1. Introduction

1A considerable body of research has shown that articles constitute an elusive grammatical category for learners of English as a second language (L2), particularly when their first language (L1) lacks a corresponding article system. For child L2 learners—successive bilinguals who begin learning L2 between the ages of 4 and 7—articles pose only temporary difficulties: at initial stages, children frequently omit articles in obligatory contexts, but eventually converge on target-like usage (e.g., Gavruseva, 2008; Morales-Reyes & Gómez Soler, 2016; Schönenberger, 2014; Zdorenko & Paradis, 2008, 2012, among others). By contrast, for adult (post-puberty) L2 learners, determiners remain persistently difficult and are among the most error-prone categories in ultimate attainment (Huebner, 1983; Derkach & Alexopoulou, 2024; Ionin, Ko & Wexler, 2004; Kim & Lakshmanan, 2009; Ko, Ionin & Wexler, 2010; Liu & Gleason, 2002; Johnson & Newport, 1989; Snape & Kupisch, 2010; Snape, Mayo & Gureli, 2013; Thomas, 1989; Trenkic, 2009, among others). An illustration of the complexity involved in English article semantics is provided in example (1):

(1) A: Have you ever been to the White House?
B: No, but I’ve been to a white house.

2The two noun phrases in (1) are nearly identical in lexical content, but differ fundamentally in meaning, a distinction signaled by the articles the and a. The definite noun phrase the White House refers to a culturally salient, unique referent, while the indefinite a white house introduces a non-specific (previously unmentioned) referent. As Hawkins (1991) observes, uses of the often rely on cultural or situational familiarity, requiring both linguistic and extralinguistic knowledge. For L2 learners who come from articleless L1 backgrounds (e.g., Japanese, Korean, or Russian), this dual reliance on linguistic form and shared knowledge presents a formidable learning task.

3In Russian, to convey the semantic contrast shown in (1), speaker B's response would need to be expressed either as 'house of white color' or the adjective would need to appear in a diminutive form to differentiate the meaning from 'the White House.' A Russian version of the dialogue is shown in (2), with two variants of B’s response indicated as (i) and (ii):

(2) A: Ty byla kogda-nibud’ v Belom Dome?

You be.Past.Sg.Fem ever in white.Loc.Sg.Masc house.Loc.Sg.Masc

B(i): Net, no ja byla v dome belogo cveta

No, but I was in house.Loc.Sg.Masc white.Gen.Sg.Masc color.Gen.Sg.Masc

B(ii) Net, no ja byla v belenkom dome (in white.DIM.Loc.Sg.Masc house.Loc.Sg.Masc)

4L2 acquisition data from learners whose L1 lacks an article system based on (in)definiteness features consistently show two types of article errors: omission and substitution. Omission errors result in ungrammatical output (e.g., '*Prius was totaled' instead of 'A Prius was totaled'), while substitution errors arguably reflect incorrect semantic categorization (e.g., using a in a definite context, or the in an indefinite one). Substitution errors occur in both directions, suggesting that learners have not yet internalized the semantic/pragmatic features that differentiate a and the. For example, a Russian-speaking learner might refer to *a father of my fiancée where the definite the father of my fiancée is required, or say the beautiful silver necklace in a context where the referent is novel and should be introduced with a (examples are from Ionin, Ko & Wexler, 2004).

6Although much prior research has investigated article acquisition with a primary focus on the semantics of definiteness and specificity, another key factor that interacts with article choice in English is grammatical noun type, i.e. whether the noun is count or mass (Chierchia, 2010; Gardelle, 2019; Rothstein, 2010). The English article system is sensitive to this grammatical distinction: a combines only with singular count nouns, the is grammatically compatible with both count and mass nouns, and mass nouns often appear with a null article (e.g., 'They sell antique jewelry,' not '*an antique jewelry'). As shown in Table 1, this creates a morphosyntactic asymmetry in the article system:

Table 1: English article use across noun types and definiteness

Noun type Indefinite Definite

Count, sg. a book the book

Count, pl. Ø books the books

Mass Ø furniture the furniture

7Because the null article co-occurs with both plural count and mass nouns, its acquisition adds an additional layer of complexity. Prior studies suggest that article errors are not uniformly distributed across contexts but are instead sensitive to the interplay of definiteness, noun type, and referential specificity. A growing body of research has begun to examine the role of the count/mass distinction in shaping L2 article acquisition, especially in interaction with (in)definiteness (Cerqueglini, 2022; Choi, Ionin & Zhu, 2018; Hermas, 2019, 2020; Inagaki, 2014; Sabir, 2019; Snape, 2008; Tang, Fiorentino & Gabriele, 2023; Yoon, 1993; Wu, 2025).

8Contributing to this line of work, the present study investigates how learners of English with Slavic L1 backgrounds (Croatian, Czech, Polish, Russian, Slovak, and Ukrainian) acquire article use in contexts where grammatical noun type (count vs. mass) and (in)definiteness interact. Specifically, we examine learners' use of a vs. the with singular count nouns and of the null article vs. the with mass nouns. We compare our participants' performance across proficiency levels and (in)definiteness contexts and examine which article contexts pose the greatest challenge, particularly in relation to noun type.

