1 Introduction

The aim of this paper is to explore the verbal expression of events of change of state (CoS) in Bulgarian. In particular, it focuses on CoS predicates in the sense of Ryle (1949) who distinguishes between two kinds of predicates expressing CoS, achievements involving punctual, instantaneous change, and accomplishments involving gradual, incremental change. In a Slavic language like Bulgarian, the verbal expression of events in general (including events of change) is related to a choice between two or three semantically related but aspectually differentiated verbs. Thus, an event of writing a letter can be expressed by any of the three members of a so-called aspectual triplet piša–napiša–napisvam ‘write’, and an event of arriving at the station can be expressed by the so-called aspectual pair pristigam–pristigna ‘arrive’. While much work has been devoted to exploring the relationship between Slavic verbal aspect and event structure, little agreement has been reached, partly due to diverging views on language specific data, and partly due to inner-Slavic differences. Little to no relevant work exists on Bulgarian, a Slavic language with a uniquely complex aspectual system comprising perfective, imperfective and secondary imperfective verbal forms and morphologically marked aspectual tenses (Aorist and Imperfect) while also differing from most Slavic languages in a number of further relevant respects. Consequently, a central objective will be to cast more light on the nature of CoS related verbal triplets in Bulgarian in terms of the relation between their morphological build-up and their lexical and aspectual properties.

Let us start with some theoretical assumptions related to the notions that the discussion will be centered around and that have been controversially discussed in the literature. In a broader sense, all dynamic predicates can be seen to involve some kind of change, as observed by Dowty (1979), who distinguishes between indefinite (activities) and definite change (achievements, accomplishments). On the other hand, Levin’s (1993: 242–248) semantic class of change of state verbs in English (e.g. break, bend, dry) represents a narrower notion that naturally falls into the classes of achievements and accomplishments. I will be assuming that CoS predicates describe the temporal structure of events as involving a gradual or punctual change from a situation where a state does not hold to a situation where the state holds. More closely, I assume Rothstein’s (2004) reanalysis of Vendler’s (1957) classification of verbal predicates, where four basic types of events denoted by verbal predicates are distinguished based on two properties, [±stages] and [±event of change], cf. Table 1 (from Braginsky & Rothstein 2008: 32).¹ The property [±stages] applies to verbal predicates denoting (sets of) events that can be seen as composed of distinguishable stages (defined as in Landman 1992). Due to this property, activities and accomplishments occur naturally with the progressive, in contrast to achievements (and states), which have the property [–stages]. Achievements denote (sets) of events which involve a direct transition between two successive moments: the last moment at which ¬ϕ holds and the first moment at which ϕ holds, where ϕ is a state determined by the lexical content of the verbal predicate. Like achievements, accomplishments involve a lexically specified change into a predefined culmination point. This change is however defined in terms of an incremental process the different stages of which are ordered by some lexically (or contextually) determined criterion (Rothstein 2004: 53). Activities differ from accomplishments in that their event stages are merely temporally ordered and lack any additional internal structure or inherent order. Thus The child ran for an hour describes an event of running consisting of a number of minimal running activities, while The child read the story of Rapunzel describes an event in which the minimal reading activities are ordered in a way that is determined by the necessary steps required to bring about the state of the story of Rapunzel becoming read (ibid.: 110).

I further adopt Rothstein’s view that while aspectual properties like [±stages] and [±change] are properties of verbs, (a)telicity is a property of VPs. This distinction accounts for the fact that in an English accomplishment VP like build a house, the meaning of the accomplishment verb determines the activity involved (a building activity) and what incremental process is associated with it (a building activity progressing gradually towards a state of a house becoming built), whereas its (a)telicity depends on the properties of the verb’s direct object (cf. Verkuyl 1972; Dowty 1979 for a different view).

Let us look next at some data. As already mentioned, the verbal expression of events in Bulgarian is related to a choice among the members of aspectual triplets (or pairs).² Table 2 shows some triplets corresponding to English verbal predicates involving events of change – achievements like die, win and accomplishments like build, eat, read. I call the triplets corresponding to Rothstein’s lexical accomplishments “accomplishment triplets” and those corresponding to achievements “achievement triplets” for short.

Each triplet consists of a basic imperfective verb (ipfv), a perfective (pfv) one derived from the ipfv by means of prefixation, and a secondary imperfective (s-ipfv) verb derived from the pfv verb by means of suffixation (most commonly with the suffix -va), or phonetic change (as in pročitam ‘read’, umiram ‘die’). Note that all Bulgarian verbs represented in the table are given in their 1^st singular present indicative form, which is the verbal citation form due to the loss of the infinitive in Modern Bulgarian. Note also that the inflectional differences in the personal form of the verbs are due to their belonging to different conjugations (a-, e- or i-conjugation) depending on their thematic vowel (cf. e.g. Manova 2007). Importantly, the prefixes that participate in the formation of these triplets are “empty” or “purely perfectivizing”, i.e. prefixes that change the aspect of the basic ipfv verb without substantially affecting its lexical meaning. (I will postpone some further important remarks on Slavic prefixes until section 2.) The imperfectivization process, i.e. the derivation of s-ipfv verbs from pfv ones, changes the aspect of a pfv verb into imperfective, while the lexical meaning of the derived verb remains unchanged (cf. e.g. Nicolova 2017: 357f). In Bulgarian, imperfectivization is a regular and productive process that applies to all pfv verbs derived by (all kinds of) prefixes.³ This is in contrast to most other Slavic languages where defectivity with respect to imperfectivization affects a number of perfective verbs with empty prefixes that do not have secondary imperfective correlates, as in Russ. pisat’–napisat’–∅ ‘write’ and Polish pisać–napisać–∅ ‘write’ (Nicolova 2017: 361).⁴

