Simplex Perfectives in Russian Verb Formation

Abstract

This paper investigates the role of simplex perfectives in the Russian aspectual system, which are known to display a number of characteristics that seem to escape a proper theoretical treatment. It is proposed that simplex perfective roots (like reš- or bros-) share with internally prefixed base predicates (like napis- or pročit-) a maximal path in their event descriptions. The two classes of predicates differ from each other, however, in that only the latter require their events to realise the path up to its limit. The underspecification of so-called simplex perfectives with respect to event maximality is resolved by the choice of the different theme vowels -a or -i. A theoretical model is developed that derives the actual verb forms in accordance with their aspectual values. It implements two different morphological cycles, with theme vowel insertion demarkating the end of the first one. Early (internal) and late (external) prefixation are defined relative to this.

1. Introduction

Slavic languages use word-forming devices to encode perfective and imperfective meanings. The present paper is concerned with Russian, the coding system of which is typically summarised as in the following quote.

Almost every verb can be classified as perfective or imperfective, with only a limited number of indeterminacies. The distinction of aspect is more a partition of the lexicon than an inflectional operation. There is no single morphological device that marks the opposition of aspect; rather, aspect is expressed by a combination of strategies. Verbs without prefixes (simplex verbs) are, as a rule, imperfective [...]. Verbs with prefixes as a rule are perfective [...] – except when an additional derivational suffix makes them imperfective [...].

(Timberlake, 2004, pp. 93–94)

The marking of verbal aspect as it is described here is illustrated in

Table 1. Aspect coding.

IPFV	PFV	IPFV
pisa-t’	napisa-t’		‘write (down)’
	zapisa-t’	zapisyva-t’	‘record’
čita-t’	pročita-t’	pročityva-t’	‘read (through)’

. From left to right, we find imperfective simplex verbs, prefixed perfective verbs, and imperfective verbs derived from the latter by means of an additional suffix.

This simple picture is an idealisation, as indicated by Timberlake. It is only ‘as a rule’ that simplex verbs are imperfective, and that prefixed verbs are perfective. Relevant exceptions are the simplex perfectives in (1a) and the prefixed imperfectives in (1b):

(1) 

a. rešit’ (‘solve, decide’), brosit’ (‘throw’), pustit’ (‘let go’), končit’ (‘end’).

b. vygljadet’ (‘look like’), soderžat’ (‘contain’), dezinformirovat’ (‘disinform’), zaviset’ (‘depend on’).

While the group of verbs in (1a) will play the main role in the present paper, verbs like those in (1b) will be ignored here (for some discussion, see A. A. Zaliznjak & Šmelev, 1997, p. 68).

The present paper asks whether we can do better than treat simplex perfectives and their derivatives as exceptional cases that need to be listed independently of the rules of aspect marking. My answer will be, not surprisingly perhaps, affirmative. I do not claim to answer all of the difficult questions raised by simplex perfectives that need to be addressed (for some recent discussion, see Tatevosov). I hope, however, that the proposal that I will make in order to assign these verbs their place in the grammatical system can contribute to our understanding of the Russian, and in the longer run, Slavic aspectual system.1

This paper consists of two parts. In the first half, I will explore the empirical data, and in the second half, I will develop an account for the data. Specifically, the paper is structured as follows. In Section 2, I point to different phenomena that are in tension with the basic rules of aspectual coding as they are stated in Table 1, focusing on the case of intermediate prefixation. This prepares for Section 3, where I present two observations made in Tatevosov (2013b) with respect to verb forms derived from the simplex perfective root reš- (‘solve, decide’). The first is that with this verb, intermediate prefixes may apply to the imperfective base reša-, giving rise to perfectives like dorešat’ (‘finish solving’). This behavior is unexpected given that intermediate prefixes usually apply before secondary imperfectivisation takes place. The second observation is that the aspectual partners of prefixed verbs based on the root reš- show different imperfective suffixes dependent on whether the prefix is internal or intermediate. These facts lead Tatevosov to conclude that the suffix -a is hierarchically closer to the root than the suffix -yva/-iva, Section 4 recapitulates the argument. In Section 5, we will go on to ask whether there are other verbs within the list of simplex perfectives that behave like reš- does. The answer will be negative, but in our search we will come across the case of bros- (‘throw’), which further complicates the picture. This root also systematically projects perfectives in -a, but this time, in contrast to the case of the perfective dorešat’, the respective perfectives such as vybrosat’ (‘throw out’) are not formed by means of intermediate prefixes. In Section 6, I propose to analyse the perfective derivatives of brosa- as event-internal pluractionals in the sense of Henderson (2017). Starting with Section 7, I will develop a theoretical account to explain the data patterns collected in the preceding sections. As the first step to this end, I propose to sort Russian base predicates (verbal roots with or without internal prefixes) into different semantic classes based on two properties, namely, whether or not the event described by the predicate follows a maximal path (path maximality), and whether or not the event is claimed to reach a maximum (event maximality). I will argue that there is one class of base predicates (namely, those corresponding to simplex perfectives) that have maximal paths, but that are underspecified with respect to event maximality. Section 8 is included to ground this proposal in the theoretical literature. In Section 9, I will make a further claim which relates to the choice of theme vowels. I will argue that the choice between vowel -a or vowel -i resolves the indeterminacy of reš-type predicates. Finally, in Section 10 and Section 11, I will show how these theoretical assumptions enable us to derive the correct Russian verb forms together with their correct aspectual values. Section 12 concludes the paper.

2. Intermediate Prefixation

Apart from the above-mentioned two exceptional cases, the simplistic picture of Table 1 is further complicated by at least the following four phenomena: (i) secondary perfectivisation, (ii) perfectivising suffixation by -n, (iii) generic suffixation, (iv) different kinds of prefixation, and (v) variation in imperfectivising suffixes. In what follows, I will not go into details of (i) to (iii), as these points do not directly bear on the topic of this paper. I will only briefly explain what these headlines stand for. After that, I will focus on (iv), as this will be our point of departure. Crucial to our concerns here is also (v), which will be taken up later in Section 3.

Secondary perfectivisation (i) refers to the possibility of perfectivising a stem that already underwent secondary imperfectivisation. Consider (2).

(2) 

a. pisa-${\mathrm{t}’}^{ipf}$→ popisa-${\mathrm{t}’}^{pf}$ (‘write a bit’)

b. tkry-${\mathrm{t}’}^{pf}$→ otkryva-${\mathrm{t}’}^{ipf}$→ pootkryva-${\mathrm{t}’}^{pf}$ (‘spent some time trying to open the door’)

In (2a), po- applies to a simplex verb (“primary perfectivisation”, if one would want to call it like that). In (2b), by contrast, po- applies to an imperfective base that itself results from the suffixation of an originally perfective base. Thus, besides secondary imperfectivisation (what Timberlake describes as the effect of an ‘additional derivational suffix’), the coding system also foresees, albeit in limited contexts, the possibility of secondary perfectivization by means of an additional derivational prefix (e.g., Mehlig, 2012).

Turning to (ii), perfective verbs may also be derived by means of suffixation, specifically by means of the suffix -n followed by the theme vowel -u.2 Such perfectives are known as ‘semelfactives’ which denote ‘instantaneous, single-state events consisting of a single point, with no associated change of state’ (Smith, 1997, p. 246; see also Mehlig, 1994). While the imperfective verbs in (3) mean ‘knock’ and ‘cough’, their perfective correlates mean ‘knock once’ and ‘cough once’. Apart from a remark that I will drop at the very end of this paper, semelfactives will be excluded from discussion altogether.

(3) 

a. ${\mathrm{stučat}’}^{ipf}$→ ${\mathrm{stuknut}’}^{pf}$ (‘knock’)

b. ${\mathrm{kašljat}’}^{ipf}$→ ${\mathrm{kašljanut}’}^{ipf}$ (‘cough’)

Generic suffixation (iii) conflicts with the simple picture of Table 1 in that the suffix, which is absolutely identical in form with the imperfectivising suffixes appearing in Table 1, may in some cases apply to a base that by itself would form an imperfective verb. So, in this case, the suffix is obviously not used for coding imperfectivity, but rather for coding genericity.

(4) 

a. ${\mathrm{pisa-t}’}^{ipf}$→ ${\mathrm{pisyva-t}’}^{ipf}$ (‘tend to write’)

b. ${\mathrm{side-t}’}^{ipf}$→ ${\mathrm{siživa-t}’}^{ipf}$ (‘tend to sit’)

The use of generic suffixes is often said to be an outdated practice (Švedova et al., 1980, but see Tatevosov, 2013b). Following Filip and Carlson (1997) and Filip (n.d.) on Czech, Biskup (2024) argues that the Russian generic -yva is an independent suffix in its own right, despite its homomorphism to the imperfective -yva.

