Chapter 1

Generic sentences
Hana Filip
Heinrich Heine University Düsseldorf

This paper focuses on the expression of generic statements in Czech (West Slavic)
by means of sentences that are formally marked with the morpheme -va-. While
cognate morphemes exist in other Slavic languages, this type of morpheme is pro-
ductive only in West Slavic languages, where it has so far not received enough
attention. In order to raise interest in this Slavic phenomenon, and to tie it to the
state-of-the-art research in genericity, I will explore its properties from the point
of view of quantificational theories of genericity (e.g., Krifka et al. 1995). It is the
main goal of this paper to argue that the morpheme -va- in Czech be best analyzed
as a generic marker sui generis (also building on (Filip & Carlson 1997; Filip 2018)
that lexicalizes a ceteris paribus condition, with epistemic/doxastic consequences
tied to reasoning with exceptions. The Czech generic marker -va- shares a num-
ber of semantic properties with similar generic (habitual) markers in typologically
unrelated languages (as Dahl (1995)) observes, and specifically with the Tlingit ha-
bitual morphology (Cable 2022), the be construction in African American English
(Green 2000), the construction with haya+participle in Modern Hebrew (Boneh &
Doron 2008)). In order to further our understanding of this Slavic phenomenon we
need to study it within a larger typological and theoretical context.
Keywords: genericity, grammatical aspect, Slavic languages, Czech

1 Introduction
West Slavic languages, Czech, Slovak and Polish in particular, provide some of
the richest sources of data for exploring form-meaning mappings in the domain
of characterizing genericity (in the sense defined in Section 2), which is of
main interest in this paper. They possess four different types of verb forms that
are in principle available for the expression of characterizing generic statements:
(i) primary (root) imperfective, (ii) secondary imperfective, (iii) perfective and

\lsCollectionPaperCitationText.
Hana Filip

(iv) a specifically generic, as I will argue here, which are formally marked with
the morpheme -va-, also known as the ‘iterative suffix’ in traditional descriptive
studies; nota bene this marker -va- must not be confounded with the imperfec-
tivizing suffix (for details see Sections 4.1 and 5.2.2). In contrast, in South and
East Slavic languages, perfective verbs are less commonly used in characteriz-
ing generic sentences, and cognate generic morphemes are also attested but are
unproductive, at least in their standard varieties.
The main focus here is on the data from Czech. In principle, in Czech, to
one and the same English verb form used in a generic context correspond four
different verb forms (for morphological details see Section 4.1). This is illustrated
by the following four sentences. Their main verbs share the same imperfective
root ‘bark’, and all correspond to one English verb form bark (3rd person plural,
present tense) in their English translations.
(1) a. Psi štěkají [primary ipf]. Czech
dogs bark
‘Dogs bark.’
b. Psi se poštěkají [pf] že ani nevíte, jak k tomu došlo.
dogs pref.bark that even neg.know how to it came
‘Dogs bark (out) at each other, and one hardly knows why.’
c. Psi poštěkávají [secondary ipf] ve spánku, vrčí a cukají packami.
dogs pref.bark.ipf in sleep, growl and twitch paws
‘Dogs bark now and then while sleeping, growl and twitch their
paws.’ [i.e., bark a little/on and off]
d. Psi štěkávají [generic] na ty, které neznají.
dogs bark.gen at those whom neg.know
‘Dogs bark at those whom they don’t know.’
However, these four different verb forms are not always freely interchangeable
in the same generic context, and to the extent that they are, they may not be
used with the same frequency and ease. The choice of the appropriate verb form
is also constrained by considerable lexical and morphological idiosyncracies of
the Czech verb system.
For instance, what is immediately striking when looking at the above exam-
ples is that only (1a) is a ‘bare’ or unrestricted generic sentence, i.e., it contains no
overt quantifiers (e.g., some, many, most, all), Q-adverbs (e.g., always, generally,
usually), modal expressions, or qualifiers that would explicitly restrict the quan-
tity of instances and circumstances under which the generalization is satisfied.
A ‘bare’ generic sentence with kind reference like (1a) instantiates a prototypical

2
 Generic sentences

characterizing generic sentence in natural languages, as many agree (see Sec-

tion 2). According to Dahl (1995), such generic sentences in all natural languages
exhibit the minimal marking tendency with respect to tense and aspect (see
Section 3): their main verb is either devoid of any tense and aspect marking, as
the English bark in Dogs bark, or corresponds to the least marked form in the
verb system, as the Czech primary (root) imperfective verb štěkají ‘they bark’,
one of the simplest verb forms in Czech, in (1a).
A ‘bare’ unrestricted generic sentence like (1a), just like the English Dogs bark,
is the most natural answer to the question ‘What sound do dogs make?’ This is
because the question concerns the characterizing property of the kind dog as a
whole without any qualifications of the conditions under which it is manifested.
The morphologically more complex forms, and specifically the generic verb with
-va-, would sound odd in an answer to the above question. They tend to be odd or
dispreferred in ‘bare’ unrestricted generic sentences, in out-of-the-blue contexts
like the above question. They sound natural and felicitous in generic sentences
with overt restrictors that qualify the relevant cases based on which the gener-
alization is made, as we see in the above examples, which is not, however, what
the above question calls for. This raises the following general question:

qestion 1: Do the different linguistic forms for the encoding of generic state-
ments differentiate different subtypes of generic sentences, each distinguished
by different clusterings of formal properties, and possibly requiring different se-
mantic/ontological commitments, and separate semantic/ontological models for
their interpretation?

This question arises for any language that has a variety of distinct morphologi-
cal and syntactic means for the expression of characterizing generic statements,
and some argued for splitting the domain of characterizing genericity into dif-
ferent subdomains or different subtypes of characterizing generic sentences (see
e.g., Pelletier 2010; Greenberg 2003). Answers to qestion 1 directly bear on the
fundamental, and still outstanding, question posed by Carlson (1977; 1995) in the
domain of characterizing genericity:

qestion 2: Can we provide a unified semantic analysis for all characterizing

generic sentences?

By referring to the Czech morpheme -va-, as in (1d), as a ‘generic’ marker, I

anticipate one of the main goals of this paper: namely, to argue that it indeed

3
Hana Filip

can be best viewed as a generic marker sui generis (also building on Filip &
Carlson 1997; Filip 2018). This is by no means uncontroversial. A good exam-
ple for the controversies surrounding its grammatical status and contribution to
the meaning of a sentence is Dahl (1995). In the context of his extensive typologi-
cal studies of tense-aspect-mood systems of natural languages, Dahl (1995) takes
the Czech marker -va-, as in (1d), to be a paradigm example of a whole class of
markers in a number of typologically diverse languages. Their common feature
is that they systematically enforce a generic interpretation of sentences in all
their occurrences, and constitute a sufficient but not a necessary condition on
the expression of generic statements. Specifically, Dahl (1995) lists 17 languages
(including Czech) out of 76 languages in his corpus: Arabic (Classical), Akan,
Catalan, Czech, Didinga, German, Guarani, Hungarian, Kammu, Limouzi, Mon-
tagnais, Sotho, Spanish, Swedish, Swedish Sign Language, Yucatec Maya, Zulu
(Dahl 1995: p. 421, fn.8). Dahl (1995) argues that they are not generic markers in
order to support his overarching hypothesis that the episodic/generic distinction,
which underpins the semantic domain of characterizing genericity, has no direct
relevance for the grammar of natural languages (see Section 3.1). Specifically,
for the Czech marker -va- and similar markers in other typologically distinct
languages, Dahl (1995) claims that

“their function is not to mark genericity per se but to serve as a kind of

quantifier over situations with, roughly, the semantics of ‘most’. Follow-
ing the terminology of Dahl (1985), let us refer to this kind of tense-aspect
markers as ‘habituals’. A relatively safe generalization is that habituals in
general exhibit a low degree of grammaticalization. They are of potential
interest to our topic mainly as possible sources of a grammaticalization
path yielding what would be the primary candidates for a gram that sys-
tematically marks genericity” (Dahl 1995: p. 421).

These are the claims that will be refuted here. The above quote encapsulates some
of the most common misconceptions about the properties of the Czech -va-, and
by the same token most likely of other morphemes that seem similar to it. Given
that the Czech marker -va- takes the center stage in Dahl’s (1995) argumentation,
it is an ideal candidate to evaluate the above claims, and ultimately, to reject
them.
The theoretical background that I presuppose comprise generic studies that
intersect the fields of semantics and pragmatics, philosophy, psychology, com-
puter science and other areas of cognitive science, mainly relying on Krifka et al.
(1995) as my point of reference when it comes to foundational assumptions and

4
 Generic sentences

terminology (see section 2). In this context, markers like the Czech -va- (see e.g.,
the list in Dahl (1995: p. 421, fn.8) given above) have so far remained largely un-
explored. In traditional descriptive grammars, verbs formed with such markers
tend to be labeled as ‘iterative’, ‘multiplicative’, ‘frequentative’ (as in Czech gram-
mars, for instance, where they are known as iterativní/násobená/frekventativní
slovesa, respectively, see e.g., Petr 1986, i.a.), and their meaning often confounded
with pluractionality, iterated semelfactivity, event-internal or event-external it-
erativity, and the like. Such designations may also be intended as merely approxi-
mative descriptive labels, given that their semantic properties are ill-understood,
and their status with respect to the categories of tense, aspect and mood is un-
clear, even if they tend to be subsumed under imperfective aspect and their
meaning assimilated to the semantics of imperfectivity, and also less commonly
aligned with tense.
As a relatively safe and uncontroversial term, some use the term ‘habitual(ity)’
for this type of marker: e.g., the Tlingit (Na-Dene) ‘habitual’ particle nooch and
the suffix-ch (Cable 2022), the be construction in African American English (Green
2000), and the construction with haya+participle in Modern Hebrew (Boneh &
Doron 2008). Such studies also directly address the important question of just
how such ‘habitual’ markers relate to characterizing genericity as it has been
delimited mainly based on the relevant form-meaning mappings in English, and
specifically how they are related to the phonologically null generic operator gen
that is taken to represent their quantificational generic force (see Section 2). Of
particular interest here is Chierchia (1995) who proposes that all languages have
a habitual morpheme Hab, which is a functional head in an (imperfective) aspec-
tual projection, which in some languages is realized by explicit aspectual mor-
phemes, while in others, as in English, it is covert, but in any case, it must be
licensed by a Q-Adverb in its Spec: namely, either some overt Q-Adverb the
gen operator, a phonologically null (modalized) universal Q-Adverb. In a sim-
ilar vein, Cable (2022) proposes that the Tlingit habitual morphemes spell out
the T(emporal)-head, which in turn is licensed just in case it is bound by a (pos-
sibly covert) Q-Adverb, hence effectively subsuming ‘habituality’ under tense.
Cable’s 2022 also raises the question whether his analysis of the Tlingit habit-
ual morphology could be adapted for the analysis of Czech generic verbs formed
with -va-, which makes sense given that at first blush Tlingit ‘habitual’ mor-
phology and the Czech -va- share a number of semantic properties (e.g., lack of
reference to particular situations, ‘actuality entailment’).
Both Chierchia (1995) and Cable (2022) provide explicit formal analyses, show-
ing how the insights of the quantificational theories of genericity (as e.g., in

5
Hana Filip

Krifka et al. 1995) could be applied to understudied, and not well-understood,

verb morphology from languages that so far not have been analyzed from the
point of view of the state of the art in the domain of genericity. For this reason, I
will discuss both in detail, even if it turns out that their adaptation to the Czech
data is not viable.
This paper is structured as follows. Section 2 will introduce the theoretical
background (a quantificational theory of genericity) and the key properties of
characterizing generic sentences, mainly relying on Krifka et al. (1995), Pelletier &
Asher (1997) and some foundational papers by Carlson, including Carlson (1995).
This will serve as a useful organizing framework for examining theoretical and
empirical issues that the Czech -va- raises, given that there is no comprehensive
analysis of all characterizing generic sentences, no theoretical framework for the
whole range of characterizing generic phenomena, but only a variety of accounts
of some facets of the meaning of generic sentences. Section 3 addresses the en-
coding of characterizing genericity which is commonly taken to be a property of
an entire sentence, without any dedicated markers, and if they exist, they are so
rare as to be insignificant in the grammar of natural languages. Here, I will raise
the issue if indeed this holds, and introduce a class of markers, including the
Czech -va- as their prominent example, which enforce only a generic (habitual)
interpretation systematically and in all their occurrences. To this goal, in Section
4 the basic morphological properties of the Czech morpheme -va- will be intro-
duced, and in Section 5 I will provide detailed arguments showing that it cannot
be treated as a marker of aspect or tense, neither can its meaning be reduced to
a ‘habitual’ quantifier over situations with the meaning similar to ‘usually’ or
‘most’, and hence invalidate Dahl’s 1995 foundation for his claim that it cannot
be a marker of genericity. I will also discuss consequences of this result for the
status of the Czech -va- as a marker of genericity, and the independence of gener-
icity (and habituality) as a category that is independent from other categories of
the tense, aspect and mood system. Section 6 will show why the analyses of
‘habitual’ morphology proposed by Chierchia (1995) and Cable (2022) cannot be
adapted for the analysis of the Czech -va-. With all this background in place,
and having mainly focused on refuting a variety of misconceptions of what -va-
is/does, in Section 7, I will finally turn to the proposal of what it is and how it
ought to be best analyzed.
The goal of this paper is not provide a full-fledged formal analysis, but rather
to raise interest in the Czech -va- as a marker of genericity sui generis, and by
the same token in cognate generic markers in other (West) Slavic languages in
Slavic linguistics and generic studies. Discussing in detail the morphological and

6
 Generic sentences

semantic properties of generic verbs that are formed with -va- from the point of
view of the quantificational theories of genericity may also prove useful for the
exploration of similar markers in unrelated languages, and perhaps even further
our better understanding of such understudied morphology.

2 Characterizing genericity
The domain of genericity is commonly divided into two independent, but related,
domains: namely, kind reference and characterizing genericity (Krifka et
al. 1995, Carlson & Pelletier 1995). Some examples are given below:

(2) a. Ravens are widespread. kind reference

b. Dodos became extinct in the 17th century.
c. Bronze was invented as early as 3000 B.C.
d. The dodo became extinct in the 17th century.

(3) a. Ravens are black. characterizing genericity

b. Dogs bark.
c. Paul plays chess on Saturdays.

There are both formal and semantic reasons for this split. The category of kind
reference is motivated by the existence of kind predicates like extinct, widespread,
invent that directly select kind denoting terms for one of their arguments (in
italics in the kind reference examples above)1 . Kind predicates attribute a prop-
erty to a whole kind and one which does not hold of its individual members. For
instance, in (2b), the property of being extinct holds of the whole kind dodo, and
it makes no sense to attribute it to any particular individual dodo or pluralities
thereof.
The main topic of this paper concerns the domain of characterizing gener-
icity. Unlike kind reference which is tied to a kind denoting term, character-
izing genericity is a property of an entire sentence. It often arises from the com-
bined meanings of the subject NP/DP and VP, and possibly also in interaction

1
A given language may impose specific requirements on the form of a kind denoting term. In
English, for instance, a kind denoting term can be realized as (i) a bare mass or plural argument,
as in (1a-c), or (ii) a singular definite, as in (1d), but not as a singular indefinite (*The dodo became
extinct in the 17th century). In contrast, in Romance languages, for instance, bare arguments
tend not to be used for kind reference: L’acqua è limpida (Italian) – ‘(The) water is clear’).

7
Hana Filip

with pragmatic, prosodic (focus), and discourse factors. Characterizing generic

sentences, as in (3a) and (3b), entail a connection between a characterizing prop-
erty and a kind, whereby the characterizing property holds of that kind as well
as of its individual members. Characterizing generic sentences also include sen-
tences like (3c) that describe a regularity of action ascribed to an ordinary indi-
vidual (Carlson 2011; Krifka et al. 1995), i.e., a habit in the proper sense of this
term, which are often taken to be paradigm examples of (weak and) descriptive
geeneralizations (e.g., Carlson 1995; some consider such habitual generics to be
less typical members of the class of characterizing generic sentences, see e.g.,
Nickel 2017.)
characterizing genericity and kind reference are independent of each
other. There are characterizing generic sentences without kind reference, as in
(3c), and kind reference may be realized in episodic sentences that make reference
to particular situations, such as Marconi invented the radio, The horse arrived in
the New World around 1500 2 . Kind reference and characterizing genericity also
co-occur in a single sentence, as in (3a) and (3b), where the subjects are kind
denoting terms.
There is no general agreement on the criteria that delimit all and only charac-
terizing generic sentences (see e.g., Dahl 1985; 1995; Nickel 2008; 2016; Pelletier
2009; Carlson 2013), although there is some agreement that prototypical exam-
ples are ‘bare generics’ like Birds fly that involvekind reference in a charac-
terizing sentence that contains no overt quantifiers (e.g., some, many, most,
all), Q-adverbs (e.g., always, generally, usually), modal expressions, or qualifiers
that would explicitly restrict the quantity of instances and circumstances under
which the generalization is satisfied. All agree that all generic sentences are
fundamentally opposed to episodic sentences. Some examples are given below:
(4) a. Fido, my neighbor’s dog, is barking at the mailman. episodic Ss
b. Paul is playing chess right now, do not disturb him.
c. Paul played chess { a few | three times every day } last week.
Episodic sentences describe particular situations, conditions or facts, which may
serve as the base for the truth conditions of the related generic sentences (e.g.,
witnessing what (4a) describes contributes to inductive evidence for the truth of
(3b)). Hence, the relation between the two provides key insights into the ground-
ing of the truth of characterizing generics.

2
Such examples of kind-reference in episodic sentences exhibit the so-called ‘avant-garde’ read-
ing, see e.g., Krifka et al. (1995); Carlson (2006).

8
 Generic sentences

Episodic sentences refer not only to particular singular situations (4b)) but
also pluralities thereof which are not a part of a pattern ((4c). A mere plural-
ity of situations, i.e., iterativity, ‘iterative aspect’, event-internal/external plurac-
tionality, multiplicativity, frequentativity, repetitivity, and the like must not be
confounded with genericity (see Section 5.1).
As many agree, following (Carlson 1977), characterizing generic (which in-
clude what is often labeled as ‘habitual’) sentences have three main properties,
which episodic sentences lack:
• All are aspectually stative (e.g., Krifka et al. 1995).

• All have an intensional (modal) import (first observed by Lawler 1973;

Dahl 1975).

• Most admit exceptions, but are also compatible with no exceptions what-
soever (e.g., Pelletier & Asher 1997; Cohen 1999; Greenberg 2007; Krifka
et al. 1995, i.a.).
First, all generic predicates are aspectually stative. They are headed by generic
predicates derived from stage-level predicates (e.g., bark, as in (3b)), or lexically
stative predicates, individual-level predicates (ILPs) like (be) intelligent, know
(French), which are taken to be inherently generic (Chierchia 1995). All aspectu-
ally stative predicates lack reference to particular situations. This is because they
express ‘tendentially stable properties’, i.e., properties of individuals that hold of
them during their whole life-time or at extended intervals as well as at any mo-
ments of such intervals. Hence, their truth is independent from any particular
situation(s), episodes, as is, for instance, shown by their oddity with adverbials
referring to specific time-points:
(5) a. ? John was intelligent on his porch at 4 p.m.
b. ? Dogs right now bark.
Turning to intensionality and exception-tolerance, they thwart any attempt at
an analysis of generic (and habitual) sentences in terms of a single extensional
quantifier (over episodic formulas containing free variables) (e.g., all, most or
some), quantity expression (e.g., in a majority of cases or in a significant number
of cases), a quantity-based criterion, or a simple statistical criterion/correlation,
no matter how contextually constrained and/or modalized they may be, or pos-
sibly also defined in vague or probabilistic terms (Pelletier & Asher 1997; Carl-
son 2013; Nickel 2013; 2017). Such notions are neither necessary nor sufficient
to ground what is characteristic, and how we infer that there is a pattern from

9
Hana Filip

our perceptions of the world. Moreover, the kind of intensionality that gener-
ics exhibit does not fit standard modal notions that are applied in the analysis
of aspect (pace Ferreira 2004; Ferreira 2016; Hacquard 2006,i.a.), and neither are
modal notions like necessity or essence (Nickel 2013; 2017 and elsewhere).
Intensionality sets generic (and habitual) sentences clearly apart from exten-
sional quantified sentences. Statements with extensional quantifiers are context
sensitive, because they may quantify over different domains in different con-
texts; the domain of quantification is the closed set of things a quantifier ranges
over. For a given thing to be in a domain of quantification (relative to a possible
world) means that it belongs to the extension of the property in question (relative
to that world). However, the truth of characterizing generic sentences does not
(just) depend on accidental facts at a given context3 , at a given world. The most
compelling argument comes from generic sentences that are true in the face of
no supporting inductive evidence, i.e., for which there are no episodic instances
in the actual world that count as evidence for their truth, such as (6a) and (6b).
(6) a. The Speaker of the House succeeds the Vice President. (Carlson 1995)
b. Tab A fits in slot B.
Context: An instruction written on a cereal box cut-out toy, which
will never be made, because the box happens to be irreparably
damaged.
The intensionality of generic sentences is also evident in how we evaluate their
truth in contrast to explicitly quantified sentences with extensional quantifiers.
In contrast to All/Most dogs bark (on its extensional interpretation), for instance,
the actual extension of the subject in a generic sentence like (3b) Dogs bark at a
given world has almost nothing to do with whether it is true (Pelletier & Asher
1997: p. 1132-3), as it does not make a claim about a closed set of existing dogs,
but about every (realistically) possible dog (Carlson 1995; Pelletier & Asher 1997).
It may be true even if at a given world all dogs happen not to bark. (3b) Dogs
bark (on its generic reading) is true based not just on inductive evidence (e.g.,
having witnessed barking dogs on a number of occasions), but also on evidence
which is not directly accessible to us, on underlying causes in the world (such as
regularities governing the biology of living organisms) from which the patterns
in the phenomena in the world can be inferred.

