1. Introduction
The concept of linguistic relativity suggests that each language has a certain impact on the cognition of its speakers. Strong variants of this hypothesis, such as linguistic determinism (Whorfianism), have been refuted, but it has become clear during the past few decades that language is indeed capable of facilitating or hindering certain cognitive processes. For example, psycholinguistic studies on color adjectives have revealed that speakers of Russian are faster in recognizing different shades of blue than English speakers, presumably because Russian has two basic color adjectives for blue: sinij ‘dark blue’ and goluboj ‘light blue’ (Winawer et al., Reference Winawer, Witthoft, Frank, Wu, Wade and Boroditsky2007).
Not only lexical phenomena are subject to this kind of investigation, but grammatical categories are also being researched regarding their impact on thinking. Here, verbal aspect, that is, the opposition of progressive and nonprogressive or imperfective and perfective verb forms, is particularly interesting and has been widely examined, in particular with respect to English.
Two lines of research can be differentiated when it comes to verbal aspect and its influence on event conceptualization: First, a number of studies have aimed to compare speakers of different languages differing in the way they encode aspectuality while working on the same task. Thus, von Stutterheim et al. (Reference von Stutterheim, Andermann, Carroll, Flecken and Schmiedtová2012) investigated target orientation, that is, how far respondents are oriented to the endpoint of a situation, in various (including Slavic) languages, finding that the speakers of languages without a grammaticized aspect are more likely to take a holistic view on motion events and refer to their endpoints than the speakers of aspect languages. Recently, Kamenetski et al. conducted a series of experiments on the influence of verbal aspect on event conceptualization, the processing of aspect temporality and event memory in English and Russian; they found processing differences between the two languages, which they attributed to different degrees of grammaticalization of the aspectual category and more or less strict correspondences between aspect forms and distinct temporal distributions (Kamenetski & Lai, Reference Kamenetski and Lai2025, Reference Kamenetski and Laiin prep; Kamenetski et al., Reference Kamenetski, Misersky, Lai and Fleckenin prep). Minor et al. (Reference Minor, Mitrofanova and Ramchand2022) on Russian and Minor et al. (Reference Minor, Mitrofanova, Guajardo, Vos and Ramchand2023) on Russian, Spanish and English took a slightly different perspective by focusing on the processing of the aspectual forms themselves and the fine-grained time course of this processing.
The second line of research is represented by studies that compare behavior within speakers of a particular language, here depending on varying linguistic or nonlinguistic (e.g., pictures, videos) input. Various studies on English have focused on the impact of the progressive and/or the simple form on event processing (Anderson et al., Reference Anderson, Matlock, Fausey and Spivey2008; Becker et al., Reference Becker, Ferretti and Madden-Lombardi2013; Feller et al., Reference Feller, Eerland, Ferretti and Magliano2019; Ferretti et al., Reference Ferretti, Kutas and McRae2007; Madden & Zwaan, Reference Madden and Zwaan2003; Magliano & Schleich, Reference Magliano and Schleich2000; Misersky et al., Reference Misersky, Slivac, Hagoort and Flecken2021). Although early studies are concerned with both progressive and nonprogressive aspects and their contribution to the construction of situation models (Madden & Zwaan, Reference Madden and Zwaan2003; Magliano & Schleich, Reference Magliano and Schleich2000; Morrow, Reference Morrow1990), in the later works, three subtopics can be identified: first, studies focusing on the progressive (taking the nonprogressive as a baseline), second, studies focusing on the nonprogressive (with the progressive as a baseline), and, third, studies focusing on the interaction of grammatical aspect (thus, both progressive and nonprogressive) and lexical aspect, that is (more or less) telicity.
Studies of the first type (‘progressive studies’) have revealed that using the progressive form causes stronger mental activation of the described situation than using the nonprogressive (Anderson et al., Reference Anderson, Matlock, Fausey and Spivey2008; Ferretti et al., Reference Ferretti, Kutas and McRae2007). As revealed in various types of experiments, participants have a more detailed memory of the components associated with a situation (e.g., location or instruments) when it is described in the progressive than in the nonprogressive aspect.
Studies of the second type (‘event studies’) have focused on the effects of the English nonprogressive simple form on event comprehension (Misersky et al., Reference Misersky, Slivac, Hagoort and Flecken2021) and event segmentation (Feller et al., Reference Feller, Eerland, Ferretti and Magliano2019); they show that the simple form is associated with holistic event representation and facilitates the discovery of temporal boundaries in narratives.
Finally, studies of the third type (‘interaction studies’) have brought together grammatical and lexical features of verbs and show that the English progressive and nonprogressive interact with the lexical aspect of a verb, that is, whether they encode accomplishments or activities. Becker et al. (Reference Becker, Ferretti and Madden-Lombardi2013) let their participants read short stories in which a target event, either an accomplishment or an activity, was presented in the progressive or nonprogressive aspect. At the reintroduction of the target event later in the story, they found a larger N400 for nonprogressive versus progressive accomplishments, but no difference after nonprogressive versus progressive activities. This means that the availability of concepts in situation models was affected by grammatical aspect, but only for accomplishments.
Few studies investigated languages other than English. A notable exception is Golshaie and Incera (Reference Golshaie and Incera2021), concerning grammatical aspect of the perfective–imperfective type in Farsi. Although ‘imperfective study’ would be the better label for this study, it comes close to ‘progressive studies’ with respect to English because their sentence material might be interpreted as eliciting a progressive reading. Golshaie and Incera partially confirmed the results on English (see the detailed discussion below).
‘Interaction studies’ on languages other than English are Yap et al. (Reference Yap, Chu, Yiu and Wong2009) and Vanek and Zhang (Reference Vanek and Zhang2024). Yap et al. (Reference Yap, Chu, Yiu and Wong2009) investigated the interaction of grammatical and lexical aspects in Cantonese by applying the perfective and imperfective aspect markers gan and zo. In contrast to Becker et al. (Reference Becker, Ferretti and Madden-Lombardi2013), they revealed slightly different results,Footnote 1 indicating that there is a need for more in-depth research. Vanek and Zhang (Reference Vanek and Zhang2024) compared L1 speakers of English and Mandarin and L2 speakers of English (with Mandarin L1) on an event segmentation task with durative and punctual events. They revealed a segmentation benefit for English speakers, presumably caused by (almost) obligatory marking of progressiveness in English in comparison to facultative marking in Mandarin.
