A corpus-based study of the acquisition of the English progressive by L1 Chinese learners: from prototypical activities to marked statives

2.1 The Aspect Hypothesis

In both L1 and L2 acquisition, learners are observed to be sensitive to the inherent lexical aspect of verbs in the acquisition of tense-aspect morphology. The Aspect Hypothesis (henceforth AH, Andersen and Shirai 1994; Bardovi-Harlig 2000) comprises four claims:^[3]

1. Learners first use past marking (e.g., as in English) or perfective marking (Chinese, Spanish, etc.) on achievement and accomplishment verbs, eventually extending its use to activities and stative verbs.

2. In languages that encode the perfective/imperfective distinction, imperfective past appears later than perfective past, and imperfective past marking begins with stative verbs and activity verbs, extending to accomplishment and achievement verbs.

3. In languages that have progressive aspect, progressive marking begins with activity verbs, then extends to accomplishment or achievement verbs.

4. Progressive markings are not overextended to stative verbs.

(Andersen and Shirai 1996: 533; Bardovi-Harlig 2000: 227; Shirai 2009: 173; originally proposed in Shirai 1991: 9–10)

The AH thus predicts that learners are strongly influenced by verbal semantics in acquiring grammatical markers of tense and aspect. That is, past perfective markers are associated with telic verbs/verb predicates (achievements and accomplishments), while general imperfective markers are associated with atelic verbs (activities and states), and progressive markers with activity verbs.^[4] Robust evidence in SLA has been found that learners acquire English past tense markings spreading from telic verbs to other verb types (Bardovi-Harlig 1998, 2000). However, mixed findings of the use of the progressive marking have been observed, especially with regard to the use of stative verbs in the progressive (Bardovi-Harlig and Bergström 1996; Fuchs and Werner 2018; Robison 1990; Rocca 2002). The present study investigates the third and fourth claims of the AH; namely, whether L2 learners’ progressive inflections in English show a strong association with activities at the earlier stages, and then later spread to other verb types, and whether learners overextend the progressive inflection to stative verbs.

Some studies have investigated whether the extension of the progressive aspect to stative verbs is related to proficiency levels, or written versus spoken mode. Housen (2002: 104) observed that only lower intermediate learners were characterized by overextension of the progressive –ing to stative verbs, whereas more advanced learners appeared to have obtained native-like semantic control over the use of the progressive -ing. Meriläinen et al. (2017) suggested, based on the analysis of the corpora of native, ESL, and EFL varieties, that less advanced learners are more likely to overextend the progressive to stative verbs than highly advanced learners.

With regard to the effect of the mode of expression, Hundt and Vogel (2011) found that the extended use of progressives by ESL learners is more typical in spoken than in written English. Tracy-Ventura and Myles (2015) noted that not all task types may be equally successful in eliciting the full range of morphological tense-aspect contrasts and found that less controlled tasks encouraged fewer instances of more advanced features, such as imperfective past in Romance languages, part of which corresponds to past progressive marking in English.

In sum, proficiency level, mode (written vs. spoken), and task type all play some role in learners’ use of tense-aspect marking. However, the interplay of these factors in the acquisition of the English progressive has not been systematically investigated.

As noted above, in the present study we focus on the third and fourth claims of the AH – whether learners’ progressive marking shows a strong association with activities and then extends to other verb types, and whether learners incorrectly extend the progressive marking to states, and the effects of factors other than lexical aspect, namely proficiency and mode (spoken vs. written).

2.2 The progressive aspect in English

Numerous linguistic studies (e.g., Kranich 2010; Smith 1983; Smitterberg 2005) present a detailed description of the English progressive aspect. Primarily, the progressive, also commonly referred to as continuous aspect, is used to express the duration and the temporary validity of the situation. In terms of how the progressive interacts with the four lexical aspect classes, activity verbs are prototypically associated with the progressive marking, which mostly describes an ongoing event, e.g., He is running.^[5] English achievement verbs with progressive markings often express a process leading up to an endpoint (e.g., He’s reaching the summit). Stative verbs by contrast are often incongruent with the progressive aspect (*She is knowing the answer), arguably due to a mismatch between the progressive aspect, which presents an ongoing or dynamic situation at the reference time, and the stative verb, which usually indicates a homogeneous, static situation that lasts timelessly. However, some stative verbs do appear in the progressive, e.g., I am feeling fine.^[6]

