This article was downloaded by: [Northeastern University]On: 17 October 2014, At: 08:24Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Scientific Studies of ReadingPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/hssr20

Exploration of theDevelopmental ComponentsContributing to ElementarySchool Children's ReadingComprehensionWilliam H. Rupley , Victor L. Willson & WilliamD. NicholsPublished online: 19 Nov 2009.

To cite this article: William H. Rupley , Victor L. Willson & William D. Nichols(1998) Exploration of the Developmental Components Contributing to ElementarySchool Children's Reading Comprehension, Scientific Studies of Reading, 2:2,143-158, DOI: 10.1207/s1532799xssr0202_3

To link to this article: http://dx.doi.org/10.1207/s1532799xssr0202_3

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views ofthe authors, and are not the views of or endorsed by Taylor & Francis.The accuracy of the Content should not be relied upon and should beindependently verified with primary sources of information. Taylor andFrancis shall not be liable for any losses, actions, claims, proceedings,demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, inrelation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

SCIENTIFIC STUDIES OF READING, 2(2), 143-158 Copyright O 1998, Lawrence Erlbaum Associates, Inc.

Exploration of the Developmental Components Contributing to

Elementary School Children's Reading Comprehension

William H. Rupley and Victor L. Willson Texas A&M University

William D. Nichols University of North Carolina at Charlotte

The purpose of this study was to compare rauding theory (Carver, 1995) to our own evolving model of reading acquisition, which supports stage and phase theories of reading development. The relations among rauding variables--cognitive power, auditory-accuracy level, word-recognition level, comprehension-accuracy level, reading-rate level, and reading-comprehension-rate level-were examined using structural equation modeling. The Kaufman Assessment Battery for Children and the Kaufman Test of Educational Achievement subtests were used to operationalize the various constructs. Carver's (1993) model was assessed at each grade in two parts, with overlapping paths allowing cross-validation of some coefficients.

For Grades 1 and 2, the fit indices were above .95, indicating a good model fit. One additional path was supported-from word recognition to reading com- prehension rate. The fit indices for Grades 3 and 4 were above .95, supporting the modified Carver model. Analyses for Grades 5 and 6 produced fit indices above .90, indicating drops in the level of association among the variables compared to earlier grades.

The results of this study offer support to Carver's (1993) rauding theory and further advance the theories that children go through stages or phases of reading develop- ment. Although word recognition is a notable component of reading development throughout the elementary grades, its contribution to comprehension begins to lessen

Requests for reprints should be sent to William H. Rupley, Educational Curriculum and Instruction (EDCI), College of Education, Texas A&M University, College Station, TX 778434232, E-mail: w-rupley @tamu.edu

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

144 RUPLEY, WILLSON, NICHOLS

at Grades 5 and 6 . There is, however, an increase with grade of the influence of cognitive power on reading comprehension, which is greater for crystallized (Pe- abody Picture Vocabulary Test-Revised) intelligence than for global (fluid and crystallized) or fluid intelligence.

Recent research suggests that, as children learn to read, they go through several distinct stages (Adams, 1990; Chall, 1996; Hoover & Gough, 1990; Stanovich, 1991; Willson & Rupley, 1993) or phases (Ehri, 1992; Ehri & Wilce, 1987) ranging from those associated with logos and environmental print to those represented by comprehension and acquisition of new knowledge. Research on these stages and phases has resulted in several models and evolutionary descriptions of reading acquisition, and, although these are not in complete agreement, they do share many similarities. For example, Chall's (1996) stages of reading development advance from Stage 0 through Stage 5. Stage 0 (prereading) coincides with the selective-as- sociation and logographic or visual cue-reading phases identified by Ehri (1992). Both of these conceptualizations of emerging reading represent the reader as one who knows little about the letter-sound system and who relies heavily on the context of print within the environment to recognize words (Adams, 1990; Ehri, 1991 ; Gough & Hillinger, 1980).

