Multifunctional Morpheme a in Czech: DM with the Superset

Abstract

This article concerns the morpheme a in Czech. It occurs in nominals, conjunctions, and various verbal predicates. In contrast to the common practice of treating such a exponents as independent, accidentally homophonous elements, it is argued that some of these as can be treated as one item. What the syncretic as have in common is pluralizing semantics. Thus, the article proposes that verbal number (specifically, plurality) is related to nominal number and conjunctions. The article addresses the questions of how the multifunctionality of morphemes—such as the Czech a—can be analyzed and which tools of lexical–realizational approaches to morphology are most suitable for the analysis. In addition to the plural interpretation, a brings about changes in the argument structure of verbal predicates and fulfills several functions in the nominal and conjunction domains. The analysis is couched in the Distributed Morphology framework. However, contrary to expectations, the multifunctional a is not treated as an underspecified marker. It is analyzed as an overspecified marker that can realize (i.e., span) several syntactic heads: the pluralization head with the pluralization operator, the voice head, plus some other heads present in verbs and nominals. It is argued that the best option for deriving the multifunctional property of a is to assume the superset principle and pre-linearization spanning.

1. Introduction: Multifunctionality of a

There is a growing body of literature on the multifunctionality of morphemes and on differences between roots, affixes, and categorizing heads (see, e.g., De Belder, 2011; Lowenstamm, 2014; Creemers et al., 2017; Simonović, 2022). This article is concerned with the multifunctional morpheme a in Czech.1 It occurs in the nominal, conjunction, and verb domains, and it can be a derivational as well as an inflectional affix. These as are usually considered to be independent elements (Petr, 1986a, 1986b; Karlík et al., 1995; Cvrček et al., 2015). In other words, the phonological identity of these elements is treated as accidental homophony. In contrast to this view, this article proposes treating some of the a exponents as one item, i.e., it argues for the syncretism perspective. There are various approaches to morphology and syncretism (see Stump, 2001). This article concentrates on the lexical-realizational approach. It addresses the question of how multifunctionality (of morphemes like a) can be dealt with. At the same time, it addresses the question of what type of elements multifunctional affixes are.

In the generative tradition, there are several approaches to multifunctional morphemes and to the way in which they are inserted at particular nodes in the structure. For discussion of the difference between accidental homophony and systematic syncretism, consider, e.g., Embick (2015). According to Distributed Morphology (Halle & Marantz, 1993, 1994; Harley & Noyer, 1999; Embick, 2015; Bobaljik, 2017), morphological exponents are inserted at terminal nodes (heads) in line with the subset principle and the specificity principle, and syncretism is treated in terms of underspecification. In contrast, in the nanosyntactic approach (Starke, 2009; Caha, 2018), exponents can also spell out non-terminal constituents, and their insertion observes the superset principle. For realizing more heads (by portmanteau morphemes), traditional Distributed Morphology uses the post-syntactic operation of fusion (Halle & Marantz, 1993), which has been criticized as being stipulative and redundant; see, e.g., Trommer (1999). An alternative is the operation of spanning, proposed, e.g., by Julien (2015), Merchant (2015), Svenonius (2016), Wiland (2019), and Blix (2021b). The spanning operation can either occur before the linearization process, or it can follow it; see Haugen and Siddiqi (2016) for a relevant discussion.

The analysis proposed here is couched in the Distributed Morphology framework and employs the operation of spanning (e.g., Haugen & Siddiqi, 2016). Comparing the relevant morphological approaches, the article argues that for apparently unrelated markers such as the multifunctional a, the superset principle is more appropriate than the subset principle. In addition, building, e.g., on data with aktionsart (superlexical) prefixes, it is argued that the spanning operation is not sensitive to linearization.

The rest of this section discusses various instances of the morpheme a in Czech and shows that it is multifunctional in several respects. Some as seem to only bring about a grammatical meaning, like the genitive ending in example (1b), which contrasts with the nominative example in (1a).

(1)	a.	les
		forest-NOM.SG.M
		‘a forest’
	b.	les-a
		forest-GEN.SG.M
		‘of a forest’

Other as clearly have a semantic import; consider, e.g., the iterative -a in example (2b) and the habitual -(v)a in (2c), which contrast with the semelfactive example (2a), denoting only one event of stabbing.2

(2)	a.	bod-nou-t
		stab-TH-INF
		‘to stab once’
	b.	bod-a-t
		stab-TH-INF
		‘to stab repeatedly’
	c.	bod-á-va-t
		stab-TH-HAB-INF
		‘to tend to stab repeatedly’

The semantic effect of a is also shown in the following example. Consider the contrast between (3a) and the conjunction example (3b), which contains a and denotes a group of two people.

(3)	a.	Tom
		Tom
		‘Tom’
	b.	Tom	a	Lucie
		Tom	and	Lucie
		‘Tom and Lucie’

This article is concerned with as that have a semantic import, specifically, with as that bring about a plural interpretation, as in (2b,c), (3b). The same effect of a can also be found in the nominal domain, as shown in example (4b). Cases like (1b) are not relevant to this article, and the genitive, singular, masculine a is considered to be distinct from the pluralizing as in examples (2), (3), and (4).

(4)	a.	kol-o
		bike-NOM/ACC/VOC.SG
		‘a bike’
	b.	kol-a
		bike-NOM/ACC/VOC.PL
		‘bikes’

Since as in examples (2), (3) and (4) have different functions and belong to words of different word classes, they are usually treated separately: the morpheme a in (4b) as a case ending, the a in (3b) as a coordinating conjunction and the as in (2b) and (2c) as a theme and an aspect marker of the verbal paradigm (see, e.g., Kopečný, 1962; Karlík et al., 1995; Šimandl, 2016). In contrast, this article proposes a unified analysis. This is in line, e.g., with the analyses of Simonović and Mišmaš (2020) and Simonović (2022), who treat certain affixes occurring in different word classes in Slovenian as the same element. Specifically, given the pluralizing effect of the a morphemes in cases like (2)–(4), the as will be analyzed as the realization of a pluralization operator (among other things).

