Abstract: The creep rupture testing of notched specimens is examined in a new light through the combined study involving creep rupture tests, metallographic observations and finite element computation. The following concepts are re-examined and revised:1. Notched creep rupture testing methods have been specified in the Standards of many countries. This test is traditionally thought of as an evaluation of the "notch sensitivitsy" of materials working in creep range. Our study points out that this test actually reflects the ability of material to resist creep crack initiation and propagation under multiaxial stress state.2. The theoretical stress concentration factor K_t is the primary factor in defining the geometric parameters of standard specimens. This study points out that the actual stress concentration in the specimen is much lower than that indicated by K_t. In fact, K_t reflects the degree of notch constraint and stress multiaxiality in the specimen rather than the initial stress concentration.3. In the Standards of many countries only cylindrical notched specimens are adopted; whereas in ASTM E 292-69, both cylindrical and flat-plate specimens are allowed. Our study points out that those two types of specimens are quite different from each other in many aspects. Thus, it is not desirable to use only cylindrical specimens for all cases, in some cases the flat-plate specimen may be closer to the service condition.