The idealised picture of aspectual coding as it is represented in Table 1 furthermore traces over important differences within the practice of prefixation—that is to say, it does not reflect the fact that different kinds of prefixation are involved in Russian verb formation (iv). As has been well established, a distinction should be drawn between internal prefixes and external prefixes; see Gehrke (2008), Ramchand (2004, 2008), Romanova (2004), and Svenonius (2004), among others. While the former participate in the creation of new verb lexemes, often giving rise to idiosyncratic meanings, the latter introduce transparent meaning components that modify an already existing lexical unit (possibly containing an internal prefix) in a semantically predictable way. The prefixes in Table 1 are all internal ones; an external prefix can be seen in (2a) (‘delimitative po-’).

Tatevosov (2007, 2011) has drawn attention to a special subclass of external prefixes in Russian for which he has coined the term intermediate prefixes. Such prefix uses contribute fully transparent meanings to the verb forms that they derive. In this respect, they behave like external prefixes do, and typically unlike internal prefixes. At the same time, they can be shown to attach lower in syntax than where external prefixes are usually expected to attach, hence the term intermediate prefixation.

Specifically, Tatevosov (2011, 2013b) identifies three formatives which are usable as intermediate prefixes, namely, the element do- when used in the ‘completive’ sense, pod- when used in the ‘attentuative’ sense, and pere- when used in the ‘repetitive’ sense. All three prefixation strategies lead to semantically transparent modifications. The completive do- derives verbs that denote a final stage of the event denoted by the base verb to which the prefix attaches. The attenuative pod- serves to derive verbs referring to a somewhat toned-down instance of the event described by the base verb. The repetitive pere-, finally, brings about verbs that denote a repetition of the event denoted by the base verb. (5) shows examples for illustration.

(5) 

a. igrat’ (‘play’) → doigrat’ (‘finish playing’)

b. solit’ (‘salt’) → podsolit’ (‘salt slightly’)

c. pečatat’ (‘print’) → perepečatat’ (‘print again’)

Within the derivational chain along which Russian verbs are built up, intermediate prefixes come into play prior to the stage at which secondary imperfectivisation is assumed to take place (Tatevosov, 2007, 2011).3 At the same time, it can be observed that intermediate prefixes apply on top of ‘ordinary prefixes’ (i.e., internal prefixes). Consider

Table 2. Intermediate prefixation.

IPFV	PFV	PFV	IPFV
taja-t’	rastaja-t’	podrastaja-t’	podrastaiva-t’	’melt slightly’
taja-t’		podtaja-t’	podtaiva-t’	’melt slightly’
pisa-t’	napisa-t’	donapisa-t’	donapisyva-t’	’finish writing’
pisa-t’		dopisa-t’	dopisyva-t’	’finish writing’
čita-t’	pročita-t’	perepročita-t’	perepročityva-t’	’re-read’
čita-t’		perečita-t’	perečityva-t’	’re-read’

for illustration. As the table also shows, intermediate prefixes combine with imperfective or perfective bases to output new perfective forms. The meanings of the simplex imperfectives in Table 2 are ‘melt’, ‘write’, and ‘read’, respectively.

Especially noteworthy are the perfective verbs podrastajat’, donapisat’, and perepročitat’ in Table 2. They show that intermediate prefixes may indeed apply on top of already internally prefixed perfectives, even if this is rare. Since there is not a single attestation of donapisat’ in the Russian National Corpus, I should perhaps provide an example. The following one can be found on the website https://mnogobukv.hse.ru/news/319964105.html (accessed on 16 December 2024). (“NLO” is for the journal “Novoe literaturnoe obozrenie” and “Venisan” is the name of a fictional city). The example is taken from an interview with Linor Goralik, a quite famous Russian-speaking author.

(6) 

Do ėtogo ja dolžna  donapisat’     stat’ju v NLO, potom – napisat’

until this I   must    finish-writing.inf.pfv article in NLO then  write.inf.pfv

vtoruju knižku pro Venisanny (ja planiruju  zakončit’  ee 7 janvarja).

second book  about Venisan  I  plan.prs.ipfv finish.inf.pfv it 7 january

‘Until then I have to finish writing an article for NLO, and write the second book

about Venisan (I plan to have it finished at the 7th of January).’

3. The Case of reš-

We now move on to (v), variation in the inventory of imperfectivising suffixes. It is often said that the Russian imperfectivising morpheme comes in different allomorphs (A. A. Zaliznjak & Šmelev, 1997; Švedova et al., 1980). According to the Academy Grammar (Švedova et al., 1980, p. 588), the imperfectivising suffix realises in one out of three ways, as -iva, as -a, or as -va (with -yva and -ja as orthografic variants of the former two morphs). (7) shows examples.

(7) 

a. vyigrat’ → vyigryvat’ (‘win’)

b. ugovorit’ → ugovarivat’ (‘persuade’)

c. rešit’ → rešat’ (‘solve’)

d. otdelit’ → otdeljat’ (‘separate’)

e. otkryt’ → otkryvat’ (‘open’)

However, it can be demonstrated that the imperfective suffixes in (7) do not form a natural class.4 One case revealing that the apparent allomorphs show no uniform distribution is the combinatorics with intermediate prefixes. Recall from above that intermediate prefixes are expected to merge below the imperfectivising suffix (Table 2). As Tatevosov (2013b) shows, however, sometimes an intermediate prefix applies to a base that already underwent imperfectivisation, thus newly producing a perfective. Importantly now, whenever this is possible, the respective suffix will be -a. A relevant (constructed) example is the following.

(8) 

Anton dorešal      uravnenie i  pošel  domoj.

Anton finish_solve.pst.pfv  equation  and  go.pst.pfv home

‘Anton finished solving the equation and went home.’

So it turns out that the rule according to which secondary imperfectivisation bleeds intermediate prefixation does not hold for the case in which imperfectivisation manifests itself in -a.5 This suggests that the element -a is, in fact, not an allomorph on a par with -va and -yva, but rather an independent formative whose status is in need of clarification (Tatevosov, 2013b, p. 73).6

The working hypothesis of the present paper is that in order to arrive at a better understanding of the suffix -a, we should conduct a closer inspection of the class of simplex perfectives and their morphological derivatives. So, let us do this, starting with reš- (‘solve, decide’), the root that Tatevosov (2013b) discusses at some length. Recall from Table 2 that intermediate prefixation systematically derives two alternative perfectives on the basis of aspectual pairs like pisat’ ∼napisat’.7

(9) 

a. ${\mathrm{napisat}’}^{pf}$ ‘write’ → ${\mathrm{donapisat}’}^{pf}$ ’finish writing down’

b. ${\mathrm{pisat}’}^{ipf}$ ‘write’ → ${\mathrm{dopisat}’}^{pf}$ ’finish writing’

The pattern in (9) repeats itself for pairs like rešat’ ∼ rešit’. I will also discuss the pair uničtožat’ ∼ uničtožit’ (‘destroy’), even though the perfective form of this verb is strictly speaking no simplex perfective. This verb is important if only because it plays such a prominent role in Tatevosov (2013b). Consider (10) and (11), where one perfective is derived from a perfective base (reši or uničtoži), while the other one is derived from an imperfective base (reša or uničtoža). According to Tatevosov (2013b, p. 67), the resulting two perfectives are ‘semantically indifferent.

(10) 

a. ${\mathrm{rešit}’}^{pf}$ ‘solve, decide’ → ${\mathrm{dorešit}’}^{pf}$ ’finish solving’

b. ${\mathrm{rešat}’}^{ipf}$ ‘solve, decide’ → ${\mathrm{dorešat}’}^{pf}$ ’finish solving’

(11) 

a. ${\mathrm{uničtožit}’}^{pf}$ ‘destroy’ → ${\mathrm{douničtožit}’}^{pf}$ ’finish destroying’

b. ${\mathrm{uničtožat}’}^{ipf}$ ‘destroy’ → ${\mathrm{douničtožat}’}^{pf}$ ’finish destroying’

Now, consider the following observation made in Tatevosov (2013b, p. 68). If the root reš- is prefixed by an intermediate prefix, as in (12), the corresponding imperfective stem will end in -iva.8 The situation is different if the root takes an internal prefix, as in (13). Then, the imperfective can only be formed by means of -a. Note also that it is the prefixed perfective in (10a), not the one in (10b), from which an imperfective partner is derived.

(12) 

a. ${\mathrm{dorešit}’}^{pf}$∼ ${\mathrm{dorešivat}’}^{ipf}$ (‘finish solving/deciding’)

b. ${\mathrm{pererešit}’}^{pf}\sim$ ${\mathrm{pererešivat}’}^{ipf}$ (‘decide again’)

(13) 

a. ${\mathrm{razrešit}’}^{pf}$∼ ${\mathrm{razrešat}’}^{ipf}$ (‘allow’)

b. ${\mathrm{otrešit}’}^{pf}\sim$ ${\mathrm{otrešat}’}^{ipf}$ (‘dismiss’)

The imperfective counterpart to douničtožit’ in (11a), which contains an intermediate prefix, is douničtožat’. Unlike dorešat’, which is perfective only, douničtožat’ is biaspectual (Tatevosov, 2013b, p. 67). There seems to be no way to combine uničtož- with an internal prefix, presumably because u- already occupies the internal prefix position. More on this case will be said below.

