Atkinson-Shiffrin’s multi-store model of memory

Atkinson-Shiffrin’s multi-store model of memory is a model for explaining human memory.

The Atkinson-Shiffrin multi-store model of memory revolves around three distinct systems: sensory memory, short-term memory and long-term memory. Each of these systems work together – indeed, their borders are rather blurred – and at the same time.

Sensory memory

Sensory memory has a short (sometimes extremely short) duration: about 0.2 seconds – 10 seconds, with echoic sensory memory (what we hear) generally lasting longer than iconic sensory memory (what we see). It is the first stage of the multi-store model, coming before short-term memory and long-term memory, in that order.

Capacity: almost unlimited
Duration: 0.2 – 10 seconds, depending on the type of sensory memory

Whilst it is true that the capacity of sensory memory is virtually unlimited (that is, there are no real restraints on how much we can see or hear), we only attend to a small percentage of what is around us. Sensory memory which we attend to moves on to short-term memory, whilst sensory memory which we do not attend to is lost from memory after its initial duration. This process is shown in the diagram below.

Sensory memory acts as a filter. Imagine having to pay attention to every single thing we see and hear!

Short-term memory (STM)

When sensory information is given attention, it moves into short-term memory. Short-term memory is the second stage of the multi-store model, coming after sensory memory, and before long-term memory. Information can also move back from long-term memory into short-term memory when it is being retrieved.

Capacity: 7 ± 2 (5 – 9 items) items
Duration: 10 – 20 seconds

Short-term memory requires rehearsal. When information is rehearsed, it remains in short-term memory. There are two types of rehearsal – maintenance rehearsal and elaborative rehearsal – which can be used to encode information into long-term memory. If information is not rehearsed, it will be lost from memory due to either decay or interference.


Maintenance rehearsal

Maintenance rehearsal is a tool used to keep information in short-term memory, or to aid in encoding it into long-term memory. Often, both occur simultaneously. Maintenance rehearsal involves simply repeating information over and over again, either in one’s head or out loud.

Elaborative rehearsal

Elaborative rehearsal tries to keep memory active at a slightly deeper level, and tends to encode new information more effectively than maintenance rehearsal. In elaborate rehearsal, the original information is changed or manipulated in some way in order to link it to information which has already been encoded in long-term memory. There are numerous ways to do this, for example:

  • Make the new information into a story or a rhyme
  • Mnemonics
  • Linking meaning
  • Salience (linking the new information to our own experiences)

Whilst rehearsal can improve the duration of short-term memory, the capacity of short-term memory can be improved through chunking:


Chunking is a tool used to improve the capacity of short-term memory, which is usually 7 ± 2 (5 – 9) items. It refers to the mental organisation of stimuli into groups, allowing numerous pieces of information to be remembered as one item, or ‘chunk.’

For example, take the following set of numbers:


It is highly unlikely that you would remember all fifteen numbers in short-term memory (remember, the usual upper bound of short-term memory is nine items). However, using chunking, the same set of numbers can be re-written:

365     247     911     000     334

Now, there are only five pieces of information to remember. In this case, chunking is likely to be even more effective, as each chunk has significance:

  • 365: the amount of days in a year
  • 247: twenty-four hours, seven days per week
  • 911: either the American emergency service number, or a reference to September 11, 2001
  • 000: the Australian emergency service number
  • 334: Sir. Donald Bradman’s highest Test score for Australia

Long-term memory

When information has been encoded into memory through rehearsal, it is stored in long-term memory. From long-term memory, information can be retrieved to use in everyday life. For example, think about what you had for dinner last night. This information was in long-term memory, but now that you are thinking about it, is back in short-term memory.

Capacity: virtually unlimited
Duration: virtually unlimited

The information in long-term memory is stored in semantic networks: mental maps which group and link information based on meaning. As long-term memory appears to have no limits, we are likely to remember significant events until the day we die, as long as there is limited decay and injury.

Below is a diagram showing some of the processes of the Atkinson-Shiffrin multi-store model of memory. Note that when sensory memory is not attended to, it is lost. Click for a larger image.


Serial position effect

The serial position effect suggests that a piece of information’s relative order of presentation impacts the likelihood of being remembered.

Primacy effect

The primacy effect suggests that stimuli presented first or at the beginning of a list are more likely to be remembered than stimuli presented at the end and, in particular, in the middle of that same list.

This may be due to the fact that the learner has had more time to rehearse the information presented first, meaning it is more likely to have been encoded into long-term memory.

Recency effect

The recency effect suggests that stimuli presented at the end of a list are more likely to be remembered than stimuli presented in the middle of that same list.

This may be due to the fact that the learner still has the final few pieces of information presented stored in short-term memory.