children's metacognition and reading comprehension[1988]

Upload: gbtmatina

Post on 03-Jun-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    1/12

    Journal of Educational Psychology1988,Vol. 80, No. 2, 131-142 Copyright 1988 by the A merican Psychological Association, Inc.0022-O663/88/S0O.75

    Developmental and Instructional Analysesof Children's Metacognition and Reading ComprehensionDavid R. CrossTexas Christian UniversitySc

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    2/12

    132 DAVID R. CROSS AND SCOTT G. PARISmetaco gnition is a complex array of knowledge and skills thatincludes dec larative knowledge (e.g., knowingthatskimmingis a reading strategy), procedural knowledge (e.g., knowinghowto skim), and con ditional knowledge (e.g., knowingwhento skim and why). It also includes the skills required toevaluate, plan, and regulate one's thinking and reading.It has been strongly suggested that metacognition is relatedto ch ildren's reading comprehension (Baker & Brown, 1984;Brown, Armbruster, & Baker, 1984) although there is rela-tively little data on the relation. Some studies have demon-strated an indirect relation between metacognition and com-prehension by showing that metacognitive status varies withage. Myers and Paris (1978), for example, interviewed 8- and12-year-old children about reading and found that older chil-dren were more aware than younger children of the effects ofmany variables on reading and the utility of strategies forcomprehension. In a study of procedural metacognitive skill,Kobasigawa, Ransom, and Holland (1980) asked 10- to 14-year-olds about skimming and found that all their subjectscould describe skimming, but that only the oldest childrencould use skimming as a strategy. Similarly, in a study ofcomprehension monitoring, Harris,Kruithof, Terwogt, andVisser (1981) found t hat 11-year-olds in their study were morelikely than8-year-oldsto identify sentences that were incom-patible with the themes of simple stories.Some studies have demonstrated a direct relation betweencomprehension and m etacognition by showing that metacog-nitive status varies with reading level. Paris and M yers (1981)compared good and poor readers matched for age, sex, andarithmetic achievement and found that good readers knewmore about reading strategies, detected errors more oftenwhile reading, and had better recall of text information. Therole of metac ognition has also been shown in studies designedto improve reading comprehension by increasing children'sawareness and use of strategies. Raphael and Pearson (1985)trained third-, fourth-, and eighth-grade students to distin-guish questions about text according to whether the answerwas available explicitly in the text or had to be generated bythe reader's inference. Trained students were better thanuntrained students at evaluating the types of questions andgiving adequate answers. Thus both awareness and compre-hension improved with metacognitive training.Palincsar and Brown (1984) taught learning-disabled juniorhigh students four strategiessumm arizing, questioning, clar-ifying, and predictingthrough reciprocal teaching. In thisprocedure students were asked to assume the roles of teachersin order to generate questions, summaries, predictions, andclarifications. Teachers and students engaged in considerabledialogue about the content of the texts as well as the value ofthe strategies they were using. Students trained with thesestrategies improved their abilities to answer comprehensionquestions and displayed positive transfer to other classroomtasks. Additional studies have also demonstrated positiveeffects due to training on cognitive strategies and awareness(e.g., Brown, Palincsar, & Arm bruster, 1984; Garn er, Wa-goner, & Smith, 1983).

    Informed Strategies for LearningInaclassroom-based project, Paris andhiscolleagues(Paris,Cross, & Lipson, 1984; Paris & Jacobs, 1984) instructed

    children abou t the existence and use of reading strategies. Ourcomprehension instruction was designed to stimulate greaterawareness of declarative, procedural, and conditional knowl-edge, while also teaching children how to evaluate, plan, andregulate their own comprehension in strategic ways. For thisreason we labeled it Informed Strategies for Learning (ISL).ISL includes several features of direct instruction (see Duffy,Sherman, Roehler, 1977) such as (a) directing children'sattention to the material to be learned, (b) generating highlevels of student involvement, and (c) providing frequentpractice and immediate feedback. In addition, instructionprogressed so that there w as a gradual release of responsibilityfrom the teacher to the student (Pearson & Gallagher, 1983).This was accomplished through (a) modeling of the targetstrategies, (b) guided practice, and (c) indepen dent applicationof strategies. Furtherm ore, we provided rationales for each ofthe strategies so that studen ts would be motivated to use themselectively and independ ently.Results from the ISL project confirm and extend previousresearch on the relation between metacognition and readingcomprehension. Paris and Jacobs (1984) report significantcorrelations between comprehension and reading awarenessfor both third graders,r(S9) =.28,p

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    3/12

    CHILDREN'S METACOGNITION 133membership as a grouping variable in the experimental design(Kareev, 1980). A related advantage is that we will be able todocument the developing relation between reading awarenessand reading performance from the beginning of third gradeto the end of fifth grade. These transitional analyses yieldpowerful statements about the development of reading skillsas a function of age and instruction.

