V 4.6.2 Inferential Statistics Results
6.3 RQ1a: What are the effects of two different forms of language teaching (TBLT and TSLT) on the development of students’ L2 speaking performance over time?
6.3.2 Effects on Syntactic Complexity
Three measures were utilised in this study to measure syntactic complexity, namely mean length of AS-unit (SMA), ratio of subordination (SRS) and mean length of clauses (SMC). The statistical results showed that neither group made any improvement across all measures from pre- to post-test; in fact, the ST group had a significant decrease for the measure of SMC. There are a few possible explanations for these findings.
First, as mentioned earlier, both pre- and post speaking tests for both groups were truncated to 200 words (tokens). It could be hypothesized that cutting down the words to 200 may have led the values of these measures to decrease especially the measures of SMA (total number of tokens / AS-units) and SMC (total number of tokens / total number of clauses). Therefore, when the number of tokens for all speaking tests was reduced to 200 words, all students would have approximately the same outcomes in which the average scores for AS-units and the total number of clauses were 15 and 28, respectively. However, the process of truncation was considered necessary as it was done to harmonise all the global measures of speech production. It could be assumed that students’ syntactic complexity was poor due to the effect of the truncation process and not because of students’ incompetence. If the speech production was not reduced to 200 words, perhaps they could have made better progress. However, the truncation process was done to prevent other variables (e.g., lexical complexity, speech length etc.) from exerting influence on the dependent variables (Treffers-Daller et al., 2016).
183 Second, the complexity of the tasks affects syntactic complexity. The AS-unit is used to measure speech and it consists of “an independent clause or sub-clausal unit, together with any subordinate clause(s) associated with either” (Foster et al., 2000, p. 365). In other words, a complex sentence is described as having two or more clauses. In the qualitative analysis (see Section 4.7.1), however, the results showed that within a single utterance, subordination was produced occasionally and mostly contained one main clause only. Therefore, students used more simple sentences which only contained one independent clause and no subordinate clause which affected the measures of SMC and SRS (total number of clauses / AS-unit). It could be assumed that the speaking tests were not complex or challenging enough to elicit students’ syntactic knowledge. The findings of this study are not in line with the study conducted by Mochizuki and Ortega (2008), which found that students who engaged in guided planning produced significantly more relative clauses in their oral narrative tasks than those who engaged in unguided planning. The study suggests that greater demands on guided planning are associated with more complexity in terms of the mean number of relative clauses per T- unit. The tasks in the previous study were guided planning oral narrative tasks which involved students working in pairs. The guided planning involved a picture storytelling task with an additional handout which briefly explained relative clauses during the five-minute preparation time. Therefore, the pair-work oral tasks with guided planning may have prompted students to formulate more complex structures resulting in more syntactically complex language with the increases of relatve clauses. However, this is not the case for this present study in which every student was assessed individually and no guidance (e.g. handout) was given to them in both speaking tests, hence the speaking tests were less effective in inducing more complex syntactic structure.
184 6.3.3 Effects on Accuracy
Two measures of accuracy the percentage of error-free clauses (APE) and number of errors per 100 words (ANE) were employed to examine whether the implementation of TSLT and TBLT would affect speaking accuracy. The descriptive statistics showed that the EIST group outperformed the ST group in the post-test on both measures of accuracy. However, although the EIST group showed improvement, the pre-post change was marginal and indeed a 2x2 ANCOVA revealed that neither group made significant progress. Thus, neither TSLT nor TBLT approaches resulted in an increase in error-free clauses. In fact, students produced more errors per 100 words. A thorough analysis was done to compare why such an effect occurred in both groups following the intervention. Several assumptions derived from these results. First, students’ primary attention was drawn to content and meaning especially for the ST group. This could be seen in the speaking outcome of the ST group which showed that the type of language used was heavily weighted towards formulaic phrases and discourse markers (see Section 4.7.1). Therefore, it could be considered that using TBLT which only employed speaking tasks without the explicit instruction led to FonM. This is in line with the study by S. Li et al. (2016) that compared TBLT and TSLT. Their findings revealed that TBLT led to limited opportunities for learning the target language and thus discouraged deeper cognitive processing of the grammatical forms and structures to take place (S. Li et al., 2016). In the present study, it could be assumed that the speaking tasks provided insufficient experience of the target structure for the ST group which resulted in little acquisition, although opportunity for learning was given. Therefore, this could lend support to the argument that noticing is a necessary condition in order to learn language form (Schmidt, 1990, 2010). Meanwhile, the non-significant statistical results for the EIST group in the present study might be caused by a trade-off effect between achieving the target structure and using the vocabulary exemplars. It could be assumed that students were ‘testing out’ a wider range of vocabulary at the post-test,
185 and thus failed to achieve the target structure, resulting in overall accuracy to decrease. Further discussion will be given in Section 6.3.5 regarding these effects on both groups in relation to the Noticing and Limited Attention Capacity Hypothesis.
