Jack M. Fletcher, University of Texas; G. Reid Lyon, National Institutes of Health; Marcia Barnes, University of Toronto; Karla K. Stuebing, University of Texas; David J. Francis, University of Houston; Richard K. Olson, University of Colorado; Sally E. Shaywitz, Bennett A. Shaywitz, Yale University
Learning Disabilities Summit: Building a Foundation for the Future White Papers
This paper is available in alternative formats: | Download Word | Download pdf |
Pages: | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
In this paper, we have reviewed federal and non-federal definitions of LD, pointing out that these definitions embed hypothetical classifications at three levels: IQ discrepancy, heterogeneity, and exclusion. We also evaluated the evidence for the hypothesis that LD can be related to constitutional factors, showing that environmental factors must be accounted for in explaining not only why a child develops LD, but also the role of instruction. Throughout the paper, we highlighted some alternative approaches to classification, reflecting different ways of thinking about LD. We focused specifically on the value of inclusionary definitions that identify specific forms of LD, leading to specific (and less time-consuming) identification practices that we believe are directly linked to intervention. We suggested a hypothetical reorganization of the types of LD identified in the 1977 operationalization of the federal definition of LD. We recognized that exclusions, with the exception of inadequate instruction, largely reflect that there are other ways of serving children including different categories in IDEA and other services, such as compensatory education. There is little evidence that children meeting these exclusionary criteria have different instructional needs or respond differently to intervention.
Kavale and Forness (2000) presented an approach to the classification, definition, and identification of LD that in many respects is the antithesis of what would be recommended based on the research reviewed in this report. It begins with the acceptance of unexpected low achievement at the first level and the notion that discrepancy sets apart a specific form of LD as a necessary but not sufficient criterion at the first level of identification. The approach recognizes the heterogeneity of LD at the second level, tying LD to achievement deficiencies in language, reading, writing, and math. These deficiencies presume the presence of IQ discrepancy. At levels III and IV, issues related to learning processes are added. At level V, children are excluded because of sensory impairment, mental deficiency, emotional disturbance, social and cultural disadvantage, and inadequate instruction.
This approach to classification hinges on the validity of IQ discrepancy as demarcating a specific form of LD that is differentiated from low achievement. The evidence reviewed in this paper shows that IQ-discrepant and low achieving groups overlap substantially in cognitive characteristics and show little difference in response to intervention and long-term outcome. Similar problems affect the use of exclusionary criteria. Consider children who have (a) an IQ discrepancy, (b) problems in reading, (c) processing difficulties, and (d) who do not meet any of the exclusionary criteria. How are they meaningfully different or have different instructional needs from children with (a) through (c), but who meet exclusionary criteria and are therefore not defined as LD? There is little evidence that this would be the case, even if the hypothesis was tested only within IQ-discrepant children. There is really no hypothesis to test, as there is no basis for imagining how such subgroups could differ if the sorting was based solely on the exclusionary criteria. Even if one argued that exclusion would be infrequent because of all the prior levels of identification, the evidence in this paper does not support the hypothesis that children excluded as non-LD are meaningfully different from those who make the cut.
We do not mean to indicate that federal and non-federal definitions that have been used to the present have not had utility. On the positive side, the evolution of the current federal definition in IDEA has successfully served as a rallying point for special interest groups and for increased funding for special education programs. The current omnibus federal definition has served well as a galvanizing force for advocacy groups in their quest to obtain funding and secured educational services support for children with LD. Current (and historical) classifications and the resultant definitions of LD should be conceptualized as hypotheses that require rigorous, ongoing evaluation. The review of evidence in this paper shows that the classifications have become obsolete and should be revised, especially if the goal is to guide and reform instruction.
A major problem is the notion that low achievement in LD is unexpected, leading to a focus on exclusion. The accumulation of research over the past 30 years shows that low achievement is expected and suggests a focus on identifying the factors responsible for poor achievement in every child. Such a shift would suggest the need to develop inclusionary definitions that build upon the cumulative research base on LD. The move from exclusionary to inclusionary definitions is the first of many steps.
In addition to the need for inclusionary definitions, we must recognize that an approach to identification based solely on test scores is not likely to be reliable and begs the question of where to put the cut-score. Achievement test scores are continuous and largely normally distributed. The tests used to measure these domains have measurement error. Any attempt to set a cut-point will lead to instability around the cut-point as scores fluctuate around the point with repeat testing, even for a decision as straightforward as demarcating low achievement. This fluctuation is not a problem of repeat testing, nor is it a matter of selecting the ideal cut-score. The problem stems from the fact that no single score can perfectly capture a student's ability in a single domain. There is always measurement error. Fluctuation will also vary across tests, depending in part on the cut-score, as tests vary in their precision at various ranges of the ability scale. This problem is more significant as the cut-point moves from the center of the distribution.
