The problem of variability of data, even in presumable homogeneous subgroupings of handicapped children, which contaminates the results of many special education curriculum evaluation studies is examined with a study of 103 second-grade learning disabled students. The distortion of statistics calculated on samples of extreme variability; coupled with polymodality, skewedness, or kurtosis, is discussed. The case study approach is offered as one way to help balance the presentation of results. An example is presented using standardized test data and longitudinal criteron-referenced measurement data from both self-contained classes and resource rooms collected in 1971-1972. Approximately two-thirds of the report consists of tables and statistical data. (Author/IM)