Learning and Instruction 19 (2009) 272e286 www.elsevier.

com/locate/learninstruc

Improving students reading comprehension skills: Effects of strategy instruction and reciprocal teaching
Nadine Spo rer a,*, Joachim C. Brunstein a, Ulf Kieschke b
a

Department of Psychology, Justus-Liebig University of Giessen, Otto-Behaghel-Strasse 10F, D-35394 Giessen, Germany b Department of Psychology, University of Potsdam, Karl-Liebkecht-Strasse 24/25, D-14476 Potsdam, Germany Received 20 December 2007; revised 18 March 2008; accepted 6 May 2008

Abstract The aim of this study was to investigate the effects of three different forms of strategy instruction on 210 elementary-school students reading comprehension. Students were assigned to any one of three intervention conditions or to a traditional instruction condition (control condition). Training students were taught four reading strategies (summarizing, questioning, clarifying, predicting) and practiced these strategies in small groups (reciprocal teaching), pairs, or instructor-guided small groups. At both the post- and follow-up test the intervention students attained higher scores on an experimenter-developed task of reading comprehension and strategy use than the control students who received traditional instruction. Furthermore, students who practiced reciprocal teaching in small groups outperformed students in instructor-guided and traditional instruction groups on a standardized reading comprehension test. 2008 Elsevier Ltd. All rights reserved.
Keywords: Reciprocal teaching; Reading comprehension; Reading strategies; Strategy instruction

1. Introduction A widespread goal of education in the elementary school is reading comprehension for all students because reading comprehension provides the basis for a substantial amount of learning in secondary school (Alvermann & Earle, 2003; Kirsch et al., 2002). In the last 20 years, a major goal of reading comprehension research has been to identify effective reading strategies that increase childrens comprehension (National Reading Panel, 2000). But as Guthrie, Wigeld, Barbosa, et al. (2004) pointed out, the evidence rests primarily on instructional research in which single cognitive strategies are taught in controlled experiments. Relatively little is known about the issue of how multiple strategies can, and should, be combined in comprehension instruction. In multiple strategies programs, strategy practice is often supported by peer-assisted learning arrangements (Brown, Pressley, Van Meter, & Schuder, 1996; Klingner, Vaughn, & Schumm, 1998; Palincsar & Brown, 1984). However, only a few investigations have addressed issues related to the identication of the effective elements inherent in multiple strategies programs. Consequently, the aim of this study was to examine the effect of strategies being taught on reading comprehension and how these strategies are practiced in relevant instruction. 1.1. Reading-comprehension strategies A substantial body of research suggests that reading-comprehension instruction should include explicit cognitive strategy instruction (Guthrie, Wigeld, Barbosa, et al., 2004). The theoretical bases for this suggestion are reading comprehension models,
* Corresponding author. Tel.: 49 641 9926194; fax: 49 641 9926199. E-mail address: nadine.spoerer@psychol.uni-giessen.de (N. Spo rer). 0959-4752/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.learninstruc.2008.05.003

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

273

such as Cromley and Azevedos (2007) direct and inferential mediation (DIME) model which in turn is based on Kintschs (1988, 1998) constructioneintegration model. The DIME model hypothesizes relationships among background knowledge, vocabulary, word reading, reading strategies, and inference that together result in reading comprehension. Reading vocabulary and background knowledge directly contribute to reading comprehension and also have effects that are mediated by inference. The model further suggests that the effect of strategies on comprehension is mediated by inference. Reading comprehension is correlated with a number of cognitive and metacognitive strategies, such as (a) activating background knowledge (Dole, Valencia, Greer, & Wardrop, 1991), (b) summarizing text (Armbruster, Anderson, & Ostertag, 1987), and (c) generating questions to capture the main idea of the passage (Rosenshine, Meister, & Chapman, 1996). Although these cognitive and metacognitive strategies have most frequently been investigated in isolation, some researchers have examined how they work together in more complex strategy packages (Brown et al., 1996; Guthrie, Wigeld, & Perencevich, 2004; Klingner et al., 1998; Palincsar & Brown, 1984; Souvignier & Mokhlesgerami, 2006). For example, reciprocal teaching (RT) is an instructional procedure developed by Palincsar and Brown (1984) to improve students text comprehension skills through scaffolded instruction of four comprehension-fostering and comprehension-monitoring strategies (Palincsar & Brown, 1984; Palincsar, David, & Brown, 1989; Rosenshine & Meister, 1994), that is, (a) generating ones own questions, (b) summarizing parts of the text, (c) clarifying word meanings and confusing text passages, and (d) predicting what might come next in the text. These four strategies are involved in RT in ongoing dialogues between a dialogue leader and the remaining students of the learning group. The dialogue leader, who can be a teacher or a student, models the use of the strategies, provides conditional knowledge about strategy use, and helps students to apply a strategy to a passage. As the students in the group become more familiar with the strategies and the procedure, dialogue leaders fade their involvement and other students take turns as discussion leaders. An underlying assumption of RT is that by applying the strategies in a group process, especially less able students can learn from their more knowledgeable peers. The overall goal is to promote, through scaffolding instruction and collaboration, the self-directed and exible use of the learned strategies. To sum, the following elements are essential to RT: instruction of the four comprehension-fostering and comprehension-monitoring strategies, application of the strategies using rich and meaningful reciprocal dialogues, and providing scaffold instruction during which teachers gradually fade their modelling of the strategies (Hacker & Tenent, 2002; Palincsar & Brown, 1984). A theoretical basis for suggesting effects of strategy instruction (which strategies are taught) and reciprocal teaching (how are strategies practiced) is Zimmermans (1998) self-regulation model. In this model, self-regulation is assumed to be organized within a learning cycle that capitalizes on three types of self-reective thoughts: (a) goal setting and strategic planning; (b) selfmonitoring of ones accuracy in implementing a selected strategy, and (c) self-assessment of strategy outcome and task performance. These processes are considered to be cyclic or recursive because each process entails information that can lead to changes in a subsequent step of the cycle. In addition, these processes qualify as self-reective cognitions in the sense that selfmonitoring of learning activities and associated corrective processes are central features of each step included in the cycle. Drawing on Zimmermans model, during reciprocal teaching students are engaged in cognitive and metacognitive activities: they alternate between prompting the use of a strategy, applying the selected strategy, and monitoring its accurate implementation. Hence, self-regulation procedures as described by Zimmerman (1998) are integral to RT. Since Palincsar and Browns (1984) seminal work, many studies have been conducted to test the effectiveness of reciprocal teaching. The procedure has been applied to different settings, age groups, and populations (Alfassi, 1998; Hart & Speece, 1998; Le Fevre, Moore, & Wilkinson, 2003; Lysynchuk, Pressley, & Vye, 1990; Palincsar, Brown, & Martin, 1987). In a meta-analysis involving 16 studies, Rosenshine and Meister (1994) reported a mean effect size of .32 for standardized test and .88 for experimenter-developed task favouring RT over control groups. Although there is clear evidence that RT promotes reading comprehension, no study analyzed the separate effects of the various aspects of RT and only a few studies have examined if RT students improved their strategies skills in terms of successfully applying a strategy to a passage. Furthermore, it is not clear if all or only one of the taught strategies is effective in fostering students reading comprehension (Rosenshine & Meister, 1994; for training college students see also Hart & Speece, 1998). So far, a training effect for summarizing only could be established. There were signicant improvements in four out of ve studies in which researchers collected summarization probes. For generating questions, ve out of six studies found no reliable difference between RT and control groups, although in all six studies RT students signicantly improved in their reading comprehension relative to control students. Even though students were taught the entire set of four strategies in 12 out of 16 studies, effects on making predictions were assessed in only one of these studies (Dermody, 1988). No study has examined students mastery of clarifying strategies. At present, it is thus unclear which strategies of RT signicantly contribute to the development of students reading comprehension skills. Besides these open questions regarding the empirical identication of effective strategies involved in RT, a number of difculties with implementing and practicing RT have been reported in the literature (Fuchs & Fuchs, 2001; Marks et al., 1993). In a qualitative analysis, Hacker and Tenent (2002) found that elementary-school teachers made many modications to adapt RT to the requirements of mainstream classroom instruction. Some teachers combined small-group activities with whole-class instruction to make the collaborative learning process easier for students as well as for the teacher. Other teachers required their students to write down their questions, answers, and summaries. Marks et al. (1993) observed that teachers sometimes changed RT in a way that elements supposedly playing a critical role in promoting deeper levels of reading comprehension, such as students

