Acquisition of the Chinese Indefinite Determiner “One + Classifier” and English Articles in Two-Way Learner Corpora

Abstract

This paper presents findings concerning use of classifiers and articles in learner corpora and the effect of learners’ native languages on their acquisition of a second language. First, we use data from the Learners’ Corpus of Chinese, an error-tagged two-way learner corpus of intermediate and advanced learners’ written production, developed by Tokyo University of Foreign Studies (TUFS) in collaboration with National Taiwan Normal University. The corpus data reveals that English L1 learners of Chinese overuse the “one + classifier” structure for indefinite reference, analogous to English indefinite articles, whereas Japanese L1 learners show underuse of this structure, despite Chinese and Japanese both being regarded as “classifier languages”. Second, data from the TUFS Learners’ Corpus of English reveals that Chinese L1 learners use the definite article in a more native-like way than Japanese L1 learners.　Third, analysis of the International Corpus of Japanese as a Second Language reveals that Chinese L1 learners of Japanese use the “one + classifier” structure more frequently than native speakers. Similarities and differences between L1 and L2 can supersede ostensible typological similarities, such as the classification of both Chinese and Japanese as classifier languages.

1 Introduction

This study investigates the acquisition of the “one + classifier” structure by Japanese native learners of Chinese (henceforth, “JLC”) and compares it with English native learners of Chinese (henceforth, “ELC”). It reveals that underuse of the “one + classifier” structure preceding nouns is more prevalent in written essays by JLC. Typical errors of omission are given below.¹

There is a park by the airport.

English native speakers, on the other hand, are less likely to omit “one + classifier”, instead displaying a slight tendency to overuse the structure, as in the following examples.

The uses of “one + classifier” can be classified into three types, based on the features of referentiality and specificity.² Table 1 shows examples of each alongside equivalent Japanese and English sentences. Example (7a) is a [+referential, + specific] usage. Example (8a) shows a [+referential, –specific] usage. Example (9a) is a non-referential usage and can be considered the most grammaticalized of the uses of “one + classifier”.

Table 1 Three types of “one + Classifier”

The Japanese sentences in (7b) and (8b) may include the form “ik-ken”, similar to the Chinese “one + classifier”. However, whereas the classifier is typically required in Chinese, this is not obligatory in Japanese. The function of “one + classifier” is less grammaticalized in Japanese than it is in Chinese; the former is used mainly to express number, which restricts the scope for positive transfer in cases where “one + classifier” is not used to express purely numeral information. The non-referential use of “one + classifier”, which is not expressed with a similar form in Japanese, is expected to be particularly difficult for L1 Japanese learners to acquire.

In contrast, the use of the “one + classifier” structure to mark an indefinite noun phrase is similar to typical uses of the English indefinite article. There is therefore the possibility for positive transfer from L1 to occur and so we predict that the “one + classifier” is relatively easy for ELC to acquire.

To summarize, we hypothesize that ELC and JLC will exhibit contrasting trends in the use of the “one + classifier” and that these differences can be plausibly explained through consideration of L1 characteristics.

In the remainder of this section, we will give an overview of functional equivalents to articles in Chinese. Chen (2004) claims that in Chinese, the demonstratives 这 zhè and 那 nà have developed definite article-like uses such as in (10a). In (10a), by using “the house”/“这栋房子”, the speaker assumes that the listener knows which house George bought. The “one + classifier” structure has also undergone some degree of grammaticalization and functions in a similar way to the English indefinite article in some cases (10b, as well as the examples in Table 1). In (10b), by using “a house” / “一栋房子”, the speaker assumes that the house George bought cannot be identified as a particular house and that the listener does not know which one he or she is talking about.

According to Chen (2004), “one + classifier” appears in five uses equivalent to the indefinite article in English (10a)–(10e).

In the analysis below, we focus on the use of “one + classifier”, its similarities with English articles, and the consequences for L2 Chinese acquisition by JLC and ELC.

2 The Present Study

2.1 Corpus Data

This paper analyzes the acquisition of the “one + classifier” structure by using the corpus of written Chinese collected by Tokyo University of Foreign Studies (TUFS corpus: https://corpus.icjs.jp). The ELC written data is provided by Taiwan Normal University and consists of essays written as part of the Test of Chinese as a Foreign Language (TOCFL). The composition of the corpus used in this paper and its size are shown in Table 2.

