Improving children’s reading comprehension and use of ... · PDF fileImproving children’s reading comprehension and use of strategies through computer-based strategy training Yao-Ting

Available online at www.sciencedirect.com

Computers in

Computers in Human Behavior 24 (2008) 1552–1571

Human Behavior

www.elsevier.com/locate/comphumbeh

Improving children’s reading comprehension anduse of strategies through computer-based

strategy training

Yao-Ting Sung a,*, Kuo-En Chang b,1, Jung-Sheng Huang b,2

a Department of Educational Psychology and Counseling, National Taiwan Normal University,

No. 162, Section 1, He-Ping East Road, Taipei, Taiwanb Graduate Institute of Information and Computer Education, National Taiwan Normal University, No. 162,

Section 1, He-Ping East Road, Taipei, Taiwan

Available online 20 August 2007

Abstract

In this study, the attention–selection–organization–integration–monitoring (ASOIM) model,revised from Mayer’s [Mayer, R. E. (1996). Learning strategies for making sense out of expositorytext: The SOI model for guiding three cognitive processes in knowledge construction. Educational

Psychology Review, 8, 357–371] SOI model of text comprehension, was used as a foundation todesign a multi-strategy based system, which was named Computer Assisted Strategy Teaching andLearning Environment (CASTLE). CASTLE aims to enhance learners’ abilities of using readingstrategies and text comprehension. The effects of CASTLE on students with different reading abilitieswere empirically evaluated. 130 sixth graders took part in an 11-week computer-based reading strat-egies course. The results show that CASTLE helps to enhance the students’ use of strategies and textcomprehension at all ability levels.� 2007 Elsevier Ltd. All rights reserved.

Keywords: Computer assisted reading; Strategies; Comprehension

0747-5632/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.chb.2007.05.009

* Corresponding author. Tel.: +886 2 23952445x509.E-mail addresses: [email protected] (Y.-T. Sung), [email protected] (K.-E. Chang), ppen@ice.

ntnu.edu.tw (J.-S. Huang).1 Tel.: +886 2 23622841x52.2 Tel.: +886 2 23622841x18.

mailto:[email protected]

mailto:[email protected]

mailto:ppen@ice.

Y.-T. Sung et al. / Computers in Human Behavior 24 (2008) 1552–1571 1553

1. Introduction

For enhancing reading ability, one of the methods most often recommended byresearchers is reading strategy instruction. Reading strategy instruction is found to beeffective. Through a deliberate design, it deals with problems on the vocabulary and sen-tence levels (Fischer, 2003; Sindelar, Monda, & O’Shea, 1990) as well as higher-level issuessuch as text comprehension (Alfassi, 2004; De Corte, Verschaffel, & De Ven, 2001; John-son-Glenberg, 2000; van Keer, 2004). In recent years researchers have tried to deduce prin-ciples of reading strategies from empirical studies (Dole, Duffy, Roehler, & Pearson, 1991;Mastropieri & Scruggs, 1997). Some of the more widely recommended approaches aredetermining the main messages (e.g., summarization), using text enhancements (e.g., illus-trations, text structure representations, and mental images), question and answer drills(e.g., self-questioning), meta-cognition (e.g., comprehension monitoring). Paris and Paris(2001) postulated that the using multiple strategies is one of the several important charac-teristics in successful reading strategy instruction.

In fact, considering the characteristics of successful strategy instructions describedby Paris and Paris (2001), one can find that it is rather difficult for teachers to createthose conditions in physical classroom environment. Take multiple-strategy instructionas an example. It takes a significant amount of training for most teachers to becomefamiliar with using the strategies and being able to teach them to their students (Duffy,1993a). In addition, preparing teaching materials for instruction simply takes too muchtime and energy (Pressley, Johnson, Symons, McGoldric, & Kurita, 1989). There arealso complexities involved in providing students with opportunities to apply the strat-egies in daily life. Perhaps the difficulties mentioned above are the major reasons whystrategy instruction is seldom implemented in classrooms (van Keer, 2004). Therefore,it is essential to create a proper environment for teaching assistance which can relieveteachers from their burden in implementing strategy instruction and help them toincreases students’ opportunities and willingness to use reading strategies outside theclassroom.

The difficulties of implementing reading strategy instruction in classrooms may bereduced with the assistance of information technology. Computer-assisted instructionhas been very popular during the last two decades, and scholars agree on the feasibilityof applying computers in reading instruction under appropriate designs (Labbo & Rein-king, 1999; Leu et al., 1998; Lungberg, 1995). There are generally several advantages ofincorporating computers in reading instruction (Lynch, Fawcett, & Nicolson, 2000;Mathes, Torgesen, & Allor, 2001). Firstly, computers can provide immediate individualfeedback based on student’s learning condition. Secondly, learning with computers allowsstudents to control the pace of learning by themselves. Thirdly, properly arranged coursesmay be operated independently with computers, thus relieving teachers from some of theburden and giving students more opportunities to learn independently. Finally, throughpresentations using different media, students’ motivation to read may be strengthened.

It is worth noting that, although the computer-assisted reading has had some impres-sive results (see Blok, Oostdam, Otter, & Overmaat, 2002; MacArthur, Ferretti, Okolo,& Cavalier, 2001, for reviews), there are some limitations as well. Regarding levels of read-ing abilities, most studies on computer-assisted reading have dealt with reading issues on amore fundamental level, such as word recognition (e.g., Frederiksen, Warren, & Roseber-ry, 1985; Wise, 1992; van Daal & Reitsma, 2000) or phonological awareness (e.g., Farmer,

1554 Y.-T. Sung et al. / Computers in Human Behavior 24 (2008) 1552–1571

Klein, & Bryson, 1992; Mathes et al., 2001; van Aarle & van den Bercken, 1999). There arerelatively few studies focusing on assisting higher-level text comprehension through com-puter technology. Currently, the majority of efforts on computer-assisted comprehensionfocus on the effects of text enhancement (such as indexes of difficult vocabularies or back-ground information for texts, Chang, Sung, & Chen, 2001; Leong, 1995; MacArthur &Haynes, 1995; Reinking, 1988) or visual or auditory aids (such as illustrations, imagesor speech, Elkind, Cohen, & Murray, 1993; Montali & Lewandowski, 1996) on enhancingtext comprehension. Those studies usually concerned with manipulating the assisting mes-sages, and dealt with the improvement of learners’ reading capabilities to a very limiteddegree. Their practical implications are still in debate (MacArthur et al., 2001).

Computer-assisted reading strategy training is a feasible approach to enhance students’capabilities to deal with texts. Regarding the training of strategies for text comprehension,there have been some studies on using computer technology to conduct strategy instruc-tion (Caverly, 1998; Chang, Sung, & Chen, 2002; Kaniel, Licht, & Bracha, 2000). Butthose strategies were chosen in a fragmented fashion, and were lack of integration. Forexample, Caverly (1998) used software for gathering strategies, arranging and presentingmaterials, while Chang et al. (2001) and Chang et al. (2002) used the concept mappingstrategy. Because currently, much emphasis is placed on using multiple-strategy instruc-tion in classroom situations (Paris & Paris, 2001; Pressley, Johnson, et al., 1989), we sug-gest it is important to find out how to design proper multiple strategies with computertechnology to facilitate text comprehension abilities. Considering that there is a limitednumber of computer environments which are directly aimed at the learner’s reading abil-ities and are equipped with a set of comprehension strategies, the first purpose of this studyis to design a reading instruction system with multiple strategies based on the componentsin the text comprehension process.

