Conclusions

In evaluating these data, some will say that we should not accept the null hypothesis and propose continued investigations of discrepancy, low achievement, and intra-individual difference models. There is little possibility that discrepancy models can yield unique subgroups of underachievers. Low achievement models do not adequately assess the construct of unexpected underachievement. Intra-individual differences models are essentially an expansion of discrepancy models, maintaining the historic emphasis on the "LD test battery." It is time for change. Such approaches have not been effective, and there are psychometric and logistical issues that make all these models difficult to implement.

Others will note that the amount of validity evidence that we are citing for response to intervention models is relatively sparse. We have not attempted to score the literature for validity studies. Nonetheless, models that incorporate response to intervention have considerable potential for identifying a unique group of underachievers. The studies that we can cite show robust differences in cognitive and achievement functions and in neural correlates. There is also evidence that different kinds of interventions are needed for non-responders, which is admittedly a somewhat circular argument. Most important, the potential for powerful demonstrations of validity and broadly understandable support for the construct of LD should be immediately apparent. In the end, measurement must assess the latent construct of LD. Previous classifications have been adopted without evidence that even the underlying measurement model was valid. No model will ever be perfect. Models that incorporate response to intervention also imperfectly assess the LD construct and we can debate the best procedures for operationalizing non- responsiveness: type of intervention, slope, intercept, number of assessment points, etc. (Fuchs, Mock, Morgan, & Young, in press). There will be no best procedure, but the potential validity is transparent and will likely transcend these measurement issues.

It is useful to conceptualize models that incorporate response to intervention as extensions of low achievement models. The construct of LD begins with underachievement and the task is to define "unexpectedness." A hybrid model in which low achievement is used as one of several criteria (along with the exclusions) and response to intervention has considerable potential and was recommended by the LD Summit (Bradley et al., 2002). In this respect, no single assessment should ever be the sole the criterion for establishing eligibility for special education. It is one thing in a research study to specify identification criteria and still another to make a complex judgment about eligibility and services. However, previously utilized classification models do not have adequate reliability or strong validity. These models incorporating response to intervention at least have the promise of operationalizing the concept of unexpected underachievement in a way that is meaningful: as a demonstrable inability to learn. The children who are identified with this classification do appear to differ from children who have other forms of underachievement. Critics of these models worry that the result of implementation efforts will be an abandonment of historic concepts of discrepancy and a loss of integrity of the LD construct (Kavale, 2002; Scruggs & Mastropieri, 2002). Neither view is correct as the model retains the concepts of unexpected underachievement and discrepancy (Fletcher et al., 2003). In fact, if the field does not move away from historic concepts that were adopted with a severely limited research base, the construct of LD will perish under the weight of the research base not supporting the reliability or validity of alternative approaches to discrepancy and the lack of effectiveness of the services associated with current approaches to identification. More research is always needed, but the critical research issue is an evaluation of what happens when hybrid models are implemented on a wide scale. Without an attempt at implementation, it will be hard to determine the viability of this model.

Previous Page | Next Page
(Response to intervention) | (References)