Chapter 42
Negation and negative polarity items in
Slavic
Mojmír Dočekal
Masaryk University
This chapter describes the Slavic expressions dependent on negation. It summarizes
three approaches to such expressions which appeared in the history of formal
linguistic Slavistics. First, syntactic analysis of negative dependent expressions
(see Progovac 1994 a.o.) is discussed, next the semantic approach of Błasczak 2001,
and last, the pragmatic stance of Crnič 2011 is introduced. The chapter shows how
Slavic data of this kind were important from the point of view of formal linguistics,
and their description mirrored the changes in the general linguistic theories.
Keywords: Negative Polarity Items, approaches to NPIs
1 Introduction
Negation and negative dependent or sensitive expressions belong to the core
grammatical constructions of natural languages. There are no known languages
that would lack expressive means to communicate negation (see Bernini & Ramat
2012 a.o., where the fact of negation being present in all natural languages is
called a “pragmatic universal”). Slavic languages are not different, of course. So
there is a plethora of phenomena (many of them already described by formal linguists)
concerning negation which are perfect candidates for inclusion into this
chapter. But of course, space and other limitations force me to narrow down the
domain described in the present chapter (but see Horn 1989 for the book aiming
at the description of all linguistic things connected to negation). So I will limit
myself to discussing various types of Slavic negative dependent expressions. And
Mojmír Dočekal. Negation and negative polarity items in Slavic. To appear in: Change Volume Editor & in
localcommands.tex Change volume title in localcommands.tex Berlin: Language Science Press. [preliminary
page numbering]
Mojmír Dočekal
I will only mention many valuable contributions which would surely deserve appropriate
space but which are beyond the horizon of negative dependent expressions.
Nevertheless, see §3 for at least some pointers to literature discussing the
topics which do not appear in this chapter.
Criteria for delimiting this chapter to Slavic negative dependent expressions
were as follows. I wanted to summarize such areas of Slavic negation where (i)
Slavic linguistics contributed to general linguistics theories (even if now the particular
approach seems to be more of a historical value) and (ii) where Slavic
data, and more specifically expressions interacting with negation or which in
some respect are negative themselves offer insights valuable from a typological
perspective. And it seems more than reasonable that Slavic negative dependent
expressions (especially negative polarity items) pass both criteria.
First, the obvious thing to note is that Slavic languages do offer much more
expressivity than the (still) most studied language, English. Consider the all-time
favorite English polarity expression any (other types of English polarity expressions
like even one will be discussed in the section §2.3). In the following English
example, any is licensed by clause-mate negation (1a), adversative predicate (1b)
(which I, for expository reasons, treat as containing negation but see §2.2.1 for
more details of adversative predicates) and, finally, by a possibility modal (1c). In
(1a) and (1b) any is usually considered as a negative polarity item (NPI), and
the example like (1c) as free choice item (FCI). Although it seems typologically
more frequent (than the opposite pattern) that one item (like English any) acts
both as the NPI and the FCI, it is not a rule.1 For the sake of exposition, let us
assume that at least in English, any is an item with roughly existential semantics,
and such a polarity item can be described by a unified theory.2 Be it as it
may, NPI and FCI are united by a licensing requirement. In all three cases, the
scope of the existential NPI/FCI must be narrow relative to the scope of negation
or existential modal. And, of course, the use of NPIs/FCIs leads to ungrammaticality
without embedding in an appropriate environment: (2). Negative polarity
items are the main topic of this chapter, and as an avid reader most probably
knows, they are the class of negative dependent expressions which are usually
discussed (at least by semanticists) the most, since their distribution can be very
1
See Horn (1972) and Haspelmath (1997) for an early observation and a typologically broader
picture, respectively. As one of the anonymous reviewers correctly points out, languages like
Latin, Romanian, and Russian are clear examples of distinguishing FCIs from NPIs series.
2
Nowadays, this seems like a standard approach, following Carlson (1980) and then Kadmon
& Landman (1993), but at least for FCIs a universal quantifier analysis was offered by Dayal
(1998); a.o. Thanks to all three reviewers for reminding me of the importance of any in research
history.
ii
42 Negation and negative polarity items in Slavic
well-expressed in terms of entailment and other related properties of the sentences
where they occur. A more formal characterization will be introduced in
§2.1.
(1) a. Peter didn’t visit anyone.
i. ¬ > ∃
ii. # ∃ > ¬
b. John refused any help.
i. ¬accept > ∃
ii. # ∃ > ¬accept
c. You can take any card from the deck.
i. ♢ > ∃
ii. # ∃ > ♢
(2) #Peter visited any student.
Compare this with the Czech translation of the three environments showing the
grammatical but also ungrammatical possibilities of three types of negative dependent
expressions (which still is a subset of possible expressions in various
ways interacting with negation, see the section §2.3 for more details).
(3) a. Petr
Petr
nenavštívil
neg.visited
{ani
even
jednoho
one
/ #byť
just
jediného
one
/ #jakéhokoliv}
any
studenta.
student.
(Intended:) ‘Peter didn’t visit {even one / a single / any} student.’
b. Petr
Petr
odmítl
refused
{jakoukoliv
any
pomoc
help
/ #ani
even
jednoho
one
studenta
student
/ pohnout
move
se
refl
byť
just
o
by
jediný
one
centimetr}.
centimeter
(Intended:) ‘Petr refused {any help / any student / to budge an inch}.’
c. Můžeš
can.2sg
si
refl
vzít
take
{jakoukoliv
any
kartu
card
/ #byť
just
jedinou
one
kartu
card
/ #ani
even
jednu
one
kartu}.
card
‘You can take {any card / even a single / not even one card}.’
(Czech)
iii
Mojmír Dočekal
The negative dependent expressions in Slavic come in many forms. Some seem
to be vanilla NPIs but are limited to clause-mate negation (like ani in (3)). Others
bear presuppositions of unlikelihood (as byť jediný in (3), which resembles English
NPIs like even one), and yet others seem to prefer free choice contexts (as
jakýkoliv from (3)). The variation of Czech (and generally Slavic NPIs) is reflected
in the patterns of their distribution. Or, to put it the other way round: the flavors
of NPIs determine in which contexts we can find them, and the varieties of Slavic
polarity items carve up the landscape in a different way from the more usually
discussed Romance or Greek polarity expression (see Laka 1990; Giannakidou
1997; a.o.).
When looking at the history of approaches to Slavic negative dependent expressions,
we can clearly distinguish three types of theories: (i) syntactic theories
of NPIs (discussed in §2.1), (ii) semantic theories (§2.2), (iii) pragmatic theories
(§2.3). The particular details of each approach will be discussed in the respective
sections. But what all of them are united by is the observation about the richness
of Slavic expressions, which offers a much clearer picture than the sparse English
any. Moreover, the various subsystems of negative dependent expressions
create beautifully intricate territories where individual subtypes compete for insertion.
Consequently, their accurate description brings new data not only for
the polarity agenda but also for our understanding of concurrence in grammar.
In summary, various approaches to Slavic NPIs will be the main topics of the
current chapter, simply because they were at the center of attention in the last
30 years of formal Slavic linguistics.
But before we begin, let us quickly address a closely related phenomenon
which I will describe just briefly in the current article (see Dočekal 2019 for full description).
Next, to semantically dependent expressions, NPIs exemplified above,
there are (in all Slavic languages) negative dependent expressions, so-called negwords.
Neg-words are at least in current standard semantic theories described
as syntactically, not semantically dependent. The theoretical distinction (negwords:
syntax, NPIs: semantics) is operationalized via the following criterion
from Giannakidou & Zeijlstra (2017):
(4) X qualifies as an neg-word iff:
a. X can be used with structures with sentential negation or other X with
meaning equivalent to ¬;
b. X provides a negative fragment answer.
Especially the second part of the criterion is the most reliable diagnostic to distinguish
NPIs from neg-words because neither of the 3 types of NPIs in (3) can be
iv
42 Negation and negative polarity items in Slavic
used as a fragmentary answer to a question. See (5B1), where all three NPIs are
unacceptable, contrasting with the acceptability of the neg-words in the same
context, see (5B2).
(5) A: Kdo
who
byl
was
dneska
today
večer
evening
na
on
náměstí?
square?
‘Who was today in the evening on the square?’
B1: # {Ani
even
jeden
one
člověk.
human
/ Byť
just
jediný
one
člověk.
human
/ Kdokoliv.}
anyone
NPIs
Intended: ‘{Not even one man / not a single man / Anyone }.’
B2: Ni-kdo.
neg-body
neg-word
‘Nobody.’
It is important to notice that both the syntactic approach of Progovac (1994) and
the semantic approach of Błasczak (2001) discussed in §2.1 and §2.2 respectively
do not accept the distinction between neg-words and NPIs as delineated above.
And it is one of the problems which caused both empirical and theoretical flaws
in both approaches. But to be fair, both approaches were published before there
was a solid framework for the treatment of neg-words in negative concord languages
(Zeijlstra 2004). Moreover, either syntactic treatment of all NPIs as in
Progovac (1993) or fully semantic treatment of neg-words (and their sub-kinds
as neg-pronouns) as in Błasczak (2001) were compatible with the linguistic state
of art valid in the time of their publication.
During this chapter, I will use the following terminology, which is based partially
on Błasczak (2001) and exemplified with Czech examples in (6).
(6) a. neg-words
nikdo ‘nobody’ (glossed as neg-body), nic ‘nothing’ (neg-thing), nikde
‘nowhere’ (neg-place)
b. a-NPIs
ani jeden ‘not even one’ (NPI-one), ani jednou ‘not even once’
(NPI-once)
c. k-pronouns
kdokoliv ‘anyone’, kdekoliv ‘anywhere’, cokoliv ‘anything’
d. b-NPIs
byť jediný (glossed depending on context ‘even one’ or ‘at least one’),
byť jednou (‘even/at least once’)
e. regular (weak) NPIs
as Czech vůbec ‘at all’
v
Mojmír Dočekal
2 Three types of approaches to NPIs (from the Slavic
perspective)
2.1 Syntactic approach of Progovac
Let us first start with the syntactic approach of Progovac (1994) (and some elaborations
like Progovac 2005 or for a compatible approach to PPIs see Szabolcsi
2004). Progovac’s (1993) approach is partially inspired by the influential work
of Linebarger (1987) where some serious problems of early semantic approaches
(like Fauconnier 1978; 1980; Ladusaw 1979) to NPIs were recognized. The syntactic
theory of NPI licensing claims that the domains and licensors of NPIs should
be described by syntactic rules: either via transformational like in the earlier approaches
of Klima (1964) or via binding type as in Progovac’s work.
Progovac’s work should be understood with the 80’s semantic theory of NPIs
as a background. In the 80s, the semantic theory of NPIs, in a nutshell, claimed
that NPIs occur in downward entailing contexts. Downward entailing contexts
are such where speakers can reason from sets to subsets (see §2.2 for more
details). Let us see this more clearly on an example: consider an example of a
downward entailing context in (7b) as opposed to an upward entailing context
in (7c) (and note the subset–superset relationship in (7a)). In (7a), the quantifier
every allows one to draw an inference about the truth of the predicate to all subsets
of dogs. The inference is valid only for the first argument of every (natural
language quantifiers have two arguments like their formalization in predicate
logic with 𝜆-abstraction, 𝜆𝑃𝜆𝑄∀𝑥[𝑃(𝑥) → 𝑄(𝑥)]), since from the truth of Every
dog barks it doesn’t follow that Every dog barks loudly. On the other hand, a quantifier
like some in (7c) allows inferences about the truth of the predicate from subsets
(e.g., dachshund) to supersets (dogs), in both its arguments. The quantifier
every is then called downward entailing (in its first argument) and some upward
entailing (in both arguments). The quantifier every is downward entailing, and
because of that (in semantic theories of NPIs), it licenses NPIs: (7d) while upward
entailing quantifier some does not: (7e). Quantifiers like every and some are then
called downward and upward monotonic, respectively, since they are monotonic
in the same sense as monotonically increasing or decreasing functions in math-
ematics.
(7) a. dachshund ⊆ dog
b. Every dog barks. ⊧ Every dachshund barks.
c. Some dogs barks. ̸⊧ Some dachshunds bark.
d. Every dog which won any medal barked.
vi
42 Negation and negative polarity items in Slavic
e. *Some dogs which won any medal barked.
The semantic approach works in simple cases like (7d), but in its basic form, it has
many drawbacks. One of the first influential critiques is Linebarger (1987), who
noticed some serious problems with the semantic approach. First, Linebarger
correctly pointed out that many NPI licensing contexts are not, in fact, downward
entailing. Her examples involved conditionals and adversative predicates;
see §2.2.1 for more details and refinement of the semantic approach, which can
deal with the majority of such problems. And her second line of attack concerned
locality restrictions between the licensors and their NPIs. One example like that
is in (8a). The example cannot have the reading (8b): it is not true that Mary
wears earrings to every party (to some yes but not to all) but only has the reading
in (8c): there are no such earrings that Mary wears to every party. According
to Linebarger, this shows that some scopal (maybe syntactic) locality constraint
(requiring the NPI to be in the immediate scope of its licensor) is needed. This
seems to be a valid objection that cannot be explained by some a theory of NPI
licensing depending only on the monotonicity of its context. But see Chierchia
(2013) for a modern pragmatic/semantic framework that covers even these supposedly
syntactic locality problems.
(8) a. Mary didn’t wear any earrings to every party.
b. #¬∀𝑦[party(𝑦) → ∃𝑥[earring(𝑥) ∧ wear at(Mary, 𝑥, 𝑦)]]
c. ¬∃𝑥[earrings(𝑥) ∧ ∀𝑦[party(𝑦) → wear at(Mary, 𝑥, 𝑦)]]
It is appropriate to understand Progovac (1993) on the background of (in its time
very important and influential) Linebarger’s (1987) theory. Progovac bases her
analysis in syntax, trying to explain NPI properties from the principles of binding
theory but applied to various types of NPIs instead of the standard pronouns
and anaphors. Next, Progovac (1993) was very influential in its time and still is
one of few systematic analyses of Slavic NPI data. Moreover, the whole analysis
does not cover only Slavic languages but also provides a general framework
designed for a typologically diverse sample of languages. But let us begin with
empirical claims. Progovac distinguishes three classes of polarity-sensitive expressions
and claims that there are two NPI licensors: negation and an operator
in Comp.3 The polarity operator is licensed either by syntactic movement or
3
With Comp I refer to complementizers, the functional category exemplified with English that
or whether. Complementizers are heads of embedded clauses (CPs).
vii
Mojmír Dočekal
by DE predicate/operator from a c-commanding clause.4 Progovac brings data
from Bosnian/Croatian/Serbian (BCS). Nevertheless, her analysis is meant for all
Slavic languages. And it is true that in the majority of Slavic languages, similar
data can be found (see §2.2 for analogous data from Polish, described in a different
framework, though). According to Progovac (from the NPI point of view), we
should distinguish the following three classes of expressions:
1. BCS pronouns which begin with the negative prefix ni, e.g., ni-(t)-ko ‘no
one’, ništa ‘nothing’, nikud ‘nowhere’ are NPIs, in our terminology negpronouns.
