A STUDY OF ENGLISH ARTICLE USAGE AMONG CROATIAN UNIVERSITY STUDENTS

: The main focus of this study was to analyse article errors in the writing of first year Croatian psychology students studying English as a foreign language as part of their study programme. Article errors among learners with different proficiency levels were also analysed; moreover, errors made by students with different essay grades were compared. The results show that the category with the highest number of errors included the omission of the indefinite article. Furthermore, it was found that proficiency levels were not a significant factor concerning the average number of errors in any of the categories. Moreover, it was also shown that essay grades did not play a significant role in the number of article errors in student writing except in the category of omission of the definite article, which was found among students with lower essay grades.


INTRODUCTION
The degree of influence of the native or first language (L1) on second language acquisition (SLA), in other words, the role of language transfer, has been an important topic in second language (L2) research (Ellis 1994;. One of the early approaches which stressed the role of the native language in SLA was Contrastive Analysis (CA). It was believed that the differences between languages could aid in discovering potential errors that learners might make in the target language. By predicting what would be easy or difficult for learners, appropriate pedagogical materials could be developed. This approach eventually lost its appeal when Noam Chomsky revolutionized the field of linguistics with the introduction of Transformational-Generative (TG) Grammar (1957Grammar ( , 1965. There was a shift in emphasis from the study of the surface forms and patterns of language to understanding the underlying rules that governed languages (Saville-Troike 2012). CA was eventually replaced by Error Analysis (Corder 1967) which considered errors important in understanding learners' L2 development. Various approaches began to influence SLA research as the field developed, including linguistic, psychological, pedagogical and social perspectives; consequently, the role of language transfer and its influence on L2 learning changed as well. One of the concerns of SLA research and L2 teaching has been on the language structures that are present in the target language but absent in L1. For example, the acquisition of English articles is a problem for many language learners whose L1 lacks this language structure. Studies in the field of English article acquisition have followed research trends in SLA research. It has been found that the influence of L1 may affect article acquisition in acquiring English as a second language. As a result, the major aim of this paper was to investigate English article usage among Croatian advanced level L2 learners whose mother tongue lacks this language structure.