The pluralizing a is multifunctional in several respects. As already shown, it occurs in more categories: in the verbal category, in the nominal domain, and in conjunctions. Further, it is multifunctional from the morphological point of view. It can be an ending, as in (4b); a stem affix, as in (2b) and (2c), and a word, as in example (3b). Moreover, it can be used both in derivation and inflection. It is also multifunctional morphosyntactically since it can take either one overt argument, as in (2b), (2c), and (4b), or two overt arguments, as in example (3b). The pluralizing a is multifunctional semantically as well. It can apply to eventive stems, as in (2b) and (2c), or to individuals, as in (3b) and (4b). It can also have more functions in one category. Besides being a (pluralizing) theme, as in (2b), or a habitual suffix, as in (2c), it can also function as a secondary imperfective marker, as shown by the verb in (5c), based on the prefixed predicate (5b).

(5)	a.	péc-tIPF
		bake-INF
		‘to bake/to be baking’
	b.	o-péc-tPF
		around-bake-INF
		‘to roast’
	c.	o-pék-a-tIPF
		around-bake-SI-INF
		‘to roast/to be roasting’

The remainder of the article is structured as follows. The next section discusses a occurrences in the three relevant domains: Section 2.1 is concerned with the pluralizing a of the nominal domain and Section 2.2 with the pluralizing a of the verbal domain. Section 2.3 discusses a contained in conjunctions. Section 3 offers a Distributed Morphology analysis, and its particular subsections deal with plurality, morphosyntax, spanning, and the superset principle. Section 4 concludes the article.

2. Three Domains with a

2.1. The Nominal Domain

The pluralizing a occurs in neuter nouns, concretely, in the nominative, accusative, and vocative plural of the first and third declensions. This is illustrated in

Table 1. Neuter nouns of the first declension.

Plural of	Město ‘Town’
NOM.PL GEN.PL DAT.PL ACC.PL VOC.PL LOC.PL INS.PL	měst-a měst měst-ům měst-a měst-a měst-ech měst-y

and

Table 2. Neuter nouns of the third declension.

Plural of	Kuře ‘Chicken’
NOM.PL GEN.PL DAT.PL ACC.PL VOC.PL LOC.PL INS.PL	kuřat-a kuřat kuřat-ům kuřat-a kuřat-a kuřat-ech kuřat-y

, with the paradigms of město ‘town’ and kuře ‘chicken’, respectively.3

There are also other as in the nominal domain, e.g., the genitive singular masculine ending in (1b) and the nominative singular -a of feminine nouns, e.g., in žen-a ‘woman’. These suffixes do not pluralize, and I take them to be distinct elements. The suffix a is also present in certain adjectives, pronouns, and participles when they co-occur with a noun like měst-a ‘towns’ (see Table 1). In such forms, the ending a functions as an agreement (concord) marker, as in the case of the deictic pronoun ta in t-a měst-a ‘the/those towns’.4

2.2. The Verbal Domain

In the verbal domain, there are several forms that contain the pluralizing a: motion verbs, multiplicative predicates, and secondary imperfectives with the iterative interpretation.

There are two main types of motion verbs in Czech (and in Slavic generally), unidirectional—also called “determinate” and “directed”—and multidirectional, also referred to as “indeterminate” or “non-directed” (see, e.g., Sussex & Cubberley, 2006). Multidirectional motion predicates denote movement that is repeated, either several times or habitually, and can contain the pluralizing theme -a, as shown in (6b). This example contrasts with (6a), which contains the theme vowel -e and has the single-action interpretation (see, e.g., Kuryłowicz, 1964; Nichols, 2010; Medová, 2012 for the claim that different theme vowels in Slavic motion verbs can bring about different interpretations).

(6)	a.	Jan	běž-e-l	k	řece.
		Jan.NOM	run-TH-PTCP	to	river
		‘Jan ran to the river./Jan was running to the river.’
	b.	Jan	běh-a-l	k	řece.
		Jan.NOM	run-TH-PTCP	to	river
		‘Jan ran to the river repeatedly.’

Nothing changes on this contrast when the directional prepositional phrase is replaced with a direct object, as demonstrated in example (7), with the accusative object maraton ‘marathon’. That the interpretation of the motion predicate with -a is iterative is obvious from the fact that the predicate cannot be modified by the adverb jednou ‘once’, in contrast to the example (6b), with the predicate containing the vowel -e.

(7)	a.	Jan	běž-e-l	maraton	(jednou).
		Jan.NOM	run-TH-PTCP	marathon.ACC	(once)
		‘Jan ran a marathon (once).’
	b.	Jan	běh-a-l	maraton	(#jednou).
		Jan.NOM	run-TH-PTCP	marathon.ACC	(#once)
		‘Jan ran a marathon repeatedly (#once).’

Multiplicative predicates also contain a and denote actions that occur repeatedly, as illustrated in example (8). (In contrast, for actions that occur only once, the semelfactive suffix is used, as shown in (2a).)

(8)	Maják	blik-a-l	(celou	noc).
	lighthouse.NOM	blink-TH-PTCP	(whole.ACC	night.ACC)
	‘The lighthouse blinked (all night long).’

Finally, let us consider secondary imperfective verbs, i.e., predicates derived by a suffix that imperfectivizes the perfective, prefixed verb. The verbs are called secondary imperfective since there are also primary imperfectives: unprefixed verbs (which are mostly imperfective and can be perfectivized by a prefix). Secondary imperfectives can also have a pluractional interpretation in Slavic (besides the other three readings, see Comrie, 1976; Dahl, 1985; Dickey, 2000; Klimek-Jankowska et al., 2025). In Czech, -a (-va in the case of stems ending in a vowel) is one of two productive secondary imperfective suffixes (along with -ova); see (9a) for an example. The four readings of the secondary imperfective in (9a) are the iterative (or more generally, pluractional) reading, with several roasting events, the progressive reading, the habitual meaning, and the general factual reading. In contrast, the perfective example in (9b), without the suffix -a, only denotes one completed event of roasting.

(9)	a.	O-pék-a-l	to.
		around-bake-SI-PTCP	it
		‘He roasted it several times.’ ‘He was roasting it.’ ‘He tended to roast it.’ ‘He already roasted it.’
	b.	O-pek-l	to.
		around-bake-PTCP	it
		‘He roasted it.’

