Lesson 3 of 11 · 18 min
Investigating the duration and capacity of short-term memory
Analyse Peterson and Peterson's (1959) duration experiment and Miller's (1956) capacity research, then design, run and interpret your own short-term memory investigations.
What you will be able to do
- Describe the aim, design, procedure and findings of Peterson and Peterson (1959).
- Explain why a distractor task is needed to measure the duration of STM.
- Describe Miller's (1956) conclusion about the capacity of STM and the role of chunking.
- Identify the variables, design and controls of a short-term memory investigation and draw conclusions from its data.
How would you measure how long something lasts in short-term memory? The obvious method fails: show someone three letters, wait 15 seconds and ask for them, and they will get them right every time, because they spent the 15 seconds repeating them. To measure STM's duration you have to stop rehearsal without filling STM with something that pushes the letters out for another reason. Peterson and Peterson's (1959) solution was a distractor task.
History
Peterson and Peterson (1959)
Aim. To find out how long information lasts in short-term memory when it cannot be rehearsed.
Participants. 24 students from introductory psychology courses (Indiana University).
Procedure. On each trial the experimenter spoke a consonant trigram (three consonants such as CHJ) followed immediately by a three-digit number. The participant counted backwards from that number in threes or fours (506, 503, 500, …), in time with a metronome, until a light signalled recall after 3, 6, 9, 12, 15 or 18 seconds. They then recalled the trigram. Each participant had 48 trials, with every interval tested in a balanced order.
Findings. Recall fell steeply as the interval grew: about 80% of trigrams were recalled correctly after 3 seconds, falling to about 10% after 18 seconds.
Conclusion. Without rehearsal, information is lost from STM within about 18 seconds. The researchers explained the loss as decay.
The shape of Peterson and Peterson's (1959) result: recall of a trigram falls steeply over the first few seconds without rehearsal.
Reading the design. This is an experiment with a repeated measures design: each participant experienced every level of the independent variable.
- Independent variable (IV): the retention interval (3, 6, 9, 12, 15 or 18 seconds).
- Dependent variable (DV): recall of the trigram, operationalised as the percentage of trigrams recalled correctly (all three letters in the right order).
- Controlled variables: consonant trigrams (not words, which are easier to remember), the same distractor task on every trial, the same presentation by the same experimenter.
The counting task is the key control. It prevents rehearsal, so the loss of the trigram can be put down to time passing in STM.
Classify each variable in a class replication of Peterson and Peterson (1959), choose the operational definitions, and pick the hypotheses.
Why did participants in Peterson and Peterson (1959) count backwards in threes or fours during the retention interval?
Hear or read a trigram, count backwards in threes on the keypad, then recall. Your result is plotted over Peterson and Peterson's. Results stay in this browser and are for learning only.
Note
Decay or interference? The evidence is open to interpretation
Peterson and Peterson concluded that the trigrams decayed. Keppel and Underwood (1962) looked again at the first few trials of experiments like this one and found that on the very first trial recall was good even after a long delay. Forgetting grew over later trials, as earlier trigrams built up. That pattern fits proactive interference (old trigrams disrupting new ones) better than decay. The same data supported two explanations, which is why the syllabus asks you to consider that memory research cannot always give a definitive answer.
Capacity: Miller (1956). Miller's paper, The magical number seven, plus or minus two, was a review: he compared the results of many earlier experiments, including digit span tasks, in which a person repeats back sequences of increasing length until they fail. Spans clustered around seven items. Miller's key insight was that the unit is the chunk: people can hold about seven chunks whatever their size, so recoding information into bigger, meaningful chunks increases how much STM can hold.
Later research suggests the true limit is smaller, about four chunks (Cowan, 2001), when people are stopped from rehearsing or chunking. For this course, state STM's capacity as 7 ± 2 items and mention chunking.