Movement from dependence on contextualized print to attention to interacting with traditional orthography occurs when readers begin to recognize the shapes, names, and sounds of letters and acquire a low level of phonemic awareness. This capability is characterized by the reader's detecting the separate sounds in the pronunciation of words and also in letter names (Chall, 1996; Ehri & Wilce, 1987; Mason, 1984). Conceptualizing and reasoning about graphemic-phonemic rela- tions of language represent children's advancement toward reading acquisition and word recognition, which facilitate and contribute significantly to comprehension and learning from text. Biemiller (1970) described these readers as beginning to exhibit a continued concern with graphic exactness. Ehri (1992), Juel (1991), and Mason (1980) referred to this increasing attention to text as the cipher stage, which is achieved when readers use their phonetic-segmentation and phonological-recod- ing skills to form complete connections that secure the entire spelling of the word in memory as a visual symbol for phonemic units in its pronunciation (Ehri, 1992; Gough, Juel, & Griffith, 1992). Common to readers at this developmental level is the ability to process words by setting up connections between letters and sequences of letters stored in memory with phonemic constituents in the pronunciation of the words.

A large body of research indicates that, in order for children to become successful in word recognition, they must successfully pass through these early stages or phases of reading. Successful progression enables readers to acquire the ability to gain word recognition automaticity, which is major factor contributing to compre- hension at the elementary school levels (Carver, 1995; Stanovich, 1991; Willson

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 45

& Rupley, 1993). Word recognition is a significant variable for children's success- ful reading comprehension in Grades 1 through 3 due to the fact that, without the ability to decode, young readers are limited in their comprehension.

Although variables other than word recognition have been investigated to better understand what contributes to children's reading, the contribution of these to a stage model of reading is still unclear. For example, from previous research (Paris, Wasik, & Turner, 1991), prior knowledge has been shown to be a significant factor in reading comprehension. Prior knowledge and background knowledge are also fundamental to various theories, such as schema theory (Anderson, Reynolds, Schallert, & Goetz, 1977), strategy-domain interactive theory (Alexander & Judy, X988), and dual coding (Sadoski, Paivio, & Goetz, 1991). However, other research conducted on prior knowledge and strategic knowledge (Miller, 1987; Paris & Meyers, 1981; Paris et al., 1991) has concluded that use of prior knowledge and strategic knowledge is limited in young children.

Some individuals (Daneman, 1991; Hoover & Gough, 1990) have argued for a simple view of reading that includes only the components of decoding and com- prehension. In this simple view, decoding is the ability to recognize and match the printed word with its mental lexicon and derive meaning at the word level, and reading comprehension is equated with linguistic comprehension, which is medi- ated through graphically represented information.

Research findings (Carver, 1993; Rupley & Willson, 1997) have challenged the idea of the simple view of reading. Such a simple view is too simple because it does not account for other salient features of reading acquisition, particularly for reading development of elementary school children. Carver has developed a theory of reading he called rauding theory. The focus of rauding theory (Carver, 1993,1995) is on "general reading ability, which can be measured by traditional standardized reading comprehension tests because they are affected by both accuracy level and rate level" (1993, p. 453). Carver (1992) discounted the roles of prior knowledge, prediction, and text type and maintained that these are trivial factors and can be disregarded when reading narrative text. In recent research, Carver (1993, 1995) a~ffirmed that the simple view of reading (including decoding and comprehension) can be merged with rauding theory to provide a theoretical representation of reading that is applicable for beginning through adult readers.

According to rauding theory, rauding-efficiency level (EL), the ability to com- prehend efficiently while reading materials at an appropriate level of difficulty for the reader, is affected by two factors-rauding-accuracy level (AL), the reading level or the level of ability to accurately comprehend while reading, and rauding-rate level (RL), the rate of reading typically associated with recreational reading of text or simple words. Causally prior components of rauding theory include (a) cognitive power (CP), an intelligence factor reflective of working-memory capacity that exerts an influence on yearly gains in rauding accuracy, (b) cognitive speed (CS), similar to naming speed and the primary factor influencing reading-rate level (RL),

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

146 RUPLEY, WILLSON, NICHOLS

and (c) auditory-accuracy level (Au~AL), a representation of listening comprehen- sion. Pronunciation level (PL) was added to a revised version of the model (Carver, 1993), which can be viewed as word recognition or the ability to use gra- phemic-phonemic relations in the pronunciation of words. PL is postulated to affect both AL and RL and is not directly influenced by either Cp or Au~AL; thus, it is an exogenous component of the reading process.

