.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Experimental evidence for neg-raising in Slavic
Mojmír Dočekal & Jakub Dotlačil
Vienna, October 14, 2015
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Introduction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Goals of the talk & Intro
1. experimental data in support for the existence of neg-raising
(NR) in Slavic languages;
2. comparison with the Boškovič & Gajewski’s (2009) claim
about the non-existence of NR in Slavic languages;
3. a scalar approach to NR (extending Romoli’s 2012, 2013 work)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ neg-raising: interpretation of negation on the embedded
predicate
(1) a. John doesn’t think that it’s raining. ⇝
b. John thinks that it’s not raining.
▶ the neg-raising interpretation isn’t predicted by the standard
semantics of propositinal attitudes (from Hintikka 1969) –
believe universal quantifier over possible worlds, restricted to
some modal base
▶ 5 classes of NR predicates: intention (want, intend, …),
obligation (advise, should, …), perception (seem, appear, …),
opinion (know, believe, …), probability (probable, likely, …)
(2) believe (p)(a)(w) = ∀w′ ∈ M(w, a)[p(w′)]
a. ¬[∀w′ ∈ M(w, a)[p(w′)]]
b. ↛ [∀w′ ∈ M(w, a)[¬p(w′)]]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Experiment
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Introduction to the debate
▶ NR generally received a lot of attention (Gajewski 2005, 2011;
Homer 2008a, 2008b, 2011 a.o.) but in the area of Slavic
languages it is relatively understudied phenomenon:
▶ there is only one article (Boškovič & Gajewski 2009), which
claims the non-existence of neg-raising in SL
▶ Dočekal (2014): limited existence of NR in SL
▶ B&G base their arguments on the lack of strict NPI licensing
in the embedded clauses of Slavic NR predicates.
▶ B&G used only one test and only for the class IV (opinion) of
NR-predicates.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(3) *Ivan ne vjeruje [da ju je Marija posjetila najmanje dvije
godine.]
‘Ivan does not believe that Mary has visited her in at least
two years.’
(4) a. *Ivan ne veril, čto Marija uedet až do zavtrašnego dnja.
Russian
b. *Jan nie wierzył, że Maria wyjedzie aż do jutra. Polish
c. *Ivan nije vjerovao da će Marija otići sve do sutra. SC
d. Az ne vjarvam/*kazah če Meri ja e poseštavala pone
ot dve godini. Bulgarian
e. *Jan nevěří, že Marie ji navštívila nejméně dva roky.
Czech
f. *Janez ne verjame, da jo je Marija obiskala že najmanj
dve leti. Slovenian
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Experiment
▶ The experiment consisted of two parts:
▶ an acceptability judgement task
▶ inference task
▶ acceptability task: participants had to judge the acceptability
of sentences with strict NPIs, ani ‘not even’ and až do ‘until’,
using the 5-point Likert scale (5=best, 1=worst)
▶ strict NPIs can be licensed
1. in negative clauses
2. in clauses embedded under negated NRs (Horn, 1989)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ 5 environments
1. a positive sentence (A)
2. a negative sentence (B)
3. a clause embedded under negated NR predicates of intention
and judgement/obligation (e.g. want, advise) (C)
4. a clause embedded under negated NR predicates of opinion
(believe) (D)
5. non-NR predicates (E)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(5) a. Ztratila
Lost
se
SE
ani
not-even
jedna
one
ovce.
sheep
‘A single sheep is missing.’
b. Neztratila
neg-lost
se
SE
ani
not-even
jedna
one
ovce.
sheep
‘Not a single sheep is missing.’
c. Nový bača v Tatrách nechce, aby se ztratila ani jedna
ovce.
new shepherd in Tatras neg-wants C SE lost not-even
one sheep.
d. Nový bača v Tatrách si nemyslí, že se ztratila ani
jedna ovce.
new shepherd in Tatras SI neg-think C SE lost
not-even one sheep
e. Nový bača v Tatrách neříká, že se ztratila ani jedna
ovce.
