Expletive Negation in Attitude Contexts
Maria-Margarita Makri
This paper examines the syntax-semantics of expletive negation in Modern and Classical Greek,
Latin, Spanish, French, Russian and Hebrew attitude complements. It shows that, contrary to
what is standardly assumed, expletive negation is not necessarily licensed in Subjunctive complements. It is licensed by predicates allowing for more than one live doxastic alternative, it
scopes above Tense, and it is in complementary distribution with epistemic modals. Based on
novel data I show that expletive negation is not semantically vacuous; I propose that it is a weak
epistemic modal that marks the contextually provided, ordered in terms of likelihood alternatives as equally likely.
1.
Introduction
The aim of this paper is to identify the distribution and the meaning of ‘expletive negation’ in
the clausal complements of attitude predicates. Expletive Negation (henceforth, simply referred
to as EN) refers to a sentential negator or a negative complementizer that does not reverse
the polarity of a proposition1 . If EN is a complementizer then it can be identical to sentential
negation (e.g. Modern Greek min (NEG), Latin ne (NEG)) or comprised of a negative marker
and a complementizer (e.g. Modern Greek mipos (NEG-that) or Latin qui-n<e> (that-NEG)).
In (1-a) below min (NEG) does not change the polarity of the embedded proposition: what the
speaker fears is that the addressee might forget something. In contrast, if min (NEG) does not
introduce the complement of fear but a purpose clause, then the clause it introduces is negative
as expected (1-b):
(1)
[Modern Greek]
a. Fovame min to
ksehasis EN
fear.1SG NEG it.CL forget.2SG
‘I fear that you might forget it.’
1
(Propositional) attitudes: mental states that we might have to propositions, e.g. belief, knowledge, suspicion,
discovery, etc.
Attitude predicates: believe, know, realize, think, discover, etc.
Proceedings of ConSOLE XXIII, 2015, 427-448
http:// www.sole.leidenuniv.nl
c Maria-Margarita Makri
428
Maria-Margarita Makri
b.
Simiose
to
min to
ksehasis ‘Real Negation’
note.IMP.2SG it.CL NEG it.CL forget
‘Write it down so that you do not forget it.’
EN is attested crosslinguistically in several constructions such as exclamatives, questions, comparatives and adjunct clauses introduced by before, until, unless, or without (for the distribution
of EN crosslinguistically cf. Appendix). This paper focuses on EN in attitude complements and
draws data from Modern Greek (MG)(1-b), Classical Greek (ClGr) (2), Latin (3), Spanish (4),
French (5), Hebrew (6) and Russian (7).
(2)
(3)
deido: me: ou tis
toi huposche:tai tode ergon. [ClGr]
fear NEG NEG somebody you promise
this task
‘I am afraid lest none should undertake for you this task.’
agebamus verentes ne quid
accideret. [Latin]
live
fearing NEG something happen
‘We live fearing lest something happens.’
(4)
Temo no le haya sucedido alguna desgracia. [Spanish]
fear NEG her have happen any misfortune
‘I am worried she (might) have suffered any misfortune.’
(5)
Elle appréhende qu’ il ne se mette à pleuvoir. [French]
she apprehend that it NEG SE put at rain
‘She fears that it might start raining.’
(6)
Gil nimna/ xadal me-ledaber im Rina. [Hebrew]
Gil refrained/ stopped NEG-talk to Rina
‘Gil refrained from/stopped talking to Rina.’
(7)
Ja bojus’, kak by on ne opzdal. [Russian]
I fear how SBJ he NEG was-late
‘I fear that he might be late.’
Homer, K39
Cic. Fam. 13, 19, 2
(Landau 2002)
(Brown & Franks 1995)
Table 1 below shows the distribution of EN in the languages mentioned above. I distinguish
four classes of predicates that license EN: emotive doxastics (verbs like fear or hope, cf. Anand
and Hacquard 2013), dubitatives, interrogative predicates and negative predicates. Yoon (2012)
based on Korean was the first to show that EN is licensed after predicates that do not have
a negative meaning (e.g. hope). As shown in Table 1 below, this also holds for Modern and
Classical Greek.
There have been several attempts to explain the meaning of EN. The previous analyses can
be classified in two major classes: the analyses that assume that EN is indeed semantically
vacuous (e.g. Espinal 1992, 2000, 2007) and those that do not. The second class can be further
subdivided into analyses in which EN is identified with ‘real’ negation and those that EN has
some semantic contribution different than that of real negation (e.g. Yoon 2011). To anticipate
the discussion, in this paper, I provide new evidence that EN is not semantically vacuous and I
argue that it is a weak epistemic modal.
Section 2 examines the distribution of EN in sentential complements and demonstrates that
429
Expletive Negation in Attitude Contexts
Table 1: EN Sentential Complements
fear compl.
hope compl.
dubitatives
hinder, resist,
refuse, delay ∗∗
interrogative
complements
∗
MGr.
+
+
+
-
ClGr.
+
+
+
+
Latin
+
+∗
+∗
Span.
+
+
-
French
+
+∗
+
+
+
+
-
-
Hebr.
+
Rus.
+
+
-
EN is licensed on condition that the matrix predicate is negated
The negative verbs that license EN may vary across languages
∗∗
EN is not licensed by Subjunctive and that it scopes above Tense. Section 3 shows that the predicates that allow for EN are predicates that have existential force and select for complements
of type <<s,t>,t>. In section 4, I examine the meaning of EN and I show three new asymmetries between sentences with and without EN negation. Based on these asymmetries and the
licensing conditions of EN identified in section 2 and section 3, I argue, in section 4, that EN
is a weak epistemic modal. Not only does this proposal capture the crosslinguistic distribution
of EN and its meaning but also it explains why EN can replace epistemic tha (will -MGr) in
counterfactuals. Finally, section 5 concludes.
2.
The distribution of EN in sentential complements
2.1.
EN and mood
Previous analyses of EN assume or argue that EN is licensed in Subjunctive complements: Abels
(2002, 2005) notes that EN in Russian is licensed if the sentence is in Subjunctive while Gaatone
(1971) and Muller (1978) argue that Subjunctive mood triggers EN in Romance. Finally, Yoon
(2012) claims that indicative predicates such as assertive, epistemic, factive and semi-factive
verbs are ungrammatical with EN, and she argues that EN is licensed by notional Subjunctive.