2.3 The stative progressive in English

The occurrence of the progressive -ing with stative verbs is often regarded as ungrammatical in ESL and grammar reference books (*He is knowing English. Quirk et al. 1985: 75). It is often seen as a non-target-like use if used by L2 learners (Van Rooy 2014). However, grammatical sentences like 1(a) and (b) are not infrequent.^[7]

Sentence (1a) conveys the meaning of temporariness of loving, which presents the speaker’s changing process of inner emotions, while sentence (1b) expresses a dynamic view of the state of looking forward to something. According to Smith (1983), the progressive sets up the expectation for change, and in this sense, it can convey the notion of temporariness, as in the example heard by Brown (1973: 323) on an airplane: Are you wanting your suitcase down? Stative verbs, as Smith (1983) noted, can be progressivized to impose viewpoint aspect to present a state as a dynamic situation.

Recent diachronic change in the use of the progressive appears to make stative progressives more frequent and acceptable in different varieties of English. Kakietek (1997: 85) concluded, based on a corpus of several British and American contemporary novels and magazines, that in current usage, despite the traditional analyses, stative verbs do not constitute a separate syntactic category, and that they can freely take progressive forms in appropriate contexts. Mair (2006) and Smitterberg (2005) reported a tendency to increasingly apply the progressive to stative verbs in the twentieth century. Recently, the subjective uses of the progressive, rather than conveying aspectual meaning, are claimed to be of increasing importance in connection with the spread of the progressive (Kranich 2010; Rautionaho 2014). Subjective progressives often portray the speaker or writer’s attitude toward a situation to convey tentativeness and vividness. Moreover, stative progressives have recently been widely reported in Inner, Outer, and Expanding Circle Englishes (Meriläinen et al. 2017; Paulasto 2014; Salles-Bernal 2015; Van Rooy 2014).

Despite the expanding use of stative progressives in English, some stative verbs are still categorically incompatible with the progressive, as exemplified by the example cited above (*He is knowing English). Since it is not easy to determine which uses of stative progressives are grammatical and which ones are ungrammatical because of the fuzzy boundaries between the two (Shirai 1994), operationally defining “overextension” has to depend on either the adult native speaker’s intuition (as in the present study, to be discussed below), which involves a certain degree of individual variation, or on how native speakers use them in corpus data (e.g., Rautionaho and Fuchs 2021).

2.4 The English stative progressive in L2 acquisition

Divergent findings concerning the use of the progressive with stative verbs have been reported in SLA. Here, we address the issue of how learners’ use of progressives is associated with the lexical aspect of verbs in terms of the AH (claim 3) and whether stative progressives are observed in L2 acquisition (claim 4), and what factors determine their presence or absence.

To our knowledge, language transfer is the first explanation offered to account for the occurrence of stative verbs in the progressive. Robison (1990) found that in the interlanguage of an L1 Spanish learner, stative verbs appeared either in the base form or in the progressive, with 22% of all stative verbs occurring with the English progressive -ing (both based on token count). Commenting on Robison (1990), Andersen and Shirai (1996) suggested that language transfer of the imperfective aspect in Spanish may be a possible explanation for overgeneralized stative progressives, because the learner may associate the progressive –ing with the imperfective aspect in their L1. Consistent with this view, Rocca (2002) found, in a bi-directional study of English and Italian child L2 acquisition, that the English L2 data from Italian children showed a high incidence of stative progressives. For example, the stative verb want occurred in bare progressives without auxiliary “be” (e.g., he wanting). Rocca argued that the stative progressives could be the result of the learners transferring a prototypical feature (i.e., stative) of the Italian imperfective (imperfetto) to the progressive in English. These studies suggest that the overgeneralization of the progressive to stative verbs may be an effect of language transfer from Romance imperfective (Spanish and Italian).

