It is well-known that biasing a learning system through the use of constraints can dramatically improve the performance of the system. We study the idea that developmental events constrain the internal representations acquired by a learning system in useful ways. We advocate the "less is more" hypothesis of Newport (1990) which states that perceptual or cognitive limitations during early stages of learning are actually helpful because they bias learners to initially acquire simple representations which are suitable building-blocks for the subsequent acquisition of more complex representations. We consider this hypothesis in two domains. One project studies the role a developmental change in visual resolution (from coarse-scale to multi-scale) may play in the acquisition of sensitivities to binocular disparities in pairs of visual images or of motion velocities in sequences of images. A second project studies the role a developmental change in free parameters of a motor control system may play in motor learning.