One problem is that the mentioned psycholinguistic studies do not take into account the full complexity of or the cross-linguistic typological differences between aspect systems. Dealing mostly with English, with a few exceptions, they are limited to the functional aspectual distinction of progressive versus nonprogressive. From a cross-linguistic perspective, aspectual differences are not limited to this distinction but are far more diverse (e.g., Comrie, Reference Comrie1976, p. 25). The aspectual category of Slavic languages represents another type, the perfective versus imperfective. In Slavic, the imperfective aspect can be used not only in progressive but also in iterative/habitual and general-factual contexts (see Section 1.2), as well as others. Then, the question arises as to how the divergent shape of the Slavic aspect opposition affects event processing and whether the mentioned effects would be associated with Slavic perfective and imperfective verbs as well.
In the current paper, we present an investigation of these issues, which follows ‘progressive studies’ and deals with the Russian imperfective in its progressive function and its hypothesized stronger mental activation effect.
The present paper is structured as follows: First, we give a more detailed overview of ‘progressive studies’ and ‘imperfective studies’. Then, we compare Russian and English verbal aspects and derive our research question and hypotheses. In Section 2, we present the material, participants and procedure of a self-paced reading experiment with 50 native Russian speakers. Section 3 presents the experiment’s results. Finally, we discuss the results in Section 4 against the background of both psycholinguistic event processing research and cognitive theories on Russian verbal aspect.
1.1. The progressive and its impact on event processing
As mentioned in the Introduction section, there is one line of research that explores the effect of stronger mental activation of an event caused by the progressive or imperfective. Because our study falls into this category, we present the experiments of Ferretti et al. (Reference Ferretti, Kutas and McRae2007) and Golshaie and Incera (Reference Golshaie and Incera2021) in more detail.
Ferretti et al. (Reference Ferretti, Kutas and McRae2007) conducted three experiments to investigate how verbal aspect influences the activation of event knowledge. Based on Madden and Zwaan (Reference Madden and Zwaan2003) and Magliano and Schleich (Reference Magliano and Schleich2000), Ferretti et al. (Reference Ferretti, Kutas and McRae2007) tested the assumption that locations involved in a situation should be more present in the mental representation of this situation if it is described in the progressive. In the first experiment – a semantic priming task – participants read aloud target nouns denoting locations after having seen verb primes in either the past perfect form (had skated) or the past progressive form (was skating). The locations were either related to the situation described by the verb (e.g., arena) or unrelated (e.g., library). Verb–location pairs were pretested for their relatedness in a norming study. The results showed that reading times for related locations were faster than for unrelated ones, but only when the prime was in the progressive form. The second experiment was a sentence completion task with sentences containing the past progressive (e.g., the cow was grazing…) or the past perfect (e.g., the cow had grazed…). Participants were asked to write continuations that sounded the most natural to them. According to the authors’ expectations, participants added locative prepositional phrases significantly more often when the beginning of the sentence contained the past progressive rather than the past perfect form. The third experiment is the most important for the present paper. In an EEG study, participants read sentences of the type the diver was snorkeling/had snorkeled in the ocean/pond. The authors measured ERP components associated with the prepositional phrases denoting locations either highly plausible (e.g., ocean) or less plausible (e.g., pond) for the aforementioned event. The authors found the effect caused by the less plausible location (ERP component N400) to be high when the situation was described in the progressive form but not when it was described in the past perfect form. N400 amplitudes to the highly and less plausible locations following the same events referenced as completed with the past perfect form were intermediate and did not vary significantly as a function of plausibility. In contrast, participants had the most difficulty (reflected by the highest N400 components) processing locations that followed progressive forms when the locations were less plausible (snorkeling in the pond). Ferretti et al. concluded that the events themselves and their locations were more activated in speakers’ situation models when those events were referenced as ongoing by the progressive form. Thus, the activation of world knowledge concerning the likelihood of events’ locations is modulated by the verb form. However, a certain problem with the Ferretti et al. study is that the authors did not properly separate tense and aspect features of their verbal material. The predicates of the sentences the diver was snorkeling in the ocean and the diver had snorkeled in the ocean differ not only in aspect (progressive versus nonprogressive) but also in tense (past versus past perfect). Thus, the observed effect might be caused by the specifics of the past perfect as well and cannot be unambiguously attributed to aspectual differences.
Golshaie and Incera (Reference Golshaie and Incera2021) found similar results; however, they applied another experimental design and transferred the research issue to another language: Farsi. Following the authors, in Farsi, the imperfective marker (mi-) is associated with three aspectual functions: ‘continuity (progressiveness), iteration, and habituality of an action’ (Golshaie & Incera, Reference Golshaie and Incera2021, p. 740). The authors stated that their ‘study will focus on the general unbounded meaning of the imperfective marker […]’ (p. 740). They conducted a mouse tracking study and asked their participants to read sentences (e.g., Sara has sliced the zucchinis or Sara is slicing the zucchinis; translation of the example sentences taken from the original study), followed by a picture of an object that might function as an instrument of the event described by the verb (e.g., a knife). Participants had to answer the question of whether the object had been mentioned in the sentence or not by pressing one of two response buttons (‘present’ or ‘absent’). For experimental sentences, the correct answer was always ‘absent’. The authors expected initiation time and reaction time (RT) to be longer and mouse trajectories to be more deviating from the ideal trajectories in the imperfective condition. This expectation was based on previous research, which had shown that the progressive causes a higher activation of the events’ components (for locations, see Ferretti et al., Reference Ferretti, Kutas and McRae2007; for instruments, see Truitt & Zwaan, Reference Truitt and Zwaan1997, as cited in Madden & Zwaan, Reference Madden and Zwaan2003). Contrary to the authors’ expectations, initiation time and RT did not show any differences as a function of aspectual variation. However, the imperfective aspect caused increased deviations of the mouse trajectories. The authors argued that this was because of the deeper mental activation of the events caused by the imperfective; thus, participants had to make a greater effort to conclude that the instrument had not been mentioned in the sentence.
In summary, both studies argued that the progressive/imperfective causes a stronger mental activation of the described event than the past perfect/perfective.
1.2. Russian verbal aspect (as compared with English)
The Slavic languages share the general notion of verbal aspect and the overall organization of the category, but they differ considerably in formal and functional details. We will concentrate on Russian in this section because our study is also dedicated to Russian.
