Cross-linguistic influence in the processing of aspect in L2 English: Slavic, but not Norwegian L1 speakers associate past simple with completion

1 Introduction

The processing of the second language (L2) is different from L1. An important role in how people acquire and process an L2 is attributed to the influence from L1. When it comes to morphosyntactic categories, availability of synonymous, transferrable structures in the L1 can help predict how analogous L2 structures are acquired and processed (e.g., Kim, 2025). Difficulties associated with non-transferrable structures are related to the necessity to internalize and learn to process a conceptually new category. With transferrable structures, issues related to competition arise, as one is tempted to map the features of an overlapping L1 structure onto the synonymous L2 structure which is not always a felicitous strategy as the features may differ.

To date, most of cross-linguistic influence (CLI) literature has been focused on the acquisition and processing of L2 morphosyntax. Less is known about CLI effects in the semantic processing of grammatical categories, such as verbal aspect. Aspect is used to convey a particular viewpoint on how an event unfolds, such as whether it is ongoing or has reached its end-point. The level of nuance associated with aspectual forms' underlying semantic representations is highly language-specific and is not outright accessible to L2 learners. As a result, appropriate usage of aspect is one of the most challenging tasks in L2.

One of the reasons for non-native-like aspect usage might be related to the mental representations of the event's temporal structure that L2 speakers (sub)consciously attribute to (non-)aspectual verb forms. To date, however, no studies have explicitly probed the online semantic interpretation of aspect in L2.

To fill this gap, we designed a study that investigated whether L1 aspectual semantics affects aspectual interpretation in L2. In the first experiment, two groups sharing an L2 (English) were selected based on the presence or absence of grammaticalized aspect in their L1: aspectual Russian and non-aspectual Norwegian. In a visual world paradigm (VWP) eyetracking experiment, participants saw ongoing and completed versions of accomplishment events while listening to their English descriptions with a past progressive or past simple verb. Online gaze preferences and offline picture selections were recorded. In the second experiment, native speakers of aspectual Polish and non-aspectual Norwegian sharing L2 English participated in an improved version of the VWP experiment. In addition to gaze preferences and picture selections, we collected quantitative information about these participants' L2 proficiency and everyday exposure to L2 English. In both experiments we asked whether L2 speakers' semantic interpretation of aspectually unmarked English past simple is affected by L1 aspect, both online and offline.

2 Aspectual interpretation in L1 English, Russian, Polish, and Norwegian

Grammatical aspect specifies the temporal perspective on an event denoted by the lexical verb. Unlike tense, aspect is not concerned with the time of the event's occurrence. Although it interacts with tense, aspect is a separate category. The most basic aspectual distinction is between perfective and imperfective, where the former conveys an external view of the event, with an emphasis on its completion or cessation; and the latter corresponds to an internal view, focusing on the event's intermediate stages rather than the completion status (Comrie, 1976). The contrast between these two viewpoints is most distinctly realized in predicates describing events with an inherent end-state and distinct stages preceding it (“accomplishments,” see, e.g., Vendler, 1957; Smith, 1991; Becker et al., 2013). Slavic languages (Russian and Polish), Norwegian, and English differ in how and to what extend they employ grammatical means to encode aspectual contrasts.

In Russian (East Slavic) and Polish (West Slavic), verbs are grammatically marked as either perfective or imperfective (1a-b, 2a-b). In episodic (i.e., non-habitual) contexts with a specific reference time point in the past, imperfective verbs typically communicate the semantics of events in progress while perfective verbs convey the meaning of event completion or cessation (Maslov, 1948; Forsyth, 1970; Dahl, 1985; Zaliznjak and Šmelev, 2000, Timberlake, 2004).

• 1. (a) Russian: Babuška pekla
• Grandma bake_{.*IMP*−*PST*}
• pirog. Imperfective past
• cake
• “Grandma was baking a cake.”
• (b) Russian: Babuška ispekla pirog. Perfective past
• Grandma bake_{.PFV − PST} cake
• “Grandma baked a cake.”
• 2. (a) Polish: Babcia piekła ciasto. Imperfective past
• Grandma bake_{.*IMP*−*PST*} cake
• “Grandma was baking a cake.”
• (b) Polish: Babcia upiekła ciasto. Perfective past
• Grandma bake_{.PFV − PST} cake
• “Grandma baked a cake.”

There are certain distinctions in the functioning of aspectual categories between Slavic languages (Dickey, 2000). A salient contrast between Russian and Polish is that Russian, but not Polish, uses imperfective to refer to (possibly completed) events at an indefinite timepoint in the past [the so-called “general factual” or “experiencial” reading, see (Grønn 2004)], as illustrated by the following examples from Dickey (2000):

• 3. (a) Russian: Ty kogda-nibud^j spotykals^ja na ulice?
• you ever trip_{.IMP − PST} on street
• “Have you ever tripped on the street?”
• (b) Polish: Czy kiedykolwiek potkna̧łeś/^*potykałeś
• if _. ever trip_{.PFV − PST}/trip_{.IMP − PST}
• siȩ na ulicy?
• self on street
• “Have you ever tripped on the street?”

Studies of aspectual processing by native speakers of Slavic languages, such as Russian and Polish, reveal striking differences in the temporal interpretation between perfective and imperfective verbs. Klimek-Jankowska and Błaszczak (2020) used ERPs to examine the processing of matches and mismatches between perfective and imperfective predicates in Polish and two types of temporal adverbials (for a minute vs. in a minute) that were either sentence-initial or post-verbal. In sentences that began with adverbials, only the perfective aspect triggered a mismatch effect. For post-verbal adverbials, mismatches elicited effects with perfective and imperfective. The presence of an effect only with the perfective in the sentence-initial adverbial condition is likely explained by the more restricted contextual distribution of the perfective as opposed to the imperfective. While the tested perfective verbs are incompatible with adverbials like for a minute in any context, imperfective verbs can combine with in a minute-type adverbials in some contexts such as the habitual. Similarly, Russian past perfective verbs are universally incompatible with habitual temporal adverbials (e.g., every Thursday), and this combination elicited a neural cost (P600) in Zeller and Clasmeier (2020), signaling an overt grammatical violation. Additionally, in a picture-sentence matching task, Russian perfective verbs elicited a P600 in sentences that followed pictures of events presented at a mismatching stage (in-progress, for perfective) (Kamenetski and Lai, 2025). Minor et al. (2022) used visual-world eye tracking to test aspect processing by Russian-speaking adults. Controlling the context of the test sentences to refer to singular episodic events at a specific time point in the past, they found that perfective verbs triggered an immediate strong preference for completed event representations, while imperfective verbs triggered an equally strong and immediate preference for ongoing events.

English grammaticalizes the distinction between progressive and non-progressive (3a–b):

• 4. (a) Grandma was baking a cake. Progressive
• (b) Grandma baked a cake. Non-progressive

This distinction overlaps with, but is not equivalent to, the distinction between perfective and imperfective aspect in Slavic languages. The English progressive has a narrower distribution than the Slavic imperfective. It is used to refer to events ongoing at a specific time point but is incompatible with stative predicates and is not used to convey the habitual interpretation, in contrast to the Slavic imperfective. The simple past has been argued to trigger a completed event interpretation when combined with telic predicates as in (3b), similar to the Slavic perfective. On the other hand, the simple past can also be used in combination with atelic activity predicates to refer to events that were ongoing for a period of time (e.g., Mary danced for a while) and with stative predicates (e.g., Mary lived in London), which are typically incompatible with perfective aspect in Slavic languages.

Experimental studies of English aspectual interpretation have yielded mixed findings. In an early study, Madden and Zwaan (2003) had English native speakers either respond to pictures of ongoing or completed events after reading past simple or progressive sentences, or read the (past simple or progressive) sentences after looking at such pictures. Faster response latencies in both experiments suggested that participants associated past simple with completed, rather than ongoing events. Progressive did not affect response latencies.