Intuitively, it is as if the availability of the chain in (10b), which licenses perfective predicates like dorešal in (8), blocks the a-suffixed form from taking on the function of expressing the imperfective aspect, with the effect of enforcing a longer form in -yva to serve this task instead. We will pick up this idea again below.

Above, we saw prefixed verbs based on the root reš- that were formed by means of the four prefixes do-, pere-, raz-, and ot-. Three more cases can be found in A. Zaliznjak (1977). All of them turn out to have internal prefixes, and all form the imperfective by means of -a.

(14) 

a. ${\mathrm{vyrešit}’}^{pf}\sim$ ${\mathrm{vyrešat}’}^{ipf}$ (‘make the final decision’)

b. ${\mathrm{porešit}’}^{pf}\sim$ ${\mathrm{porešat}’}^{ipf}$ (‘decide’)

c. ${\mathrm{predrešit}’}^{pf}\sim$ ${\mathrm{predrešat}’}^{ipf}$ (‘decide in advance’)

The general moral that we can draw by now is that prefixed verbs of one and the same root do not have to adhere to one and the same imperfectivising morphology. The case of reš- shows that, depending on whether the prefix is internal or intermediate, the imperfective marker will be either -a or -yva. Starting from this observation, we should ask whether there are further simplex perfectives that distribute the imperfective suffixes -a and -iva over the contexts of internal and intermediate prefixes.

4. Suffix -a Is Lower in Structure Than -yva

Cross-classification of two kinds of prefixations with two different suffixes yields four logically possible configurations. In the present section, we will look for whether each of these four options is attested by at least one simplex perfective instantiating it. This recapitulates the argument that Tatevosov (2013b) provides to show that the suffix -a is hierarchically closer to the root than the suffix -yva/-iva.

There are perfective roots, to begin with, where the internally prefixed perfective (15b) and the intermediately prefixed perfective (15c), unlike the non-prefixed base verb (15a), both have imperfective partner verbs in -yva/-iva.

(15) 

a. ${\mathrm{končit}’}^{pf}\sim$ ${\mathrm{končat}’}^{ipf}$ (‘end sth.’)

b. ${\mathrm{prikončit}’}^{pf}\sim$ ${\mathrm{prikančivat}’}^{ipf}$ (‘kill ultimately’)

c. ${\mathrm{dokončit}’}^{pf}\sim$ ${\mathrm{dokančivat}’}^{ipf}$ (‘finish ending’)

It is also possible to provide examples of intermediately prefixed and internally prefixed imperfectives both choosing -a, i.e., both choosing the imperfective marker that already the non-prefixed base verb uses.

(16) 

a. ${\mathrm{pustit}’}^{pf}\sim$ ${\mathrm{puskat}’}^{ipf}$ (‘let go’)

b. ${\mathrm{vypustit}’}^{pf}\sim$ ${\mathrm{vypuskat}’}^{ipf}$ (‘set free’)

c. ${\mathrm{dopustit}’}^{pf}\sim$ ${\mathrm{dopuskat}’}^{ipf}$ (‘let go toward’)

One might want to object that the do-prefixed item (16c) involves no intermediate prefix because it does not denote the final stage of the event of letting someone or something go, but rather a letting go in the direction of some implied target.9 As the following contextualisation shows, however, (16c) does allow for a final-stage-reading, and if it does, as can be seen, the imperfective remains ending in -a.10

(17) 

Mne kažetsja, on prosto  ne  dopuskal    do konca mysli,  čto  ona

me   seems   he simply not allow.pst.ipfv until end  thought that she

vzapravdu možet vot tak umeret’.

really    can   prt so  die.inf.pfv

‘It seems to me that he simply did not allow completely the thought that she might in fact die right away.’

The following could be another example illustrating the same pattern.

(18) 

a. ${\mathrm{oščutit}’}^{pf}\sim$ ${\mathrm{oščuščat}’}^{ipf}$ (‘sense, notice’)

b. ${\mathrm{dooščutit}’}^{pf}\sim$ ${\mathrm{dooščuščat}’}^{ipf}$ (‘infer/complete information by sensing’)

c. ${\mathrm{pereoščutit}’}^{pf}\sim$ ${\mathrm{pereoščuščat}’}^{ipf}$ (‘sense/notice sth. again’)

The lexeme (18b) is exemplified by (19) taken from Nikitivič (2010), where the prefix do- in dooščutit’ is arguably no intermediate prefix because the resulting predicate does not mean ‘finish sensing’ or ‘finish noticing’. If this is correct, we will have to count it as an internal prefix.

(19) 

To est’  zritel’   dolžen dodumat’,       dooščutit’

that exists viewer must  add-by-thinking.inf.pfv add-by-sensing.inf.pfv

dal’nejšie sobytija?

further  events

‘Does this mean that the viewer has to anticipate upcoming events by thinking and feeling?’

It is possible without difficulty to transpose the infinitives in the scope of the modal into the imperfective aspect. If we do that, as in (20), we can see that the imperfective partner of internally prefixed dooščutit’ is dooščuščat’, as stated in (18b).

(20) 

To est’  zritel’  dolžen dodumyvat’,       dooščuščat’

that exists viewer must    add-by-thinking.inf.ipfv add-by-sensing.inf.ipfv

dal’nejšie sobytija?

further  events

‘Does this mean that the viewer has to anticipate upcoming events by thinking and feeling?’

The lexeme (18c) is exemplified by (21), again from Nikitivič (2010). Given the meaning of the verb pereoščuščat’, the prefix pere- is an instance of ‘repetitive pere-’ in (21)—i.e., it is an intermediate prefix in the sense of Tatevosov (2011). As can be seen, the stem uses -a for marking the imperfective. In sum, we may conclude that (18) again shows a symmetric strategy by which I mean that both internally and intermediately prefixed verbs go for one and the same coding option. While -yva was used throughout the prefixed verbs in (15), the common marker is -a in (18), as it was in (16).

(21) 

Budu pereoščuščat’   oščuščennoe, ibo   narušilas’  kartina mira.

will re-sense.inf.ipfv feeling   because break.pst.pfv picture world

‘I will feel what I felt again because the world-view has collapsed.’

Besides these two symmetric coding strategies, there are, as we already know from Section 2, also cases of asymmetric coding, namely, the combination of intermediately prefixed imperfectives in -yva/-iva and internally prefixed imperfectives in -a (Tatevosov, 2013b; see above).

(22) 

a. ${\mathrm{rešit}’}^{pf}$∼ ${\mathrm{rešat}’}^{ipf}$ (‘solve/decide’)

b. ${\mathrm{razrešit}’}^{pf}$∼ ${\mathrm{razrešat}’}^{ipf}$ (‘allow’)

c. ${\mathrm{dorešit}’}^{pf}$∼ ${\mathrm{dorešivat}’}^{ipf}$ (‘finish solving/deciding’)

Crucially, however, there seems to be no root based on which intermediately prefixed perfectives were imperfectivised in -a, while internally prefixed perfectives were imperfectivised in -yva. In view of this gap, Tatevosov (2013b, p. 68) draws the following generalisation:11

(23) 

Descriptive Generalisation. There are no contexts in which a perfective stem derived from a verbal root by means of an internal prefix would choose the imperfective marker -yva, while a perfective derived from the same root by means of an intermediate (or external) prefix would choose the marker -a.

The available coding strategies are summarised in

Table 3. Prefix-dependent coding strategies: -a or -yva.

	Internal Prefix	Intermediate Prefix	Attested?
I	-a	-a	Yes
II	-yva	-yva	Yes
III	-a	-yva	Yes
IV	-yva	-a	No

. Since internal prefixes by definition merge lower (closer to the root) than intermediate (external) prefixes, Tatevosov (2013b) understandably concludes that the formative -a likewise resides closer to the root than the formative -yva.

5. Are There More Examples Like reš-?

Above, we recapitulated the analysis of Tatevosov (2013b) that there can be imperfective suffixes (-a or -yva) that are different from each other and dependent on what sort of prefix a root carries (internal or intermediate). Tatevosov presents exactly one example showing this, namely, reš-. In the present section, we ask the following: Is it possible to come up with further instances of strategy III, i.e., with more examples like reš- in which -a is used for internally prefixed imperfectives, while -yva is used for intermediately prefixed imperfectives? To answer this question, let us have a more systematic look at the derivatives of simplex perfectives.

Švedova et al. (1980, p. 590) list 18 non-prefixed perfectives altogether. Four have imperfective counterparts in -yva or -va. Two out of these four are biaspectual with additional imperfective forms in -yva, namely, biaspectual atakovat’ (‘attack’, with imperfective atakovyvat’) and biaspectual arendovat’ (‘rent’, with imperfective arendovyvat’). The two others are perfective dat’ with imperfective davat’ (‘give’), and perfective arestovat’ with imperfective arestovyvat’ (‘arest’).12 Here, we are interested in the remaining 14 simplex perfectives, which all have imperfective partner verbs marked in -a, among them rešit’ (see

Table 4. Simplex perfectives with imperfective partners in -a (from Švedova et al., 1980).