Progovac, following Laka (1990), claimed that neg-words in natural
languages with negative concord should be analyzed as NPIs, which
is one of the possible positions but definitely not the only one (see §1).
Neg-pronouns are claimed to be anaphoric, subject to Principle A: they
have to be bound by negation in their governing category. Data supporting
anaphoric nature of NPIs are in (9),(10). Grammaticality/ungrammaticality
of examples like (9)/(10) is then simply accounted as syntax locality licensing
or, more specifically, as a failure to find an antecedent that would be
local enough.5
(9) Milan
Milan
*(ne)
not
vidi
sees
ni-šta.
neg-thing
‘Milan cannot see anything.’ (BCS; Progovac 1994: 41)
(10) * Milan
Milan
ne
not
tvrdi
claims
[da
that
Marija
Mary
poznaje
knows
ni(t)-koga].
no-one.acc
Intended: ‘Milan doesn’t claim that Mary knows anyone.’
(BCS; Progovac 1994: 41)
2. Progovac further claims that another class of BCS pronouns, beginning
with the prefix i, e.g., i-(t)-ko ‘anyone’, išta ‘anything’, ikud ‘anywhere’,
and the like are polarity items as well, in our terminology k-pronouns. Kpronouns
are, according to Progovac, anaphoric pronominals. They are
subject to Principle B: they need to be free in the local clause but bound in
4
What falls under syntactic movement is the licencing of NPIs in questions, conditionals, and
some other constructions where inversion or movement can license the existence of the polarity
operator in Comp.
5
In Slavic languages, the local domain for the majority of anaphors (governed by Principle A)
is a tensed clause, TP. See Büring 2005 for a good textbook on Binding Theory, which also
discusses cross-linguistic data.
viii
42 Negation and negative polarity items in Slavic
the sentence (licensed from a higher position) – after Progovac (1994: 64).
Environments allowing only k-pronouns, not neg-words, are (according
to Progovac): questions, conditionals, adversative predicates, restrictive
clauses of universal quantifiers, and superordinate negation. For all these
environments, she offers a syntactic explanation of the possible occurrence
of k-pronouns (and the impossibility of neg-words): operator in Comp (in
all five environments) is supposed to license k-pronouns but is too nonlocal
for neg-words. Some examples (after Progovac 1994) are shown in
(11), the explanation is the same for them all (but see Progovac 1994: 64 for
a full range of examples).
(11) a. Da
that
li
Q
Milan
Milan
voli
loves
i(t)ko-ga
anyone.acc
/ *ni(t)-koga?
no-one.acc
(Intended:) ‘Does Milan love anyone?’
b. Akor
if
Milan
Milan
povredi
hurts
i(t)koga
anyone.acc
/ *ni(t)-koga,
no-one.acc
bi-će
be.fut
kažnjen.
punished
(Intended:) ‘If Milan hurts anyone, he will be punished.’
c. Sumnja-m
doubt.1.sg
da
that
Milan
Milan
voli
loves
i(t)koga
anyone.acc
/ *ni-(t)koga.
no-one.acc
‘I doubt that Milan loves anyone.’
d. Svako
everyone
(t)ko
who
povredi
injures
i(t)koga
anyone.acc
/ *ni(t)-koga,
no-one.acc
mora
must
bit
be
kažnjen.
punished
‘Everyone who injures anyone must be punished.’
(BCS; Progovac 1994)
3. Finally, Progovac analyzes BCS pronouns beginning with the prefix ne, e.g.,
ne-(t)-ko ‘someone,’ nešto ‘something’ as positive polarity items (PPIs).
PPIs (ne(t)ko, nešto, etc.) are according to her pronominals (in the sense of
binding theory). They are subject to Principle B and have to be free in their
governing category. Data supporting PPI analysis of ne pronouns come
from the grammaticality of these pronouns in positive sentences. Their
ability to occur in the scope of superordinate negation and their possible
narrow scope (added formalization, MD) with respect to a superordinate
negation but their obligatory wider scope than their clausemate negation,
after Progovac (1994).
ix
Mojmír Dočekal
(12) Milan
Milan
je
is
uvredio
insulted
ne(t)ko-ga.
someone.acc
‘Milan has insulted someone.’
(13) a. Marija
Mary
ne
not
smatra
thinks
da
that
je
has
Milan
Milan
uvrdio
insulted
ne(t)koga.
someone
‘Mary does not think that Milan insulted someone.’
b. ¬think(Mary, 𝑝) ∧ 𝑝 = ∃𝑥[person(𝑥) ∧ insulted(Milan, 𝑥)]
(14) a. Ne(t)ko
someone
nije
not-is
došao.
come
‘someone has not come.’
b. i. Available: ∃𝑥[person(𝑥) ∧ ¬come(𝑥)]
ii. Unavailable: ¬∃𝑥[person(𝑥) ∧ come(𝑥)]
To take stock: Progovac analyzes three classes of BCS expressions: neg-words, kpronouns,
and indefinite pronouns. She explains their various polarity sensitivity
as a syntactic phenomenon: basically as a need of an expression to be bound either
locally (neg-words) or non-locally (k-pronouns). In the third case (indefinite
pronouns) as an obligation to be free locally (indefinite pronouns must be locally
free but might be bound out of their governing category). The negation acts as
a binder for the first (and to some extent second class). The operator in Comp
is the long-distance binder for the second class; she offers a similar analysis for
English, though with parametric changes.
Progovac distinguishes only two classes of English expressions: (i) PPIs and (ii)
NPIs. The items she considers uncontroversial PPIs for English are some or the
temporal adverbial already. English PPIs (analogically to Slavic data discussed
above) have to be locally free (Principle B). NPIs, like any, have to be bound
locally (Principle A). There is an obvious difference between any and ni- BCS
pronouns – any can be licensed across the clausal boundary (either by negation
or by another DE operator). Such possibility to be licensed non-locally (against
Principle A) is, according to Progovac, the result of a language variation: Englishtype
of NPIs can raise at LF, BCS NPIs cannot.
For English then, Progovac predicts that any can be bound either by a clausemate
negation: (15a) or by an OP licensed from, e.g., a root clause with predicates
that are not upward-entailing: (15b) – be sorry is a downward entailing predicate:
if you are sorry that people kill animals, you are sorry that people kill pigs. This
x
42 Negation and negative polarity items in Slavic
is not possible for Slavic neg-words since they cannot raise at LF; their clausemate
negation must license them. English-type NPIs cannot be licensed only in
a clause without negation and with a verb which is upward entailing like (15c);
if you forgot keys, you forgot an object but not the other way round.6 This all is
correct and seems like a beautiful cross-linguistically sensitive syntactic theory
of NPIs.7
(15) a. Peter didn’t write any letter.
b. Peter is sorry that he said anything.
c. * Peter forgot anything.
But there are some problems. Progovac’s approach predicts (as correctly pointed
out already by Krifka 1995: 212–213) that the English-type NPI can never be licensed
in the non-clausal argument of a non-negated root clause (which works
as an explanation of the ungrammaticality of (15c)). There is no Comp for an
operator in the root clause, and there is no negation to license it either. But
this prediction is wrong both for English and for Slavic languages. Consider the
English example from Krifka (1995: 213), which was later modified by Błasczak
(2001) to show that Slavic-type NPIs (neg-words) can be licensed without any
proper syntactic licensor in Progovac’s sense, (16a). The same point can be made
for k-pronouns as well: they are licensed by adversative predicates, usually licensing
NPIs; see the naturally-sounding Czech example in (17). (17) is a positive
root sentence, so according to Progovac’s approach, the k-pronoun should be
free in this domain. Nevertheless, it is precisely the main predicate of the sentence,
which allows the grammaticality of the k-pronoun. But if the predicate is
the licensor here, we have a Principle B violation pace Progovac (1993). At least
in Czech, the set of predicates that license k-pronouns by themselves seems to
6
I follow here Krifka (1995) and his slightly Janus-face approach to forget: for Krifka, the clausal
argument of forget is (at least) not upward entailing since “If Mary forgot that a woman came
yesterday, she might not have forgotten that a person came yesterday” (Krifka 1995: 212), but
the NP argument of forget is upward entailing since “if Mary forgot a poem by Goethe, then
she forgot a poem, but not necessarily vice versa” (Ibidem). That explains correctly why NPIs
in the clausal arguments of forget are licensed (I forgot to bring anything), unlike in (15c). But
this simply seems like a restatement of NPI licensing behavior in terms forget’s entailment
profile, as one of the anonymous reviewers correctly remarks. It would be best to thoroughly
test NPI-licensing patterns of forget across languages and then hopefully come up with the
right theory of forget, but that has to remain a project for future work.
7
I simplify here a bit. Progovac, in fact, observes that there are some cases of acceptable any
in the non-clausal argument of non-negated root sentences, but she classifies such examples
as FCI usage of any. Her explanation was then questioned by Horn & Lee (1995); Hoeksema
(1996). Thanks to one of the anonymous reviewers for pointing out the debate.
xi
Mojmír Dočekal
overlap with adversative predicates. Most frequent collocations of positive verbs
occurring with k- pronouns – from Czech National Corpus – are verbs as zakázat
‘forbid,’ odmítnout ‘refuse’ and vyhýbat ‘avoid.’ But there are also many
examples of licensing by positive existential modals or imperatives. Such cases
are not explainable as “inherently negative” in the style of adversative predicates;
see Strachoňová (2017) for many valuable data points).
(16) a. John lacks any sense of humor.
b. John came without any present.
(17) Petr
Petr
odmítal
refused
jakoukoliv
k-pronoun
pomoc.
help
‘Petr refused any help.’ (Czech)
In this section, a historically very influential syntactic theory of NPIs licensing
was discussed. There are obvious shortcomings of the theory, discussed partially
above and furthermore in section §2.2. Maybe the most serious problem is the
attempt to unify three Slavic classes of expressions under the roof of the binding
theory. Today it seems more reasonable to treat neg-words really in syntax. One
of today’s standard theories of neg-words is Zeijlstra (2004) – different in formalization
but similar to Progovac (1994) in taking locality constraints seriously as
syntactic in nature. And as for k-pronouns and indefinite pronouns, they ought
to be explained via some appropriate semantic or pragmatic theory (already the
problems of adversative predicates pointed out the difficulty of finding a reasonable
syntactic licensor for k-pronouns or English NPIs in a purely syntactic
framework like Progovac 1994). But despite these problems, Progovac (1994) is
the first serious attempt to deliver a theory of various NPI classes, starting with
Slavic data but aiming at a general framework of cross-linguistically conceived
polarity licensing.
2.2 Semantic theories
The semantic approach to NPI licensing is still the standard theory of the semantic
negative dependent expressions. There are various ways of approaching the
idea that NPIs are licensed purely semantically. The most widely accepted reasoning
is based on the downward entailing approach of Ladusaw (1992), but the
idea can be found in influential works of Heim (1984); Ladusaw (1992); Kadmon
& Landman (1993); Krifka (1995); Giannakidou (1997); Lahiri (1998) too. It usually
starts with the observation (introduced quickly already in section §1) that next
xii
42 Negation and negative polarity items in Slavic
to negation, there are many other prototypical NPI licensing contexts. Such contexts
include downward entailing quantifiers, antecedents of conditionals, the
scope of the exclusive particle only, adversative predicates, comparatives, and
superlatives. All the contexts are claimed to share the property of reversing the
direction of entailment. The monotonicity reasoning is demonstrated in the predicate
logic implications in (18). In the non-negated formula (18a) (corresponding
to a natural language positive sentence under (18b)), the entailment goes from a
subset (intersection of P and Q) to a superset (union of P and Q). In a negated
formula (18c) (and corresponding natural language sentence under (18d)), the entailment
is reversed and proceeds from a superset (union) to its subset (intersection).
The same entailment reversal can be observed in the difference between
upward-entailing quantifiers as some versus downward entailing quantifiers like
few demonstrated in (19).
(18) a. ∃𝑥[𝑃(𝑥) ∧ 𝑄(𝑥)] → ∃𝑥[𝑃(𝑥) ∨ 𝑄(𝑥)]
b. ∃𝑥[red(𝑥) ∧ wine(𝑥)] → ∃𝑥[red(𝑥) ∨ wine(𝑥)]
John likes red wine. → John likes wine.
John likes wine. ↛ John likes red wine.
c. ¬∃𝑥[𝑃(𝑥) ∨ 𝑄(𝑥)] → ¬∃𝑥[𝑃(𝑥) ∧ 𝑄(𝑥)]
d. ¬∃𝑥[red(𝑥) ∨ wine(𝑥)] → ¬∃𝑥[red(𝑥) ∧ wine(𝑥)]
John doesn’t like wine. → John doesn’t like red wine.
John doesn’t like red wine. ↛ John doesn’t like wine.
(19) a. Some people drank red wine. → Some people drank wine.
b. Some people drank wine. ↛ Some people drank red wine.
c. Few people drank red wine. ↛ Few people drank wine.
d. Few people drank wine. → Few people drank red wine.
The semantic approach is very successful: it explains the distribution of so-called
weak NPIs (like English any; more on different types of NPIs in the section §2.2.2)
via a unified semantic property. The property of NPI licensing is that they occur
in downward entailing (entailment reversing) environments (but see §2.2.1 for
many qualifications). Despite some problems discussed below, the DE-based explanation
of NPI licensing is still considered a benchmark of semantic reasoning
and appears in standard textbooks on formal semantics (Portner 2005; Coppock
& Champollion 2024; among many others). The proper formalization of DE licensing
can be stated as (20), and DE is defined in (21), both after von Fintel
(1999). The explanation of weak NPI acceptability in a sentence like John didn’t
xiii
Mojmír Dočekal
drink any wine yesterday (contrasted with ungrammaticality of *John drank any
wine yesterday) is as follows: (22) states the entailment pattern for red wine and
wine (x and y from (21)), subset relation on sets corresponds to the entailment in
(21) as ∀𝑥[[red(𝑥) ∧ wine(𝑥)] → wine(𝑥)]. But negation (F from (21)) reverses
the sub-set superset relationship: ∀𝑥[¬wine(𝑥) ⇒ ¬[red(𝑥) ∧ wine(𝑥)]]. And
because of this DE entailment reversal, any is licensed as stated in (20).