The lexical/semantic correspondence between the English CoS verbs and aspectual triplets of the type described above raises the question of the aspectual characteristics of the triplet members with respect to the properties outlined in Table 1. Do all members of a triplet corresponding to an accomplishment verb like build or achievement verb like win have the properties of accomplishments and achievements respectively? Especially the existence of two imperfective triplet members is somewhat puzzling. In the literature, the aspectual status of s-ipfv verbs in Bulgarian is far from settled, with views ranging from their expressing grammatical/Viewpoint aspectual distinctions (e.g. Jetchev & Bertinetto 2002), atelicity (e.g. Łazorczyk 2010), telicity (e.g. Slabakova 2005), or a special “iterative” aspectual kind as a middle ground between pfv and ipfv verbs (e.g. Ivančev 1971; Chakyrova 1998). At the same time, the temporal properties of the different members in a Bulgarian aspectual triplet have to my knowledge never been systematically studied. It has been widely assumed that basic ipfv verbs in general denote activities or states (and are thus atelic), cf. Nicolova (2017: 349); Slabakova (2005); Łazorczyk (2010); Rivero & Slavkov (2014), while prefixed forms are mostly telic, denoting accomplishments and achievements (cf. e.g. Rivero & Slavkov 2014), based on the common assumption that most Slavic prefixes are resultative and derive telic predicates, i.e. achievements and accomplishments, from the atelic ipfv ones (e.g. Brecht 1985; Arsenijević 2007; Gehrke 2008).⁵ A tacit assumption that follows from this view is that while all ipfv verbs in Table 2 denote activities, both the pfv and s-ipfv verb corresponding to an accomplishment like read would denote events with the properties [+stages, +change], and those corresponding to an achievement like win would denote the properties [–stages, +change].

Much work on Slavic languages like Russian, Polish and Czech has addressed the question about the relationship between Slavic verbal aspect in terms of the (im)perfectivity distinction and event structure in terms of Vendlerian classes explicitly, such as Padučeva (1996); Braginsky & Rothstein (2008); Filip (2008); Ramchand (2008); MacDonald (2008), among others. Thus Padučeva (1996) argues that activities and states in Russian are manifested by (basic and secondary) ipfv verbs and achievements are realized by pfv ones. Accomplishments are realized by both pfv and imperfective verbs forming “bounded pairs”, where the perfective member denotes a process that was completed after reaching some inherent goal, cf. pfv pročitat’ in Ivan pročital knigu (‘Ivan read a book’), whereas the imperfective member denotes the same goal-oriented process that has however not yet reached its inherent goal but is still in progress, thus resembling progressive accomplishments in English, cf. čitat’ ‘read’ in Ivan čital knigu (‘Ivan was reading a book’). This view is supported in Braginsky & Rothstein (2008) who provide an analysis of the two types of accomplishments in the framework of Rothstein (2004).

As we shall see, this account is not applicable to Bulgarian, where ipfv verbs like četa ‘read’ do not entail completion even though they may be associated with an incremental theme. Besides, it does not account for the status of s-ipfv verbs in some Russian achievement and accomplishment triplets, as in the case of the s-ipfv counterpart (pročityvat’) of the ipfv verb čitat’. Moreover, it is not clear how this analysis applies to lexical classes of verbs other than verbs of creation and consumption, such as verbs of movement like run and degree achievements like warm, which denote a change along a path or scale when heading telic VPs, and thus behave similar to verbs with incremental theme arguments (cf. e.g. Hay et al. 1999). As I will show, however, in a motion verb triplet like tičam–iztičam–iztičvam ‘run’, or a degree predicate triplet such as toplja–stoplja–stopljam ‘warm’, only the s-ipfv member can be seen to denote gradual change along a path or scale, whereas the basic ipfv member denotes an activity event without an inherent culmination point, and the pfv member denotes instantaneous change. The ipfv verbs tičam and toplja would immediately be accounted for by Braginsky & Rothstein (2008: 15) in terms of activities under the assumption that “non-stative imperfective verbs denoting events with duration can be divided into those which are incremental (by hypothesis, accomplishments) and those which are not (uncontroversially, activities)”. However, it remains unclear how the respective s-ipfv triplet members (which are neither incremental nor plain activities) fit this distinction.

A different view is advocated in Filip (2008: 21ff.) who argues that most underived Slavic verbs are formally imperfective and semantically non-maximal (where most of them belong neither to the incremental nor the scalar class) and therefore have sets of unordered eventualities in their denotation. In contrast, Vendlerian accomplishments are realized by underived pfv verbs,⁶ which have as part of their semantics a maximality operator requiring that the understood upper bound of the described event be reached. She further argues that only a few Slavic root verbs qualify as Vendlerian achievements, mentioning Ru. najtí/Cz. najít ‘find’ and Ru. vstretit’/Cz. potkat ‘meet’, which she claims are odd with incremental adverbials like gradually, while paradigm examples of English stem verbs denoting achievements are expressed by derived perfectives in Slavic, as in Czech posnat ‘get to know’ derived from znát ‘know’. These Slavic achievement-like verbs are argued to be unable to shift from singular events into plural events when combined with indefinite plural or mass arguments. Instead, Filip contends, the corresponding ipfv verb must be used in order to express plurality of instantaneous events, thus effectively taking the stance that achievement verbs in Slavic have both pfv and ipfv realizations.

Although Filip potentially offers an account of the contrast between pfv and s-ipfv verbs in achievement tuples, it is not directly applicable to Bulgarian either. One problem is that the expression of plurality of instantaneous events in Bulgarian is not restricted to s-ipfv tuple members, since, as we shall see, also pfv verbs express plurality of events when used in the present or imperfect tense.⁷ A further challenge for Filip’s account comes from Polish, where according to Młynarczyk (2004) underived perfective verbs like kupic ‘buy’ denote culminations (understood as instantaneous changes of state), thus corresponding to Vendler’s achievements, whereas their s-ipfv “twins” refer to the culminating event in terms of the preparatory process that leads up to the culmination (both in episodic and iterative contexts).