Table 2 Composition and size of the corpora

2.2 Methodology

Instances of the “one + classifier” structure were manually extracted from the corpus. Correct and incorrect example sentences were then distinguished and categorized based on the type of error, the linguistic context in which the error was produced, and the learner’s proficiency level. The following sections will discuss possible causes of learners’ under- and overuse of the “one + classifier” structure.

In Sect. 3, we provide quantitative and qualitative analysis of errors in the use of classifiers by Chinese L2 learners and show how L1 appears to affect L2 acquisition. We then provide supporting evidence in the form of case studies of English L2 article use and Japanese L2 “one + classifier” use in Sects. 4 and 5 respectively.

3 Results and Discussion

This section presents results of analyzes of error trends in the use of “one + classifier” in L2 Chinese. Quantitative and qualitative findings for JLC are reported in Sects. 3.1 and 3.2 respectively. Section 3.3 briefly highlights characteristic errors by ELC. These errors additionally display parallels with the L2 English article use that we cover in Sect. 4.

3.1 Quantitative Analysis of the Use of “One + Classifier” by Japanese and English L1 Learners of Chinese

Instances of “one + classifier” produced by JLC and ELC were extracted then compared using adjusted frequencies (per 10,000 words). The results are shown in Table 3.

Table 3 Comparison of the output of “one + classifier” by Japanese/English native learners

Table 3 shows that ELC produced the “one + classifier” structure 74.77 times per 10,000 words, which is significantly higher than the 17.78 times produced by JLC. This suggests that Japanese learners avoid the use of “one + classifier”, and/or English learners overuse the “one + classifier”. In order to confirm the above hypotheses, errors were categorized as “Underuse” (i.e., omission of “one + classifier” where it is required), “Overuse” (i.e., use of “one + classifier” where it cannot appear), or “Replace” (using the wrong classifier). The results are shown in Tables 4 and 5.

Table 4 The correct use and misuse of the “one + classifier” by JLC

Table 5 The correct use and misuse of the “one + classifier” by ELC

As shown in Table 4, JLC at all three proficiency levels exhibit low levels of accuracy in the use of “one + classifier”. The proportion of correct use in fact decreases slightly with increasing proficiency level. Regarding error type, we observe significant underuse of “one + classifier” at all levels. Instances of underuse account for more than 50% of the errors, and this proportion does not change significantly with increased proficiency or length of language study.

The patterns of correct use and misuse of the “one + classifier” structure by ELC as shown in Table 5 are notably different. The overall frequency of misuse by English L1 speakers is low, and the proportion of errors decreases as learners’ proficiency level increases. By CEFR-B2 level, ELC can be said to have acquired the “one + classifier” structure. The breakdown by error type also differs from the JLC data. The majority of errors made by ELC are of the “replace” type, but its proportion also decreases as proficiency increases. There are also slightly more instances of overuse by ELC.

In summary, JLC and ELC exhibit contrasting trends in the acquisition of “one + classifier”. JLC have difficulty in acquiring “one + classifier”. Errors of omission of “one + classifier” are prevalent and do not reduce significantly with increasing proficiency. In contrast, acquisition of “one + classifier” by ELC occurs more smoothly. ELC already achieve higher accuracy levels at CEFR A2 level and accuracy further improves as proficiency rises to the B1 and B2 levels. They do, however, experience some difficulty in selecting the appropriate classifier. These error trends are predictable, given that JLC lack functional equivalents to “one + classifier”, whereas ELC possess functional equivalents to “one + classifier” (the indefinite article) but do not have a highly developed system of classifiers in their native language.

3.2 Qualitative Analysis of the Use of “One + Classifier” by Japanese L1 Learners of Chinese

In the previous section we showed the contrasting use of the “one + classifier” structure by JLC and ELC. Use of the “one + classifier” structure appears to be a more problematic and persistent issue for JLC than it is for ELC. In this section, we therefore focus on underuse of “one + classifier” by JLC, considering the structure’s different functions.

