In the recent decade statistical literature is enriched by a good number of adaptive design methods which are proposed particularly for applications of clinical trials. Of these designs, adaptive selection designs appear to be increasingly seen in regulatory applications, such as selection of a subgroup of patients to target for statistical testing, selection of dosing regimens for study. The selection may be based on some marker(s) or clinical endpoint(s). This presentation will pinpoint the essential features of such designs and their potential utilities and pitfalls. In fact, such designs are special cases of multi-stage designs. A number of case examples will be used for illustration.