The success of a self-generated identification code for linking longitudinal questionnaire data was examined. The matching procedure developed for linking questionnaires, including a simple technique to compensate for nonidentical codes, yielded a high success rate (92% linkage of cases over a one-month interval and 78% over a one-year interval) and very few incorrectly linked cases. The procedure worked equally well with elementary and high school students, and the resulting samples were representative of the student population on a wide range of measures. Some suggestions are offered regarding the elements comprising self-generated codes.