It is now widely recognized that the nonlinear kinematic hardening rul
e of Armstrong-Frederick (AF) type (1966), although proven very accura
te for description of most first order effects associated with cyclic
plasticity, is deficient in its representation of second order cyclic
strain accumulation (ratchetting) under asymmetric loading with mean s
tress ([Clement & Gionnet, 1985]; [Bower, 1989]; [Bower & Johnson, 198
9]; [Chaboche, 1989, 1991]; [Chaboche & Nouailhas, 1989]; [McDowell &
Lamar, 1989]; [Lacey, 1991]; [Gionnet, 1992]). This inadequacy is mani
fested by unrealistic unloading-reloading behavior associated with sub
cycle events under even uniaxial conditions [Chaboche, 1989]. In parti
cular, the reloading compliance predicted by the AF rule significantly
exceeds that observed experimentally. Likewise ratchetting rates are,
in general, overpredicted by this rule. In this paper, the nonlinear
dynamic recovery modification of the AF rule recently proposed by Ohno
& Wang [1991a, 1991b] is considered. Several important criteria are d
iscussed for models capable of representing stress state and amplitude
dependence of ratchetting behavior. Experimental results obtained on
a carbon rail steel subjected to various uniaxial and nonproportional
loading conditions are presented and correlated with an extension of t
he Ohno & Wang model which accounts for a broader range of stress stat
e and amplitude effects.