I will propose that in Bulgarian aspectual triplets like those in Table 2, the empty prefixes mark the culmination, or the result state of an event, and that this is the case for both pfv and s-ipfv triplet members. This explains the fact that both the pfv and s-ipfv triplet members naturally head telic VPs. However, they differ in that while the pfv members denote an instantaneous change of state, their s-ipfv counterparts denote an event of change that is gradual. Basic imperfective triplet members on the other hand denote an activity and naturally head atelic VPs. I show that this distribution of aspectual properties holds independently of the lexical properties of the verbal roots, extending the analysis from accomplishment and achievement triplets to motion and scalar verb triplets. At the same time, I take into account the obvious parallels in the temporal properties of ipfv, pfv and s-ipfv verbs across the various semantic verbal classes, and capture both the aspectual and lexical properties of the triplet members in terms of Rothstein’s (2004) event type templates. I further propose that the (a)telicity (in terms of Rothstein’s atomicity) of the VP headed by each of the triplet members may vary depending on the aspectual tense operators (and possibly the temporal modifiers) the verb interacts with. Thus, while pfv verbs naturally head telic VPs, they head atelic VPs when used in the Present and Imperfect, where they get a habitual/iterative interpretation. Similarly, s-ipfv verbs which I argue to naturally head telic VPs in the Present historical and the Aorist, head atelic VPs on habitual and episodic readings in the Present and Imperfect. These results altogether support a view on which aspectual verbal triplets in Bulgarian (and the perfective/imperfective distinction in general) pertain to inner aspect (taking sides with e.g., Rivero & Slavkov 2014, among others) and pave the way for a future account of the complex interplay between the two levels of aspect in Bulgarian that still awaits a comprehensive in-depth analysis.

The paper is structured as follows. After providing some further background on the nature of the prefixes involved in the triplets, as well as on Bulgarian tense and its interaction with the perfective/imperfective distinction, in the next section I take a closer look at the triplets in Table 2 in terms of the aspectual properties [±stages] and [±change]. In section 3, I present my proposal. Section 4 discusses how the distinction between the property of having (or not) an inherent endpoint relates to Rothstein’s notion of telicity in the context of the Bulgarian verbal triplets investigated. Section 5 presents conclusions and open questions.

2 The aspectual properties of CoS related triplets

After introducing some background on Slavic prefixes and on the relation between tense and aspect in Bulgarian, I will examine the aspectual properties of verbal triplets corresponding to lexical accomplishment and achievement verbs in English. I do this by applying standard tests that have been used in the literature, following essentially Rothstein (2004) and Mittwoch (1991; 2019). The tests show that the pfv members of both accomplishment and achievement triplets have the same temporal properties, and that also basic and s-ipfv triplet members behave uniformly independent of the lexical properties of the verbal roots. I subsequently look at triplets formed from verbal roots that belong to the classes of scalar and motion verbs and find the same pattern of distribution of temporal properties across the members of the verbal tuples.

3 Proposal

I propose that both the aspectual and lexical properties of Bulgarian CoS triplets can be captured in Rothstein’s (2004) framework in terms of her event type templates:

• the ipfv triplet members have the structure of Rothstein’s activities independent of their semantic class

• the pfv members have the structure of Rothstein’s achievements and denote a change of state brought about by the activity denoted by the ipfv root

• s-ipfv members have the structure of Rothstein’s accomplishments and denote complex events consisting of a subevent of change, and an activity subevent denoted by the ipfv root

In order to spell out the proposal, let us look at the structure of an English lexical accomplishment like read which Rothstein (2004: 102) specifies in terms of the following template:

Informally, an accomplishment verb denotes a complex event (composed via a sum operation on events) consisting of an activity subevent e₁ and a temporally extended BECOME subevent e₂. Both the activity and the BECOME subevents are lexically specified by the verb, which is reflected in the subscripts <X> and <Y>. The BECOME subevent is an event of change that is incremental in the sense that each of its stages has a distinguishable upper bound and the stages have an inherent order. This is formally implemented in terms of a contextually determined incremental chain C(e₂) that is imposed on the event of change e₂ (via the stage-of relation that holds between its subevents defined in Landman 1992), to the effect that the incremental chain breaks down the event of change into its contextually relevant stages. Finally, the activity subevent e₁ is related to the incremental structure of the BECOME subevent e₂ via the incremental relation INCR which imposes the same incremental structure on the activity event. This is done by means of a function μ that maps each element of the incremental chain onto the part of the activity event that has the same running time. The incremental relation thus captures the intuition that the change of state event “measures out” the progress of the activity and thus of the event as a whole. Culmination is defined as the final minimal event in an incremental process, and more precisely, its final part or upper bound. The content of the BECOME event is derived from the structure of the incremental relation in the sense that it is “a change of state in which the participant “undergoes” the activity event” (ibid.: 109). In other words, the BECOME event associated with build a house will be BECOME-BUILT(e₂) ∧ Arg(e₂)=Th(e₁), the BECOME event associated with read a book will be BECOME-READ(e₂) ∧ Arg(e₂)=Th(e₁) (ibid.). Consequently, an accomplishment like read will have the interpretation below (cf. Rothstein 2004: 110):

In contrast, the ipfv verb četa ‘read’ has an incremental theme but this theme is not the argument of an incremental BECOME event. Rather, četa must be accounted for in terms of an activity, defined as in (19). I propose that the same structure can be assumed for ipfv triplet members with non-incremental theme arguments like pečelja ‘win’, bârša ‘wipe’ and toplja ‘warm’, cf. (20a), as well as for intransitive verbs like mra ‘die’, cf. (20b) and motion verbs like tičam ‘run’ which are intransitive but may be modified by measure phrases or goal arguments.²⁶

I further propose that the meaning of the pfv triplet members has the structure in (21) which is the structure Rothstein (2004) assigns to lexical achievements, where the instantaneous BECOME event will be determined by the lexical content of the verb.²⁷ This will result in the following representations of the events denoted by consumption/creation verbs like (21a), verbs with non-incremental themes like spečelja ‘win’, scalar predicates like (21c), and verbs like umra ‘die’ (where the theme argument experiencing the change is the subject):²⁸

Finally, I propose that the meaning of an s-ipfv verb like pročitam has the same structure as Rothstein’s lexical accomplishment read defined in (18):.