First, we review the syntactic positions in which “one + classifier” may appear in a sentence and its function in each instance. When placed in the subject position, “one + classifier” can marks either a referential specific (12a) or non-referential (in this case, generic) noun phrase (12b).

When placed in the object position, “one + classifier” marks a referential specific noun phrase, as in the following examples.

Finally, when used as a predictive component, “one + classifier” is used non-referentially, as in (14).

We divided the correct uses of “one + classifier” by JLC following the above categories. The results are shown in Figs. 1 and 2. Figure 1 shows that the correct “one + classifier” structures produced by Japanese L1 learners are mainly found in cases where the indefinite noun phrase appears in object position, or in cases where it is used as a predictive element after the copula “是shi”. The proportion of the former increases as proficiency increases. At all levels of proficiency, there are few instances of “one + classifier” appearing in subject position.

Fig. 1

Distribution of the correct use of “one + classifier” by Japanese L1 learners at each proficiency level

Fig. 2

Distribution of the instances of omission of “one + classifier” by Japanese L1 learners at each proficiency level

Figure 2 shows misuse of “one + classifier”. By comparing the proportions of each use in Figs. 1 and 2, we can ascertain which uses prove to be relatively difficult at each level. First, we will consider uses of “one + classifier” in the subject position. At CEFR A2 level, these represent 7% of correct uses (Fig. 1) but 17% of omissions. At B1 and B2 levels, the proportions of correct use and omission are virtually the same. This suggests that the JLC had difficulty correctly including “one + classifier” in subject position at A2 level, but subsequently acquired this use.

Next, we will consider non-referential uses of “one + classifier”. At A2 level, these uses represented a higher proportion of errors (47%, Fig. 2) than the proportion of correct use (38%, Fig. 1). As with “one + classifier” in subject position, the non-referential use, therefore, appears to be problematic at A2 level. At B1 and B2 level, the proportions are reversed, suggesting that non-referential uses of “one + classifier” become relatively unproblematic as proficiency increases.

Finally, we consider the uses of “one + classifier” in object position. In this position, “one + classifier” is used to express specific or non-specific referential noun phrases (see Table 1 above for examples). It can perhaps be said to be the most prototypical of the uses of “one + classifier”. This use represents a larger and larger proportion of total use, and misuse fluctuates around the 50% mark.

In this section we gave a breakdown of the types of sentence where “one + classifier” is used or omitted by JLC. In the following section, we turn to characteristic errors appearing in the ELC data.

3.3 Qualitative Analysis of the Use of “One + Classifier” by English L1 Learners of Chinese

In this section we briefly analyze errors made by English L1 learners of Chinese (ELC). As demonstrated in Sect. 3.1, ELC seem able to use “one + classifier” with greater ease than JLC. At intermediate (i.e., CEFR B1 and B2) levels in particular, ELC use “one + classifier” correctly in over 90% of instances. This contrasts with JLC, where the proportion of correct uses is low in elementary (A2) level learners and remains largely unchanged regardless of increases in proficiency.

Nonetheless, a small number of errors do occur in ELC production. Strikingly, errors of overuse are more prevalent than errors of omission. Examples are provided below. Superfluous uses of “one + article” are shown in brackets.

As is evident from the English translations of each sentence, ELC use “one + article” where an indefinite article would be required in English. Overall, this is an effective strategy, as “one + article” and the indefinite article are functionally equivalent in many cases, but it is inappropriate in (15–18). Note that none of these examples express realis events. (15–16) are negative sentences and (17–18) express future possibilities that may or may not occur. All four examples are incompatible with the individualizing function of “one + classifier”. English articles are not affected by similar semantic considerations, so it is perhaps unsurprising that a one-to-one mapping of “one + classifier” and the indefinite article in the interlanguage of ELC would lead to overuse of the kind shown in (15–18).³

The results of this study are complementary to those of Crosthwaite et al. (2017), who analyzed the expression of definite discourse-new (so-called “bridging”) reference by English, Korean and Japanese learners of Chinese. “Bridging” refers to situations where a new referent can be linked to a previous referent in the discourse. For example, in (19a), “waiter” can be marked with the definite article because its existence is implied by “restaurant” in the preceding sentence. Chinese does not mark definite discourse-new reference morphologically (19b), in contrast to standard uses of “one + classifier” to mark discourse new, noninferable reference, e.g., in (7–9) and (11).

