The term memory span refers to the maximum length of a sequence of items that can be reproduced from memory following a single presentation. Scientists have been interested in memory span since the publication of the first important study of memory, nineteenth-century German experimental psychologist Hermann Ebbinghaus's monograph in 1885. Using himself as his only subject, Ebbinghaus determined the number of presentations necessary for an error-free reproduction of a sequence of items; he found that this number decreased dramatically as the length of the sequence decreased until the sequence included only seven items, at which point only a single presentation was needed. Ebbinghaus showed no particular interest in this finding, but others did. Within two years, memory span was shown to increase systematically during childhood and to be appreciably shorter for the mentally impaired. Within a decade, memory span was firmly established in what was then an emerging field of mental abilities testing, where it has remained ever since.
Most often the procedure for testing memory span calls for the recall of the items in the order in which they were presented. Sometimes the order in which the items have to be recalled is entirely unconstrained; sometimes the items have to be recalled in their reverse order of presentation. But only rarely does the procedure call for more difficult transformations, such as recall of the items in alphabetical or numerical order. Such transformations would draw on what is often referred to as working memory and, interestingly, would probably provide a more valid measure of mental ability. Incidentally, the measure of choice for this ability has been a variant of memory span known as working memory span, which is the number of items that can be retained while performing some other cognitive task.
Measuring Memory Span
Order of reproduction aside, there has never been a fixed procedure by which memory span is measured. Most of the methods reviewed long ago by J. Paul Guilford and Karl Dallenbach (1925) are still extant. Sometimes the sequence of items is deliberately set somewhat too long for perfect reproduction, and memory span is defined as the number of items from this sequence that are recalled. This method is quick to administer but too problematic to be considered anything more than rough and ready. One problem is that the usual requirement that the items be recalled in their exact order of presentation has to be modified to allow for imperfect recall. Another is that, whatever the criterion regarding recall order, the number of items recalled is likely to vary according to the number presented, even in the supraspan range. For such reasons, memory span is usually determined by presenting lists of several different lengths and ascertaining the maximum length for which recall is perfect. In the procedure typical of most mental abilities tests, list length is at first so short that perfect reproduction is virtually certain, and then is gradually increased until errors are made.
Like any other psychological measure, memory span is not entirely reliable, and is therefore a statistical concept—the sequence length for which there is an even chance of perfect reproduction. For this reason, most tests include two or three lists of each length, a specific stopping rule, and a specific averaging procedure. More precise estimates, as may be needed for certain research purposes, may be obtained with the staircase, or up-and-down, method. This involves the presentation of a series of lists, the length of any given list being one more than that of the immediately preceding list if the latter was correctly reproduced and one less if it was not. Memory span is given by averaging the list lengths. In determining this average the first few lists should be disregarded because they will reflect the arbitrary length of the first list. Also, the length of what would have been the next list in the series, as given by the length and outcome for the final presented list, should be included.
Ebbinghaus's finding of seven items as the maximum length for a reproducible list provides a first approximation of memory span. Closer approximations will vary with the age of the rememberer, for span increases through childhood and declines in old age. Moreover, span varies among individuals of the same age.
Factors in Memory Span
In addition to varying among individuals, memory span varies for a given individual according to a considerable number of factors. For example, span can be increased by presenting the items at an irregular rate, so that they appear temporally grouped. Also, span for verbal items tends to be slightly greater with auditory presentation than with visual presentation. Of particular interest is the effect of the nature of the list item. The most common kind of item, especially in mental abilities testing, is the digit. Digit span is roughly one item greater than letter span, which in turn is roughly one item greater than word span. Also used as list items have been nonsense syllables (which is what Ebbinghaus used), geometric designs, and pictures of objects. One finding that has emerged from a comparison of memory span for different kinds of items is an impressive linear relation between memory span and a hypothetical "search rate," operationalized in terms of the slope of the roughly linear function that relates the time to decide whether a test, or probe, item was included in a just-presented short sequence to the length of the sequence. Specifically, a relatively small increment in decision time is incurred by increasing the length of a sequence of items of the kind that yields a relatively large memory span. Although the significance of this finding remains uncertain, it clearly raises the question of how memory span should be conceptualized.