LANGUAGE TRANSFER
One of the earliest approaches which considered the role of L1 in L2 acquisition, that is, the influence of language transfer, was Contrastive Analysis (CA). The leading figure in CA was Robert Lado (1957), who argued that a comparison of similarities and differences between L1 and L2 would allow researchers to predict and describe structures that learners would find easy and difficult to learn. Under the influence of structural linguists, attempts were made to compare and contrast the surface forms of the L1 and L2 language systems (Saville-Troike 2012). CA was influenced by the behaviourist theory of learning, which assumed that languages are acquired through habit formation based on a process of Stimulus-Response-Reinforcement (S-R-R). Language is learned by imitation and repetition of structures which are positively reinforced until they become habits. In L2 acquisition it was also assumed that transfer from the L1 would occur. Positive or facilitating transfer would take place if the same structure exists in L1 and L2, while negative transfer or interference would result if the structure is different (Gass and Selinker 2007). The CA approach to L2 learning resulted in numerous studies in which two languages were compared with the aim of discovering differences and thereby difficulties for learners of L2 (Mitchell and Myles 2004). In Croatia, for example, one of the major projects led by R. Filipović (Filipović 1972) was The Yugoslav Serbo-Croatian-English Contrastive Project. CA proved to be an important influence in L2 language and teaching research at the time (Prebeg-Vilke 1977).
The strong version of CA began to be heavily criticised and it eventually lost its appeal when the behaviourist language learning theory was challenged by Chomsky (1957 as cited in Lightbown andSpada 2006), who argued that children know more about the structure of their language than they have heard or have been taught. Chomsky stressed that language learning was rule-governed behaviour; moreover, children have an innate capacity to learn languages, which is biologically driven. Consequently, language came to be viewed in terms of structured rules rather than habits; similarly, learning was considered a process that involved active rule formation as opposed to imitation. Later, Chomsky (1981) reconceptualised the construct of Universal Grammar (UG) within the Principles and Parameters framework and suggested that principles "are properties of all languages in the world…(moreover), some of these principles contain parameters, or points where there is a limited choice of settings depending on which specific language is involved" (as cited in Saville-Troike 2012: 50). From the UG perspective, individuals reset the parameters from their L1 when learning an L2. Influenced by Chomsky's internal focus of language learning, studies began to show that not all errors that learners made were the result of interference from the L1, nor were errors always equated with the level of difficulty (Mitchell and Myles 2004).
The shift in focus from the surface forms and patterns of a language to underlying rules, that is, an emphasis on the creative capacity of learners to learn languages, led to Error Analysis (EA). In contrast to the behaviourist view of language learning, Corder (1967) argued that errors could show the system of language which the learner is using at a specific point in the process of L2 development, that is, the learners' transitional competence in the development of the L2. According to Corder (1967), errors could be categorised as interlingual (the result of negative transfer or interference from the L1), or intralingual (developmental errors within a language that could be the result of incomplete learning of L2 rules, or by over-generalising rules). Building on the concepts of EA, Larry Selinker (1972) introduced the term Interlanguage (IL), which described the intermediate states (or interim grammars) of a learner's language as it develops toward the target L2. The language of the L2 learner was considered to be a creative process that is systematic, dynamic, and variable (Saville-Troike 2012). Although a learner's IL was considered to be influenced by both L1 and input from L2, it was recognised as a third language system which was different from both the native and target language.
During the 1970s the minimalist position emerged, which downplayed the role of L1 in L2 acquisition, and instead stressed the importance of universal processes in language learning (Ellis 2008). Studies led by Dulay and Burt (1973) showed that the order of morpheme acquisition in English L2 learning was similar to English L1 acquisition, which suggested the existence of a natural order in the grammatical development of learners which was similar for both languages. Other studies also showed that learners follow a regular sequence in the acquisition of some syntactic constructions (Bailey et al. 1974). The implication was that transfer from L1 is of secondary importance since all language development, both L1 and subsequent language acquisition, is based on universal principles of language learning.
The minimalist position, however, was criticised with some researchers claiming that it undervalued the role of L1. Studies focusing on the crosslinguistic effects among languages on the level of production, discourse, and reception have shown that transfer is an important element of L2 acquisition (Ellis 2008). According to Ellis (2008), researchers have turned to verifying various constraints which may influence transfer, including social factors, markedness, prototypicality, language distance and psychotypolology, and developmental factors. Today, it is difficult to distinguish a single coherent theory of language transfer since crosslinguistic influences are too numerous and complex (Odlin 2003).

Research on the L2 acquisition of articles
The influence of L1 on L2 becomes an interesting problem when there is an absence of a similar structure in L1. For example, the acquisition of English articles poses a major problem for Croatian learners of English. Research focused on the acquisition of English articles has developed in a similar way as has SLA research (Ekiert 2007). According to Butler (2002), early studies in article usage focused on article errors which were attributed to inadequate acquisition of certain grammatical norms of L2. There was also an assumption that IL develops through a number of stages or developmental sequences. In the 1980's, researchers (e.g. Huebner 1979Huebner , 1983 began to be interested in the semantics of IL and thus there was a shift in article research from the acquisition of morphology as form to a focus on morphology as the surface realization of an underlying semantic and conceptual system (Ekeirt 2007).
Since then, there have been many different approaches that have attempted to explain the process of the acquisition of articles. Huebner (1983) argued that it is possible that there is a semantic discord between the L1 and L2 with regard to how concepts such as definiteness and/or specificity are encoded in the L1 and L2. However, a problem occurs when the learner's L1 does not have an equivalent form. Other researchers have argued that although learners whose L1 does not have an article system might have problems in article acquisition, they can, and often do, have access to Universal Grammar (UG) features such as definiteness, specificity and partitivity (Ionin, Ko and Wexler 2004). Ionin (2004) describes definiteness as a discourse-related semantic feature related to the knowledge state of the speaker and hearer, while the notion of specificity refers to the knowledge understood only by the speaker. These notions are important for the Article Choice Parameter (Ionin 2007), by which the author explains the variability in the L2 learners' production of articles that leads to the Fluctuation Hypothesis. Under this formalist hypothesis, L2 learners are predicted to fluctuate between the two settings until the input leads them to the right option.
In terms of how articles are acquired, many studies have shown that definiteness is acquired earlier than indefiniteness (Chaudron and Parker 1990;Lardiere 2004). Research has also shown that many learners have the most difficulty in acquiring indefinite articles, especially learners whose L1 lacks a corresponding article system (Roberston 2000;Jarvis 2002: White 2003. Moreover, some research has shown that proficiency is a factor in article acquisition (Young 1996), while others have shown that L2 level is not a factor in article acquisition (Ionin 2003).
The development of the proficient use of articles in the interlanguage of English as a foreign language (EFL) for learners has been found to be a very long process with pronounced variability (Jarvis 2002;Ko, Ionin and Wexler 2004;Zdorenko and Paradis 2008). This is even more so for the learners with an article-less L1 (Trenkić 2002;Zergollern-Miletić 2008Balenović and Medved Krajnović 2013;. L1 interference certainly plays a significant role (Liu & Gleason 2002), although Ionin, Zubizarreta and Maldonado (2008) showed that fluctuation overrides language transfer with L2 learners whose L1 does not contain articles (e.g. Russian), while transfer overrides fluctuation with L2 learners whose L1 has an article system (e.g. Spanish). In other words, learners transfer article semantics from their L1 to their L2.