Intuitively, a pročitam (‘finish reading’) event is the sum of the activity četa ‘read’ and the BECOME-event pročeten (‘become read’). Similarly, a stopljam (‘finish warming’) event is composed of the activity topja ‘warm’ and the BECOME-event stoplen (‘become warmed’), and an iztičvam (‘getting at a destination by running’) event is the sum of the activity tičam ‘run’ and the BECOME-event of reaching the destination specified by the goal argument. Thus, a general template representing the structure of s-ipfv verbs will need to specify the semantic relation between the triplet members. I propose that this can be done as in (23), where the subscripts indicate that the activity is lexically specified by the root corresponding to the ipfv triplet member, and the event of change is lexically specified by the corresponding pfv member:

Summing up, I proposed that while ipfv triplet members describe a dynamic event without specifying its inherent endpoint, their pfv counterparts specify the event’s inherent endpoint, and the s-ipfv members additionally specify the activity leading to the endpoint. An obvious suggestion is that in triplets, the BECOME event is encoded in the empty prefix. This has been proposed in Gehrke (2008: 163) who argues that internal prefixes in Russian and Czech represent resultativity markers that form a complex predicate with the verb they attach to, such that verbs with internal prefixes encode a BECOME event and thus fall into the classes of accomplishments and achievements. More specifically, she suggests that internal prefixes directly mark the culmination of a BECOME event in terms of the upper bound of Rothstein’s incremental chain.²⁹ However, Gehrke’s proposal (which remains at a general level as it is not worked out in detail) does not take into account that achievements, in contrast to accomplishments, are not complex predicates, and that their BECOME events do not involve incremental processes. A more specific proposal that would also account for the Bulgarian data is the following:³⁰

• empty prefixes in pfv triplet members mark an instantaneous BECOME event, and the activity denoted by the root is understood as a preparatory activity

• in combination with an imperfectivizing suffix, the empty prefix marks an incremental BECOME event, and the activity denoted by the root is the activity involved in the incremental BECOME event denoted by the s-ipfv triplet member

In terms of morpho-syntax, a hierarchical representation of the relation between the internal prefix and the imperfectivizing suffix -va (and its allomorphs) could look like the one in Figure 1, where, following Rivero & Slavkov (2014); Svenonius (2004); Gehrke (2008); Žaucer (2009), I assume that the empty internal prefix heads a small clause complement of the verb and encodes resultativity, while the imperfectivizing suffix scopes over the prefix, modifying the prefixed pfv verb by introducing an activity subevent. An adequate morpho-syntactical account is however outside the scope of the present study.

Figure 1: The morpho-syntax of Bulgarian CoS triplets.

This kind of proposal runs against more traditional assumptions that Slavic prefixes affect grammatical aspect. It also runs against proposals like Filip (2008) who argues that Slavic prefixes cannot be viewed as telic operators since they do not always enforce the maximal interpretation of the event, drawing on s-ipfv verbs which on her account denote non-maximal events in spite of being prefixed. She argues therefore for a separation between maximality as a culmination requirement and the “culmination condition” as an ordering criterion on events. While the former is covertly encoded in (underived and derived) perfective verbs, the latter comes from the prefixes. Thus, although predications with s-ipfv verbs like Czech dopisovat ‘finish writing’ derived from incremental verbs (as well as s-ipfv verbs derived from closed or open scale root adjectives like vyprazdňovat ‘empty’) have sets of partially ordered events in their denotation due to the contribution of the prefix, the imperfective suffix on the verb explicitly suspends the requirement that the verb only has maximal events in its denotation. This view is however contradicted by the Bulgarian data presented in the previous sections, and also seems to run against data from Polish. Thus, comparing degree achievements with closed scales like clean and open scale ones like bend in Polish (among other languages), Martínez Vera (2021) concludes that culmination (in terms of event maximalization) is reached only in the case where the Polish verb has a prefix. Similarly, Kardos (2016) compares her findings on the role of Hungarian aspectual particles and resultative/locative expressions in marking telicity with data from Polish, suggesting that verbal prefixes play a similar role as means for morphologically encoding telicity.