Rauding theory has several features that are similar to our past research efforts associated with elementary school children's stages of reading development. We have found that word recognition accounts for a significant portion of the variance associated with elementary school students' reading comprehension-R' values ranging from .90 for Grades 1 and 2 to .55 for Grades 5 and 6 (Rupley & Willson, 1997). However, we had not considered the rate variables for either word recogni- tion or comprehension. Furthermore, we had not included the variables of Cp and Cs in our previous investigations of factors contributing to children's' reading comprehension. Using Carver's construct for reading suggests that measures asso- ciated with rate, cognition, and accuracy would account for most of the variance in children's reading comprehension of narrative text.

The purpose of the present study was to theoretically model an adapted repre- sentation of rauding theory built on an assimilation of our previous research. We wanted to advance our knowledge about stage models that have emerged in our past inquiries by expanding our consideration of variables to include cognitive, rate, and auditory components. Specifically, we were interested in applying features of rauding theory in exploring children's acquisition of reading capabilities from the perspective of stages and phases of reading development. In addition, we wanted to compare our adapted model of rauding theory to the relations obtained for each data set across Grades 1 through 6 for Carver's rauding theory.

Two changes and one omission were made to Carver's original model. We assumed that Cp and AU&L should directly affect PL based on the research cited earlier relating IQ to word recognition and reading comprehension. Although we would predict that the path from Cs to AL would disappear with this change, we left the C ~ A L path in because it had been proposed as a primary mechanism (Carver, 1993). The omission was the lack of a suitable measure of Cs. As its only theoretical influence is on Rh and it was not included in Carver's (1993) diagnostic package, we felt this was not a critical part of testing his model.

The modified theoretical model is shown in Figure 1. This model is identical to that presented by Carver (1995) except for the two added paths.

We deliberately did not construct measures identical to those used by Carver. A major purpose of our investigation was to determine if Carver's model was robust with respect to the constructs that he proposed. That is, Carver's model is of little use if the only application is with the specific variables measured exactly as he originally defined them. A general theory should reasonably fit different operationalizations of the same constructs, and the departures in model

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 47

Carver's Model

This path is not in Carver's I

Word Recognitio '1' a-cg> Reading Rate

FIGURE 1 Modified theoretical rauding model in structural form.

fit may give indications of the limitations of the model application. As reading researchers use a great number of different measures all called reading compre- hension, word identification, and the like, Carver's model should be general enough to account for the various findings established for them, even though hie has argued that rauding is an extension of theories such as simple view and does not fit into them. We believed that we had sufficient data and variables to nnodel rauding theory and advance the credibility of stagewise perspectives of reading development. If the model were supported by our representation, it would provide evidence for both the robustness of Carver's model and the role that the specified variables have in reading development at elementary school levels. If the model were not supported, the conclusion would be problematic because the result might be attributable to inappropriate operationalization rather than inappropriate modeling.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

148 RUPLEY, WILLSON, NICHOLS

One difficulty we encountered in attempting to model rauding theory was that we did not have adequate variables in a single data set to represent all constructs, but they all were available in two different data sets that had identical variables for parts of the model. It was not coincidental that the exogenous and intermediate variables were available in an individual intelligence test data set, and the interme- diate and dependent endogenous variables were available in a school achievement test battery data set. Thus, the model was evaluated with two separate analyses that estimated the contributions of the variables to their respective parts, with the availability of overlap in evaluation of the relations among Cp, PL, and AL thus providing internal evidence of model consistency based on two independent samples.

METHOD

Participants

The standardization samples for the Kaufman Assessment Battery for Children (KABC; Kaufman & Kaufman, 1983) and the Kaufman Test of Educational Achievement (KTEA; Kaufman & Kaufman, 1985) were used as the data sources for this study. There were 1,085 students-541 took the KABC (181 in Grades 1 & 2; 176 in Grades 3 & 4; and 184 in Grades 5 & 6), and 544 took the KTEA (183 in Grades 1 & 2; 177 in Grades 3 & 4; and 184 in Grades 5 & 6). These elementary school students were selected to reasonably mirror the 1980 census.