274

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

continuous engagement in reciprocal teaching, were completely dropped from the instruction. As Fuchs and Fuchs (2001) stated, there are two main reasons why it may be difcult for teachers to implement RT in naturally constituted classrooms: First, many teachers are unfamiliar with the procedure of reciprocal teaching. Second, the instructional technique for helping children to develop responsibility for strategic behaviour is challenging. Also, RT may be inappropriate for elementary-school children because students of all ability levels are assigned the role of the group leader and thereby have to take on responsibility for the groups learning. For young children, this may result in a cognitive overload (Rosenshine & Meister, 1994). One possibility of implementing RT in regular classroom lessons without losing major features of the program is to combine RT with peer-assisted learning arrangements in which students read in pairs (Fuchs & Fuchs, 2001; Fuchs, Fuchs, Mathes, & Simmons, 1997). Peer-assisted learning lessons consist of a set of structured activities, such as partner reading, summarizing, and predicting, and students are taught to enact these activities independently. At the beginning, teachers use a set of briey scripted lessons including teacher presentations, student practice, and teacher feedback. Later on, two students are paired to share and practice reading activities (McMaster, Fuchs, & Fuchs, 2006). Similar to Palincsar and Browns (1984) RT program, students roles can be reciprocal so that both students in a pair serve as tutor during each lesson. Compared to RT, reading in pairs has several advantages regarding the implementation of strategy instruction in regular classrooms. First, in their pairs students keep track of their reading activities on score cards that serve as external metacognitive guides and, thus, facilitate structured working in pairs (McMaster et al., 2006). Second, reading in pairs is more similar to instructional procedures teachers often adopt in reading lessons (e.g., partner reading) and, thus, is more likely to be accepted by teachers (Fuchs & Fuchs, 2001). Since teachers strongly rely on their beliefs and knowledge about instruction when adding new practices to their teaching repertoire (Borko & Putnam, 1996), Hacker and Tenent (2002) argued that researchers have to consider that teachers, too, need to take ownership of their learning by constructing their own understanding of new curricula and methods using their prior knowledge. Third, students who work in pairs have more opportunities to practice the use of reading strategies than students who work together in small groups. 1.2. The present study In our study, we adopted and further advanced the argument that there is a need for identifying effective elements of a multiple strategies program, namely RT (Rosenshine & Meister, 1994). Accordingly, our rst aim was to examine if both strategy instruction and reciprocal teaching contribute to the acquisition of reading strategies and, thus, to the development of young students reading comprehension. Our second aim was to examine the effects of a potentially more classroomappropriate intervention and so we created a condition in which RT was practiced in pairs. Altogether, we examined three intervention conditions and a traditional instruction condition (control condition) in terms of their effectiveness. In the rst intervention condition (RT condition), in which students practiced traditional RT (Palincsar & Brown, 1984), we integrated methods of direct instruction with cognitive modelling and phases of independent reciprocal teaching to help students acquire the four reading strategies of summarizing, questioning, clarifying, and predicting. To manipulate the presence versus absence of reciprocal teaching as an integral part of the strategy training we created a second strategy condition labelled instructorguided reading condition (IG condition). In this condition, a small group of 4e6 students was guided by a graduate assistant (the instructor) during the course of the intervention. The task of the instructor was to model the four reading strategies, ask students to apply a strategy and give feedback about the quality of strategy used. In a third strategy condition, labelled reciprocal teaching in pairs (RTP condition), students were rst taught the four reading strategies of summarizing, questioning, clarifying, and predicting, and then led to continue practicing these strategies in pairs. Similar to RT, these pairs engaged in reciprocal dialogues while they practiced the use of the learned strategies (see also Table 1 for a comparison of the three intervention conditions).

Table 1 Characteristics of intervention conditions. Instructional elements Conditions IG Strategy instruction Instructors use explicit instruction and cognitive modelling of reading strategies Practice in strategy use Students apply a strategy to a passage Scaffolding instruction and reciprocal teaching Instructors fade their involvement and students take turns as discussion leaders Recording reading activities on worksheets Students write down words to clarify, questions a teacher might ask, a summary, and a prediction RT reciprocal teaching; RTP reciprocal teaching in pairs; IG instructor-guided reading. RT RTP

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

275

We applied three criteria to evaluate the four conditions: (a) the effectiveness of the conditions (use of reading strategies at posttest), (b) the maintenance of strategies across time (use of reading strategies at follow-up test); and (c) the transfer of the learned strategies to experimenter-developed task (near transfer) and standardized reading comprehension test (far transfer). So, outcome measures included both experimenter-developed task and standardized comprehension test. As Rosenshine and Meister (1994) stated, experimenter-developed comprehension tasks may be easier to answer because compared to standardized tests, text passages are longer and organized in a topic-sentence-and-supporting-detail format. Furthermore, answering experimenter-developed questions usually requires less background knowledge and searching of the text. Therefore, to examine the level of generalization, both an experimenter-developed task as indicator of near transfer and a standardized test as indicator of far transfer were administered. The use of the four strategies: clarifying, summarizing, questioning, and predicting was assessed to analyze if differences in reading comprehension could be accounted for by differences in students strategy acquisition. The hypotheses guiding this investigation were as follows: First, drawing on previous reading comprehension research (Brown et al., 1996; Guthrie, Wigeld, Barbosa, et al., 2004; Klingner et al., 1998; Palincsar & Brown, 1984) and emphasizing the inuence of explicit reading strategy instruction on reading comprehension, we predicted that compared to the control condition, the three intervention conditions would be more effective in fostering the acquisition of reading strategies at post- and follow-up test as well as reading comprehension assessed with near-transfer task, that is, experimenter-developed task (Hypothesis 1). Second, as regards far-transfer tests, we expected the combination of strategy instruction with RT to be more successful than the control condition and the condition in which students were taught strategies for reading in the absence of RT, that is, in the IG condition (Hypothesis 2). Complying with earlier research highlighting the potential benets of incorporating self-reective, metacognitive practices into the training of cognitive strategies (Elliott-Faust & Pressley, 1986; Fuchs et al., 2003), both RT and RTP conditions should be especially effective in terms of far transfer. Third, drawing on reading comprehension models (Cromley & Azevedo, 2007) we assumed that RT effects on standardized reading comprehension tasks (for transfer tasks) would be mediated by the extent to which reading strategies were used correctly at the end of the intervention (Hypothesis 3). Finally, comparing RT and RTP conditions between them, we expected that the two conditions are similarly effective in fostering comprehension strategies and reading comprehension. Hence, instead of formulating a hypothesis describing differences between the two RT conditions, we posed the following question: To what extent does RTP differ from RT in inuencing reading strategies and reading comprehension?

2. Method 2.1. Participants e design Participants were 210 third- to sixth-graders from two elementary schools serving middle-class neighbourhoods in a mediumsized German town. Both schools are public half-day schools, without a special prole, teaching students from Grade 1 to 6. Public schools are not stratied at this stage. So students of all ability levels are instructed. The primary language of the children was German (86%). Their ethnic identication was predominantly (97%) Caucasian. Because we were not allowed to collect data about parents household income and education level, as indicator of socioeconomic status we asked the children how many books their family had at home. Most of the children (47%) indicated that their family had 26e100 books at home. According to independent ANOVA and chi-square tests, respectively, there were no signicant differences ( p > .05) between conditions in demographic data (see Table 2). The schools had no obligation to participate in the study, and the implementation of the reading intervention was completely voluntary. Students were randomly assigned to the different conditions in two steps. First, one school was randomly assigned to the traditional instruction condition as control group, whereas the other school was assigned to the intervention. Second, we randomly assigned students of each intervention class to the three different intervention conditions. Students of the intervention conditions were taught by instructional assistants in groups of 4e6 students. In the RT condition eight groups (with a total of 42 students, one group per class) were instructed. Another eight groups (42 students, one group per class) were instructed in the IG condition. Finally, 14 groups (60 students, depending on class size, one or two groups per class) were instructed in the RTP condition. In the control condition, students were instructed in reading comprehension by their regular teachers in German language lessons. In contrast, in each intervention class instruction was provided in small groups after regular lessons by graduate students. To control for instructional time, teachers of intervention classes provided no reading instruction during the course of the training. Therefore, even though control students received reading instruction during regular lessons, while intervention students received after their regular lessons, the total reading instruction time was comparable across conditions. The study involved a pretest, posttest, and follow-up test design. Pre- and posttest materials were administered one week before and after the intervention. Follow-up test was conducted 12 weeks after the posttest.