(20) Fauconnier-Ladusaw’s Licensing condition: An NPI is only grammatical if
it is in the scope of an 𝛼 such that 𝛼 is DE. (von Fintel 1999: 100)
(21) A function 𝑓 of type ⟨𝛿, 𝜏⟩ is downward entailing iff for all 𝑥, 𝑦 of type 𝛿
such that 𝑥 ⇒ 𝑦: 𝑓 (𝑦) ⇒ 𝑓 (𝑥) [‘⇒’ stands for cross-categorial entailment].
(von Fintel 1999: 100)
(22) {𝑥 ∶ 𝑥 is a red wine} ⊆ {𝑦 ∶ 𝑦 is a wine}
2.2.1 Some problems of purely semantic theories
Even if the criterion in (20) is very successful, during years of NPI exploration,
there appeared many problems. The problems lead to some ramifications (and
in some cases abandonment or replacement of the criterion; see Giannakidou
1997; Linebarger 1987). Very famous problems arise when we scrutinize monotonic
properties of only, conditionals, superlatives, and adversative predicates.
All these expressions seem to license NPIs but do not fit the downward entailing
defining property, at least at first sight. A proper exposition of such problems
goes beyond the scope of this chapter; see von Fintel (1999) for the classic reference
and Gajewski (2011) among many others for a recent refinement.8 But let
me demonstrate the nature of the problems of simple DE theory with an example.
Despite the validity of the sentential logic tautology in (23a), conditionals in
natural language seem not to allow the strengthening of antecedents. Consider
counter-intuitive reasoning from (23b) to (23c), but even if such reasoning can
sound paradoxical when viewed intuitively, it would be formalized in classical
logic as the correct strengthening in (23a). As a consequence, it seems that natural
8
As one of the anonymous reviewers correctly notices, von Fintel (1999) is just one of the attempts
(even if very influential and empirically successful) to overcome the limitations and
problems of a purely semantic approach to NPIs. Historically sorted, at least the following
researchers – Heim (1984); Kadmon & Landman (1993); Israel (1996) – proposed various pragmatic
relativizations of the entailing conditions necessary for NPI licensing, be it contextual
entailment, fixation of contexts, or integrating background knowledge of communication participants
into the reasoning important for NPI licensing.
xiv
42 Negation and negative polarity items in Slavic
language conditionals do not exhibit downward entailing property, but despite
that, they are one of the prototypical NPI licensors. Similar problems were observed
with the contexts mentioned above (adversative predicates, superlatives,
…) too. Even if the environments are not intuitively downward entailing, they
license NPIs like any: see (24).
(23) a. (𝑝 → 𝑞) → ((𝑝 ∧ 𝑟) → 𝑞)
b. If [you pet a dog] it will be happy. ↛
c. If [you pet a dog and kick it] it will be happy.
(24) a. If you pet any dog, it will be happy.
b. I am sorry to give you any trouble.
c. Only Peter drank any tea.
The analysis of the problems concerning conditionals (and other problematic
environments) proceeds in the following way (after von Fintel 1999). First, the
basic notion of downward entailment is replaced by Strawson downward entailment
(SDE), and SDE adds to the standard argument a presupposition of
the sentence. Second, the added presupposition plus the at-issue meaning of the
sentence then qualify for the downward entailing property as described in (21).9
For conditionals like (23b), the added presupposition would concern the current
modal horizon. Modal horizons are sets of worlds that are compatible with everyday
reasoning about conditionals. For (23b), possible worlds where an agens
both pets and kicks a dog are beyond the current modal horizon. Consequently,
adding a presupposition to the effect that the current modal horizon is compatible
only with worlds where only petting a dog would cause happiness of a dog
would save the downward monotonicity of conditionals and explain the grammaticality
of an NPI in (24).
This gives some impressionistic introduction to SDE reasoning, but let us
demonstrate the SDE on a simpler pattern since a proper treatment of reasoning
in conditionals is surely beyond the scope of this article. Gajewski (2016)
offers exactly such a type of example; he discusses the cases of NPI licensing by
plural definite NPs like in (25). But the entailment in (26) is invalid since knowing
that the supremum of (in the context) salient students drank beer does not
tell us anything about their nationality. Exactly a similar problem as discussed
above with respect to the pattern in (24) is encountered – expressions that are
not DE do license NPIs.
9
The formal definition of SDE is in (27).
xv
Mojmír Dočekal
(25) The students who have any beer are sharing it. (after Gajewski 2016: ex. 2)
(26) The students drank beer. ↛ The German students drank beer.
The exact definition of Strawson downward entailment is in (27). In (26), the
conclusion wasn’t entailed by the premise. But according to the (27), the presupposition
of the conclusion (the existence and uniqueness presupposition of the
definite description in conclusion) should be granted as a premise in the argument
as demonstrated in (28). Then the argument becomes valid, although the
price is obvious: the entailment must take into account pragmatic notions like
presupposition. Despite that, SDE was successfully applied to many cases of NPI
licensing in environments that simply are not DE, so it is a useful linguistic tool.
(27) Strawson Downward Entailingness (after von Fintel 1999: 104)
A function 𝑓 of type ⟨𝜎, 𝜏⟩ is Strawson-DE iff for all 𝑥, 𝑦 of type 𝜎 such that
𝑥 ⇒ 𝑦 and 𝑓 (𝑥) is defined: 𝑓 (𝑦) ⇒ 𝑓 (𝑥).
(28) The students drank beer.
There are salient German students.
⊧ The German students drank beer.
But even if the solution via Strawson downward entailment is widely accepted
today, there still remain problems that are not well-understood. One of them
(important later – see section §2.3.3) is the following one: von Fintel (1999) and
his SDE predicts that adversative predicates as sorry, regret, amazed, difficult,
or refuse can license NPIs because adding a (usually factive) presupposition unmasks
their hidden DE properties. But predicates like want, glad, or like cannot
license NPIs because they are not even SDE (they are usually analyzed as upward
monotonic but see von Fintel (1999) for qualifications). The prediction seems to
be mostly right: compare the contrast between (29a) and (29b). But there is a
special type of context (discussed by Kadmon & Landman 1993 already) where
glad can license NPIs. Such contexts permitting NPIs under glad have a special
interpretation called “settle for less” by Kadmon & Landman (1993). The core description
of such contexts demonstrated in the example (30) is the following: the
speaker wanted an alternative (higher) number of tickets, but because he considered
the situation so dire that even lousy tickets were the only possibility, he
settled for less than he would normally want. For some reasons (not totally clear,
but see Alonso-Ovalle 2016 for some recent insights), it seems that other adversative
predicates like want, hope, or wish at least in English do not allow the
settle-for-less reading and consequently cannot license NPIs. Nevertheless, to
xvi
42 Negation and negative polarity items in Slavic
sum up, the current sub-section: the semantic approach (if enriched with a pragmatic
notion like SDE) can deal with the majority of more complicated cases
where allegedly non-downward entailing expressions license NPIs (and which
were correctly pointed out already by Linebarger (1987) as not simply downward
entailing).
(29) a. I regret there is any idea like that.
b. # I am glad there is any idea like that.
(30) I am glad that we got any tickets.
2.2.2 Types of NPIs
Another step in broadening the empirical coverage of semantic NPI theories was
a seminal work of Zwarts (1998) where various types of NPIs are, for the first time
systematized, and a purely semantic explanation of their differing distribution is
offered. This classification of NPIs was especially important from the perspective
of Slavic languages as it inspired the most thorough semantic approach to Slavic
NPIs: Błasczak (2001) and some following works of her like Błaszczak (2003) and
Błaszczak (2008); Błasczak’s contributions are discussed in §2.2.3. Consider the
first examples of the two most prominent classes of NPIs: weak in (31) and strong
in (32). As evident, only negation and negative quantifiers license strong NPIs
(such as until midnight in (32)), while for the licensing of weak NPIs as any, the
downward entailing environment is enough: (31)
(31) a. John didn’t leave any message.
b. No teacher left any message.
c. Few teachers left any message.
d. At most ten teachers left any message.
e. # Some teachers left any message.
(32) a. John didn’t leave until midnight.
b. No teacher left until midnight.
c. # Few teachers left until midnight.
d. # At most ten teachers left until midnight.
e. # Some teachers left until midnight.
xvii
Mojmír Dočekal
The logical property which licenses strong NPIs share is a strengthened form
of entailment reversal and usually is named anti-additivity.10 The definition
of anti-additivity is in (34): it represents a half of the well-known deMorgan’s
laws of conjunction/disjunction transformation via negation (Tarski 1994: 46).
Standard logical negation obeys two deMorgan’s laws in (33).
(33) a. ¬(𝑝 ∨ 𝑞) ↔ ¬𝑝 ∧ ¬𝑞
b. ¬(𝑝 ∧ 𝑞) ↔ ¬𝑝 ∨ ¬𝑞
Anti-additivity picks one of them, so we can view anti-additivity as a subset of
negation properties. (35) illustrates the anti-additivity (the quantifier no is antiadditive
since negation is always anti-additive, as is clear from deMorgan’s laws).
But DE quantifiers like few in (36) are not anti-additive – imagine a scenario with
10 students, three of them drinking and three of them smoking, then ∨ part of
(36) is false while ∧ part of (36) is true. That explains why no is licensed by the
strong NPI until midnight in (32) and not licensed by just downward entailing
quantifier few.11
(34) Anti-additive function: 𝐹(𝑥 ∨ 𝑦) ↔ 𝐹(𝑥) ∧ 𝐹(𝑦)
(35) No student smokes or drinks.
↔ No student smokes and no student drinks.
(36) Few students smoke or drink.
↮ Few students smoke and few students drink.
2.2.3 Semantic theory applied to Slavic data
Błasczak (2001) and other works by her apply this classification of NPIs to Slavic
data. First, Błasczak observes that generally: k-pronouns (Polish pronouns with
10
A popular alternative explanation of strong NPIs, and their behavior can be found in Gajewski
(2011). Gajewski describes their stricter distribution via downward entailing properties but
checked both in at-issue meaning and in the presupposition/implicature part of the meaning.
More pragmatic approaches to NPI licensing will be presented in section §2.3.
11
As one of the reviewers correctly remarks, anti-additivity as presented in this section and its
contrastive logical behavior with respect to DE is just a part of the whole picture, presented
carefully and systematically in Zwarts (1998). Zwart’s main idea then is that there is a hierarchy
of negative strength: downward entailing < anti-additive < anti-morphic. Anti-morphicity
is defined as: F is anti-morphic iff 𝐹(¬𝑝) ↔ ¬𝐹(𝑝). Classical verbal negation is, of course,
anti-morphic, since ¬(¬𝑝) ↔ ¬¬𝑝 but universal quantifiers, which are anti-additive, since
∀𝑥((𝑃(𝑥) ∨ 𝑄(𝑥)) → 𝑅(𝑥)) ↔ ∀𝑥(𝑃(𝑥) → 𝑅(𝑥)) ∧ ∀𝑥(𝑄(𝑥) → 𝑅(𝑥)), are not anti-morphic
because ¬∀𝑥(𝑃(𝑥) → 𝑄(𝑥)) ↮ ∀𝑥(𝑃(𝑥) → ¬𝑄(𝑥)). The relative strength of the licensor is, for
Zwarts (1998), picked up by weak, strong, and superstrong classes of NPIs, respectively.
xviii
42 Negation and negative polarity items in Slavic
the suffix -kolwiek) are licensed in prototypical NPI contexts. Such contexts include
higher clause negation, questions, conditionals, the scope of downward entailing
quantifiers, comparatives, adversative predicates, before-clauses, etc. Negpronouns
(Polish pronouns usually beginning with n-) are generally licensed
only by clause-mate negation: see (37) and (38).
(37) a. * Ewa
Eve
nie
neg
chciała,
wanted
żeby
that.sbjv
Jan
John
nikogo
neg-pronoun
zapraszał.
invited
Intended: ‘Eve didn’t want John to invite anyone.’
(Polish; Błaszczak 2003: 2)
b. Ewa
Eve
chciała,
wanted
żeby
that.sbjv
Jan
John
nie
neg
nikogo
neg-pronoun
zapraszał.
invited
‘Eve wanted John not to invite anybody.’
(Polish; Błaszczak 2003: 2)
(38) a. Czy
whether
widziałeś
saw.2sg
tam
there
kogokolwiek?
anybody
‘Have you seen anybody there?’
b. Jeśli
if
ktokolwiek
k-pronoun
przyjdzie,
comes
daj
let
mi
me
znać.
know
‘If anyone comes, let me know.’
c. Ewa
Eve
nie
neg
chciała,
wanted
żeby
that.sbjv
Jan
John
kogokolwiek
k-pronoun
zapraszał.
invited
‘Eve didn’t want John to invite anyone.’
(Polish; Błaszczak (2003: 2))
This is a familiar observation concerning the partial complementary distribution
of two classes of Slavic pronouns, which was discussed already in the section §2.1
dedicated to the syntactic approach of Progovac (1994). But Błasczak’s (2001) solution
is different, and her critique of Progovac (1993) is correct. Błasczak (2001)
criticizes syntactic approach of Progovac (1993) mainly on empirical grounds.
One of her main arguments concerns the licensing of neg-pronouns by a preposition
bez ‘without.’ The grammaticality of (39) is unaccounted by the syntactic
approach of Progovac since there is no negation or polarity operator in Comp.
But since (at least in Polish) neg-pronouns are fully licensed in the complement
of bez ‘without,’ Błaszczak concludes that the semantic approach is more in accordance
with the Slavic data. For more general problems of Progovac’s syntactic
theory of NPI licensing, see §2.1.
xix
Mojmír Dočekal
(39) Został
was-left.3sg
bez
without
nikogo.
neg-pronoun
‘He was left without anyone.’