Contrasting evidence comes from Minor et al. (2022), 2023) who used a VWP to compare online aspectual interpretation in Russian (imperfective vs. perfective), English (past progressive vs. past simple), and Spanish (imperfect vs. preterit). Spanish grammaticalizes the aspectual distinction in the past tense contrasting the imperfect form with the perfective preterit. Unlike Slavic perfective, Spanish preterit conveys the meaning of event cessation, but does not necessarily entail that the event reached its result state. Pragmatically, this can implicate event completion with telic predicates. L1 speakers listened to imperfective (progressive) and perfective (preterit/past simple) sentences describing telic events in their respective L1s while looking at pairs of pictures representing ongoing and completed versions of the event. In all three groups offline judgments and online gaze fixations indicated an overwhelming preference for ongoing events when listening to imperfective (progressive) sentences. Differences emerged in the perfective (preterit) condition. Russian speakers associated perfective with completed events both offline (95%) and online, where a clear preference for completed events was significant between 550 and 3,000 ms post verb onset. In Spanish, completed events were also preferred in the preterit condition, but the magnitude was smaller offline (84%), while online gaze preferences for completed events reached significance around 750 ms post verb onset. Finally, English native speakers showed no preference either offline (54% completed events) or online in the past simple condition. The absence of online preference for completed events with English past simple has been replicated in the same paradigm using web-based eyetracking by Vos et al. (2022).

These findings are in line with Jeschull (2007) who tested English-speaking children and adults on a sentence-video matching task. The participants saw two videos representing different characters performing the same action (e.g., drinking a bottle of coke). One of the characters completed the action (e.g., drank the whole bottle), while the other did not. The participants were then asked a question involving either a past simple verb (Who drank his coke?) or a past simple particle verb (Who drank his coke up?). The study found that adult speakers selected the incomplete event video 51.1% of the time when the question included a past simple verb without a particle, but only 9.1% of the time when the question involved a particle verb. This suggests that the past simple form on its own does not induce a strong preference for completed events. In a slightly different design, Arunachalam and Kothari (2011) showed adult English speakers videos of either completed or incomplete telic events (e.g., an actor eating all of or half of the cookie), and asked them to judge a description of the video as true or false. All descriptions involved a past simple verb (e.g., ate a cookie). Participants accepted the descriptions as true 97% of the time for fully completed events, and 46.9% of the time for partially completed events, suggesting that past simple forms of telic predicates do not necessarily entail event completion. The response pattern observed for English past simple was similar to what the study found for the Hindi simple perfective form, which has been argued not to entail event completion (Singh, 1991, 1998).

Misersky et al. (2021) used ERPs to investigate the effect of English aspect on mental representations of object states. English native speakers read sentences describing accomplishment events with past progressive and past simple predicates (e.g., The chef chopped/was chopping the onion). Each sentence was followed by either a picture of the object in an initial or affected state (e.g., a whole onion or a chopped onion) or an unrelated object. An exploratory analysis revealed an effect time-locked to the presentation of an affected object after reading past simple sentences compared to past progressive, suggesting that mental representations triggered by past progressive descriptions were a worse match for affected states compared to past simple descriptions. In a similar vein, Ferretti et al. (2009) investigated coreferential processing following sentences describing transfer-of-possession events in English past progressive (e.g., was giving) or past simple (e.g., gave). In an ERP study, past simple activated the goal of the transfer more than past progressive (e.g., Jeremy in Caroline pitched a ball to Jeremy). This finding was corroborated by the story continuation task where participants continued past simple sentences with a reference to the goal of the transfer event more often than past progressive sentences.

In sum, all the discussed results are consistent with the English past simple being more compatible with representations of completed events/result states compared to the past progressive. However, findings differ in the extent to which past simple is also compatible with ongoing/incomplete event representations. These differences can be at least partially attributed to methodological considerations. For example, different cognitive processes are ostensibly involved in aspectual interpretation while looking at two possible interpretations of the event (VWP) as opposed to the sentence-picture matching paradigm. Choice of predicates may also play a role: both Minor et al. (2022) and (Arunachalam and Kothari 2011) report large item-level variation in response patterns. Nevertheless, most relevantly to our study, direct comparison between English past simple and Slavic perfective suggests that Slavic perfective verbs are more closely associated with event completion and result state (Minor et al., 2022, 2023).

Finally, Norwegian does not grammatically encode aspectual distinctions on its verbs, employing a single preterit form in the past tense:

• 5. Norwegian: Bestemor bakte en kake. Preterit
• grandma bake_.PST a cake
• “Grandma baked a cake.”

In contrast to English, Norwegian does not possess a grammaticalized progressive form. Instead, lexical and periphrastic means have to be used to specifically convey the progressive meaning, e.g.,:

• 6. (a) Bestemor holdt på å bake en kake.
• grandma hold_.PST on to bake a cake
• “Grandma was baking a cake.”
• (b) Bestemor var i gang med å bake en kake.
• grandma be_.PST in process with to bake a cake
• “Grandma was baking a cake.”

Empirical investigations of tense and aspect processing in Norwegian are extremely limited. Unpublished data from our research group, however, suggests that Norwegian preterit may be even less associated with event completion than English past simple. In an unpublished VWP study using an identical design and visual stimuli as in Minor et al. (2023) and Vos et al. (2022), Norwegian native speakers (n = 23) listened to 24 Norwegian sentences describing events either using a verb in the preterit form or a periphrastic progressive construction (holdt på å + verb). Results indicated that in both conditions participants preferred ongoing events when selecting the best matching picture (periphrastic progressive: 99.6% ongoing events, preterit: 72.5% ongoing events), and this preference was also reflected in the gaze patterns (periphrastic progressive: significantly more looks to ongoing events 950–3,000 ms after verb onset; preterit: significantly more looks to ongoing events 1,700–2,650 ms after verb onset).

3 Grammatical aspect acquisition and processing in L2

L2 aspect acquisition research can be broadly divided into functional and formal approaches. Functional approaches focus on how L2 learners establish aspectual distinctions. One of the most prominent views in this framework follows the principles of the Aspect Hypothesis (AH) (Andersen and Shirai, 1994), which highlights the effect of inherent predicational aspect (e.g., González, 2013).

The concept of inherent predicational aspect is based on the aspectual features of an entire predicate (e.g., to eat an apple) rather than the lexical verb alone (to eat). Predicates can be categorized in terms of economical two-way distinctions based on temporal distributions associated with them. For example, in a study of the perfective-imperfective distinction in L2 Spanish, native speakers of (non-aspectual) Dutch relied on the terminative–durative inherent aspectual distinction (González and Quintana Hernández, 2018). Terminative (or telic) predicates are those referring to discrete events with inherent boundaries (e.g., to eat an apple). Durative (or atelic) predicates refer to atelic events without a clear completion boundary (e.g., to eat apples). In contrast, Domínguez et al. (2013) found that L1 English speakers learning L2 Spanish were more likely to ground their aspect use in event dynamicity: while dynamic, non-stative predicates (e.g., to eat an apple/apples) were more likely to be used in perfective, the non-dynamic, stative ones (e.g., to like apples) were more frequent in imperfective. Nevertheless, in addition to this pattern, L1 English speakers overused perfective in both Domínguez et al. (2013) and (González and Quintana Hernández 2018) as was predicted for an unmarked, more frequently used form to which learners were likely exposed outside of the classroom.

Input frequency effects in L2 acquisition were also pointed out by Wulff et al. (2009): in L2 English acquisition, verbs first learned in progressive are both atelic and frequently used in progressive by L1 English speakers. Likewise, verbs first learned in past simple are typically telic and most frequently used in past simple (p. 366). These observations, however, ignore the (potential) effect of L1 transfer. In a cloze task used by Izquierdo and Collins (2008), English L1 speakers overused perfective in L2 French and relied on predicational aspect more than Spanish L1 speakers, who transferred their L1 aspectual concepts and distinguished French perfective and imperfective even in non-prototypical contexts.

In sum, existing findings point to the role of an event's inherent temporal contour in L2 aspect acquisition as well as provide evidence for CLI from the L1, particularly in cases of substantial overlap between L1 and L2 aspectual systems.

While the AH generally does not focus on L1 influence, this is not the case for all function-focused approaches. For example, von Stutterheim et al. working within Slobin's Thinking-for-Speaking framework (e.g., Slobin, 1996), have emphasized the importance of L1 grammaticalized morphology for L2 event conceptualization. The underlying mechanism is related to the sensitization to the function of grammaticalized categories (e.g., Von Stutterheim, 2003). As grammaticalized categories are obligatorily used, language users are sensitized to the phenomena they denote. For example, an aspectual L1 speaker would be sensitized to phasal qualities of events which affect aspect choice. By the same token, when certain categories are not grammaticalized, native speakers are not sensitized to corresponding phenomena, even if the distinction can be conveyed lexically. The key difference is that lexical distinctions are optional, while grammatical ones are obligatory.