9Learners from these Slavic backgrounds are treated here as a unified group, since their languages are fundamentally similar in core structural properties of the nominal domain: (i) absence of obligatory determiners comparable to English a/the; (ii) agreement in number, gender, and case between demonstratives/adjectives and the noun; (iii) reliance on information-structure distinctions (theme or old information/rheme or new information) and word order for (in)definiteness interpretation; and (iv) the presence of a count–mass distinction in the nominal system (Siewierska & Uhlířová, 2010; Zlatić, 2014). Taken together, these shared features justify treating our L2 sample as a coherent group whose article-related challenges in L2 English stem from comparable characteristics of their respective L1s.

10The remainder of this article is structured as follows. Section 2 reviews the relevant background literature on article acquisition and the role of noun type. Section 3 describes the design and implementation of the comprehension experiment, followed by a discussion of the results. Section 4 discusses the findings in light of previous research. Section 5 concludes the article.

2. Background

11A growing body of research in L2 acquisition has addressed the importance of noun type, specifically, the count/mass distinction, in shaping L2 learners’ article use. Yoon (1993) was one of the first to investigate how L1-Japanese/L2-English learners perceive noun countability and how this perception influences article use. Using a cloze task in which learners supplied articles in an essay context, she examined the interaction between perceived noun countability and article choice (null vs. indefinite a). Her results showed that nouns perceived as noncount (e.g., appreciation, burden, defiance) elicited consistent use of the zero article by L2 learners. However, native English speakers patterned differently: for these same nouns, they typically supplied the indefinite article a in the essay context, showing sensitivity to how discourse context can shift perceived noun type. Yoon concluded that L2 learners had difficulty tracking such contextual influences on noun countability and therefore diverged from native speakers in their use of indefinite articles. This early work suggested that noun type must be treated as a core grammatical variable, highlighting L2 learners' failure to recognize that some English nouns can be both count and mass.

12Similarly, Wu (2025) investigated Chinese-speaking learners’ acquisition of flexible nouns (e.g., chocolate, ribbon, rope) that can shift between count and mass interpretations depending on context. In the quantity judgment task, intermediate and advanced learners performed at chance level with flexible nouns, but showed high accuracy with count nouns and atomic mass nouns (e.g., furniture, luggage). These findings suggest that even when learners grasp core count–mass distinctions, semantic flexibility remains a challenge, influenced by L1 conceptualization patterns in Chinese.

13Inagaki (2014) explored L1-Japanese learners’ acquisition of the mass–count distinction in English, not through article use, but via quantity judgment tasks involving proportional reasoning. Learners showed native-like patterns when interpreting canonical count nouns and substance-mass nouns (e.g., mustard), basing their judgments on number or volume as appropriate. However, their performance declined when nouns were ambiguous between mass and count uses (e.g., string vs. strings in English). Learners did not reliably shift their interpretation according to the morphosyntactic context, suggesting persistent difficulty in integrating syntactic cues with semantic interpretation. Although not a study of article use per se, these findings support the broader claim that mass noun representation is problematic for L2 learners and may contribute to article-related challenges.

14Recent studies have increasingly explored the role of atomicity—the semantic notion that some mass nouns denote discrete, countable units or ‘atoms’—as a potential universal influencing how L2 learners acquire the count/mass distinction in English. In this framework, atomic mass nouns (e.g., furniture, advice) refer to sets of individuated entities, whereas non-atomic mass nouns (e.g., water, patience) are construed as undifferentiated substances or abstract notions. Choi, Ionin, and Zhu (2018) investigated how L1-Korean and L1-Mandarin Chinese learners of English acquire the semantic property of atomicity within mass nouns. The study probed whether learners would be sensitive to the distinction between atomic mass nouns (e.g., furniture) and non-atomic mass nouns (e.g., water) in terms of their use of plural morphology. The results showed that learners from both L1 backgrounds correctly applied plural –s to canonical count nouns. However, they tended to overextend plural marking to atomic mass nouns like furniture, but not to non-atomic mass nouns like water. These findings suggest that sensitivity to atomicity (a semantic universal) guides L2 learners' morphosyntactic choices even when their L1s differ in how they encode number and individuation.

15Similarly, Tang, Fiorentino, and Gabriele (2023) contributed to this line of research by examining how L1-French and L1-Chinese learners of English handle mass nouns that differ in atomicity. Tang et al. found that both L1-French and L1-Chinese learners exhibited greater difficulty with atomic mass nouns encoding abstract concepts (e.g., advice, vocabulary), suggesting that learners draw on the concept of atomicity when interpreting the count/mass distinction in English. Importantly, the study also identified a proficiency-related pattern: only advanced learners demonstrated sensitivity to grammaticality judgments involving atomic abstract mass nouns, suggesting that the influence of semantic universals such as atomicity becomes more pronounced with increased L2 proficiency. Together, these studies point to an emerging consensus that atomicity functions as a universal semantic guide in L2 acquisition of nominal semantics and morphosyntax.