Another limitation of current available studies on computer-assisted reading is thatmost studies conducted on elementary school children focus on students with below-aver-age learning ability (MacArthur et al., 2001). In contrast, there are fewer studies using stu-dents of average or above-average abilities as participants. Some studies on strategyinstruction indicated that children of lower learning ability benefited less than those withaverage or above-average abilities (Garner, 1982). Nevertheless, some other researchersfound that students with poorer abilities still managed to benefit from strategy instruction(Brown, Pressley, Van Meter, & Schuder, 1996; De Corte et al., 2001; Duffy, 1993b; Pal-incsar & Brown, 1984). There are few studies related to issues of whether strategy instruc-tion in a computer environment produces different instructional effects on children withdifferent abilities. Thus, the second purpose of this study is to compare the benefits forchildren with different abilities using computer-assisted reading strategies.

2. Method

2.1. Participants

The participants in this study were sixth-grade students from four classes in an elemen-tary school of Taoyuan County, Taiwan. There were 65 students in the experimentalgroup (35 males and 30 females) and 65 students in the control group (34 males and 31females). The majority of the children were from middle-class families. The age of the stu-dents ranged from 12 to 13, with an average of 12 years 3 months.


2.2. Design

This study employed a quasi-experimental design. The independent variables in theexperiment included group (experimental or control) and reading ability (high or low).Each class was randomly assigned into the experimental group or the control group.The students were given the Reading Comprehension Screening Test (Ko, 1999) priorthe experiment. Students whose scores were higher than the average of all sixth-grade stu-dents were categorized to the high reading ability group, while those with scores lowerthan the average went into the low reading ability group. The experimental group had31 high-ability and 34 low-ability students. The control group had 35 high-ability and30 low-ability students. The dependent variables included a reading comprehension scoreof narratives, a reading comprehension score of exposition, and a score of using readingstrategies, which was further divided into five categories, self-questioning, error detection,inference blank-filling, summarization, and prior-knowledge integration. The covariatewas the students’ scores in their language arts course of the previous school year.

2.3. Experimental tool

This study developed a Computer Assisted Strategy Teaching and Learning Environ-ment (CASTLE) as the major tool for implementing the strategy instruction. To find asound base for designing strategies for CASTLE, the researchers of this study examinedvarious models of reading comprehension (e.g., Just & Carpenter, 1987; Kintsch, 1998;Mayer, 1996). Finally, the Mayer’s (1996) Selection–Organization–Integration (SOI) modelfor text comprehension, which we believed was a suitable framework, was utilized in choos-ing and designing reading strategies. In Mayer’s model, selection means that in reading anarticle, one needs to know which parts of the content are more important and which onesare less important. Organization means that one has to re-organize the selected messages inthe short-term memory, and then to form a coherent, interconnected overall concept. Inte-gration represents linking the knowledge organized in short-term memory with formeracquired knowledge into long-term memory. However, Dole et al. (1991) pointed out thatthe readers’ attention and whether they know how to monitor their own comprehension areimportant factors influencing comprehension. Although the SOI model has been validatedby several studies about text-based reading (see examples in Mayer, 1996) and multimedia-based reading (Harp & Mayer, 1998), the SOI model does not sufficiently reflect the com-ponents of attention and monitoring in reading. So, this study expanded Mayer’s SOImodel by adding the components of attention and monitoring. The attention–selection–organization–integration–monitoring (ASOIM) model (see Fig. 1) was constructed as theframework for selecting and designing comprehension strategies in this study. These strat-egies then formed the basis for designing a multiple-strategy reading assistance system,CASTLE. The examples, functions, and workflow of CASTLE are explained in details inthe Procedure section. The following explains the detailed design of each strategy.

Strategies for attention. The strategies of self-questioning and error detection were usedto help readers concentrate on reading texts. Self-questioning meant that learners askedthemselves questions while reading an article. For example, after readers had read an arti-cle, they asked messages of the article regarding who (characters), what (events), where(settings), when (time) and why (reasons), and then they would answer to those questions.For expository texts, questions regarding what, how, why, or relationships were more

Fig. 1. An ASOIM model for text comprehension.


emphasized. Andr’e and Anderson (1978–1979) and Anderson (1978) proposed that byusing this strategy, students would have to constantly focus their attention on the materialthey were reading. A number of studies (Clark, Deshler, Schumaker, Alley, & Warner,1984; King, 1989; Short & Ryan, 1984) also found that this strategy helped to enhancecomprehension. Another aspect of strategies for attention was error detection whichmeant that readers should find explicitly or implicitly contradictory messages in the con-tent of the article they were reading (Markman, 1979). In the process of error detectionreaders must concentrate to find any contradictions. This study incorporated it as oneof the strategies for attention.

Strategies for selection. This study used strategies of concept map blank-filling andhighlighting to train the participants to select important messages. Novak and Gowin(1984) argued that when teaching materials were presented through concept maps, stu-dents could easily find the interrelation of the concepts. To avoid the difficulty of produc-ing concept maps from scratch, this study applied the concept map blank-filling strategyby Chang et al. (2001), asking students to select corresponding important messages to sup-plement the missing parts of the concept maps. The highlighting strategy was implementedby researchers to confirm key concepts and to enhance reading comprehension (Shaugh-nessy & Baker, 1988; Wade & Trathen, 1989).

Strategies for organization. This study also adopted the strategies of concept map cor-rection and inference blank-filling to train readers how to organize the messages in an arti-cle. The concept map correction method worked as follows. After the students hadfinished reading an article, they would be given a complete concept map in which someconcepts were wrong. The students needed to select the correct concepts from a list of con-cepts provided by the system and then filled them in. Because the students must under-stand the organization and relationship of the main concepts in the entire concept mapbefore he could judge which concept nodes were correct or incorrect, the student had toorganize and link the concepts in the entire article. This process might help students toorganize the ideas in the text (Chang et al., 2002). Inference blank-filling requires the


readers to resort their own abilities in order to supply the missing important messages inthe context of the article. The system left out some blanks in an article. The readers neededto make judgments according to the context of the article, chose the correct words orphrases from a pull-down menu provided by the system, and then filled them in the blanks.

Strategies for integration. This study used proposition-combining and summarizationstrategies to train the readers to integrate knowledge. Proposition combining was derivedfrom the sentence-combining strategy (Fusaro, 1992; Neville & Searls, 1991). The differ-ence between proposition combining and sentence combining is that combining sentencesformed paragraphs, while combining propositions used concepts (arguments) as units andcombined them into propositions. In the strategy of proposition combining, importantconcepts were picked out from an article, and then mixed with new concepts related orunrelated to the old propositions that were not found in the present article (the conceptswere chosen from textbooks previously used by the readers). Then, the system providedrelation links (predicates) for the readers to integrate the concepts. Summarization wasa strategy preferred by many researchers. Anderson and Hidi (1989) pointed out that asummary was a brief narrative of condensed information to represent the original article,which could help students comprehend the reading material and organize the article theywere reading. In fact, it is not easy to re-organize the important concepts in an article with-out a fair amount of prior knowledge as foundation. Therefore, summary to a certaindegree should involve integrating new and old knowledge (Palincsar & Brown, 1984). Thisstudy adopted a method based on Brown and Day (1983) and Dole et al. (1991), whichasked readers to select important content from an article, to delete the trivial and redun-dant details, and to combine the remaining important sentences.