As mentioned in the Introduction section, Russian and English represent different types of aspectual opposition (Comrie, Reference Comrie1976, p. 25). While the English aspect is based on the opposition of progressive versus nonprogressive, the Russian aspectual category belongs to the perfective versus imperfective type. Perfective and imperfective verbs with identical lexical meaning form oppositions, referred to as aspectual pairs. (Almost) each verb can be assigned either to the group of perfective or to the group of imperfective verbs.Footnote 2 Usually, two verbs from each group form an aspectual relationship, for example the perfective pridumat’ and the imperfective pridumyvat’, both meaning ‘(to) come up with sth, (to) invent’. Aspectual semantics and the question of which verbs form an aspectual pair is a matter of enduring and controversial debate.
Most importantly for our research question, Russian imperfective verbs can occur in several different contexts or, in other words, can have different functions, with progressive (traditionally called concrete-processual in Slavic linguistics) being one of these, cp. the imperfective letel ‘was flying’ in the sentence V tot moment kogda ja letel otvesno vniz golovoj, ja požaluj ničego drugogo ne želal […] ‘when I was flying upside down, I wasn’t wishing for anything else’ (example from the Russian National Corpus (RNC)). Unrestrictedly iterative, general-factual and stative situations, which are expressed in English by the nonprogressive form, are expressed by Russian imperfective verbs as well. Unrestrictedly iterative predicates refer to indefinitely often repeating situations, cp. vyxodil ‘went out’ in the sentence […] i každyj den’ k semi utra, kak na dežurstvo, vyxodil ja na ulicu Seny, […] ‘[…] and every day around seven o’clock in the morning I went out on Sena Street, as if I was on guard duty […]’ (example from the RNC). Verbs with general-factual meaning refer to nonepisodic, that is, temporally indefinite situations (e.g., rešali ‘have been resolved’ in the sentence Semušin vsё znaet. Voprosy oplaty s vami uže rešali? ‘Semušin knows everything. Have payment issues been resolved with you yet?’; see below for more details). Stative situations refer to ‘states’ (Vendler, Reference Vendler1957), cp. sootvetstvuet in the sentence Takim obrazom čislo golov sootvetstvuet čislu strun na lire Orfeja‚ ‘Thus the number of heads corresponds to the number of strings on Orpheus’ lyre’ (example from the RNC). Correspondingly, the functional array of the perfective is much smaller than that of the English nonprogressive. Mainly, perfective verbs occur in concrete-factual contexts, referring to single, episodic situations that are presented as events, that is, as having reached an endpoint. However, in specific contexts, they can carry further particular meanings (e.g., the summative meaning, referring to a certain number of iterations of a situation, being presented as one single event).
Because it is the leading cognitive approach to Slavic verbal aspect, we base our assumptions and argumentation concerning aspectual semantics on Stephen Dickey’s ‘East–West theory of Slavic aspect’ (Dickey, Reference Dickey2000, Reference Dickey2018, Reference Dickey2024). Following Dickey, the East Slavic (which includes Russian) perfective expresses temporal definiteness of the denoted situation, which means, following Leinonen (Reference Leinonen1982), that the situation is ‘uniquely locatable in a context, i.e., if it is viewed as contiguous in time to qualitatively different states of affairs’ (Dickey, Reference Dickey2000, pp. 19–20). In contrast, the imperfective allows ‘no assignment of the situation to a unique point in time’ (qualitative temporal indefiniteness; Dickey, Reference Dickey2000, p. 24).
According to Dickey, three components are important for characterizing the Russian perfective: First, temporal definiteness is the basic meaning of the Russian perfective, which is usually attested by the sequentiality of events.Footnote 3 Second, Dickey argued that ‘in the construal of a great many instances of the Russian perfective’, the goal of the action plays a crucial role (Dickey, Reference Dickey2024, p. 11). Third, De Wit and Dickey (Reference De Wit, Dickey, De Wit, Brisard, Madden-Lombardi, Meeuwis and Patard2024) developed the ‘epistemic approach’ for the cross-linguistic analysis of the perfective and applied it to English and Russian. This approach is consistent with the concept of temporal definiteness but adds the perspective that ‘the essence of perfective categories is the speaker’s full knowledge of a given situation at speech time’ (Dickey, Reference Dickey2024, p. 13). This means that, in contrast to situations described by imperfective verbs, the beginning and result of the perfective situation are completely present in the speaker’s knowledge set. For example, in the sentence Ja ešče krivljalas’ ipf i prygala ipf vokrug stola, kak vdrug Pavlik zaplakal pf ‘I was still making faces and skipping around the table, when suddenly Pavlik started crying’ (Dickey, Reference Dickey2024, p. 14), the speaker has full knowledge about the beginning of Pavlik’s crying (referred to by the perfective form); however, the making of faces and skipping (described by imperfectives) is a temporally undefined background activity.
From a pragmatic point of view, Dickey’s three fundamental components of the basic meaning of the Russian perfective initiate particular inferences that are elicited with each use of a perfective verb.
Thus, perfectives are characterized by the presence of certain features (temporal definiteness, goal orientation, and full knowledge of the situation), while, as a whole category, imperfectives can be determined only by a missing element, that is, the absence of these features. Instead, their basic semantic content (without further context) is argued to be the type-specification of the denoted situation. When an imperfective verb is used without further context, we can hardly say anything more about its meaning than that the type of situation is specified and the inferences brought in by the perfective form are avoided. From the perspective of cognitive grammar, perfectives involve a foregrounding of time (temporal components), whereas the imperfective is associated with a foregrounding of the situation (activity, event) itself (Dickey, Reference Dickey2024).
Importantly, relations of formal and functional complexity are distributed differently in Russian and English. From a formal perspective, the English progressive is more complex than the nonprogressive. In Russian, formal complexity relations to a certain degree vary with lexical features of the verb, namely telicity, which is also referred to as ‘lexical aspect’ (Lehmann, Reference Lehmann, Anstatt and Norman2010). Prefixed telic verbs usually form perfective basic forms and imperfective partners derived from them via secondary suffixation (e.g., perfective (pf.) pridumat’ vs. imperfective (ipf.) pridumyvat’ ‘(to) come up with sth, (to) invent’). The basic forms of atelic verbs, conversely, are usually imperfective and form their perfective partners via prefixation. The status of the last-mentioned atelic pairs as aspectual pairs is a matter of debate in Slavic linguistics because, formally, verbal prefixation is widely used for lexical derivation as well and prefixes with aspectual functions in one verb can have lexical functions in others. Functionally, it has been argued that aspectual prefixes for the perfectives of atelic verbs are not ‘purely aspectual’ but add some information, for example the imperfective plakat’ ‘to cry’ vs. the perfective verbs zaplakat’ ‘to start to cry’, otplakat’ ‘to finish crying’, proplakat’ ‘to cry for a specific amount of time’ and poplakat’ ‘to cry for some time’.Footnote 4 However, we follow Lehmann’s functional and cognitive approach to Russian aspect (2010) and acknowledge at least po-prefixed perfectives as aspectual partners for atelic imperfectives.