    The goal of clarifying the relation between awareness, stra-tegic reading, and reading comprehension will be achieved inthree steps. First, significant subgroups from the ISL projectwill be identified in the third and fifth grade at both the pretestand posttest. The corresponding cluster profiles will identifypatterns among measures of metacognition and reading per-formance. Second, the longitudinal aspect of the design, in-cluding measurements at the fall pretest and the spring post-test, will allow the analysis of transitions between the clustersidentified in Step 1. Of particular interest is the way in whichthese transitions vary for children in the treatment and controlgroups. The third step will combine the within-grades, longi-tudinal information with the information derived from thebetween-grades, cross-sectional aspect ofth edesign. Thus thedesign resembles a time series with four points: third grade/pretest, third grade/posttest, fifth grade/pretest, and fifthgrade/posttest. This time series of multivariate measurementson reading awareness, strategic reading, and reading compre-hension provides a rare opportunity to observe the developingrelation between metacognition and performance in the do-main of reading.

    Method

    Subjects and SettingThe subjects were 87 third graders (mean age 8 years, 5 months)an d 84fifth graders (meanage = 10years,5 months) from eightintact classes. Two third-grade(n=42) and two fifth-grade (n= 42)classes received training,and the remaining four classes servedascontrols. Separate schools were chosenforexperimental and controlclassesinordertoprevent teachersand students from sharing theirknowledge gained from training w ith the control groups. Asaresult,one third-gradeand onefifth-gradeclass from eachoffour schoolswere randomly assignedtoeitherthetreatmentor thecontrolcon-dition. Prior to assigningthe experimental conditions,theschoolswere matched roughlyondemographicandachievement character-istics. Each classroomhadnearly equal numbersofboysandgirls,and classrooms had similar ethnic representations, approximately65% Caucasian and 35% Black, Asian, and Native American.Measures

    The assessment battery included tasks designed to assess children'sreading comprehension and awareness about reading. Reading com-prehension wasmeasured with the Comprehension subtestof theGates-MacG initie Reading Tests (M acGinitie, 1978). Strategic read-ing was assessed w ithacloze task an danerror detection task, whichrequire childrento usecognitive strategiestosupply missing wordsand to mo nitor m eaning in text. Reading awarenesswasassessed witha structured interview (Paris Jacobs, 1984)and astrategy ratingstask (Paris Myers, 1981). During the interview, children werequestioned about e valuation, planning, and monitoring strategiesforreading. The ratings task required childrentoevaluate strategies thathad either negativeorpositive effects on comprehension.

    Reading comprehension.The Com prehension subtest of the Gates-MacGinitie Reading Tests consistsof a seriesof short paragraph-length stories followed bytwo, three,or four multiple-choice ques-tions designedtoassess children's understandingofeach story. TestLevel C (designed for third-grade students)wasadministered to third-grade classes,andtest LevelD(designedfor fourth- through sixth-grade students) wasadministered tofifth-grade classes. Form 1 ofeach levelwasused at the pretest, and Form 2 was used at the posttest.Each child's extended scale score was derived according to proceduresdescribedinthe teacher's manua l and usedinsubsequent analyses.

    Strategicreading. Four different versionsof the cloze task wereconstructed,one foreach gradeateachofthetwotesting sessions.The passages used to co nstruct the different versions were taken fromthe Classroom Reading Inventory (Silvaroli, 1973). Second-andfourth-grade level passages were usedfor thethird-and fifth-gradeversions, respectively. Withtheexception ofthe firstandlastsen-tencesof the passages, every fifth wordwas deleted from thetext.This created 13 blank spaces that children w ere requiredtofillinwith single words. Each paragraph was typed in large print on asinglepage withatitle and picture.Children's cloze responses were scored accordingto thefollowingprocedure: (a) responses thatwereboth semantically and syntacticallyappropriate to the missing word were awarded 2 points; (b) response sthat were either semantically or syntactically appropriate, but notboth, were awarded1point;(c)blanks and responses thatwereneithersemantically nor syntactically appropriate were awarded 0 points.Semantic appropriatenesswasjudgedinrelation to the text meaning,and syntactic appropriateness was judged in relation to sentenceconstruction. With this procedure,a child's total score could rangefrom0 to26 points.Children's reading was also measuredbytheir abilitiestodetectincomprehensible information intext. Error detectionisoften con-sidered to be an index of comprehension monitoring because childrenare directedtosearch for and to underline anomalous portionsoftext.At both pretest and posttest, children read twopassagesand

