Most of the information that gets into sensory memory is forgotten, but information that we turn our attention to, with the goal of remembering it, may pass into short-term memory. In short-term memory (STM), small amounts of information can be temporarily kept for more than a few seconds but usually for less than one minute (Baddeley, Vallar, & Shallice, 1990). Information in short-term memory is not stored permanently but rather becomes available for us to process. The processes that we use to make sense of, modify, interpret, and store information in STM are known as working memory.

Although it is called memory, working memory is not just a store of memory like STM but is also a set of memory procedures or operations. A prominent theory regarding the nature of working memory was proposed by Baddeley & Hitch (1974) and has been expanded on since then.

Baddeley’s model of working memory

Based on experiments demonstrating connections between LTM and STM, as well as experiments indicating that STM consists of more components, Alan Baddeley and Graham Hitch proposed a multi-component working memory model in 1974. The new term working memory was supposed to emphasize the importance of this system in cognitive processing. Baddeley and Hitch suggested working memory is composed of three parts: the central executive, a system that controls the phonological loop (a subsystem for remembering phonological information such as language by constant refreshing through repetition in the loop), and the visuospatial sketch pad (a subsystem for storing visual information).

This model was later revised and improved by Baddeley but also contributed by other authors, which resulted in additional component of episodic buffer in year 2000 and more detailed functions and analysis of other components, as described in table below.

Table 1.
Central executive	It is still unclear whether it is a single system or more systems working together. Central executive’s functions include attention and focusing, active inhibition of stimuli, planning and decision-making, sequencing, updating , maintenance and integration of information from phonological loop and visuospatial sketchpad. These functions also include communication with long-term memory and connections to language understanding and production centers.
Episodic buffer	Episodic buffer has the role of integrating the information from phonological loop and visuospatial sketchpad, but also from long-term memory. It serves as the storage component of central executive, or otherwise information integration wouldn’t be possible.
Phonological loop	According to Baddeley, phonological loop consists of two components: a sound storage which lasts just a few seconds and an articulatory processor which maintains sound information in the storage by vocal or sub vocal repetition. Verbal information seems to be automatically processed by phonological loop and it also plays an important, maybe even key role in language learning and speech production. It can also help in memorizing information from the visuospatial sketchpad. (For example, repeating “A red car is on the lawn.”)
Visuospatial sketchpad	This construct according to Baddeley enables temporary storing, maintaining and manipulating of visuospatial information. It is important in spatial orientation and solving visuospatial problems. Studies have indicated that visuospatial sketchpad might actually contain two different systems: one for spatial information and processes and the other for visual information and processes.

To gain a sense of how we require and utilize such processes, imagine that you are asked to participate in a task such as this one, which is a measure of working memory (Unsworth & Engle, 2007). Each of the following questions appears individually on a computer screen and then disappears after you answer the question:

Is 10 × 2 – 5 = 15? (Answer YES OR NO) Then remember “S”

Is 12 ÷ 6 – 2 = 1? (Answer YES OR NO) Then remember “R”

Is 10 × 2 = 5? (Answer YES OR NO) Then remember “P”

Is 8 ÷ 2 – 1 = 1? (Answer YES OR NO) Then remember “T”

Is 6 × 2 – 1 = 8? (Answer YES OR NO) Then remember “U”

Is 2 × 3 – 3 = 0? (Answer YES OR NO) Then remember “Q”

To successfully accomplish the task, you have to answer each of the math problems correctly and at the same time remember the letter that follows the task. Then, after the six questions, you must list the letters that appeared in each of the trials in the correct order (in this case S, R, P, T, U, Q).

To accomplish this difficult task you need to use a variety of skills. You clearly need a temporary memory store (something like STM), as you must keep the letters in storage until you are asked to list them. But you also need a way to make the best use of your available attention and processing. For instance, you might decide to use a strategy of repeat the letters twice, then quickly solve the next problem, and then repeat the letters twice again including the new one. Keeping this strategy (or others like it) going is the role of working memory’s central executive — the part of working memory that directs attention and processing. The central executive will make use of whatever strategies seem to be best for the given task. For instance, the central executive will direct the rehearsal process, and at the same time direct the visual cortex to form an image of the list of letters in memory. You can see that although STM is involved, the processes that we use to operate on the material in memory are also critical.

