on
seeing it, used to give it food'
flowever, it is possible to see why this should be treated as irrealis. Although
past time reference is usually treated as realis, what is relevant here is that the
habitual past does not relate to specific actions in the past, but to a tendency
to act. Indeed, Given (1994: 323) suggests the habitual is a 'hybrid modality',
sharing some features of realis (higher assertive certainty) and some of irrealis
('lack of specific temporal reference; lack of specific evidence; . . .').
Moreover, there are similar examples in other languages. As noted in 2.4,
there is an evidential suffix in Kashaya, that indicates 'Remote' - an action
(habitual?) in the remote past (Oswalt 1986: 40):
men si-yi?ci?-thi-miy
this do-PL+HAB-NEG-REM
'They never used to do that in the old days'
It may also be relevant that Tolkapaya Yavapai has a marker that looks like
irrealis to indicate what was previously true but no longer is. This was
discussed in 6.6.5.
It may also be recalled that in Manam the prefix for habitual events is
realis, although the prefix for 'sequences of customary or habitual activities'
is irrealis (6.5.1) and that what might be regarded as irrealis in Dyirbal (6.6.1)
may refer to either future or habitual events. Moreover, as Chung and
Timberlake (1985: 221) point out, English uses a modal verb to express
habitual action:
We would go for a walk most weekends
However, this view has been challenged (Bybee et al. 1994: 236-40; Bybee and
Fleischman 1995b: 9-10), and there will be a great deal more discussion in 7.3.

6.7 Event modality

6.7.1 Imperative and jussive

As noted in 3.4 (see also 5.4), two types of commands are often distinguished
jrammatically. Those that are directed at the addressee or addressees alone

179
Realis and Irrealis

are usually treated as imperatives; in that sense there are only second per s
imperatives. All others are treated as jussives, though the term 'hortative' '
also used. These are usually introduced in English by Let and are either fi
or third person.
In Central Pomo, imperatives and jussives occur with irrealis, as is show
in the examples (6.3.2):
qh hni-?el d--hi m'tuya-1
water bread-the make-SEMEL-SAME+iRR 3PL-PAT
qa'-w-'-ka-m
biting-gO-IMPFV+PL-CAUS-IMP
'Make the water bread and invite them to eat it'
ya-ka hy -d--ma-hi
1+PL+AG-INF wood drag-carry-iNCH-coop-SAME+iRR
qha-kay -d--ma-w-?khe
water-too dragging-carry-iNCH-coop-PERFV-FUT
'Let's get some wood and haul in some water'
In Takelma (see 6.3.1) there are two imperatives, a present imperative for
commands to be carried out immediately and a future imperative for action
at a later time (for other examples see 3.4), and both are marked as irrealis.
By contrast in Maricopa (Yuman, Arizona - Gordon 1986a: 25) the
imperative occurs with realis:
k-tpuy-m
IMP-kill-REAL
'Kill it!'
The same is true of Caddo (Chafe 1986: 358)
yah?-yibahw
1 +AG+REAL-see
'Look at it'
Chafe sees the marking of Imperative as realis as 'an inconsistency' and
offers two possible explanations. This was discussed in 6.6.1.
It is instructive to note that 'prohibitives' (negative imperatives) are
marked as irrealis in both Caddo and Central Pomo, but for different
reasons. In Caddo, although the imperative is marked as realis (see 6.1), there
is a prohibitive marker that occurs with irrealis:
ka-sah?-yibahw
PROH-2+AG+IRR-See
'Don't look at it'
In Central Pomo the imperative is marked as irrealis and is still irrealis when
the negative also occurs:

180
 6.7 Event modality
h
da'wi ? --hi khy sw-lan?khe-thn ?e ma
on road stop-SAME-iRR game play-NEG it.is you
'Don't stop and play on the way home'
in Caddo, then, the negative imperative is irrealis because of the negative,
w hile in C. Porno it is irrealis because of the imperative. Again, Mithun
/1995: 383) offers an explanation in terms of scope (see 6.6.3, 6.6.4):

Central Pomo NEGATIVE (-REALIS (IMPERATIVE)

Caddo -REALIS (NEGATIVE (IMPERATIVE))

However, in Kiowa (Oklahoma - Watkins 1994: 172), only the 'prohibi

tive' is marked as irrealis, the imperative having no mood marker (repeated
from 6.5.3):
py t' mn-tl-t'
PROH all 2sG+DUAL-tell-IRR
'Don't tell them everything'
-th-?-t' g -tl
2sG/sG-find-iRR and+SAME 2sc/so-send-iMP
'If you find him, send him here'
By contrast, in Sursurunga (Papuan - Bugenhagen 1994: 16) thene "is
irrealis marking in imperatives, but not in'prohibitives':
u-na rumrum i mama-m mai kaka-m
2SG-IRR respect OBJ mother-2so and father-2sG
'Respect your father and mother'

6.7.2 'Strong' and 'polite' commands

A number of languages make a distinction between 'strong' and 'polite'
commands (imperatives and jussives) by means of realis/irrealis marking.
Thus, in Alamblak (Roberts 1990: 390-1, quoting Bruce 1984: 137, 140),
both the imperative and the hortative (i.e., the jussive) are used without an
irrealis marker. Examples, repeated from 6.5.2, are:
nuat wa-ya-n-t
sago.patty iMP-eat-2so-3F.SG+OBj
'Eat the sago patty!'
nuat a-ya-nem-t
sago.patty HORT-eat-1 PL-3F.SG+OBJ
'Let's eat the sago patty'
However, both the imperative and the 'hortative' may also be used with a
barker that is interpreted as exhibiting the cumulative exponents of future,
lr
realis and 'hortative'/imperative:

181
Realis and Irrealis

(nik) wa-roh-twa-k
(2PL) IMP-St-FUT+IRR+IMP/HORT-2PL
'Would you be seated'
(rr) a-roh-twa-r
(3SG) HORT-St-FUT+IRR+IMP/HORT-3M.SG
'May he be seated'

Here, then, marking with irrealis creates a politer command (with both
imperative and jussive).
Similarly, in Nakanai (Johnston 1980: 62) both imperatives and 'prohibi
tions' (negative imperatives) may be expressed both with no mood marker
and with the 'non-imminent' irrealis marker ge (see 6.6.5). The translations
suggest that the unmarked sentences express a stronger imperative:
tola egite so-ma
call 3PL to-here
'Call them here'
amutou ge lolo-a
2PL IRR hear-3sG
'You are to listen'
umala kokue-a
don't hit-3sG
'Don't hit him'
eme umala ge kokue-a
2SG don't IRR hit-3so
'You must not hit him'
Slightly differently, in Jamul Diego (Yuman - Miller 1990: 119) there are
two types of command, neither of them with an imperative marker. The
'strong' form for politeness has an irrealis suffix, while the other has a special
second person prefix and no irrealis suffix:
nya-m-mpa-pu m-rar-x-s
INDEF-2-NOM+Want-DEM 2-do-IRR-EMPH
'Do whatever you want'
k-naw
2-run
'Run'
These examples would seem to have something in common with the
situation in European languages, where polite commands are expressed by
the subjunctive. This has been illustrated for Italian and Spanish (see 5.4.2),
where even greater politeness is achieved by using the third person form, a
well as the subjunctive: ..,-..

182
pp 6.7 Event modality

Italian
entri pure
enter+3sG+PRES+suBJ if.you.please
'Please come in'
Spanish
Tome su libro
take+3so+PRES+suBj your book
'Take your book'
In Amele (Roberts 1990: 384) both strong and polite commands are
marked as irrealis, but in rather different ways. There is a clear contrast in
the final verb in that with a strong command the final verb is marked with
the imperative, while with a polite command the final verb is simply marked
a s irrealis (with no imperative marker):

h-og-a "'"'"'"''

come-2sG-iMP
'Come!'
ho-ho-m
SIM-COme-2SG+DS+IRR
'Would you come'
However, even with the strong imperative (itself not marked as irrealis) the
medial verb in a linked clause is marked as irrealis, as in (repeated from
6.3.2):
ho bu-basal-en age qo-ig-a
pig siM-run.out-3sG+DS+iRR 3PL hit-3PL-MP
'Kill the pig as it runs out'
There is a similar situation in another Papuan language, Bargam (Roberts
1990: 384, quoting an unpublished manuscript):
ni bol
2SG come+iMp
'Come!'
ni bol-eq
2SG come-ss+iRR
'Would you come'

6.7.3 Obligative
In Caddo, there is an obligative marker that occurs with irrealis (repeated
fr
om 6.1):
kas-sa-ny? aw
OBL-3+AG+IRR-Sng ' ' ' ''
'He should/is obliged to sing'

183
Realis and Irrealis

There is also a marker for weak obligation in Central Pomo (Mithun 1990.
177), but this is also a conditional marker (see 6.6.8):
c-w-htow ?'y-yo-hi twhal da-'-?le
house-LOC-from away-go-SAME+iRR work handling-catch-coND
'He should go home and get a job'
In Kiowa (Oklahoma - Watkins 1984: 172), irrealis alone may indicate weak
obligation, though it is also used for future and for what might have occurred
but didn't (repeated from 6.5.3):
k'yakomda m'-o.'-d kyd -thy-d'-t'
life indeed-good-because longtime lPL+SG.OBJ-with-hold-FUT
'Because life is so enjoyable, we ought to hang on as long as possible'
In the Papuan languages, it appears that obligation is usually expressed by
a marker of obligation with a complement (see 6.3.3).

6.7.4 Abilitive
In Takelma (see 6.6.2) there is a paradigm referred to as 'potential', which is
also for dynamic ability and for general possibility.
In the Papuan languages, ability, like obligation, appears usually to be
expressed by a form indicating ability plus a complement, but an example
from Muyuw (Bugenhagen 1994: 19) might be interpreted as a grammatical
marker used jointly with irrealis (see 6.3.3):
kadiloka b-i-vag
ABILITY IRR-3SG-do
'He can do it'

184
 7
Subjunctive and irrealis

This chapter will deal with a number of issues relating to both subjunctive
and irrealis that have not been addressed, or not fully addressed, in the
preceding two.

7.1 Similarities and differences

The decision to use different terminology for indicative/subjunctive and to
deal with them in separate chapters rests upon a number of considerations.
It is partly a result of different traditions. The terms 'indicative' and
'subjunctive' are the traditional terms used in the description of the classical
and modem languages of Europe, though, not unreasonably, they have been
used by many scholars for other languages. According to Bybee et al. (1994:
236) the earliest use of the terms 'realis' and 'irrealis' noted in an extensive
corpus is to be found in the description of the Australian language Maung in
Capell and Hinch (1970). These terms have been preferred by scholars
working in the Native American languages and in the languages of the
Pacific, particularly those of Papua New Guinea. However, since it is unlikely
that the scholars working in these languages were unaware of the classical
tradition, it is probable that this decision was influenced by perceived
differences in the nature of the data being investigated.
One possible difference that might be suggested is that mood in the
European languages is a morphosyntactic category closely integrated with
person, number, tense and voice. The four categories are not independently
marked, the form being simultaneously the marker of all the grammatical
categories, as well as of the lexical item. Each form is what Matthews (1991:
174) calls a 'cumulative exponent' of person, number, tense, mood and voice.
Such cumulative exponents of grammatical categories (and usually, but not
always, of the lexical item), are of course, the essential characteristic of
languages that are called 'inflectional'.
Mood in the languages where it is characterized as 'realis' and 'irrealis' is
often marked by single words or individual affixes and clitics, as was
185
Subjunctive and irrealis
illustrated in 6.3.1. However, cumulative exponents are also to be found
was shown in that section for Caddo (Oklahoma - Chafe 1995: 352-3) u
it was noted that there are prefixes that are cumulative exponents of r
person - first, second, third and 'defocussing', (ii) markers of the grammati
relations (see Palmer 1994), agent, patient and beneficiary, as well as ("\
realis and irrealis. A few examples are (repeated from 6.3.1):
Agent Patient Beneficiary
Realis prefixes
1st person ci- ku- ku-
2nd person yah?- si- si-
Irrealis prefixes
1 st person ta-/ti- ba- ba-
2nd person sah?- sa?a- sa?u-
There is even some syncretism, a characteristic of inflection, such as that
shown by the formal identity of the first three pairs of patient and beneficiary
above. Two other languages with similar features, Takelma and Alsea, were
also discussed in 6.3.1, where it was noted that Takelma has a morphological
system very like that of Latin.
One similarity is that both subjunctive and irrealis markers are often
redundant, in that the notional irrealis feature is already marked elsewhere in
the sentence. Yet there is also a major difference. The subjunctive is generally
redundant only in subordinate clauses, where the subordinating verb clearly
indicates the notional feature, as in the Italian:
Gli hanno ordinato che tacesse
to.him they.have ordered that be.quiet+PAST+suBj
'They ordered him to be quiet'
After a verb of ordering with the conjunction che, the use of the subjunctive
is obligatory. By contrast, an irrealis marker commonly occurs in main
clauses, where it co-occurs with a grammatical marker that is notionally
irrealis, as with the marker of obligation in Caddo (Oklahoma - Chafe 1995:
356):
kas-sa-ny?aw
OBL-3+AG+IRR-sing
'He should/is obliged to sing'
Syntactically, the structures in which there is redundancy are very different.
At first sight, the notional features associated with subjunctive and with
irrealis may appear different, but only if their functions in main clauses are
considered. For the notions associated with the subjunctive in main clauses
do not include question, denial or futurity (except rarely), while these are
186
 7.2 A binary system

m r n o n l y associated with irrealis. However, in subordinate clauses these

notional features are often associated with subjunctive, often redundantly -
s e e 5.2.3. 5.2.4, 5.2.6.
Perhaps one of the most striking features is that both irrealis and
ubjunctive are used to indicate what is presupposed - known to both
sneaker and hearer (though again with the issue of subordination). Thus the
subjunctive is used in the subordinate clause of the Spanish examples:
Sp. Lamento que aprenda
It. Mi displace che impari
I regret that learn+3sG+PRES+SL)Bj
'I regret that he learns/is learning'
Irrealis similarly occurs to indicate what is presupposed with the 'admirative'
in Caddo (Oklahoma - Chafe 1995: 357 (see 6.6.7)):
hs-ba?a-sayi-k'awih-sa?
ADM-1 +BEN+lRR-name-know-PROG
'My goodness he knows my name'

7.2 A binary system

It has been suggested several times in this book that, basically (prototypi-
cally), the Realis/Irrealis distinction is binary as illustrated by the contrasts
provided by indicative/subjunctive and realis/irrealis. However, there are
obvious examples that seem to be counterexamples and these must now be
considered.
The issue of a binary Realis/Irrealis contrast is not crucial to modal
systems, though it can often be established, and is relevant in that it links
modal systems to mood in the overall category of modality. For instance,
there is a contrast in English between an unmarked Declarative and the
constructions with modal verbs. Similarly in many of the languages with
evidential systems there is an unmarked form and a number of marked terms
that form the modal system. In these languages, it can be suggested that there
is the contrast between Realis and Irrealis. Yet it is not crucial to the analysis,
as shown by the fact that in Tuyuca there appears to be no unmarked form,
in stark contrast to the situation in Central Porno (see 2.7.1).
With the indicative/subjunctive contrast there are two problems. The first
is that there are also forms for jussive and imperative. However, jussives and
imperatives are not as fully inflected as indicative and subjunctive in that
they are never (hardly ever?) marked for tense. The second problem is that
Classical Greek has an optative mood in addition to the other two. However,
t can be argued that the distinction between subjunctive and optative is one
f tense ('modal-tense') rather than mood (see 8.2). >

187
Subjunctive and irrealis
The situation is less clear with languages that have realis/irrealis marki
Yet, there is a binary contrast in most of the languages with joint marki
e.g., Caddo (6.1). Particularly interesting is the situation in Amele, where it '
true that realis/irrealis marking does not extend over the whole language h
is restricted to linked clauses with different subjects (6.3.2). But what '
important is that, where the marking is required, it is binary. A little mo
problematic is the situation where there are unmarked forms as well as thos
marked as realis and irrealis (6.5.2), but, again, where there is marking it j
binary. Yet it must be admitted that, in some languages, the forms labelled
'realis' and 'irrealis' are merely two terms in a wider system (6.5.3). Yet even
there it can be argued that the notional features are essentially those
associated with Realis and Irrealis elsewhere, so that there is, within the
wider system, a Realis/irrealis contrast. Alternatively, such systems can be
treated in terms of modal systems rather than mood.

