Memory for Serial Order

Memory for serial order is important in many aspects of daily life, including language comprehension. Despite the existence of a large and stable empirical data base, theoretical progress has lagged behind experimental work. Some of the theoretical notions to date have included chaining of associations, nonassociative storage in bins, and reverberatory loops. These concepts have been valuable for application to individual paradigms, but a unified approach has been lacking. We present an extension to Murdock's Theory of Distributed Associative Memory (TODAM) based on associative chaining between items. A distributed memory system has a number of a priori advantages; it can sustain local damage without complete failure and requires no memory search for retrieval. We applied TODAM to a number of serial order phenomena-among them serial list learning, delayed recall effects, partial report effects, and buildup and release from proactive interference-and found that the theory provided a good quantitative description of the data with a small set of parameters.