ELC showed a tendency to use Chinese demonstratives and classifiers analogously to English articles. In other words, they used “demonstrative + classifier + noun” for bridging reference, and “one + classifier” for nonbridging reference (Crosthwaite et al., 2017: 644).⁴ While article-like use of “one + classifier” may result in native-like use as attested in Sect. 3, in irrealis situations (the current study) or bridging situations (Crosthwaite et al.’s study), over-generalization of the functional equivalence between articles and “one + classifier” results in infelicitous use of the latter.

This section has shown how both the high level of overall acquisition of “one + article” and its overuse in specific circumstances can be explained if we assume that ELC associate “one + article” with the indefinite article in their L1. Whether this phenomenon is due to conscious strategies by individual learners, the result of L2 pedagogy or an unconscious association is a task for further research.

3.4 Analysis of the Use of “One + Classifier” by Korean L1 Learners of Chinese

In this section, we discuss the data relating to the use of “one + classifier” in essays written by Korean native learners of Chinese (henceforth, “KLC”). Table 6 shows that among instances of misuse of “one + classifier” (underuse: 78.40%, overuse: 16.00%, replace: 5.60%), the percentage of underuse (196: 78.40%) is remarkably high in the essays by KLC. This high proportion of underuse errors is similar to the results for the JLC data shown in Sect. 3.1. This may be due to the fact that Korean is typologically similar to Japanese.

Table 6 The correct use and misuse of the “one + classifier” by KLC

That is, Korean, as well as Japanese, does not have a determiner position (DP) in syntax, which may have affected the underuse of “one + classifier”. The following are some specific examples for misuse of underuse in KLC data.

The percentage of correct use is considerably higher in the KLC data than in the JLC data (KLC: 81.82%, JLC: 40.95%). This difference is presumably due to differing proficiency levels of the KLC and JLC learners. The Korean learners belonged to the Chinese language department of a university in China. In other words, they are learning Chinese not only in the classroom but also in their living environment, which means their Chinese level is likely to be higher. In contrast, Japanese learners were studying Chinese as a foreign language in Japan.

However, in spite of the higher Chinese level of KLC, the proportion of overuse is higher in the data of KLC (2.91%) than in the data of JLC (1.18%), which may be due to individual differences among learners. These errors are different from those of JLC, such as use of “one + classifier” in negative sentences, and are a potential area of future research.

4 L2 English Article Use by Chinese L1 and Japanese L1 Learners

In this section we focus on the acquisition of articles in L2 English. In Sect. 4.4 we will refer back to error examples in Sect. 3.3, demonstrating how the presence or absence of a realis/irrealis distinction in the use of determiners affects L2 English as it does L2 Chinese. First, we introduce our data set and more general findings.

4.1 Background

In Sect. 3.1 we demonstrated how similarities and differences between L1 and L2 appear to contribute to contrasting trends in the use of the “one + classifier” in L2 Chinese. Specifically, the functional similarity between the English indefinite article and the Chinese “one + classifier” structure appears to be more conducive to the acquisition of the L2 form than the morphological similarity between “one + classifier” in Chinese and Japanese. In the following sections, we offer a preliminary investigation of similar processes in L2 article use by Chinese learners of English (CLE) and Japanese learners of English (JLE).

Research on the acquisition of English L2 articles is voluminous, with the majority of studies being based on forced elicitation tasks, self-paced reading tasks, or other experimental designs. Research has particularly focused on native speakers of “article-less” languages, including Chinese (Díez-Bedmar & Papp, 2008; Robertson, 2000; Snape, Leung, & Ting, 2006; Xu et al., 2016; Yang & Ionin, 2009) and Japanese (Butler, 2002; Hawkins et al., 2006; Snape, Leung, & Ting, 2006; Ogawa, 2007; Kume, 2016; Yamada, 2019).⁵ Learners of these languages are believed to find the acquisition of articles problematic due to an absence of equivalent features in L1. Nonetheless, there is little consensus about which particular uses of articles learners struggle with the most, and the underlying causes. It has been argued that learners’ choices of article can be affected by factors including definiteness, specificity, countability, and reference salience. Some research has argued that individual learners whose native languages lack articles fluctuate in their L2 article use, although results are not uniform between or within languages. The differing results of previous studies have numerous potential causes, including mode- or task-related effects, learner proficiency, and even different analytical frameworks.⁶