Another possible explanation discussed in the literature for the absence or occurrence of the stative progressive is the effect of instruction. Bardovi-Harlig (1998) suggested that the reason why learners in Robison (1990) and Rohde (1996) produced ungrammatical stative progressives is that they had little or no English instruction. Other studies (e.g., Bardovi-Harlig 2012a; Bardovi-Harlig and Bergström 1996; Bardovi-Harlig and Reynolds 1995) found that learners who were tutored in English used progressives with stative verbs only infrequently. Bardovi-Harlig and Reynolds (1995) conducted a cross-sectional study with 182 adult classroom ESL learners (mixed L1s) at six levels of proficiency. In a completion task, learners were given the base form of the verb and asked to supply the missing word(s) in cloze passages. Results showed that stative verbs were mostly inflected in the present tense, and activity verbs in the progressive. Only negligible use of stative progressives was reported. Bardovi-Harlig and Bergström (1996) used a film retell task to compare the written narratives from L2 English learners (mixed L1) with narratives collected from L2 French (English L1) learners, which revealed the effect of lexical aspect on the distribution of tense-aspect morphology in support of the AH in both target languages. Regarding the L2 English group (mixed L1s), the lowest-proficiency learners used either present progressives or bare progressives. The intermediate level learners’ past progressive emerged with activity verbs. Unlike the adult-learner in Robison (1990), who had very limited formal English instruction, these instructed learners did not overextend the progressive to stative verbs.

In contrast to the mixed L1 learners of English studied by Bardovi-Harlig and associates, some instructed learners show extensive use of stative progressives. Collins (2002) conducted two cross-sectional studies with undergraduate ESL learners who were native speakers of Canadian French, following the methodology in Bardovi-Harlig and Reynolds (1995). The results were generally consistent with the AH: the past morphology was robustly associated with telic verbs and the progressive with activity verbs. However, a substantial number of stative verbs (e.g., look, smell, and think) were incorrectly used in the progressive aspect by the L1 French instructed learners.^[8] This may suggest that Romance L1 is a good predictor of the presence of stative progressives, consistent with the L1 Spanish and Italian learners reviewed above.

Although earlier studies that reported results relevant to L2 stative progressives are mostly on ESL learners, more recent research has focused on EFL learners, and in these studies, proficiency appears to influence the use of stative progressives. Housen (2002) conducted a cross-sectional corpus-based study of L1 Dutch- and L1 French- EFL learners and native English speakers as a baseline. Although at the lower level of proficiency, learners’ initial use of -ing is mainly restricted to activity verbs, the –ing form gradually extends to achievements and states and increasingly functions as an imperfective aspect marker. At the higher level, learners’ use of stative progressive resembles that of native speakers. Specifically, Housen grouped learners’ transcripts into four proficiency levels in terms of lexical and grammatical indices. Results show the overextension of –ing to highly stative verbs like seem, know, want, and be initially appear in the low proficiency level. The learners in the high proficiency group no longer overextend it to stative verbs of cognition and perception, though the percentage of stative progressive is 16% and its pattern of distribution closely resembles that of the native speaker group (15%). The author suggested that this distributional bias of verb morphemes towards a certain type of verb in learner language reflects a similar distributional bias in the speech of native speakers, which serves as input for the language learners.^[9] Fuchs and Werner (2018) investigated the frequency of stative progressives in L2 English in the beginning and lower intermediate instructed EFL learners of three L1s with progressive marking (Mandarin/Cantonese, Japanese, and Spanish) and three L1s without (Polish, German, and Hebrew) in a written corpus (International Corpus of Cross-linguistic Interlanguage; Tono and Díez-Bedmar 2014). Results show that beginning/intermediate level learners, regardless of their L1, rarely use stative progressives. However, stative progressives are found mostly in contexts that are grammatical in native usage, especially by learners with a progressive in their L1, supporting Claim (4) of the Aspect Hypothesis.

Advanced German learners of English in Dose-Heidelmayer and Götz (2016) significantly underused the progressive in the spoken language (LINDSEI-GE subcorpus; Gilquin et al. 2010) and redundantly used the progressive in non-required contexts, compared with the native speakers (LOCNEC; cf. De Cock 2004). Further analysis showed that the misuse of the progressive was related to some verbs describing habitual states and activities.

