Influence of topic-prominent L1s on the use of L2 English copula be: a corpus-based study

2.1 Copula be and topic marking

Be can be divided into auxiliary and copula.^[1] They differ from each other in formal and functional aspects. Auxiliary be denotes aspect and/or voice of the verb phrase when used with a present participle (progressive; e.g., he is running) or a past participle (passive; e.g., it is cancelled). Copula be represents a relation between a subject and a predicate (e.g., I am a doctor), serving as a ‘linking’ verb (Halliday 2014; Stowell 1981) with little content meaning (Partee 1986).

While there are a number of proposals on the notion of ‘topic’ (among many others, Reinhart 1981; Gundel 1988), ‘givenness’ and ‘aboutness’ encompass all these proposals (Roberts 2019, p. 406): (a) A topic must be something that is either familiar (or given) itself or is an identifiable member of some familiar set of entities (givenness); (b) a topic is what the utterance in which it occurs is about (aboutness). Notably, there is a clause type whereby copula be should follow a topic, as acknowledged in Mikkelsen (2005, p. 161): “almost all researchers agree that the subject of a specificational clause is topic.”^[2] See (1a-b) for examples.

Mikkelsen claims that the subject of a specificational copula clause should be topic based on the observation that the subject NP requires a definite NP subject (the + NP; as in 1b), but not an indefinite NP subject (a + NP; as in 1a). This observation is particularly interesting for this study. Exploring which article – the or a – occurs with a subject NP may indicate the topicality of the subject NP in part, given that (a) topic can be defined in terms of givenness, which is related to the familiarity to the hearer (e.g., hearer-old, hearer-new; Prince 1981, 1992), and (b) that the hearer-old information is expressed with the and the hearer-new information is with a or zero-articles (except generic expressions) (Bickerton 1981; Hawkins 2015; Huebner 1983).

Given that ‘comment’ is what is being said about the topic (Gundel 1974), copula clauses containing ‘topic’ subjects may reflect a topic-comment structure. Den Dikken (2005) supports this idea, claiming that standard English, along with Belfast English which shows a strong tendency to use the topic-comment structure (Finlay 1988, p. 691), exploits the topic-comment structure for some copula constructions (see O’Neill 2015; Partee 2010 for a similar account). He then provides the syntax for this topic-comment structure in which the first constituent is “the topic, occupying the specifier position of a TopP” whereby the head takes the second constituent, “the comment, as its complement” (p. 708).

The topic-comment structure is in fact a prominent feature in East Asian languages that are categorized as topic-prominent languages, such as Korean, Japanese, and Chinese (Li and Thompson 1976). In these languages, there are topic markers widely used in various types of topic-comment structures. For example, in Korean, the topic marker -(n)un is used between a topic and a comment, as in (2), where the topic marker is placed between the topic (teacher) and the comment (very busy).

Notably, the topic marker shares some superficial similarities with the English copula be. It typically follows the topic NP; it does not carry specific content meaning, typically not affecting the propositional meaning of the sentence. Moreover, since topic is universally grammaticized as subject most strongly (Talmy 1983), topic markers – placed between topics (usually the subject) and comments – appear parallel to copulas, which are placed between the subject and predicate. These similarities are also found with the Japanese topic marker -wa, as in (3), where the topic marker follows the topic NP sensei ‘teacher’.

In Chinese, another topic-prominent language that employs a topic-comment structure, topics are generally marked by initial placement. Notably, the copula shì can also function similarly to Korean -n(un) and Japanese -wa in certain circumstances, as it is typically placed after the topic (Chan 2004; Hsieh 2009) and it should follow the topic in specificational constructions (Tham 2008), as shown in (4a).^[4] Unlike Korean -(n)un and Japanese -wa, however, the Chinese shì cannot take an Adjective Phrase (AP) as its complement (Hsieh 2009). For example, (4b) is ungrammatical as shì cannot take the AP hěn máng ‘very busy’ as its complement.