One further factor, which will be the focus of the present study, is the influence of learners’ native language. Specifically, we investigate whether L1 Chinese and L1 Japanese learners of English exhibit different patterns in their use and misuse of English articles. We introduce one study that is particularly pertinent to this research question below.

Crosthwaite (2016a) is a corpus study comparing article use by L1 Mandarin, L1 Korean and L1 Thai learners of English. While all three of these languages are regarded as “article-less”, Crosthwaite (2016a: 78) asserts that in L1 language use Chinese speakers “appear to use overt syntactic means to signal (in)definiteness (e.g., overt or deliberately omitted numeral + classifiers, demonstratives) more often and in more clearly differentiated article contexts than Korean and Thai speakers”. As a result, “the potential for positive L1 transfer of certain form/function relationships associated with the English article system appears to be greater” for Chinese L1 learners.

In Crosthwaite’s (2016a) study, Chinese L1 learners indeed exhibited more target-like use of articles than Korean L1 and Thai L1 learners. Furthermore, Chinese L1 learners exhibited similar levels of accuracy for zero, definite and indefinite articles, in contrast to Korean and Thai L1 learners, for whom definite and indefinite articles proved to be more challenging than zero articles. Crosthwaite (2016a: 33) concludes that this phenomenon can be regarded as the effect of positive transfer, and as evidence that Chinese “does, in fact, have an article-like system”. In summary, Crosthwaite’s study demonstrates the practical effects on second language acquisition of the article-like uses of “one + classifier” and demonstratives noted by earlier studies (Chen, 2004; Gundel et al., 1993; Snape, Leung & Ting, 2006). It also demonstrates that differences among learners whose L1 lack articles can be observed not only in experimental contexts but also in natural language use as captured in learner corpora.

In the remainder of Sect. 4, we examine our own data for patterns similar to those observed by Crosthwaite (2016a) and parallel to the L2 Chinese data detailed in previous sections.

4.2 Data Set

The data set referred to in the rest of Sect. 4 comprises another subsection of the three-way learner corpus developed at Tokyo University of Foreign Studies. There are two important caveats regarding this particular data set. First, we lack proficiency data of the type referred to above for the Chinese L2 data. The Japanese L1 learners were first-year English majors at the time of data collection, whereas the Chinese L1 learners were fourth-year English majors. Second, the data is taken from a translation task, in which learners translated equivalent texts from their L1 into English. The task was thus controlled for content but not for proficiency, so the results presented below cannot be compared directly to the analysis in Sect. 3.1. They can, however, be considered as another potential instantiation of L1 effects on the acquisition of L2 forms, and represent a task type that has not, to our knowledge, been employed in L2 acquisition studies concerning English articles.

4.3 Quantitative Analysis

Tables 7 and 8 show article use by CLE and JLE respectively. Raw figures have been used because both data sets consist of the same number of translations of the same source text (n = 40). Shaded cells represent correct use of either the zero, indefinite or definite article.⁷

Table 7 Article use by Chinese L1 learners

Table 8 Article use by Japanese L1 learners

Error patterns appear to be largely similar in the two groups of learners, with omissions of the indefinite article being most prominent, followed by omissions of the definite article. Examples of each error type are shown in (24) and (25) below.⁸

Furthermore, overuse of articles (i.e., where the correct form is zero) represent a very low proportion of total errors (approximately 12% for CLE and 6% for JLE). Examples for the definite article are given in (26) for reference.

The error trends illustrated above are expected, given that both Chinese and Japanese are regarded as article-less languages. Note, however, that JLE exhibit a higher total frequency of errors than CLE, and that the proportion of errors of omission is greater for JLE (69%) than it is for CLE (58%). This difference appears to mirror Japanese native speakers’ omission of “one + classifier” in L2 Chinese, and also suggests that Japanese learners of English may be closer to Korean and Thai learners of English than they are to Chinese learners of English, in terms of their frequent omission of articles.

