A projective method is viewed in the present work as a psychological test, by which is understood a standardized task that is set before a person, the solution of which is used for making inferences about psychological quantities. The present research was designed to shed light on several aspects of the usefulness of projective methods in longitudinal research, particularly with regard to reliability and to the distinction between measurement and theory-free empirical-comparison data. Data come from an ongoing longitudinal study of 212 Swedish children who were born in 1955-1958. Interrater agreement was good or excellent for all or almost all variables studied. An extended kappa coefficient for calculating overall chance-corrected interobserver agreement among the multivariate ratings of several judges was presented. Some temporal stability was found for many of the test phenomena and scores studied; however, difficulties in the interpretation of stability were pointed out. Part of the study concerned global personality ratings based on the Rorschach, which at face value seemed to consistently capture important personality characteristics; however, problems with their integrity were noted and steps toward their validation were suggested. Further results suggested that recoding of archival Rorschach records may be a fruitful procedure for obtaining scores that can be compared with current research findings, when the inherent limitations are respected. Finally a theory-based measurement model was fitted to empirical codings of children's human figure drawings. In doing so, this study joined a long-existing stage-development theory with a measurement model for the first time. The implications of these results were discussed, and suggestions for further research were made.