One of the most influential approaches in studying information systems in the cognitive sciences is the one proposed by David Marr. Marr asserts that giving a full account of an information system implies: (i) explaining what the system does and why, (ii) describing representations on which it operates, and (iii) determining physical mechanisms involved in these operations. The idea underlying this approach is that the levels of description constrain each other, resulting in convergence of explanations towards a holistic account of the system. The systems’ behavioral flexibility is taken as a criterion by which systems operating on mental states can be distinguished from those operating on other kinds of states. Flexibility of the system is manifested in its ability to establish characteristic relations between the input states and the representational states. On the computational and the representational levels of description, flexibility of the system is manifested in the abilities of designation and interpretation; on the implementation level this characteristic of the system is realized through perceptive learning and similar neurophysiological mechanisms.