7.3 The typological status of Realis/irrealis

A fundamental question is whether the distinction between Realis and Irrealis
as exemplified here in terms of mood (though it is applicable to modal
systems as well) is a coherent and homogeneous category that has typological
validity. This is disputed by Bybee et al. 1994, (see also Bybee and
Fleischman 1995b: 9-10, and Bybee 1998). There appear to be four main
arguments.
(i) The distinction is rarely realized in a language as a simple binary
morphological distinction,
(ii) Irrealis and Subjunctive markers are often semantically
redundant in that the meaning is carried by some other element
in the context,
(iii) There is great variation in the notional features marked by them,
which makes it difficult to circumscribe a focal meaning for
them.
(iv) Some of the notional features appear to be wholly inappropriate.
The first argument that 'realis/irrealis is rarely realized in a language as a
binary morphological distinction' (Bybee et al. 1994: 237-8) is an interpret-
ation of the statement by Foley, in his book on Papuan languages (1986:
158) that the 'basic distinction . . . is a binary one . . . but few languages
express it in just this way'. Foley goes on to show that many languages make
'a number of distinctions along the continuum from real to unreal.
Nevertheless there is a significant number of languages in which the binary
distinction is found, and many of them have been well illustrated in the
 7.3 The typological status of Realist'Irrealis
receding two chapters. Roberts (1994: 7-8) provides a large list of
languages in which a binary distinction is marked, and more are to be found
jnBugenhagen(1994).
The second argument, like the first, is rather an overstatement. It is
nerfectly true that both subjunctive and irrealis markers are often redundant.
Rybee et al. (1994: 10) say 'it is difficult to determine in many instances
whether the modal meaning of an utterance is contributed specifically by the
so-called Irrealis marker or by some other element (lexical or morphosyn-
tactic), in which case the Irrealis marker is functionally redundant'. Yet there
are plenty of examples in the two preceding chapters where this is not so.
Thus the subjunctive in main clauses generally contrasts with the indicative
(see 5.1.1). Even in subordinate clauses there is often a choice between them,
e.g. with verbs of belief in Italian and Spanish, depending on the degree of
hesitancy about the proposition, or the degree to which the proposition is
hypothetical (see 5.2.1). Similarly, Mithun (1995: 379) has a very clear and
interesting example from S. Pomo to show that the irrealis marker is not
redundant, since 'realis' and 'irrealis' futures can be distinguished. With realis
the future indicates that the event is portrayed as being certain to happen,
whereas the irrealis marker suggests some uncertainty. Examples (repeated
from 6.6.1) are:

te'nta-lil w-n-hi ?' q-be-w-?khe

town-to go-iMPF-SAME+iRR 1+AG toward-carry-PERF-FUT
'I'll go to town and bring it back'
?' h-w-da m ?b-'-n-i-w-?khe
1+AG nOt-PERF-DIFF+SIM+REAL 2+AG Suffer-DUR-REFL-PERF-FUT
'When I am no longer here, you will suffer'
Even more importantly, there can be redundancy only where there is joint
marking (6.4). Where the marking is non-joint it is clearly not redundant
(6.5), and may distinguish past-present from future time
The facts seem then to weaken or refute the first two arguments. The third
is a little more difficult to deal with, since it must be admitted that there is
great variation among languages in what is treated grammatically as Realis
or Irrealis, especially in the languages in which mood is treated in terms of
realis/irrealis. Thus Chung and Timberlake (1985: 241) say that 'languages
differ significantly as to which events are evaluated as actual (and expressed
morphologically by the realis mood) vs non-actual (and expressed by the
"realis mood)'. Bybee et al. (1994) state that there is considerable misalign
ment between 'the irrealis as defined and the actual distribution of forms in
languages'. Indeed, they say that they did not find a single language in which
189
Subjunctive and irrealis

a single grammatical marker 'could be adequately described as marking fl-

ail the irrealis territory'. However, as Givn (1994: 323) points out th
degree of uniformity apparently demanded by Bybee et al. represents
extreme approach, and there is considerable cross-language variation with
other generally recognized categories. He goes on to illustrate the complexjt
involved with passive. It may be necessary, then, to admit that the notion I
features associated with Realis/Irrealis are more complex and varied tha
those of, say, tense or aspect, but that should not preclude the recognition of
mood as a typologically valid category.
Bybee et al. are particularly concerned with the fact that, in Maune
(Australia) as analysed by Capell and Hinch (1970: 67), negative imperative
is realis and positive imperative is irrealis. This may be a little surprising, but
notionally, both are potentially irrealis. There is probably a historical reason
for the present situation, as there must be for the fact that the words for 'sun'
and 'moon' are masculine and feminine respectively in French, but feminine
and masculine in German, and there are plenty of similar examples - see
Palmer (1984: 34-40). Most cases of variation, however, can be explained, in
the sense that there is some motivation for treating them as either Realis or
Irrealis. This is fairly obvious with statements about the future - they can be
seen as either realis, as assertions, differing only from statements about
present or past in terms of time, or they can be seen as irrealis, because,
unlike the present and past, the future is unknown. The variation with
Interrogative and Negative can similarly be explained - they can be seen as
questions about, or denials of, asserted propositions or simply as non-
assertions (see 6.6.3 and 6.6.4).
It should also be noted that Bybee et al. (1994: 237) actually say 'If one sits
and thinks about what negative, possibility, hypothetical, and imperative
have in common, one could well come up with the idea that they all describe
situations that are unreal or non-actual in some sense.' Moreover, they also
appear to accept the relevance of assertion. This seems to be an admission
that there could well be a notional basis for the distinction.
The fourth argument is based on the fact that the habitual past is treated
as irrealis in Bargam (see 6.6.9). Bybee et al. (1994: 238) say 'if this binary
distinction differs so much across languages that a past aspect, which is
usually considered one of the prototypical cases of realis (Foley 1986: 158ff)>
can be considered irrealis in some languages, this binary distinction is not
cross-linguistically valid'. This is not as strong an argument as might be
supposed, for past is often marked as irrealis if there is negation, interroga-
tion, conditionality, etc. It is only reference to simple acts in the past that is
always marked as realis. Moreover, as suggested in 6.6.9, the past
habitual
190
 7.4 Alternative subordinate constructions

does not refer to actual events but simply to a tendency. As such it is not
entirely surprising (as Givn 1994: 322 says) that it is treated as irrealis in
some languages. It is not the fact of being past that may account for the
irrealis marking, but the fact that it is habitual. Indeed, Givn (1994: 323)
sugg e s t s the habitual is a 'hybrid modality', sharing some features of realis
higher assertive certainty) and some of irrealis ('lack of specific temporal
reference; lack of specific evidence;. . .'). (See 6.6.9.)
A further point is that there is at least one other example of irrealis
marking of habitual past, the evidential suffix in Kashaya (2.4), which has an
evidential modal system (2.2.3). To this might be added the use of would in
English, which was cited by Chung and Timberlake (1986: 221) in relation to
'non-actual mood', but not in connection with the debate about Bargam. The
reply of Bybee et al. (1994: 239) is that 'The past tense of willan "to want" in
Old English was already in use for habitual past . . . centuries before the
future uses of will or the conditional use of would developed. Taking these
two unrelated uses of would and trying to construct some common semantics
does not improve our understanding of linguistic categories.' But this is not
as powerful an argument as they appear to think. For, in the first place, the
Old English usage is obviously to be explained in terms of 'wanting' being
interpreted as 'tending' to act, precisely the explanation given for the Bargam
example. Secondly, since the modern form has lost so much of the semantics
of the Old English, it may well be asked why the habitual sense survived
when the modal auxiliary was formed. A plausible answer is that it did so
because it was semantically consonant with the uses of the new auxiliary.
When thinking about semantic change, it is also important to ask why some
things have not changed. The uses of WILL are not as unrelated as Bybee et
al. suggest. In particular, they share the potentially irrealis meaning of
tending to act, which may, quite easily, be interpreted as either conditional
future (conditional upon other events, see Palmer (1990: 98)) or as habitual.
Roberts (1994: 23) also points out that the 'future tense' in Dyirbal (Dixon
1972: 55) is similarly used for habitual actions and is thus better viewed in
terms of modality than tense. The argument about would, therefore, is not
disproved, nor is the interpretation of the Bargam habitual past, nor, at an
even higher level, is the view that mood in terms of realis and irrealis is a
v
alid typological category.

7.4 Alternative subordinate constructions

It was seen in Chapter 5 that the subjunctive is often used to express irreality
wi subordinate clauses. There are, however, alternatives, and the contrast of
these with the subjunctive can be interesting.

191
Subjunctive and irrealis

7.4.1 Non-finite clauses

The use of the accusative and infinitive construction as an alternative to
construction with subjunctive has been mentioned three times in Chapter 5
in the sections on report (5.2.2), directives (5.3.1) and wishes, fears, etc
(5.3.3).
Latin uses the 'accusative and infinitive' construction for reported
statements:

Dicit se de Gallis . . . postulare triumphum (Liv. 36.4o 31

he.says self from Gauls to.demand triumph
'He says he claims a triumph from the Gauls'

Classical Greek uses a similar construction, but also has a construction with
a finite complement (but with the indicative, not the subjunctive):

oesthe khalkidas te:n hllada so:sein (Dem 9.74)

you.know Chalcidean+PL+ACC the Greece save+FUT+iNF
'You know that the Chalcideans will save Greece'
legon hti K:ros . . . tthne:ke, (Xen. An. 2.1.3)
they.said that Cyrus die+3sc+PLUP+iNDic
'They said that Cyrus had died'
There is a contrast with verbs of wishing and hoping. In Spanish and
Italian both occur with a conjunction plus the subjunctive (5.3.3), but in
Latin an accusative construction is possible for wish, and is the only
construction for hope:
sperabam tuum adventm appropinquare (Cic. Fam. 4.6.3)
I.hoped your arrival come-near+PRES+iNF
'I hoped your arrival would be soon'
These differences may reflect the fact that hopes are more likely to be realized
than wishes - and so less likely to be marked as Irrealis.
For reported commands, Latin generally uses a finite complement with the
subjunctive (see 5.4.2), but with IUBEO T order' it uses the non-finite
construction:
cum . . . eos. . . suum adventm expectare (Caes. B.G. 1.27)
since them his arrival wait.for+PRES+iNF
iussisset
he.had.ordered
'since he had ordered them to wait for his arrival'
There is thus a contrast between the two verbs IUBEO and IMPERO, both
translated as T order' (and there is a further difference in that the notional
object is in the accusative with the first and in the dative with the second):

192
 7.4 Alternative subordinate constructions

Eum iubet venire

he+so+ACC he.orders come+PRES+iNFiN
'He orders him to come'
Ei ut veniat imperat
he+so+DAT that come+3sG+PRES+suBj he.orders
'He orders him to come'
In Latin the choice of construction is determined by the superordinate verb,
but in Italian there is freedom of choice with many verbs, as in:
Gli hanno ordinato di tacere
to.him they.have ordered PREP be.quiet+iNFiN
Gli hanno ordinato che tacesse
to.him they.have ordered that be.quiet+PAST+suBJ
'They ordered him to be quiet'
Among the verbs that take either construction are ABBAIARE 'bark' and
R1NGHIARE 'growl'!
It is not at all clear whether there is a notional difference between the two
Latin verbs, or whether there is a difference between the two Italian
constructions. However, Givn (1994: 281-3) offers an explanation for the
fact that in Spanish some verbs can take either the finite or the non-finite
constructions, while others take only the finite construction. He makes two
suggestions.
First, he compares what he calls the deontic 'manipulation verbs' and
claims that the infinitive invariably signals stronger manipulation than the
subjunctive. Examples are:
Le mandaron seguir-les
him they.ordered follow+iNFiN-them
'They ordered him to follow them'
le mandaron que les-siguiera
him they.ordered that them-follow+3sG+PRES+suBj
'They told him that he should follow them'
te prohibo cantar ;'
you I.forbid sing+iNHN
T forbid you to sing'
te prohibo que cantes
you I.forbid that sing+2sG+PRES+suBj
'I forbid you to sing' (*'I forbid you that you sing')
The second suggestion is that there is a scale of manipulation and that there
will be a changeover to obligatory use of the subjunctive (a 'cut-off point') at
some place on the scale. Thus while both constructions are used with the
verbs in the examples above, only the subjunctive is allowed with:

193
Subjunctive and irrealis

le dijeron que les-siguiera

him they.told that them-follow+3sG+PRES+suBJ
'They told him that he should follow them'
NOT
*le dijeron seguir-les
him they told follow+iNFiN-them

Telling, he suggests, is much weaker manipulation than ordering. The

deontic verbs that he illustrates are:

Stronger, with either construction: MANDAR 'order', PROHIBIR 'prohibit'

OBLIGAR 'make'
Weaker, with subjunctive only: DECIR 'tell', PEDIR 'ask', QUERER 'want'

He also suggests (Given 1994: 283; cf. Givn 1971: 75) that, in respect of
the stronger manipulation signalled by the infinitive, there is a similar
situation in a Bantu language, Bemba. He compares three constructions, the
non-finite construction with an object of the main verb, the finite construc-
tion with an object and the finite construction with no object as in:

n-a-mu-ebele uku-ya
I-REM-him-order iNFiN-go
T told him to leave'
n--mu-ebele (ukuti) a-y-e
1-REM-him-order (that) he-go-suBJ
T told him that he should leave'
n--ebele ukuti a-y-e
I-REM-order that he-go-suBJ
T said that he should leave'

This is very similar to the Spanish sentences with MANDAR and DECIR, but
represented by a single verb, with three grades of manipulative force. He also
considers another verb of ordering with higher force and again finds three
degrees of manipulative force, but with a higher cut-off point.

n--mu-koonkomeshya uku-ya
I-REM-him-order INFiN-go
'I forced him to leave' (He left)
n--mu-koonkoomeshya (ukuti) a-y-e
I-REM-him-order (that) he-go-suBJ
'I ordered him to leave' (He may or not have left)
n--koonkoomeshya ukuti a-y-e
I-REM-order that he-go-suBJ
'I ordered that he should leave'

194
 7.4 Alternative subordinate constructions

Here the non-finite construction has the sense of giving an order that is
carried out (and there is no translational equivalent in English), but there are
again three degrees of force.
Givn also points out that in English there are different possibilities with
ORDER and with TELL:

They ordered him to follow them

Not *They ordered him that he (should) follow them
They told him to follow them
They told him that he should follow them
Of course, this is different from Spanish. In English, ORDER does not allow
the 'subjunctive' (should - see 7.6), because, Givn says, it is too strong a
manipulative verb, while TELL allows both. This might suggest that the
difference between Spanish and English is a matter of the scale of manipula-
tion. The higher verb in Spanish allows both constructions and the lower
verb allows only the infinitive, while in English the higher verb has only the
infinitive and the lower verb allows both constructions, as shown in:

English Spanish
order Infin. mandar Infin. Subj.
forbid Infin. prohibir Infin. Subj.
make Infin. obligar Infin. Subj.
tell Infin. Subj. decir Subj.
ask Infin. Subj. pedir Subj.
However, the situation in English is rather more complicated than this would
suggest. In all the examples above, the noun phrase following the verb is its
object semantically and syntactically. One orders, forbids, makes, tells, asks
somebody to do something. But there are verbs where the only object is the
entire subordinate clause. Thus with / ordered John to come, the likely
meaning is that I ordered John, i.e., that I gave the order to John, but with /
wanted John to come, I did not, in any sense, want John - what I wanted was
'John-to-come'. Now there are many verbs like this in English; they take
either the infinitive or the subjunctive construction. Surprisingly, perhaps,
ORDER is among them, as can be clearly seen where a passive is used in the
subordinate clause:

He ordered the flag to be raised

He ordered that the flag be raised/should be raised
It is obvious that orders were not given to the flag. However, this is rather
curious, since PREFER and REQUIRE allow both constructions, WANT and
EXPECT allow only the infinitive construction, while SUGGEST and PROPOSE
permit only the subjunctive, and TELL does not fit in here at all:

195
Subjunctive and irrealis
I prefer/require the flag to be raised
I prefer/require that the flag (should) be raised
I want/expect the flag to be raised
*I want/expect that the flag (should) be raised
I suggest that the flag (should) be raised
*I suggest the flag to be raised
1 propose that the flag (should) be raised
*I propose the flag to be raised
*I told the flag to be raised
*I told that the flag (should) be raised
These cannot be directly compared with Spanish sentences, because, in this
type of construction, Spanish permits only the finite form with subjunctive
Yet Givon's scale of manipulation may still be relevant in that all the
stronger verbs allow the infinitive and only the weaker verbs, SUGGEST and
PROPOSE, do not.
Yet there is a problem with TELL. Why does it not fit in here at all? Why is
it not like ORDER? The most obvious reason is that it always requires an
animate object when it has a complement. One always has to tell someone.
Another reason may be that it is mainly used for reported statements rather
than reported commands. This is clear enough in:
I told them that I was coming
But this may also be true of:
They told him that he should follow them
Should here is not the 'subjunctive' should within an indirect command, but
the modal verb used for obligation within an indirect statement. The evidence
for that is that the simple form as 'subjunctive' cannot be used in this
construction:
!He told them that they follow him
This cannot mean 'He told them to follow him', but only 'He told them that
they (in fact) follow him.'
However, in Italian, the verb DIRE 'tell/say' is one of the many verbs that
take either construction - it has none of the restrictions of DECIR in Spanish
and TELL in English.
The choice between the two constructions is not, then, clearly motivated
by notional features, though there is a tendency for the subjunctive to signal
a greater degree of irrealis.

