Classification of Learning Disabilities: An Evidence-Based Evaluation

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 |

INTRODUCTION

The purpose of this paper is to review research on the classification of learning disabilities (LD). We begin by briefly reviewing the nature of classification research. Then we discuss the evolution of definitions of LD, making explicit the classification hypotheses from which these definitions derive. An extensive review of the evidence for these hypotheses will be provided for the three components of classification implicit in the federal definition of LD: discrepancy, heterogeneity, and exclusion. We will show that classification hypotheses involving discrepancy and exclusion as embedded in federal (and state) policy have at best weak validity, often representing inaccurate and outdated assumptions about LD. There is evidence for heterogeneity of LD, but some reorganization of the types of LD identified in the federal definition may be necessary. Throughout the paper we identify alternative approaches to classification and identification, including weaknesses in any psychometric approach to the identification of LD. We suggest that classifications based on inclusionary definitions that specify attributes of different forms of LD are more desirable than current exclusionary definitions. Inclusionary definitions permit a focus on identification procedures that are intervention oriented as well as a focus on prevention, both of which are desirable and could contribute to improved results in remediating LD.

WHAT IS CLASSIFICATION?

Classification is the process of forming groups from a large set of entities based on their similarities and dissimilarities. It is not the same as identification, which is the process of assigning entities to an established classification. Valid classifications can be differentiated according to variables not used to form the groups. They are also reliable and have adequate coverage, i.e., permit identification of the majority of entities of interest. In classification research, groups are formed and evaluated for reliability, validity, and coverage. All classifications are hypotheses about the independent variables. Classification researchers evaluate the reliability, validity, and coverage of a hypothetical grouping of interest (Fletcher, Francis, Rourke, Shaywitz, & Shaywitz, 1993; Morris & Fletcher, 1988; Skinner, 1981).

Classification is fundamental to science and practice. It is virtually impossible to identify components of science or practice, regardless of the discipline and epistemological orientation, that do not involve classification. Although ubiquitous, classifications are often implicit and not explicitly identified. As part of science, however, all classifications are hypotheses that need to be empirically evaluated. Whenever a set of dependent variables is compared in relation to a set of independent variables (e.g., memory performance in children with and without LD), there is an explicit test of the hypotheses motivating the dependent variables (e.g., memory is weaker in LD), but also an implicit test of the independent variables (i.e., criteria for identifying children with and without LD) that derive from a hypothetical classification (Morris & Fletcher, 1988).

Even classifications that seem more straightforward, such as those used for defining children with and without traumatic brain injury, represent hypotheses at the level of the independent variables. To continue the memory performance example, if groups with and without traumatic brain injury differ in memory performance, evidence accumulates for the hypotheses that (a) memory is impaired in children with traumatic brain injury and (b) the criteria for defining traumatic brain injury are valid. The latter evidence would support the hypothetical classification of children along dimensions of brain injury (loss of consciousness, duration of coma, neuroimaging findings). Such evidence could be used to expand the classification towards hypothetical definitions of levels of severity (mild, moderate, severe); this classification and the criteria that lead to identification of children into severity groups could also be systematically evaluated along multiple dimensions: cognitive functions, prognosis, and response to intervention. The capacity of the classification to account for all children with traumatic brain injury (coverage) and to validly discriminate traumatic brain injury from other forms of brain injury (e.g., strokes, tumors) could also be evaluated. The keys are to recognize that there is a classification, to make it explicit, and to evaluate its reliability, validity, and coverage. When variation occurs in cognitive function, prognosis, or response to intervention among individuals with different levels of severity of traumatic brain injury, we can establish that the hypotheses leading to selection of these dependent variables were valid, but also that (a) the classification of injury severity has validity, and (b) the criteria used to operationalize the definitions of injury severity have validity (Fletcher et al., 1993).

In the area of LD, classification occurs at multiple levels: in identifying children as LD or typically achieving; as LD versus mentally deficient; within LD, as reading versus math impaired. Across classes of putative childhood conditions that produce underachievement, LD is identified as a particular type of "unexpected" low achievement and is distinguished from types where low achievement is expected due to emotional disturbance, social or cultural disadvantage, or inadequate instruction (Kavale & Forness, 2000). From a classification perspective, these levels of classification and the notion of LD as a form of low achievement that is unexpected represent hypotheses that should be evaluated.

That there are multiple underlying classifications of LD that are essentially hypotheses has not been consistently recognized. When the criteria for identifying LD began to evolve into policy in the 1960s, there was little research on which to base the underlying classifications and resultant definitions. This situation has gradually changed over the past 30 years, but the research that has emerged has had little impact on policy at the federal, state, and local levels. Indeed, the persistence of common assumptions about LD, its classification, and the perpetuation of resultant identification procedures are surprising given what has been learned about these disorders (Lyon et al., 2001). As we turn to research on the classification of LD, the question of how classifications should change as knowledge advances will emerge as a challenge to the field.

Pages: | 1 | 2 | 3 | 4 | 5 | 6 | 7 |

Return to LD Summit papers table of contents