ArticlePDF Available

The dual nature of irrealis in complementation

Authors:

Abstract

This article explores the connection between irrealis mood and type of complement clause. It is argued that irrealis performs different functions in propositional and state-of-affairs complements, reflecting either irreality (non-actualisation in a situation that often has a concrete location in time) or lack of temporal and situational anchoring. While these distinct functions could be viewed as different realisations of a general irrealis meaning, one must keep in mind that general irrealis meanings are generalisations of more specific meanings arising in particular contexts of grammaticalisation. It is argued that the functions defined above are useful intermediate generalisations enabling a coherent account of irrealis use in complementation. However, the distinction between propositional and state-of-affairs complements is not neat: functions characteristic of the propositional domain also occasionally spill over into the state-of-affairs domain (in the form of counterfactuality or non-factuality marking) and vice versa (as the unanchoring function of subjunctives with evaluative predicates).
John Benjamins Publishing Company
This is a contribution from Studies in Language 44:1
© 2020. John Benjamins Publishing Company
This electronic file may not be altered in any way.
The author(s) of this article is/are permitted to use this PDF file to generate printed
copies to be used by way of offprints, for their personal use only.
Permission is granted by the publishers to post this file on a closed server which is
accessible only to members (students and faculty) of the author's/s' institute. It is not
permitted to post this PDF on the internet, or to share it on sites such as Mendeley,
ResearchGate, Academia.edu.
Please see our rights policy on https://benjamins.com/content/customers/rights
For any other use of this material prior written permission should be obtained from the
publishers or through the Copyright Clearance Center (for USA: www.copyright.com).
Please contact rights@benjamins.nl or consult our website: www.benjamins.com
The dual nature of irrealis
in complementation
Axel Holvoet
Vilnius University
This article explores the connection between irrealis mood and type of
complement clause. It is argued that irrealis performs dierent functions in
propositional and state-of-aairs complements, reecting either irreality
(non-actualisation in a situation that oen has a concrete location in time)
or lack of temporal and situational anchoring. While these distinct func-
tions could be viewed as dierent realisations of a general irrealis meaning,
one must keep in mind that general irrealis meanings are generalisations of
more specic meanings arising in particular contexts of grammaticalisation.
It is argued that the functions dened above are useful intermediate gener-
alisations enabling a coherent account of irrealis use in complementation.
However, the distinction between propositional and state-of-aairs comple-
ments is not neat: functions characteristic of the propositional domain also
occasionally spill over into the state-of-aairs domain (in the form of coun-
terfactuality or non-factuality marking) and vice versa (as the unanchoring
function of subjunctives with evaluative predicates).
Keywords: irrealis, mood, complementation, propositional complements,
state-of-aairs complements, factive predicates, factivity
1. Introduction
This article deals with the functions of irrealis moods in complementation. Irre-
alis has been used as a cover term for marked moods like ‘subjunctives’ and ‘opta-
tives’, but also to refer to a notional domain assumed to be relevant to various
aspects of language structure. It has, for instance, been invoked in discussing dif-
ferent types of clausal complements, a distinction I will be focusing on in this arti-
cle. My main emphasis will, however, be on irrealis mood rather than on irrealis
as a notional domain that may be relevant also beyond the sphere of mood. More
specically, I will be concerned with the functions of irrealis mood in comple-
mentation.
https://doi.org/10.1075/sl.18057.hol
Studies in Language 44:1 (2020), pp. 165–190. issn 0378-4177 |eissn 1569-9978
© John Benjamins Publishing Company
The structure of the article is as follows. I will briey comment, rst, on the
notions of realis and irrealis in the study of mood, and on the use of the notions
of realis and irrealis with reference to complementation types. Then I will discuss,
rst, the domain of propositional (truth-valued) complementation, in which irre-
alis may be relatively uncontroversially said to be associated with degrees of real-
ity (in dierent but related senses); and secondly, the domain of state-of-aairs
(non-truth-valued) complementation, where irreality distinctions seem to have a
lesser explanatory value and the notion of ‘unanchoring’ (suspension of temporal
and situational reference) is introduced. I will also show that though mood basi-
cally functions dierently in the two domains of complementation, there is a cer-
tain overlap between them as far as mood functions are concerned.
2. Irrealis mood: Notional and denitional problems
The usefulness of the notion of irrealis in explaining the use of mood forms is noto-
riously controversial; for an overview of the discussion cf. Nikolaeva (: –).
The literature provides us with a list of contexts and marking types where irrealis
may, though need not, appear, cf. Elliott (). The notion of irrealis has, in this
sense, a certain explanatory value, as there is probably some very abstract general
feature shared by all the instances of irrealis. But it clearly has no predictive value:
there is no recurrent set of uses in which we may expect irrealis forms to appear in
any given language, which is only natural because an irrealis form, like any other
grammatical form, is always the result of a grammaticalisation history in a partic-
ular language. As Bybee (:) points out, “there is not one case in which a
grammatical distinction corresponds directly to the notional distinction between
real and unreal situations. The argument is taken up by Cristofaro (), who
points out that a lot of diculties in operating with the notion of irrealis arise from
the mixing-up of dierent levels of linguistic analysis – descriptive categories of
individual languages, notions relevant to diachronic processes, and notions used
in explanatory hypotheses. Irrealis might be a kind of comparative concept of the
kind characterised by Haspelmath (), invoked by linguists but not belonging to
the linguistic knowledge of the speakers of any particular language. Finally, irrealis
forms always function as part of a series of constructions (counterfactive, purpo-
sive, desiderative, etc.) where they may have constructional meanings; the general
meanings may be epiphenomenal if they are meant to account for the facts of indi-
vidual languages. A somewhat more hopeful note is struck by De Haan (), who
pleads for a bottom-up approach: instead of squeezing mood forms into the mould
of aprioristic realis vs. irrealis distinctions, one may nd more useful lower-level
generalisations by looking at particular languages. The present article is also about
lower-level generalisations, more particularly in complementation.
166 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
It is clear that irrealis moods do not originate as such, as a means of encoding
abstract general meanings like ‘irreality’. They are always, as shown by Bybee,
Perkins & Pagliuca (), the outcome of a series of grammaticalisation
processes in particular contexts, where they initially encode much more concrete
meanings such as ‘purpose’, ‘counterfactual condition, etc. But the rise of an irre-
alis category also involves extension; for instance, an irrealis form grammati-
calised in one type of subordinate clause (for instance, purpose clauses) may
come to encode other types of subordinate clauses as well (for instance, comple-
ment clauses with volitional verbs). It is these extensions that may give us a clue
as to certain intermediate-level generalisations operating in processes of exten-
sion of irrealis forms beyond the original context of grammaticalisation, because
these extensions involve, by denition, some degree of generalisation. It is on such
extensions that I will, to a large extent, concentrate in this article.
On a purely terminological level, irrealis is, as pointed out by Van der Auwera
& Schalley (), a convenient cover term in view of the lack of a principled way
of dierentiating between the notions of ‘subjunctive’, ‘optative’ and ‘conditional’
used in the grammatical traditions of individual languages.1I will here follow Van
der Auwera & Schalleys proposal. In glossing, however, I will retain the tradi-
tional terms ‘subjunctive’ and ‘conditional’ for Romance as well as ‘subjunctive
and ‘optative’ for Classical Greek, while using ‘irrealis’ for languages like Slavic
and Baltic, which have only one instantiation of the category of irrealis.
3. Irrealis in complementation
In discussions on realis vs irrealis distinctions, irrealis in complementation and,
more generally, in subordination naturally looms large. Many uses of irrealis
forms are restricted to subordinate clauses, and the Latin terms subiunctivus and
coniunctivus highlight this domain of use to the exclusion of all others. However,
the specic import of irrealis in complementation remains, in my view, insu-
ciently understood.
1. In the literature on many languages (e.g., Baltic), the same category may be referred to by
different authors as ‘subjunctive’, ‘optative’ or ‘conditional’. To be noted is iero’s proposal
(:–) of reserving the term ‘conditional’ (or ‘western conditional’) for what is basically
a kind of future-in-the-past with modal overtones (French je viendrais ‘I would come’, etc.); it
is not quite clear whether such forms belong to mood or to tense. ese western-style condi-
tionals seem to play no role specifically connected to complementation in the sense that they
are not used as what will here be characterised as mood as a complementation strategy (French
uses its conditional as a kind of evidential strategy, but does this also in main clauses), so that
the problem will not be considered here.
The dual nature of irrealis in complementation 167
© 2020. John Benjamins Publishing Company
All rights reserved
What lends a particular interest to irrealis functions in complementation is that
thenotions ofrealis andirrealis havealso, somewhat confusingly,been used to char-
acterise complement-clause types. The broader use of the notions of realis and irre-
alis, referring to whole clauses, is already seen in Cro (), who invokes ‘modus
irrealis contexts’, a usage that seems to have inspired Haspelmath (: , passim)
to introduce the notion of ‘irrealis-nonspecic’ as a subtype of indeniteness. Apart
from that, the notion of irrealis is on record also in the literature on clausal comple-
mentation. Mithun (:) mentions from Alsea (Penutian, Oregon) dierent
complementisers for ‘realis clauses’ and ‘irrealis clauses. The term ‘irrealis comple-
mentisers’ is used by Ammann & Van der Auwera (:) to account for
Romanian (< Lat. si), the complementiser of verbs like ‘want, as opposed to (<
Lat. quia), the complementiser of verbs like ‘say’. That is to say, we have a notion of
irrealis clauses, possibly introduced by irrealis complementisers. We might perhaps
expect such clauses to contain, at least optionally, irrealis verb forms.
Where two dierent strains of research meet, the question arises whether the
notions used in them converge. Here problems present themselves because the
notion of irrealis commonly used in the literature is rather broad, as it is meant
to capture a large array of construction types. Palmer (:) uses the following
formulation proposed by Mithun: “The realis portrays situations as actualised, 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.” In order to examine the interaction between mood and a
particular domain of usage, in this case complementation, it is useful to have more
precise notions. In particular, while the meaning of ‘actualised’ is more or less
clear, ‘non-actualised’ may mean several things. It may refer to a state-of-aairs
contrary to what actually occurred – so-called counterfactive use, as in if you had
listened to me; or it may refer to the object of an act of volition that may yet be
actualised, as in if you’ll only listen to me. It is obvious that these two instances of
‘non-actualisation’ (which a language could mark with the same irrealis form) will
fall into dierent classes if we oppose predicates like ‘say’ to predicates like ‘want’.
() French
Je
I
ne

