As explained in the introduction, cognition is one of the most important topics for multimodal systems. Without it, data can not be interpreted sufficiently by a computer, and the meaning and intention of the user's actions will not be recognized by the machine. Therefore, cognition is the key issue to separate multimodal research from the more technically oriented multimedia and VR research. We start this chapter with a short overview on cognition in humans, followed by a part on learning in humans as well as in computers. In the last section, we will review the concept of intelligent agents which is today's most interesting approach to realize `intelligent', autonomous, and intentional behavior.