3
Following Montague (1974), and also Kaplan (1989), contexts are identified with indices, fea-
tures of the situation needed to determine extension. An index contains a speaker, an ad-
dressee, a time, a place, a possible world, as well as other elements like“indicated object” coor-
dinates to determine the referents of demonstrative pronouns; see also Stalnaker (1998).

10
 Generic sentences

Weak and descriptive generalizations like John drinks beer also have an inten-
sional interpretation. I.e., this sentence may be understood as meaning that John
is not averse to drinking beer; he may or may not drink it, if you give it to him
(disposition), or as a regularity that John drinks beer when a suitable occasion
arises. To take another example, This machine crushes oranges can have a purely
intensional intepretation concerning its intended design function, and be true
by virtue even if it is still packed in its original box, and has not yet crushed any
oranges; or it may also be true by virtue of having been already used to crush
oranges. The intensional meaning component of generic sentences follows from
the nature of generalizations. They transcend particular situations, states of af-
fairs, or facts, in short, they transcend our immediate experiences of the world.
Generic sentences specify not only what actually obtains at given worlds and
times, but they are law-like in so far as they capture ‘non-accidental’ properties,
rather that what merely contingent ones, and predict what is possible, and pos-
sibly may never ever be realized.
Finally, being law-like generalizations, characterizing generic sentences may
be true in the face of exceptions (e.g., Pelletier & Asher 1997; Cohen 1999; Green-
berg 2007; Krifka et al. 1995, i.a.). For instance, we judge (3b) Dogs bark to be true,
even if it is part of our general world knowledge that there are non-barking dog
species like Basenjis and grey hounds, and individual dogs that may not bark due
to injury or illness. We treat them as exceptional individuals that do not invali-
date (3b) Dogs bark. Similarly, we judge (3c) as being true even if there are (some
or even many) Saturdays on which Paul does not play chess.
Exception-tolerance of generic sentences is perhaps their most intriguing and
challenging property (Pelletier & Asher 1997; Pelletier 2009). We judge a sentence
like (3b) Dogs bark as true despite knowing that not all the dogs bark, which
has the consequence that we cannot validly infer for any particular, arbitrary
individual dog denoted by its subject argument of that it will have the property
of barking.
Moreover, different types of generic sentences admit a different number and
type of exceptions or non-confirming cases while still remaining true. How many
exceptions and what kind a given characterizing generic sentence will admit de-
pends on a variety of factors which include not only objective facts of the world,
but also our kind-specific expectations as well as subjective judgments about
what we view as significant, salient or striking in some way. What makes (3b)
true, among others, is that most dogs satisfy the characteristic property of bark-
ing. However, majority satisfaction is neither sufficient nor necessary for the
truth of all exception-tolerating generics. Printed books are paperbacks is false,

11
Hana Filip

despite the majority of printed books being paperbacks. What is characteristic

of a kind need not be prevalent among its members. it may hold of only a minor-
ity subkind, and even of a very small fraction of its members, provided we view
it as significant in some way (see Krifka et al. 1995, Leslie 2008). For instance,
Mosquitoes carry the West Nile virus is true even if a vast majority of members
(99%) fail to carry the virus. Given that the virus is for us potentially highly dan-
gerous, we judge as ‘significant’ to us in some way, because it is a feature of
mosquitoes that is particularly striking, impressive, salient, recognizable, mem-
orable, subjectively impressive, and the like.
What is ‘significant’ depends not only on objective facts about the world, as
we know it, but also on our subjective, affective-evaluative, psychological stance
to our experiences with the world or our practical goals and needs. While such
subjective factors are occasionally acknowledged in the literature on genericity
(e.g., Krifka et al. 1995), they largely remain at a pretheoretical level, and it is
unclear how they can be analyzed and formally represented (see Nickel 2017).
Compatibility of characterizing generics with a different number and kind of
exceptions raises the question about what is ‘characteristic’, ‘typical’ or ‘normal’,
on the one hand, what we view as failing to conform to it and why, on the other
hand.
Leslie’s (2007, 2008) approach to genericity is couched within cognitive psy-
chology, and emphasizes the insight (previously suggested elsewhere, see e.g.,
Krifka et al. 1995) that the truth of generic sentences, besides depending on ob-
jective facts about the world, as we know it, is also determined by psychological
(subjective, affective-evaluative) factors stemming from our specific kind-related
expectations. What we view as characteristic of a kind might be based on judging
it as significant in some way, because it is particularly salient, noticeable, strik-
ing or impressive like a lion’s mane, or we perceive it as particularly harmful to
us, such as a virus or an animal that can be potentially deadly to us. It is precisely
this subjective salience that also motivates why what we view as characteristic
of a kind need not be prevalent among its members, and may even hold for only a
small fraction of its members. This is illustrated with the sentences given below:
(7) a. Mosquitoes carry the West Nile virus (though 99% do not).
b. Sharks attack people (though most do not).
As a matter of fact, only about 1% of mosquitoes carry the West Nile virus. Al-
though sharks are known for attacking humans, it is rare for them to do so. Only
a few Russian generals died young in Stalin’s times, but those how did, often died
under particularly egregious, mysterious or sudden circumstances.

12
 Generic sentences

The view of characterizing genericity which has just been briefly sketched is
widespread in contemporary formal semantics, philosophy and psychology, and
other areas of cognitive science (such as AI and computer science) with which
they intersect. There is no shortage of promising analyses that focus on some
key facet of the meaning of generic (and habitual) sentences (see summaries in
Krifka et al. 1995; Pelletier & Asher 1997; Nickel 2016; 2017, i.a.), however, none
has so far offered a comprehensive analysis that can be equated with the analy-
sis of all generic (and habitual) sentences. Moreover, while exception-tolerance,
and its extensional consequences, tend to garner most attention, intensionality,
however, often remains at a pretheoretical level and so largely unaccounted for
(Pelletier & Asher 1997).
For this reason, when it comes to representational assumptions, as a useful
point of reference and an organizing principle for examining various theoretical
and empirical issues associated with the meanings of characterizing generic sen-
tences, we may adopt a logical representation in terms of a tripartite structure
with a phonologically null, modalized dyadic gen operator. It has the following
general form (adapted from Krifka et al. 1995: p. 32 (57), see also Krifka 1987):

(8) gen [x1 … xi ; y1 … yj](Restrictor [x1 … xi]; Matrix [{x1} … {xi}, y1 … yj])
x1 … xi: variables bound by gen
y1 … yj: variables bound existentially, with scope just in matrix
{x1} … {xi}: means x1 … xi may or may not occur in matrix

Just like an overt Q-Adverb, gen takes two arguments, a Restrictor (which states
the restricting cases relevant to the Matrix/Nuclear Scope) and a Matrix (which
makes the main assertion of the sentence), which are separated by ‘;’ in the tri-
partite formula: If the conditions stated in the Restrictor hold, other conditions
stated in the Matrix will also hold.4 gen represents the requisite generic force,
that is, intuitively, it stands for a certain close ‘characterizing’ relation between
the conditions specified in the Restrictor and the Matrix, i.e., the conditions that
specify the episodic base of the generalization.
The base for the generalization may be (i) instances of particular individuals
(as in (3a), where it is an instance of a particular raven that is in the state of being
black), (ii) instances of singular/particular situations (as in (3c)), or (iii) instances

4
gen just as overt Q-Adverbs are taken to bind the variables introduced in the Matrix, un-
less they are bound by existential closure. Moreover, as is now commonly assumed, overt
Q-Adverbs and gen are not unselective binders (in contrast to their early treatments, as in
papers in Krifka et al. 1995, for instance), and which variables can be bound in a given logical
structure, depends on a number of factors, including context and topic-focus structure.

13
Hana Filip

of both situations and individuals, as in (3b), i.e., ‘mixed’ cases of kind reference in
characterizing sentences (Krifka 2009), are often taken to represent prototypical
cases of characterizing generic sentences (e.g., Dahl 1995).5
For instance, the logical structure of (3c), repeated below for convenience,
which expresses a generalization over situations (the case (ii) above), roughly
amounts to (see also Krifka et al. 1995):

(9) Paul plays chess on Saturdays. [ = (2c)]

gen [s,x;] (x=Paul and x in s and s is_on_Saturday; x plays_chess in s)
‘Given a situation s such that it is on Saturday and Paul is in s, Paul plays
chess in s.’

In Krifka et al. (1995), all characterizing generic sentences that quantify over
situations, also occasions or cases, as Lawler (1972) originally proposed, and akin
‘cases’ of Lewis (1975) (Krifka et al. 1995: p. 30). 6 are referred to as ‘habitual’
sentences:
(10) A sentence is habitual iff its semantic representation is of the form
gen[ … s … ; … ] (Restrictor [ … s … ]; Matrix [ … s … ]),
where s is a situation variable (Krifka et al. 1995: p. 32 (56))
The term ‘habitual’ sentence is also used in the narrowest sense for characterizing
generics like (3c)) that describe ‘habits’ proper, i.e., attribute a regularity of action
to a specific individual of an ordinary sort (see e.g., Pelletier & Asher 1997).
The main burden of the analysis, and a matter of many controversies, is on
defining the meaning of gen.7 When we try cutting through the many con-
troversies that surround the evasive truth conditions of characterizing generic
5
They involve a ‘double generalization’ (in Carlson’s 2011; see also Pelletier & Asher 1997:
p. 1131). (3b) is true by virtue of (a) individual dogs that realize the kind dog having the char-
acterizing property of barking, and also by virtue of (b) particular situations of barking by a
stage of an individual dog.
6
The notion of ‘situation’, also similar to ‘cases’ of Lewis (1975), plays a role similar to that of
‘stages’ in Carlson (1977, and elsewhere) which is tied to Carlson’s distinction between stage-
level (episodic) predicates (SLPs) and individual-level (stative) predicates (ILPs). Lewis’ cases
are tuples of situations (or events) and individuals (which take part in the situations), and
possibly also including times. “If a case is a ‘tuple of persons plus a time-coordinate, we may
take it generally that the persons must be alive at the time to make the case admissible. Or if a
case is a ‘tuple of persons plus an event-coordinate, it may be that the persons must take part
in the event to make the case admissible. It may also be required that the participants in the
’tuple are all different, so that no two variables receive the same value. (I am not sure whether
these restrictions are always in force, but I believe that they often are.)” (Lewis 1975: p. 12).
7
Any proposed generic quantifier (whether realized as a nominal determiner or as a frequency
adverbial), unlike other quantifiers, is not conservative (Barwise & Cooper 1981), see Cohen

14
 Generic sentences

sentences, one useful way to do was proposed by Carlson (1995), who suggests
that various proposals may be divided into two main fundamental perspectives.
One perspective is roughly aligned with an inductive model, on which generic
sentences express inductive generalizations that are based on some observed (or
unobserved) set of instances or particulars in the (actual) world from which we
infer a general rule. Paradigm cases are sentences like (3c) that express gener-
alizations over situations of the kind described by their corresponding episodic
predicate like (4b). On this perspective, characterizing generics would (mainly)
express descriptive (weak) generalizations. One obvious disadvantage of the in-
ductive model is that it is burdened by the problem of induction.
The second main perspective corresponds to what Carlson (1995) dubs a rules-
and-regulations model. On this perspective, generics express generalizations that
do not (necessarily) depend on particular episodic conditions in the actual world,
but rather on some causal structure or underlying forces in the world, which give
rise to episodic instances (if there are any), and motivate the patterns that we rec-
ognize in the actual world. Besides the ontology presupposed by the inductive
model, the rules-and-regulations model assumes some such underlying factors
for the grounding of truth of generic sentences. Paradigm cases are characteriz-
ing generic sentences expressing various rules, such as game rules like Bishops
move diagonally, instructions, customs and the like, and which we may learn di-
rectly (rather than inferring them from episodic conditions in the actual world).
Other examples include job descriptions, functions, dispositions, abilities, insti-
tutional regulations (6a), customs, instructions (cooking recipes), social norms,
as well as universal laws of physics, and also non-universal (ceteris paribus) laws
of biology, psychology, geography and economics. In contrast to the inductive
model, the rules-and-regulations model can also handle characterizing generic
sentences that are independent of particular instances or situations like Sam is
a bachelor (lexical characterizing sentences), and also those for which there are,
have been, and may never be verifying instances, such as (6a) and (6b). On the
rules-and-regulations model of genericity, we judge generic sentences to be true
or false with respect to a set of rules (or a finite list of propositions), which are
considered to be irreducible entities, and that support the truth of possibly an un-
limited number of propositions. A generic sentence is then true with respect to a
(1999: 53–54); Nickel (2008). A quantificational relation is conservative when its first argu-
ment can be intersected with its second argument without a change in truth conditions, e.g.,
All ravens are black iff ‘All ravens are ravens that are black’. In other words, the truth of
a conservative quantifier depends on nothing outside of the extension of its first argument.
However, for generics we have: Lions have manes it does not hold that it is true iff ‘Lions are
lions that have manes’.

15
Hana Filip

set of rules iff that set of rules entails the proposition expressed by that sentence.
These two fundamentally different perspectives lead Carlson (1995) to pose the
fundamental question posed at the outset and here repeated for convenience:

qestion 2: Can we provide a unified semantic analysis for all characterizing

generic sentences?

As a null hypothesis, it is preferable for generic sentences to constitute a single

class for which a unified semantic analysis is possible. Carlson (1995) proposes
that the rules-and-regulations model is a better candidate than the inductive one
to do this job, and one that might ground the semantics of the phonologically null
gen operator. However, a rules-and-regulations model is not a good fit for gener-
ics like (3c) Paul plays chess on Saturdays “which lend the most prima facie plau-
sibility to the inductive model” (Carlson 1995: p. 237). A rules-and-regulations
model would require us that we identify some rule, causal factor or underlying
condition that gives rise to or ‘regulates’ Paul’s chess playing behavior, which,
however, seems rather far-fetched.
Here, I pursue a rather modest goal: namely, to explore some formal means
that are dedicated to the expression of characterizing genericity. Of main interest
is the Czech morpheme -va- that enforces a characterizing generic reading, sys-
tematically in all its occurrences, and which constitutes a sufficient condition for
a generic reading of a sentence or a clause. This goal is motivated by the general
idea that one of the most reliable strategies in trying to evaluate the claims about
the semantics of a given domain is to examine the properties of formal means
that are associated with that domain. Let me, therefore, next turn to the encod-
ing of characterizing genericity in order to set up the stage for the exploration of
the Czech morpheme -va-. It is one of the main claims of this paper that it might
best be viewed as a generic marker sui generis.

3 Encoding of characterizing genericity

3.1 A property of a whole sentence
Those who study genericity from the point of view sketched in the previous sec-
tion (Section 2), i.e., many semanticists, philosophers, computer scientists and
researchers from other related fields, implicitly or explicitly understand charac-
terizing genericity as a category in its own right, independent from other gram-
matical categories of natural language, and specifically from the categories of the

16
 Generic sentences

tense-aspect-mood system. This view finds compelling support in the explana-

tory role played by the semantic generic/episodic distinction, and the closely
related individual-level/stage-level distinction (Carlson 1977), in the semantics
of natural languages in the past nearly fifty years.
At the same time, however, the status of characterizing genericity as a cate-
gory sui generis is not uncontroversial, given the widespread assumption that
it appears to lack dedicated markers8 , or if there are such markers, they are so
rare as to constitute a marginal phenomenon at best (Dahl 1995). As observed
above, characterizing genericity is taken to be a property of an entire sentence,
arising from the combined meanings of the subject NP/DP and VP, and possibly
also interacting with pragmatic, prosodic (focus), and discourse factors (see e.g.,
Krifka et al. 1995; Pelletier 2010). Natural languages have a variety of morpho-
logical or syntactic means available to signal or facilitate a generic interpreta-
tion of a sentence, including adverbs of quantification (or Q-adverbs for short),
such as seldom, often, usually 9 , and a variety of means that are tied to the fi-
nite verb (ibid.). They include constructions like the English used to construction,
verbal complexes with auxiliaries (e.g., bruka in Swedish, see Dahl 1995), and
most prominently, the generic interpretation is often indirectly reflected in the
choice of grammatical markers of tense and aspect. In English, for instance, the
simple (non-progressive) present tense form has predominantly a generic inter-
pretation, while the progressive an episodic one, as the contrast between (3b)
and (4a) shows.
Hence, philosophers and occasionally linguists (Dahl 1975, Dahl 1985) use the
term ‘generic tense’ or ‘gnomic tense’ for the simple present tense in English,
suggesting that genericity is a member of the tense system of a language. Less
commonly, genericity is taken to belong to the system of modals. The most
widespread claim is that genericity is a part of the aspectual system of natural
language, and specifically a subcategory of imperfectivity (Comrie 1976: p. 26ff.
Comrie 1985: p. 40; Dahl 1995, i.a.).
Based on such observations and a corpus study of about 76 languages, Dahl
(1995) argues that the episodic/generic distinction, and by extension character-

8
Leslie (2008), for instance, observes that the absence of dedicated ‘generic’ determiners or
quantifiers should make the acquisition of generic sentences more difficult than the acquisition
of sentences with explicit quantifiers like all and some, and yet they are easier for children
to master than sentences with overt quantifiers. The acquisition ease might perhaps be also
related to how ubiquitous generic sentences are in everyday speech and how fundamental they
are for our reasoning about the world around us.
9
There are proposals that subsume genericity under the system of frequency adverbs, e.g.,
Farkas & Sugioka (1983).

17
Hana Filip

izing genericity, has no direct relevance for the grammar of natural languages,
but it is only indirectly reflected in users’ choices between grammatical mark-
ers of aspect and tense, ”however sad this may be for those of us to whom the
episodic/generic distinction is dear” (Dahl 1995: p. 425). Moreover, with respect
to tense and aspect, prototypical characterizing generic sentences (e.g., Birds fly
in English, on his view) exhibit the following encoding strategy:
The minimal marking tendency of genericity: In the majority of nat-
ural languages, prototypical generic sentences either lack overt tense-
aspect marking (e.g., fly in Birds fly) or use the least marked form in the
tense-aspect category system in a given language (Dahl 1995: p. 415ff.).

I would like to shed doubts on this view, and try to at least raise the possibility,
if not establish, that natural languages have dedicated markers of characterizing
genericity, albeit often constituting a sufficient (not a necessary) condition for a
characterizing generic interpretation of a sentences. If so, this would also provide
additional support in favor of characterizing genericity being a category in its
own right.

3.2 Generic markers on verbs: The Czech marker -va- as a case in

point
Many languages have verb affixes, preverbs or non-bound particles on verbs that
for the most part constitute a sufficient, but not a necessary, condition for the
characterizing generic interpretation. Krifka et al. (1995) mention the Swahili ver-
bal prefix hu-, and Dahl (1995) the Czech (West Slavic) morpheme -va- (using its
traditional, but misleading, label ‘iterative suffix’, see further below), and Cable
(2022) the particle nooch and the suffix -ch in Tlingit (Na-Dene; Alaska, British
Columbia, Yukon):

• Swahili (Bantu): verbal prefix hu-

(11) a. Wanafunzi wanasoma Kiswahili.10 non-generic
students pres.read/study Kiswahili.
‘Students are { reading/studying | read/study } Kiswahili.’
b. Wanafunzi husoma Kiswahili. generic
students hu.read/study Kiswahili.
‘Students (generally/usually/as a rule) read/study Kiswahili.’

10
http://www2.ku.edu/ kiswahili/pdfs/lesson_48.pdf

18
 Generic sentences

• Czech (West Slavic): morpheme -va-, ‘iterative suffix’ Petr (1986); Dahl (1995)
(12) a. Moře je IPF v tuto dobu vyhřáté na příjemných 28℃.11 non-generic
sea be.pres in this time warmed-up on pleasant 28℃
‘At this time the sea is warmed up to a pleasant 28℃.’
b. Moře bývá v tuto dobu vyhřáté na příjemných 28℃. generic
sea be.gen.pres in this time warmed-up on pleasant 28℃
‘At this time the sea tends to be warmed up to a pleasant 28℃.’
• Tlingit (Na-Dene): ‘habitual particle nooch’ Cable (2022), ex. (12)
Scenario: Some dogs are barking outside. You want to remark on this.
(13) a. Yeedát gáanx’ áwé asháa wé keitl. non-generic
now outside.at foc imprv.3s.bark dem dog
‘Dogs are barking outside now.’ (C)15
b. (#)Yeedát gáanx’ áwé asháa nooch wé keitl. generic/‘habitual’
now outside.at foc imprv.3s.bark hab dem dog
‘Some dogs often/always/regularly bark outside.’ (C)
Speaker Comment: “Nooch means ‘sometimes’.” (SE)
The above (b) sentences minimally differ from the (a) sentences by the presence
of markers that enforce their generic or ‘habitual’ interpretation. They are suffi-
cient, but not necessary, for the expression of characterizing genericity, because
their unmarked counterparts, may also have contextually determined generic
readings, apart from episodic ones. From the Slavic point of view, it is worth
immediately mentioning that the Czech marker -va-, as in (12b), must not be
confused and confounded with the imperfectivizing suffix. I will address their
differences in detail further below.
As mentioned at the outset, in Section 1), Dahl (1995) takes the Czech marker
-va-, as in (1d) or (12b), as a paradigm example of a whole class of markers in a
number of typologically diverse languages, and argues that it does not qualify as
a generic marker at all, and by proxy all the other markers in other languages
that appear similar to it. This argument is of key importance to Dahl, as it serves
to support his overarching hypothesis that the episodic/generic distinction has
no direct relevance for the grammar of natural languages (see Section 3.1). In
arguing that the Czech marker -va- does not qualify as a generic marker, Dahl
11
The abbreviations to be used here in glosses are as follows: 1/2/3 = 1st/2nd/3rd person; ACC =
accusative; COP = copula; DAT = dative; GEN = genitive; INF = infinitive; IPF = imperfective;
NEG = negation; NOM = nominative; PF = perfective; PL = plural; PREF = prefix; PST = past;
REFL = reflexive; SG = singular.