If we look at the crosslinguistic distribution of EN in sentential complements, it is clear that
there is not a correlation between EN and mood. As Table 2 shows, EN is not licensed exclusively in Subjunctive complements: it can be licensed in clausal complements with Indicative,
Subjunctive and Optative even Infinitival complements. Data in (8) show an example, taken
from Classical Greek:
(8)
[ClGr]
a. skopeite me:
doke:sin eichete
ek theo:n
consider NEG.NONIND. fancy have.IND.PST.IMP.2PL by gods.GEN.
‘Consider whether you had some fancy, sent by the gods.’
Eur. Hel. 119
b. ephobe:the:san me:
kai epi
sphas ho stratos
fear.PST.PERF. NEG.NONIND. and against them the army
430
Maria-Margarita Makri
Modern
Greek
Classical
Greek
Latin
French
Spanish
Hebrew
Russian
c.
d.
e.
Table 2: Mood Distribution of EN complements
Emotive
Dubitatives
Interrogatives
Negative
Doxastics
Predicates
Indicative
Indicative
Indicative
Subjunctive
Indicative
Indicative
Indicative
Subj/ve
Subjunctive
Subjunctive
Subjunctive
Optative
Optative
Optative
Infinitive
Infinitive
Infinitive
Subj/ve
Subjunctive
Subjunctive
Subjunctive
Subj/ve
Subjunctive
Subjunctive
Subj/ve
Infinitive
Subj/ve
Subjunctive
Infinitive
cho:re:se:i.
march.SBJ.PST.PFV
‘they feared lest the army marches against them.’
Thuc. II 101,2.
edeisan de me:
lutta tis ho:s-per
kusin humin
fear
PCL NEG.NONIND. rage some like-exactly dogs you.PL
empepto:koi
fall-upon.OPT.PRS.PRF.
‘they feared lest some kind of madness might prove to have fallen upon you exactly
like upon dogs.’
Xen. Anab. V 7,26.
ou-te
he: parousa eudhaimonia pareschen oknon me:
NEG.IND-and the present happiness provide fear NEG.NONIND.
elthein
es ta deina
come.PST.PRFV.INF. to the danger
‘nor the existing prosperity could dissuade them from affronting danger.’ Thuc. III
39,3
phobeisthai
to
me:-te
fear.V.INF.PRS.MP the.D.N.ACC. NEG.NONIND.-and
epenegkein
pseude:
timo:rian
bring.upon.PST.PFV.INF. wrongful.F.ACC punishment.F.ACC
‘. . . to dread bringing upon him a wrongful punishment . . . ’
Plat. L. 12.943d
On the other hand, craindre (fear -French) can select Subjunctive or Infinitival Complements.
However, EN is grammatical only in the Subjunctive complements of craindre (9). So, if EN is
compatible with infinitives in Classical Greek, there is a puzzle that remains unsolved: If EN is
not incompatible with infinitives (cf. Table 2 and (8)), why is EN ungrammatical with French
and Spanish infinitival complements?
(9)
[French]
431
Expletive Negation in Attitude Contexts
a.
b.
Je crains qu’ il ne vienne.
I fear that he NEG come.SBJ.
‘I fear that he might come.’
Je crains d’ (*ne) apprendre une mauvaise nouvelle.
I fear of NEG learn.INF a bad
news
‘I am afraid of hearing some bad news.’
In section 2.2, I will show that the difference between French and Spanish infinitives on the one
hand and Classical Greek, Russian and Hebrew, on the other, is that the first class of infinitives
always has anaphoric Tense while the second (can) have free or dependent tense.
2.2.
EN and Tense
In this section, I examine the Tense properties of complement clauses with EN. To this end,
I adopt the notion of (in)dependent and anaphoric tense as in Picallo (1984); Iatridou (1993);
Landau (2004) a.o. The Semantic Tense of an embedded clause can be:
Independent or free: The embedded Tense can have any Tense specification, e.g. indicative
clauses. In these clauses, morphological tense corresponds to semantic tense. A sentential
complement with independent tense refers to an event that is not related to the matrix
predicate (e.g. verba dicendi et sentiendi).
Dependent Tense: The embedded Tense alternations are constrained by the matrix predicate,
e.g. some of the Balkan subjunctives. In these clauses, morphological tense alternations
reflect aspectual alternations and the semantic tense is future oriented/irrealis.
Anaphoric or empty Tense: The event of the embedded clause is not distinct from the event of
the matrix predicate, e.g. started to play. In these clauses, the morphological Tense of the
embedded predicate is always [-Pst]. In this paper, I follow Grano (2012) who argues that
obligatory morphological nonpast tense (i.e. anaphoric semantic tense) indicates the absence of tense altogether. In other words, sentential complements with ‘anaphoric Tense’
lack a Tense projection.
Three criteria are used for characterising Tense:
1. the availability of a matrix and an embedded temporal adverb with distinct reference:
(10)
Včera
oni bojalis’
kak by
mafija ix zavtra ne
Yesterday they feared.PST how MOD mafia them tomorrow NEG
našla.
[Russian]
find.PST.PRF
‘Yesterday they feared that tomorrow he might find them.’ (N. Radkevic p.c.)
2. licensing of an overt embedded subject;
432
Maria-Margarita Makri
(11)
[Hebrew]
a. Ha-bikus ha-acum mana
me-ha-mexirim laredet.
the-demand the-huge prevented from-the-prices to-fall.INF.
‘The huge demand prevented the prices from falling.’
b. Ha-kaba’im man’u
me-ha-es
le’hitpaset.
the-firemen prevented from-the-fire to-spread.INF.
‘The firemen prevented the fire from spreading.’
(Landau 2002: 469)
3. availability of morphological tense alternations.
(12)
[ClGr]
a. apistountes te
me:
einai
tous
disbelieve and.PCL NEG.NONIND. be.PRS.INF them.ACC.PL.
paradontas
tois
tethneo:sin
surrender.PCPL.M.ACC.PL. the.M.DAT.PL. die.PCPL.M.DAT.PL
omoious
same
‘people could scarcely believe that those who had surrendered were of the
same stuff as the fallen . . . ’
Thuc. 4.40
b. apistountes
auton me:
he:ksein
disbelieve.PRS.PCPL.PL him.PR. NEG.NONIND. come.FUT.INF.
‘not believing that he would come.’