Having completed the verbal domain, let us now turn to conjunctions.

2.3. Conjunctions

As shown by example (3b), repeated for convenience as (10a), a can function as a monomorphemic conjunction. The examples in (10b) and (10c) demonstrate that the morpheme a can also be part of polymorphemic conjunctions.

(10)	a.	Tom	a	Lucie
		Tom	and	Lucie
		‘Tom and Lucie’
	b.	Tom	a-ni	Lucie
		Tom	and-not	Lucie
		‘neither Tom nor Lucie’
	c.	malý	a-le	pěkný
		small	and-PRTC	nice
		‘small but nice’

The data in (10) show that the conjunction a as well as the polymorphemic conjunctions take two syntactic arguments. The conjunctions conjoin individuals, as in (10a) and (10b), and predicates, as in example (10c). The morpheme a again pluralizes. It returns the sum of the denotations of its arguments. It can be pure plurality, as in (10a), or plurality with some other meaning, e.g., adversative, as in example (10c), or negative polarity, as in example (10b).

3. Analysis

3.1. Plurality

It was shown in Section 2 that there is a pluralizing morpheme a, which occurs in the verbal, nominal, and conjunction domains. That the as of the three domains are related is supported by the fact that the domains display an interesting exponent overlap with respect to plurality. Generally, the three domains only use the phonemes /a/, /ɛ/, /ɪ/ (and their long counterparts) for marking plurality. There are no back vowel phonemes /o/ and /u/ in this function.

In the nominal domain, neuter nouns use endings /a/, /ɛ/, and /iː/ for the plural nominative, accusative, and vocative, as demonstrated for /a/ in Table 1 and Table 2. Feminine nouns use /ɛ/ and /ɪ/ for the plural nominative, accusative, and vocative, while masculine nouns employ the endings /ɛ/, /ɛː/, and /ɪ/. For reasons of space, I only show the paradigms with -a, as in Table 1 and Table 2. Note that grammar books and descriptive works on the declension of Czech nouns mostly assume fourteen declension classes (paradigms) for the three genders. Long hard adjectives and certain pronouns use for the plural nominative, accusative, and vocative: /aː/ for the neuter gender, /ɛː/ for the feminine gender, and /ɛː/ and /iː/ for the masculine gender. Soft adjectives use /i:/ for all endings in all genders.5

The verbal domain also uses only the phonemes /a/, /ɛ/, and /ɪ/. For the iterative meaning of secondary imperfectives and for the habitual meaning, /a/ and /ɛ/ are used, as illustrated, e.g., by the habitual example (2c).6 Multiplicative verbs employ the suffix /a/, as shown in examples (2b) and (8), and non-directed verbs use /a/ and /ɪ/, as demonstrated in (6b) and (7b).

In the conjunction domain, there is also /a/ and /ɪ/, as shown by (10a) for the conjunction /a/ and by Tom i Lucie ‘Tom and even Lucie’ for the conjunction /ɪ/ (see also example (37)). For some reason, the phoneme /ɛ/ is not present.

Thus, the pluralizing a should realize some operator in the three domains that returns a sum of individuals, events, etc., as will be demonstrated below. In this respect, I follow the mereological approach of Link (1983, 1998), Landman (1991), and Florio and Nicholas (2021) and assume that a spells out a head that can host the pluralization operator *. I will apply various tests below in order to better map the nature of this element. The pluralization operator directly captures the semantic effect of plural endings such as -a. Specifically, the application of the pluralizing a to ‘town’ (see Table 1 again) results in any sum of towns; consider the meaning of the nominative, accusative, and vocative plural form of města in (11).

(11)	[[města]] = λx.*town(x)

Example (12) shows that města must denote at least two towns (or two town kinds). The plural noun pluralizes the corresponding predicate in (12a), marked with the plural agreement; hence, there will be more events of living up. There cannot be just one event of living up, as demonstrated by the ungrammatical (12b), with the singular form of the verbal predicate (oživne is, of course, compatible with the singular měst-o ‘town‘).

(12)	a.	Měst-a	oživn-ou.
		town-PL.NOM	liven.up-3.PL
		‘Towns will liven up.’
	b.*	Měst-a	oživn-e.
		town-PL.NOM	liven.up-3.SG
		Intended: ‘Towns will liven up.’

The pluralization operator * is also used in the case of event plurality in the verbal domain (see, e.g., Kratzer, 2008). When the pluralizing a applies to a motion predicate like ‘to run’, it returns a sum of running events, as illustrated in (13) (under the assumption that the agent argument is introduced by the head voice merged later in the derivation).

(13)	[[běhat]] = λe.*run(e)

That there is a sum of running events was demonstrated by example (7b) (repeated for convenience as (14b)), concretely by the incompatibility of the multidirectional motion predicate běh-a-l with the modifying adverb ‘once’. In contrast, if the pluralizing -a is not present, as in (14a), the predicate has a singular event interpretation and is compatible with ‘once’.

(14)	a.	Jan	běž-e-l	maraton	(jednou).
		Jan.NOM	run-TH-PTCP	marathon.ACC	(once)
		‘Jan ran marathon (once).’
	b.	Jan	běh-a-l	maraton	(#jednou).
		Jan.NOM	run-TH-PTCP	marathon.ACC	(#once)
		‘Jan ran marathon repeatedly (#once).’

In the same vein, a can realize the pluralization operator * in secondary imperfectives like opék-a-t ‘to roast’, which results in a sum of roasting events, as shown in (15a). Since secondary imperfective predicates also have interpretations that are compatible with ‘once’ (as the progressive one, shown in (9)), it is not possible to use an argument analogous to (12) and (14). It is, however, possible to deliver a positive argument, showing that secondary imperfective predicates with the pluralizing -a are compatible with iterative adverbs specifying the cardinality of the event set, such that it is greater than one. For instance, example (15b), with the iterative ‘twice’, demonstrates that the cardinality of the sum of roasting events can be two.