7,4.2 Direct speech

Another alternative to the subjunctive in a subordinate clause is the use of
direct speech, the reporting of the actual words of the original utterance.

196
 7.4 Alternative subordinate constructions

Direct speech is often used for reporting statements (where the use of the
subjunctive is not very common - see the opening paragraph of the last
section), but it can also be used for reporting questions and commands
(where the use of the subjunctive is fairly common):
He said 'I'm coming'
I asked 'Are you coming?'
She said 'Come!'

Direct speech is almost certainly to be found in all languages, either as an

optional alternative to indirect speech, or as the only way of reporting. Some
examples of direct speech (all for reporting commands) may be taken from
Kobon (New Guinea - Davies 1981: 2), Yidiny (Australia - Dixon 1977: 524)
and Syrian Arabic (Cowell 1964: 450):
nipe ip hag-p ne fiel ud-ag-0 QUOTE g-op
3SG ISG say-PERF+3sc 2so firewood take-NEG-2sG a do-PERF+3sG
'He said to me "Do not take the firewood" '
damarirjgu budiiji bana: bucji diga mamba
Damari+ERG tell+PAST water+LOC put+imp pour.water.on+iMp sour+ABS
'Damari told him "Put it in the water, pour water on it, it's bitter" '
marra w-marrtn ?alt-allo l talab battari
time and-time I.told-to.him not play+iMP in.the.street
'Time and time I've told him "Don't play in the street"'

Rather more interestingly, there are languages that employ direct speech
constructions much more widely than in English. Thus in spoken Amharic
(Semitic, Ethiopia - Cohen 1936: 363), an expression of fear can be
represented with the verb 'fear' plus the verb 'say' followed by a direct
quotation, even though no actual words were spoken as in:

mangadum aagarmaum bsyye afarallauh

the.road he.will.not.find.it I.saying I.fear
'I am afraid he won't find the road'

It is even possible to express 'refusal' by an inanimate object in the same way:

saga albasbmm ala
meat.I.will.not.cook said
'The meat refused to cook'
It has been claimed for a number of languages that they have no indirect
speech, but it may be more accurate to say that indirect speech is rarely used
in some of the languages. For instance, it appears that generally the
Australian languages employ only direct speech, but consider this example
from Ngiyambaa (Australia - Donaldson 1980: 280):
197
Subjunctive and irrealis
rjadhu-na niyiyi girma-li ninu:
I+NOM-3ABS say+PAST wake-puRp you+OBL
'I told her to wake you'

N ow the actual words would have been 'Wake him', so that this is clearly not
direct speech. In fact, 'deictic shift' - changing pronouns, as here (and also
other deictics such as time markers in some languages, e.g. English), is an
essential feature of indirect speech. For English compare two sentences in
which four deictics have been changed - pronouns, tense and temporal
pronouns:

He said 'I'm coming tomorrow'

He said he was coming the next day

7.5 Speaker commitment

There is a further point to be made about indirect speech. Some languages
have devices to indicate whether the speaker believes the reported utterance
to be true. In English, there may be no deictic shift involving change of tense,
if a reported statement is accepted as true by the secondary speaker:
I said 'I'm coming on Tuesday'
I said that 1 was coming on Tuesday
I said that I'm coming on Tuesday
The third example suggests that it is still the speaker's intention to come. A
very clear example is given by Jespersen (1909-49: iv,156):
The ancients thought that the sun moved round the earth: they did not
know that it is the earth that moves round the sun
The point here, quite clearly, is that the first proposition, with moved, is not
accepted by the speaker, while the second, with moves, is accepted as true by
him.
This can, obviously, be regarded as notionally a modal feature, and, in
German, grammatical mood is involved, in that although - as noted in 5.2.2
- with past tense verbs, literary German uses the subjunctive in indirect
speech, the indicative can be used to indicate confidence by the speaker:
Er sagte, er schwimmt gern
he said he swim+3so+pRES+iND with.pleasure
'He says he likes swimming'
There is a contrast in:
Ich glaubte dass er ware krank
I thought that he be+3sc+lMPF+suBj ill
'I thought he was ill'
198
 7.5 Speaker commitment
Ich glaubte dass er krank war
1 thought that he ill be+3so+iMPF+iND
'I thought he was ill'
gut the choice of mood depends also on stylistic and sociolinguistic factors
(see Hammer 1983: 265-8). Notice that there is no change of tense as there is
in English, only a change of mood.
Similarly, Classical Greek uses the optative mood with past tense verbs of
reporting, but may also use the indicative, with the original (direct speech)
tense:
e:peile:s' hod . . . badioime:n (Ar. PI. 88)
I.threatened that go+lsG+FUT+OPT
'I threatened that I would go'
e:nglthe aut:i hti Mgara aphsteike (Thuc. 1.114)
it.was.reported to.him that Megara revolt+3sG+PERF+iND
'News came to him that Megara had revolted' (not 'has revolted')
There are some examples with both constructions in the same sentence, with,
presumably, some indication of speaker commitment, as in:
legon hti K:ros mn tthneike,
they.said that Cyrus on.the.one.hand die+3sc+PLUP+iND
Ariaos d pepheug:s . . . ee: (Xen. An. 2.1.3)
Ariaeus on.the.other having.fled be+3sc+PRES+0PT
'They said that Cyrus had died, but that Ariaeus had fled'
In some languages, e.g. Japanese (Kuno 1973: 261), deictic shift does not
affect tense - the reported speech retains the tense of the actual utterance.
However, Japanese has three different conjunctions, no, koto and to (Kuno
1973: 213-22). It appears that the choice of no or koto rather than to depends
on the degree of the speaker's commitment to the truth of the proposition
(Suzuki 1994, in a reexamination of Kuno's arguments). According to Kuno
(1973: 216-17), there is a choice of conjunctions in the following:
John wa Mary ga sinda to sinzinakatta
John Mary died that not-believed
'John did not believe that Mary was dead' (She might or might not have
been)
John wa Mary ga sinda koto o sinzinakatta
John Mary dead that not-believed
'John did not believe that Mary was dead' (She was)
For this reason 'to forget' cannot occur with to, but only with koto (or no):
John wa Mary ga tunbo de aru koto/no (*to) o wasureteita
John Mary deaf is that had.forgotten ,
'John had forgotten that Mary was deaf

199
Subjunctive and irrealis

In contrast, a 'wrong notion' can occur only with to (Suzuki 1994: 526):
John wa Mary ga shinda to (*koto/no) o gokaishita
John Mary died that formed.a.wrong.notion
'John formed the wrong notion that Mary had died'
Similarly, in Kinyarwanda (Bantu, Rwanda - Givn 1982: 26-32, quotin
Givn and Kimenyi 1974), there are three conjunctions ko, ngo and konso
each expressing different kinds of commitment by the (actual) speaker:
ya-vuze ko a-zaa-za
he+PAST-say that he-FUT-come
'He said that he'd come' (and I have no comment)
ya-vuze ngo a-zaa-za
he+PAST-say that he-FUT-come
'He said that he'd come' (but I have direct evidence which makes me
doubt it)
ya-vuze kongo a-zaa-za
he+PAST-say that he-FUT-come
'He said that he'd come' (but I have indirect/hearsay evidence which
makes me doubt it)
Factive, non-assertive, complementizers such as 'regret', 'forget' require ko,
the marker of the strongest degree of commitment.
Yet another language to make a similar distinction is the Jacaltengo
dialect of Jacaltec (Mayan, Guatemala - Craig 1977: 268). There is a contrast
between chubil ('corresponds to a factive " t h a t " ' ) and tato ('introduces an
expected, supposed or believed fact'). Both can be used with a verb of saying
to indicate the reliability or unreliability of what is said:
xal naj tato chuluj naj presidente
said he that will.come the president
'He said that the president is going to come'
xal naj alcal chubil chuluj naj presidente
said the alcalde that will.come the president
'The alcalde said that the president is going to come'
The first indicates an unreliable source of information, the second a reliable
source.
In some languages, the speaker's doubt about the truth of what is reported
may be indicated by a particle, e.g. (according to Frajzyngier (1985: 244-5))
Polish, Czech and Mapun (Chadic, Nigeria). Polish and Mapun examples
are:
Powiedzial, e niby jest chory
he.said that BE+3SG DOUBT sick
'He said that he is sick, but I am not sure that it is true'
200
r wur sat mun ni din paa
7.6 The subjunctive in English
yol muan
3SG say IPL that 3SG+COREF DOUBT go trip
'He told us that he went on a trip, but maybe he didn't go'
In Polish, but not Mapun, the particle may be used where the subject of the
main verb is first person, to indicate that the reported statement is a lie (by
the speaker):
Powiedsiatem, e niby jestem chory
I said that be+3sc DOUBT sick
'I said that I am sick, but I am not'
Slightly differently, in Lega (Bantu, E. Zaire - Botne 1997: 512-14) a
particle that is commonly used as Reported may be used to express doubt
about a statement (see 2.2.4):
nkumgwg (bono), mbo bazongo b kulyg merende
I.hear that whites QUOT eat frog
'I hear that Westerners eat frogs (though I find that unlikely)'
It expresses disagreement when the statement is about the speaker:
bbol bono mbo nne nekobkes j
they.tell that QUOT I tricked
'They say that I tricked them (though I take exception to it)'

7,6 The subjunctive in English

There are three constructions that have been thought to be subjunctive in
English.
One is the use of were in If I were rich, but this is very marginal, partly
because it is restricted to conditional sentences, but also because, strictly,
there is nothing unusual about were. With all other verbs there is just one past
tense form - came, liked, walked, etc., and this is also true of were in
conditional sentences: / were, you were, they were, etc. In that sense were is
quite regular. What is irregular is the use of was in the singular for all other
past tense forms, for no other verb has different forms for singular and plural.
A second construction is with the simple form in sentences such as
I demand that he come tomorrow. This use is becoming rare in English,
especially in British English, though it seems to be alive in written American
English, at least, although grammarians bewail its demise. But it is not a
distinct form, like the subjunctive in other languages: it is merely the simple
form of the verb, which has several functions in English, particularly those of
the infinitive and the imperative. Little is gained by calling it 'subjunctive'
rather than acknowledging that it represents one function of the simple,
uninflected, form.

201
Subjunctive and irrealis

A third possibility is the use of should in, for instance, It's odd that you
should say that. This, obviously, is the use of a modal verb, and as such
comparable with the subjunctive. However, formally, this is the past tense of
a modal verb, not a distinct paradigm like the subjunctive of this chapter
(For a similar reason there is, strictly, no future tense in English - see
Huddleston (1995).)
One striking similarity between should and the subjunctive is its use in
constructions where the proposition is presupposed, e.g. in the Italian
example from 5.2.5 and one of its English translations:
mi sorprende che tu dica questo
me it.surprises that you say+2so+PRES+iND that
'It surprises me that you say that'
'It surprises me that you should say that'
It should be added that should can refer to the present or the past. The
second English sentence above is equivalent to either of the two:
'It surprises me that you say that'
'It surprises me that you said that'
Quirk et al., (1985: 1014) refer to this use as 'putative', and suggest that it
refers to putative situations that are recognized as possibly existing or
coming into existence. They compare:
I'm surprised that he should feel lonely
I'm surprised that he feels lonely
They go on to suggest that the first questions the loneliness, while the
second accepts it as true. Similarly, they suggest that there is a non-factual
basis when the meaning is close to that of a conditional clause:
It's a pity that they should be so obstinate
It's a pity if they are so obstinate
Another use of should is in conditional sentences, such as:
If John should come, Bill will leave
Cf. If John comes, Bill will leave
The use of should here expresses some uncertainty. (This is not an example of
an unreal conditional, the topic of 8,3. It is quite different from If John came,
Bill would leave. See 8.3.4.)
It is not very important whether or not these uses of should Are called
'subjunctive'. They have some parallel with subjunctives in other languages,
but, on the other hand, the term 'subjunctive' is generally used to indicate an
inflectional category. Should is formally merely the past tense form of a
modal verb.

202
I

8
Past tense as modal

It was mentioned in 1.4.4 that past tense often has a modal function. There
is, in fact a considerable body of literature that discusses this use of past
tense forms to express 'unreality'.

8.1 'Real' and 'unreal'

'Unreality' in this sense has much in common notionally with irrealis, in that
it indicates some degree of lack of confidence by the speaker, but is best
treated as a different feature for several reasons. First, it is marked differently
- by tense. Secondly, it co-occurs with markers of irrealis, in both mood and
modality systems; in fact, 'real' and 'unreal' in this sense mark a further
distinction, another parameter within a wider field of modality. Not surpris-
ingly, its functions are often quite different from those of irrealis.

8.2 Modal-tense
'Unreality' is not a satisfactory term, since it is too like 'irrealis' and,
unfortunately, the terms 'realis' and 'irrealis' have been used in traditional
grammars to refer to 'unreal conditionals' (below, 8.3). Instead the terms that
will be used are 'modal-past' and, where relevant, 'modal-present' and
'modal-tense'.

