“THE ROLE OF THE FIRST LANGUAGE (TURKISH) IN

SECOND LANGUAGE (ENGLISH) ACQUISITION OF ARTICLES”

Kübranur Toplar -1808344-

THREE STUDY • “Use of English Articles by Speakers of Turkish in the

EFL Setting” by GÜLİN DAĞDEVİREN

• “Second Language Acquisition of the Turkish Article System by Turkish Learners: The Role of Semantic Notions” by ZEYNEP ATAY

• “The Role of the First Language in Child Second Language Acquisition of Articles” by TATIANA ZDORENKO AND JOHANNE PARADIS

In the study by Dagdeviren• The data collection instrument: A fill-in-the-blanks test which

included six types of sentences.• The test was administered to 30 low-proficiency and 30 high-

proficiency level participants to evaluate their levels’ effect on the acquisition on articles in English.

• The results of the fill-in-the-blanks test showed that accuracy rates in six types of sentences were higher than the overuses. In other words, the participants of the study did not associate article “the” with [+specific] contexts and “a” with [-specific] contexts.

• It was also found that the overuse of article “the” did not reach its maximum level in [+partitive, +specific] contexts.The study also investigates the article choice of the participants of different proficiency levels, it is clear that the results of the current study are in the line with the previous literature which evidences the role of proficiency in the article choice in L2-English.

In the study by Atay• Participants : 40 elementary, 40 intermediate and 40 upper-intermediate

students• Data collection instrument : A forced-choice elicitation task. 40 short and

contextualized dialogues.• Each dialogue is missing an article and learners were asked to fill the gap with

an appropriate article; “a/an”, “the” or “Ø”.• Dialogues belong to four different contexts; i.e. definite/specific, definite/non-

specific, indefinite/specific and indefinite/non-specific.• Each context has 10 dialogues with four different contexts that are

randomized. Data were analyzed using SPSS 17 packet program (descriptive analysis and one-way ANOVA).

• The results: Intermediate level learners exhibited fluctuation between definiteness and specificity in (+definite/-specific) and (-definite/+specific) contexts. Elementary level learners were more accurate in these contexts exhibiting article omission errors in definite contexts. Upper intermediate level students were quite successful in article assignment in defined contexts. This revealed that there is a positive correlation between article system acquisition and proficiency.

In the study by Zdorenko and Paradis• They used elicited narratives as the material for studying the use

of articles in L2 children.• They created a longitudinal corpus of stories from 16 children

who are at the age of 4 and 5 at the outset and acquire English as a second language.

• The children were followed for 2 years. • They divided the learners into two groups, a group with [–article] in the first language and a group with [+article] in the

first language. • The study was designed to determine whether the initial state in

child’s second language acquisition shows evidence for transfer of the first language with respect to articles; to investigate whether the second language learners from both the [+article] and [–article] groups acquire articles in the same sequence; and to compare the rate of article overuse in the two groups.

LITERATURE OVERVIEW• Acquisition of the English article system is one of the

biggest challenges in this process, especially for the learners whose native language has no article system.

• Articles are unstressed function words, hence perceptually non-salient and semantically light-weight.

• Second language learners, mostly seek one-to-one form-meaning relationship while acquiring the language, so the concepts and the rules become easier to understand and internalize. However, the article system does not have one-to-one form-meaning connection in discourse because it is context dependent and too complex to be identified via simple and one way form-meaning association.

• The studies mainly focus on the acquisition of English article system which defines definiteness, specificity and partitivity as semantic features.

• Definiteness means that a determiner phrase is definite if its referent is known to both speaker and hearer, and is unique in the contextually relevant domain.Otherwise, the phrase is indefinite.

• In other words, the knowledge of a definite noun phrase is shared by the speaker and the hearer. If a determiner phrase of the form [D NP] is [+specific], then the speaker intends to refer to a unique individual in the set denoted by the noun phrase, and considers this individual to possess some noteworthy property. While definiteness is marked by using the, a and zero article, specificity is not encoded in English.

No matter they are specific or not , they receive an article on the basis of definiteness:

• a. My brother is looking for a book- It is on archaeological ethics.

• b. My brother is looking for a book. I don’t know what it is about.

In (2a) the speaker and the hearer share the knowledge of a salient book and this referent has a noteworthy property. Therefore, the referent is specific. However, in (2b) noteworthiness and uniqueness of a referent are not provided, thus it is interpreted as nonspecific.As seen in the sentences given below, specificity is not marked in English. No matter they are specific or not , they receive an article on the basis of definiteness.

Presuppositionality is the other semantic feature. The definition of presuppositionality based on that If DP of the form [D NP] is [+presuppositional], then the speaker assumes that the hearer shares the presupposition of the existence of a unique individual in the set denoted by NP. It is asserted that presuppositionality is established either by introducing a set that the referent belongs to or by mutual world knowledge. It is defined that the first type of presuppositionality as partitivity which is similar to previous mentioned definiteness except for uniqueness.

• a.Before going to beach trip she had to buy many things: bikinis, hats, towels, sunglasses and suncreams. But she was able to buy a hat as she was nearly broke!

• Hat belongs to a set that includes bikinis, hats, towels, suncreams and this set was introduced before; thus it has a partitive reading.

RESEARCH QUESTIONS The study of Dagdeviren aims to examine the

choice of English article system by the speakers of Turkish to compare with the previous studies. The participants’ article choice is also taken into consideration on the basis of proficiency levels. The study addresses the following research questions:

• 1. How do L1-Turkish speakers mark [+specific] and [-specific] contexts?

• 2. How do L1-Turkish speakers mark [+partitive, +specific] contexts?

• 3. Does accuracy of article use vary in all types of sentences with respect to proficiency level?