A second problem with the typical use of cut-scores concerns their arbitrary nature. A cut-point on a norm-referenced test is an arbitrary, relative standard of performance. The arbitrariness of the standard does not mean that a cut-point does not indicate a problem. Rather, arbitrariness reflects the meaninglessness of distinctions between, for example, the 15th and 20th percentile (or the 20th and 21st percentile). The problem with arbitrariness is not so much with the use of norm-referenced tests for establishing cut-points, but reflects difficulties inherent in any approach that would make critical decisions based on a single indicator. A single assessment at a single point in time is not psychometrically adequate for deciding placement. The flexibility in IDEA that allows interdisciplinary teams to go beyond test scores and encourages clinical judgment is necessary because of these issues. But the basis for clinical judgment should include performance on psychometric tests that involve achievement and cognitive performance. Inclusionary definitions based on patterns on these types of tests may be especially useful.
Such an approach would dramatically reduce the reliance on IQ tests for the identification of LD. Although there may be a role for IQ tests in determining mental deficiency, even here the more important concept is adaptive behavior, and there are difficulties establishing the upper range that distinguishes mental deficiency from LD (MacMillan, Siperstein, & Gresham, 1996).
The problems that we observed above in setting cut-points also apply to IQ distributions. There is no natural subdivision that demarcates mental deficiency from LD. Even with the stipulation of mental deficiency, there is no need to give every child referred for special education an IQ test. For LD, the information has limited relevance, particularly for intervention. The concept of IQ as it is applied to LD is outmoded and reflects an obsolete practice. The use of IQ tests reflects a focus on compliance as opposed to results that must shift if placement in special education as LD is to benefit the person so designated. IQ tests do not measure aptitude for learning or provide an index of response to intervention. The processes that contribute to performance on an IQ test may well be an outcome of the same processes that led to the LD. Dropping IQ from the LD definition would shift the focus to achievement/cognitive processes and also result in more efficient, less expensive evaluations.
Related to the issue of the obsolete role of IQ for LD is the notion of the slow learner, or garden-variety poor learner. These terms are also used to refer to children with low achievement at levels consistent with their IQ. There are clearly children who have impairments in multiple cognitive and academic domains who obtain lower scores on IQ tests. Many of these children represent what we described earlier as the comorbid RD-MD group.
Although it is commonly assumed that IQ is an indicator of the slow learner, this does not appear to be the case. It is difficult to identify an IQ cut-point, even in the non-mentally deficient range, that would differentiate specific LD and garden-variety LD. IQ scores do not reliably differentiate children with different types of LD. To illustrate, McFadden (1990) completed a cluster analysis to determine whether level of IQ was associated with different types of LD. McFadden found that (1) children with IQs between 70 and 80 were generally represented in all clusters of children with learning disabilities; (2) many children with low IQs exhibited similar patterns of cognitive difficulties relative to children defined as having learning disabilities by discrepancy criteria; (3) although a WISC full scale IQ cut-off of 80 reduced the number of children with low IQs in learning disabilities clusters, several subtypes still contained children with approximately 20 percent lower IQs; and (4) children with low IQs were apparent in clusters of children with learning disabilities and, within such clusters, differences occurred in level but not shape. These results question the validity of differentiating learning disabilities according to IQ cutoffs of 80 and above, but do not identify appropriate cutoffs (if any).
In another cluster analytic study, Morris et al. (1998) were able to distinguish children with specific RD who had cognitive problems relatively restricted to the phonological domain from those who had more generalized difficulties in multiple cognitive domains (e.g., vocabulary, speech production, attention). On average, children with non-specific RD had lower scores on IQ tests than children with specific RD, but IQ ranged considerably within each subtype. The differences between specific and non-specific subtypes were most reliably indexed by the child's vocabulary development and could be understood as the consequences of the child's poor language development, which in turn produced lowered IQ scores.
The notion that low achievement is expected in garden-variety RD and unexpected in specific RD is also specious. The basis for reading difficulties was associated with phonological processing in all subtypes with word reading problems. Some specific subtypes read as poorly as the non-specific subtypes, but the groups did not differ qualitatively in language characteristics related to reading (i.e., in the phonological domain). The garden-variety group may well have a poorer prognosis and need different types of instruction. The garden-variety group may even show different neurological characteristics. But would we really want to restrict our concept of LD or eligibility for services to children with specific types? Schools are interested in serving the lowest achievers as these are most difficult to teach (MacMillan et al., 1996).
Research on children with LD has not progressed to the point where we can say definitively that children with specific and garden-variety subtypes need different interventions, have different prognoses, or respond differently to treatment. This reflects in part the preoccupation with concepts of LD based on unexpectedness, IQ-discrepancy, and anxiety over the role of underachievement in LD. The consolidating issue is that the concept of underachievement and the linking of LD to an academic deficiency (e.g., reading, math) are necessary to the concept of LD. They are not sufficient and it is essential to include the concept of process (e.g., language, perceptual skills) as necessary to the concept of LD (Kavale & Forness, 2000). It is also essential to drop notions of "potential," "ability," and their operationalization in measures of IQ and to move towards attributes or components that are measurable and linked to intervention. Thus, we would move the concept of LD from a disorder that is unexpected because of discrepancies between ability potential and achievement to one in which underachievement is expected because of impairment of key cognitive processes. These processes are measurable and can be directly linked to intervention.
