The Structure of Short-Term Memory and Ways to Improve It

Short-term memory, or STM, has been referred to as the workbench for memory, learning, and perception processes (Ellis & Hunt 1993). In 1968, Richard Atkinson and Richard Shiffrin proposed their ideas regarding the human memory system, which are still applied today. According to their work, incoming information is first captured by sensory structures, primarily the eyes and ears (Loftus 1980). This information is then directed by attention and focus to the short term memory system, which strives to manipulate the new information and incorporate it into the long term memory (Loftus 1980). The precise processes involved in this transfer of information and locations of storage for new information remain controversial, but the fact stands that short term memory involves an intricate toolbox of manipulations and that short term and long term are extremely different in structure and behavior (Zechmeister & Nyberg 1982).

The accepted model for Short term memory structure, known as "working memory," revolves around three different structures. First, the "central executive" functions through attention as the sorter of information, distributing incoming messages to the other two structures, depending on the type of information they can most efficiently process (Baddeley 1993). One of these structures is the phonological loop, which is often compared to a tape which constantly repeats in the mind (Jones 2002). The phonological loop is useful when the central executive is faced with a test like strings of numbers, letters, or words, since such items are best remembered by automatic and constant repetition in order of the items seen (Baddeley 1993). Researchers suggested that the fact that strings of information are remembered best by mental repetition must be because STM follows an acoustic (speech-based) code, where repetition by sound is necessary for keeping the new information intact. This is an interesting note, since long term memory has been found to follow a semantic, or meaning-based, code where items of similar meaning can be easily remembered, associated, and recalled (Baddeley 1993). Opposite the phonological loop, the visuo-spatial sketch pad is responsible for holding and manipulating visual information (Baddeley 1993). For example, the mental multiplication of large numbers requires the use of this sketch pad since the actual act of multiplication requires a visualization of the activity (here, the numbers being multiplied, "saved," and added). The three parts that make up short-term memory work together cooperatively and are necessary for proper and efficient memory (Baddeley 1993).

In 1956 George Miller published a popular theory suggesting that the average STM capacity appears to be seven items, plus or minus two (Jones 2002). This means that a typical STM system will be able to hold between 5 and 9 items. The "seven plus/minus two" hypothesis was based on results using a numerical STM test (Ellis & Hunt 1993). However, the experimenters assumed that short term memory treats all types of items (letters, numbers, words) in the same manner (Jones 2002). The test failed to observe for other influences on STM, like ones of language diversity, memory processes, and memory techniques (Jones 2002).

The fact that STM follows a speech and sound-based code suggests that language diversity may influence the performance of STM (Jones 2002). From recent experiments with Chinese and Welsh subjects, the number of items that could be remembered and recalled was significantly higher and lower, respectively. Chinese subjects could recall 9.9 items, while their Welsh counterparts recalled only 5.8 items (Jones 2002). The root of this difference lies in the acoustic aspect of language: The length of time needed to repeat a single item (a number, letter, or word) mentally is different for each language since the equivalent translated word varies in the number of syllables (Jones 2002). Therefore, it appears that the capacity for the acoustic-based STM stands at a timeframe rather than an actual item capacity when applied universally. The STM can hold approximately 2 seconds of "recorded" sound, like a series of items like numbers (Jones 2002). Therefore, subjects speaking a different language can repeat a different number of items in the 2 second time period.

Other factors that influence the performance of STM include the natural processes associated with memory (Ellis & Hunt 1993). For example, the process of rehearsal is crucial to the storage of new information. Rehearsal involves the review and repetition of the information that a person needs to learn in order to keep the material active. Without rehearsal, the decay of information is staggering (Ellis & Hunt 1993). In one experiment where subjects were not allowed to practice rehearsal by having to perform a "distraction" activity (like counting backwards) for a certain length of time, the subjects often recalled one item or none at all after only 18 seconds (Kellett 1980). The lifespan of unrehearsed bits of information stands at around 30 seconds (Ellis & Hunt 1993). This connects to the theory of interference, which suggests that the length of an intervention or distraction in memory processes affects its recall, not the nature of the disruption (Kellett 1980).

Certain memory techniques like chunking and coding also drastically affect the performance of STM. The technique of chunking involves the breakdown of an entire string of items into smaller groups (Ellis & Hunt 1993). Instead of remembering a single digit sequence like "2," "5," and "6" chunking allows the string to become a group of "256." This consolidates the available STM space for the items and allows for more new pieces to be remembered. Chunking is useful in phone numbers and social security numbers, for example (Kellett 1980). The idea of chunking also relates to the observation that strings of letters are often remembered better than strings of numbers (Kellett 1980). Although there are many more letters than numbers, subjects tend to naturally "chunk" when presented with the string of letters since an odd word or sound can be created. With numbers, such an option does not exist (Kellett 1980)

Another memory technique, "coding," involves a series of connections between short term memory and long term memory, or LTM (Ellis & Hunt 1993). In this case, the information that needs to be remembered is split into pieces that are related to a piece of accessible long term memory. For example, a sequence like 191419391991 can be split into 1914-1939-1991 by making the connection with long term knowledge that the numbers can represent years in which major wars began. In coding, a sort of deciphering rule exists where the LTM dictates how the STM can be recalled easily (Ellis & Hunt 1993). Experiments have proven that when memory techniques like chunking and coding are removed, the STM memory capacity decreases to one of about 3 to 5 items, instead of the 5-9 suggested by Miller in 1956 (Jones 2002).

Although the exact structure of short term memory may not be fully understood, it is clearly a crucial step in the process of memory and learning. The exact background of STM is not as valuable as its function and mechanics. The work of STM may be perhaps the most influential entity in human memory, since the information that enters the STM is pure from any form of modification and the processes of STM dictate what will be remembered and what will be discarded.

A. Baddeley, "Your memory: a user's guide." Avery Publishing Group, Inc.

H. Ellis and R. Hunt, "Fundamentals of Cognitive Psychology." Brown & Benchmark Publishers.

D. Jones, "The 7 +/- 2 urban legend.". Knowledge Software.

M. Kellett, "Memory power." Sterling Publishing Co., Inc.

E. Loftus, "Memory." Addison-Wesley Publishing Company.

E. Zechmeister and S. Nyberg, "Human Memory: An Introduction to Research

and Theory." Brooks/Cole Publishing Company.