16Building on Hua & Lee's (2005) research, Snape (2008) conducted an experimental study to investigate L2 learners’ sensitivity to the count/mass distinction and their interpretation of definite descriptions. The study employed two tasks: a grammaticality judgement task to assess learners’ ability to recognize acceptable mass and count noun uses, and a forced elicitation task that tested three types of definite expressions: (a) referential definites based on shared knowledge between speaker and hearer, (b) encyclopaedic definites rooted in culturally unique references (similar to 'the White House' example in (1)), and (c) larger situation definites, which presuppose situational familiarity without reference to any specific entity. Crucially, noun type (count vs. mass) was manipulated across these conditions. The results revealed an asymmetry: while Japanese learners generally performed well on count noun items, they frequently misjudged grammatical mass noun uses and were more likely to accept ungrammatical mass plurals (e.g., *few sunshines). Spanish learners demonstrated stronger performance with mass nouns than their Japanese counterparts, but showed similar difficulty when interpreting plural mass nouns in ungrammatical contexts. These findings suggest that even learners from typologically different L1s may exhibit persistent challenges with the mass noun category, particularly when the interaction of plurality and definiteness is manipulated.

17Hermas (2019) explored how L1-Moroccan Arabic speakers acquire mass generic noun phrases in L3 English, with special attention to the influence of non-facilitative transfer. The findings revealed a clear developmental trajectory: at the pre-intermediate level, learners inappropriately used definite articles with mass generics (indefinite contexts). Intermediate learners began to alternate between definite and bare mass noun forms, suggesting an intermediate interlanguage stage in which competing grammatical representations coexisted. Advanced learners demonstrated convergence with native speaker norms, using bare nouns appropriately to express genericity. The study thus highlights the evolving role of transfer and the gradual refinement of grammatical representations in multilingual acquisition.

18Sabir (2019) also focused on Arabic learners of English and confirmed that mass nouns present a notable challenge for this population. In indefinite contexts, learners frequently used the where the null article was required, suggesting an over-reliance on definiteness and a failure to distinguish between syntactic and discourse-driven cues. Using a context-based acceptability judgment task, Sabir found that L1-Arabic learners showed a tendency to over-accept definite singular forms with English mass nouns (e.g., 'The rice contains iron'), a response pattern consistent with L1 transfer. However, contrary to expectations, L2 learners demonstrated relatively accurate judgments for bare singular mass nouns (e.g., 'Rice contains iron'). Sabir suggests that while L1 influence plays a role in shaping learner response patterns, access to Universal Grammar may support restructuring toward target-like forms, especially with increased proficiency. Taken together, Hermas’s and Sabir’s findings highlight the dual role of L1 transfer and universal constraints in the development of article use and mass noun interpretation. Both studies underscore the persistent challenge that mass nouns pose for learners whose L1 encodes definiteness and countability differently than English.

19In aggregate, these studies converge on a key insight: article acquisition in English is sculpted not only by semantic features such as definiteness and specificity, but also by the morphosyntactic properties of the noun. Mass nouns emerge as a consistent source of difficulty, especially in indefinite generic contexts where the null article is required. The present study contributes to this growing literature by focusing on L1-Slavic speakers learning English. We aim to systematically compare their article use in contexts involving mass and count nouns across definite and indefinite domains.

3. The study

20Building on the findings from previous research, which underscore the challenges that noun type and (in)definiteness pose for L2 learners of English articles, the present study aims to investigate these dimensions more closely. While prior studies have highlighted the influence of semantic features such as specificity and atomicity, fewer studies have systematically examined how count versus mass nouns interact with article use across proficiency levels. This study aims to bridge this gap by exploring article choice in pragmatically rich dialogue contexts. Our study was guided by the following research questions:

1-How accurately do L1-Slavic/L2-English learners use articles in syntactic and semantic contexts involving different noun types (count vs. mass) and levels of definiteness (indefinite vs. definite)?

2-Are there systematic differences in article use across learner proficiency levels?

3-Which combinations of noun type and definiteness are most prone to article omission or substitution errors?

21To address these questions, we designed a comprehension-based online survey, using brief conversational dialogues to elicit article judgments. The following sections describe the experimental design, report the results, and discuss their theoretical and empirical implications.

3.1 Participants

22Forty-one (41) non-native speakers (NNSs) and fourteen (14) native speakers (NSs) of English were recruited to participate in the study. Prior to the experiment, participants completed a language background questionnaire that elicited information about: (i) age at the time of testing, (ii) native language (defined as the language acquired from birth), (iii) any additional language(s) spoken in early childhood, (iv) age of first exposure to English, (v) age of arrival in the U.S., and (vi) total length of residence in the U.S.

23All NNS participants had begun to learn English as a second language in their country of birth and were residing in the United States at the time of testing. Their native languages included Russian (n = 27), Slovak (n = 7), Ukrainian (n = 4), Croatian (n = 1), Czech (n = 1), and Polish (n = 1). Five participants reported bilingual exposure to another Slavic language during childhood, with language pairs consisting of Ukrainian/Russian (n = 4) and Slovak/Czech (n = 1).

24Non-native participants also provided a self-assessment of their English proficiency by selecting one of the following categories: native-like, advanced, high-intermediate, intermediate, low-intermediate, or beginner. Based on these self-reports, participants were grouped into three proficiency levels: Native-like (n = 16), Advanced (n = 9), and Intermediate (n = 16). Table 2 summarizes participant demographics and language background data for each group.