new shepherd in Tatras neg-say C SE lost not-even one
sheep
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Items
▶ in all environments indicative/subjunctive complementizers
were tested without any impact on grammaticality
▶ 20 items tested with ani jeden ‘not-even one’ strict NPIs, 20
items tested with až do ‘until’ strict NPIs
▶ intention/judgement/obligation verbs (C): chtít ‘want’, hodlat
dopustit ‘allow’, mít v úmyslu ‘have intention’, přát si ‘wish’,
vyžadovat ‘require’, potřebovat ‘need’, usilovat ‘strive’, radit
‘advice’, doporučovat ‘recommend’, navrhovat ‘propose’, …
▶ opinion verbs (D): myslet ‘think’, věřit ‘believe’, předpokládat
‘suppose’, představovat ‘imagine’, očekávat ‘expect’, uvažovat
o ‘speculate’, domnívat se ‘assume’, soudit ‘judge’, spoléhat
se ‘count on’, …
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ non-NR verbs (E): říkat ‘say’, slyšet ‘hear’, tvrdit ‘assert’,
rozhlašovat ‘rumour’, naznačit ‘indicate’, prozradit ‘reveal’,
sdělit ‘tell’, zavolat ‘call’, napsat ‘write’, způsobit ‘cause’,
vyrozumět ‘inform’, nutit ‘force’, číst ‘read’, chápat
‘understand’
▶ in most cases the sentences were conceived as ‘small scenarios’
to help subjects construct plausible context, pseudo-Czech:
(6) a. New librarian neg-want/believe/say
that-IND/that-SUBJ strict-NPI book get lost.
b. The writer of detective novels neg-need/think/reveal
that he is praised by strict-NPI journalist.
c. Karel’s doctor neg-recommend/believe/say that Karel
miss strict-NPI cure.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(7) a. Mother neg-wants/imagine/hear that-IND/that-SUBJ
father arrives until Christmas.
b. Strict teacher neg-intend/guess/indicate that he will
release students until twilight.
c. Klara’s parents neg-allow/think/tell that he will marry
her until summer.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ the second part:
1. neg-raising is intuitively valid (¬NR[P] ⇝ NR[¬P])
(8) a. John doesn’t think that it’s raining. ¬NR[P] ⇝
b. John thinks that it’s not raining. NR[¬P]
2. cyclic neg-raising is valid (¬NR1[NR2[P]] ⇝ NR1[NR2[¬P]])
(9) a. I don’t believe Bill wanted Harry to die.
¬NR1[NR2[P]] ⇝
b. I believe Bill wanted Harry not to die. NR1[NR2[¬P]]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3. existential wide scope is valid
(¬∀xNR1[NR2[P]] ⇝ ∃xNR1[NR2[¬P]])
(10) a. Not every student thinks that Mary passed.
¬∀xNR1[NR2[P]] ⇝
b. There are some students who think that Mary didn’t
pass. ∃xNR1[NR2[¬P]]
▶ only NRs of intention were used in this part of the experiment
▶ examples items (pseudo-Czech):
(11) a. Old duke would neg-be glad if the best wine would get
lost from the cellar. ⇝
b. Old duke would be glad if the best wine would neg-get
lost from the cellar.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(12) a. Servants neg-believe that old duke would be glad if the
best wine would get lost from the cellar. ⇝
b. Servants believe that old duke would be glad if the
best wine would neg-get lost from the cellar.
(13) a. Not all servants believe that old duke would be glad if
the best wine would get lost from the cellar. ⇝
b. Some servants believe that old duke would be glad if
the best wine would neg-get lost from the cellar.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ 40 exp. items in part 1 and 20 exp. items in part 2: 260
tested sentences
▶ each part 30 fillers, 60 Czech native speakers, ≈ 1 hour
▶ the experiment online in Ibex: link
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Results of acceptability task
▶ all participants passed control fillers (uncontroversially
grammatical/ungrammatical)
▶ acceptability task: modeled by mixed-effects ordered probit
regression
▶ Condition C as the reference level
▶ negated sentences, Condition B, were judged as better than
NRs (β = 1.84, z = 23, p < .001)
▶ positive sentences, Condition A, were judged as worse than
NRs (β = −1.1, z = −15, p < .001)
▶ sentences with negated non-NR (E) predicates worse than any
NR (β = −0.65, z = −9, p < .001)
▶ evidence for:
1. treating ani ‘not even’ and až do ‘until’ as a strict NPIs
(otherwise, Condition E should be acceptable, contrary to the
facts)
2. Czech has a class of NR verbs.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0
100
200
300
1 2 3 4 5
Acceptability
count
Condition
Condition A
Condition B
Condition C
Condition D
Condition E
Acceptability task
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Results of inference task
▶ the analysis of the inference task: using mixed-effects
logistic regression (with 1=inference follows, 0=inference does
not follow)
▶ one fixed factor: Condition I as the reference level
▶ condition I was significantly higher than a chance, prob=0.5,
“inference follows” was preferred
▶ Condition II and III had a significantly smaller preference for 1,
in fact, they did not significantly differ from prob=0.5
▶ unexpected in previous accounts of NRs (all three inference
tasks should be possible) but
▶ the difference is likely a consequence of the higher complexity
of Condition II and III (i.e., more clauses, more complex
sentences): Condition II & III 2-3 times longer reaction time
than Condition I
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Analysis
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Scalar approach to NR
Roadmap:
▶ explain the acceptability and the inference tasks
▶ especially w.r.t. strict NPI licensing:
0. against syntactic (reconstruction) treatment – see (14) (type
E) were tested both with indicative/subjunct. embedded CP
and in both environments the strict NPIs were ungrammatical
1. why are NR (C and D) better than non-NR (E)
2. but worse than simple negated sentences (B)
(14) Petr neřekl, *že/*aby ani jeden student prošel.