(15)	a.	[[opékat]] = λxλe.*roast(x)(e)
	b.	O-pék-a-l	to	dvakrát.
		around-bake-SI-PTCP	it	twice
		‘He roasted it twice.’

It has been argued that the pluralization operator * has specific properties, e.g., that it can be iterated (Sternefeld, 1998). This also holds for the pluralizing morpheme a, as illustrated in (16), repeated for convenience from (2b) and (2c).

(16)	a.	bod-a-t
		stab-TH-INF
		‘to stab repeatedly’
	b.	bod-á-va-t
		stab-TH-HAB-INF
		‘to tend to stab repeatedly’

The pluralization operator * has been argued to be optional (e.g., Sauerland, 1998), which also holds true for the pluralizing a, as shown in (17), partially repeated from example (5). The pluralizing secondary imperfective a can—but does not have to—attach to the predicate (17a). The same is also true for the habitual a in (16b).

(17)	a.	opéc-t
		around.bake-INF
		‘to roast’
	b.	opék-a-t
		around.bake-SI-INF
		‘to roast/to be roasting’

Another typical property of the pluralization operator * is that it is cross-categorial (e.g., Kratzer, 2008). This also holds for the pluralizing a since it was shown in preceding sections that the pluralizing a occurs in different categories, such as nouns, verbs, and conjunctions.

There is a distinction between nouns and verbs on one side and conjunctions on the other side. When the pluralizing a is a suffix, i.e., it is a head of the pluralization projection in nouns and verbs, then it takes one morphosyntactic argument: its sister constituent. In contrast, in the conjunction domain, a is (part of) a word and takes two morphosyntactic arguments, as shown in (18) for the example Tom a Lucie ‘Tom and Lucie’. The pluralization head is spelled out by a; its complement is Lucie, and the second argument, Tom, occupies the specifier position of *P.

(18)	[*P [Tom] [*’ [* a] [Lucie]]]

I assume that there is a relation between wordhood (prosody), morphosyntax (number of arguments), and semantics (plurality). The connection between semantics and prosody is mediated by syntax and is established in the following way. First, phases are mapped onto prosodic domains. Second, it has been argued that prosody can have a direct effect on syntactic structure (see, e.g., Richards, 2016, and references in Elfner, 2018). Third, there is a selection relation between the type of the pluralization head and the prosodic properties of its complement. The first type of the pluralization head selects prosodic words and has two arguments. The second type selects one argument that is not a prosodic word.

Concretely, under the assumption that nouns are phases and are mapped onto a prosodic word, in (18), the first type of the pluralization head merges with a prosodic word (Lucie), and also needs to have a prosodic word specifier (Tom). This pluralization head a cannot attach to the prosodic word Lucie as a suffix. This correlates with the semantic property of the head that it forms a sum from the denotations of the two syntactic arguments (Tom and Lucie) at the semantic interface. That is, the head semantically functions as the binary summation operator (⊕ in the mereological approach, see, e.g., Florio & Nicholas, 2021).

In contrast, in the verbal and nominal domains, the second type of the pluralization head is present, which selects one non-prosodic-word argument. Since the verbal and nominal stems are not prosodic words, the pluralizing head a can attach to them. Semantically, this pluralizing head a forms a sum from the denotation of its verbal or nominal complement. This amounts to saying that plurality is either intransitive—in the verbal and nominal domains—or transitive, as in the domain of conjunctions.

This proposal is not surprising in the light of analyses by Link (1983, 1998), Lasersohn (1995), and Landman (2000). For instance, Lasersohn (1995) argues for the unified treatment of conjoined noun phrases and plural noun phrases: both denote groups of individuals. He also extends the analysis to conjoined verbs and sentences; they denote groups of events, analogously to predicates with the pluractional meaning.

3.2. Morphosyntax

Besides realizing the head of the pluralization projection, a also has other functions. In the verbal domain, it verbalizes the acategorial root, as shown in (19) and (20). The roots běh and let, which can occur in both nouns and verbs, are verbalized by the pluralizing a in examples (19b) and (20b). In this respect, I make the standard assumption that theme vowels can realize the verbalizing head v in Slavic (Svenonius, 2004; Biskup, 2019; Klimek-Jankowska & Błaszczak, 2022; and Simonović et al., 2023).

(19)	a.	běh
		run.NOM.SG.M
		‘a run’
	b.	běh-a-t
		run-TH-INF
		‘to run repeatedly/in various directions’

(20)	a.	let
		fly-NOM.SG.M
		‘a flight’
	b.	lét-a-t
		fly-TH-INF
		‘to fly repeatedly/in various directions’

One might propose that it is the inner aspectual projection that a realizes in cases like (19) and (20). Indeed, there are approaches with an aspectual phrase inside the verbal domain in Slavic (e.g., Romanova, 2004; Łazorczyk, 2010; MacDonald & Markova, 2010). The first problem is that the aspectual head inside the verbal domain is often assumed for the secondary imperfective suffix (Romanova, 2004; Łazorczyk, 2010), but the pluralizing a can co-occur with the imperfectivizing suffix, as in poplác-á-va-t ‘to slap repeatedly for a short time several times’ (see also (30c)). The second problem of the inner aspectual analysis is that the iterative a does not have an aspectual effect in terms of telicity, quantization, or resultativity. It can occur in imperfective, atelic predicates, such as běhat ‘to run’, as well as in perfective, telic verbs, such as vyběhat ‘to get rid of sth. by running’, and in atelic secondary imperfectives, such as vyběhávat ‘to (be) get(ting) rid of sth. by running’. In all these examples, it is either the prefix or the secondary imperfective suffix that brings about (a)telicity, (non-)quantization, or resultativity. This is obvious, e.g., from the fact that the prefixed vyběhat is compatible with in-adverbials like za hodinu ‘in an hour’ but incompatible with durative for-adverbials like hodinu ‘for an hour’. In contrast, the unprefixed verb běhat behaves in the opposite way with respect to this telicity test. The same also holds for the verb vyběhávat, derived from the perfective verb by the secondary imperfective suffix; it is also compatible with durative for-adverbials like hodinu ‘for an hour’ and incompatible with in-adverbials like za hodinu ‘in an hour’.7

It was shown in example (9) that a can form secondary imperfective predicates with the pluractional meaning. According to Biskup (2024, 2025, n.d.), secondary imperfective suffixes can head the progressive and iterative projections, and the iterative projection responsible for the pluractional meaning is placed above the delimitative prefix po- in the clausal hierarchy. In contrast, the pluralizing a of motion verbs and multiplicative predicates seems to be structurally lower than the delimitative po-, as shown in examples (21) and (22).

