1. We analyzed the light-evoked responses of retinal neurons by means of a white-noise technique. Horizontal and bipolar cells produced a modulation response that was linearly related to a modulation of the mean luminance of a large field of light. The first-order kernels were capable of reproducing the cells' modulation response with a fair degree of accuracy. The amplitude as well as the waveform of the kernels changed with the change in the mean luminance. This is a parametric change and is a form of field adaptation. As the time constant of the parametric change was much longer than that of the modulation response (memory), neurons were assumed to be at a dynamic steady state at a given mean luminance. 2. With the presence of a steady annular illumination, the first-order kernel derived from stimulation with a small spot of light became faster in peak response time and larger in amplitude. For horizontal-cell somas and bipolar cells, the surround also linearized their modulation response. This surround enhancement has been seen in all the cone-driven retinal cells except the receptor and horizontal cell axon, in which a steady surround decreased the amplitude of the spot-evoked kernel but shortened the peak response time. 3. A change in the modulation depth did not affect either the amplitude or the wave-form of the first-order kernels from the horizontal and bipolar cells. In the amacrine and ganglion cells, on the other hand, the amplitude of kernels was related inversely to the depth of modulation. These cells were more sensitive to the modulation of a small modulation depth. 4. A static nonlinearity appeared when signals were transmitted to the amacrine cells. The nonlinearity was first produced in the type-C amacrine cells by a process, which could be modeled by squaring the bipolar cell response. A gamut of more complex second-order nonlinearities found in type-N amacrine cells could be modeled by a band-pass filtering of the type-C cell response. Linear components in the bipolar cells and nonlinear components in the amacrine cells are encoded into spike trains in the ganglion cells. Thus, under our simple stimulus regimen, the ganglion cells transformed the results of the preganglionic signal processing into a spike train without much modification. 5. We propose a tentative diagram of the signal flow in the cone-driven catfish retinal neurons based on this and previous studies.(ABSTRACT TRUNCATED AT 400 WORDS)