The aim of the study of Atay is to investigate the article system acquisition of L1 Turkish learners of English and to seek the effects of semantic universals to this process. Depending on the purposes, the research questions of this study are;

• 1) What are the systematic errors of L1 Turkish learners on the course of English article system acquisition?

• 2) What are the reasons of these systematic errors observed in L1 Turkish learners’ data?

• 3) What are the developmental features of Turkish learners’ acquisition of English articles? Does proficiency have an effect in the correct use of articles?

• 4) What kind of pedagogical implications can be drawn from the findings? How do the findings help teaching?

PREDICTIONS: • She predicted that in her study she will find out L1 Turkish

learners of English will associate the definite article with specificity.

• Definite article will be overused in –definite / +specific contexts.

• Indefinite article will be overused in +definite/-specific contexts.

• Fluctuation frequency will differ according to the proficiency level: as learners become more proficient, they will start to set the correct parameter for the article choice and assign the articles of the target language in the desired way.

• Upper-intermediate level learners will fluctuate less than intermediate counterparts. Similarly intermediate level students will fluctuate less than elementary level students.

In their study, Zdorenko and Paradis examined longitudinal data from English L2 children from [+article] and [–article] L1 backgrounds to address the following questions:

• (1) Does L1 background play a role in the acquisition sequence? In other words, do L2 learners from both L1 backgrounds acquire definite before indefinite articles?

• (2) Do L2 learners from [–article] and [+article] L1 backgrounds omit articles to the same extent? Do [–article] L1 learners acquire correct use of articles more slowly?

• (3) Is the overuse a common error in both [+article] and [–article] L1 groups?

EXPECTATIONS: • The same pattern in both groups, since definite

before indefinite is the sequence documented in both L1 and adult L2 acquisition.

• The [–article] L1 group to omit articles more frequently than the [+article] group, and acquire them more slowly as L1 transfer is likely to underlie this difference since the learners from both groups are exposed to the same input.

• Both groups of children would show this error, since it is also found in L1 acquisition.

METHOD - PARTICIPANTSStudy by Zdorenko and Paradis:• Sixteen children learning English as a L2 in

Edmonton, Canada, were studied every six months for approximately 2 years.

• Children were from new Canadian families and had little or no exposure to English before regular attendance at a preschool or school program.

• Onset of exposure to English was determined by children’s entry into such a program. The children’s mean age was 4 and 5 and mean exposure to English was 9 months at the onset of the study.

• Mean ages and months of exposure at each round of testing are given in Table 1:

Table 1. Average ages and months of exposure (MOE) at each round of data collection

The children were selected for the study in order to form two groups, roughly equivalent in size, with [+article] and [–article] L1 backgrounds. The [+article] group included 7 children whose L1s were Arabic, Romanian, and Spanish, and the [–article] group included 9 children whose L1s were Cantonese, Japanese, Korean, and Mandarin.

Study by Atay:• The participants were METU Basic English Department preparatory class

students who had proficiency test and according to the results of the test their proficiency levels were determined. Their ages ranged from 18-20. None of them were bilingual.

• For the study, learners from three different levels were tested; 40 elementary students, 40 intermediate level students and 40 upper-intermediate level students.

• They were given half an hour to complete the task, which is composed of 40 contextualized mini-dialogues.

• The questionnaire was piloted with fifteen participants from TOBB University of Economics and Technology. They were from different proficiency levels: 5 from elementary, 5 from intermediate and 5 from upper-intermediate level.

• To test the validity of the data collection instrument, task was applied to a test group that consists 5 adult native speakers of English living in England. They were students at Nottingham Trent University. The test group was accessed via e-mail. The elicitation task was sent back via e-mail, again.

Study by Dağdeviren:• The participants of the study were L1-Turkish speakers

of L2-English. They were college students studying at an English-medium university in Turkey. The test was administered to low- and high-proficiency groups, each including 30 participants. The mean age of all participants is 20.45 and the groups included 27 males and 33 females.

• All participants took part in the study on a voluntary basis; none of them was paid or received academic credit for participating.

DATA COLLECTION INSTRUMENT

Study by Zdorenko and Paradis:• In each of the 5 testing sessions, the children told stories, following two

picture books, to the experimenter who could not see the pictures. This was done to ensure that the child could not resort to gestural communication, i.e., pointing, and could not assume mutual speaker/hearer knowledge based on joint attention to the same picture.

• In the analysis of the use of articles in the stories, two contexts of article use were set apart: first and subsequent mentions contexts were regarded as indefinite and definite contexts respectively, in order to make it clear which article was the target form.

• The analysis was limited only to referring expressions that were used for the characters and concrete objects in the stories, i.e. four animate characters and three objects in each picture book. Articles used were then coded according to their appropriateness in the context as ‘correct’ or ‘incorrect’. Article omission was coded as ‘null article’ (which was always an error, since proper names were not included in the analysis).

Examples of data coding are provided below. • (1) Incorrect the in indefinite context *EXP: how do you start? *CHI: # mm # the elephant throw the ball. • (2) Correct a, incorrect null article in indefinite contexts *CHI: a little elephant in # in the pool *CHI: Ø giraffe see it too.• (3) Correct the in definite context *CHI: first there are two cross-eyed animals at the

pool. *CHI: one was elephant … who had a very fat body. *CHI: and the elephant, the female elephant was

bouncing up and down her ball.

Study by Atay: 1. Forced-Choice Elicitation Task • A forced choice elicitation task is used to elicit data. The task consists of 40

contextualized mini dialogues which belong to four different contexts (definite-specific, indefinite-specific, definite-nonspecific and indefinite-nonspecific) Each context has 10 mini dialogues and in total there were 40 conversations.

