This study examines the impact of manipulation on the performance of 3.5- and 4-year-old children in the Give-N task with concrete representations (sets of bottle caps and pictures of dots) and spoken number words. In this task, children were asked to give a certain number of items (cookies) using the three representational formats. Children were assigned to one of two conditions: manipulation (children could physically interact with the objects and pictures) or observation (children watched as the experimenter presented the representations). Results show that manipulation had different effects according to the age of the children. For 3.5-year-olds, manipulation with objects but not with pictures consisted in task-irrelevant actions, and hindered performance. For 4-year-olds, however, manipulation with both concrete representations were task related and did not disrupt performance. The mere contemplation of the objects and pictures, without physically interacting with them, facilitated performance in both age groups, in comparison with number words. Emerging cardinal comprehension seems to be influenced not only by the kind of representations used by children, but also by the type of physical exploration afforded during the task.