Thuc. 2.101
c. . . . an tis
. . . apistoie: me:
genesthai
ton stolon
if
somebody disbelieve NEG.NONIND. be.PST.PFV.Inf the navy
tosouton hoson hoi poie:tai eire:kasi
that-big as-big the poets say.PRS.PRF
‘If somebody disbelieved that the navy was as big as the poets have said.’
Thuc.1.10
d. apistoumen
me:
kata
te:n oligarchoumene:n
disbelieve.PRS.1PL NEG.NONIND. according-to the oligarchical
polin omoiote:ti
ton
pheido:lon te
city resemblance.SG.DAT the.M.SG.ACC and.PCL and
kai
chrematiste:n
tetachthai?
money-making.M.SG.ACC classify.INF.PRS.PF.
‘Have we any further doubt as to the correspondence and resemblance between the thrifty and money-making man and the oligarchical state?’ Plat.
Rep. 8.555a-b
The availability of morphological tense alternations is sufficient to assume that a clause does not
have anaphoric tense, but this is not biconditional: the infinitival clauses of a language can bear
free or dependent Tense without being morphologically marked for that. Russian and Hebrew
infinitives are not morphologically marked for tense, though based on the former two diagnostics (temporal adverbs, subject licensing) the diagnostics indicate that they have an active tense
433
Expletive Negation in Attitude Contexts
operator.
Table 3 below summarizes the tense properties of EN complements. It is evident that EN
is not licensed in sentences with anaphoric tense or put differently, it is not licensed in complements that lack a Tense projection. This indicates that EN scopes above Tense, otherwise it
would also be grammatical in clauses with anaphoric tense.
Table 3: Tense of EN sentential complements
Emotive
Dubitatives Interr/ves NEG entailing
Doxastics
Verbs
Modern Greek
Free
Free
Free
Classical
Free
Free
Free
Free
Greek
Dependent
Dependent
Latin
Dependent Dependent Dependent
Dependent
French
Dependent Dependent
Dependent
Spanish
Dependent
Hebrew
Free
Russian
Free
Free
How can the newly established correlation between EN and T solve our previous puzzle, i.e
why is EN ungrammatical in French infinitives? As shown above, EN is licensed only in Tensed
clauses. The difference between French and Classical Greek infinitives, therefore, lies in that the
latter but not the former have an active Tense operator. This also explains the difference between
(9-a) and (9-b) above: French Subjunctive complements (in contrast to infinitival ones) are used
in obviative environments and inflect for tense. This indicates that they have dependent Tense;
thus EN can be licensed. On the other hand, French infinitival complements have anaphoric
Tense; thus EN is ungrammatical.
Table 4: Comparative Table of infinitival complements
ClGr Hebrew Russian Latin French
Does the language have Yes
Yes
Yes
No
No
infinitival Cs?
Is the EN infinitive intro- No
Yes
Yes
N/A
N/A
duced by an overt C?
Categorial Status of the CP
CP
CP
N/A
N/A
EN infinitival clause
Categorial status of the CP
TP
TP
CP/VP
CP
infinitives
Can the infinitives license Yes
Yes
Yes 2
(Yes)
No
their own subject?
Do the infinitives have Yes
Yes
Yes
(Yes)
No
their own tense domain?
2
Though the use of EN with infinitives in Russian is not very common, it is possible:
Spanish
No
N/A
N/A
CP
No
No
434
Maria-Margarita Makri
As far as French is concerned, this correlation between EN and Tense leads us to one more
prediction: EN in attitude complements is only found in high register French where tensed
Subjunctive is still used. On the other hand, it is not expected to be found in low and middle
registers of French as Tensed Subjunctive is not used in these varieties anymore.3 Indeed, EN
in attitude complements is nowadays used only in high registers. This also leads to a second
prediction: that EN would be productive in earlier stages of French when tensed Subjunctive
was still used.
Additionally, the correlation between Tense and EN put forward in this section makes relevant an analysis of control in EN constructions. There is a debate whether control is dependent
on the Tense specification of the embedded clause or not. Sevdali (2006 et seq.) has argued that
control is not dependent on Tense whereas Landau (2009) argues that there is a well established
correlation between (non) availability of obligatory control and the Tense specification of the
embedded clause. If Landau is correct, another prediction is made: EN complements are never
obligatorily controlled.
In this section, I argued against the dominant thesis that EN is licensed by Subjunctive and
I showed that it is licensed only in ‘tensed’ clausal complements, which means that EN scopes
above Tense. In the next section I will identify another necessary condition for EN licensing:
the selecting predicate must allow for more than one live doxastic possibility, in other words, to
have existential force and select for complements of type <<s,t>,t> .
3.
The semantics of the licensing predicates
This section focuses on the semantics of the predicates that select for EN. These predicates have
been described as ‘verbs of fear . . . and verbs of doubt, which are all non-veridical with respect
to the proposition expressed in the subordinate clause’ (Espinal 2000: 61), ‘verbs or other lexical elements with “negative import” ’ (Wouden 1994), ‘negative entailing . . . and weak implicatives in the sense of Pesetsky (1991)’ (Landau 2002: 476, fn.11), ‘adversative predicates/fearcomplements’, ‘hope’ and ‘dubitatives’, ‘negative predicates licensing a negative implicature’
Yoon (2011). As mentioned in the introduction Yoon (2011) was the first to observe that negativity is not the common link amongst the predicates that license EN, however, she assumes that
the only distinctive property of predicates selecting EN is non-veridicality4 .
In the remainder of this section, I will examine the properties of each semantic class and
(i)
Včera
oni bojalis’ kak by
im zavtra
ne opzdat’
na poezd. [Russian]
yesterday they fear
how MOD them tomorrow NEG be.late.INF.PRF. on train
‘Yesterday they feared that they might miss the train.’ (N.Radkevic p.c.)
3
4
The reader is referred to Iatridou (2000) for a discussion of the loss of French Past Subjunctive.
(Non)veridicality for propositional operators (Giannakidou 2006)
a. A propositional operator F is veridical iff F p entails or presupposes that p is true in some individual’s epistemic
model ME (x); otherwise F is non-veridical.
b. A nonveridical operator F is ANTIVERIDICAL iff F p entails that NOT p in some individual’s epistemic model:
F p → ¬p in some ME (x).
Expletive Negation in Attitude Contexts
435
propose that EN is licensed only by predicates that have existential force and select for complements of type <<s,t>,t>.