276 Table 2 Participant characteristics by conditions. Characteristics

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

Conditions RT RTP 60 22 38 16 16 18 10 49 11 6 10 23 5 16 IG 42 16 26 11 10 12 9 35 7 2 6 19 9 6 Control 66 31 35 18 18 18 12 59 7 1 7 38 9 11 Total 210 85 125 57 55 57 41 180 30 9 31 99 30 41

Number of participants Sex Male Female Grade 3 4 5 6 Language primarily spoken at home German Other language Books at home 0e10 11e25 26e100 101e200 More than 200

42 16 26 12 11 9 10 37 5 0 8 19 7 8

RT reciprocal teaching; RTP reciprocal teaching in pairs; IG instructor-guided reading; Control control condition (traditional instruction).

2.2. Intervention conditions 2.2.1. Reading strategy instruction Reading strategy instruction was delivered in 14 lessons (two lessons per week) each consisting of a 45-min lesson. In the rst phase of the training, students in the three intervention conditions received the same collaborative, interactive, and scaffolding instruction in three stages of strategy instruction: discussing, modelling, and practicing the strategies. Explicit teaching was chosen as instruction form (Palincsar et al., 1987). In the rst six training lessons students were introduced to the four strategies, one by one, using worksheet activities led by the instructor. The rst lesson served to familiarize students with the instructor. This was done by discussing why reading strategies are important and which strategies students already knew, outlining the upcoming training lessons, and working out behaviour rules during training lessons. Furthermore, students received folders for storing materials and passages. In the second lesson, students were introduced to the strategy of making predictions. Following Hart and Speece (1998), this strategy was described as a means of determining what might be discussed next by the author in order to help the reader to think about what he or she already knows about a topic in preparation for what might be coming next. The third lesson was devoted to instruction of question generating and to recapitulate the strategy prediction. Here, students were informed about the types of questions teachers might ask: questions about details, questions about cause and effect, questions that compare and contrast, questions about the main idea, and questions that require inference. In the next two lessons summarization and clarication were practiced. The following rules were used to teach summarization: (a) delete minor and unimportant information, (b) combine similar ideas into categories, (c) state the main idea when the author provides it, and (d) invent the main idea when author does not provide it (Kintsch & van Dijk, 1978). In the fth lesson, students were taught the clarication strategy by identifying either words, concepts, or both, that may be obscure, ambiguous, or hard to understand. In the sixth and concluding lesson of the explicit teaching part, students practiced collaboratively as well as independently the four reading strategies and received a bookmark with the most important information about the four reading strategies. At the end of the explicit teaching phase, students had acquired knowledge about the four strategies but still had not applied the strategies to longer reading passages. In the second phase of the training, students practiced applying strategies to reading passages through different forms of teaching which are described next. (Complete reading passages, lesson plan, and instructional materials for the three treatments are available from the rst author upon request.) 2.2.2. Traditional reciprocal teaching During the second part of the training (Lessons 7e14) students in the traditional reciprocal teaching (RT) groups received the intervention described by Palincsar and Brown (1984). Initially, the instructor demonstrated how the reading strategies were to be applied to a paragraph. She rst read the title of the passage and made a prediction about the content of the text. After reading aloud the rst paragraph, she modelled how word meanings or confusing passages could be claried and how the paragraph could

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

277

be summarized. She then formulated 2e3 questions and predicted what might come next in the text. Next, the instructor asked a student to apply one of the four strategies and provided praise and feedback. At the beginning of Lesson 9, the instructor stated that it would now be a students task to lead the dialogue. At this stage, as students were encouraged to provide instructional support for each other, the instructor explained the function of praise and feedback and helped the dialogue leader to formulate appropriate comments. Students kept on practicing reciprocal dialogues until the end of Lesson 14. 2.2.3. Reciprocal teaching in pairs In the RTP condition students were taught in Lessons 7 and 8 how to apply the four reading strategies. Initially, the instructor demonstrated how to apply the reading strategies to a paragraph. Similar to the traditional RT condition the instructor modelled how to apply a strategy to a paragraph, encouraged students to proceed, and praised students for the correct use of a strategy. In Lesson 9 students started working in pairs and used reciprocal dialogues to practice strategies. For this purpose, the instructor requested one student of each pair to lead the dialogue in the rst paragraph. Different from the traditional RT condition, pairs recorded their reading activities on worksheets. After the other student of the pair had read aloud the rst paragraph of the passage, he or she was asked to identify words the meaning of which was unclear. Students wrote down these words on a worksheet, tried together to clarify the words meanings and, if necessary, asked the instructor for help. Next, the dialogue leader asked the other student to formulate a question a teacher might ask. The leader provided feedback and together the two students discussed which question they wanted to write down. Then, the dialogue leader asked the student to summarize the paragraph. Again the leader provided feedback and together the two students formulated the summary and noted it on the worksheet. Finally, the leader asked the other student to generate a prediction, gave feedback and wrote down what they thought the best prediction would be. Then, the two students switched their roles. The pair discussed if the prediction came true and started to apply the strategies to the paragraph. Students kept on practicing reciprocal dialogues in pairs until the end of Lesson 14. The task of the instructor was to monitor the dialogues of the pairs and to provide assistance upon request. 2.2.4. Instructor-guided reading As in the above-described conditions the instructor modelled how to apply the strategies to a text paragraph. Then she asked a student to apply one of the strategies. In each of the remaining lessons she guided students activities by using prompts (What question did you think a teacher might ask?), instructions (Remember, a summary is shorter than the paragraph.), and modications (If you are having a hard time formulating a summary, why dont you rst state the main idea of the paragraph?). Different from both RT and RTP students, the IG students were never assigned the role of the dialogue leader. The instructor decided which of the students applied a strategy, asked other students to help, and gave praise and feedback. 2.3. Control condition During the course of the training, control students were instructed in reading comprehension by their regular teachers in two German language lessons per week with traditional instruction. Traditional instruction consisted of an extensive amount of text interaction with age-appropriate reading materials. Strategies such as activating background knowledge, clarifying, and predicting were taught implicitly as appropriate to the text. Whole-class reading as well as reading in small groups was used for practicing reading. 2.4. Instructional assistants and treatment integrity Instructors were six female graduate assistants who had gained in pilot work extensive experience in teaching reading skills to elementary-school children. The instructors were randomly assigned to student groups. Every instructor taught at least one group in each of the three intervention conditions. To prevent instructors from creating their own expectancies concerning the differential effectiveness of the three intervention procedures, we informed them that all conditions would be effective in fostering reading comprehension. Four months before the rst training lesson the six assistants met with the rst author. Each instructor received a manual describing in detail the strategies, exercises, materials, and instructions to be taught and assigned to students in each lesson included in the respective intervention condition. Instructors were required to model each lesson until they demonstrated a high level of prociency in modelling strategies, providing temporary guidance to students, fading instructional support, providing criterion-referenced feedback, encouraging children, and praising them for good work. Treatment integrity was assessed over time and by lesson (Gresham, MacMillan, Beebe-Frankenberger, & Bocian, 2000). In their logs, instructors checked each step of a lesson as it was completed and jointly discussed intervention progress in weekly staff meetings. During the rst phase of the training (Lessons 1e6) estimated time of activities was compared with instructors needed time. In the rst six lessons of the three strategy instruction conditions 100% of the steps were completed. Hence, although our strategy interventions were time-based, it was ensured that at the end of the explicit-teaching phase, all students had completed the assigned activities. For the second phase of the training, instructors had no guidelines how many paragraphs students should read per lesson. To facilitate reading comprehension, instructors encouraged students to apply the four strategies and discuss each