4 Telicity

As pointed out in the introduction, I follow Rothstein (2004) in assuming that the temporal properties [±stages] and [±change] of events denoted by verbal predicates allow for predictions as to whether a verbal predicate can head a telic or atelic VP. Thus, Rothstein argues that while accomplishments naturally head telic VPs, they may also head atelic ones in case their direct object (or incremental theme) is a bare plural or mass noun. On the other hand, while activities naturally head atelic predicates, they may also head telic ones. They differ from accomplishments in that the telicity of VPs headed by activities does not depend on the properties of the direct object. Rothstein defines telicity in terms of atomicity, arguing that telicity has to do with counting, which in turn amounts to “picking out individuals according to a particular criterion for individuation” (ibid.: 167). In the domain of events this amounts to identifying atomic events. Thus a VP is telic if it provides criteria for individuating atomic events, and atelic otherwise. A sentence like Mary ran provides no criteria allowing to determine what counts as a single running event. On the other hand, in Mary ran a mile or Mary ran to the store, the measure phrase and the path argument of the running event provide a criterion for measuring the event, and the VP is telic. Accomplishments and achievements are naturally telic because of their association with (lexically specified) events of change, the BECOME events, which determine that the event is over when the change has taken place and thus provide criteria for atomicity.³¹ However, when the direct object of an accomplishment is a bare plural or mass noun which does not specify how many singular BECOME events occurred, as in Mary wrote magazine articles, it is left open how long the plural event took and the VP is atelic, and the theme is not atomic. Rothstein argues further that the atomicity of a theme argument is overtly ‘marked’ in English in terms of a lexically realized determiner (ibid.: 156), assuming that when a determiner is complemented by a bare plural that denotes a set of singular entities, the denotation of the nominal is shifted to a set of entities that count as atoms, which she calls an atomic set, and the VP denotes sets of atomic events (ibid.: 173–174). Thus a predicate like build 3 houses on its plural (distributive) reading denotes pluralities of singular events in which a house got built, involving a plural set the members of which are atomic events of building a house. Following Landman’s (2000) theory of plurality, Rothstein derives a plural set from a singular set via the summing operation, and construes plural events as sums of singular events, assuming that if the events they are summed out of are atomic, they will be sums of atomic events. Since 3 houses is a plural of a set of atomic houses, the predicate will distribute down to atomic accomplishments and thus atomic BECOME events, providing a measure for individuation or atomicity for the whole VP, and this in turn will result in a telic VP, in accordance with the telicity principle that Rothstein (2004: 158) defines as follows: “A VP is telic if it denotes a set of events X which is atomic, or which is a pluralization of an atomic set (i.e., if the criterion for individuating an atomic event in X are fully recoverable).”

Achievements, on the other hand, head telic VPs even when the argument to which the change happens is a bare plural or mass noun, since the atomicity of the BECOME event does not depend on the atomicity of its argument. This is due to the nature of the change lexicalized in an achievement, which is “the shortest possible change” and does not allow the event to be broken into smaller changes, thus making the BECOME event atomic.

Rothstein’s account of the relation between aspectual classes and telicity leads to the following predictions for the Bulgarian verbal triplets: (i) pfv triplet members will always head telic VPs, (ii) ipfv ones will naturally head atelic VPs but may head telic ones in some contexts, where telicity will not hinge on the properties of the theme, and (iii) s-ipfv triplet members will head telic VPs provided that they have an incremental theme that is atomic, and otherwise head atelic VPs. However, we must also take into account that telicity in Slavic is widely assumed not to depend on aspectual composition but to be located on the verb (cf. e.g. Mittwoch 2019: 43). Thus, Filip (2008) argues that in languages like Russian, Czech and Polish, the aspectual properties of the verb determine the interpretation of the incremental theme object, and not vice versa. In these languages, which lack definite articles, bare plural and mass noun objects of imperfective verbs are interpreted as indefinite (jel hleb ‘ate (some) bread’), whereas the ones of perfective verbs are interpreted as definite (sjel hleb ‘ate the bread’) (Mittwoch 2019: 13). Moreover, Filip (2008) argues that the direct object in telic Slavic verbs has a different role than in English, since the telic verb constrains the interpretation of its argument. Thus, if the argument of the pfv verb is a bare plural or mass noun, it is interpreted as referentially specific, as in Polish On zjadl^P kaszę/oliwki ‘He ate up all the porridge/olives’, since, she argues, the only maximal interpretation available for bare plural and mass nouns is to view them as a definite group of objects.³² In contrast, an imperfective verb does not constrain the interpretation of its direct object, as in On jadl^I kaszę/oliwki, where the direct object can be interpreted as a kind, as weak indefinite, or as definite, depending on the context. On the other hand, she argues that while incremental themes realized as singular count nouns and quantified DPs are also constrained by pfv verbs to have a maximal interpretation, they are not constrained to a referentially specific interpretation, and that arguments that are non-incremental themes are not constrained by a perfective verb to a maximal interpretation. As for imperfective verbs, Filip argues, when the VPs they head contain some material specifying an ordering criterion, the VP is still atelic and the maximality effects that are sometimes observed are due to the linguistic context and/or world knowledge.

In Bulgarian, which (next to Macedonian) is the only Slavic language that possesses a definite article and a (singular and plural) indefinite determiner (Geist 2013), the situation is different. On the one hand, the aspect of the verb seems to constrain the incremental theme object in terms of its reference, quantized vs. cumulative (in the sense of Krifka 1998). Thus, pfv and s-ipfv verbs require quantized themes, realized by numerals or definite NPs (cf. izjade^P/izjažda^SI (edna) jabâlka/*jabâlki/dve jabâlki/*hljab/hljaba ‘ate an apple/apples/two apples/bread/the bread’),³³ whereas basic ipfv verbs do not pose such requirements on their incremental theme which can be a count noun, a bare plural or a mass noun (cf. jade^I (edna) jabâlka/jabâlki/hljab ‘ate an apple/apples/bread’).³⁴ On the other hand, a pfv or s-ipfv verb does not in general seem to constrain its non-incremental object in terms of maximality or referential specificity (cf. Peter nameri^P/namira^SI zlato/zlatoto. ‘Peter found gold/the gold’), and this holds independent of tense. In view of Rothstein’s insight about the atomic nature of incremental theme arguments with lexically realized determiner, this would suggest that both pfv and s-ipfv verbs encode the requirement that their incremental theme is atomic, and in particular that pfv verbs will always have themes that are atomic. At the same time, it has been claimed in Slabakova (2005: 8) that pfv verbs with internal prefixes force a telic interpretation even in cases where the incremental theme is a bare plural noun, as in Peter narisuva kartini za pet časa (‘Peter painted pictures in five hours’), while pfv verbs with external prefixes allow both a telic and atelic interpretation (cf. Peter prerisuva kartini (za) pet časa ‘Peter repainted (some) pictures in/for five hours’). Moreover, s-ipfv verbs seem to behave differently than in Polish and Czech, since they do seem to require that their incremental themes are referentially specific.

