Under conditions where an object is inside another object and only a single face is visible, is there a bias to assume smooth continuation of the surface straight back into the object? To examine the ability to estimate how features progress into a volume, participants viewed 16 pictures of everyday objects (rocks, food, wood) presented with only a single face visible (see Figure 1). Participants reported whether a highlighted region of a picture was present on the surface or extended into the object. If they perceived the region as extending in, they positioned a rod to indicate the angle. Surface responses were rare and instead participants’ readily perceived 3D forms from 2D views. Inspection of frequency histograms revealed a systematic bias to estimate the angle of extension in the 80-110° range. This type of completion process suggests constraints on models of visual completion and has implications for STEM education, in particular, how students deal with ambiguity.