    underlined those wordsorsentences thatdid notmake sense. Eachpassage contained two nonsense wordsandtwo clauses with scram-bled word orders. The eight passages were derived from grade-appro-priate storiesin theDiagnostic Reading Scales (Spache, 1972).Pas-sagesforthird graders were 9or10 sentences long; passagesfor fifthgraders ranged from 11to 16 sentences.Each child's score was the numberof errors correctly underlined(i.e.,hits) minus the square root of the number of correct portions oftext underlined (i.e., false alarms). The square root of the false alarmswas used because the distribution was positively skewed. A fewchildren had as many as 30or40 false alarms, although the numberof false alarmsformost children was less than eightthe numberofpossible hits. Although the false alarms were skewed, the distributionof the difference scores formed by subtracting the square root of thefalse alarms from thehitswasnearly norm al. Theerror detection

    scores ranged from -5.5 to 6.3 at thepretest,andfrom 7.0to 6.6at the posttest.Awarenessaboutreading. Children's metacognitive reading skillswere assessed using two different measure s. The R eading Aw arenessInterview was designedto assess children's awareness about readingin three areas: evaluationoftask difficulty and one's own abilities,planning to reachagoal,andmonitoring progress towards the goal.Although the interview contained 33 Likert-scale items and 19 open-ended questions, only 15ofthe open-ended questions w ere usedtoconstruct an interview score. The15questions comprisingthereadingawareness scores are describedinParis and Jacobs (1984).Although the children c ould respond as often as they wished to theopen-ended questions, only their first responses were scored becausefewer than 20% ofthechildren gave more than one response. Theseresponses were assignedtocategories which wereinturn aw arded0,1,or 2 points based on thelevel of conceptual awareness about

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    4/12

    134 DAVID R. CROSS AND SCOTT G. PARISreading (see Paris & Jacobs, 1984, for information about interraterreliability and examples of the scoring system). Zero meant that noresponse was given or tha t the response was inappropriate. One pointwas awarded for appropriate yet nonstrategic responses that reflected(a)the child's use of external aids suchasother people, (b) complexityofthem aterials, or (c) reading activities such as practice and payingattention. Two points were awarded for answers that were evaluative,planful, or strategic.

    Children's awareness about reading was also measured using thestrategy rating task constructed by Paris and Myers (1981). Thestrategies were read aloud to each child individually, and the childrenwere asked to indicate o n a 7-point scale wha t effect each strategy hason reading com prehension, ranging from hurts a lot (I), tomakes nodifference (4), to helps a lot (7). The children were presented withpractice items, and they could indicate their responses on a bar chartspecially constructed to represent the scale alternatives. The task itselfconsists of 25 different strategies or activities that can influencereading. Five items e ach were inter nal positive (e.g., imaging the storylike a picture in m y m ind), internal negative (e.g., saying the wordsover and over), external positive (e.g., underlining the importantparts), external negative (e.g., watching TV while I read), and neutral(e.g., having red hair).

    Inspection of item histograms showed that the children tended touse the end- and midpoints (Le., 1, 4, and 7) more than the otherpoints. Furtherm ore, this tendency was slightly more pronounced forthe poorer readers than for the better readers. In order to eliminatethis bias, and in addition to improve the distribution of the scaleresponses, the 7-point scale was collapsed into a3-pointscale acco rd-ing to the following scheme: 1 = (1 , 2), 2 = ( 3, 4, 5), and 3 = (6, 7).Using the collapsed scale, a strategy rating score was constructed bysubtracting the mean of the 10 negative items (5 internal and 5external) from the mean of the 10 positive items (5 internal and 5external). The obtained score could range from2(if all the negativestrategies were ratedhelpsa lotand the positive strategies were ratedhurts a lot)to +2 (if all the negative strategies were rated hurts a lotand the positive strategies were ratedhelpsa lot).Procedure

    With the exception of the Reading A wareness Interview, all taskswere administered to children in their classrooms as intact groups;the Reading Awareness Interview was administered individually asan oral interview. Of the group tests, the Gates comprehension is atimedtest,but the other tests were unconstrained. Pretests were givenin October, and posttests were administered in April and May, onemonth after the last ISL lessons were taught. The tests were admin-istered by research assistants, who were pro vided with specific instruc-tions regarding the order of tests, timing, and proctoring to ensureuniformity in procedures across testers. After the instructions weregiven, the children worked independently on the Gates comprehen-sion, cloze task, and e rror detection. The items on th e strategy ratingswere read aloud to ensure that students comprehended the task.