While we discussed some similarities among English be, Korean -(n)un, Japanese -wa, and Chinese shì with regards to topic marking, English has noteworthy differences from the other topic-prominent languages. One such difference is regarding so-called double subject constructions (Li and Thompson 1976). The double subject construction is considered to have two subjects: one is a topical subject, and the other is the subject of a clause-level comment, as demonstrated in the Korean sentence (5), where the topic marker -nun is placed between the topic khokkili ‘elephant’ and a clause-level comment consisting of a subject kho ‘nose’ and a predicate kil-ta ‘long’. Notably, the English translation equivalent in (5) does not allow copula be between the topic and the clausal comment.^[5] Instead, when the comment is a clause, a separate topic marking expression, such as as for, should be used.

Second, while the topic markers in Korean and Japanese and shì in Chinese can be followed by a verbal predicate, as in (6a-c), respectively, English copula be does not allow a verbal predicate as its complement, as in (6d).

If certain features are expressed through overt morphology in L1, but with zero morphology in L2, L2 learners may assume the target language also has overt morphology to express those features (Lardiere 2008; Montrul 2001). In other words, L2 learners may look for substitute L2-specific lexical items to host these features. In this regard, previous research has focused on be as a substitute for L1 topic markers. If L2-English learners with topic-prominent L1s indeed use copula be as a substitute for the topic marker, the linguistic analysis thus far yields the following predictions:

1. L2-English learners with topic-prominent L1s will use copula be productively in the course of representing topic-comment structure; in turn, hearer-old NPs will likely appear in the subject position of copula be more frequently than hearer-new NPs. One way to test this is, as discussed with Mikkelsen’s (2005) examples (1a-b), by analyzing whether the definite article the (as opposed to the indefinite a(n)) appears with the subject NP in a sentence containing copula be.

2. When L1-Chinese learners use copula be as a substitute for Chinese shì, the use may be restricted to NP complements, compared to learners of other L1-topic prominent languages.

2.2 L2 usage of be verb

Numerous studies have explored how a common feature shared across different L1s similarly influences L2 acquisition (Cadierno 2010; Stam 2006). It is widely acknowledged that L2 learners are similarly influenced by the same feature across different L1s, with this impact being pronounced when the L1 and L2 differ significantly in those features (Jeon and Sung 2023). These collective L1 effects have also been observed in L2 usage of English verb be. A myriad of studies, mostly based on the analysis of L2 learner corpus, have highlighted how L2 usage varies depending on the topic-prominence of the learners’ L1s, and investigated the nontargetlike use of be (e.g., John is want pizza) by L2 learners whose L1 belongs to topic-prominent languages (e.g., Chinese, Japanese, Korean, Lao). They suggest that such nontargetlike be serves as a topic-marker that creates a link between a topic and a comment, as in L1 (Hahn 2000, among many others).^[7] For example, Shin (2000) speculated that Korean learners’ erroneous production of she is like music reflects be as being used as a topic marker, meaning as for her, she likes music. Similarly, Huebner (1983) found that his Lao learner of English used is or isa, as in (7), to distinguish presupposition from assertion, or topic from comment, and concluded that is(a) is a topic-comment boundary marker.

Given that the topic-prominence of L1 affects nontargetlike be in L2 English, similar effects may exist for targetlike be in L2 English. In particular, there are notable similarities among topic markers and copula be placing between a topic and a comment, as discussed in Section 2.1. It has been suggested that copula be may have topic-marking properties transferred from topic-prominent L1s (Hahn 2000; Nam 2020; Shin 2000). For example, Nam (2020) conducted an elicited production task in which participants were asked to complete a sentence following a prompt word, as exemplified in (8). The task was administered for three groups (native speakers of English, L1-Russian learners, and L1-Korean learners), and it was found that only the Korean learners frequently used copula sentences, as in (9), comprising 12 % of their responses.

In (8), the question introduces two topical nouns tennis and ping pong. In (9), the L1-Korean participant puts is between each topical noun and its comment. This topic-comment structure appears quite similar with the topic-comment structure in the Korean translation equivalent in (10), in which the topic marker -nun is placed between a topic and a comment. Based on the finding, Nam (2020) speculated that tennis is every Sunday might not be a typical copula sentence, but rather a topic-comment structure.