dis
say..1
pas

que
that
tu
you
aies
have..2
commis
commit.
une

erreur.
mistake
‘I’m not saying you committed an error.
() Je
I
ne

veux
want..1
pas

que
that
tu
you
commettes
commit..2
une

erreur.
mistake
‘I don’t want you to commit an error.’
168 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
The complement of ‘say’ in () refers or may refer to an actual situation, precisely
located in time, where a certain type of event is said not to have taken place,
whereas the potential error in (), of which it is not known whether it will take
place, is not anchored in an actual situation. Whereas the ‘non-actualisation’ as
illustrated in () shows an inherent connection with the semantics of ‘want’ as a
complement-taking verb, there is no logical connection between the irrealis in ()
and what the semantics of ‘say’ as a complement-taking verb is assumed to imply.
We could even speak of a mismatch between irrealis marking and the semantics of
the complement-taking verb in the case of evaluative or factive higher predicates.
The complements of such predicates are marked with irrealis in certain languages,
though these predicates (unlike the factuality-neutral ‘say’) presuppose the reality
of the situation described in the complement:
() Italian
È
be..3
peccato
pity
che
that
io
1
di
of
cantonese
Cantonese
e
and
mandarino
Mandarin
non

ne
thereof
sappia
know...1
una

parola.
word
‘It’s a pity I don’t speak a word of Cantonese or Mandarin.2
Such sentences are so conspicuously at odds with the notion of ‘irreality’ that they
have inspired reformulations of the function of irrealis (subjunctive) to the eect
that what is involved is non-assertability rather than non-reality (cf. Lunn ,
endorsed by Palmer :–, –).
Intheir broadersense,not necessarily pertainingto mood, the terms‘realis’and
‘irrealis’ are suciently widespread, as shown by the above examples. In a discus-
sion of irrealis mood, however, there is an obvious advantage in avoiding notional
confusion by using dierent terms for types of clausal complements. There is no
shortage of such terms: ‘truth modality’ and ‘action modality ’ (Ransom :–),
‘fact type’ and ‘potential type’ (Dixon : –), etc.; here I will adopt the terms
‘propositional’ and ‘state-of-aairs’ (Kehayov & Boye b: –, invoking Dik
& Hengeveld ). The dierence consists in that “propositions evoke concepts
construed as having a (situational) referent, whereas S[tates]o[f ]A[airs] evoke
concepts not construed as having a referent” (Kehayov & Boye b:). Other
useful terms are ‘truth-valued’ and ‘non-truth-valued’ (Kehayov & Boye a:),
though they need to be qualied. As the content of a complement clause is not
asserted, the complement must be extracted from its embedding in order for its
truth value to be established; otherwise only the notions of factivity, non-factivity
and counterfactivity (formulated in terms of presuppositions) will apply. For
2. https://www.edizionieo.it/review/
The dual nature of irrealis in complementation 169
© 2020. John Benjamins Publishing Company
All rights reserved
instance, I’m not saying you made a mistake does not assert You made a mistake, and
does not even imply it as ‘say’ is inherently non-factive, but by itself, extracted from
its embedding, the proposition You made a mistake is truth-valued (and, as I will
argue below, languages have means of precluding a factive reading of such truth-
valued embedded propositions).
4. Formal problems with irrealis in complementation
State-of-aairs complements may be formally distinguished from propositional
complements in dierent ways. Some languages have specialised complementis-
ers, as mentioned above for Romanian; another example is Latvian, with ka for
propositional complements and lai for state-of-aairs complements. In view of
this lexical dierentiation, mood is not necessary as a dierentiating device here:
() Latvian
Zinu,
know..1
ka

skumšu
long..1
pēc
after
mājām
home..
un
and
Rīgas.
Riga.
‘I know I will long for home and Riga..
() Gribu,
want..1
lai

ziloi
elephant..
atgriežas
come.back..3
Rīgā.
Riga.
‘I want the elephants to come back to Riga.
Latvian’s close cognate Lithuanian has a dierent system, with the same comple-
mentiser, kad, for propositional and state-of-aairs complements but a dierence
in mood: state-of-aairs complements (as in ()) are marked with irrealis:
() Lithuanian
Kas
who.
sakė,
say..3
kad

gido
guide..
darbas
work..
yra
be..3
nuobodus?
boring...
‘Who said that a guide’s job is boring?
() Vyras
husband..
nori,
want..3
kad

važiuotume
go..1
gyventi
live.
į
to
Norvegiją.
Norway.
‘My husband wants us to go and live in Norway.
In this case, irrealis mood, rather than the choice of a complementiser, is used
as part of a complementation strategy, viz. as a means of encoding state-of-aairs
complements. It must be added, for the sake of completeness, that not every irre-
alis in a complement clause is part of a complementation strategy. An irrealis can
be used for reasons that are independent of complementation, e.g.,
170 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
() Latvian
Es
1.
domāju,
think..1
ka
that
vii
3...
būtu
be.
apmierināti
satisfied...
ar
with
rezultātiem.
result..
‘I think they would be satised with the results.
Here we have an irrealis form in the complement clause, but its use reects the
fact that the content of this clause is viewed as contingent on an implicit condi-
tion (e.g., if they took the test). That is, the use of the irrealis is not specically
connected with complementation here, and nothing would change if we extracted
this clause from its embedding and used it as a simple clause. This situation could
be referred to as internal mood marking in the complement clause, as opposed to
complementising mood, which is part of a complementation strategy, as in ().
There are also instances where an irrealis marker is used as part of a comple-
mentation strategy, but it does not appear on the verb. This may be seen in Slavic.
In East and West Slavic the irrealis marker, which is enclitic, has accreted to the
complementiser, giving rise to what the grammars now usually describe as a dis-
tinct complementiser:
() Russian
Ja
1.
xoču,
want..1
čto-by
-
vsë
everything.
stalo
become.-
kak
like
ran’še.
before
‘I want everything to become as it used to be.
As sequences like stalo by ‘would become’ still function as an analytic irrealis
mood form in Russian, it would seem possible to analyse čto-by as the comple-
mentiser čto ‘that’ serving as a host for the enclitic mood marker grammatically
associated with the verb. But constructions like () are clearly opposed to con-
structions in which the irrealis marker occurs in the vicinity of the verb and
marks what I have described above as ‘internal mood marking in the complement
clause’:
() Russian
Ja
1.
dumaju,
think..1
čto

