It is known that visual direction representation is more accurate for cardinal directions compared to oblique, a phenomenon named the "oblique effect". It has been hypothesized that there are two sources of oblique effect, a low level one confined to vision and a high level one extending to different modalities and corresponding to higher cognitive processes. In this study directional error (DE) was measured when normal individuals tried to align the direction of an arrow presented in the center of a computer monitor to the direction of a peripheral target located in one of 32 directions equally spaced on an imaginary circle of 60 mm radius. Task difficulty was manipulated by varying arrow length (15, 30, 45 and 60 mm). By measuring mean DE and its variance we identified two independent sources of the oblique effect. A low level oblique effect was manifested in higher accuracy or equivalently lower variance of DE in the alignment for cardinal orientations compared to oblique. A second oblique effect was manifested measuring mean DE resulting in space expansion in the vicinity of cardinal directions and space contraction in the vicinity of oblique directions. Only this latter source of oblique effect was modulated by arrow length as predicted from a theoretical model postulating that this oblique effect is produced by a cognitive process of 2-D space categorization.