278

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

paragraph in detail. Instructors noted in their logs of Lessons 7 to 14 how many paragraphs were read by groups and pairs, respectively. 2.5. Training materials For the rst phase of the training, which contained the explicit teaching of the four reading strategies, worksheets for each strategy were handed to the students. These worksheets were originally developed for English speaking students by Brady (1990) and were translated by Demmrich (2005) for a training study with German elementary-school students. Each student received a bookmark that depicted the name and a symbol of each of the four reading strategies. Eight reading passages for third- and fourth-grade students and eight passages for fth- and sixth-grade students, comprising topics of age-appropriate science and social studies, were selected for student practice of reading strategies during the second phase of training. Third- and fourth graders and fth- and sixth-graders, respectively, received the same passages in the same order (one passage per lesson). Each expository passage was chosen on the basis of its possible appeal to a diverse student population. Subjects and texts for reading passages were obtained from workbook and magazine sources. Passages for third- and fourth-grade students consisted of between 179 and 312 words in length (M 260, SD 44). Passages for fth- and sixth-grade students consisted of between 236 and 368 words in length (M 330, SD 39). All passages were visibly divided into paragraphs. Each of the paragraphs consisted of at least three sentences building a meaningful unit of the text to ensure that a summary could be made. 2.6. Procedure Testing sessions lasted 90 min. In each class one of the six research assistants collected data in whole-classroom arrangement. Students were not permitted to use any external aid throughout the testing sessions. Written measures were used to assess students acquisition of reading strategies as well as reading comprehension. At each session, students were rst asked to read a passage, apply all four reading strategies step by step and answer a number of comprehension questions. After a 10-min break, students reading comprehension skills were assessed with a standardized scholastic achievement test. For each grade different testing passages were selected from German workbook and magazine sources. The length of the passages varied between 254 and 481 words in length (Grade 3: M 255, SD 1; Grade 4: M 311, SD 40; Grade 5: M 423, SD 5; Grade 6: M 465, SD 14). To ensure that passages and comprehension questions were equal in difculty passages were tested in the following way: Four months before the training started, for each grade, four passages with 10 reading comprehension questions each were developed and administered to 106 third- to sixth-graders from other classes. To avoid overload each student received a reading set consisting of two out of four passages. Pairwise contrasts among within-subjects means revealed no signicant differences between reading comprehension scores ( p > .05). On the basis of students reading comprehension scores three passages with nine questions each were chosen for the training studys pretest, posttest, and follow-up test. Furthermore, with the teachers of the participating classes we discussed each passage and questions and they judged the nally selected passages and questions to be equivalent in interest value. Since the pilot testing ensured equal difculty across texts, the reading passage sets were presented in the same order for each student across the three measurement points. None of these passages was assigned to any of the participating students during the instructional period. 2.6.1. Scoring Before scoring students responses to the open-ended questions, each questionnaire was assigned a code number so that the scorers, the six female graduate assistants, would be unaware of the testing session, intervention condition, student, and school. Scorers were trained to ensure reliability and accuracy in each measure. A 3-h training session was conducted including the presentation of procedures, rubrics for scoring each measure, controlled practice for each measure, and independent scoring of each measure. For each measure 20% of the assessments were randomly selected for a reliability check and independently rated by a second scorer. Interrater reliability (Pearsons coefcient) was computed for each measure and testing period. 2.7. Measures 2.7.1. Strategy acquisition Experimenter-developed assessments of strategy acquisition were modelled after Hart and Speeces (1998) study in which students were asked to apply all four strategies to a passage. For each grade, different passages were administered at pretest, posttest and follow-up test. To assess students ability to make predictions, only the title and the rst paragraph of the text were printed on the rst page of the questionnaire. Students were asked to read the paragraph and to write down how the text might continue. On the second page of the questionnaire the whole passage was printed. After students had read the passage they were asked to identify words or concepts that might need some further clarication. Then they were requested to generate questions. Finally, students were asked to write a summary about the passage.

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

279

These assessments were evaluated by a scorer on a 6-point scale adopted from Hart and Speece (1998) which was specic to each strategy (available from the authors). Predictions were evaluated as follows: 0 (no response), 1 (a response but untied to the passage), 2 (a response based on an unimportant detail of the paragraph), 3 (a response based on one feature of the paragraph), 4 (a response based on two or more features of the paragraph), 5 (a response based on two or more features of the paragraph clearly demonstrating the link between the paragraph and what probably would come next). For each testing time, interscorer reliability was >.85. Clarication was evaluated as follows: 0 (no response), 1 (a word whose meaning is stated in the text), 2 (a concept whose meaning is stated in the text), 3 (a difcult word), 4 (a difcult concept), 5 (a difcult word and a difcult concept), where difcult meant that a word or concept was neither directly nor indirectly explained in the passage. For each testing time, interscorer reliability was >.84. Questions were evaluated as follows: 0 (no response), 1 (a question of detail using a sentence from the text), 2 (a question of detail using own language), 3 (a question based on a main idea using a sentence from the text), 4 (a question based on a main idea using own language), 5 (an inference, comparison, evaluative, or cause and effect question). For each testing time, interscorer reliability was >.85. Summaries were evaluated as follows: 0 (no response), 1 (only topic sentences from the text are used, minor details are included), 2 (inclusion of topic sentences as well as of invented sentences, minor details are included), 3 (invented sentences are used, some minor details are included, does not quite capture the gist of the passage), 4 (invented sentences are used, no minor details are included, does not quite capture the gist of the passage), 5 (invented sentences are used, no minor details are included, completely captures the gist of the passage). For each testing time, interscorer reliability was >.88. 2.7.2. Reading comprehension Students performance in reading comprehension was measured with both experimenter-developed task and standardized test. At pretest, posttest, and follow-up test, students reading comprehension skills were tested with experimenter-developed tasks. These self-constructed tasks were administered as follows: After students read the passage and applied step by step all four reading strategies to the text (see Section 2.7.1), they were requested to indicate that they had nished by raising their hands. When a student indicated that s/he had nished reading, the text was removed. Students were then presented with nine comprehension questions. Three questions were designed to tap single pieces of information about the text (scored with one point for each correct answer) and six were designed to cover main ideas of the text and evoke longer, complex answers. Each of the latter questions was scored as follows: 0 (incorrect answer), 1 (a meaningful answer covering details of the text paragraph), 2 (a meaningful answer covering the main idea of the text paragraph) and, therefore, the range of responses was 0e15. For each testing time, interscorer reliability was >.91. Furthermore, at pretest and follow-up test all students completed Nauck and Ottes (1980) reading comprehension test from the Diagnostischer Test Deutsch [Diagnostic Test German]; Cronbachs alpha was .74 at pretest and .68 at follow-up test. This standardized reading comprehension test has two parallel forms. Consequently, we administered the one test form at pretest and the other at the follow-up test to assess transfer effects of strategy instruction. Each test form consists of two texts: Version A comprises a letter from a friend of 146 words and a short story about wasps of 203 words; Version B comprises a fable about a farmer of 142 words and a fable about a bishop of 236 words. Students were given 16 min to answer 20 (Version A) and 18 (Version B) multiple-choice questions, respectively. This test was more difcult than the experimenter-developed task because passages were not organized in a topic-sentence-and-supporting-detail format and answering the questions required re-reading the text and combining different sentences. For fourth- to sixth-graders, Nauck and Otte (1980) reported the correlation of students reading comprehension scores in this standardized test inventory with a measure of uid intelligence to be .35. Performance on this test was in standard scores, with M 50 and SD 10. 2.7.3. Social validity At posttest students were asked questions regarding their own perception of the effectiveness of the intervention. Specically, students responded to eight items reecting their motivation and involvement in the training as well as their enjoyment of working in a group. Example item was How much fun did you have? Cronbachs alpha for the internal consistency of the eight items was .85. They indicated their response to each item on a 1 (no/not at all) to 4 (very much) point scale. In addition, students were asked to respond to three open-ended questions: (a) What have you learned since we started working together? (b) What did you like most in the lessons? (c) What would you change in the lessons? 3. Results 3.1. Preliminary analyses and overview of statistical procedures The unit of analysis was each students individual score. Although students in the intervention conditions worked together in xed groups, control students did not. Consequently, groups of students were not used as unit of analysis.