vse
everything
stalo
become..
by

kak
like
ran’še.
before
‘I think everything would (then) become as it used to be.
On the other hand, čtoby can now function independently from the form in -l-
with which it forms the irrealis, and combine, e.g., with innitives:
The dual nature of irrealis in complementation 171
© 2020. John Benjamins Publishing Company
All rights reserved
() (Simanovskij, NKRJa)Russian
Nam
1.
podali
give...
signal
sign..
čto-by
that-
ne

streljat’.
shoot.
‘They gave us a sign not to shoot.
As can be seen, we have a cluster of descriptive problems here. Dobrušina
(:–) oers a detailed discussion of cases where by clusters with the verb
and acts as an irrealis verb form, and those where it clusters with a complemen-
tiser or subordinator or seems to modify the whole clause. She arrives at the con-
clusion that there is no very sharp line of division between the dierent types of
situations. This conclusion suggests that it would not be correct simply to regard
čtoby as something that is completely lexicalised as a distinct complementiser and
that is historically but not synchronically related to irrealis marking. Indeed, if this
were the case, one would probably expect this new complementiser to combine,
at least in some cases, with irrealis forms of the verb; actually constructions like
() are impossible:
() Russian
*Ja
1.
xoču,
want..1
čto-by
-
vse
everything
bylo
be..
by

kak
like
ran’še.
before
intended meaning: ‘I would have liked everything to be as it used to be.
This suggests that in čtoby, the component by is still an irrealis marker. If it cannot
be viewed as being part of the verb, it can be viewed as a clausal irrealis marker. I
will return to this further on. As to the verb form in sentences like (), it looks like
a past tense, but Russian grammars describe it as ‘the l-form, phonetically coincid-
ing with the past tense but not to be identied with it (e.g., Švedova .i: );
using Arono ’s term, we could describe it as the ‘l-morphome’. I gloss it corre-
spondingly in this article.3
The Russian construction with čtoby is used mainly in state-of-aairs comple-
ments, as in (). However, it also occurs in the propositional domain in sentences
similar to (), in which the content of the complement is rejected as unreal.
() Russian
Ja
1.
ne

dumaju,
think..1
čtoby

èto
this
imelo
have..
značenie.
significance..
‘I don’t think this has any signicance.
3. Historically, the ‘l-form’ is a participle that, in combination with enclitic present tense forms
of the auxiliary ‘be’, derived a past tense (originally a perfect) as well as an irrealis. As the auxil-
iary was dropped in Russian, the ‘l-form’ now functions as a past tense by itself, and in practical
grammars of Russian the complementiser čtoby is said to combine with the past tense.
172 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
This seems logical as in () the irrealis is undoubtedly an instance of complemen-
tising mood, that is, mood as part of a complementation strategy, as its use is con-
tingent upon negation in the matrix clause. The ‘internal-mood’ reading (‘I do not
think this would have any signicance’) would require a dierent position of the
irrealis marker, in the vicinity of the verb and not as a constituent part of the com-
plementiser. For this reason, we cannot regard čtoby as a distinct complementiser
for the state-of-aairs domain: it consists of čto plus clause-level irrealis marking.
The value of this irrealis marking depends on the domain in which it operates; it
is to this question I will now turn.
5. Irrealis for irreality
By denition, an irrealis form in complementation would qualify for being viewed
as reecting irreality. Even with reference to situations like that illustrated in ()
above, one could argue that the realisation of an event desired by a person is not
guaranteed, and that this is reected in the use of irrealis. This interpretation is
problematic, as will be argued further on. For the time being, however, I will con-
centrate on a domain where reality and irreality (truth and falsehood) would seem
most relevant, viz. the domain of truth-valued, or propositional, complementa-
tion. The following examples are from French, where propositional complements
oen contain a subjunctive. This occurs, for example, with verbs of propositional
stance like ‘think’, ‘believe’ etc. under raised negation, and with verbs inherently
expressing disbelief, like ‘doubt’:
() French
Je
I
crois
think..1
qu’
that
ils
they
sont
be..3
contents
happy.
des
with.the
résultats.
result.
‘I think they’re happy with the results.
() Je
I
ne

crois
think..1
pas

qu’
that
ils
they
soient
be...3
contents
happy.
des
with.the
résultats.
result.
‘I don’t think they’re happy with the results.
() Je
I
doute
doubt..1
qu’
that
ils
they
soient
be...3
contents
happy.
des
with.the
résultats.
result.
‘I doubt whether they’re happy with the results.
Here irrealis appears to express the rejection of a claim, or uncertainty about the
reality of a situation. In some languages this use of the irrealis in propositional
complements occurs mainly if the content of the complement clause is rejected. In
The dual nature of irrealis in complementation 173
© 2020. John Benjamins Publishing Company
All rights reserved
some cases this requires an explicitly negative epistemic evaluation (‘I don’t think’,
‘I doubt’), whereas in others it occurs also in complements of verbs expressing a
lack of certainty about the reality of a situation, even though the epistemic evalu-
ation is rather positive:
() Italian
Credo
think..1
che

sia
be...3.
un

ottima
excellent
idea.
idea
‘I think it’s an excellent idea.
One way of formulating the motivation for such uses of the irrealis is to say that a
certain state of aairs is represented as unreal, or with no guarantee of reality. Of
course, the above-mentioned feature ‘lack of assertion’ (invoked, e.g., to account
for the use of irrealis in sentences like ()) would also capture this. A type that
the notion of ‘unreal situations’ would naturally account for but for which lack of
assertion seems less convincing is the use of irrealis for negated events. In Euro-
pean languages only raised negation as in () seems to induce irrealis, but else-
where it appears under direct negation as well:
() (Caddoan; Chafe :–)Caddo
núy- t’a-yibahw
--see
‘I don’t see him.
There is an obvious dierence between ‘I assert that I do not see him’ and ‘I do
not assert that I see him’ (which would mean something like ‘I’m not sure I see
him’), so that an account in terms of lack of assertion does not seem satisfactory.
It should also be noted that irrealis marking in the propositional domain can add
an additional shade of irreality to a proposition that is already non-assertive. This
can be seen in the following examples, taken from Latvian.
() Latvian
Neviens
nobody.
ne-teica,
-say..3
ka
that
via
3...
ir
be..3
arestēta.
arrest.....
‘Nobody said she was arrested.
() Neviens
nobody.
ne-teica,
-say..3
ka
that
via
3...
būtu
be.
arestēta.
arrest.....
‘Nobody said she was arrested.
() can have two dierent readings; on one of them there is a truth-conditional
dierence between the two sentences whereas on the other there is not. Whereas
() can (but need not) be taken to mean the person in question was in fact
174 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
arrested but nobody wanted to say this, () would normally suggest that nothing
is known about the person’s being arrested as nothing was said about this. A
speaker knowing that the person in question had in fact been arrested could use
only (). The main motive for the use of the irrealis in () seems to be to pre-
clude this interpretation, that is, to exclude a factive reading of the complement in
(). As no assertion is involved in either case, I have referred to the function of
the irrealis in () as ‘emphatic non-factivity’ (Holvoet :).
There are thus, in the propositional domain, types of clausal complements
where irrealis use can be said to encode the perceived irreality of a situation. This
irreality may be realised as doubt or lack of certainty about the reality of a situa-
tion, counterfactivity or emphatic non-factivity (e.g., irrealis marking to exclude a
factive reading of a complement).
6. State-of-aairs complements
In many languages, as in Lithuanian Example (), we may assume irrealis marking
to be part of the complementising strategy, that is, it marks a state-of-aairs com-
plement as such. The question is why – does it encode irreality? The object of
a volitional verb is not regarded as unreal; in fact, the speaker may well regard
its realisation as likely. It is known that, in the domain of states-of-aairs, lan-
guages oen behave in contradictory ways where irrealis marking is concerned.
Imperatives are marked as irrealis in some languages and as realis in others, as
noted by Mithun (:–). Mithun conjectures that this might be associ-
ated with “diering expectations of actuation” (:), but adds, as another
possibility, that the dierence might simply rest on diachronic factors, reecting
dierent paths of extension, e.g., from indicative to imperative clauses. The latter
is also argued, on the basis of typological data, by Mauri & Sansò (). These
two explanations are not necessarily contradictory, as, whatever the grammatical-
isation source, subsequent generalisations can take place along the lines of “expec-
tations of actuation. We must ask, therefore, whether there is any evidence for
such generalisations in complement clauses of the state-of-aairs type with, say,
volitional verbs.
The answer seems to be that distinctions based on anticipated degree of irre-
ality are, in principle, possible in the state-of-aairs domain, but they probably
are not a very common situation. I will illustrate this with examples from Latvian.
Latvian does not automatically use irrealis forms in state-of-aairs complements
with, say, desiderative verbs. The reason is that it has a specialised state-of-aairs
complementiser that species the complement type, as shown in (). This com-
plementiser, however, was originally a hortative particle, with a function similar
The dual nature of irrealis in complementation 175
© 2020. John Benjamins Publishing Company
All rights reserved
to English let in let him come here. This hortative particle combines with a (realis)
present tense form:
() Latvian
Lai

