Most research on infants' development of object and spatial concepts is based on object search tasks or visual tracking studies. In this review, it is concluded that many phenomena, particularly the tracking data, can be accounted for in terms of objective spatial perception in early infancy. However, some phenomena, particularly the stage IV search error, cannot be explained in these terms. A theoretical framework is put forward that integrates aspects of theory of direct perception with constructionist theory. First, it is proposed that objective properties of the basic spatial array are directly perceived, but that because perception is derived from action in the environment, there are initial limits to the scope and coherence of perception that diminish as the infant gains effective motor control. Second, it is proposed that in attempts to decipher the structure of social interchanges over objects, infants construct functional categorizations of objects that allow them to overcome limited awareness of their physical properties. Finally, it is suggested that these functional concepts provide an adequate means of interpreting the world, so that concepts based on the physical properties of objects, including their permanence, are relatively late developments.