The core of Błasczak’s (2001) proposal concerning Polish NPIs is the following:
1. K-pronouns are licensed in downward entailing or anti-additive environments.
So in contrast to the syntactic binding-theoretic analysis of Progovac
(1993), we see the standard semantic treatment of NPIs, which is more
in accordance with the variability of NPI licensors: quantifiers, polarity operators,
various types of embedding, etc., which cast doubt on any purely
syntactic approaches to NPIs.
2. According to Błasczak (2001), neg-pronouns can occur only in anti-morphic
environments.12 The definition of the anti-morphic functor (adding one
condition on top of anti-additivity) is in (40). An anti-morphic function is
a classical negation because (deMorgan’s laws) it obeys both requirements
of (40). And it seems that only clausal negation fits such a requirement in
a natural language. And because of its nature, it also satisfies another condition
in (41): first negation before the equation sign in (41b) corresponds
to f in (41a). But notice that anti-morphic are also negated truth-value ascribing
predicates and neg-raising predicates: like it is not true that … or
x does not believe that …. Both contexts will be discussed at the end of the
current section.
Applied to examples in (37): a negated modal verb does not license neg-pronouns
in its embedded clause, since from ¬∀[𝑃] it does not follow ∀[¬𝑃]. Nevertheless,
the ungrammaticality of (37) is not caused by the interfering modal since modals
generally do not interrupt NPI licensing. Verbal negation, of course, licenses the
clausemate neg-words. For k-pronouns in (38), the semantic analysis predicts
that they should be grammatical in anti-additive conditionals (they are, in fact,
Strawson anti-additive, see Gajewski 2011) or in a simple downward entailing
context like negated embedding modal verb.
(40) A functor F is anti-morphic iff:
a. 𝐹(𝑥 ∧ 𝑦) ↔ 𝐹(𝑥) ∨ 𝐹(𝑦)
12
As one of the reviewers correctly points out, it is not clear whether this claim is valid even
for sentences with ≥ 2 occurrences of neg-words, since the scopally lowest neg-word is in the
anti-additive environment, but the sentence is still grammatical.
xx
42 Negation and negative polarity items in Slavic
b. 𝐹(𝑥 ∨ 𝑦) ↔ 𝐹(𝑥) ∧ 𝐹(𝑦)
c. ¬(𝑥 ∧ 𝑦) ↔ ¬(𝑥) ∨ ¬(𝑦)
d. ¬(𝑥 ∨ 𝑦) ↔ ¬(𝑥) ∧ ¬(𝑦)
(41) a. 𝑓 (¬𝑥) = ¬𝑓 (𝑥)
b. ¬(¬𝑥) = ¬¬(𝑥)
The semantic approach is much more appropriate than the syntactic approach,
especially in its treatment of k-pronouns. The set of environments where kpronouns
appear forms a natural semantic class. But there are some problems
with the semantic approach too. Some of them were observed in the recent general
linguistic works, which usually argue for a semantic and pragmatic approach
to NPIs instead of a purely semantic one (see Krifka 1995; Gajewski 2011; Chierchia
2013 among many others). Nevertheless, the standard approach today is to
treat neg-words as a separate class of negative dependent expressions licensed
in syntax (see already Krifka 1995 for some insightful remarks and Zeijlstra 2004
for the current de facto standard theory of neg-words). But let us discuss some
empirical problems which are particularly interesting and concern Slavic data:
1. As (42a) shows, only NPIs are licensed in neg-raising contexts; real negwords
are not acceptable there (see Dočekal & Dotlačil 2016a for the experimental
data confirming this claim). But neg-raising predicates are antimorphic
even in their embedded clauses, so a purely semantic approach
predicts that both NPIs and neg-words should be licensed there (contrary
to facts). This asymmetry raises serious problems for any semantic treatment
of neg-words.13
13
One of the anonymous reviewers correctly remarks that the status of neg-raising predicates
as either anti-morphic or as anti-additive is, to some extent, theory-dependent. In approaches
where the negation would really be present syntactically and semantically in the embedded
clause, they would be anti-morphic, but for Gajewski (2005; 2007), the embedded clauses of
neg-raisers are just anti-additive. Gajewski (2007: 305) empirically supports his classification
with the following entailment pattern where (for him) the first entailment is valid, but the
second one is not. And since, for him, the crucial test for anti-morphicity is the entailment
𝑓 (𝑥 ∧ 𝑦) → 𝑓 (𝑥) ∨ 𝑓 (𝑦), he argues for the mere anti-additive status of neg-raising embedded
clauses.
(i) John doesn’t think Mary left and John doesn’t think Bill left.
→ John doesn’t think Mary left or Bill left.
(ii) John isn’t certain that Mary left and John isn’t certain that Bill left.
↛ John isn’t certain that Mary left or Bill left.
xxi
Mojmír Dočekal
2. The same point holds for truth-value ascribing predicates: (42b) and (42c)
are equivalent, but only (42e) allows negative concord items. Again purely
semantic theory of neg-words is the problem here, while syntactic locality
constraints fare better.14
(42) a. Petr
Petr
nechce,
neg.wants
aby
comp
ani
even
jeden
one
student
student
/ #nikdo
neg-person
propadl.
failed
(Intended:) ‘Petr doesn’t want even one student/neg-person to fail.’
(Czech; MD)
b. It’s not true that Peter was sleeping. ↔
c. It’s true that Peter wasn’t sleeping.
d. * Není
neg.is
pravda,
true
že
that
nikdo
neg-person
přišel.
came
Intended: ‘It’s not true that neg-person came.’
e. Je
is
pravda,
true
že
that
nikdo
neg-person
nepřišel.
neg.came
‘It’s true that neg-person didn’t come.’
(Czech; MD)
Let us summarize the current section. The relationship between verbal negation
and the morphological negation on neg-words is a syntactic phenomenon; trying
to explain it in semantics leads to empirical problems. There are NPIs in
Slavic languages, though, which seem to behave in the way described by Błasczak
(2001), namely as anti-morphic NPIs (Czech ani jeden is an example of such an
NPI). But the Slavic neg-words are (it seems from today’s perspective) better delegated
to purely syntactic rules.
2.3 Pragmatic theories
After summarizing two types of NPI theories applied to Slavic data in the previous
sections (a syntactic theory in §2.1 and a semantic one in §2.2), I will foI’m
not sure whether his argument is very strong since the predicate logic implication
¬∀𝑥(𝑃𝑥 ∧ 𝑄𝑥) → (¬∀𝑥𝑃𝑥 ∨¬∀𝑥𝑄𝑥) is valid. Recent work on necessity modals (Agha & Jeretič
2022) opens up a possibility to formalize certain necessity modals not as universal quantifiers.
Going in this direction would weaken Gajewski’s argument even more. Nevertheless, following
this route of argumentation, would be intriguing but would lead us too far away from the
merit of the current chapter.
14
Gajewski (2007: Appendix to Sect. 2) offers a semantic treatment of the fact that strict NPI is
not licensed in the clauses embedded under it’s not true that predicates. The proposal works
with the idea of presupposition-canceling properties of true; see Gajewski (2007: 308–310) for
details. Thanks to one of the anonymous reviewers for reminding me of this passage.
xxii
42 Negation and negative polarity items in Slavic
cus on the last type of approach to NPIs – a pragmatic one. The pragmatic approach
has not been systematically applied to Slavic languages (at least not to
the same extent as semantic and syntactic theories). But there is some work in
this direction (as Tomaszewicz 2013, gajic2016coordination; Gajić 2022, Dočekal
& Dotlačil 2016c or Dočekal & Šafratová 2019). But let us start with the generallinguistic
properties of this approach. Currently, it seems that the pragmatic theories
of polarity constraints are de facto standard, one of the recent formalizations,
namely Chierchia (2013), is probably the most widely used framework in
the area. One of the applications of Chierchia (2013) to Slavic data can be found in
gajic2016coordination. The shared idea of all pragmatic theories of NPI is that in
describing polarity effects, we have to consider both the semantic properties of
the expressions under consideration and general pragmatic mechanisms as well.
The formalizations usually agree on the idea that NPIs introduce alternatives,
and polarity effects emerge when something goes wrong during the computation
of alternatives and their inclusion into the truth conditions. The execution
of these ideas can then proceed either via the postulation of exhaustification operators
present in syntax (like Chierchia 2013) or in the neo-Gricean spirit where
the pragmatic mechanisms are more related to general rules of rational commu-
nication.
In particular, I will follow the neo-Gricean pragmatic approach to NPIs as
introduced by Krifka (1995) and followed by many others (especially the work of
Crnič 2011 and Alonso-Ovalle 2016 will be of most relevance). But for the present
purposes, nothing hinges too much on the adopted framework, and it seems to
me that it would be possible to reformulate the present section in the framework
of Chierchia (2013). Nevertheless, let us first illustrate the typical problems pragmatic
theories (unlike syntactic or semantic) can deal with. Consider the Czech
scalar particles (SP) i and ani, which have a complementary distribution to
some extent and resemble the complementary patterns described by Progovac
(1993) and Błasczak (2001). Until now, I have described the polarity properties
of NPIs, but from now on, I will focus on scalar particles, which represent an
extension of the polarity landscape we discussed so far. Czech scalar particle ani
behaves to some extent like a strong NPI requiring negation, but simple antiadditivity
is not enough to license its appropriate occurrence. In the context of
the familiar three books of Tolkien’s Lord of the Rings, consider its acceptability
(for experimental data supporting the following judgments, see Dočekal &
Šafratová 2019). As (43a) shows, ani cannot be licensed in downward entailing
contexts only, but even the anti-additive operator is not able to license all possible
occurrences of ani: (43b). As the example shows, ani associates only with
the bottom of the scale conveyed by the scalar expression (the contextual scale
xxiii
Mojmír Dočekal
being ⟨first volume, second volume, third volume⟩). Association with the top of
the scale leads to ani’s unacceptability. The need for a pragmatic explanation, intuitively
put, is the following: reading the first volume is the most likely case, the
second less likely, and reading the third one is the least likely. Ani then requires
not only a particular type of monotonicity but also ranking over the probability
of alternatives where alternatives (required by ani) come from focus computation.
Similarly, as can be seen from the contrast in (43c), i shows (at first sight)
just the opposite pattern to ani: it occurs in positive sentences, and pragmatically
it can associate only with the top element of a contextual scale. When it occurs
in a negated sentence, it cannot associate with the top or bottom elements of a
scale; see (43d).15 Again, monotonicity alone is not enough to license its grammaticality.
In summary: there are both semantic and pragmatic requirements of
both i and ani. Technical explanation of these kinds of patterns follows in §2.3.1.
(43) a. # Málo
few
lidí
people
přečetlo
read
ani
not.even
{první
first
/ třetí}
third
díl
volume
Pána
Lord
prstenů
Rings
Intended: ‘Few people have even read the first/third volume of the
Lord of the Rings.’
b. Petr
Petr
nepřečetl
neg.read
ani
not.even
{první
first
/ #třetí}
third
díl
volume
Pána
Lord
prstenů.
Rings
‘Petr didn’t read even first/#third volume of the Lord of the Rings.’
c. Petr
Petr
přečetl
read
i
even
{#první
first
/
/
třetí}
even
díl
third
Pána
Lord
prstenů.
Rings
‘Petr read # even the first / third volume of Lord of the Rings.’
d. # Petr
Petr
nepřečetl
neg.read
i
even
{první
first
/ třetí}
third
díl
volume
Pána
Lord
prstenů.
Rings
Intended: ‘Petr have even read the first/the third volume of Lord of
the Rings.’
(Czech; MD)
2.3.1 Crnič’s decompositional theory of scalar particles
I will introduce a pragmatic theory of Crnič (2011) where the constraints on scalar
particle distribution and their semantic properties are explained via a pragmatic
theory of presuppositions. The presuppositions are contributed by NPIs or scalar
particles as i/ani. The description of scalar particles is related to Crnič’s (2011)
15
(??) is unacceptable under the scalar interpretation. But since i can be used also as an additive
particle, there can be non-scalar contexts where (43d) would be acceptable. Thanks to Radek
Šimík for this point.
xxiv
42 Negation and negative polarity items in Slavic
pragmatic approach to English even as a polarity item. But especially in chapter
4 and chapter 5, where he describes scalar particles, he pays close attention to
cross-linguistic variation and discusses many intriguing properties of Slovenian
scalar particles. So unlike in §2.1 and §2.2, where I summarized two contributions
to polarity research explicitly focused to (mostly) Slavic data, in this section, I will
introduce a more general framework. The framework was only partially applied
to Slavic data more as a form of demonstration. This, of course, has some consequences;
the nice one is that Crnič’s framework is an explicit formal theory of
presupposition and polarity effects of NPIs/scalar particles. The worse one is that
the approach still lacks proper empirical meat, and naturally concerning Slavic
data (even if it offers many insights) there are many points where it has to be
fine-tuned or slightly changed to be in accordance with the data. Be it as it may,
another strong aspect of Crnič’s theory is its further component which describes
the distribution of scalar particles as determined partially by the spectrum of the
relevant scalar particles in the particular language. In other words, there is a competition
between various types of scalar particles, and their behavior is a result of
two factors. The first factor is the presuppositions calculation. The second factor
is a competition between various expressions available for the lexical insertion.
2.3.2 Three basic types of scalar particles
I will start by introducing fundamental dichotomies in the territory of scalar
particles. I will deal mostly with so-called weak and strong scalar particles (e.g.,
German expressions sogar and auch nur). And I will only scratch the surface of
concessive scalar particles like Slovenian magari or Spanish siquiera.
2.3.2.1 Nondiscriminating scalar particles The first type of scalar particles is
the one associating both with strong and weak elements in their immediate surface
scope. The most famous linguistic example is English even: its strong association
is free, (44a), and it can associate with weak expressions in non-upward
monotonic embedding too: (44b). There are no known Slavic examples of such
undiscriminating type of even, Crnič (2011: 129) cites French même as a second
example, next to English even.16
16
It remains to be worked out whether there is no example of Slavic nondiscriminating scalar
particles or whether it so far remained hidden. One of the anonymous reviewers suggests that
Polish nawet ‘even’ can be used both with strong and weak associated expressions. Similarly,
in Czech the scalar particle dokonce ‘even’ seem to associate with both ends of the scale. Neverthelless,
as natural occurences suggest, nawet is out of the blue more acceptable with weak
elements, for strong elements it usually requires some intervening modal or other operator.
xxv
Mojmír Dočekal
(44) a. John drank even TEN beers.
b. Paul didn’t drink even ONE beer.