3.1.
Emotive Doxastics (fear, hope) (modified from Anand & Hacquard 2013)
In this section, I will examine the semantics of predicates meaning ‘fear’ or ‘hope’. These predicates fall into the class Anand & Hacquard (2013) call ‘emotive doxastics’. They are classified
as such because of their hybrid nature: they present doxastic alternatives of an attitude holder
as doxastics do (e.g. believe) but they also involve a preference component like desideratives/
directives. Their hybrid nature is also reflected in epistemic modal licensing: emotive doxastics
license possibility epistemics as doxastics do, but they do not license necessity epistemics as
subjunctive selecting predicates.
(13)
John fears that Mary is pregnant.
Presupposes: John is not sure whether Mary is pregnant or not.
Asserts: John considers possible both that Mary is pregnant and that she is not. &
John considers likely that Mary is pregnant. &
Mary being pregnant is less desired than not being pregnant.
(14)
John hopes that Mary is pregnant.
Presupposes: John is not sure whether Mary is pregnant or not.
Asserts: John considers possible both that Mary is pregnant and that she is not. &
John considers likely that Mary is pregnant. &
For John, Mary being pregnant is more desired than not being pregnant.
This meaning of fear can be formally expressed as in (15) (modified from Anand & Hacquard
2013):
(15)
J fearC that ϕKc,w,S,g =
λx: ϕ-verifiers in S 0 6= ∅ & ϕ-falsifiers in S 0 6= ∅
(uncertainty cond.)
If defined =1 iff
0 0
∃w0 ∈ S 0 : [JϕKc,w ,S ,g = 1]∧
(doxastic assertion)
(preference assertion)
ϕ-verifiers <Desx,w ϕ-falsifiers ∧
ϕ-verifiers >P robx,w ϕ-falsifiers
(preference assertion)
where S 0 =DOXx,w and ϕ-verifiers in S’= λS 00 .S 00 ⊂ S 0 & ∀S 000 ⊂ S 00 : [∀w0 ∈
0 000
S 000 [JϕKc,w ,S ,g = 1]]
ϕ-verifiers in S 0 = ¬ϕ-falsifiers in S 0
In (15) above, S stands for an information state (Yalcin 2007), namely a body of knowledge
or evidence of type < s, w >. According to Yalcin, an attitude verb quantifies over a set of
possibilities compatible with the attitude state and shifts the value of the information state parameter s to that set of possibilities. Anand & Hacquard (2013) define an information state S 0
as a ϕ-verifier iff ϕ is true relative to S and all of its subsets. As is also shown in the formula
(15) above, ϕ-verifiers and ϕ-falsifiers are in complementary distribution in S 0 .
An emotive doxastic therefore introduces a bipartition of possible worlds into those that the
436
Maria-Margarita Makri
embedded proposition ϕ is true and those that ϕ is false. It asserts that the embedded proposition
can be true and that ϕ is more likely to be true than false. 5 Finally, verbs meaning fear and verbs
meaning hope introduce a desirability scale on which the embedded proposition being true is
less or more desired respectively than being false (cf. (13) and (14) above).
3.2.
Dubitatives (e.g. doubt, suspect) (Anand & Hacquard 2013)
The meaning of dubitatives is very similar to the meaning of emotive doxastics: they presuppose
that the attitude holder is not certain about the truth of the embedded proposition and they have
a double assertion. They assert that the attitude holder considers the embedded proposition
possible (representational component) and that the embedded proposition is more likely to be
false than true. This is exemplified in (16) and formalised in (17) below:
(16)
John doubts that Mary is pregnant
Presupposes: John is not sure whether Mary is pregnant or not.
Asserts: John considers possible that Mary is pregnant. &
John considers likely that Mary is not pregnant.
(17)
J doubtC that ϕKc,w,S,g =
λx: ϕ-verifiers in S 0 6= ∅ & ϕ-falsifiers in S 0 6= ∅
uncertainty cond.
If defined =1 iff
0 0
∃w0 ∈ S 0 : [JϕKc,w ,S ,g = 1]∧
doxastic assertion
preference assertion
ϕ-verifiers <P robx,w ϕ-falsifiers
where S 0 =DOXx,w and ϕ-verifiers in S’= λS 00 .S 00 ⊂ S 0 & ∀S 000 ⊂ S 00 : [∀w0 ∈
0 000
S 000 [JϕKc,w ,S ,g = 1]]
ϕ-verifiers in S 0 = ¬ϕ-falsifiers in S 0
So dubitatives, similar to emotive doxastics, involve an uncertainty condition, a weak doxastic
assertion and an ordering relation.
3.3.
Interrogative predicates
The predicates that select interrogative complements can be divided into three classes as shown
in (18) Lahiri (2002):
5
Anand and Hacquard (2013) propose that the preference component of emotive doxastics consists only of the
desirability ordering of ϕ-verifiers and ϕ-falsifiers. The addition of the likelihood scale is an iteration from their
proposal.
437
Expletive Negation in Attitude Contexts
(18)
Predicates that take interrogative Complements
Rogative
wonder, ask, depend on
investigate, examine
Responsive
Veridical Responsive
know, rembember, tell
forget, be aware, be surprised
Non-veridical Responsive
be certain (about), agree (on),
conjecture (about)
Lahiri (2002)
From these three classes of predicates only rogative and negated veridical responsive predicates
can select for EN.
3.3.1.
Veridical Responsive Predicates
Veridical responsive predicates select for interrogative complements introduced by whether/if
or declarative complements introduced by that. They do not select for EN, unless they are
negated (19)-(20-b):
(19)
Thimame pos/ an/ *mi-pos
itan i Maria pu eklise tin porta. [MG]
remember that/ if/ lest.NEG-that was the Mary that closed the door
‘I remember that/ whether/*lest Mary shut the door.’
(20)
[MG]
a. Dhen thimate pos/ an i Maria eklise tin porta.
NEG remember that/ if the Mary closed the door
‘(S)he does not remember that/if Mary shut the door.’
b. Dhen thimate mipos itan i Maria pu eklise tin porta.
NEG remember lest was the Mary that closed the door
‘(S)he does not remember lest Mary shut the door.’
Uegaki (2012) proposes the following semantics for know6 :
(21)
JknowKw = λQ ∈ D<st,t> : [∃p ∈ Q[p(w) = 1]] ∧ λx.∀p ∈ Q[p(w) = 1 →
DOXx,w ⊆ p].
