Psi man mascot

The Atkinson-Shiffrin Model

Memory is not one thing. Rather, it is any process that allows us to use previously stored information. Such processes may be widespread in the brain, and each major brain system may have its own form of memory.

This insight occurred gradually to modern psychologists. It represented a major shift of emphasis in memory research. Ebbinghaus and other early memory researchers tended to think memories were stored whole, and there was only one form of memory.

By the mid to late 1960s, psychologists still thought of memory as a single system, but they were getting comfort­able with the idea of human information processing. They realized that humans, like computers, had to acquire and organize information before remem­bering it.

In 1965 Atkinson and Shiffrin suggested that human memory was organized as a system with three stages. They elabo­rated on their idea in 1968, producing the following model of memory.

diagram of the 3-box model
The classic Atkinson & Shiffrin model

The diagram suggests that information from the environment first enters a sensory storage system, which Atkinson and Shiffrin called the sensory registers. Information in this system (the first box) is preserved for a brief period so the brain can process it.

What were the three stages of the memory system, according to Atkinson & Shiffrin?

Next the information enters a second box or memory system, labeled short-term memory. This box represents ongoing activity of the brain. It includes whatever is in the thought process.

In the 1890s, William James called this primary memory. In the 1970s it was called short-term memory (abbreviated STM). Now it is more commonly called working memory.

The word attention refers to activity in this memory system. It holds information so we can focus on it.

As time goes on, researchers make more subtle distinctions between various sorts of short-term storage processes: a phonological loop for sounds, a visual sketch pad, and other systems linked to different forms of information processing.

Finally, information enters long-term memory. As you can see, researchers in the mid-1960s (who used this scheme) thought of memory as a flow of infor­mation through different stages.

They did not yet think of memory as many different, interacting processes. That would change.

Criticisms of the Three-Box Model

The Atkinson-Shiffrin model was simple and clear. Everybody knew it was a simplification, and that helped the science of memory research advance. People amended it, criticized its shortcomings, and argued for distinctions between different forms of memory not shown in the model.

Here are some common criticisms of the three-box model of memory:

1. The sensory stores are sensory systems, not memory systems. For example, they do not support voluntary recall.

2. The three-box model shows nothing between short-term activation and long-term storage. However, evidence shows that information can be warmed up but outside of attention. In other words, intermediate levels of activation are possible.

3. The three-box model implies there is just one short-term system and just one long-term system. In reality, there are many memory systems operating in parallel (different systems for vision, language, and odor memory, among others). Each has short-term and long-term operations.

4) The Atkinson-Shiffrin model does not give enough emphasis to unconscious processes. Modern researchers find that unconscious and implicit forms of memory are more common than consciously directed memory processes.

What were objections to the Atkinson-Shiffrin model?

The old Atkinson-Shiffrin model had limit­ations. However, it was fruitful. Every researcher became familiar with it and used it as a foil (a sort of negative reference point) when proposing new models of memory.

Iconic Memory

In the classic Atkinson & Shiffrin model, the first box is labeled sensory registers. These are more commonly called the sensory stores today.

The sensory stores are like brief delay systems associated with each sensory system. They preserve a pattern of stimulation before it enters attention.

The sensory stores are sensory systems. They are also memory systems because they preserve information after the external stimulus is gone. Other names for the sensory stores are sensory buffers or very short term stores.

What does "icon" mean? What are characteristics of iconic memory?

Iconic memory is the sensory store for vision. The term icon means form or image. Ulric Neisser (1967) proposed this label to convey the idea that iconic memory preserves an exact duplicate of the image falling on the retina.

Iconic memory was investigated by George Sperling (1963). Sperling tested subjects by flashing several rows of letters on a screen for a split second to see how many letters they could read after very short exposures.

What is a tachistoscope? What did Sperling do?

Sperling used a tachistoscope (pro­nounced tuh-KISS-tuh-scope). This instrument was invented by Volkmann in 1859 to replace the previous method­ology of using electric sparks to produce brief visual exposures.

The tachistoscope used a shutter to flash a picture onto a screen for a brief time measured in milliseconds (thousandths of a second). In Sperling's experiment, subjects saw an array of letters flashed very briefly on a screen:




Subjects were asked to read as many letters as possible during the brief flash. Usually they could read only 3 or 4 letters.

Next Sperling tried a variation called the partial report method. After he flashed the letters he sounded a high, medium, or low tone. Depending on which tone was sounded, the subject read the high, medium, or low row of letters.

When was the tone sounded, in the "partial report" experiment?

The tone came 250 msec after the flash of letters. You might think it would not do the subjects any good.

However, Sperling found that as long as the tone was sounded within a quarter second of the flash, subjects could report 3 out of 4 letters from any row. This was proof they preserved a memory of the entire image for a quarter second.

Why does this memory system exist? Eye movements take about a quarter second, and during the eye movements called saccades, visual information from the eye to the brain is interrupted.