8.2.1 Modal verbs

There was a discussion in 2.1.5 and 3.2.3 of the use of past tense forms of the
niodals to provide weaker, more tentative, forms of the judgments and
directives. These are, obviously, what are now being called 'modal-past'.
Examples of the past tense forms of MAY and WILL as tentative epistemic
judgments (Speculative and Assumptive) are:
He might be there (Cf. He may be there)
He would be on holiday now (Cf. He will be on holiday now)
MUST has no past tense form, but ought to and should appear to fill the gap:
203
Past tense as modal
He ought to/should be there by now (Cf. He must be there by now)
Formally should is the past tense form of SHALL, but it expresses tentative
epistemic necessity (Deductive) and so is notionally one of the moda]-n a s t
forms of MUST.
With deontic modals there is one-to-one correspondence only with WILL-
I'd do that for you (Cf. I'll do that for you)
Yet while may and can give permission, the corresponding past tense forms
might and could are normally used to make suggestions:
You might/could ask your father
Ought to and should are again available for MUST:
You ought to/should come tomorrow (Cf. You must come tomorrow).
Commissive SHALL has no (notionally) corresponding modal-past form.
Because of these anomalies many scholars have argued might, could and
should are not the past tense forms of MAY, CAN and SHALL. But the relation-
ship in terms of tense can be demonstrated in two ways. First, by the
'sequence of tense' rule, present tense forms change to past tense after a past
tense verb of reporting. This applies to the modals as well as to full verbs:
'I am coming'
He said that he was coming
'He may be there'
She said that he might be there
'You may come in'
He said that I might come in
'I'll do it for you'
She said that she'd do it for me
Secondly, these past tense forms are used with exactly the same relationship
to the present tense forms in unreal conditionals, the subject of 8.3.
This use of tense is not confined to modal verbs. It is also to be found in
polite, tentative, requests and questions, such as:
I wanted to ask you a question
Did you want to speak to me?
However, as Bybee (1995: 506) points out, WANT is notionally modal. This
tense usage is not found with other types of verb.

8.2.2 Greek optative

In addition to the subjunctive, Classical Greek has an optative mood. One of
its functions is in the expression of wishes:

204
 8.2 Modal-tense

i pa gnoio patrs eutukhsteros (Soph. Aj. 550)

O child become+2sG+AOR+opT of.father luckier
'My child, mayst thou be luckier than thy father'
It is also used in Homeric Greek for suggestions, rather like English might:
tu epois Akhil:i
these say+2sG+AOR+OPT to.Achilles
'You might tell Achilles this'
Since there are, apparently, three moods in Classical Greek, this might
seem to be an argument against the binary nature of realis and irrealis, but,
in fact, the optative can be seen as a modal-past form of the subjunctive.
Morphologically, there is evidence for this. To begin with, the subjunctive
has similarities to the present tense, while the optative has similarities to the
imperfect, as can be seen in the following forms (first person, middle voice):

Present indicative luomai

Subjunctive luoimai
Past indicative eluome'n
Optative luoimem

More importantly, a distinction is traditionally made between 'primary'

and 'historic' tenses, which, with slight complications, is essentially the
distinction between present and past tense, and this is of syntactic relevance
(see below). The present and imperfect tenses are primary and historic
respectively, but they differ only in the indicative - there is only one
subjunctive and one optative paradigm associated with the pair of them. The
same is true of the perfect and the pluperfect, which can be seen as differing
from the present and perfect in aspect, not tense, with the pluperfect being
the past of the perfect. (The traditional terminology is misleading. The
present and imperfect would be better seen as the present and past of the
imperfect, and the perfect and pluperfect as the present and past of the
perfect.) With these tenses, then it could well be argued that the subjunctive
and optative also differ in tense, the subjunctive being present and the
optative being past:

Imperfect
Indicative Subjunctive
Present Past Present Past
'present indicative' 'imperfect indicative' 'present subjunctive' 'present optative'
Perfect
Indicative Subjunctive
Present Past Present Past
'perfect indicative' 'pluperfect indicative' 'perfect subjunctive' 'perfect optative'

205
Past tense as modal

The suggestion that Classical Greek has perfect and imperfect aspects
within which there are tense distinctions, is not new, but it has not been
generally recognized that formally there are two other categories that belonp
to the same paradigm as imperfect and perfect. These are the aorist and the
future, so that it could be argued that there are four, not two, aspects
(though 'aspect' might not be the most suitable term). The reason for arguing
that there are four 'aspects' is that each has its own infinitive and participles
(not differentiated for tense), plus the fact that they also have subjunctives
and optatives.
There are, however, some differences with the aorist and the future. The
aorist does not have morphologically distinct present and past forms, yet it
has both a subjunctive and an optative. However, syntactically, the aorist
indicative can be either primary or historic, and so could be said to have a
primary (present) subjunctive and a historic (past) optative:

Aorist
Indicative Subjunctive
Present Past Present. Past
'aorist' ('primary') 'aorist'('historic') 'aorist subjunctive' 'aorist optative'
The future has only an optative, so that the pattern would seem to be:
Future
Indicative Subjunctive
Present Past Present Past
'future indicative' none none 'future optative'
However, the absence of a subjunctive may be thought to be due to the close
association of the future with the subjunctive (and the fact that both are
notionally irrealis).
The syntactic argument for Primary/Historic and so for the optative as
modal-past is that the optative is required instead of the subjunctive in some
types of subordinate clause if the superordinate verb is Historic. This is
clearly so with purpose clauses, as illustrated by:
tn kakn dei kolzein hn' amenon
the bad it.is.necessary to.punish in.order.that better
e:i (Plat. Legg. 944D)
be+3sG+PRES+suBj
'We must punish the bad man to make him better'
pempsa h:s pthoito
I.sent that learn+3sG+AOR+OPT
'I sent him that he might learn'
In the first example the main verb (dei) is present and primary, and the

206
 8.3 Real and unreal conditions

subordinate verb is in the subjunctive. In the second example the main verb is
aorist and historic, and the subordinate verb is in the optative.
The optative is also used to report a present tense form in the original
utterance:
legon hti pants ksia lgoi sethe:s (X. An. 8.3.13)
they .said that of.all worthy say+3sc+PRES+0PT seuthes
'They said that Seuthes's words were all-important'
The original words would have contained a present indicative (lgei, 'says').
English similarly would use a past tense after a past tense verb of reporting,
as the translation shows, where were reports an original are.
Similarly the optative is used for a reported cause, after a historic main
verb (here an aorist participle), where Latin would use the subjunctive (see
5.2.2):
ostha epainsanta autn tn Agammnona ho:s basileste te
you.know praise+AOR+PART he the Agamemnon as king both
ee: gathos, kraters t aikhme:t:s (Xen. Symp. 4.6)
be+3sG+PRES+oPT good, mighty and warrior
'You know he praised Agamemnon because (he said) he was both a
good king and a powerful warrior'
There is no need, therefore, to recognize optative as another mood,
different from both indicative and subjunctive. It can be seen as a modal-past
subjunctive.

8.3 Real and unreal conditions

The most striking and most widespread use of past tense for unreality is in
conditional sentences. In a large number of languages a clear distinction can
be made between real and unreal conditions, the latter being used where the
speaker indicates some kind of negative attitude towards the condition
expressed in the protasis (the '//"-clause'). In some languages this is marked
with tense alone, in others by a combination of tense and mood.

8.3.1 Marking by tense

In English (see, e.g., Palmer 1990: 169) there is a contrast between:
If John comes, Bill will leave
If John came, Bill would leave
These are examples of a real and an unreal conditional sentence. Notionally
the difference is that, with the unreal conditional, the speaker indicates some
doubt about the likelihood of the event indicated in the protasis (John's
coming), whereas, with the real conditional, the possibility is simply left
open. Formally, the difference between them lies in the tense marking in both

207
Past tense as modal

clauses, the real conditional having present tense verbs, the unreal condi-
tional having past tense verbs in both protasis and apodosis (the main
clause). To convert the real into the unreal, comes is replaced by came and
will by would.
There is a corresponding unreal conditional in the past, which is:
If John had come, Bill would have left
Here past is marked twice, once for past time, once for unreality, the markers
being past tense had, would and HAVE, since HAVE functions in English as both
a marker of perfect aspect and of past time (for details see Palmer 1990: 170).
Conditionals of this type, which may be called 'predictive' since they
predict, in terms of cause and effect, the occurrence of one event on condition
of another, always require a m o d a l verb in the apodosis (both real and
unreal), e.g.:
If John comes, Bill can leave
If John came, Bill would leave
If John had come, Bill might have left
Non-predictive conditionals are such as If John came, Mary left, which
merely indicate that if the protasis is true, so is the apodosis, without
necessarily implying that there is cause and effect. There are no restrictions
on the tense or modality of such sentences.
Examples of this use of tense for unreality can be found in other languages.
Thus in Classical Greek the imperfect indicative is used in both clauses for
present unreal conditional, and the apodosis also contains the particle n:
ei touto epoeis, he:mrtanes n
if this do+2sG+MPF+iNDic err+2sG+lMPF+iND:c n
'If you were doing this, you would be wrong'
For past unreal conditions, it uses the aorist, instead of the imperfect:
ei touto epoe:sas, himartes n
if this do+2sG+AOR+iNDic err+2sG+AOR+iNDic n
'If you had done this you would have been wrong'
However, where there is clear future reference, real conditionals use the
subjunctive (or the future) in the protasis and the future in the apodosis, and
unreal conditionals use the optative in both:
ei touto epoeiis, hamartiseis
if this do+2sc+PRES+suBj err+2sG+PRES+FUT
'If you do this, you will be wrong'
ei touto epoioe:s, hamartnois n
if this do+2sG+PRES+OPT err+2sG+PRES+0PT n
'If you did this, you would be wrong'

208
 8.3 Real and unreal conditions

Although mood is required here, the difference between real and unreal is
marked by subjunctive and optative. Essentially, then, in the light of the
arguments in 8.2.2, the distinction is still one of tense.
There is a similar use of tense to mark unreality in the Ethiopian Semitic
languages Tigre and Tigrinya, spoken in Eritrea (personal research). Tigre
has a past tense (identified with the Semitic 'imperfect'), plus the particle wa
in both clauses:
London wa-gasko, wa-matko
London coND-go+1 SG+PERF coND-die+1 SG+PERF
'If I went to London, I would die'
Tigrinya has the present (imperfect) in the protasis, but a past (perfect) in the
apodosis:
'ant taharm-o, ma-mote
if hit+2sG+iMPF-him coND-die+3sG+PERF
'If you hit him, he would die'
In a neighbouring Cushitic language, Bilin (Palmer 1957), unreality is also
marked by past tense but in a very different way. In this language the verbs
have a large number of paradigms (inflected for person, gender and number),
but these paradigms can be classed in terms of certain features of vowel
quality and tonal patterns into two types, A and B. With the verb class
:
llustrated below, the vowel immediately after the stem is open and without
igh tone in A, but front close and with high tone in B:
A B
jabkw he buys jabixw he bought
jabaxw which he buys jabxw which he bought
jabat (says) that he buys jabet (says) that he bought
As these examples show, many of the paradigms are paired, B referring to
past time, A to present time. There are two forms for conditionals (used in
both protasis and apodosis). These are quite different in form, but the
paradigm used for real conditions is of Type A and the paradigm used for
unreal conditions is of Type B (compare the vowels after the stem and the
tones):
jaban 'will buy/if he buys'
jabndik 'would buy/if he bought'

8.3.2 Marking by tense and mood

Often unreal conditions are marked by both mood and past tense. For unreal
conditionals in the present, Latin uses the imperfect subjunctive in both
clauses:

209
Past tense as modal

Si hoc faceres, errares

if this do+2sG+iMPF+suBj err+2so+iMPF+suBj
'If you did/were doing this you would be wrong'
However, Latin, like Classical Greek, has a different set of forms for unreal
conditionals with clear future reference - it has present subjunctive in both
clauses:
Si hoc facias, erres
if this do+2sG+PRES+suBJ err+2sc+PRES+suBj
'If you did/were to do this you would be wrong'
German is a little more complex. It has the imperfect subjunctive in the
protasis but either the imperfect subjunctive or the 'conditional tense' in the
apodosis (Hammer 1983: 253). However the 'conditional' tense is, in fact,
formed by the imperfect subjunctive of the modal WERDEN, which often
marks the future. Examples are:
Wenn ich ihm heute schriebe, bekme
if I to.him today write+lSG+iMPF+suBJ get+3sG+lMPF+suBj
er den Brief morgen
he the letter tomorrow
Wenn ich ihm heute schriebe, wtirde
if 1 to.him today write+lsG+iMPF+suBJ be+3sG+iMPF+suBJ
er den Brief morgen bekommen
he the letter tomorrow to.get
'If I wrote to him today, he would get the letter tomorrow'
Spanish also uses 'conditional tenses' and has either the imperfect sub-
junctive or the conditional in the protasis, but the conditional or the
conditional subjunctive in the apodosis:
Si yo tuviese/tuviera bastante dinero,
if I have+lSG+lMPF+suBj/have+lsG+coND enough money,
comprara/comprara otro automvil
buy+lsG+iMPF+suBj/buy+lSG COND+SUBS other car
'If I had enough money, 1 would buy another car'
However, Spanish, unlike English and German, does not employ a modal
verb, but rather its (notionally very similar) future tense, and combines this
with past.
For unreal conditionals in the past, Latin simply uses a more remote past
marker, the pluperfect instead of the imperfect, but still with the subjunctive:
si hoc fecisses, errasses
if this do+2sG+PLUP+suBj err+2sc+PLUP+suBJ
'If you had done that you would have been wrong'

210
 8.3 Real and unreal conditions

German uses its 'have' form (HABEN), like English, to add the extra past tense
(yielding what are, predictably, called the 'pluperfect subjunctive' and the
'conditional perfect'):

Wenn ich ihm heute geschrieben htte,

if I to.him today written have+lsG+iMPF+suBj
htte er den Brief morgen bekommen
have+so+iMPF+suBJ he the letter tomorrow got
Wenn ich ihm heute geschrieben htte,
if I to.him today written have+lsG+iMPF+suBj
wiirde er den Brief morgen bekommen haben
be+SG+MPF+suBj he the letter tomorrow got to.have
'If I had written to him today, he would have got the letter tomorrow'

In a similar way, past unreal conditionals in Spanish simply involve the

addition of HABER (hubiese and habria).

8.3.3 Interaction of modality and unreality

Not surprisingly, perhaps, there is some interaction between modality and
(conditional) unreality. This is most striking in Latin, but is most easily
explained by first considering some English examples.
It has been argued that there are differences in the uses of could have in:

He could have jumped six feet, if he'd trained hard

He could have jumped six feet, if he'd wanted to

The first example says that training hard would have resulted in the ability to
jump six feet, whereas the second example says that wanting to would have
resulted in jumping six feet, not in the ability to do so, for that ability was
already there. In other words, could have indicates 'I was able to, and would
have' not '1 would have been able to'. Austin (1956: 163 [1961: 164]) suggests
that / could have can be either a 'past indicative' or a 'past conditional' and
may be equivalent either to Latin potui (T was in a position to') or potuissem
(T should have been in a position to').
A clear example of the pluperfect subjunctive is:

quod certe si essem interfectus, accidere

which certainly if be+lsG+iMPF+suBj killed to.happen
non potuisset (Cic. Sest. 22.49)
not can+3sG+PLUP+suBJ
'Which certainly could not have happened if I had been killed'

In contrast the perfect indicative is used in:

211
Past tense as modal

deleri totus exercitus potuit, si fugientis

be.destroyed whole army can+3sG+pERF+iND if those.fleeing
persecuti victores essent (Liv. 32, 12 61
pursue+3PL+PLUP+suBj victors (discontinuous with persecuti)
'The whole army could have been destroyed, if the victors had pursued
those who were fleeing'

The sense is that it was possible for the whole army to be destroyed (that the
victors were in the position of being able to destroy it), not that it would have
been possible. Handford (1947: 132) sees the choice in terms of what is
'logically preferable'.
This phenomenon is found with other expressions that are notionally
modal, such as LICITUM esse 'be allowed', MALLE 'prefer', DEBERE 'ought',
ocCASio ESSE 'be a chance' (Handford 1947: 131). An example of the last is:

quern si interfere voluisset, quantae quotiens

whom if to.kill wish+3sc+Pnjp+suBj how.many how.often
occasiones fuerunt (Cic. Mil. 38)
occasions be+3PL+perf+iND
'If he had wished to kill him, how many chances, how often, there were!'