According to (21) know presupposes that in w there is some proposition p that is a true answer
to the embedded question Q and that for all true propositions p in w, the doxastic alternatives
of the attitude holder (matrix subject) are a subset of p. Even though this seems to capture
correctly the truth conditions of sentences in which know has selected a that complement, it
falsely predicts that a clause with know entails the embedded proposition even when that is
introduced by whether.
6
Know and remember pattern alike with respect to complement selection (Lahiri 2002) and EN licensing therefore I treat them the same.
438
Maria-Margarita Makri
As suggested by the translations in (19) above and shown in (22) and (23) below, veridical
responsive predicates are implicatives only if they select a that-complement; with a whethercomplement they are not.
(22)
John knows that Mary closed the door.
Presupposition: Mary closed the door
Assertion: John believes only that Mary closed the door.
(23)
John knows whether Mary closed the door.
Assertion: John knows the correct answer to the question ‘Did Mary close the door?’.
Based on this data I will revise Uegaki’s proposal as in (24):
(24)
a.
b.
Jknow that pKw = λQ ∈ D<st,t> : [∃p ∈ Q[p(w) = 1]] ∧ λx.∀p ∈ Q[p(w) = 1 →
DOXx,w ⊆ p].
Jknow whether pKw = λQ ∈ D<st,t> : λx.∀p ∈ Q[p(w) = 1 → DOXx,w ⊆ p].
The pair in (24) captures the fact that the verb know entails the truth of the embedded proposition
only if it selects complementizer that. 7
3.3.2.
Rogative predicates
Rogative predicates (e.g. ask, wonder, investigate, examine) select for questions or EN complements:
(25)
Rotisa ean/ mipos/
*pos /*na hriazosun tipota.[MG]
Asked if/ lest.NEG-that/ that/ SBJ needed anything
‘I asked if/whether/*that/*to you needed anything.’
As Uegaki (2012) argues both veridical responsive and rogative predicates select for questions
(<< s, t >, t >). As he shows, the difference between the two lies in that rogative predicates
carry a ‘non triviality presupposition’ defined as:
(26)
JwonderKw : (Q)(x) is defined iff x can believe both of the following:
a. λw.∃p ∈ Q[p(w) = 1] In prose, there is a true proposition in Q.
b. λw.∃p ∈ Q[p(w) = 0] In prose, there is a false proposition in Q. (Uegaki 2012)
This non-triviality presupposition ensures that there is a bipartition in the doxastic alternatives
of the attitude holder, such that they include any possible answer to the question and not only
the true one. Based on (26) the meaning of wonder is (27):
(27)
JwonderKw : λQ ∈ D<st,t> : p(w) = 1 ⊂ DOXx,w ∧ p(w) = 0 ⊂ DOXx,w .λx.∃p ∈
Q[p(w) = 1 →DOXx,w ⊂ p]
7
This distinction suggests that it is the complementizer that asserts (or not) the truth of the embedded proposition. However, an elaborate analysis of the semantics of Cs lies beyond the scope of this thesis.
439
Expletive Negation in Attitude Contexts
In the light of the analysis of question embedding predicates provided in this section, it is
evident that EN is only licensed by predicates that presuppose or assert that there is not only
one doxastic possibility in the doxastic alternatives of the attitude holder. In rogative predicates
it is presupposed that the attitude holder does not know the correct answer whereas in negated
responsive veridical predicates it is asserted that the attitude holder does not know the correct
answer 8 .
If we compare the non-triviality presupposition of rogative predicates proposed by Uegaki
(2012) and the uncertainty condition defined by Anand & Hacquard (2013) for dubitatives and
emotive doxastics, it is evident that they are the same.
3.4.
Negative Predicates
Lastly, EN is licensed in what I have so far called negative predicates in ClGr, French and
Hebrew. This class consists of a variety of predicates like prevent, hinder, refuse, deny, forbid,
retract an opinion, oppose, alter one’s plans, change one’s mind. As might be evident from
the name of the class these predicates are what Karttunen (1971) called ‘negative implicatives’.
The great variability of the meanings of these predicates, especially in ClGr, would prove any
attempt of exhaustively describing their semantics superfluous. For this reason, I will focus on
two aspects of their meaning that are relevant to EN-licensing.
All of these predicates presuppose a bipartition of the doxastic possibilities of the attitude
holder and a negative assertion against one of the two. In that sense, these predicates are the
negative equivalent of know in that they embed a set of propositions and that they entail the falsity of the embedded proposition. This proposal is actually in accordance with the experimental
findings of White et al. (2014) who identify that predicates with some kind of negative component form a semantic class and thus are distinct from the rest of attitude verbs. As White et
al. (2014) report the negative meaning had a very strong effect on participants’ semantic judgements, something which was not replicated in their syntactic judgements (the test was conducted
in English). The results of White et al. (2014) of the semantic clustering of predicates are shown
in Figure 1. In that respect, the examination of EN predicates is important for one more reason:
it provides further evidence that syntax and especially complementation is sensitive to semantic
cues.
A third model in which semantic and ‘weighed’ syntactic factors where jointly considered,
showed that there is a class of ‘negative verbs’ which consists of verbs like worry, doubt, forbid,
deny etc. (Figure 2). This class of ‘negative verbs’ overlaps with the class predicates that license
EN and especially the class of negative predicates.
3.5.
Interim Summary
In the previous sections, I examined the semantic properties of the predicates that license EN. To
summarize them, I showed that emotive doxastics and dubitatives involve an uncertainty con8
EN (complementizers) selection is not the only case where the polarity of the Matrix Predicate affects its
selectional properties (c.f. Adger & Quer 2001; Roussou 2010)
440
Maria-Margarita Makri
Figure 1: Semantic Clusters of Attitudes (White et al. 2014)
Figure 2: Model’s most likely guesses for similar worlds; shading represents classes found by
Walktrap algorithm (White et al. 2014) [dark grey class (left to right): worry, doubt, forget (the
more remote verb of the class), forbid (in the center), bother, hate, deny)]
Expletive Negation in Attitude Contexts
441
dition and a probability scale (emotive doxastics also a desirability scale). Interrogative predicates all select for questions but they can further be split into two subclasses. the first subclass is
Lahiri’s (2002) rogative predicates, which according to Uegaki involve a ‘non-triviality presupposition’ (which is actually similar to the uncertainty condition proposed by Anand & Hacquard
(2013) for emotive doxastics and dubitatives). The second subclass consists of negated responsive predicates which actually assert the same split of the attitude holder’s doxastic domain as
the split triggered by the presupposition in dubitatives, emotive doxastics and rogatives.