• The order of the items in the task was truly random. In the task, in each dialogue, the target sentence was missing an article. The learners were asked to choose among the articles given (a /an /the /Ø) basing their responses on the proceeding context.

• All the task items were contextualized because it is known that the interpretation of nouns or NPs (hence the required article) may change depending on the contexts.

• Data collection instrument was prepared by Atay. To test the reliability KR-21 test were applied and all task items were proved as reliable. To test the content validity and be sure about the reliability the task had been applied to a test group. To ensure about the face validity task items were applied to a piloting group.

• Before the task was distributed, the participants were informed that the purpose is not to test their accuracy but their intuitions about the article choice. So they were asked to choose the article which first comes to their mind and not to change their answers.

Here are example dialogues from each context;• Definite/specific• At a restaurant• A: Hey! Did you see ____ (Ø / a/ an / the) waiter?• B: Yes, but what’s so surprising about him?• A: He is my sister’s fiancé.• Definite/non-specific • Definite/non-specific• Two friends are chatting • A: Did you hear what happened? Someone broke into Mrs. Romney’s flat and stole her• jewelry. • B: Oh! Did the police catch ____ (Ø / a/ an / the) thief? • A: Not yet, they have no idea about his / her identity, but they are investigating.• Indefinite/specific• Phone conversation between siblings• Julia: Hi! It’s Julia. How are you doing?• Gary: Good Julia thanks but this is the wrong time to call. I must go now because I’m going

to meet ____ (Ø / a/ an / the) friend who is very special Sorry!• • Indefinite/non-specific• Mother and son are chatting in the kitchen• Mother: How was the birthday party?• Son: Everything was marvelous, mum. Alan’s girlfriend, Catharine, told us that Alan’s

father bought him ____ (Ø / a/ an / the) expensive sports car.

For the research two different analysis techniques have been performed:

• The first one is the descriptive analysis of the SPSS 17.0 packet program. This analysis shows us at what percentages learners assign the target article and make errors.

• The second technique is one way ANOVA. This technique is applied in order to compare the means of participants from different levels and to see how significant the participants’ mean scores are in four different contexts with respect to their proficiency levels.

In order to see the significance of the findings, Tukey and Post Hoc tests have been examined.

Study by Dağdeviren:• The data collection method is a kind of short-answer test. Due

to the fact that short answers tests prevent learners from recognizing the answer from selected-response items and answering correctly by merely guessing, a fill-in-the-blanks test was administered to participants.

• The number of the items used in the current test was 40, which included 30 critical items and 10 fillers. The 30 critical items represented 6 semantic contexts which were embodied in 5 test items for each. These semantic contexts are [+definite, -partitive, +specific], [+definite, - partitive, -specific], [-definite, -partitive, +specific], [-definite, -partitive, -specific], [-definite, +partitive, +specific] and [-definite, +partitive, -specific].

• Filler items of the test targeted a range of words such as pronouns, auxiliaries.

• The Data Coding: The coding process was started by marking sentence types and fillers in the test. The 6 sentence types defines 6 semantic contexts which are

Type 1 sentences [+definite, -partitive, +specific]Type 2 sentences [+definite, -partitive, -specific]Type 3 sentences [-definite, - partitive, +specific]Type 4 sentences [-definite, -partitive, -specific]Type 5 sentences [- definite, +partitive, +specific]Type 6 sentences [-definite, +partitive, -specific].

The statistical analysis of the test was made by using SPSS 11.5(Statistical Package for the Social Sciences). Answers elicitedfrom participants were coded in four labels which are 1.00 =“the”, 2.00 = “a/an”, 3.00 = “dash” and 9.00 = “other”. Thevariables of the test were also included sex, age, proficiencylevel of English, department, foreign language and proficiencylevel in this language.

RESULTS

Study by Zdorenko and Paradis: 1. Accuracy in the use of definite and indefinite articles

The questions they asked for the analyses of the children’s accuracy in the use of articles were as follows:

(1) Are accuracy rates higher with the than they are with a, and when do each reach over 90%?

(2) (2) Are the rates of acquisition different for children in the [–article] than in the [+article] L1 group?

Children’s accurate use of a was calculated as a proportion of uses of “a” out of

all indefinite contexts where a was expected. Similarly, children’s accurate use

of “the” was calculated as a percentage of uses of the in all definite contexts

involving common nouns as referents. The resulting percentage correct use of

“a” and “the” for each round are shown in Figure 1a for the [+article] L1 group

and Figure 1b for the [–article] L1 group.

Figure 1a: Mean percent correct use of articles for [+article] L1 group

• For both L1 groups, accuracy rates were consistently higher at all rounds with the in definite contexts than they were with a in indefinite contexts. Above 90% accuracy for the use of the in definite contexts was reached by round 3 by both groups, i.e. in about 21 months from the onset. For a use in indefinite contexts, accuracy was still below 90% at the last round.

• Results of a three-way mixed ANOVA with L1 background as a between-subjects factor ([+article] group and [–article] group), and round (5 levels) and article type (definite and indefinite) as within-subjects factors, revealed that both round (F(4,48) = 10.68, p = .000) and article type (F(1,12) = 31.723, p = .000) main effects were significant, but L1 background was not (F(1,12) = 2.030, p = .180).

• The significant main effect for the two-level factor, article type, confirms that the children were more accurate with the definite than indefinite article, as the means for the former were consistently higher.

• Applying planned independent sample t-tests on the correct use in context of each article type between each L1 group at each round, we found just one significant result in all the pair wise comparisons: The [–article] group had lower accuracy for the definite article at round 1 than the [+article group] (60.2% vs. 92.2%,t(13) = -3.869, p = .002). Over all, L1 background did not exert much influence on children’s accuracy with article choice in context, with the exception that the [–article] group were lagging behind in accuracy at round 1 with the. Put differently, article type context was a more important factor than L1 background in determining these children’s accurate use of English articles.

