Causal Modeling of Reading Acquisition

Abstract

). For the simple 2W2V design, where two variables (x, y) are measured at two points in time (1, 2), the purpose of the cross-lagged panel procedure is to infer the direction of causality by comparing the cross-lagged correlations [i.e., r(xl, y2) vs. r(yl, x2)]. In path analysis, a model is constructed which hypothesizes the causal relationships among a set of variables. A multiple linear regression equation is derived for each dependent variable in the model, and consists of those variables believed to influence that particular dependent variable. The estimates of these influences are known as standardized path coefficients. Structural equation modeling is a form of path analysis where there are multiple measures or indicators of each unmeasured latent variable, instead of equating indicator variables with latent variables. Even with the recent popularity of such techniques, some researchers still believe that the intent of causal modeling is to make causal inferences from correlational data. In actuality, the intent is to test the plausibility of alternative theoretical models by gathering evidence useful for evaluating the possible causal relationships among the relevant and important variables.Given these introductory remarks, four points will be discussed which relate to the use of these specific methods, particularly with respect to the modeling of reading acquisition. The first point concerns the causality interpretation of cross-lagged panel correlation (CLPC) analyses. Rogosa (1979) has shown that even when a strong causal relationship between two variables is known, the cross-lagged correlations may not be different or may indicate the opposite direction of the causal influence. Thus the CLPC method may not uncover a "true" causal relationship and instead may indicate either that there is no relationship or that there is an inverse relationship. The method merely compares sample bivariate correlations and does not yield estimates of the causal parameters that we are really interested in. For this reason the CLPC technique is not recommended.The second point concerns the measurement error assumption in the path analysis (PA) method. An implicit assumption of PA is that the variables are measured without error, that is, the variables are perfectly or infallibly measured. Considerable measurement error in a path model will have an effect upon the path coefficients. When the variables are based on fallible measure-

Intelligence, Cognitive Skills, and Early Reading Progress

“…Performance on the phonological oddity task made a significant independent contribution to predicting pseudoword naming time, and RPM performance was a significant predictor of phonological oddity scores. Thus, within this model, decoding speed and listening comprehension appear to have direct influences on reading comprehension, but the effect of phonological awareness appears to be indirect, through its influence on decoding speed (see Lomax, 1982, for a similar finding).…”

“…Phonological awareness also emerged as a separate predictor, explaining components of reading skill that were independent of those accounted for by general verbal comprehension ability. It is likely that phonological awareness underlies the ability to segment and analyze the speech stream, and the latter processes are important in determining early success at decoding unknown words (Calfee et al, 1972(Calfee et al, , 1973Fox & Routh, 1980;Goldstein, 1976;Gough & Hillinger, 1980;Helfgott, 1976;Liberman, 1973;Liberman, Shankweiler, Fischer, & Carter, 1974;Lomax, 1982). However, in addition to the mere knowledge of the spelling-to-sound correspondence rules, it appears to be important to develop the ability to rapidly apply the rules, because decoding speed emerged as a separate predictor independent of phonological awareness.…”

“…At the early stages of reading acquisition there are other independent and equally important determinants of initial reading progress, factors that may initiate a causal chain of achievement deficits. For example, phonological awareness is probably causally related to the early acquisition of decoding skills (Bradley & Bryant, 1978, 1983Fox & Routh, 1976;Goldstein, 1976;Lomax, 1982;Mann, 1981;Perfetti, Beck, & Hughes, 1981;Stanovich, 1982aStanovich, , 1982cTreiman & Baron, 1981, 1983. Inadequate decoding skills result in unrewarding early reading experiences that lead to less involvement in reading-related activities (a substantial practice differential between skilled and less skilled readers is apparent as early as January of the first-grade year, see Biemiller, 1977Biemiller, -1978.…”

Facilitating Reading Comprehension Through Spatial Skills Training

“…Recent developments have suggested the possibility that far from being the spurious outcome of a hidden third variable (cf. Crano & Brewer, 1986;Kenny, 1975), the covariation of spatial skills and higher order cognitive development might provide an important, albeit previously ignored, key to understanding a basic aspect of skilled reading acquisition (e.g., Lomax, 1982). The present study was designed to explore this possibility.…”

Using Computer Guided Practice to Increase Decoding Fluency in Learning Disabled Children: A Study Using the Hint and Hunt I Program

“…However, there is good evidence (Groff, 1978;Lesgold & Resnick, 1982;Lomax, 1982) that inefficient decoding is causctlly related to poor reading comprehension. Put simply, if children cannot read many of the words on the page, or if they read them only with great difficulty, they will be strictly limited in their ability to use whatever other knowledge they possess in constructing the meaning of the passage.…”