The data in Tables 7 and 8 were then used to calculate the Target Language Use (TLU) as proposed by Pica (1983) for the definite, indefinite, and zero articles. TLU takes into account both overuse and underuse of a target form and has been repeatedly adopted in previous studies on article use (Crosthwaite, 2016a, 2016b; Díez-Bedmar & Papp, 2008). TLU is calculated using the formula shown below. A TLU score of 1 represents entirely “target-like” use. The lower the TLU, the more problematic the form for learners.

(27) ${\text{TLU}} = \frac{{{\text{no. of suppliances in obligatory contexts}}}}{{{\text{(no. of obligatory contexts)}}}+{{\text{(no. of suppliances in non-obligatory contexts)}}}}$.

The TLU for each article is shown in Table 9. The patterns can be summarized as follows. First, for both groups of learners, TLU was lowest for the zero article and highest for the indefinite article. Comparison with Tables 7 and 8 shows that the low TLU for the zero article was due to “overuse” of the zero article, in other words, omission of the indefinite article and, to a lesser extent, omission of the definite article. Again, such errors of omission are expected given that both Chinese and Japanese lack articles. Despite this trend to omit the indefinite article, in the majority of cases learners’ selection of the indefinite article was in fact appropriate. As a result, the indefinite article showed the highest TLU, for both CLE (TLU = 0.81) and JLE (TLU = 0.79).

Table 9 TLU of articles by Chinese L1 and Japanese L1 learners

Second, the TLU for the zero article and indefinite article was almost identical for CLE and JLE. In other words, the functional similarity between “one + classifier” and the indefinite article did not have an obvious positive effect on L2 article acquisition by CLE in the current data set.

In contrast, the TLU for the definite article was notably higher for CLE (TLU = 0.77) than for JLE (TLU = 0.70). The precise reason for this difference is not entirely clear, but it reflects a greater overall use of the definite article by CLE. Table 10, calculated from the figures in Tables 7 and 8, shows the proportion of articles used by each group of learners. In the Chinese L1 data, the definite article accounts for over 40% of overall article use, the highest of the three possible forms, whereas in the Japanese L1 data, the percentage is less than 30%, the lowest of the three possible forms in article position. Table 10 also shows that the L1 Japanese data exhibits a greater proportion of zero article use (32.8%) than that seen in the L1 Chinese data (23.6%). These figures refer to overall use irrespective of whether the usage is correct or incorrect, but they are suggestive of a tendency among JLE to avoid articles more generally.

Table 10 Proportion of use of each article

Taken together, the data in tables from Tables 7, 8 and 9 and 10 suggest that CLE show greater mastery of articles in general and the definite article in particular compared to JLE. This may ultimately be due in part to the functional equivalence between the definite article and determiner-like uses of demonstratives in Chinese, though it is not clear why a similar phenomenon does not occur with the indefinite article in the current data set.⁹ There may be task-related issues, so further studies should be conducted using different data sets to enable data triangulation.

4.4 Qualitative Analysis

This section briefly discusses some concrete examples of article errors and suggests how these may have been influenced by learners’ L1. As mentioned above, the current data set is a translation task and so L1 influence can potentially occur not only through learners’ interlanguage, but explicitly through the source text. However, the text in question features few uses of either “one + classifier” or determiners, so L1 influence is likely to occur more generally. Below, we consider how the absence of “one + classifier” in L1 Chinese may lead to omission of articles in L2 English.

The examples in (28) show omission of articles by CLE, with the appropriate article added in brackets. These examples feature negation and are found almost exclusively in the CLE data. This is a reflection of the fact that the “one + classifier” structure is not required or even allowed in irrealis sentences such as negatives and conditionals. Crucially, this mirrors the erroneous use of “one + classifier” structure by ELC in irrealis sentences illustrated in Sect. 3.3. In other words, although “one + classifier” and articles have functional similarities and appear to be sources of positive transfer overall, restrictions on the use of the former to realize contexts appear to contribute to overuse of “one + classifier” by ELC and underuse of the indefinite article in particular contexts by CLE.