Strategies for monitoring. The purpose of the monitoring strategy in this study was totrain students to understand how well they are executing the strategies in the four processesdescribed above. By doing this, we tried to foster the reader’s awareness in their own com-prehension and employment of strategies. The method used in this study was based on theinformed strategies for learning by Paris, Gross, and Lipson (1984), which asks readers toremind themselves and to evaluate their state of learning after completing each strategylearning session through the verification of sentences related to the use of strategies.

After the selected strategies were implemented in CASTLE system, this study invited 52fifth graders to participate a 20-h pilot study. And the opinions from these students andtheir teachers were integrated for the revision of the interface and the procedure of CAS-TLE system. Through the current version of CASTLE, teachers might replace the textsused in CASTLE with any texts they think are suitable for their classes. More functionsfor teachers’ selection and design of other types of reading strategies fitting their needsare under implementation.

Teaching materials. The teaching materials were 62 articles which were selected andmodified jointly by the researchers and an elementary school teacher. Of these, 33 werenarratives and 32 were expository texts. For the experiment group, these articles wereincorporated into CASTLE, and each reading strategy instruction unit contained abouteight articles for practice.

2.4. Measures

Reading comprehension tests. The Expository Text Comprehension Test (ETCT) andthe Narrative Text Comprehension Test (NTCT, Lin & Su, 1991) were adopted as post-


test tools to evaluate the students’ comprehension abilities. The ETCT included five piecesof expository writings from both scientific and social science domains. Each article hadabout 200–400 Chinese characters. The NTCT contained five stories, each story includedabout 200–350 Chinese characters. Both tests had 25 multiple-choice questions con-structed from those articles or stories. The students scored 1 point for each correct answer.A pilot test on 196 fifth and sixth graders had Cronbach a coefficients of 0.69 and 0.71 forNTCT and ETCT, respectively. The criterion-related validities using the scores of lan-guage arts courses in the previous school year as the criterion were 0.50 and 0.49 forNTCT and ETCT, respectively.

Use-of-strategy test. The use-of-strategy test was developed for understanding the stu-dents’ abilities of using strategies after the strategy instruction. The strategy-use test leftout strategies that needed operations requiring long practice sessions, such as conceptmapping. There were five subtests. In each strategy subtest, one narrative and one expos-itory article were given. The articles, titles, word count, guiding remarks, and score rangesof the strategies are shown in the appendix. The format and rationale for the constructionare as following:

For the self-questioning, the students were required to write down five questions andanswers to those questions after reading each article. The standards for scoring were basedon the self-questioning strategy instruction by Clark et al. (1984) and the story grammarstrategy proposed by Short and Ryan (1984). If the questions posed by the students wererelated with the characters, setting, events, time, reasons, how, or relationships in the arti-cle, then they earned 1 point. If students’ answer for the posed question was correct thenthey got one more point.

For error detection, students had to detect the explicit and implicit contradictions,based on Markman (1979), in the article. Each article contained six errors, and each errordetected could score 1 point.

For inference blank-filling, the content was designed according to the cloze test by Sch-mitt (1987). Each article contained 10 blanks related to the context. The students couldearn one point for each correct blanks.

For text summarization, one narrative and one expository articles were provided. Thestudents were required to summarize each article in about 250 words. According to Garner(1982), summarizing efficiency represented students’ summarizing abilities. Summarizationefficiency scores ranged from 0 to 1, the fewer words used to represent the main ideas of anarticle, the higher the summarization efficiency. The idea units, which correspondedroughly to the major clauses of the sentences (Lorch & Lorch, 1996), in the summary werelabeled level 1, 2 or 3 according to their importance. 1 represented the main concepts thatshould be appeared in the summary, while 2 and 3 meant those secondary and unimpor-tant concepts that should not be appeared in the summary. The number of main conceptunits was divided by the total number of characters in the summary to get an indicator ofsummary efficiency. There were 13 and 14 idea units in the narrative and expository arti-cles, respectively. If students’ summaries had any level 1 idea units which were as same asthe idea units in the articles, then they earned one point.

For prior-knowledge integration, this test was designed according to the elaborationquestions proposed by Gagne, Yekovich, and Yekovich (1993). It asked the readers toelaborate their ideas on the article after completing the reading, such as giving moredetailed description of the situations in the text, citing examples, telling the possible fol-lowing scenarios, or making analogies. If the students’ elaborations were belong to any


of the four categories mentioned above and were related to the text, then one point wasassigned. Each correct cited example and analogy could score one point (maximum 3).

Ninety-nine sixth and fifth graders were sampled to get the reliability coefficients for thesubtests. For the narrative article, the Kuder–Richardson reliabilities were .68 and .64 forthe error detection and inference subtests, respectively. The inter-rater reliability from twograduate students scoring the self-questioning, summarization, and prior-knowledge inte-gration were .83, .84, and .82, respectively. For the expository article, the Kuder–Richard-son reliabilities were .65 and .63 for the error detection and inference subtests, respectively.The inter-rater reliability from two graduate students scoring the self-questioning, summa-rization, and prior-knowledge integration were .80, .78, and .79, respectively. In thepost-test of this study, any inconsistency in scoring self-questioning, summarization,and prior-knowledge integration subtests was solved by further negotiation by two raters.

2.5. Procedure

Both of the experimental and control groups took the screening test in a 20-min sessionone week before the formal experiment began. This study utilized the Reading Comprehen-sion Screening Test (Ko, 1999) to distinguish the students of high reading ability from thoseof low reading ability. The content of the test included one short article and 20 multiple-choice questions with four choices each. Each correct answer was scored one point. A pilottest administered on 175 sixth-grade students had a Cronbach a coefficient of 0.80. Its cri-terion related validity with the revised Peabody graphic vocabulary test (Lu & Liu, 1994)was 0.61 (p < .001). In the formal experiment, the experimental group was administeredinstructions on using CASTLE. The 50-min instructions were given twice a week duringthe 11-week instruction experiment, so there were 22 sessions in total. For session 1 to ses-sion 12, one strategy was instructed and completed in each session. Further, when two strat-egy-training sessions are completed, the following session was used for review (using newarticles). For session 13 to session 22, each strategy was reviewed in one or two sessions.

In each session, the students of the experimental group logged on to CASTLE systemthrough a browser on a local computer, and retrieved data of the teaching materials froma database on the remote server. CASTLE used an agent to guide the users. The agentgave voice instructions to provide guidance on three major procedures: the work proce-dure, the user’s interface, and feedback. The guidance on the work procedure was pre-sented to the users on the system startup. For guidance on the user’s interface,instructions were given to the users regarding the operations of the system interface.For guidance on feedback, users were given proper feedback according to the results afterany strategies were completed. Fig. 2 shows how the agent and the toolbox worked.

