We report five psychophysical experiments that employed a cue conflict paradigm to investigate integration by the human visual system of surface shape information from stereo, texture and outline cues. The experiments used convex parabolic and triangular three-dimensional ridge stimuli, with amplitudes (base to peak) in the range 3-9 cm, viewed from 57 cm. The observers' task was to judge ridge amplitude using a scale of two-dimensional drawings of ridge profiles. Cue integration was studied using both vertically and horizontally oriented ridges and both real ridges and stereograms of ridges. The main findings were: (a) stereo strongly dominated all horizontal ridge stereograms; (b) texture and outline cues strongly dominated low (3-6 cm) but not high (9 cm) amplitude vertical ridge stereograms; (c) stereo strongly dominated all real ridge stimuli. These results are evidence against explanations of the vertical/horizontal stereo anisotropy which propose that it derives from stereo mechanisms being tuned only to disparity cues with non-zero second-order spatial derivatives or to disparity discontinuities. They also show that radically different results can be obtained when stereo mechanisms are explored using stereograms and real surfaces and possible reasons for this are discussed.