A considerable amount of research demonstrates that people perceive cardinal orientations (horizontal and vertical) more accurately than other orientations; this is termed the oblique effect. We investigated the interaction of this effect with the degree of elongation of the stimulus. Our stimuli were ellipses with a wide range of aspect ratios, varying from a circle (aspect ratio = 1) to a line (aspect ratio = 123.5). The task was to set a probe line in the same orientation as the long axis of the ellipse. In our first experiment, we determined that performance is degraded as the aspect ratio decreases; furthermore, the bias and response variability are linearly related to a transformation of aspect ratio (roundness). We found significant individual differences; the results show high within-subjects correlations and low between-subjects correlations. In our second experiment, we had observers judge the orientation of circles randomly mixed in with ellipses of low aspect ratio. The observers demonstrated intrinsic preferences and generated reproducible distributions of orientation settings with idiosyncratic profiles. These distributions predict the influence on the response to ellipses with an aspect ratio higher than one and can be considered as the beholder's share in the perception of shape orientation.