The same feature is found with present unreal conditionals, as in English

with could:

He could jump six feet, if he trained hard

He could jump six feet, if he wanted to

So in Latin with LICITUM ESSE, with the present indicative instead of the
imperfect subjunctive:
licitum est si velles (PI. 7>/. 566)
allowed be+3so+PRES+iND if WS1I+2SG+IMPF+SUBJ
'You could if you wished'

In other examples there are different combinations of tense, but all with the
subjunctive in the protasis and the indicative in the apodosis:

idque si nunc memorare hie velim . . . vere

and.that if now to.remember this wish+lsG+pRES+suBj truly
possum (Ter. Hec. 471)
can+ ISG+PRES+IND
'If I now chose to recall this, I could do so with truth'
si ita sententia esset . . . tibi servire
if thus opinion be+3sc+iMPF+suBj to.you to.serve
malui (PI. Miles 1356)
prefer+ lSG+perf+iND
'If it were your view, I would have preferred to serve you'
212
 8.3 Real and unreal conditions

eum contumeliis onerasti quem patris loco, si ulla in te

him with.insults you.loaded whom of.father in.place if any in you
pietas esset, colere debebas (Cic. Phil. 2.38.99)
affection be+3sc+iMPF+suBj to.honour ought+2sc+iMPF+rND
'You loaded with insult the man whom, if you had any affection in you,
you ought to have honoured as a father'

However, with some examples the 'logical' explanation is not available.

Handford (1947: 132) comments that (with POSSE at least) 'a Latin author's
choice of mood depended largely on personal taste, the whim of the moment,
metre or clausula'. He quotes:

non potuit . . . fieri sapiens, nisi

not can+3sG+PERF+iND become wise unless
natus esset (Cic. fin. 2.31.103)
born be+3sG+iMPF+si>BJ
'He could not have become wise, if he had never been born'

here is one passage in which there is a contrast between the pluperfect

ubjunctive and the perfect indicative (in the apodosis) in two consecutive
entences. The first is:

in qua quid facere potuissem, nisi turn consul

in which what to.do can+lso+PLUP+suBJ unless then consul
fuissem?
be+1 SG+PLUP+SUBJ
'What could I have done in that (crisis) if I had not then been consul?'

The second is:

Consul autem esse qui potui,

consul however to.be who can+lsG+PERF+lND
nisi eum vitae cursum tenuissem? (Cic. Sest. 22.49)
unless that of.life course hold+lsG+PLUP+suBj
'But how could I have been consul, if I had not held to that manner of
life?'

Certainly the 'logical' explanation does not explain the second sentence, for
the manner of life was the cause of the ability to become consul. However,
Cicero was, in fact, able to become consul, and it may be that this is the
reason for the choice of the indicative.
Giulio Lepschy (personal communication) has pointed out that there is
something similar (but not quite the same) in Italian. The imperfect indicative
potevo (= Latin potui) may be preferred to the 'past conditional' (formed by
the conditional form of the verb 'to have' plus the past participle) if there is
an implication that the ability was already in place, as in:

213
Past tense as modal

Potevi dir-me-lo!
can+2so+iMPF+iND tell-me-it
'You might have told me!'
(The 'past conditional' Avresti potuto dirmelo is less natural.) The same is
true of the protasis of past unreal conditional. Thus the imperfect indicative
is possible in the first example, but almost impossible in the second:
potevo farlo se mi aiutavano
can+lsG+iMPF+iNDic do-it if me you.had.helped
'I could have done it if you had helped me'
avrei potuto farlo se fossi stato piii vecchio
can+lso+PAST.coND do-it if I.had been more old
'I could have done it if I had been older'
The essential point here is that, with the first, there is an implication that I
had the ability and merely needed help to fulfil it, but with the second it is
clear that I did not have the ability, but would have had it if (impossibly) I
had been older.
However, the Latin examples show that there is not always a 'logical'
explanation, although the use of the indicative is more likely when there is an
implication that the modal feature was already-in-place ability. The explana-
tion may be simpler - that, since modal expressions are irrealis, it may seem
redundant to mark them as unreal as well. There is, in fact, some support for
this in English, in that, as Bybee (1995: 506) points out, it is possible for
notionally modal expressions to be used in the apodosis of unreal condi-
tionals, instead of a modal verb as in:

I wanted to
I was gonna
If I saw Judy \ , ,. tell her the news
1
I was supposed to
I intended to
(But Bybee admits that some of these are not 'highly grammaticized'.)

8.3.4 Further possibilities

Not every language makes the distinction between real and unreal condi-
tionals. According to Akatsuka (1985: 627), this is the case with many East
Asian languages, such as Japanese, Chinese, Korean, Mongolian, Semai and
Thai. An example from Japanese is:
musoko ga ikite i-tara ii noni naa!
son SUB alive be-if good though EXCL
'If my son is alive, I'll be so happy'
'If my son were alive, I'd be so happy' . ,

214
 8.3 Real and unreal conditions

Some languages make the distinction between real and unreal conditionals,
but make further distinctions as well. Thus, in Southern Agau (Ethiopian
Cushitic - Hetzron 1969: 25), there are specific conditional paradigms
(marking person, number and gender) for both the protasis and the apodosis,
but tense (imperfect and perfect) is marked in the protasis and reality/
unreality in the apodosis. There is, however, a further distinction between
'indefinite' and 'definite' that is also marked in the form used in the apodosis,
and which indicates further degrees of speaker confidence.
Hetzron illustrates six 'main combinations'. For real conditions, the
conditional imperfect is used in the protasis and the regular imperfect, either
definite or indefinite, in the apodosis:
'an desni ncye ygc
I study+lsG+cond.iMPF work find+lsG+iMPF.lNDEF
'If I study I shall find work'
'an desni ncye ygcyc '" '
I study+lSG+cond.iMPF work find+lsc+iMPF.DEF
'If I study I shall find work'
For unreal conditionals either the conditional imperfect or the conditional
perfect is used in the protasis and the conditional indefinite or the conditional
definite in the apodosis, giving four possibilities:
'an desni yizg desyw
I study+lSG+coND.iMPF my.relative study+3sG+C0ND.iNDEF
'If I study my relative would (probably) study'
'an desni yizg deswyw '
|, I study+lsG+C0ND.iMPF my.relative study+3sc+coND.DEF
'If I study my relative would (certainly) study'
'an desni yizg desyw
I study+lsG+coND.PERF my.relative study+3sG+coND.iNDEF >-.,:
'If I had studied, my relative would (probably) have studied'
'an desni yizg deswyw ^
I study+lsG+C0ND.PERF my.relative study+3sG+C0ND.DEF
'If I had studied, my relative would (certainly) have studied'
Less striking, there are two further possibilities with conditional sentences
in English, the use of should (see 7.6) and were to. The following two
sentences may be compared with the first two given in 8.3.1 (in brackets
here):
If John should come, Bill will leave
If John were to come, Bill would leave
(If John comes, Bill will leave)
(If John came, Bill would leave) .'

215
Past tense as modal

The choice of will in the first and of would in the second clearly shows that
these are real and unreal conditionals respectively. They appear to differ
from the bracketed sentences only in suggesting slightly greater likelihood of
the event referred to in the apodosis (John's coming). This, in combination
with the real/unreal contrast, gives four degrees of the expression of
likelihood.
There is a similar situation in Swahili (Bantu - Salone 1983), where what
Salone calls the 'imperfective' is the usual marker of a real conditional. This
is an infix that occurs in the place of tense-aspect markers, as in:
Mtumwa a-ki-taka ku-ondoka-na minyaro . . .
slave suB-iMPFV-want iNFiN-leave-RECip shackles
i-na-m-lazimu a-fany-e mapambano
suB-PRES-oBJ-necessary suB-do-suBj struggle
'If a slave wants to rid himself of his shackles, he must struggle'
However, the form ikiwa or iwapo followed by a verb with full tense-aspect
marking may be used to suggest that the condition is unlikely to be fulfilled:
ikiwa walawezi wote wa-na-fikiria hivyo ni-ta-jiuzili la
if settlers all suB-PRES-think this.way lsG-FUT-resign but
yuko kalika kundi la-wachache
he in group of.few
'If all the settlers think that way, I shall resign, but he is in the minority'
However, it must be noted that ikiwa contains the infix -ki- and is literally 'if
it is' (and iwapo is a similar formation with a locative marker). It may not be
surprising that an interpretation in terms of'If it is the case t h a t . . .' suggests
some degree of doubt, usually towards a statement made by someone else.

8.4 Wishes
Wishes also use past tense forms, often along exactly the same lines as the
protases of unreal conditions. Thus Classical Greek uses the optative, the
imperfect and the aorist for future, present and past wishes respectively
(compare the conditionals in 8.3.1). They are introduced by ei gr, and it is
no coincidence that ei is the conjunction ('if') of conditionals:
ei gr genomein tknon, ant so
oh that become+lsG+AOR+opT son, instead.of you
nekrs (Eur. Hipp. 1410)
corpse
'O that I might be a corpse, my child, instead of you!'
ei gr tosatein dnamin ekhon (Eur. Ale. 1072)
oh that such strength have+lsG+iMPF+iNDic
'Had 1 such strength'

216

J
 8.5 Russian 'subjunctive'

ei gr m' hup g:n . . . hiken (Aesch. P. V. 152)

oh that me below earth send+3sG+AOR+iNDic
'Would he had sent me under the earth'
Similarly, Latin uses present, imperfect and pluperfect subjunctive, often
with utinam:
Atque utinam ipse Varro incumbat in
but that self Varro apply.self+3so+PRES+suBJ in
causm (Cic. An. 3. 15)
cause
'But if only Varro would apply himself to the cause'
Modo valeres! (Cic . An. 11.23)
only be.well+2sc+iMPF+suBj
'If only you were well'
Utinam ne . . . tetigissent litora puppes (Catull. 64. 171)
that not touch+3PL+PLUP+suBj shores ships
'Would that the ships had not touched the shore'
English, like Classical Greek, sometimes uses an expression that contains
the conditional conjunction - if only. More commonly it uses the lexical verb
WISH. As with conditionals, past tense forms are used, but for wishes for the
future, would or could are normally used:
If only/I wish John would/could speak tomorrow
If only/I wish John had spoken yesterday
The use of would or could in future conditionals suggests that there is some
willing (or able) agent, for wishes usually imply that there is some such agency.
For wishes about the present, a past tense form is used, as for conditionals
/ wish John liked curry. This also explains the less likely wish for the future
If only II wish John spoke tomorrow, for this is related not to John will speak
tomorrow, but to John speaks tomorrow, with the meaning that it is arranged
or planned for John to speak.
There are various ways in which other languages express wishes. Thus, for
Fula, Arnott (1970: 299) offers:
njuutaa balde
be.long+2so+suBj in.days
'May you live long!'
Particles are used for the same purpose in Serrano (Hill 1967: 88), and in
Huichol together with a 'desiderative' suffix (Grimes 1964: 60-1).

8.5 Russian 'subjunctive'

There are other ways in which the past tense is associated with the
subjunctive. In Russian, for instance, what is often called 'subjunctive'

217
Past tense as modal

consists in fact of the particle -by plus the past tense. This is used in several
constructions where the subjunctive would be used in other languages. It i s
used, for instance, after the verbs 'desire', 'demand', 'insist' (whereas the
accusative/dative and infinitive construction is used with 'order', and
'suggest' takes either construction):
Ja zhelaju/trebuju/nastaivaju chto-by ona ushla
I desire/demand/insist thaX-by she go+F.sG+pAST
'I desire/demand/insist that she should go'
On poprosil/prikazal emu/emu uiti
he asked/ordered him/to.him go+iNFiN
'He asked/ordered him to go'
On predlozhil chto-by my ushli
he suggested that-by we go+PL+pAST
'He suggested we should go'
On predlozhil nam uiti
he suggested to.us go+iNF
'He suggested we should go'
It is also used after the negative verbs of thinking, etc., where a Romance
language would use the subjunctive (see 5.2.2):
Ja ne dumaju chto-by on byl glup
I not think that-y he be+M.SG+PAST stupid
T don't think he's stupid'
Compare the Italian:
Non credo che sia Corelli
no I.think that be+3sG+PRES+suBJ Corelli
'I don't think that it's Corelli'
Russian also uses this 'subjunctive' in purpose clauses, another type often
associated with the subjunctive proper (see 5.3.2):
cht-by nikt ne znal ob tom, nado
that+by no.one not know+3M.sc+pAST about this necessary
molch
be.silent
'So that no one may know, we must keep silent'
This Russian 'subjunctive' seems to show, once again, the connection
between past tense and unreality.

8.6 Explanation
In 8.3 there were examples from a number of languages of past tense forms
being used to indicate unreal conditions. It is easy enough to find further
examples. Steele (1975: 200) mentions Garo (Tibeto-Burman), Chipewyan
 8.6 Explanation
(N. America, Athabaskan) and old Marathi (India, Indo-Iranian). James
(1982: 376) adds Old Irish to the Western European languages exemplified,
and also mentions the Bantu languages Tonga and Haya and further North
American languages, Cree (Algonquian) and Nitinaht (Wakashan). Steele
also argues, perhaps a little speculatively, that the protolanguage of Uto-
Aztecan (N. America) had a morpheme reconstructed as ta, which combined
the notions irrealis and past; the evidence for this is found, in languages
within the family, in morphemes which can be regarded as reflexes of ta.
Naturally, this modal function of past tense can occur only in languages
that have a tense system that differentiates past from present or non-past. It
does not occur in languages such as those of Papua New Guinea (e.g., Amele
- see 6.4) that deal with time relations within the category of Realis/Irrealis
mood. Bhat (1999: 144) sees the use of past tense for unreal conditionals as a
characteristic of 'tense prominent' languages, as distinguished from those
that are 'aspect prominent' or 'mood prominent'. However, although it is
obvious that it does not occur in 'mood prominent' languages, it must be said
that with 'aspect prominent' languages the categories used in conditionals
appear to be restricted to perfective and imperfective, the categories that
have a close association with past and present time. Bhat gives an example
i from Gujerati:
| tuN wakhstsar awe to apNe bahar jaiej
1 you on.time come+PERFV then we out might.go
I 'If you come on time, we might go out'
w e saurN kam karlo hOt to praphesar
H he good work do+iMPFV then professor become
'If he were doing good work, he would (have) become a professor'
K It would seem that, strictly, the relationship is with markers associated with
past time rather than with grammatical past tense.
An obvious and important question is 'Why is past tense (or, more strictly,
forms associated with past time) so widely used to express this type of
I modality?' Why is the future not used instead, since the future often seems to
I be a kind of modality (Lyons 1977: 816) and is often treated as irrealis?
i The relation between past and unreality has often been noted, but the
I explanations seem to be largely circular. Joos (1964: 121-2), for instance,
i suggests that the essential common feature is remoteness, in time or reality.
I Similarly, James (1982: 396) speaks of 'remoteness from reality' and
I Langacker (1978: 855) uses the label 'distal'. But this may be no more than
giving a single label to two quite different meanings, and so may be
uninformative.
Steele (1975: 217), argues for the 'semantic primitive of disassociative',
219
Past tense as modal

past time being disassociated from present time and unreality from reality.
Of particular relevance is her interpretation of the use of past tense for polite
requests, as in:
Would you pass the salt?
cf. Will you pass the salt?
She suggests that the past tense abstracts the speaker from the request. But
why? A possible answer is that the use of the past leaves open the possibility
that the willingness does not extend into the present, though it may imply
that it does. Thus / want to speak to you demands attention, whereas / wanted
to speak to you indicates only past desire and leaves it open for the hearer to
make an excuse such as T haven't the time right now . . .' She also notes
Hale's (1969: 22) suggestions that intentions in the past are usually un-
achieved and that unachieved intentions are 'a specific sort of irrealis'.
An argument along these lines is presented in much more detail and more
convincingly by Bybee (1995: 506-8). She suggests that, with a modal verb
(i.e. a notionally modal verb) such as WANT, there is a rather tenuous relation
between the agent and the main predicate. / wanted to can be used 'both in
contexts in which the desire was carried out and in contexts in which the
desire was not carried out'. Moreover, as stative verbs the past forms of such
'modals' 'assert that a state existed before the moment of speech, but they do
not say whether that state exists in the present or not'. There are two areas of
vagueness: '(i) whether or not the predicate event was completed; and (ii)
whether or not the modality remains in effect'. With a modal in past time,
there is the possibility that some conditions on the completion of the main
event were not met, as with:

I wanted to help you (but didn't)

She then considers the 'polite' use of past tense in:
I wanted to ask you a question
This, she suggests implies that there might be conditions that are unmet,
including the question of whether the addressee wants to be asked a question.
She could have added that it also leaves open the implication that the
modality may no longer remain in effect. This seems even more important -
for by suggesting that it is not necessarily the case that I still want to ask the
question, I greatly weaken my insistence on the question being answered.
This explanation works for the dynamic modals CAN and WILL in sentences
such as:
I could do that for you
I would do that for you ;>

220
 8.6 Explanation
Here the use of present tense forms can and will might be taken as an
instruction rather than an offer. For can may be used in (a mother to a
child):
You can stay in your room until you behave
Similarly will may be used to instruct in:
You will do as I say
By contrast the past tense forms, for the reasons given above, can be
interpreted as mere offers or suggestions.
It is worth recalling, at this point (see 3.3.2), that we cannot normally say:
*I ran fast and could catch the bus
*I asked him and he would come
For a past tense modal may, as argued above, imply that the event was not
completed. It is, therefore, inappropriate when there is a clear implication
that it was completed.
A somewhat similar explanation can be given for the use of might and
could for suggestions, and of epistemic might. Yet it will not account in full
for all the past tense modal forms. It does not, for instance, explain the use of
should as the 'unreal' form of MUST rather than SHALL.
Moreover, this explanation does not seem to be applicable to the use of
past tense verbs in unreal conditionals, especially where there is no modal
verb, as in Classical Greek and Tigre (8.2.1). Bybee attempts to extend her
analysis to If she saw Judy, she would tell her the news. She says 'The past in
both clauses . . . signals that the conditional relation held in past time. The
modality in the f/jew-clause signals that the relation is still in effect. (Cf. If
Mike saw her he told her the news.)'. This is not entirely convincing, for
although it explains why a past tense form might be interpreted as unreal, it
does not explain why a past tense form should be used in the first place, since
it is not actually true that the conditional relation held in past time.
Finally, and, unfortunately, negatively, it must be said that nothing in this
section seems to account for the use of past tense in the Russian 'subjunctive'.
Not everything in language is capable of rational explanation!

