Memory span is the most commonly used measure of short-term memory and refers to the number of words a person can recall in order immediately after hearing them. The study of memory span has a long history in experimental psychology. As early as 1885 Ebbinghaus noted that he could recite a list of up to seven unrelated words at the first attempt, but longer lists took more trials; the longer the list, the greater the number of trials before he could recall it correctly. Jacobs (1887) was the first psychologist to have used a standard memory span test with a wide range of individuals. Jacobs gave individuals a series of lists of increasing length to determine memory span, defined as the longest list that could be recalled in correct order. He made three main observations, which still stand today: memory span varies according to the material used; span is less in children than it is in adults; span varies according to “natural ability” (intelligence). It was soon established that memory span shows a gradual increase

with age. Memory span was also quickly accepted as a useful index of mental capacity. A memory span test (using sequences of digits of increasing length) was included in the first test of mental development devised by Binet and Simon, and a digit span test is still a part of modern IQ and mental ability tests, such as the Wechsler Intelligence Scale for Children (WISC), the British Ability Scales (BAS), and the Illinois Test of Psycholinguistic Abilities (ITPA). Digit span is included in such tests

because there is a gradual, steady increase in performance on the test from the age of 4 to 5 years to around 16 years. At age 4 the individual can recall on average just 3 digits in order, but performance normally improves steadily to about 6 digits at age 10 or 12; then there is a more gentle increase up to age 16 when digit span plateaus at around 7 to 8 items. When the Wechsler Intelligence Scale for Children was revised in

1976 the digit span sub-test was found to correlate moderately with the full-score IQ based on 10 sub-tests (r = 0.43). Not surprisingly, the correlation of digit span with the other five language sub-tests (r = 0.45) was higher than with the six performance tests (r = 0.34). These figures are based on the performance of 1,100 normal children aged 6 to 16 used in the standardization and revision of the WISC (Wechsler, 1976). Bachelder and Denny (1977) reviewed a number of studies in which a wider IQ range was included (normal children and those with learning difficulties). They found that when a broader IQ range is sampled, the correlation coefficient between digit span and other IQ sub-tests increases to figures in the 0.6 to 0.8 range. Just as the increase in memory span is a consistent feature of normal

development, so it has often been noted that memory span is impaired in groups with learning difficulties. In the 1880s, Galton measured memory span in “feeble-minded” individuals (i.e. those with learning difficulties), and found that although span varied amongst those classed as feeble-minded, their memory span was lower than that of normal individuals. More recent studies have shown that digit span is poor in those with learning difficulties, even in relation to many other tests of ability (Baumeister & Bartlett, 1962; Belmont, Birch, & Belmont, 1967; Marinosson, 1974). These apparently stable relationships between memory span and age,

and memory span and IQ, have been known and accepted for nearly a century. We must now consider the nature, and importance, of these relationships. One possible explanation draws on the idea that span reflects some general limitation on the human information processing system. Before considering this idea in more detail we will first consider, from an historical perspective, theories of short-term memory.