19
Hana Filip

(1995) attributes to it a number of specific properties, which as it turns out, it

does not have, as I will show. Given that the Czech marker -va- serves as one of
the key data points for Dahl’s (1995) hypothesis, it is ideally suited for showing
why it is based on false premises when it comes to the role of the Czech -va- in
his argumentation.
Before setting out to do so, let me first clarify a terminological issue. The com-
mon label ‘iterative suffix’, or also ‘multiplicative suffix’ (násobenost ‘multiplica-
tivity’ in Czech), for the Czech marker -va- that is used in descriptive reference
grammars of Czech (see e.g., Petr 1986), and also by Dahl (1995), is problematic,
and will be eschewed here. Its status as a suffix is not uncontroversial, and some
argue that it is an infix (e.g., Kučera 1981). Trying to resolve this ‘suffix-infix’ con-
troversy would go well beyond the scope of this paper, so instead I will hence-
forth use the neutral terms ‘morpheme’ or ‘marker’ -va-.
More importantly, the label ‘iterative’ or ‘multiplicative’ is a misnomer at best,
and in fact, as I will show in Section 5.1, it is a semantic category error. Iterativity
amounts to a mere plurality of situations, i.e., a series of particular episodic situ-
ations of the same type at a given world and time, which can be counted. This,
however, is not the function of the Czech marker -va- at all (see Section 5.1 for de-
tails). This also raises the question about a similar misanalysis of corresponding
markers in other languages (including those listed by Dahl 1995: p. 421, fn.8) and
which are variously labeled as ‘iterative’, ‘multiplicative’, ‘frequentative’, ‘repet-
itive’, ‘pluractional’, ‘habitual’ and the like in traditional, descriptive or standard
reference grammars. It is one of the main goals of this paper to establish that the
Czech marker -va- be best viewed as a marker of characterizing genericity (also
building on Filip 1993; Filip & Carlson 1997; Filip 2009a,b,c and Filip 2021). To
anticipate, I will henceforth refer to it as the generic morpheme or marker.
In what follows, I will introduce basic morphological and semantic properties
of the Czech generic marker -va-, and then show that it does not have the prop-
erties that Dahl (1995: p. 421) (see above) attributes to it: it is fully independent
from tense and aspect, both formally and notionally, it does not function as a
kind of quantifier over situations with meaning reduced roughly to the seman-
tics of ‘most’ or ‘usually’, and neither is it a marker of ‘habituality’. If -va- fails to
have these properties, then Dahl’s (1995) claim that it is not a marker of genericity
has no empirical grounding, and must be rejected. If the Czech marker -va- is a
paradigm example for similar markers in other languages, as Dahl (1995) claims,
then one might also question the validity Dahl’s claim that they are not markers
of genericity.

20
 Generic sentences

4 Characterizing genericity in Czech: Basic morphological

facts
4.1 The productivity and systematicity of the Czech morpheme -va-
The Czech –va- applies to past tense imperfective stems (Petr 1986: Vol. 2, p. 184),
which are realized in non-generic imperfective verbs that freely alternate be-
tween episodic and generic interpretations depending on context, and derives
generic verbs, which exclude any episodic interpretations. As we see in Figure 1,
there are two types of non-generic imperfective stems to which -va- applies:

(i) primary (root) imperfective (formally unmarked for ipf),

(ii) secondary imperfective (formally marked with the ipf suffix), and

Figure 1: Derivation of generic verbs from non-generic imperfective verbs

imperfective: episodic/generic generic

(i) primary
psát → psávat
write.inf write.gen.inf
‘to (tend to) write’ or ‘to be writing’ ‘to tend to write as a rule /
seldom / on and off / usually, …’

(ii) secondary
(a) dávat → dávávat
give.ipf.inf give.ipf.gen.inf
‘(to tend) to give’ or ‘to be giving’ ‘to tend to give as a rule /
seldom / on and off / usually, …’

(b) přepisovat → přepisovávat

iter.write.ipf.inf iter.write.ipf.gen.inf
‘(to tend) to rewrite / (to tend) to copy’ ‘to tend to rewrite (or copy) /
or ‘to be rewriting / to be copying’ as a rule / seldom / on and off /
usually, …’
Secondary imperfective stems are formed by means of the imperfectivizing
suffix, which is one of the most common and productive verb formation patterns
in Slavic languages. The imperfectivizing suffix uniformly applies to a perfec-

21
 Hana Filip

tive base, which is either simplex, illustrated in Figure 2, or prefixed, shown in

Figure 3, and forms its imperfective counterpart with the same lexical meaning
and argument structure. Hence, some classify the imperfectivizing suffix as an
inflectional suffix (Isačenko 1960; Sperling 1996; Dahl 1990).

Figure 2: Derivation of secondary imperfectives from simplex perfectives

simplex pf → secondary ipf

dát dávat
give.inf give.ipf.inf
‘to give’ ‘to give’ or ‘to be giving’

Figure 3: Derivation of secondary imperfectives from prefixed perfectives

primary ipf → prefixed pf → secondary ipf

psát přepsat přepisovat
write.inf iter.write.inf iter.write.ipf.inf
‘to write’ or ‘to be ‘to rewrite’ ‘(to tend) to rewrite’
writing’ or ‘to copy’ or ‘to copy’
‘to be rewriting’ or
‘to be copying’

As in other Slavic languages, primary (root) imperfectives or secondary imper-

fectives (marked with the imperfectivizing suffix) are ‘general imperfectives’ in
the sense of Comrie (1976): they cover the whole semantic domain of imperfec-
tivity, as they have contextually determined episodic and generic readings. Put
differently, they are formally unmarked for genericity. A number of languages
have such ‘general imperfective’ forms (e.g., Romance languages, Modern Greek,
Georgian, see Comrie 1976). In Slavic languages, they serve as input into the
derivation of specifically generic verbs by means of the generic morpheme like
-va- in Czech. When it is applied to their (past) stems it predictably restricts their
meaning to a generic interpretation. Such formally marked generic verbs are
typically subsumed under imperfective aspect, which, however, is not entirely
unproblematic, as I will discuss in the next section.
Minimal pairs like those in Figure 1, which only differ in the presence of –va- in
the more complex forms, instantiate a highly regular and productive morpholog-
ical pattern12 . Verbs with the morpheme -va- are common in all styles of speech
12
There are also irregular generic verbs: e.g., číst → čítat, vidět → vídat, slyšet → slýchat, sedět

22
 Generic sentences

(Kučera 1981: p. 177; Petr 1986). Moreover, it is also reported that children use
them early in their language acquisition, even before they master the grammati-
cal perfective/imperfective distinction (Chmelíčková 2005).
The high productivity of the generic morpheme -va- in Czech can be also seen
in the fact that it is reduplicated, as we see in Figure 4:

Figure 4: Reduplication of the generic morpheme -va- (see Bıĺ ý 1986)

ipf generic generic generic

mít → mívát → mívávát → mívávávát
have have.gen have.gen.gen have.gen.gen.gen
‘to have’ ‘to tend to ‘to tend to ‘to have once in a
have, sporad- have, very spo- great while’
ically/very radically/very
often’ often’
Verbs with the reduplicated generic morpheme -va- are not outliers, and are fairly
common, as the attested examples below show. Reduplication is often associated
with the speaker’s subjective judgment concerning the regularity of instances
that form the base for the generalization; it affectively connotes their strikingly
high or low frequency.

(14) a. Říkávává (say.gen.gen.3sg.pres) se, že po bitvě je každý generál.

‘As the saying goes, after the battle, everybody is a general.’ 13
b. Bude to z jiného konce, než bývává (be.gen.gen.3sg.pres) zvykem.
‘It will be approached from a different angle than is customary.’
While all Slavic languages have cognate generic morphemes, they are pro-
ductive only in West Slavic languages: namely, in Czech, Slovak (e.g., Kučera
1981; Bıĺ ý 1986, i.a.), and to a lesser degree also in Polish (Bıĺ ý 1986, i.a.). In East
and South Slavic languages, cognate generic markers tend to be unproductive,
at least in their standard varieties (see e.g., Širokova 1963: p. 62; Comrie 1976:
p.27; Kučera 1981; Bıĺ ý 1984; Bıĺ ý 1986; Petr 1986, among others). For instance, in
standard spoken and written Russian, the generic marker only occurs on some
verbs that are marginal in use (Comrie 1976: p. 27,fn. 2, who cites Isačenko 1962:
pp. 405–7), and often used in fixed collocations (e.g., v žizni tak byvaet - ‘that’s
how life goes’). The generic marker tends to be treated as a part of a lexicalized

→ sedat, ležet → léhat.

13
https://www.vidlakovykydy.cz/clanky/maginotova-linie-ceskoslovenske-opevneni accessed
13/12/2021

23
Hana Filip

combination with a verb base. Verbs with this marker are taken to be a feature of
a colloquial (‘substandard’) speech (Isačenko 1962: pp. 405–7; Forsyth 1970: p. 28
and 168-1971; Comrie 1976: p. 27; Kučera 1981: p. 177; Bıĺ ý 1984; Vinogradov 1986:
pp. 413–4; Petr 1986; Bıĺ ý 1986), where some of them are in fact quite frequent:
e.g., byvat’ (< byt’ ‘to be’), znavat’ (< znat’ ‘to know’), govarivat’ (< govorit’ ‘to
speak’), pivat’ (< pit’ ‘to drink’), siživat’ (< sidet’ ‘to sit’) and xoživat’ (< xodit’ ‘to
go’). Interestingly, the generic marker similar to that in West Slavic languages is
productive in some Northern Russian dialects (Barnetová 1956).
Cross-linguistically speaking, the formation of generic verbs by means of the
Czech morpheme -va- follows a pattern that is attested in a number of typo-
logically unrelated languages that have specifically generic forms, i.e., forms
that enforce a characterizing generic interpretation, and are at least a sufficient
condition for its expression: namely, generic forms are derived, morphologi-
cally or syntactically complex; they are the marked case relative to non-generic
forms, which are morphologically or syntactically simpler and alternate between
episodic and generic interpretations. Most importantly, we do not find generic
forms that are basic and episodic forms derived from them (Carlson 1995; 2005;
2012, and elsewhere).

4.2 The generic versus non-generic subsystem among Czech verbs

In traditional and structuralist linguistics, the Czech verb system is taken to have
four main types of verb forms, and their relations are structured by two main
distinctions: (i) perfective/imperfective and (ii) non-generic/generic in my terms,
aka ‘iterative/non-iterative’ (alternately ‘multiplicative/non-multiplicative’), in
traditional terms, which divides the imperfective domain. This is captured in Fig-
ure 5, which is adapted from the Czech Academy grammar (Petr 1986: Vol. 2, p. 180),
and examples for each type are given below Figure 5, along with their derivational
relations:

Figure 5

24
 Generic sentences

(b) štěkatIPF primary → (a) po.štěkatPF

bark.inf pref.bark.inf
‘to (be) barking’ ‘to bark (a little)’; metaphorically:
‘to bicker’, ‘to squabble’

(a) po.štěkatPF primary → (c) po.štěkávatIPF secondary

pref.bark.inf pref.bark.ipf.inf
‘to bark (a little)’; ‘to (be) bark(ing) (a little)’;
metaphorically: ‘to bicker’, … metaphorically: ‘to (be) bicker(ing)’,…

(c) po.štěkávatIPF secondary → (d) po.štěkávávatIPF generic

pref.bark.ipf.inf bark.ipf.gen.inf
‘to (be) bark(ing)/bicker(ing) ‘to (tend to) bark/bicker a little
(a little), on and off’ regularly, seldom, often, usually, …

(b) štěkatIPF primary → (d) štěkávatIPF generic

bark.inf bark.gen.inf
‘to (be) barking’ ‘to (tend to) bark regularly, seldom,
often, usually, …

In Figure 5, the input/output of the generic -va- is here indicated with arrows
pointing from non-generic imperfectives, primary (root) (b) or secondary (c) to
generic forms in (d). The four examples above share the same imperfective stem
‘bark’. The simplest form is the primary (root) imperfective in (b). The perfective
verb in (a) is formed from the stem ‘bark’ by adding to it the prefix po-, which may

25
Hana Filip

be here understood as contributing an attenuative meaning to the stem, roughly

amounting to ‘bark (a little)’; this perfective verb also has a metaphoric sense
of ‘to bicker’, ‘to squabble’. Adding the imperfectivizing suffix to the prefixed
perfective stem of (a) yields the secondary imperfective verb in (c).
According to standard reference grammars of Czech like Petr (1986: Vol. 2, p. 180)
which is reflected in Figure 5, the non-generic/generic distinction in my terms
(‘iterative/non-iterative’ or ‘multiplicative/non-multiplicative’ distinction in tra-
ditional terms) subdivides the imperfective domain. This means that the generic
member (‘iterative’, ‘multiplicative’) is just a different member of the same imper-
fective category system, on par with the non-generic member (say like the past
and present tense are two different members of the same tense category system).
If it were so, then both should pass most/all standard (formal and semantic) test
for Slavic imperfectivity. But the generic, unlike the non-generic, member does
not, and hence it does not neatly fit the Slavic imperfective paradigm. Subsuming
it under imperfectivity must, therefore, be at least questioned, if not rejected.
Specifically, generic (traditionally ‘iterative’ or ‘multiplicative’) verbs straight-
forwardly pass only one standard imperfectivity test, namely compatibility with
verbs that only select for imperfective verbs as their complements: phasal verbs,
as we see in (15), and the future auxiliary verb (see (20c) further below):

(15) Až si na tenhle druh komunikace zvykne třeba ti sám od sebe začne

psávat. 14
when refl on this type communication accustomed perhaps you.dat
alone from self begins write.gen.inf …
‘When he gets accustomed to this type of communication, perhaps he
will on his own start writing to you … ’ [i.e., regularly, on and off, as a
matter of habit …]

Most importantly, generic verbs fail one key test for Slavic imperfective verbs:
namely, the availability of the reference to particular situations, including those
that are ‘on-going’, ‘in-progress’. In contrast, all non-generic imperfective verbs
(including secondary ones formed with the imperfectivizing suffix) pass this test.
Generic -va-marked verbs fail this test, because the morpheme -va- with which
they are formed systematically precludes reference to particular situations. This
property is well-known in Bohemistics as its hallmark ‘non-actuality’ or ‘atem-
porality’ property, at least since Kopečný (1948) who coined it. (See also Kopečný
1962; 1965; Kopečný 1966: p. 259.) His paradigm example shows that –va-verbs

14
https://www.zpovednice.eu/detail.php?statusik=898813, accessed 13/12/2021

26
 Generic sentences

cannot be felicitously used in answers to questions about what is actually on-

going at the speech time. In contrast, their imperfective counterparts without
the morpheme -va- can (glosses and translations are mine):
(16) A: ‘What are you doing right now?’

B: # PsávámIPF jeden naléhavý dopis. ‘#‘ infelicitous, unacceptable

write.gen.1sg.pres one well-known letter
# ‘I write one urgent letter.’
I.e., as a rule / as a habit / often /rarely/ usually …

B’: PíšuIPF jeden naléhavý dopis.

write.1sg.pres one urgent letter
‘I am writing one urgent letter.’
The same contrast between imperfective verbs and specifically generic (or ‘habit-
ual’) verbs derived from them is also reported in Tlingit by Cable (2022). Unlike
imperfective verbs, however, habitual-marked verbs with nooch lack reference
to particular situations, i.e., they cannot be used to describe ongoing situations,
be they episodic or stative (denoted by lexically stative predicates).
The ‘atemporality’ or ‘non-actuality’ feature of the morpheme -va- along with
its productivity and compositional transparency (see Section 4.1) lead Kopečný
(1948) to treat it as a grammatical marker of a ‘third aspect’, which he labels ‘atem-
poral aspect’, which is independent of the perfective/imperfective distinction. If
so, as Kopečný further argues, then Indo-European languages have a grammati-
cal category of ‘atemporality’ in its own right, contrary to Koschmieder (1929)15 .
Recast in terms of the generic/non-generic distinction, Kopečný’s ‘third as-
pect’ proposal might be schematically represented as in Figure 6.

Figure 6

15
Kopečný (1948): “Není tedy zcela správné tvrzení Koschmiedrovo v jeho Nauce o aspek-
tech (33), že indoevropské jazyky nemají [153] gramatickou kategorii atemporálnosti (mi-
močasovosti, „pozaczasowości“).” [Koschmieder’s claim that Indo-European languages have
no grammatical category of atemporality (extra-temporality, „pozaczasowości“ [Polish]) is not
entirely correct.]

27
Hana Filip

While the ‘non-actuality’ or ‘atemporality’ property of verbs that are formally

marked with the generic morpheme -va- (aka ‘iterative’ or ‘multiplicative’ in tra-
ditional, but misleading, terms) is well-known in Bohemistics, it was Kopečný
(1948) (also Kopečný 1962; 1965; 1966) who properly recognizes it as their hall-
mark property, and draws the wide-reaching consequence from it that such verbs
should be viewed as independent from the main Slavic perfective/imperfective
distinction.
Put in contemporary terms, the ‘non-actuality’/‘atemporality’ property cor-
responds to the lack of reference to particular situations, which makes verbs
formed with the morpheme -va- aspectually stative (see e.g., Carlson 1977; Krifka
et al. 1995), and aligned with generic predicates:

Semantic Property 1: The Czech morpheme -va- derives verbs that

denote predicates that are aspectually stative, which is manifested in
the lack of reference to particular situations, and makes them aligned
with generic predicates.

Moreover, the independence of verbs that are formally marked with the generic
morpheme -va- from the main perfective/imperfective distinction, as proposed
by Kopečný (1948), can be supported by the observation that they do not pass
the standard tests for Slavic imperfectivity, unlike bona fide imperfective verbs
(both primary (root) and secondary derived the imperfectivizing suffix -va-), and
they clearly fail all perfectivity tests. I will address this point in detail in Section
5.2. To the extent that such insights can be viewed as bolstering the status of -va-
as a marker of genericity in its own right, as I argue, can also provide support for
the independent status of characterizing genericity with respect to the categories
of the TMA system.
To conclude this section, we can also dispense with two hurdles (among oth-
ers), which, according to Dahl (1995), prevent the Czech marker -va- from being

28
 Generic sentences

treated as a marker of genericity. Dahl (1995) claims that they exhibit a low de-
gree of grammaticalization (see Section 3.2). However, this would seem to be
contradicted by the view common in Bohemistics that the high productivity, reg-
ularity and semantic transparency of the Czech marker -va- indicates that it is
a morpheme with a high degree of grammaticalization, and one which enforces
an ‘atemporal’ interpretation of sentences, systematically in all its occurrences.
It is precisely the high productivity and regularity of the morpheme -va- that
Kopečný (1948) (and elsewhere) adduces in support of his argument that it en-
codes an independent grammatical category of ‘atemporality’ in Indo-European
languages.
Furthermore, Dahl (1995) relies on a highly restrictive, and implausible, de-
limitation of what it means to be a marker of genericity that the Czech marker
-va-, and similar markers in other languages, fails to satisfy: namely, a generic
marker occurs in all and only characterizing generic sentences (Dahl 1995: p. 421).
Only 3 out of the 76 languages in his corpus then qualify as ‘generic markers’ in
Dahl’s sense, Wolof (Niger-Kordofanian), Iṣẹkiri (Niger-Kordofanian) and Maori
(Austronesian) (Dahl 1995: p. 420, 424). There is, however, no good reason for us
to accept this restrictive definition of a generic marker, when in fact even Dahl
(1995) himself acknowledges that “it is indeed rare for a grammatical marker to
have a domain which is neatly delineated by a single semantic distinction. At
best, the distinctions that interest semanticists are relevant in the straightfor-
ward cases of a grammatical opposition, but even then secondary readings may
spoil what appeared to be nice generalizations” (Dahl 1995: p. 425).16

5 Some fallacies and bad attitudes

5.1 The fallacy of iterativity, a plurality of situations, events, ‘cases’
Genericity cannot be confused or confounded with iterativity. In contrast to
generic predicates, iterative predicates denote a mere plurality of situations, i.e.,
a series of particular situations of the same type at a given world/time that can
be counted. A closed set of plural (episodic) situations of the same type at a given
world and time does not amount to a generalization, a regularity, or a pattern.
16
While it is true that the process of grammaticalization is associated with increasing obligatori-
ness of linguistic structures, many scholars agree that absolute obligatoriness is not necessary
for grammaticalization to take place (see e.g., Lehmann 1995 [1982]: 12, Heine & Kuteva 2002;
2007, and others). For instance, in Lehmann (1995 [1982]: 12), obligatoriness is a useful crite-
rion, but it is not “an absolute one”, because “[s]omething is obligatory relative to the context;
i.e., it may be obligatory in one context, optional in another and impossible in a third context”.