During an eye movement, the iconic memory system preserves information from the last place where the eye stopped. This is called the previous fixation point.

Therefore iconic memory helps maximize useful information available to the visual system. It preserves information from one eye fixation while the eye moves to the next fixation point.

What function does iconic memory serve?

Sperling's subjects were unaware of the gap between the flash of letters and the tone. They simply waited for a tone and read the appropriate row.

The subjects believed the image of letters was still showing on the screen when the tone sounded. Actually they were reading the letters from their iconic image when the tone sounded.

When did Sperling's subjects think the tone was sounded?

The iconic image is complete, containing all the image available from the eye. However, it lasts only a fraction of a second and cannot be conjured up voluntarily at a later time. Probably the location of the iconic image is the circuitry of the retina itself.

The visual areas of the brain can also preserve images or pictures for about five minutes. This information is not quite as complete as iconic imagery, and most people have limited access to this information.

Echoic Memory

Just as the eye has a delay system to cling onto sensory information, so does the ear. The auditory information store, dubbed echoic memory by Neisser (1967), lasts one or two seconds. Echoic memory can also be called the auditory store or auditory sensory register.

What is echoic memory and how long does it last? How did Guttman and Julesz test the duration of echoic memory and what did they determine?

One creative experiment designed to measure echoic memory was carried out by Guttman and Julesz (1963). They used a computer to generate repeating segments of white noise.

White noise is composed of all fre­quencies randomly mixed together. It sounds like "shhhh" and cannot be described or memorized. The computer made it possible to put together a repeating pattern of white noise with no gap between repetitions.

The subjects had no clue that a sound was being repeated. Guttman and Julesz instructed subjects to put on head­phones, listen to the noise, and report what they heard.

If the repeating segment of white noise lasted longer than a few seconds, the subjects never realized it was repeating. They heard a continuous whooshing sound with no pattern.

If the segment of white noise was less than two seconds long, the subjects realized they heard a repeated sound. They still could not describe the sound (other than saying "shh, shh, shh") but they knew it was being repeated.

To detect a repeating pattern of random frequencies, subjects must use a memory system capable of preserving an exact copy of the noise from one repetition to the next. This is what echoic memory does: it preserves the exact pattern of sound for one or two seconds.

How does the "Why did you say?" phenomenon illustrate echoic memory?

A less scientific demonstration of echoic memory is the "What did you say?" phenomenon, which goes like this:

Person #1: "What time is it?"

Person #2: "What did you say? Oh, 2:30."

The second person hears the question after asking, "What did you say?" This is due to echoic memory, which holds the sound of the question for a second or two.

Even if you were not paying attention to the words when they were uttered, you can "hear" them when you turn your attention to them. This can be annoying to the person who starts repeating the question only to be interrupted by an answer.

Development of brain scanning tech­nology made it possible to observe echoic memory activity in the brain. Using MEG (magneto­encephalography), the team of Lu, Williamson, and Kaufman (1992) were able to show activity in a portion of the auditory cortex (part of the cerebral cortex which responds to sound) lasting two to five seconds after a sound stimulus.

Taste Flashes

Sensory stores exist in realms other than vision and hearing. Kelling and Halpern (1983) used a special apparatus to study "taste flashes."

The equipment designed by the re­searchers applied salt or saccharin solution to the tongue for 100, 200, 300 or 1000 milliseconds. Then the solution was quickly washed away with distilled water.

This was supposed to be analogous to the flash of letters in the Sperling task. The stimulus ended before the subject could report it.

How was the "taste flash" experiment carried out, and what did it reveal?

Taste recognition accuracy was 68% for salt. It was 94% for saccharin even at the shortest pulse duration: 100 msec.

There must be some kind of "gustatory sensory register" that preserves the taste information from such a short pulse, otherwise the taste would have disap­peared before subjects could identify it.


Atkinson, R. C. & Shiffrin, R. M. (1968). Human memory: A proposed system and its control. Psychology of Learning and Memory, 2, 89-193.

Guttman, N. & Julesz, B. (1963) Lower limits of auditory periodicity analysis. JASA, 35, p.610.

Ishai, A. & Sagi, D. (1995) Common mechanisms of visual imagery and perception. Science, 268, 1772-1774.

Kelling, S. T. & Halpern, B. (1983). Taste flashes: Reaction times, intensity, and quality. Science, 219, 412-414.

Lu, Z. L., Williamson, S. J., & Kaufman, L. (1992). Behavioral lifetime of human auditory sensory memory predicted by physiological measures. Science, 258, 1668-1669.

Neisser, U. (1967). Cognitive Psychology. Englewood Cliffs, NJ: Prentice Hall.

Sperling, G. (1963). A model for visual memory tasks. Human Factors, 5, 19-31.

Write to Dr. Dewey at

Don't see what you need? Psych Web has over 1,000 pages, so it may be elsewhere on the site. Do a site-specific Google search using the box below.