Composite measurement scales (CMSs) have contributed to improving the measurement of complex medical phenomena such as physical and psychological functioning or health-related quality of life. However, their use in patient care and research is often limited by their length and excessive respondent burden. In such situations, short instruments should be made available. Efforts to develop short instruments have largely focused on shortening existing instruments. To investigate the methodology currently used in the shortening of CMS, we assessed 42 studies reported in medical, psychological, and educational journals between 1984 and 1994. A number of methodological and statistical considerations important in the CMS shortening procedure were found to have been ignored or neglected by authors developing short forms from existing CMS. Serious flaws appear mainly to result from inadequate conceptualization of the shortening process, and inappropriate use and excess credit given to statistical techniques used to select items to be retained in short forms. When performed, the assessment of measurement properties of the short form was often inappropriate, and cross validation studies were seldom conducted. We propose recommendations for shortening existing CMS, to help authors and investigators develop and choose, respectively, shortened measurement instruments. These recommendations address the preliminary choice of the original CMS to be shortened, and the two successive phases to be considered in the development of short forms: the shortening process itself, where items are selected, and the validation of the shortened CMS, which should be conducted independently using independent subject samples.