Daylit architecture is perceived as a dynamic luminous composition, yet most existing performance metrics were designed to evaluate natural illumination for its ability to adequately illuminate a two-dimensional task surface and avoid glare-based discomfort. It may be argued that task-driven approaches based on surface illumination and glare ignore the likelihood that contrast can provide positive impacts on our visual perception of space. Advances in these metrics to accommodate climate-based sky conditions and occupant behavior have improved our ability to evaluate task illumination and glare, yet the same attention has not been paid to evaluating positive perceptual responses to daylight. Existing studies have attempted to link subjective ratings of composition to simple global contrast metrics without reaching consensus. More advanced metrics have been developed in computational graphics and vision fields, but have not been applied to studies in qualitative lighting research. This paper introduces the results from an online experiment where subject ratings of daylight composition are compared to quantitative contrast measures across a series of renderings. This paper will identify which measures correlate to subjects’ ratings of visual interest, and introduces a modified contrast algorithm, which can be used as a novel prediction model for visual interest in daylit renderings.