In order to be able to arrive at a more in-depth account of the telicity facts with respect to the CoS related triplets under investigation, I next test the triplet members for compatibility with the time frame and time span adverbials in α time and for α time. These adverbials provide tests for distinguishing between telic and atelic VPs, where telic VPs can be modified by in α time and atelic by for α time. They are also related to Rothstein’s theory of atomicity in the following way: in α time measures the length of atomic events, i.e. it assigns a time-frame within which an atomic event took place and thus modifies sets of atomic events, whereas for α time imposes a temporal measure criterion on singular events which are not already atomic (Rothstein 2004: e.g. 177, 181). On the other hand, while in α time does not usually modify atelic VPs (cf. #Mary pushed the cart in half an hour), if a contextually determined criterion for atomicity is provided, the adverbial forces a telic interpretation (as in John ran in an hour). In addition, the modifier may co-occur with bare plurals (as in Mary built houses in 3 weeks), where it forces the bare plural to be interpreted as the pluralization of an (contextually determined) atomic set. However, this shift occurs at the embedded level, while at the higher level, the plural predicate is atelic and can be modified by for α time, as in Mary built houses in 3 weeks for three months.

We start with the pfv triplet members which were claimed to have the properties of English lexical achievements, and which are predicted to lead to telic interpretations, while encoding the requirement that their incremental theme arguments are atomic. This prediction seems to be borne out, since both pfv verbs with incremental theme arguments like postroja ‘build’ and stoplja ‘warm’ and such with non-incremental theme arguments like spečelja ‘win’ and iztičam ‘run to’ in (24) are compatible with the time frame adverbial in α time and incompatible with the time span adverbial for α time.

With respect to the properties of the arguments (both incremental and non incremental), we should note that while in all cases the pfv verb is in general compatible with both count and mass nouns, bare plurals are awkward and seem to need the addition of the indefinite plural determiner edni, or a modifier.³⁵ This fact can be explained in terms of Rothstein, on the assumption that the determiner and the modifier contribute a criterion for specifying atomic sets and thus picking out atomic events.³⁶ Thus, the pfv triplet members seem to encode next to the BECOME event also the requirement that their themes are atomic, in the same way as achievements do on Rothstein’s account.

Also when combined with the Present and Imperfect, where they yield an iterative/ habitual interpretation, the pfv triplet members are only compatible with the time frame adverbial and incompatible with the time span adverbial. However, at a higher level, the plural predicate becomes atelic, which is an effect that must be attributed to the interaction between the verbal temporal structure and the semantics of the habitual operator, which according to Verkuyl (1999: 151, 156) has the effect of universal quantification over instances.

Turning to the ipfv triplet members, the prediction is that they will naturally head atelic VPs since they do not lexicalize a BECOME event, but may in some contexts head telic ones, where however telicity will not hinge on the properties of the theme. We can see that the prediction is borne out. Thus, in (present and past) episodic contexts, the ipfv incremental četa ‘read’ and the scalar toplja ‘warm’ head atelic VPs, cf (26).

However, when used in the Present or Imperfect with a direct object that has a lexically realized determiner – a definite or indefinite³⁷ determiner or a numeral, the sentence gets a habitual or generic interpretation, while the VPs headed by the ipfv triplet members are telic at the level of the single events. This is predicted by Rothstein’s assumption that the determiner provides a criterion for picking out atomic events. In addition, it seems that this interpretation is not possible without the time frame adverbial. This suggests that the telic interpretation at the level of the single events is forced by the adverbial, which serves as an external measure providing a criterion for individuation/atomicity and requiring an atomic theme. The same holds for a motion verb like tičam ‘run’, cf. (27). Note also that this reading is not available in the Aorist for these verbs, which suggests that interaction with tense also plays a crucial role in yielding the observed interpretation.

Finally, we turn to the s-ipfv triplet members, where the prediction is that they should head telic VPs provided that they have an incremental theme that is atomic, and otherwise atelic. However, we can observe that the same restrictions that hold for the themes of the pfv triplet members hold also for their s-ipfv mates, cf. (28) where the s-ipfv verbs are not compatible with bare plural and mass nouns as themes. The examples also show that in present historic contexts, the s-ipfv triplet members head telic VPs, as the compatibility with the time frame adverbial shows.

Also when used in present and imperfect habitual contexts, the same restrictions hold for the theme arguments,³⁸ and the s-ipfv triplet members are only compatible with time frame adverbials, which indicates that at the level of the single events, they head telic VPs, cf. (29). At the same time, it seems that the presence of the adverbial is crucial for the telicity to hold at the level of the individual events.

At the outer level, however, the predicate denotes a plural, non-atomic event. The VP headed by the s-ipfv verb is atelic and can be modified by a time span adverbial, cf. (30).

Also in present and imperfect episodic contexts, as well as in the Aorist, where the interpretation is episodic, the s-ipfv verbs are incompatible with bare plural and mass nouns, cf. (31) and (32), respectively. However, despite the atomicity of their arguments, in both cases they are only compatible with the time span adverbial and incompatible with time frame adverbials. At the same time, while the sentences in (31) are clearly atelic,³⁹ the predicates in (32) do not seem to be atelic despite the presence of the time span adverbial, but entail completion, as indicated by the continuation sentences.

