The objective of this research was to determine fatigue behavior of SAE
1045 steel subjected to very high tensile mean stress for unnotched, mildly notched, and sharply notched test specimens, and to determine if common S-N
f and ε-N
f mean stress fatigue life models are applicable. High tensile mean stress fatigue tests for R ratios of 0.8 and 0.9 were conducted using unnotched and notched, K
t=1.65 and K
t=3.65, axial loaded SAE
1045 steel specimens with hardness levels of Rc=10, 37, and 50. The monotonic notch strength ratio, NSR, for 5 of 6 test conditions was greater than 1, which allowed many notched cyclic test values of S
max or S
m to exceed the unnotched ultimate tensile strength. Much notched specimen fatigue resistance at these high R ratios was superior to that of unnotched specimens. However, cyclic creep/ratcheting, particularly for Rc=10 and 37, was a predominant cause of failure. Scanning electron microscopy examination revealed many cyclic failures were similar to monotonic failures consistent with cyclic creep/ratcheting measurements. Specimens failed due to cyclic creep/ratcheting, to an interaction between cyclic creep/ratcheting and fatigue, or to surface fatigue crack nucleation and growth only. Fatigue life calculations based on S-N
f and ε-N
f models were up to 5 orders of magnitude in error, and thus completely unreasonable at these high R ratios. For a given test condition, this inaccuracy was due to cyclic creep/ratcheting, scatter, notch strengthening, material properties used, inherent inaccuracy of S-N
f constant life diagrams at high R ratios, and/or extrapolation of the Ramberg-Osgood equation beyond S
u and σ
f.