In CASTLE, the workflow for reading strategy instructions was designed according tothe steps of reading strategy instruction recommended by Taylor, Harris, and Pearson(1995). First of all, what was in the strategy(what) and why the strategy was important(why) were explained. The next step was to demonstrate how to use the strategy (how).After that, in what situation this strategy may be used (when) should be explained to read-ers. And then, an opportunity should be provided to guide the readers to practice usingactual articles. Finally, the readers would be asked to practice independently throughreading articles.

Incorporating the steps mentioned above, the procedures CASTLE executed were asfollows. On startup, CASTLE communicated with the users through text and voice

Object Illustration and explanation 1. The Agent has three functions: guidance for the work procedure, guidance for the user interface, and guidance for feedback

Agent

2. Right-click on the Agent and the system will provide the users with some explanation of the functions

System toolbox buttons Press the Read button to make the article selected by

the user appear.

Press the Underscore button to underscore the highlighted text.

Press the Undo button to remove highlighting.

Press the Voice button to convert the highlighted text into speech.

Press the Help button to bring up a help window with the agent.

Press the Add button to include new selection items.

Press the Delete button to delete selected items.

Pressing the OK button causes the system to provide feedback based on the operations of the users.

Press the Menu button to return to the Reading Strategy selection menu.

Fig. 2. Functions in the toolbox of CASTLE.


instruction by the agent to inform the users about the meaning of the selected readingstrategy. Then, the system provided an example, in which the agent operated and demon-strated the instruction procedures of the reading strategy. To help the users to get familiarwith that reading strategy, they were asked to choose another article to read, and actuallyused the strategy. After the operation had completed, CASTLE provided the users withfeedback, and then the users verified the processes of using the strategy. The four processesdescribed above formed a complete cycle. After reading one article, the readers needed togo through all the processes before he could go to the next article.

Regarding the explanation of a reading strategy and practice it on an example, CAS-TLE informed the users about the meaning of the reading strategy through text and voicenarration by the agent, and asked the users to practice it on an example. The users neededto practice successfully on the example before they were allowed to choose an article toread (see Fig. 3a).

For reading an article and using a strategy, after the readers had read the article, theycould use the mouse to highlight the sentences in the article, and then pressed the ‘‘under-score’’ button to transform the highlighted sentences into blue characters to signify thesemantic contradictions. The readers could also use the mouse to highlight the words orsentences he did not know and then pressed the ‘‘voice’’ button to ask the agent to pro-nounce or read aloud (see Fig. 3b).

Fig. 3a. An example of the interface for reading strategy instruction and example practice (using error detection).

Fig. 3b. An example of interface for reading the article and operating the strategy.


In the third step of giving feedback according to the reader’s operations, CASTLE pro-vided voice and text feedback according to the reader’s results of using the strategy (Fig. 3c).

The final step was verification of the reading process. When the readers had completedthe strategy training on one article, the system provided the readers with a six-questionreading process verification checklist to fill out. For example, the questions may include:I have used the detection strategy in reading the article (1, not at all; 4, very frequently);For the error detection strategy, I am: (1, not familiar with; 4, quite familiar with); What isthe meaning of error detection: (four answers to choose from) (see Fig. 3d).

The researchers and the monitoring teachers of the classes answered the students’ ques-tions regarding the operation of computers. All of the students’ operations and the timethey took were recorded by the computer, which also monitored the progress.

Fig. 3c. An example of the interface for feedback according to the operation process.

Fig. 3d. An example of the interface of the reading process checklist.


The control group used the self-study approach to learn the strategies. In each session,students were given the same reading materials to read. They were also given the paper-based workflow of the CASTLE. The main differences between the two groups were thatstudents in the control group could not see the demonstration of using the strategies by theagent. They could not get feedback when they were practicing paper-based exercises, andtheir progresses of learning were not monitored and used as a basis for feedback.

The post-tests were administered one week after the instruction experiment ended. Thetests came in three parts at different times. The first part was the ETCT and the summariza-tion test of expository text in the use-of-strategy test, each of which lasted 20 min. The secondpart was the NTCT and the summarization test of narrative text in the strategy-use test, eachof which lasted 20 min. The third part was a 35-min session for the self-questioning, errordetection, inference, and prior-knowledge integration subtests from the use-of-strategy test.

3. Results

This study collected the students’ scores of reading comprehension and strategy-usetests (shown in Tables 1 and 2). We conducted two two-way multivariate analyses ofcovariance (MANCOVA) to study how the students in different groups with different abil-ities differ in the various dependent variables.

3.1. Comparison of reading comprehension and strategy-use scores in the narrative text

The students’ scores of the Narrative Text Comprehension Test (NTCT) and scores ofusing strategies with narrative text, such as self-questioning, error detection, inference,summarization efficiency, and prior-knowledge integration, were submitted to a 2 (exper-imental or control groups) by 2 (high and low abilities) multivariate analyses of covariance(MANCOVA), in which the six scores mentioned above were used as dependent variablesand the language arts score from the students’ previous semester was used as the covariate.Prior to the MANCOVA analysis, we tested the assumption of the homogeneity of regres-sion coefficients of the covariate for different levels of groups and abilities, and the resultwas not significant, the Wilk’s Lambda statistic = .04, S = 3, M = 1, N = 57 1/2, p > .05.

Table 2The means and (standard deviations) of the scores of the strategy-use tests in narrative and expository articles

Test Experiment Control

High Low High Low

Self-questioning

Narrative 5.06 4.32 2.80 2.10(1.84) (1.64) (1.67) (0.88)

Expository 4.29 3.50 2.74 2.16(2.08) (1.86) (1.37) (1.11)

Error detection

Narrative 4.77 4.52 4.02 2.60(1.11) (1.23) (1.52) (1.52)

Expository 3.41 3.02 2.42 1.56(1.40) (1.16) (1.57) (1.33)

Inference

Narrative 7.09 7.26 6.91 6.30(1.24) (1.30) (1.42) (1.70)

Expository 8.00 7.41 7.00 5.66(1.31) (1.41) (1.60) (2.24)

Summarization efficiency

Narrative 0.037 0.033 0.019 0.022(0.0087) (0.0099) (0.0070) (0.0074)

Expository 0.040 0.036 0.028 0.025(0.0093) (0.0095) (0.0091) (0.0092)

Prior-knowledge integration

Narrative 2.35 1.82 1.68 1.27(1.02) (0.97) (0.79) (0.58)

Expository 1.71 1.15 1.29 1.133(1.04) (0.02) (0.67) (0.73)

Table 1The means and standard deviations of the scores of the reading comprehension tests in narrative and expositoryarticles

Experiment Control

M SD M SD

Narrative

High 16.06 3.51 15.49 3.60Low 13.79 3.36 12.27 3.12

Expository

High 15.06 3.27 14.09 3.50Low 13.35 3.35 11.30 3.28


This suggested a common regression coefficient was appropriate for the covariance portionof analysis.