However, the situation is more complex, as there are also morphologically simple verbs that are telic (e.g., ipf. stroit’ ‘to build’). Moreover, a number of morphologically simple verbs are not telic or atelic per se but depending on the context (called ‘diffuse verbs’ by Lehmann, Reference Lehmann, Anstatt and Norman2010). More precisely, the diffusivity of the situation type is a characterizing feature of many imperfective base forms, for example pisat’ ‘to write’. In a context like pisat’ pis’mo ‘to write a letter’, the verb is telic (Vendler’s accomplishment), but in pisat’ nekotoroe vremja ‘to write for some time’, it is atelic (Vendler’s activity). When it comes to the perfectivization of diffuse imperfectives, diffusivity is dismantled: The perfective partner of atelic pisat’ is popisat’ ‘to write for some time’; for telic pisat’, it is napisat’ ‘to write sth’, which is built by a verb-specific prefix.
To sum up, the formal complexity of Russian aspectual forms depends to a certain degree on telicity: In (lexically prefixed) telic verbs, the imperfective is the more complex form because of its aspectual suffix. In contrast, in atelic and diffuse verbs, perfectives are more complex because of their aspectual prefix. In this interaction of grammatical and lexical aspects at the formal level, Russian considerably differs from English.
Their relationships to functional complexity differ as well: While in Russian the perfective has a narrower spectrum of meanings than the imperfective, in English, the progressive is the semantically more specialized form. The nonprogressive ‘has such a wide spectrum that it is quite justifiable to call it “general aspect”’ (Maslov, Reference Maslov and Maslov1985, p. 33). In Russian, this ‘generality’ is on the side of the imperfective. Following Dickey (Reference Dickey2024), the basic meaning of the imperfective is merely a type-specification of the denoted situation. In contrast, the perfective brings into play at least three semantic components (temporal definiteness, goal orientation, and the speaker’s full knowledge of the situation). Thus, the English nonprogressive and Russian imperfective are characterized by ‘generality’, while the English progressive and Russian perfective are functionally more specific and complex.
1.3. Research question and hypotheses
With the present study, we transfer the question of how aspectual forms affect event processing to Russian, where the correspondence between aspectual forms and functions differs from English in that the ongoing action is one but not the only context for the imperfective aspect to occur. Thus, we address the following research question: Does the Russian imperfective aspect in the progressive function cause an effect of stronger mental activation, as in the studies on English (Ferretti et al., Reference Ferretti, Kutas and McRae2007) and Farsi (Golshaie & Incera, Reference Golshaie and Incera2021)?
To investigate this question, we conducted a self-paced reading experiment with a design based on the third experiment reported in Ferretti et al. (Reference Ferretti, Kutas and McRae2007): the atypical location study. Based on the previous findings, we expected to observe the effect of stronger mental activation in the progressive function of imperfective Russian verbs. We hypothesized that reading times should be slower for atypical than for typical locations (as measured either on the locative noun itself or on the following word), and because of the effect of stronger mental activation, this effect should be larger in imperfective than in perfective sentences. Thus, we expected an interaction of the two factors: typicality of location and aspect.
2. Materials and methods
2.1. Participants
We conducted the experiment with 50 native Russian speakers living in Germany. They received compensation for their participation in the experiment.
Participants living in Germany were chosen for practical reasons. All respondents had knowledge of German, and some had native knowledge of other languages. Even though all of them were linguistically socialized in a Russian-speaking environment, we cannot completely rule out the possibility that linguistic transfer (with German being a language without grammaticalized aspect) or language attrition had some influence on the results. However, to minimize this possibility, we excluded two participants who rated their reading competence in Russian only at level 3 on a six-point scale (0 = no mastery; 5 = perfect mastery). Of the 48 participants remaining in the sample, 47 rated their reading competence at the highest level of 5, and one participant at the second highest level of 4. The participants in the final sample were between 18 and 63 years old (mean = 40.9; SD = 13.4; median = 43) and had attended a school with Russian as the language of instruction for between 8 and 12 years (mean = 10.4; SD = 1.0; median = 10; one missing answer). They had immigrated to Germany between 16 and 52 years of age (mean = 29.7; SD = 9.8; median = 30) and had lived in Germany for less than 1 to 31 years (mean = 11.2; SD = 11.7; median = 5).
2.2. Materials
The tested material included 32 experimental item sets, each consisting of four conditions and 52 filler items.
The experimental sentences were all constructed following the same scheme:
All sentences were initiated with a temporal adverbial consisting of a preposition and a noun referring to one day of the week, followed by the subject, which was always represented by a first name. At the fourth position, the predicate occurred in the imperfective (conditions 1 and 2) or perfective (conditions 3 and 4) form. The predicate was followed by a locative adverbial consisting of a locative preposition (consistent in each item set, mostly v ‘in’ or na ‘at, on’) and a noun expressing either a location typical for the described action (conditions 1 and 3) or possible but atypical for this action (conditions 2 and 4). Verb–location pairs were pretested in a typicality judgment task (conducted at SoSci-Survey) with 44 native speakers of Russian who judged the typicality of the two locations (one typical and one atypical for the action denoted by the verb) on a 1–5 Likert scale. In all verb–location combinations used in the experiment, the mean difference between the judgments for typical and atypical locations was at least 1.
The location adverbial was followed by a comma and the conjunction a ‘and’, and after this, either the temporal adverbial potom ‘then’ (perfective conditions 3 and 4) or a new subject (again a first name) occurred (imperfective conditions 1 and 2). Finally, the sentence was completed by another predicate and an object or instrument. In imperfective conditions 1 and 2, the sentences expressed parallel actions conducted by two different subjects. In such contexts, the imperfective form is supposed to express ongoingness, that is, the progressive function. Perfective conditions 3 and 4 presented sequences of events conducted by the same subject. This difference (same vs. different subjects) between the imperfective and perfective conditions was unavoidable because, otherwise, either conditions 1 and 2 or conditions 3 and 4 would not express the intended aspectual function. However, the second part of the sentence – especially the second verb and the object – was of minor interest for our analysis because our hypotheses were related to reading times for the location adverbial and the following (in all sentences identical) conjunction and for the predicate of the first clause.