280

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

Scores for strategy acquisition and reading comprehension measures were tested for signicant differences between conditions using ANOVAs. First, using condition as between-subjects factor, one-way ANOVAs were conducted for each measure at pretest to evaluate differences between conditions prior to the instruction. No statistically signicant differences were found (each p > .05). Next, correlations among pretest, posttest, and follow-up test measures were calculated (see Table 3). Except for Clarifying, a relationship was detected in the bivariate correlations of strategy acquisition scores and reading-comprehension scores. Although treatment differences at pretest were statistically nonsignicant, for three reasons we used an analysis of covariance (ANCOVA) approach of posttest and follow-up test data: (a) to reduce the probability of a Type II error, (b) to increase power by reducing the error variance, and (c) to control for variability in the pretest (Huck, 2000). Pretest scores did not interact with intervention condition ( p > .05), indicating that (a) our data met the ANCOVA assumption of homogeneous regression slopes and (b) pretest levels of reading skills did not moderate treatment effects on the dependent variables. To estimate the practical signicance of treatment effects, we computed partial eta-squared as a measure of the variance accounted for by intervention condition in the dependent variable of interest. Signicant ANCOVAs were followed by Bonferroni-adjusted pairwise contrasts among between-subjects means with the appropriate pretest as the covariate. To calculate effect sizes, we adopted a procedure that paralleled the ANCOVA approach: We rst covaried the pretest from the dependent variable and then used the residualized means and standard deviations to estimate the size of the effect. These effects are reported in standard deviation units (Cohens d; Cohen, 1988). To further explore treatment effects on students reading behaviour, we analyzed for each condition separately changes in repeated assessments of the reading strategies and comprehension measures. For this purpose, we adopted a multiple dependentsample t test procedure preceded by a multivariate F test for the Condition Time of Assessment effect. Because of the number of comparisons, Bonferroni correction of alpha level was applied and alpha was set at .006 (.05/8 .00625). To test whether training effects on far transfer (on the standardised reading comprehension test) were mediated by the correct use of reading strategies at the end of the intervention regression analyses were run (Baron & Kenny, 1986). Further, we conducted the Sobel test (Sobel, 1982), a statistically based method by which mediation can be formally assessed (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). We were also interested in examining the extent to which potential differences in the instructional effectiveness of our research assistants might have inuenced the outcomes of each of the interventions. For each intervention separately, we thus analyzed the reading-comprehension measures in a series of ANCOVAs, using posttest and follow-up test scores as the dependent variables and the respective pretest as the covariate. Except for one instructor effect ( p .04) on RTP students reading comprehension score (experimenter-developed task) at posttest, all effects for instructor were nonsignicant ( p > .05). In keeping with our observations from the treatment delity check, these results suggest that each of the three interventions had been implemented properly with a high degree of homogeneity across instructors. Means, standard deviations, adjusted means, effect sizes, and gain t-values with Bonferroni-adjusted signicance levels for each condition by measure are presented in Table 4. Student gender and grade did not modify any of the inferential analyses reported below. Therefore, these variables are not discussed further.

Table 3 Correlations among measures of pretest, posttest, and follow-up test. Measures Pretest 1. Clarifying 2. Summarizing 3. Questioning 4. Acquisition of predicting 5. Reading comprehension (ED) 6. Reading comprehension (ST) Posttest 7. Clarifying 8. Summarizing 9. Questioning 10. Predicting 11. Reading comprehension (ED) Follow-up test 12. Clarifying 13. Summarizing 14. Questioning 15. Predicting 16. Reading comprehension (ED) 17. Reading comprehension (ST) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

.04 .03 .02 .15 .10 .19 .01 .01 .06 .08 .01 .11 .06 .10 .01 .03

.34 .46 .29 .38 .18 .34 .30 .24 .29 .25 .36 .28 .36 .38 .22

.28 .21 .35 .09 .32 .25 .18 .30 .08 .25 .38 .25 .36 .26

.34 .16 .02 .29 .26 .25 .24 .07 .29 .26 .30 .37 .22

.22 .02 .10 .23 .20 .23 .02 .15 .10 .20 .26 .15 .09 .19 .29 .20 .40 .05 .28 .28 .26 .43 .44

.08 .14 .13 .05 .21 .01 .01 .01 .09 .08

.42 .38 .41 .14 .42 .35 .38 .45 .29

.36 .48 .01 .28 .30 .24 .38 .37

.35 .05 .30 .26 .31 .30 .19 .11 .42 .33 .24 .50 .42

.11 .05 .08 .09 .12

.44 .42 .55 .45

.35 .43 .32

.47 .30

.51

ED experimenter-developed task (near transfer); ST standardized test (far transfer). Testeretest correlations are in bold. Correlations >.13 are statistically signicant ( p < .05).

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

281

Table 4 Means, standard deviations, and adjusted means (with pretest scores as covariate), effect sizes (Cohens d ) and gains t for strategy measures by testing occasion and condition. Conditions Pretest M Clarifying RT RTP IG Control Contrasts Summarizing RT RTP IG Control Contrasts Questioning RT RTP IG Control Contrasts Predicting RT RTP IG Control Contrasts 0.75 0.75 0.93 1.16 ns 1.45 1.40 1.63 1.79 ns 1.97 2.66 2.15 2.17 ns 1.46 1.58 1.85 1.61 ns SD 1.04 1.09 1.09 1.17 Posttest M 1.41 1.29 1.48 0.90 ns SD 1.29 1.21 1.40 1.11 Adjusted M 1.45 1.33 1.48 0.84 d .80 .70 .67 Gains t 3.31* 3.03* 1.98 1.47 Follow-up test M 1.29 1.25 1.00 0.87 ns SD 1.28 1.37 1.22 1.22 Adjusted M 1.29 1.25 1.00 0.87 d .70 .66 .31 Gains t 2.06 2.31 0.26 1.39

1.50 1.48 1.58 1.52

2.40 1.38 3.03 1.69 2.62 1.60 1.77 1.64 Control < RT, RTP, 3.31 1.48 3.40 1.29 2.96 1.59 2.48 1.58 Control < RT, RTP 2.62 1.22 2.46 1.17 2.66 1.25 1.80 1.15 Control < RT, RTP, 6.63 2.70 6.29 2.58 6.56 2.90 4.67 2.05 Control < RT, RTP,

2.45 3.10 2.60 1.68 IG ( p < .05) 3.38 3.31 2.98 2.50 ( p < .05) 2.65 2.47 2.60 1.80 IG ( p < .05) 6.82 6.39 6.55 4.47 IG ( p < .05)

.63 1.02 .63

3.73* 6.32* 4.50* 0.12

2.52 1.46 2.57 2.48 1.39 2.55 2.15 1.44 2.13 1.90 1.48 1.82 Control < RT, RTP ( p < .05) 3.20 1.33 3.31 3.01 1.58 2.87 2.59 1.49 2.63 2.36 1.39 2.39 Control < RT ( p < .05) 2.95 1.26 3.01 2.37 1.42 2.38 2.55 1.46 2.47 2.06 1.37 2.07 Control < RT ( p < .05) 6.72 2.44 6.55 2.95 6.39 2.86 5.16 2.41 Control < RT, RTP, IG 6.93 6.66 6.38 4.97 ( p < .05)

.65 .66 .27

4.03* 4.84* 1.95 0.60

1.53 1.51 1.72 1.54

.67 .31 .32

5.03* 3.50* 2.51* 1.26

.74 .12 .17

4.76* 1.49 1.94 0.91

1.24 1.18 1.42 1.28

.81 .59 .54

5.00* 3.92* 3.92* 1.16

.79 .25 .17

7.08* 3.32* 3.36* 2.47

Reading comprehension (ED) RT 2.84 2.51 RTP 3.14 3.03 IG 3.51 2.90 Control 4.13 1.94 Contrasts ns Reading comprehension (ST) RT 47.26 9.05 RTP 46.93 9.32 IG 46.48 9.63 Control 45.65 7.40 Contrasts ns

1.44 1.09 1.05

6.87* 7.63* 6.46* 1.83

1.24 .91 .74

8.49* 7.25* 6.18* 3.19*

52.62 6.13 52.36 49.92 7.26 49.77 48.79 5.89 48.80 47.44 7.12 47.73 Control, IG < RT ( p < .05)

.57 .19 .10

4.68* 2.48 1.64 1.92

RT reciprocal teaching, RTP reciprocal teaching in pairs, IG instructor-guided reading, Control control condition (traditional instruction). ED experimenter-developed task (near transfer), ST standardized test (far transfer). Before group contrasts were tested for signicance, variations in pretest scores were partialled out. Cohens d effect size in standard deviation units (control vs. each treatment group). *p < .006 (Gains t Bonferroni-adjusted alpha level .05/8).