The results from the MANCOVA revealed a significant multivariate test of the groupfactor, Wilk’s Lambda statistic = .87, S = 1, M = 2, N = 57 1/2, p < .05. The univariatetest indicated that students in the experimental group scored higher in NTCT, adjustedM = 14.92 (F(1, 122) = 4.38, g2 = .04, p < .05, MSe = 6.82); self-questioning, adjusted


M = 4.48 (F(1, 122) = 60.01, g2 = .34, p < .01, MSe = 1.86); error detection, adjustedM = 4.72 (F(1,122) = 21.33, g2 = .19, p < .01, MSe = 2.97), summarization, adjustedM = 0.036 (F(1,122) = 83.44, g2 = .41, p < .01, MSe = 0.02), and prior-knowledge inte-gration, adjusted M = 2.01 (F(1, 122) = 11.45, g2 = .11, p < .01, MSe = 3.46) than stu-dents in the control group, in which the adjusted Ms of the NTCT, self-questioning,error detection, summarization, and knowledge integration were 13.11, 2.22, 3.31, 0.022,1.38, respectively. The difference of the inference subtest between the two groups wasnot significant (F(1, 122) = 2.88, g2 = .03, p > .05, MSe = 5.27).

There was also a significant multivariate test of the ability factor, Wilk’s Lambda statis-tic = .89, S = 1, M = 2, N = 57 1/2, p < .05. The univariate test indicated that students withhigher reading ability performed better in NTCT, adjusted M = 15.89 (F(1, 122) = 12.78,g2 = .11, p < .01, MSe = 2.82); self-questioning, adjusted M = 4.01 (F(1, 122) = 3.86,g2 = .02, p < .05, MSe = 3.37), error detection, adjusted M = 4.20 (F(1,122) = 7.54, g2 =.06, p < .01, MSe = 2.24), and the prior-knowledge integration, adjusted M = 2.21(F(1,122) = 4.92, g2 = .05, p < .05, MSe = 2.11) than the lower ability group, in which themeans of NTCT, self-questioning, error detection, and knowledge integration are 12.25,3.11, 3.21, 1.12, respectively. The two groups were not different in the subtests of inference,adjusted Ms = 6.89 and 6.01 in high- and low-ability groups (F(1,122) = .71, g2 = .00,p > .05, MSe = 3.98) and summarization efficiency, adjusted Ms = 0.026 and 0.020 in high-and low- ability groups (F(1, 122) = .02, g2 = .00, p > .05, MSe = 0.20).

Regarding the interaction of group and ability, the multivariate test indicated that theWilk’s Lambda statistic = .92, S = 1, M = 2, N = 57 1/2, p < .05. The univariate test indi-cated that the interaction between group and ability was significant in the error detection(F(1,122) = 5.21, g2 = .04, p < .05, MSe = 2.20) and summarization (F(1,122) = 7.83,g2 = .06, p < .01, MSe = 0.09) subtests, but not in the subtests of self-questioning(F(1,122) = .13, g2 = .00, p > .05, MSe = 4.43), inference (F(1,122) = .1.24, g2 = .01,p > .05, MSe = 3.78), and prior-knowledge integration (F(1, 122) = .17, g2 = .00, p > .05,MSe = 4.42). A simple main effect test for the error detection subtest showed that group(experiment or control) had significant effects in both the high- and low-ability readers,and the treatment effect in low-ability readers (F(1,122) = 28.02, p < .01, MSe = 1.13)was more significant than that in high-ability readers (F(1,122) = 4.32, p < .05,MSe = 2.01). A simple main effect test for summarization indicated that the effect of abil-ity in the experiment group was not significant (F(1,122) = 2.43, p > .05, MSe = 0.12).However, it was significant in the control group (F(1,122) = 6.01, p < .05, MSe = 0.06),where students with low reading ability got higher scores in the summarization subtestthan those with high reading ability.

3.2. Comparison of reading comprehension and strategy-use scores in the expository text

The students’ scores of the Expository Text Comprehension Test (ETCT) and scores ofusing strategies with expository text, such as self-questioning, error detection, inference,summarization efficiency, and prior-knowledge integration, were submitted to a two-way MANCOVA, in which the design was the same as above section. Prior to the MAN-COVA analysis, we tested the assumption of the homogeneity of regression coefficients ofthe covariate for different levels of groups and abilities, and the result was not significant,the Wilk’s Lambda statistic = .10, S = 3, M = 1, N = 57 1/2, p > .05. This suggested acommon regression coefficient was appropriate for the covariance portion of analysis.


The results from the MANCOVA revealed a significant multivariate test of the group fac-tor, the Wilk’s Lambda statistic = 1.2, S = 1, M = 2, N = 57 1/2, p < .05. The univariate testindicated that students in the experimental group scored higher in ETCT, adjusted M =14.22 (F(1,122) = 5.32, g2 = .06, p < .05, MSe = 5.81); self-questioning, adjustedM = 3.66 (F(1,122) = 20.99, g2 = .14, p < .01, MSe = 2.83), error detection, adjustedM = 3.20 (F(1,122) = 21.78, g2 = .15, p < .01, MSe = 2.21), inference, adjusted M = 7.26(F(1, 122) = 16.82, g2 = .12, p < .01, MSe = 4.80), and summarization, adjustedM = 0.036 (F(1,122) = 44.19, g2 = .25, p < .01, MSe = 0.02). The difference of the prior-knowledge integration subtest in the experimental group, adjusted M = 1.33, was notsignificantly higher than the control group (F(1, 122) = 1.22, g2 = .01, p > .05, MSe =6.82). The adjusted means of the ETCT, self-questioning, error detection, inference, summa-rization, and knowledge integration of control groups were 12.66, 2.12, 1.99, 6.02, 0.022,and 1.11, respectively.

There was also a significant multivariate test of the ability factor, the Wilk’s Lambdastatistic = 1.23, S = 1, M = 2, N = 57 1/2, p < .01. The univariate test indicated that stu-dents with higher reading ability performed better in ETCT, adjusted M = 15.71(F(1, 122) = 10.68, g2 = .09, p < .05, MSe = 3.83) and inference subtests, adjustedM = 7.30 (F(1,122) = 3.56, g2 = .03, p = .052, MSe = 5..01), but not in the subtests ofself-questioning, adjusted M = 3.60 (F(1,122) = 2.11, g2 = .02, p > .05, MSe = 7.22), errorcorrection, adjusted M = 2.90 (F(1, 122) = 2.60, g2 = .02, p > .05, MSe = 6.77), summari-zation, adjusted M = 0.031 (F(1,122) = 0.94, g2 = .00, p > .01, MSe = 0.15), and prior-knowledge integration, adjusted M = 1.40 (F(1, 122) = 1.57, g2 = .00, p > .05,MSe = 7.83). The adjusted means of the ETCT, self-questioning, error detection, infer-ence, summarization, and knowledge integration of control groups were 12.35, 2.69,2.21, 6.18, 0.030, and 1.12, respectively.

Regarding the interaction of group and ability, the multivariate test indicated that theWilk’s Lambda statistic = .11, S = 1, M = 2, N = 57 1/2, p > .05. The univariate test indi-cated that there was no significant interaction between group and ability in all the depen-dent variables with the expository text.

