ABSTRACT
From our stead-state flicker data, Kelly's (1971) model correctly predicts the transient thresholds for rectangular pulses of variable duration when the (flickering or flashed) stimulus is a 4 cycle/deg grating, but the same prediction fails for a uniform (8 degrees) field. However, if we augment the model with a "hard" nonlinearity, we can fit both types of transient thresholds as well as the steady-state thresholds. The most plausible embodiment of this essential nonlinearity is an asymmetric recifier, which seems to represent the behavior of retinal ganglion cells. Unlike the symmetric models of Roufs and Rashbass, this asymmetry also correctly predicts that the decrement thresholds for some stimuli are smaller than the corresponding increment thresholds.
