Common Error Types of Iranian Learners of English

This paper aimed at obtaining a clear understanding of Iranian EFL learners’ L2 writing error types. To develop such an understanding, a research question was formulated to see whether there is any significant difference between the participants’ language proficiency level and their error types in writing. To this end, a sample version of the structure and reading sections of a paper-based TOFEL test, and a Test of Written English (TWE) were administered to 103 university students majoring in English. The statistical analyses revealed that a) there were statistically significant differences among proficiency groups on overall error types they made in their compositions, and b) frequency of occurrence of error types in each group was different.


Introduction
The study of second language learners' errors has emerged as one the focal concerns of second language studies since 1970s.Since then, many researchers have devoted their time to explore the nature and the cause of the errors second language learners make in their production of the second language.Many studies aimed at identifying the frequency of error types among either EFL or ESL learners.Among these studies, some have tried to observe errors which are mainly caused by the L1 influence (Chen, 2006;Kao, 1999;Lin, 2002;Tseng, 1980;Ying, 1987), other studies have dealt with detecting those error types which are general among all EFL and ESL learners (Burt & Kiparsky, 1972;Fitikides, 2002;Richards, 1974).However, most of these studies have not considered error types which are peculiar to different proficiency levels.And this insufficiency can be traced in the studies conducted on the error types of Iranian-speaking English learners' error types (for example, Rooshanzamir, 1995;Shakeri, 1993).In fact, the present study seeks to observe the performance of Iranian speakers of English regarding the error types they make in their writing.For that reason, the study aims to answer the following research question: Are there any significant differences in the type of writing errors of Iranian EFL learners at different proficiency levels?

Significance of Studying Learner Errors
One strategy to address the problem of how much the standard of English in non-native production has declined is to first look at the type of errors that learners make.We can cite evidence that the analysis of errors provides researchers with valuable information on learner language, and helps teachers improve instruction.(Burt & Kiparsky, 1972;Corder, 1967;Dulay & Burt, 1972;Ellis, 1983;Richards, 1974;Selinker, 1972;Silveira, 1999).As Ellis (2003) notes there are good reasons to study learners' errors.First, they tell us why learners make errors and provide us with useful information on learner language.Secondly, the types of errors learners make can help teachers.Thirdly, "paradoxically, it is possible that making errors may actually help learners to learn when they self-correct the errors they make" (p.15).
Studying any kind of deviation, from a selected norm of language performance, regardless of its cause(s), paves the way for remedial actions in the process of error correction (Corder, 1981).It also helps teachers and learners in finding the areas of weakness in which learners have difficulty in producing the second/foreign language (Dulay, Burt, & Krashen, 1982).Brown (2000) believes that the occurrence of errors in L2 learners' production is inevitable.He adds that if learners neither make errors nor receive any feedback on their errors, their acquisition process will be impeded.These errors, deemed meaningful and systematic, are of outmost importance to researchers and teachers of L2 writing (Benson, 1980;Chen, 2006;Li & Chan, 1998;Ferris & Hedgcock, 2005;Leki, 1992, Frodesen, 1991;Hirokawa, 1986;Kao, 1999;Lin, 2002;Neumann, 1977;Tseng, 1980;Silva, 1993;Ying, 1987;Yu & Atkinson, 1988).Matsuda, Canagarajah, Harklau, Hyland, and Warschauer (2003) draw attention to the significance of the study of student text in teaching effectively.They, referring to some studies (Biber & Reppen, 1998;Granager, 1998;Milton, 1999), believe that this helps "teachers target students' more frequent and intractable errors" (p.166).

