Fatigue tests were performed using circumferentially-notched, round bar specimens with a stress concentration factor, Kt, of 6.6 for Type 304, metastable, austenitic stainless steel. The tests were carried out in ambient air and in 0.7 MPa hydrogen gas at room temperature. In a relatively higher stress amplitude regime (i.e., the amplitude resulting in N-f < 10(5)), the hydrogen gas environment caused a marked degradation in fatigue life. In contrast, in a relatively lower stress amplitude regime (i.e., the amplitude resulting in N-f > 10(5)), it appeared that fatigue life did not differ between air and hydrogen gas. It was also confirmed that the fatigue limit appeared not have been degraded in the hydrogen environment though there was a slight difference between the data obtained in two environments. The fatigue life curve and fatigue limit were predicted by assuming that the notch was equivalent to a circumferential crack. Consequently, there was a significant disparity between the prediction and the experimental results. As a result of microscopic observations of the fracture process in combination with elastic-plastic finite element analyses, these discrepancies were attributed to (i) complex cyclic plastic deformation behavior under large-and small-scale yielding conditions within the vicinity of the notch root, (ii) the retardation of crack initiation in the finite life regime, and (iii) the absence of non-propagating cracks at the fatigue limit, all of which are typical characteristics of metastable austenitic stainless steel.