Thus, while the sentences in (31) intuitively receive an interpretation of “unsuccessful attempt” on which the reading/building event did not reach its culmination, the ones in (32) yield what we could call a “success upon multiple attempts”-interpretation, on which the building/warming/running event took a certain amount of time, possibly involving several attempts, but by the end of this period, all the houses/soups/children were completed/warm/at their destination. The former interpretation is not unexpected as it resembles the effect of the Progressive on accomplishment verbs in English (known as the imperfective paradox), and the atelicity in (31) must be attributed to a shift induced by the episodic Imperfect (since the adverbial is not necessary for the reading to arise), much in the same way in which the Progressive induces a shift from a telic to an atelic VP in English. The interpretation in (32) is more puzzling (although not entirely unexpected in view of the aspectual meaning of the Aorist), as it does not allow for a time frame adverbial in spite of the atomicity of the VP. A similar example where a time span adverbial applies to a telic VP is (33), which Rothstein (2004: 182) explains in terms of the linguistic context providing a temporal boundary for the habitual event.

It is conceivable that in (32), it is the Aorist that provides such a boundary. At the same time, it seems that the presence of the adverbial is also crucial for the multiple-attempt reading to arise. Rothstein (2016: 71) analyses the meaning of for α time in sentences like John ran/ate apples for an hour in terms of it applying to a predicate P to yield a set of events in P with running times specified by the adverbial, such that at all subintervals of the running times of these events, an event in the denotation of P was going on. A similar situation seems to apply to the sentences in (32), where at every relevant subinterval of the running time of e.g. the house building event, an attempt of building the houses was going on. Thus, while the predicates in (32) are homogenous at the embedded level, cf. (34), they are telic at the higher level, possibly due to the Aorist which closes off the event as a whole.

Summing up, we can conclude that Rothstein’s generalization about the relationship between BECOME events, atomicity and telicity can in general account for the behavior of the Bulgarian CoS related triplets under investigation. The finding that pfv triplet members in general head telic VPs follows from the inherently atomic nature of the BECOME event they lexicalize, and atomicity is in addition required by them in terms of the selectional restrictions they pose on their theme arguments. At the same time, we saw that the telicity of a VP headed by a pfv triplet member can be shifted due to interaction with Viewpoint aspectual operators like the Bulgarian Imperfect. On the other hand, we saw that while ipfv verbs head atelic VPs and do not pose such selectional restrictions on their arguments, they may head telic VPs, requiring the presence of time frame adverbials and atomic themes and resulting in habitual interpretations.⁴⁰

As for the s-ipfv triplet members which were argued to lexicalize an incremental BECOME event, we saw that while they seem to require atomic theme arguments just as their pfv counterparts, the telicity of the VP they head can be easily shifted due to interaction with Viewpoint aspectual operators and/or temporal modifiers.⁴¹

Taken altogether, these results suggest that telicity in Bulgarian, at least in the triplets investigated, does not hinge on the nature of the argument, but rather on the interaction with Viewpoint aspectual operators, and invites an account of this interaction in terms of aspectual shifts. Rothstein (2004) deals with shifts like the one involved in English progressive achievements in terms of shifts between aspectual classes (where the Progressive shifts the achievement to an accomplishment which then combines with the Progressive). A more adequate approach with view of the assumed morphological encoding of aspectual classes in Bulgarian would be one that locates such shifts at the level of Viewpoint aspect (AspP). A thorough account of the complex interaction between the two layers of aspect in Bulgarian must however be postponed for the future.

5 Conclusion

I argued that verbs in Bulgarian aspectual triplets formed with empty prefixes and expressing events of change can be characterized in terms of denoting Rothstein’s basic event types representing constraints on the kinds of events that can be individuated by verbal predicates. More specifically, I argued that while the ipfv members of such triplets denote dynamic events of non-change, their pfv and s-ipfv counterparts denote events of change of different types: simple events of change (pfv) and events of change with a complex structure (s-ipfv). These correspondences between aspectual type and event type were shown to hold independent of the lexical semantics of the verbs which may equally belong to the classes of incremental, scalar and motion verbs. I proposed that simple events of change are marked in the triplets by means of empty prefixes, while complex events of change are marked by a combination of empty prefix and imperfectivizing suffix. I further showed that a distinction between a predefined endpoint as a property of verbal predicates and telicity as a property of VPs is useful also in view of Bulgarian data in spite of the different nature of the relation between the verb and its theme.⁴²

Some cross-linguistic and theoretical consequences follow from this proposal. First of all, it shows that virtually the same constraints that underlie the distinction between event types in English hold for the division between aspectual types in Bulgarian CoS-triplets, and this is arguably quite different from the majority of the Slavic languages. On the other hand, it shows that in contrast to English accomplishments, the telicity of the VP headed by s-ipfv accomplishments does not depend on the type of its object but on the interaction with Viewpoint aspectual operators. This further suggests a division of labor between the members of the aspectual triplets investigated, where basic and secondary ipfv verbs denote different event types and behave in rather complementary ways in terms of argument constraints and interaction with aspectual tenses. Since this different behavior is intimately connected to the presence of definite articles and aspectual tenses that most other Slavic languages lack, it may offer an explanation of the productivity of the imperfectivization process in Bulgarian, which is unique among the Slavic languages. In addition, the account presented here provides evidence for a strategy of expressing aspectual distinctions in which aspectual properties are not directly tied to the roots of verbal predicates expressing events of change but are partly morphologically and partly lexically encoded. It also shows that a language employing this strategy may provide different conceptualizations of one and the same real world event, viewing it optionally as an unstructured activity, a resultant state, or an incremental process. At the same time, it provides cross-linguistic evidence for Rothstein’s position on the independent status of aspectual classes as basic event types, as well as for her theory of telicity in the unique context of a Slavic language that possesses a definite article.