viš
3...
atnāk
come..3
šeit.
here
‘Let him come here.
This use of the present tense is retained when lai becomes a complementiser, as
seen in (). Such a complement with a realis present tense can also combine with
main clauses in the past tense, as in ():
() Latvian
Sākot
start.
skolas
school..
gaitas,
walk..
vecāki
parent..
gribēja,
want..3
lai

rakstu
write..1
ar
with
labo
right...
roku.4
hand..
‘When I started school my parents wanted me to write with my right hand.
Though () has no irrealis marking, the present tense form rakstu has certain
features that set it apart from usual present-tense forms. It is temporally rigid, that
is, cannot be replaced with any other tense form. No past tense form could be used
in (), though the main clause is in the past tense, nor could a future be used
although the object of volition is, as mentioned, necessarily posterior to the act
of volition. This present is also frequently derived from perfective verbs, whereas
such verbs do not have a real present tense.
() Latvian
Izdevējs
publisher..
gribēja,
want..3
lai

es
1.
uzrakstu
.write..1
kaut ko
.
vēl
still
trakāku.
crazy...
‘The publisher wanted me to write something more crazy still.5
I have called this the ‘subjunctival present’ (Holvoet :–). It is known
that present tense forms ousted, in their basic function, by competing (e.g., pro-
gressive) forms are oen retained in subordination and become subjunctives; this
has been noted for the subjunctive in some branches of Indo-European, such
4. http://www.la.lv/vai-viegli-but-kreilim-labrocu-pasaule-%E2%80%A9stasta-specialisti-un-
slavenibas-kreili-3/ accessed 2017 12 02
5. https://www.diena.lv/kd/intervijas/intervija-ar-dianu-civli-un-aivu-rozenbergu-pasaule-
pagalma-?p=
176 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
as Indic and Greek (Kuryłowicz : –) and is echoed by the Armen-
ian present (Bybee, Perkins & Pagliuca :–); on this process cf. also
Haspelmath (). In Latvian this has not happened, as the present has not
been ousted in its basic function, but the subjunctival present is a distinct type
of use, characteristic of state-of-aairs complements. If the present is completely
ousted in its original function, the subjunctival present takes over the function of
a subjunctive, without irrealis functions in the sense of marking of irreality being
involved. In such cases, suspension of tense features or, to put it in a more gen-
eral way, temporal and situational unanchoring, appears to be an essential factor
in the rise of an irrealis form.
The Latvian subjunctival present will not be called an irrealis not only because
it has not yet become dissociated from the present proper, but also because it faces
competition from the inherited irrealis, the use of which in state-of-aairs comple-
ments is also possible, though not obligatory as in Lithuanian:
() Latvian
Arī
also
vecāki
parent..
gribēja,
want..3
lai

es
1.
ne-vazātos
-drag..
apkārt
around
pa
about
stūriem
corner..
un
and
ne-nodarbotos
-busy..
ar
with
visādām
various...
stulbām
silly...
lietām.
things..
Also my parents didn’t want me to roam around all corners and to busy myself
with all kinds of silly things.6
In most cases it would be hard to formulate a dierence between the constructions
with realis and irrealis forms; ‘expectations of actuation’ are dicult to measure.
But certain facts point to the possibility of associating the distinction, in principle,
with degrees of irreality. Latvian has another state-of-aairs complementiser apart
from lai, viz. kaut. Its distinguishing feature is that it conveys a nuance of irreality.
We could qualify it as an irrealis complementiser in another sense than that in
which the term is used by Ammann and Van der Auwera (), viz. as a com-
plementiser conveying irreality. True, this complementiser is only semantically
irrealis, as it contains no formal marker of irrealis. It combines only with irrealis
forms of the verb:
6. http://sportacentrs.com/hokejs/-galvins_loti_daudziem_no_dinamo_nepatik_s
The dual nature of irrealis in complementation 177
© 2020. John Benjamins Publishing Company
All rights reserved
() (Sandra Kalniete)Latvian
Tik
so
oti
much
gribētu,
want.
kaut

es
1.
prastu
know.how.
aizraujoši
stirringly
stāstīt
narrate.
par
about
šo
this..
aizgājušo
bygone...
vēlēšanu
election..
laiku, …
time.
‘I wish so much I were able to narrate in a stirring way about this bygone elec-
tion time.
In simple sentences, kaut functions as a marker of counterfactual wishes, and this
is perhaps the original function, in which case the irrealis marking would be a
trace of its grammaticalisation history. But the use of irrealis with lai cannot be
explained in this way and must be accounted for by extension. This Latvian exam-
ple shows that in state-of-aairs complements as well, some languages may use
their irrealis forms with proper irrealis meaning, to convey low expectations of
actuation. Mutatis mutandis, this echoes the dierences between () and (),
() in the propositional domain. Perhaps it simply copies these dierences – as
Mauri & Sansò () argue, distinctions along the reality dimension in directives
are basically carried over from domains where such distinctions (in the form of
counterfactivity, etc.) are naturally at home.
Not all languages with subjunctival presents make use of the possibility of
marking realis-irrealis distinctions. Among the South Slavic languages, Slovenian
has an opposition between subjunctival present, in the sense just dened for Lat-
vian, and irrealis (on the use of moods and tenses in complementation in Sloven-
ian see Browne ):
() Slovenian
Želim,
wish..1
da

vsi
all..
izvedo
learn..3
[da je bila izjemno dobra oseba].
‘I want everybody to know [that she was an exceptionally good person].’
() Slovenian
Žele-l
wish-.
bi,

da

bi

se

Turčija
Turkey.
približevala
draw.close..
Evropi,
Europe.
namesto
instead
da

se

oddaljuje.
move.away..3
‘I wish Turkey would draw closer to Europe instead of driing away from it.7
7. http://reporter.si/clanek/svet/juncker-turciji-noce-zapreti-vrat-za-vstop-v-eu-a-pravi-ce-
turcija-uvede-smrtno-kazen-?_comment_id=_
178 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
On the other hand, Bosnian-Croatian-Serbian seems to use almost exclusively the
subjunctival present:8
() Bosnian-Croatian-Serbian
Želio
wish..
bih
.1
da
that
se