    TrainingThe purpose of ISL is to inform children about strategies thatfacilitate reading comprehension. Strategy training was provided forfour months and included three modes of instruction: direct expla-nation in wholeclasslessons, bulletin board m aterials, and suggestionsfor classroom teachers on how to use the strategies. The classroominstruction was the most critical component and involved 30-mingroup lessons taught twice each week by Professor Marjorie Lipson.During these lessons, children were stimulated to think about reading

    strategies bydirect questions and group discussions. The lessons were

    intended to make the children reflect upon various goals of readingand the actions they could take to reach them . Questions, dialogues,analogies, and modeling were all used to stimulate awareness. Thegroup lessons included four characteristics of instructional explana-tion identified by Duffy, Roehler, Meloth, and Vavrus (1986): (a)information sharingisinteractive and responsive;(b)enhanced aware-ness is an instructional objective; (c) declarative, procedural, andconditional knowledge about reading strategies are provided explic-itly; and (d) teachers assist students by explaining things graduallyand sequentially with consideration for the student's perspective.Each lesson also included worksheets that required children to readhigh-interest material and to apply the instructed strategy. Discussionfollowed the worksheets and provided feedback about options forselecting and employing strategies.

    The second vehicle of instruction was the bulletin board thataccompanied each lesson. In order to m ake the lessons concrete andeasy to understand, metaphors were used to represent each strategyor concept. These metaphors were illustrated graphically on largebulletin board displays and were included as themes within eachlesson. For example, the lesson on task evaluation used the me taphor Be a reading detective to examine clues to the passage topic, length,and difficulty. Each bulletin board colorfully illustrated the metaphorand included several focal questions that directed children to thinkabout how, why, and when to apply the strategy. The m etaphors andquestions were incorporated into the worksheets and daily lessons.The third m ode of instruction was a brief description of each m odulethat was given to the classroom teachers. These three-page han doutsincluded descriptions of the strategies, rationales for their importanceto reading, and suggestions for incorporating each one into otherclassroom activities. The purpose of the handouts was to shareknowledge with the teachers and to encourage them to include thestrategies in their own teaching of reading and other content areas.These features of the training remained the same across all ISLmodules and provided a learning set for students and teachers.However, the content of the lessons was organized into a sequence of

    instruction tha t was generally a top-down or conceptual approach toteaching comprehension strategies. Training was divided into threephases that each lasted 5-6 weeks. The first phase addressed thegeneral issue ofawarenessof reading goals, plans, an d strategies. Thesecond phase focused on specific strategies related to comprehendingtext meaning. The final phase emphasized comprehension monitor-ing; children learned about specific strategies for evaluating andregulating their reading. The topics and metaphors for the 14 lessonsare described more completely in Paris et al. (1984).Finally, the control classrooms were not neglected during thetraining period; more time was actually spent in these classroomsthan in the experimental classrooms. The experimenters providedtutoring, showed movies, an d taught group lessons on topics unrelatedto reading. The se activitieswereprovided accordingtothe preferencesof individual teachers so that all children in the study were provided

    with useful educational experiences.

    ResultsThe results for the third- and fifth-grade children are pre-sented separately. For each grade there are three subsections:(a) presentation of the pretest cluster analysis, (b) presentationof the posttest cluster analysis, and (c) analysis of the transi-tions between pre- and posttest clusters. Collectively, thecluster profiles obtained at the four time pointsthird grade/pretest, third grade/posttest, fifth grade/pretest, and fifthgrade/posttestgraphically describe the growth of the readingcomponents measured in this study. The transition analyses

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    5/12

    CHILDREN'S METACOGNITION 135illustrate po tential grou p differences in the effectiveness of theISL instruction.Cluster analysis requires two choices; one must choose analgorithm for performing the analysis and a stopping rule fordetermining the number of clusters. The clustering algorithmused in our research was Ward's method with Euclideandistance (Ward, 1963; Wishart, 1969), and the stopping rule 1was a variation of the scree slope test proposed by Thorndike(1953; Aldenderfer & Blashfield, 1984). We chose W ard'smethod because it has been shown to w prk well in situationsthat are typical of social science research (Aldenderfer &Blashfield, 1984). For exam ple, in conditions w here clustersare not well separated but tend to overlap due to errors ofmeasurement or the use ofscaleswhich are designed to yieldcontinuous distributions, Ward's method has been shown tooutperform most other clustering methods (Bayne, Beau-cham p, Begovich, & Kane, 1980).