The above review suggests that (1) L1 Romance learners consistently show the use of stative progressives (Robison, Rocca, Collins), except for Fuchs and Werner’s L1 Spanish learners and Housen’s L1 French learners, (2) Instructed learners do use stative progressives if the L1 is Romance (Collins), (3) Whether the learners’ L1 has progressive marking is not an important predictor (Fuchs and Werner), and (4) proficiency appears to be a contributing factor.

The tasks or text types have been argued to be responsible for the different patterns of use of the progressive. Bardovi-Harlig (1998) reported that the overextension of progressive to states was not observed in her elicited narrative production study nor in Bardovi-Harlig and Reynolds (1995), which was a study based on an elicited task of cloze passages. Bardovi-Harlig (1998) suggested that no extension of the progressive to states can be attributed to the elicited task which does not draw on speakers’ personal experience in the same way that a personal narrative might, since learners have little opportunity or need to report habitual (background) activities or states (Bardovi-Harlig 1998: 500). Therefore, text type as a factor may be important.

In sum, the accounts based on language transfer, instructional effect, learners’ proficiency, and task type have been proposed to influence the use of the progressive marking to stative verbs in L2 acquisition. However, the picture of how these factors interact is not clear yet. For example, the effect of task, including the mode (spoken/written), and proficiency are not investigated systematically, rendering stative progressives still under-investigated in SLA.

The present study investigates the development of progressive marking in L1 Chinese EFL learners, with particular emphasis on the effect of lexical aspect, to address Claims (3) and (4) of the Aspect Hypothesis. Three research questions were asked:

1. What is the overall frequency of the English progressive marking used by Chinese-speaking learners and native speakers in spoken and written corpora?

2. What is the general distribution pattern of the progressive marking in relation to lexical aspect and how do they differ in learners of different proficiency levels and in native speakers?

3. What are the distribution patterns of stative verbs with progressive marking in different modes of production (spoken vs. written), and how do they differ in learners of different proficiency levels and in native speakers?

3.1 The data

The learner corpus analyzed in the present study is The Spoken and Written English Corpus of Chinese Learners (SWECCL; Wen et al. 2008), consisting of data from English tests (see more details below). It contains a spoken subcorpus (SECCL) and a written subcorpus (WECCL) from L1 Chinese college students studying English as a foreign language in mainland China (age range 18–23). They received at least six years of formal instruction of English in China. The corpus totals approximately 100 million words. https://heep.unipus.cn/product/book.php?BookID=427.

A subset of transcribed texts in the spoken subcorpus on five topics (see Appendix 1) of TEM 8 (Test of English Majors, Band 8), a standardized speaking test for English majors widely used in mainland China, was retrieved. Topics of the data sampled for the current study involve making a comment on a given topic, such as “your view about China’s employment market challenged by more graduates”, “your suggestions for 2008 Olympic Games”, and “your view about whether city citizens should keep pets at home or not” (see Appendix 1 for the complete list). To ensure that both spoken and written data have measurable differences in proficiency between the learners representing intermediate and advanced groups, we had native raters classify learner data sampled for the current study into CEFR proficiency levels. Two native speakers of English with extensive EFL teaching experience in China independently rated the transcribed speech data (from 906 learners) on the basis of descriptors for speaking of the CEFR (Common European Framework of Reference) for Languages (Council of Europe 2001). 462 texts, totaling 145,149 words, which were rated B1 and B2, and 444 texts, totaling 138,129 words, which were rated C1 and C2 are grouped as two proficiency levels of “intermediate” and “advanced” respectively (see Table 1). To test the inter-rater reliability of coding, Cohen’s kappa was calculated. The resulting scores of κ = 0.967 and p < 0.001 suggest that the reliability of the two raters’ classifying texts into B1/B2 versus C1/C2 was very high. The two raters resolved the 15 cases of disagreement through discussion. These 906 texts were the subsample analyzed in the current study.