dvije
two..
zemlje
country..
međusobno
mutually
dogovore.
agree..3
‘I wish the two countries would settle the matter between them.’9
For Latvian, there is thus an indication that the use of the irrealis in state-of-aairs
complements is associated with expectations of actuation: the irrealis comple-
mentiser kaut combines only with irrealis, and lai optionally combines with irre-
alis though it is not motivated by the grammaticalisation source. Perhaps a similar
situation holds in Slovenian, as the grammaticalisation source does not lead us to
expect irrealis (and we don’t nd it used at all in closely related Bosnian-Croatian-
Serbian). But there is probably no reason to assume that Lithuanian uses mood for
the purpose of marking reality distinctions: having no specialised state-of-aairs
complementiser, it uses mood to dierentiate the two types of complements. If the
marking of expectations of actuation is just a marginal factor in the use of irre-
alis forms in states-of-aairs complements, then the use of irrealis forms in such
complements could perhaps (apart from grammaticalisation history) be associ-
ated with another property of irrealis forms, viz. that of being untensed; it is what
I have called the unanchoring function.
One could object that this notion cannot be veried any more than the irreal-
ity function can be veried. If, in sentences like (), the irrealis establishes itself (as
a result of its grammaticalisation history) as the only possible form used to mark
this type of complement clause, there seems to be no way of proving either that
it is motivated by ‘irreality’ (low expectation of actuation) or ‘unanchoring’ (sus-
pension of features). Is there positive evidence of cases where the generalisation
of irrealis forms is driven by the feature of temporal and situational unanchoring?
I want to suggest that there are indeed such situations, and that the use of irrealis
for presupposed content as in () is an instance of this.
The mechanism invoked by Bybee, Perkins & Pagliuca (:–) for
irrealis in complementation is the ‘harmony principle’, illustrated with the exam-
ple of English should:
8. Szucsich (:), however, claims that the irrealis, though infrequent, is not quite mar-
ginal.
9. https://direktno.hr/eu-i-svijet/eu/juncker-zelio-bih-da-se-dvije-zemlje-medusobno-
dogovore-/
The dual nature of irrealis in complementation 179
© 2020. John Benjamins Publishing Company
All rights reserved
() He should resign.
() His opponents demanded that he should resign.
The ‘harmony’ consists in that the same deontic (volitional) meaning is both
inherent to the meaning of the higher predicate (demand) and marked in its com-
plement clause (should). At the stage where English should extends to other types
of complement clauses beyond deontic higher predicates, it may appear to be
semantically empty. But this is probably misleading: this ‘subordinating’ should
(that is, the use of should as a marker of a certain type of clausal complement
rather than in its original deontic meaning) is carried over to certain other higher
predicates, but not to all:
() It is necessary that he should resign.
() It is strange that he should resign.
() It is clear that he should resign.
While () is grammatical, it is clear that in its complement clause should has its
original deontic value and is not the ‘subordinating’ should we nd in (). We
must therefore look for some semantic rationale explaining why should is carried
over to evaluative predicates but not just any higher predicate. The same applies
to irrealis forms in Romance languages: assuming the subjunctive originates with
deontic predicates, the question is why it spreads to evaluative predicates like
French il est dommage ‘it’s a pity, whose complement, expressing a presupposed
state-of-aairs, one would expect to behave like that of il est clair ‘it is clear’:
() Il
it
est
be..3
dommage
pity
qu’

elle
she
soit
be...3
partie
depart.
‘It’s a pity that she has le.
() Il
it
est
be..3
clair
clear
qu’

elle
he
est
be..3
partie
depart.
‘It is clear that she has le.
As mentioned, several authors have attempted to explain subjunctives with fac-
tive higher predicates by invoking the ‘non-assertive’ status of the irrealis. Lunn
() argues that the use of the Spanish subjunctive for discourse-old or pre-
supposed content reects non-assertability. Assertability, she points out, is deter-
mined by relevance (in the sense of Sperber & Wilson ). Both counterfactual
content and presupposed or discourse-old content are communicatively low in
relevance, and both are therefore opposed (though each for slightly dierent rea-
sons) to the prototype of assertion, which is based on the relevance of what
is being conveyed. Lunn’s account is coherent and convincing as a synchronic
180 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
account of the use of the subjunctive in modern Spanish, but what does it imply
in terms of grammaticalisation? The feature of non-assertability could certainly
have served as a rationale for the spread of subjunctives to the domain of presup-
posed content from some other domain, and Lunn’s reasoning suggests it could
have been from the domain of irreality or counterfactivity; one does not see how
it could have originated, in terms of a grammaticalisation process, in the domain
of what is presupposed.
For English ‘emotional should’ as in (), Jespersen (:) certainly seems
to invoke the irreality dimension. He observes that “It is strange that he should
exercise so great inuence […] lays more stress on the strangeness by using the
imaginative should in the clause”. But this strictly explains only the use of should
with predicates evaluating the intrinsic likelihood of an event, not with those eval-
uating their intrinsic desirability, like it’s a pity. One would have to assume sec-
ondary extension from predicates like it is strange to predicates like it’s a pity. This,
however, cannot be correct, as the starting point must have been (as pointed out
by Bybee, Perkins & Pagliuca ) the deontic use of should, and the mechanism
involved must have been the above-mentioned principle of ‘harmony’. We do not
know exactly what was behind the spread of subjunctives to complements of eval-
uative predicates in Romance, but state-of-aairs complements with deontic pred-
icates seem to be a plausible source.
What seems to drive the carrying over of irrealis forms to factive predicates
is a shi towards the evaluation of intrinsic likelihood or desirability. The event
evaluated as strange has, in most cases, already occurred, so it cannot be qualied
as likely or unlikely, but it can still be evaluated in terms of intrinsic likelihood.
In a similar way, what has already occurred cannot be the object of an act of voli-
tion, but can be evaluated in terms of intrinsic desirability. There is thus a shi
away from a concrete event with situational anchoring towards a type of event.
This is reected in an encoding strategy stripping the complement of its temporal
anchoring, that is, its tense features. This eect can be achieved by using a sub-
junctive, but also, for example, by nominalisations. Compare the French example
with the subjunctive in () and the English construction that could be used to
render it in ():
() (Gide, Les faux-monnayeurs)French
Je
I
comprends
understand..1
maintenant
now
que
that
tu
2.
veuilles
want...2
te
2.
trouver
find.
à
at
son
3.
arrivée.
arrival
‘Now I understand [why] you want to be present at his arrival.
() I now understand your wanting to be present at his arrival.
The dual nature of irrealis in complementation 181
© 2020. John Benjamins Publishing Company
All rights reserved
Here, in opposition to the indicative, the subjunctive and the nominalisation both
suggest that not the fact of the addressee entertaining the wish but the psycholog-
ical motive behind it is the object of understanding. The reason why the subjunc-
tive in () performs a similar function as the nominalisation in () seems to be
that it is untensed. This claim may appear strange because, according to grammat-
ical terminology, subjunctives may show tense oppositions. French, for instance,
used to have the past subjunctive in cases where the main clause verb is in the
past, cf. (). In modern French, this past subjunctive has been replaced with the
present subjunctive, cf. ().
() (obsolescent)French
Il
he
voulait
want..3
que
that
je
I
lui
him
rendisse
give.back...3
l’

argent.
money
‘He wanted me to give him the money back.
() French
Il
he
voulait
want..3
que

je
I
lui
him
rende
give.back...3
l’