Error Taxonomies
Following Dulay et al. (1982), four criteria for descriptive classification of errors can be established: linguistic taxonomy, surface strategy taxonomy, comparative analysis taxonomy, and communicative effect taxonomy.A linguistic taxonomy includes categories which are based on descriptive grammar of the target language.Such a grammar emphasizes on categories concerning "basic sentence structure, the verb phrase, verb complementation, the noun phrase, prepositional phrase, adjuncts, coordinate and subordinate constructions and sentence connection" (Ellis & Barkhuizen, 2005, p. 60).Surface strategy taxonomy is based on the ways surface structures are changed.Learners may omit necessary items or add unnecessary ones; they may misform items or misorder them.To develop a comparative taxonomy, a researcher should classify the error types based on "comparisons between the structure of L2 errors and certain other types of constructions" (Dulay et al., 1982, p. 163).Dulay et al. (1982) presents four categories in comparative taxonomy: developmental, interlingual, ambiguous, and other errors.The first two error categories are major and the second two ones are drawn from the first two error categories.Developmental errors are errors comparable to those made by children leaning the target language as their first language.Finally, communicative effect taxonomy focuses on the effect of errors on the listener or reader.Based on such a category, there exist "global" errors which significantly hinder the flow of the communication, and "local" errors which do not (Burt & Kiparsky, 1972).
However, it should be noted that researchers have developed different systems of error categories.For example, Weltig (2004) adapted a system for error classification of his study from Sachs (2003), which was adapted in turn from Polio (1997) and Kroll (1990).He tried to develop a linguistic category, and focused on more linguistic and lexical errors (verb tense, verb voice, verb formation, preposition, lexical choices).Albeit classified in terms of linguistic units and dealt with sentence-level elements of discourse, error types were ranked from those errors which hinder communication to those which did not.Otoshi (2005) developed a linguistic taxonomy of grammatical errors, focusing on five major error categories: verb errors, noun ending errors, article errors, wrong word, and sentence structures.He claims that these five categories have been considered as major errors in much of the second language writing literature.Chen (2006) developed a taxonomy based on a structured-linguistic error taxonomy.His error classification was devoted to 15 major categories including subcategories for each.Major error types compromising this taxonomy included: errors in the use of nouns, articles, pronouns (incorrect case forms, missing possessives), verbs (tense, subject-verb agreement, auxiliary, verb omitted), prepositions (prepositions omitted, wrong prepositions, unnecessary prepositions), and conjunction (coordination, subordination, missing).Kao (1999) studied 169 compositions written by Taiwanese students to find out their English learning deficiencies.Of total of 928 errors found grammatical errors displayed the greatest frequency (66 %), followed by semantic errors with 18%, and then lexical errors with 16% (as cited in Chen 2006, p. 79).Neumann (1977) made an attempt to investigate common errors shared between intermediate and beginning learners by identifying and analyzing errors made in compositions of 158 students.The most frequent errors of the intermediate students in Neumann's (1977) study, as reported in Hatch (1983, p. 103), were errors in the uses of noun modification, verb, preposition, lexical option, number agreement, noun, adverb of time/place, and negation.
A focal point, worth mentioning, is that most of these studies have barely paid due attention to different proficiency levels at which structural and/or lexical errors creep.The present study, thus, aimed at filling the gap.However, as to limitations of the study, factors such as L1 background knowledge, psychological sources of errors, participants' psychological and emotional status, which deserve independent research studies, were not taken into account.Moreover, gender showed itself a problem as far as statistical theories were concerned.In Iran, male English learners are few compared to female ones due to socio-economic reasons, making researchers unable to find equal number of students for their sample.

Participants
The participants who took part in the present study were 103 Iranian BA students majoring in English at various proficiency levels.They were studying at two universities in Iran -University of Tehran, and Islamic Azad University, Rasht Branch.The sample included 78 females (75.7%) and 25 males (24.3%).

Instrument
Two instruments were utilized to collect data for this study: (1) The structure and reading sections of a paper-based TOFEL test selected from the "TOEFL Test Preparation Kit, second edition" ( 2003), (2) A Test of Written English.The topics for the essay were selected from TWE section of the 'TOEFL Preparation Course ( 2001) by Longman Publication.