Table 2: Participant demographics and English exposure history by proficiency group

	Number (M/F)	Mean age at time of testing	Age of exposure to English in the country of birth	Age of arrival in the US	Total years of residency in the US
Native mean (SD) range	14 (3/11)	39.9 (17.5) 18–72	--	--	--
Native-like mean (SD) range	16 (2/14)	30.4 (8.67) 22–44	9.1 (3.9) 5–19	20.6 (8.56) 8–42	8.8 (6.4) 3–23
Advanced mean (SD) range	9 (3/6)	38.4 (8.9) 26–51	9.2 (3.2) 5–15	23.8 (6.3) 15–35	13.6 (7.8) 2–23
Intermediate mean (SD) range	16 (7/9)	47 (11.1) 22–59	10.3 (2.1) 6–15	32.6 (3.8) 22–38	14.4 (9.5) 4–30

25Native-like group. The Native-like group consisted of 16 participants (2 male, 14 female), with a mean age of 30.4 years. These participants were first exposed to English in their country of birth at a mean age of 9.1 years, arrived in the United States at a mean age of 20.6 years, and had resided in the U.S. continuously for an average of 8.8 years. (Only two participants arrived in the U.S. before the onset of puberty, at ages 8 and 11.)

26Advanced group. The Advanced group included 9 participants (3 male, 6 female), with a mean age of 38.4 years. Their mean age of first exposure to English was 9.2 years, and they arrived in the U.S. at a mean age of 23.8 years. Their average length of continuous residence in the US was 13.6 years.

27Intermediate group. The Intermediate group included 16 participants (7 male, 9 female), with a mean age of 47 years. These participants were first exposed to English at a mean age of 10.3 years, arrived in the United States at a mean age of 32.6 years, and had the longest average duration of residence in the U.S., at 14.4 years.

3.2 Materials

28The experimental stimuli consisted of 48 short written dialogues, evenly distributed across four conditions based on (in)definiteness and noun type: Indefinite Count-A, Indefinite Mass-Null, Definite Count-The, and Definite Mass-The. Each dialogue took the form of an informal exchange between two speakers and contained a target noun phrase embedded within a sentence. Each target noun appeared with a blank space where participants were asked to choose the most appropriate article from the following multiple-choice menu: a, the, or no article needed. Table 3 provides a list of the target nouns used in each condition.

Table 3: Target nouns by experimental condition and noun type

Condition	Target Nouns
Indefinite Count (a)	book, trip, package, job, shop, lamp, house, notebook, coat, bike, necklace, wallet
Indefinite Mass (null)	homework, weather, evidence, success, information, land, rice, wealth, furniture, mail, happiness, sushi
Definite Count (the)	magazine, bus, party, job, website, flash drive, car, laptop, house, key, answer, exit
Definite Mass (the)	homework, weather, evidence, success, information, land, rice, cash, water, milk, patience, butter

29Each condition was matched for key structural properties: (i) all target noun phrases were first mentions (i.e., they had no explicit, prior antecedent in the dialogue), (ii) they occurred as syntactic objects of transitive predicates, (iii) they were modified by restrictive relative clauses, and (iv) each condition featured the same set of verbal predicates (e.g., look for, hope for, plan, wait for, have, own, keep, know, get, find, lose, buy) to ensure consistency in lexical environments.

Table 4: Sample experimental dialogues illustrating each article/noun type condition

(In)definiteness	Noun type
	Count-A	Mass-Null
Indefinite	Person A: I heard that John just interviewed with JP Morgan Chase! Person B: Yes, he is hoping for ___ job that could make him rich. Person A: Something in finance? Person B: Exactly!	Person A: Is Linda teaching this semester? She looks stressed. Person B: Yes, that’s because she works too much. For tomorrow, she is planning __ homework that would keep her students busy for two days! Person A: What’s it about? Person B: Irregular verbs!
	Count-The	Mass-The
Definite	Person A: What are you doing here so late? Person B: I’m waiting for ___ bus that goes by my house. It’s line 38. Person A: I don't think it comes very often. Come on, I can walk you home!	Person A: Our new police officer isn’t good, is he? Person B: No, he isn’t. Look at him. He is digging through his desk, looking for ___ evidence that we put together for him last night. Person A: Oh, he’s going to get in trouble for that.

30In addition to the 48 experimental items, we constructed 48 distractor dialogues. The distractors also consisted of short dialogues and offered a three-option multiple-choice menu.

3.3 Procedure

31The experimental task was administered using Qualtrics, a secure web-based survey platform. All participants were recruited via the author’s professional and personal network and received an anonymous link to the survey by email. They were instructed to complete the survey in a single sitting and in a quiet environment to ensure focused participation. The survey began with a demographic and language background questionnaire, which served as the first block of items. This was followed by a set of instructions introducing the task. Participants were asked to imagine the dialogues as representative of informal, everyday conversation.

32Before beginning the main task, participants completed a brief practice item to familiarize themselves with the test format. They then proceeded to the full set of experimental and distractor dialogues. The presentation order of both the dialogue items and the multiple-choice response options was randomized using Qualtrics’ built-in randomization function, ensuring that each participant received a unique item and response sequence. Upon completion of the survey, participants received a $10 digital gift card as compensation for their time.

3.4 Results

33Descriptive statistics. We begin by presenting the distribution of article use across the four experimental conditions (Count-A, Count-The, Mass-Null, and Mass-The) for each participant group: native speakers, native-like, advanced, and intermediate learners. The bar charts in Table 5 illustrate the proportion of target and non-target article choices, along with variability within groups, providing a detailed picture of accuracy and error patterns across conditions.

