A number of parts-based theories of 3D object representation (e.g., Marr & Nishihara, 1977; Biederman, 1987) hold that simple object parts can be modeled as generalized cones (GCs). GCs are the volume created by sweeping a cross-section along an axis as, for example, when a circle is moved along a straight axis to produce a cylinder. Different volumes can be produced through variations along different dimensions, such as axis curvature, the size variation of the cross-section during its sweep, and aspect ratio. GC dimensions may be coded independently, as shown by Stanckiewicz (2002), who reported that discrimination of noisy variations in one GC dimension could be performed independently of the noise level on another GC dimension. Kayaert et al. (2005) showed that 95% of the variance of the firing of macaque IT cells to 2D shapes could be accounted for by independent representation of the GC dimensions. (Sweeping a line along an axis will produce a 2D GC shape, such as a rectangle.) Because both the psychophysical subjects and the monkeys live in environments replete with geometrically regular shapes, one could argue that the sensitivity to GC dimensions reflects familiarity rather than being a function of robust statistics of images that would characterize any natural environment. The Himba, a semi-nomadic people in a remote area of Namibia, have little exposure to developed-world artifacts. They performed a texture discrimination task in which they had to judge whether the boundary between two texture regions composed of macaroni-like shapes differing in aspect ratio (narrow-wide) and axis curvature (low-high) was vertical or horizontal. For both Westerners and the Himba, when the boundary was defined by variation in a single shape dimension, performance was markedly superior to when the boundary was defined by variation in two dimensions.