To match the spoken data, written essays on five topics that are similar to the ones chosen for the spoken data were retrieved from WECCL (see Appendix 1). Prompts used for the writing tasks are about learners’ view on a topic, such as, “your view about whether education is a life-long process”, “your view about whether university education is to prepare students for employment”, and “your view about whether the animals should be treated as pets.” The same two native raters were asked to rate the essays based on the descriptors for writing of the CEFR for Languages. 579 texts that were rated B1 and B2, totaling about 149,791 words, qualify as intermediate learners’ written data, and 538 texts that were rated C1 and C2, totaling about 141,214 words, qualify as advanced learners’ written data. Cohen’s kappa value suggested the agreement between the two raters to be of a satisfactory level (κ = 0.970, p < 0.001). The two raters resolved the 17 cases (out of 1,117 texts) of disagreement through discussion.

To compare the learner data with native-speaker data, we used the Corpus of Contemporary American English (COCA, https://www.corpusdata.org/). To maximize comparability with the learner data, in selecting the native-speaker data, we tried to ensure similarities in terms of topic and discourse type.^[10] With regard to topic, for native-speaker written data, we selected writings from COCA magazine (mostly from 1995 to 2014) with topics similar to the learners’ written data (education, animals, employment and education, child education, students and campus living) totaling 159,723 words in 194 texts. For the native spoken data, we selected speech from COCA spoken (mostly from 1995 to 2014) on topics similar to the spoken learner data (spaceships/space exploration, employment and education, the Olympics, fireworks and festival celebration, pets) totaling 147,250 words in 248 texts. In terms of discourse type, both written and spoken data in the L2 corpus consist of speech or writing in which the learners express their opinions on a specific topic. Furthermore, in choosing the data from COCA spoken data, which mostly consists of unscripted,^[11] TV/radio interviews, we chose to analyze segments where a monologue of 300 words or longer is included. This is because the learners’ spoken tasks elicited monologues, which typically were about 300–400 words in length. Although this is not ideal, it enhances the comparability of the interview data, which is more interactional than the monologue data. The complete list of the native data, both spoken and written, chosen for analysis is in Appendix 2.

These corpora are comparable across modes (spoken vs. written): the spoken and written data from both native speakers and learners are argumentative, expressing one’s own ideas on certain topics. The learner data are also comparable across proficiency levels (intermediate vs. advanced), rated on the basis of the same criteria (CEFR), in that both levels of learners were given comparable topics within modes (spoken vs. written).

3.2 Data coding

The spoken and written data were tagged using the Tree Tagger software, a tool for automatic annotation of text corpora with part-of-speech and lemma information. Then all instances of VVG, which is the POS tag of lexical verbs in the gerund or present participle form in the Tree Tagger set, were retrieved with AntConc 3.5.0. and entered into Microsoft Excel spreadsheets for manual checking to make sure they were true progressives.

The following criteria were used to determine what we considered as “true” progressives. First, periphrastic forms (e.g., be going to V) were excluded.^[12] Second, repeated verbs in one sentence were counted only once to avoid skewing the results. For example, in the sentence “First, with the society’s highly development, people’s work is getting more and more, getting more and more busy”, the verb “getting” is counted only once. Third, if two sentences were adjacent and repeated, only the second sentence was analyzed. For example, only the second lagging is counted in “It was lagging. It is lagging far behind the developed countries <country>.” Fourth, sentences with unknown intended meaning were excluded. For example, “…the city is holding will spend a large sum, err, on construction”. Progressive forms without the auxiliary (bare progressive), such as “he working” were included, because they indicate learners’ attempts to use the progressive form. After applying these criteria, 407 progressive tokens from SECCL-I (Spoken-intermediate), 552 from SECCL-A (Spoken-advanced), 354 from WECCL-I (Written-intermediate), 467 from WECCL-A (Written-advanced), 980 from COCA-Spoken, and 653 from COCA-Written were obtained (altogether 3,413 tokens). Following the criteria proposed in earlier studies (Dowty 1979; Shirai 1991; Vendler 1957), we classified verb predicates used with progressive marking into four lexical aspect types: activity (ACT), accomplishment (ACC), achievement (ACH), and state (STA). For example,

Specifically, two native speakers of English, both linguists with background knowledge in tense-aspect research, independently coded all 3,413 verbs following the system of verb classification in Shirai (2013, see Appendix 3). Cohen’s Kappa was calculated to assess the inter-rater reliability of the two raters’ classification. The kappa value was 0.984 and p < 0.001, which indicates a satisfactory level of reliability. In case of disagreements (31 cases out of 3,413), the two raters resolved them through discussion.