The argument is based on scope properties of—hence c-command relations between—the pluralizing a and the delimitative po-. Firstly, that the pluralizing a is lower than po- is obvious from the fact that the attachment of the delimitative prefix in (21b) perfectivizes the imperfective predicate containing the pluralizing a. This argument is again based on the fact that prefixes make verbs perfective (with a few well-defined exceptions). Secondly, it is evident from the fact that the prefix temporally restricts the whole sum of running (sub)events and does not restrict the particular events in the sum. In other words, if po- attached lower than the pluralizing a, one would expect an interpretation, according to which there are short-running (sub)events that are repeated (pluralized). However, such an interpretation is not available for (21b); only the reversed scope interpretation is possible: There are temporally unrestricted (sub)events of running in various directions and all together happened for a while.

(21)	a.	běh-a-tIPF
		run-TH-INF
		‘to run repeatedly/in various directions’
	b.	po-běh-a-tPF	si
		DEL-run-TH-INF	self
		‘to run repeatedly/in various directions for a while’

Multiplicative predicates display an analogous scope behavior, as demonstrated by example (22), with the verb bodat. The head with the pluralization operator, realized by a, is again in the scope of the delimitative prefix, as shown in (22b). The perfective predicate denotes a sum of stabbing events that are all together delimited, with the result that the repeated stabbing events do not lead to death.

(22)	a.	bod-a-tIPF
		stab-TH-INF
		‘to stab repeatedly’
	b.	po-bod-a-tPF
		DEL-stab-TH-INF
		‘to stab repeatedly but not to death’

From this, I conclude that a can spell out the verbalizing head v, the head of the internal pluralization projection (present in multiplicatives and motion verbs), and the head of the external pluralization projection, which is present in secondary imperfective predicates.

Furthermore, a also marks changes in argument structure. In Slavic, including Czech, impersonal constructions are formed with the help of the reflexive se ‘self’. Unaccusative predicates do not form impersonal constructions in contrast to unergative and transitive verbs (see, e.g., Fehrmann et al., 2010). This is illustrated in example (23a). However, when the predicate is imperfectivized, and the pluralizing a is attached, se ‘self’ is licensed, as shown in (23b). Notice that it does not mean that an agent is added.

(23)	a. *	Na	pitevně	se	o-mdl-e-l-o.
		on	dissecting.room	self	about-faint-TH-PTCP-SG.N
	b.	Na	pitevně	se	o-mdl-é-va-l-o.
		on	dissecting.room	self	about-faint-TH-SI-PTCP-SG.N
		‘In the dissecting room, people fell into a faint.’

Now consider the pair of motion verbs: letět ‘to fly’, which is unidirectional, and létat ‘to fly’, which is multidirectional. If these verbs are prefixed, i.e., a complement expressing a result state is added (see, e.g., Ramchand, 2004; Gehrke, 2008 for the resultative function of prefixes), they behave differently. The unidirectional letět prefixed with od- can form the unaccusative -lý adjective, as in (24a), but cannot derive the transitive -ný adjective, as shown in (24b) (see Kosta & Frasek, 2004; Medová, 2012; Biskup, 2019 for the unaccusative versus transitive behavior of -lý adjectives and -ný adjectives, respectively).8

(24)	a.	odlétlý	kamínek
		fell.off	stone
		‘a stone that fell off’
	b.*	odletěný	kamínek
		fallen.off	stone
		Intended: ‘a stone that fell off’

On the contrary, the multidirectional létat, with the pluralizing -a, can derive the transitive -ný adjective, as in (25b), but cannot derive the unaccusative -lý adjective, as demonstrated in example (25a).

(25)	a.*	odlétalé	kilometry
		flew.away	kilometers
		Intended: ‘kilometers that were flown’
	b.	odlétané	kilometry
		flown.away	kilometers
		‘kilometers that were flown’

These facts show that odlétat ‘to reach the end by flying’ is transitive and agentive, whereas odletět ‘to fly away’ is unaccusative. In other words, -a in multidirectional verbs like létat and odlétat is related to agentivity, whereas the suffix -e in unidirectionals such as letět and odletět is related to unaccusativity. The other pair of motion verbs with identical theme vowels—the unidirectional běž-e-t ‘to run‘ and the multidirectional běh-a-t ‘to run‘—behaves alike (the other pairs of common motion verbs use different themes).9 This is in accord with the claim that in Russian, unidirectional motion verbs are unaccusative and multidirectional predicates are unergative (Schoorlemmer, 1995; Harves, 2002; Romanova, 2004).

This discussion suggests that a can also spell out the voice head. In the case of multidirectional motion verbs like létat ‘to fly’ and běhat ‘to run’, the voice is agentive, whereas in the case of unaccusative predicates like omdlévat ‘to faint’ in (23), the voice head is of the expletive type and licenses the reflexive se ‘self’ (see Alexiadou et al., 2015).

3.3. Spanning

Given that a should realize several heads, I use the operation of spanning; see, e.g., Abels and Muriungi (2008), Julien (2015), Merchant (2015), Caha and Ziková (2016), Svenonius (2016), Wiland (2019), Bleotu (2019), and Blix (2021b). Building on the discussion in the previous section, motion verbs, such as běhám ‘I run’, are analyzed as illustrated in (26). Example (26b) shows the structure of the motion verb (26a) after head movement. The colors indicate which portion of the structure of the complex head the particular morphemes realize. The structure before head movement looks like (27).