Lee and Huang (2004) report more subtle effects of L1 topic-prominence features as they examine the use of be by L1-Chinese 4th graders (n = 270) in a primary school in Hong Kong. They collected production data from a story-writing task and found that the children performed better on the use of copula be with a noun (e.g., I am a king) than with an adjective (e.g., I am good). Assuming that the Chinese shì corresponds to copula be in their interlanguage, the gap in the use of copula be before adjectives and nouns may reflect the transfer of the L1 configurational regulation, which disallows shì before adjectival comments but allows it before nominal comments (for a similar finding, see Hsieh 2009).

Although these previous studies have suggested the effects of topic prominent L1s on the use of be in interlanguage, several gaps still remain. Firstly, previous research mainly examines nontargetlike be (e.g., she is like music) to investigate the transfer of topic prominence. Only a few studies have discussed the possibility that both copula be and nontargetlike be have the same function, a topic-marker property in the interlanguage grammar of L1-topic-prominent languages. If be serves as a topic marker in interlanguage due to L1 transfer, then copula be and the nontargetlike be may not be essentially different for these learners. Secondly, most of the previous studies claiming an L1-effect (e.g., use of be as an L1-topic-marker equivalent, frequent use of NP complements by L1-Chinese learners) are solely based on L1-interlanguage similarities. Such ‘L1-interlanguage similarities’, however, are not sufficient to conclude L1 transfer (Jarvis 2000) rather than universal L2 strategies if there are no ‘inter-L1 group differences’. Finally, previous studies exploring be in L2 speaking tend to examine a small number of L2 learners, which may limit the generalizability of findings. Therefore, the investigations on use of be in a large collection of spoken data (e.g., a learner corpus) are needed to overcome this limitation.

To address these limitations, the current study explores a topic-marker property in both copula be and nontargetlike be. To understand comprehensively how the topic-prominence of the L1 affects the use of these two types of be in interlanguage, learner corpora from three topic-prominent L1 (Chinese, Japanese, Korean) learner groups and a non-topic-prominent L1 (Filipino) learner group are compared.

Based on the linguistic similarities of copula be and topic markers in topic-prominent L1s, as discussed in Sections 2.1 and 2.2, the research questions of this study are formulated as follows.

1. Do the learners from topic-prominent L1s use copula be more frequently than those from non-topic-prominent L1s to express topic-comment relations?

2. Does the use of be and its variants by the learners from topic-prominent L1s show topic-marking features?

3. Does the topic-marking variation among the topic-prominent L1s affect the use of copular be?

With the hypothesis that learners from topic-prominent L1s are likely to associate be verbs with topic-coding due to L1 influence, the expected results are as follows. First, learners from topic-prominent L1s are anticipated to show a higher proportion of copula be relative to auxiliary be, as the latter primarily encodes aspect or voice rather than topic-comment relations. Second, the proportion of the-NP subjects versus a(n)-NP subjects is expected to be higher in topic-prominent L1 groups, along with more nontargetlike uses of be, such as double-subject constructions or be-insertion before finite thematic verbs due to the transfer of L1 topic-comment structures. Finally, within the topic-prominent L1 groups, L1-Chinese learners are expected to show a higher frequency of NP complements due to L1 transfer effects on the complementizer type used with the copula.

3.1 Corpora

The present study analyzed the International Corpus Network of Asian Learners of English (ICNALE: Ishikawa 2019). This corpus contains a variety of written and spoken production data such as essays, monologues, and dialogues. Among them, we specifically selected the native monologue corpus and four learner monologue corpora – viz., Chinese, Japanese, Korean, and Filipino – to collectively examine the impact of L1-topic-prominence on spontaneous production of be verbs. Other learner corpora in the ICNALE were not considered in this study due to data-related constraints such as their small size (e.g., Hong Kong corpus with only 50 participants), the inclusion of participants with multiple L1s (e.g., L1-Urdu and L1-Punjabi participants in the Pakistan corpus), or regional variations of the Chinese language (e.g., Singapore and Taiwan corpora), which would compromise the consistency and comparability required for this analysis.

The speech data of the five target corpora were collected using the same procedure. Participants gave a self-introduction and made four 60-s argumentative speeches: (1) the first speech about the first topic, (2) the second speech about the first topic, (3) the first speech about the second topic, and (4) the second speech about the second topic. The repetition of the same topics was based on the observation that a single recording does not provide enough data to represent one’s speaking performance (Ishikawa 2014).