    Third-Grade PretestThe profiles for the five groups uncovered by the clusteranalysisareshown in Figure 1. T he variables used to constructthe profile for each group were the Gates-MacGinitie com-prehension subtest (Gates), cloze task (cloze), error detectiontask (ED), reading awareness interview (RAI), and strategyratings (SR). Children's scores on the five measures weretransformed to Z-scores (within grades) so that the meanscould be plotted on a single scale. The five groups in Figure1 were labeled according to the following convention. If theaverage of the three performance measuresGates, cloze,and EDwas more than one-half standard deviation abovethe grand m ean (zero in standard units), then the first portionof the group label wasgood (G). If the average of the threeperformance measureswaswithin one-half standard deviationof the grand mean, then the first portion of the label wasaverage(A); and if the performance average was more thanone-half standard deviation below the grand mean, then thefirst portion waspoor (P). The second portion of the grouplabels was determined similarly using the average of the tworeading awareness measures. It should be noted, however, thatthe cluster analysis does not necessarily identify groups ofsubjects at each of these three levels. Indeed, not all ninecombinations of these factors (3 performance x 3 awareness)were evident at either grade.Each cluster identified in the analysis is shown in Figure 1with levels of performance/awareness indicated. In Figure 1the G/G and G /P groups have above average (> 0.5) readingperformance means, the P/G and P/P groups have belowaverage (< 0.5) reading performance means, and the A/Agroup has performance means arou nd zero. Similarly, the G /G and P/G groups have above average (> 0.5) reading aware-ness means, the G/P and P/P groups have below average (

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    6/12

    136 DAVID R. CROSS AND SCOTT G. PARIS

    1.5-1

    8

    2 0.5 4

    1 . 5

    2Gates Cloze ED RAI

    VariableSR

    ClusterO Good/Good Good/PoorD Avg /Avg Poor/GoodA Poor/Poor

    Figure 1. Cluster profiles for the third-grade pretest. (ED = error detection task; RAI readingawareness interview; SR = strategy ratings; AVG = average.)

    Descriptive statistics are show n in Table 2. As was the caseat the pretest, cluster variability is similar, with the ratios oflargest to smallest standard deviation s ranging from 1.5 forthe error detection to2.3 forthe Gates comprehension subtest.There also appears to be good separation between clusters,withFratios(dfsare 4 and 72) of 17.2 for the Gates compre-hension, 51.4 for the cloze task, 20.1 for the error detection,31.4 for the reading awareness, and 22.7 for the strategyratings.Table 1DescriptiveStatistics for Five C lusters:Third-Grade Pretest

    MeasureGates comprehen-sionM

    SDCloze taskMSDError detectionMSDReading awarenessMSDStrategy ratingsMSDn

    G /G

    524.464.920.83.6

    3.11.522.41.40.940.3214.0

    G / P

    498.868.619.13.4

    1.81.913.83.80.730.2612.0

    ClusterA /A

    452.248.212.16.0

    0.91.819.42. 20.550.2721.0

    P / G

    417.036.9

    5.54.4-0 .81.519.82. 7

    1.250.2311.0

    P / P

    327.651.6

    3.22. 2-2 .01.314.73.60.680.4214.0

    Note.G = good; P = poor; A = average.

    TransitionsIn this section, we investigate the patterns of transitionsbetween pretest cluster mem bership and po sttest cluster mem -bership as a function of the ISL intervention. Transition

    frequencies are shown separately for the experimental andcontrol groups in Tables 3 and 4. The entries in these tablesare too sparse for comparison of the table interiors using asignificance test, but it is possible to com pare th e p retest andposttest marginal distributions, which indirectly tests for thetreatm ent main effect. U singachi-square test of homogeneity,the distribution of pretest cluster marginals are not signifi-cantly different for the experimental and control groups, % 2(4,N = 64) = 1.47, p > .80. However, there is a significantdifference between the pos ttest cluster marginals, x 2 (4 ,N -64) = 14.12,p

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    7/12

    CHILDREN'S METACOGNITION 137

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    8/12

    13 8 DAVID R. CROSS AND SCOTT G. PARISTable 4Transition Frequencies Between Pretest andPosttestClusters: Third-Grade Control

    Pretest clusterPosttestcluster G/G G/P A/A P/G P/P TotalG /GG/AA /PP /GP /PTotal

    0 16 21 40 00 0

    0 04 13 00 14 211 4

    113911135

    Note.G = good; P = poor; A = average.

