We investigated how two surface properties--gloss and bumpiness--interact in judgments of the perceived gloss and the perceived bumpiness of 3D textures. An ideal observer for gloss would ignore changes in bumpiness and an ideal observer for bumpiness would ignore changes in gloss. We used a conjoint measurement design to determine how much each property influences the judgment of the other. Methods: Stimuli were 3D textures composed of 400 overlapping ellipsoids. Ellipsoid heights were drawn from a uniform distribution whose range determined surface bumpiness. Each surface had one of five levels of bumpiness and one of five levels of gloss (i.e., five sets of parameters of the Ward reflectance model used to create the images). We rendered stereo images of the 25 possible surfaces under a rectangular light source at a fixed location relative to the surface. Six observers viewed the 325 possible pairings of surfaces and judged which surface of each pair was bumpier; six observers viewed the same pairs but judged which of each was glossier. We fit an additive conjoint measurement model to each observer's choice data using a maximum-likelihood criterion. Results: Increasing surface gloss made surfaces appear bumpier. Gloss increased perceived bumpiness by about 0.14 (ranging from -0.11 to 0.16 across observers) of our range of bumpiness. Increasing surface bumpiness also led to an increase in perceived surface gloss. Bumpiness increased perceived gloss by about 0.27 (ranging from 0.03 to 0.68 across observers) of our range of gloss. In both conditions, four out of six observers showed a significant effect of one cue on judgments of the other cue. While the visual system clearly attenuated the effect of each cue in judging the other, there is still significant interaction.