Children 6 and 10 years of age and adults were asked how fast a toy car had to be to catch up with another car, the latter moving with a constant speed throughout. The speed change was required either after half of the time (linear condition) or half of the distance (nonlinear condition), and responses were given either on a rating scale (judgment condition) or by actually producing the motion (action condition). In the linear condition, the data patterns for both judgments and actions were in accordance with the normative rule at all ages. This was not true for the nonlinear condition, where children’s and adults’ judgment and also children’s action patterns were linear, and only adults’ action patterns were in line with the nonlinearity principle. Discussing the reasons for the misconceptions and for the action-judgment dissociations, a claim is made for a new view on the development of children’s concepts of time and speed.