Several criteria had to be fulfilled by the predicates preceding the location adverbial: As mentioned above, we included mostly atelic (23 verbs) or diffuse (6) verbs (according to Mende et al., Reference Mende, Born-Rauxeneker, Brjugeman, Dipong, Kukla and Lehman2011) with their imperfective simple forms and po-prefixed perfective partners because telic intransitive verbs did not fit in with the chosen sentence model.Footnote 5 No transitive verbs were considered because an additional object would interfere with the plausibility of the aspectual categories (Anstatt, Reference Anstatt2003). However, three of our verbs were reflexive (brit’sja ‘to shave oneself’, myt’sja ‘to wash oneself’ and krasit’sja ‘to put on makeup’) and telic (according to Mende et al., Reference Mende, Born-Rauxeneker, Brjugeman, Dipong, Kukla and Lehman2011). Finally, for the rationale of the experiment, the verbs’ semantics had to show an affinity to a certain location, given, for example, in molit’sja ‘to pray’, but not in dumat’ ‘to think’.Footnote 6
The perfective verbs inevitably differed from the imperfective verbs not only in their aspectual value; they contained one morpheme more (the prefix) and accordingly two, in one case (ždat‘ – podoždat’ ‘to wait’) four, letters more. In addition, the perfective verbs had a lower corpus frequency according to the RNC (accessed on 18/05/25) (ipm for imperfective verbs: 1.1–454.9; mean = 73.8; SD = 118.9; median = 20.3; ipm for perfective verbs; 0.0–48.3; mean = 7.4; SD = 11.8; median = 2.5; difference per pair: 1.0–439.4; mean = 66.4; SD = 110.2; median = 16.1).
In addition to the 32 experimental item sets, 52 filler items were created to prevent the participants from getting used to the atypical location adverbials. Filler sentences were semantically inconspicuous but of a similar length and macrostructure as the experimental sentences (e.g., Včera večerom Alena zanimalas’ dlinnym ėsse, no ne smogla ego zakončit.’ ‘Yesterday evening Alena dealt with a long essay, but couldn’t finish it.’). However, the initiating temporal adverbials and conjunctions varied.
Half of both experimental and filler sentences (43) were followed by comprehension questions (e.g., regarding the example sentences from Table 1: Jana molilas’ v cerkvi? ‘Did Jana pray in the church?’). In 50% of the cases, the expected answer was ‘yes’ and in 50% ‘no’. Comprehension questions with the expected answer ‘no’ that followed experimental sentences contained divergent locations or referred to the second clause of the sentence. ‘No’-questions that were preceded by filler sentences included divergent verbs or objects (e.g., regarding the example sentence above: Nina udalila ėsse? ‘Did Nina delete her essay?’). The purpose of these comprehension questions was to ensure that the participants paid attention throughout the experiment and did not particularly focus on the predicates or location adverbials. Finally, we prepared four experimental lists, each containing the 52 filler items and 32 experimental items but with each verb–location pair in only one of the four conditions. The complete sentence material is available in the repository zenodo: doi 10.5281/zenodo.14234472.
Examples of experimental sentences in conditions 1–4
2.3. Procedure
We conducted a self-paced reading experiment in a noncumulative linear format with word-by-word segmentation (moving window format) (Jegerski, Reference Jegerski, Jegerski and VanPatten2014). The participants were instructed in Russian and told that they would be reading Russian sentences presented word by word and that they would decide how fast they would read these sentences. Furthermore, they were asked to carefully read all sentences and to answer the questions occurring at irregular intervals after some of the sentences. Questions had to be answered by pressing a button.
After giving the instructions, the participants were presented with three practice trials. During the experiment, each participant read 84 items in pseudo-randomized order that were split into two blocks. The experimental procedure was programmed using Eprime 3.0 (Psychology Software Tools, Pittsburgh, PA). After the experiment, the participants were asked to complete a questionnaire concerning their language biography, visual capacities and other factors that might influence the results.
2.4. Data preprocessing
Following Jegerski (Reference Jegerski, Jegerski and VanPatten2014), outliers were removed in two steps. First, all responses with a RT of less than 200 ms or more than 3000 ms were removed. In a second step, the average and standard deviation for the RT per subject, condition and window were calculated for the remaining responses, and those that deviated more than 2 SD from the respective mean were excluded. The RT was measured for each window, but only two windows were analyzed statistically: window 6 (noun denoting the location) and window 7 (the conjunction a directly after the location). For these time windows, 1429 (93.0%) responses remained for the location, and 1437 (93.6%) for the conjunction.
2.5. Statistical methods
While the use of linear mixed-effects models (LMMs) has now become established as a method for analyzing psycholinguistic data, there is debate regarding the analysis of skewed RT data (Lo & Andrews, Reference Lo and Andrews2015). While the analysis of raw RT fits well with psychological models (the ‘mental chronometry’ approach, Posner, Reference Posner1978), they usually do not fulfill the statistical assumptions. Therefore, some recommend nonlinear transformations of RT (Baayen, Reference Baayen2008). This approach is problematized by Lo and Andrews (Reference Lo and Andrews2015), as the transformation distorts the ratio scale, making it less suitable for psychological models and therefore less suitable for answering the research question.
Lo and Andrews therefore recommend the use of generalized linear mixed models (GLMM), which do not require a transformation of the dependent variable and are not dependent on the statistical assumptions of linear models. We therefore calculated LMM of the raw, non-transformed data, the log-transformed data and GLMM and discuss the advantages and disadvantages in the Results section.