PFV	IPFV		PFV	IPFV
reši-t’	reša-t’	‘decide,solve’	vnuši-t’	vnuša-t’	‘meddle’
konči-t’	konča-t’	‘end’	zaščiti-t’	zaščišča-t’	‘defend’
liši-t’	liša-t’	‘deprivate’	izumi-t’	izumlja-t’	‘perplex’
prosti-t’	prošča-t’	‘forgive’	oščuti-t’	oščušča-t’	‘sense,notice’
javi-t’	javlja-t’	‘display’	voskresi-t’	voskreša-t’	‘resurrect’
pusti-t’	puska-t’	‘let go’	oskorbi-t’	oskorblja-t’	‘insult’
brosi-t’	brosa-t’	‘throw’	vstreti-t’	vstreča-t’	‘meet’

).

The verbs in the right half of Table 4 contain in their morphological structure former prefixes which, according to Švedova et al. (1980, p. 590), are no longer recognised as such due to lexicalisation. For these verbs, it is difficult to find derivatives formed by means of a (second) internal prefix. The only candidate verb in A. Zaliznjak (1977) is the archaic form predoščutit’ (‘sense/notice in advance’), which has predoščuščat’ as its imperfective counterpart. The items that I provided above in (18) are not listed in standard dictionaries, but they are, like many more, attestable in the Russian language as it is used on the internet, albeit at a comparatively low frequency (Nikitivič, 2010).

Within the verbs on the left-hand side of Table 4, rešit’ turns out to be outstanding in showing asymmetric coding of imperfectivity (strategy III). We saw in (15) that končit’ follows strategy II, using -yva for both internally and intermediately prefixed imperfectives. The verb lišit’ derives no prefixed forms at all. As shown in (16), pustit’ goes for the symmetric coding strategy I using -a. The same could be noted for javit’ (‘display’), from which the prefixed verb ob”javit’ (‘announce’) can be derived, with ob”javljat’ as imperfective partner. I was not able to find attestation of the use of the repetitive pere- in order to create perejavit’ (‘display again’). If that was possible, the correct form would probably be perejavljat’, stated in (24c). Support for such a conclusion comes from the fact that this verb is listed in Dal’s 19th century dictionary in the meaning ‘display again’ (javit’ snova)—see Dal’ (1978–1980).

(24) 

a. javit’ ∼ javljat’ (‘display’)

b. ob”javit’ ∼ ob”javlajt’ (‘announce’)

c. ?perejavit’ ∼ ?perejavljat’ (‘display again’)

The verb prostit’ likewise follows strategy I. Here, again, internally prefixed lexemes form imperfectives in -a, while intermediately prefixed lexemes would also use -a.

(25) 

a. prostit’ ∼ proščat’ (‘forgive’)

b. uprostit’ ∼ uproščat’ (‘simplify’)

c. ?pereprostit’ ∼ ?pereproščat’ (‘forgive again’)

The last example listed in Table 4 is the perfective verb brosit’ (‘throw’) with its imperfective partner brosat’. This verb deserves special attention. If a perfective is derived from it by means of an internal prefix, the marker -yva will be chosen for suffixing the imperfective counterpart, e.g., vybrosit’ ∼ vybrasyvat’ (‘throw out’). The alternative imperfective form that might have been expected, vybrosat’, does not exist. Interestingly, however, there is a perfective verb vybrosat’, meaning something like ‘throw out one after the other until every object from a limited set of objects available has been thrown out’. A. Zaliznjak (1977) in fact includes 15 prefixed perfectives in -a derived from brosit’. Apart from vybrosat’, these are the ones in

Table 5. Prefixed perfectives in -a based on bros- (from A. Zaliznjak, 1977).

verb.PFV	Translation
vbrosat’	‘throw in one by one’
zabrosat’	‘throw at one by one’
otbrosat’	‘meet a target by throwing one by one’
pobrosat’	‘let fall one by one’
ponabrosat’	‘conciously let fall one by one’
pribrosat’	‘drop one by one’
razbrosat’	‘throw in different directions one by one’
dobrosat’	‘add to by throwing one by one’
nabrosat’	‘accumulate by throwing one by one’
perebrosat’	‘throw over one by one’
podbrosat’	‘throw onto one by one’
porazbrosat’	‘throw in different directions one by one’
probrosat’	‘throw through one by one’
sbrosat’	‘throw from one by one’

. It should be noted that some of them are barely accepted by native speakers today (for the sake of exhaustiveness, I will also list these).

In this section, to sum up so far, we have first checked for whether other simplex perfectives besides rešit’ instantiate the asymmetric coding strategy III from Table 3. We couldn’t find none. The second example to generate forms instantiating strategy III provided in Tatevosov (2013b) is uničtožit’, whose status as a simplex perfective is debatable (more on that below). On our search, we then stumbled across the case of brosit’, which resembles rešit’ and uničtožit’ in so far as it is capable of deriving perfectives in -a (Table 5). The following section is dedicated to these special kinds of morphologically complex perfectives.

6. Event-Internal Pluractionals

The meanings of the perfectives in Table 5 suggest that all of these verbs entail the existence of a series of events, with each event of the series falling under the non-prefixed base predicate bros- and with all events being lined up on a common upper-bounded path introduced by the respective prefix. In the case of vybrosat’, for instance, the common path is a scale that measures the amount of thrown objects located in the out-area relative to the coordinates of the thrower, their number increasing with every throwing event. Here is a nice example from Valentin Rasputins novel Izba (which was first published in 1999).

(26) 

Letom  chodili  za jagodoj i    prodavali, zimoj  iskali       melkich

summer go.pst.ipfv for berries and  sell.pst.ipfv winter search.pst.ipfv soft

poručenij: vody s  berega na čaj prinesti    [...], vybrosat’      iz stajki

todos  water from bank on tea bring.inf.pfv  out-throw.pst.pfv  out barn

iz-pod   korovy  ševjaki, otgresti   sneg.

from-under  cows   shit  clear.inf.pfv   snow

‘In the summer they went for berries to sell them, in the winter they were looking for small errands: to bring water from the shore for tea, to throw out of the barn the pats from under the cow, to shovel the snow..’

The path along which the throwing events are ordered can be depicted as in (27): n is the maximal number of cowpats in the barn, matching the number of events of throwing a cowpat out. The task of cleaning the barn will be completed as soon as n cowpats are thrown out.13

(27) 

vybrosat’ iz stajki iz-pod korovy ševjaki

The function of the path (or scale, see Kagan, 2015) which is introduced by the prefix is to integrate the plurality of events into a single whole. Perfectives like those in Table 5 can be viewed as instances of event-internal pluractionality in the sense of Henderson (2017). According to Henderson, predicates encoding event-internal pluractionality denote atomic entities (‘single wholes’) associated with a plurality of events by means of a trace function. The trace can be partitioned such that each part of it is by itself the trace of an event satisfying some lexically given predicate. The partition has to have a relatively large cardinality, and the eventive parts must form a (temporal) cluster. For details, see Henderson (2017, p. 190ff.); the notion of temporal cluster has been made precise in Wągiel (2023).14 Russian verbal predicates like vybrosat’ seem to satisfy all of the relevant properties. As argued above with respect to (27), such predicates denote single events with their trace function (path) connecting a considerable number of constitutive events, each one satisfying the base predicate of the prefixed verb (here: bros- ‘throw’). In other words, they are analysable as denoting atomic events of throwing something out of something, associated with a partitioned trace, whereby each partition itself is the trace of a throwing event. In the wording of Wągiel (2023, p. 285), the events denoted by vybrosat’ are higher-order units, each representing a collection of events/acts forming a temporal cluster.

The example that we came across in the beginning of this paper, (8), repeated here for convenience, is of the same kind. Relative to that example, n would correspond to the number of calculation steps necessary to solve the equation in question.

(28) 

Anton dorešal      uravnenie i  pošel  domoj.

Anton finish_solve.pst.pfv  equation and go.pst.pfv home

‘Anton finished solving the equation and went home.’

Let us wrap up the discussion. We saw above that the ‘imperfectivising suffix’ -a has two properties that set it apart from its apparent allomorphs -yva and -va.

First, as shown in Tatevosov (2013b), a-suffixed imperfectives may undergo intermediate prefixation. Like the perfective dopisat’ (‘finish writing’) may be derived from the imperfective pisat’ (‘write’), the perfective douničtožat’ (‘finish destroying’) may be derived from the imperfective uničtožat’ (‘destroy’), and the perfective dorešat’ (‘finish solving’) may be derived from the imperfective rešat’ (‘solve’). If we would analyse uničtožat’ and rešat’ as being derived from the perfective base uničtožit’ and rešit’, respectively, as is frequently done, we would have the case of an intermediate prefix applying to a secondary imperfective. Other secondary imperfectives do not allow for that (but see footnote 15 ), which leads us to the conclusion that -a is situated lower in syntactic structure than the other imperfective suffixes.

Secondly, by inspecting the derivatives of the root bros-, we determined that the suffix -a frequently participates in the formation of what we called event-internal pluractionals: perfective predicates denoting events made up of a series of microevents (denoted by the non-prefixed base) located on a common path (denoted by the prefix). It should be noted that, in the case of bros-, the capacity to derive event-internal pluractionals is not limited to one of the classes of intermediate prefixations established by Tatevosov (2013b), while this seems to be so with reš- and uničtož-. This is an important observation that we will have to address later again.