Table 5: Accuracy rates by experimental condition: percent correct with 95% CI

34The charts illustrate a general trend: accuracy decreases as noun type shifts from count to mass, and as proficiency declines from Native-like to Intermediate. Native speakers performed at ceiling across all four conditions, with accuracy rates consistently at or above 90%. Their responses show categorical use of the target article in each context: a in Count-A, the in Count-The and Mass-The, and null in Mass-Null. Native-like and Advanced L2 learners perform well in the count noun conditions but exhibit greater variability in mass noun contexts. Intermediate learners show the lowest accuracy overall, with particularly sharp declines in the Mass–Null and Mass–The conditions. These group-level accuracy patterns reflect the participants’ self-rated proficiency, supporting the validity of the self-assessment categories used in the analysis.

35In the Count–A condition, Native-like learners demonstrated high accuracy (93%), with minimal substitution errors. Advanced learners performed less accurately (78%), and Intermediate learners showed the lowest accuracy (71%). Both Advanced and Intermediate learners exhibited substitution of the for a, at rates of 22% and 26%, respectively.

36In the Count–The condition, Advanced learners performed best (88%), followed by Native-like learners (79%) and Intermediate learners (63%). Substitution of a for the was observed across all non-native groups, with rates of 21% for Native-like, 12% for Advanced, and 33% for Intermediate learners.

37Performance diverged more strongly in the Mass–Null condition. Native-like learners achieved 75% accuracy, whereas Advanced learners dropped to 58%, and Intermediate learners to 35%. Intermediate learners also showed a strikingly high rate of substitution with a (40%). Substitution with the was present in all groups, with error rates of 17% for Native-like, 28% for Advanced, and 25% for Intermediate learners.

38Finally, in the Mass–The condition, accuracy declined progressively across groups, with 70% for Native-like, 61% for Advanced, and 41% for Intermediate learners. Only Intermediate learners showed frequent substitution of a for the (25%). All non-native groups also exhibited elevated substitution of the null determiner in this condition, with rates of 29% (Native-like), 35% (Advanced), and 34% (Intermediate). These rates are higher than those observed in the Mass–Null condition, suggesting an overextension of the null determiner to definite mass contexts.

39ANOVA analyses. To determine whether the observed patterns in article accuracy across conditions and proficiency groups were statistically significant, we conducted a series of repeated measures analyses of variance (ANOVAs). The results of these analyses are summarized in Table 6.

Table 6: Results of repeated measures ANOVA on article accuracy by proficiency level, noun type, and condition

Effect	F-statistic and p-value
proficiency level condition level x condition noun type (count vs. mass) level x noun type definiteness level x definiteness	F (3, 47) = 39.281, p < 0.001 F (3, 146) = 16.052, p < 0.001 F (9, 146) = 2.473, p < 0.01 F (1, 154) = 44.210, p < 0.001 F (3, 154) = 4.585, p < 0.001 F (1, 155) = 0.242, p < 0.623 (ns) F (3, 155) = 1.434, p < 0.235 (ns)

40The ANOVA revealed a significant main effect of proficiency level (p < .001), confirming that article accuracy varied systematically across the four proficiency groups. There was also a significant main effect of condition (p < .001), as well as a significant level × condition interaction (p < .01). These results indicate that the effect of condition was not uniform across groups, with certain conditions being more difficult for some proficiency levels than others.

41We also tested the effect of noun type (count vs. mass) and found a significant main effect (p < .001), as well as a significant level × noun type interaction (p < .001). These findings suggest that mass nouns are more difficult overall and that this difficulty increases with lower proficiency. In contrast, the effect of definiteness (definite vs. indefinite) was not significant (p = .623), nor was the level × definiteness interaction (p = .235), indicating that definiteness alone did not significantly impact accuracy or interact with proficiency level.

Table 7: Poisson regression results: effect of self-rated proficiency group on the count of correct responses (baseline = native speakers)

==========================================================

Dependent variable:

---------------------------------------------------------------------------

Correct_Count of Responses in Each Condition

Count–A Count–The Mass–Null Mass–The

---------------------------------------------------------------------------

Native-like -0.044 -0.118 -0.165 -0.266**

(0.110) (0.118) (0.119) (0.120)

Advanced -0.225 -0.006 -0.416*** -0.398***

(0.136) (0.133) (0.152) (0.149)

Intermediate -0.312*** -0.339*** -0.915*** -0.794***

(0.118) (0.125) (0.148) (0.140)

Constant

(intercept) 2.459*** 2.362*** 2.362*** 2.391***

(0.081) (0.085) (0.085) (0.084)

----------------------------------------------------------------------------------------------

Log Likelihood -121.366 -131.807 -122.799 -125.723

AIC 250.732 271.613 253.598 259.447

==========================================================

Note: Standard errors are in parentheses. Significance levels: **p<0.05; ***p<0.01

43As shown in Table 7, Native-like learners did not differ significantly from native speakers in the Count–A, Count–The, or Mass–Null conditions, but they did show a statistically significant difference in the Mass–The condition (β = –0.266, p < .05). This result confirms that even highly proficient non-native speakers continue to diverge from native-like performance when using the definite article the with mass nouns, a pattern suggested by their lower accuracy rates (70%) and elevated substitution of the null article in that condition (29%).