Within this approach, the influence of grammaticalization in L1 is taken beyond mere use of aspectual forms. Instead, the focus is on more global consequences of habitually attending to specific characteristics of events. For example, (Schmiedtová and Sahonenko 2008) compared how L1 Czech and L1 Russian speakers conceptualized visually presented events in non-aspectual L2 German. Although both Czech and Russian grammaticalize perfective and imperfective, they differ subtly in certain contexts. This difference is most obvious in the use of aspectual forms in the present tense to describe events that progress toward a culmination, but are not seen reaching it. While L1 Russian speakers tended to describe such events using imperfective verbs, L1 Czech speakers used perfective most of the time. These differences may be related how native speakers of these two languages conceptualized events in L2 German. Specifically, L1 Czech speakers mentioned end-points of events in German, taking a holistic, “perfective” perspective, more often than the L1 Russian group.

Based on this approach, we expect that L1 speakers of aspectual languages would pay more attention to the way temporal characteristics of events are encoded in the L2 compared to speakers of non-aspectual languages. L1-like event construal is predicted to persist even in advanced L2 speakers, as building a new sensitivity, or overcoming sensitivity that one has been accustomed to since childhood, is an extremely challenging cognitive task. Specifically, when acquiring the simple past vs progressive distinction in L2 English, L1 Polish and Russian speakers would be expected to map perfective and imperfective onto past simple and progressive forms of telic predicates, respectively, and persist in this mapping even at more advanced levels of L2 mastery. Norwegian L1 speakers, on the other hand, whose L1 does not grammaticalize aspect, may be less sensitive to the semantic distinction between the English past simple and progressive forms. Evidence from L2 English learners with non-aspectual L1 (e.g., German) shows that English progressive can be successfully acquired and used to describe ongoing events at higher levels of proficiency, likely due to its high frequency, regularity and transparency (e.g., Yoshioka and Hilberink-Schulpen, 2012; Von Stutterheim and Carroll, 2006; Von Stutterheim, 2003). Therefore, difficulties with associating progressive with ongoing events would not be expected for highly proficient learners. However, Norwegians could adhere to interpreting the English past simple similarly to the Norwegian preterit, gravitating toward ongoing events.

Formal approaches to CLI focus on the acquisition of overt morphological marking and the development of learners' knowledge underlying the use of such forms. Much of the discussion focuses on L1 influence and the transfer of its properties. Within this approach, some work takes a generalist position: the very availability of aspect as a grammatical category in L1 predicts its acquirability in L2, reflected in both online and offline measures. In comparison, when grammatical aspect is not available in L1, its full acquisition in L2 is more problematic. (Roberts and Liszka 2013, 2021) investigated the influence of aspectual and non-aspectual L1s on aspect-related processing in L2 English. In both studies native speakers of aspectual French demonstrated higher sensitivity to English aspectual properties during online sentence processing, in comparison to non-aspectual German native speakers. In fact, German native speakers demonstrated metalinguistic knowledge of aspectual distinctions in offline judgments (Roberts and Liszka, 2013), but no online sensitivity to aspectual manipulations (Roberts and Liszka, 2013, 2021). Although grammatical aspect in French and English is configured differently, successful aspectual processing by the French group is hypothesized to stem from “preconfigured” sensitivity to aspectuality due to their experience in processing grammatical aspect in their L1.

Within this view, Polish and Russian native speakers, by virtue of having grammaticalized aspect in their L1, could have an advantage in processing English aspect, in comparison with Norwegians. This would result in more target-like processing of both forms by the Polish and Russian groups both online and offline. The Norwegian group, in contrast, might not be sensitive to English aspect online, yet still demonstrate metalinguistic knowledge in their offline performance.

The Feature Reassembly Hypothesis (FRH) introduced a more nuanced approach to L1 transfer (Lardiere, 2008, 2009). FRH suggests that the acquisition of L2 functional morphology and associated features is supported by the process of reconfiguration of syntactic features that are shared by L1 and L2. The situations of structural overlap between L1 and L2 functional morphology represent the main locus of difficulties associated with L2 acquisition, particularly when similar functional categories are configured differently in the two languages. In this case, the learner has to disassociate what had been assembled in their L1 and reassemble it in a new way that matches the L2. This process can involve both adding and removing features.

Within the FRH framework, prior work has pointed out an asymmetry in the degree of difficulty of adding and removing features in overlapping functional categories. Guo (2022) investigated the acquisition of grammatical aspect in L2 Mandarin Chinese by L1 English speakers. The results of a battery of offline measures indicated that advanced L2 learners successfully added semantic features to overlapping functional categories, but struggled with discarding features transferred from the L1 (see Slabakova, 2009 for a discussion of the FRH and scenarios of grammatical acquisition that are predicted to be difficult vs. easy in terms of feature reassembly).

In sum, empirical evidence suggests that discarding features is more challenging than adding them. Therefore, the FRH predicts that the Polish and Russian groups would map perfective aspect onto English past simple of telic verbs and subsequently struggle with disassociating it from the restrictive perfective semantics linked to event completion. The Norwegian group, by contrast, may not experience substantial difficulties with the acquisition of broad temporal semantics associated with past simple by way of mapping it onto the Norwegian preterit. Following our unpublished findings, though, Norwegians might transfer the association of preterit with ongoing events. The acquisition timeline would thus require this L1 group to disassociate past simple from ongoingness.

Regardless of whether one assumes a formal or functional view of L2 aspect acquisition, the main contrast in terms of predictions for our study is concerned with whether L1 aspect influences aspectual interpretation in L2, and if so, how. From the point of view of the AH, L1 would not play a role. Instead, prototypical associations between predicational and grammatical aspect would predict the patterns in L2 speakers, stronger at lower L2 proficiency levels and weaker with increasing proficiency. On the other hand, views advocating for L1 influence would predict distinct interpretation patterns between groups with different L1 aspectual systems, with L1-like patterns weakening as L2 proficiency increases. Furthermore, approaches supporting L1 influence predict a possible discrepancy between online and offline behavior. Offline, a more target-like behavior in both groups might reflect metalinguistic knowledge of semantic interpretation associated with aspectual forms in English, while online behavior might reveal a more persistent L1 influence. More generalist approaches to CLI would predict that the Slavic L1 groups, having life-long experience with aspectual distinctions in their L1s, would be more successful than L1 speakers of aspectless Norwegian in replicating the target-like aspect processing patterns in L2 English, both online and offline. Finally, the FRH would make a more nuanced prediction whereby Russian and Polish L1 speakers may persist in associating past simple forms of telic predicates with event completion even at higher levels of English proficiency. In contrast, Norwegian L1 participants would map the Norwegian preterit form onto the English past simple, and tend to associate it with ongoing events, in line with their preference in the L1.

4 Current study

To investigate the role of L1 in aspectual interpretation of aspectually ambiguous English past simple, we collected online and offline measures using a web-based eye tracking VWP experiment combined with a sentence-picture matching task. Additionally, we collected L2 proficiency and exposure measures from native speakers of aspectual Slavic languages (Russian and Polish) and non-aspectual Norwegian. Our research questions were:

• (1) How do native speakers of Slavic languages (Russian, Polish) and Norwegian interpret the temporal semantics of English past simple? (Experiments 1 and 2)
• (2) How do L2 proficiency-related characteristics, specifically L2 proficiency and the amount of exposure to L2 in everyday life, modulate the way native speakers of Slavic languages (Russian, Polish) and Norwegian interpret English past simple? (Experiment 2)

In each trial participants saw a picture dyad while listening to a sentence preceded by a temporal context. The temporal context was introduced to constrain the interpretation to events at a definite timepoint in the past. In this type of context general factual readings are excluded and aspectual categories in both Russian and Polish function similarly. In experimental trials the picture dyad presented the ongoing and completed version of the same event. Each sentence was a description of the event following the SVO word order with the predicate consisting of a transitive verb, a modifier, and an object. The verb was either in past progressive or past simple (e.g., Grandma baked/was baking a vanilla cake.). All predicates corresponded to Vendlerian accomplishments.

In addition, to address the second research question, participants in Experiment 2 completed a LexTALE task in English (Lemhöfer and Broersma, 2012) as well as the Language and Social Background Questionnaire (LSBQ) (Anderson et al., 2018). LexTALE is a well-established measure of vocabulary size that is closely correlated with general proficiency (Ferré and Brysbaert, 2017; Nakata et al., 2020). LSBQ is an attested instrument for calculating a composite score indicative of weighted relative L1/L2 language use at home and in typical social situations, regardless of specific languages involved. The balance between social and home use of a bilingual's languages is a characteristic that, in addition to L2 proficiency, provides qualitatively different information about the amount of exposure to both languages in daily life. Independently from L2 proficiency, participants with more opportunities to use L2 in daily life may be more sensitized to ways in which L2 is used in real-life situations, including by its native speakers.