Petr didn’t-say *that_IND/*C_SUBJ not-even one student
passed.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ the scalar approach to NR (originally Horn, today esp. Romoli
2012,2013)
▶ NR predicates (beside the assertion – (15-b)) contribute the
excluded middle (EM) to the semantic composition ((15-a)):
▶ alternatives grow through the semantic composition of the
sentence until they hit the proposition taking EXH – (16-a)
▶ it applies to the proposition and negates the excludable –
(16-b) – alternatives
(15) a. Alt(NR) = {λpλx.□x[p], λpλx.[□x[p] ∨ □x[¬p]]}
b. P = λpλx.□x[p]
(16) a. EXH(Alt(p))(p)(w) = p(w) ∧ ∀q ∈
Excl(p, Alt(p))[¬q(w)]
b. Excl(p, Alt(p)) = {q ∈ Alt(p) : λw[¬q(w)] ∩ p ̸= ∅}
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ application to the Czech NR (C) predicate chtít ‘want’ in
(17)– assertion
▶ (18-a) the alternatives, (18-b) the exhaustification
▶ such exhaustification proceeds only if it is an answer (to and
implicit or explicit) QUD like What does the new shepherd
want his sheep to do?
(17) ‘New shepherd in Tatra mountains doesn’t want even one
sheep to be missing.’ ¬wants[p] .
(18) a.
Alt(¬wants[p]) = {¬wants[p], ¬(wants[p]∨wants[¬p])}
b. EXH (¬wants[p]) =
¬wants[p] ∧ ¬¬(wants[p] ∨ wants[¬p]) |= wants[¬p]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ explanation of the inference task: neg-raisers (C,D) do have
the excluded middle alternative unlike non-NR (E)
▶ and it explains the acceptability distinction between
neg-raisers (C,D) and non-NR (E)
▶ the scope of negation of neg-raisers is low → strict NPIs are
licensed in the embedded clause
▶ with non-NR predicates like be certain, the alternative is
existential (or missing – say, …) → explanation of the
acceptability difference between NR (C,D) and non-NR (E)
▶ the existential alternative blocks the DE environment (parallel
to every: *I doubt that every student ever solved that
problem)
(19) a. John isn’t certain that Mary will arrive.
b. ⇝ It’s possible for John that Mary will arrive.
c. ¬□j[p] ∧ ⋄j[p]
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
NR suspension – NR vs. simple negated sentences
▶ the limited NR inferences ability of SL as a result of NR
inferences suspension (for English observed by Gajewski 2007:
John DOESN’T think that Fred left. He isn’t sure)
▶ for the suspension in SL: the verbal negation is interpreted
(systematically ambiguous between the propositional and
constituent scope) with constituent scope
▶ the interpretation marks the sentence: an answer to a
different QUD → the EM alternative becomes irrelevant
(20) Petr nepřečetl tři Kunderovy romány, on je zhltl s vášní.
‘Petr didn’t READ three Kundera’s novels, he devoured
them with passion.’
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ scalar implicatures are context dependent – only relevant
alternatives produce scalar implicatures and the relevance of
the alternatives depends on QUDs
▶ e.g. exact scalar implicature for numerals is valid only for
QUD like (21-a) but not for (21-c)
▶ So an answer to QUD like (22-a) creates a partition like (22-b)
▶ the excluded middle (c3) is relevant, yielding the NR inference
(21) a. How many beers did John drink?
b. John drank 3 beers. ≈ #x = 3
c. Who did drink 3 beers?
d. John drank 3 beers. ≈ #x ≥ 3
(22) a. What does the new shepherd want his sheep to do?
b. Q = {c1 = wants[p], c2 = wants[¬p], c3 =
¬(wants[p] ∨ wants[¬p])}
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ but different stress e.g. on think in (23) leads to a QUD like
(24-a)
▶ it’s a predicate verum focus negation
▶ which create partitions like (24-b)
▶ The excluded middle proposition is not relevant in such
partitioning because it doesn’t distinguish beween the
partitions but only within c2.
▶ Therefore the NR inference is suspended.