4. Discussion

It is widely agreed by many researchers that comprehension strategies enhance com-prehension ability. However, the practical application of comprehension strategies inreading instruction leaves much to be investigated (Alfassi, 2004; De Corte et al.,2001; Pressley, Johnson, et al., 1989). Based on considerations for improving the acces-sibility of comprehension strategies, increasing the opportunities for students to prac-tice these strategies (Paris & Paris, 2001), and relieving teachers’ burdens ofdesigning strategy instruction (Pressley, Johnson, et al., 1989), this study selected com-prehension strategies that have more potential to produce benefits. According to theASOIM model of text comprehension and recommendations from scholars in paststudies (Mastropieri & Scruggs, 1997; Pressley, Johnson, et al., 1989), strategies withboth verbal and visual features, such as self-questioning, error detection, inference,summarization, and concept mapping are implemented on the Internet based CASTLEsystem.

We found that this type of design can effectively enhance students’ abilities to applystrategies. In this study, the participants in the experimental group performed better than


the control group in applying the majority of strategies corresponding to the text compre-hension process. This finding conforms with the discovery by researchers conducting sin-gle-strategy instruction in classroom situations, such as error detection training by Parisand Jacobs (1984) and Paris and Oka (1986), inference blank-filling training by Schmitt(1987) and Tregaskes (1987), and summarization training by Chang et al. (2002) and Jiten-dra et al. (2000). It is also consistent with the findings of researchers conducting studies onmultiple-strategy instruction, such as De Corte et al. (2001). The positive results of apply-ing the CASTLE system support the argument that strategy instruction should be reason-ably feasible in a computerized environment.

But this study also finds that the advantages of strategy training as described above arenot unconditional, and the performance varies with the style of articles being used. Forexample, in applying the inference blank-filling strategy, the students in the experimentalgroup only outperform those in the control group in expository articles, but not in the nar-rative articles. One of the reasons for this may be that in narrative articles, it is easier tofind relevant clues (such as the story grammar) through prior knowledge about the story.As a result, it is easier to complete the story and the difference between groups is less dis-tinguished. On the other hand, expository articles have a much more complex structureand contain far more related messages. Therefore, it takes more inference to find connec-tions between the concepts, and the scoring is more difficult. At this time the benefits ofstrategy training become more apparent.

In addition, in applying prior-knowledge integration strategy, the students in the exper-imental group only outperform the control group in narrative, but not in expository arti-cles. Several possible reasons emerge after analysis. First, the prior-knowledge integrationtest for expository article is more difficult to answer. Since the article used in this study ismainly scientific in nature (‘‘How Animals Breathe’’), the students may have difficulty inelaborating related concepts by describing the details, citing examples and continuing thearticle without prior knowledge or exposure to related articles. As a result, even if the stu-dents have the intention and ability to use the strategy of integrating prior knowledge withnew concepts, they may not be able to put this strategy into practice. On the contrary, thecontent of the narrative (‘‘the Brave Eskimo Child’’) may provide messages that are moreconcrete and closer to living and to give the readers more room for imagination, so thatthe integration with prior knowledge and the elaboration of the concepts in the textmay be easier.

Another noteworthy finding of this study is that in the control group, students of lowreading ability performed better than those with high reading ability in summarizingnarrative articles. While investigating the reason, we find that if we use the summarizingefficiency formula by Garner (1982), in which summarizing efficiency equals to the num-ber of main concepts in the summary divided by the total word count in the summarycontent, the summary word count may lead to lower summarizing efficiency if studentswrite down lots of words in their summary or higher efficiency if students write downvery few words in their summary, in case they have the same number of correct ideaunits. In this study, the control group students with low reading ability tend to use lesswords (202 on average) in writing the summary, while the control group students withhigh reading ability use more words (227 on average). Further, there are more students(12) who used less than 60 words in the lower ability group than those (four) in thehigher ability group. This is a possible cause for lower ability students got higher sum-marizing efficiency scores.


Many previous studies used reading strategy instruction programs to help studentsenhance their reading comprehension abilities, and had satisfactory application results(Brown et al., 1996; De Corte et al., 2001; Palincsar & Brown, 1984). This study also foundthat after the students’ reading strategies improved, their reading comprehension was alsoimproved. Past studies related to computer-assisted reading comprehension did not reachdecisive conclusions regarding its effect on improving reading comprehension (MacArthuret al., 2001). The positive effects produced by our computer-based strategy instruction maybe due to several reasons. First, the mechanisms for computer-assisted reading comprehen-sion in the past mostly focused on providing assistance for contents of articles themselves,like synthesized speech (Elkind et al., 1993; Farmer et al., 1992) or related text enhancementssuch as background knowledge, indexes for difficult vocabularies, summaries, etc. (Leong,1995; Reinking, 1988; MacArthur, Graham, Schwartz, & Schafer, 1995). These mechanismsdo not intervene with a learner’s own ability, so in the testing stage when there was no assist-ing message given, improvement in a reader’s achievement may not be easy to continue.Compared to the weaker approaches mentioned above, this study trained the readers inthe comprehension strategies that are directly related to their comprehension abilities. Thisinvolves a more robust intervention. Moreover, since the duration of strategy training in thisstudy was 11 weeks long with 22 sessions, which was much longer than the average com-puter-assisted reading programs, the effects may be more pronounced.

During the past two decades, more emphasis has been placed on cognitive process-ori-ented strategy instruction, and the arrangement and selection of strategy instruction havebeen growing in diversity (Dole et al., 1991). Process oriented and multi-strategy basedinstruction may be an important factor in making strategy instruction beneficial to learn-ers with poorer learning abilities (Brown et al., 1996; De Corte et al., 2001; Klinger & Vau-ghn, 1996; Palincsar & Brown, 1984), rather than limited to just those with high learningabilities. Our results indicate that the CASTLE system, which is also a process orientedand multi-strategy based teaching environment, again supports the finding that the multi-ple-strategy instruction can benefit students with low reading ability. Another possible rea-son for this benefit might be attributed to the characteristics of computer-basedinstruction, such as individual instruction, monitoring and immediate feedback. CASTLEallows personalized progress by placing a learner in an independent and threat-free situa-tion in which enhances the effects of practice. This study finds that these characteristicsmay also have a high potential for helping learners with lower abilities in their acquisitionof higher-level skills, and not only for lower-level reading skills such as word-recognitionand phonological awareness.

This study found that while integrating computer technology with strategy instructionhas a potential for reading instruction design, there are some points worthy of investiga-tion. To elaborate the design of systems like CASTLE, researchers may consider providinga wider variety of appropriate reading strategies for learners or teachers to choose from, orarrange strategies according to their own needs or preferences.

Acknowledgements

The evaluation study of the CASTLE system was supported by a grant from theNational Science Council, ROC (Contract Numbers NSC91-2520-S-003-023 andNSC93-2520-S-003-004) and the Center for Research of Educational Evaluation andDevelopment (95E0012-C-01-01).