In addition, to investigate whether learners’ stative verbs in progressives are acceptable uses or overextensions, both of the two native speakers judged the acceptability of all 109 tokens of stative progressives. The inter-rater reliability of the two raters was 99.1% (i.e., only one case of disagreement, which they resolved through discussion).^[13]

4.1 Overall frequency

The first analysis presents the overall frequency distribution of the English progressive marking. Table 2 provides an overview across the six (sub)-corpora in terms of verb types, verb tokens, corpus size, and relative verb token frequency per million words (pmw).

Table 2 indicates that native speakers use the progressive more frequently than advanced learners, who use it more often than the intermediate learners, in both the spoken and written corpora. Significant differences^[14] exist between the three groups in the spoken subcorpora (Spoken-intermediate and Spoken-native: χ² (1) = 228.84.00, p < 0.01; Spoken-advanced and Spoken-native: χ² (1) = 93.88, p < 0.01; Spoken-intermediate and Spoken-advanced: χ² (1) = 29.47, p < 0.01, and in the written subcorpora (Written-intermediate and Written-native: χ² (1) = 70.40, p < 0.01; Written-intermediate and Written-advanced: χ² (1) = 22.68, p < 0.01; Written-advanced and Written-native: χ² (1) = 12.13, p < 0.01.^[15] Progressives are much more frequent in spoken than in written corpora, as suggested in previous research (e.g., Leech et al. 2009). Chi-square tests indicate that the differnce in the overall frequency of progressives in the spoken corpora compared to the written corpora is significant for the intermediate learners (χ² (1) = 5.39, p < 0.05), for the advanced learners (χ² (1) = 8.94, p < 0.01), and for the native speakers (χ² (1) = 94.93, p < 0.01).

4.2 Distribution of progressives across lexical aspect classes

Second, we examined the effect of lexical aspect (states, activities, accomplishments, and achievements) on the use of progressive marking. Table 3 displays the results of the analysis based on tokens (but type counts also provided) of the progressive –ing across lexical aspectual classes in the spoken subcorpora of two proficiency levels of learners and native comparison data (spoken), and Table 4 shows the results for the written subcorpora and native comparison data (written).

The tables show that the progressive does not evenly distribute over lexical aspect categories. The spoken and written data exhibit similar distributional patterns of the progressive by learners and native speakers in that the token frequencies of activity verbs (49.14% by the intermediate learners, 66.30% by the advanced learners, and 66.02% by the native speakers in the spoken corpus; 47.74% by the intermediate learners and 64.88% by the advanced learners and 62.02% by the native speakers in the written corpus) are much higher than the other verb classes. The four lexical aspectual classes are ordered: activity > accomplishment > achievement > state in all three groups.^[16]

The effect of lexical aspect on the use of progressive marking presents a slightly different picture across proficiency and mode (spoken vs. written). The data indicate that the association of the English progressive -ing with activities is stronger among more advanced learners. A cross-tabulation analysis was performed to examine the differences in the three groups as a function of lexical aspect in spoken subcorpora (Table 3). Pearson chi-square, with Bonferroni correction applied for multiple comparisons, revealed a significant difference among the three groups as a function of lexical aspect (χ² (6) = 73.18, p < 0.01). In the spoken corpora, Chi-square tests show a significant difference between the two learner groups (χ² (1) = 27.83, p < 0.01), and between the native speakers and the intermediate group (χ² (1) = 33.76, p < 0.01), but no significant difference between the native speakers and the advanced group (χ² (1) = 0.003, p > 0.05) in terms of the progressive used with activity verbs. The advanced learners, like native speakers, associate the progressive with activity verbs more strongly than the intermediate learners in speech. We also investigated the differences in the three groups as a function of lexical aspect in written subcorpora using cross-tabulation analysis (Table 4). Pearson chi-square, with Bonferroni correction applied for multiple comparisons, revealed a significant difference in the three groups as a function of lexical aspect (χ² (6) = 49.91, p < 0.01). In writing, there is also a significant difference between the intermediate and advanced learners (χ² (1) = 23.52, p < 0.01), and between the native speakers and the intermediate learners (χ² (1) = 18.52, p < 0.01), but no significant difference between the native speakers and the advanced learners (χ² (1) = 0.84, p > 0.05). This indicates that advanced learners, like native speakers, are more sensitive to the association of progressive with activity verbs. This sensitivity is unaffected by mode; no significant difference is observed between speech and writing in intermediate learners (χ² (1) = 0.10, p > 0.05), advanced learners (χ² (1) = 0.17, p > 0.05), or native speakers (χ² (1) = 2.56, p > 0.05). The data show that both intermediate and advanced learners exhibit similar patterns of association of the English progressive -ing with lexical aspect, both in speech and writing.