3.2. Strategy acquisition We analyzed posttest and follow-up test scores of strategy acquisition with four separate one-way ANCOVAs using the pretest as the covariate. Except for Clarifying at follow up, F(3, 205) 1.38, p > .05, all condition effects turned out to be signicant, although the effect sizes were small to moderate: for Clarifying at posttest, F(3, 205) 3.39, p < .05, partial h2 .05; for Summarizing at posttest, F(3, 205) 9.86, p < .001, partial h2 .13; for Summarizing at follow-up test, F(3, 205) 4.16, p < .01, partial h2 .06; for Questioning at posttest, F(3, 205) 4.53, p < .01, partial h2 .06; for Questioning at follow-up test, F(3, 205) 4.24, p < .01, partial h2 .06; for Predicting at posttest, F(3, 205) 6.76, p < .01, partial h2 .09; and for Predicting at follow-up test, F(3, 205) 4.40, p < .01, partial h2 .06. Contrast analyses revealed the following results (see Table 4): (a) Summarizing: At posttest, students in each of the three intervention conditions wrote better summaries than control students. At follow-up test IG students failed to create better summaries than control students, but RT and RTP students did. (b) Questioning: At posttest, RT and RTP students formulated better questions than control students. At follow-up test, only RT students outperformed control students. (c) Predicting: Similar to the acquisition of summarizing skills, at posttest students in the intervention conditions made better predictions than control students. At follow-up test, RT students continued to outperform control students. Except for the small effect of RTP students

282

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

questioning skills, all effect sizes were in the medium to upper range. For Clarifying, signicant differences between conditions were not found at either posttest or follow-up test. 3.2.1. Interaction of condition with time of assessment The Condition Time of Assessment on strategy acquisition scores in the mixed design ANOVA with Condition as between subjects factor and Time of Assessment as within subjects factor were signicant in the following cases, although the effect sizes were small to moderate: for Clarifying, Wilkss lambda .92, F(6, 410) 2.97, p .01, partial h2 .04; for Summarizing, Wilkss lambda .86, F(6, 410) 5.48, p < .001, partial h2 .07; for Questioning, Wilkss lambda .94, F(6, 410) 2.20, p < .05, partial h2 .03; for Predicting, Wilkss lambda .91, F(6, 410) 3.43, p < .01, partial h2 .05. Across testing periods, within-subjects analyses yielded the following results (see Table 4 for t-values comparing pre- and post-/follow-up-test scores, all p < .006): Except for Clarifying at follow-up test, RT students improved in all strategy acquisition measures from pre- to posttest and surpassed their pretest scores at follow-up test. The quality of RTP students strategy acquisition improved from preto posttest but at follow-up test only the quality of their summaries and predictions was signicantly better than at pretest. Finally, IG students strategy acquisition improved from pre- to posttest but relapsed at follow-up test (except for Predicting). The control students did not display a signicant change in strategy acquisition neither from pre- to posttest nor from pretest to follow-up test. 3.3. Reading comprehension Both at posttest, F(3, 205) 11.22, p < .001, partial h2 .14, and at follow-up test, F(3, 205) 6.90, p < .001, partial h .09, the effect of condition with the pretest as the covariate on the experimenter-developed task (near transfer) was signicant. At post- and follow-up tests, students in each of the three intervention conditions outperformed control students. Both at post- and at follow-up tests, the magnitude of the effects reecting the superiority of the intervention conditions relative to the control condition was large. The effect of condition on reading comprehension as assessed with the standardized readingcomprehension test (far transfer) at pretest and follow-up test was signicant, with moderate effect size, F(3, 205) 5.19, p < .01, partial h2 .07. Pairwise contrasts revealed that RT students outperformed IG students (Cohens d .55), and control students (Cohens d .57).
2

3.3.1. Interaction of condition with time of assessment The Condition Time of Assessment effect on reading comprehension scores was signicant for the near transfer test, Wilkss lambda .82, F(6, 410) 6.97, p < .001, partial h2 .09, but not for the far transfer test, Wilkss lambda .98, F(3, 206) 1.66, ns. The multiple dependent-sample t tests yielded the following pattern of results: (a) At posttest, students in each of the three intervention conditions improved their reading comprehension as assessed with the near transfer test compared to the pretest p .006. (b) Intervention and control students displayed a signicant change in the near transfer reading comprehension measure from pretest to follow-up test, p .006. (c) Only RT students improved in their reading comprehension (assessed with the far transfer test) from pretest to follow-up test, p .006. Table 4 shows the complete results of the within-subjects analyses. 3.4. Mediational analyses of far transfer test Except for Clarifying, all follow-up test scores of strategy acquisition (a) were signicantly ( p < .05) correlated with the far transfer test scores (Summarizing: r .42; Questioning: r .32; Predicting: r .30) and (b) were reliably predicted by intervention condition. This pattern of results met the conditions required by a mediational analysis according to the principles specied by Baron and Kenny (1986). Therefore, for follow-up test, we examined if variations in far transfer test scores might be accounted for by variations in the respective strategy variables with pretest reading-comprehension scores controlled for. In doing so, we adopted a hierarchical regression approach using a dummy variable to code intervention condition (0 control and IG, 1 RT), excluding RTP students because they had failed to outperform control and IG students on the far transfer test. To test the statistical signicance of the hypothesized mediational relationships, we used the formula given by Sobel (1982). In three separate regression analyses, Summarizing, Questioning, and Predicting each had a (nearly) signicant ( p .06) mediational effect on the relationship between intervention condition and reading comprehension at follow-up test (see Table 5, Models 2e4). When the three aforementioned strategies were entered simultaneously, as a set, into one and the same regression equation, the variance accounted for by the dummy-coded treatment factor in students reading comprehension declined from 7.6% to 5.4% but still remained signicant (see Table 5, Model 5). So, adding strategies into the model reduced the variance accounted for by treatment by 29%. Notably, in this multiple-mediator analysis, Summarizing constituted the only strategy that had a signicant unique effect on changes in reading comprehension from pretest to follow-up test.

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

283

Table 5 Hierarchical regression analysis with the follow-up standardized reading comprehension scores as dependent variable and the pretest standardized reading comprehension scores (Pretest RC), intervention condition, and strategy acquisition as predictors. Predictors Model 1 b Pretest RC Condition Summarizing Questioning Predicting R2 F .44*** .28*** T 6.36 3.97 Model 2 b .35*** .24** .30*** T 5.12 3.53 4.30 Model 3 b .39*** .24** .19* .29 (2, 147) 30.00, p < .001 DR2 .08 DF (1, 146) 18.47, p < .001 (1, 146) 6.67, p < .05 (1, 146) 7.44, p < .01 (3, 144) 6,92, p < .001 .03 .03 .09 .37 (3, 146) 28.54, p < .001 .32 (3, 146) 22.99, p < .001 T 5.43 3.36 2.58 .20** .32 2.73 Model 4 b .39*** .24** T 5.58 3.38 Model 5 b .33*** .21** .24** .07 .08 .38 T 4.67 3.09 2.96 0.93 1.05

(3, 146) 23.36, p < .001

(5, 144) 17.60, p < .001

*p < .05; **p < .01; ***p < .001. Clarifying was not added to the regression analysis because it was not signicantly correlated with the standardized reading comprehension measure and hence did not meet the conditions required by mediational analysis (Baron & Kenny, 1986). Condition dummy was coded (0 control and instructor-guided reading, 1 reciprocal teaching).