2. Error type distribution in definite and indefinite contexts• Previous research suggests that the overuse of the is a much more frequent

error than the overuse of a. • The accuracy analyses in the previous section support the prediction that

accuracy is higher in definite contexts, but do not directly indicate if the overuse is the most common error, since the overuse would constitute the use of the in indefinite contexts.

• Furthermore, previous research has indicated that the overuse of null articles in L2 English was also documented in learners.

• The questions we asked for the analyses of error types were (1) What is the relative distribution of the overuse, a overuse and null article

overuse among children’s errors?(2) Is the overuse the most common error?(3) Does L1 background interact with error types, for example, do the [–article]

children display more null article overuse than the [+article] children?First, the proportion of incorrect use of the, a and null articles was calculatedfrom the total number of incorrect uses at each round across all children, anddivided into L1 group and definite and indefinite contexts. The results of theseanalyses are presented in Figures 2a and 2b.

• The results show that the overuse is clearly the dominant error type for both the [+article] and

• [–article] groups, and that null articles are an error type specific to the [–article] group, since they are negligible in the [+article] group data but appear in the [–article] group data in both definite and indefinite contexts. Furthermore, null articles begin to disappear even for the [–article] group after round 3, indicating that these errors are more frequent during the early stages of English L2 acquisition. Independent sample t-tests performed on the means collapsed across time support the distributional data in Figures 2a and 2b. The only significant group differences are for null article use in definite (t (15)=2.375, p=.036) and indefinite contexts (t(15)=2.638, p=.019), with the [–article] children having a higher proportion of these errors.

• Study by Atay:• 1. +Definite/+Specific Context• Table 5 gives one way ANOVA scores for + definite /

+ specific contexts according to proficiency level. The results show that upper intermediate level students’ mean score in “+ definite / + specific” contexts (X= 9,300) is respectively higher than intermediate level (X= 8, 0750) and elementary level (X=7, 2500) students’ mean scores. Stated differences among mean scores are statistically significant; F (2-117) =31.381 p<, 001.

• The findings indicate that students’ achievement in + definite / + specific contexts change according to proficiency level.

• In order to find out exactly which means are significantly different from which other ones, Post Hoc tests have been calculated and examined.

• According to the Tukey test, mean differences between all groups are statistically significant (F (2-117) =31.381 p<, 001). This means that students’ achievement on “+ definite / + specific” contexts increases, while their proficiency in English language improves.

• Following the ANOVA, descriptive analysis has also been carried out to see the article choice percentages of L1 Turkish learners of English in definite specific contexts.

• When the data is analyzed for the +definite/+specific context it is found out that Turkish learners assign the target item correctly at a considerably high percentage. The upper-intermediate group’s performance is significantly better than that of intermediate and elementary level. They are able to supply the correctly at a rate of 93, 0%, while intermediate and elementary level learners fall behind with 80,8 % and 72,5 %, respectively. Table 5 illustrates overall results of the analysis for the +definite/+specific context.

Table 2: Article Choice Percentages in +Definite/+Specific Context

• Another significant finding is that L1 Turkish learners of English tend to omit articles in certain items of this context. Especially elementary level students has high rates for article omission when compared to intermediate and upper-intermediate counterparts (17, 3 % = elementary, 9, 8 %= intermediate, 6, 0 %= upper-intermediate).

• The rates of article omission are given in Table 5. Looking at the results it is fair to state that there is an inverse proportion between the proficiency level and article omission. Article omission rate in the elementary group is noteworthy with a rate 17.3 %; however in intermediate and upper-intermediate groups omission rates are not that much significant.

2.Definite/+Specific ContextTable 7 gives one way ANOVA scores for - Definite / + Specific contexts according tothe proficiency level.

• According to the results upper intermediate level students’ mean score in - definite / + specific contexts (X= 8, 8750) has a higher proportion than that of the intermediate level (X= 5, 7500) and that of the elementary level (X=7, 5750) students. Interestingly, however, elementary level students’ mean score is higher than intermediate level students’ (X=7, 5750 vs. X=5, 7500 respectively). Differences among mean scores are statistically significant F (2-117) =51,746 p<, 001. The findings show that students’ achievement in - definite / + specific contexts changes according to proficiency level. The results of the descriptive statistics indicated that in –definite/+specific context Turkish learners are observed the overuse.

• Table 8 displays the overall results for this context.

Table 8: Article Choice Percentages in -Definite/+Specific Contexts

• The table reveals that in indefinite specific context particularly the intermediate group fails to assign the correct article.

• The group use the instead of a in -definite/+specific contexts at a rate of %36, 3 which is a considerably high percentage. Interestingly however, contrary to my predictions elementary level learners perform significantly better than intermediate counterparts with the 15, 0 % rate of substitution. This rate still shows that they are observed the overuse; however compared to upper-intermediate group whose substitution rate is 11, 0 %, elementary group’s performance is undeniably good.

• With respect to upper-intermediate level learners, it is fair to assert that although they have the least overuse rate among groups (11, 0 %,), they still demonstrate significant percentage of systematic substitution error. Compared to +definite/+specific context, upper-intermediates are less successful in assigning correct article in –definite/+specific context.

• In this context there is an item which is formed with the relative clause structure. In this item the target article is a because the item is –definite/+specific. However, relative clause structure leads learners to confusion because they are taught that before the relative clause structure definite article must be used. As a result most of the learners use definite article without considering the context’s itself. 62.5 % of intermediate level students assign definite article in this item instead of the indefinite one. This problematic item is given below:

• 3. Phone conversation between siblings• Julia: Hi! It’s Julia. How are you doing?• Gary: Good Julia, thanks but this is the wrong

time to call. I must go now because I’m going to meet ____ (Ø / a/ an / the) friend who is very special Sorry!

