Entrepreneur: Start & Grow Your Business

Reading comprehension exercises online: the effects of feedback, proficiency and interaction.


by Murphy, Philip

This paper describes an ongoing project to create an online version of a reading programme, a custom-designed English language proficiency course at a university in Japan. Following an interactionist view of second language acquisition, it was hypothesised that comprehension of a reading passage could be enhanced by online materials promoting interaction between students as they completed a multiple-choice reading comprehension exercise. Interaction was promoted: (a) through pair work at a single computer and (b) by providing Elaborative feedback in the form of hints about incorrect answers as a means of stimulating discussion about corrections. Students were randomly selected from upper and lower levels of English proficiency, as determined by the Kanda English Proficiency Test (Bonk & Ockey, 2003), to receive either Elaborative feedback or Knowledge of Correct Response feedback (which supplies the correct answers). Within these groups, some students worked in pairs and some alone. Quantitative results show that the interaction between Type of feedback and Manner of study (individual or pair work) was statistically significant; students performed best on a follow-up comprehension exercise when in pairs and having been provided with Elaborative feedback. Furthermore, qualitative analysis of transcribed interactions also shows that Elaborative feedback was conducive to quality interaction.

INTRODUCTION

Advancing the design and use of computer-assisted language learning (CALL) activities is a key concern for researchers. As Chapelle (1997, pp. 19-22) explains, critical questions need to be answered about how CALL can be used to improve instructed second language acquisition (SLA). Two such questions are:

1) What kind of language does the learner engage in during a CALL activity?

2) How good is the language experience in CALL for L2 learning? Chapelle (1997) describes how the answer to the second question is dependent upon beliefs concerning

what types of language use are expected to be beneficial for second language development. For those espousing an interactionist view of SLA (Lantolf, 2000; Lightbown & Spada, 1999; Long, 1981; Pica, 1996; Van Lier, 1996), there is an assumption that L2 acquisition is facilitated by learners' interaction in the target language, thereby providing opportunities to comprehend message meaning. Accordingly, to ensure that L2 tasks meet such assumptions, and to facilitate SLA, researchers need to specify ideal observable features of learner language, such as signals that focus attention on language and features that may elicit a repetition or an expansion of previously acquired language.

In line with Chapelle's recommendations, a key concern for research is how these ideal features and appropriate tasks can be incorporated into an experimental reading programme. This concern is relevant due to the two goals of the current course, namely:

1) to provide students with the choice of an alternative and principled mode of online study and

2) to promote learner autonomy (Benson, 2001).

Throughout this paper, focus is placed on the first exercise that the students meet in the course, a reading comprehension exercise. It was hypothesised that increased interaction could be facilitated by requiring students to collaborate in pairs at a single computer (Beatty & Nunan, 2004; Stevens, 1992), and by providing Elaborative feedback in the form of hints to promote discussion as students self-correct errors. This type of feedback was provided as an alternative to Knowledge of Correct Response (KCR) feedback, which replicates traditional paper-based answer sheets by providing correct answers. It was also hypothesised that increased interaction through pair work with Elaborative feedback would be an effective method for promoting comprehension of a reading text. Results are analysed both quantitatively and qualitatively.

Context of the Study

During the second term of the reading course (see Murphy & Imrie, 2003 for a description), students are encouraged to choose from a series of activities and create, with guidance, an individualized syllabus. Students select, complete and then check their answers to the exercises with answer papers provided by the teachers. However, this procedure has proven to be problematic on the paper-based course for the following reasons:

1) when correcting answers, it is uncertain whether students: a) fully understand their errors and b) actively engage in the process of self-correction and,

2) for those students who choose to work outside of lesson time, there is potentially a wait of up to one week between lessons (and longer during holidays) before they can check their answers.

In a bid to overcome these challenges, this research focuses on the contribution that computer-mediated feedback can make. A key question that arises is: how and what kind of feedback maximizes comprehension? It is towards this issue that the following discussion is directed.

