4.5 Feedback studies in CALL
4.5.4 The use of ICALL feedback by learners
Heift (2001a) conducted a similar study focusing on the use learners made of error- specific feedback generated by E-Tutor for grammar and vocabulary exercises. The study was conducted on 33 beginner learners of German at university level. The results show that learners reacted in five different ways to system feedback (2001a: p. 103):
• They corrected the error(s) explained by the system.
• They corrected an error in the sentence,but not the one explained by the system. • They changed a correct structure.
• They resubmitted the same sentence. • They requested the correct answer(s).
Heift (2001a: p. 107) showed that for the vast majority of sentences students at- tended to system feedback and corrected only the errors that were highlighted by the system. Heift interprets this as an indicator that students do read the feed- back messages. Moreover, Heift (2001a: p. 108) found that students attended to metalinguistic feedback and corrected their output accordingly, even if they had the opportunity to look at the correct answer. These findings indiicate a willingness by learners to be informed about errors.
In a later study, Heift studied the correlation between types of feedback and learner takeup among beginner learners of German at the university level with respect to the following three types of feedback (2004: p. 419):
• Metalinguistic feedback.
• Highlighting4 and metalinguistic feedback.
• Highlighting and repetition.
4Highlighting consists in using a graphical strategy to draw the learner’s attention to the location
The study concludes that in those circumstances experiment subjects were more likely to correct their mistakes with the feedback that combined highlighting and metalinguistic deployment, while they were less likely to correct them if the feedback combined highlighting and repetition. Heift found differences with respect to the variables gender and skill level, but these are not statistically significant.
4.6
Chapter summary
In this chapter, we presented fundamental aspects of the research and practice in SLA, FLTL and CALL. In particular, we introduced the concept of Communicative Language Teaching, an approach that emphasises the need of teaching and learning languages as a means for interaction. Within CLT, we focused on Task-Based Lan- guage Teaching, a methodology that relies on tasks simulating aspects of real-life communication settings to elicit from learners those linguistic and communicative elements that we expect learners to be competent on.
As we described, the consensus of SLA and FLTL researchers is that a well- justified attention to form, as in FFI, is not only positive but also needed in com- municative approaches to language learning – thus, the existence of more language- oriented tasks and more communication-oriented tasks. TBLT theorists and practi- tioners propose classifying tasks according to their communicative nature. Estaire and Zan´on (1994) and Ellis (2003)’s propose distingushing among tasks that purely focus on form and tasks that purely focus on meaning, but Littlewood (2004) pro- poses a continuum between the two with several stages that allows for a gradual and fine-grained classification. Given that there is no intrinsic goodness or badness in the nature of tasks (Brown, 2007: pp. 18 and 241, Ellis, 2003: Ch. 8), our aim will be to be able to learn the characteristics of the FL learning activities that are suitable for NLP-based assessment, that is, those that are part of the viable processing ground. We also presented Estaire and Zan´on (1994)’s framework for the design of TBLT- driven materials. This framework requires a thorough specification of the pedagogical and linguistic goals of each learning activity foreseen. These specifications provide the information that helps identify relevant linguistic features in the domain of ap- plication in terms of NLP.
Another important concept introduced is the assessment of learner production. We presented in detail Bachman and Palmer (1996)’s framework to characterise language tests in terms of target language use setting. Such a framework facilitates a detailed specification of the pedagogical goals and the communicative and linguistic properties of the emulated communicative setting. It includes strategies to describe activity instructions, expected learner responses, and the reciprocal influences that activity instructions have on the expected learner responses, which serve as seeds to not only determine the language to be learnt, but also the language to be processed in an ICALL context. These concepts were suggested as very relevant by Bailey and Meurers (2009) for the characterisation of the viable processing ground and will be further investigated in this thesis.
Finally, we discussed the importance of feedback and feedback types, and a series of studies in CALL and ICALL that investigated the effects of feedback on learning
gains, as well as on the use of feedback by learners. The importance of guaranteeing the learner the control over the access to feedback is emphasised by two studies, Pujol`a (2001) and Heift (2001b). Moreover, Nagata (1993, 1995), and Heift (2004) found that feedback strategies including metalinguistic feedback increased the per- formance of learners doing ICALL tasks in form-focused instruction.
Part III
ICALL tasks
[A] well-defined task design with its clear set of relevant language con- structions facilitates the restriction to a linguistic domain which is ‘man- ageable’ for a system’s natural language processing modules.
“Taking Intelligent CALL to the Task” Task-Based Language Learning and Teaching with Technology Mathias Schulze (2010: p. 79)
Chapter 5
Methodological considerations
This chapter presents key methodological considerations regarding ICALL instruc- tion settings. We present the different ways in which assessment occurs in ICALL settings as opposed to the way in which it occurs in face-to-face instruction. Af- ter that, we explore the relation between activity instructions, learner language, the language processing module and the feedback generation module.
These methodological considerations allow us to focus the object of study of our research. Particularly, we will consider the activity, the learner responses, the language analysis module, and the feedback generation module as four of the elements whose characterisation is critical for the specification of the linguistic properties of the expected learner responses, as well as for their assessment criteria. Such linguistic properties and assessment criteria turn into implementation requirements for the NLP-based feedback generation solutions.
5.1
Teaching and learning in an ICALL setting
As a particular kind of FL learning material, ICALL activities differ crucially from other kinds of learning materials in that, while learners do them, they interact with a virtual tutor enhanced with automatic assessment functionalities.