When young children copy a vertical line that protrudes from the midpoint of a pre-drawn oblique baseline, they tend to perpendicularise this line. We measured this bias for standard figures and for symmetrical angular figures. The 5- and 6-year-old participants produced the expected bias for the standard figures. However, these same children showed no such bias when tested with angular figures. This suggests that the mechanisms involved in coding angular information directly are not prone to the perpendicular bias.