The study examined the process by which diagnostic data on children thought to have learning disabilities (LD) are generated and used. Emphasis was placed on the validity of test instruments in common use in the field. Since part of the practical validity of any instrument used to identify an LD child rests on the validity of the LD construct itself, current definitions of LD were examined for their implications for evaluative measures. Fifteen tests were reviewed: four intelligence measures (Wechsler Intelligence Scale for Children-Revised, Denver Developmental Screening Test, Raven's Coloured Progressive Matrices, Slosson Intelligence Test); two tests of psycholinguistic abilities (Illinois Test of Psycholinguistic Abilities, McCarthy Scales of Children's Abilities); five visual and auditory perception tests (Bender-Gestalt Test, Slingerland Screening Test for Identifying Children with Specific Language Disability, Frostig Developmental Test of Visual Perception, Beery-Buktenica Developmental Test of Visual Motor Integration, Wepman Auditory Discrimination Test); and four standardized achievement tests (Wide Range Achievement Test, Canadian Test of Basic Skills, Gates-MacGinitie Reading Tests, Peabody Individual Achievement Test). Among common components of definitions in use were a disparity between some measure of potential and some measure of performance and an assumption of an underlying dysfunctonal learning process. The majority of tests in use had not been normed on an LD population, and no single test used alone was found to be a valid assessment tool. Visits to selected Ontario School Boards revealed that definitions and assessment procedures were not uniform across the province. (SBH)