One way of conceptualizing memory span is as a measure of the capacity of what the American psychologist and philosopher William James called primary memory. In other words, memory span can be considered as the number of items that can be held in conscious mind at any given instance. Aside from its intuitive appeal, this interpretation is supported by evidence that memory span varies according to certain characteristics of the list items that are salient in conscious experience during performance of the task. In particular, the representation of the list items in mind usually takes the form of inner speech, and memory span has been found to depend on factors that can reasonably be regarded as relevant to the spoken form of the items. Thus, span is shorter when the list items are phonemically similar to one another than when they are phonemically dissimilar (e.g., shorter for GBVDPTZ than for GMRKSQY), when the list items are phonemically lengthy than when they are phonemically short (e.g., shorter for lists of polysyllabic words than for lists of monosyllabic words), and when subjects engage in irrelevant vocalization during list presentation—an activity that is likely to suppress covert naming of the items. Other support for the idea of memory span as the product of primary memory derives from evidence that memory span for one kind of item predicts fairly well memory span for another kind of item but is a poor predictor of performance on supraspan tasks that clearly cannot be performed on the basis of primary memory.
Such supportive evidence notwithstanding, the idea that memory span measures primary memory does little to shape twenty-first-century theorizing. One reason for this is that virtually all of today's memory theorists adopt an information-processing perspective, and they give little or no consideration to the conscious realization of the mechanisms they hypothesize. A more particular reason is that primary memory has been recast as a short-term store, the capacity of which has been estimated at three or four items. This estimate, which was based on the number of items recalled from toward the end of a supraspan sequence, is appreciably smaller than memory span. One way to account for the discrepancy is to make the plausible assumption that the capacity of primary memory varies with the nature of the task and that, unlike supraspan recall tasks, the memory span task stretches primary memory to its maximum. Another possibility, supported by recent evidence, is that memory span represents primary memory supplemented by secondary memory—that is, by items that have dropped out of conscious mind and have to be recollected.
A conception of memory span more in keeping with the zeitgeist, though not inherently incompatible with the primary memory interpretation, is that it represents the amount of information that can be articulated in a certain time, variously estimated as between 1.5 and 2.0 seconds. And, indeed, items that yield long spans, such as digits, tend to be those that are articulated relatively quickly. There are, however, exceptions to this rule, leaving memory span still without a satisfactory interpretation.
Brener, R. (1940). An experimental investigation of memory span. Journal of Experimental Psychology 26, 467-482.
Brooks, J. O., III, and Watkins, M. J. (1990). Further evidence of the intricacy of memory span. Journal of Experimental Psychology: Learning, Memory, and Cognition 16, 1,134-1,141.
Cavanaugh, J. P. (1972). Relation between immediate memory span and the memory search rate. Psychological Review 79, 525-530.
Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences 24, 87-114.
Dempster, F. N. (1981). Memory span: Sources of individual and developmental differences. Psychological Bulletin 89, 63-100.
Guilford, J. P., and Dallenbach, K. M. (1925). The determination of memory span by the method of constant stimuli. American Journal of Psychology 36, 621-628.
Miyake, A., ed. (2001). Individual differences in working memory. Journal of Experimental Psychology: General 130, 163-237.
Schiano, D. J., and Watkins, M. J. (1981). Speech-like coding of pictures in short-term memory. Memory & Cognition 9, 110-114.
Schweikert, R., and Boruff, B. (1986). Short-term memory capacity: Magic number or magic spell? Journal of Experimental Psychology: Learning, Memory, and Cognition 12, 419-425.
Underwood, B. J., Boruch, R. F., and Malmi, R. A. (1978). Composition of episodic memory. Journal of Experimental Psychology: General 107, 393-419.