29
Hana Filip

The marker -va- in Czech is commonly referred to as an “‘iterative’ suffix -va-”

(see e.g., Petr 1986, Dahl 1995, i.a.). Its status as a suffix is not uncontroversial
(see above), and the ‘iterative’ (iterativní in Czech) label is a misnomer at best
(Kučera 1981, Bıĺ ý 1986), and a semantic category error at worst (Filip 1993 and
elsewhere), as I will show in this section. The same holds mutatis mutandi for
‘multiplicativity’ (násobenost in Czech), another notion that is commonly associ-
ated with it.
A mere plurality of situations, implied by both iterativity and multiplicativity,
is the wrong way to think of the meaning of the Czech morpheme -va-. If it were
an iterative (or a multiplicative) operator, it should be compatible with iterative
adverbials like ‘three times’, ‘several times’, which count particular situations
(that are not a part of a larger pattern). But it is not, as the following semantically
odd, uninterpretable example shows:
(17) ⁇ Pavel hrával třikrát šachy s dědou.
⁇ Včera Paul play.gen.pst 3.times chess with grandpa.
⁇ ‘Paul used to play chess with grandpa three times yesterday.’
Intended meaning: A single plurality of three chess games.
The oddity of the above sentence must be due to the clash between the iterative
adverbial ‘three times’ and the generic marker -va-, because if -va- is dropped, we
get the corresponding non-generic sentence with a primary (root) imperfective
verb which is perfectly acceptable:
(18) Včera Pavel hrálIPF třikrát šachy s dědou.
Paul play.pst 3.times chess with grandpa yesterday.
‘Paul played chess with grandpa three times yesterday.’
Secondary imperfective verbs formed with the imperfectivizing suffix are also
straightforwardly compatible with iterative adverbials like ‘three times’:
(19) Sice jsem již třikrát vyhrávalaIPF , ale nikdy jsem nevyhrála.
while aux already 3.times pref.play.ipf.pst but never aux neg.win.pst
‘Although I was already winning three times, but I never won.’
The contrast between (17) and (19) shows that the generic morpheme -va- must
not be confounded with the imperfectivizing suffix, i.e., the morpheme impli-
cated in the derivation of the complex verb in (17) cannot be the imperfectivizing
suffix, but a different morpheme separate from it, which makes a different se-
mantic contribution to the meaning of a sentence. (I will return to this point
further below.)

30
 Generic sentences

The Czech generic -va- derives verbs that are incompatible with iterative ad-
verbials, as we have just seen (17), which is one manifestation of lack reference
to particular situations. But this means that verbs derived with it have one of the
key properties of generic predicates, namely being aspectually stative. This is
because, notionally speaking, generic predicates transcend particular situations,
conditions or facts. In contrast, iterativity amounts to a mere plurality of situa-
tions, i.e., a series of particular situations of the same type at a given world/time,
which can be counted; such closed sets of plural (episodic) situations do not con-
stitute a generalization.
Generally, incompatibility of generic (and habitual) predicates with iterative
adverbials clearly indicates that genericity (and habituality) must not be con-
founded with iterativity, ‘iterative aspect’, and other notions that presuppose a
plurality of cases, situations or events, such as event-internal or event-external
iterativity, frequentativity, pluractionality, multiplicativity, iterated semelfactiv-
ity, and the like. However, this confounding is fairly common in the analyses
of generic, and habitual, sentences (e.g., Nef (1986); Vlach (1993); Van Geen-
hoven (2001; 2004); Scheiner (2003); Rimell & Lue (2005); Bittner & Trondhjem
(2008)), and especially in analyses of generic (and habitual) readings of imperfec-
tive forms.
For instance, one common strategy is to add to the imperfective operator a
parameter related to a number requirement on its application to capture the dif-
ference between the episodic and generic (‘habitual’) interpretations of imperfec-
tive forms. A predicate of singular eventualities corresponds to an episodic (pro-
gressive) interpretation, predicate of plural eventualities, to a generic (‘habitual’)
interpretation. This idea is implemented in Ferreira (2016), also Ferreira (2004)
and Ferreira (2005)), who also assumes that the imperfective operator uniformly
introduces existential quantification over time intervals and universal (modal)
quantification over possible worlds, which is based on the modal (intensional)
semantics developed for the progressive (Dowty 1979; Landman 1992; Portner
1998).17 Hacquard (2006) proposes a similar analysis of the progressive (episodic)

17
The differences between progressive (episodic) and generic interpretations of imperfective
forms are reduced to the number parameter of time intervals, on the assumption that VPs
denote sets of time intervals, and VPs like NPs can be singular or plural. So we have SG(⟦VP⟧) =
{ e1, e2, e3, … }; PL(⟦VP⟧) = e1 ⊕ e2, e2 ⊕ e3, e1 ⊕ e3, e1 ⊕ e2 ⊕ e3, ….. The progressive (episodic)
and generic (and habitual) interpretations share the same temporal component (existential
quantification over time intervals) and modal component (universal (modal) quantification
over possible worlds, which is based on the modal(intensional) semantics developed for the
progressive (Dowty 1979; Landman 1992; Portner 1998). Under the progressive interpretation, a
VP denotes singular time intervals at which it holds, or a set of “singular” (sic) P-eventualities:

31
Hana Filip

and generic interpretations of imperfective aspect.

To the extent that such analyses of the generic (habitual) reading of imper-
fective forms depend on a plurality of situations, i.e., on a purely quantitative
criterion, they are inadequate, because it is neither sufficient nor necessary for
a generic interpretation of sentences. It is not sufficient, because a plurality of
situations can also be a part of the meaning of what clearly are non-generic sen-
tences, be it a sequence of temporally close situations that constitute a single
complex situation (Last night, he knocked three times on the door to be let in), or a
few situations spread over a whole life-span of an individual (During her whole
life Kim flew only three times).
Moreover, a plurality of situations is not necessary for the truth of generic sen-
tences, because there are generics whose truth is not inferable from any array of
situations at all. First, some generics (e.g., (6a) The Speaker of the House replaces
the Vice President) are true even if there are no actualized situations or cases that
count as evidence for their truth, and what is more, may/will never be any. Sec-
ond, there are generic sentences that are true based on the occurrence of one and
only one particular situation that serves as a precondition for the characterizing
property/state they denote: e.g., own (presupposes one situation of acquisition,
such as buying), be a student (one situation of registration), be married (one mar-
riage ceremony) (Krifka et al. 1995: p. 38). Finally, some generic sentences are
best viewed as independent of particular situations or conditions in the actual
world, because it makes no sense to try to identify them or because there are
none: e.g., Bob is a bachelor (Carlson 1995).
Most importantly, an analysis of genericity (or habituality) that relies on a
plurality of events (as in (Ferreira 2004, Ferreira 2005, Ferreira 2016, and Hac-
quard 2006) leaves aside its three hallmark properties (see Section 2). It does
not address aspectual stativity (manifested in the lack of reference to particular
situations), exception-tolerance, and intensionality tied to its law-like or gnomic
force. Assimilating the intensionality (modality) component of genericity (habit-
uality) to that of progressivity hardly makes any sense, given that the two have
entirely different sources and are of different modal nature. The intensionality
of the progressive is motivated by the problem of incomplete particular singular
situations, and also of incomplete particular objects (‘the imperfective paradox’
coined by Dowty 1979, see also Parsons 1990, Zucchi 1999, i.a.). A progressive sen-
tence asserts that a stage of a situation denoted by its base predicate occurred,
or is occurring, at some reference point (world-time pair), and it is true even if

∃e: … sg(P)(e) … . Under the generic (or habitual) interpretation, a VP denotes plural time
intervals, or a set of “plural” (sic) P-eventualities: ∃e: … pl(P)(e) …. (Ferreira 2016: p. 358).

32
 Generic sentences

some ‘fatal interruption’ prevents that stage from developing further into larger
stages and reaching the end-point/final stage of a situation specified by the base
telic predicate (e.g., John was crossing the street when a bus hit him, and killed him
on the spot.) The intensionality of generic and habitual sentences is of a different
kind. It stems from their gnomic, law-like or predictive import: namely, they
express generalizations that are true not just by virtue of what might have al-
ready occurred with some regularity (and what may inductively projected from
observed instances), but also by virtue of some underlying causes and forces, and
they predict what might be possible in the future, and support counterfactuals.

5.2 The Czech marker-va- is a marker of tense or aspect

In this section I will provide arguments that the Czech morpheme -va- cannot
be subsumed under markers of either tense or aspect, that is, it is not a kind of
‘tense-aspect marker’, contrary to Dahl (1985; 1995), neither can it be analyzed
in terms of notions developed for the analysis of tense and aspect markers. To
the extent that similar markers in other languages, for which -va- stands as a
paradigm example, according to Dahl (1995), cannot be treated as markers of
tense and/or aspect, we may seriously consider the possibility that they, just like
the Czech morpheme -va-, be analyzed as markers of genericity.
If this is correct, then we would have to acknowledge that characterizing gener-
icity has direct grammatical reflexes in natural language, pace Dahl (1995), and
its status as a category in its own right, independent from both tense and aspect,
would then be based on a firmer ground. While previous arguments for its inde-
pendent status with respect to other categories of natural language were mainly
made on semantic grounds and based on notions developed in philosophy, artifi-
cial intelligence and computer science (disposition, law-like regularity, capacity,
probability, and the like) (see e.g., Carlson 1977; 1982; Krifka et al. 1995; Pelletier
& Asher 1997), here the relevant arguments are also based on the formal means
for the encoding of genericity, as opposed to tense and aspect.

5.2.1 Genericity versus tense

Let us start with the straightforward argument that the Czech -va- cannot be
viewed as a tense marker. Notionally, genericity and tense are independent of
each other, tense is a deictic category, while genericity is non-deictic. In tensed
languages like English, characterizing generic sentences can be in any tense form,
and refer to present, future or past time: e.g., Corruption starts at the top, Men were
deceivers ever, The poet will go to any end to make a rhyme. Moreover, characteriz-

33
Hana Filip

ing generic sentences are compatible with any sufficiently large temporal inter-
val, (contextually) located in the present, past or future; they may also express
generalizations that are restricted in time and located in the present, past or fu-
ture (e.g., The current President eats broccoli, Dinosaurs (usually) ate kelp, Starting
next Monday this office will be open only from 2 p.m. to 4 p.m.).
Formally speaking, genericity and tense markers are independent. There are
tensed languages with generic markers (e.g., Czech, arguably), without generic
markers (e.g., English), and tenseless languages with specific generic markers
(e.g., American Sign Language) and without them (e.g., Chinese), in which case
they convey generic statements by other means (e.g., in Dyirbal and Burmese by
means of a modal distinction between realis and irrealis; Comrie 1985: p. 51).
In Czech, the morpheme -va- and overt markers of tense are fully orthogonal
to each other: they co-occur on the same verb, or occur independently of each
other. E.g., -va- occurs on non-finite forms like the infinitive and imperative (for
examples see Filip & Carlson 1997), and it also freely co-occurs with all the overt
tense markers (past, present and future) on the same verb or in the same verb
complex, as we see in the examples below.

(20) a. Karel hrával hokej. past tense

Charles play.gen.past hockey
‘Charles used to play hockey.’ [remote past]
b. Karel hrává hokej. present tense
Charles play.gen.present hockey
‘Charles tends to play hockey.’ [i.e., regularly, often, seldom, … ]
c. Karel bude hrávat hokej. future tense
Charles be.aux.fut play.gen.inf hockey
‘Charles will tend to play hockey.’ [i.e., regularly, often, seldom, … ]

In the past tense, generic sentences formally marked with -va- have a remote past
time reference, whereby its reduplication entails a high degree in the past tem-
poral remoteness (Kučera 1981; 1983; Bıĺ ý 1986); in both the cases, -va- triggers a
conversational implicature that the described situation no longer holds.
(21) Vídávával jsem ho tam ve čtvrtky. (Chmelíčková 2005)
see.gen.gen.1sg.past aux him there on Thursdays ‘I used to see him
there on Thursdays once in a while / regularly … a very long time ago.’
The fact that the marker -va- does not stand in complementary distribution to
overt tense markers is the most compelling argument against its classification
as a tense marker, contrary to Dahl (1995). The argument relies on the standard

34
 Generic sentences

assumption about what it means to be a ‘formal member of a category system’:

namely, it must stand in complementary distribution to other formal members of
the same category, morphologically and syntactically. For instance, the present
and past tense morphemes in English are formal members of the tense category.
They are in complementary distribution, because the formal expression of one
precludes the expression of the other on the same verb: e.g., *she laugh-s-ed. The
markers of tense and progressive aspect in English can co-occur, because they
are not formal members of the same grammatical category: e.g., she was/is/will
be laughing. They also each make an independent semantic contribution to the
meaning of a sentence, which also holds for overt markers of tense and the
generic morpheme -va- in Czech.

5.2.2 Genericity versus aspect

Let us now turn to constructing a parallel argument for the independence of
the marker -va- from aspect, and also continuing the argument against Dahl’s
(1995) claim that it, along with similar markers in other languages, is a kind of
‘tense-aspect’ marker. The ‘aspect’ part of Dahl’s claim is to be understood as
meaning that such markers fall specifically under imperfective aspect, say on a
par with other bona fide markers of imperfectivity like the imperfectivizing suf-
fix in Slavic languages, or Romance imperfective suffixes. This follows given that
Dahl (1995) presupposes the widespread view that genericity (and habituality) is
a subcategory of imperfectivity, and the use of perfective verbs in generic sen-
tences is a marginal phenomenon in natural languages. However, as I will show
here, empirical facts do not support either of these two claims.
One of the well-known versions of this erroneous view is Comrie’s (1976) clas-
sification of aspectual oppositions (Comrie 1976: p. 25, Table I and 26ff.). Com-
rie (1976) just like many others empirically motivates the treatment of generic-
ity (and habituality) as a subcategory of imperfectivity by the observation that
a number of languages (e.g., Romance, Slavic, Georgian, Modern Greek) have
imperfective forms that freely alternate between episodic and generic (habitual)
readings depending on context. By way of clarification, let me mention that Com-
rie (1976) uses the term ‘habituality’, rather than genericity, but his examples in-
dicate that his ‘habituality’ notionally corresponds to ‘characterizing genericity’
in the sense of Krifka et al. (1995) (see Section 2). This widest use of the term
‘habituality’ is widespread in aspect studies, and Comrie (1976) is often cited in
this connection.
The key issue that arises for Slavic languages is the delimitation of the imper-
fectivizing suffix, which is productive in all of them, from the generic (‘habitual’)

35
 Hana Filip

morpheme, which is productive in West Slavic languages, while in South and

East Slavic languages it is attested, though less or not productive (see above Sec-
tion 4.1). This separation has general implications for the separation of genericity
(habituality) from imperfective aspect. Now, to the extent that one finds plausi-
ble the above argument for the independence of the Czech generic morpheme
-va- from tense markers, one should also accept a parallel argument for the inde-
pendence of -va- from the imperfectivizing suffix, which is an indisputable im-
perfective marker, based on the same logic of complementarity of paradigmatic
oppositions. We start with the observation that the generic morpheme -va- and
the imperfectivizing suffix co-occur on the same verb, as illustrated in (22), which
contains the verb given in Figure 1, in (ii). Such morphologically complex verbs
by no means constitute some kind of outlier case or oddity of the Czech system,
and natural examples are easy to find.

(22) Bývalý mi taky dávávalIPF nádherně a originálně vázané kytice, přímo

umělecká díla, …
ex me.dat also give.ipf.gen.pst wonderfully and originally bound
bouquets directly art works, …
‘My ex also used to give me beautifully bouquets in original
arrangements, real art works, …’

Given that the generic marker -va- and the imperfectivizing suffix co-occur on the
same verb, they do not stand in a paradigmatic opposition to each other, unlike
the present and past tense morphemes in English, for instance. Therefore, they
cannot be two different markers of the same grammatical category system, unlike
the present and past tense morphemes in English, but rather each must be a
marker of a different system. If the imperfectivizing suffix is a bona fide marker of
imperfectivity, as all agree, the generic morpheme -va- cannot be another marker
of imperfectivity.
The formal independence of the generic marker -va- and the imperfectivizing
suffix in Czech fits the general pattern of orthogonality of markers of genericity
and aspect (imperfectivity) in natural languages. All their four possible encod-
ing combinations are attested. Languages with grammatical aspect may have
specific generic (and habitual) morphemes (Czech) or not (French). Languages
with no grammatical aspect may have specific generic morphemes (Guarani, see
Dahl 1985), or lack both aspect and generic morphology (German) and express
genericity by other means.
Returning to the argument in support of the separation of the imperfectiviz-
ing suffix and the generic marker -va-, one might object that what is taken to be

36
 Generic sentences

a sequence of these different markers on the same verb, as in (22) above, is in

fact a reduplication of one and the same imperfectivizing suffix, or some such
proposal along these lines. For instance, as (Cable 2022) proposes, the Czech
generic (habitual) morpheme could be interpreted as “etymologically an instance
of imperfective-aspect combining with another, lower imperfective-form of the
verb, a kind of ‘doubly aspectualized’ verb”. So the first (lower) instance forms
a non-generic, or ‘general’, secondary imperfective counterpart from the perfec-
tive base (see Figure 2 and Figure 3), while the second (higher) instance derives
a specifically generic verb and restricts the denotation of a whole sentence to a
characterizing generic one (see Figure 1, (iia) and (iib)).
There are at least three counterarguments against this reduplication rebuttal.
First, it seems to imply some economy principle that it would be superfluous to
reduplicate one and the same imperfectivizing suffix to do the same aspectual job
of marking the whole domain of imperfectivity. This, however, is not compelling.
After all reduplication without a change in the truth conditions of sentences is
common in natural languages (see also (⁇)).
Second, there are verbs with only a generic interpretation, despite having just
a single occurrence of the alleged imperfectivizing suffix. This is the case when
the input is an imperfective simplex that carries no overt imperfective marking.
An example is given in Figure 1 (i), repeated here for convenience: psát ‘to (tend
to) write’/ ‘to be writing’ (episodic/generic interpretation) → psávat (generic in-
terpretation) ‘to tend to write, to write as a rule, regularly, on and off, seldom,
often …’ . One might try to rescue the reduplication view by proposing that
the imperfective simplex is not morphologically ‘simplex’ at all, as standardly as-
sumed, but rather contains a phonologically null (covert) ipf suffix. The requisite
reduplication (if it can still be called that), would then involve one covert ipf suf-
fix and one overt. But this would force on us a highly undesirable consequence of
having to radically change the analysis of all imperfective simplex verbs in Czech
(and possibly also in all Slavic languages) and postulating that they all introduce
a null ipf suffix with no independent empirical motivation, but to obviate the
existence of a generic marker for purely theory-internal reasons. Arguably we
fare better by acknowledging its existence as a generic marker of its own kind,
as it is taken for granted in standard reference grammars like Petr (1986) (albeit
they do not use the label ‘generic’ morpheme for it, but instead, most often, the
misleading ‘iterative’ label).
Third, if verb forms that exclusively have a generic interpretation were formed
with the alleged imperfectivizing suffix, then they should pass the standard tests
for imperfectivity in Slavic languages, just like secondary imperfectives formed

37
Hana Filip

with the uncontroversial imperfectivizing suffix do. But this prediction is not
borne out. For example, secondary imperfectives, as well as primary imperfec-
tives, are used with reference to an on-going situation at some reference time
(a ‘progressive’ reading). However, generic verbs formed with -va- pattern with
generic predicates in so far as they lack the reference to particular situations,
which in Bohemistic studies is known as their ‘atemporality’ or ‘non-actuality’
property (see Section 4.2).

(23) a. Včera v deset hodin, když ji dával pusu na dobrou noc, …

yesterday at ten o’clock when her.dat give.ipf.past kiss.sg.acc for
good night, …
‘Yesterday at 10 p.m., when he was kissing her good night, … ’
b. ⁇ Včera v deset hodin Pavel hrával šachy s dědou.
⁇ yesterday at ten o’clock Paul play.gen.past chess with grandpa.
? ‘Paul used to play chess with grandpa yesterday at 10 p.m.’

In sum, we have seen that the generic morpheme -va- and the imperfectiviz-
ing suffix do not stand in complementary distribution, and therefore cannot be
members of the same category system. Each has a different input requirement,
imperfective base and perfective base, respectively (see Section 4.1 above). Each
makes a different contribution to the meaning of sentences. The chief semantic
difference is that generic sentences formally marked with -va- are aspectually
stative, as they lack reference to particular situations, while secondary imperfec-
tives formally marked with the imperfectivizing suffix freely alternate between
episodic and generic readings, hence are non-generic and not aspectually sta-
tive. All of the above leads to the unequivocal rejection of Dahl’s claim that the
Czech morpheme -va- is a marker of imperfective aspect (say on a par with the
imperfectivizing suffix).
The formal and semantic separation of the generic morpheme -va- from the
imperfectivizing suffix has broad implications that go well beyond the nature
of the Slavic verb system. It provides an additional support to the separation
of genericity (and habituality) from imperfective aspect. A further supporting
argument for this separation comes from the fact that a number of languages
have perfective forms that are also used for the expression of generic (habitual)
statements. But this means that genericity (and habituality) and grammatical
aspect cross-classify, and so they, and their subcategories, are independent of
each other. I will turn to this point next.

38
 Generic sentences

5.2.3 Perfective verbs in generics: ‘A case of deviant behavior’

The view that genericity (and habituality) is a subcategory of imperfectivity,
which I argue against, is often accompanied by the implicit or explicit assump-
tion that perfectivity formally and semantically is incompatible with genericity
(and habituality). As mentioned above, such ideas inform Comrie’s (1976) in-
fluential classification of aspectual oppositions (Comrie 1976: p. 25, Table I and
26ff.). What Comrie’s classification ignores, and indeed excludes, is the fact that
in a number of languages not only imperfective forms, but also perfective forms
freely alternate between episodic and generic (habitual) readings. For instance,
in Romance languages like French and Italian, passé simple and passato remoto,
respectively, are commonly used in sentences with a generic (habitual) reading.
Consider the following examples (the Italian example is taken from Bonomi 1997:
p. 508):

(24) a. Sempre, quando mi videPF , il custode aprìPF la porta. italian

always when me saw the janitor opened the door
‘Always when the janitor saw me, he opened the door.’
b. Toujours, quand il parlaPF , il déterminaPF le vote du conseil. french
always when he spoke he determined the vote of the Board
‘Always, when he spoke, he determined the vote of the Board.’