By this closer examination of the semantics of predicates that license EN I showed that EN is
licensed after predicates that presuppose or assert the existence of more than one live doxastic
possibility. The semantic properties of the aforementioned predicates are summarized in Table
5 below:
Table 5: The semantics of predicate classes without EN
Presupposition Modal
Selected
Scalar
Force
Complements
Assertion
Propositional
∀
sets of
No
Selecting
propositions
Predicates
< s, t >
e.g believe
Responsive
∀
<< s, t > t >
No
Predicates
functions from sets
e.g. know
of propositions
to truth values
Rogative
non-triviality
∃
<< s, t > t >
No
Predicates
presupp.
functions from sets
e.g. ask
uncertainty cond.
of propositions
to truth values
Emotive
uncertainty
∃
<< s, t > t >
Yes
Doxastics
condition
functions from sets (>des ), > prob
e.g. fear, hope
of propositions
Dubitatives
to truth values
e.g. doubt
Negative
non-triviality
<< s, t > t >
No
Predicates
presupp.
functions from sets
e.g. refuse
uncertainty cond.
of propositions
to truth values
Desideratives
<t>
Yes
e.g. want
propositions
(>des )
442
Maria-Margarita Makri
4.
The meaning of EN
In this section, I show that EN is not semantically vacuous, and I propose that EN is an epistemic
modal 9 that marks the lack of evidence on the part of the attitude holder about the probability
of the different alternatives, rendering them equally probable.
4.1.
4.1.1.
Asymmetries between Declarative and EN-complements
Epistemic Modals are not licensed in EN-complements
Expletive Negation is in complementary distribution with epistemic modals. As I propose in this
paper this is because EN is actually an epistemic modal therefore there cannot be two epistemic
modals in a single sentence. This is illustrated in the minimal pair in (28) below:
(28)
[MG]
a. Fovame pos mpori na fighun ta pedhia simera.
Fear
that might SBJ leave the children today
‘I fear that children might leave today.’
b. *Fovame mi-pos
mporei na fighun ta pedhia simera
Fear
lest.NEG-that might SBJ leave the children today
‘I fear lest children might leave today.’
4.1.2.
Answering Questions: No doxastic assertion in EN complements
In section 3.1 I followed Anand & Hacquard (2013) who suggested that in a sentence like [α
fears that ϕ] there is an assertion that in some of the doxastic worlds of the attitude holder ϕ is
true. A diagnostic indicating that an emotive doxastic with a that-complement asserts that the
proposition in the complement clause is more probable is indicated by the fact that the emotive
doxastic clause can function as an answer in a question asking ϕ like (29) below:
(29)
— Erhete o Nikos? — Fovame pos erhete. [MG]
Come the Nikos
Fear.1SG that comes.
‘— Is Nikos coming? — I fear that he is coming.’
The mininmally different reply with EN would nonetheless be an infelicitous answer:
(30)
— Erhete o Nikos? —# Fovame mi-pos
erhete.[MG]
Come the Nikos
Fear.1SG lest.NEG-that comes.
‘— Is Nikos coming? — I am afraid of the possibility that he might come.’
9
The semantic contribution as well as the syntactic position of EN could also indicate that EN is an evidential
marker. In this paper, I adopt Mathewson’s (2012) thesis that all evidentials contribute epistemic modal semantics,
and all epistemic modals contribute evidential semantics. I leave to future research an investigation of whether
a finer distinction between epistemic modals and evidentials is necessary to account for the whole array of EN
constructions.
443
Expletive Negation in Attitude Contexts
This indicates that the clause with EN in (30) does not have a doxastic assertion as (29) does.
Additionally, the answer in (30) does not seem to make any statement about the likelihood
of Nikos’s coming. Both possibilities (ϕ and ¬ϕ) are equally likely. The meaning of (30) is
formalized in (31) below:
(31)
J fearC NEG-that ϕKc,w,S,g =
λx: ϕ-verifiers in S 0 6= ∅ & ϕ-falsifiers in S 0 6= ∅
uncertainty cond.
If defined =1 iff
ϕ-verifiers <Desx,w ϕ-falsifiers ∧
preference assertion
ϕ-verifiers ≥Probx,w ϕ-falsifiers
preference assertion
where S 0 =DOXa,w and
0 000
ϕ-verifiers in S’= λS 00 .S 00 ⊂ S 0 & ∀S 000 ⊂ S 00 : [∀w0 ∈ S 000 [JϕKc,w ,S ,g = 1]] =
P ow(S 0 ∩ p)
ϕ-verifiers in S 0 = ¬ϕ-falsifiers in S 0
Why is the answer with EN infelicitous? The sentence with EN is less informative than the
sentence in (29) and thus it triggers a scalar implicature that the embedded proposition being
true is equally likely to being false (32):
(32)
Scalar Implicature of EN:
ϕ-verifiers =P robx,w ϕ-falsifiers.
This asymmetry holds also in Russian, where a sentence with EN like (33) is an infelicitous
answer to a question ‘Will he break the car?’
(33)
ja bojus’ kak by
on ne razbil
mašinu. [Russian]
I fear how MOD he NEG break.PST.PRF car.ACC
‘I fear that he might break the car.’ (N.Radkevic p.c.)
4.1.3.
Scope of Negation
Another piece of evidence that EN is not semantically vacuous comes from the scope of matrix
negation in sentences with emotive doxastics. Matrix negation can scope over the doxastic component of the emotive doxastic regardless of the presence or absence of EN ((35-a) and (34-a)
respectively). However, it can scope over the desirability scale only if the embedded clause has
EN; compare (34-b) to (35-b).
(34)
[MG]
a. Dhen fovame pos kseri tin alithia. Ime sighuros pos ehi mavra mesanihta.
NEG fear
that know the truth. Am sure
that has black midnight
‘I do not fear that he knows the truth. I am sure he is totally ignorant.’
b. Dhen fovame pos kseri tin alithia. #Vasika to
elpizo kiolas.