Finally, we mention some other areas of difficulty for learners. Article errors are concentrated in three main areas: a. use of the definite article for bridging reference, b. non-referential use of the indefinite article, and c. inappropriate use of the definite article in bridging relations.

First, learners appear to have difficulty with the definite article used for bridging reference, i.e., where a new referent can be inferred from another referent. The translation task includes such pairs as “bed” and “futon”, “box” and “lid”, and “restaurant” and “menu”. Learners frequently failed to mark “futon”, “lid” or “menu” with the definite article. In fact, these errors occur almost exclusively among Japanese learners. This is unsurprising, as Japanese neither marks bridging relations morphologically, nor distinguishes these from new referents (which are also unmarked morphologically). Therefore, even JLE who successfully use the definite article for previously mentioned referents may struggle with bridging reference. Instead, they use the indefinite article or omit the article altogether (29).

Second, both Chinese and Japanese learners tend to omit the indefinite article in non-referential situations. The irrealis sentences in (28) can be regarded as non-referential. In addition, there are frequent omissions of the indefinite article in contexts like those shown in (30). The example in (30a) refers to Chinese restaurants in general and the examples in (30b-d) do not to refer to actual concrete objects.¹⁰

Finally, we comment briefly on errors where learners selected the definite article instead of the indefinite article. This represents the third most common error type among both groups of learners, following omission of the indefinite article and omission of the definite article (see Tables 7 and 8). This error type largely represents inappropriate use of the definite article in bridging relations. In other words, learners use the definite article for new referents despite their being no clear implication of the referents existence. For example, in (31a) there is no reason to assume the existence of a teacup, let alone one known to the reader, based on the text up to that point. Learners are therefore marking an indefinite, specific referent as if it were a definite, specific referent.

Why, then, does this error trend emerge? We suggest that the presence of modifying elements (underlined in the examples above) gives learners the impression that there is enough information for the reader to identify the referent.¹¹ While this is an incorrect application of the English definite article, such behavior has been observed in previous studies. The current study, therefore, supports the idea that learners confuse the features of definiteness and specificity when choosing the appropriate English article.

Section 4 has provided some partial supporting evidence for the assertion that learners’ native language affects acquisition of L2 forms. L1 influence is not uniformly positive or negative but can lead to both native-like and erroneous use of L2 forms, depending on a range of other factors. In the next section, we briefly turn to the acquisition of L2 Japanese.

5 Use of Japanese “One + classifier” in Compositions by L1 Chinese and English Learners

Fig. 3

Five-frame cartoon by I-JAS

This section examines the use of “one + classifier” in Japanese compositions by Chinese L1 learners (CLJ) and English L1 learners (ELJ). The data used in this section is taken from the “Story writing 1” section (SW1)¹² of I-JAS.¹³ We examined the use of modifying elements preceding the noun “inu” (dog) in the SW1 data for CLJ and ELJ and compared each to the use of modifying elements by native speakers of Japanese on the same task. The types of modifying elements used by L1 speakers of each language are summarized in Table 11. The table reveals the following two points.

Table 11 Modifier elements and frequency of occurrence with “inu” (dog)

First, Chinese L1 learners show a significantly higher frequency in the use of “one + classifier” (Chinese L1: 47.53, Japanese native: 0.00, G² = 5.03, p = 0.025) and of the indefinite element “aru” (Chinese L1: 43.87, Japanese native: 0.00, G² = 4.64, p = 0.031) as compared to Japanese native speakers. Second, in contrast to Chinese L1 learners, English L1 learners, like Japanese native speakers, do not exhibit overuse of “one + classifier” or “aru”. Indeed, neither ELJ nor Japanese native speakers show any use of “one + classifier” or “aru”.

This phenomenon suggests that although both Chinese and Japanese possess the “one + classifier” structure, its function is different in both languages, otherwise we would expect to see similar patterns of use between Japanese native speakers and CLJ. While Japanese possesses the “one + classifier” structure, in practice it is not selected by native speakers in the current context, where conveying numeral information is not necessary.

Next, we will discuss “one + classifier” from the perspective of “boundedness”¹⁴ in cognitive structure and “the function of introducing new information” in informational structure.