We calculated (generalized) linear mixed-effects models using the lme4 package (Bates et al., Reference Bates, Maechler, Bolker and Walker2015b) and lmerTest package (Kuznetsova et al., Reference Kuznetsova, Brockhoff and Christensen2017) in R (R Core Team, 2024) for the RT for two windows: the locational noun (window 6) and the word after the location (the conjunction a ‘and/but’, window 7) with subject (n = 48) and item (n = 32) as random effects and aspect (ipf. vs. pf.), typicality of the location (typical vs. atypical) and the interaction between aspect and typicality as fixed effects. To determine main effects, aspect and typicality were deviation-coded. Regarding the random slopes, we started with the full random effects structure following Barr (Reference Barr2013) and Barr et al. (Reference Barr, Levy, Scheepers and Tily2013), including random slopes for the aspect x typicality interaction. If these full models did not converge, we used principal component analysis (PCA) using the rePCA function in the lme4 package to detect whether these models were overparameterized. Following Bates et al. (Reference Bates, Kliegl, Vasishth and Baayen2015a), we first excluded the random slopes for the interaction and then planned to reduce the random effects structure after inspecting the variance of the random slopes until the models converged and did not result in singular fits and the PCA did not hint at overparameterization. However, all models converged at least after excluding the random slopes for the interaction. Significance of the interactions was tested by means of likelihood-ratio tests using the anova function in R. To analyze interactions, we calculated post hoc comparisons (contrasts) with the help of the emmeans package (Lenth, Reference Lenth2023) using the Bonferroni correction for multiple comparisons. The R code and raw data are available in the repository zenodo: doi 10.5281/zenodo.14234472.
3. Results
3.1. Task performance
The participants answered an average of 41.4 of the 43 decision questions correctly (36–43; median = 42; SD = 1.6).
3.2. Reaction time
Figure 1 shows the mean RT and confidence intervals for the mean (mean +/− 1.96 SE) for the first 10 windows for sentences with imperfective (a) and perfective (b) verbs, comparing sentences with typical (black) and atypical (red) locations.
Mean RT and confidence intervals for the first 10 windows for sentences with imperfective (a) and perfective (b) verbs, comparing sentences with typical (black) and atypical (red) locations. 4 is the position of the first predicate (V), 5 is the position of the location adverbial’s preposition (P), 6 is the position of the location adverbial’s noun (L), and 7 is the position of the conjunction (C) after the adverbial.
Table 2 shows the mean and standard deviation for the two critical windows:
Descriptive statistics
In window 6 (location noun) and in window 7 (conjunction), unexpected locations after perfective verbs were accompanied by the longest RT, followed by unexpected locations after imperfective verbs.
3.2.1. RT at the location (window 6)
We calculated three models: a) a LMM with the raw RT as the dependent variable, b) a LMM with the log-transformed RT as the dependent variable and c) a GLMM with the raw RT as the dependent variable, using the inverse-gaussian family. All three models showed main effects for aspect (model a: b = 54.0 ms, SE = 22.1 ms, t(30.68) = 2.45, p = 0.020; model b: b = 0.068, SE = 0.025, t(32.13) = 2.71, p = 0.011; and model c: see below) and for typicality of the location (model a: b = 127.9 ms, SE = 33.3 ms, t(44.29) = 3.85, p < 0.001; model b: b = 0.138, SE = 0.032, t(36.02) = 4.31, p < 0.001; and model c: see below) as well as an interaction between aspect and typicality (model a: χ2(1) = 7.38, p = 0.007; model b: χ2(1) = 4.36, p = 0.037; and model c: see below). However, model a violated the statistical assumptions: Residuals were not normally distributed and the assumption of homoscedasticity was not met, with larger residuals for higher fitted values. With model b, residuals were still not normally distributed, but visual inspection did not indicate a violation of the assumption of homoscedasticity. Model c did not rely on these assumptions and also allowed the full random effects structure to be included, whereas in models a and b, no random slopes for the interaction could be included since the models did not converge.
We therefore report model c, the GLMM using the inverse-gaussian family, which does not rely on the statistical assumptions, does not require a transformation of the original scale and has the most complex random effects structure, that is, random intercepts for subject and item and random slopes for the interaction of aspect and typicality for item and subject. We found main effects for aspect, with longer RT after perfective verbs (b = 57.1, SE = 18.65, t = 3.06, p = 0.002), and for typicality, with atypical locations causing longer RT than typical locations (b = 142.9, SE = 20.4, t = 7.00, p < 0.001). We could not directly compare this model with the same model without the interaction because the latter did not converge. However, we found a significant interaction term for aspect * typicality (b = 110.5 ms, SE = 26.4 ms, t = 4.18, p < 0.001), and the model was significantly better than a model without the interaction and without random slopes for the interaction (χ2(9) = 67.74, p < 0.001).
Post hoc comparisons showed that the response to an atypical location took longer than the response to a typical one both after perfective verbs (typical vs. atypical: b = −198.1 ms, SE = 27.0 ms, z-ratio(Inf) = −7.34, p < 0.001) and after imperfective verbs (typical vs. atypical: b = −87.7 ms, SE = 21.3 ms, t-ratio(Inf) = −4.11, p < 0.001). For a typical location, we found no difference depending on whether it appeared after an imperfective or perfective verb (ipf. vs. pf.: b = −1.9. ms, SE = 18.4 ms, z-ratio(Inf) = −0.10, p = 1.00). For an atypical location, the RT was longer after a verb in the perfective aspect than after a verb in the imperfective aspect (ipf. vs. pf.: b = −112.3 ms, SE = 26.5 ms, z-ratio(Inf) = −4.23, p < 0.001). Figure 2 shows the RT predicted by the model and confidence intervals for the four combinations of the two factors.
Predicted reaction times (fixed effects only) and confidence intervals for the location adverbial’s noun for the four combinations of the factors aspect and typicality of the location.
3.2.2. RT at the conjunction (window 7)
Again, we calculated three models: (a) a LMM with the raw RT as the dependent variable, (b) a LMM with the log-transformed RT as the dependent variable and (c) a GLMM with the raw RT as the dependent variable, using the inverse-gaussian family.
Model a included the full random effects structure, that is, random slopes for the interaction. In models b and c, this was not the case, that is, they included only random slopes for aspect and typicality for subject and item, respectively. All three models showed main effects for typicality of the location (model a: b = 27.4 ms, SE = 9.1 ms, t(36.65) = 3.02, p = 0.005; model b: b = 0.055, SE = 0.017, t(29.90) = 3.31, p = 0.002; and model c: see below), but no main effect for aspect (model a: b = 4.1 ms, SE = 9.6 ms, t(38.39) = 0.42, p = 0.67; model b: b = 0.013, SE = 0.018, t(34.73) = 0.74, p = 0.46; and model c: see below). Model a did not show an interaction between aspect and typicality (χ2(1) = 2.08, p = 0.15), but model b (χ2(1) = 6.29, p = 0.012) and model c (see below) did. However, model a violated the statistical assumptions: Residuals were not normally distributed and the assumption of homoscedasticity was not met, with larger residuals for higher fitted values. With model b, residuals were still not normally distributed, but visual inspection did not indicate a violation of the assumption of homoscedasticity. Model c did not rely on these assumptions.