argent.
money
‘He wanted me to give him the money back.
It is clear that in such cases the ‘tense form’ of the subjunctive does not reect
tense dierences but just mechanically copies the tense feature of the main-clause
verb. Modern French, which usually has tense backshiing in sentences with
main-clause past-tense forms, no longer extends it to subjunctives with desider-
ative verbs because the clauses involved are untensed.10 Those past subjunctives
which are not due to backswitching mark counterfactivity, not tense, as in the fol-
lowing Italian example:
() Italian
Se
if
sapessi
know...1
cosa
what
fare,
do.
lo
it
farei.
do..1
‘If I knew what to do, I would do it.
The use of the subjunctive in state-of-aairs complements thus basically reects
suspension of tense features. This predisposes it for the expression of evaluation
in terms of intrinsic desirability and intrinsic likelihood. This claim is consistent
10. Modern German, which usually does not require subjunctive with desiderative verbs (the
German subjunctive in complementation being mainly associated with indirect speech) shows
a similar choice between past and present indicative in sentences with a past-tense main verb:
Sie wollte, dass er mitkam / mitkomme ‘She wanted him to come along. e choice is between a
past tense imposed by tense agreement and a present subjunctive reflecting ‘untensedness’ – we
basically have a kind of subjunctival present tense here.
182 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
with Lunn’s analysis of the subjunctive in Spanish to the extent that an event
evaluated in terms of intrinsic desirability or likelihood is usually presupposed.
But whereas Lunn assumes a connection between presupposed content and irre-
alis through the shared feature of non-assertability, I think it is the unanchoring
property of the irrealis that is involved. For the Romance subjunctive we have no
access to grammaticalisation history as the subjunctive is inherited from Latin.
The parallel of English should therefore provides important evidence, and it
suggests relevance of the unanchoring function of the irrealis (or equivalent
expressions like English should) spreading from state-of-aairs complements used
originally with deontic predicates (according to the harmony principle invoked
by Bybee ).
7. Two irrealis functions
The facts discussed above suggest that the function of irrealis marking in state-
of-aairs complementation might be dierent from what it is in propositional
complementation. In state-of-aairs complements, the irrealis seems to be a form
stripping the predicate of its temporal anchoring. It is therefore conceivable that
propositional and state-of-aairs complements reveal two dierent dimensions of
irrealis mood. This notion is not quite new: Cleary-Kemp () suggests that
the label ‘irrealis’ is applied to two dierent categories, viz. verbal forms encoding
non-reality and verbal forms encoding temporal non-specicity. Cleary-Kemp’s
claim is, however, not based on a detailed analysis of irrealis uses in dierent
contexts. She examines the clusters of features cited in the literature as being
served, in dierent languages, by the irrealis, nds only one cluster instantiated
in the Oceanic language Koro, and concludes that ‘irrealis’ actually refers to
two dierent categories: a cluster comprising negation, interrogative, presupposi-
tion and uncertainty as representing a ‘non-reality’ category, and one comprising
future/prospective, jussive modalities, conditional and past habitual as represent-
ing a ‘temporal non-specicity’ category.11 It is highly doubtful, however, whether
Cleary-Kemp’s notion of two distinct categories unduly swept under one rug is
correct. Rather than being disjoint, the two types of functions can perfectly well
coexist within one language. Still, it is worthwhile singling out these two aspects
11. e assignment of presupposed content to the irreality cluster is probably wrong. It is based
on the fact that in Koro, the language investigated by Cleary-Kemp, irrealis marking “is not
triggered by negation, presupposition, uncertainty, or interrogative” (:). is negative
evidence is, however, insufficient. e author adduces no positive evidence in the form of a lan-
guage with an exclusively irreality-based irrealis also extending to presupposed content.
The dual nature of irrealis in complementation 183
© 2020. John Benjamins Publishing Company
All rights reserved
of irrealis meaning. The subjunctives and similar marked moods of the Euro-
pean languages seem to straddle the two types of irrealis functions, with the irre-
ality function represented in propositional complements and the unanchoring
(feature-suspension) function in the states-of-aairs domain. We could perhaps
represent the irrealis domain as comprising two foci, which can co-occur in one
language while it is also possible for one of these foci to be more strongly repre-
sented in a language’s instantiation of the irrealis than the other.
Of course, the logical possibilities are not two but three: one can also think
of languages with two instantiations of irrealis, each centering around a dierent
semantic focus; here, situations like that of the optative and subjunctive in Classi-
cal Greek, or the conditional and subjunctive in Romance, come to mind. The t
seems accurate for Classical Greek with its subjunctive and optative, the ‘mood of
anticipation’ and the ‘mood of the fancy’, as Gildersleeve (:, ) calls them
respectively. The Greek optative covers the irreality dimension, whereas the sub-
junctive is associated with unanchoring. The subjunctive is mainly restricted to
subordinate clauses like purpose or fear, with possible actuation always posterior
with regard to the content of the main clause; in independent clauses (in Home-
ric Greek), the default realisation is future. The optative is mainly associated with
counterfactivity in wishes (óloito ‘would that he perished’) or what I have called
emphatic non-factivity, in so-called potentialis (ei génoito ‘if it should happen’). A
curious case is that of optatives replacing indicatives and subjunctives in subordi-
nation in the past-tense domain, a point I will address further on. The situation
is more complex in the Romance languages, where the so-called conditional is
partly a posterity tense (future in the past) and partly a mood with irreality-based
(counterfactive) functions; the subjunctive is split, with both kinds of functions
in complementation, as shown above (in fact counterfactivity can be expressed
both ways, cf. French j’aurais été heureux and, in a more formal register, j’eusse été
heureux ‘I would have been happy’).
8. Spillovers
Though the two irrealis functions singled out here are basically distributed along
the propositions vs states-of-aairs divide, each of them extends, to a certain
extent, beyond the spheres thus dened. Two instances have already been dis-
cussed, but more can be found. As mentioned, the irreality dimension of the irre-
alis extends into the states-of-aairs domain in the form of distinctions reecting
dierent ‘expectations of actuation’. But the Latvian and Slovenian use of irre-
alis forms to reect ‘low expectation of actuation’ is not the only incursion of the
irreality dimension into the state-of-aairs domain. The notion of ‘expectation of
184 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
(Letters of John Keats to his Family and Friends, https://ebooks.adelaide.edu.au
/k/keats/john/letters/letter.html)
actuation’ belongs to the domain of the future, but distinctions along these lines
can be transferred to past-tense situations, with hindsight being suspended and
the likelihood of realisation of an event being evaluated from the point of view of
some moment in the past. But hindsight can be applied to descriptions of events
planned or intended in the past, and it may result in explicit marking of counter-
factivity or emphatic non-factivity. This tendency is reected, for instance, in the
use of anteriority forms of the innitive in complements of desiderative verbs in
English, leading to apparently illogical constructions as in ():12
() I intended to have written immediately on my return from Scotland […], but
then I was told by Mrs. W. that you had said you would not wish any one to
write till we had heard from you.
Here the compound innitive clearly does not mark anteriority, as the reference
time for anteriority would have been the time of the act of volition, which would
result in a contradiction. What is actually marked is counterfactivity or non-
factivity, a function of compound (especially pluperfect) tense forms prominent,
e.g., in modal constructions (should have said,ought to have done so). This mark-
ing is opposed to a construction with a simple innitive susceptible of a factive
reading (he wrote the story though he did not intend to write it back then). The
same type of non-factivity marking with the aid of a compound tense is observed
in Italian constructions with an apparently ‘illogical’ past conditional in comple-
ments of past-tense verbs:
() Italian
A.
A.
disse
say...3
che
that
sarebbe
be..3sg
arrivata
arrive..
alle
at...
5.
5o’clock
(example from Squartini : )A. said she would arrive at .’.
As Squartini observes, this past conditional has ousted an original present condi-
tional used in other Romance languages and still used in Italian up to the th cen-
tury. The fact that this past conditional does not have its usual anteriority function
(she said she would have arrived by  o’clock) is evident from the lack of a choice
between simple and past conditional. It follows that we are dealing with non-
factivity marking, as in English. Both types of marking exploit anteriority forms
rather than mood, but there is a priori no reason why mood should not be used
for the same purpose (note that typically counterfactive forms like pluperfects can
12. One finds this construction condemned time and again by prescriptive grammarians of the
past, see, e.g., Hodgson (:), which suggests it was actually frequently used.
The dual nature of irrealis in complementation 185
© 2020. John Benjamins Publishing Company
All rights reserved
be sources for irrealis, as Sičinava  argues for Slavic).13 It is perhaps in this
light that we should view the oblique optative of Classical Greek, i.e., the use of
the optative in complements of past-tense verbs (in constructions basically requir-
ing subjunctive or indicative), as can be seen in () as compared to ():
() dédoika
fear..1
m

epilathmetha
forget...1
ts
the...
oíkade
home
hodoũ
way..
‘I fear that we may forget the way home.
() ephobeĩto […]
fear..3
m