INTERACTION IN THE READING PROCESS

CALL researchers have turned to the work of interactionist SLA researchers when evaluating the quality of learner language. As Chapelle (1997) explains, the linguistic form of: "... a good interaction is hypothesized to occur when the normal interactional structure has been modified because the learner has requested, for example, a repetition, clarification or restatement of the original input" (pp. 25-26). This modified interaction is thought to be good because it can function to promote both the negotiation of meaning of the input (Beatty, 2003; Chapelle, 2001; Long, 1985; Nunan, 1993; Pica, 1994) and greatly contribute to language acquisition (Ellis, 1998; Krashen, 1985; Van den Branden, 2000). From a reading proficiency perspective, Larsen-Freeman and Long (1991) note:

Modification of the interactional structure of conversation or

of written discourse during reading ... is a [good] candidate

for a necessary (not sufficient) condition for acquisition.

The role it plays in negotiation for meaning helps to make

input comprehensible while still containing unknown linguistic

elements, and, hence, potential intake for acquisition. (p.

144)

Following research that points to the importance of comprehensible output to the acquisition of the target language (Chapelle, 1997; Swain, 1985), a conscious effort was made in this study to investigate the effects of feedback designed to promote negotiation of meaning, form and / or content in situations similar to those described by Swain and Lapkin (1995):

In producing the L2, a learner will on occasion

become aware of (i.e., notice) a linguistic

problem (brought to his / her attention either

by external feedback [e.g., clarification

requests] or internal feedback). Noticing a

problem 'pushes' the learner to modify his / her

output. (p. 373)

Although the importance of both negotiation of meaning and comprehensible output is well documented, few studies have investigated the effects on reading comprehension (Van den Branden, 2000); nevertheless, the design of this study was informed by research that was available and specifically by studies that point to the usefulness of promoting reading proficiency through interaction (Grabe & Stoller, 2002; Shanker & Ekwall, 2003). Despite the fact that the studies such as Eldredge and Butterfield (1986), Koskienen and Blum (1986) and Nes (2003) were carried out in non-computer-mediated environments, they provide positive implications for promoting interaction through paired online reading activities.

Quality Student Interaction Around Computers

As with non-computer-mediated environments, it is important to consider the interaction that is generated in computer-based tasks (Beatty, 2003; Stevens, 1992), and the type of interaction that is desirable for promoting comprehension, learning and language acquisition around computers. Based on findings from Fisher's (1992) study, students working on tutorial software exhibited the same IRF (Initiation, Response, Follow-up / Feedback) discursive structure. However, researchers have attempted to increase levels of interaction between students in various ways. For example, Wegerif and Mercer (1996) proposed a transformation to an IDRF (Initiation, Discussion, Response, Follow-up / Feedback) structure by including a discussion stage. Furthermore, software can also be developed to replicate techniques which teachers use to stimulate interaction, notably: (a) eliciting knowledge from students, (b) responding to what students say (confirmations, repetitions, elaborations and reformations) and (c) describing significant aspects of shared experiences ('we' statements) (Mercer, 2004).

When considering the quality of interaction around computers, two key features are particularly desirable: (1) learners need to be actively involved (Van den Branden, 2000); and (2) learners need to produce Exploratory talk in which partners engage critically and constructively with each other's ideas (Mercer, 1995). Regarding the former, Mercer (2004) explains how it is helpful for the analyst to perceive the degree in which students in joint activities are: "(a) behaving cooperatively or competitively and (b) engaging in the critical reflection or in the mutual acceptance of ideas" (p. 146). As for promoting Exploratory talk among learners, Wegerif, Mercer, & Dawes (1998), having been influenced by findings of research into effective collaborative learning (summarized in Wegerif & Mercer, 1996), note the importance of: sharing relevant information, reaching agreement, expecting reasons and challenges, discussing alternatives and encouraging peers. A key concern for research, however, is how quality interaction and reading comprehension can be promoted through computer-based activities.

