Abstract: This study is part of a larger study on feedback and it was aimed at discovering the most prevalent errors Iranian EFL learners have in their writing task two for IELTS. To this end, 60 writing responses for each of different proficiency levels were studied to find out about the error gravity. The participants were at three different proficiency levels according to the IELTS bands namely, band 5, 6 and 7. The results suggested that syntactic errors were more common than lexical and mechanical ones among all the three different proficiency levels. Analysis of Chi-square was used to find significant difference among the groups. However, the differences were not significant. Finally, the Some pedagogical implications are finally drawn from this study.
INTRODUCTION
For quite some time, a lot of research has been carried out on corrective feedback (however, it is called) and this is not only devoted to a limited period of time (Atai and Alipour, 2012; Bitchener et al., 2005; Evans et al., 2010; Hanaoka and Izumi, 2012; Hyland, 2010). Teachers present feedback since they wish to see their students advance in their written production. However, they may not be fully aware of how much feedback they should present and on what type or whether the feedback they employ is effective in the long run (Truscott, 2004, 2007, 2009; Truscott and Hsu, 2008) although some cast doubt on the view (Hyland, 1998, 2000).
Robb et al. (1986) did not approve of spending time on correcting mechanical errors. (Cohen, 1987; Ferris, 1995) believed that students require that their writing be corrected formally. In other words, they wanted their writing to be corrected in terms of grammar and mechanics. In their research (Ferris, 2006; Zamel, 1985) discovered that teachers also focused more on local errors than global errors. Although global errors affect meaning and might cause a communication breakdown, these studies do not suggest whether feedback on the global errors could also be as fruitful as feedback on local ones. Furthermore, they did not mention which error types are more prevalent.
Knowledge of the most frequently occurring errors could give us good justifications as to what errors should be treated and need immediate and focused attention in certain corrective feedback types. Different research has been carried out in the realm of error gravity (Dorri et al., 2008; Robb et al., 1986; Semke, 1984; Vann et al., 1984; Sheorey and Ward, 1984) and most mentioned that mechanical errors are the most prevalent errors. Most of the researches done in the error gravity area was on the formal part of the language and none focused on the proficiency level of the learners and how the errors could vary along the proficiency continuum.
The purpose of this study was therefore to find out about the formal errors learners of English have across three different proficiency, band 5, 6 and 7, on an IELTS score-based scale. The study was limited to these band strategies only and did not include other proficiency levels.
Based on the gap in research and the purpose of this research, the following research question was proposed.
What are the most prevalent errors in different-ability groups?
METHODOLOGY
Participants, data collection and procedures: The students were asked to write a composition. They were checked on their writing ability and placed into one of the three categories, modest, competent or good user, i.e., band 5, 6 and 7, respectively based on the band descriptors of IELTS as proposed by the University of Cambridge. The students’ writing responses served two purposes: (a) It helped find out what errors were more prevalent and (b) What the participants’ writing ability proficiency levels were. The written pieces, therefore, had to be corrected.
The students’ writings were rated three times by two raters. The first rater rated them twice with a time interval of two weeks. As displayed in Table 1, the intra-rater reliability for the first rater’s two ratings is 0.78 (p<0.05). Based on these results it can be concluded that the two ratings of the first rater enjoy statistically significant intra-rater reliability.
The inter-rater reliability between the ratings of the second rater with the mean ratings of the first one is 0.94 (p<0.05). Based on these results, it can be concluded that the two ratings enjoy statistically significant inter-rater reliability (Table 2).
Then, the compositions were corrected once again by the researcher to find out about the error gravity. Initially, 30 compositions in each group were corrected and the inaccuracies were tallied under 31 categories. Later, some new compositions were also corrected to have 60 papers in each category. The linguistic inaccuracies which were tallied were later boiled down into three major categories: Mechanical, lexical and grammatical inaccuracies. These categories appear in data analysis section where the research question is analyzed.
