In this paper we describe four experiments concerned with spatial-interval discrimination. The first experiment demonstrates that perturbations in the internal light distribution of two bright bars alter the perceived separation of the bars without influencing the precision of spatial-interval discrimination. Specifically, alterations in the centroid of the luminance distribution of the bars were found to influence their perceived separation systematically. In the second experiment we investigated separation discrimination for a pair of bright lines, a pair of dark lines, and a pair of lines with opposite polarity (a bright-dark pair). For large interline separations, the thresholds for separation discrimination of the three stimuli were equivalent. However, for small separations (less than 4 arcmin) the thresholds for the opposite-polarity pair were significantly higher than those for either bright or dark lines. The third experiment showed that separation discrimination is also impaired when the direction-of-brightness change of adjacent edges is the same. In the fourth experiment, random luminance perturbations were introduced in the interval between two bright bars separated by 1.5 to 8 arcmin by the addition of a dim line within the interval. The effect of these perturbations was also to alter the perceived separation of the bars with no effect on the precision of spatial-interval discrimination. Increasing the intensity of the perturbating line (to equal that of the outer lines) elevated thresholds for small separations (less than 4 arcmin) to approximately 6 arcsec but had no effect on the threshold for larger separations. Taken together, these results suggest that large and small separations are processed differently and place strong constraints on models for spatial-interval discrimination.