• In terms of article omission, elementary level Turkish learners have the highest ratio with the rate of 9, 3%. For intermediate and upper –intermediate groups, article omission rate is not significant; 6, 3% and 0, 3% respectively.

3. + Definite / - Specific ContextOverall ANOVA results for definite nonspecific contexts are given in Table 9.

• The results in the table show that in + definite / - specific context upper intermediate level students’ mean score (X= 9, 4000) has the highest ratio. Interestingly again in this context elementary group has a higher mean score (X= 5, 8750) than intermediate group (X= 7.2500).

• Differences between mean scores are statistically significant F (2-117) =32, 798 p<, 001. This significance points out that students’ achievement on + definite / - specific contexts change according to proficiency level.

• The data analysis clearly shows us that the upper-intermediate level is, again, the most successful group in assigning the correct article in the +definite/-specific context. In the same vein with the above mentioned section, however, it is not the intermediate group which follows that upper-intermediates but the elementary group. This finding tells us that elementary level students are more successful in article use than intermediate level students in definite / - specific context.

• The results of the descriptive statistics support ANOVA results.

• As foreseen, in the +definite/-specific context intermediate level students substitute the target item with the indefinite article a and a overuse is observed with the rate of 30, 3%.

• Elementary level learners are more accurate in their target article choice when compared to intermediates. Their rate of article substitution falls behind the intermediate group with the rate 16, 0%.

• The predictions were that upper-intermediate students will be the most successful group because proficiency will help learners in time. As expected it is obvious in the results that upper-intermediate students have very slight misuse; their substitution rate is 0,8 %. Overall results are illustrated in the table 10.

• Considering the article omission rates, elementary and intermediate level learners tend to omit articles at the rate 11, 5 % and 11, 0 % respectively.

• When compared to the omission rates in the -definite / + specific context tendency to omit articles is higher in + definite / - specific context.

• This suggests that Turkish learners tend to omit articles more in definite contexts rather than indefinite contexts. That is to say, article omission for Turkish learners is definiteness-sensitive rather than specificity. Here is the +definite/-specific item which most of the learners omit articles:

A phone conversation• Susan: Hi, Mrs. Shepherd. Can I talk to Alice?• Mrs. Shepherd: Sorry Susan, but Alice is out. She went to

____ (Ø / a/ an / the) school library to work on her project.

4. - Definite/-Specific ContextANOVA results of –definite / -specific contexts are summarized in table 11.

• The mean score of the upper intermediate level group (X= 9, 6750) is higher than the elementary and intermediate groups. (X=8, 8750 and X= 8, 3000 respectively).

• Differences between mean scores are statistically significant F (2-117) =18,986 p<, 001. It means that students’ achievement on + definite / - specific contexts change according to the proficiency level.

• Results of the descriptive statistics for indefinite non-specific contexts tell us that all groups perform better than other contexts (88, 8 %= elementary, 83, 0 % intermediate, 96, 8 % upper-intermediate).

• Elementary level learners supply the target article at the rate 88, 8 % which is quite a satisfactory percentage for this level. Article substitution is observed only at the rate of 4, 5 %. Their article omission rate is a little higher than article substitution rate but the percentage (6, 8 %) is too low to consider it as important. Table 12 illustrates the results for the –definite/-specific contexts.

Table 12: Article Choice Percentages in -Definite/-Specific Contexts

• As for the intermediate level learners the results reveal the fact that with the 83,0%accuracy rate, they fall behind the elementary group whose accuracy rate is 88,8 %. Interms of article omission intermediate group is not remarkably better. They tend toomit articles at the rate of 6, 5 % which is quite close to the elementary group; 6, 8 %.Example illustrates the definite/-specific item in which L1 Turkish learners of Englishmostly omit the article. Mother and father are talking in the kitchen just before the dinner Mother: Jane will not be with us tonight, honey. Father: Why not? Mother: She told me that she is going to wait for ____ (Ø / a/ an / the) client.• More important than the article omission rate of the intermediate group thesubstitution rate is quite high for this context. They tend to use the instead of a at therate 10, 5%.• The upper intermediate group is the most successful group. They are able to assignthe target article at a considerably high rate; 96, 8 %. This percentage shows us thatthe upper intermediate group shows native like performance in article selection in theindefinite nonspecific context. With respect to substitution and omission errors theupper-intermediate group’s error rates are remarkably low. They tend to omit articlesat the rate 4, 9 % and use the instead of a at the rate 1, 8.

Study by Dağdeviren:In order to analyze the use of English articles by the speakers of Turkish, the frequencydistribution of the articles supplied by the participants was given.