Instrumentation

We used KABC variables to operationalize A u ~ A L , CP, PL. and AL. A u ~ A L was defined by the KTEA Riddles subtest. Reynolds (personal communication, October 3, 1996) indicated that the Riddles subtest most closely assessed listening ability. The student is asked to name a concrete or abstract concept when given a series of characteristics orally. Thirty-two items are used, and split-half reliabilities for the ages studied were all over .85.

The KABC Mental Processing score (total IQ score) defined CP. Carver has suggested that Cp is a fluid intelligence construct such as measured by Raven's Progressive Matrices (RPM). We suggest that a more global intelligence test should functionally define Cp as well. This score is based on eight subtests individually administered; it functions much as other major individual IQ tests with comparable reliability and validity, with split-half reliabilities over .95 for the ages studied.

PL is virtually identical with most word-identification tests. The KABC Reading Decoding subtest (an achievement test) presents words to the student to pronounce.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 49

Words were selected from more than 10,000 samples of texts at Grades 3 to 9. Word frequency and linguistic features were used to select the 38 words used in the test. Split-half reliabilities were all over .90 for all ages investigated.

AL was assessed from the KABC Reading Understanding subtest. In this subtest, children are presented, on an easel, a series of tasks or commands that they must read and then perform. This format is quite different from the typical reading-com- prehension test and minimizes prior knowledge and verbal fluency. Because children read at their own rate on this test, and because this test has only easy words, it reasonably fits Carver's rauding-accuracy level. Split-half reliabilities were all over .90 for the ages studied.

The KTEA was used as the second data source for this investigation. Identically developed subtests, Reading Decoding and Reading Understanding (comprehen- sion), were available. (For each age grouping used, Reading Decoding split-half reliability was greater than .95, and Reading Understanding split-half reliability was greater than .90.) The KTEA is individually administered and provides norms from age 6 years to age 18 years, 11 months. In addition, the testing time for each participant on each subtest was available. This allowed us to construct measures similar to Carver's RL and EL by dividing the number of items correct on each subtest by the amount of testing time. Although these measures are clearly not identical to Carver's (1993) formulation, they do represent rate measures of decoding simple words (RL is the rate of comprehending simple text) and reading efficiency (the rnost difficult material that is comprehended at a rate comparable to the material). We considered alternative formulations (e.g., using our EL measure instead as a rauding-rate level); however, the intent of EL seems to be better framed as a working definition of efficiency-that is, how well comprehension occurs per unit of time. Because the Reading Understanding subtest assesses maximum difficulty of com- prehension of actual text rather than words, it theoretically better represents the component of rauding efficiency.

The KTEA Reading Understanding subtest uses paragraphs of varying lengths tlo sample comprehension. Students respond orally to literal and inferential ques- tions that are asked following the reading of each passage. Students are allowed to respond orally or gesturally depending on the difficulty of the passage.

For the KTEA analysis, to assess Cp, we used the Peabody Picture Vocabulary Test-Revised (PPVT-R; Dunn & Dunn, 1981). The PPVT-R is a power test of listening vocabulary. It has a median correlation of .62 with the Stanford-Binet Intelligence Scale, .58 with the KABC Mental Processing score, and .64 with the Wechsler Intelligence Scale for Children. The test is administered individually, and tlhe student selects one of four pictures that best represents the meaning of the stimulus word. It has a split-half reliability median of .8 1 for school-age children and a delayed-retest alternate-forms reliability coefficient median of .77. Based on our previous research and on our theoretical modeling of rauding theory research findings, we considered the PPVT-R to be a valid measure of cognitive power,

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

150 RUPLEY, WILLSON, NICHOLS

particularly for verbal, crystallized components of intelligence. As both the more global KABC Mental Processing score and the PPVT-R score were available for investigating the same relations, it allowed us to compare their relative utility for rauding theory. Carver (1993) suggested that fluid intelligence as represented by RPM most directly corresponds to Cp. Both measures used here have greater verbal, crystallized intelligence components than the RPM. A direct comparison using the KABC Simultaneous Processing score was available, but preliminary analyses indicated virtually no difference in the models investigated. We discuss this point later.

