Multiple-baseline design in instructional research: pitfalls of measurement and procedural advantages.

The multiple-baseline design has utility for evaluating the instructional programs used with mentally retarded persons; however, there are several pitfalls of measurement that may be encountered in using this design. Subjects' performance may be inadvertently altered by (a) repeated testing during baseline, (b) a procedural contrast between training and testing, and (c) inaccurate generalization during testing. Procedures to mitigate the effects of these problems were recommended. The pitfalls may arise in part because of the direct interpersonal nature of measuring the dependent variable in instructional research. Also, the pitfalls are not inherent in the multiple-baseline design per se but are issues of measurement that may occur in other designs as well. Finally, the advantage of using the multiple-baseline design to study covariation of responding was highlighted.