This paper reviews research on the classification of students with learning disability (LD). It first examines the evolution of LD definitions and the evidence for three components of LD classification: discrepancy, heterogeneity, and exclusion. It suggests that classification hypotheses involving discrepancy and exclusion have weak validity and often represent inaccurate and outdated assumptions about LD. The paper next reviews some alternative classification hypotheses and argues that inclusionary definitions that specify attributes of different forms of LD are more desirable than exclusionary definitions and that inclusionary definitions permit identification procedures that are treatment oriented and focus on prevention. Finally, the paper discusses some difficult issues in relating policy and practice with research, essential for improved identification and treatment of LD. The paper urges the development of a consensus process to devise a new classification of LD with inclusionary definitions, the possibility of prevention, the use of response to intervention as a major factor in evaluating the disability, and an emphasis on results and outcomes. (DB)