▶ simple negated sentences (B) don’t need implicature
calculation for strict NPI licensing → the difference between B
and C,D
(23) Nový bača si NEMYSLÍ, že se ztratila *ani jedna ovce.
New shepherd DOESN’T think that even-one one sheep
dissapeared.
(24) a. Does the new shepherd think that one sheep
dissapeared?
b. Q = {c1 = thinks[p], c2 = ¬thinks[p]}
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Quick comparison with Gajewski’s approach
▶ basic ingredients of Gajewski’s theory:
1. excluded middle is a presupposition:
(25) P = λpλx : [P(p)(x) ∨ P(¬p)(x)].P(p)(x)
2. NR are soft presupposition triggers ↔ context dependent:
suspension by ignorance:
(26) I don’t know whether Peter enrolled in the race, but if he
won, he is in the bar now.
▶ but soft presupposition triggers need contextual information
to be suspended
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
▶ if excluded middle was a presupposition, it should project; but
it does not
(27) a. Peter thinks that the sheep is missing.
b. Peter has an opinion as to whether the sheep is
missing.
c. Peter doesn’t think that the sheep is missing. (⇝
projects (27-b))
d. Does Peter think that the sheep is missing? (̸⇝
projects (27-b))
e. If Peter thinks that the sheep is missing, we should call
police. (̸⇝ projects (27-b))
f. Perhaps Peter thinks that the sheep is missing. (̸⇝
projects (27-b))
▶ in the same environments the exhaustification of scalar
implicatures follows similar pattern:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(28) a. Peter didn’t find every sheep. (⇝ found some)
b. Did Petr find every sheep? (̸⇝ found some)
c. If Peter found every sheep, we’ll celebrate. (̸⇝ found
some)
d. Perhaps Peter found every sheep. (̸⇝ found some)
▶ maybe the presuppositions are cancelled by indicative like
Homer’s example: the indicative triggers the presupposition
that the speaker holds true the complement clause → NPIs
mai ‘ever’ are un-licensed:
(29) Context: Maria has visited Paris several times.
a. *Gianni
Gianni
non
NEG
pensa
thinks
che
that
Maria
Maria
e
be.IND
mai
ever
andata
gone
a
to
Parigi
Paris
‘Gianni doesn’t think that Maria has ever been to
Paris.’
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(30) Gianni
Gianni
non
NEG
pensa
thinks
che
that
Maria
Maria
sia
be.SUBJ
mai
ever
andata
gone
a
to
Parigi
Paris
‘Gianni doesn’t think that Maria has ever been to Paris.’
▶ highly improbable for Czech data: the same items with NR C
in subjunctive and NR D in indicative were judged similarly,
e.g.
(31) a. Nový vedoucí katedry nehodlá dopustit, aby studenti
vynechali ani jednu přednášku.
‘New chief of the department doesn’t want C-SUBJ
students miss even one lecture.’ (SUBJ)
b. Nový vedoucí katedry nevěří, že studenti vynechali ani
jednu výběrovou přednášku.
‘New chief of the department doesn’t believe that
students missed even one class.’ (IND)’
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
References
[1] Abusch, Dorit. “Lexical Alternatives as a Source of Pragmatic Presuppositions.” Semantics and Linguistic
Theory (SALT) 12 (2002). http://semanticsonline.org/pragmatics/protected/abusch.pdf.
[2] Bartsch, Renate. “Negative Transportation” Gibt Es Nicht.” Linguistische Berichte 27, no. 7 (1973).
[3] Bošković, Željko, Gajewski, Jon. “Semantic Correlates of the NP/DP Parameter.” In Proceedings of NELS 39,
2009.
[4] Chierchia, Gennaro. Logic in Grammar: Polarity, Free Choice, and Intervention. Vol. 2. Oxford University Press,
2013.
[5] Gajewski, Jon R. “Licensing Strong Npis.” Natural Language Semantics 19, no. 2 (2011): 109–48.
[6] Gajewski, Jon Robert. “Neg-Raising: Polarity and Presupposition.” Massachusetts Institute of Technology, 2005.
http://gajewski.uconn.edu/papers/thesis.pdf.
[7] Hintikka, Jaakko. “Semantics for Propositional Attitudes.” In Philosophical Logic, 21–45. Springer, 1969.
[8] Homer, Vincent. “Disruption of NPI Licensing: The Case of Presuppositions.” In Proceedings of SALT,
18:429–46, 2008.
[9] Horn, Laurence. A Natural History of Negation. Chicago: University Of Chicago Press, 1989.
[10] Romoli, Jacopo. A scalar implicature-based approach to neg-raising. Linguistics and Philosophy 36, 2013:
291-353.