Intuitively it may seem likely that orientation-modulated (OM) and frequency-modulated (FM) textures are processed utilizing the first-order channels that are most responsive to the first-order (luminance) information contained in the textures. This assumption would imply that the detection or segmentation of OM or FM textures is accomplished by second-order mechanisms that receive their first-order input from neurons tuned to either the center, or to the peaks in the orientation and spatial-frequency distribution of the texture. Here we show that at low depths of modulation this is not the case. Using an adaptation paradigm, we show that the first-order filters involved in the perception of OM and FM textures are those which maximize the differential response between the different texture regions. Our explanation of this result is similar to that made by Regan and Beverley [J. Opt. Soc. Am. 73 (1983) 1684; J. Opt. Soc. Am. A 2 (1985) 147] for simple grating stimuli. However, we show that whereas Regan and Beverley's results could be accounted for on the basis of the tuning functions of the putative mechanisms involved, our results can be explained in terms of the characteristics of the textures themselves. Some implications of our finding are discussed.