The two speech topics were college students’ part-time jobs and a ban on smoking in restaurants. These topics were presented orally using the following prompts: (1) It is important for college students to have a part-time job, and (2) Smoking should be completely banned at all the restaurants in the country. The preparation time was 20 s for the first speeches and 10 s for the second speeches. The audio-recorded speeches were transcribed and tagged with parts of speech (hereafter, POS) on the Sketch Engine System, which uses the grammar rule of the Penn Treebank Project (Taylor et al. 2003).

The five target corpora consisted of 2,600 recordings produced by 650 participants. The native corpus included 600 recordings by 150 native English adult speakers, totalling 91,967 words, while the learner corpora had 2,000 recordings produced by 500 English learners, totalling 185,761 words.

The learner participants were classified by the proficiency levels of Common European framework of reference for languages, namely the CEFR (Council of Europe 2020): the proficiency classification was based on an L2 vocabulary size test (Nation and Beglar 2007) and their scores in standardized proficiency tests such as TOEFL and TOEIC (Ishikawa 2019). The CEFR describes six reference levels for L2 proficiency, from A1 to C2, which can be grouped into three types of language users: basic (A1 and A2), independent (B1 and B2), and proficient user (C1 and C2). All the learners were as A2 (waystage), B1 (threshold), or B2 (vantage) level on the CEFR proficiency scale.

Since controlling for L2 proficiency is crucial when testing a particular between-group factor in L2 acquisition studies (Schwartz and Sprouse 2000), we focused primarily on analyzing data from the B1 group, the group with largest data (Table 1).

According the CEFR scale for overall oral production, B1-level learners can “fluently sustain a straightforward description of one of a variety of subjects within their field of interest” (Council of Europe 2020, p. 62).

3.2 Analysis

Every case of be and its variants was coded to one of the three major functions: auxiliary, copular, and existence. We extracted every case of be verbs in the native and the four learner corpora, using the six relevant POS tags of the Sketch Engine System (Taylor et al. 2003): VB (be), VBP (am, are), VBZ (is), VBD (was, were), VBG (being), and VBN (been). Then, we removed inaccurately tagged cases, such as mistaking the adverb maybe for the modal verb may and the verb be. This left us with a total of 14,733 instances of be verbs across all five corpora.

However, many instances were challenging to categorize into one of the three major functions. Most of them were incomplete be verb phrases as in self-correction (e.g., he is, he has a ball) or a be verb followed by in its repetition (e.g., he is, is tall). These cases were coded as “incomplete”. We also found many cases of nontargetlike be verbs, a common error in L2 English (e.g., he is make money), so we used a separate code for these nontargetlike be.

The coding of be and its variants was based on the neighboring words and their POS tags. For example, a be verb that is followed by present participles (e.g., is running) or past participles (e.g., be cancelled) was coded as auxiliary be, while be and its variants that are followed by adjectives (e.g., is happy) or determiners (e.g., am a) were coded as copular be.

This automatic classification was complemented by manual inspection to guarantee high accuracy in verb coding. For example, the verb is in the sentence it is designated area should be coded as copula be, but it was automatically coded as auxiliary be because it was followed by a past participle. To correct this sort of coding error, a considerable amount of manual coding was conducted by looking into the entire sentences. This enabled us to correct the automatic coding errors in approximately 500 cases of be verbs.

The coding of subject articles preceding copula be was also based on the neighboring words and their POS tags. Firstly, singular nouns preceding copula be were identified, and then the a and the uses preceding up to four words away from them were coded. If articles were not adjacent to nouns, the intervening words were restricted to nouns (e.g., part-time), adjectives (e.g., good), adverbials (e.g. frequently), and present/past participles (e.g., smoking, known).

The coding of be and other elements was primarily carried out by the first author, with each coding instance reviewed by the third author. The coding process was iterative, and revisions continued until both authors reached full agreement. The data file can be found at https://osf.io/ru3tg/?view_only=36b63a6a7f594292b3da22b46551c092.

After the preliminary analysis of be usage in the four corpora, frequency-based analyses were conducted to determine whether and to what extent different topic-prominent L1s affect English learners’ use of be verbs.

1. To investigate the frequency distribution of the be verb, the proportion of copula be as opposed to auxiliary be was compared across the groups.