17
2.3.2.2 Strong scalar particles The second class of particles, strong scalar
particles, comprises of such particles which associate with strong scalar elements.
This class is well documented for many natural languages, Crnič (2011:
130) discusses German sogar and Slovenian celo. We can add Czech scalar particle
i introduced already above: see a grammatical association of i with a strong
element in (45) where an unlikelihood presupposition of i requires the prejacent
to be less likely than the alternatives (drinking 𝑛 < 10 beers) which intuitively is
the case (formalizations below).
(45) (V
in
těch
those
letech)
years
Petr
Petr
vypil
neg.drank
i
even
DESET
ten
piv.
beers
‘(In those years) Peter drank even ten beers.’ (Czech)
In Crnič’s theory, strong scalar particles seem to be superficially stuck in their
base position when they would appear in a downward entailing environment.
Otherwise, they would be grammatical contrary to natural language data, see
ungrammatical (46a) with the un-acceptability explaining formalization in (46b).
If the scope of i would be above negation ([i [not [Petr drank one 𝐹 beer]]]), the
unlikelihood presupposition of i would be satisfied.
(46) a. * (V
in
těch
those
letech)
years
Petr
Petr
nevypil
neg.drank
i
even
JEDNO
one
pivo.
beer
Intended: ‘(In those years) Petr didn’t drink even one beer.’
(Czech)
b. [not [i [Petr drank one 𝐹 beer]]]
Czech dokonce behaves just the other way round, it occurs most naturally with strong elements.
In summary, the proper classification of promissing candidates for Slavic nondiscriminating
scalar particles remains a future project.
17
(44a) out of context can sound a bit strange in English but as the following sentence from
Sketchengine (Kilgarriff et al. 2014) English Web 2021 shows, English even associates with the
strong scalar elements: Some workers even drank 3-4 liters of the contaminated water after mine
managers failed to warn them.
xxvi
42 Negation and negative polarity items in Slavic
2.3.2.3 Scalar particles that may associate only with weak elements The last
type of particles I will write about, namely weak scalar particles, are particles
which associate with weak elements. Unambiguous examples are German auch
nur, einmal, English so much as and Slovenian niti. English even can be used as
a weak particle too, but in that case, it must be embedded under non-upwardentailing
operator (other weak scalar particles can appear only in non-upward
entailing contexts too but they – unlike English even – are limited to such environments).
Linguists approaching scalar particles from the typological perspective
(like Crnič 2011: chapters 4 & 5 and Gast & Van der Auwera 2011) sub-classify
weak scalar particles according to their possible scope under negation or other
non-upward-entailing operators. Most constrained are scalar particles like German
einmal and Slovenian niti: they appear only in the immediate scope of negation.
The second subtype comprises scalar particles which occur in the scope
of the non-upward-entailing operator but may not occur under negation, German
auch nur or Slovenian tudi are scalar particles of this sort. Last, there are
in-discriminating weak scalar particles, like English so much as which may occur
both in the immediate scope of negation and in other non-upward-entailing
environments as well.
Now, let us illustrate the classification at three types of Czech scalar particles
corresponding to strong, weak, and negation requiring weak particles respectively.
Following the introductory discussion of i, let us consider its type: i like
German sogar requires association with strong elements: if the contextual scale
is 1 to 10 beers, then only numerals contextually vaguely around 10 would make
an appropriate focus associate of i, see (47) with acceptable strong associate and
un-acceptable weak associate.
(47) Petr
Petr
vypil
drank
i
even
{#JEDNO
one
pivo
beer
/ DESET
ten
piv}.
beers
‘Petr drank even #one/ten beers.’ (Czech)
The second type of Czech scalar particles, are expressions of type byť jediný ‘at
least/even one,’ they generally appear in Strawson-downward entailing contexts
like an antecedent of conditionals, under super-ordinate negation or generally in
the scope of non-upward-monotonic operators like imperatives and some adversative
predicates. Depending on their embedding their correct glossing to English
is either ‘at least one’ or ‘even one.’ The expression seems to be frozen in cardinality,
so is pre-destined for the weak association, consider (48), unlike i which
(depending on the type of scale) can associate with different kinds of expressions.
Nevertheless, the expression byť jediný ‘even one’ is not an idiom as witnessed
xxvii
Mojmír Dočekal
by various sortal version of the numeral included in it: it can count cardinality
of objects like in (48) but also cardinality of events (byť jednou ‘even once’) or
cardinality of a sequence of occurrences like byť poprvé ‘even for the first time’.
(48) Jestli
if
Petr
Petr
vypije
drink
byť
just
{jediné
one
pivo
beer
/ #deset
ten
piv},
beers
tak
then
bude
will
tančit.
dance
‘If Peter will drink just one beer/#just ten beers he will dance.’ (Czech)
The distinction between i and byť jediný follows the typology of strong and weak
scalar particles assumed in Crnič (2011); Gast & Van der Auwera (2011) and can be
formalized as lexical items competing for lexical insertion using a pattern in (49).
The scale in (49) uses semantically interpreted features formalized below
but intuitively understandable as follows: even is a feature responsible for the unlikelihood
presupposition in examples like (45). solo is a feature formalizing the
presupposition of likelihood in sentences like (48), where the weak scalar item is
embedded in a Strawson downward entailing environment. The presuppositions
of even and solo are contradictory: the first requires unlikelihood against alternatives,
the second one dictates the likelihood of the prejacent. Their conflict can
be resolved only in environments reversing entailment.
(49) a. ⟨i, byť jediný⟩
b. ⟨[even], [even][solo]⟩
The features even and solo are semantically interpreted in (50) and (51) following
Crnič (2011: 134). The part of both formulas, 𝑝 ◁ 𝐶 𝑞, means that in the
context C (focus induced alternatives) the proposition p is at most as likely as
the proposition q. The mechanism works under the assumption introduced to
linguistics by Lahiri (1998): that there is a one-way relationship between logical
strength and entailment. The relationship is: logically stronger propositions
cannot be more likely than logically weaker propositions. Intuitively from the
conjunction of two propositions It rained, and it hailed any of the propositions
follows ((𝑝∧𝑞) → 𝑝/(𝑝∧𝑞) → 𝑞). The conjunction of the two propositions is logically
stronger and less likely than any of its atomic propositions. More formally
is the relationship between likelihood and entailment stated in (52).
(50) even 𝑔,𝑐(𝐶, 𝑝, 𝑤) is defined only if ∃𝑞 ∈ 𝐶[𝑝 ◁𝑐 𝑞].
If defined, even 𝑔,𝑐(𝐶, 𝑝, 𝑤) = 1 iff 𝑝(𝑤) = 1
(51) solo 𝑔,𝑐(𝐶, 𝑝, 𝑤) is defined only if ∀𝑞 ∈ 𝐶[𝑞 ≠ 𝑝 → 𝑞 ◁𝑐 𝑝].
If defined, solo 𝑔,𝑐(𝐶, 𝑝, 𝑤) = 1 iff 𝑝(𝑤) = 1
xxviii
42 Negation and negative polarity items in Slavic
(52) If 𝑝 → 𝑞 then 𝑝 ◁𝑐 𝑞.
I will now apply Crnič’s (2011) pragmatic framework to Czech with a simple
illustration of possible derivations. First, strong scalar particles as Czech i in an
appropriate context like (47) would obtain a correct LF like (53) (in pseudo-Czech).
The sentence is appropriate in all contexts where drinking ten beers is less likely
then drinking alternative 𝑛 < 10-number of beers. If the context is set up for a
scale 1, … , 10, the presupposition of even is satisfied and i can be spelled out in
the structure as (53).
(53) [even C1] Peter drank ten 𝐹 beers.
Now, let us consider weak scalar particles, for a weak scalar particle in the Strawson
downward entailing context like (48) a possible LF in (54) is obtained. The
presupposition of solo is satisfied: to drink one beer is more likely than to drink
𝑛 > 1 beers. The presupposition of even (with a wider scope than the antecedent
of implication) is satisfied too. Drinking one beer as a cause of dancing is less
likely than drinking 𝑛 > 1 beers as a reason for dancing intuitively. More formally:
as usually, the antecedent of a conditional is Strawson downward entailing,
so the weak prejacent becomes logically strong. If drinking one beer is a sufficient
condition for dancing, then drinking 𝑛 > 1 beers is a sufficient condition
for dancing as well. All the contextual alternatives are entailed by the prejacent
and due to (52) are less likely.
(54) [even C1][[if [solo C0] Peter drinks one 𝐹 beer] he will dance]
The scale in (49) can be interpreted as the following morphological rules in
(55a)/(55b) formalizing both the idea that i bears un-likelihood presupposition,
while byť jediný bears two contradictory presuppositions, and the idea that the
items compete for insertion. But before I tackle the competition part, let us again
consider the details of weak scalar particles formalization. Weak scalar particles
like Czech byť jediný bear both the un-likelihood presupposition of strong i and
the likelihood presupposition formalized as the feature solo. The contradictory
presuppositions of byť jediný explain its limited distribution. While even in (54)
scopes over an entailment reversing operator, solo scopes under it and consequently
both presuppositions can be satisfied. Note as well, that this explains the
ungrammaticality of (56). There is no entailment reversing operator over which
even would scope and the conflicting presuppositions of the two features (solo
and even) clash with each other necessarily. The prejacent cannot be both less
likely and more likely than its alternatives (in any context).
xxix
Mojmír Dočekal
(55) a. [even] ↔ i
b. [even][solo] ↔ byť jediný
(56) # Petr
Petr
včera
yesterday
přečetl
read
byť
even
jedinou
one
knížku.
book
Intended: ‘Petr read just one book yesterday.’
2.3.2.4 Scalar particles and competition The morphological rules in (55a)/(55b)
do not explain the restricted distribution of i on their own (recall that i avoids
weak contexts like (47)). But in the currents state of formalization, we predict
that an ungrammatical sentence like (57a) would get assigned a correct LF like
(57b). Such LF would be principally spellable as a sentence containing i. The LF
(57b) would produce a sensible presupposition of even because it scopes over
the entailment reversing operator (the antecedent of conditional). In such a situation,
the weakest scalar expression entails all the other contextual alternatives
and becomes less likely than the alternatives.
(57) a. # Jestli
if
Petr
Petr
vypije
drink
i
even
jedno
one
pivo,
beer
tak
then
bude
will
tančit.
dance
Intended: ‘If Petr will drink even one beer he will dance.’
b. [even C3] [[if [even C3] Peter drinks one 𝐹 beer] he will dance]
But such a spell out ignores the scale ⟨i, byť jediný⟩ where the elements compete
with each other for insertion. This is quite a frequent situation in linguistics
where competition between elements dictates their distribution. There are many
possible formalizations of the old intuition that only the most specific item can
be inserted and eventually blocks insertion of less specific competitors (called
Panini’s Principle by many, see Booij 2012 among others). Crnič uses Heim’s
(1991) Maximize Presupposition reformulation of the Panini’s Principle. But I
will rely on a neo-Gricean approach essentially following Krifka (1995) and his
approach to NPIs (both approaches are well suited for working with competing
items differing only in terms of non-at-issue meaning). In the case at hand: there
is a competing LF like (58) which is more feature-specific (and would be spelled
out as a sentence containing byť jediný):
(58) [even C3][[if [even C3] [solo] Peter drinks one 𝐹 beer] he will dance]
There is only one more presupposition of (58) that in the LF (57b): it is triggered
by solo in its base position and requires its prejacent (drinking one beer) to be
xxx
42 Negation and negative polarity items in Slavic
more likely than contextual alternatives (drinking 𝑛 > 1 beers). Both presuppositions
(even and solo) of (58) and the even presupposition of (57b) are satisfied
in natural contexts. Moreover, the assertion impact of both LF is the same, so
the structures denote contextually equivalent propositions. Via the usual neoGricean
reasoning: a speaker communicating (57b) would indicate that she is
not in a position to assert (58) but as (58) and (57b) are contextually equivalent,
we arrive to a contradiction caused by the competition of elements for insertion.
But a speaker communicating (58) does not indicate anything about the
non-assertability of (57b), since byť jediný (the spell-out of the features in (58))
is logically stronger and its assertion entails the logically weaker alternative (i).
In simple words: if contextually more feature-specific byť jediný can be inserted,
it blocks the insertion of the feature-poorer i. This doesn’t happen in the case of
strong contexts like (45) though. The solo component of byť jediný would produce
a contradictory presupposition there. Consequently, byť jediný and i do not
compete for insertion in contexts like that.
2.3.3 Scalar particles and negation
Recall that we have distinguished three classes of weak even SP: (i) indiscriminating
(English even), (ii) weak particles requiring non-upward-monotonic environment
(Czech byť jediný or Polish chociaż/choċ/choċ ‘just’), and finally (iii) weak
scalar particles limited to the immediate scope of negation. Crnič formalizes the
distinction between the negation requiring particles (like Czech ani or Polish
ani) and weak particles like byť jediný via a formal uninterpretable feature
(following essentially Zeijlstra 2004 and Penka & Zeijlstra 2005).18 For Czech
the morphological rules and corresponding scale would be as in (59) (similar to
Crnič’s (2011) analysis of Slovenian tudi and niti).
(59) a. [even][solo] ↔ byť jediný
b. [even][solo][uNEG] ↔ ani
c. ⟨byť jediný, ani⟩
This neatly explains the limited occurrence of ani type of scalar particles to (in
the majority of cases) sentences with negated predicates (but see Dočekal & Dotlačil
2016a,c for qualifications, especially concerning neg-raising contexts). Moreover,
it explains some cases of byť jediný incompatibility with negation; Crnič
shows this “avoid negation” constraint in the following Slovenian sentence.
18
Thanks to one of the reviewers for suggesting that Czech and Polish ani and Czech byť jediný/Polish
chociaż/choċ/choċ do not behave differently in this respect.
xxxi
Mojmír Dočekal
(60) # Janez
Janez
ni
not
prebral
read
tudi
even
ENE
one
knjige.
book
‘John did not read even one book.’ (Slovenian; Crnič 2011: 139 )
The direct translation of (60) into Czech sounds distinctly odd, and there are other
cases of byť jediný type of scalar particles which are incompatible with negation
(see Crnič 2011: chap. 5 for German, Spanish and Greek data). The blocking is also
familiar in Slavic competition of indefinites (or free choice items) and neg-words
(see Pereltsvaig 2004). Moreover, the “avoid negation” rule follows the same logic
of competition induced above to explain the blocking of strong scalar particles
by weak ones. This led Crnič to postulate the following typological implicational
relation, which is supposed to be operative in all negative concord languages.
