Laboratory simulations of legal decisionmaking present greater problems of generalizability and external validity than does real-world research, but the choice of a method in real-world research determines the effectiveness with which it deals with the same problems. In most cases of real-world research, an archival analysis is to be preferred to interviews, questionnaires, and the observation and coding of the hearing. An archival analysis consists of the coding of at least two kinds of written materials: the transcript of the hearing and the file containing a variety of documents pertaining to the case. In archival analysis, the coding categories used should be similar to those used by the participants in the real-world system rather than derived from currently popular social-psychological theories. Further, coding should be as exhaustive as possible. The statistical analyses should examine the effects on the criterion decision of various combinations of predictors, with a relatively large number of predictors in each predictive set, so that both main effects and interactions can be captured. Prior to archival analyses, a sufficient amount of background research should have been done by the investigators concerning the daily operation of the system so as to leave no doubt that the file being coded is at the disposal of the decisionmaker prior to the targeted decision. The categories of legal decisionmaking most attractive for research are those that account for the greatest percentage of variance in the processing of cases through the system as a whole. A flow chart indicates decisionmaking points in criminal justice processing and the percentage of variance at each point. Fifty-one references are listed.