Short-term memory is limited in both the length and the amount of information it can hold. Peterson and Peterson (1959) found that when people were asked to remember a list of three-letter strings and then were immediately asked to perform a distracting task (counting backward by threes), the material was quickly forgotten (see figure below), such that by 18 seconds it was virtually gone.

 

One way to prevent the decay of information from short-term memory is to rehearse it. Maintenance rehearsal is the process of repeating information mentally or out loud with the goal of keeping it in memory. We engage in maintenance rehearsal to keep something that we want to remember (e.g., a person’s name, email address, or phone number) in mind long enough to write it down, use it, or potentially transfer it to long-term memory.

If we continue to rehearse information, it will stay in STM until we stop rehearsing it, but there is also a capacity limit to STM. Try reading each of the following rows of numbers, one row at a time, at a rate of about one number each second. Then when you have finished each row, close your eyes and write down as many of the numbers as you can remember.

019

3586

10295

861059

1029384

75674834

657874104

6550423897

If you are like the average person, you will have found that on this test of working memory, known as a digit span test, you did pretty well up to about the fourth line, and then you started having trouble. You probably missed some of the numbers in the last three rows, and did pretty poorly on the last one.

Originally, the digit span of most adults was thought to be between five and nine digits, with an average of about seven. The cognitive psychologist George Miller (1956) referred to “seven plus or minus two” pieces of information as the “magic number” in short-term memory. But if we can only hold a maximum of about nine digits in short-term memory, then how can we remember larger amounts of information than this? For instance, how can we ever remember a 10-digit phone number long enough to dial it? Additionally, we tend to find that the “magic number” can differ across types of information (e.g. numbers, letters, words, colors, and musical notes) and circumstances so that it is quite a bit more variable than previously thought.

So what are some ways we can improve the capacity of our STM? One main insight in this domain involves the idea of recoding – or converting information from one format to another to aid in processing and retrieval. This often involves grouping information into meaningful or related chunks. But, as we will discuss, meaning can vary depending on experience.

Chunking

One way we are able to expand our ability to remember things in STM is by using a memory technique called chunking. Chunking is the process of organizing information into smaller (ideally meaningful) groupings (chunks), thereby increasing the number of items that can be held in STM. For instance, try to remember this string of 12 letters:

XOFCBANNCVTM

You probably won’t do that well because the number of letters is more than the magic number of seven.

Now try again with this one:

CBSNBCABCHBO

Would it help you if I pointed out that the material in this string could be chunked into four sets of three letters each? I think it would, because then rather than remembering 12 letters, you would only have to remember the names of four television stations. In this case, chunking changes the number of items you have to remember from 12 to only four.

So, we find that in such tasks, capacity limits can expand or even disappear when participants are able to find some higher-order meaning in the stimuli (such as above). As another example, one famous subject in a random decimal digit memorization experiment found he could remember more digits at a time by mentally recoding them as mile times since he was an avid runner (Ericsson, Chase, & Faloon, 1980). The ability to find higher-order meaning then can depend greatly on someone’s experience.

Expertise and meaningful information

In the real world, people are constantly recoding stimuli, so it can be difficult to define precisely what a chunk is. For example, “Experts” in a field rely on chunking to help them process complex information across many contexts. Herbert Simon and William Chase (1973) showed chess masters and chess novices various positions of pieces on a chessboard for a few seconds each. The experts did a lot better than the novices in remembering the positions because they were able to see the “big picture.” They didn’t have to remember the position of each of the pieces individually, but chunked the pieces into several larger layouts. But when the researchers showed both groups random chess positions — positions that would be very unlikely to occur in real games — both groups did equally poorly, because in this situation the experts lost their ability to organize the layouts (see figure below). The same occurs for basketball: basketball players recall actual basketball positions much better than do nonplayers, but only when the positions make sense in terms of what is happening on the court, or what is likely to happen in the near future, and thus can be chunked into bigger units (Didierjean & Marmèche, 2005).

We will return to the idea of recoding information when we discuss factors that influence our initial encoding of information, or the way in which we take information into our system.