Perfective verbs are also used in generic sentences in Slavic languages, com-
monly in West Slavic (Czech, Slovak, Polish), and less often in East (e.g., Russian)
and South Slavic (e.g., Serbo-Croatian) languages (Forsyth 1970; Fortuin & Kam-
phuis 2015; Wiemer & Seržant 2017). Some examples are given below:

(25) a. Ja codziennie przepalęPF F 20 papierosów. polish

I every.day smoke 20 cigarettes
‘I smoke 20 cigarettes every day.’ Lenga 1976: p. 46
b. Nastojaščij drug vsegda pomožetPF . russian
real friend always helps
‘A real friend will always help you.’
c. Svako jutro popijemPF čašu rakije. serbo-croatian
every morning drink glass brandy
‘Every morning I drink a glass of brandy.’ Mønnesland 1984: p. 62

In Czech, even kind predicates are expressed by perfective verbs and construc-
tions, i.e., predicates like ‘become extinct’, ‘die out’, ‘invent’ that select for kind-
referring terms in one of their argument positions (see Section 2):

39
 Hana Filip

(26) a. Mamuti vymřeliPF . czech

mammoth.pl.nom died.out
‘Mammoths died out.’
b. Jízdní kolo bylo vynalezenéPF v roce 1817.
riding.wheel was invented in year 1817
‘The bicycle was invented in 1817.’
Moreover, generic and perfective markers freely co-occur in a single verb form
and are used for the expression of generic statements: e.g., Lithuanian ‘frequenta-
tive past’ (Dambriunas, Klimas & Schmalstieg 1966), Awa (New Guinea) (Loving
& McKaughan 1964), among others, and given that perfective verb forms are com-
monly used for the expression of characterizing genericity (see Filip & Carlson
1997). Of particular interest for the arguments in support of the independence of
genericity (habituality) and grammatical aspect is Tlingit (Na-Dene) (Cable 2022):
namely, it has a ‘habitual perfective-mode’ of a verb that is realized via the com-
bination of certain conjugation-class morphemes with a dedicated habitual suffix
-ch, as well as a ‘habitual imperfective-mode’ of a verb that is formed with the reg-
ular imperfective mode form followed by the habitual particle nooch. If both im-
perfectivity and perfectivity are straightforwardly compatible with generic (and
habitual) interpretations, and generic markers can co-occur with/on both imper-
fective and perfective forms, then we must acknowledge that genericity (and ha-
bituality) and grammatical aspect cross-classify, and so are clearly independent
of each other.
This should suffice to show that there is no formal and semantic incompati-
bility between perfectivity and genericity (and habituality), contrary to Comrie
(1976), and also contrary to Dahl (1995: p. 420) who assumes that the expression
of generic statements by means of perfective forms is so highly unusual in natu-
ral languages as to not even make the list of exceptions to his minimal marking
tendency of genericity (see Section 3). Specifically, in Slavic languages, it is
yet another “case of deviant behavior” (Dahl 1995: p. 420) of their aspect systems.
This claim is plain wrong.
The use of perfective verbs for the encoding of generic statements in Slavic
languages is their systemic, regular feature, rather than a bug, or an aberration.
However, the frequency as well as the functional range of perfective forms in
generic statements differs across Slavic languages. There is a divide between
West Slavic languages, where perfective verbs are common in generic statements,
while in East and South Slavic languages, they are more rarely used (Forsyth
1970; Fortuin & Kamphuis 2015; Wiemer & Seržant 2017), though perhaps not
as rarely as is generally thought (Wiemer 2023). In Czech (West Slavic), for in-

40
 Generic sentences

stance, perfective verbs are preferred to the other verb forms in certain types of
generics, such as cooking recipes, assembly instructions, game rules like rules
of chess, proverbs, to name just a few, and are also naturally used in colloquiual
speech (e.g., To se jeden nalítá (pf) v úřadovnách prezidentského paláce - ‘What a
runaround one must take in the offices of the presidential palace.’).18 .

5.2.4 Summary: Independence of the -va- from tense and aspect

If the Czech morpheme -va- is not a marker of tense or (imperfective) aspect, as
I have just shown, and given that the prima facie semantic properties of verbs
it forms aligns them with generic predicates (e.g., stativity), it is plausible to hy-
pothesize that it is a good candidate for a marker of genericity. If -va- is also a
paradigm example of a class of similar markers attested in a number of typologi-
cally distinct languages, as Dahl (1995) suggests, the questions for future research
are: ‘What is the status of similar so-called ‘tense-aspect’ markers (as Dahl 1995
labels them) with respect to the categories of the tense and aspect systems? If
they cannot be subsumed under tense or aspect, should they also be better treated
as generic markers?’
The independence of the Czech marker -va- from tense and aspect is here also
used to make a general point about the independence of genericity (habitual-
ity) from both tense and aspect. Acknowledging this insight is not to deny that
there are notional and formal affinities between genericity (habituality) and the
categories of tense and imperfective aspect. For instance, in English the simple
(non-progressive) present tense has predominantly generic (and habitual) inter-
pretation. There are also languages in which dedicated morphemes that enforce a
generic interpretation of sentences (which in standard reference grammars are of-
ten misleadingly labeled as ‘habitual’, ‘frequentative’, ‘iterative’ and the like) are
obligatory or highly preferred only in some tenses. In Haida (a North American
Native language), for instance, they are obligatory in the past, but not present,
tense (Enrico 2003). There are also semantic affinities between characterizing
genericity (habituality) and imperfectivity. For instance, basic aspectually sta-
tive predicates in Slavic languages are expressed by imperfective verbs, and as-
pectual stativity is a key semantic property of generic predicates. Moreover, from
the grammaticalization point of view, the expression of characterizing generic-
ity (habituality) is tied to markers that become grammaticalized on the path from
(present) progressive markers to imperfective ones (Bybee & Dahl 1989).

18
https://www.kosmas.cz/oko/ukazky/430121/o-panovnici-elite-prezidentu-povysenci-a-
polarnikovi-v-nervoznim-pohybu/, accessed July 9, 2023

41
Hana Filip

5.3 ‘Habitual’ quantifiers over situations

As mentioned above, Dahl (1995) proposes that the Czech marker -va- and by the
same token, according to him, a number of similar markers in other languages
for which -va- stands as a proxy, are ‘habitual’ markers that function as quanti-
fiers over situations. The view that -va- quantifies over situations or events only
is implicit in its standard ‘iterative suffix’ or ‘multiplicative suffix’ label used in
standard reference grammars of Czech (e.g., Petr 1986). This idea is also in com-
pliance with the way in which similar markers in other languages are viewed:
e.g., the Swahili prefix hu- (see (11b) in Krifka 1995), the Tlingit ‘habitual’ parti-
cle nooch (Cable 2022), the be construction in African American English (Green
2000), and the construction with haya+participle in Modern Hebrew (Boneh &
Doron 2008).19
However, this proposal must be rejected for the Czech marker -va- for two
main reasons. First, its domain of quantification is not restricted to situations
or events. It commonly directly quantifies over individuals only. This is shown
in the following example, which is a paradigm example of a ceteris paribus (cp)
generalization, and in which -va- quantifies over individual members of the kind
raven:
(27) a. Vrány bývají černé.
ravens be.gen.3pl.pres
‘Ravens tend to be black.’ / ‘Ravens are black, ceteris paribus.’
b. va[x](raven[x]; black[x])
Indeed, the Czech marker -va- functions as a quantifier that ranges over the
same three main ‘cases’ (in the sense of Lewis 1975, see Section 2 above) that the
null generic quantifier gen and overt Q-Adverbs do:

(i) situations, as in (12a),

(ii) individuals, as in (27a), and
(iii) both situations and individuals, as in (31) (further below).

As Filip (2009a) shows, -va- introduces a quantifier into the logical representa-
tion that has variable-binding properties that pattern with the null generic quan-
tifier gen, which, according to Chierchia (1995), gen in turn shares with overt
Q-adverbs:

19
“We represent habituality by means of an operator Hab, an intensional summation of events,
which is distinct from the generic operator Gen.”

42
 Generic sentences

(28) Variable-binding properties of the generic morpheme -va- in Czech:

(i) a situation variable
(ii) variables provided by singular indefinites and bare plurals
(iii) variables provided by kind-denoting definites
(iv) more than one variable
(v) relatively freely select the arguments for its binding, modulo context.
It is also worth mentioning that -va-, perhaps surprisingly, also formally marks
kind reference sentences (see Section 2), i.e., in sentences that contain kind
predicates like widespread that directly select kind-denoting terms for one of
their arguments, as in (29a). It is also used in characterizing generic sentences
that predicate generic properties of kind-denoting terms like ‘man’, as in (29b):
(29) a. Rohozub nachový bývá rozšířený u lidských sídlišť.
ceratodon purpureus be.gen.3sg.pres widespread at human dwellings
‘Fire moss tends to be widespread close to human dwellings.’
b. Člověk se k stáru měnívá.
man refl toward old.age change.gen.3sg.pres
‘Man changes as he grows old.’ (Čapek, Ordinary Life, 1934)
Second, the label ‘habitual(ity)’ is problematic as a designation for the whole
range of the functions/uses of -va-, and should be best avoided. The main reason
for this is that in contemporary genericity and aspect studies, the label ‘habit-
ual(ity)’ is not used consistently for the same phenomenon, and some even view
it as separate from genericity (though again the term ‘genericity’ may not be well
defined or understood). In its narrowest sense, ‘habitual(ity)’ is reserved for char-
acterizing generic sentences like (3c) Paul plays chess on Saturdays that describe
habits proper, i.e., regularities of action predicated of an ordinary individual, typ-
ically human agents, but not of kinds (Pelletier & Asher 1997; Pelletier 2010). In
a wider sense, ‘habitual(ity)’ is used for all characterizing generic sentences that
“express generalizations over situations that are specified by the corresponding
episodic verbal predicate” Krifka et al. (1995: p. 32,[56), which covers both gener-
alizations over situations with ordinary individuals, such as (3c) Paul plays chess
on Saturdays, and also with kinds like (3b) Dogs bark (see Section 2). Neither of
these two uses of ‘habitual(ity)’ are adequate for -va-, because they fail to cover
its function as a quantifier over individuals only.
In its widest sense, ‘habitual(ity)’ covers all the uses of (imperfective) sentences
that are not episodic, as in Comrie (1976), for instance, and others who follow his
suit. Judging according to Comrie’s examples, ‘habituality’ in his sense is co-
extensional with ‘characterizing genericity’ in Krifka et al. (1995) (see Section 2).

43
Hana Filip

So if this sense of ‘habitual(ity)’ were meant, labeling the Czech marker -va- as
‘habitual’ would not be incorrect, but it would invite confusion with used in the
other two narrower senses. Therefore, ‘habitual(ity)’ should be avoided in favor
of ‘(characterizing) genericity’.

5.4 Irreducibility to ‘usually’ and to any other single quantifier

In standard reference grammars of Czech, the meaning of the marker -va- is often
characterized in terms of ‘usuality’ (uzuálnost in Czech) (Petr 1986: Vol. 2, p. 184,
i.a.), which is also in the spirit of Dahl’s (1995) claim that it semantically functions
as a quantifier over situations with the meaning of approximately ‘most’ or the
Q-Adverb ‘usually’ (see Section 3.2). Dahl’s (1995) reductionist ‘most’/‘usually’
proposal is based on two rather flimsy data points: namely, “the glosses given in
Kučera (1981)” (Dahl 1995: p. 421) for the relevant Czech examples that contain
‘usually’ or ‘most’, and a perfunctory observation about the tendential occur-
rence of the Czech marker-va-, and corresponding markers in other languages,
in Dahl’s two diagnostic “‘generic’ contexts in the wide sense” (Dahl 1995: p. 420-
421):

(i) [Q: What your brother usually do after breakfast? A:] He write letters.
(ii) [A: My brother works at an office. B: What kind of work he do? A:] He write
letters.

Dahl (1995) seems to imply that the Czech marker -va- is optional in the transla-
tion of (i) with the Q-Adverb ‘usually’, and typically absent in (ii)20 .
While the above observations are true, they are insufficient as empirical evi-
dence for analyzing the meaning of -va- in terms of a quantifier over situations
with the meaning of approximately ‘most’ or ‘usually’, contrary to Dahl (1995).
It is fairly easy to show. First, the Czech marker -va- freely co-occurs with virtu-
ally any expression of quantification or quantity (see, for instance, corpus studies
of Širokova 1963: p. 62, 81; Širokova 1965 and Danaher 2003). Danaher’s (2003)
corpus study shows that -va- occurs less often with obvykle ‘usually’ than with
Q-Adverb(ial)s denoting a low frequency like občas ‘from time to time’, někdy
20
The context (i) tends to elicit generics that are roughly aligned with Carlson’s 1995 inductive
model of genericity (weak descriptive generalizations), while (ii) with his rules-and-regulations
model (see above Section 2). The latter concerns a job description, and hence what matters for
its truth is the knowledge about the skill that the office job in question requires, and it may be
true even if it remains merely hypothetical, if the job holder has not yet had the opportunity
to put it to use.

44
 Generic sentences

‘sometimes’, málokdy ‘rarely’, tu a tam ‘here and there’, vzácně ‘rarely’21 . Any of
these Q-Adverb(ial)s can fill the [q-adverb(ial)] slot of (30) (which is a modified
version of Dahl’s (1995) generic ‘usually’ context (i):

(30) Po snídani Tomáš [q-adverb(ial)] psává dopisy.

after breakfast Thomas [q-adverb(ial)] write.va.pres letters
‘Thomas [q-adverb(ial)] writes letters after breakfast.’

In this respect, the Czech generic -va- patterns with Tlingit’s ‘habitual’ markers,
namely the perfective habitual-mode suffix -ch and the imperfective habitual-
mode particle nooch, which, as Cable (2022) observes, are compatible with adver-
bial expressions of any quantity, both with e.g., ‘always’ and ‘sometimes’.
Second, the most compelling argument against analyzing the Czech -va- in
terms of ‘most’ or ‘usually’ is that it formally marks generic sentences that are
true even if most relevant instances do not satisfy the generically-predicated
property, as we see in (31),
(31) Za Putina, ruští oligarchové umírávali pádem z okna.22 true
during Putin Russian oligarchs die.gen.past fall from window
‘In Putin’s times, Russian oligarchs tended to die by falling out of
windows.’
What is more, adding ‘most’ or ‘usually’ to formally generic sentences like (31)
renders generics that are false, as we see in (32) and (33):

(32) Za Putina, většina ruských oligarchů umírávalo pádem z okna. false

during Putin Russian most oligarchs die.gen.past fall from window
‘In Putin’s times, most Russian oligarchs died by falling out of windows.’

(33) Za Putina, ruští oligarchové obyčejně umírávali pádem z okna. false

during Putin Russian oligarchs usually die.gen.past fall from window
‘In Putin’s times, Russian oligarchs usually died by falling out of
windows.’

21
Danaher’s (2003) corpus comprises 376 attested examples of verbs marked with the marker
–va-, collected from a wide variety of text genres (e.g., contemporary literary essays, fiction,
memoirs, journalistic prose and scholarly writing).
22
This example is tailored on Kučera’s (1981, 1999) example: Za Stalina ruští generálové umírá-
vali v mladém věku. [during Stalin Russian generals died.gen in young age] ‘In Stalin’s times,
Russian generals tended to die young.’ Kučera translates this sentence as ‘Most generals died
young in Stalin’s times’, but this does not seem to be correct, given that what it describes is
factually false; most Russian generals did not in fact die young in Stalin’s times, as we know.

45
 Hana Filip

The truth of (31) does not depend on most Russian oligarchs dying by falling out
of windows. In fact the quantity of the referents of subject argument has almost
nothing to do with its truth, but rather that we judge the property of falling out
of windows as so highly appalling, striking, unexpected, or memorable when it
comes to Russian oligarchs that we treat it as a characterizing property of the
kind russian oligarchs in Putin’s times. This kind of subjective evaluation fol-
lows the same psychological mechanism that motivates the truth of generic sen-
tences like Mosquitoes carry the West Nile Virus, mentioned above. This sentence
is judged true, even if only about 1% of mosquitoes carry the virus.
In sum, when it comes to the quantificational properties of the Czech marker
-va-, just like the null generic operator gen, its meaning cannot be reduced to
any known extensional quantifier or an expression of quantity (see Section 2):

Semantic Property 2: The Czech generic morpheme -va- on its own

contributes no requirement on a specific quantity of situations or
instances that count as evidence for the truth of generic sentences
formally marked by it.
corollary: It cannot be analyzed in terms of any single extensional
quantifier (like ‘usually’ or‘most’), any single expression of quantity, or
any single quantitative criterion or quantity-based measure.

5.5 Summary and questions

We can safely reject Dahl’s (1995) claim (see Section 3.2) that the Czech marker
-va- is not a marker of genericity, because all his arguments in its support are in-
valid. Contrary to Dahl 1995, the Czech marker -va- (i) is independent from tense
and aspect, (ii) is not a ‘habitual’ marker, and (iii) does not have meaning that can
be reduced to a quantifier over situations with the meaning of roughly ‘usually’
or ‘most’, as is often mistakenly assumed, also in standard reference grammars
of Czech. Rather, in its function as a quantifier, it ranges over the same ‘cases’ (in
the sense of Lewis 1975) as the null generic quantifier gen does (see Section 2 and
(28) above). We may hypothesize that the Czech marker -va-, and by the same to-
ken cognate markers in West Slavic languages (Slovak and Polish, in particular)
where they are also productive, are good candidates for markers of genericity in
their own right, which justifies the following main questions:

qestion 3a: What kind of generic sentence does the generic morpheme in West
Slavic languages delimit?

46
 Generic sentences

qestion 3b: What is the relation of the West Slavic generic morpheme to the
generic gen operator, does it introduce an operator into the logical representa-
tion of sentences that is distinct, but possibly related, to gen?

6 Generic morphology, (c)overt q-adverbs and gen

As I observed at the outset, markers like the Czech -va-, which enforce a char-
acterizing generic (habitual) interpretation of sentences, have so far remained
largely unexplored in contemporary genericity studies that combine insights
from formal semantics, philosophy, psychology and other closely related do-
mains of cognitive science, as outline in Section 2. Among the few proposals
that there are, two stand out, Chierchia (1995) and Cable (2022), as they most
directly bear on qestions 2a and 2b. I will turn to them next.

6.1 gen and overt Q-Adverbs as lincensors of (c)overt ‘Hab’

According to Chierchia (1995: p. 197), “genericity manifests itself overtly in the
aspectual system of a language” (most likely the imperfective system might be
meant), whereby all languages have a habitual morpheme Hab, which is a func-
tional head in an aspectual projection. In some languages, Hab is realized by
explicit aspectual morphemes, presumably like the Swahili prefix hu- (see Krifka
et al. 1995), while in others, as in English, it is covert. In English, the simple
present tense has a predominently generic interpretation, which is aspectually
imperfective (Chierchia 1995: p. 197), and the simple past and future can also
have generic interpretations. The semantically relevant property of Hab is the
agreement feature [+Q] (Hab[+Q]) which requires the presence of a Q-Adverb
in its Spec: namely, either some overt Q-Adverb (Chierchia 1995: p. 197-8) or the
gen operator, which is a phonologically null (modalized) universal Q-Adverb
(Chierchia 1995: p. 210). Inspired by the analysis of NPIs (Ladusaw 1979), both
types of Q-Adverbs stand in a local licensing relation to Hab[+Q] (Chierchia 1995:
p. 212ff). which effectively makes genericity a polarity phenomenon.
The main innovation is the treatment of ILPs as inherent generics, capitaliz-
ing on the assumption that polarity phenomena are rooted in the lexicon (with
effects on sentences). The lexical entries of ILPs have an abstract habitual mor-
pheme Hab[+Q], whereby its agreement feature [+Q] can be locally licensed only
by gen, and never by overt Q-Adverbs. ILPs are also taken to involve generic
quantification over situations. Their Davidsonian situation variable s is bound

47
 Hana Filip

by gen making it unavailable for binding by overt Q-Adverbs. This straightfor-

wardly motivates why ILP denotations cannot be spatio-temporally located (e.g.,
⁇ John was intelligent in his car / at 4 p.m.), and why ILPs are semantically incom-
patible with overt Q-Adverbs (e.g., That is an apple versus That is ? often / ? always
an apple), as well as their behavior with respect to perceptual reports, existential
sentences, and bare plurals. There are no such constraints for SLPs (e.g., ⁇ John
was happy in his car / at 4 p.m., John often / always makes jokes). On this account,
for instance, John is intelligent is associated with the tripartite structure Gen s
[in’(j,s)] [intelligent’(j,s)], which approximately amounts to a generalization over
situations in which John appears, and, generally, a situation involving John will
be also a situation in which John is intelligent23 .
Chierchia’s (1995) proposal is, however, flawed in several respects (for details
see Fernald 2000), but most importantly for us here, in its unmodified form at
least it is unsuitable for the analysis of the Czech -va-, which at face value should
be a good candidate for an overt realization of the covert Hab[+Q]. Whatever
covert habitual Hab[+Q] is implicated in the logical structure of generic sen-
tences on Chierchia’s (1995) analysis, it cannot be of the same kind as the overt
morpheme -va- in Czech. If -va- were the spell-out of the covert functional head
Hab[+Q], then it should have the same semantic and pragmatic properties as its
covert counterpart, and both should have the same agreement feature [+Q] that
must be locally licensed by the covert gen or some overt Q-Adverb. However,
as we have seen, generic sentences that are formally marked by the morpheme
-va- differ in a number of respects from formally unmarked sentences on their
generic reading, and in which the Hab[+Q] is covert.
Specifically, two main observations can be adduced in support of this point.
The Czech -va-, as I propose, in addition to its generic quantificational force,
tracks reasoning with exceptions, and it can be viewed as a ceteris paribus hedge
operator; it directly concerns exception-tolerance of generics (e.g., Birds fly is
true even if not all do, penguins, for instance, do not), which is one of their hall-
mark properties (see Section 2). In Chierchia (1995), exception-tolerance is not
in the focus, and the obvious tension between exception-tolerating generics and
gen in their logical structure that is treated as a universal quantifier is resolved
in an ad hoc way via the context variable C (Chierchia 1995: p. 196), without
elaborating on its properties.
Second, what Chierchia (1995) does not take into account are ‘interruption’

23
As Fernald (2000) observes, this may seem counterintuitive, because sentences like John is
intelligent do not seem to be generalization over situations in which John appears, but rather
are most naturally understood as attributing a property to John.

