Presented is a procedure that could be used by a school system to identify by the end of first grade those children who are likely to underachieve (achieve or overachieve) in reading by the third grade. The procedure consists of (1) Defining underachievers, achievers, and overachievers through use of regression and its standard error, (2) Obtaining the measurements to be used in constructing a discriminant model for predicting underachievers, achievers, and overachievers, (3) Constructing an "optimal" model through use of the stepwise discriminant procedure (BMDO7M), (4) Validating the model through use of an independent sample, (5) Using the model to make predictions, and (6) Updating the model periodically. For this study, an underachiever was a child who had not gained in reading at a rate comparable to others at his reading level. Table II identifies the predictor variables utilized by the discriminant model and the coefficients and constants which enable the model to discriminate among the groups of interest. It is pointed out that if one can identify potential underachievers, intervention programs may be designed to attempt to overcome the underachieving tendency. (CK)