    The means and standard deviations for the four clusters arepresented in Table 5. The four clusters are comparable interms of within-clusters variation, with the ratios of largest tosmallest standard deviation s ranging from 1.4 for the errordetection to 2.4 for the reading awareness interview. Theclustersalsoappear to be d istinct, as indicated by the between-clustersF ratios (3 and 59 degrees of freedom): 40.2 for theGates comprehension, 37.2 for the cloze task, 18.4 for theerror detection , 14.2 for the reading awareness, and 14.8 forthe strategy ratings.Fifth-GradePosttest

    The cluster profiles obtained at the fifth-grade posttest areshown in Figure 4. Although similar to those found at thethird-grade posttest, the profiles in Figure 4 do not reveal theextreme incongruencies present in the third grade (i.e., thereis no P/G cluster). In addition, whereas there were four

    Table 5Descriptive Statistics for Four Clusters: Fifth-Grade Pretest

    ClusterMeasure G/G A/G A/P P/A

    Gates comprehen-sionM 592.5 493.3 533.0 453.7SD 39.0 28.9 53.0 41.7Cloze taskM 19.1 15.4 19.4 7.7SD 4,0 2.1 2.2 5.0Error detectionM 2.5 1.4 0.5 -1 .6SD 2.1 1.9 1.9 1.5Reading awarenessM 23.4 22.2 18.1 18.6SD 1.6 2.9 3.9 3.0Strategy ratingsM 1.09 1.32 0.52 0.87SD 0.20 0.31 0.26 0.37n 20.0 9.0 12.0 23.0Note.G = good; A = average; P = poor.

    clusters at the fifth-grad e pretest, there are now five clusters.Because there were five clusters obtained in all of the clusteranalyses except the fifth-grade pretest, and because there weresubstantially fewer cases with complete da ta at the fifth-gradepretest, our finding four clusters as opposed to five at thattime point is probably not of substantive interest, but merelya consequ ence of our difficulty in collecting comp lete data onall of the fifth graders at the fall pretest.The means and standard deviations for the five posttestclusters are presented in T able6.The variation w ithin clustersis again com parable across th e five clusters, with the ratios of

    ClusterO Good/Good Avg./GoodD Avg./Poor Poor/Avg.

    Gates Cloze ED RAIVariable

    SR

    Figure3. Ou ster profiles for thefifth-gradepretest. (See Figure1 for explanation of abbreviations.)

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    9/12

    CHILDREN'S METACOGNITION 139

    ClusterO Good/Avg. Avg./GoodD Avg./Poor Poor/Avg.A Poor/Poor

    Gates Cloze EDRVariable

    Figure4. Cluster profiles for thefifth-gradeposttest. (See Figure1orexplanation of abbreviations.)

    the largest to smallest standard deviations ranging from 2.1for the reading awareness to 3.3 for the cloze task. Further-more, it is again the case that between-clusters variability islarge relative to within-clusters variability. The F ratios are23.0 for the Gates comprehension, 49.3 for the cloze task,36.1 for the error detection, 22.5 for the reading awareness,and 27.5 for the strategy ratings.Transitions

    The fifth-grade transition frequencies are shown separatelyfor the experimental and control groups in Tables 7 and 8.Table 6DescriptiveStatistics for Five C lusters:Fifth-Grade Posttest

    Cluster

    As before, the entries in these two tables are too sparse forcomparison of the table interiors using a significance test, sothe pre- and posttest marginal distributions are comparedusing thechi-square test of homog eneity. There isasignificanttreatment group difference in the pretest cluster marginaldistributions, x2(3 ,N = 60) = 12.57,p .20,although there are still more experimental children in the A/G cluster and slightly mo re control children in the P /P cluster.In contrast to the third grade, where the experimental group

    MeasureGates comprehen-

    sionMSDCloze taskMError detectionMReading awarenessMSDStrategy ratingsMSDn

    G / A

    586.242.023.9

    i ?i /4.11.2

    24.61.70.850.1923.0

    A /G

    530.949.222.9

    Z D

    3.21.724.21.6

    1.370.2119.0

    A / P

    494.932.419.6

    0.91.120.72. 80.650.2514.0

    P /A

    436.872.415.3

    7J t

    -0 .90.823.01.3

    1.280.138.0

    P / P

    431.671.28.2

    - 1 . 52.418.42.10.740.389.0

    the posttest,

    Table 7

    the effect of the treatment

    Transition Frequencies Between PretestClusters:Fifth-GradeE xperimentalPosttestcluster

    G /AA / GA / PP /AP/PTotal

    G/G340007

    Pretest clusterA / G

    52300

    10

    A / P030003

    IIUVl\iVU *U11WUUMin the fifth grade was

    an dPosttest

    P / A06232

    13

    Total8

    1553233

    Note.G = good; A = average; Ppoor. Note.G = good; A = average; P = p oor.