48
 Generic sentences

construals of ILPs (in the sense of Fernald 2000) that are triggered by modifica-
tion with Q-Adverbs, and in Czech also by the morpheme -va- (alone, and also
together with Q-Adverbs). On such construals, ILPs have truth-values that vary
across evaluation times/worlds, as they may hold of one of their arguments (typ-
ically subject) intermittently, on and off. This fluctuation in truth-value over
evaluation times provides the requisite plurality of ‘cases’ that serve as the base
for the generalization. Put differently, they satisfy the Plurality Condition on the
application of overt A-Qdverbs and also on the covert gen (see e.g. Krifka et
al. 1995,Fernald 2000, also following related suggestions by Swart 1991), which
requires that they presuppose that both the Restrictor and Matrix are satisfied
by multiple ‘cases’, or situations. Some English and Czech examples are given
below.

(34) a. Max is sometimes a California resident. (Fernald 2000)

b. Francis is occasionally blond.

(35) a. Františka je občas blond’atá.

Francis be.3sg.pres occasionally blond.
‘Francis is occasionally blond.’
b. Františka bývá občas blond’atá.
Francis be.gen.3sg.pres occasionally blond.
‘Francis tends to be occasionally blond.’

(36) a. Každopádně nejste sám, kdo si to někdy myslívá.

in.any.case neg.be.2pl alone who refl it sometimes
think.gen.3sg.pres
‘In any case, you are not the only one who occasionaly thinks like
this.’
b. Takový makléř bývá chytrý, s lidmi jedná obratně a se šarmem.
such realtor be.gen.3sg.pres smart with people deals skillfully and
with charm
‘Such a realtor tends to be smart, he deals with people in a skillful and
charming way.’
c. … občas i k slovesu ‘milovat‘ patřívá zájmeno zvratné. Karel Kryl
… occasionally even to verb love belong.gen.3sg.pres pronoun
reflexive
‘ … occasionally even the verb ‘love’ does the reflexive pronoun
belong’

49
Hana Filip

The use of ILPs in generic sentences under the ‘interruption’ construal, both in
English and Czech, might look like a kind of coercion of an ILP into a SLP in-
terpretation, which is triggered by the generic marker -va- or a Q-Adverb. If
they involved coerced SLP interpretations, then they would have not only truth-
values that vary across evaluation times/worlds, but also denote ordinary situ-
ations that can be spatio-temporally located. However, Carlson (1977), Kratzer
(1988) and Fernald (2000) propose that such ILPs do not shift into a SLP interpre-
tation, but rather still have arguments that are of the individual sort, and denote
stative properties, albeit having truth-values that vary across evaluation times,
i.e., are true on and off, but which are nonetheless stative in so far as they are
not spatio-temporally locatable.
Chierchia’s (1995) analysis of ILPs wrongly predicts that the examples (34a),
(34b) and (35a) all should be ungrammatical, because all overt Q-Adverbs should
be semantically incompatible with ILPs. However, Chierchia (1995) does not mo-
tivate just why this should be the case, why gen, a sort of null Q-Adverb, is the
only Q-Adverb that can license the covert habitual morpheme Hab[+Q] of ILPs.
Neither does Chierchia’s (1995) analysis of ILPs motivate the behavior of the
Czech -va- in generic sentences with verbs denoting ILPs, if we were to seriously
consider the idea that -va- is indeed an overt exponent of the covert habitual
morpheme Hab[+Q] introduced in the lexical entries of all ILPs. First, such an
analysis wrongly predicts that -va- should be compatible with all ILPs, but this
is not borne out. It is only compatible with those ILPs that allow for an ‘interrup-
tion’ construal, as in (35b), (36a)-(36c). It is not compatible with ILPs that denote
properties that necessarily hold of their arguments at any evaluation time/world,
without interruptions. Good examples are ILPs that denote defining, constitutive
properties of individuals like ‘be in 3/4 time’:

(37) Valčík je / ?bývá ve tříčtvrtečním taktu.

waltz be.3sg.pres / be.gen.3sg.pres in three.quarter time
‘A waltz is / ? tends to be in three quarter time.’

Second, no ILP should be compatible with an overt Q-Adverb, be that ILP ex-
pressed by a verb that is formally unmarked for genericity or formally marked
with -va-24 . However, this is not borne out, as the contrast between (35a) and
(35b), on the one hand, and (38), on the other hand, show:
24
This would follow if -va- were an overt realization of the covert Hab[+Q] of ILPs, and taking the
proposal of Chierchia (1995) to its logical conclusion, because then it should be only licensed
by gen, which binds the situation variable of ILPs making it unavailable for binding by overt
Q-Adverbs

50
 Generic sentences

(38) ? Valčík je/bývá občas ve tříčtvrtečním taktu.

waltz be.3sg.pres / be.gen.3sg.pres occasionally in three.quarter time
‘? A waltz is /tends to be occasionally in three quarter time.’

6.2 Generic (‘habitual’) morphology bound by temporal quantifiers

introduced by (implicit) Q-Adverbs
Based on his analysis of Tlingit (Na-Dene), Cable (2022) hypothesizes that “the
Czech habitual morphology is etymologically an instance of imperfective-aspect
combining with another, lower imperfective-form of the verb, a kind of ‘doubly
aspectualized’ verb (Filip 2018). One might wonder, then, whether that higher
‘imperfective’-morphology might at all synchronically be a realization of a (bound)
T-head, while the lower aspectual morphology is a realization of (non-modal)
[IMPRVOG].”
The Tlingit ‘habitual-mode’ forms share a number of semantic features in com-
mon with generic verbs formally marked with the morpheme -va- in Czech, judg-
ing according to the Tlingit examples given by Cable (2022). Nonetheless, his
analysis of Tlingit data, which seems to work well for the Tlingit data at hand, is
unsuitable for the analysis of Czech generic forms marked with -va-. As already
argued in Section 5.2.2, the idea of a ‘doubly aspectualized’ verb, or a kind of
reduplication of imperfective morphology, is not viable.
Neither can the generic -va- be treated as a spell-out of a T-head that is li-
censed just in case it is bound by a (possibly covert) Q-Adverb, which is what Ca-
ble (2022) proposes for the two ‘habitual-mode’ forms in Tlingit (Na-Dene). The
‘habitual perfective-mode’ of a verb is formed with the dedicated ‘habitual’ suffix
-ch, while the ‘habitual imperfective-mode’ with the habitual particle nooch. The
morphosyntactic structure for sentences that are formally marked with these
markers, as opposed to those that are formally unmarked, contains a (possi-
bly covert) Q-Adverb, treated as a temporal quantifier, which binds the T-head
which is spelled out by -ch or nooch. This is motivated by an intuitive relation
between Tlingit ‘habitual’ morphology and temporal quantification expressed by
Q-Adverbs (e.g., ‘always’ or ‘sometimes’). In this respect, Cable (2022) exploits
semantic affinities between overt Q-Adverbs and generic/habitual morphology,
and his analysis resembles Chierchia’s 1995) in so far as it assumes a licensing
relation between habitual morphology and (possibly covert) Q-Adverbs; albeit
Cable (2022), unlike Chierchia (1995), also aims at accounting for the non-modal
(‘actuality entailment’) uses of habitual constructions. While Chierchia (1995)
effectively subsumes genericity/habituality under (imperfective) aspect, Cable
(2022) subsumes Tlingit habitual morphemes under tense.

51
Hana Filip

One of Cable)’s innovations is the proposal that Tlingit sentences that are
overtly marked with -ch and nooch are assigned a bi-partite morphosyntactic
structure, consisting of a (higher) habitual component and a (lower) aspectual
component. Crucially, the habitual component is not (directly) contributed by
the habitual morphology itself, but instead by a (covert) temporal Q-Adverb,
which has a purely extensional temporal semantics, and does not introduce modal
(intensional) quantification over possible worlds. Any modality of a generic (ha-
bitual) sentence is contributed by some other material it may contain, such as
the lower imperfective aspectual heads. Verbs bearing habitual imperfective
mode have the modal quantification that allows the sentence to describe non-
actualized, merely hypothetical generalizations. Habitual constructions combin-
ing with a non-modal Asp-head (e.g., constructions with verbs bearing habitual
perfective mode) lack this modal quantification, which accounts for their ‘actu-
ality entailment’ that there is a set of events in the actual world verifying the
expressed generalization.
One fundamental reason why the Czech morpheme -va- cannot be a realiza-
tion of a T-head standing in a binding relation to a (covert) Q-Adverb is that
it is used to formally mark a wider swaths of and also perhaps different types
of generic sentences than the two Tlingit ‘habitual modes’ do, again judging ac-
cording to the examples discussed in Cable (2022). First, one of the key results of
Cable’s analysis is that ‘habitual-mode’ forms should be incompatible with ILPs.
However, the Czech -va- is compatible with ILPs, as we see in (27a) and (29a),
and also in the examples discussed in Section 6.1.
Second, the ‘habitual’ (generic) meaning of the markers -ch or nooch seems to
be reduced to the purely temporal, extensional, semantics of a (c)overt Q-Adverb
which contributes the specification of the frequency of events of the kind de-
scribed by the VP (e.g., ‘always’, ‘sometimes’), or put differently, to a plurality of
events, which may be actualized or non-actualized, modulo the (non-)modality
of the lower aspectual heads in the bi-partite structure. However, the meaning
of sentences formally marked with the Czech generic -va- does not (just) depend
on a mere plurality of events, situations or ‘cases’ of the kind described by the
base predicates to which -va- is applied, as 29b) and (29a). illustrate.
Even when the truth of a generic sentence formally marked with -va- depends
on (un)observed array of actual situations (a plurality of situations) in the real
world (the ‘actuality entailment’), it does not follow that it lacks a modal (in-
tensional) component; neither can the modality (intensionality) of characteriz-
ing genericity be derived from that the imperfective aspect (see Section 5.2.2).
Rather, it follows from its gnomic, law-like force, which all episodic sentences

52
 Generic sentences

lack (Carlson 1977; Krifka et al. 1995; Pelletier & Asher 1997, i.a.). However, this
is in principle excluded on Cable’s analysis.
Being law-like, generic sentences formally marked with -va-, also support
counterfactual inferences Lewis (1973). For instance, the truth of the following
generic sentence, formally marked with -va-, comes with the commitment that it
is not an accidental generalization, based on a mere summation of situations of
various children actually getting good grades, but rather that there is some under-
lying cause(s) for this (e.g. the quality of instruction, parents’ attention to their
children, etc.). While we may not determine the particular cause or causes at
work here, they give rise to the counterfactual inference, which that transcends
observable real world episodic conditions or actual circumstances we have ac-
cess to, that if a child were to attend this school, that child too would most likely
or probably get good grades:
(39) Děti v této škole mívají dobré známky.
children in this school have.gen.3pl good grades
‘Children in this school tend to have good grades.’
Tied to the (g)nomic, law-like character of generic (and habitual) sentences is not
only their intensionality (modality), but also another of their key characteristics:
namely, they may be true in the face of exceptions. This feature, which, however,
is not a factor in Cable’s analysis of the Tlingit data, is one of the most prominent
and intriguing features of the morpheme -va-. It is incompatible with exception-
less generalizations, and its uses largely depend on the speaker’s reasoning about
generalizations with exceptions, as I will propose.

7 Proposal
7.1 Generalizations with Ceteris Paribus Condition
As established above, the Czech generic morpheme -va- derives verbs exhibit-
ing two semantic properties that are signature properties of generic predicates
(Krifka et al. 1995; Nickel 2017, i.a.):

• Aspectual stativity (Semantic Property 1)

• No requirement of a quantitative criterion or quantity-based measure for events,
particulars or cases counting as evidence for their truth (Semantic Property 2)

53
Hana Filip

What differentiates -va- from the null generic operator gen, but aligns it with
similar markers in other languages that systematically enforce a generic (habit-
ual) interpretation of sentences, is the so-called ‘actuality entailment’, which will
be introduced further below as Semantic Property 4, and here can be generally
stated as follows:

• ‘Actuality entailment’: Morphological markers or syntactic constructions that

are dedicated to the expression of generic (and habitual) statements trigger the
‘actuality entailment’ (Bhatt 1999), i.e., the uncancellable inference that there are
verifying instances in the actual world that count as evidence for its truth. (See
e.g., the Czech marker -va-, the Tlingit ‘habitual’ morphology (Cable 2022), the
be construction in African American English (Green 2000), and the construction
with haya+participle in Modern Hebrew (Boneh & Doron 2008).)
The key semantic property of the Czech morpheme -va-, and one which so far
has not been properly acknowledged, can be formulated as follows, and will be
motivated below (see Semantic Property 3):

• Generalizations with Ceteris Paribus Condition: The generic morpheme -va-

expresses an indeterminate ceteris paribus condition, in addition to contributing
a generic quantificational force to the meaning of sentences. Depending on con-
text, it conveys the inference that the speaker is either

(i) certain that there are exceptions to the expressed generalization:

certainty inference: Ks ¬[SG(p)], or
(ii) uncertain whether all exceptions can be categorically excluded:
ignorance inference: ¬Ks ¬[SG(p)];

where
Ks[p] stands for ‘speaker knows that p’
SG(p) stands for ‘p expresses an exceptionless, ‘strong generalization’
Ks[SG(p)] stands for ‘speaker knows that p expresses an exceptionless,
‘strong generalization’
It is an open question whether similar markers in other languages that systemat-
ically enforce a generic (habitual) interpretation of sentences also introduce the
ceteris paribus condition into the logical structure of generic (habitual) sentences,
i.e., put in the simplest terms, that the speaker denies the commitment that there
are no exceptions to the expressed generalization. The ceteris paribus condition

54
 Generic sentences

is essentially about how we reason with exceptions, how we epistemically deal

with exceptions. Assuming that -va- formally marks generic sentences that are
semantically analyzed in terms of a quantificational tripartite structure, we may
provide a relatively straightforward account of the indeterminate ceteris paribus
conditon expressed by -va-, as constituting a part of the Restrictor clause in the
tripartite structure (see 8). Intuitively, ceteris partibus qualifiers have pragmatic
consequences in so far as they bring to focus the existence of exceptions that
the speaker thinks they should not or cannot safely ignore, by virtue of some
possibly implicit background understanding of what count as salient reasons for
(or against) the correctness of applying the generalization to a given ‘case’ (in
the sense of Lewis 1973). This proposal finds support in the observation that the
acceptability of -va- in generic sentences may depend on explicit qualifiers that
restrict the conditions under which the generalization can be appropriately ap-
plied to specific cases, and so explicitly draw a line between cases to which the
generalization applies and those that it excludes. In effect then, generic sentences
formally marked with -va- are endowed with two layers of modality (intension-
ality): one that is tied to their law-like, nomic force, and the other layer tied to
reasoning with exceptions, which is best characterized in terms of epistemic and
doxastic modality.
In what follows I will adduce additional supporting evidence for this proposal,
which will be structured by the following guiding question:

qestion 4: What are the conditions under which the generic morpheme -va- in
Czech may, must and must not be used?

7.2 Further supporting evidence

7.2.1 When -va- may be used: Substitution salva veritate of generic -va-
verbs and non-generic verbs
The Czech generic marker -va- may be used in generic sentences that have excep-
tions of the kind that we safely ignore. For instance, (40) is true, despite ravens
that are not black.

(40) Ravens are black.

≈ ‘Ravens tend to be black’ or ‘Ravens are black, ceteris paribus.’
where the ceteris paribus clause can be understood as ‘other things being
equal / unless prevented / in the absence of disturbing factors … ’

55
Hana Filip

(40) is a paradigm example of a ceteris paribus (cp) generalization, which may be

paraphrased with ‘Ravens tend to be black’ (see e.g., Nickel 2016: p. 24). To the
English sentence (40) naturally correspond two sentences in Czech, one with a
non-generic verb, formally unmarked for genericity, just like its English counter-
part, and the other with a generic verb formally marked with -va- that arguably
explicitly asserts what is taken to be the proper paraphrase of the English (40):
(41) a. Vrány jsou černé.
ravens be.3pl.pres
‘Ravens are black.’
b. Vrány bývají černé. [ = (27a)]
ravens be.gen.3pl.pres
‘Ravens tend to be black’ or ‘Ravens are black, ceteris paribus.’
All the above sentences, (40), (41a) and (41b), have the same truth conditions. We
judge them to be true because the majority of ravens are black, which we in turn
may infer from our experiences with ravens, but also, and perhaps more impor-
tantly, based on our knowledge of some underlying causal factors like a stable
coloration mechanism that this kind developed. Generally, in the context of sen-
tences that express cp generalizations with known tolerable exceptions, generic
verbs marked with -va- are substitutable salva veritate with their non-generic
counterparts without -va-. This immediately raises the question ‘How do we
motivate the use of a formally marked generic form to express a characterizing
statement, when such a statement can also be expressed by its corresponding
unmarked non-generic form?’
I propose that the generic morpheme -va-, apart from its quantificational generic
force, expresses a ceteris paribus qualifying condition, albeit open-ended or inde-
terminate, akin to ‘other things being equal’, ‘unless prevented’, ‘in the absence
of disturbing factors’, and by doing so, it explicitly constrains the application of
the generalization to only a subset of cases or instances of the kind. The spe-
cific restricting condition may be retrieved from the extra-linguistic context, it
may be based on some implicit consensus of what might be typical examples of
disturbing factors, or it may be explicitly specified by cp clauses. For instance,
our examples above might be continued with ‘even though some ravens have no
pigment due to a genetic defect’.
If making a statement publicly commits the speaker to its content25 , then by
25
“[A] speaker is committed to a statement when he makes it himself, or agrees to it as made
by someone else, or if he makes or agrees to other statements from which it clearly follows”
(Hamblin 1971: p. 136, also Hamblin 1987).

56
 Generic sentences

formally marking a generic sentence with -va- the speaker publicly commits to
the ceteris paribus condition, which in turn invites inferences on the part of the
hearer about the speaker’s epistemic commitment to exceptions to the expressed
generalization. For instance, by using (41b), instead of (41a), the speaker most
likely intends to convey that they know that there are exceptional ravens that
are not black, and given that is part of our general world knowledge, also intends
to shift focus of the asserted generalization to the existence of such known and
tolerable exceptions, whatever the reason for this might be. In contrast, such in-
ferences about speaker’s reasoning about exceptions are lacking in (41a) without
the marker -va-. I propose that the ceteris paribus condition is part of the truth-
conditional content of -va- stored in its lexical entry, giving rise to inferences
about the speaker’s epistemic stance with regard to exceptions to expressed gen-
eralizations:
Semantic Property 3: The generic morpheme -va- expresses an in-
determinate ceteris paribus condition, in addition to contributing the
generic quantificational force to the meaning of sentences. Depending
on context, it conveys the inference that the speaker is either (i) certain
that there are exceptions to the expressed generalization (certainty
inference) or (ii) uncertain for lack of sufficient information whether
all exceptions can be categorically excluded (ignorance inference).
corollary: In generic sentences that express generalizations with
generally known tolerable exceptions, generic verbs marked with -va-
are substitutable salva veritate with their non-generic counterparts
without -va-.

7.2.2 When -va- must not be used: Exceptionless and purely hypothetical
generic sentences
The Czech marker -va- must not be used for the expression of two types of
generalizations (see also Filip 1993; 1994; 2019):

• generalizations that are commonly known to permit no exceptions in all

the known conditions, e.g., universal laws of nature, analytic truths; and

• generalizations with no verifying instances in the actual world, which

have a purely dispositional (intensional) interpretation (Lawler 1972, Lawler
1973, Dahl 1985).

