Ninety-eight percent of all states, and most federal agencies servicing adolescents and adults, have followed the lead of federal law and have in-
3It is important to note that space limitations do not allow a detailed explanation of all caveats related to the use of the different WJ III discrepancy procedures in the "art and science" of LD decision making and classification. Comprehensive models that encompass a broader array of variables are required. A particularly interesting comprehensive model grounded in the CHC theory has recently been outlined by Flanagan, Ortiz, Alfonso, and Mascolo (in press).
corporated the notion of a learning disability being defined by a discrepancy between a person's actual (measured) and expected achievement (usually predicted from general intellectual ability) (Flanagan et al., 2001). The WJ III provides reliable and valid procedures that can contribute useful information for use in these procedures. The GIA Ability-Achievement Model and the Predicted Achievement Model are both presented in Figure 14.3.
It is important to note that neither the WJ III GIA or PA discrepancy procedures were designed for the diagnosis of LD, if the intent is to identify a specific learning disorder. The GIA and PA approaches are intended to answer the question, "Given the person's present cognitive abilities, is he/she achieving as well as could be expected?" The WJ III GIA Model is straightforward. Either the differentially weighted WJ III GIA-Std or GIA-Ext g standard scores are used to provide a prediction of what a person's achievement standard score would be, given their level of general intellectual ability and age or grade. As portrayed in Figure 14.3, the individual's expected achievement (e.g., predicted Basic Reading Skills standard score) is then subtracted from the person's actual Basic Reading Skills standard score, producing an ability-achievement standard score discrepancy. The WJ III is unique with respect to three major features of this model.
• The WJ III provides a true g score for use in the ability-achievement calculations. Other major intelligence batteries rely on an arithmetic average of test scores, an average that implies an equally weighted general intelligence score. Thus, a more theoretically sound general ability index is used in the WJ III GIA ability-achievement discrepancy procedures.
• The WJ III predicted (sometimes referred to as "expected") achievement score accounts for regression-to-the-mean in a manner that captures the developmental changes in ability— achievement correlations. The regression-to-the-mean effect is greatest for predicted achievement scores that diverge the farthest
WJ III GIA (g) and predicted achievement ability-achievement discrepancy
WJ III GIA (g) and predicted achievement ability-achievement discrepancy from the mean. This occurs because the correlation between any combination of ability and achievement measures is less than perfect.
• The WJ III capitalizes on the co-norming of the WJ III COG and ACH to provide slightly different regression adjustments across age or grade when generating the predicted achievement score. The developmental changes in the ability-achievement correlations observed in the WJ III norm data are incorporated into the prediction equations. A common practice, particularly when an intelligence test is not co-normed with the administered achievement test, is to select a single point correla tion (e.g., .65) and use this value to correct for regression effects via a formula for all ages. The predicted score in the WJ III GIA Ability-Achievement Model is based on a devel-opmentally sensitive regression-to-the-mean adjustment procedure.
• The WJ III GIA ability-achievement discrepancy scores are compared against real discrepancy norms (end box in Figure 14.3; McGrew, 1994; McGrew et al., 1991; McGrew & Woodcock, 2001; Mather & Schrank, 2001; Woodcock, 1978). Predicted achievement scores in all relevant achievement domains were generated for all norming subjects. Each subject's predicted and actual achievement difference scores in the same domain produced ability-achievement discrepancy scores for the WJ III norm file. The distributional characteristics of each ability-achievement discrepancy score distribution were used to produce the WJ III GIA ability-achievement discrepancy norms just as age- or grade-specific norms are determined for any test or cluster.
In summary, whenever a WJ III GIA ability— achievement discrepancy score is calculated, this score incorporates the developmentally appropriate degree of regression-to-the-mean and is then compared against the actual distributions of discrepancy scores in the norming sample. The WJ III GIA ability-achievement discrepancy scores may be interpreted in three different metrics (standard score discrepancy, percentile rank, and standard deviation units; McGrew & Woodcock, 2001) as the examiner chooses.4
Was this article helpful?