Procedures
The instruments were administered in a two-phase session.In phase one, participants should spend 90 minutes to answer to the Structure, Reading Comprehension sections of the TOEFL test.In phase two, started immediately after phase one, students were required to spend 60 minutes on witting an essay about one of the topics presented on their TWE sheets.The scoring scheme appeared in the TOEFL Preparation Kit (2003) was applied to scores the participants obtained by responding to structure and reading comprehension tests.
Once the proficiency scores were defined and ranked, a descriptive statistics was conducted in order to classify the participants into three groups (high, med, low).To do so, the mean (472.18) and the standard deviation (90.47) of the participants' proficiency scores were used as the criteria for grouping the participants (Table 1).Accordingly, those who scored higher than 'mean-plus-one half' standard deviation were assigned to high group (30 people-29.1%),those who stood within the 'mean-plus-one half' standard deviation and the 'mean-minus-one half' standard deviation range were assigned to mid group(36 people-35%), and those who scored below the 'mean-minus-one half' standard deviation were assigned to low group (37 people-35.9%).
The next step was an attempt to identify the error types on compositions.As for the compositions, a number of selected grammatical and lexical features were examined for specific frequency counts as well as correct or incorrect use.The basis for marking and categorizing errors was the Weltig's (2004) error classification.Furthermore, the criterion for conducting our error analysis was grammaticality.It means errors found were defined as "breach of the rule of the code" (Corder, 1971, as cited in Ellis & Barkhuizen, 2005, p. 56).However, issue of acceptability relating to researcher's sense in "making stylistic judgments" (Ellis & Barkhuizen, 2005) was not considered in determining errors.Finally, once the error types were thoroughly detected, a frequency count regarding the occurrence of error type in each text was performed.

Results
The initial step in data analysis was counting the errors in the writing of the participants.Total errors found were 4109 in number, of which 1939 (47.2%) were observed in low group, 1579 (38.4%) in middle group, and 591 (14.4%) in high group.The minimum number of each participant's total error frequency count equaled 16 in low group, 7 in middle group, and 0 in high group.According to Table 2, the maximum number of total error frequency for each participant was 112 in low group, 96 in middle group, and 55 in high group.Among the participants, only one student in high group made no errors, while another student in low group made 112 errors which showed the maximum total number of errors, regardless of their types, among participants.Furthermore, as shown in Table 3, a frequency analysis was used to identify the proportion and percentage of error types in three proficiency groups (low, middle, and high), and to explore the most frequent error types among groups.The first ten most frequent error types among the three proficiency groups were: errors in punctuation, lexical/phrase choice, spelling, article, verb formation, the use of plurals (singular for plural), preposition, verb tense/ aspect, clause structure (aberrant clause), and subject/verb agreement.
To answer the question addressed in the present study, data were analyzed through a Kruskal-Wallis test on overall error frequencies as well as a series of Kruskal-Wallis test on error frequencies within individual error categories.
The Kruskal-Wallis test on error frequencies within individual error categories revealed a significant change across levels in the mean ranks of groups on 22 error types.They include: aberrant clause, verb/object complement, sentence fragment, run-on sentence, word order, tense/aspect, voice, verb formation, subject-verb agreement, quantifier-noun agreement, ambiguous reference, lexical/phrase choice, idiom, word form, comparative & superlative adjectives, singular for plural, plural for singular, preposition, article, punctuation, spelling, and possessive (Table 4).Unlikely, no statistically significant difference was found among the performances of three proficiency groups on 22 other error types.As shown in Table 4, they include: subject formation, verb missing, dangling modifier, parallel structure, relative clause, extraneous words, missing word, wrong model, noun-pronoun agreement, epenthetic pronoun, wrong case, noun phrase morphology, genitive, native language equivalent, mixed construction, coordinating conjunction, conditional, question form, regularization, abbreviation, meaning duplication, and word duplication.
Further, the results of Kruskal-Wallis test on overall error frequencies (x 2 = 39.306,df= 2, P< .05)revealed that there are significant differences among proficiency groups on overall error types they made in their compositions (Figure 1 and 2).

Conclusion
One of the most immediate implications of the present study would be considered productive for L2 writing teachers, in particular those opting for direct feedback (Ferris & Roberts, 2001;Ferris, Cheyney, Komura, Roberts, & McKee, 2000;Komura, 1999;Rennie, 2000).They will find errors of different nature and types naturally occurring in compositions of students at varying proficiency levels in heterogeneous classes where groups may diversely struggle with verb tense, sentence structure, word order, word choice and so forth.However, it is immature to claim that errors found in this study should only be targeted since students, depending on their learning experience, make different errors in a special context.

Table 1 .
Sample analysis

Table 3 .
Hierarchical frequency list of each error type in all groups

Table 4 .
The Kruskal-Wallis test on error frequencies within individual error categories