The first constraint is illustrated by the following examples:

57
Hana Filip

(42) a. Země se točí / #točívá kolem slunce.

earth refl revolve.3sg.pres / revolve.gen.3sg.pres around sun
‘The Earth revolves / # tends to revolve around the Sun.’
b. Valčík je / #bývá ve tříčtvrtečním taktu.
waltz be.3sg.pres / be.gen.3sg.pres in three.quarter time
‘A waltz is / # tends to be in three quarter time.’
The incompatibility of -va- with generalizations that are exceptionless in all the
known conditions like the Earth revolving around the Sun, or the waltz being in
three quarter time straightforwardly follows from the Semantic Property 3. If -va-
has as part of its lexical meaning an indeterminate ceteris paribus condition, and
triggers either the certainty or ignorance speaker-related inferences about
exceptions, then this clearly clahes with any exceptionless generalizations.
Generally, our judgments about the truth and acceptability of generic sen-
tences are influenced by facts about the world, our general world knowledge,
and also by a speaker’s information state. Facts about the world comprise ‘brute
physical facts’ (in the sense of Searle 1995) like those captured by universal laws
of nature, as in (42a), and also ‘social facts’ (Searle 1995) that are made possible
by constitutive rules, as in (42b), and regulatory rules. The speaker may very well
have false beliefs about the facts of the actual world, which is what makes the use
of -va- infelicitous, odd (semantically uninterpretable) in the above examples.
In Czech, formally unmarked generic sentences are ‘exception-tolerating’ in
the most general sense in which this term is standardly understood in the generic
literature (see e.g., Krifka et al. 1995; Pelletier & Asher 1997): their truth is compat-
ible with a different number and type of exceptions (e.g., (41a)), but they are also
compatible with no exceptions whatsoever (e.g., (42a)). In stark contrast to this,
generic sentences that are formally marked with -va- which are incompatible
with generalizations that have no exceptions, as we have just seen.
This is also confirmed by the observation that -va- clashes with overt universal
quantifiers in the same clause, which also follows from the Semantic Property 3.
Some examples are given below:
(43) a. # Každou sobotu / vždycky / nikdy sedává Honza v hospodě.
each Saturday / always / never sit.gen.3sg.pres Honza in pub
# ‘Every Saturday / always / never Honza usually tends to sit in a pub.’
b. Každou sobotu / vždycky / nikdy sedí Honza v hospodě.
each Saturday / always / never sit.3sg.pres Honza in pub
‘Every Saturday / always / never Honza sits in a pub.’
In the above examples, the universal quantifiers introduced by ‘every Saturday’,

58
 Generic sentences

‘always’, ‘never’ are odd (semantically uninterpretable) with the generic mor-
pheme -va-, but are perfectly acceptable in the absence of -va-. For instance,
(43a) is odd, because it suggests that the speaker holds inconsistent or contradic-
tory beliefs. On the one hand, the speaker makes a universal assertion that Honza
is in a pub every Saturday, for instance, but on the other hand, and contrary to
this, also asserts, by virtue of the ceteris paribus condition lexically introduced
by the morpheme -va- that they cannot publicly commit to it, because they either
know for sure that there are exceptions to what would be an exceptionless gen-
eralization (‘certainty inference’) or that they fail to have enough information
to be in the position to be committed to no exceptions whatsoever (‘ignorance
inference’).
There are apparent counterexamples to the claim that -va- is incompatible with
overt universal quantifiers in the same clause. Some perfectly felicitous attested
examples are given below:
(44) a. Mládež ve Vídni se zabývala Hebblem — já jsem vždycky býval
(was.gen) skeptický k takovým módním proudům.
‘Viennese youth were all reading Hebbel — I was usually always
skeptical about these fashionable influences.’ (Čapek 1990: p. 57)
b. “Je to divný,” pokračovala pak rychlým a věcným šepotem, “jeden
šuplík má zamčenej, a nikdy ho nemívá (neg.has.gen) zamčenej. A
nepasuje mi do něj žádnej klíč.”
“It’s strange,” she continued in a quick and matter-of-fact whisper,
“one of his desk drawers is locked and he never has it locked. And
none of my keys fit the lock.” (Bělohradská 1992)
Universal Q-Adverbs like vždycky ‘always’ and nikdy ‘never’, as in the above
sentences, do not clash with the generic morpheme -va- just in case they are
used for ‘intensification of the strength’ of the expressed regularity which albeit
allows for exceptions, as Danaher (2003: p. 45) observes, and not as quantifiers
with their inherent universal quantificational force. Such ‘-va- + universal Q-
Adverb’ combinations in Czech are similar to felicitous combinations like usually
always or usually never in English, where arguably the universal Q-Adverb also
lacks its inherent universal quantificational force:
(45) a. I am usually always happy, but today I feel really depressed.
b. I am usually never neurotic about being messy and keeping things
tidy, but I can’t seem to go to sleep if clothes are hanging up to dry in
my room.

59
Hana Filip

Related to the claim that -va- is incompatible with overt universal quantifiers
in the same clause, and apparent counterexamples to it, we also observe that the
generic morpheme -va- tends to be semantically anomalous, dispreferred or odd
when it is applied to verbs denoting ILPs (e.g., ‘be intelligent’, ‘be in three quarter
time’, ‘know’, ‘believe’, ‘have long arms’). The reason for this is that they denote
properties that are non-temporary, atemporal, or ‘tendentially stable’ over time
Chierchia (1995), i.e., they typically hold over most of an individual’s life time, or
at a relatively large interval, and also at any and all of its moments; or alternately,
they ‘spread’ to all the situations of an individual’s life. Hence, in this sense, ILPs
involve a kind of implicit universal quantification. This is at odds with the ceteris
paribus condition lexically associated with -va-, as is illustrated by (42b), for in-
stance, unless the properties denoted by ILPs can be construed as holding with
interruptions of their arguments (see also 6.1). Some examples of such imperfec-
tive ILPs are given below:

Figure 6: The generic morpheme -va- applied to ipf verb stems denoting ilps

ipf: episodic/generic generic

myslit (si) ‘to think’ → myslívat (si) ‘(to tend) to think’ (see (36a)),
patřit ‘to belong (to)’ → patřívat ‘(to tend) to belong to’ (see (36c))
věřit ‘to believe’ → věřívat ‘(to tend) to believe’
mít ‘to have’ → mívat ‘(to tend) to have’

Generally, a generic verb formally marked with -va- is dispreferred, or even

unacceptable, for the expression of a characterizing property that is thought of
as ‘tendentially stable’, or necessarily ‘spreading’ over the whole life-time of an
individual for which it holds. Consider the following example:

Question: Jaké má Marie povolání? - ‘What is Mary’s profession?’

(46) a. Učí na střední škole.
teach.3sg.pres on middle school
‘She teaches at high school.’
Inference: ‘Mary is a high-school teacher.’
b. # Učívá na střední škole.
teach.gen.3sg.pres on middle school
# ‘She teaches at high school on and off / occasionally / often …’
Inference: ‘Being a high-school teacher is not Mary’s main profession,
main occupation.’

60
 Generic sentences

The question about Mary’s profession is best answered with (46a), which is headed
by a non-generic imperfective verb, and is paraphrasable with ‘Mary is a high-
school teacher.’ Having a profession as a teacher means that one has some requi-
site qualification (e.g., having completed an accredited training program), which
is a stable property holding without interruptions during one’s life, and even
when one happens not to be under a teacher’s contract, or during times off work
(weekends, holidays, or absence due to illness). In contrast, (46b) is formally
marked with -va-, and it would be odd or infelicitous as an answer to the above
question, because -va- conveys the inference that the situation of teaching, func-
tioning as a teacher, holds only intermittently or occasionally, i.e., Mary teaches
on and off. Intuitively, this follows given that any generic quantifier, and also
arguably -va-, presupposes that the Restrictor and the Matrix clause denote a
multiplicity of cases (see 8), or have the potential to do so. But this amounts
to Mary being a temporary or a substitute teacher, and that she may also have
other alternative jobs or occupations. In short, (46b) carries the inference that
teaching is not Mary’s profession, because the property of teaching only holds of
some situations of Mary’s professional life, which is at odds with the question.
The second constraint on when -va- must not be used has to do with another
of its key semantic properties, namely what is commonly known as the ‘actuality
entailment’ (Bhatt 1999):26

Semantic Property 4: A generic sentence formally marked with the

generic morpheme -va- has the ‘actuality entailment’ (Bhatt 1999), i.e.,
the uncancellable inference (as part of its truth-conditional content) that
there are verifying instances in the actual world that count as evidence
for its truth.
Consider the following example that is set in the context in which there is no set
of actual world situations of the machine crushing oranges:

(47) scenario: We have just bought a new machine for crushing oranges, but
we have not yet used it, and it is still packed in its original box.
Tento stroj { drtí | # drtívá } pomeranče.
this machine { crush.3sg.pres | # crush.gen.3sg.pres } oranges
‘This machine crushes oranges.’

26
‘Actuality entailment’ is a term originally coined by Bhatt (1999) to describe the implicative
inference that arises with ability modals in perfective aspect.

61
 Hana Filip

Under this scenario, and understanding ‘this machine’ as referring to the specific
machine token in the domain of discourse, the sentence can only have a purely
intensional (dispositional) interpretation. What makes it true is the yet not re-
alized function of the machine to crush oranges. Such truth conditions are only
compatible with a sentence that contains a non-generic verb, here the primary
imperfective form drtí ‘(it/(s)he crushes’). In contrast, the generic verb marked
with -va- makes the whole sentence false, because -va- carries the cancellable
inference that situations denoted by its imperfective base were actualized in the
real world, that is, it requires that generic sentences it formally marks express
generalizations that are inductively projected from (observed) instances in the
real world. This ‘actuality entailment’ is part of the truth-conditional content of
-va-, it is not cancellable, as also evidenced by the observation that the generic
verb on which it occurs cannot be substituted salva veritate with its non-generic
counterpart in the above sentence under the given scenario.
In so far as the Czech generic marker -va- contributes the ‘actuality entailment’
to the meaning of a sentence, it patterns with generic (or habitual) markers in
other languages, such as the Tlingit ‘habitual-perfective mode’ morphology (Ca-
ble 2022), the be construction in African American English (Green 2000), and the
construction with haya+participle in Modern Hebrew (Boneh & Doron 2008).

7.2.3 When -va- must be used: ‘Positive counterinstances’ and plausible

deniability
Many generic sentences are true in the face of exceptions to them, which is one of
their hallmark properties, as observed above. For instance, we judge Birds fly to
be true despite birds (e.g., penguins, ostriches or emus) that do not fly. However,
there are generic sentences expressing generalizations that seem false in the face
of exceptions to them:
(48) a. Books are paperback. false
b. Knihy jsou IPF brožované. false
books be.3pl.pres paperback
‘Books are paperback.’
Why should this be the case? It turns out that for a generic statement to be true
it matters how exactly the exceptional or non-confirming individuals of a certain
kind fail to have the requisite property that characterizes it, as Leslie (2008) ar-
gues. If they fail it in virtue of having an equally salient, memorable, and striking
positive property, the generic sentence is unlikely to be judged true. Hardback
books, e-books, audiobooks constitute subkinds of the kind book which fail to

62
 Generic sentences

satisfy the above generalizations. However, they do not simply fail to be pa-
perbacks, but possess a salient distinctive feature which is subjectively for us
as salient, memorable, significant or striking as having the positive property of
paperback. In this sense, they are alternative ‘positive counterinstances’ to pa-
perbacks (in the sense of Leslie 2008). They are cognitively salient, a part of our
kind-related expectations, which is what makes them exceptions that are partic-
ularly hard for us to ignore when it comes to our truth-value judgment of (48a)
and (48b), which is why, according to Leslie (2008), we tend to judge them as
false.27
If, however, individuals that fall under a certain kind simply lack the requi-
site characterizing property of that kind, that is, if we do not attribute to them
any salient alternative positive properties, they constitute merely ‘negative coun-
terinstances’ (Leslie 2008), then we are far more likely to judge a generic sentence
involving that characterizing property to be true. For example, non-flying birds
offer no alternative positive property that we pit against the positive character-
izing property of flying that would be equally (subjectively) salient or striking as
flying. Such negative counterinstances to Birds fly can be safely ignored when
we judge this sentence as true.
Adding to ‘bare’ or unrestricted generic sentences like (48a) and (48b) that are
false in the face of positive counterinstances a qualifier that explicitly restricts
their application to only a subset of members of the kind satisfying them renders
generic sentences that are true. Such qualifiers include non-universal Q-Adverbs
like usually, rarely, seldom, generic adverbs like typically, normally, or quantity
expressions like in the majority of cases, in some cases, and, as I propose, also the
Czech generic morpheme -va-:

(49) a. Typically/Usually, books are paperback. true

b. Knihy jsou IPF { obyčejně | nejčastěji } brožované. true
books be.3pl.pres { usually | most.often.comp } paperback
‘Books are { usually | most often } paperback.’
c. Knihy bývají brožované. true
books be.gen.3pl.pres paperback
‘Books tend to be paperback.’

The truth of the generic sentence formally marked with -va- (49c), in contrast
27
Other examples of generic sentences that are judged false, because of the positive counterin-
stances to the generalization they express (even if the (vast) majority of members of a kind
have the characterizing property) are: Bees are sterile, Prime numbers are odd, Nurses are fe-
male, Crocodiles die before the age of two weeks, People have brown eyes.

63
 Hana Filip

to its unmarked ‘bare’, or unrestricted counterpart (48b) that is judged false,

straightforwardly follows from the Semantic Property 3 given above, i.e., assum-
ing that -va- has as part of its lexical meaning the indeterminate ceteris paribus
condition. Given that (48b) and (49c) minimally differ only by the presence of
-va- in the latter, I would like to propose that the Czech generic marker -va- must
arguably be used as a hedge to safeguard the truth of generic sentences that oth-
erwise would be false. Based on such data and observations, I propose to update
the Semantic Property 3 as follows:

Semantic Property 3 (revised): The generic morpheme -va- expresses an

indeterminate ceteris paribus condition, in addition to contributing the
generic quantificational force to the meaning of sentences. Depending
on context, it conveys the inference that the speaker is either (i) certain
that there are exceptions to the expressed generalization (certainty
inference) or (ii) uncertain for lack of sufficient information whether
all exceptions can be categorically excluded (ignorance inference).
corollary 1: In generic sentences that express generalizations with
generally known tolerable exceptions (i.e., ‘negative counterinstances’
in the sense of Leslie 2008), generic verbs marked with -va- are substi-
tutable salva veritate with their non-generic counterparts without -va-.
corollary 2: The generic morpheme -va- safeguards the truth of
generic sentences that would otherwise be false (e.g., due to ‘positive
counterinstances’ in the sense of Leslie 2008), in the absence of overt
restrictors that restrict the appropriate application of the characteriz-
ing property to the conforming subset of a given kind that satisfies them.

The ceteris paribus condition entailed by -va- is exploited as a hedge not only to
warrant the truth of generic sentences, which otherwise would be false, but also
as a hedge to safeguard a generalization against refutation by states of affairs of
which the speaker is ignorant, and so to ensure its plausible deniability. Consider
the following examples adapted from Bıĺ ý (1986):

(50) a. Knihkupectví má tu knihu ve výloze (ale ne vždycky).

bookstore have.3sg.pres book in shop.window (but not always)
‘The bookstore has the book in the shop window (but not always).’
b. Knihkupectví mívá tu knihu ve výloze.
bookstore have.gen.3sg.pres book in shop.window
‘The bookstore tends to have the book in the shop window.’

64
 Generic sentences

I.e., “I am not sure if the book still is in the shop-window, it was there
quite recently, but I do not dare draw the conclusion that it is still
there.” Bıĺ ý (1986: p. 29)

Let us suppose that I am interested in a unique antiquarian book, and ask my

friend whether our favorite used bookstore has it for me to see. However, it
is after hours and shops are closed. If my friend responds with (50a), and I go
to the bookstore, but the book is not in the shop window, I could blame my
friend for misleading me at least. The proposition conveyed by the utterance of
(50a) would not be false in this situation, because generally a formally unmarked
generic sentence allows for exceptions, as also the perfectly fine continuation
of (50a) with ‘… but not always’ confirms. However, I reason that by choosing
a formally unmarked generic sentence (50a), which allows for exceptions, but
is also compatible with no exceptions, rather than one formally marked with -
va- (50b) (or with some other explicit hedging device) that flags incompatibility
with no exceptions, I may infer by pragmatic strengthening that there are no
exceptions to the book being in the shop window. If, however, my friend utters
(50b), it is true and felicitous under the same scenario, because -va- here triggers
the inference that my friend is either certain that the book is not always in the
shop window, or is uncertain whether such exceptions to the book being in the
shop window can be completely excluded.
The plausible deniability, and the public commitment to at least the possibility,
if not a certainty, of exceptions to a given regularity, that can be inferred from
the use of -va- is also useful in legal discourse, in the formulation of various rules,
instructions and the like. It is, therefore, unsurprising that we find -va- formally
marking generic sentences that fit Carlson’s (1995) rules and regulation model
for the truth of generic sentences (see section 2 above). Recall that on this model,
generic sentences are true by virtue of some underlying rule, principle or cause,
which we may directly learn, rather than (just) real world particulars. Using a
generic sentence marked with -va- for such regularities is useful as a hedging de-
vice, because it gives us a certain leeway in the application of a rule, for example,
and also allows for unprecedented or unexpected exceptions to it, in cases when
the use of a formally unnmarked generic sentence without -va- would allow a
stricter or a more absolute interpretation, precisely because it is compatible with
no exceptions, and unless the context explicitly indicates otherwise, it invites
pragmatic strengthening to an exceptionless generic statement. Some attested
examples are given below. They concern game rules (51a), institutional regula-
tions (51b), job descriptions (51c), customs or instructions (51d):

65
Hana Filip

(51) a. Zápas se hrává na více vítězných her. 28

match refl play.gen.3sg.pres on more victorious sets
‘The match tends to be played in several winning sets.’
b. Členské schůze svolávává výbor aspoň jednou měsíčně.
member meetings convene.gen.3sg.pres board at.least once monthly
‘The board tends to convene member meetings at least once a month.’
c. Listonoši, potažmo pošťáci, nosívali hlavně dopisy a pohlednice.
letter.carriers, or postmen, carry.gen.3pl.pst mainly letters and
postcards
‘Letter carriers, or postmen, used to deliver mainly letters and
postcards.’
d. Štika se prodávává ve formě živé tržní ryby, výrobků z ryb (chlazené i
mražené), … 29
pike refl sell.gen.3sg.pres in form lifesg.gen marketsg.gen
fish.sg.gen, products fish.pl.gen (refrigerated and frozen) …
‘The pike tends to be sold alive on wet markets, in the form of fish
products (refrigerated and frozen) …
The above examples show that -va- formally marks sentences that well fit the
rules and regulation model of genericity in terms of Carlson (1995) (see Section 2).
However, as we have also seen, it is excluded in generic sentences that provide
the most compelling support for this model of genericity, which concern what is
merely hypothetical and lacking the ‘actuality entailment’, as in (47). At the same
time, -va- is an appropriate generic marker for generic sentences that express
(weak) descriptive generalizations, also including habits proper (i.e., regularities
of action by ordinary individuals), that are paradigm examples for the other main
model of genericity, the inductive model (Carlson 1995).
In short, the class of generic sentences that are formally marked with the
generic morpheme -va- in Czech cuts across these two main models for the truth
of generic sentences, discussed by Carlson (1995), and while it does not cover
the whole range of generics that fit the rules and regulations model, it does
not neatly fit just the inductive model. Generic sentences with overt markers
that seem similar to the Czech generic -va-, e.g., the Tlingit ‘habitual’ particle
nooch and the suffix -ch (Cable 2022), the be construction in African American
English (Green 2000), and the construction with haya+participle in Modern He-
brew (Boneh & Doron 2008), are often analyzed based on examples that express
28
https://cs.wikipedia.org/wiki/Kuličky, accessed on April 9, 2022
29
http://www.cz-ryby.cz/produkce-ryb/chovane-ryby/esox-lucius, accessed on April 9, 2022

66
 Generic sentences

(weak) descriptive generalization, and so appear to best fit the ‘inductive model’
for characterizing genericity (in the sense of Carlson 1995). This possible data
sparsity issue raises the question about the whole range of different types of
generic sentences that such markers can indeed formally mark, besides habitual
ones in the strict sense of ‘habitual’, and whether they can also be use in generics
describing various rules and regularions like (51b)-(51d), or also for propensities
of natural phenomena like (12b) above, for instance.

7.3 Restricted versus unrestricted generic sentences

As briefly observed at the outset of this paper, ‘bare’ generic sentences with kind
reference that are formally unmarked for genericity and without any overt re-
strictors or quantifiers, like (1a), repeated below as (52), constitute what many
consider to be prototypical characterizing generic sentences. In contrast, ‘bare’
generic sentences that are formally marked for genericity with -va- and without
any overt restrictors or quantifiers like (53a) seem odd or dispreferred in out-of-
the-blue contexts. They sound natural and felicitous in generic sentences with
overt restrictors that qualify the relevant cases based on which the generalization
is made, as we see in (53b) and (53c):
(52) Psi štěkají. [ = (1a)]
dogs bark.3pl
‘Dogs bark.’

(53) a. (#)Psi štěkávají.

dogs bark.gen.3pl
‘Dogs tend to bark / bark now and then / often / seldom / usually …’
b. Psi štěkávají na ty, které neznají. [ = (1d)]
dogs bark.gen at those whom neg.know
‘Dogs bark at those whom they don’t know.’
c. Psi štěkávají, s výjimkou některých druhů psů jako chrti.
dogs bark.gen.3pl.pres with exception some breeds dogs like
greyhounds
‘Dogs tend to bark, except for some dog breeds like greyhounds.’