• As seen in Table (1), the participants of the study provided the at a rate of 73.0% in Type 1sentences, the at a rate of 70.0% in Type 2 sentences, a/an at a rate of 69.33% in Type 3sentences, a/an at a rate of 74.66 % in Type 4 sentences, a/an at a rate of 68.66 % in Type 5sentences, a/an at a rate of 74.33 % in Type 6 sentences.• A significant difference between high-proficiency and low-proficiency groups was found in alltypes of sentences. Sentence types which are [+specific] are Type 1 [+definite, -partitive,+specific], Type 3 [-definite, -partitive, +specific] and Type 5 [-definite, +partitive, +specific]sentences in this study. Apart from Type 1 sentences which evidently receive the as the NP isdefinite, Type 3 and Type 5 sentences were taken into consideration. In Type 3 sentences, theparticipants of the study provided the at a rate of 10.0% (30 uses out of 300) and a/an at a rateof 69.33% (208 uses out of 300). The words -apart from the articles- provided by theparticipantsconstituted 7.33% (22 uses out of 300) of the responses. The rate of missing slots in the test was10.0% (30 uses out of 300). Similarly, the participants of the study supplied the target articlea/an at a rate of 68.66% (206/300) in Type 5 sentences which include partitive and specific NPs.• The participants substituted the at a rate of 14.66 % (44/300), zero article at a rate of 3.0%(9/300) and other words at a rate of 4.33% (13/300). In this study the sentence types whichinclude [-specific] NPs are Type 2 [+definite, -partitive,-specific], Type 4 [-definite, -partitive, specific] and Type 6 [-definite, +partitive, -specific] sentences. Given that Type 4 and Type 6sentences naturally receive a/an as they are indefinite, Type 2 sentences were taken intoconsideration. • According to the results of the study, participants’ use of article “the” for Type 2 sentencesamounted to 70% (210 uses out of 300 contexts). The participants used the article a/an at a rateof 5.33% (16 uses out of 300 uses) and zero article at a rate of 7% (21 uses out of 300 uses). Theparticipants produced answers which were coded as “other” at a rate of 7.33% (22 uses out of300 contexts) and left sentences unanswered at a rate of 10.33% (31 slots out of 300) contexts.

DISCUSSION AND CONCLUSION

By Atay:The study yielded quite important findings related to

the acquisition of English Article System by Turkish learners.

• First of all, in the study article choices of learners from three different proficiency levels were examined and effects of definiteness and specificity to their article choice were investigated.

1. Overuses and Fluctuation• First research question was about the systematic errors of Turkish learners on the course of

English article system acquisition. As an answer for this question the most important finding in the study is that L1 Turkish learners overused certain articles in certain contexts.

• In +definite/-specific and –definite/+specific contexts especially intermediate level learners overused the indefinite article a and definite article the respectively.

• Taking related literature and theoretical knowledge as the background it is argued that like other L2 learners of English from article-less languages, L1 Turkish learners of English associate the definite article with specificity instead of definiteness.

• The reason of this association is that specificity distinctions are more basic than definiteness distinctions in the sentence. Moreover, according to Kim and Lakshmanan (2009), the selection of specificity setting for the definite article may also be triggered by the input (92). It means that definiteness is less transparent in the input; it is hard to infer the meaning of definiteness and definites are more frequently specific in the input. As a result of the frequency bias in the input, L2 learners of English associate the definite article with specificity. When the context is specific, they perceive it as definite. This misinterpretation causes overuses in the contexts where definiteness and specificity have contrastive values (when one of them is (+) the other is (-)). In +definite/-specific contexts learners tend to use the indefinite article because the specificity feature has (-) value and in –definite/+specific contexts they use the definite article as the context has (+) specific value. Some of the time they assign articles on the basis of specificity but some of the time definiteness is the setting for the article choice. This fluctuation lasts until the input leads them to set the right parameter for the article choice in the target language. However, setting the right parameter takes time because input triggers are discourse based and discourse and pragmatics are acquired quite late and through experience.

• This factor makes the article system acquisition process quite challenging for learners. However, once the learners have control over the discourse and reach the input triggers, they start to master the article system. In her third research question was about the developmental features of the English article system acquisition process of Turkish learners and I asked whether the proficiency have an effect on the process or not. Obtained results clearly answered these questions. In my data, the highest fluctuation is observed in –definite/+specific contexts and the highest fluctuation percentage belongs to the intermediate group. This finding is contrary to my predictions. I expected that elementary level learners would be more confused in these contexts with contrasting values and be less accurate in assigning the target article. This unexpected result finds support from literature. The possible reason of this unexpected accuracy might be that at this stage of acquisition learners start to become aware of the syntactic properties of definiteness and indefiniteness in English, so they are confused and fluctuate between definiteness and specificity. Article choice of lower proficiency level learners is influenced by the rules taught by their teachers and written in their text books. These rules are stored and used without clear understanding of definiteness or specificity. As there is rules and formulas in their minds, there is no confusion, or let’s say no fluctuation. When they become intermediate, they recognize their errors. This effect of recognition leads learners to confusion in their article choice. They start to assign articles with their newly acquired awareness and it brings learners to adopt a number of temporary, ad hoc hypotheses for choosing articles. As they improve in terms of their language proficiency reliance on rule-based hypotheses decrease and learners realize that they should also consider speaker-hearer knowledge and the context in which the given article is used. It means that in time they set the parameter for the article choice appropriately. The results of this study largely support this assertion because upper-intermediate level learners’ fluctuation rates are quite minor both in –definite/+specific and +definite/-specific contexts in my data. However, they still do not show native-like performance in these contexts. This finding takes us to the fact that the acquisition of the article system in English is a difficult process which requires quite a long time. It takes time because in order to choose the correct article, learners should evaluate the discourse and find out the input triggers which will lead them to the correct choice. However, input triggers are discourse based and discourse and pragmatics are acquired quite late and through experience. So eliciting cues from the input to choose the right setting is not possible due to the current level of L2 learners. As learners have not fully acquired or mastered the acquisition of discourse and pragmatics yet, they manifest fluctuation.

2. Omission Errors• As well as overuses and fluctuation, article omission errors are also observed in the gathered data. All groups tend to omit articles

in [+definite/+specific] context at considerable rates. However, the biggest rate of omission belongs to the elementary group. Omissions are also observed in +definite/-specific and -definite/-specific contexts. Results related to omissions are in line with previous researches.

• When these omissions are observed item by item, some consequences and results can be deduced. First of all, it is important to state that omissions are mostly observed in definite contexts whether it is specific or not. There are two items from the +definite/+specific context which most of the learners show a tendency to choose no article option.

