A recurrent theme in the study of human memory over the last 20 years has been the question of whether it should be regarded as a unitary system, or as a collection of two or more subsystems. After a few years of relative quiescence, this topic has begun to move back into the theoretical limelight and, as has often been the case in this area, the stimulus for theoretical development has come from the study of clinical evidence from patients with memory deficits. Before going on to discuss these

more »

... recent developments, it would perhaps be helpful to outline the earlier background to the controversy. (A more detailed discussion is given in Baddeley, 1983a, chapters 9-11.) Up to and including the 1950s, the study of memory was largely concerned with the learning of lists of words or paragraphs of prose and their retention over periods ranging from a few minutes to a few years. Memory was, implicitly at least, regarded as a single unitary system. During the latter half of the 1950s, however, a number of investigators began to observe that even small amounts of material would show signs of being forgotten over a matter of seconds, provided that the subject was prevented from continually rehearsing it, and they postulated a separate short-term memory system to account for their results (e.g. Broadbent, 1958) . The 1960s saw the development of a controversy as to whether it was, in fact, necessary to assume separate systems to account for these new results, or whether existing theories could explain both sets of data (Melton, 1963) . This produced a great deal of experimental work attempting to look for clear empirical evidence for or against two separate memory systems. Many apparent differences were observed, with the following three being among the more theoretically cogent. EVIDENCE FOR A DICHOTOMY A number of investigators noted that many memory tasks appear to behave as though they comprise two quite separate components, one relatively durable and the other quite labile. The most commonly quoted instance of this is the task known as free recall in which subjects are presented with a list of unrelated items-for example, twenty words, one at a time-and then asked to recall as many of the items as they can in any order they wish. Typically, subjects recall the last few items extremely well, and earlier items considerably less well. The tendency for the last few items to be well remembered, the so-called recency effect, disappears, however, if recall is delayed for a few seconds during which the subjects are distracted by requiring them to count or do a simple sum. In contrast, the effect of this delay on earlier items is negligible. A simple interpretation of this result is to suggest that the most recent items are being held in a brief temporary 'short-term memory' store, and are forgotten during the delay (Glanzer, 1972) . 2 A second apparent difference between this brief temporary storage system and the longer-term system emerges when the role of encoding in memory is studied. If subjects are required to hear and immediately recall a string of unrelated letters or words, they appear to rely heavily on the spoken characteristics of the material. Hence items that are phonologically similar to each other, such as the consonant string B V T C P D or the word sequence man cad mat map can, are much less likely to be recalled accurately than otherwise equivalent but dissimilar sequences, such as K W Y Q R L