Research on the L2 acquisition of articles in Croatia
There have been several recent studies on the acquisition of English articles by Croatian EFL learners Balenović and Medved Krajnović 2013;Zergollern-Miletić 2008;2009;2010; whose L1 lacks an article system to express the category of definiteness/indefiniteness. The lack of articles in learners' L1 supposes difficulties in their acquisition, since learners are not aware of the existence of the category in their L1, even though this category is often seen as a linguistic universal (Chomsky 2000;Silić 2000) expressed with different linguistic means in different languages. In the Croatian language, Silić (2000) considers this category as a part of determiners linked to adjectives, while Pranjković (2000) links it to nouns rather than adjectives. Thus, linguists explain the concepts of definiteness and indefiniteness in the Croatian language in different ways.
Zergollern-Miletić (2008) conducted a study among advanced L2 speakers of English (Croatian university majors in English) with the purpose of examining their use of articles. The research results showed the wrong use of articles with abstract nouns, the omission of articles when the noun was defined by an adjective, as well as article substitution (definite for indefinite and vice versa). Thus, the author stressed the importance of developing learners' awareness of the existence of the concept of definiteness/indefiniteness in their first language, in spite of the fact that the Croatian language lacks an article system. Balenović ( , 2015 conducted a study on the acquisition of English articles among primary school EFL learners within the framework of a larger research project titled Early acquisition of English as a foreign language: The analysis of the learner's interlanguage (for details see Mihaljević Djigunović and Medved Krajnović 2015). The longitudinal study was based on the recordings of individual oral production tasks of Croatian EFL learners (mean age 7-11), over a period of three years. The purpose of the study was to analyse the process of article acquisition among Croatian primary school EFL learners. The research results showed frequent omission of the use of the indefinite article in front of nouns mentioned for the first time, or article substitution. Furthermore, the indefinite article was found to be used with higher accuracy than the definite article, which is in contrast with some other studies (e.g. Ionin and Wexler 2004) which showed that the definite article is commonly acquired before the indefinite article. The use of the number one instead of the indefinite article was also found, probably due to the influence of learners' L1. The author also found the omission of the indefinite article at the beginning of a sentence and the correct use of the definite article in the same sentence. This finding implies that in early EFL learning learners cannot concentrate on more than one linguistic problem within one utterance.
Zergollern-Miletić (2015) conducted a study within the above mentioned research project, based on oral and written production tasks of Croatian EFL learners who started learning English in the fourth grade. The study was conducted longitudinally, over a period of four years, and included primary school learners from grades 5 to 8. The author expected a general omission of articles. The research results showed that learners omitted articles more in the written tasks as compared to the oral ones, but not in as many cases as assumed. Article substitution was also found, for example, the definite article for the indefinite. The author concluded that learners had internalized the notions of familiarity, countability and uniqueness to a certain extent; however, young Croatian EFL learners had not entirely mastered the use of English articles, such as the use of articles before abstract nouns. In short, it may be suggested that the English article system consists of high levels of lexico-grammatical complexity and multifunctionality (Hawkins & Mayo 2006), which may be difficult to acquire even for L1 users (Thomas 1989).