Second, the feedback was given to the students in both groups. Although the main focus of the study was mainly on the two different forms of language teaching (TSLT and TBLT), feedback was given as it is a part of the course requirement to evaluate and give students feedback on their performance. However, feedback was not done rigorously to avoid overshadowing the effects of the different forms of language teaching (TSLT and TBLT). The nature of the feedback in both groups was the same in the during-task stage when the speaking tasks involved pair work activity and group discussion. During this stage, both groups received confirmation checks without providing them with any correction. Confirmation checks were given spontaneously when the researcher was observing the class, group by group. Meanwhile, in the post-task stage, additional feedback was given. In this stage, both groups were required to do a group presentation from what had been discussed in the during-task stage. Feedback was given after all groups had presented rather than during presentation to avoid repeated interruption which could impede the flow of the presentation. A few errors were selected randomly from the students’ presentation and explicitly explained to the classroom as a whole, rather than as an individual oral assessment since time was limited. S. Li et al. (2016) revealed that groups that received attention to the form (explicit instruction and recasts feedback) acquired the target language (passive verb forms) effectively and explicit instruction had a greater effect on students’ language accuracy. However, the findings of the previous study do not corroborate those of the present study. The students in both groups of the present study did not make progress in accuracy.
Third, the characteristics and conditions of the task. The lack of improvement in accuracy might be arguably be because the nature of the tasks used in this study, which was different
186 compard to the previous study. In S. Li et al. (2016), dictogloss tasks were conducted by the teacher and students were required to listen to the narrative presented, work in pairs, rehearse with no time limit and retell to the class. However, in the present study, the time given for planning and completing the speaking tests was short. In addition, the speaking tests were conducted and assessed individually, thus students did not receive any guidance nor did they have time rehearse. Although the feedback given during the intervention did not fully benefit students’ accuracy in the post-speaking test, it did somehow give students opportunities to reflect on and modify their linguistic knowledge especially for the EIST group when they made more reformulations over time (see Section 6.3.4 for discussion in relation to reformulation and students’ fluency outcomes). Therefore, the explicit instruction and feedback may have prompted students in the EIST group to notice the gap between their existing knowledge and target-like knowledge, hence prompting them to reconstruct existing knowledge towards the target-like representations.
Fourth, fossilisation may have occurred. The qualitative results showed that errors concerning morphology (e.g., verb tense, subject-verb agreement and word form) constituted more than half of the total number of errors. Moreover, these errors seemed to be persistent and frequent over time, hence providing evidence of possible fossilisation in specific language areas in spite of the intervention. The findings of this study are not in harmony with the study of Van de Guchte et al. (2017), which revealed that the FonL group (i.e. focus on language, given written instructions directed to language) were able to achieve the target structure more accurately than the FonC group (i.e. focus on content, given written instructions directed to content). The contradictory findings of both previous and present study might be because of the students’ age and language experience. The average age of the students in the previous study was 14 and the German language was newly introduced to them as a foreign language in secondary school (two hours per week for about 17 months). However, the average age of the students in the
187 present study was 21 and English had been taught to them for nearly 13 years since primary school. It is possible that students in the present study, who had learned English grammar for some time, had committed errors during English lessons in the past which were not properly corrected, leading to fossilisation, whereas students in the previous study just had started to acquire the new language.