Intensity and colour variations in images are known to be of high importance to the human visual system. These pictorial variations can be due to many different causes (e.g., shape, material, illumination). We seek to precisely formulate the different contributions to observable shading gradients and quantify their interactions (mixing, masking, reinforcing...). This will provide new computer graphics applications, as well as knowledge of how pictorial variations are composed, which provides new insights into how the visual human system decomposes images into different causes.  In order to achieve this we will use a first-order (gradient) analysis of the rendering equation, directly in the picture plane. Deriving a formula for the gradient of reflected radiance in terms of gradients of its constituents will provide the basic foundations for computer graphics applications and perceptual experiments in collaboration with other partners. The components of the rendering equation may be represented in a variety of ways, and we will investigate different models and basis functions to represent materials and illumination respectively. To validate our approach, we will compare our analytical derivations with numerical estimates in synthetic 3D scenes. Once we will have understood pictorial variations we will be able to make use of this knowledge to design novel lighting environments targeted to specific goals such as legible shape and material depiction, or exaggerated and stylized rendering.