Analyses

The methodology chosen for this study is based on structural equation modeling (SEM), also termed LISREL. This methodology represents the most technologically advanced quantitative approach yet developed, including measurement theory, factor analysis, path modeling of underlying factors, and linear modeling of relations. SEM is a statistical method that subsumes both regression analysis and measurement theory (Bollen, 1989). It has been developed over the last 30 years to become the most general method for quantitative analysis in the social sciences. The conceptual origin of SEM lies in path analysis developed by Wright (see Bollen, 1989). Under different restrictions, SEM can be shown to reduce to multivariate and univariate analyses of variance (ANOVAs), canonical correlation, and multiple regression. It differs in approach from these methods in that the focus of SEM is on covariance between variables. All parameters are estimated simultaneously, and each is tested with a t-statistic. In addition, various tests are available for the complete model hypothesized--called goodness-of-fit indices (GFIs). A test of either removing a parameter (giving it a zero value) or freeing a parameter that has been fixed (usually to be zero) is available in SEM and is termed a modification index. It has associated with it a chi-square statistic. This is extremely helpful in exploratory model development.

As regression and ANOVA are embedded in SEM, one might ask why one should bother with SEM. The answer is that models more complex than are permitted with regression or ANOVA can be hypothesized and tested. The inclusion of the measurement model alone-the basis for reliability and generalizability theory-makes SEM superior, as multiple indicators of a construct can be formally represented in a model as factors (the term as used in test theory). Also, the representation of variables simultaneously as causally prior (independent) to some others and causally dependent on yet others is simply not possible in ANOVA, regression, or their multivariate generalizations.

SEM has some limitations. Notably, it is not a small-sample method. Although statisticians currently argue the lower bound for reasonable estimation and hypothe-

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 5 1

sis testing, sample sizes of at least 100 or 200 are advocated. Also, the number of parameters to be estimated is a serious consideration, as both variances and covariances for all variables are potential parameters to be estimated. That is, any relation between two variables can be represented as a covariance. Error variances must also be accounted for. The problem is identical with the degree-of-freedom problem in regression: There can be no more predictors than data points available (less the intercept). Finally, SEM estimation in general uses the maximum-likeli- hood method, which is computationally complex, requiring more complex com- puter packages. Fortunately, such programs are now much more available and, for example, are included in packages such as SPSS and SAS.

All analyses used here were manifest-variable path models, as the factor struc- tures of the subtests were assumed to follow those represented in the technical manuals of the KABC and KTEA. Based on the model presented in Figure 1, we examined the path models for Grades 1 and 2, 3 and 4, and 5 and 6 in order to explore the application of the constructs of rauding theory to each of these grade levels and to determine the fit of each model to our theoretical model. It should be noted that a covariance of zero was assumed between Cp andAudAL consistent with Carver (1993). A liberal alpha level was applied to testing hypothesized paths (.05), ,whereas a stringent criterion ( p < .001) was applied to add a new path using Lagrange multiplier tests (Bollen, 1989). Two GFIs were exarnined-the adjusted (GFl (AGFI) and the normed fit index (NF1)-based on Tanaka's (1993) discussion of criteria for use and of various studies of their robustness across various model assumptions. In general, the subtests met well the multivariate normality assump- tion underlying SEM analysis.

RESULTS

A great deal of statistical output is typically available from the program we used-PROC CALIS (SAS Institute Inc., 1989)-so only relevant aspects are discussed here. As measures are comparable across the two analyses performed at each pair of grades, covariances are shown in the figures accompanying the analyses. Significance levels are all beyond p < .05 for hypothesized paths and are beyond p < .001 for added paths. In cases in which the path is not significant, the coefficient value is followed by a notation of ns.

As all data are the normative samples for the national testing, means and standard deviations are unimportant, but it is important to note that they are the reported means for the tests (M = 50, SD = 10) except for the Mental Processing score (M = 90, SD = 13), which was adjusted for AudAb For Grades 1 and 2, the rate measures are RL (M = 4.73, SD = 3.94) and EL (M = .163, SD = .185); for Grades 3 and 4, they are RL (M = 7.77, SD = 5.09) and EL (M = .326, SD = .199); for Grades 5 and 6, they areRL(M=9.01, SD=5.17) and EL(M= .375, SD= .190).