• For the first example which is the 17the item in the task 7 largely elementary and intermediate level learners chose no article option to a large extend. The reason of the omission might be that elementary and intermediate level learners could not notice the reduced relative clause in the sentence which requires the target article. I argue that this is a strong reason because upper-intermediates did quite well in this item. Upper-intermediates’ doing much less substitution and omission errors stand for the fact that proficiency level positively affect article choice on the way of article system acquisition. It means that proficiency help fluctuation and omission errors and participants do better in time. This result goes hand in hand with the expectations and is the answer of my third research question. For the next +definite/+specific item, which is the 18th one in the task, the reason of omission finds its support from literature. Here, the definite article is on the scene as an example for the cultural use of the. In Liu and Gleason (2002) the authors argued that the definite article in English have four nongeneric uses and these four uses pose different levels of difficulty for ESL learners and are acquired in different times. The results indicate that the last acquired use of the is cultural use and our 18th item is a quintessence example of this late acquired cultural use. Because of this, learners from all levels face with difficulty while trying to assign an article for this item and, at last, chose no article option. Similarly, the 36th item in the task is yielded a high rate of article omission. In this item the possible reason might be that the NP (library) just before the target article is modified by the word school so a compound word is formed; school library. Trenkic (2009) indicates her study that L2 English speakers show a tendency to omit articles in premodified contexts (132). She argues that L2 learners take definite article use to be “based on a pragmatic principle akin to Grice’s maxim of quantity; use a referential form that is sufficiently informative for your purpose but not more informative than necessary” (128).), Robertson (2000) also states that if the use of definite article is pragmatically redundant, it is more likely to be omitted. The situation in this item might be an example for these explanations. L1 Turkish learners of English might have thought that the NP, library, had already been modified by the modifier school and there is no need for the to be more informative. Apart from the findings related to omission and fluctuation, the study has revealed the fact that in the - definite/- specific context proved to be the most successful context for all learners whatever their proficiency level is. This result has been reported in earlier studies (Ekiert, 2007; Thu, 2005). However, there are some other studies which found that learners from article-less languages show better results in definite article marking when compared to indefinite article marking (P.G. Mayo, 2009; Bergeron and Matoba, 2007; Zdorenko and Paradis, 2007a) . The reason why L1 Turkish learners are better in the indefinite nonspecific context is that in Turkish, there is an indefinite article; bir. As a result, Turkish learners have transferred their already existing parameter values to their interlanguage and this transferred knowledge has helped them to assign the target article successfully.

• Summary of the Discussion• To sum up, the results of the study largely supported my expectations. The

intermediate level fluctuated between definiteness and specificity as foreseen because they associate definite article with specificity. As a result of this, overuses were observed in the data in –definite/+specific and +definite/-specific contexts at considerable rates. Contrary to my predictions; however, elementary level learners fell behind the intermediates in these problematic contexts and they did not fluctuate between definiteness and specificity as much as their intermediate counterparts. This does not necessarily display the elementary level learners’ mastery over article system. The reason of this temporary success is that at this level students are not confused as they have not reached the awareness of the underlying target language parameters ; they just assigned articles according to the rules they are taught. So they did not show fluctuation at considerable rates but still made errors. One of the most striking errors was the omission error. Elementary level learners tended to omit articles to a high extent in the +definite/+specific context. In definite contexts, other learners also showed omission errors. Finally, L1 Turkish learners of English were better in assigning the indefinite article when compared to the definite article. The reason of this mastery is that Turkish has an indefinite article and there is a way for learners to transfer their already existing parameter values related to the indefinite article to their interlanguage. So they could show mastery over this context.

• Conclusion• In the overall sense the aim of my study is to contribute to the literature of English Article System acquisition

by Turkish learners and widen the spectrum of the researches. At a specific level the purpose is to examine the second language acquisition of the English article system by Turkish learners and investigate the role of definiteness and specificity during the acquisition process. It is also aimed to find out the role of the proficiency level on the course of article system acquisition. The data for the thesis was collected from the preparatory class students at M.E.T.U. Students were from 3 different learner groups from 3 different proficiency levels were tested. Before the data collection, the elicitation task was piloted with a group of preparatory class students at TOBB University of Economics and Technology. They were from different proficiency levels; 5 from elementary, 5 from intermediate and 5 from upper-intermediate. In addition to the test group, there was also a control group for the task. They were 5 native speakers of English. The soft version of the task was sent to them as an e-mail. . The elicitation task was sent back via e-mail, again. As the data collection instrument, a forced choice elicitation task was prepared. This task consisted of 40 short dialogues which depended on four different contexts; +definite/+specific, +definite/-specific, -definite/+specific, -definite/-specific. Collected data was analyzed by means of SPSS 17 package program using the descriptive analysis and ANOVA techniques. The results of the study were mostly in line with the predictions. It was expected that especially elementary and intermediate level learners would fluctuate between definiteness and specificity in +definite/-specific, -definite/+specific contexts because they would associate the definite article with specificity instead of definiteness. This fluctuation would last until they became proficient enough and input leaded them to assign the right article. As expected, intermediate level learners fluctuated between definiteness and specificity and overused “a” and “the” in [+definite/-specific], [-definite/+specific] contexts. However, contrary to my predictions, elementary level learners performed better than intermediate counterparts. From upper-intermediates’ side the results were quite satisfactory; they did well nearly in all contexts. It proved that proficiency level helps article system acquisition and learners can perform better in time on the course of English Article System acquisition. As well as fluctuation, omission errors were also observed. Especially elementary level learners tended to omit articles in [+definite/+specific] context. Articles omission rates mostly observed in definite contexts. In general all learners were better in [-definite/-specific] context and worse in [+definite/-specific] and [+definite/+specific] contexts. The reason might be that Turkish has an indefinite article and learners might have transferred this already existing parameter to their interlanguage so performed better.