In the next section, I want to show that we can make some sense of these at first glance confusing data if we assume with Beavers and Koontz-Garboden (2012, 2020) and Beavers et al. (2021) that roots can carry semantic entailments of change of state. This view opposes the not infrequently expressed assumption in Slavic linguistics that roots per se are acategorial, i.e., that the meaning of change, if present, is always introduced by syntactic structure above the root. Contrary to this assumption, I will propose below that the choice of Russian theme vowels, among them the theme vowel -a, is partly sensitive to semantically different classes of roots, and that the class of reš-type predicates plays a special role in this.

7. Semantic Classes of Roots

Verbal predicates are linguistic forms that, broadly speaking, provide partial descriptions of situations. Instead of ‘situation’ (e.g., Comrie 1976), one also encounters the terms ‘eventuality’ (e.g., Bach, 1986) or simply ‘event’ in the theoretical literature. The kinds of events described by verbal predicates may be sorted into different semantic classes of grammatical relevance. Respective classifications go under various names such as Vendler classes (Vendler, 1967), actionality classes (Tatevosov, 2002), situation types (Smith, 1997), eventuality types (Filip, 1999), taxonomic categories (Padučeva, 1996), event kinds (Gehrke, 2019), or lexical aspects (Dickey, 2020). At a very general level, a first such class distinction may be drawn between states and dynamic situations:

With a state, unless something happens to change that state, then the state will continue [...] With a dynamic situation, on the other hand, the situation will only continue if it is continually subject to a new input of energy. [...] To remain in a state requires no effort, whereas to remain in a dynamic situation does require effort, whether from inside or outside.

(Comrie, 1976, p. 49)

From now on, instead of speaking of ‘situations’, I will use the term ‘event’, which is familiar from event semantics (Maienborn, 2011). It seems intuitive to relate the requirement for ‘input of energy’ to motion (kinesis). A plausible way of thinking about dynamic events is accordingly to treat them as motion events, which means to identify within their description a path along which the event develops: ‘If e is a motion event, then trace(e) is the path followed by the theme of e’ (Zwarts, 2005, p. 755; see also Krifka, 1998; Gehrke, 2008). Building on this idea, I will argue that two properties are of particular relevance to the grammar of the Russian verb, both relating to the path structure of the (dynamic) event predicate built up from the respective root. The two relevant properties are whether or not the event described by the verbal predicate follows a maximal path, and whether or not the event is claimed to reach a maximum provided by its path. The function $TRACE$ exploited in the following definition is borrowed from Zwarts (2005, p. 755). It is a function that maps dynamic events (i.e., motion events) to their unique path.

(29) 

Path structure of event predicates.

Every predicate of dynamic events e is associated with a set of paths $\mathbb{Q}$ such that $\mathbb{Q}$ = {q | q ≤ TRACE(e)}.

Against the background of (29), we may define the entailment of a maximal path as follows.

(30) 

Path-maximality (PMax).

Let P be event predicates, e events, and q paths.

If $P\left(e\right)\wedge PMax\left(P\right)$, then $\exists q^{\prime }\in \mathbb{Q}:\forall q.q^{\prime }+q\to (q^{\prime }+q)\nleqq TRACE\left(e\right)$.

This says that a dynamic predicate P is path-maximal (= satisfies the predicate $PMax$, which I will abbreviate as [+pmax] in what follows) if its path structure $\mathbb{Q}$ includes a path $q^{\prime }$ that cannot be concatenated by some other path q with the resulting path $q^{\prime }+q$ still belonging to $\mathbb{Q}$.

So, the first relevant property, path-maximality, is about whether the event’s path has a maximum. The second relevant property is about whether the event will be realised up to this maximum. To define it, we use Krifka’s temporal trace function: ‘The temporal trace function ${\tau }_E$(e) maps events to their run time, the time at which an event is going on’ (Krifka, 1998, p. 206).

(31) 

Event-maximality (EMax).

Let P be event predicates, e events, and t time intervals.

If $P\left(e\right)\wedge EMax\left(P\right)$, then $\exists t^{\prime }.t^{\prime }\subseteq \tau \left(e\right):\forall t.t^{\prime }+t\to (t^{\prime }+t)⊈\tau \left(e\right)$.

According to (31), a predicate P is event-maximal (= satisfies the predicate $EMax$, which I will abbreviate as [+emax]) if the run time of an event e denoted by it includes an interval $t^{\prime }$ that cannot be prolonged by another time interval t with the resulting interval $t^{\prime }+t$ still belonging to the run time of e.

With these definitions at hand, I now propose to set up the following axioms.

(32) 

a. Roots like čit-, pis-, ... denote event properties that have paths, but no maximal path.

b. Roots like reš-, bros-, uničtož-, ... denote event properties that have a maximal path, and that leave it open as to whether the event is maximal.

c. Lexical bases resulting from a spatial (internal) prefix attaching to a root, such as pročit-, napis-, podpis-, ..., denote event properties that provide a maximal path and characterise maximal events.

d. Roots like zna-, ... denote non-dynamic properties (i.e., properties that have no path structure).

Table 6. Path properties of roots and base predicates.

Lexical Base	Dynamic	Pmax	Emax
pis-	+	–	–	’write’
napis-	+	+	+	’write down’
podpis-	+	+	+	’sign’
reš-	+	+	+/–	’solve’
razreš-	+	+	+/–	’allow’
bros-	+	+	+/–	’throw’
čit-	+	–	–	’read’
pročit-	+	+	+	’read through’
uničtož-	+	+	+/–	’destroy’
zna-	–	Not Applicable	Not Applicable	’know’

gives an overview. The left column is headed by ‘lexical base’ because what is listed there are not only bare roots, but also prefixed roots. The next column (‘dynamic’) indicates whether or not the respective base predicate involves a path structure. The third one (‘pmax’) asks for whether the path structure of the predicate contains a maximal path, defined as in (30). The column ‘emax’, finally, assigns + to predicates denoting exclusively maximal events, defined as in (31).16

Our main concern in this paper is with the class of dynamic predicates which have in their path structure a maximal path, but which do not require event maximality. Members of this class have the feature specification [+pmax,+/–emax] in Table 6. The class is represented by the three items reš-, bros-, and uničtož-.

8. Grounding the Present Approach in the Literature

This section is inserted to integrate the specific assumptions made above into the theoretical literature on aspectual semantics. It should be clear without saying that I am not the first to exploit path structures for the modeling of verbal aspect—see Krifka (1998), Gehrke (2008), Kagan (2015), and Biskup (2019), among others. Beyond this general remark, the proposed distinction between path maximality and event maximality makes the present analysis belong to the family of theories that, in one way or another, distinguishes between potential boundaries and actual boundaries in the meaning of Slavic verbs. This family turns out to be quite large as we can relate ‘potential boundaries’ to the traditional aspectological notion of predel’nost’ (≈the having of boundaries) and ‘actual boundaries’ to the traditional notion of celostnost’ (≈the being a whole entity).

The Russian Academy grammar (Švedova et al., 1980, p. 583), for instance, describes perfective verbs as denoting ‘whole events constrained by a boundary’ (ograničennoe predelom celostnoe dejstvie). It should be obvious that the constrained-by-a-boundary component can be related to path-maximality, and the whole-event component to event-maximality. In a similar way, Filip (2008, 2017) draws a distinction between predicates containing a maximality condition and predicates entailing a maximality requirement. A predicate will be perfective, on that approach, if it comes with a requirement for event maximality. What all these proposals share is that they assume a hierarchical two-level theory of the semantics of perfectivity. Before I go on to discuss Filip’s theory, let me still mention Barentsen (1995, 1998), who, with respect to Russian, even argues for a three-level semantics of perfective aspect.

According to Barentsen’s theory, Russian perfective verbs entail not only boundedness (predel’nost’) and wholeness (celostnost’), but, on top of that, they impose a third constraint on interpretation which he calls connectedness (sekventnaja svjas’). Building on that model, Stunová (1991, 1993) shows that the topmost level of connectedness is present in Russian aspect, but absent in Czech. Dickey (2000, 2015) picks up on connectedness by considering temporal definiteness to be the semantic feature which is distinctive for Eastern slavic perfectivity. In Mueller-Reichau (2018), I provide further arguments supporting the view that the presence of connectedness (or target state relevance, as I call it following Grønn, 2004; see also Mueller-Reichau, n.d.) is what makes aspect use in Russian different from aspect use in Czech.

In the present paper, I make the tacit assumption that, in order to explain the morphology of Russian verbs, it suffices to consider the first two conditions that a perfective predicate has to satisfy, predel’nost’ and celostnost’, which I operationalise with reference to a predicate’s path structure as [+pmax] and [+emax], respectively.