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    10/12

    14 DAVID R. CROSS AND SCOTT G. PARISTable 8TransitionFrequencies Between Pretesta ndPosttestClusters: FifthGrade Control

    PosttestclusterG/AA/GA /PP/AP/PTotal

    G / G83100

    12

    Pretest clusterA / G

    000000

    A / P121116

    P /A111249

    Total10633527

    Note.G = good; A = average; P = poor.

    to bring the experimental group up to the same level of skillas the control group, and perhaps a little bit ahead.A more precise evaluation of the treatment effect in thefifth grade can be made by comparing the columns of Tables7 and 8. The most d ramatic treatme nt group difference occursfor pretest cluster P/A, where 8 of 13 experimental childrenin this pretest cluster moved into the posttest A/G and A/Pgroups, whereas 6 of9control children remained in P/A andP/P groups at the posttest. There is also the suggestion of atreatment difference for the children in pretest cluster A/P,where all three of the experimental children moved intoposttest cluster A/P, but the control group children werealmost evenly divided among the five posttest clusters. Nocomparison can be made in regard to pretest cluster A/G,because there were no co ntrol children in this cluster. Finally,there was a greater tendency for the control group children inpretest cluster G/A to belong to posttest cluster G/A asopposed to posttest cluster A/G, whereas the experimentalchildren were evenly split among these two posttest clusters.It appears tha t in the fifth grade,the instruction haditsgreatesteffect on the poorest readers (i.e., pretest cluster P/A), andless of an effect on the better reade rs.

    DiscussionThis instructional study was designed to investigate therelation between children's reading awareness and readingperforman ce. C luster analysis was used to identify significantsubgroups of children with markedly different profiles ofreading skill. These analyses were conducted at four timepoints: third-grade pretest (fall), third-grade po sttest (spring),fifth-grade pretest (fall), and fifth-grade posttest (spring). Be-cause of the temporal nature of this design, developmentaltrends in the cluster profiles across the four time points werereadily discernible. It was also possible to analyze transitionfrequencies between pretest and posttest within each grade.These transition frequencies were useful for identifying apti-tude-by-treatment interactions in response to the ISL training.Three features of the results deserve special attention. First,there is the incongruence found between reading awarenessand reading performance at the third-grade pretest. Two ofthe clusters were either high on performance and low on

    awareness (G/P ), or low on performance and high on aware-

    ness (P/G). The discrepancy between the performance andawareness profiles in these two groups is largea differenceof about one standard deviationand these two groups rep-resent nearly one third of the third-grade sample. In terms ofthe magnitude of the differences and the number of childreninvolved, the finding of incongruence between reading per-formance and reading awareness is not a trivial result. Thisresult is consistent with modest correlations between meta-cognition and performance observed by others for youngchildren (Garner, 1987).A second feature ofthesedata is the increasing congruencebetween awareness and performance observed between 8- to10-year-olds. To our knowledge, there are no other longitu-dinal or time-series data on the patterns of relations betweenmetacognition and reading performance. Our data revealstronger correlations among measures of metacognition andmeasures of reading comprehension from third to fifth grade(Paris Jacobs, 1984). Even more compelling, though, is theincreasing congruence on multivariate measures that emergesfrom the beginning of third grade to the end of fifth grade.Among the 84 fifth graders, there was no group of childrenwho exhibited w idely divergent pattern s of reading awarenessand comprehension. Thus it appears that children integratetheir unders tanding a bout reading strategies with their readingperformance throughout Grades 3, 4, and 5. However, con-gruence between metacognition and performance may alsoincrease among both younger and older children on otherstrategies that are appropriate for beginning or advancedreaders.