Reading, Computers and the Internet

Despite the fact that models and guidelines recommending pedagogically sound practices for incorporating Internet-based materials exist (Berry, 2000; Brandl, 2002; Chun & Plass, 2000), a major concern is that the number of such examples remains limited. Likewise, guidelines for offering a reading course via the Internet (Caverly & McConald, 1998; Jones & Wolf, 2001; Mikulecky, 1998) are similarly few. However, evidence exists to support the assumption that integrating reading with computer-mediated support improves ESL students' reading skills (Chun & Plass, 1996; Hong, 1997; Stakhnevich, 2002; Williams & Williams, 2000). A common theme in studies such as these is that learners benefit from facilities offering support and assistance in web learning environments, for instance, from online dictionaries, glosses, graphics, blogs, bulletin boards or chat rooms. As a further example of a potentially advantageous facility that can also be offered through computer-assisted language learning (CALL), the following section comprises a discussion of the merits of computer-mediated feedback.

COMPUTER-MEDIATED FEEDBACK

One area in which computers are playing an increasingly important role in SLA concerns the identification of students' errors and the subsequent provision of appropriate feedback (Brandl, 1995; Clark & Dwyer, 1998; Tsutsui, 2004). However, many software products opt for a generic form of feedback and rarely go above the level of indicating whether an answer to a question is correct or incorrect (Sales, 1993). Beatty (2003) explains:

A CALL program is likely implicitly to state

'I have the answers to your questions; just

click here.' A teacher is more likely to say,

'What do you think the answer might be?' or

'Why do you ask this question?' ... it is

difficult for computers to deal with ambiguous

learner input, but this is an area of research

that needs to be further investigated. (p. 138)

Clariana (2000), who has published extensively on the topics of computer-mediated feedback, provides a succinct summary of the traditionally investigated types of feedback in CALL:

Knowledge of response (KR) that states "right" or "wrong" or otherwise tells learners whether their response is correct or incorrect; Knowledge of correct response (KCR) that states or indicates the correct response; and Elaborative feedback that includes several more complex forms of feedback that explains, directs, or monitors (Smith, 1988). Elaborative feedback includes the forms listed below:

1) Explanatory feedback provides additional explanations, such as why a learner's error response is incorrect or perhaps why a correct response is correct and various types of additional remedial screens that may amount to new instruction (Merrill, 1985, 1987; Spock, 1987).

2) Directive feedback may provide prompts, hints, or cues to assist the learner in determining the correct response (Nielson, 1990). Answer until correct (AUC, Pressey, 1926) is a common form of elaborative feedback where the learner is directed to respond until correct.

3) Monitoring feedback, also referred to as advisement, lets the learner know how they are doing overall. (p.1)

As can be seen, several forms of feedback exist. Accordingly, numerous researchers have tried to identify the most effective forms in different contexts. This issue is addressed in the next section.

How Effective is Feedback?

It is difficult to say which type of feedback is best for SLA as results are mixed (Clariana, 2000; Mory 1994). For example, summarizing findings by Bangert-Drowns, Kulik, Kulik, and Morgan (1991), AUC and Elaborative feedback are considered to be the most effective:

KR < no feedback < KCR < AUC = Elaborative feedback

However, following a review of 30 studies, Clariana's (1993) findings, which are consistent with both Schimmel's (1983) meta-analysis of 15 studies and also Kulhavy and Wager's (1993) research, show feedback has proven more effective than no feedback:

No feedback < KR = KCR = Directive feedback = Multiple try feedback

(AUC)

In addition to investigating the most effective type of feedback, various other aspects of computermediated feedback have also been the focus of research. For example:

1) Clariana and Koul (2006) investigate the effects of multiple-try immediate feedback for questions differing in difficulty

2) Clariana and Koul (2005) compare multiple-try feedback to other types of feedback also for questions of differing difficulty

3) Clariana (2003) and Clariana and Lee (2001) investigate the effects of recognition (multiple-choice) and recall (constructed-response) study tasks with feedback

4) Clariana, Wagner and Murphy (2000) consider the timing of feedback and

5) Nagata (1996) and Mory (1994) investigate the effects of Elaborative feedback.