Filip (2000) proposes that Slavic verbal prefixes are word formation devices that furnish the lexical meaning of a predicate with a maximality condition. To adopt a perfective meaning, the predicate should on top of that become associated with a maximality requirement. At first glance, one might believe that the requirement for maximal events is logically stronger than the condition describing a maximal event. One might think, in other words, that providing the description of a maximal event would be a necessary prerequisite for the denotation of maximal events. This is in fact the reasoning in-built in the hierarchical semantics of Russian perfectivity advocated by Barentsen (1995): every predicate satisfying celostnost’ must at the same time satisfy predel’nost’. As has been shown in Filip (2000), however, this is not true. The presence of a maximality condition (an upper bound in the event description) is not mandatory for denoting maximal events (whole events). There may be, in my terms, [–pmax,+emax] predicates. Such predicates are famously known as delimitatives, formed in Russian by means of the prefix po-.

The independence of the two notions that I call here path-maximality and event-maximality has been established in the literature as the distinction between telicity on the one hand and perfectivity on the other hand. It is precisely observations relating to perfectives formed by the delimitative po- (and perdurative pro-) that lead Borik (2002, p. 70) to the conclusion that telicity (≈ path-maximality) is neither a sufficient nor a necessary condition for perfectivity (≈ path-maximality). The same observation is made in Schoorlemmer (1995) and Gehrke (2008). Note that delimitative predicates are not represented in Table 6.

Tatevosov (n.d.) demonstrates that internally prefixed predicates are of higher event-structural complexity than simple predicates, which he relates to the presence of a result state argument in their semantics. In terms of the present approach, the higher complexity of the denoted events may be related to the fact that internal prefixes furnish the predicate with a maximal path. This converges with Tatevosov’s analysis in so far as a path with a predetermined maximum implies a result state.

The distinction between path maximality and event maximality that I propose in this paper is even more directly relatable to Filip’s view noted above. The presence of a maximal path in the path structure of a predicate (30) corresponds to the presence of a maximality condition, while the limitation of the predicate’s denotation to maximal events (31) amounts to a maximality requirement. Since we associate the presence of a requirement for event maximality with perfective forms, the value [+emax] will indicate in our system that the respective verb is perfective. The [+emax]-predicates of Table 6 are all perfective, in other words. Predicates carrying the meaning [+pmax, –emax], on the other hand, cannot be used as perfectives, because they do not fulfill the maximality requirement. They introduce a maximality condition in that they provide a maximal path, but this alone does not imply a maximal event. According to Table 6, reš-, bros-, and uničtož- are underspecified with respect to event maximality, which is expressed as [+/–emax]. Dependent on whether or not this indeterminacy will be settled as [+emax] or [–emax], their usage as a predicate may or may not come with a requirement for maximal events.

Hoping to have shown how the presented approach is grounded in the theoretical literature, I return to the main line of argument. What have we achieved so far? By drawing lexical distinctions among the verbal roots of Russian, we arrived at a classification of verbal predicates, exploiting the properties of path maximality and event maximality, as defined in Section 7. The lexical assumptions made are summarised in (32). They give rise to the following predictions.

(33) 

Unless additional information comes into play, ...

a. ... a predicate whose stem matches lexical bases like čit- or pis- cannot be used to express perfectivity.

b. ... a predicate whose stem matches lexical bases like reš-, bros-, or uničtož- can be used to express perfectivity, but does not have to.

c. ... a predicate whose stem matches lexical bases like pročit-, napis-, or podpis- cannot but be used to express perfectivity.

d. ... a predicate whose stem matches lexical bases like zna- cannot be used to express perfectivity.

I claim grammatical relevance for the proposed semantic classification of verbal roots or, to be more precise, lexical base predicates. In the next section, it will be shown why this claim is licit.

9. Theme Vowel Selection Disambiguates Underspecified Base Predicates

According to (33b), the roots reš-, bros- and uničtož- can be used as perfective or imperfective predicates. This seems correct, the two predicted options being formally distinguished by two different theme vowels that the root combines with. To form a perfective predicate, the root adopts the theme vowel -i, whereas to form an imperfective, it will take the vowel -a. The following constructed example contrasts these two possibilities with each other.

(34) 

Ja rešal ėtu problemu tridcat’ minut, no tak i ne rešil.

I solve.pst.ipfv this problem thirty minutes but so and not solve.pst.pfv

‘I was engaged in solving the problem (= I was trying to solve the problem) for 30 min, but I never solved it.’

In terms of the feature specification from above, this implies that each of the two theme vowel alternatives -i and -a signifies one of the two event maximality values. Theme vowel -i indicates that the value is [+emax], whereas theme vowel -a signals that it is [–emax]. What I claim here, in other words, is that the Russian theme vowels -a and -i serve the function of disambiguating the meaning of predicates like reš-, bros-, or uničtož-. As we have seen, these predicates are special in that they provide a maximal path, at the same time being underspecified with respect to whether the events in their denotation set will at some point reach the maximum introduced by the path.

(35) 

Disambiguation by vowel insertion-a∼-i. If the meaning of the input predicate is path-maximal and underspecified with respect to event maximality, the theme vowel -i will be chosen to indicate [+emax], while the theme vowel -a will be used to signal [–emax].

Note that this proposal implies that the vowel -a in the imperfective verb rešat’ does not express the undoing of former perfectivity—i.e., the suffix -a does in this case not function as a marker of secondary imperfectivisation. Instead, the vowel -a marks the resolution of an underspecified meaning. For a root the meaning of which does neither require nor deny event maximality, in other words, the choice of the theme vowel will serve a tie breaker for verbal aspect. The use of -i will install the maximality requirement of perfectivity.

Now, let us see what this implies for the derivation of specific prefixed verbs. A plausible derivation for razrešit’ seems to be (36).

(36) 

reš- [+pmax, +/–emax] > reši- [+pmax, +emax] > razreši [+pmax, +emax]

The first derivational step in (36) represents the disambiguation of the underspecified event maximality-value. As proposed above, the theme vowel -i indicates [+emax]. In line with (32c), the prefixation in the second derivational step then outputs [+pmax, +emax] as values (which, in the given case, match the input values). So far, so good, but now see that while this approach seems to work for razrešit’, it runs into problems with imperfective razrešat’:

(37) 

*reš- [+pmax, +/–emax] > reša- [+pmax, –emax] > razreša- [+pmax, +emax]

The hypothetical chain would be the one given in (37). Disambiguation by means of -a in the first step amounts to a [+pmax, –emax]-predicate, which is the correct value for an imperfective verb. Prefixation in the second step produces a [+pmax, +emax]-predicate, thus perfectivising it. This is not what we want, however, because verbs whose stem is razreša- are not perfective. (The same issue arises with imperfective otrešat’.)

This problem can be fixed by adopting two assumptions. First, we should let the prefixes apply to the root before the theme vowel is added. Secondly, we should assume that prefixing a [+pmax;+/–emax]-predicate does not change these values. The first assumption is an option that we are free to adopt, as far as I can see. The second one deserves some comment, however. To see that it is a plausible assumption, we have to reflect on what it means that the root reš- is underspecified with respect to event maximality. It means that the root denotes both maximal and non-maximal events. This should then be considered in conjunction with the impact of path maximality: all events in the denotation of the predicate develop along maximal paths.

(38) 

reš-:

a. [+pmax, +emax]

b. [+pmax, –emax]

(38a) shows the abstract contour of a maximal event in the denotation set of reš-, while (38b) shows the contour of a non-maximal event in it. Consider (38a) first. The maximal events all traverse along a maximal path (indicated by the arrow), with the path maximum demarcating the point at which the event maximum has been reached (indicated by a vertical line). Turning to (38b), we see that, since the root is restricted to path-maximal events, the denotation likewise contains only events traversing along upper-bounded (maximal) paths. This time, however, since the predicate denotes non-maximal events, these path-maximal events line up to form an open series.17

The described peculiarity of [+pmax;+/–emax]-predicates has an interesting consequence when it comes to prefixation. If a prefix attaches to such a predicate, it will not newly introduce path maximality, because the base predicate is already [+pmax]. The semantic effect of the prefix reduces to a mere qualitative modification concerning the lexical description of the event.

Our assumption that the semantic effect of prefixing a predicate like reš- corresponds to a change in the kind of maximal path, but not to the introduction of a maximal path itself, is confirmed by dictionary descriptions. According to Kuznecov (2020), rešit’/rešat’ has a polysemous lexical semantics subsuming the following meanings: 1. drawing a decision; 2. together arriving at a decision; 3. finding a solution; 4. solving a task when fulfilling a plan; 5. determining the further course of events (about an event). With respect to razrešit’/razrešat’, the same source gives the following lexical meanings: 1. giving permission; 2. deciding which option is the correct or best one; 3. settling a dispute. As is clear from these lexical descriptions, the prefixed predicate razreš- has a more specific meaning compared to the simplex reš-. While the meanings of reš- can be united under the umbrella ‘deciding an issue’, the meanings of razreš- all seem to instantiate the general meaning ‘deciding an issue by selecting among contextually available candidate solutions’.

In the next section, I want to show that, furnished with (32) and (35), we are able to correctly derive the inventory of morphologically complex forms based on reš-, uničtož- and pis-.