Theoretically-driven predictions were outlined in the previous section. We predicted that group differences would emerge only in temporal interpretation of past simple (see specific predictions below). In brief, the biggest predicted contrast is related to whether the participants' online and offline behavior would resemble that associated with their corresponding L1s. Of the proposed theoretical approaches, only those following the AH suggest that participants' behavior might be driven by universal predicate characteristics, resulting in prototypical associations between past simple and completed events in both groups. This association is further predicted to be affected by L2 proficiency and exposure, with a more English native-like pattern (i.e., no preference for either completed or ongoing events) associated with higher L2 proficiency and, by the same token, greater exposure to L2.

By contrast, approaches supporting CLI from L1 predict two distinct patterns, especially associated with lower levels of L2 proficiency and exposure: the Slavic-speaking group would be expected to associate past simple with completed events, mirroring strong completion-focused semantics of Slavic past perfective (see Section 2) while the L1 Norwegian group would be inclined to associate it with ongoing events, similar to the preference observed for Norwegian preterit (as described in Section 2). Again, these associations are predicted to weaken and give way to a more English native-like pattern in more proficient L2 speakers. Furthermore, these approaches also suggest a possible discrepancy between online and offline behavior. Specifically, online behavior might resemble L1-like preferences in contrast with offline judgments, which are more likely to be native-like due to metalinguistic knowledge of English aspect. Such persistence of L1 influence online might hold even at higher levels of L2 exposure and proficiency.

5 Experiment 1

5.1 Participants

N = 20 native Russian speakers (M_*age* = 21.4 years) and n = 21 native Norwegian speakers (M_*age*= 25.1 years) were recruited among student population at UiT - The Arctic University of Norway (Tromsø, Norway) and The Northern Federal Arctic University (Arkhangelsk, Russia), and were reimbursed 10-15 Euro for participation. All participants reported either Norwegian or Russian as their only L1 and English as their L2. All participants gave informed consent after reading the experiment's description and instructions. Demographic information, including self-reported L2 English proficiency and English proficiency test scores, is summarized in Supplementary Table S1. Groups did not statistically differ on either measure [self-reported English proficiency: t (29) = −0.38, p = 0.71; English proficiency test scores: t (34) = 0.45, p = 0.65].

5.2 Materials

The experiment included n = 24 experimental and n = 24 filler items. Each test item included a visual display, a spoken preamble sentence, and a spoken target sentence (both recorded in a sound-proof booth by a female native speaker of British English). The preambles were used to locate the narrative in the past (e.g., It was already late afternoon) as well as to restrict past progressive and past simple to their narrative readings. All of the experimental items involved a sentence describing the scene with the verb in the past tense. We manipulated the aspectual form of the verb (n = 12 past progressive, n = 12 past simple). Each participant heard either a past progressive or a past simple version of each experimental sentence. Each experimental visual display contained two pictures located side by side, representing two distinct stages of the same event: one depicted its ongoing and the other, a completed version (Figure 1; see Supplementary Data Sheet 1 for a full list of experimental items).

All experimental sentences had the Subject-Verb-Object (SVO) structure. The subject always referred to one of the four characters: grandma, grandpa, the girl, or the boy. The object always involved a combination of an adjective and an inanimate noun. The two versions of the same target sentence in each item involved verbs with the same lexical content, but different aspectual forms: one was past simple and one past progressive (see examples 7a–b).

• 7. a. The boy constructed a tall tower.
• b. The boy was constructing a tall tower.

In each filler item, the visual display involved two pictures corresponding to different events in various combinations of their ongoing and completed versions. Both pictures included the same character (grandma, grandpa, the girl, or the boy). The preamble sentences were identical to those used in the experimental items. To counteract the experimental bias that would arise if hearing the auxiliary was would always uniquely correspond to the ongoing event picture, half of the filler items included a construction with was that described the completed event (e.g., The boy was tired when he finally sealed the envelope). The other half of fillers included a past tense verb describing the initial or intermediate stages of an event (e.g., The girl concentrated on tying her shoelaces).

The design and materials used in this web-based experiment closely matched the design and materials of L1 aspect processing studies conducted by our research group (Minor et al., 2022, 2023). In addition, the web-based version of this design had been validated in a replication of Minor et al. (2023) in-lab study with L1 English speakers by Vos et al. (2022).

The web-based eye tracking experiment was programmed using the jsPsych library (De Leeuw, 2015), and was hosted on a JATOS server operated by UiT - The Arctic University of Norway (Lange et al., 2015). The jsPsych library was chosen because it incorporates webcam-based eye-tracking functionality based on WebGazer.js (Papoutsaki et al., 2016). WebGazer uses the participant's webcam to detect their face and employs prediction models to infer the participant's gaze locations based on eye position in real time. A series of recent studies have successfully demonstrated the use of WebGazer to replicate processing effects that were previously investigated with infrared eye trackers (e.g., Yang and Krajbich, 2021; Vos et al., 2022).

5.3 Procedure

The study was entirely web-based. All participants performed the tasks unsupervised. The entire study was carried out in English.

Participants gave informed consent to participate in the study before proceeding to the eye-tracking task. They were instructed that the experiment would take 15 min during which they would need to focus on the task. They were asked to ensure that their face was well lit and that they were seated comfortably. Participants were not explicitly instructed to ensure a particular distance from the screen. However, during the calibration phase, participants needed to position themselves in a way that would correspond to the optimal distance from the screen and the position of the face.

The experimental procedure began with an initial calibration described in Supplementary Data Sheet 2. After a successful initial calibration, instructions for the eye-tracking experiment were presented. Participants were informed that they would listen to recordings and would need to click on pictures that best corresponded to the sentences they heard.

Each trial began with a dot at the center of a white screen. Participants had to click on the dot to begin the trial. After the dot was clicked, the preamble sentence was played. After the preamble sentence was presented, the mouse cursor disappeared and two pictures appeared on the screen. After a 500 ms delay intended to allow participants a brief preview of the pictures, the experimental sentence audio was played (see Supplementary Table S2 for a description of auditory stimuli). Once the audio finished, the mouse cursor reappeared cuing the participant to click on one of the pictures. After the participant responded, the next trial began.

Each participant was randomly assigned to one of four lists. In each list each experimental item was presented once, either with a past progressive or a past simple experimental sentence. Across the four lists, each item occurred twice with a past simple experimental sentence and twice with a past progressive one. The position of the ongoing and completed event pictures on the screen was counterbalanced across the four lists. The order of items within each list was pseudorandomized to avoid clustering of trials that might lead to habituation.

Once participants finished the eye-tracking task, they were directed to a language background questionnaire and English proficiency test. The language background questionnaire asked the participant to assess their level of English proficiency using the CEFR proficiency scale (A1–C2). The English proficiency test was an abridged version of the Oxford Placement Test that consisted of 18 multiple choice questions covering vocabulary and grammar knowledge at the B1–C2 level.

The study was approved by the Norwegian Center for Research Data (NSD).

5.4 Statistical analyses

5.4.1 Offline judgments

Offline judgments were analyzed using a logistic generalized linear mixed effects model (GLMM) with picture choice coded as the binary dependent variable (1 if the ongoing event picture was selected, 0 if the completed event picture was selected). The following predictors were used in the model: Condition (Past Progressive, Past Simple), Group (Russian, Norwegian), and their interaction. Self-reported proficiency levels and proficiency test scores, both centered and scaled, as well as the interaction between the proficiency measures and Condition were included as covariates in the model. Self-reported proficiency was coded on a scale from 1 to 5 corresponding to CEFR levels A2 to C2. Categorical predictors Condition and Group were sum-coded to report main effects. The initial random effects structure included random intercepts for participants and items (corresponding to the event regardless of stage) as well as a by-participant random slope for Condition and by-item random slopes for Condition and Group as well as their interaction. The random effects were simplified until the model converged or reached a non-singular fit (Barr et al., 2013). This resulted in random intercepts for participants and items as well as by-participant random slopes for Condition. To address convergence issues the optimizer bobyqa was used for all models. Variance Inflection Factors (VIF) were calculated to perform a multicollinearity check.

The model was followed up with planned Bonferroni-corrected pairwise comparisons to explore the interaction between Group and Condition (simple effects: L1 Russian vs. L1 Norwegian for each aspectual condition).