(61) Implicational relation for weak scalar particles
There is a scalar particle that may only be weak and that only occurs in
the immediate scope of negation in the language. → No other weak scalar
particle that may only be weak occurs in the immediate scope of negation
in the language. (after Crnič 2011: 131)
But even if the generalization is linguistically reasonable and supported by the
data above, I have serious doubts about its validity. First, let us look at the empirical
facts. It is easy to find many natural occurrences of both Czech byť jediný
and Slovenian tudi weak scalar particles in negated sentences. Consider (62) and
(63); the Czech examples are from Czech National Corpus (Křen et al. 2016).
(62) a. Nemají
neg.have
byť
even
jediný
one
věšák
stand
na
for
kabáty.
coats
‘They lack even one coat-stand.’
b. Nikomu
neg-person
ze
from
soutěžících
competitors
nepoložil
neg.ask
byť
even
jedinou
one
otázku.
question
‘He didn’t ask any competitor even one question.’
c. Nikdy
neg-time
by
sbjv
se
refl
neodvážily
neg.dare
byť
even
jediným
one
slovem
word
projevit
manifest
svou
their
nespokojenost.
discontent
‘They didn’t dare by even one word to manifest their discontent.’
(Czech)
(63) Ni
neg
izpustil
pass
tudi
even
ene
one
same
single
dirke.
race
‘He didn’t pass even one race.’ (Slovenian; Lanko Marušič, p.c.)
xxxii
42 Negation and negative polarity items in Slavic
Even if it is true that in some configurations negating byť jediný type of weak
scalar particles leads to unacceptability, there is definitely no general ban on
negating weak scalar particles even in negative concord languages like Czech
or Slovenian (against Crnič’s 2011 typological rules as (61)). There are, of course,
many possible explanations why the putative blocking between the items on
scale (59) does not happen in cases like (62) or (63). A proper investigation of this
issue lies beyond the scope of this chapter, though, but I will at least hint at one
probable solution. As observed already by Kadmon & Landman (1993) for certain
usages of any and further scrutinized by Crnič (2011); Alonso-Ovalle (2016) for
weak even carries in many contexts peculiar semantics which is usually termed
settle for less (see also section §2.2.1).
Let me demonstrate the semantics on example in (64a). The example has the
following meaning ingredients: (i) the speaker of the sentence would prefer to
buy more than one ticket; (ii) but he settled for one ticket. Slavic b-NPIs like
Czech byť jediný (or Slovenian tudi) can have a settle-for-less reading, at least in
some contexts. But this is not a general feature of weak scalar particles; compare
(64b) with a-NPI ani (64b), which is compatible with a speaker who did not care
about how many tickets he bought at the end, ani would be strange in a settlefor-less
context. Consequently, at least some Slavic languages with b-NPIs seem
to carry the settle-for-less semantics, unlike weak a-NPIs.19
(64) a. Nakonec
finally
jsem
aux
nekoupil
neg.bought
byť
even
jediný
one
lístek.
ticket
‘At the end, I didn’t buy even one ticket.’
b. Nakonec
finally
jsem
aux
nekoupil
neg.bought
ani
even
jeden
one
lístek.
ticket
‘At the end, I bought not even one ticket.’ (Czech)
2.3.4 Short note on concessive scalar particles
In the previous section, we observed that Slavic b-NPIs (unlike a-NPIs) could
carry the settle-for-less meaning. The main idea of explaining the lack of block-
19
One of the anonymous reviewers notices that if byť jediný in examples like (62) can be substituted
with the a-NPI ani without any meaning difference, the competition story offered in
§2.3.4 would be weakened. Intuitions of Czech native speakers I consulted support the subtle
difference discussed in this section, although (for obvious reasons) it is not easy to demonstrate
them on (62). Nevertheless, the difference can be made more visible in the following scenario:
imagine a man who always wanted to stay a bachelor without children; in such a context, he
can say Vyšlo to, nemám ani jedno dítě ‘It worked. I don’t have even one child’ (a-NPI) but to
say Vyšlo to, nemám byť jediné dítě ‘It worked. I don’t have even one child’ (b-NPI) is conceived
by native speakers as incoherent.
xxxiii
Mojmír Dočekal
ing between ani and byť jediný under negation would be the following: it seems
reasonable that b-NPIs (byť jediný) can spell out a semantics unavailable for aNPIs
(ani); if they do, like in (64a), they are not less specific than a-NPIs, and
the usual blocking is gone. In cases like (60) used by Crnič, there is no plausible
settle-for-less semantic component, which would be spelled out by b-NPIs and
blocking kicks in. In other words, I claim that weak b-NPIs sometimes behave
like concessive scalar particles. But to get this idea working, we have to make a
short detour to proper concessive particle territory.
Concessive scalar particles are expressions like Greek esto, Spanish aunque
sea, siquiera, Slovenian magari and Czech alespoň. They occur in downward entailing
contexts, in questions, and in some modal contexts. Moreover, there is a
partial overlap of concessive scalar particles with weak scalar particles.20 Consider
desire predicates like want in Czech (65a). Note, that in both items (byť
jedenkrát and alespoň jedenkrát) are glossed with at least (contrast this with (48)
where byť jediný was glossed as even one). And finally, both weak and concessive
scalar particles are ungrammatical under doxastic/epistemic predicates as think,
know: in (65b) and in the grammatical sentence (65a) (in the scope of the desire
predicate) their associate is low on the pragmatic scale (this is a necessary, not
sufficient condition though: it is low even in the example (65b), but the example
is still ungrammatical).21
20
As one of the anonymous reviewers notes, there are approaches to sub-types of NPIs that
resemble the reasoning in the current section. One such example is Rullmann (1996), where
two different types of Dutch NPIs are described via two different theoretical approaches. The
first via the scalar theory of any from Lee & Horn (1994) and the second via the non-scalar,
domain-widening approach to any from Kadmon & Landman (1993). Despite some common
properties, the approach in this section is different from Rullmann (1996) since both concessive
and weak scalar particles are described as scalar.
21
Both weak scalar particles and concessive particles are grammatical as well in the scope of
desire predicates and positive emotive factives like in (i). And as one of the anonymous reviewers
correctly remarks, this is a non-trivial problem since glad is not SDE and desire predicates
like want are Strawson upward entailing (see von Fintel 1999). Crnič (2011: section 4.2.5)
offers a technical solution to this problem which (in bare essentials) rests on the free choice
interpretation of concessive scalar particles and its exhaustification interpretation leading to a
non-entailing set of alternatives. Discussing Crnič’s solution would lead us too far away from
the goals of the current chapter, but all the details of Crnič’s analysis should carry on to the
cases of Czech concessive particles in contexts like (i).
(i) Petr
Petr
je
is
rád,
glad
že
that
{byť
even
jedenkrát
once
/ alespoň
at.least
jednou}
once
navštívil
visited
Grónsko.
Greenland
‘Petr is glad that he visited Greenland at least once.’
xxxiv
42 Negation and negative polarity items in Slavic
(65) a. Petr
Petr
chce,
wants
aby
that
Karel
Karel
navštívil
visited
Grónsko
Greenland
{byť
even
jedenkrát
once
/ alespoň
at.least
jedenkrát}.
once
‘Petr wants Karel to visit Greenland at least once.’
b. * Petr
Petr
ví,
knows
že
that
{byť
even
jedenkrát
once
/ alespoň
at.least
jednou}
once
navštívil
visited
Grónsko.
Greenland
Intended: ‘Petr knows that he visited Greenland at least once.’
So far, we have discussed only the properties which are shared both by weak
and concessive scalar particles. But there is an obvious difference between them:
concessive scalar particles are additive, so they contribute a meaning of multiplicity,
unlike the weak scalar particles (even if the additive presupposition can
be masked in some contexts). The difference between them can be clearly observed
in some contexts where concessive scalar particles are licensed but weak
particles are not, see (66). The core meaning of (66) is the necessary condition:
obtaining 𝑛 ≥ 1 scout badges is sine qua non for participating in a scout camp.
This means that earning 2, 3, or more scout badges is even more positive; the
additive component stems from the concessive scalar particle alespoň.
(66) (Abys
in.order.to
mohl
be.able
jet
participate
na
in
skautský
scout
tábor),
camp
musíš
have.to
získat
obtain
{#byť
even
jediného
one
/ alespoň
at.least
jednoho}
one
bobříka.
beaver
‘(For you to be able to participate in a scout camp) you have to obtain
even one / at least one scout badge.’
Based on the observations above, I postulate that Slavic b-NPIs (Slavic byť jediný,
Slovenian tudi) can realize some of the concessive scalar particles’ meaning.
More specifically, I will use the feature formalization of concessive particles from
Crnič (2011: 109), below as (67). There are two components of [at-least]: (i) the
familiar presupposition part: the prejacent has to be more likely than all its alternatives
(this part is shared with the presupposition of solo, see (51)); (ii) the
assertive meaning (different both from even and solo since both even and solo
are vacuous in at-issue meaning; see (50) and (51)): there is at least one alternative
to the prejacent which is at most likely and is true (can be the prejacent itself).
Let us see the working of the formalization demonstrated for adversative predicate
regret in (68). For Czech alespoň, I postulate a morphological rule in (69).
xxxv
Mojmír Dočekal
And because both presuppositions of (68) are satisfied in natural contexts: (i) the
scalar presupposition triggered by at-least: it is more likely to visit Greenland
once than 𝑛 > 1 times; (ii) the scalar presupposition of even: because regret is
SDE, the weak associate becomes strongest (if somebody regrets that he visited
Greenland once, he regrets that he visited it 𝑛 > 1 times); alespoň can be inserted.
The additive multiplicity part of the at-issue contribution is in the scope of the adversative
predicate, so correctly not projecting, unlike the presupposition part of
at-least. But recall that (68) is LF of (65a) where both alespoň and byť jedenkrát
were acceptable with the same interpretation. This then shows that, at least in
some cases, the b-NPI can spell out features like (69).
(67) at-least 𝑔,𝐶
= 𝜆𝐶.𝜆𝑝 ∶ ∀𝑞 ∈ 𝐶[𝑝 ≠ 𝑞 → 𝑞 ◁ 𝐶 𝑝].𝜆𝑤.∃𝑞 ∈ 𝐶[𝑞 ◁ 𝐶 𝑝 ∧ 𝑞(𝑤) = 1]
Crnič (2011: 109)
(68) [even C2][Petr regrets [that [at-least C0] he visited Greenland once 𝐹 ]]
(69) [even][at-least] ↔ alespoň
And if b-NPIs can spell out at-least features, we get the formal implementation
of the idea from the beginning of the current section. Consider modified morphological
rule for byť jediný in (70a) and repeated morphological rule for ani
in (70b). In terms of features, the competition between the two items is then in
equilibrium. Both items realize the same number of features; neither of them is
more specific, stronger, and blocking the other. That explains the occurrences
of b-NPIs under negation in examples like (62) – when b-NPIs realize the same
number of features as a-NPIs, they can be inserted even in the immediate scope
of negation. In cases like (60), the blocking works as proposed by Crnič (2011:
chap. 5). Of course, the proper scrutiny of the proposed leak in the generalization
(61) is beyond the scope of this chapter, but I hope to find more supporting
data in future work.
(70) a. [even][solo][at-least] ↔ byť jediný
b. [even][solo][uNEG] ↔ ani
The empirical claims concerning Czech scalar particles and the features they realize
are summarized in Table 1. This section has introduced the pragmatic mechanism
of NPI licensing. It followed the seminal work of Krifka (1995) where the
licensing of NPIs was rethought from the standard semantic downward entailing
explanation to a pragmatic theory. In particular, the description of Czech
xxxvi
42 Negation and negative polarity items in Slavic
scalar particles was formalized in terms of even-based theory of NPI licensing,
following Crnič (2011) (where more details and also references to other literature
can be found). As was stated at the beginning of this section, the pragmatic turn,
where the environments licensing NPIs are basically characterized via the logical
requirements on the focus alternatives for the prejacent, can be cashed out
in different frameworks, one widely adopted being Chierchia (2013). See see also
gajic2016coordination for its application to Slavic data. The pragmatic explanation
should be viewed as a continuation of the semantic theories, with which it
shares the core idea that NPI licensing is a matter of semantic strength (or, entailment
patterns; see (52)). Moreover, the pragmatic theory offers more flexibility
and also a handle on non-monotonic licensing of NPIs (see again Crnič 2011 for
details).
Table 1: Table of features for Czech scalar particles
Czech expression features
i [even]
byť jediný [even][solo]
byť jediný [even][solo][at-least]
alespoň [even][at-least]
ani [even][solo][uNEG]
3 Topics related to negation (but not discussed in this
chapter)
The previous sections delivered a story about NPIs and their treatment by (mostly)
Slavic formal semanticists. The phenomenon of NPIs is a very active and massive
topic in the current formal semantics. Consequently, even the summary of their
linguistic history in Slavic formal semantics was not short. Despite that, due to
limitations of space or scope, some significant Slavic contributions to polarity
and negation were unable to be incorporated into this summary. It is acknowledged
that these works would enhance the comprehensiveness of the current
Handbook. But instead of summarizing them, let me at least enumerate them and
shortly classify their areas of interest. But after I list the topics and contributions
which are more or less directly related to negation, I will dedicate a special part
xxxvii
Mojmír Dočekal
to the genitive of negation. The genitive of negation has a special place among
the topics related to negation for many good reasons; the most important is the
substantial one: in the Slavic genitive of negation, many core semantic and pragmatic
mechanisms (as shifts of meanings, presuppositions, semantic bleaching,
etc.) are manifested. The second important reason is historic: genitive of negation
is in the center of both west and east Slavicists since Jakobson 1936; Babby
1980; Pesetsky 1982 to Padučeva 1997; 2006 and a recent semantic approach of
Borschev & Partee 2002a,b,c, Partee et al. 2011; Borschev et al. 2010 which successfully
tries to reconcile both stands of research and came up with a formal
semantic approach to the phenomenon. I will rely heavily on this synthetic approach
for its merits and also because it focuses on the semantic and pragmatic
part of the genitive of negation debate. But first, let us list the topics which are
not directly connected with the genitive of negation.