Appendix. The articles, score range, word count and instruction used in the strategy-use tests

Strategy Article title Wordcount

Guiding remarks and examples

Self-questioning(0–10)a

1.b SmartFarragut

472 Guidance: This is a self-questioning test. Afterreading the articles presented, pleasewrite as many questions in the articleas you think are important as possi-ble in the limited time, then writedown the right answer

Example: Yue-fei lived in the Sung Dynasty. Hefought wars with the Jing troops andavenged the shame of our country.He won people’s respect

Question: Who is this article mainly about?Answer: Yue-fei

2. Penguin 365

Error detection(0–6)

1. Brave Jackson 260 Guidance: There are two articles in this test.Both of the articles contain somestatements that are not consistentwith the context. Please read thetwo articles carefully, and put circlesaround the statements that youthink are wrong

Example: Tom forgot to bring his homework toschool this morning and was punishedby the teacher. He was very happy .What should we change happy into?

2. Why areflowersfragrant?

333

Inference blank-filling (0–10)

1. A white dogthatappreciatespaintings

280 Guidance: There are two articles in this test.There are some blanks in the articles.Please fill in the blanks with one ortwo appropriate words according tothe title and the context

Example: The ingredients of air include at leastoxygen and carbon dioxide. Plantsoften release . . . the morning to makethe morning air especially fresh.The blank . . . should be filled with‘‘oxygen’’

2. Howdemoisturizerswork

300

Summarization(0–1)a

1.b The legend ofLanternFestival

798 Guidance: This is a summarizing ability test.Please use 20 min to carefully readthe article presented; and when youhave finished, find out the contentyou think is important and write asummary of no more than 250words. You may also cross out theunimportant content in the articleand then use the remaining content

2. Koala 760

Appendix (continued)

Strategy Article title Wordcount

Guiding remarks and examples

Prior-knowledgeintegration (0–5)

1. TheEskimoHunter

230 Guidance: There are two articles in this test. Eachone is followed by several questionsrelated to the article. After you have fin-ished reading each of the articles, pleaseanswer the following essay questions

Example: Taking the Eskimo Hunter as example,the questions are:

1. The story mentions that Kish’s father is agreat hunter. He brings home lots ofmeat and distributes the meat fairly.Besides that, try to write a more detaileddescription of Kish’s father

2. The story mentions that the chief held avillage council with the hunters. Nowcite some examples of when we needto hold meetings

3. If the story continues developing, whatdo you think will happen next?

2. Howanimalsbreathe

221

a Score range.b 1: Narrative style; 2: expository style.


References

Alfassi, M. (2004). Reading to learn: Effects of combined strategy instruction on high school students. Journal of

Educational Research, 97, 171–184.Anderson, T. H. (1978). Another look at the self-questioning study technique. Technical Report No. 6. Urbana:

University of Illinois, Center for the Study of Reading.Anderson, V., & Hidi, S. (1989). Teaching students to summarize. Educational Leadership, 46, 26–28.Andr’e, M. E. D. A., & Anderson, T. H. (1978–1979). The self-questioning study technique. Reading Research

Quarterly, 14, 605–623.Blok, H., Oostdam, R., Otter, M. E., & Overmaat, M. (2002). Computer-assisted instruction in support of

beginning reading instruction: A review. Review of Educational Research, 72, 101–130.Brown, A. L., & Day, J. D. (1983). Macrorules for summarizing texts: The development of expertise. Journal of

Verbal Leaning and Verbal Behavior, 22, 1–4.Brown, R., Pressley, M., Van Meter, P., & Schuder, T. (1996). A quasi-experimental validation of transactional

strategies instruction with low-achieving second-grade readers. Journal of Educational Psychology, 88, 18–37.Caverly, D. (1998). Techtalk: GAP, a reading strategy for multiple sources. Journal of Developmental Education,

22, 38–39.Chang, K. E., Sung, Y. T., & Chen, S. F. (2001). Learning through computer-based concept mapping with

scaffolding aids. Journal of Computer Assisted Learning, 17, 13–27.Chang, K. E., Sung, Y. T., & Chen, I. D. (2002). The effect of concept mapping to enhance text comprehension

and summarization. Journal of Experimental Education, 71, 5–24.Clark, F. L., Deshler, D. D., Schumaker, J. B., Alley, F. R., & Warner, M. M. (1984). Visual imagery and self-

questioning: Strategies to improve comprehension of written material. Journal of Learning Disabilities, 17,145–149.

De Corte, E., Verschaffel, L., & De Ven, V. (2001). Improving text comprehension strategies in upper primary

school children: A design experiment. British Journal of Educational Psychology, 71, 531–559.Dole, J. A., Duffy, G. G., Roehler, L. R., & Pearson, P. D. (1991). Moving from the old to the new: Research on

reading comprehension instruction. Review of Educational Research, 61, 239–264.


Duffy, G. G. (1993a). Teachers’ progress toward becoming expert strategy teachers. The Elementary School

Journal, 94, 109–120.Duffy, G. G. (1993b). Rethinking strategy instruction: Four teachers’ development and their low achievers’

understanding. The Elementary School Journal, 93, 231–247.Elkind, J., Cohen, K., & Murray, C. (1993). Using computer-based readers to improve reading comprehension of

students with disabilities. Annals of Dyslexia, 43, 238–259.Farmer, M., Klein, R., & Bryson, S. (1992). Computer-assisted reading: Effects of whole-word feedback on fluency

and comprehension in readers with severe disabilities. Remedial and Special Education, 13, 50–60.Fischer, C. (2003). Revisiting the reader’s rudder: A comprehension strategy. Journal of Adolescent & Adult

Literacy, 47, 248–256.Frederiksen, J. R., Warren, B. M., & Roseberry, A. S. (1985). A componential approach to training reading skills:

Part 1. perceptual units training. Cognition and Instruction, 2, 91–130.Fusaro, J. A. (1992). Meta-analysis of the effect of sentence-combining on reading comprehension when the

criterion measure is the test of reading comprehension. Perceptual and Motor Skills, 74, 331–333.Gagne, E. D., Yekovich, C. W., & Yekovich, F. R. (1993). Cognitive psychology of school learning. Boston, MA:

Little Brown and Company.Garner, R. (1982). Efficient text summarization: Cost and benefit. Journal of Educational Research, 75, 275–279.Harp, S. F., & Mayer, R. E. (1998). How deductive details do their damage: A theory of cognitive interest in

science learning. Journal of Educational Psychology, 90, 414–434.Jitendra, J. A., Hoppes, M. K., & Xin, Y. P. (2000). Enhancing main idea comprehension for students with

learning problems: The role of a summarization strategy and self-monitoring instruction. Journal of Special

Education, 34, 127–139.Johnson-Glenberg, M. C. (2000). Training reading comprehension in adequate decoders/poor comprehensions:

Verbal versus visual strategies. Journal of Educational Psychology, 92, 772–782.Just, M. A., & Carpenter, P. A. (1987). The psychology of reading and language comprehension. Boston: Allyn &

Bacon.Kaniel, S., Licht, P., & Bracha, P. (2000). The influence of metacognitive instruction of reading and writing

strategies on positive transfer. Gifted Education International, 15, 45–63.King, A. (1989). Effects of self-questioning training on college student’s comprehension of lectures. Contemporary