The main model was fitted using the R package lme4 (Bates et al., 2015). Post-hoc analyses were conducted using the package emmeans (Lenth, 2022). Multicollinearity checks were carried out using the package car (Fox and Weisberg, 2019).

5.4.2 Gaze preferences

To analyze the participants' gaze preferences, we removed all looks to white space and calculated proportions of looks to the ongoing event picture for each Group and Aspect condition in forty 50 ms time bins starting from 200 ms post lexical verb onset (200 ms were added to account for the time needed to plan and execute a saccade). We then conducted a cluster-based permutation (CBP) analysis (cf. Minor et al., 2022; Maris and Oostenveld, 2007; Hahn et al., 2015; Yang et al., 2020; Oakes et al., 2013) comparing looks to the ongoing event picture between the Groups separately in the Past Simple and Past Progressive conditions. First, we fitted a mixed-effects logistic regression which estimated the log odds of looks to the ongoing event picture in each time-bin with Group included as a fixed effect. The random effects structure included random intercepts for participants and items. Clusters were identified based on adjacent time-bins where the p-value of the Group effect was below 0.08. We selected a higher threshold for including time bins in a cluster to capture potential weak but long-lasting effects and avoid artificial fragmentation of clusters. This choice does not affect the overall validity of the CBP test (Maris and Oostenveld, 2007; Hahn et al., 2015). The z-values for all the time bins within a cluster were summed up to generate the sum statistic for the cluster. Next, we randomly permuted the participants' Group labels (Russian vs. Norwegian), i.e., for each participant the Group label was permuted with 0.5 probability. We then ran the same procedure as described above to identify effect clusters in the permuted data, calculating the sum statistic for each cluster and storing the largest sum-statistic. This process (permutation of the data and calculation of the largest sum statistic) was performed 1,000 times, generating a distribution of sum statistics under the null hypothesis that Group labels do not have an effect on gaze preferences. It was then possible to assess the sum statistics obtained for the clusters in the original dataset against the null-hypothesis distribution, and obtain p-values for these clusters. The significance level for these p-value was set at 0.05.

Analyses were carried out in R (R Core Team, 2025) using packages multcomp (Hothorn et al., 2008), doParallel (Folashade et al., 2022a), foreach (Folashade et al., 2022b), itsadug (van Rij et al., 2022). R package tidyverse (Wickham et al., 2019) was used for data preparation and package ggplot2 was used for data visualization (Wickham, 2016). Experiment data are available at https://osf.io/kyvar?view_only=652cd42634d64b2999e0c32711528d6a.

5.5 Results

5.5.1 Offline judgments

Descriptively (Supplementary Table S3), both groups were highly accurate in their judgments of filler items. This confirmed diligent participation in the task. Turning to the experimental items, both groups strongly preferred ongoing event pictures in the Past Progressive condition. In contrast, group differences were apparent in the Past Simple condition, where the L1 Russian group preferred completed events, selecting ongoing event pictures in only 7.1% of the trials, while the L1 Norwegian group preferred ongoing events, selecting them in 75.4% of the trials (Figure 2).

GLMM analysis (for model summary and plots see Supplementary Table S4 and Supplementary Figure S5) revealed significant main effects of Condition and Group as well as a significant Condition by Group interaction. The effects of Proficiency score and Self-reported proficiency, as well as their interactions with Condition, were not significant. Follow-up pairwise comparisons between Groups within Conditions indicated that L1 Norwegian speakers were significantly more likely to select the ongoing event picture in the Past Simple condition compared to L1 Russian speakers (Δ log-odds = 4.98, SE = 0.65, z = 7.69, p < 0.0001). In the Past Progressive condition, the difference between the groups was not significant (Δlog-odds = −1.64, SE = 1.2, z = −1.37, p = 0.17).

5.5.2 Gaze preferences

Plots of the participants' gaze preferences suggest that both groups preferentially looked at the ongoing event picture when hearing a verb in the past progressive form (Figure 3, top panel). When hearing a past simple verb, however, Russian speakers looked more at the completed event picture, while Norwegian speakers displayed a slight preference for the ongoing event picture (Figure 3, bottom panel).

We used CBP analysis to compare gaze preferences between the Groups within each Condition. In the Past Progressive condition, we did not find significant differences between the groups, suggesting that both groups processed the past progressive form similarly. In the Past Simple condition, however, we found that the L1 Russian speakers exhibited significantly fewer looks to the ongoing event picture (and correspondingly, more looks to the completed event picture) compared to the L1 Norwegian group, corresponding to a cluster from 1,150 to 2,000 ms after verb onset + 200ms (sum z = −55.81, p = 0.001).

5.6 Experiment 1 discussion

In the Past Progressive condition, both groups demonstrated similar patterns of offline responses and gaze preferences associating past progressive forms with ongoing events, as expected. Notably, two distinct group-specific patterns emerged with past simple: the Russian group strongly associated past simple with completed events, while the Norwegian group preferred ongoing events, as revealed in both the offline responses and online gaze patterns. These preferences resemble those reported in these groups' respective L1s suggesting that the observed difference between the groups stems from L1 influence.

6 Experiment 2

The goal of Experiment 2 was to test if the results of Experiment 1 would be replicated with a different group of Slavic speakers while increasing both the size of the participant samples and the number of test items. Furthermore, we wanted to investigate whether L2 proficiency and the amount of L2 exposure modulates the way native speakers of Slavic and Norwegian interpret the English Past Simple.

6.1 Participants

We recruited two new groups of participants: L1 Polish and Norwegian speakers. L1 Polish speakers were recruited via Prolific (http://www.prolific.co) and were reimbursed 4 GBP for participation. To define the pool of potential participants, the following filters were used: Location (Poland), First Language (Polish), Fluent languages (English), Raised only with their native language, Bilingual status (native language plus one other language). Standard sample was selected for study distribution. In total, 63 people completed the study. Data from six participants were discarded either due to failure to complete the second part of the study (the demographic questionnaire, the LSBQ questionnaire, and the LexTALE task) or low eye tracking data quality (absence of analyzable datapoints). All remaining 57 participants (M_*age* = 24.1 years) reported Poland as both their birth country and current country of residence, as well as the country of school education and of most recent degree completion. Thirty-one reported knowledge of at least one other language in addition to Polish and English (German, Spanish, Russian, Italian). In addition, 12 reported knowledge of a fourth foreign language (German, Hindi, French, Spanish, Russian). Six participants reported a history of prolonged stay outside of Poland (range one to three years).

L1 Norwegian speakers were recruited via a combination of Prolific, recruitment of university students for course credit, and in-person recruitment from the pool of participants of other studies carried out at UiT - The Arctic University of Norway in Tromsø (Norway). Prolific participants (n = 6) were reimbursed 4 GBP for participation. To define the pool of potential participants, the following filters were used: Location (Norway), First Language (Norwegian), Fluent languages (English), Raised only with their native language, Bilingual status (native language plus one other language). Students (n = 29) received course credit. Participants recruited in-person from study pools (n = 5) were reimbursed 200 NOK. In total, 40 people completed the study. Data from eight participants were discarded due to poor quality of eye tracking data (absence of analyzable datapoints). One participant's data were discarded due to non-Norwegian background reported in the demographic questionnaire. Of the remaining 29 participants (M_*age* = 26.9 years), 28 reported Norway as their birth country; one person reported China as their birth country but indicated that they received school education as well as the most recent degree in Norway, and did not report knowledge of languages other than Norwegian and English. Thus, we classified this participant as a Norwegian native speaker. All 29 reported Norway as the country of their school education and their most recent education degree. Nine participants reported knowledge of at least one additional language other than Norwegian and English (German, Swedish, Spanish, French, Finnish). One participant also reported knowledge of a fourth foreign language (Thai). Six participants reported a history of prolonged stay outside of Norway (up to 1 year).

Demographic information, including LSBQ composite factor scores and LexTALE scores, is summarized in Supplementary Table S6.

6.2 Materials

Materials for Experiment 2 had been selected based on a norming study (for a detailed description, see Supplementary Data Sheet 3), which yielded 30 experimental items and 30 fillers. During data processing it was discovered that one of the experimental items (experimental sentence Grandpa was building a big snowman) appeared only with a progressive description in all lists. It was removed from all analyses due to inability to form a pair with a past simple counterpart. Thus, the number of experimental items included in the statistical analyses was reduced to 29 per condition. The structure of the test items and the fillers was identical to Experiment 1 (see Supplementary Table S7 for a description of auditory stimuli).