221
 References

Akatsuka, Noriko. 1985. Conditionals and the epistemic scale. Language 61: 625-39.
Aksu-Koc, Ayhan A. and Slobin, Dan I. 1986. A psychological account of the
development and use of evidentials in Turkish. In Chafe and Nichols 1986:
159-67.
Anderson, J. 1971. The grammar of case: towards a Vocalist theory. Cambridge:
Cambridge University Press.
Andrews, J. Richard. 1975. Introduction to Classical Nahuatl. Austin: University of
Texas Press.
Arnott, D. W. 1970. The nominal and verbal system ofFula. Oxford: Clarendon Press.
Asher, R. E. 1982. Tamil. (Lingua Descriptive Series 7.) Amsterdam: North Holland.
Austin, J. L. 1956. Ifs and cans. Proceedings of the British Academy 42: 109-32.
Reprinted in Austin 1961: 153-80.
1961. Philosophical papers. Oxford: Clarendon Press.
Austin, Peter. 1981. A grammar of Diyari, South Australia. Cambridge: Cambridge
University Press.
Bar-Hillel, Y. 1970. Aspects of language. Jerusalem: Magnus.
Barnes, J. 1984. Evidentials in the Tuyuca verb. International Journal of American
Linguistics 50: 255-71.
Bavin, Edith. 1995. The obligation modality in Western Nilotic languages. In Bybee
and Fleischman 1995a: 107-33.
Bhat, D. N. S. 1999. The prominence of tense, aspect and mood. (Studies in language
companion series 49.) Amsterdam and Philadelphia: Benjamins.
Bliese, L. F. 1981. A generative grammar of Afar. (Publications in linguistics 65.)
Arlington: Summer Institute of Linguistics and University of Texas at Arlington.
Bloomfield, Leonard. 1933. Language. New York: Holt; (1935) London: Allen and
Unwin.
Boas, Franz (ed.). 1922. Handbook of American Indian languages (Vols, i and n).
Washington: Government Printing Office.
1933-8. Handbook of American Indian languages (Vol. m). Columbia: Columbia
University Press.
Bolinger, Dwight L. 1968. Postposed main phrases: an English rule for the Romance
subjunctive. Canadian Journal of Linguistics 14: 3-33.

222
 References

1978. Y e s - N o questions are not alternative questions. In Hiz 1978: 87-105.

Botne, Robert. 1997. Evidentiality and epistemic modality in Lega. Studies in language
21: 509-32.
Bromley, H. M. 1981. A grammar of Lower Grand Valley Dani (Pacific Linguistics
C 63.) Canberra: Department of Linguistics, Australian National University.
Brown, Keith. 1991. Double modals in Hawick Scots. In Trudgill and Chambers 1991:
74-103.
Bruce, L. 1984. The Alamblak language of Papua New Guinea {Pacific Linguistics
C 81.) Canberra: Department of Linguistics, Australian National University.
Buckley, Eugene. 1988. Temporal boundaries in Alsea. Papers from the sixteenth
meeting of the Berkeley Linguistic Society 16: 10-22.
Bugenhagen, Robert D. 1994. The semantics of irrealis in the Austronesian languages
of Papua New Guinea. In Reesink 1994: 1-39.
Butt, J. and Benjamin, C. 1988. A new reference grammar of modern Spanish. London:
Edward Arnold.
Bybee, Joan. 1985. Morphology: a study of the relation between meaning and form
(Typological studies in language 9.) Amsterdam & Philadephia: John Benja
mins.
1995. The semantic development of past tense modals in English. In Bybee and
Fleischman 1995a: 503-17.
1998. Trrealis as a grammatical category'. Anthropological Linguistics 40: 257-71.
Bybee, Joan and Fleischman, Suzanne (eds.). 1995a. Modality and grammar in
discourse (Typological studies in language 32.) Amsterdam and Philadelphia: John
Benjamins.
1995b. Modality in grammar and discourse: an introductory essay. In Bybee and
Fleischman 1995a: 1-14.
Bybee, Joan L., Pagliuca, William and Perkins, Revere D. 1991. Back to the future. In
Traugott and Heine 1991 Vol. n: 17-58.
Bybee, Joan L., Perkins, Revere D. and Pagliuca, William. 1994. The evolution of
grammar: tense, aspect and modality in the languages of the world. Chicago:
University of Chicago Press.
Capell, A. and Hinch, H. E. 1970. Maung grammar: texts and vocabulary. The Hague
and Paris: Mouton.
Chafe, Wallace. 1986. Evidentiality in English conversation and academic writing. In
Chafe and Nichols 1986: 261-72.
1995. The realis-irrealis distinction in Caddo, the Northern Iroquoian languages,
and English. In Bybee and Fleischman 1995a: 349-65.
Chafe, Wallace L. and Nichols, Joanna. 1986. Evidentiality: the linguistic coding of
epistemology. Norwood, N. J.: Ablex.
Chung, Sandra and Timberlake, Alan. 1985. Tense, aspect and mood. In Shopen 1985
Vol in: 202-58.
Coates, Jennifer. 1983. The semantics of the modal auxiliaries. London: Croom Helm.
Cohen, M. 1936. Trade de langue amharique. Paris: Institut ethnologie.

223
References

Cole, P. 1982. Imbabura Quechua. (Lingua Descriptive Series 5.) Amsterdam: North
Holland.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
1985. Tense. Cambridge: Cambridge University Press.
Cormack, Annabel and Smith, Neil. Forthcoming. Modals and negation in English.
Cowell, M. W. 1964. A reference grammar of Syrian Arabic. Washington, D.C.:
Georgetown University Press.
Craig, C. G. 1977. The structure of Jacaltec. Austin and London: University of Texas
Press.
Croft, William. 1995. Modern syntactic typology. In Shibatani and Bynon 1995:
85-144.
Curme, G. O. 1905. A grammar of the German language. London: Macmillan; (1960
rev. edn) New York: Frederick Ungar.
Davidsen-Nielsen, Niels. 1990. Tense and mood in English: a comparison with Danish.
Berlin and New York: Mouton de Gruyter.
Davies, J. 1981. Kobon. {Lingua Descriptive Series 3.) Amsterdam: North Holland.
De Haan, Ferdinand. 1997. The interaction of modality and negation: a typological
study (Outstanding disertations in linguistics.) New York: Garland Publishing.
Deibler, E. W. 1976. Semantic relationships of Gahuhu verbs. Studies in Linguistics
(Norman, Oklahoma, Summer Institute of Linguistics) 48.
Delancey, Scott. 1986. Evidentiality and volitionality in Tibetan. In Chafe and
Nichols 1986: 203-14.
Derbyshire, D. C. 1979. Hixkaryana. (Lingua Descriptive Series 1.) Amsterdam: North
Holland.
Dixon, R. M. W. 1972. The Dyirbal language of North Queensland. Cambridge:
Cambridge University Press.
1977. A grammar of Yidi/i. Cambridge: Cambridge University Press.
Donaldson, T. 1980. Ngiyambaa: the language of the Wangaaybuwan. Cambridge:
Cambridge University Press.
Ehrman, M. 1966. The meaning of the modals in present-day American English. The
Hague: Mouton.
Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge
University Press.
Frachtenberg, Leo. 1920. Yakonan (Alsea) grammar. Unpublished Ms. in the Franz
Boas collection of the American Philosophical Society Library, Philadelphia.
1922. Coos. In Boas 1922 Vol. n: 297-429.
Frajzyngier, Zigmunt. 1985. Truth and the indicative sentence. Studies in Language 9:
243-54.
Frawley, William. 1992. Linguistic semantics. Hillsdale, N.J.: Lawrence Erlbaum.
Friedman, Victor A. 1981. Admirativity and confirmativity. Zeitschrift fur Balkano-
logie 17: 12-28.
1986. Evidentiality in the Balkans: Bulgarian, Macedonian, and Albanian. In Chafe
and Nichols 1986: 168-87.

224
 References

Fries, C. C. 1927. The expression of the future. Language 3: 87-95.

Gary, J. O. and Gamel-Eldin, S. 1982. Egyptian colloquial Arabic. (Lingua Descriptive
Series 6.) Amsterdam: North Holland.
Giridhar, P. P. 1994. Mao Naga grammar. Mysore: Central Institute of Indian
languages.
Givn, Talmy. 1971. Dependent modals, performatives, factivity, Bantu subjunctives
and what-not. Studies in African Linguistics 2: 61-81.
1982. Evidentiality and epistemic space. Studies in Language 6: 2349.
1994. Irrealis and the subjunctive. Studies in Language 18: 265-337.
Givn, T. and Kimenyi, A. 1974. Truth, belief and doubt in Kinyarwanda. Papers
from the fifth annual conference on African Linguistics (Studies in African
Linguistics, Supplement 5): 95-113.
Gordon, Lynn. 1986a. Maricopa morphology and syntax. (University of California
Publications in linguistics 108.) Berkeley: University of California Press.
1986b. The development of evidentials in Maricoba. In Chafe and Nichols 1986:
75-88.
Grimes, J. 1964. Huichol syntax. (Janua Linguarium, Series Practica 11.) The Hague:
Mouton.
Hale, K. 1969. Papagoli-m. Cambridge, Mass.: Massachusetts Institute of Tech
nology (mimeo).
Hall, R. A. 1964. Introductory linguistics. Philadelphia and New York: Chilton
Books.
Halliday, M. A. K. 1970. Functional diversity in language as seen from a considera
tion of mood and modality in English. Foundations of Language 4: 225-42.
Hammer, A. E. 1983. German grammar and usage. (Reprinted edition with corrections
and Supplement.) London: Edward Arnold.
Hamp, E. P. 1982. Latin ut/ne and ut (. . . non). Glotta 60: 115-20.
Handford, S. A. 1947. The Latin subjunctive: its usage and development from Plautus to
Tacitus. London: Methuen.
Hardy, H. K. and Gordon, L. 1980. Types of adverbial and modal constructions in
Tolkapaya. International Journal of American Linguistics 46: 183-96.
Haugen, E. 1976. The Scandinavian languages: an introduction to their history.
London: Faber and Faber.
Hensarling, G. 1982. Evidentials in Kogi. Ms.
Hetzron, Robert. 1969. The verbal system of Southern Agaw. (University of California
Publications, Near Eastern Studies 12.) Berkeley and Los Angeles: University of
California Press.
Hewitt, B. G. 1979. Abkhaz. (Lingua Descriptive Series 2.) Amsterdam: North
Holland.
Hill, K. C. 1967. A grammar of the Serrano language. Unpublished dissertation,
University of California, Los Angeles.
Hiz, H. (ed.) 1978. Questions. Dordrecht: Reidel.
Hockett, C. F. 1968. A course in modern linguistics. New York: Macmillan.

225
References

Hoijer, Harry. 1931/83. Tonkawa: an Indian language of Texas. In Boas 1938-8:

1-148.
Hooper, J. B. 1975. On assertive predicates. In Kimball 1975: 91-124.
Hope, E. R. 1974. The deep syntax of Lisu sentences: transformational case grammar
(Pacific Linguistics, B 34.) Canberra: Department of Linguistics, Australian
National University.
Horton, A. E. 1949. A grammar of Luvale. Johannesburg: Witwatersrand University
Press.
Huddleston, Rodney. 1976. Some theoretical issues in the description of the English
verb. Lingua 40: 331-83.
1995. The case against a future tense in English. Studies in Language 19: 399-446.
Huisman, R. D. 1973. Angaataha verb morphology. Linguistics 110: 43-54.
Jacobsen., William H. 1986. The heterogeneity of evidentials in Makah. In Chafe and
Nichols 1986: 3-28.
James, D. 1982. Past tense and the hypothetical: a cross-linguistic study. Studies in
Language 6: 375-403.
Jarvella, R. J. and Klein, W. (eds.). 1982. Speech, place and action: studies in deixis and
related topics. New York: Wiley.
Jespersen, O. 1909-49. A modern English grammar. 1 vols. Heidelberg: Karl Winter;
Copenhagen: Einar Munksgaard.
1924. The philosophy of grammar. London: Allen and Unwin.
Johnston, Raymond Leslie. 1980. Nakanai of New Britain. The grammar of an Oceanic
language. (Pacific Linguistics, B 70.) Canberra: Department of Linguistics,
Australian National University.
Joos. M. 1964. The English verb: form and meanings. Madison and Milwaukee, Wise:
The University of Wisconsin Press.
Kastovsky, D. and Szwedek, A. (eds.). 1986. Linguistics across historical and geo
graphical boundaries: in honour of Jacek Fisiak. The Hague: Mouton de Gruyter.
Kayne, R. S. 1975. French syntax: the transformational cycle. Cambridge, Mass., and
London: MIT Press.
Kimball, P. (ed.). 1975. Syntax and semantics 4. New York: Academic Press.
Kiparsky, P. and Kiparsky, C. 1971. Fact. In Steinberg and Jakobovits 1971: 345-69.
Klein, F. 1975. Pragmatic constraints in distribution: the Spanish subjunctive. Papers
from the 11th Regional Meeting of the Chicago Linguistic Society: 353-65.
Koshal, S. 1979. Ladakhi grammar. Delhi: Motilat Banarsidass.
Kuno, S. 1973. The structure of the Japanese language. Cambridge, Mass., and
London: MIT Press.
Lakoff, R. T. 1968. Abstract syntax and Latin complementation. Cambridge, Mass.:
MIT Press.
Langacker, R. W. 1978. The form and meaning of the English auxiliary. Language 54:
853-82.
Langdon, Margaret. 1970. A grammar of Diegueo: the Mesa Grande dialect. Berkeley:
University of California Press.