44Advanced learners showed significantly lower accuracy than natives in the Mass–Null (β = –0.416, p < .001) and Mass–The (β = –0.398, p < .001) conditions. Intermediate learners were significantly different from native speakers in all four conditions, with the strongest effects observed in Mass–Null (β = –0.915, p < .001) and Mass–The (β = –0.794, p < .001). These findings strengthen the claim that article acquisition in L2 English is highly sensitive to both noun type and (in)definiteness, with mass nouns posing the greatest challenge, particularly for less proficient learners.

45Overall, the Poisson model confirms the robustness of proficiency-level distinctions found in the descriptive and ANOVA analyses, confirming the difficulty of article use with mass nouns for both native-like and lower-proficiency learners. The results also show a developmental trajectory in which learners progress from general inaccuracy (intermediate) to partial convergence (advanced) and near-native accuracy in count noun contexts (native-like), while mass noun contexts, particularly with definite reference, remain the most resistant to acquisition.

46Error analysis in the Mass-Null contexts. We now turn to an item-level error analysis in the native group and in the L2 sample, focusing on how accuracy rates vary from lower to higher proficiency. In order to assess whether some mass nouns were especially prone to errors, we examine overgeneralization patterns, such as the extension of a to mass contexts, substitution of the null article in definite contexts, and the use of the in indefinite contexts. We begin with Table 8, which summarizes the results from the native speakers.

Table 8: Distribution of article use with indefinite mass nouns among native speakers

	weather	evidence	furniture	land		info	homework	wealth	rice	mail	happiness	sushi	success
a	0	0	0		0	0.00	7.69	7.69	0.00	0.00	0.00	0.00	0.00
the	0	0	0		0	7.69	7.69	7.69	15.38	15.38	15.38	15.38	38.46
null	100	100	100		100	92.3	84.62	84.62	84.62	84.62	84.62	84.62	61.54

47Overall, native speakers showed highly accurate use of the null article with indefinite mass nouns. Uses of a were minimal, while most errors involved substitution with the, suggesting that some indefinite contexts were interpreted as definite. The item with success was particularly prone to this interpretation. Despite these occasional definiteness intrusions, the null article was applied correctly in the vast majority of cases (88% average accuracy).

48Table 9 summarizes the item-level analysis in the native-like group.

Table 9: Distribution of article use with indefinite mass nouns among native-like learners

	sushi	furni ture	happiness	mail	homework	evid ence	rice	land	weather	info	success	wealth
a	0	0	0	12.5	18.75	18.75	6.25	12.5	12.5	0	6.25	6.25
the	0	6.25	12.5	0	0	6.25	18.75	18.75	25	37.5	37.5	43.75
null	100	93.75	87.5	87.5	81.25	75	75	68.75	62.5	62.5	56.25	50

49The item-level results show that uses of a were relatively infrequent overall, with the highest substitution rates occurring only with homework and evidence (19%). By contrast, substitution with the emerged as the more typical error type. These errors were particularly pronounced with nouns such as weather, information, success, and wealth, with the rates ranging between 25–44%. Recall, however, that Poisson regression revealed no statistically significant differences between native-like learners and native speakers in the Mass–Null context overall, suggesting that these item-level effects do not amount to systematic divergence at the group level.

50Table 10 summarizes the item-level analysis in the advanced group.

Table 10: Distribution of article use with indefinite mass nouns among advanced learners

	wealth	happiness	sushi	rice	mail	land	success	furni ture	homework	weather	evid ence	info
a	22.22	0.00	22.22	0.00	22.22	0.00	11.11	11.11	33.33	22.22	22.22	0.00
the	0.00	22.22	0.00	33.33	11.11	33.33	33.33	33.33	22.22	33.33	44.44	66.67
null	77.78	77.78	77.78	66.67	66.67	66.67	55.56	55.56	44.44	44.44	33.33	33.33

51In the Advanced group, the most frequent error type was substitution with the, a pattern that extended to nearly all nouns except wealth and sushi (and only minimally to mail). The highest substitution rates with the were found for evidence (44%) and information (67%), while the remaining nouns ranged between 22–33%. This pattern suggests that indefinite mass contexts were often misinterpreted as definite, in contrast to indefinite count nouns, highlighting an asymmetry between mass and count contexts. Accuracy in the Mass–Null condition was statistically lower than that of native speakers in the Poisson regression analysis, suggesting that this pattern represents a developmental stage in the acquisition trajectory. Overextension of a was also observed, though at much lower rates, affecting most nouns, except happiness, rice, land, and information.

52Table 11 summarizes the item-level analysis in the intermediate group.

Table 11: Distribution of article use with indefinite mass nouns among intermediate learners

	rice	sushi	land	furni ture	success	wealth	happiness	mail	info	homework	weather	evid ence
a	37.5	37.5	31.25	43.75	31.25	25	25	62.5	43.75	43.75	43.75	50
the	6.25	6.25	18.75	12.5	25	31.25	31.25	6.25	31.25	43.75	43.75	43.75
null	56.25	56.25	50	43.75	43.75	43.75	43.75	31.25	25	12.5	12.5	6.25

53In the Intermediate group, all mass nouns show strong overextension of a, suggesting that learners frequently miscategorized mass nouns as count. This likely reflects L1 influence since in the learners’ L1s mass nouns are grammatically singular and therefore more easily associated with indefinite singular marking. The rates of a-errors are striking, ranging from 43–44% for weather, homework, information, furniture, to 50% for evidence, and peaking at 63% for mail. Overall, this indicates that overgeneralization of a was the dominant error type at this proficiency level, surpassing misuse of the. Poisson regression confirmed that the Intermediate group’s performance was statistically different from native speakers in the Mass–Null context.

