The detection of minimal residual disease (MRD) represents a major goal in leukemias with great potential applications: identification of patients at high risk of relapse, earlier detection of relapses, possible improvement of autologous bone marrow transplantation (determination of the best time for bone marrow harvest and assessment of its quality). Theoretical requirements for a model of MRD detection exist and are recalled. Numerous methods (cytology, immunology, cytogenetics, clonogenic assays, molecular biology) have been used. All share limitations which may concern sensitivity, specificity, reproducibility, conveniency. Progresses will be brought by more sophisticated methods in immunology (double staining with or without flow cytometry) and above all molecular biology (revolutioned by polymerase chain reaction applications).