226
 References

Lavandera, B. R. 1978. Analysis of semantic variation: the Spanish moods. (Mimeo.)

Leman, Wayne. 1980. A reference grammar of the Cheyenne. (Occasional Publications
in Anthropology, Linguistics Series 5.) Museum of Anthropology, University of
Northern Colorado, Greeley, Colorado.
Lepschy, A. L. and Lepschy, G. 1977. The Italian language today. London: Hutchison.
Leslau, W. 1941. Documents Tigrinya. Paris: Klincksieck.
1945. A short grammar of Tigre. (Publications of the American Oriental Society
Offprint Series 18.) New Haven, Conn.: American Oriental Society.
Levinsohn, S. H. 1975. Functional perspective in Inga. Journal of Linguistics 11: 1-37.
Lewis, G. L. 1967. Turkish grammar. Oxford: Clarendon Press.
Li, C. N. and Thompson, S. A. 1981. Mandarin Chinese: a functional reference
grammar. Berkeley, Los Angeles and London: University of California Press.
Lichtenberk, Frantiek. 1983. A grammar of Manam. (Oceanic Linguistics Special
Publication 18.) Hawaii: University of Hawaii Press.
1995. Apprehensional epistemics. In Bybee and Fleischman 1995a: 293-327.
Lightfoot, D. 1979. Principles of diachronic syntax. Cambridge: Cambridge University
Press.
Loeweke, E. and May, J. 1980. General grammar of Fasu (Namo Me). Work Papers
in Papua New Guinea Linguistics 27: 5-106.
Lowe, I. 1972. On the relation of the formal and sememic matrices with illustrations
from Nambiquara. Foundations of Language 8: 360-90.
Lunn, Patricia V. 1995. The evaluative function of the Spanish subjunctive. In Bybee
and Fleischmann 1995: 419-49.
Lyons, J. 1968. Introduction of theoretical linguistics. Cambridge: Cambridge Univer
sity Press.
1977. Semantics. 2 vols. Cambridge: Cambridge University Press.
1982. Deixis and subjectivity: loquor ergo sum! In Jarvella and Klein 1982: 101-24.
Malone, Terrell. 1988. The origin and development of Tuyuca evidentials. Inter
national Journal of American Linguistics 54: 119-40.
Matthews, G. H. 1965. Hidatsa syntax. The Hague: Mouton.
Matthews, P. H. 1991. Morphology (2nd edition). Cambridge: Cambridge University
Press.
Merlan, F. 1982. Mangarayi. (Lingua Descriptive Series 4.) Amsterdam: North
Holland.
Miller, Amy. 1990. A grammar of Jamul Digueo. Ph.D dissertation, University of
California, San Diego.
Mithun, Marianne. 1986. Evidential diachrony in Northern Iroquoian. In Chafe and
Nichols 1986: 89-112.
1995. On the relativity of irreality. In Bybee and Fleischman 1995a: 367-88.
1999. The languages of Native North America. Cambridge: Cambridge University
Press.
Moore, R. W. 1934. Comparative Greek and Latin syntax. London: Bell.
Munro, Pamela. 1976. Topics in Mojave syntax. New York: Garland Press.

227
References

Muslin, liana. 1997. Direct speech and evidentiality in Macedonian. Papers from the
thirty-third regional meeting of the Chicago Linguistic Society (CLS 33), Papers
from the main session: 287-330.
Newmark, Leonard, Hubbard, Philip and Prifti, Peter. 1982. Standard Albanian: a
reference grammar for students. Stanford: Stanford University Press.
Okell, John. 1969. A reference grammar of colloquial Burmese. 2 vols. London: Oxford
University Press.
Osborne, C. R. 1974. The Tiwi language. (Australian aboriginal studies 55, Linguistic
series 21.) Canberra: Australian Institute of Aboriginal Studies.
Oswalt, Robert L. 1986. The evidential system of Kashaya. In Chafe and Nichols
1986: 29-45.
Palmer, F. R. 1957. The verb in Bilin. Bulletin of the School of Oriental and African
Studies 19: 131-59.
1962. Relative clauses in Tigrinya. Journal of Semantic Studies 7: 36-43.
1974. The English verb. London and New York: Longman.
1979. Modality and the English modals. London and New York: Longman.
1983. Review of Coates 1983. Australian Journal of Linguistics 3: 287-93.
1984. Grammar (2nd edition). Harmondsworth: Penguin.
1986. Mood and modality. Cambridge: Cambridge University Press.
1987. The English verb (2nd edition). London and New York: Longman.
1990. Modality and the English modals (2nd edition). London and New York:
Longman.
1994. Grammatical roles and relations. Cambridge: Cambridge University Press.
1995. Negation and the modals of possibility and necessity. In Bybee and
Fleischman 1995a: 453-71.
1997. Negation and modality in the Germanic languages. In Swan and Westvik
1997: 133-49.
Parks, Douglas. R. 1976. A grammar of Pawnee. (Garland studies in American Indian
linguistics.) New York: Garland Publishing.
Perkins, Michael R. 1982. The core meaning of the English modals. Journal of
Linguistics 18: 245-73.
Picallo, M. Carme. 1990. Modal verbs in Catalan. Natural language and linguistic
theory 8: 285-312.
Plank, F. 1984. The modals story retold. Studies in Language 8: 305-64.
Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey and Svartvik, Jan. 1985. A
comprehensive grammar of the English language. London and New York:
Longman.
Rastorgueva, V. S. 1963. A short sketch of Tajik. (Indiana Research Center in
Anthropology, Folklore and Linguistics Publication 28. International Journal of
American Linguistics 29.)
Reesink, Ger. P. (ed.) 1994. Topics in descriptive Austronesian linguistics. (Samaan 11.)
Leiden: Vakgroep Talen en Culturen van Zuidoost-Azi en Ocenie, Rijksuniver-
siteit Leiden.

228
 References

Rivero, M.-L. 1975. Referential properties of Spanish noun phrases. Language 51:
32-48.
1977. Specificity and existence: a reply. Language 53: 70-85.
Roberts, John R. 1990. Modality in Amele and other Papuan languages. Journal of
linguistics 26: 363-401.
1994. The category 'irrealis' in Papuan medial verbs. Notes on Linguistics 67: 5 - 4 1 .
Robins, R. H. 1952. Noun and verb in universal grammar. Language 28: 289-98.
Rojas, N. 1977. Referentiality in Spanish noun phrases. Language 53: 61-9.
Ruwet, N. 1967. Introduction a la grammaire generative. Paris: Plon.
Salone, Sukari. 1983. The pragmatics of reality and unreality in conditional sentences
in Swahili. Journal of pragmatics 7:311 - 2 4 .
Sapir, Edward. 1922. The Takelma language of Southwestern Oregon. In Boas 1922
Vol. n: 7-296.
Schlichter, Alice. 1986. The origins and deictic nature of Wintu evidentials. In Chafe
and Nichols 1986:46-59.
Schubiger, M. 1965. English intonation and German modal particles: a comparative
study. Phonetica 12: 65-84.
Searle, J. R. 1983. Intentionality. Cambridge: Cambridge University Press.
Semeonoff, A. H. 1958. A new Russian grammar (12th revised edition). London: Dent.
Shell, O. A. 1975. Cashibo modals and the performative analysis. Foundations of
Language 13: 177-99.
Shibatani, Masoyoshi and Bynon, Theodora (eds.). 1995. Approaches to language
typology. Oxford: Clarendon Press.
Shipley, William F. 1964. Maidu grammar. (University of California Publications in
Linguistics 41.) Berkeley and Los Angeles: University of California Press.
Shopen, Timothy (ed.). 1985. Language typology and syntactic description. 3 vols.
Cambridge: Cambridge University Press.
Silva-Corvaln, Carmen. 1995. Contextual conditions for the interpretation of'poder'
and 'deber' in Spanish. In Bybee and Fleischman 1995a: 67-105.
Steele, S. 1975. Past and irrealis: just what does it all mean? International Journal of
American Linguistics 41: 200-17.
1997. The future of typology. Papers from the 32nd Regional meeting. The Chicago
Linguistic Society (CLS), Papers from the panels 33: 287-330.
Steele, Susan, with Alemajian, Adrian and Wadow, Thomas. 1981. An encyclopedia of
AUX; a study in cross-linguistic equivalence. (Linguistic Inquiry Monograph 5.)
Cambridge, Mass.: MIT Press.
Steere, Edward. 1943. A handbook of Swahili language as spoken at Zanzibar (4th
edition revised by A. B. Hellier) London: The Sheldon Press.
Steinberg, D. D. and Jakobovits, L. A. (eds.). 1971. Semantics. Cambridge: Cam-
bridge University Press.
Suzuki, Satoko. 1994. Is that a fact? Reevaluation of the relationships between
factivity and complementizer choice in Japanese. Proceedings of the twentieth
annual meeting of the Berkeley Linguistic Society (BLS 20): 521-31.

229
References

Swan, Toril and Westvik, Olaf Jansen. 1997. Modality in Germanic languages;
historical and comparative perspectives. (Trends in linguistics: studies and mono-
graphs 99.) Berlin and New York: Mouton de Gruyter.
Sweetser, Eve. E. 1982. Root and epistemic modals: causality in two worlds. Proceed-
ings of the eighth annual meeting of the Berkeley Linguistic Society (BLS 8):
484-507.
1990. From etymology to pragmatics; metaphorical and cultural aspects of semantic
structure. Cambridge: Cambridge University Press.
Terrell, Tracy and Hooper, Joan B. 1974. A semantically based analysis of mood in
Spanish. Hispania 57: 484-94.
Thiagarayan, K. 1981. Modal systems of English and Tamil. Unpublished disserta-
tion, University of Madras.
Traugott, Elizabeth Closs. 1989. On the rise of epistemic meanings in English: an
example of subjectification in semantic change. International Journal of American
Linguistics 55: 31-55.
Traugott, E. C. and Heine, B. (eds.) 1991. Approaches to grammaticalization. 3 vols.
Amsterdam and Philadelphia: John Benjamins.
Trudgill, Peter and Chambers, J. K. (eds.) 1991. Dialects of English; Studies in
grammatical variations. London: Longman.
Tucker, A. N. and Mpaayei, J. T. O. 1955. A Maasai grammar. London: Longman.
Watkins, Laurel. 1998. A grammar of Kiowa. Lincoln: University of Nebraska Press.
West, D. 1983. Wojokeso sentence, paragraph and discourse analysis. (Pacific linguistics
B28.) Canberra: Department of Linguistics: Australian National University.
Willett, Thomas. 1988. A cross-linguistic survey of the grammaticalization of evidenti-
ality. Studies in Language 12: 51-97.
von Wright, E. H. 1951. An essay in modal logic. Amsterdam: North Holland.

230
 Language index

Abkhaz,48-9, 88 Diyari, 134

Acholi, 128 i ; . . Dyirbal, 83, 84, 170, 179, 191
Afar, 81
Alamblak, 161,163, 174, 181 English, 2, 4, 6-12, 14-15, 18, 25-6, 28,
Albanian, 74 30-4, 47, 49, 51, 53-6, 60, 64-8,
Alsea, 151, 155, 186 70-6, 78-80, 86, 88-95, 97-9, 101,
Amele, 2, 4, 18, 145, 153, 154, 156, 103-5, 126-7, 133,136,142,144,
158-9, 161, 168, 174, 183, 188,219 160, 170, 173, 175, 179-80, 187, 191,
Amharic, 137-8, 141, 197 195-9, 201-2, 205, 207-8, 211-12,
Angaataha, 158 214-15,217
Anglo-Saxon, 104
Anjam, 158-9 Fasu, 38,40-1,45
Arabic, 88, 96, 138, 141, 197 French, 10, 71, 78, 96, 101-2, 104-5,
Assamese, 94 117, 170,190 - '
Fula, 13, 108, 110, 132, 143,217
Baluchi, 88, 112,127
Bargam, 55,158-9,168,179,183,190,191 Gahuka, 158
Bilin, 209 Garo, 218
Burmese, 170, 171 German, 4, 9, 15, 18, 24, 31-3, 40, 42, 52,
60,71,76,78,93,97, 101, 107,
Caddo, 11-12, 18, 146-7, 150, 152, 157, 113-16, 119-20,190, 198,210-11
169, 170, 172-3, 175, 177, 180-1, Greek, see Classical Greek, Homeric
183, 186-8 Greek, Modern Greek
Cashibo, 58, 59, 103 Gujerati, 219
Catalan, 95
Cayuga, 19, 31,41 Hanis Coos, 52
Central Pomo, 4-9, 15, 36, 42, 44, 50, 66, Haya, 219
104, 148, 152, 157, 169, 173-5, 178, Hidatsa, 37,41,66, 69, 239
180-1, 184, 187 Hixkaryana, 40, 41, 54, 67, 104, 165, 168,
Cheyenne, 82 172, 177
Chinese, 60, 78, 102, 103, 126, 214 Homeric Greek, 124, 129, 205
Chipewyan, 218 Huichol, 54, 68, 217
Classical Greek, 13, 81, 108, 116, 124-5,
128-9, 133, 137-8, 140, 187, 192, Imbabura, 12, 15, 19, 27, 53, 68
199, 204-6, 208-9, 216-17, 221 Irish, 219
Cree, 219 Italian, 11, 13,26,31,71,89,96-7,
Czech, 200 101-2, 104-5, 109, 112-14, 117-18,
120-2, 124, 126-7, 129, 132-3,
Dani, 162, 171 138-40, 170, 177, 182-3, 186, 189,
Danish, 26, 40, 71,94-5, 107 192-3, 196,202,213,218

231
Language index

Jacaltec, 200 North Frisian, 71

Jamul Diego, 182 Norwegian, 97
Japanese, 199, 214
Oneida, 41
Kashaya, 30-1, 44-5, 50-1, 55-6, 67-8,
179, 191 Polish, 200-1
Khezha, 55 Portuguese, 132
Kinyarwanda, 96, 200
Kiowa, 150, 163, 164, 168,171,175,181, Quechua, 12, 18,27
184,241
Kobon, 197 Romance, 18, 26, 76, 101-2, 104-5, 107,
Kogi, 54, 62 116,118-19,218
Korean, 214 Russian, iii, 131, 137, 217-18, 221

Ladakhi, 26, 38, 46 Semai, 214

Lao, 88 Seneca, 31,41, 162, 171
Latin, 13, 20, 31, 83, 104-8, 110, 112, Serrano, 54, 67, 104, 167, 168, 169, 172,
115, 119,121, 123,125-33,135-9, 177,217
142, 151, 186-7, 192-3,207, Sherpa, 41,48, 67
209-14,217 Southern Agau, 215
Lega, 15,42,49,201 Spanish, 2-4, 13, 18, 31, 99, 101-2, 105,
Lisu, 72, 77-8, 89 107, 112-13, 116-22, 124-5, 127,
Luiseflo, 57 129, 132-3, 135, 138-40, 149, 170,
Luvale, 108, 111, 116,140 177, 182-3, 187, 189, 192-6,
210-11
Maasai, 144 Sursurunga, 155, 178, 181
Maidu, 78, 81-2 Swahili, 78, 105, 108, 111, 140, 144, 216
Makah, 31,45, 50, 67
Manam, 18, 147, 149, 155, 160, 178, 179 Tajik, 48
Mangarayi, 143 Takelma, 19, 82, 151, 157, 168, 171, 180,
MaoNaga, 165, 176 184, 186
Mapun, 200-1 Tamil, 18, 27, 28, 72, 87
Marathi, 219 Thai, 214
Maricopa, 45-6, 150, 164, 165, 180 Tigre, 141,209,221
Maung, 148, 185, 190 Tigrinya, 130, 141,209
Menomini, 52, 56 Tiwi, 53, 75
Mesa Grande Diegueo, 173 To'aba'ita, 134
Modern Greek, 26, 71, 86, 92-3, 95-6, Tolkapaya Yavapai, 152, 175, 179
103, 105, 128 Tonga, 219
Mohawk, 31,41 Tonkawa, 78, 81,82
Mongolian, 214 Turkish, 47
Muyuw, 145, 156, 161, 168, 174, 184 Tutatulabal, 87
Tuyuca, 6, 24, 29, 36, 40, 41, 43, 51, 55,
Nahuatl, 142 56,57,65,66, 187
Nakanai, 176, 182
Nambiquara, 63 Welsh, 95, 106
Ngiyambaa, 16, 27, 35-6, 40, 43, 47, 53, Wintu, 29, 67
59,65,68,70,73,81,83,87, 131, Wojokeso, 158-9
149, 197 Wyandot, 31
Nitinaht, 219
Nobonob, 158-9 Yidiny, 83, 84, 136, 197

