Because stereoscopic depth cues are not present in an ordinary photograph, matching a two-dimensional (2-D) photograph to a three-dimensional (3-D) face requires establishing a correspondence between information derived from monocular depth cues and ignoring discrepant information from
stereopsis. To examine the implication of this transfer of information across 2-D–3-D representations, we had participants perform identification tasks where only one of the two images for training and test contained stereo information. We compared identification performance for this
incongruent condition of dimensionality to two congruent conditions where images for both training and test were shown with or without stereo information. Results revealed that performance in the incongruent condition is overall 4–5% less accurate for both a recognition task (Experiment
1) and a sequential matching task (Experiment 2). This small cost of depth cue inconsistency suggests a degree of encoding specificity for 2-D and 3-D information in face recognition.