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

152 RUPLEY, WILLSON, NICHOLS

Grades 1 and 2

The two analyses are shown in Figure 2, with KABC path coefficients first and KTEA coefficients second, for the overlapping analyses. The theoretical model based on rauding theory had AGFIs over .90 and NFls over .96. One path (from PL to EL) was indicated to be added, with changes of AGFI to ,951 and NFI to .992. At Grades 1 and 2, the path from CP to PL was significant for the KABC Mental Processing measure of Cp, but it was not significant for the PPVT-R. The path from Cp to AL exhibited the inverse: Cp was not significant for KABC but was for PPVT-R. Our hypothesized path from AU&L to PL, not in Carver's original formulation, was significant. All other paths hypothesized by Carver were signifi- cant. Covariance values were quite similar for overlapping analyses, even though,

tory Accuracy Level A u ~ A L

,851 .I 13

+ Reading Rate

FIGURE 2 Structural equation modeling analyses of rauding model for Grades 1 and 2.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 53

in several cases, one coefficient was significant, whereas the other was not for a given path. The strongest relations were between AudAr. and PL and between PL and AL

Grades 3 and 4

' f i e second two analyses are shown in Figure 3, with KABC path coefficients first and KTEA coefficients second, for the overlapping analyses. The theoretical model based on rauding theory had AGms over .95 and NFIs over 97. No paths were indicated to be added. The path from CP to AL showed that CP was not significant for KABC but was for PPVT-R. Our hypothesized path from A u ~ A L to PL, not in Carver's original formulation, was significant. Two other paths hypothesized by

Auditory Accuracy Level

,675

I -.000ns

Reading Rate

FIGURE 3 Structural equation modeling analyses of rauding model for Grades 3 and 4.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

154 RUPLEY, WILLSON, NICHOLS

Carver (AL to EL and PL to RL) were not significant; the rest were. Covariance values were less similar for overlapping analyses than in Grades 1 and 2, and, as before in one path (CP to AL), one coefficient was significant, but the other was not. The strongest relations were between A u ~ A L and PL and between PL and Ah with the latter stronger than in Grades 1 and 2.

Grades 5 and 6

The third two analyses are shown in Figure 4, with KABC path coefficients first and KTEA coefficients second, for the overlapping analyses. The theoretical model based on rauding theory had AGFIs over .90 and NFIs over .95. No paths were indicated to be added. Our hypothesized paths from Cp to PL and from AudAL to PL, not in Carver's original formulation, were significant for all analyses. Two paths

Auditory Accuracy Level

.604 ,312

+

Reading Rate

FIGURE 4 Structural equation modeling analyses of rauding model for Grades 5 and 6.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 55

hypothesized by Carver (AL to EL and PL to RL) were not significant. Covariance values were less similar for CP to PL and CP to Ar, than in Grades 1 and 2 and were stronger than in Grades 3 and 4. The PPVT-R measure of Cp exhibited a stronger effect of AL than the KABC measure of CP. The strongest relations were between AU&L and PL and between PL and A& with the latter weaker than in either Grades 1 and 2 or Grades 3 and 4. Finally, there was support for a path between CP and EL at this level, which might substitute for Carver's hypothesized path of AL to Er, which was not significant. This model was not explored further because there is no current theoretical justification.

SUMMARY AND CONCLUSIONS

The results of this study for Grades 1 and 2 affirm Carver's (1993, 1995) rauding theory concepts and the research cited earlier reporting the major contribution that word recognition makes to elementary school children's reading comprehension. The theoretical model we proposed, based on an adaptation of rauding theory and on our own research on stages and phases of reading development factors contrib- uting to young children's reading comprehension, had an NFI of .992. The only difference noticed in the analyses of Grades 1 and 2 was the addition of a path from PL (word recognition) to EL (reading-comprehension rate). For these young chil- dren, this suggests that their reading-comprehension rate is directly related to their word-recognition capability and that their accuracy level in reading comprehension is directly influenced by their rate of processing text. Slower rates of word recognition would directly affect comprehension and inhibit chunking of informa- tion into meaningful information units. Thus, both comprehension of new informa- tion and expansion and elaboration of existing knowledge would be affected by elementary school children's rate and accuracy in processing information. The acquisition of new information through reading is associated with later stages or phases of reading development; for example, Chall (1996) proposed that children do not use reading as a primary means of knowledge acquisition until the third stage (about Grade 4).

