A procedure which may be used to project the frequency distribution of one test onto that of another test is described and illustrated. The procedure is useful when a test developer wishes to construct an alternate form with preferred distributional characteristics. For example, the test developer may wish to construct a new test form with a completely different set of items but which yield comparable measures of central tendency, dispersion, skewness and kurtosis, reliability (KR-20, KR-21, Kappa), norms, pass-fail rates, etc. Rather than constructing alternate parallel tests, the test developer may wish to construct alternate forms which are substantially easier or harder. The procedure is also applicable in that situation. The technique assumes that items on both tests have been previously calibrated and equated with the one-parameter Rasch model. The three-parameter logistic model may also be used when number-right rather than pattern scoring is used. A basic skills multiple choice test consisting of 21 items or 32 items provides the illustration. (Author/GDC)