Short Term Memory

Short Term Memory

Variety of Attributes

(1) Duration:

how long do short term memories last?

With rehearsal: indefinitely (in theory)

Without rehearsal: info lost very quickly

Can be studied with the Brown-Peterson technique (1959) – subjects receive a three letter stimulus and then asked to do something immediately after stimulus (counting backwards by 3s). The task immediately following stimulus in a memory experiment is called a distractor task. It requires full attention of the subject (like shadowing) and doesn’t allow rehearsal.

Memory for the letters is gone in as little as 20 seconds!!!

How can such quick forgetting be explained?


Petersons: simple decay function (items not rehearsed vanish as time passes)

Waugh and Norman (1965): interference, not decay, produced by the distractor task.

Decay or interference

Image via Wikipedia

Waugh and Norman (1965)

examined the effect different distractor tasks.

If forgetting is caused by decay, then different tasks shouldn’t matter (only that should matter is the time that goes by).

But different distractor tasks do affect memory differentially. (linguistic tasks interfere the most) (tapping a table doesn’t interfere)

Conclusion: interference is the reason STM fades quickly

(2) Capacity:

How much information can STM hold?

Miller (1956)

Review of previous research on how much you can hold in STM. He found that research was fairly consistent ranging from 5 to 9 items. He sold the idea by naming his paper “the Magical number 7, plus or minus 2”

Demonstration: people can’t remember 12 digit number. But can remember 149217762011 But this exceeds the magical number.

Miller proposed that we can chunk information in STM. Chunking is organizing information into meaningful units.

The second number (the one we can remember) can be chunked into three meaningful dates.


STM does have fixed capacity but it’s 7+/- 2 chunks, not items.

Miller said to think of STM like a purse that can hold just seven coins but they can be copper, or they can be gold.

(3) Coding:

In what form does STM hold the information in? (the form that information makes in STM)

Many different ways to hang on to the same thing.

Variety of Codes

Chair {visual code} – how the word looks

Picture of Chair {analog code} – mental image

sounding it out {acoustic code} – how the word sounds

definition of word {semantic code} – what the word means

Recoding – changing information from one code into another.  analog ——->acoustic

Does STM have a preferred code?

Conrad 1964:

Presented subjects with letters (visually).

Later asked to recall them.

Subjects confused letters S, F, and X. (similar sounds)

Rarely confused S and B. (similar shapes)

S,F, and X sound alike, but don’t look alike

Conclusion: STM likes acoustic coding


Later researchers found evidence for other codes in STM. STM seems to employ a variety of codes. Some take longer to form.

(4) Retrieval:

How is STM searched?

Sternberg paradigm:

Image via Wikipedia

Subjects memorize a list of 1 to 6 letters (or words or pictures of an object). (This is the memory set.) Allowed to study list as long as they need to. Subjects then would see a probe letter. They were asked whether the probe letter was part of the original memory set or not. They respond yes and press a button on keyboard or no button is pressed meaning the probe letter was not in memory set. Speed and accuracy are important.

Ind. var = memory set size

Dep var. = Reaction time


Search strategies:

Sternbergs paradigm allows different possibilities to be assessed:

(1) serial vs. parallel – do we search through the memory set one letter at a time (serial) or do we check them all at once (parallel)?

If all items can be checked at once (in parallel) reaction time won’t be affected by increases in memory set size.

If items must be checked one at a time (serially) then reaction time will increase as memory set size increases.

(2) Self-terminating vs exhaustive – (self-terminating = search through the memory set until we find a match, then stop) (Exhaustive = checking every item in the set, even if a match has been found).

If all the items in the memory set are checked regardless of a match then the results are the same as a serial set.

If the system can stop checking once a match is found then the yes button will take less time than no responses because fewer items will need to be checked. So you aren’t working through the entire list. On average you will work through half the list.

Therefore, the yes responses will have half the slope of the no responses. On average you will have to work through half the list.

Sternberg’s data shows evidence for exhaustive serial searching. It is also high speed. Slope of line is just .38 msec.  So every additional item is only .38 msec.

STM is serial, exhaustive, and high speed.

It’s serial because as the set size increases, reaction time increases.

It’s exhaustive because the yes and no responses take the same amount of time.

It’s high speed because each additional item in the memory set takes an additional 38 msec to check or about 26 items per second.

Why exhaustive?

An exhaustive search requires less attention than a self-terminating search. So a “fast, dumb” search may be the best compromise: speedy, highly accurate process but doesn’t tie up limited attentional resources.

Related articles


Leave a Reply

Your email address will not be published. Required fields are marked *