In the simple view of reading, Gough and Tunmer (1986) described the process of reading as the ability to decode and comprehend. Reading research findings of the past, which are further supported by the results of this study, noted that word recognition plays an important if not key role in reading comprehension. Conven- tional literacy begins when readers begin to recognize specific words. The acqui- sition of literacy is primarily the acquisition of word-recognition skills (Gough et al., 1992). With word recognition's playing such a substantial role in the early development of reading comprehension, it is noteworthy that it maintains a salient role throughout Grades 1 through 6.

In addition to seeing the primary role that word recognition plays, we found that the KABC measure of cognitive power made only a small contribution to reading

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

156 RUPLEY, WILLSON, NICHOLS

comprehension. The cognitive power ( C p ) variable associated with the PPVT-R was generally larger but still small until Grades 5 and 6. This is consistent with the conclusions of Stanovich, Cunningham, and Freeman (1984) and Rupley and Willson (1997) that cognitive power operates primarily through decoding and does not directly influence reading comprehension in young readers. Cognitive power, however, is theorized to be a factor influencing yearly gain in students' compre- hension. Support for this was found at Grades 3 and 4 and Grades 5 and 6. A direct path emerged at Grades 3 and 4 from cognitive power (PPVT-R) to both word recognition and comprehension, and, at Grades 5 and 6, both measures of cognitive power directly affected both decoding and comprehension. The strength of the paths increased from Grades 3 and 4 to Grades 5 and 6. The slight increase with grade of the effect of cognitive power on reading comprehension is greater for crystallized (PPVT-R) intelligence than for global (fluid and crystallized) or fluid intelligence. As mentioned earlier, preliminary analyses with nonverbal, fluid intelligence (KABC Simultaneous Processing score) produced virtually identical results.

Concomitant with the increase in cognitive power was a decrease in the asso- ciation of word recognition with comprehension at the higher grades and no effect of comprehension on comprehension rate. Our finding that neither word-recogni- tion rate nor comprehension rate had paths with other variables for Grades 3 and 4 and Grades 5 and 6 results in the failure to advance the credibility of these variables in contributing to reading comprehension, although they did covary with each other at all levels. However, this finding does not differ from Carver's (1993, 1995) assertion in discussing his "simple view 11" of reading. His data analyses support the hypotheses that word recognition (decoding) and listening ability alone have a direct causal effect on improvement in reading comprehension for elementary school children. Our data would support his hypothesis.

Not only do the results of this study offer support to Carver's (1993) simple view I1 of reading, they further advance the conclusion that children go through stages or phases of reading development as they refine and acquire reading capabilities. Although word recognition remains a salient component throughout the elementary grades, its contribution to comprehension begins to lessen at the higher grade levels. The decreased effects of word recognition on comprehension with increased age and reading proficiency supports the idea that, as word recognition becomes automated, its relation to reading comprehension becomes less apparent as readers' competency increases.

REFERENCES

Adams, M. J. (1990). Beginning to read. Cambridge, MA: MIT Press. Alexander, P. A,, & Judy, J. E. (1988). The interaction of domain-specific and strategic knowledge in

academic performance. Review qfEducationa1 Research, 58, 375404.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

EXPLORATION OF THE DEVELOPMENTAL COMPONENTS 1 57

Anderson, R. C., Reynolds, R. E., Schallert, D. L., & Goetz, E. T. (1977). Frameworks for comprehend- ing discourse. American Educational Research Journal, 14,367-382.

Biemiller, A. (1970). The development of the use of graphic and contextual information as children learn to read. Reading Research Quarterly, 6, 75-96.

Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Carver. R. P. (1992). Effect of prediction activities, prior knowledge, and text type upon amount

comprehended: Using rauding theory to critique schema theory research. Reading Research Quar- terly, 27, 165-174.

Carver, R. P. (1993). Merging the simple view of reading with rauding theory. Journal of Reading Behavior, 4, 439-454.

Carver, R. P. (1995, April). Causal model for reading achievement. Paper presented at the meeting of the Society for the Scientific Study of Reading, New York.

Chall, J. S. (1996). Stages of reading development. New York: McGraw-Hill. Daneman, M. (1991). Individual differences in reading skills. In R. Barr, M. Kamil, P. Mosenthal, &

P. Pearson (Eds.), Handbook of reading research (Vol. 2, pp. 512-538). White Plains, NY: Longman.

Dunn, L. M., &Durn, L. M. (1981). Peabody Picture Vocabulary Test-Revised: Manual. Circle Pines, MN: American Guidance Service.