232
 General index

ability, Abilitive, 10, 15, 47, 70, 76-80, Counter-assertion, 17, 59

84-7,155-6,160,171,184,211, counterfactual, 2, 53, 148, 149, 155,
213-14 157-60, 163,178,232
accusative and infinitive, 133, 139, 192 cumulative exponents, 150-2, 181, 185-6
admirative, 11, 146, 157, 187
affirmative, 27-8 debitive, 27-8, 87
agent-oriented, 84-5 declarative, Declarative, 7, 20, 44, 45, 47,
aspect, 27, 64, 102, 137, 144, 190, 205-6, 51,58-9,64-8,81-2,87
208,216,219 deductive, Deductive, deduced,
assertion, 3, 17, 48, 54, 56, 57-8, 59, 64, deduction, 6, 8-9, 15, 20, 22, 24-31,
66,68-9,109-11, 122, 173, 190 33-5, 38, 44, 47, 49, 51-2, 57, 63-5,
assumptive, Assumptive, 6, 8, 20, 22, 67-8,71-2,89, 149, 166, 167,
24-5,28-30,51-2,57,67,78, 105, 170-2, 204
170, 203 deontic, Deontic, 7-10, 13-15, 20, 22-3,
auditory, Auditory, 6, 8-9, 15, 22, 29, 28, 38, 64, 70-81, 84-99, 101-3,
43-6,50-1,56 109-11, 127-8, 134-5, 138, 162,
193-4, 204
belief clitics, 16-17,59 desiderative, Desiderative, 22, 131, 156,
binary, 1, 4-5, 9, 137, 146-7, 160-1, 159, 165,217,232
163-4, 169, 171, 187-8, 189-90, 205 diachronic, 169, 170
direct and indirect evidence, 47
categorical, 2 direct speech, 142, 196-9
Categorical assertion, 17, 59, 68 directives, 10, 70, 75, 80, 84, 110, 128,
certainty, 4, 31, 52, 67, 98, 167, 179, 191 130,192,203
clitics, 6, 17-19, 36, 53, 59, 68, 150, 185 discourse, 58
commands, 2, 80, 82-3, 106, 138, 141-2, doubt, 2-3, 5, 12, 18, 25, 27, 31-2, 49,
148, 160, 169, 179-83, 192, 196-7 52,54,60,62-3,99, 110,117-19,
Commissive, 10, 22, 70, 72-3, 81, 105, 167, 174,176,200-1,207,216
204 dubitative, Dubitative, 25, 149, 163, 167,
commitment, 31, 34-5, 64, 98, 198-200 169, 172, 232
complement clause, 126, 131, 154-7 dynamic, 8-10, 14-15, 20, 22-3, 70, 72,
concessive, Concessive, 11, 27, 31, 76-9, 85-6, 89, 91, 96, 99, 103, 109,
109-10 111, 184,220
conditional, Conditional, 14, 21-2, 32,
74, 78, 104, 125, 132, 144, 152, 154, epistemic, 7-9, 12-15, 20, 22, 24-5,
157, 163, 174, 177-8, 184, 191, 201, 27-9, 32-5, 38, 52, 56, 64-5, 68,
207-9,221,232 70-3, 76-7, 79, 81, 83, 86-94, 96,
confidence, 9, 15, 34, 35, 52, 198, 203, 98-9, 101-5, 109-12, 134-5, 143,
215 149, 162, 170,203-4,220
conjecture, 12, 27 epistemological, 8

233
General index
event modality, 8, 22, 70, 81, 84, 86-7, interrogative, Interrogative,
89, 111, 116, 126 interrogation, 11-13, 20-2, 52-5,
evidence clitics, 16-17 57-60, 82, 114, 120-1, 147, 157,
evidential, evidential modality, 6, 8-9, 168, 170, 172-5, 190
IS, 18-19, 22, 24-6, 29-30, 35-9, irrealis, Irrealis, 1, 2-7, 11-13, 16,
44-7, 49, 50-2, 56-7, 62-3. 65-8, 18-23,40,55,62,64-5,69,73,
86, 99, 104, 144, 162, 165-8, 176, 79-80,104,106-8,110,116,119,
179, 187, 191 124, 126, 129, 133, 136, 142-92, 196,
Exclamatory, 82 203, 205-6, 214, 219, 220
experiential, 38, 47
joint, 40, 145-7, 148-9, 152, 156-61,
factive, 11,105, 121,200 168, 188-9
fears, 13,22, 111, 129, 131, 134-5, 192 judgment, 2, 6-9, 12, 15, 24-7, 29, 31-3,
first-hand, 6, 8, 12, 55, 57, 67-8 36-7, 40,49, 65, 67-8, 73, 76, 88,
folklore, 40-1 99, 113, 170,203
future, Future, 2-3, 5, 8, 14, 18, 20-2, jussive, Jussive, 5, 21-2, 80-2, 108-11,
26,28,62,73,78,81, 103-6,111, 127, 129, 136-8, 140-3, 161,
124-5, 132, 135, 144-5, 147-9, 151, 179-82, 187
154-65, 167-74, 176, 178-81, 184,
189-91, 202, 206, 209-10, 216-17, Knowledge clitics, 16-17
219
linked clauses, 152-4, 157, 188
grammatical relations, 150, 186 logical, 3, 34,48, 89, 90, 91,92,93, 213,
214
habitual, 2, 22, 39, 51, 55, 79, 80, 81, 147,
148, 156, 158-9, 160, 168, 170, 176, manipulation, 135, 193-6
179, 190, 191 modal system, i, 4, 6, 7, 8, 10, 12-15, 19,
habitual past, 2, 22, 55, 80-1, 158-9, 179 22, 52, 54, 60, 64-5, 67, 69, 80-1,
harmonic combinations, 35 86, 88, 90, 98-9, 103-4, 106, 136,
hearsay, Hearsay, 6, 8, 21, 38, 40-1, 42, 144, 146, 149-52, 162-3, 165,
47-8, 52, 67, 114, 164, 166, 177,200 167-8, 170, 172,177, 187-8, 191
hedges, 35 modal verb, 4, 6-7, 9, 11, 14-15, 18-19-,.
hierarchy, 51 25, 34-5, 42, 47, 60, 64-5, 71-2, 74,
historic, 205-7 78, 80, 86-7, 90, 95-6, 99, 100-4,
hortative, 2, 81, 158-9, 161, 180-1 106,126,179, 187, 196,202,204,
hypotheticals, 178 208, 210, 214, 220-1
modality, 2, 4, 6-11, 13, 15, 19-20,
imminent, 176, 182 22-4, 27-8, 33, 38, 47, 52, 56, 58,
imperative, Imperative, 2, 5, 16, 20-1, 64, 70, 72-4, 76, 77-8, 84, 86-7,
53, 64-5, 80-2, 84, 100, 108, 89-92, 95-9, 102-3, 105-6, 109,
110-11, 116-17, 127-9, 136-8, 112, 126, 144, 168, 179, 187. 191,
141-2, 147, 151-2, 154, 156-9, 161, 203,208.211,219-21
164-5, 169, 179-83, 187. 190. 201 modal-past, 14, 31-2, 127, 203-7
indicative, 2-5, 19-21, 23, 53, 68, 81, 99, modal-tense, 203
101,104,107-8,111, 113-14,115, mood, 2. 4-5, 7, 11-13, 19, 22-3, 37, 40,
117-19, 121-6, 129, 130, 135-7, 55, 64-6, 67, 81-2, 88, 104, 106-8,
142, 144-5, 148-9, 151, 154, 164, 113,119,124,134,137,142-3,145-7,
185, 187, 189, 192, 198-9,205-8, 149,150,154-7,159-63,165,168,
211-14 172-4,176-8,181-2,185,187-91,
inference, inferential, 6, 8-9, 24-6, 198-9,203-4,207,209,213,219
29-35, 38-9, 46-52, 105, 165, 167 morphology, 137, 150
inflectional, 19, 103, 105, 185,202
intentive, Intentive, intention, 78, 82, 105, necessity, 8, 28, 72-3, 83, 86, 89-98, 102,
131, 141, 147, 156-7, 159, 169, 198 106, 128, 204

234
 General index

negative, Negative, negation, 2, 11-13, 144-5, 147, 151, 153, 157-61, 164,
2 0 - 2 , 2 7 - 8 , 34-5, 51-4, 6 0 - 1 , 73, 168, 170, 172-4, 180, 189-90,
7 9 - 8 0 , 8 4 , 9 0 - 1 0 3 , 106, 116-21, 202-10,212,216-17,219-20
129-30, 133-6, 138-9, 141, 146, presupposed, Presupposed,
152, 157, 163-4, 168-9, 173-5, presupposition, 4, 11, 18, 21-2, 31,
180-2, 190,207,218 56,59, 109-12, 121-4, 143, 177,
negative commands, 106, 138 187,202
negative imperative, 20, 116, 117, 129, primary, 15, 205-6
164, 180-2, 190 probability, 2 7 - 8 , 3 2 - 3 , 113, 170
NICE properties, 60, 99 prohibitive, prohibition, 20, 98, 146, 156,
non-assertive, 11-12, 53, 78, 120, 200 157-9, 180-1
non-finite, 192 propositional modality, 7-9, 22, 2 4 - 5 ,
non-joint, 145-50, 159-61, 168, 189 51,56,86,89, 111-12, 168
non-visual, Non-Visual, 9, 36, 43, 44, 51 purposive, Purposive, purpose, 9, 16, 22,
notional, 1-2, 4 - 7 , 15, 18-21, 25-6, 58, 6 5 , 7 0 , 7 3 , 8 3 - 4 , 126,128-31,
65, 92, 95-6, 98, 140, 145, 148-9, 133-6, 141-2, 149-50, 154-6, 159,
159, 163, 170,186-8, 190, 192-3, 206,217-18
196
question, 2, 13, 20, 49, 53-5, 58-61,
Obligative, obligation, 7, 9, 10, 22, 71-6, 6 7 - 8 , 7 4 , 9 1 , 105,111, 118, 121, 128,
8 3 - 5 , 8 8 - 9 , 109-10, 127-8,131, 135, 137-8, 159, 165, 167, 169,
138, 1 4 6 , 1 5 5 - 7 , 1 6 2 - 3 , 165, 178, 172-3, 186, 188, 190, 197, 202, 204,
183-4, 186, 196 219-20
optative, 81-2, 116, 141-2, 162, 187, quotation marks, 20-1
199,204-7,209,216 quotative, Quotative, 21, 37, 4 0 - 1 , 5 0 - 1 ,
58, 167, 177
participant, 62
particles, 15, 19, 41, 49, 52, 60, 62, real, 1, 3, 5, 14, 125, 147-9, 151-3, 155,
159-61, 166-7 158-9, 161-2, 164-5, 169, 171,
past tense, 13-16, 23, 31, 32-4, 67, 73, 173-4, 177-8, 180, 188-9,203,
76,79-80, 100, 113-14, 131,138, 207-8,214-16
142, 159, 168, 179, 191, 198-9, realis, Realis, 1-6, 20-2, 40, 6 4 - 5 , 69,
201-11,216-19,220-1 107, 119, 145, 147, 149, 151-2,
perfective, 4 4 - 5 , 50, 67, 157, 165, 173, 157-8, 161, 186-90,219
219 redundant, redundancy, 5, 148, 151, 157,
performative, 33, 50, 51, 80, 127, 137 160, 186, 189,214
permissive, Permissive, permission, 7, reliability, 200
9-10, 1 5 , 2 2 , 2 7 - 8 , 7 1 - 2 , 7 4 - 7 , remote past, 2 - 3 , 55, 158-9, 168, 174,
7 9 - 8 1 , 84-9, 98, 102-3, 110, 128, 179,210
160,204 report, Report, reported, Reported, 1, 2,
personal experience, 6, 8, 48, 50 8-9, 13, 15, 2 1 - 2 , 24, 2 7 - 9 , 33, 35,
polite, politeness, 32, 71, 74, 82, 110, 127, 37-44, 47-52, 5 6 - 8 , 63-4, 67, 76,
138, 181-3,204,220 100, 110, 112-16,120-1,126, 128,
possibility, Possibility, 8, 10, 20, 28, 32, 146, 148, 159, 164, 170, 171-2, 174,
34, 49, 53, 7 2 - 3 , 77, 79, 84, 86, 176-7, 192, 196, 198-201,207
89-96, 98, 102-3, 106, 110, 112-13, result, 48, 79, 83, 84, 129, 135-6, 149,
133, 143, 151, 162, 176, 184, 190, 185
202, 207, 220
potential, 8, 54, 70, 75, 86, 98, 101, 151, sensory, Sensory, 9, 15, 17, 22, 24, 30,
158-60, 162, 167,171-3, 184 35-6,43,47,49-51,56,67
preference, 13 sequential, 152-3
present tense, 2, 5, 14, 26, 3 2 - 3 , 47, 52, simulative, 146, 157, 177
5 5 - 6 , 6 5 , 7 5 , 7 6 , 8 1 , 101, 103,114, simultaneous, 19, 152-3, 174
119, 120, 125, 127, 132, 135, 142, speaker-oriented, 8 4 - 5

235
General index
speculative. Speculative, speculated, 6, Timitive, 22, 131
8-9, 15, 18, 20, 22, 24-8, 31, 38, 49, typology, typological, 1-2, 5-6, 8, 10,
52,65,71-2,89, 109-10, 112, 168, 14, 18-22, 29, 102, 107, 145, 149,
170-2,203 188, 191
subjective, subjectivity, 33-4, 75-6, 80
subjunctive, 2-5, 11, 13, 18-21, 23, unmarked, 5, 7, 45, 50, 53-4, 64-8, 146,
31-2,42,53,81-3,88,99, 104, 152, 157, 161-2, 164, 174. 182,
106-7, 108-45, 148-9, 151, 154, 187-8
156,177, 182, 185-98,201-2, unreal, unreality, 1, 3, 13-14, 136, 149,
204-13,217-18,221 155, 157, 175, 177-8, 188. 190,
subordinate, subordination, 5, 11, 13, 80, 202-4,207-16,218-220
83,95,99,101,108,111-17, 121-3, unrealized, 3, 13, 109, 111, 134, 159
128-44, 154-7, 159-60, 186-7, 189,
191,195-6,206-7 visual, Visual, 6, 8-9, 15, 21-2, 29-30,
suffixes, 19, 38, 39, 46, 55, 67, 72, 81, 87 36, 38, 43-6, 50-1, 56-7, 63, 65-7,
165
tense, 1, 5, 13-16, 19-20, 23. 31-4, 48, Visual-Auditory, 46
64,67,70,73,76,79-81, 100-1, Volitive, 10, 22, 76, 78, 109-10, 131
103-5,108, 110,113-14, 119-20,
125,131, 137-8, 142, 144,151,155, wanting, 13, 78, 126, 131-3, 135, 144,
158-9, 185, 187, 190-1, 198-9, 155, 157, 191,211
201-12,215-21 warning, 61, 148, 160, 167
tentative, tentativeness, 13-14, 31-2, 54, willingness, 10, 76, 78-80, 220
73-4, 203-4 wishes, 13,22, 111, 116, 131-4, 142, 175,
terminology, 6, 19, 20, 89, 108, 149, 153, 192,204,216-17
163,185,205 witnessed, 27-8, 67

236