NEG fear
that know the truth. basically it.CL hope even
‘I do not fear that he knows the truth. In fact, I even hope it.’
(35)
[MG]
444
Maria-Margarita Makri
a.
b.
Dhen fovame mi-pos
kseri tin alithia. Ime sighuros pos ehi mavra
NEG fear
lest.NEG-that know the truth. Am sure
that has black
mesanihta.
midnight
‘I do not fear that he knows the truth. I am sure he is totally ignorant.’
Dhen fovame mi-pos
kseri tin alithia. Vasika to
elpizo kiolas.
NEG fear
lest.NEG-that know the truth basically it.CL hope even
‘I do not fear that he knows the truth. In fact, I even hope it.’
This asymmetry is predicted if we adopt the proposal that EN ‘acts’ on the probability scale introduced by the emotive doxastic. In the non-EN cases, matrix negation acts on the probability
assertion of the matrix predicates and reverses the ordering of the alternatives. In the EN examples (35-a)-(35-b), it may be inferred that ϕ-verifiers =P ROBa,w ϕ-falsifiers. In that case, the
matrix negation can target the second preference assertion and reverse the desirability ordering.
Table 6 below summarizes the properties of the predicates with complements containing/introduced
by EN.
Table 6: The semantics of predicate classes selecting EN
Presupposition Modal
Selected
Scalar
Force
Complements
Assertion
Negated
¬∀
<< s, t > t >
No
Responsive
functions from sets
Predicates
of propositions
e.g. know
to truth values
Rogative
non-triviality
∃
<< s, t > t >
No
Predicates
presupp.
functions from sets
e.g. ask
uncertainty cond.
of propositions
to truth values
Emotive
uncertainty
∃
<< s, t > t >
Yes
Doxastics
condition
functions from sets (>des ), ≥ prob
e.g. fear, hope
of propositions
Dubitatives
to truth values
e.g. doubt
Negative
non-triviality
<< s, t > t >
Yes
Predicates
presupp.
functions from sets
≥prob
e.g. refuse
uncertainty cond.
of propositions
to truth values
4.2.
Extensions: Greek counterfactuals
In this section, I proposed that EN acts on a likelihood scale introduced by the predicate and
asserts that the proposition being true is equally likely as being false. Further evidence about
the status of EN as an epistemic modal comes from another environment: counterfactuals.
Expletive Negation in Attitude Contexts
445
Iatridou (2000:fn.4) points that a future morpheme (or might) is necessary in the consequent
of a counterfactual in many (but not all) languages. On the other hand, Roussou (2000), while
discussing the structural position and meaning of tha (will) in Modern Greek, points out that tha
(will) is not a Tense particle but a modal. She argues for that based on the fact that tha (will) has
an exclusively epistemic reading if combined with a [+Past,+Perf] predicate and that it forms
a counterfactual if combined with a [+Past,-Perf] predicate. So taking into account Roussou’s
proposal we can revise Iatridou’s observation and state that an epistemic modal (will or might)
is necessary to form Greek (and English) counterfactuals. If the proposal that EN is an epistemic
modal is correct, then we expect EN to be licensed in the consequent of counterfactuals. Indeed,
this prediction is borne out:
(36)
An epine
afto to siropi, mi-pos ghinotan
kala. [MG]
If drank.PST.IMP. that the syrup, NEG-that be.PST.IMP well.
‘If s/he had drank that syrup, s/he might recover.’
Notice that (36) conveys that recovering would be possible if the subject had drunk that syrup.
The minimally different (37) indicates that in all the worlds that the subject drank the syrop they
recover.
(37)
An epine
afto to siropi, tha ghinotan
kala. [MG]
If drank.PST.IMP. that the syrup, will be.PST.IMP well.
‘If s/he had drank that syrup, s/he would recover.’
5.
Summary and Conclusions
In this paper, I examined the distribution and semantic contribution of EN in complements of
attitude predicates. Through a comparison of Modern Greek, Classical Greek, Spanish, French,
Latin, Russian and Hebrew sentential complements with EN I showed that EN does not correlate
with mood but with Tense and that it scopes above Tense. I also examined the predicates that
license EN and I showed that what they have in common is that they allow for a set of more
than one live doxastic possibility. As far as the distribution of EN is concerned, I also showed
that it is in complementary distribution with epistemic modals.
Based on asymmetries between complements with and without EN, I demonstrated that EN
is not semantically vacuous. I proposed that EN denotes the attitude holder’s lack of evidence
regarding the probability of the different alternatives, rendering them to equally probable. Based
on the meaning of EN as well as its distribution I proposed that EN is a weak epistemic modal.
446
Maria-Margarita Makri
Appendix
Table 7: Crosslinguistic distribution of EN
fear compl.
hope compl.
dubitatives
hinder, resist,
refuse, delay ∗∗
questions
rhetorical questions
interrogative
complements
exclamatives
free relatives
degree comparatives
metalinguistic
comparatives
before-clause
until-clause
unless-clause
without
∗
MGr.
+
+
+
-
ClGr.
+
+
+
+
Latin
+
+∗
[+∗ ]
Span.
+
+
-
French
+
+∗
+
Hebr.
+
Rus.
+
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
+
-
-
+
-
+
-
+
+
-
+
-
+
-
+
-
+
+
-
+
+
+
+
+
-
-
-
+
+
-
-
EN is licensed on condition that the matrix predicate is negated.
∗∗
The negative verbs that license EN may vary across languages. For presentational purposes, the categories of these verbs are collapsed in this table.
447
Expletive Negation in Attitude Contexts
Abbreviations
Cic. Fam.
Eur. Hel.
Plat. L.
Plat. Rep.
Th.
Xen. Anab.
M. T. Cicero, Epistulae ad Familiares
Euripides, Helena
Plato, Leges
Plato, Respublica
Thucydides
Xenophontis, Anabasis
Acknowledgements
This work was supported by the A.G.Leventis Scholarship Foundation and the Economic and
Social Research Council [grant number ES/J500215/1]. I would also like to thank Nina Radkevic for being my informant for the Russian data.
Maria-Margarita Makri
University of York
[email protected]
https://york.academia.edu/MariaMargaritaMakri
References
Abels, K. (2002). Expletive (?) Negation. Jindrich, T. (ed.), Proceedings of Formal Approaches to Slavic Linguistics, Michigan Slavic Publications, Bloomington, ID, vol. 10, pp. 1–20.