We again report model c, the GLMM, having in mind that model a had the more complex random effects structure. The final model included random intercepts for subject and item and random slopes for aspect and typicality per item and per subject, but no random slope for the interaction of aspect and typicality. This model showed a main effect for typicality, with atypical locations causing longer RT than typical locations (b = 26.13 ms, SE = 11.12 ms, t = 2.35, p = 0.019), no main effect for aspect (b = −5.61 ms, SE = 11.70 ms, t = −0.48, p = 0.63), but an interaction between aspect and typicality (χ2(1) = 9.52, p = 0.002).
For a typical location, contrasts showed no difference depending on whether it appeared after an imperfective or perfective verb (ipf. vs. pf.: b = 18.5 ms, SE = 12.3 ms, z-ratio(Inf) = 1.51, p = 0.79). There was also no difference between atypical locations after imperfective and perfective verbs (ipf. vs. pf.: b = −7.3 ms, SE = 12.5 ms, z-ratio(Inf) = −0.58, p = 1.00). However, the response to an atypical location took longer than the response to a typical one after perfective verbs (typical vs. atypical: b = −39.0 ms, SE = 11.9 ms, z-ratio(Inf) = −3.27, p = 0.006), but not after imperfective verbs (typical vs. atypical: b = −13.2 ms, SE = 11.8 ms, z-ratio(Inf) = −1.12, p = 1.00). Figure 3 shows the RT and confidence intervals predicted by the model for the four combinations of the two factors.
Predicted reaction times (fixed effects only) and confidence intervals for the conjunction after the location adverbial for the four combinations of the factors aspect and typicality of the location.
4. Discussion and conclusion
We conducted a self-paced reading experiment with 48 native speakers of Russian to gain insights into the effect of Russian verbal aspect on event processing, more precisely on processing typical versus atypical locations preceded by perfective versus imperfective verbs. We expected reading times to be slower for atypical locations, and because of the hypothesized effect of stronger mental activation (Ferretti et al., Reference Ferretti, Kutas and McRae2007; Golshaie & Incera, Reference Golshaie and Incera2021), this effect was predicted to be larger in imperfective sentences.
Directly on the noun denoting the location, we found a significant interaction of the two factors typicality (of the location) and aspect. While atypical locations were processed slower both after imperfective and after perfective verbs, contrary to our expectations, the slowing effect of atypical locations was larger for perfective than for imperfective sentences. For the window directly after the noun (the conjunction a in all conditions), we found longer RT after atypical locations only after perfective verbs. The effect of atypical locations was thus stronger and lasted longer after perfective verbs.
As we developed the design of our study using Ferretti et al. (Reference Ferretti, Kutas and McRae2007), who observed an effect for the English progressive, we will now first discuss possible explanations for the missing effect of stronger mental activation caused by the imperfective in our data on Russian: One reason may be that we were not successful in eliciting the progressive function of the imperfective aspect with our sentence material. As we explained in Section 2.2, constructing sentences that worked equally well with the imperfective and with the perfective form of the verb was anything but easy. For the imperfective condition, we decided to build sentences that could express temporally parallel actions with two imperfective verbs and two different subjects (V pjatnicu Jana molilas’ipf v cerkvi, a Oleg vygulivalipf sobaku. ‘On Friday Jana prayed in the church, and Oleg took the dog out for a walk.’). For the perfective condition, we used the same verbs and built sequences of events that shared the same subject (V pjatnicu Jana pomolilas’pf v cerkvi, a potom vyguljalapf sobaku. ‘On Friday Jana prayed in the church and then took the dog out for a walk.’). When the whole sentence is considered, these two functions (parallel action and sequence of events) can be unambiguously differentiated. However, the effect of stronger mental activation was expected to appear at the sixth word of the sentence: the location adverbial. When this (typical or atypical) location occurred, the participants were not yet aware of the second part of the sentence. Theoretically, the sentence in the imperfective condition might have continued with, for example, neskol’ko raz ‘several times’, which would also be congruent with the imperfective form of the verb but in its iterative function. By initiating the sentence with the temporal adverbial ‘on Friday’ (and other days of the week), we tried to prevent the readers from interpreting the imperfective verb as expressing a repeated action (the chosen construction unambiguously refers to a single day, cf. the iterative po pjatnicam ‘on Fridays’) or generally stating a fact not assigned to a unique point of time (general-factual, see above). However, this might not have worked out for all participants. Typically, the progressive function of an imperfective verb is accompanied by more explicit markers of ‘ongoingness’ (adverbial phrases, conjunctions, etc.), which was not an option because we had to use the same sentence material (at least the first part) in the perfective condition as well. With a follow-up experiment using a different design, we might be able to verify whether explanation 1 holds true, that is, whether, in a more unambiguous progressive context, the effect of stronger mental activation would occur.
A second explanation for the missing effect from a theoretical standpoint would be that the progressive function of the imperfective aspect has been successfully elicited, but the effect of stronger mental activation does not exist for Russian imperfectives in the progressive function. Finally, there is a third possible explanation that assumes that the progressive function has been successfully elicited and that the effect does exist but that it is superimposed by a stronger effect caused by the po-prefixed pf. verbs. Because explanations 2 and 3 are interconnected with the interpretation of the observed effect of the perfective aspect, we now turn to the discussion of this main result.
In the main, we argue that the observed effect, that is, longer reading and processing times for atypical locations when preceded by perfective verbs, is caused by the functional complexity of the perfective verbs. In short, we argue that functionally less specialized aspectual categories connect better with unusual features of the situation than more specialized ones. As stated in Section 1.2, functional complexity relations are complementarily distributed in English and the mostly atelic and diffuse Russian verbs used in our study.
In English, the progressive is functionally more specialized than the nonprogressive, but in Russian, the perfective is more specialized than the imperfective. The Russian perfective is associated with the basic meaning of temporal definiteness, which implies the sequentiality of events, whereas the imperfective is characterized ‘negatively’, thus merely by the absence of temporal definiteness (‘qualitative temporal indefiniteness’, and thus simple reference to the type of situation; Dickey, Reference Dickey2000, Reference Dickey2024). The imperfective has a much broader spectrum of very different contextual functions; thus, it is impossible to formulate a basic meaning that includes all these functions. In this respect, the Russian imperfective is similar to the English nonprogressive.