First, very close to the polarity licensing is the description of free choice items,
their licensing, properties, and interaction with negation – works discussing
Slavic data in this direction include (among others) the following ones: Pereltsvaig
(2004), Yanovich (2005), Pereltsvaig (2008), Błaszczak (2008), Dočekal & Strachoňová
(2014a), Denic (2015), Strachoňová (2017).
Next, there is an old and influential tradition describing negation and its interaction
with information structure; see Hajičová (1974) as a classic reference, and
other older works are referenced there. For a newer and partially corpus/semantic
oriented work, see Dočekal & Strachoňová (2014b).
Further, there is a couple of articles discussing the interaction of negation with
various types of Slavic conjunctions, see Abels (2005), Jasinskaja & Zeevat (2009),
Arsenijević (2011), Tiskin (2017).
In recent years there also appeared experimental studies, focussing on Negraising
and scalar particles, such as Dočekal & Dotlačil (2016b), Dočekal & Dotlačil
(2016d).
Lastly, let me point out many papers discussing various phenomena at syntax/semantics
interface like modals, weak islands, constituent negation, and others,
see Veselovská (1995), Bailyn (1997), Billings (1997), Przepiórkowski & Kupść
(1997), Brown (1999), Przepiórkowski (1999), Kosta (2001), Perissutti (2003), Borschev
et al. (2006), Dočekal & Kučerová (2013), Yanovich (2013), Gruet-Skrabalova (2013),
Dočekal (2017).
Genitive of negation The alternation between structural cases (nominative
and accusative) and genitive is grammaticalized in Slavic and Baltic languages.
The alternation is diachronically changing; it is optional in contemporary Russian,
obligatory in Polish, and practically non-existent in modern Czech, although
xxxviii
42 Negation and negative polarity items in Slavic
it was obligatory in Old Czech at least until 18th century (Lamprecht et al. 1986).
I will focus here (as many other researchers) on subject genitive of negation,
demonstrated with (71a)/(71b) from Partee et al. 2011: 136 and object genitive of
negation as in (72a)/(72b) from Partee et al. 2011: 137. There are various related
genitive alternations, like genitive of intensionality and partitive genitive, but
discussing them is clearly beyond the scope of the current chapter (but see Kagan
2007).
(71) a. Otvet
Answer-nom.m.sg
iz
from
polka
regiment
ne
neg
prišel.
arrived-m.sg
‘The answer from the regiment has not arrived.’
b. Otveta
Answer-gen.m.sg
iz
from
polka
regiment
ne
neg
prišlo.
arrived-n.sg
‘The answer from the regiment has not arrived.’
(72) a. Ja
I
ne
neg
zametil
noticed
vodku
vodka-acc
na
on
stole.
table
‘I didn’t notice the vodka on the table.’ (presuppositional: vodka was
there)
b. Ja
I
ne
neg
zametil
noticed
vodki
vodka-gen
na
on
stole.
table
‘I didn’t notice any vodka on the table.’ (non-presuppositional: maybe
there was none)
In both subject and object genitive of negation, the negated sentence allows
the Nom-Acc/Gen alternations, which are triggered by semantic and pragmatic
factors. The traditional description following the original Jakobson’s insight is
(for subject GenNeg) Babby 1980, who claims that the Nom is outside the scope of
negation while Gen has a narrow scope with respect to negation. He furthermore
correlates this scope reasoning with the Topic/Focus structure and claims that
the Nom subject is a topic (outside of the negation scope) while Gen is inside the
negation scope (and with an empty topic). Borschev et al. 2010: ex.(6)-(7) offer
conclusive evidence that it is not the Topic/Focus articulation that is the decisive
factor of GenNeg since it is easy to find topics realized as GenNeg, which goes
against the Jakobson/Babby information structure hypothesis.
While Partee & Borschev 2002; Borschev et al. 2010; Partee et al. 2011 disagree
with the particular realization of the idea, they follow the main intuition of the
previous research, namely that GenNeg is a pragmatically triggered alternation,
they come with a new name for it Perspectival Structure. For them, the choice
xxxix
Mojmír Dočekal
of the case signals the perspectival center: in (72a), the perspectival center is
vodka, but in (72b) genitive signals a demotion of vodka and the perspective center
becomes the location (the table). They furthermore claim that the perspectival
center is presupposed to exist (relative to a location). The demoted genitive
is then not a center, and they cache this observation as a type shifting of the Gen
NP. The type shift involves the shift of the argumental NP (of type ⟨𝑒⟩ if it is Acc
or Nom) to the predicative type ⟨𝑒, 𝑡⟩ (see also Kagan 2007). The shift of the NP
is accompanied by the shift of its verb. Let me demonstrate their reasoning with
an example toy derivation of (parts) of (72a)/(72b) truth conditions. My formalization,
of course, follows the spirit of Partee et al. 2011; Borschev et al. 2010, but
I simplify and use the 𝜎 definite determiner to emulate their presupposition idea.
(73) contains the truth-conditions for the Acc marked object which promotes
vodka as the presupposed perspective center (𝜎-bound property in (73b) which
results in type ⟨𝑒⟩), the verb in (73a) remains at the basic ⟨𝑒⟩ type function with
respect to its object.
(73) a. notice1 = 𝜆𝑙𝜆𝑥𝜆𝑦[notice(𝑥, 𝑦, 𝑙)] …x notices y at location l
b. vodka 𝐴𝑐𝑐 = 𝜎𝑧.vodka(𝑧) …⟨𝑒⟩ type
c. notice1( vodka 𝐴𝑐𝑐 ) = 𝜆𝑙𝜆𝑥[notice(𝑥, 𝜎 𝑧.vodka(𝑧), 𝑙)]
But in genitive version, (74b), vodka becomes a property of type ⟨𝑒, 𝑡⟩, the
verb is type-shifted (Borschev et al. 2010 speculate about semantic incorporation),
(74a), and the resulting truth-conditions are in (74c). The negation would
semantically scope wider than all the arguments, which is in accordance with the
syntactic observation that Slavic verbal negation has wide scope over the whole
sentence since it licenses neg-words in all argument and adjunct positions. The
apparent wide scope of Acc/Nom with respect to negation then results from the
presupposition (the perspectival) center. This seems like a reasonable and empirically
well-motivated explanation using minimum ad-hoc stipulations.
(74) a. notice2 = 𝜆𝑙𝜆𝑥𝜆𝑃[∃𝑦(𝑃(𝑦) ∧ notice(𝑥, 𝑦, 𝑙))] …x notices a property
P at location l
b. vodka 𝐺𝑒𝑛 = 𝜆𝑧.vodka(𝑧) …⟨𝑒, 𝑡⟩ type
c. notice2( vodka 𝐺𝑒𝑛 ) = 𝜆𝑙𝜆𝑥[∃𝑦(vodka(𝑦) ∧ notice(𝑥, 𝑦, 𝑙))]
This was a core claim of the research, including at least Borschev & Partee
2002a,b,c, Partee et al. 2011; Borschev et al. 2010 concerning GenNeg. It derives
nicely (and by using independently widely accepted mechanisms) the crucial
properties of GenNeg: the genitive marked subject or object is not presupposed
xl
42 Negation and negative polarity items in Slavic
and is semantically and pragmatically demoted. Furthermore, it explains some
other properties of the whole construction: the verbal semantics is bleached in
GenNeg (in many cases close to the presentational sentences) because of the typeshift
of the verb. There (of course) remain some partially open issues discussed at
length in Partee et al. 2011; Borschev et al. 2010 such as the relationship between
subject GenNeg and object GenNeg, various degrees of optionality, the relation
of GenNeg to other genitive alternations, etc. But be it as it may, the semantic
theory of Partee and colleagues is an excellent example of a careful data analysis
combined with a very reasonable semantic/pragmatic approach that illuminates
the complicated data patterns with up-to-date formal linguistic tools.
4 Summary
In this chapter, I summarized syntactic, semantic, and pragmatic theories of (not
only) Slavic negative polarity items and scalar particles. I tried to summarize the
history of ideas in formal Slavistics pertaining to the description and explanation
of Negative Polarity Items (and scalar particles). The chapter has shown how a
better understanding of Slavic polarity-dependent expressions improves our understanding
of the division of labor between syntax, semantics, and pragmatics.
Abbreviations
neg negation
not negation
npi Negative Polarity Item
acc Accusative
fut futurum
de downward entailing
comp complementize
even unlikelihood scalar presup-
position
at-least scalar inference (also) in at-
issue
SE clitic se
sg singular
neg-body neg-word for persons
Q question particle
sde Strawson downward entail-
ing
subj subjunctive
aux auxiliary
solo likelihood scalar presuppo-
sition
𝑢 𝑁 𝐸𝐺 uninterpretatble syntactic
negative feature
Acknowledgements
The work on this chapter benefited from discussions and correspondence with
many people. First, I would like to thank the three anonymous reviewers. Their
xli
Mojmír Dočekal
comments and hints helped the chapter to be more thorough and insightful. Next,
I would like to thank people who read my chapter and gave me comments and
ideas, namely Hana Strachoňová and Iveta Šafratová. Last, I am very grateful to
Berit and Radek, the editors, for their patience and help.
References
Abels, Klaus. 2005. ’Expletive Negation’ in Russian: A Conspiracy Theory. Journal
of Slavic Linguistics 13. 5–74.
Agha, Omar & Paloma Jeretič. 2022. Weak necessity modals as homogeneous
pluralities of worlds. In Semantics and linguistic theory, 831–851.
Alonso-Ovalle, Luis. 2016. Are All Concessive Scalar Particles the Same? Probing
into Spanish” Siquiera”. In Semantics and linguistic theory, vol. 26, 185–204.
DOI: 10.3765/salt.v26i0.3820.
Arsenijević, Boban. 2011. Serbo-Croatian coordinative conjunctions at the syntaxsemantics
interface. The Linguistic Review 28(2). 175–206. DOI: 10.1515/tlir.2011.
005.
Babby, Leonard Harvey. 1980. Existential sentences and negation in Russian. Ann
Arbor: Caroma Publ. DOI: 10.2307/413563.
Bailyn, John F. 1997. Genitive of negation is obligatory. In Annual Workshop on
Formal Approaches to Slavic Linguistics: The Cornell Meeting 1995.[= FASL 4],
84–114.
Bernini, Giuliano & Paolo Ramat. 2012. Negative Sentences in the Languages
of Europe: A Typological Approach. In B. Comries G. Bossong (ed.), Negative
Sentences in the Languages of Europe. Berlin: De Gruyter Mouton. DOI: 10.1515/
9783110819748.
Billings, Loren. 1997. Negated prepositional phrases in Slavic. In E. Wayles Browne
et al. (ed.), Formal approaches to slavic linguistics 4, vol. 4, 115–134. Ann Arbor:
Michigan Slavic Publications.
Błasczak, Joanna. 2001. Investigation into the interaction between the indefinites
and negation. vol. 51: Studia Grammatica. Berlin: Akademie Verlag.
Błaszczak, Joanna. 2003. Negative Polarity items in Polish: Towards a uniform account.
In C. Maeinborn (ed.), (a)symmetrien – (a)symmetries, 25–56. Tübingen:
Stauffenburg.
Błaszczak, Joanna. 2008. The puzzle of kolwiek-pronouns in Polish. In Workshop
on free choiceness: facts, models, and problems, 3–12.
Booij, Geert. 2012. The grammar of words: An introduction to linguistic morphology.
Oxford: Oxford University Press.
xlii
42 Negation and negative polarity items in Slavic
Borschev, Vladimir, Elena V. Padučeva, Barbara H. Partee, Yakov Testelets & Igor
Yanovich. 2010. On Semantic Bleaching and Compositionality: Subtraction or
Addition? (On the bleaching of ”lexical verbs” in Russian negated existential
sentences). In Yehuda N. Falk (ed.), Proceedings of Israel Association for Theoretical
Linguistics. Jerusalem: Bar Ilan University.
Borschev, Vladimir, Elena V Paducheva, Barbara H Partee, Yakov G Testelets
& Igor Yanovich. 2006. Sentential and constituent negation in Russian BEsentences
revisited. In Formal approaches to slavic linguistics: the princeton
meeting 2005 (fasl 14), 50–65.
Borschev, Vladimir & Barbara H. Partee. 2002a. O semantike bytijnyx predloženij
(On the semantics of existential sentences). Semiotika i informatika 37. 59–78.
Borschev, Vladimir & Barbara H. Partee. 2002b. The Russian genitive of negation
in existential sentences. Prague linguistic circle papers 4. Publisher: John
Benjamins Publishing, 185–250.
Borschev, Vladimir & Barbara H. Partee. 2002c. The Russian genitive of negation:
Theme-rheme structure or perspective structure? Journal of Slavic Linguistics
10. Publisher: JSTOR, 105–144.
Brown, Sue. 1999. The syntax of negation in Russian: A minimalist approach. Bloomington:
Indiana University. (Doctoral dissertation). DOI: 10.2307/417189.
Büring, Daniel. 2005. Binding theory. Cambridge: Cambridge University Press.
Carlson, Greg N. 1980. Polarity any is existential. Linguistic Inquiry 11(4). 799–
804.
Chierchia, Gennaro. 2013. Logic in grammar: Polarity, free choice, and intervention.
Oxford: Oxford University Press. DOI: 10.1093/acprof:oso/9780199697977.001.
0001.
Coppock, Elizabeth & Lucas Champollion. 2024. Invitation to formal semantics.
https://eecoppock.info/bootcamp/semantics-boot-camp.pdf.
Crnič, Luka. 2011. Getting even. Cambridge, MA: Massachusetts Institute of Technology.
(Doctoral dissertation).
Dayal, Veneeta. 1998. Any as inherently modal. Linguistics and Philosophy 21(5).
433–476.
Denic, Milica. 2015. Minimizing Scope Ambiguity Hypothesis. In Proceedings of
esslli 2015 student session (2015), 54–65.
Dočekal, Mojmír. 2017. Upper bounded and un-bounded ‘no more’. eng. In Eva
Dekany (ed.), Acta Linguistica Academica, 213–231. Budapest: Akadémiai Kiadó.
DOI: 10.1556/2062.2017.64.2.3.
Dočekal, Mojmír. 2019. N-words and NPIs: between syntax, semantics and experiments.