10. Early Prefixation (Before Vowel Selection)

We have seen that the attachment of the prefix raz- to the root reš- will lead to a more narrow meaning of the verb, but it will not alter the values for the properties of path maximality and event maximality. If we assume that reš- is [+/–emax], as we did in (32b), we have no reason not to assume the same value for razreš-. The semantic effect of prefixing raz- boils down to the indication of contextually salient alternatives to choose from. The feature representation of razreš- is thus [+pmax, +/–emax], as stated for the respective item in Table 6 (see footnote 18).

(39) 

reš- [+pmax, +/–emax] > razreš- [+pmax, +/–emax]

Discussing the phenomenon within his theory, Tatevosov (n.d., p. 25) reasons that with prefixed simplex perfectives ‘a result state can somehow come on top of a complex eventuality description for which its own result state has been specified’. Consequently, razreš- should be analysed as being associated with two result states. As far as I can see, this is well compatible with the proposal made here, which basically says that the result states in the denotation of razreš- form a subset of the result states in the denotation of reš-. For instance, the set of states of having decided something by selecting one contextually available alternative solution forms a subset of the set of states of having decided something. Or, to use the example discussed in Tatevosov (n.d.), otrešit’, the set of states of having decided to fire someone forms a subset of the set of states of having decided something.

Now see that, if we adopt the two assumptions argued for above, we will correctly derive the perfective verb razrešit’ and the imperfective verb razrešat’ by appealing to (35).

(40) 

a. reš- [+pmax, +/–emax] > razreš- [+pmax, +/–emax] > razreši- [+pmax, +emax]

b. reš- [+pmax, +/–emax] > razreš- [+pmax, +/–emax] > razreša- [+pmax, –emax]

Basically the same story could be told about the virtual root ničtoz- (‘stop existing’). Like reš-, it is [+pmax] and [+/–emax]. The root is virtual in the sense that it is not capable of forming a stem on its own, as there is no Russian verb ničtozit’ or ničtozat’. There are two prefixes with which it can combine, namely, u- and iz-. Both derive perfective and imperfective verbs depending on whether the theme vowel is -i or -a. Note that due to the absence of the verb ničtozit’ or ničtozat’, the proposed analysis is equivalent to one that treats uničtoz- and izničtoz- as roots.19

(41) 

a. (ničtož- [+pmax, +/–emax] >) uničtož- [+pmax, +/–emax] > uničtoži- [+pmax, +emax] (‘destroy’)

b. (ničtož- [+pmax, +/–emax] >) uzničtož- [+pmax, +/–emax] > uničtoža- [+pmax, –emax] (‘destroy’)

(42) 

a. (ničtož- [+pmax, +/–emax] >) izničtož- [+pmax, +/–emax] > izničtoži- [+pmax, +emax] (‘erase’)

b. (ničtož- [+pmax, +/–emax] >) izničtož- [+pmax, +/–emax] > izničtoža- [+pmax, –emax] (‘erase’)

Above, I have argued for prefixation taking place before the theme vowel is determined. Together with the idea that the choice between -a and -i disambiguates predicates that were otherwise underspecified with respect to event maximality, this assumption made it possible to explain the properties of the verbs razresit’ and razresat’ as well as uničtozit’ and uničtozat’. Once we allow for early prefixation with these [+/–emax]-roots, we should perhaps assume early prefixation also for other cases. There is in fact no reason not to assume the derivation noted in (43) for [-pmax, –emax]-roots.

(43) 

pis- [–pmax, –emax] > dopis- [+pmax, +emax]

In the next step, the theme vowel can be added, and since the base predicate dopis- is unequivocally specified for event maximality, (35) does not apply. Theme vowel selection is void of semantic significance in these cases (at least as far as I can see), and the choice in the given case randomly falls on -a.

(44) 

dopis- [+pmax, +emax] > dopisa- [+pmax, +emax]

To derive an imperfective counterpart to this perfective stem, the suffix -yva is used. The impact of yva- is to switch [+emax] to [–emax].20

(45) 

dopisa- [+pmax, +emax] > dopisyva- [+pmax, –emax]

The step from pis- to dopis- implies a change in lexical meaning (from ‘write’ to ‘finish writing’). The step from pis- to napis- does not, or at least not obviously, which leads to the well-known characterisation of na- as an empty prefix. This difference notwithstanding, the derivation of napis- can be analysed on a par with the derivation of dopis-.

(46) 

pis- [-pmax, –emax] > napis- [+pmax, +emax] > napisa- [+pmax, +emax] > napisyva- [+pmax, –emax]

For those who believe that napis- does not allow for secondary imperfectivisation, I add an actual example from 2023 (https://www.instagram.com/nemolchi.uz/p/Cn6pwxltljj/ accessed on 16 December 2024):

(47) 

Vse ėto vremja on napisyval i napisyval, o tom kak ja emu

all this time he write.pst.ipfv and write.pst.ipfv about this how I him

nravljus’. Ja ničego daže podozritel’nogo ne zamečala.

please I nothing even suspicious not notice.pst.ipfv

‘All this time he’s been writing and writing about how much he likes me. I didn’t even notice anything suspicious.’

11. Late Prefixation (After Vowel Selection)

By postulating semantic classes of verb roots that are different from each other, it was possible to correctly derive the forms rešit’, rešat’, razrešit’, razrešat’, uničtozit’, and uničtozat’ in accordance with their actual aspectual values. The next step is now to come up with a correct generation of the verbs which are based on the root bros-. To begin with, consider the following morphological chain, which correctly derives the perfective verb form vybrosit’ (‘throw out’). In line with what was said above, the prefix applies early, i.e., prior to vowel suffixation, with the resulting predicate inheriting the indeterminate event maximality value of the base. Suffixation by -i will then, in a second step, delete the non-maximal events from the denotation set.

(48) 

bros- [+pmax, +/–emax] > vybros- [+pmax, +/–emax] > vybrosi- [+pmax, +emax]

What was said above also licenses the following chain, i.e., the derivation of imperfective vybrosat’. This chain is not licit, however, as there is no imperfective verb vybrosat’, and the question is why.

(49) 

*bros- [+pmax, +/–emax] > vybros- [+pmax, +/–emax] > vybrosa- [+pmax, –emax]

As already anticipated in Section 2, my answer is that the potential imperfective verb vybrosat’ is blocked by the availability of the perfective verb vybrosat’, which is arguably derived along the following lines.

(50) 

bros- [+pmax, +/–emax] > brosa- [+pmax, –emax] > vybrosa- [+pmax, +emax]

(50) would be a case of late prefixation in the sense that the prefix applies only after the theme vowel has been added. The result is a perfective event-internal pluractional; recall (27). I propose, in other words, that the existence of (50) uses up, so to speak, the form vybrosat’, making it no longer available as an imperfective. As a consequence, the perfective vybrosit’ in (48) has to look for an alternative aspectual partner. Such an alternative can be found in vybrasyvat’, resulting from canonical secondary imperfectivisation.

(51) 

vybrosi- [+pmax, +emax] > vybrasyva- [+pmax, –emax]

How to theoretically pin down the suggested blocking mechanism? To set the stage, we first of all note in (52) what was already said in the discussion above.

(52) 

Early and late prefixation. Prefixes may apply prior to or after the stage at which the theme vowel is inserted.

In a next step, let us take seriously the well-known fact that external prefixes, including intermediate prefixes and unlike internal prefixes, always make a semantic contribution which is fully transparent. External prefixes satisfy compositionality: the meaning of the prefix always modifies the meaning of the base to which it attaches in a predictable way. For this to be possible, plausibly enough, the input expression of the prefix has to have a predictably modifiable meaning.

I assume that, in general, a compositional semantic operation has to operate over the extensions of the two expressions that are semantically related toward each other. A predictably modifiable meaning has to be, therefore, an extensional meaning, i.e., one that delivers a set of entities relative to some world and time. It is arguably only with the insertion of a theme vowel that a string of morphemes turns into a predicate satisfying this requirement. Russian verb formation involves, in other words, a first morphological cycle in which ‘proto-predicates’ are formed. This cycle ends when a theme vowel is added to turn the proto-predicate into what I want to call a ‘lexical core predicate’.21

(53) 

Theme vowels complete the formation of the lexical core predicate. There is a first cycle of morphological structure building producing a base predicate to express what is the lexical core meaning of a verbal predicate. This process is completed with the addition of a theme vowel.

Given (52) and (53), we can now state the blocking rule that we need.

(54) 

Prefer simple cores. If two derivational chains arrive at one and the same form, the chain that uses less operations during the first morphological cycle (i.e., the chain that creates the predicate’s core meaning with fewer steps) will win over the other chain.

Against this background, compare (49) and (50). Both chains derive the form vybrosat’. The former reaches the point at which the theme vowel is added after two steps. The latter needs only one derivational step to reach that point. Therefore, according to (54), the latter will win the battle for vybrosat’ against the former.

Principle (54) enables us to explain Tatevosov’s observation discussed in Section 2. Recall the following two perfective predicates that we came across.