    A third highlight of the results is the treatment groupdifferences in the distribution of pre- to posttest transitions.Clearly, direct explan ations ab out reading strategies increasedchildren's awareness an duse of the strategies for most chil-dren. In the third grade, the instruction benefited all of thepretest groups except the pretest P/P cluster. More important,children in the pretest P/G cluster benefited from the instruc-tion, whereas children in the P/P cluster did not. Either someinitial reading awareness is necessary for less skilled readersto improve their reading performance in response to instruc-tion such as ISL or more intervention is needed to have aneffect. In th e fifth grade, it wasonlythe children in the pretestP/A group w ho gained dramatically as a result of ISL instruc-tion. Because many of the experimental ch ildren in the pretestP/A group moved into the A/G group at the posttest, themetacognitive instruction improved their reading perform-ance skills as well as their reading awareness skills.Wh at are the instructional im plications of this research? Tobegin w ith, these data corroborate other training studies (e.g.,Bereiter Bird,1985;Duffy, Roehler, M eloth, Vavrus, Book,et al., 1986; Palincsar & Brown, 1984) and lend support tothe important role attributed to metacognition in theories oflearning and development (Baker & Brown, 1984; Paris &Lindauer, 1982). The pragmatic value of this project is thedemonstration that group instruction can be used to informchildren about reading strategies and, more important, thatchildren can be co nvinced to use these strategies on their ow n.Children in the experimental classes learned declarative, pro-cedural, and conditional knowledge about reading, and theirincreased performance on strategic reading tasks suggests that

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    11/12

  • 8/12/2019 Children's Metacognition and Reading Comprehension[1988]

    12/12

    14 2 DAVID R. CROSS AND SCOTT G. PARISabout reading.Journal ofEducationalPsychology, 70 , 680-690.Palincsar, A. M., & Brown, A. L. (1984). Reciprocal teaching ofcomprehension-fostering and comprehension-monitoring activi-ties.Cognitionand Instruction, 1,117-175.Paris, S. G., Cross, D. R., & Lipson, M. E. (1984). Informed strategiesfor learning: A program to improve children's reading awarenessand comprehension.Journal ofEducationalPsychology, 76 , 1239-1252.Paris, S. G., & Jacobs, J. E . (1984). The benefits of informed instruc-tion for children's reading awareness and comprehension skills.ChildDevelopment, 55 ,2083-2093.Paris, S. G., Und auer, B. K. (1982). The development of cognitiveskills during childhood. In B. Wolman (Ed.),Handbook ofdevel-opmentalpsychology (pp. 333-349). Englewood Cliffs, NJ: Pren-tice-Hall.Paris, S. G., Lipson, M. Y., & Wixson, K. K. (1983). Becoming astrategic reader. Contemporary EducationalPsychology, 8, 293-316.Paris, S. G., & Myers, M. (1981). Comprehension m onitoring, mem-ory, and study strategies of good and poor readers. Journal ofReadingBehavior, 13 ,5-22.Paris, S. G., & Oka, E. R. (1986). Children's reading strategies,metacognition, and motivation.DevelopmentalReview, 6,25-56.Pearson, P. D., & Gallagher, M. S. (1983). The instruction of readingcomprehension. Contemporary EducationalPsychology, 8,317344.Raphael, T. E., & Pearson, P. D . (1985). Increasing students' aware-ness of sources of information for answering questions. AmericanEducational ResearchJournal, 22 ,217-236.

    Resnick, L. (1981). Instructional psychology.AnnualReviewof Psy-chology, 32 ,659-704.Ryan, E. B. (1981). Identifying and remediating failures in readingcomprehension: Toward an instructional approach for poor com-prehenders. In T. G. W aller .G. E. MacKinnon (Eds.),Advancesinreading research(Vol. 3, pp. 223 -261). New York: Academic.Samuels,S.J., Kamil, M . L. (1984). Models of the reading process.In P. D . Pearson (Ed.),Handbook ofreadingresearch. New York:Longman.Silvaroli, N. J. (1973). Classroom readinginventory (2nd ed.). Du-buque, IA: William C. Brown Co.Spache, G. D . (1972).DiagnosticReadingScales(rev.ed.).Monterey,CA: CTB/McGraw-Hill.Stanovich, K. E. (1980). Toward an interactive-compensatory modelof individual differences in th e developm ent of reading flue ncy.ReadingResearchQuarterly, 16 ,32-71.Thorndike, R. L. (1953). Who belongs in a family? Psychometrika,18 , 267-276.Tufte, E. R. (1983). The visual display ofquantitative information.Cheshire, CT : Graphics Press.Ward, J. (1963). Hierarchical grouping to optimize an objective

    function Journal ofthe AmericanStatistical Association, 58,236-244.Wishart, D. (1969). An algorithm for hierarchical classifications.Biometrics, 25, 165-170.Received April 1 1, 1986Revision received July 27, 1987Accepted Novem ber 3, 1987