This conjecture of a common mechanism can be combined with Atkinson and Shiffrin's (1968) distinction between short- and long-term memory. In this modal view, unidimensional stimuli are not stored in long-term memory. Instead, they can only be held in short-term memory (e.g., Marley & Cook, 1984). Poor unidimensional absolute identification performance directly reflects the capacity limits of short-term memory. With these stimuli, there is no longterm improvement with practice because there is no opportunity to store stimulus-specific information (i.e., the perceived value of a stimulus on a relevant dimension) in long-term memory.
As in the other new study, Todd and Marois found that performance declined with increasing set size, levelling off at between three and four items when visual memory capacity was reached. The imaging data revealed that activity within a single bilaterally symmetric region of the intraparietal and intraoccipital sulci correlated with the number of objects encoded, reaching a plateau by set size four. Crucially, the general effects of task difficulty could be ruled out because, beyond set size four, accuracy decreased and reaction times increased, yet activity in this region remained constant, as did the number of items successfully encoded.
When people are asked to distinguish between different tones, if the number of tones presented is over about 5, their accuracy at this task decreases rapidly. When asked to recall a series of unrelated words or numbers, people fail when the size of the series increases to 6 or 7. In other words, the span of immediate memory imposes severe limitations on the number of items we are able to receive, process and remember.
In fact the actual number varies according to the type of information you're storing (numbers, words or images). To see how fragile working memory is, ask someone to remember just three consonants - for example, G, X, S. Test them in a few minutes time but instead of allowing them to think about these letters and commit them to memory, make them count backwards from 100 in jumps of 3 (100, 97, 94, 91 and so on) … they'll find it much hard to remember the letters than if they had a few minutes to concentrate undisrupted.
"Clustering" is another effective technique. Seven, according to experts, is the magic number for short-term, or working, memory. That's roughly how many things we can consciously hold in the mind at one time. But we can trick it into holding more by inventing seven or so main categories and then grouping several things under each.
I wish I knew why the brain has self-imposed limits on memory. But I suspect that those limits exist in order to allow attention and focus to thrive. If we could absorb into memory a lengthy list of ideas or independent items, we would be flitting from one thing to another, giving no particular emphasis to any given idea. We would become robotic.
"Whoever developed the current postal-code system must have been out of his mind," he said. "To call B.C. 'V' is stupid - it should have been 'B'. To increase the number of digits from six to 10 would be disastrous."
He said there was a famous paper about memory written about 30 years ago by an American academic called The Magical Number Seven Plus Or Minus Two.
"That means most people can remember up to seven pieces of information without too much problem," he said. "The plus or minus two depends on the type of information - whether it's grouped or includes numbers and letters."
This is why phone numbers are seven digits long. "Ten digits," you say? Yes and no. You are supposed to have the area code stored in long-term memory so that you say, "New York is 212." Then you hold in short-term memory the other seven digits that somebody is rattling off until you successfully dial the number. So if you don't already know the area code, you'd better get a pen.
Short term memory is only what you hold in your mind at the moment. If you don't elaborate on it - find some way to make it stick - then as soon as you stop repeating the information to yourself, it will be gone.
Miller's 7±2 categories correspond to 2.81 bits/stimulus minus 0.49 bits/stimulus and plus 0.36 bits/stimulus. MacRae (1970, p. 112) noted, Miller's (1956a, b) work noting that "The quantitative correspondence between quite different sensory modalities with this technique made it seem likely that some fundamental aspect of performance was being measured". Many further attempts to measure channel capacity followed the publication of Miller's paper.
"The Magic Number 7, Plus or Minus 2," a frequently cited, and now often debated, paper published in 1956 by psychologist George A. Miller, suggested that people are immediately capable of storing seven "chunks," or pieces of arbitrary information, give or take two items. But that has been refuted by scientists over the years, and the number has variously ranged from three to six items.