Additionally, we used a demographic questionnaire to collect participants' general demographic information; a LexTALE task (Lemhöfer and Broersma, 2012) to measure participants' L2 English proficiency; and a Language Social and Background Questionnaire (LSBQ) (Anderson et al., 2018) to calculate a composite score serving as an index of the amount of L2 English used in participants' daily social and home life relative to their L1. LSBQ was minimally adapted to account for participants' respective L1s (Polish or Norwegian) and L2 (English).

6.3 Procedure

The experimental procedure was identical to Experiment 1. The study was entirely web-based and the participants performed the tasks unsupervised. The entire study was carried out in English.

The experiment was programmed using Gorilla Experiment Builder (http://www.gorilla.sc) (Anwyl-Irvine et al., 2020) and used the webcam-based eye tracking package WebGazer.js (Papoutsaki et al., 2016). Participants gave informed consent to participate in the study before proceeding to the eye tracking task. They were instructed to find a quiet, well-lit room where they would be undisturbed for at least 20 min. Participants were not explicitly instructed to ensure a particular distance from the screen. However, during the calibration phase (described in Supplementary Data Sheet 2), participants needed to position themselves in a way that would correspond to the optimal distance from the screen and the position of the face.

After the successful initial calibration, participants were instructed that they would listen to recordings and would need to click on pictures that best corresponded to the event description that they heard. Each trial began with a blank white screen that had a “Continue” button at the bottom. Participants had to click the button to begin the trial. The button was placed at the bottom of the screen to force the participants to leave the cursor at the bottom of the screen before the trial to avoid distractions. After the “Continue” button was clicked, a fixation cross appeared in the middle of the white screen and simultaneously the preamble sentence was played. The fixation cross remained on the screen for 3,000 ms. One hundred ms after the fixation cross disappeared, two pictures, placed horizontally next to each other, appeared on the screen. After a 500 ms delay, intended to allow participants a brief preview of pictures, the experimental sentence audio was played. When the audio finished, a question mark appeared in the middle of the blank space between the pictures, serving as a cue to respond by clicking on either picture. Once the participant responded, the next trial began. If no response was received, the next trial automatically advanced after 3,000 ms.

Each participant was randomly assigned to one of four lists. The structure of the lists was identical to Experiment 1. Once participants finished the eye tracking task, they were redirected to the survey portion of the study. The surveys were presented in the following order: the general demographic questionnaire, LSBQ, LexTALE.

The study was reviewed and approved by the Norwegian Center for Research Data (NSD, https://www.nsd.no/en).

6.4 Statistical analyses

6.4.1 Offline judgments

To compare participants' offline judgements on the picture selection task we fit a logistic mixed effects model (GLMM) with picture choice as the dependent variable (1 if the ongoing event picture was selected, 0 is the completed event picture was selected). The following predictors were used in the model: Condition (Past Progressive, Past Simple), Group (Polish, Norwegian), LSBQ Composite Factor Score and the LexTALE score, two-way and three-way interactions between Condition, Group and LSBQ Composite Factor Score, and between Condition, Group and LexTALE score. Categorical predictors Condition and Group were sum-coded to report main effects. Random effects included random intercepts for participants and items as well as by-item and by-participant random slopes for Condition.¹ This random effects structure resulted from gradual simplification of the maximal random effects structure until the model converged without a singular fit warning. To address the potential issue of multicollinearity, LSBQ Composite Factor Score and LexTALE score were scaled and centered. In addition, VIFs were calculated to check for multicollinearity.

The model was followed up with planned Bonferroni-corrected pairwise comparisons to explore significant interactions of interest, specifically the two-way interaction between Condition and Group (simple effects: L1 Polish vs. L1 Norwegian for each condition separately, for RQ1) and the three-way interactions between Group, Condition, and LexTALE score, and between Group, Condition, and LSBQ Composite Factor Score (RQ2).

The main model was fitted using the R package lme4 (Bates et al., 2015). Post-hoc analyses were conducted using the package emmeans (Lenth, 2022), and ggeffects (Lüdecke, 2018) was used for modeled data visualization. Multicollinearity checks were carried out using the package car (Fox and Weisberg, 2019).

6.4.2 Gaze preferences

To compare gaze preferences between the groups, an identical CBP analysis procedure was conducted as in Experiment 1.

In addition, we fitted a logistic GLMM to explore how the participants' gaze preferences may be modulated by exposure and proficiency in English, as estimated by the LSBQ Composite Factor Score and the LexTALE score. To reduce the effects of autocorrelation in the gaze data, for each trial we calculated proportions of looks to the ongoing event picture within four consecutive 500 ms time-windows starting from 200ms after lexical verb onset. Most of the resulting proportions were either 0 or 1 (81.8% of the data). The remaining proportions were binarized by replacing all proportions >0.5 with 1 and all proportions <0.5 with 0. Proportions equal to 0.5 were removed (2% of the data). The resulting binary values were included as the dependent variable in a logistic mixed effects model. To assess if sensitivity to the past progressive vs. past simple manipulation was differentially modulated by L2 proficiency and exposure, we first fitted a maximal model including all main effects and two-way interactions among Condition, Group, the LSBQ score, and the LexTALE score, and three-way interactions among Condition, Group, and the LSBQ score, and Condition, Group, and the LexTALE score. After simplification to ensure convergence, the random effect structure included random intercepts for participants, items, and time-window, and by-item random slopes for Condition. Continuous predictors (LSBQ score and LexTALE score) were scaled and centered; categorical predictors (Condition and Group) were sum-coded. To assess whether the Condition by Group interaction was modulated by L2 proficiency and/or exposure, we compared the full model to a reduced model without the three-way interactions via a likelihood-ratio test (LRT). Since the test was non-significant [χ²(2) = 1.34, p = 0.51], we removed the three-way terms and interpreted the Condition by Group interaction from the reduced model.² The model was followed up by Bonferroni corrected pairwise comparisons to investigate significant interactions.

6.5 Results

6.5.1 Proficiency and exposure

Overall, the Norwegian group scored higher on the LexTALE task [mean score = 8.28, SD = 0.93 (6.72,9.6)] than in the Polish group [mean score = 5.71, SD = 0.99 (3.58,7.55)], which was confirmed by a t-test (p < 0.001). Exposure to L2 English, measured by the LSBQ Composite Factor Score, was comparable between the groups [Norwegian: M = 7.32, SD = 4.37 (−0.44, 17.07), Polish: M = 6.51, SD = 4.25 (−6.39, 17.82), p = 0.4]. In both groups, the LSBQ composite factor score positively correlated with the LexTALE score (Polish group: r² = 0.10; Norwegian group: r² = 0.27).

6.5.2 Offline judgments

Overall, the pattern of responses on the picture selection task was similar to that observed in Experiment 1 (Figure 4). Both groups showed a strong preference for ongoing event pictures in the Past Progressive condition (93.3% of the trials in the Norwegian group and 90.1% in the Polish group). In the Past Simple condition, the groups exhibited contrasting preferences: Norwegian speakers chose the ongoing event picture in 71.9% of the trials, while Polish speakers chose the ongoing event in only 11.3% of the trials. Both groups had close-to-ceiling accuracy on the filler trials (L1 Norwegian: 97.6%, L1 Polish: 95.5%), indicating diligent task performance.

GLLM revealed significant main effects of Condition, Group, and their interaction, as well as significant two-way interactions between Condition and LexTALE score, Group and LexTALE score, Group and LSBQ Composite Factor Score, and a significant three-way interaction between Group, Condition, and LexTALE score (see Supplementary Table S8 and Supplementary Figure S9, for model output summary). Post-hoc comparisons between Groups within Conditions revealed that the Norwegian participants were significantly more likely to select the ongoing event picture in the Past Simple condition compared to the Polish participants (Δlog-odds = 5.6, SE = 0.62, z = 9.09, p < 0.0001), while in the Past Progressive condition the difference between groups was not significant (Δlog-odds = 0.07, SE = 1.04, z = 0.06, p = 0.95).