It is essential to introduce the student's response to well-designed instruction and remediation programs as a major component of the identification of LD. This introduction should be made in the context of early identification and prevention programs that are seen as fundamental to general and special education. Children who do not benefit from early and intensive interventions will require even more powerful remediation programs as well as educational accommodations as they proceed through their schooling. The information on how well the child responded operationalizes the "inadequate instruction" component and those who do not respond to increasingly intense interventions may indeed be disabled. In addition, continuous monitoring of progress will be helpful not only for instructional planning, but also for identifying those who do not respond to adequate instruction (Fuchs & Fuchs, 1998).
To do justice to the need for a classification of LD that yields inclusionary definitions with the features we have identified as being desirable, we call for the development of a consensus process. As part of this process, the relevant federal agencies responsible for research and practice involving people with LD should work together to synthesize the available research. The principles and goals of a new overarching classification should be explicitly articulated, with specification of boundaries and overlaps with other classifications of childhood disorders (e.g., mental deficiency, emotional disorders). The possibility of comorbid associations should be incorporated. Working groups could be assembled to formulate definitions of different types of LD. Definitions should be formulated only for those types of LD where there is clear evidence of their nature and correlates. The classification and definitions should be treated as hypotheses. Research to evaluate the resultant classifications is desirable and should be supported. Plans to periodically update and revise the definitions should be made. Such a process would hopefully permit the development of specific procedures for identifying different types of LD that are efficient, that do not waste resources, and that lead to specific interventions.
Thus, we propose careful assessment of academic skills and their cognitive correlates as part of the implementation of inclusionary definitions. These assessments should be completed to address prevention/intervention needs with a goal of evaluating the instructional needs of the child. Adding continuous monitoring of progress and response to intervention as considerations in this process may go a long way towards the ultimate goal of helping as many children as possible master academic skills and return to regular education. A consensus process would help ensure that the last available evidence from research and the best available presentations were marked in a re-formulation of the federal classification of LD. Any changes must take into account the need for improved teacher preparation in general and special education, especially if the federal classification is changed.
Some researchers have confused the IQ-discrepancy hypothesis with the concept of LD, which is not appropriate. For example, Aaron (1997) stated that "when the discrepancy formula disappears from the educational science, so will the concept of LD" (p. 489). Similarly, Kavale and Forness (1994) stated that "... the notion of discrepancy ... has led to a confounding ... most clearly seen in the suggestion that there are more similarities than differences between LD and low achieving students. Such a suggestion calls into question the very notion of LD" (p. 43).
This conceptualization could be shown to be unreliable and invalid with no consequences for the validity of the concept of LD. This is clearly indicated in Figure 7, which shows that children with RD, MD, and RD-MD can be differentiated from those with no LD, even when ADHD is involved. Patterns of performance differentiate types of LD, while both level and patterns differentiate those with and without LD. Children identified with either an IQ discrepancy or LD definition are disabled, need to be identified, and respond similarly to appropriate educational interventions. What is being questioned is the validity of classifications of LD based on the presence or absence of IQ discrepancy and exclusion, not the reality of LD. Both definitions validly identify LD with or without the exclusions. As a classification, coverage, reliability, and validity are not adequate.
When the original federal definition of LD was proposed, there was little research that supported the discrepancy, heterogeneity, and exclusionary components of LD classifications. Since then research has accumulated suggesting that the discrepancy and exclusion components have (at best) weak validity and may be harmful and represent an obstacle to effective intervention. The goal should be to close the achievement gap for students identified as LD relative to their peers. Unfortunately, this happens all too infrequently by virtue of placing students with LD in special education (Lyon et al., 2001). One part of the solution is to revise the federal definition of LD and develop new classifications that are linked to research. New definition and identification practices will emerge, so that those who serve children with LD can focus on early identification, prevention, and effective remedial strategies. Eligibility and compliance presently consume excessive fiscal and emotional resources; this consumption should be redirected to intervention and special education should be re-oriented towards results, which means truly remediating children and returning them to the educational mainstream. These are the ultimate purposes of classifying a student as LD and the reasons that such classifications were developed. Such purposes must guide the reworking of the federal classification of LD essential to ensuring that all children can learn and reach their full potential in our society.
Funding for NRCLD is provided by the U.S. Department of Education, Office of Special
Education Programs. Award No. H324U010004. Opinions expressed herein are those of the National
Research Center on Learning Disabilities and do not necessarily represent the position of the U.S.
Department of Education.
© 2007 National Research Center on Learning Disabilities
Last Updated: November 19, 2007
Contact: nrcld@ku.edu