Educational Psychology, 14, 366–381.Kintsch, W. (1998). Comprehension: A paradigm of cognition. Boulder: Cambridge University Press.Klinger, J. K., & Vaughn, S. (1996). Reciprocal teaching of reading comprehension strategies for students with

learning disabilities who use English as a second language. Elementary School Journal, 96, 275–293.Ko, H. W. (1999). Reading comprehension screening test. Psychological Testing, 46, 1–14.Labbo, L. D., & Reinking, D. (1999). Negotiating the multiple realities of technology in literacy research and

instruction. Reading Research Quarterly, 34, 478–492.Leong, C. K. (1995). Effects of on-line reading and simultaneous DECtalk auding in helping below-average and

poor readers comprehend and summarize text. Learning Disabilities Quarterly, 18, 101–116.Leu, D. J., Hillinger, M., Loseby, P. H., Balcom, M. L., Dinkin, J., Eckels, M. L., et al. (1998). Grounding the

design of new technologies for literacy and learning in teachers’ instructional needs. In D. Reinking, M. C.McKenna, L. D. Labbo, & R. D. Kieffer (Eds.), Handbook of literacy and technology (pp. 203–220). Mahwah,NJ: Lawrence Erlbaum Associates.

Lin, C. S., & Su, Y. F. (1991). The effect of metacognition training to enhance the metacognitive abilities andreading comprehension of the sixth grade students. Technical Report, NSC78-0301-H003-12. NationalScience Council, ROC.

Lorch, R. F., Jr., & Lorch, E. P. (1996). Effects of organizational signals on free recall of expository text. Journal

of Educational Psychology, 88, 38–48.Lu, L., & Liu, H. S. (1994). The revised Peabody picture-vocabulary test. Taipei: Psychological Press.Lungberg, I. (1995). The computer as a tool for remediation in the education of students with reading disabilities:

A theory based approach. Learning Disabilities Quarterly, 18, 89–100.Lynch, L., Fawcett, A., & Nicolson, R. I. (2000). Computer-assisted reading intervention in a secondary school:

An evaluation study. British Journal of Educational Technology, 31, 333–348.MacArthur, C., & Haynes, J. (1995). Student assistant for learning from text (SALT): A hypermedia reading aid.

Journal of Learning Disabilities, 28, 150–159.MacArthur, C. A., Ferretti, R. P., Okolo, C. M., & Cavalier, A. R. (2001). Technology applications for students

with literacy problems: A critical review. Elementary School Journal, 101, 273–301.


MacArthur, C. A., Graham, S., Schwartz, S. S., & Schafer, W. D. (1995). Evaluation of a writing instruction

model that integrated a process approach, strategy instruction, and word processing. Learning Disability

Quarterly, 18, 278–291.Markman, E. (1979). Realizing that you do not understand: Elementary school children’s awareness of

inconsistencies. Child Development, 50, 643–655.Mastropieri, M. A., & Scruggs, T. E. (1997). Best practices in promoting reading comprehension in students with

learning disabilities: 1976 to 1996. Remedial and Special Education, 18, 197–213.Mathes, P. G., Torgesen, J. K., & Allor, J. H. (2001). The effects of peer-assisted literacy strategies for first-grade

readers with and without additional computer-assisted instruction in phonological awareness. American

Educational Research Journal, 38, 371–410.Mayer, R. E. (1996). Learning strategies for making sense out of expository text: The SOI model for guiding three

cognitive processes in knowledge construction. Educational Psychology Review, 8, 357–371.Montali, P., & Lewandowski, L. (1996). Bimodal reading: Benefits of a talking computer for average and less

skilled readers. Journal of Learning Disabilities, 29, 271–279.Neville, D. D., & Searls, E. F. (1991). A meta-analytic review of the effect of sentence-combining on reading

comprehension. Reading Research and Instruction, 31, 63–76.Novak, J. D., & Gowin, D. B. (1984). Learning how to learn. Cambridge, London: Cambridge University Press.Palincsar, A. S., & Brown, A. L. (1984). Reciprocal teaching of comprehension-monitoring activities. Cognition

and Instruction, 1, 117–175.Paris, S. G., Gross, D. R., & Lipson, M. Y. (1984). Informed strategies for learning: A program to

improve children’s reading awareness and comprehension. Journal of Educational Psychology, 79,1239–1252.

Paris, S. G., & Jacobs, J. E. (1984). The benefits of informed instruction for children’s reading awareness and

comprehension skills. Child Development, 55, 2083–2093.Paris, S. G., & Oka, E. R. (1986). Children’s reading strategies, meta-cognition, and motivation. Developmental

Review, 6, 25–56.Paris, S. G., & Paris, A. H. (2001). Classroom applications of research on self-regulated learning. Educational

Psychologist, 36, 89–101.Pressley, M., Johnson, C. J., Symons, S., McGoldric, J. A., & Kurita, J. A. (1989). Strategies that improve

children’s memory and comprehension of text. Elementary School Journal, 90, 3–32.Reinking, D. (1988). Computer-mediated text and comprehension differences: The role of reading time, reader

preference, and estimation of learning. Reading Research Quarterly, 23, 484–498.Schmitt, M. C. (1987). The effects of an elaborated directed reading activity on the metacomprehension skills of

third grades. Unpublished doctoral dissertation, Purdue University, West Lafayette, IN.Shaughnessy, M. F., & Baker, B. (1988). Learning strategies: Teaching students how to learn (ERIC Document

Reproduction Service No. ED 300 357).Short, E. J., & Ryan, E. B. (1984). Metacognitive difference between skilled and less skilled readers:

Remediating deficits through story grammar and attribution training. Journal of Educational Psychology,

76, 225–235.Sindelar, P. T., Monda, L., & O’Shea, L. (1990). Effects of repeated readings on instructional- and mastery-level

readers. Journal of Educational Research, 83, 220–226.Taylor, B., Harris, L. A., & Pearson, P. D. (1995). Reading difficulties: Instruction and assessment (2nd ed.). New

York: McGraw-Hill.Tregaskes, M. R. (1987). Metacognitive strategies: Their effects on reading comprehension of sixth grade social

studies students. Unpublished doctoral dissertation, Brigham Young University.van Aarle, E. J. M., & van den Bercken, J. H. L. (1999). The development of knowledge-based system supporting

the diagnosis of reading and spelling problems (II). Computers in Human Behavior, 15, 693–712.van Daal, V. H. P., & Reitsma, P. (2000). Computer-assisted learning to read and spell: Results from two pilot

studies. Journal of Research in Reading, 23, 181–193.van Keer, H. (2004). Fostering reading comprehension in fifth grade by explicit instruction in reading strategies

and peer tutoring. British Journal of Educational Psychology, 74, 37–70.Wade, S. E., & Trathen, W. (1989). Effects of self-selected methods on learning. Journal of Educational Psychology,

81, 40–47.Wise, B. (1992). Whole words and decoding for short-term learning: Comparisons on a ‘talking computer’ system.

Journal of Experimental Child Psychology, 54, 147–167.

Documents

Improving children’s reading comprehension and use of ... · PDF fileImproving children’s reading comprehension and use of strategies through computer-based strategy training Yao-Ting