AIM OF THE STUDY
Although contrasting results have been shown regarding the acquisition of English articles among Croatian EFL learners, more research is needed to substantiate current findings. This study aims to investigate article usage among an under-investigated sample, in particular, non-philological university majors taking an English language course as part of their study programme. Specifically, the study examined the types of article errors that occur in L2 essay writing and whether there are differences in the number of errors based on learners' proficiency levels and essay grade levels. The following research questions were addressed: 1. What kinds of errors in article usage occur in student essay writing? 2. Are there differences between the number of article errors among students with different proficiency levels? 3. Are there differences between the number of article errors among students with different essay grades?

Participants
The sample included the written work of 29 first year psychology students from the University of Zadar, who were taking English classes as part of their study programme. The participants included 26 females and 3 males. Moreover, the number of years of studying English ranged from 4 to 14 (Chart 1), while the average number of years of studying English was 10.3 (Table 1). Students were tested at the beginning of the year with regard to their proficiency level using the Oxford Placement test 2 (Allan 2002). It was found that the majority of students (16) were at the upper-intermediate level ( Table 2). The average essay grade was 3 on scale of 1 to 5, with 5 being the highest grade (Table 3)

Corpus
The corpus consisted of a total of 29 essays which were written by first year psychology students from the University of Zadar as part of their final English written exam at the end of the first semester. The exam entailed reading, vocabulary, grammar and writing sections, which the students had to complete in 2 hours. The writing task involved a 'for and against' essay of approximately 250 words. Students had practiced writing this type of essay in their English class during the semester; moreover, it should be noted that Croatian students are required to write this type of essay for the university state entrance exam, thus they had plenty of practice with this type of writing.

Data analysis
The essays were analysed for article mistakes by a native speaker of English and discussed with the authors of the paper. It was found that mistakes in article usage fell into seven categories, which are listed below. The results were then compared with student proficiency level scores and with essay grades using the Kruskal-Wallis test.

Categories of article errors
After analysing the essays it was found that incorrect usage of articles fell into seven categories: 1) omission of the, 2) the used instead of Ø, 3) a used instead of the, 4) a used instead of Ø, 5) omission of a, 6) the used instead of a, 7) a used instead of an. Table 4 shows the results of the distribution of mistakes with definite and indefinite articles, Mean (M) (the average number of errors in each category for this sample, N=29), and Standard Deviation (SD) scores. As shown in the above table, the majority of mistakes with article usage included the omission of the indefinite article a (N=103). This category clearly outnumbered the second most frequent category of mistakes which included the omission of the definite article the (N= 38). This is followed by overuse of the definite article (N= 20), and overuse of the indefinite article (N= 18).

Comparison of article errors among proficiency levels
The next step in the analysis was focused on the differences in mistakes in article usage among students with various proficiency levels. A choice was made to use a non-parametric test due to the fact that the descriptive statistics showed that the data was not normally distributed, and also because non-parametric tests are usually used with frequency data (Mackey & Gass 2005). The descriptive statistics showed that various proficiency level groups had less than 5 samples; consequently, several proficiency groups had to be combined in order to satisfy the minimum criterion for standard statistical procedures. The proficiency and advanced groups were put into group 1 (A), the upper-intermediate group into group 2 (B), and the intermediate and elementary groups were put into group 3 (C). Since the independent variable (proficiency levels) included 3 or more groups, the Kruskal-Wallis test was used. The results show no statistically significant differences between the number of errors in categories 1, 2, 4, 5 and proficiency scores: Category 1 (H (2) = 5.1, p = .08) with a mean rank of 9.7 for group A, 18 for group B, and 13 for group C; Category 2 (H (2) = 2.2, p = .33) with a mean rank of 11.4 for group A, 16.6 for group B, and 14.4 for group C; Category 4 (H (2) = 3.1, p = .22) with a mean rank of 11.2 for group A, 17.1 for group B, and 14 for group C; Category 5 (H (2) = 5.8, p = .06) with a mean rank of 7.7 for group A, 17 for group B, and 17 for group C. Error scores for categories 3, 6, and 7 numbered 3 or less; consequently, no comparisons could be made for the three proficiency groups.

