DescriptionLine drawings lack direct 3D depth information, yet human vision easily perceives the 3D shapes from the contours. This dissertation investigates the mechanisms underlying the 3D shape inference from 2D line drawings. Here, four psychophysical experiments and a computational model for the 3D shape inference are discussed. Experiment 1 shows that human responses in depth judgments for line drawings reflect an underlying uncertainty of the perceived 3D shape, which is based on the complex interaction of local and global depth cues propagated from the contours. The computational model estimates the posterior probability of possible 3D surfaces from the contours of a line drawing in a Bayesian framework. The comparison of the model predictions and human depth responses for the line drawings from Experiment 1 demonstrates that the model accounts for the probabilistic 3D shape interpretation of line drawings by human vision. Experiment 2 shows that the reliability of a contour segment in a line drawing as a meaningful depth cue is conditional to the complex global context. Experiments 3 and 4 show that the certainty of depth difference perceptions from partial line drawings increases as more non-local visual cues are available. The experiments and the model offer a new perspective on 3D shape perception from line drawings as an inference based on the probability over possible 3D shapes given the contour cues, providing a broader understanding on the mechanisms of human vision.