A great variety of noise models has been proposed for nonlinear circuit elements. The topic is unfortunately in some intellectual disorder. I would suggest it might be helpful to examine existing models to determine if they predict circuit behavoir consistent with accepted thermodynamic principles. Only models that pass this test should be kept as candidates for representing particular devices.

Three types of white noise models will be examined in detail:

• the Nyquist-Johnson linear resistor model,
• proposed generalizations of the above for nonlinear resistors, and
• a shot-noise model.

I hope to convince the audience that the Nyquist-Johnson model passes this test entirely, in that arbitrary nonlinear LC circuits driven by this model display all the equilibrium and nonequilibrium behavoir thermodynamics leads us to expect. The proposed extensions to the nonlinear case fail the test conclusively. The test leads to a unique shot-noise model for any given v-i curve. The shot-noise model passes an elementary version of the test. The mathematical techniques needed for a more general test of the shot-noise model are still being worked out.