We examine the shape of hand and forearm occlusion on a multi-touch table for different touch contact types and tasks. Individuals have characteristic occlusion shapes, but with commonalities across tasks, postures, and handedness. Based on this, we create templates for designers to justify occlusion-related decisions and we propose geometric models capturing the shape of occlusion. A model using diffused illumination captures performed well when augmented with a forearm rectangle, as did a modified circle and rectangle model with ellipse "fingers" suitable when only X-Y contact positions are available. Finally, we describe the corpus of detailed multi-touch input data we generated which is available to the community.