Hysteresis loops were recorded intermittently but periodically during completely reversed, strain-controlled low cycle fatigue tests of four kinds of steel : a mild steel and three high strength steels. The longitudinal strain amplitude ε_±a was controlled over the gauge length of 12.5 mm on 8 mm diameter test section of the smooth, solid cylinder type specimen. After the conversion of the loops to true stress versus natural strain hysteresis loops, cyclic characteristics were studied on true stress amplitude, hysteresis energy and hysteresis loop shape. Main conclusions obtained are as follows : (1) The transition strain amplitude ε^* _ta , on which neither cyclic hardening nor cyclicsoftening is occurred, is expressed as a function of ultimate tensile strength σ_u (kgf/mm²). log ε^* _ta =0.0478 σ_u -4.544 (2) Although plastic strain hysteresis energy per cycle Δ W behaves irregular and remarkable change in the beginning stage of cycling, that is approximately less than 2 percent of the crack initiation life, the change of Δ W is neglegible order after the beginning stage. The fatigue toughness W_f (kgf/mm²) is expressed as a function of ultimate tensile strength σu and visible crack initiation life N _c. W _f = (94.93-0.347σ_u) N _c^0.335 (3) The ratio of Δ W to the product of true stress range and plastic strain range is not constant during the fatigue cycling, and the mean value of the ratio over the total cycling for a specimen is expressed as a function of the total strain amplitude. (4) The constants a and n in the n-th power strain hardening formula are expressed as a function of N and ε _ta for the nearly total range of cycling excluding the beginning stage. An empirical formula was found between a and 1/ n . 1/ n = (log a +1.623+0.376log σ_u)/(0.1539-1.169 logσu) (5) An estimation procedure for crack initiation life was proposed on the basis of the fatigue toughness, showing an estimation error less than 28 percent.