4.3 Distribution of stative verbs in progressives

The third analysis focused on the distribution of stative verbs in progressives used by learners and native speakers. Notice that progressive -ing with stative verbs does appear in this study. Tables 3 and 4, presented earlier, show the distribution of the progressive -ing used with stative verbs, which ranges from 2.26% (intermediate, written, token count) to 9.60% (advanced, spoken, token count) of all progressives used. Chi-square tests were conducted to determine if any difference between speech and writing is statistically significant. The following analyses are based on token counts. For intermediate learners, a significant difference (χ² (1) = 4.74, p < 0.05) was found between the spoken and written data, indicating that the intermediate learners use more stative progressives in speech (5.65%) than in writing (2.26%). The difference between the distribution of stative progressives across speech and writing among advanced level learners also turns out to be significant (χ² (1) = 5.87, p < 0.05). In speech, advanced learners use more stative progressives (9.60%) than in writing (5.35%). Native speakers also use more stative progressives in speech (10.10%) than in writing (6.13%) and the difference is significant (χ² (1) = 7.45, p < 0.01).

To determine the difference between the two proficiency levels, Chi-square tests were conducted, and the results indicate that the advanced learners use more stative verbs in progressives than the intermediate learners in speech (χ² (1) = 4.48, p < 0.05). In writing as well, the advanced learners use more stative verbs in progressives than the intermediate learners (χ² (1) = 4.22, p < 0.05). The effect of increasing L2 proficiency is clearly shown by comparing the native speaker data with the two levels of learners. Significant differences exist between the intermediate learners and the native speakers both in speech (χ² (1) = 6.56, p < 0.05) and in writing (χ² (1) = 6.73, p < 0.01), while no significant difference exists between the advanced learners and native speakers, either in speech (χ² (1) = 0.05, p > 0.05) or in writing (χ² (1) = 0.17, p > 0.05).

Unlike most previous studies, the present study coded for the overextension of the progressive to stative verbs produced by learners (see Table 5). In speech, 16 out of 23 (69.57%) stative progressive tokens by the intermediate learners, and 21 out of 53 (39.62%) by the advanced learners were used incorrectly. In writing, 3 out of 8 (37.50%) stative progressives used by intermediate learners, and 4 out of 25 (16.00%) used by advanced learners, were incorrect. This means that the advanced learners used stative progressives more often than intermediate learners, but that intermediate learners more often incorrectly used stative progressives than advanced learners. For both groups, stative progressives are more frequent in speech than in writing.

Examples of overextended stative progressives are given below.

These examples may suggest that learners overextend the progressive aspect on stative verbs for special communicative motivations.^[17] In (3a) and (3b), the learners seem to be using the progressive marking on the stative verb “need” and “want” to present it as a dynamic situation, suggesting that the state of needing and wanting is a temporary one. In a similar vein, (3c)–(3d) add a dynamic and temporary meaning to the proposition. These sentences were judged as incorrect by native English speakers. However, some stative verbs in the progressive are acceptable. Examples are as follows.

These lexically stative verbs appear in the progressive because they are used non-statively, expressing temporariness rather than permanence (Smith 1983).