The Role of Memory in Brain, Values, and Choice
Bennet B. Murdock Department of Psychology University of Toronto Toronto, ON M5S 3G3 Canada Telephone: (416) 978-3175/FAX: (416) 978-4811 E-mail: firstname.lastname@example.org
The paper discusses a memory model TODAM that presents a particular point of view on the relation between data and brain formalisms. TODAM, a theory of distributed, associative memory is based on the holographic approach of Pribram ( 1971) and the convolution-correlation formalism of Borsellino and Poggio ( 1973). It attempts to explain the storage and retrieval of item, associative, and serial-order information, and is tested by data from psychological studies of short-term human episodic memory. A brief account of the application of the model to these three areas and illustrative experimental data will be presented.
The general theme of the Appalachian conferences is the relation between data and brain formalisms. The theme of Appalachian V is brain, values, and choice. This chapter will deal with the role of memory in brain, values, and choice. I will focus my remarks on the understanding of memory per se, and leave the broader questions to others.
Memory is not a single entity; rather, there are different kinds of memory. The basic partitioning is in turns of the types of information stored and retrieved. At least for simple episodic memory there are three types of information: item information, associative or relational information, and serial-order or positional information. Item information allows us to recognize objects and events--names, faces, sounds, the myriad objects and events that make up the world around us. It underlies the feeling of familiarity when we see or hear them. Associative information allows us to remember the relation between two objects or events: lightning and thunder, the taste and smell of a particular food, or the names and faces of our friends. Serial-order information allows us to remember temporal order: the days of the week, the letters of the alphabet, or how to spell words. Experimental evidence for these distinctions may be found in Murdock ( 1974).
Until recently the standard model for memory was a node-and-link model where the items were stored in the terminal nodes of a semantic network and the associations were the links between the nodes. Serial order could be represented as the order in which the links were traversed, and search was a universal metaphor for retrieval. The nodes were organized on the basis of semantic similarity, and spreading activation accounted for retrieval ( Collins & Loftus, 1975).