When we see an optical pattern that has a gradient of the size and/or density of its texture elements, we often perceive a surface that is slanted in depth. Our inquiry was to ask whether the effect of a texture gradient depends on the direction of the gradient (ground, ceiling, and sidewall patterns) or on the position of the observer’s head (upward, forward, or downward). In Experiments 1 and 2, a total of 63 observers judged the apparent slant of polka-dot, grid, or flagstone patterns; regardless of head position, the ground patterns were judged to be closer to the frontal plane than were the other patterns. This means that there is a visual anisotropy in head-centric slant perception. To explain this result, we assumed accumulated positional effects of size contrast—that is, a tendency to perceive the size of the upper part of visual space to be larger than the size of the lower part. This hypothesis was examined in two subsequent experiments by reducing the size contrast among the texture elements. When 23 observers viewed regularly arranged same-sized-dot patterns with gradients of the interdot interval and with linear perspective of the dotted lines, anisotropic effects were still detected. When 22 observers viewed dynamic random-dot patterns with gradients of velocity, the anisotropic effects mentioned above were removed in many cases, and the ceiling patterns were sometimes judged to be less slanted than the other patterns. These results partially support the size contrast hypothesis and were compared with the predictions from other hypotheses.