I am dismayed about current trends in robot architecture discussions. For example, counting the number of layers in a system is like arguing over the number of angels that can dance on the head of a pin - don't we have bigger fish to fry?
Why must there be "one final solution"? While I am personally happy with the way hybrid architectures have progressed - (AuRA merged a traditional AI hierarchical planner with a behavioral controller back in 1986 [1,2,3]) - it seems that this movement is starting to suffer from the same elitism that was a defining characteristic of both the hierarchical  and purely behavioral approaches  (with apologies to both Jim and Rod) from which they descended. Single-minded (read close-minded) solutions are often blind to benefits that can arise from elsewhere.
Ecological robotics  calls for developing a thorough understanding of the robot's targeted task/environment first, and then calling in the architects second. The adage "if all you have is a hammer to work with, then everything looks like a nail" it seems is perilously true for robot system designers.
Biology offers diversity for different task/environments - ants differ from dogs differ from humans in terms of anatomical structure and organization. Pipe crawling robots should differ from lunar rovers from roboticised personal assistants in terms of structure and organization. Guidelines for the design of systems is one thing - dogma (e.g., "this is the fundamental principle of intelligence", "this is the right number of layers") is another. A suitable guideline is ecological fitness of the robot to its task and environment - how well can this system perform its task and survive both physically and economically. It is not the task of the robot to prove that the underlying dogmatic assertions of the designer living in his or her research environment are correct.
Different task/environments call for different architectures. Sure, you could build an igloo in the Sahara, or you could place a grass hut in Lappland, but it's not a good idea. One size does not fit all, at least comfortably.
 Arkin, R.C., "Path Planning for a Vision-based Autonomous Robot", Proc. SPIE Conference on Mobile Robots}, Cambridge, MA, pp.~240-249, 1986.
 Arkin, R.C., "Motor Schema Based Navigation for a Mobile Robot: An Approach to Programming by Behavior", Proceedings of the 1987 IEEE Conference on Robotics and Automation, Raleigh, NC, pp. 264-271.
 Arkin, R.C., "Towards Cosmopolitan Robots: Intelligent Navigation in Extended Man-made Environments", Ph.D. Thesis , COINS Technical Report 87-80, University of Massachusetts, Department of Computer and Information Science, 1987.
 Albus, J., "Outline for a Theory of Intelligence", IEEE Transactions on Systems, Man and Cybernetics, Vol. 21, No. 3, May-June 1991, pp. 473-509.
 Brooks, R., "Intelligence Without Reason", A.I. Memo No. 1293}, MIT AI Laboratory, April 1991.
 Arkin, R.C., Cervantes-Perez, F., and Weitzenfeld, A., "Ecological Robotics: A Schema-Theoretic Approach", in Intelligent Robots: Sensing, Modelling and Planning, eds. R.C. Bolles, H. Bunke, and H. Noltemeier, World Scientific, 1997.