Predicted probabilities of ongoing event picture choice by Group, Condition, and LexTALE score are illustrated in Figure 5. Planned post-hoc comparisons indicated that the Polish group's association between past progressive and ongoing events increased significantly with higher LexTALE score [β = 1.9; 95% CI (0.96,2.85)], and conversely the association between past simple and ongoing events decreased significantly with higher LexTALE score [β = −0.96; 95% CI (−1.56,−0.35)]. Norwegian participants were also significantly less likely to select the ongoing event picture in the Past Simple condition with higher LexTALE scores [β = −0.93; 95% CI (−1.76,−0.09)]. The same trend was observed in the Past Progressive condition, although it was not significant [β = −0.9; 95% CI (−2.42,0.61)]. Furthermore, the analysis revealed a significant difference in the effect of LexTALE score between the two groups in the Past Progressive condition (Δlog-odds = −2.81, SE = 0.92, z = −3.04, p = 0.002), but not in the Past Simple condition (Δlog-odds = 0.03, SE = 0.52, z = 0.06, p = 0.95).

The predicted effects of LSBQ Composite Factor Score on ongoing event picture selection probability are illustrated in Figure 6. Post-hoc comparisons revealed that overall, LSBQ score had a significantly larger positive effect on the probability of ongoing event selection in the Norwegian group compared to the Polish group (Δlog-odds = 0.77, SE = 0.31, z = 2.46, p = 0.01). This effect was driven by a significant difference of LSBQ score effects between the Groups in the Past Simple condition (Δlog-odds = 0.7, SE = 0.31, z = 2.25, p = 0.02). The trend was similar in the Past Progressive condition but was not significant (Δlog-odd = 0.8, SE = 0.58, z = 1.45, p = 0.14).

6.5.3 Gaze preferences

In line with the offline results, CBP analysis of the participants' gaze preferences revealed that the Norwegian group exhibited significantly more looks to the ongoing event picture in the Past Simple condition compared to the Polish group (Figure 7). This effect corresponded to a cluster from 1,150 to 1,550 ms after lexical verb onset+200ms (sum z = −21.1, p = 0.019). Another cluster from 400 to 700 ms was marginally significant (sum z = −13.82, p = 0.057). No significant differences between the Groups were found in the Past Progressive condition.

GLMM revealed significant main effects of Group, Condition, LexTALE score and LSBQ score, as well as a significant interaction between Group and Condition (see model summary in Supplementary Table S10 and Supplementary Figure S11). Interactions involving LexTALE score and LSBQ score were not significant. Post-hoc comparisons revealed that the Norwegian group was significantly more likely to look at the ongoing event picture compared to the Polish group in the Past Simple condition (Δlog-odds = 0.63, SE = 0.16, z = 4.03, p < 0.001). In the Past Progressive condition the difference between the groups was not significant (Δlog-odds = 0.26, SE = 0.15, z = 1.74, p = 0.08). Furthermore, both groups were more likely to look at the ongoing event picture than the completed event picture in the Past Progressive condition [Norwegian group: 0.63 probability, 95% CI (0.56,0.69); Polish group: 0.57 probability, 95% CI (0.52,0.61)]. In the Past Simple condition, the Norwegian group was also more likely to look at the ongoing event picture than the completed event picture [0.6 probability, 95% CI (0.53,0.67)], while the Polish group was less likely to look at the ongoing event picture than the completed event picture [0.45 probability of looks to ongoing event, 95% CI (0.4,0.49)].

6.6 Experiment 2 discussion

Overall, the results of Experiment 2 replicate the findings from Experiment 1. Similarly to L1 Russian speakers, L1 Polish speakers were less likely to look at and select the ongoing event picture in the Past Simple condition compared to Norwegian speakers, while in the Past Progressive condition the groups behaved similarly. Furthermore, we found that L2 proficiency had an effect on the participants' offline responses: in both groups, participants with higher proficiency were more likely to select the completed event picture in the Past Simple condition. The interpretation of this effect may be different in the two groups, however. In the Norwegian group, participants become more L1-like in their responses with increased proficiency, moving closer to the approx. 50% completed event preference pattern observed for L1 English speakers in Minor et al. (2023). In the Polish group, on the other hand, participants appeared to move away from the L1-patterns with increased proficiency, toward a ceiling preference for the completed event picture.

Additionally, groups differed in the effect of L2 English proficiency on offline responses in the Past Progressive condition: more proficient L1 Polish participants tended to select the ongoing event picture more often, while no such effect was observed in the Norwegian group. However, this pattern may be due to the fact that our Norwegian sample did not include any participants within in the lower proficiency range where the corresponding effect was primarily observed in the Polish group (see Figure 5).

Finally, the analysis indicated a difference in the effect of L2 English exposure on offline responses in both groups: in the Norwegian group, higher levels of exposure were associated with higher overall probability of selecting the ongoing event picture, particularly in the Past Simple condition, while no such effect was observed in the Polish group. In view of the positive (albeit relatively weak) correlation between LSBQ and LexTALE scores, it is surprising that in the Norwegian group a higher LexTALE score predicted a decrease in the preference for ongoing events in the Past Simple condition, while an increase in the LSBQ score had an opposite effect. First, a potential explanation is related to the unbalanced distribution of data in our sample: while the groups did not significantly differ in the distribution of LSBQ scores, the Norwegian group's LexTALE scores were significantly higher than those of the Polish group, with lower scores absent in the dataset. It might well be that our L2 proficiency and exposure data do not fully capture the relationship between these two variables. Second, speculatively, in our sample Norwegians with greater exposure to L2 English in daily life might predominantly interact with other non-native English speakers, whose associations between English aspect and event stage could well differ from those of L1 English speakers. This could have led to the observed non-L1-English-like pattern.

7 General discussion

We used web-based eye tracking combined with a sentence-picture matching task to investigate the influence of aspectual L1 Polish and Russian as well as non-aspectual L1 Norwegian on the interpretation of past simple and past progressive forms in L2 English. Slavic and Norwegian L1 speakers performed similarly with past progressive, strongly associating this form with ongoing events online and offline. Past simple revealed group differences. Slavic groups strongly associated past simple with completed events online and offline, while Norwegians tended to associate past simple more with ongoing events.

These effects in offline responses were modulated by L2 proficiency. For Polish L1 speakers, the link between past simple and completed events strengthened with an increased L2 English proficiency. On the other hand, the Norwegian speakers' association between past simple and ongoing events weakened with increased English proficiency.

Overall, our findings are consistent with theoretical accounts of L2 aspect acquisition that predict grammatical CLI from the L1. However, they also add further nuance with respect to L1-specific trajectories of such influence. The Polish group demonstrated a strong L1 influence that strengthened even more with higher L2 proficiency. The Norwegian group, by contrast, showed a pattern that suggested a tendency toward weakening of L1 influence with higher L2 proficiency.

We did not find support for approaches focused on the primary role of lexical aspect, such as the Aspect Hypothesis (AH). Such approaches would predict that L1 does not play a substantial role in the interpretation of L2 aspect. Following this view, all groups in our study would follow a similar trajectory, associating past simple accomplishments with completed events, especially at lower levels of L2 proficiency. With increasing L2 proficiency, this association should become less categorical. This was not the case in our data: although a somewhat similar pattern was observed in the L1 Polish group in Experiment 2, the pattern detected in the Norwegian group was strikingly different. These distinct associations held even though we controlled for participants' L2 proficiency and exposure in both online and offline models. Furthermore, with an increase in L2 proficiency both groups tended to strengthen their preference for the completed events. The presence of statistically significant differences in the processing patterns between the L1 groups, especially at lower levels of L2 proficiency, indicates that the source of non-target-like performance was not the same across the groups.

Nevertheless, one might argue that our findings are not incompatible with AH's predictions. Given the overall lower L2 proficiency of the Polish group, its preference for completed events might stem from the prototypical association of past simple accomplishments with completion, which would be strongest at lower L2 proficiency levels. Following this line of reasoning, Norwegians' tendency to assimilate past simple with ongoing events at lower levels of L2 proficiency could be simply a reflection of L2 development, given the group's higher L2 proficiency relative to the L1 Polish group. However, we find this reasoning implausible. First, in Experiment 1 the groups were balanced on proficiency, yet the same pattern of visual preferences emerged. Second, regression analyses with L2 proficiency and exposure used in Experiment 2 allowed us to capture the predicted variation both at higher and lower levels of L2 proficiency in both groups. If AH's predictions explained the pattern of results found in our study, the statistical models would predict a similar tendency of associating past simple with completed events at lower levels of proficiency regardless of L1. This was not the case. In both groups, the tendencies at the lower end of L2 proficiency unequivocally point to completed events for the Polish group and ongoing events for the Norwegian group, a pattern most consistent with the transfer of aspectual representations from respective L1s.