3.5. Social validity At posttest, students in all intervention conditions were asked to evaluate the intervention. The magnitude of the means reecting how much students liked the intervention was large for all three conditions (for RT students, M 2.95, SD 0.44; for RTP students, M 3.22, SD 0.59; for IG students, M 3.25, SD 0.40). A single factor (condition: RT, RTP, IG) ANOVA was conducted. There was a statistically signicant main effect, F(2, 141) 4.80, p < .01, partial h2 .06 (moderate effect size). Pairwise contrasts revealed effects in the medium range for RTP students (Cohens d .52) and IG students (Cohens d .71) when compared with RT students. Furthermore, students had to respond to three open-ended questions. From 144 students 92 responded to these questions. First, we asked students what they had learned in the lessons. Most of them (65%) stated that they had learned reading strategies, 25% answered that they had learned how to understand difcult texts and 10% indicated that they learned nothing at all. When asked what they liked most in the lessons, students replied that they liked reading different texts and applying the strategies (50%), the atmosphere during the lessons (25%), and cooperative learning and being the tutor (16%). Only 9% liked nothing at all. In the nal question, students were asked to tell what they would change in the lessons. Of them, 29% responded that they would change nothing in the lessons, 20% replied that they did not like a particular strategy or text and thus would leave it out, while 33% did not like working together with a particular classmate and 12% did not like to ll out worksheets. Finally, 6% of the students did not like the training at all. According to chi-square tests, there were no signicant differences between groups in students responses. In summary, students from all intervention conditions found their training useful for improving reading comprehension. 4. Discussion The main objectives of our study were (a) to investigate the effects of explicit instruction of reading strategies on third- to sixth-graders strategy acquisition and reading comprehension achievement and (b) to study the differential impact of practicing the strategies in reciprocal small group (RT) and pair (RTP) activities as compared to instructor-guided (IG) activities and traditional instruction (control group). The following three major results emerged from this study. First, compared with the control students, students in the intervention conditions were better able to use the strategies of summarizing, questioning, and predicting when reading a text at the posttest. Our ndings indicated that especially RT students who practiced strategies in reciprocal small group activities beneted in the short as well as the long run from training lessons. Medium to large effect sizes were obtained when RT students were compared with control students. Furthermore, RTP and IG students outperformed control students at posttest but did not maintain their superior performance at the follow-up test. Second, relative to control students, students in the three intervention conditions scored higher on the near transfer test of reading comprehension both at the posttest and at the follow-up test. The associated effect sizes were large and exceeded the effect sizes for near transfer measures of reading comprehension reported in the reciprocal teaching literature (Rosenshine & Meister, 1994). Control students showed very limited improvement in their reading comprehension over the course of this study.

284

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

Hence, the ndings of the present study conrmed the efcacy of explicit reading instruction as a feasible tool to enhance students reading comprehension. Thus, Hypothesis 1 was conrmed. Third, there was a positive improvement of performance in the far transfer test only for RT students when compared with IG and control students. Hypothesis 2 was partially conrmed. Although this far transfer test was administered not immediately but 12 weeks after training, the obtained effect sizes were larger than the median effect size (.32) reported by Rosenshine and Meister (1994) for reciprocal teaching interventions. Researchers and practitioners are in agreement that strategic processing of text is critical to reading comprehension (Cromley & Azevedo, 2007). In the present study, we could show how the RT strategies contribute to reading comprehension. A mediational analysis revealed that differences in the far transfer test at the follow-up test were accounted for by differences in students acquisition of the comprehension-fostering strategies of summarizing, questioning, and predicting. Thus, Hypothesis 3 was conrmed. Furthermore, our ndings supported the viewpoint that especially summarizing skills play a central role in mediating the effects of the RT method on the improvement of reading comprehension performance. Compared to IG and control students, RT students had learned to summarize text paragraphs more concisely and this growth in strategic reading helped them to better understand difcult, unfamiliar texts. To summarize, students who participated in one of the three intervention conditions showed near transfer in the sense that they reached higher reading comprehension scores as assessed with the experimenter-developed task. However, only students who practiced reciprocal teaching in small groups showed far transfer in the sense that they got higher reading comprehension scores as assessed with the standardized test. Compared to IG students, RT students had the opportunity to lead the dialogue. RT groups showed lasting effects of strategy acquisition. These results add to the extant knowledge about strategy instruction and reciprocal teaching (Palincsar & Brown, 1984; Rosenshine & Meister, 1994) in the sense that the relative advantage of RT indicates how multiple strategies can, and should, be combined in comprehension instruction to achieve lasting effects of reading instruction. Both the instruction of the comprehension-fostering (i.e., summarizing) and comprehension-monitoring (i.e., questioning) strategies and the application of the strategies using reciprocal dialogues in small heterogeneous groups were identied as effective elements of RT. In addition, students from all intervention conditions found their training useful for improving reading comprehension. Relative to IG and RTP counterparts, RT students, however, reported the lowest social validity scores. The ndings corroborate the view that for elementary students traditional RT is challenging, but worthwhile. With respect to RT students growth in reading comprehension, the role of the teacher in encouraging students to provide instructional support for each other, therefore, is to be emphasised. 4.1. Limitations and implications for future research Before closing, several limitations of our study should be acknowledged. First, we assumed that students taught to monitor and regulate their reading behaviour through reciprocal teaching (RT and RTP) would display superior performance in the standardized reading comprehension measure compared to IG students. It turned out that RT but not RTP students improved in the standardized reading comprehension measure from pretest to follow-up test. Not only in the domain of reading but also in the domains of writing (Harris, Graham, & Mason, 2006), and mathematics (Fuchs, Fuchs, Phillips, Hamlett, & Karns, 1995), research on cognitive strategy instruction suggests that peer support procedures are particularly effective in stimulating metacognitive activities while students work together and share their ideas. So, why did students who practiced reciprocal teaching in small groups (but not students in pairs) perform better after the training lessons? A possible explanation is that RT students took greater benet from small group activities because they got more room for discussing a paragraph and exchanging their ideas. In pairs, communication is naturally limited to two students. As the research assistants observed, the pairs lled out their worksheets systematically. However, working in pairs was more a completing of given tasks than a lively discourse. Given that for elementary students writing is very challenging, discussing and writing at the same time might be too difcult to handle. On the contrary, reading without writing may have made it easier for the dialogue leader in RT groups to give full attention to metacognitive skills, such as monitoring and regulating students comprehension (Zimmerman, 1998). However, to ensure the ecological validity of small group procedures, the instructional approaches need further renement so that they can be implemented by teachers in naturally constituted classrooms. For instance, Hacker and Tenents (2002) qualitative research in mainstream classes showed that the observed elementary-school teachers encountered different obstacles while implementing and practicing RT in their classes and therefore made many modications to strategy use and teaching. More research is needed to identify instructional settings and techniques that are both effective and feasible when strategies for fostering reading comprehension are to be integrated by teachers into the daily routine of classroom lessons. Component analysis of RT and RTP procedures could provide insight into the unique contributions of reciprocal teaching and writing components to learning, generalization, and follow-up test. Second, a main aim of our study was to nd out which strategies are most effective in improving reading comprehension. Whereas a strong relationship existed between reading comprehension and the strategies of summarizing, questioning, and predicting, no such relation was found for clarifying. Furthermore, no signicant correlation could be obtained between clarifying and the other three strategies. Before arguing that clarifying is a less effective strategy, it should be taken into consideration that

