Learning and Contemporary Views Memory

Typically, structuralists are concerned not only with delineating the number of memory systems but also in locating them within the brain. Advocates of the systems approach identify five major memory systems: working memory, semantic memory, episodic memory, the perceptual representation system, and procedural memory.

Working memory is a system for the temporary maintenance and storage of internal information and can be thought of as the place where cognitive "work" is performed. It has multiple subsystems, most notably one for the processing of verbal or speech-like stimuli and another for the processing of visual-spatial information. Each features a store (the phonological store and the visual-spatial sketchpad) and a maintenance process that offsets decay (articulatory control process and the visual scribe). Unlike short-term memory, the capacity limit is based on a trade-off between the rehearsal process and decay.

Semantic memory refers to memory for general knowledge and includes facts, concepts, and vocabulary. The term is somewhat misleading, however, as it implies storage of only semantic information, which is not the case. Rather, the distinguishing feature of this system is the lack of conscious awareness of the learning episode. In contrast, episodic memory is the system that does provide this awareness of the learning episode and thus enables the individual to "travel back" mentally into his or her personal past, to use Endel Tulving's term. As an example of the difference, consider the fact—picked hopefully for its obscurity—that Samuel Beckett is the only first-class cricket player to win a Nobel Prize (he played for Dublin University and won the 1969 Nobel Prize for Literature). Imagine that two weeks from now you happen to participate in a conversation about Nobel Prizes, and you recall the above information about Samuel Beckett. If you remember reading the information in this book, where you were, whether it was hot or cold, or anything else about the episode other than the mere facts, then the information is said to be in episodic memory. If you remember only the fact itself, then the information is said to be in semantic memory.

All three of these systems are part of declarative memory, which is contrasted with procedural memory, the fourth proposed major memory system. Declarative memory is concerned with knowing "that" rather than with knowing "how." For example, if you know that two plus two equals four, the information is said to be in declarative memory. By contrast, if you know how to ride a bicycle, that information is said to be in procedural memory. One distinction is that you can usefully communicate declarative information but not procedural information: simply telling someone how to ride a bicycle (explaining balance, pedaling, and steering) does not work.

Finally, the perceptual representation system is a collection of non-declarative domain-specific modules that operate on perceptual information about the form and structure of words and objects. Although currently the dominant approach, the structuralist conception of memory is not without its problems.

First, there is little agreement on what the criteria should be for a memory system, let alone agreement on the number of systems. In particular, there are no criteria that result in producing just the five systems named above. The most well-specified criteria for determining whether two systems are separate were proposed by David Sherry and Daniel Schacter in 1987: (1) functional dissociations: an independent variable that has one effect on a task thought to tap system A either has no effect or a different effect on a task thought to tap system B; (2) different neural substrates: system A must depend on different brain regions than system B; (3) stochastic independence: performance on a task that taps system A should be uncorrelated with performance on a task that taps system B; and (4) functional incompatibility: a function carried out by system A cannot be performed by system B.

Henry Roediger, Randy Buckner, and Kathleen McDermott have shown how recall and recognition meet the first three of these criteria: the variable word frequency produces functional dissociations (high-frequency words are recalled better than low, but low-frequency words are recognized more accurately than high); neuropsychological dissociations between recall and recognition are observable in amnesiac subjects; and numerous studies have found essentially no correlation between performance on the two tests. A thought experiment suggests that recall and recognition might also demonstrate functional incompatibility: memory for odors is a function that is well supported by recognition (consider the case of Marcel Proust) but not very well supported by recall.

Second, the proposed systems have not yet been anatomically isolated (although it remains an empirical question whether they will be), and even when isolation is found, the results are often equivocal. For example, many studies show that retrieval in episodic tasks involves right prefrontal activation, whereas retrieval of semantic information involves left prefrontal activation. This neuropsychological distinction, however, reverses when the two tasks are equated for difficulty: retrieval of semantic information now produces larger amounts of right frontal activation than the episodic task.

Third, this view does not account for the general pattern of decline seen when normally healthy people get older. One would expect a systematic relationship between tasks that shows a decrement in performance and the underlying system. For example, one might expect that episodic memory might deteriorate faster than semantic memory. In fact one finds no such relationship. Performance on some episodic tasks is unaffected, whereas performance on some semantic tasks is quite impaired. Rather, the pattern of performance loss is better described based on the processes, especially those at retrieval.