Although our data are consistent with L1 influence, they are not equally compatible with all accounts supporting CLI from L1. Specifically, in Experiment 2, the Polish group's offline judgments strengthened the preference for completed events, an L1 Polish-resembling pattern, with an increase in L2 proficiency. Such strengthening of L1 influence is not easily accounted for by theoretical models.

Functional accounts following the Thinking-for-Speaking hypothesis predict persistent L1 influence, especially in online performance that taps subconscious processing. Such influence, stemming from life-long sensitization (or lack thereof) to temporal characteristics of events, is predicted by the mere presence or absence of grammaticalized aspect in the L1. Both Slavic groups were indeed sensitive toward the distinct temporal stages of events presented in the experimental trials: Slavic participants categorically associated past progressive with ongoing events and past simple with completed events. This dual mapping strongly resembles the perfective-imperfective opposition in Slavic (Polish and Russian) aspect. By contrast, Norwegians were substantially less sensitive to the aspectual contrast, associating both past simple and past progressive with ongoing events, although to a different extent. This pattern is consistent with the ongoing event–preterit association found in L1 Norwegian using the same experimental design (cf. Section 2).

Although Thinking-for-Speaking does not provide a clear prediction with respect to L2 proficiency effects, we can deduce that increased L2 proficiency might sensitize L2 speakers to conceptualization patterns specific to L2, resulting in more target-like performance. In Experiment 2, a trajectory toward more English native-like performance can be observed in the Norwegian group's offline judgments, but not in the Polish group. In fact, in the Polish group, L1 influence appeared to strengthen with higher L2 proficiency in offline judgments. Although persistence of L1 influence is, generally speaking, not incompatible with this view, the approach within the Thinking-for-Speaking framework does not provide an adequate account of its increase at higher levels of proficiency.

Persistent L1 influence, based on mere presence or absence of a grammaticalized aspectual distinction, is also consistent with the formal generative view that takes a “generalist” approach to L1 influence. In this view, the presence of grammaticalized aspect in L1 should give an advantage when processing aspectual information in L2, even if L2 aspect is configured differently from the L1. For example, in (Roberts and Liszka 2013, 2021), French, but not German, native speakers showed native-like sensitivity during online processing of aspectual and aspect-related violations in L2 English. Reduced sensitivity to the aspect manipulation in the Norwegian groups is in line with predictions for speakers of a non-aspectual L1. It is notable, however, that our Slavic groups were not merely sensitized to the aspectual manipulation in English, but overgeneralized the perfective–imperfective distinction, strongly associating the English past simple form with completed events, in contrast to English L1 speakers (Minor et al., 2022, 2023). Experiment 2 indicated that this overgeneralization was especially pronounced in offline judgments at higher L2 proficiency. This pattern suggests that a generalist account is insufficient as it does not explain language-specific effects. Clearly, the dual aspectual paradigm in our Slavic groups' L1s interfered with sensitivity toward aspectual properties in the L2.

The Feature Reassembly Hypothesis (FRH) offers the most adequate account of our results. According to the FRH, mapping of L1 categories onto the L2 is a natural step in L2 acquisition. It is followed by fine-tuning (reassembling) the features that are not identical between L1 and L2. This arduous process can be prolonged and is especially challenging when one has to disassemble L1-relevant features, in comparison with adding new ones. First, all groups demonstrated mapping L1 perfective (Russian, Polish) or preterit (Norwegian) onto L2 English past simple, as revealed both in online gaze preferences and offline responses on the picture selection task. Second, the asymmetry in the groups' performance is expected considering different reassembly needs of the groups. Slavic groups were faced with the task of disassociating the perceived dichotomy of progressive-simple aspect in L2 from the perfective–imperfective dichotomy in their L1. This involves reducing the strong association between the past simple and event completion (typical for the Slavic perfective), which these groups did not seem to master successfully even at higher levels of proficiency. In contrast, the Norwegian group did show improvements in reducing the association of past simple with ongoing events with increased English proficiency. We hypothesize that this was possible because this process did not require feature disassembly. On this approach, the stronger association of the English past simple with completed events compared to Norwegian preterit is due to pragmatic competition with a grammaticalized progressive form which specifically encodes ongoing events. Norwegian lacks such a form. Thus, it is expected that with increased proficiency in English, Norwegian speakers would more strongly internalize the opposition between past simple and past progressive, which would then result in a reduced association between past simple and ongoing event representations. In other words, the developmental path of the Norwegian learner can be understood in terms of a pragmatic shift in past simple interpretation triggered by increased sensitivity to the past simple vs. past progressive opposition.

The Polish group's strengthening offline preference for completed events with English past simple with increased proficiency may also be connected to metalinguistic awareness, and specifically a conscious attribution of perfective semantics to the English past simple, resulting from life-long habituation to the Polish aspectual system with a dual opposition between perfective and imperfective. The reliance on metalinguistic awareness of aspectual oppositions is not unheard of in L2 aspect acquisition literature. For example, Izquierdo and Collins (2008) reported that Spanish L1 speakers explicitly relied on L1 perfective–imperfective opposition from their L1 when using aspect in L2 French, in contrast with English L1 speakers who showed greater reliance on predicational aspect. In the case of Polish and English, the performance of the Polish group could be seen as “overinterpretation” of an aspectually underspecified form in L2 when one's L1 encodes a full perfective-imperfective opposition.

8 Conclusion

This study investigated the role of L1 influence on the interpretation of aspectual semantics in L2 English, focusing on how native speakers of aspectual Slavic languages (Polish and Russian) and non-aspectual Norwegian process English past simple and past progressive. Using web-based eye-tracking and an offline sentence-picture matching task, we found that L1-specific patterns significantly influenced both online and offline interpretations of English aspect.

Slavic speakers consistently associated English past simple of telic predicates with completed events, reflecting a transfer of the perfective-imperfective distinction from their L1. Strong CLI effects were observed despite clear differences in the way aspect is morphologically encoded in Slavic languages and English. Interestingly, the association of English past simple with event completion—a non-canonical interpretation in comparison with L1 English speakers (Minor et al., 2022, 2023)—persisted and even strengthened with higher L2 proficiency for L1 Slavic speakers. In contrast, Norwegian speakers associated English past simple with ongoing events, mirroring the semantics of Norwegian preterit. However, this association weakened with increased L2 proficiency, indicating a gradual shift toward a more native-like English pattern.

Our findings support theoretical accounts of cross-linguistic grammatical influence, particularly the Feature Reassembly Hypothesis, which posits that L2 learners map existing L1 representations onto L2 forms and face challenges in reassembling features that differ between the two languages. Under this account, Slavic speakers struggled to disassociate English past simple from the restrictive perfective semantics of their L1, while Norwegian speakers demonstrated a more flexible trajectory, likely due to the absence of grammaticalized aspectual features in their L1.

The results challenge approaches like the Aspect Hypothesis, which predict universal prototypical associations between grammatical and lexical aspect in L2 acquisition. Instead, the data highlight the persistent and nuanced role of L1-specific grammatical systems in shaping L2 aspectual interpretation, even in high-proficiency L2 learners.

Overall, this study underscores the complexity of L2 aspect acquisition, showing that L1 influence is not uniform but varies depending on the grammatical features of the L1. These findings contribute to a deeper understanding of how cross-linguistic differences in first languages shape L2 processing and provide valuable insights for theoretical models of CLI and L2 acquisition. In our study, we focused on the interpretation of the past simple and past progressive forms for a specific group of predicates: telic accomplishment events. Each predicate was presented in the context of a single sentence, following a preamble that set the narrative in the past. Building on the results reported here, future research should investigate the interpretation of English simple past tense forms across various predicate types: telic as well as atelic, and within a wider range of contexts—including sequence of events and diverse narrative types. Such research would offer a more comprehensive understanding of the factors that might influence the interpretation of verbal tense and aspect forms in English by L1 as well as L2/Ln English speakers with different background languages.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/flang.2026.1756472/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Matthew W. Crocker, Saarland University, Germany

Reviewed by: Onur Özsoy, Leibniz Center for General Linguistics (ZAS), Germany

Vladislava Warditz, University of Cologne, Germany

Footnotes

Model formula:

Picture_choice ~ Condition^*Group^*(LSBQ_score + LexTALE_score) + (1+Condition | Participant) + (1+Condition | Item).

Model formula:

Looks_to_OE ~ Condition^*Group^*(LSBQ_score + LexTALE_score) – Condition:Group:LSBQ_score – Condition:Group: LexTALE_score + (1 | Participant) + (1+Condition | Item) + (1 | time-window).