Abels, K. (2005). “ Expletive Negation ” in Russian : A Conspiracy Theory. Journal of Slavic Linguistics 13:1, pp.
5–74.
Adger, D. & J. Quer (2001). The Syntax and Semantics of Unselected Embedded Questions. Language 77:1, pp.
107–133.
Anand, P. & V. Hacquard (2013). Epistemics and attitudes. Semantics and Pragmatics 6:8, pp. 1–59.
Brown, S. & S. Franks (1995). Asymmetries in the Scope of Russian Negation. Journal of Slavic Linguistics 3:2,
pp. 239–287.
Cicero, M. (1901). Letters to his friends. Purser, L. C. (ed.), Epistulae ad Familiares, E Typographeo Clarendoniano
Oxford University Press, Oxonii, p. 562.
Espinal, M. T. (1992). Expletive Negation and Logical Absorption. The Linguistic Review 9:4, pp. 333–358.
Espinal, M. T. (2000). Expletive Negation, Negative Concord and Feature Checking. Catalan Working Paper in
Linguistics 8, pp. 47–69.
Espinal, M. T. (2007). Licensing Expletive Negation and Negative Concord in Catalan and Spanish. Floricic, F.
(ed.), La négation dans les langues Romanes, John Benjamins, Amsterdam, pp. 49 – 74.
Euripides (1913). Helene. Murray, G. (ed.), Euripidis Fabulae, Tomvs III, E Typographeo Clarendoniano Oxford
University Press, Oxonii.
Gaatone, D. (1971). Étude descriptive du système de la négation en français contemporain., vol. 114. Librairie
Droz, Genève.
Giannakidou, A. (2006). Only, emotive factive verbs, and the dual nature of polarity dependency. Language pp.
575–603.
448
Maria-Margarita Makri
Grano, T. A. (2012). Control and restructuring at the syntax-semantics interface. Phd, University of Chicago,
Illinois.
Homer (1920). Iliad. Allen, T. W. (ed.), Homeri opera, E Typographeo Clarendoniano Oxford University Press,
Oxonii.
Iatridou, S. (1993). On Nominative Case Assignment and a Few Related Things. MIT Working Papers in Linguistics
pp. 175–196.
Iatridou, S. (2000). The Grammatical Ingredients of Counterfactuality. Linguistic Inquiry 31:2, pp. 231–270.
Karttunen, L. (1971). Implicative Verbs. Language 47:2, pp. 340–358.
Lahiri, U. (2002). Questions and Answers in Embedded Contexts. Oxford University Press, Oxford.
Landau, I. (2002). ( Un ) interpretable Neg in Comp. Linguistic Inquiry 33:3, pp. 465–492.
Landau, I. (2004). The Scale of Finiteness and the Calculus of Control. Natural Language and Linguistic Theory
22:4, pp. 811–877.
Landau, I. (2009). Commentary on Sevdali’s paper. Halpert, C., J. Hartman & D. Hill (eds.), Proceedings of the
2007 Workshop in Greek Syntax and Semantics at MIT, MIT Working Paper in Linguistics, MIT, Cambridge,
MA, pp. 267–273.
Mathewson, L. (2012). Evidence about evidentials: Where fieldwork meets theory. Britta, S. & S. Featherston
(eds.), Empirical Approaches to Linguistic Theory: Studies of Meaning and Structure, Empirical Approaches to
Linguistic Theory: Studies of Meaning and Structure, Mouton de Gruyter, Berlin.
Muller, C. (1978). La négation explétive dans les constructions complétives. Langue Française 39, pp. 76–103.
Pesetsky, D. (1991). Zero syntax II: an essay on infinitives. Manuscript, MIT .
Picallo, C. M. (1984). The Infl Node and the Null Subject Parameter. Linguistic Inquiry 15:1, pp. 75–102.
Plato (1902). Leges. Burnet, J. (ed.), Platonis Opera, Tomvs V, E Typographeo Clarendoniano Oxford University
Press, Oxonii.
Plato (1903). Respublica. Burnet, J. (ed.), Platonis Opera, Tomvs IV, Tetralogia VIII, E Typographeo Clarendoniano
Oxford University Press, Oxonii.
Roussou, A. (2000). On the left periphery. Modal Particles and complementizers. Journal of Greek Linguistics 1,
pp. 65–94.
Roussou, A. (2010). Selecting Complementizers. Lingua 120:3, pp. 582–603.
Sevdali, C. (2006). Infinitival Clauses in Ancient Greek: Overt and Null Subjects, The role of Case and Focus.
Ph.D. thesis, University of Cambridge, Cambridge, UK.
Thucydides (1942). Thucydidis Historiae. Jones, H. S. (ed.), History of the Peloponesian War, E Typographeo
Clarendoniano Oxford University Press, Oxonii.
Uegaki, W. (2012). Content nouns and the semantics of question-embedding predicates. Aguilar-Guevara, A.,
A. Chernilovskaya & R. Nouwen (eds.), Proceedings of Sinn und Bedeutung 16, July, MIT Working Papers in
Linguistics, Cambridge, MA, pp. 1–14.
White, A. S., R. Dudley, V. Hacquard & J. Lidz (2014). Discovering Classes of Attitude Verbs Using Subcategorization Frame Distributions. Huang, H.-L., E. Poole & A. Rysling (eds.), Proceedings of NELS, Graduate
Linguistics Student Association, Amherst, MA, vol. 43.
Wouden, T. V. D. (1994). Polarity and ’Illogical Negation’. Kanazawa, M. & C. J. Pinon (eds.), Dymanics, Polarity,
and Quantification, CSLI, Stanford, pp. 17–45.
Xenophon (1904). Anabasis. Marchant, E. C. (ed.), Xenophontis Opera omnia, Volume III, Expeditio Cyri, Anabasis., E Typographeo Clarendoniano Oxford University Press, Oxonii.
Yalcin, S. (2007). Epistemic Modals. Mind 116:464, pp. 983–1026.
Yoon, S. (2011). ’Not’ in the Mood: The Syntax, Semantics, and Pragmatics of Evaluative Negation. Ph.D. thesis,
The University of Chicago, Chicago.
Yoon, S. (2012). Rhetorical comparatives : Polarity items , expletive negation , and subjunctive mood. Journal of
Pragmatics 43:7, pp. 2012–2033.