ou

dýnaito
be.able...3
exeltheĩn
get.away.
(Rijksbaron :)‘He was afraid that he would not be able to get away.
As a parallel for () we could adduce (), with the counterfactual or emphati-
cally non-factive would have died:
() […] and so home, where I nd my poor father newly come out of an unexpected
t of his pain, that they feared he would have died.
(Samuel Pepys, Diary, Sunday  May )
The problem of the use of the optative in past-tense contexts is still debated in
Greek scholarship and many accounts have been suggested. Its use was, as is
known, optional. Rijksbaron (:) characterises the optative as representing
the narrator’s perspective; this is, in a way, reected in Lillo (), who speaks
of an evidential function. While such explanations may be correct, the functions
they assume can only be secondary as the optative does not have the said func-
tions outside the domain of subordination in past-tense contexts. In order to con-
nect this use to other functions of the optative, which is basically counterfactual,
and to explain the restriction to past-tense contexts, it makes sense to think of this
optative use as a kind of counterfactivity marking in complements of past-tense
verbs as it is attested, in a dierent form, in other languages.
Extensions in the opposite direction also occur. As mentioned, the unanchor-
ing function of the irrealis seems to be involved when irrealis forms occur in the
complements of evaluative predicates, which are basically propositions referring
to situationally and temporally located events.
13. De Haan () argues for including into the discussion on irrealis certain forms that can-
not be formally classified with irrealis but display similar functions; the example he discussed is
that of English will.
186 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
9. In conclusion
The above thoughts on the functions of irrealis mood were prompted by the
twofold use of the notions of ‘realis’ and ‘irrealis’ – one associated with mood dis-
tinctions and the other with putative types of clauses and with complementis-
ers introducing them. As a category concerned with reality status, irrealis mood
should be sensitive to clause type, and in this article I have explored this relation-
ship. Assuming a distinction between propositions and states-of-aairs in clausal
complementation (widely recognised in the literature though the terminology
varies), I have argued that the notion of ‘non-actualisation’ associated with irrealis
has dierent meanings in the two clausal types, viz. ‘irreality’ and ‘unanchoring’.
This is unproblematic if one assumes an invariant feature of ‘non-actualisation’
realised in dierent ways according to the context, or if one describes irrealis
semantics as a kind of network with two distinct foci. But whatever reality such
invariants or networks may have, they are, in any case, just generalisations of
much more concrete meanings arising in particular grammaticalisation contexts.
This process must involve dierent intermediate generalisations, which follow dif-
ferent principles in the two types of clauses outlined here, but may also occa-
sionally cross the borderline between the two domains (as in the case of factive
predicates, which belong to the propositional domain but in some languages show
irrealis uses imported from the states-of-aairs domain).
Acknowledgements
I am indebted to two anonymous reviewers for their incisive and constructive criticisms, and
to Wayles Browne for his useful comments and for improving my English. For the remaining
shortcomings of the article I am solely responsible.
Abbreviations
 accusative
 aorist
 adverb
 comparative
 complementiser
 conditional
 converb
 dative
 denite
 feminine
 future
 genitive
 hortative particle
 imperfect
 indenite
 innitive
 imperfect
 irrealis
 masculine
 neuter
The dual nature of irrealis in complementation 187
© 2020. John Benjamins Publishing Company
All rights reserved
 negative
 nominative
 object
 optative
 perfective
 plural
 possessive
 past participle
 perfect
 present
 past
 reexive
 singular
 subject
 subjunctive
References
Ammann, Andreas & Johan van der Auwera. . Complementizer-headed main clauses for
volitional moods in the languages of South-Eastern Europe: A Balkanism? In
Olga Mis
eska Tomic
(ed.), Balkan Syntax and Semantics, –. Amsterdam: John
Benjamins.
Auwera, Johan van der & Ewa Schalley. . From optative and subjunctive to irrealis. In
Frank Brisard, Michael Meeuwis & Bart Vandenabeele (eds.), Seduction, Community,
Speech. A Festschri for Herman Parret, –. Amsterdam-Philadelphia: John
Benjamins. https://doi.org/10.1075/pbns.127.08auw
Browne, Wayles. . Verbal forms with da in Slovenian complement, purpose and result
clauses. In Neka mu e večna slavata. Studies Dedicated to the Memory of Zbigniew Gołąb
= Balkanistica , –.
Bybee, Joan S. . ‘Irrealis’ as a grammatical category. Anthropological Linguistics ().
–.
Bybee, Joan S., Revere Perkins & William Pagliuca. . The Evolution of Grammar. Tense,
Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press.
Chafe, Wallace. . The realis-irrealis distinction in Caddo, the Northern Iroquoian
Languages, and English. In Joan S. Bybee & Susan Fleischman (eds.), Modality in
Grammar and Discourse, –. Amsterdam: John Benjamins.
https://doi.org/10.1075/tsl.32.15cha
Cleary-Kemp, Jessica. . Irrealis as verbal non-specicity in Koro (Oceanic). Berkeley
Linguistics Society , –. https://doi.org/10.3765/bls.v40i0.3131
Cristofaro, Sonia. . Descriptive notions vs. Grammatical categories: Unrealized states-of-
aairs and ‘irrealis’. Language Sciences (). –.
https://doi.org/10.1016/j.langsci.2011.08.001
Cro, William. . Quantier scope ambiguity and deniteness. Berkeley Linguistics Society
. –.
Dik, Simon C. & Kees Hengeveld. . The hierarchical structure of the clause and the
typology of perception-verb complements. Linguistics (). –.
https://doi.org/10.1515/ling.1991.29.2.231
Dixon, Robert M.W. . Complement clauses and complementation strategies in
typological perspective. In Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds.),
Complementation, –. Oxford: Oxford University Press.
188 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
Dobrušina, Nina. . Soslagatel’noe naklonenie v russkom jazyke. Opyt issledovanija
grammatičeskoj semantiki [The Subjunctive in Russian. An Investigation into
Grammatical Semantics]. Prague: Animedia.
Elliott, Jennifer. . Realis and Irrealis: Forms and concepts of the grammaticalisation of
reality. Linguistic Typology (). –. https://doi.org/10.1515/lity.2000.4.1.55
Gildersleeve, Basil L. . Syntax of Classical Greek from Homer to Demosthenes. New York:
American Book Company.
Haan, Ferdinand de. . Irrealis: Fact or ction? Language Sciences (). –.
https://doi.org/10.1016/j.langsci.2011.06.016
Haspelmath, Martin. . Indenite Pronouns. Oxford: Clarendon Press.
Haspelmath, Martin. . The semantic development of old presents: New futures and
subjunctives without grammaticalization. Diachronica (). –.
https://doi.org/10.1075/dia.15.1.03has
Haspelmath. . Comparative concepts and descriptive categories in cross-linguistic studies.
Language (). –. https://doi.org/10.1353/lan.2010.0021
Hodgson, William B. . Errors in the Use of English. Edinburgh: David Douglas.
Holvoet, Axel. . Mood and Modality in Baltic. Cracow: Wydawnictwo Uniwersytetu
Jagiellońskiego.
Jespersen, Otto. . A Modern English Grammar on Historical Principles. Part iv. Syntax
(Third Volume). London: George Allen & Unwin.
Kehayov, Petar & Kasper Boye. a. Complementizer semantics – an introduction. In:
Kasper Boye & Petar Kehayov (eds), Complementizer Semantics in European Languages,
–. Berlin: De Gruyter Mouton. https://doi.org/10.1515/9783110416619004
Kehayov, Petar & Kasper Boye. b. Complementizer semantics in European languages:
Overview and generalizations. In Kasper Boye & Petar Kehayov (eds.), Complementizer
Semantics in European Languages, –. Berlin: De Gruyter Mouton.
https://doi.org/10.1515/9783110416619023
Kuryłowicz, Jerzy. . The Inectional Categories of Indo-European. Heidelberg: Carl
Winter.
Lillo, Antonio. . On the oblique optative in Herodotus’ completive sentences: An
evidentiality mark in Ancient Greek. In Felicia Logozzo & Paolo Pocett (eds.), Ancient
Greek Linguistics. New Approaches, Insights, Perspectives, –. Berlin: De Gruyter.
https://doi.org/10.1515/9783110551754325
Lunn, Patricia V. . Spanish mood and the prototype of assertability. Linguistics (),
–. https://doi.org/10.1515/ling.1989.27.4.687
Mauri, Caterina & Andrea Sansò. . The reality status of directives and its coding across
languages. Language Sciences (), –. https://doi.org/10.1016/j.langsci.2011.08.002
Mithun, Nancy. . On the relativity of irreality. In Joan S. Bybee & Susan Fleischman (eds.),
Modality in Grammar and Discourse, –. Amsterdam: John Benjamins.
https://doi.org/10.1075/tsl.32.16mit
Nikolaeva, Irina. . Analyses of the semantics of mood. In Jan Nuyts &
Johan van der Auwera (eds.), The Oxford Handbook of Modality and Mood, –.
Oxford: Oxford University Press.
Palmer, F. R. . Mood and Modality, nd edn. Cambridge: Cambridge University Press.
https://doi.org/10.1017/CBO9781139167178
Ransom, Evelyn. . Complementation: Its Meanings and Forms. Amsterdam: John
Benjamins. https://doi.org/10.1075/tsl.10
The dual nature of irrealis in complementation 189
© 2020. John Benjamins Publishing Company
All rights reserved
Rijksbaron, Albert. . The Syntax and Semantics of the Verb in Classical Greek, nd edn.
