Are young children's drawings canonically biased?

In a between-subjects design, 4- to 6-year-olds were asked to draw from three-dimensional (3D) models, two-and-a-half-dimensional (212D) models with or without depth cues, or two-dimensional (2D) models of a familiar object (a saucepan) in noncanonical orientations (handle at the back or at the front). Results showed that canonical errors were produced mainly by the youngest children in the 3D models copying condition. A large proportion of errors did not, however, reflect biases toward the participants' "own" view of the object; rather, these errors reflected biases toward their "best" view of the object. Drawing from models having fewer than three dimensions resulted in a significant increase in view-specific productions in all children for the back orientation. When a more complex front orientation of the object was proposed, sensitivity to model type developed gradually with age. Results are discussed in terms of the flexibility of the mental model that mediates between children's perception of a model object and the depiction decisions.