A key question is whether computer-based feedback can offer advantages over traditional paper-based answer papers. Of particular interest, therefore, is the fact that Nagata (1996) found ongoing intelligent computer feedback to be more effective than simple workbook answer sheets for developing grammatical skill in producing Japanese particles and sentences. However, Clariana (2000, p. 2) draws on research to show how, in contrast to Nagata (1996) and Bangert-Drowns et al. (1991), Elaborative forms of feedback often produce no significant improvement over KR feedback despite requiring considerable development and implementation cost (Merrill, 1987). Nevertheless, Ferris (2003) explains how indirect feedback, or Elaborative feedback from a CALL perspective, is generally thought to be conducive to long-term student development; it forces students to think about their own errors and self-correction, thereby leading to: " ... increased student engagement and attention to forms and problems" (p. 52). Indeed, de Bot (1996) explains how students need to be active when producing language to discover what they can and cannot do. Noticing a problem, possibly through feedback, may be the incentive learners need to reengage with information in the input, thereby providing an opportunity for learning. Therefore, by placing the onus on the students to identify and correct their own errors, it would seem that the potential for interaction and negotiation of meaning, form and / or content is increased.

THE NEED FOR RESEARCH

When providing feedback, what messages should be supplied to the students on the screen? If feedback is presented in the form of KR or KCR feedback, it is simple to envisage what is displayed (a message such as right or wrong, a highlighted answer or a mark indicating the correct answer); however, with Elaborative feedback, what is it exactly that should be displayed to promote interaction and comprehension? This is the crucial question for anyone creating software to provide such feedback. Among the extensive literature related to the theme of feedback, the number of studies researching Elaborative feedback is relatively small; however, notable examples exist (Brandl, 1995; Clark and Dwyer, 1998; Mory, 1994; Nagata, 1996; Van der Linden, 1993). Even more difficult to find are examples of guidelines for its presentation. For example, Van der Linden (1993) notes: "Long feedback (exceeding three lines) is not read and for that reason is not useful" (p. 56). Van der Linden concludes that: "... feedback, in order to be consulted, has to be concise and precise." Mory (1994) recommends that isolated feedback should be avoided as it may provide little context for revision of an erroneous response. Consequently, Mory advises designers to: "... include the learner's answer and other alternative choices on the same screen as the feedback" (p. 287). As Chapelle (2001) explains, therefore, further research is vital: "What is needed are theoretically and empirically based criteria for choosing among the potential design options and methods for evaluating their effectiveness for promoting learners' communicative L2 ability" (p. 2). Accordingly, the following study comprises an investigation into the effectiveness of Elaborative feedback with the aim of identifying guidelines for researchers creating software to provide such feedback.

Research hypotheses

The literature shows that feedback has the potential to promote comprehension of a reading text; however, as to which type of feedback is more effective, results have been varied. The literature also shows that interaction between students can promote comprehension; however, both (a) what kind of interaction is generated through pair work as a result of Elaborative feedback, and (b) whether the interaction is sufficient to promote comprehension need investigating. The effect of a student's English proficiency level also needs to be determined. Stemming from this discussion, the following hypotheses were formed:

Hypothesis 1: Elaborative feedback will be more effective for promoting comprehension of the reading text than KCR feedback.

Hypothesis 2: Pair work will be more effective for promoting comprehension of the reading text than individual work.

Hypothesis 3: Students with a higher level of English proficiency will demonstrate higher levels of comprehension of the reading text than those with a lower level.