Amsterdam: Gieben.
Sičinava, Dmitrij V. . K probleme proisxoždenija slavjanskogo uslovnogo naklonenija [On
the problem of the origin of the Slavic conditional]. In Jurij A. Lander,
Vladimir A. Plungjan & Anna Ju. Urmančieva (eds.), Irrealis i irreal’nost’ [Irrealis and
Irreality], –. Moskva: Gnozis.
Sperber, Dan & Deirdre Wilson. . Relevance: Communication and Cognition. Cambridge
MA: Harvard University Press.
Squartini, Mario. . Mood in Italian. In Björn Rothstein & Rolf Thiero (eds.), Mood in the
Languages of Europe, –. Amsterdam: John Benjamins.
https://doi.org/10.1075/slcs.120.13squ
Szucsich, Luka. . Mood in Bosnian, Croatian and Serbian. In Björn Rothstein &
Rolf Thiero (eds.), Mood in the Languages of Europe, –. Amsterdam: John
Benjamins. https://doi.org/10.1075/slcs.120.22szu
Švedova, Natalija Ju. (ed.). . Russkaja grammatika [Russian Grammar] i–ii. Moskva:
Nauka.
Thiero, Rolf. . Moods, moods, moods. In Björn Rothstein & Rolf Thiero (eds.), Mood
in the Languages of Europe, –. Amsterdam: John Benjamins.
https://doi.org/10.1075/slcs.120.01thi
Address for correspondence
Axel Holvoet
Vilnius University
Institute for the Languages and Cultures of the Baltic
Universiteto 
LT- Vilnius
Lithuania
190 Axel Holvoet
© 2020. John Benjamins Publishing Company
All rights reserved
... Communication is carried out by the addressee with the addresser, with both the addressee and the addresser being intelligent and subjective, according to the definition, but not with the intelligible world of objects as things-in-themselves (Holvoet, 2020). Therefore we do not agree with either structuralists (for example, with R. Jakobson, who objectified the subject of communication 1 ), or with postmodernists for example, Deleuze and Guattari (2004), who eliminate the idea of the subject from their systems). ...
Conference Paper
Full-text available
This paper investigates metaphysical character of communication since the main difficulty in determining it lies in the way information passes from one subject to another without their direct interaction in a metaphysical way. Communication models developed by phenomenologists, behaviourists and postmodernists represent communication as subjectless and objective, thus depriving the participants of intelligibility. However, as argued in this paper, the participants in communication are intelligible. On this basis, we assert that any communication is not reproductive but creative in its primary origin for each participant in communication creates a certain ideal sign action (semiosis), so that the signal is a consequence of the semiosis of the speaker and the cause for the semiosis of the listener. Since semiosis is thought to be relation between the signifier and the signified and this relation cannot be represented either as a substance (inclusion), or as a process (implication) the only way to avoid accidentality of the components of semiosis following in the material form of time, is by introducing the ideal, which would connect them, and therefore make their relation necessary. It allows us to consider communication as relation of two semiosis combined by a signal. Following from it and from definitions of components of semiosis we give trascendentalistic definition of communication and create a diagram demonstarting its properties.
... Communication is carried out by the addressee with the addresser, with both the addressee and the addresser being intelligent and subjective, according to the definition, but not with the intelligible world of objects as things-in-themselves (Holvoet, 2020). Therefore we do not agree with either structuralists (for example, with R. Jakobson, who objectified the subject of communication 1 ), or with postmodernists for example, Deleuze and Guattari (2004), who eliminate the idea of the subject from their systems). ...
Article
In this article a parallel is drawn between certain functions of irrealis and imperfective in evaluating contexts in Russian and Polish. The functions of irrealis in complementation are twofold: while in propositional complements it reflects irreality, in state-of-affairs complements it reflects temporal and situational unanchoring. This unanchoring function manifests itself also in complements of evaluative (commentative) predicates, where it extracts an event from its situational setting for the purpose of evaluating its intrinsic properties – intrinsic likelihood or desirability. It is argued that the representation of event tokens as event types by means of imperfective verbs in Slavonic performs a similar unanchoring function in evaluative contexts. The data of Slavonic languages, where both unanchoring devices cooccur, enable a coherent explanation of certain hitherto not fully understood phenomena in the domains of mood and aspect and shed a new light on the long-standing problem of the Romance so-called “factive” subjunctive.
Article
This paper develops a framework capable of analysing modal qualifications coded by both canonical modal markers and grammaticalized expressions within a semiotic grammar approach (McGregor 1997). We focus on grammatical constructions that developed from complement constructions containing the shell nouns need, way, chance, doubt, question and wonder, often preceded by a negative quantifier. We range the different types of modal qualification in a hierarchy. The top half subsumes mirativity and epistemic modality, which scope over propositions, and interpersonal deontic modality, which scopes over processes. The bottom half contains representational deontic and dynamic modality, which are internal to the proposition. We also address the neglected issue of the presence of polarity choices in the modal qualifications and in the propositions and processes. In particular, we argue that epistemic modality ascribes inherently positive degrees of likelihood to either positive or negative propositions, thus assessing the probability of occurrence versus non-occurrence of temporally located processes. This allows us to explain not only the semantic equivalence involved in NEG-raising, as in I think he did nothing wrong and I don’t think he did anything wrong, but also, more broadly, between pairs like there’s no chance she will cry and there’s no doubt she will not cry. Finally, the proposed semantic model puts us in a position to propose structural analyses of the interpersonal and representational qualifications of propositions and processes within the broad outlines of McGregor’s (1997) Semiotic Grammar.
Preprint
Full-text available
Though mood and modality are usually mentioned in close conjunction, as one research domain, the notion of modality has recently become increasingly prominent at the expense of mood. Yet the notion of mood, mostly restricted to the realis-irrealis distinction, has not been discarded. In this article an attempt is made to understand the enduring viability of the notion of mood by looking at the features prototypically associated with it. These features are partly formal, as mood is mainly associated with inflectional exponence, but also partly semantic, as formations diverging from prototypically inflectional mood tend to be classified with mood to the extent that they reflect the more abstract semantic features associated with the realis-irrealis distinction. In the final part of the article the question of the relationship between mood and modality is once more addressed.
Research
Full-text available
The article attempts to provide an answer to the puzzle of the so-called factive irrealis, a type of irrealis attested in the Romance languages and consisting in the use of irrealis forms with predicates presupposing that the content of their complements is real. It is argued that this irrealis use reflects the historical process of transfer of irrealis from the states-of-affairs type of clausal complements to the propositional type, and involves a certain reappropriation of the characteristic feature of irrealis in state-of-affairs complementation, viz. suspension of tense features. In evaluative contexts this suspension of tense features performs an unanchoring function, i. e., it extracts an event from its situational setting for the purpose of evaluating its intrinsic properties-intrinsic likelihood or desirability. It is argued that the representation of event tokens as event type by means of imperfective verbs in Slavonic performs a similar unanchoring function in evaluative contexts, which provides independent evidence for the relevance of the notion of temporal unanchoring in accounting for the factive irrealis. This, in turn, shows the relevance of temporal unanchoring as one of the features constituting the semantic domain of the irrealis, alongside counterfactivity or nonfactivity.
Chapter
The oblique optative in Herodotus’ completive sentences is the marked element from the point of view of Evidentiality; the indicative and subjunctive have no such marks, since these forms indicate pure actions, within their corresponding temporal and modal coordinates. Unlike these forms, the oblique optative presents actions from the viewpoint of the source of information, actions that the speaker knows by verification or inference. In consequence, oblique optative would not be merely a mark of subordination, but a modal use to mark the source of the information and, consequently, it would be a mark of Evidentiality.