(26)	a.	běh-á-m
	b.	[[[[[[[√běh] v] *int] Voice] Asp] Tpres] Agr1sg]

Building on the discussion of the previous section, the root běh (in red in (26)) is verbalized by the head v, realized by the theme vowel a (in violet in (26)). This is in accord with the standard assumption that categorizing heads are adjacent to the root. vP merges with the internal pluralizing head *int, which is present in multidirectional verbs like běhat ‘to run’. It was shown in Section 2.2 that a has the pluralizing effect in multidirectional predicates like běh-a-t ‘to run’, in contrast to -e of the unidirectional běž-e-t ‘to run’. Thus, the pluralizing head *int is also spelled out by a (in violet in (26)). Notice that the internal pluralizing head is missing from predicates that do not have the pluractional meaning. In such cases, depending on the particular root, either a different theme is used, as in the case of the unidirectional běž-e-t ‘to run’ in (6), or the theme a spells out (a portion of) the structure without the head *int, as in děl-a-t ‘to do’. The internal pluralizing projection is below the voice projection in (27) since it is in the scope of superlexical (i.e., aktionsart) prefixes like the delimitative prefix po- (see Section 3.2), which in turn are lower than the voice projection with the agent (e.g., Tatevosov, 2015; Biskup, 2023).

It was demonstrated in Section 3.2 that a affects argument structure and licenses the agent argument. For this reason, a also spells out the voice head, as shown in violet in (26). The aspectual head is realized by a null exponent (see, e.g., Tatevosov, 2015; Biskup, 2023; Klimek-Jankowska & Błaszczak, 2023) and its value [perfective]/[imperfective] is determined by Agree with the closest aspectual affix (Biskup, 2023; Klimek-Jankowska et al., 2025). The exponent a is lengthened by a floating mora in (26a), which realizes the head T with the [present] feature (and generally attaches to the closest preceding vowel). If the feature [present] is missing, there is no mora in T that could lengthen the preceding vowel. Therefore, the infinitival běh-a-t ‘to run’ only contains the short a in contrast to běh-á-m ‘I run’ in (26a). Finally, the agreement marker -m is inserted at the Agreement head with the features [first, singular].

A similar representation is obtained in the case of secondary imperfective predicates with the pluractional meaning, like opékám ‘I roast’ in (28a), based on example (5c). The exponent a again spans (realizes) three heads (in violet in (28b)), but the difference is that in such verbs, a realizes the external pluralization head in addition to the verbalizing v and the voice head. Moreover, in opékám, a lexical prefix is present, which is represented as the incorporated preposition P in (28b).10 The structure before head movement is shown in (29).

(28)	a.	o-pék-á-m
	b.	[[[[[[[[P]√pék] v] *ext] Voice]Asp] Tpres] Agr1sg]

Since spans are argued to be head-complement sequences (e.g., Merchant, 2015; Svenonius, 2016), we expect intervening heads to block the spanning operation. This is correct, as shown by the verb containing both the internal and external pluralization projections in example (30c). Since aktionsart (superlexical) prefixes like the delimitative po- intervene between the internal and the external pluralization projection, two -(v)as are necessary for realizing the relevant piece of structure.

(30)	a.	plác-a-tIPF
		slap-TH-INF
		‘to slap repeatedly’
	b.	po-plác-a-tPF
		on-slap-TH-INF
		‘to slap repeatedly for a short time’
	c.	po-plác-á-va-tIPF
		on-slap-TH-TH-INF
		‘to slap repeatedly for a short time several times’

That the projection realized by the prefix po- is located between the two pluralization projections is evident from the aspectual and interpretational properties of example (30). As shown in (30b), the prefix perfectivizes and temporally delimits the base verb with the pluractional meaning. In (30c), the external pluralization -(v)a scopes over the delimitative po-, i.e., it imperfectivizes the predicate again and pluralizes the predicate temporally delimited by the prefix. Although the delimitative head realized by po- structurally intervenes between the internal and the external pluralization head, as shown in (31b), the exponent po-, being specified as a prefix, is linearized to the left of the root, as illustrated in yellow in (31a).

(31)	a	po-plác-á-vá-m
	b.	[[[[[[[[[√plác] v] *int] Del] *ext] Voice] Asp] Tpres] Agr1sg]

Thus, given the intervening delimitative head Del, the first -a (in violet) spans the verbalizing head v and the internal pluralization head, whereas the second -(v)a spells out the external pluralization head and the agentive voice head. That means that the spanning operation cannot be based only on linear adjacency or cannot follow linearization (as proposed, e.g., by Haugen & Siddiqi, 2016) since in the linearized verb, as in (31a), the prefix does not occur between the two pluralization morphemes.

As shown in (2c), Czech has a habitual -(v)a (see also Nübler, 2017). It has been proposed to analyze habituality in terms of plurality (e.g., Ferreira, 2016; but see also Filip, n.d.). Consequently, the habitual -(v)a can be taken to realize a pluralization head, as illustrated in (32), with the habitual bodávám ‘to tend to stab repeatedly’, based on example (16b).

(32)	a	bod-á-vá-m
	b.	[[[[[[[[√bod] v] *int] Voice] Asp] *] Tpres] Agr1sg]

The root bod is verbalized by -á, which also realizes the internal pluralization head and the voice head (in violet). Since the head with the habitual interpretation is located above the aspectual head in Czech (e.g., Biskup, 2023), there cannot be just one span realized by -á. Because of the intervening Asp, the second (-v)-á occurs, which spells out the habitual pluralization head (in blue in (32)). If this analysis is on the right track, then the consequence is that it is not linear adjacency that is relevant to the spanning operation (see also the discussion of (31)). What is relevant is hierarchical adjacency: Although the aspectual head is realized by a null exponent, it blocks the spanning operation. In other words, spanning can be a pre-linearization operation (contrary to the post-linearization proposal by Haugen & Siddiqi, 2016). In order to clarify whether languages can differ in this respect, further research is necessary.

3.4. Subset Principle or Superset Principle

In preceding sections, I argued against the accidental homophony position and maintained that the specific occurrences of a should be viewed as syncretic. What they have in common is the plural semantics. Recall also that a needs to realize not only several heads in the verbal domain but also heads of the nominal and conjunction domains. Thus, the as of the three domains are taken to form a natural class since they share the plural semantics, which will be modeled via the presence of the feature [*] in this section.