Hypothesis 4: Students studying in pairs and receiving Elaborative feedback will demonstrate higher levels of comprehension of the reading text than other students.

Hypothesis 5: Students with higher proficiency receiving Elaborative feedback will demonstrate higher levels of comprehension of the reading text than other students.

Hypothesis 6: Students with higher proficiency studying in pairs will demonstrate higher levels of comprehension of the reading text than other students.

Hypothesis 7: Students with higher proficiency studying in pairs and receiving Elaborative feedback will demonstrate higher levels of comprehension of the reading text than other students.

RESEARCH METHODOLOGY

Participants

The participants are first-year English majors at Kanda University of International Studies in Japan, who are streamed according to a test of global proficiency (Bonk and Ockey, 2003). In the 2005-6 academic year, 407 students (15 classes) were assigned to one of four bands: advanced (three classes), upper intermediate (four classes), intermediate (four classes) or lower intermediate (four classes). For the purpose of this study, the top two bands were grouped together and the bottom two bands were also grouped together.

Materials developed for this study were trialed with 162 of the 407 first-year students. 14 others were late or absent for lessons, and the main study was conducted with the remaining 231 students. Six of these 231 students (two pairs from the Elaborative feedback condition and one pair from the KCR feedback condition) scored 100% on the first comprehension exercise on the first attempt. Due to the fact that these students made no errors, and, therefore, received no feedback to promote comprehension (the focus of this research), these records were omitted from the study. Therefore, the statistical data analysis was performed on the data from the remaining 225 students. Videos were recorded of 12 volunteer students (six pairs) as follows: (a) four pairs were recorded during the trials and (b) two pairs were recorded during the main study. From analysis of the transcripts of these recordings, qualitative data is used to support / reject the various research hypotheses. All students agreed to participate in the study.

Materials: Reading Materials

Materials comprised one reading text (see Appendix A for an excerpt of the text) and two multiple-choice comprehension exercises, each with 15 questions (see Appendices B and C for example questions). Therefore, the maximum score on each exercise was 15 points. While the questions in the two comprehension exercises were different, the same content points were covered by corresponding questions.

Materials: Feedback Treatment

The methodology for displaying feedback is based upon the results of a study undertaken by Murphy (2005). Research was conducted into identifying the types of errors that students made in response to multiple-choice comprehension questions about a reading text and how students changed their answers following Elaborative feedback with an Answer-until-correct methodology. Students were allowed to change their answers until they answered correctly, until they gave up or until they ran out of time. Each question was associated with one piece of Elaborative feedback. Whenever a question was answered incorrectly, the corresponding Elaborative feedback was supplied irrespective of how many times the answers had been checked. Based upon interviews with students and an analysis of the way in which answers were changed as a result of feedback, recommendations for displaying Elaborative feedback are summarised as follows:

* Students should answer all questions before receiving any feedback

* Students should be allowed to get the computer to check their answers a maximum of four times by clicking on a 'Check answers' button on the screen at the end of the exercise.

* After checking answers, Elaborative feedback should be provided for each of the errors. If necessary, there should be up to three rounds of different Elaborative feedback before the KCR feedback. For those who make errors, the opportunity for random guessing of answers is minimised by not identifying incorrectly answered questions by question number; instead, students are encouraged to read the feedback, reengage with the materials, locate any errors by themselves and then change any answers they feel are incorrect. Furthermore, in order not to identify errors by question number, students are not supplied the correct answers until they either finish the exercise by answering every question correctly or receive the KCR feedback after the three rounds of Elaborative feedback.

* Following the first check of answers, the first round of Elaborative feedback should address the issue of silly errors such as selecting an incorrect answer by mistake. In this situation, students may not need a detailed explanation of the error if they are already able to comprehend. Instead, they just need an indication that an error has been made. Therefore, Elaborative feedback for each error should only direct students back to a key area in the text, for example: Please have