Researchers and program evaluators would often like to use a particular instrument, but do not because it is too long or would require too much testing time. Having a validated set of objective procedures for reducing the size of an instrument could improve many research and evaluation efforts. This paper reports the results of test reduction or shortening procedures devised and used in the development of parallel short forms of the Crandall Social Desirability Test for Children (CSDTC), a 48-item scale that measures a person's tendency to give socially desirable answers to questions rather than the subject's true views or feelings. The first procedure was to draw four random samples of 12 items (25% of the CSDTC) to produce parallel short forms based on a rationale from sampling theory. These samples were assessed through responses of 356 fifth and sixth graders in a Boston (Massachusetts) middle school to the full scale and shortened versions. A factor analytic approach was also used to make parallel short forms by taking the four items that loaded most highly on each of the three major factors from the full scale administration. Cross validation with 181 sixth and eighth graders; a test-retest reliability study with 55 seventh graders; and a validity study with 254 seventh, eighth, and ninth graders supported the short form derived through factor analysis. Results demonstrate empirically that long instruments can be effectively reduced, and that factor analytic strategies are superior to sampling strategies for this purpose. Six tables present form items and results from the validity studies. Four references are listed. (SLD)