(55) 

a. ${\mathrm{dorešat}’}^{pf}$ (cf. (10b))

b. ${\mathrm{douničtožat}’}^{pf}$ (cf. (11b))

(55a) will be derived along the lines of (56), thus creating an event-internal pluractional. The derivation of the verb in (55b) works analogously to this.

(56) 

reš- [+pmax, +/–emax] > reša- [+pmax, –emax] > doreša- [+pmax, +emax]

Tatevosov’s observation that the imperfective counterpart of the perfective dorešit’ is dorešivat’ rather than dorešat’ may be explained by appealing to (54). While (56) settles the lexical core meaning in one step, (57) needs two steps to do so.

(57) 

*reš- [+pmax, +/–emax] > doreš- [+pmax, +/–emax] > doreša- [+pmax, –emax]

Since (57) is ruled out by the availability of (56), the mechanism of secondary imperfectivisation is called for to create an imperfective partner verb for dorešit’.

(58) 

reš- [+pmax, +/–emax] > doreš- [+pmax, +/–emax] > doreši- [+pmax, +emax] > dorešiva- [+pmax, –emax]

Let me end by mentioning two apparent counterexamples to the theory developed here. The dictionary A. Zaliznjak (1977, p. 690) includes a perfective predicate pererešat’ and an imperfective one of the same form. The meaning of the former is indicated as ‘decide many issues’ (rešit’ mnogoe), and the meaning of the latter is ‘decide again’ (rešat’ zanovo).

(59) 

a.	${\mathrm{pererešat}’}^{pf}$ ‘decide many issues’	(event-internal pluractional)
b.	${\mathrm{pererešat}’}^{ipf}$ ‘decide again’	(imperfective with intermediate prefix)

According to the analysis proposed in the present paper, these two forms result from the two derivations below.

(60) 

reš- [+pmax, +/–emax] > reša- [+pmax, –emax] > perereša- [+pmax, +emax]

(61) 

reš- [+pmax, +/–emax] > perereš- [+pmax, +/–emax] > perereša- [+pmax, –emax]

Given Principle (54), we would now expect that the generation of pererešat’ along the lines of (61) will be blocked because, in comparison to the chain in (60), the chain of (61) has a more extended first morphological cycle involving more steps until the point of theme vowel insertion is reached. So, if we accept the information in Zaliznjak’s dictionary at face value, this expectation is not borne out and we face a counterexample here. Note, however, that (61) is marked by Tatevosov (2013b) by a question mark, i.e., as being dispreferred (see footnote 22 ).

The morphological chain that actually follows from the rules argued for in the present paper is (62). Here, the repetitive pere- applies after the first cycle, as it is supposed to be with intermediate prefixes. In line with Tatevosov’s judgement, this chain derives pererešivat’.

(62) 

reš- [+pmax, +/–emax] > reši- [+pmax, +emax] > perereši- [+pmax, +emax] > pererešiva- (‘decide again’) [+pmax, –emax]

A second potential counterexample is douničtožat’. As mentioned in Section 2, this form can be perfective or imperfective.

(63)

a.	${\mathrm{douničtožat}’}^{pf}$ ‘finish destroying’	(event-internal pluractional)
b.	${\mathrm{douničtožat}’}^{ipf}$ ‘finish destroying’	(imperfective with intermediate prefix)

The two derivations that give us the forms in (63) are the ones shown in (64).

(64) 

a. uzničtož- [+pmax, +/–emax] > uničtoža- [+pmax, –emax] > douničtoža- [+pmax, +emax]

b. uničtož- [+pmax, +/–emax] > uničtoži- [+pmax, +emax] > douničtoži- [+pmax, +emax] > douničtoža- [+pmax, –emax]

Note that none of the two items blocks the other one by (54), as both involve the same amount of steps in the first morphological cycle. What is somewhat unexpected is something else, namely, the final derivational step in (64b). It would have been possible to produce douničtoživat’, by analogy with (62). This would have avoided the homonymy in (63), but for some unclear reason, this option was not chosen.

12. Conclusions

In this paper, attention was drawn to some facts about Russian verb formation that seem to ‘impurify’ the simple and elegant picture according to which prefixation triggers perfectivisation of an imperfective base, and subsequent suffixation triggers the re-imperfectivisation of the perfective. A theoretical analysis was developed to integrate simplex perfectives into the picture. It was shown how it is possible to correctly derive the morphological and aspectual characteristics of these verbs and their derivatives without treating non-prefixed perfectives as exceptions standing outside of the grammatical system. What needs to be assumed for the proposed analysis to work are four things.

1. We have to distinguish between different lexical classes of verbal roots, based on two semantic criteria: whether or not the predicate requires the event to develop along a maximal path, and whether or not it requires the event to actually reach its maximum.

2. We have to consider theme vowels to be boundary markers that close a fundamental first morphological cycle at which the core meaning of a predicate is determined. If the base to which the theme vowel attaches is underspecified with respect to event maximality, theme vowel selection gains a special significance in that it specifies the event maximality value as either maximal (by selection of -i) or non-maximal (by selection of -a).

3. We should expect prefixes to attach prior to or after the boundary marking theme vowel, i.e., inside or outside the first cycle. In the former case, the prefix will serve as an internal prefix contributing to the core lexical meaning; in the latter case, it will function as an intermediate prefix, which takes the core meaning as input and transforms it in a predefined way.

4. We have to take into consideration a pragmatic blocking mechanism which says that if two identical forms are derived with a differing number of operations in the first morphological cycle, the one with the smaller number will win.

One important implication that the proposal made in this paper has for aspectological theory is that the inventory of prefixes that may apply after theme vowel insertion (i.e., external to the first cycle) but prior to secondary imperfectivisation is considerably larger than Tatevosov’s class of intermediate prefixes (Tatevosov, 2011). What all the prefixes possibly applying at the ‘intermediate’ stage of verb formation share is that they operate over the extension of the predicate derived during the first morphological cycle, i.e., over the set of events denoted by the lexical core predicate, as it was called above. Taking as input the set of events characterised by the core predicate, they return a new set of events which relates to the input set in a systematic and transparent way. The new denotation may be the set of events representing final stages of the input events (the impact of completive do-), or the set of events representing diminished clones of the input events (the impact of attenuative pod-), or the set of events which are single repetitions of the input events (the impact of repetitive pere-). Besides that, a prefix may apply at the intermediate stage to return the set of events consisting of a multitude of repetitions of input events, temporally ordered such that their successive realisation will measure out the path introduced by the prefix. Respective events in the output denotation are event-internal pluractionals.

Restrictions on the combinability of ‘intermediate prefixes’ and core predicates seem to be pragmatic by nature. For to allow the derivation of completive, attenuative, or repetitive prefix verbs, the lexical base should provide an input meaning suitable to the content of the respective operation (completive, attenuative, or repetitive). Similar pragmatic restrictions apply to the derivation of an event-internal pluractional (Wągiel, 2023, p. 279). As for the cases discussed in this paper, the path delivered by the prefix and the kind of event delivered by the base should harmonise. While the meaning of the verb bros- (‘throw’) is well compatible with a broad range of different kinds of spatial paths/prefixes in this respect, the meaning of the verb reš- (‘solve’) seems to lend itself for an event-internal pluractional only in connection with the path provided by the completive prefix do-.

The present paper reemphasises, moreover, that it is too simple to count the suffix -a as an allomorph of the secondary imperfectivising morpheme across-the-board. The suffix -a may appear as an alternative form to -yva, and it is not quite clear by now when it does and when it does not. Recall from above the case of the imperfective douničtožat’. Also, why is it prikančivat’, and not prikančat’ on the one hand, but vypuskat’, and not vypuskivat’ on the other hand? However, according to the analysis outlined above, the element -a may likewise appear as an imperfectivity-marking theme vowel, as in rešat’ and brosat’, for instance.

If the treatment of theme vowels as boundary markers of the (internal) first cycle of morphological structure building is correct, this implies not only the possibility of early (internal) and late (external) prefixation. We should also expect other morphological operators to be able to appear before (‘early’) and after (‘late’) theme vowel insertion. And indeed, it seems that also the semelfactive suffix may apply early or late. In some cases, this gives rise to perfective doublets like those in (65)—see A. A. Zaliznjak & Šmelev (1997, p. 101).

(65) 

a. ${\mathrm{tolkat}’}^{ipf}$ (‘push’) ∼ ${\mathrm{tolknut}’}^{pf}$ (‘push once’); ${\mathrm{tolkanut}’}^{pf}$ (‘push once heavily’)

b. ${\mathrm{mazat}’}^{ipf}$ (‘smear’) ∼ ${\mathrm{maznut}’}^{pf}$ (‘smear once’); ${\mathrm{mazanut}’}^{pf}$ (‘smear once heavily’)

It is tempting to consider the stems of the imperfective tolkat’ and the perfective tolknut’ to be formed within the first cycle, but the perfective tolkanut’ resulting from n(u)- attaching after the theme vowel -a has been established. The idea is sketched in (66).

(66) 

a. tolk- > tolka-

b. tolk- > tolkn- > tolknu-

c. tolk- > tolka- > tolkan(u)-

Like many other issues only touched upon in this paper, this raises a lot of new questions that call for independent studies.