54Taken together, these findings suggest that Intermediate learners represent a developmental stage in which most mass nouns are systematically miscategorized as count nouns. By the advanced level, this type of misanalysis somewhat abates, but a new pattern emerges: once mass nouns are correctly analyzed as mass, learners tend to overuse the in indefinite contexts. (In other words, substitution of the for the null article becomes the dominant error type.) This suggests that learners face persistent difficulties in mapping indefiniteness features onto mass referents, often treating them as uniquely identifiable or specific when they are not. Future research should explore why assigning an indefinite interpretation to mass nouns poses greater challenges than with count nouns in otherwise comparable discourse contexts (recall that our experimental stimuli controlled for discourse conditions across both noun types).

55Error analysis in the Mass-The contexts. We now turn to item-level results in the Mass-The condition. Table 12 summarizes the patterns in the control group.

Table 12: Distribution of article use with definite mass nouns among native speakers

	success	weather	rice	water	butter	evidence	h/w	info	land	milk	patience	cash
a	0	0	0	0	0	0	0	0	0	0	0	0
null	0	0	0	0	0	15.38	7.69	15.38	15.38	15.38	15.38	23.08
the	100	100	100	100	100	84.62	92.31	84.62	84.62	84.62	84.62	76.92

56In the native speaker group, accuracy rates in the definite mass noun contexts were uniformly high, with most nouns at or above 85% correct use. The one exception is cash, where accuracy dropped to 77%, suggesting that this one context was sometimes interpreted as indefinite. The only source of error is occasional substitution of the for the expected null article.

57Table 13 presents the results for the native-like group.

Table 13: Distribution of article use with definite mass nouns among native-like learners

	success	h/w	weather	info	butter	water	rice	land	cash	patience	evidence	milk
a	6.25	0	0	0	0	0	0	0	0	0	0	0
null	6.25	12.5	18.75	18.75	18.75	18.75	25	31.25	43.75	50	50	56.25
the	87.5	87.5	81.25	81.25	81.25	81.25	75	68.75	56.25	50	43.75	43.75

58Accuracy in the native-like group ranged between 69–88% for 8 out of 12 nouns. Importantly, there was virtually no overgeneralization of a, with the exception of success, which yielded the highest accuracy rate (87.5%). Substitution errors primarily involved the null article, and these varied considerably across items. Three nouns—patience, evidence, and milk—showed the highest substitution rates (50–56%); cash and land followed with 44% and 31%. For the remaining nouns, substitution errors ranged between 19–25%, with the lowest substitution rates observed for success and homework. These patterns suggest that while native-like learners are largely successful in supplying the, they continue to misanalyze some definite contexts, treating them as indefinite. Moreover, Poisson regression revealed a statistically significant difference from native speakers in this condition, indicating that even at native-like levels, learners experience difficulty in mapping definiteness features onto mass referents.

59Table 14 presents the results from the advanced group.

Table 14: Distribution of article use with definite mass nouns among advanced learners

	h/w	evid ence	rice	butter	weather	info	land	water	success	cash	pati ence	milk
a	0	11.11	0	0	11.11	0	0	0	11.11	0	0	11.11
null	11.11	11.11	22.22	22.22	22.22	33.33	33.33	33.33	33.33	66.67	66.67	66.67
the	88.89	77.78	77.78	77.78	66.67	66.67	66.67	66.67	55.56	33.33	33.33	22.22

60The advanced group displayed accuracy between 67–89% for 8 out of 12 nouns. No consistent overextension of a was observed, apart from a minimal 11% with evidence, weather, success, and milk. The dominant error type was substitution with the null article, which reached particularly high rates for milk, patience, and cash (67% each). This distribution highlights item-specific vulnerabilities, with some mass nouns contexts more prone to misanalysis as indefinite. These patterns indicate that advanced learners, just as native-like learners, often failed to map definiteness onto some mass referents.

Table 15: Distribution of article use with definite mass nouns among intermediate learners

	evid ence	info	rice	h/w	land	water	weather	cash	success	butter	pati ence	milk
a	25	25	18.75	18.75	25	12.5	43.75	12.5	31.25	37.5	6.25	37.5
null	6.25	12.5	25	31.25	25	37.5	12.5	50	43.75	37.5	75	56.25
the	68.75	62.5	56.25	50	50	50	43.75	37.5	25	25	18.75	6.25

61For the Intermediate learners, performance in definite mass contexts shows a marked struggle compared to the more advanced groups. Overgeneralization of a is frequent and widespread. All nouns show some substitution of a, with rates ranging from 6% (patience) to 44% (weather), pointing to a tendency to misclassify mass nouns as countable. Null article substitution was also high, with the highest rates found for patience (75%), milk (56%), and success (44%). Other nouns (cash, butter, water, homework, and land) also showed elevated substitution rates of 31–50%, while nouns like evidence/information (12.5%) and weather (6.25%) were less affected.

62Although some nouns such as evidence (69%) and information (63%) showed moderately high accuracy rates with the, overall accuracy was low and inconsistent across nouns, dropping as low as 6% for milk. Taken together, these results indicate that intermediate learners are at a developmental stage distinct from advanced and native-like speakers. Their stage is characterized by instability in both the mass/count distinction and failure to map definiteness features onto the majority of mass referents.