2. To explore the topic-marking features of be and its variants, the proportions of “the + singular NP” subjects and “a(n) + NP” subjects before copula be were analyzed across groups. In addition, to identify distinctive uses of be variants presumably due to the transfer of topic-prominent L1, a total of 133 cases of errors with nontargetlike be were found and analyzed.

3. Finally, to answer whether the intra-topic-prominent-L1 variation may affect L2 use of copula be, the uses of NP and non-NP complements after copula be were compared in the three learner corpora with topic-prominent L1s (viz., Chinese, Japanese, and Korean).

4.1 Frequency of copula be

Table 2 presents the frequency of be verbs across the native and four learner corpora, along with their standardized frequencies assuming each corpus had a word count of 100,000. The standardized frequency of be verbs was highest in the native corpus (=5,830), and lowest in the Chinese-B1 corpus (=4,386). The chi-square test based on the frequency of be verbs and the corpus size indicated a significant difference among the five corpora, χ ²  = 171.87, p < .001.

Table 3 shows the frequency distributions of auxiliary be and copula be across the corpora. The proportion of copula be was higher than that of auxiliary be in every group. In particular, there were greater gaps observed among the topic-prominent L1 groups, where the frequency of copula be was over three times greater than that of auxiliary be (e.g., 1,560 versus 424 in the Chinese group). Meanwhile, the frequency ratio of copula be and auxiliary be was about two-to-one in the native and Filipino groups.

According to the chi-square test, there was a significant difference in the frequency proportions of copula be and auxiliary be among the five corpora (χ ²  = 244.32, p < .001). Post-hoc analyses of pairwise comparisons with Bonferroni correction of a = .005 (=0.05/10) revealed that the topic prominence-contrastive-L1 pairs (viz., a pair of a non-topic-prominent L1 and a topic-prominent L1; e.g., native speakers versus Chinese learners) had greater chi-square statistics than the intra-topic-prominent-L1 pairs (e.g., native speakers versus Filipino learners), as shown in Table 4.

All six topic-prominence-constrastive-L1 pairs showed significant differences in the frequency distribution of copula be and auxiliary be at a < .0001 (=0.001/10). In contrast, only two of the four intra-topic-prominent L1 pairs showed significant differences, mostly due to the higher frequencies of copula be in the Japanese B1 and Korean B1 corpora. The mean chi-square value for the significant topic-prominence-contrastive-L1 pairs was 86.75, while that of the significant intra-topic-prominent L1 pairs was 12.58. These results indicate that the use of copula be was similar among groups whose L1 is topic-prominent, while the use of copula be varied among groups with non-topic-prominent L1s.

4.2 Topic-marking features of be and its variants

The relationship between be and topic-marking was examined by comparing the proportions of copula be and auxiliary be between topic-prominent L1 groups and the others. Nontargetlike uses of be were also identified and analyzed in each group.

4.3 Intra-topic-prominent L1 group variations in complements of copular be

To examine the effect of the intra-topic-prominent-L1 variation concerning the complements copular be can take, the use of NP and non-NP complements by learners from three topic-prominent L1s was analyzed, as shown in Table 7.

L1-Chinese learners used an NP complement for copula be more frequently than L1-Japanese and L1-Korean learners. The chi-square test revealed that the frequency distributions of NP and non-NP complements for copula be were significantly different between Chinese and Korean (χ ²  = 10.13, p = .0015 < .01) and between Chinese and Japanese corpora (χ ²  = 18.79, p < .001), but not between Korean and Japanese corpora (χ ²  = 0.49, p = .48). In other words, Chinese learners used NP complements after copula be significantly more frequently than Japanese and Korean learners.

For additional evidence, we took a look at the proportions of NP versus non-NP complements after copula be at a lower level, viz., A2. In the Chinese learner data, the proportion of NP complements decreased as their proficiency increased, from 31.7 % at A2 to 28.3 % at B1. In contrast, the other two learner corpora showed the opposite trend, with higher-level learners using NP complement more frequently. These results suggest that the Chinese learners’ heavier reliance on NP complement after copula be, compared to the Japanese and Korean learners, may be attributable to L1 transfer; however, this pattern can gradually disappear as they become more proficient in L2 English.