Comparison of article errors and essay grades
The Kruskal-Wallis test was also used to measure for statistically significant differences between article errors and students' essay grade levels. Once again, in order to satisfy standard statistical procedures, the essay grade level categories were combined: Group 1 (A) included essay grades of 4 and 5, Group 2 (B) included essay scores of 3, and Group 3 (C) included essay scores of 2. The results show no statistically significant differences between the number of article errors from categories 2, 4, 5 among students with different essay grades: Category 2 (H (2) = 1.8, p = .41) with a mean rank of 15.7 for group A, 16.0 for group B, and 11.4 for group C; Category 4 (H (2) = 3.0, p = .23) with a mean rank of 13.9 for group A, 13.7 for group B, and 19.7 for group C; Category 5 (H (2) = 4.2, p = .13) with a mean rank of 10.1 for group A, 16.1 for group B, and 18.8 for group C. However, there was a statistically significant difference in Category 1 (H (2) = 6.4, p = .04) with a mean rank of 8.9 for group A, 17.0 for group B, and 18.1 for group C. The descriptive statistics show that the majority of errors in omission of the definite article were made by students with an essay grade of 2. Once again, since error scores for categories 3, 6, and 7 numbered 3 or less, no comparisons could be made for the three essay grade level groups.

DISCUSSION
The results show that errors with the indefinite article by far outweigh other article errors. This is consistent with many other studies (Roberston 2000;Jarvis 2002;White 2003) which also found the omission of the indefinite article to be the greatest source of problems related to article usage. Many early studies in the acquisition order of grammatical morphemes, including the acquisition of definite and indefinite articles have shown that definite articles are acquired before indefinite articles (Hakuta 1976;Huebner 1979;1983). On the other hand, Balenović and Medved Krajnović's (2013) study on the acquisition of English articles in early EFL learning in Croatia found that the indefinite article was used with higher accuracy than the definite article. The authors attributed the results to the influence of the context of learning, that is, the teaching materials, and the fact that Croatian young EFL learners are exposed more to the indefinite than to the definite article. For instance, new words are introduced as whole phrases, such as in the following example: 'This is a book'. It could be possible that Croatian learners have acquired the indefinite article and are still in the process of acquiring the definite article in their IL.
However, when a comparison was made with proficiency levels, it was found that there were no statistically significant differences between article errors in any of the categories among students with different L2 proficiency levels. Ionin (2003) also found this to be the case with Russian ESL learners. Nevertheless, the results of this study must be taken with a note of caution. Namely, the sample size was relatively small (29) and the number of errors within each group was also small, which reduces the statistical weight of the results. A descriptive analysis of the number of errors made by students who were at the advanced and proficiency levels of L2 competence shows that fewer articles errors were made by these students; however, a larger sample size is needed in order to verify the actual significant statistical differences.
In addition, no statistically significant differences were found between article errors among students with different essay grades in category 2 (the used instead of Ø), category 4 (a used instead of Ø) and category 5 (omission of a). There was a statistical difference in category 1 (omission of the) among students with lower essay grades, which might suggest that students have not acquired this category at this stage. Nevertheless, this result must also be taken with some reserve. The small sample size in each category, as well as the small number of errors, reduces the effect of this statistical difference.
Notwithstanding the statistical limitations noted above, certain general trends in article usage can be noted. Students continue to make article mistakes in their writing. Many studies have shown that if an article system is lacking in L1, more difficulty is encountered by learners. For instance, Roberston (2000) found this to be the case with Chinese ESL learners; Jarvis (2002) (2008) with Syrian Arab native speakers. However, although all these languages lack an equivalent article system, it was found that the most frequent article errors were related to the indefinite article. This study also shows that errors related to the indefinite article were the most frequent. Ekiert (2007) argues that L2 acquisition can be thwarted by the lack of an equivalent conceptual and semantic system in the L1. Specifically, a learner whose L1 lacks an article system may have difficulty in conceptualizing notions such as indefiniteness (Ekiert 2007: 32). Thus, it may be suggested that learners generally have problems with articles due to the lack of a similar system in their L1, but it might also be possible that, for these language learners, notions of definiteness are easier to acquire than notions of indefiniteness.

CONCLUSION
This study has shown that Croatian EFL learners make article errors in their writing, especially with regard to the indefinite article. It is possible that the acquisition of English articles among Croatian EFL learners is related to a natural order of acquisition; moreover, it may also be possible that the omission of the indefinite article shows learner difficulties in the concepts of indefiniteness. Descriptive statistics has shown that advanced level students have, on average, fewer errors than lower proficiency level students (the proficient level students had no article errors), which might indicate that Croatian learners eventually do manage to acquire English articles. This might lend credibility to UG-based explanations for L2 acquisition, whereby learners rely on the principles and parameters of UG in the L2 acquisition process.