Shen (1995) suggests that the function of classifiers in Chinese is to “embody the opposition between bounded and unbounded” in the human cognitive structure. We analyze the use of “one + classifier” by CLJ in the I-JAS data making reference to this concept of “boundedness”. Errors related to classifiers have been corrected, with corrections shown in brackets. Unrelated errors have been left uncorrected.

“Locative Structure”.

¹⁵

“Verb-Complement Structure”.

Examples of Japanese sentences written by Chinese L1 learners are shown in (35–37). The equivalent sentences in Chinese are provided for comparison. As the Chinese sentences require “one + classifier” to realize boundedness in each case, the overuse of “one + classifier” in L2 Japanese by Chinese L1 learners in I-JAS may be due to L1 transfer. On the other hand, the overuse of classifiers is also related to factors concerning information structure. In Japanese, information structure is typically expressed by marking sentence elements with either the case marker “ga” or topic marker “wa”.

In (35b), which is the Chinese translation of (35a), the subject “gou” (dog) is an indefinite noun expressing new information and so must be marked with “one + classifier”. Because of this stipulation in L1, it is assumed that Chinese L1 learners will overuse “one + classifier” in Japanese sentences like (35a). In I-JAS, the sentences including “one + classifier” produced by Chinese L1 learners basically co-occur with “-ga” (“-ga”: 10 cases, “-wo”: 2 cases, “no particle”: 1 case) and there are no cases of co-occurrence with the particle “-wa” in particular.

Table 12 shows a comparison of the use of “-wa” and “-ga” added to “inu” (dog) in the two scenes of the story writing task. While Japanese native speakers almost exclusively use “-ga” in both scenes, CLJ use “-wa” more frequently (35% in scene 1, 31% in scene 2). This may be caused by the fact that learners have not fully acquired the distinction between “-wa” to mark old information and “-ga” to mark new information.

Table 12 Use of “wa” and “ga” by Japanese native speakers and Chinese L1 learners

6 Implications for Chinese Teaching

Based on the three case studies presented above, in this section we outline the implications of our findings for teaching Chinese as a foreign or second language.

Elementary and intermediate Chinese language teaching materials currently in use in Japanese universities typically provide little or no explanation of the “indefinite” use of the “one + classifier” structure. Classifiers are treated as “units to count objects” and are usually only brought up in relation to the range of classifiers used to mark objects with different properties and shapes. Japanese does not have a grammatical form expressing indefiniteness and lacks obligatory marking of the (in)definiteness of noun phrases. As such, Japanese learners are predicted to struggle to acquire the “one + classifier” structure unless they are taught it explicitly.

We propose that the striking tendency for JLC to omit “one + classifier” is caused by such differences. Likewise, this may explain why ELC, whose native language shows a grammatical distinction between definite and indefinite, did not tend to omit “one + classifier”, instead overusing the structure on occasion.

The implications for Chinese language pedagogy aimed at L1 Japanese learners can be summarized as follows. First, the “indefinite” use of the “one + classifier” structure should be introduced from the elementary or intermediate level. Second, it may be effective to introduce the “one + classifier” structure through reference to the English indefinite article, to which all university level learners will have been exposed to. In this way, L2 (English) knowledge could potentially aid L3 (Chinese) acquisition.

7 Conclusion

This paper has introduced three case studies examining the possible influence of learners’ native language on the acquisition of L2 forms. The findings can be summarized as follows. In L2 Chinese, functional similarities between the indefinite article and “one + classifier” had a beneficial effect on L2 acquisition for English native speakers, whereas morphological similarities between “one + classifier” in Japanese and Chinese led to the omission of the target form by Japanese native speakers. The same morphological similarities also contributed to the overuse of “one + classifier” in Chinese native speakers’ L2 Japanese, although other factors including information structure also appear to be at play. Finally, functional similarity may also contribute to greater and more accurate use of English articles by Chinese native speakers in some contexts. Overall, this paper’s findings suggest that when teaching grammatical forms there is a need for nuanced consideration of characteristics of learners’ native languages, especially given that superficial morphological similarities may be just as likely to lead to errors than to native-like use.