Let us now consider the question of how the vocabulary item of a should be defined. The Distributed Morphology approach (e.g., Halle & Marantz, 1993, 1994; Arregi & Nevins, 2013; Embick, 2015) assumes the subset and specificity principles, i.e., a vocabulary item can be inserted into a head if features of the item form a subset of the features in the head, and a more specific item blocks insertion of a more general item. Thus, the approach treats syncretism and multifunctional markers like the a under discussion in terms of underspecification. Given the broad distribution of a and given the fact that, e.g., in the verbal domain, there are also other theme vowels (such as -e and -i), a should be (fully) underspecified. The competing, distributionally more restricted themes should be more specific.

Let us assume for the sake of argument that a is fully underspecified, and let us consider the theme vowel -i, occurring, e.g., in motion verbs like nos-i-t ‘to carry’. Given that -i also verbalizes the root, it could be specified for the feature [v]. It could also have some other feature, e.g., the [Scale] feature, as proposed by Milosavljević and Arsenijević (2022) for the Serbo-Croatian theme -i. The [Scale] feature can be used for Czech, too, since the Czech conjunction i ‘even’ is scalar. If insertion of vocabulary items happens only once and only at one terminal, as originally assumed in Distributed Morphology, there will be problems with portmanteau morphemes. Concretely, because of the subset principle, features of i like [v] or [Scale] should also be present on the higher heads that are to be spelled out by i (e.g., on the voice head of the agentive type in nos-i-t ‘to carry’). The standard Distributed Morphology approach uses the post-syntactic operation of fusion (Halle & Marantz, 1993); however, it has been argued that this operation is unnecessary and that the result can be achieved in more standard ways (see, e.g., Trommer, 1999; Caha, 2018). An alternative solution is to spell out bigger portions of the structure (phases); consider, e.g., Julien’s (2015) spans. However, there will still be a conceptual problem with the nominal and conjunction domains if the subset principle is used, since, e.g., the verbal feature [v] should be present on the relevant heads in nouns and conjunctions.

A way out is to combine the operation of spanning with the superset principle used in Nanosyntax. With apparently unrelated exponents, such as the pluralizing a, the superset principle fares well since the incongruent features are part of vocabulary items, not of syntactic structure itself. In other words, it is not necessary to manipulate (enlarge the set of) features of syntactic heads to allow insertion of a specific marker. Notice that there is an important difference: In standard generative models, features of syntactic structure are sent to LF, which means that they are relevant to interpretation. Consequently, the incongruent features—like the verbal feature [v]—could affect the interpretation of unrelated elements, e.g., heads in nouns and conjunctions. In contrast, features of vocabulary items are relevant to vocabulary insertion, which happens in the mapping from syntax to phonological form (at PF). Hence, it does not bring any interpretational problem when a vocabulary item contains incongruent or even contradictory features (if the set of its features is enlarged to satisfy the superset principle). The features of the relevant syntactic heads—which are relevant to interpretation—remain unchanged in such a case. Thus, with respect to apparently unrelated exponents, the enlargement of the feature set of vocabulary items is preferred over the enlargement of the set of features of syntactic heads. Having said this, I define the multifunctional a as shown in (33).

(33)

a

↔ {v, *, Prog, Voice, Nom, Acc, N}

That is, a can realize the verbalizing head v, the pluralization head present in the three domains, the voice head, and the nominative, accusative, and neuter heads.11 Given that a can also realize the progressive head in secondary imperfective predicates, the vocabulary item also contains the feature [Prog]. It is important to note that the set of the vocabulary item is not ordered since there are variable orderings, e.g., the voice head is lower than the pluralization head * in habitual predicates, but it is higher than * in multiplicatives and secondary imperfectives.

For instance, in the habitual example (32), repeated as (34), the left (violet) á realizes the verbalizing head v, the internal pluralization head, and the voice head. I assume that these heads have features [v], [*], and [Voice], respectively. The comparison with the a vocabulary item in (33) shows that all three features are present in the set of features (plus features [Prog], [Nom], [Acc], and [N]). That means that (33) is in the superset relation with the syntactic context. Hence, á can realize the three adjacent heads: v, *_int, and Voice.

(34)	a.	bod-á-vá-m
	b.	[[[[[[[[√bod] v] *int] Voice] Asp] *] Tpres] Agr1sg]

Assuming that the aspectual head bears the features [Asp] and [imperfective], á cannot realize this head since the set of the vocabulary item (33) does not contain [Asp] and [imperfective]. Therefore, the pluralizing á can only span the three heads. At the habitual pluralization head, with the feature [*], the second á is inserted, again in accordance with the superset principle since the feature [*] is present in (33).

3.5. Conjunctions

Given the superset principle and the feature specification of the vocabulary item in (33), it is obvious that a can also realize the pluralization head with the feature [*] in conjunctions like a ‘and’ and ani ‘nor’ in (35), repeated from example (10).

(35)	a.	Tom	a	Lucie
		Tom	and	Lucie
		‘Tom and Lucie’
	b.	Tom	a-ni	Lucie
		Tom	and-not	Lucie
		‘neither Tom nor Lucie’

As discussed in Section 3.1, example (35a) can be analyzed as (36a), with the pluralization head realized by a, with Lucie in the complement position and the second argument Tom in the specifier position of *P. Ani ‘nor’ in example (35b) is a negative polarity item and needs to be licensed by sentential negation. Since it is -ni that makes a conjunction with the negative polarity property (more accurately, negative concord property) from the conjunction a, I assume that -ni is the head of a negative polarity phrase, as shown in (36b). Given that the negative interpretation comes from sentential negation, the [negative] feature of the NegPol head is uninterpretable and enters into an agree relation with the negation head of the clause, containing the interpretable [negative] feature (see, e.g., Zeijlstra, 2004, for an agree analysis of negative concord phenomena).