4. Discussion

63The findings of this study confirm that both noun type and proficiency level significantly influence article accuracy in the L2 English of L1-Slavic learners. As Figure 1 shows, accuracy declines in a gradient pattern: first as proficiency decreases, and second as noun type shifts from count to mass. This pattern is consistent with previous research suggesting that mass nouns pose unique challenges for L2 learners (Inagaki, 2014; Snape, 2008; Yoon, 1993).

64In the Count–A condition, learners at all proficiency levels show relatively high accuracy. While substitution of the for a increases slightly at lower proficiency levels (22% and 26% in Advanced and Intermediate learners, respectively), the target indefinite article a remains dominant. This suggests that count-singular contexts with indefinite reference (existential and non-existential) are learned earlier and more robustly, possibly due to their frequency in input and clearer morphosyntactic marking.

65The Count–The condition yielded high overall accuracy among the non-native groups (63% to 88% range). Nonetheless, a substitution pattern emerged: learners across all levels occasionally used a instead of the, with the highest rates observed in Intermediate learners (33%). These results are consistent with previously reported bidirectional substitution patterns in article choice with singular count nouns (e.g., Ionin, Ko & Wexler, 2004; Tryzna, 2009).

66Mass noun conditions produced the lowest accuracy scores across the L2 groups, underscoring the difficulty of integrating (in)definiteness with noun type in article use. In the Mass–Null condition, Intermediate learners were particularly inaccurate, correctly selecting the null article only 35% of the time. Their most frequent error – substituting a for the null article – reflects overgeneralization of the singular indefinite marker to contexts where it is ungrammatical. This error pattern is most likely a consequence of miscategorization of some mass nouns as count nouns. Overuse of the was also observed across all proficiency levels. These findings align with earlier studies showing that L2 learners often misapply article forms in mass noun contexts, particularly when semantic individuation is unclear (Hermas, 2019; Sabir, 2019; Snape, 2008).

67The Mass–The condition emerged as the most diagnostic in distinguishing native-like from native performance. While native speakers achieved 92% accuracy, Native-like learners fell to 70% and substituted the null article nearly a third of the time. Error patterns in the mass noun contexts were bidirectional: learners of all proficiency levels substituted the null determiner for the and the other way around. In addition, the overgeneralization of a to (in)definite mass noun contexts in the Intermediate group suggests miscategorization of some mass nouns as countable, a pattern consistent with Yoon’s (1993) and Wu’s (2025) findings.

68Statistical analyses support these descriptive trends. The ANOVA revealed significant main effects of proficiency level, condition, and noun type, along with two key interaction effects: level × condition and level × noun type. These findings confirm that the difficulty of article selection is not uniform across contexts and is sensitive to mass/count distinctions, especially at lower proficiency levels. By contrast, (in)definiteness as a main effect was not significant, suggesting that learners may rely more on noun type and referentiality cues than on (in)definiteness alone.

69The Poisson regression analysis further clarifies these effects. When native speakers were used as the baseline, Intermediate learners were significantly different in all four conditions, with the largest divergences in the Mass–Null and Mass–The conditions. Advanced learners showed selective difficulty in mass contexts, particularly with Mass–The (p < .001) where substitution of null article for the was common (35%). Even Native-like learners differed significantly from native speakers in the Mass–The condition (p < .05), confirming that this context remains a persistent challenge even at native-like levels.

70This progression highlights a developmental trajectory in which learners move from broad misanalysis of mass nouns at the intermediate stage to a largely correct grammatical treatment of mass nouns as mass at the advanced stage. At this point, however, errors shift toward bidirectional substitution between the and the null article, revealing persistent difficulty in mapping (in)definiteness onto mass referents. By the native-like level, these substitution errors become more restricted, occurring primarily in Mass-The contexts, while performance in the Mass-Null condition becomes statistically indistinguishable from native speakers. Taken together, these findings suggest that the acquisition path involves an initial restructuring of mass/count distinctions, followed by a refinement of (in)definiteness mapping, with native-like performance reached only once both dimensions are stabilized.

5. Conclusion

71The current study provides new evidence that L2 learners’ article acquisition is shaped by complex interactions among noun type, (in)definiteness, and proficiency. Mass nouns – especially in definite contexts – pose enduring challenges even for advanced learners. These findings have implications not only for theories of L2 article acquisition (e.g., the Fluctuation Hypothesis) but also for instructional approaches: pedagogical interventions should focus more explicitly on article use with mass nouns and integrate semantic-pragmatic features that underlie their distribution. More targeted instructional studies could assess whether explicit teaching of article use in mass noun contexts improves learner outcomes, particularly in higher-proficiency groups where bidirectional substitution errors (the for null and null for the) persist.

72Future research should continue to explore how semantic and pragmatic features interact to shape article use with mass nouns in L2 English. While the current study highlights clear proficiency-related patterns, further work is needed to disentangle what types of mass nouns (e.g., concrete vs. abstract or atomic vs. non-atomic) are more prone to misclassification and article errors. Additionally, experimental designs incorporating both comprehension and production tasks could offer a more complete view of how learners represent and use articles in mass noun contexts. Cross-linguistic comparisons involving learners from article-less languages versus those with article systems could also help refine models of transfer and universal semantic influence.