Screening refers here to the process of making an initial selection of those who need further diagnostic evaluations. A screening process which effectively selects true learning disability (LD) cases from a population reduces the number of incorrectly selected non-LD children who would be needlessly given costly psychoeducational evaluations. Although there are several LD screening instruments in wide use, empirical evidence of the validity of these instruments is sparse and often misdirected. Discussion of the merits of these instruments typically focuses on the relation of the screening device to the construct of learning disabilities. Instruments which have most successfully emulated the components of a full diagnosis receive favorable evaluations. Although this is problably a wise approach in instrument development it is not the most important consideration in the evaluation of the effectiveness of the screening instrument in actual use. The empirical validity of a screening instrument is a function of correct classification, and this validity can be demonstrated by a simple 2 x 2 classification matrix. Validity data on Myklebust's Pupil Rating Scale is re-analyzed to demonstrate that high content validity does not automatically provide high discriminative validity. (Author/BW)