In Petra Mišmaš Marušič Franc & Rok Žaucer (eds.), Formal description
of slavic languages 12.5, 15–51. Berlin: Language SciencePress.
xliii
Mojmír Dočekal
Dočekal, Mojmír & Jakub Dotlačil. 2016a. Experimental evidence for Neg-Raising
in Slavic. Linguistica 56(1). 93–109. DOI: 10.4312/linguistica.56.1.93-109.
Dočekal, Mojmír & Jakub Dotlačil. 2016b. Experimental evidence for Neg-Raising
in Slavic. eng. Linguistica 56(1). 93–109. DOI: 10.4312/linguistica.56.1.93-109.
Dočekal, Mojmír & Jakub Dotlačil. 2016c. When is not-believing believing that
not? In Robert Truswell, Chris Cummins, Caroline Heycock, Brian Rabern &
Hannah Rohde (eds.), Sinn und bedeutung, edinburgh. Edinburgh: University
of Edinburgh.
When is not-believing believing that not? 2016d. eng.
Dočekal, Mojmír & Ivona Kučerová. 2013. On Degree Questions, Scope of Negation,
and the Lack of Weak Islands in Czech. eng. In Steven et al Franks (ed.),
Annual workshop on formal approaches to slavic linguistics, 49–63. Ann Arbor:
Michigan Slavic Publications.
Dočekal, Mojmír & Iveta Šafratová. 2019. Even Hypothesis of PPIs Licensing :
An Experimental Study. eng. In Joseph Emonds, Markéta Janebová & Ludmila
Veselovská (eds.), Language Use and Linguistic Structure. Proceedings of
the Olomouc Linguistics Colloquium 2018, 253–273. Olomouc: Palacký University.
https://anglistika.upol.cz/olinco2018proceedings/.
Dočekal, Mojmír & Hana Strachoňová. 2014a. Don’t Scope Your Universal Quantifier
over Negation! eng. In Cassandra Chapman, Olena Kit & Ivona Kučerová
(eds.), Annual Workshop on Formal Approaches to Slavic Linguistics. The Mc Master
Meeting 2013, 67–88. Ann Arbor: Michigan Slavic Publications.
Dočekal, Mojmír & Hana Strachoňová. 2014b. Freedom to choose alternatives.
eng. In G. Zybatow, P. Biskup, M. Guhl, C. Hurtig, O. Mueller-Reichau & M.
Yastrebova (eds.), Slavic grammar from a formal perspective, 131–146. Frankfurt
am Main: Peter Lang.
Fauconnier, Gilles. 1978. Implication reversal in a natural language. In Formal
semantics and pragmatics for natural languages, 289–301. Springer. DOI: 10 .
1007/978-94-009-9775-2_10.
Fauconnier, Gilles. 1980. Pragmatic entailment and questions. In Speech act theory
and pragmatics, 57–69. Springer. DOI: 10.1007/978-94-009-8964-1_3.
Gajewski, Jon. 2016. Another Look at NPIs in Definite Descriptions: An Experimental
Approach. en. In Pierre Larrivée & Chungmin Lee (eds.), Negation and
Polarity: Experimental Perspectives (Language, Cognition, and Mind), 307–327.
Cham: Springer International Publishing. DOI: 10.1007/978-3-319-17464-8_13.
https://doi.org/10.1007/978-3-319-17464-8_13 (1 July, 2020).
Gajewski, Jon R. 2011. Licensing strong NPIs. Natural Language Semantics 19(2).
109–148. DOI: 10.1007/s11050-010-9067-1.
xliv
42 Negation and negative polarity items in Slavic
Gajewski, Jon Robert. 2005. Neg-raising: Polarity and presupposition. Cambridge,
MA: Massachusetts Institute of Technology. (PhD Thesis).
Gajewski, Jon Robert. 2007. Neg-Raising and Polarity. en. Linguistics and Philosophy
30(3). Company: Springer Distributor: Springer Institution: Springer
Label: Springer Number: 3 Publisher: Springer Netherlands, 289–328. DOI: 10.
1007/s10988-007-9020-z. https://link.springer.com/article/10.1007/s10988-007-
9020-z (21 July, 2020).
Gajić, Jovana. 2022. Negative coordination. Goettingen: University of Göttingen.
(PhD Thesis).
Gast, Volker & Johan Van der Auwera. 2011. Scalar additive operators in the languages
of Europe. Language 87(1). 2–54. DOI: 10.1353/lan.2011.0008.
Giannakidou, Anastasia. 1997. The landscape of polarity items. Groningen: Rijk-
suniversiteit.
Giannakidou, Anastasia & Hedde Zeijlstra. 2017. The landscape of negative dependencies:
negative concord and n-words. In Martin Everaert & Henk van
Riemsdijk (eds.), The wiley blackwell companion to syntax, 1–38. Oxford: John
Wiley & Sons, Inc. DOI: 10.1002/9781118358733.wbsyncom102.
Gruet-Skrabalova, Hana. 2013. Verbs and particles in minimal answers to yes-no
questions in Czech. In Formal description of slavic languages 10, 197–215. DOI:
10.3726/978-3-653-05335-7/22.
Hajičová, Eva. 1974. Negace a presupozice ve významové stavbě věty [Negation and
presupposition in the meaning structure of a sentence]. Prague: Academia.
Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Clarendon Press.
Heim, Irene. 1984. A note on negative polarity and downward entailingness. In
Proceedings of NELS, vol. 14, 98–107.
Heim, Irene. 1991. Articles and Definiteness Definiteness. In A. V. Stechow &
D. Wunderlich (eds.), Semantics: an international handbook of contemporary
research, 487–535. Berlin: De Gruyter.
Hoeksema, Jack. 1996. Progovac, Ljiljana. Negative and Positive Polarity: A Binding
Approach. Studies in Language 20(1). 196–207.
Horn, Laurence. 1989. A natural history of negation. Chicago: University of Chicago
Press.
Horn, Laurence R. 1972. On the semantic properties of logical operators in English.
Los Angeles: UCLA. (Doctoral dissertation).
Horn, Laurence R. & Young-Suk Lee. 1995. Progovac on polarity Review of Progovac’s
(1994) book. Journal of Linguistics (31). 401–424.
Israel, M. 1996. Polarity sensitivity as lexical semantics. en. Linguistics and Philosophy
19(6). 619–666. DOI: 10.1007/bf00632710. http://link.springer.com/10.
1007/BF00632710 (13 July, 2020).
xlv
Mojmír Dočekal
Jakobson, Roman. 1936. Beitrag zur allgemeinen Kasuslehre. Gesamtbedeutingen
der russischen Kasus. Travaux du CercleLinguistique de Prague 6. 240–88. DOI:
10.1515/9783110873269.23.
Jasinskaja, Katja & Henk Zeevat. 2009. Explaining conjunction systems: Russian,
English, German. In Proceedings of sinn und bedeutung, vol. 13, 231–245.
Kadmon, Nirit & Fred Landman. 1993. Any. Linguistics and Philosophy 16(4). 353–
422. DOI: 10.1007/bf00985272.
Kagan, Olga. 2007. On the Semantics of Structural Case. Jerusalem: Hebrew University.
(PhD Thesis).
Kilgarriff, Adam, Vít Baisa, Jan Bušta, Miloš Jakubíček, Vojtěch Kovář, Jan Michelfeit,
Pavel Rychlý & Vít Suchomel. 2014. The sketch engine: ten years on. Lexicography
1. 7–36.
Klima, Edward. 1964. Negation in English. The structure of language, ed. by JA
Fodor and JJ Katz, 246-323. In J. Fodor & J. Katz (eds.), Thestructure of language,
246–323. Englewood Cliffs, NJ: Prentice-Hall.
Kosta, Peter. 2001. Negace a větná struktura v češtině. In Petr Karlík & Hladká
Zdeňka (eds.), Univerzália a specifika, vol. 3, 117–138. Brno: MUNI press.
Křen, M, V Cvrček, T Čapka, A Čermáková, M Hnátková, L Chlumská, T Jelínek,
D Kováříková, V Petkevič, P Procházka, H Skoumalová, M Škrabal, P Truneček,
P Vondřička & A Zasina. 2016. In Proceedings of the tenth international conference
on language resources and evaluation (lrec’16), 2522–2528. Portorož: ELRA.
Krifka, Manfred. 1995. The semantics and pragmatics of polarity items. Linguistic
analysis 25(3-4). 209–257.
Ladusaw, William. 1979. Polarity Sensitivity as Inherent Scope Relations, Texas:
University of Texas at Austin. (Doctoral dissertation).
Ladusaw, William A. 1992. Expressing negation. In Chris Barker & David Dowty
(eds.), Semantics and linguistic theory 2, 237–260. Ohio State University. DOI:
10.3765/salt.v2i0.3030.
Lahiri, Utpal. 1998. Focus and negative polarity in Hindi. Natural Language Semantics
6(1). 57–123. DOI: 10.1023/A:1008211808250.
Laka, Itziar. 1990. Negation in syntax–on the nature of functional categories and
projections. Cambridge, MA: Massachusetts Institute of Technology. (Doctoral
dissertation).
Lamprecht, Arnošt, Dušan Šlosar & Jaroslav Bauer. 1986. Historická mluvnice
češtiny. Praha: Státní pedagogické nakl.
Lee, Young-Suk & Laurence Horn. 1994. Any as indefinite plus even. Unpublished
Ms. Yale Univerity.
Linebarger, Marcia C. 1987. Negative polarity and grammatical representation.
Linguistics and Philosophy 10(3). 325–387. DOI: 10.1007/bf00584131.
xlvi
42 Negation and negative polarity items in Slavic
Padučeva, Elena V. 1997. Roditel’nyj subjekta v otricatel’nom predloženii: sintaksis
ili semantika?[Genitive subject in a negative sentence: syntax or semantics?]
Voprosy jazykoznanija 2. 101–116.
Padučeva, Elena V. 2006. Genitiv dopolnenija v otricatel’nom predloženii. Voprosy
jazykoznanija 6. 21–43.
Partee, Barbara H & Vladimir Borschev. 2002. Genitive of negation and scope
of negation in Russian existential sentences. In Annual Workshop on Formal
Approaches to Slavic Linguistics: the Second Ann Arbor Meeting 2001.
Partee, Barbara H., Vladimir Borschev, Elena V. Paducheva, Yakov Testelets &
Igor Yanovich. 2011. Russian Genitive of Negation Alternations: The Role of
Verb Semantics. en. Scando-Slavica 57(2). 135–159. DOI: 10.1080/00806765.2011.
631775. http://www.tandfonline.com/doi/abs/10.1080/00806765.2011.631775
(9 July, 2020).
Penka, Doris & Hedde Zeijlstra. 2005. Negative indefinites in Dutch and German.
In 20th comparative germanic syntax workshop, tilburg.
Pereltsvaig, Asya. 2004. Negative polarity items in Russian and the ’bagel problem’.
In Adam Przepiorkowski & Sue Brown (eds.), Negation in slavic, 153–17.
Bloomington: Slavica Publishers.
Pereltsvaig, Asya. 2008. Russian nibud’-series as markers of co-variation. In Proceedings
of the 27th West Coast Conference on Formal Linguistics, 370–378.
Perissutti, Anna Maria. 2003. Determinátory neurčitosti v češtině. Napoli: Univ.
degli Studi di Napoli” L’Orientale”.
Pesetsky, David Michael. 1982. Paths and categories. Cambridge, MA: Massachusetts
Institute of Technology. (PhD Thesis).
Portner, Paul. 2005. What is meaning?: Fundamentals of Formal Semantics. Oxford:
Blackwell.
Progovac, Ljiliana. 1994. Negative and positive polarity: A binding approach. Vol. 68.
Oxford: Cambridge University Press. DOI: 10.1017/cbo9780511554308.
Progovac, Ljiliana. 2005. Negative and positive polarity: A binding approach. Vol. 68.
Cambridge: Cambridge University Press. DOI: 10.1017/cbo9780511554308.
Progovac, Ljiljana. 1993. Negative polarity: Entailment and binding. Linguistics
and Philosophy 16(2). 149–180. DOI: 10.1007/bf00985178.
Przepiórkowski, Adam. 1999. On Negative Eventualities, Negative Concord, and
Negative” Yes/No” Questions. In Semantics and linguistic theory, vol. 9, 237–
254. DOI: 10.3765/salt.v9i0.2828.
Przepiórkowski, Adam & Anna Kupść. 1997. Negative concord in Polish. Warsaw:
IPI PAN.
Rullmann, Hotze. 1996. Two types of negative polarity items. In Proceedings of
NELS, vol. 26, 335–350. University of Massachusetts, GLSA.
xlvii
Mojmír Dočekal
Strachoňová, Hana. 2017. Česká free choice indefinita. Brno: Masarykova univerzita,
Filozofická fakulta, Brno. (Disertační práce). https://is.muni.cz/th/gmdxp/.
Szabolcsi, Anna. 2004. Positive polarity–negative polarity. Natural Language &
Linguistic Theory 22(2). 409–452.
Tarski, Alfred. 1994. Introduction to logic and to the methodology of the deductive
sciences. Vol. 24. Oxford: Oxford University Press.
Tomaszewicz, Barbara. 2013. Aż/čak–the scalar opposite of scalar only. In Formal
Description of Slavic Languages: The Ninth Conference. Proceedings of FDSL,
vol. 9.
Veselovská, Ludmila. 1995. Phrasal movement and X-morphology: Word order parallels
in Czech and English nominal and verbal projections. Olomouc: UPOL.
(Doctoral dissertation).
von Fintel, Kai. 1999. NPI licensing, Strawson entailment, and context dependency.
Journal of Semantics 16(2). 97–148. DOI: 10.1093/jos/16.2.97.
Yanovich, Igor. 2005. Choice-functional series of indefinite pronouns and Hamblin
semantics. In Semantics and linguistic theory, vol. 15, 309–326. DOI: 10.
3765/salt.v0i0.2921.
Yanovich, Igor. 2013. Four pieces for modality, context and usage. Cambridge, MA:
Massachusetts Institute of Technology. (Doctoral dissertation).
Zeijlstra, Hedde. 2004. Sentential negation and negative concord. Amsterdam: LOT/ACLC.
Zwarts, Frans. 1998. Three types of polarity. In F. Hamm & E. Hinrichs (eds.),
Plurality and quantification, 177–238. Dordrecht: Springer. DOI: 10.1007/978-
94-017-2706-8_5.
xlviii