A class investigation of capacity. A teacher tests the digit span of 12 Year 12 students: each student hears sequences of digits, one a second, starting at three digits and adding one each time, until they fail two sequences of the same length. Their spans were:
Check the mean, median and standard deviation against your own working.
Calculate the mean digit span of the 12 students, to two decimal places. Is it consistent with Miller's (1956) capacity of STM?
Worked example
Analyse a class replication of Peterson and Peterson (1959)
A Year 12 class replicated Peterson and Peterson (1959). Each of 10 students heard 12 consonant trigrams, 4 at each of three retention intervals (3, 9 and 18 seconds) in a random order, counting backwards in threes during each interval. The table shows the number of trigrams each student recalled correctly at each interval.
| Student | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| 3 s | 4 | 3 | 4 | 4 | 3 | 4 | 3 | 4 | 4 | 3 |
| 9 s | 2 | 1 | 3 | 2 | 2 | 1 | 2 | 3 | 1 | 2 |
| 18 s | 1 | 0 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 1 |
(a) Identify the experimental design. (b) Calculate the mean percentage of trigrams recalled at each interval. (c) Describe the trend. (d) Draw a conclusion and give one limitation of the evidence.
Step 1: (a) Design
Every student was tested at every interval: a repeated measures design. Mixing the intervals in a random order controls order effects such as practice and fatigue.
Step 2: (b) Mean percentages
Totals: 3 s: ; 9 s: ; 18 s: , each out of trigrams.
3 s: ; 9 s: ; 18 s: .
Each student had the same number of trials at each interval, so the percentage from the totals equals the mean of the students' percentages.
Step 3: (c) Trend
Recall fell as the retention interval increased, from 90% at 3 s to 47.5% at 9 s and 12.5% at 18 s. The fall was steepest over the first few seconds.
Step 4: (d) Conclusion
Information that cannot be rehearsed is quickly lost from short-term memory, almost entirely within about 18 seconds, consistent with Peterson and Peterson (1959).
Step 5: (d) Limitation
With 12 trials in a row, earlier trigrams may have interfered with later ones (proactive interference), so the loss may not be due to decay alone. The sample of 10 students from one class is also small and not random, which limits generalisation.
Try it yourself first, then check the working one step at a time.
Answer
(a) Repeated measures. (b) 90%, 47.5% and 12.5%. (c) Recall fell steeply as the interval increased. (d) Unrehearsed information is lost from STM within about 18 s; limitation: proactive interference across trials (or the small convenience sample).
Method
Total the correct responses for each condition and divide by the number of trials in it.
Describe a trend with its direction and the values at the ends.
Conclude about the theory, then qualify the conclusion with a limitation.
Watch out for
Dividing the total by 10 (students) instead of 40 (trials) for a percentage.
Concluding that decay is proven: interference explains the same pattern.
Exam tip
Data questions on short-term memory
- Calculate: show the formula with the numbers in it, then the answer with its unit (digits, %, s).
- Identify a trend: give the direction and the size with values from the data: "recall fell from 85% at 3 s to 15% at 18 s".
- Draw a conclusion: link the trend to the model ("information that cannot be rehearsed is lost from STM within about 18 seconds"), and say how far it can be trusted: the sample, the artificial material, and whether interference could explain it too.
Summary
Peterson and Peterson (1959): consonant trigrams, a counting-backwards distractor task, retention intervals of 3 to 18 seconds; recall fell from about 80% to about 10%, so unrehearsed information lasts only about 18 seconds in STM.
The distractor task prevents rehearsal: it is the control that makes the experiment measure duration.
The design is repeated measures; IV retention interval, DV percentage of trigrams recalled correctly.
The loss may be due to proactive interference rather than decay (Keppel & Underwood, 1962): the data are open to interpretation.
Miller (1956) reviewed digit span and other research: STM holds about 7 ± 2 chunks, and chunking increases how much information it holds.
Finished the lesson?
Try the 2 quick checks above and the lesson completes itself, or mark it complete now.