Abstract: A horizontal plexiglas tube containing a food-reward was presented to four naive tufted capuchins and suitable sticks were provided to push the reward out. Three monkeys out of four spontaneously used the tools and showed very different styles of solving the task. In more complex conditions, in which the sticks needed to be combined or actively modified in order to become effective, the monkeys were always successful; however, their performance was loaded with errors which did not disappear throughout the trials. Evidence of a difference between success in solving the problem and its understanding was found. This suggests that although capuchins can discover new means through active experimentation, they do not mentally represent the characteristics necessary for a tool to be effective, nor do they modify the tool appropriately beforehand. At this level, a major difference with chimpanzees emerges.