(36)	a.	[*P [Tom] [*’ [* a] [Lucie]]]
	b.	[NegPolP [NegPol ni] [*P [Tom] [*’ [* a] [Lucie]]]]

There are several ways to achieve the order Tom a Lucie. First possibility is to move the head a to the negative polarity head -ni. After that, Tom would move to the specifier of the negative polarity phrase. The second option is to assume the generalized head movement by Arregi and Pietraszko (2018), with the pronunciation of the lower copy of ani. It is also possible to achieve the correct order without head movement, e.g., via lowering; see Embick and Noyer (2001). Since it is not crucial for the main argument which option we choose, I leave the decision for future research. Possibly, -ni could also be decomposed into the negative polarity -n and the scalar head -i, as assumed for the conjunction i in the following paragraphs.12

As mentioned in Section 3.1, there is also the scalar conjunction i in Czech; consider example (37). It is obvious from the discussion thus far that to a large extent, the pluralization analysis can be applied to the morpheme i since it has the pluralizing function in multidirectional motion verbs, in feminine and masculine nouns, and in conjunctions.

(37)	Tom	i	Lucie
	Tom	even	Lucie
	‘Tom and even Lucie’

The syntactic structure of example (37) is shown in (38). Building on the discussion of the theme vowel i in the verbal domain in Section 3.4, I assume that i can spell out the scale head present in (38).

(38)	[ScaleP [Scale’ [Scale] [*P [Tom] [*’ [*] [Lucie]]]]]

In addition, i should also realize the pluralization head of the conjunction in (38). Given the superset principle, the vocabulary item of i is defined as shown in (39). It states that i can realize the verbalizing head v with [v], the scale and voice heads with the corresponding features [Scale] and [Voice], the pluralization head present in the three domains, and the nominative, accusative, masculine, and feminine heads with the corresponding features in the nominal domain.

(39)	i ↔ {v, *, Scale, Voice, Nom, Acc, M, F}

The vocabulary item in (39) allows i to span the two heads, * and scale, in (38). Since specifiers and adjuncts are not visible for the operation of spanning, Tom in the specifier of *P does not block the span of i. At the same time, since the vocabulary item of a does not contain the feature [Scale], as shown in (40), repeated from (33), a cannot span the pluralization and the scale head. In addition, given that i is more efficient (it can realize both heads), it wins over the potential form ai, with a realizing the pluralization head and i spelling out the scale head.

(40)	a ↔ {v, *, Prog, Voice, Nom, Acc, N}

Concerning the question of why a is inserted in conjunction structures in (36) instead of i, the comparison of (39) and (40) shows that the vocabulary item of a is less complex than the vocabulary item of i. Given this, it is the candidate with the fewest unused features. Therefore, it is a that realizes the (part of) conjunctions in (35a) and (35b).

3.6. Nominals

It has been argued that there is a case hierarchy (see, e.g., Blake, 1994). I follow Caha (2009), who treats case hierarchy in terms of containment and argues that non-structural cases are more complex than structural cases, such as nominative and accusative. Let us consider the relevant part of the case hierarchy in (41).

(41)	[[[[[X] Nom] Acc] Gen] Dat ]…

Given the superset principle and the vocabulary item (40), a cannot realize non-structural cases and feminine or masculine forms. It can only spell out neuter plural forms in nominative (vocative) and accusative.

For instance, for the accusative example města ‘towns’ (42a), taken from Table 1, I assume structure (42b), in which the head of the neuter gender is structurally lower than the pluralization head, which in turn is lower than case projections of nominative and accusative. These hierarchical relations are in accord with the (more or less) standard view; see, e.g., Kramer (2015), Wiltschko (2021), and also with the nanosyntactic perspective (Caha, 2021). There are various nanosyntactic approaches to nominal phrases and their declension—e.g., the peeling approach by Caha (2009), the backtracking analysis in Caha (2021), or the partial overwrite proposal by Blix (2021a)—and the structure of města could be more elaborate than (42b), including, e.g., the nominalizing head n above the root. Abstracting away from the technical details, it is primarily important for the current analysis that a can spell out the pluralization head of the specific declensions.

(42)	a.	měst-a
	b.	[√town [N [* [Nom [Acc
		‘towns’

In the case under discussion, a (in red) spans, i.e., spells out the contiguous set of four heads—the neuter head, the pluralization head, and the nominative and accusative heads—in accordance with the superset principle and the vocabulary item (40).

The answer to the question why i does not occur in (42) is easy. The presence of the feature [Neuter] in (42b) blocks the insertion (hence, spanning) of the i exponent since the set of the vocabulary item (39), repeated as (43), does not include [Neuter], i.e., it does not form a superset of the feature specification in (42b).

(43)	i ↔ {v, *, Scale, Voice, Nom, Acc, M, F}

4. Conclusions

The main goal of this article was to analyze the multifunctional morpheme a in Czech. It shows that multifunctional affixes can function as inflectional affixes as well as (categorizing) derivational affixes and that they also can realize various functional features. Specifically, a has the verbalizing function and can spell out several “derivational” heads in verbal predicates, e.g., the progressive and voice heads. It also has the pluralizing function in the verbal, nominal, and conjunction domains, as both a derivational and inflectional affix. The pluralizing function is found in various predicates in the verbal domain: multiplicative, habitual, motion, and secondary imperfective. While in multiplicatives and multidirectional motion verbs, a realizes the internal pluralization head, in secondary imperfectives, it spells out the external pluralization head, and in habitual predicates, the pluralization head is located above AspP. Moreover, a functions as an inflectional marker in the nominal domain, where it realizes case and gender features besides the plural number.

Contrary to the standard practice, I have analyzed the pluralizing a of the verbal, nominal, and conjunction domains as one vocabulary item. This item spells out the head hosting the pluralization operator responsible for forming a sum. In this way, I relate verbal plurality to nominal plurality and to conjunctions. In order to accommodate the multifunctional property of a, I have proposed a Distributed Morphology analysis which employs the superset principle (well-known from Nanosyntax) instead of the subset principle. The analysis treats a as an overspecified marker.

I have argued that for apparently unrelated exponents—such as the pluralizing a occurring in the three domains in Czech—the superset principle is more suitable than the subset principle because (incongruent) features of vocabulary items do not affect interpretation in contrast to features of syntactic structure itself. Building on data with scopally intervening elements such as the delimitative prefix po-, I have argued that the operation of spanning can already happen before linearization because it is not sensitive to the linearized form.