The light sources in a scene can drastically affect the pattern of intensities falling on the retina. However, it is unclear how the visual system represents the light sources in a scene. One possibility is that a light source is treated as a scene component: an entity that exists within a scene and interacts with other scene components (object shape and object reflectance) to produce the retinal image. The aim of this thesis was to test two key predictions arising from a perceptual framework in which light sources and the objects they illuminate are considered to be scene components by the visual system. We begin examining the first prediction in Chapter 3, focusing on the role of a dynamic shape cue in the interaction between shape, reflectance, and lighting. In two psychophysics experiments, we show that the visual system can "explain away'" alternative interpretations of luminance gradients using the information provided by a dynamic shape cue (kinetic depth). In subsequent chapters, the research focus shifts to the second prediction, investigating whether multiple objects in a scene are integrated to estimate light source direction. In Chapter 4, participants were presented with scenes that contained 1, 9, and 25 objects and asked to judge whether the scenes were illuminated from the left or right, relative to their viewpoint. We found that increasing the number of objects in a scene worsened, if anything, discrimination sensitivity. To further understand this result, we conducted an equivalent noise experiment in Chapter 5 to examine the contributions of internal noise and integration to estimates of light source direction. Our results indicate that participants used only 1 or 2 objects to judge light source direction for scenes with 9 and 25 objects. Chapter 6 presents a shape discrimination experiment that required participants to make an implicit, rather than explicit, judgement of light source direction. Consistent with the results reported in Chapters 4 and 5, we find that shape discrimination sensitivity was comparable for scenes containing 1, 9, and 25 objects. Taken together, the findings presented here suggest that while object shape and reflectance may be represented as scene components, lighting seems to be associated with individual objects rather than having a scene-level representation.