Ehri, L. C. (1991). Development of the ability to read words. In R. Barr, M. Kamil, P. Mosenthal, & P. Pearson (Eds.), Handbook of reading research (Vol. 2, pp. 383-417). White Plains, NY: Longman.

Ehri, L. C. (1992). Reconceptualizing the development of sight word reading and its relationship to recoding. In P. Gough, L. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 107-144). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Bhri, L. C., & Wilce, L. S. (1987). Cipher versus cue reading: An experiment in decoding acquisition. Journal of Educational Psychology, 79, 3-13.

Gough, P. B., & Hillinger, M. L. (1980). Learning to read: An unnatural act. Bulletin of the Orton Society, 30, 179-196.

Gough, P. B., Juel, C., & Griffith, P. L. (1992). Reading, spelling, and the orthographic cipher. In P. Gough, L. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 35-48). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, andreadingdisability. Remedial and Special Education, 7, 6-10.

Hoover, W. A., & Gough, P. B. (1990). The simple view of reading. Reading and Writing, 2, 127-160. Juel, C. (1991). Beginning reading. In R. Barr, M. Kamil, P. Mosenthal, & P. Pearson (Eds.), Handbook

of reading research (Vol. 2, pp. 759-788). White Plains, NY: Longman. Kaufman, A. S., & Kaufman, N. L. (1983). Kaufman Assessment Battery for Children. Circle Pines,

MN: American Guidance Service. Kaufman, A. S., & Kaufman, N. L. (1985). Kaufman Test of Educational Achievement: Comprehensive

form manual. Circle Pines, MN: American Guidance Service. Mason, J. M. (1980). When do children begin to read: An exploration of four year old children's letter

and word reading comprehension. Reading Research Quarterly, 25, 203-224. Mason, J. M. (1984). Early reading from a developmental perspective. In R. Barr, M. Karnil, P.

Mosenthal, & P. Pearson (Eds.), Handbook of reading research (pp. 505-504). White Plains, NY: Longman.

Miller, G. E., (1987). The influence of self-instruction on the comprehension monitoring performance of average and above-average readers. Journal of Reading Behavior, 19,303-317.

Paris, S. G., & Meyers, M., 11. (1981). Comprehension monitoring, memory, and study strategies of good and poor readers. Journal ofReading Behavior, 13, 5-22.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014

158 RUPLEY, WILLSON, NICHOLS

Paris, S. G., Wasik, B. A., &Turner, J. C. (1991). The development of strategic readers. In R. Ban, M. Kamil, P. Mosenthal, & P. D. Pearson (Eds.), Handbook qf rending research (Vol. 2, pp. 609-640). White Plains, NY: Longman.

Rupley, W. H., & Willson, V. L. (1997). Relationship between reading comprehension and components of word recognition: Support for developmental shifts. Journal qf Research and Development in Education, 30, 255-260.

Sadoski, M., Paivio, A., & Goetz, E. T. (1991). A critique of schema theory in reading and adual coding alternative. Reading Research Quarterly, 26, 463-484.

SAS Institute Inc. (1989). SAT/STAT user's guide: Version 6 (4th ed.). Cary, NC: Author. Stanovich, K. E. (1991). Word recognition: Changing perspectives. In R. Barr, Kamil, P. Mosenthal, &

P. Pearson (Eds.), Handbook of rending research (Vol. 2, pp. 418-452). White Plains, NY: Longman.

Stanovich, K. E., Cunningham, A. E., &Freeman, D. .I. (1984). Intelligence, cognitive skills, and early reading progress. Reading Research Quarterly, 29, 278-303.

Tanaka, J. S. (1993). Multifaceted conceptions of fit in structural equation models. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 1G39). Newbury Park, CA: Sage.

Willson, V. L., & Rupley, W. H. (1993). Structural components of single word recognition: Activation of orthographic, meaning, and phonological processors. Reading Research and Instruction, 32, 33-45.

Manuscript received February 10, 1997 Final Revision received October 21, 1997 Accepted October 28, 1997

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

24 1

7 O

ctob

er 2

014