DATA ANALYSIS
To find out what the most prevalent errors are, the researcher randomly chose 60 writing responses within each category. The compositions were tallied when a problem was detected in one of the categories (Table 3).
| Table 1: | Intra-rater reliability index |
| Table 2: | Inter-rater reliability index |
| Table 3: | Frequency table for errors in different ability groups |
The list was not pre-determined and it evolved as more and more errors were detected. Once almost all categories evolved, the compositions were tallied from the beginning. Then the categories were divided into three major categories, named mechanical, lexical and grammatical. All these categories along the different band scores appear in Table 3.
To answer this question, the researcher employed an analysis of chi-square to compare the types of errors the learners in the three ability groups made when writing in English. As displayed in Table 4, the fifth and seventh band scores have made fewer mechanical errors than the sixth band scores group as shown through the negative Std. residuals of -1.5 (5th band score) and -1.5 (7th band score). The sixth band score with a Std. residual of 3 made more mechanical errors which is also statistically significant because the Std. residual of 3 is higher than 1.96.
Although none were significantly different, the seventh band score (Std. residual = 1.1) made more lexical errors than the fifth band score whose Std. residual is -1.4. The sixth band score group’s lexical errors are negligible (Std. residual = 0.6).
| Fig. 1: | Percentage categories of writing errors by band scores |
| Table 4: | Descriptive statistics for writing errors by band scores |
| Table 5: | Chi-square analysis |
| 0 cells (0.0%) have expected count less than 5. The minimum expected count is 336.73 | |
The sixth band score group made significantly fewer grammatical mistakes (Std. residual = -2.2) than the fifth band score (Std. residual = 1.8). The number of grammatical inaccuracies for the seventh band score group is negligible (Std. Residual = 0.2).
Based on these results, it can be concluded that the fifth and seventh band score groups make fewer mechanical errors while they make more lexical errors. The fifth band score group has more grammatical inaccuracies.
The chi-square observed value of 24.87 (p = 0.000<0.05) indicates that the differences observed in Table 3 were statistically significant. Thus the null-hypothesis is rejected (Table 5).
Figure 1 and 2 display the percentages and Std. residuals as appeared in Table 3.
| Fig. 2: | Standard residuals categories of writing errors by band scores |
The Cronbach’s alpha reliability indices are 0.53 for both groups, 0.49 for teachers and 0.53 for students.
CONCLUSION
Regarding the research question, the findings could remove some of the doubts some researchers have as to what grammatical error to correct in that they pinpoint some of the most prevalent errors across different language ability learners.
Instantly, when one hears more errors, he or she might think that this area needs more intensive care and even worse some might think more errors might mean less proficiency. The findings of this research suggest that this is not true, though. Mechanical errors are more prevalent in bands five and six compared to band seven students. This observation might suggest that band six students are not much better in their use of punctuation marks, capitalization and spelling than band five candidates. However, it should be borne in mind that this interpretation is too superficial because as we progress towards higher levels of proficiency, we resort to more advanced structures and more advanced vocabulary items. Consequently, learners are more likely to make mistakes because of the complexity of the structures and vocabulary items used. For instance, using transitional adverb appears more in band six than in band five. Most learners who use these cohesive ties, however, are negligent in terms of punctuation. This in turn could lead to a large proportion of errors in this category. Nonetheless, band 5 learners tend to use simple sentences more than complex ones. Hence, they make fewer punctuation mistakes.
For the same reasons mentioned above and percentage-wise, there are more mistakes in the lexical section of band 7 learners than band 5 and 6 ones (7 = 22.8%>6 = 22%>5 = 20.2%). This might mean that band 7 students take more risks than band 6 students. This could also mean that having more lexically wrong counts does not necessarily mean a less effective performance. Of course, this claim should also be made with caution because the final score is a collective score and encompasses all the different band descriptors, namely, task response, cohesion and coherence, grammatical range and accuracy and lexical resources. This could mean we can have a higher score in one and a lower one in another or the others.
The findings of this study could also open a new line in research on error. Some studies could be focusing on the different types of errors except for the formal ones. Another study could be on other proficiency levels both at lower and higher levels in writing and speaking.