An interesting experiment in the design of intelligent agents was conducted by Vere and Bickmore [Vere and Bickmore, 1990]. They argued that the enabling technologies for intelligent agents are sufficiently developed to be able to construct a prototype autonomous agent, with linguistic ability, planning and acting capabilities, and so on. They developed such an agent, and christened it HOMER. This agent is a simulated robot submarine, which exists in a two-dimensional `Seaworld', about which it has only partial knowledge. HOMER takes instructions from a user in a limited subset of English with about an 800 word vocabulary; instructions can contain moderately sophisticated temporal references. HOMER can plan how to achieve its instructions, (which typically relate to collecting and moving items around the Seaworld), and can then execute its plans, modifying them as required during execution. The agent has a limited episodic memory, and using this, is able to answer questions about its past experiences.