ABSTRACT
The Current Status of Multistore Models
Change in the Context of Normal Science The expectations of memory theorists were that the Atkinson-Shiffrin model would be extended and the characteristics of the stores firmed up. This did not happen, and the central characteristics of multistore formulations are no longer the primary interest of memory researchers. This change took place within the context of normal science, not revolution.
The Fall of Theory Theory can be abandoned without change in the six facets of a paradigm. This is a gradual process, which can be tracked in scientific journals as articles appear criticizing the theory followed by a lack of interest in either the theory or its criticism. To explain why this happened to Atkinson and Shiffrin's theory, we consider what attracted researchers to it, and how scientists decide to retain or abandon theory.
Three Dimensions of the A tkinson Shiffrin Model Theory consists of a conceptual core, a formal implementation, and a domain of intended applications. The conceptual core of the Atkinson-Shiffrin model comprised the three stores and their properties; the formal implementation was drawn from mathematical learning theory; and the domain of intended applications was to a range of traditional verbal-learning data. Each of these aspects of the model attracted a constituency.
What Influences an Experimental Psychologist's Choice of Theory? Psychologists differ. Some are highly committed to successful experimental testing; others are more committed to sufficiency. In the case of Atkinson and Shiffrin, both types of psychologist found reasons for dissatisfaction with the model.
Experimental Problems With Multistore Models
Coding Although each store was supposed to accept a unique type of code, experiments indicated that they were more flexible than that.
258 Forgetting Functions Researchers were unable to establish stable estimates of the forgetting rates from the stores.
Capacity Capacity estimates also turned out to be unpalatably variable.
Conceptual Problems With Multistore Models Some psychologists were dissatisfied with the model because it was so tied to verbal-learning experimentation. They made the metatheoretical judgment that it could not be extended to the kind of meaningful human information processing that interested them. They would have abandoned it on these grounds, however the experiments came out. Some turned to semantic memory research.
Alternatives to Multistore Models of Human Episodic Memory
The Current State of Episodic Memory Research
What Remains of the Multistore Approach? The data that multistore models were intended to organize still compels episodic memory research: iconic memory and limited capacity in particular. However, current approaches emphasize flexibility.
The Heterogeneous Nature of the Contemporary Field No single theory has come to replace multistore models. At present, there are a number of loosely related problem areas in episodic memory research. We describe those that involve a relatively large number of researchers, transcend particular laboratory tasks, are relatively theory-oriented, and are most likely to have a continuing impact on cognitive psychology.
Levels of Processing
The "Levels" View of Memory Levels of processing were introduced by Craik and Lockhart (1972) as a metatheoretical alternative to stores or stages. Since then, there has been an effort to develop a precise theoretical account of levels.
Some Typical Experiments
Theoretical Extensions Besides attempts to systematize "levels" into a coherent theory, there have been demonstrations of its extensibility to sentences, paragraphs, and theories of semantic memory. Nevertheless, "levels" is most successful in the context of episodic memory studies. It remains to be seen if it will generalize successfully.
Encoding Specificity
The Encoding Specificity Phenomenon It is easier to remember something if the same cues are present when one tries to remember as were present when the event initially occurred.
The Generality of Encoding Specificity Although it was introduced in word memory studies, parallel outcomes have occurred in studies of memory for sentences, faces, and tones.
Further Research Questions Suggested by Encoding Specificity Several researchers are attempting to determine what makes a particular cue "salient" at encoding and retrieval.
Analogue Memory Although much episodic memory research involves words, letters, and digits, a major memory function is the storage of perceptual (e.g., visual) experience. What is the coding format of such "analogue" memory, and what is its relation to more abstract memory function?
Evidence for Analogue Memory People claim to be able to form mental images, and yet for a long time psychology ignored this fact. There is now good experimental evidence that people can use images to help them remember and that they can manipulate these images mentally.
259 Multiple Coding Paivio (1969, 1971) has suggested that people have two codes available: one for perceptual inputs, one for linguistic inputs. Much research evidence supports this idea. However, the characterization of the perceptual code as an "image" is controversial. Many theorists consider this metaphor too photographic.
Extensions of Multistore Models
Both Atkinson and Shiffrin have developed models that are built on the foundation of the original multistore theory. However, both emphasize the flexibility of human episodic memory, and the three stores are not necessarily essential to either one.
The Atkinson- Juola Theory The conceptual core of this theory retains the three stores, but the role of short-term storage is different and there has been considerable development of the nature of the long-term store. Two kinds of codes are considered available: perceptual and conceptual. The major focus of the model is how people compare input information to stored memory information in a memory-scanning task similar to that of Sternberg. The Atkinson-Juola account incorporates the mathematics of signal detection theory.
The Shiffrin-Schneider Theory The new model authored by Shiffrin and Schneider emphasizes two fundamental capabilities: automatic detection and controlled search. They hope to bring together under one theoretical umbrella the research on memory scanning and other episodic memory tasks, and existing work on attention. The domain of intended applications, and the type of experiments offered by the authors, is primarily a modified version of Sternberg's memory-scanning tasks. The theory is potentially of great significance.
Conclusions