If -va- has as part of its meaning a ceteris paribus condition, a kind of an indeter-
minate qualifier, as I propose, then the contrast between (53a), on the one hand,
and (53b) and (53c), on the other hand, straightforwardly follows. Intuitively, -va-
brings to focus the existence of exceptions or restrictors on the application of the
expressed generalization that the speaker thinks they should not or cannot safely

67
Hana Filip

ignore, which raises the issue what exactly they are. Now, it is generally known,
a part of our general world knowledge, that there are dogs and dog breeds that
do not bark. These are exceptions that we safely ignore, ‘negative counterin-
stances’ (in the sense of Leslie 2008), when we judge the sentence Dogs bark as
true, and mutatis mutandi also the corresponding formally unmarked (52)/(1a)
in Czech. The formally unmarked, simpler form in Czech tolerates the expected
exceptions, ‘negative counterinstances’, that are also known to both the inter-
locutors. When the speaker uses the formally marked sentence with -va-, which
is a more complex form, the hearer may reason that the speaker intends to bring
to focus the existence of other exceptions, other disturbing factors, or restrictors
that are less usual or unexpected. Therefore, the acceptability of -va- is enhanced
when it co-occurs with overt restrictors that explicitly specify the cases against
which the characterizing statement is evaluated.
Across different languages, overt restrictors are often necessary or preferred
to ensure the acceptability of generic sentences, which otherwise would be odd,
unacceptable or even false. For instance, in French (and also in Italian), indefi-
nite singular kind denoting terms are odd or unacceptable, unless we add explicit
restrictors (e.g., Q-Adverbs, overt expressions of a point of view like ‘in my opin-
ion’, temporal clauses). According to Mari (2008), they introduce the speaker’s
‘point of view’ or ‘respect’ (Nunberg & Pan 1975; Ross 1977) against which the
expressed characterizing statement is evaluated; on her analysis, ‘respects’ are
implemented in a modal framework (Lewis 1973; Kratzer 1989) as restrictions on
the worlds in the modal basis.

(54) a. ⁇ Un chien est intelligent. (French).30 (Mari 2008)

‘A dog is intelligent.’
b. En cas de danger / De mon point de vue, un chien est intelligent.
‘In case of danger / From my point of view, a dog is intelligent.’

(55) a. ⁇ Un chien mange des croquettes.

‘A dog eats croquettes.’
b. En Autriche, un chien mange des croquettes.
‘In Austria, a dog eats croquettes.’

30
Krifka et al. (1995: p. 15) classify English generic sentences with a kind denoting singular
indefinite like A bird flies, A potato contains potassium or A lion as a mane as characterizing
sentences with nonspecific non-kind reference. Such characterizing sentences are taken to
involve an indirect reference to a natural kind, due to the indefinite noun phrase. That is,
a potato in A potato contains potassium denotes the individual specimens of a natural kind
POTATO.

68
 Generic sentences

8 Conclusion
The Czech marker -va- can be plausibly analyzed as a kind of generic operator sui
generis, with a lexicalized indeterminate ceteris paribus condition, which brings
to focus the existence of exceptions that should not or cannot be safely ignored
(for whatever reason). It is thus intimately tied to reasoning with exceptions. It is
an open question what kind of theoretical framework is best suited for its formal
analysis among those that specifically focus on the exception-tolerance of char-
acterizing generics (e.g., relevant quantification, abstract Objects, prototypes,
stereotypes, non-monotonic logic (default reasoning approaches), probabilistic-
based, normality‐based approaches). Among the impulses for future research is
the question whether similar markers in other languages, which share with -va-
the ‘actuality entailment’ also exhibit the same proclivity towards or necessity of
reasoning with exceptions. Moreover, given that the ‘actuality entailment’ and
the ceteris paribus condition are associated with a single generic form, formally
marked with -va-, it may not be a matter of accident, which begs the question
about their common denominator, a unified concept that underpins both.
In Czech we have linguistically different means for the expression of character-
izing generic statements: verb forms that are formally marked for genericity with
-va- and verb forms that are formally unmarked, tied to two different, but over-
lapping, models of interpretation. But these two models cannot be neatly aligned
with either the inductive or rules and regulations models of genericity (Carlson &
Pelletier 1995), as both exhibit features that fit both these models. This then leads
to the conclusion that there may not be a single model for the interpretation of
all characterizing generic sentences in Czech, which would then confirm the in-
sights of others (e.g., Greenberg 2007; Pelletier 2010; Boneh & Doron 2008; Cable
2022,i.a.) that different linguistic forms for the encoding of generic statements
differentiate different subtypes of generic sentences, each requiring different se-
mantic/pragmatic commitments, and separate models for their interpretation.

Abbreviations

Acknowledgements

References
Barnetová, Vilma. Násobenost slovesného děje. Kapitoly ze srovnávacı.́

69
Hana Filip

Barwise, Jon & Robin Cooper. Generalized quantifiers and natural language. Lin-
guistics and Philosophy. 159–219.
Bělohradská, Hana. Poslednı ́ večeře. Československý spisovatel.
Bhatt, Rajesh. Covert Modality in Non-Finite Contexts. University of Pennsylvania
Ph.D. Thesis.
Bittner, Maria & Naja Trondhjem. Quantification as Reference: Evidence From Q-
Verbs. In Lisa Matthewson (ed.), Quantification: a cross-linguistic perspective, 7–
66. Emerald.
Bıĺ ý, Milan. Preliminaries to an Aspect Heuristic That Makes Sense to Learners
of Slavic Aspect System. X Nordiska Slavistmötet 13-17 augusti 1984, Åbo. 13–26.
Bıĺ ý, Milan. Iterative Verbs in West Slavic Languages Especially in Czech. Slovo:
Journal of Slavic Languages and Literatures (32). 21–37.
Boneh, Nora & Edit Doron. Habituality and the habitual aspect. Theoretical and
crosslinguistic approaches to the semantics of aspect 110. 321–348.
Bonomi, Andrea. Aspect, Quantification and When-Clauses in Italian. Linguistics
and Philosophy 20(5). 469–514.
Bybee, Joan & Östen Dahl. The Creation of Tense and Aspect Systems in the
Languages of the World. Studies in Language 13. 51–103.
Cable, Seth. Two paths to habituality: The semantics of habitual mode in Tlingit.
Semantics and Pragmatics 15.
Čapek, Karel. Obyčejný život [Ordinary life]. Olomouc: Otakar II.
Čapek, Karel. Hovory s T. G. Masarykem. Prague: Československý spisovatel.
Carlson, Gregory. The meaningful bounds of incorporation. Non-definiteness and
plurality 95. 35–50.
Carlson, Gregory N. A unified analysis of the English bare plural. Linguistics and
Philosophy 1(3). 413–457.
Carlson, Gregory N. Generic Terms and Generic Sentences. Journal of Philosoph-
ical logic 11. 145–181.
Carlson, Gregory N. Truth conditions of generic sentences: Two contrasting views.
In G.N. Carlson & F.J. Pelletier (eds.), The Generic Book, 224–237. Chicago: Uni-
versity of Chicago Press.
Carlson, Gregory N. Generics, Habituals and Iteratives. In Alex Barber (ed.), En-
cyclopedia of language and linguistics. Elsevier.
Carlson, Gregory N. Genericity. In Claudia Maienborn, Klaus von Heusinger &
Paul Portner (eds.), Semantics: An International Handbook of Natural Language
Meaning, 2–1153. De Gruyter Mouton.
Carlson, Gregory N. Suspicious Minds. Talk given at Yale Symposium on Generics.

70
 Generic sentences

Carlson, Gregory N. Suspicious Minds. Yale Symposium on Generics November

16, 2013.
Carlson, Gregory N. & Francis Jeffry Pelletier. The Generic Book. Chicago: Uni-
versity of Chicago Press.
Chierchia, Gennaro. Individual level predicates as inherent generics. In G.N. Carl-
son & F.J. Pelletier (eds.), The Generic Book, 176–223. Chicago University Press.
Chmelíčková, Petra. Určité a neurčité tvary slovesa v dětské řeči. Masaryk Univer-
sity Bachelor Thesis.
Cohen, Ariel. Think Generic!: The Meaning and Use of Generic Sentences. Stanford:
CSLI.
Comrie, Bernard. Aspect: An introduction to the study of verbal aspect and related
problems. Vol. 2. Cambridge university press.
Comrie, Bernard. Tense. Cambridge: Cambridge University Press.
Dahl, Östen. On Generics. In E.L. Keenan (Hg.) (ed.), A Companion to the Philos-
ophy of Language, 99–111. Cambridge University Press.
Dahl, Östen. Tense and Aspect Systems. Oxford: Blackwell.
Dahl, Östen. Contextualization and de-contextualization. In Rudi Studer (ed.),
Natural language and logic, 62–69. Berlin, Heidelberg: Springer Berlin Heidel-
berg.
Dahl, Östen. The Marking of the Episodic/Generic Distinction in Tense-Aspect
Systems. In G.N. Carlson & F.J. Pelletier (eds.), The Generic Book, 415–425. Chicago:
University of Chicago Press.
Dambriunas, Leonardas, Antanas Klimas & William R Schmalstieg. Introduction
to Modern Lithuanian. Brooklyn: Franciscan Fathers Press.
Danaher, David S. The Semantics and Discourse Function of Habitual-Iterative Verbs
in Contemporary Czech. Munich: Lincom.
Dowty, David R. Word meaning and Montague grammar: the semantics of verbs
and times in generative semantics and in Montague’s PTQ (Synthese language
library). Dordrecht, Boston: D. Reidel Pub. Co.
Farkas, Donka F & Yoko Sugioka. Restrictive if/when clauses. Linguistics and Phi-
losophy. 225–258.
Fernald, Theodore B. Predicates And Temporal Arguments. Oxford University Press.
Ferreira, Fernanda. Production-Comprehension Asymmetries. Behavioral and Brain
Sciences 27(2). 196–196.
Ferreira, Fernanda. Psycholinguistics, formal grammars, and cognitive science.
The Linguistic Review 22(2-4). 365–380.
Ferreira, Marcelo. The semantic ingredients of imperfectivity in progressives, ha-
bituals, and counterfactuals. Natural Language Semantics 24(4). 353–397.

71
Hana Filip

Filip, Hana. On Genericity: A Case Study in Czech. In Proceedings of the Nine-

teenth Meeting of the Berkeley Linguistic Society, 125–142.
Filip, Hana. Quantificational morphology. In S. Franks S. Avrutin & L. Progovac
(eds.), Formal Approaches to Slavic Linguistics, Vol. ii, the MIT Meeting 1993,
vol. 2, 144–177. Ann Arbor, MI: Slavic Publications.
Filip, Hana. Habituals and Q-adverbs. Conference on Genericity: Interpretation
and Uses. École normale supérieure, Paris, France.
Filip, Hana. Imperfectivity and Genericity: A Non-Reductionist Analysis. Workshop
on Imperfective Form and Imperfective Meaning. Yale University.
Filip, Hana. Word-internal and phrasal semantic relations. Invited Talk. Prague
Charles University invited talk Charles University, Faculty of Arts, Institute of
Czech Language and Theory of Communication, Prague, Czech Republic.
Filip, Hana. Characterizing Genericity and Epistemic Commitments. The workshop
on Formal Semantics and Its Interfaces, Faculty of Humanities, Zhejiang Uni-
versity, Hangzhou, China.
Filip, Hana. Suspicious Minds: Genericity with Epistemic Effects. Invited lecture.
University of Ostrava, Czech Republic.
Filip, Hana. Countability in natural language. Cambridge University Press.
Filip, Hana & Gregory N Carlson. Sui Generis Genericity. In Proceedings of the
Twenty-First Annual Penn Linguistics Colloquium, 91–110. University of Penn-
sylvania, Philadelphia, USA.
Forsyth, James. A grammar of aspect: Usage and meaning in the Russian verb. Cam-
bridge University Press.
Fortuin, Egbert & Jaap Kamphuis. The typology of slavic aspect: a review of the
East-West Theory of Slavic aspect. Russian Linguistics 39. 163–208.
Green, Lisa. Aspectual be–type constructions and coercion in African American
English. Natural Language Semantics 8. 1–25.
Greenberg, Yael. Manifestations of Genericity. New York: Routledge.
Greenberg, Yael. Exceptions to generics: Where vagueness, context dependence
and modality interact. Journal of Semantics 24(2). 131–167.
Hacquard, Valentine. Aspects of modality. Massachusetts Institute of Technology
PhD thesis.
Hamblin, Charles Leonard. Mathematical models of dialogue 1. Theoria 37(2). 130–
155.
Hamblin, Charles Leonard. Imperatives. Oxford: Basil Blackwell.
Heine, Bernd & Tania Kuteva. World lexicon of grammaticalization. Cambridge:
Cambridge University Press.

72
 Generic sentences

Heine, Bernd & Tania Kuteva. The genesis of grammar: A reconstruction. Vol. 9.
Oxford University Press, USA.
Isačenko, Aleksandr Vasil’evič. Grammatičeskij stroj russkogo jazyka v sopostavlenii
s slovackim – čast’ vtoraja: morfologija. Bratislava: Izdatel’ stvo akademii nauk.
Isačenko, Alexander V. Die russische Sprache der Gegenwart. Teil 1: Morphologie.
Munich: Hueber.
Kaplan, David. Demonstratives: An Essay on the Semantics, Logic, Metaphysics
and Epistemology of Demonstratives and Other Indexicals. In Joseph Almog,
John Perry & Howard Wettstein (eds.), Themes From Kaplan, 481–563. Oxford
University Press.
Kopečný, František. Dva přıś pěvky k vidu a času v češtině. Slovo a slovesnost 10(3).
151–158.
Kopečný, František. Zur Entstehung der Futurbedeutung beim perfektiven Präsens
im Slavischen. Scando-Slavica 8(1). 171–181.
Kopečný, František. Zur sogenannten Elementarverwandtschaft. Linguistics 3(19).
80–101.
Kopečný, František. Ještě ke gramatické, neaktualizaci’ českého slovesa. Slovo a
slovesnost 27(3). 258–261.
Koschmieder, Erwin. Zeitbezug und Sprache: Ein Beitrag zur Aspekt- und Tempus-
frage (Wissenschaftliche Grundfragen), philosophische abhandlungen, herausgeg.
v. r. hönigswald, h.xi.) Teubner, Leipzig-Berlin.
Kratzer, Angelika. Stage-level and individual-level predicates. In M. Krifka (ed.),
Genericity in Natural Language, 247–84. Berlin, New York: University of Tübin-
gen.
Kratzer, Angelika. Stage-level and individual-level predicates. Occasional Papers
in Linguistics 12(10).
Krifka, Manfred. Nominal reference and temporal constitution: Towards a seman-
tics of quantity. In Proceedings of the 6th amsterdam colloquium, university of
amsterdam, 153–173.
Krifka, Manfred. In Gregory N. Carlson & Francis Jeffry Pelletier (eds.), The Generic
Book, 238–264. Chicago: University of Chicago Press.
Krifka, Manfred. Definitional Generics as Second-Order Predications. Talk given at
Conference on Genericity: Interpretation and Use, Institut Nicod, Paris.
Krifka, Manfred, Francis Jeffry Pelletier, Gregory Carlson, Alice Ter Meulen, Gen-
naro Chierchia & Godehard Link. Genericity: an introduction. In G.N. Carlson
& F.J. Pelletier (eds.), The Generic Book, 1–124. Chicago: University of Chicago
Press.

73
Hana Filip

Kučera, Henry. Aspect, Markedness, and t in Tense and Aspect. In Philip Tedeschi
& Annie Zaenen (eds.), Syntax and semantics ann arbor, mich, vol. 14, 177–189.
Kučera, Henry. A semantic model of verbal aspect. In M. Flier (ed.), American
contributions to the ninth international congress of slavists, vol. 1 (Linguistics),
171–184. Columbus, Ohio.
Kučera, Henry. In the Beginning Was the Verb Markedness in Grammatical Cate-
gories. Prague Linguistic Circle Papers: Travaux du cercle linguistique de Prague
nouvelle série 3. 109.
Ladusaw, William Allen. Polarity Sensitivity as Inherent Scope Relations. The Uni-
versity of Texas at Austin.
Landman, Fred. The Progressive. Natural Language Semantics 1(1). 1–32.
Lawler, John. Studies in English Generics. University of Michigan Ph.D. Thesis.
Lawler, John M. Generic to a Fault. In Papers from the 8th regional meeting, chicago
linguistic society (cls 8).
Lehmann, Christian. Thoughts on grammaticalization (Arbeiten des Kölner Universalien-
Projekts 1). Institut für Sprachwissenschaft, Universität Köln.
Lehmann, Christian. Synsemantika. In Joachim Jacobs, Arnim von Stechow, Wolf-
gang Sternefeld & Theo Vennemann (eds.), 2. Halbband. An International Hand-
book of Contemporary Research, 1251–1266. Berlin, New York: De Gruyter Mou-
ton.
Lenga, Gerd. Zur Kontextdeterminierung des Verbalaspekts im modernen Polnisch.
Peter Lang International Academic Publishers.
Leslie, Sarah-Jane. Generics and the structure of the mind. Philosophical perspec-
tives 21. 375–403.
Leslie, Sarah-Jane. Generics: Cognition and Acquisition. The Philosophical Review
117(1). 1–47.
Lewis, David. Counterfactuals and comparative possibility. In IFS: Conditionals,
Belief, Decision, Chance and Time, 57–85. Springer.
Lewis, David K. Adverbs of Quantification. In Edward L. Keenan (ed.), Formal
semantics of natural language, 3–15. Cambridge University Press.
Loving, Richard & Howard P McKaughan. Awa verbs part i: The internal struc-
ture of independent verbs. In Alan Pence (ed.), Verb studies in five New Guinea
languages, vol. 10 (Summer Institute of Linguistics Publications in Linguistics
and Related Fields), 1–30. Norman: Summer Institute of Linguistics of the Uni-
versity of Oklahoma.
Mari, Alda. Analyticity under Perspective: Indefinite Generics in French. In Atle
Grønn (ed.), Proceedings of Sinn und Bedeutung 12, vol. 12, 414–429. Oslo.

74
 Generic sentences

Mønnesland, Svein. The Slavonic frequentative habitual. In Casper de Groot &

Hannu Tommola (eds.), A Voyage into the Realm of Germanic, Slavonic and
Finno-Ugrian Aspectology, 53–76. Berlin, Boston: De Gruyter Mouton.
Montague, Richard. Formal Philosophy: Selected Papers of Richard Montague. R.
Thomason (ed.). New Haven,: Yale University Press.
Nef, Frédéric. Sémantique de la référence temporelle en français moderne. Vol. 32.
Herbert Lang Et Company Ag.
Nickel, Bernhard. Generics and the ways of normality. Linguistics and Philosophy
31(6). 629–648.
Nickel, Bernhard. Dynamics, Brandom-Style. Philosophical Studies 162(2). 333–
354.
Nickel, Bernhard. Between Logic and the World: An Integrated Theory of Generics.
Oxford, GB: Oxford University Press UK.
Nickel, Bernhard. Generics. In A Companion to the Philosophy of Language, chap. 18,
437–462. John Wiley & Sons, Ltd.
Nunberg, Geoffrey & Chiahua Pan. Inferring quantification in generic sentences.
CLS 11. 412–422.
Pelletier, Francis Jeffry. Are All Generics Created Equal? In F. J. Pelletier (ed.),
Kinds, Things, and Stuff: Mass Terms and Generics, 60–79. Oxford University
Press.
Pelletier, Francis Jeffry. Mass Terms: A Philosophical Introduction. In Kinds, Things,
and Stuff: Mass Terms and Generics, 161–169.
Pelletier, Francis Jeffry & Nicholas Asher. Generics and Defaults. In J. van Ben-
them & A.T. Meulen (eds.), Handbook of Logic and Language, 1125–1177. Ams-
terdam.
Petr, Jan. Mluvnice češtiny [Grammar of Czech] (Mluvnice češtiny 2). Prague:
Academia.
Portner, Paul. The progressive in modal semantics. Language. 760–787.
Rimell, Laura & Thomas Lue. VP-preposing and relative scope. University of Penn-
sylvania Working Papers in Linguistics 11(1). 19.
Ross, John Robert. Remnants. Studies in Language. International Journal sponsored
by the Foundation “Foundations of Language” 1(1). 127–135.
Scheiner, Magdalena J. Temporal anchoring of habituals. In Joanna Sio Marjo van
Koppen & Mark de Vos (eds.), Proceedings of ConSole XI, 1–15. Leiden: Univer-
sity Centre for Linguistics.
Searle, John R. The construction of Social Reality. Simon & Schuster.
Širokova, Aleksandra G. O Kategorii Mnogokratnosti v Češskom Jazyke, Issledovanija
po Češskomu Jazyku. Moscow: Akademija nauk SSSR.

75
Hana Filip

Širokova, Aleksandra G. Osnovnoe značenie mnogokratnyx glagolov v češskom

jazyke. Voprosy jazykoznanija (2). 73–84.
Sperling, Annie Joly. The Semantic Markings of Russian Verbal Suffixes. Andrews,
Edna. 159–98.
Stalnaker, Robert. On the representation of context. Journal of Logic, Language
and Information 7. 3–19.
Swart, Henriette. Adverbs of Quantification: A Generalized Quantifier Approach.
Rodi.
Van Geenhoven, Veerle. Noun incorporation. Glot international 5(8). 261–271.
Van Geenhoven, Veerle. For-adverbials, frequentative aspect, and pluractionality.
Natural language semantics 12(2). 135–190.
Vinogradov, Viktor V. Russian Language: (the Grammatical Doctrine about a
Word). Moscow: Vysšaja Škola.
Vlach, Frank. Temporal adverbials, tenses and the perfect. Linguistics and Philos-
ophy 16(3). 231–283.
Wiemer, Björn. Between analytical mood and clause-initial particles – on the
diagnostics of subordination for (emergent) complementizers. Zeitschrift für
Slawistik 68(2). 187–260.
Wiemer, Björn & Ilja A Seržant. Diachrony and typology of Slavic aspect: What
does morphology tell us. Unity and diversity in grammaticalization scenarios
16. 239.

76