From a pragmatic point of view, this distribution implies a different number of inferencesFootnote 7 that are triggered in the reader when processing a Russian perfective versus imperfective verb. We have already mentioned that the perfective is associated with an expected sequence of events. Moreover, following Dickey, the use ‘of a great many instances of the Russian perfective’ (2024: 11) includes inferencing the action’s goal as a crucial element in event construal, and this is ‘relevant primarily for agentive resultative verbs; however, it can also be relevant for others, for example delimitatives in po-’ (Dickey, Reference Dickey2024, p. 11). Finally, De Wit and Dickey (Reference De Wit, Dickey, De Wit, Brisard, Madden-Lombardi, Meeuwis and Patard2024) outlined a cross-linguistic approach to aspect based on the idea that the essence of the perfective is the speaker’s full knowledge of the denoted situation at speech time. Thus, we are aware of at least three concrete inferences that come along with the use of a Russian perfective verb: sequentiality of events, goal orientation and full knowledge of the situation.
The basic semantic content of the Russian imperfective (without further context) is argued to be essentially the type-specification of the situation denoted. Hence, its use is motivated by the desire to avoid the perfective event construal and the inferences arising from it (Dickey, Reference Dickey2024).
Based on Dickey’s approach, we propose that inferences brought into the situation by the perfective aspect of the predicate cause expectations about the continuation of the sentence in the reader’s mind. Processing of a perfective verb would trigger the expectation that the sentence will focus on the action’s goal or continue with a subsequent event.
We would argue that typical locational information is congruent with this expectation since typical locations just make explicit the location the reader had already activated as a default location for the action described by the verb and are not in conflict with the expectation regarding the relevance of the event’s goal. However, in the atypical condition, the expectation triggered by the perfective verb causes a profound irritation. Atypical locational information must pragmatically be understood as being the central information of the sentence, which is in conflict with the expectation elicited by the perfective aspect. We thus argue that locational information is not in general less expected after perfective verbs, but only such information that due to its atypicality clashes with the expectation regarding the pragmatic prevalence of the event’s goal.Footnote 8 Furthermore, if the imperfective involves a foregrounding of the situation (activity, event) itself (Dickey, Reference Dickey2024), details about atypical circumstances of the situation should be not unexpected with the imperfective aspect. This might explain the observed significant effect of longer reading times for the atypical location adverbial when preceded by a perfective predicate. However, as this argumentation is speculative, it needs to be evaluated in future research with an appropriate experiment.
Finally, a drawback of our study is that the majority of our experimental verbs were atelic or diffuse regarding their situation type, and thus, the results and interpretation are limited to this class of verbs. This is important because functional complexity at the lexico-morphological level, just like formal complexity, depends to a certain degree on telicity. Dickey’s postulated basic meanings for the East Slavic perfective and imperfective are independent from situation type and refer to the functioning of perfective and imperfective verbs at the discourse/textual level. It is undisputed that, at this level, all perfectives, whether they are telic or atelic, are functionally more specified than imperfectives. However, at the lexico-morphological level, functional complexity depends on telicity: Following Lehmann’s theory of alpha and beta verbs (1993, 2010), each aspectual relationship consists of one alpha verb, which is congruent with respect to situation type and grammatical aspect (perfective telic verb, e.g., otkryt’ pf ‘to open’, and imperfective atelic verbs, e.g., plakat’ ipf ‘to cry’) and at least one beta verb, which is not congruent in this respect (imperfective telic verbs, e.g., otkryvat’ ipf ‘to open’, and perfective atelic verbs, e.g., poplakat’ ‘to crypf’). Nowadays, there is broad empirical and cross-linguistic evidence for this differentiation: Alpha verbs have been shown to precede beta verbs in language acquisition (Stoll, Reference Stoll1998, Reference Stoll2001; Gagarina, Reference Gagarina2000, Reference Gagarina, Bittner and Gagarina2004; van Hout, Reference Hout, Lidz, Snyder and Pater2016), to be retained longer in the mental lexica of bilinguals (for Russian, cf. Clasmeier, Reference Clasmeier, Fischer, Krumbholz and Lazar2010; Polinsky, Reference Polinsky, Corbett and Noonan2008) and to be more frequent in corpora (Anstatt & Clasmeier, Reference Anstatt and Clasmeier2012, Reference Anstatt and Clasmeierin prep.; Lehmann, Reference Lehmann1993).
Based on these theoretical assumptions, functional aspectual complexity on the lexico-morphological level is higher for beta verbs because they combine incongruent lexical and grammatical features: Atelic verbs continue to be atelic, that is, without inherent limit of the action (e.g., crying), in the perfective form (cf. poplakat’) because telicity is a lexical feature that does not vary with grammatical changes. However, the perfective form adds an ‘external’ limit to the action. On the other hand, telic verbs continue to be telic in the imperfective form (cf. otkryvat’); however, the initially single-phase event (e.g., of opening) is secondarily split up into multiple phases so that it may be used in the progressive function as well.
Returning to our study, the majority of the po-prefixed perfectives of the atelic or diffuse verbs we used were beta verbs, thus being functionally more complex than their imperfective partners at the sentence and at the morphological levels.Footnote 9 In the next step, these two levels should be differentiated to gain insights into their respective impact on event processing. Therefore, in a follow-up experiment, we would like to investigate the potential differences between perfective beta verbs being functionally complex at both levels and perfective alpha verbs being morphologically less complex than their imperfective partners.
Another drawback of our study is that the frequency of perfective verbs was lower than that of imperfective verbs. We did not control for frequency, as we expected the effect of atypical locations to be in the other direction, that is, a stronger effect for the more frequent verbs. An alternative explanation would therefore be that the processing of typical locations is not influenced by the aspect of the verb, but that atypical locations are more difficult to process after less frequent verbs than after more frequent verbs. Even if this seems less plausible to us, we cannot rule out this possibility. In conclusion, the present study contributes to the growing body of aspect and event conceptualization research, demonstrating the value of including languages like the Slavic that show aspectual systems of other types than the English progressive/nonprogressive. Cross-linguistic comparison and thorough differentiation of formal and functional factors influencing speech processing reveal a complete and more appropriate picture of how such a complex grammatical category as verbal aspect impacts event conceptualization.
Acknowledgements
We thank Vladimir Arifulin, Maria Stoll and Sviatlana Tesch for helping us with constructing the sentence material and Natascha Knjasew and Maria Stoll for carrying out the experiment. We would like to thank the two anonymous reviewers for their in-depth engagement with our study and for their numerous, highly valuable comments on earlier versions of this paper.
Competing interests
The authors declare none.
Open access