By Zdorenko and Paradis:In this study the sentence types which include [-specific] NPs are Type 2 [+definite, -partitive,-specific], Type 4 [-definite, -partitive, -specific] and Type 6 [-definite, +partitive, -specific]sentences. Given that Type 4 and Type 6 sentences naturally receive a/an as they areindefinite, Type 2 sentences were taken into consideration. According to the results of thestudy, participants’ use of article the for Type 2 sentences amounted to 70% (210 uses out of300 contexts). The participants used the article a/an at a rate of 5.33% (16 uses out of 300uses) and zero article at a rate of 7% (21 uses out of 300 uses). The participants producedanswers which were coded as “other” at a rate of 7.33% (22 uses out of 300 contexts) and leftsentences unanswered at a rate of 10.33% (31 slots out of 300) contexts. In this study the sentence types which include [-

specific] NPs are Type 2 [+definite, -partitive,-specific], Type 4 [-definite, -partitive, -specific] and Type 6 [-definite, +partitive, -specific]sentences. Given that Type 4 and Type 6 sentences naturally receive a/an as they areindefinite, Type 2 sentences were taken into consideration. According to the results of thestudy, participants’ use of article the for Type 2 sentences amounted to 70% (210 uses out of300 contexts). The participants used the article a/an at a rate of 5.33% (16 uses out of 300uses) and zero article at a rate of 7% (21 uses out of 300 uses). The participants producedanswers which were coded as “other” at a rate of 7.33% (22 uses out of 300 contexts) and leftsentences unanswered at a rate of 10.33% (31 slots out of 300) contexts. In this study the sentence types which include [-

specific] NPs are Type 2 [+definite, -partitive,-specific], Type 4 [-definite, -partitive, -specific] and Type 6 [-definite, +partitive, -specific]sentences. Given that Type 4 and Type 6 sentences naturally receive a/an as they areindefinite, Type 2 sentences were taken into consideration. According to the results of thestudy, participants’ use of article the for Type 2 sentences amounted to 70% (210 uses out of300 contexts). The participants used the article a/an at a rate of 5.33% (16 uses out of 300uses) and zero article at a rate of 7% (21 uses out of 300 uses). The participants producedanswers which were coded as “other” at a rate of 7.33% (22 uses out of 300 contexts) and leftsentences unanswered at a rate of 10.33% (31 slots out of 300) contexts.

By Dağdeviren:The results of the current study are partly in line with the previous studies. Firstly, the articlechoice of the participants of this study is similar to the ones observed in the studies whichfocused on the article choice of L1-Turkish speakers. Yılmaz (2006) concluded that semanticdistinctions among different contexts could be made by Turkish learners of English. She alsosuggested that L1-Turkish learners could acquire the English article system despite theabsence of an article system in their native language. Similarly, Önen (2007:102) found thatthe accuracy of article use varied in respect to proficiency levels and task types. The currentstudy supports the evidence that semantic distinctions are discriminable by the speakers ofTurkish.Secondly, on the basis of article choice in L2-English by the speakers of different levels ofproficiency, the present study showed notable similarities to previous studies. According tothese studies, there is a significant difference in article choice between proficiency levels(Master, 1997; Thomas, 1989; Yoon, 1993; Murphy, 1997; Mizuno, 1999; Robertson, 2000;Lu, 2001; Goto-Butler, 2002; Jarvis, 2002; Liu & Gleason, 2002; Ekiert, 2004; amongothers).Thirdly, according to the specificity distinction proposed by Ko et al. (2008), the L2-Englishlearners were expected to overuse the in [+specific] contexts and a in [-specific] contexts. In their study they

supported the hypothesis with the evidence by L1-Serbo-Croatian and L1-Korean learners of English. However; in the current study the participants’ overuse of the inType 3 [-definite, -partitive, +specific] and Type 5 [-definite, +partitive, +specific] sentencesdid not outnumber the rate of accuracy. Similarly, Ko et al.’s (2008) hypothesis predicting theoveruse of a in [-specific] contexts was not supported by the results of this study. Theaccuracy rate in Type 2 [+definite, -partitive, -specific] sentences was found higher than theoveruse of a.

As the current study aims to investigate the role of proficiency on article choice, the articleuses of the two proficiency levels were also analyzed on the basis of Ko et. al.’s (2008) hypothesis.

According to the results of the current study, neither the low- nor the high proficiency level participants’ overuse of the and a were found higher than the accuracy rates in Type 3, Type 5 sentences and Type 2 sentences, respectively. However, when looked at the performance of proficiency levels, it is seen that there is a significant difference between the low- and the high-proficiency level participants on the basis of providing a/an to Type 3 [- definite, -partitive, +specific] and Type 5 [-definite, +partitive, +specific] sentences and the to Type 2 [+definite, -partitive, -specific] sentences. The results showed that the highproficiency level participants were more successful than the low-proficiency participants at providing the correct articles to [+definite] and [-definite] contexts.

Fourthly, Ko et al.’s (2008: 123-124) prediction on the maximal use of the with indefinites in [+partitive, +specific] contexts was also analyzed. In the current study, Type 5 sentences

define [-definite, +partitive, +specific] context in which the maximal overuse of the wasexpected. However, according to the results of the study participants used a/an morefrequently than of the in [-definite, +partitive, +specific] context. Additionally, the findings onthe perfomance of the proficiency levels revealed that high-proficiency level participantsperformed better at providing a/an to Type 5 sentences.This study was conducted to investigate the article choice in L1-Turkish learners’ L2-English.It can be said that the limitations of this study may have affected the interpretation of theresults. Further research which adopts test instruments of both spoken and writtenperformance may produce different results concerning article use in L2-English of L1-Turkishspeakers. This might especially be true if the test instruments include more test items per task,and if a greater number of participants are included in the study.