ABSTRACT
The preceding chapters were concerned with the processing and structuring of individual pieces of information as they are input into long-term memory. Chapter 5 displayed the structure of propositional trees acceptable to and recorded in HAM's memory, and referenced the parsing system by which inputs are translated into our internal representations. In the present chapter we begin to examine the matter of accessing the information stored in memory. Here the concern is with how incoming information accesses and makes contact with those portions of longterm memory which are relevant to it. This is the central problem of "recognition." It is carried out in HAM by a process called MATCH which tries to find the best matching tree in memory corresponding to an input tree. The MATCH process is used not only during decoding (i.e., in answering questions), but also in encoding (i.e., comprehending and learning new statements). In particular, the MATCH process enables HAM to distinguish already known information in an input from novel information, and so enables HAM to restrict its efforts at encoding to the new information. In this way, HAM can make use of known information to reduce the task of encoding new information.