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

285

implications might be limited by the measurement of the acquisition of clarifying. Following Hart and Speece (1998), we assessed the acquisition of clarifying skills by asking students to note words or concepts that needed clarication. In this case, no response (coded with 0) could represent that a student felt no need to clarify a word or concept and, thus, wrote nothing down. Hence, a lack of correlation between reading comprehension and clarifying might have been due to the ambiguity of categories. Additional measures and rating categories are needed to tap more exactly the acquisition of clarifying. Third, it might be useful to consider motivation variables (e.g., a students interest in reading; see Guthrie, Wigeld, Barbosa, et al., 2004) as further potential constituents of a mediation model that could explain in greater detail than we did it here how reading strategies translate into reading comprehension. As Guthrie, Wigeld, Barbosa, et al., (2004) found out, an instructional framework combining motivation support and strategy instruction is more successful in increasing students reading comprehension than strategy instruction alone. Even though motivation support was not an explicit part of the training, it may be that during training lessons the role of the dialogue leader in RT groups was especially motivating for students and therefore increased engaged reading and reading comprehension. 4.2. Conclusion Despite these limitations, the results of our study corroborate the view that explicit instruction of multiple reading strategies is a feasible tool to enhance students reading comprehension and that third- to sixth-graders beneted most from explicit reading instruction supplemented with practice in small groups reciprocal teaching activities. Although it may be challenging for teachers to practice RT in the traditional way (that is, providing scaffolding for the four strategies, applying strategies in small groups while using reciprocal dialogues), it is worthwhile because this form of reciprocal teaching was the most effective. Above all, it is important to mention that after a relatively short time of reading comprehension instruction students had become self-regulated readers. Explicit strategy instruction and reciprocal teaching as part of the overall curriculum appear to be a promising procedure to get this process off to a good start. References
Alfassi, M. (1998). Reading for meaning: the efcacy of reciprocal teaching in fostering reading comprehension in high school students in remedial reading classes. American Educational Research Journal, 35, 309e332. Alvermann, D., & Earle, J. (2003). Comprehension instruction. In A. P. Sweet, & C. Snow (Eds.), Rethinking reading comprehension (pp. 12e30). New York: Guilford. Armbruster, B. B., Anderson, T. H., & Ostertag, J. (1987). Does text structure/summarization instruction facilitate learning from expository text? Reading Research Quarterly, 22, 331e346. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173e1182. Borko, H., & Putnam, R. T. (1996). Learning to teach. In D. C. Berliner, & R. C. Calfee (Eds.), Handbook of educational psychology (pp. 673e708). New York: Macmillan. Brady, P. L. (1990). Improving the reading comprehension of middle school students through reciprocal teaching and semantic mapping strategies. Unpublished doctoral dissertation. University of Oregon-Eugene. Brown, R., Pressley, M., Van Meter, P., & Schuder, T. (1996). A quasi-experimental validation of transactional strategies instruction with low achieving secondgrade readers. Journal of Educational Psychology, 88, 18e37. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Cromley, J. G., & Azevedo, R. (2007). Testing and rening the direct and inferential mediation model of reading comprehension. Journal of Educational Psychology, 99, 311e325. Demmrich, A. (2005). Improving reading comprehension by enhancing metacognitive competences: an evaluation of the reciprocal teaching method. <http://opus. kobv.de/ubp/volltexte/2005/524>. Accessed 28.02.07. Dermody, M. (February, 1988). Metacognitive strategies for development of reading comprehension for younger children. Paper presented at the American Association of Colleges for Teacher Education, New Orleans, Louisiana. Dole, J. A., Valencia, S. W., Greer, E. A., & Wardrop, J. L. (1991). Effects of two types of prereading instruction on the comprehension of narrative and expository text. Reading Research Quarterly, 26, 142e159. Elliott-Faust, D. J., & Pressley, M. (1986). How to teach comparison processing to increase childrens short- and long-term listening comprehension monitoring. Journal of Educational Psychology, 78, 27e33. Fuchs, D., & Fuchs, L. S. (2001). Peer-assisted learning strategies in reading: extensions for kindergarten, rst grade, and high school. Remedial & Special Education, 22, 15e21. Fuchs, D., Fuchs, L. S., Mathes, P. G., & Simmons, D. C. (1997). Peer-assisted learning strategies: making classrooms more responsive to diversity. American Educational Research Journal, 34, 174e206. Fuchs, L. S., Fuchs, D., Phillips, N., Hamlett, C., & Karns, K. (1995). Acquisition and transfer effects of classwide peer-assisted learning strategies in mathematics for students with varying learning histories. School Psychology Review, 24, 604e620. Fuchs, L. S., Fuchs, D., Prentice, K., Burch, M., Hamlett, C. L., Owen, R., et al. (2003). Enhancing third-grade students mathematical problem solving with self-regulated learning strategies. Journal of Educational Psychology, 2, 306e315. Gresham, F. M., MacMillan, D. L., Beebe-Frankenberger, M. E., & Bocian, K. M. (2000). Treatment integrity in learning disabilities intervention research: do we really know how treatments are implemented? Learning Disabilities Research and Practice, 15, 198e205. Guthrie, J. T., Wigeld, A., Barbosa, P., Perencevich, K. C., Taboada, A., Davis, M. H., et al. (2004). Increasing reading comprehension and engagement through concept-oriented reading instruction. Journal of Educational Psychology, 96, 403e423.

286

N. Spo rer et al. / Learning and Instruction 19 (2009) 272e286

Guthrie, J. T., Wigeld, A., & Perencevich, K. C. (Eds.). (2004). Motivating reading comprehension: Concept-oriented reading instruction. Mahwah, NJ: Erlbaum. Hacker, D. J., & Tenent, A. (2002). Implementing reciprocal teaching in the classroom: overcoming obstacles and making modications. Journal of Educational Psychology, 94, 699e718. Harris, K. R., Graham, S., & Mason, L. (2006). Improving the writing, knowledge, and motivation of struggling young writers: effects of self-regulated strategy development with and without peer support. American Educational Research Journal, 43, 295e340. Hart, E. R., & Speece, D. L. (1998). Reciprocal teaching goes to college: effects of postsecondary students at risk for academic failure. Journal of Educational Psychology, 90, 670e681. Huck, S. W. (2000). Reading statistics and research (3rd ed.). New York: Longman. Kintsch, W. (1988). The role of knowledge in discourse comprehension: a construction-integration model. Psychological Review, 95, 163e182. Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cambridge, England: Cambridge University Press. Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363e394. Kirsch, I., de Jong, J., LaFontaine, D., McQueen, J., Mendelovits, J., & Monseur, C. (2002). Reading for change: Performance and engagement across countries. Paris: Organisation for Economic Co-operation and Development. Klingner, J. K., Vaughn, S., & Schumm, J. S. (1998). Collaborative strategic reading during social studies in heterogeneous fourth-grade classrooms. Elementary School Journal, 99, 3e22. Le Fevre, D. M., Moore, D. W., & Wilkinson, I. A. G. (2003). Tape-assisted reciprocal teaching: cognitive bootstrapping for poor decoders. British Journal of Educational Psychology, 73, 37e58. Lysynchuk, L. M., Pressley, M., & Vye, N. J. (1990). Reciprocal teaching improves standardized reading-comprehension performance in poor comprehenders. Elementary School Journal, 90, 471e484. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83e104. McMaster, K. L., Fuchs, D., & Fuchs, L. S. (2006). Research on peer-assisted learning strategies: the promise and limitations of peer-mediated instruction. Reading & Writing Quarterly, 22, 5e25. Marks, M., Pressley, M., Coley, J. D., Craig, S., Gardner, R., DePinto, W., & Rose, W. (1993). Three teachers adaptations of reciprocal teaching in comparison to traditional reciprocal teaching. Elementary School Journal, 94, 267e283. Nauck, J., & Otte, R. (1980). Diagnostischer test Deutsch. [Diagnostic test German]. Goettingen, Germany: Hogrefe. National Reading Panel. (2000). Teaching children to read: An evidence-based assessment of the scientic research literature on reading and its implications for reading instruction. Washington, DC: National Institute of Child Health and Human Development. Palincsar, A. S., & Brown, A. L. (1984). Reciprocal teaching of comprehension-fostering and comprehension-monitoring activities. Cognition and Instruction, 1, 117e175. Palincsar, A. S., Brown, A. L., & Martin, S. M. (1987). Peer interaction in reading comprehension instruction. Educational Psychologist, 22, 231e253. Palincsar, A. S., David, Y. M., & Brown, A. L. (1989). Using reciprocal teaching in the classroom: a guide for teachers. Unpublished manuscript. Rosenshine, B., & Meister, C. (1994). Reciprocal teaching: a review of the research. Review of Educational Research, 64, 479e530. Rosenshine, B., Meister, C., & Chapman, S. (1996). Teaching students to generate questions: a review of the intervention studies. Review of Educational Research, 66, 181e221. Sobel, M. (1982). Asymptotic condence intervals for indirect effects in structural equation models. In S. Leinhart (Ed.), Sociological methodology (pp. 290e312). San Francisco: Jossey-Bass. Souvignier, E., & Mokhlesgerami, J. (2006). Using self-regulation as a framework for implementing strategy-instruction to foster reading comprehension. Learning & Instruction, 16, 57e71. Zimmerman, B. J. (1998). Academic studying and the development of personal skill: a self-regulatory perspective. Educational Psychologist, 33, 73e86.