Some of the main issues that must be left open is the question of how (a)telicity at the VP-level interacts with Viewpoint aspectual operators introduced by the aspectual tenses in Bulgarian, as well as how the telicity shifts resulting from this interaction can be represented in semantic and morpho-syntactic terms. Further open issues are related to the empirical coverage of the proposal. Do other types of triplets, e.g. those formed by means of the perfectivizing suffix -n, such as sedja–sedna–sjadam ‘sit (down)’ involving basic ipfv verbs denoting states, or čukam–čukna–čukvam ‘knock’, involving semelfactives, behave differently? Do underived pfv and s-ipfv verbs forming aspectual pairs such as the ones dealt with in the Appendix behave in the same way when it comes to selectional restrictions or the interaction with tenses? A look at the pair namerja–namiram ‘find’ suggests that this may not be the case, as both verbs can take bare plural and mass nouns as their (non incremental) theme (as in Ivan nameri/namira zlato/pari ‘Ivan found gold/money’). On the other hand, this only holds for the s-ipfv members of aspectual pairs with non-empty internal prefixes, such as podpiša–podpisvam ‘sign’, cf. Ivan *podpisa/podpisva dokumenti tsjala nošt (‘Ivan was signing documents all night long’). These and further related issues must be left for future work.

Appendix: A note on pairs

Some (degree) achievements in Bulgarian come not in triplets but in pairs (cf. Table 5), where a basic ipfv verb is lacking due to one of the following reasons:

1. the pfv verb is underived. The group of underived pfv verbs comprises about 50 verbs, many of which denote transfer of posession, such as dam ‘give’, as well as verbs like kaža ‘say’ and reša ‘decide’ whose English counterparts count as achievement verbs

2. the pfv verb is derived from a root that has lost its morphological independence or its transparency in the course of language change, as in složa ‘put’ < Proto-Slavic *lažiti (causative form of *ležati ‘lie’), stigna ‘reach’ < Proto-Slavic *tegam ‘move in some direction’, cf. Georgiev (1971–2010)

3. the pfv verb is derived from a basic ipfv verb whose lexical meaning has been modified as a result of prefixation with a non-empty internal prefix, as in namerja ‘find’ < merja ‘to measure/weigh’ or inchoatives like zaspja ‘fall asleep’ (from spja ‘sleep’), or by the suffix -n, as in sedna ‘sit down’ (from sedja ‘sit’), legna ‘lay down’ (from leža ‘lie’)

4. the pfv verb is derived by means of a secondary perfectivization process, as in pristigna ‘arrive’, which is derived from the pfv verb stigna ‘reach’

5. the pfv verb is directly derived from a degree adjective, cf. *studja–izstudja–izstudjavam ‘to cool’ (from adj. studen ‘cold’)⁴³

It can be easily shown that the pfv and s-ipfv members of these verbal pairs display the same aspectual properties as those in the triplets discussed in section 2.

I propose to represent the meaning of the pfv members of the pairs in the same way as those in triplets:⁴⁴

For s-ipfv verbs in pairs, where the pfv verb cannot be analysed as being derived from an activity denoting root, I propose a structure that is slightly different from the one suggested for s-ipfv triplet members. This structure corresponds to the one proposed in Rothstein for treating progressive achievements. According to this analysis, progressive achievements like in Mary is arriving at the station have the structure of accomplishments. However, the characteristics of the process stage are not lexically specified but must be contextually determined, in contrast to a lexical accomplishment (or activity) where the characteristics which identify its process stages are determined by the meaning of the verb (Rothstein 2004: 49):

The argument of the DO predicate in (38), α, is a free variable. Unlike the activity subevent, the culmination event is in the denotation of the lexical achievement. This raises the denotation of the VP into the structure of an accomplishment whose culmination is given by the lexical VP, cf. for instance, the structure of the derived accomplishment in Marry is arriving at the station, where the culmination is in the denotation of arrive at the station, Cul(e) is characterized as a near-instantaneous event part of e, and the argument of the culmination event is the incremental theme of the matrix verb:

In contrast, the culmination of an English lexical accomplishment is determined by what happens to its theme. Thus, Mary built a house culminates at the point at which a house becomes built. Note that progressive achievements result from a shift operation triggered when the progressive is applied to an achievement-headed VP. This shift converts the achievement into an accomplishment whose culmination is in the denotation of the achievement (Rothstein 2004).⁴⁵ A similar structure can be assumed for an s-ipfv verb like namiram ‘find’ or pristigam ‘arrive’, since the activity subevent involved in the complex event they denote is not lexically specified. At the same time, Rothstein (2004: 55) reasons that a progressive achievement is used “only to make an assertion about an event stage which is relatively close to the telic point”. Whether or not similar considerations apply for the s-ipfv verbs, to which I attribute the same semantics, is a question that must be addressed on a different occasion.

In terms of selectional restrictions and their relation to telicity, achievement pairs seem to behave more ununiformly as compared to the triplets discussed in section 4. Thus while pairs like pristigna–pristigam ‘arrive’ seem to require a subject that has a lexically realized determiner, pairs like namerja–namiram ‘find’ seem to allow for bare mass nouns. At the same time, both pfv and s-ipfv members of pairs like namerja–namiram ‘find’ seem to be compatible with time span adverbials and to head VPs that are telic at the level of the singular events but atelic at the level of the plural event. A detailed exploration of the properties of such pairs is however a topic for future work.

Abbreviations

1 = first person, def = definite, indf = indefinite, ipf = Imperfect, ipfv = imperfective, aor = Aorist, pfv = perfective, pl = plural, prs = present tense, ptcp = participle, refl = reflexive, sg = singular, s-ipfv = secondary imperfective

Acknowledgements

The author would like to thank the organizers and the audience of the DGfS 2022 workshop Change of State Verbs – Empirical and Theoretical Perspectives, Torgrim Solstad, as well as three anonymous reviewers for their valuable suggestions and comments.

Competing interests

The author has no competing interests to declare.