# How much can users hold in their short term memory

I recently undertook a questionnaire and was asked a question about short term memory. One that I just can't find an answer to at all as it seems to be so contradictory.

This got me curious. How much can a user generally remember when viewing a website? There are various research papers on this including Heyer and Barrett (1974) and Shulman (1970).

The question was:

Research has shown that short-term memory is fragile. Which of the statements below is more accurate in describing how much information we can remember and for how long?

1. The amount you can remember is equal to about what you can say in 20 seconds, you will be able to hold this remembered amount in short-term memory for about 40 seconds.
2. The amount you can remember is equal to about what you can say in 30 seconds, you will be able to hold this remembered amount in short-term memory for about 20 seconds.
3. The amount you can remember is equal to about what you can say in 20 seconds, you will be able to hold this remembered amount in short-term memory for about 10 seconds.
4. The amount you can remember is equal to about what you can say in 20 seconds, you will be able to hold this remembered amount in short-term memory for about 20 seconds.

Does anybody know the definitive answer to this or, like me,

-

## migrated from ux.stackexchange.comJun 3 '13 at 16:54

This question came from our site for user experience researchers and experts.

Generally, don't assume you always have full attention. Design your UI for people using it while watching TV and juggling nuclear fuel rods with the other hand. – peterchen Jun 3 '13 at 12:02

This is a very simple question with no single answer, I'm afraid. The other classic paper quoted in this context is Miller's "The Magical Number Seven, Plus or Minus Two" (e.g. http://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_or_Minus_Two). In the case of a user using a website, memory will also be affected by:

• Familiarity with the subject. Is the website about a topic the user is intimately familiar with, or something completely new (unlikely) or somewhere in between?
• Available contextual information. Can the user quickly get to supporting information? In which case, the load on memory may effectively be reduced
• The extent to which the information presented on the site is related. The key message of Miller's paper is that people group ('chunk') information in order to reduce memory load

Not sure if your question relates to a specific problem, or if it's a general query, but things like memory limits are hard to define absolutely, hence testing should always take place in as realistic a context as possible.

-
Thanks Peter - good enough answer although was specifically after an answer 1-4 really - the questionnaire is from HFI and must be respectable. What would you put if you had to? – David M Jun 5 '13 at 10:45

Sorry David :) I genuinely think the question isn't very meaningful: 20 seconds of speech could be information dense, or information poor. For instance, if we're talking about a website we could assume the listener knows context, or not. So with context "The header had primary and utility navigation, and a company logo in the top left.". Without context, we'd have to add that the header is at the top of the page, which would add to the time taken.

People have explored different ways of encoding memory: acoustically, semantically and visually. http://www.s-cool.co.uk/a-level/psychology/human-memory/revise-it/short-term-memory looks at the merits of each theory. At the end is quotes a study saying that in the absence of rehearsal, short term memory can only hold on to information for 15-30 seconds. So I'd go with 4, but:

• I don't think the question is reasonable, for the reasons already stated, plus the likely error bounds in any memory study
• Memory isn't just one thing - it's not just speech, or pictures. It's also affected by how we as humans choose to encode it. Simple encodings work well for random lists, but websites aren't seen or interpreted like that

Obviously I've not seen the questionnaire! I'd be interested to know what HFI come back with - would be great if you could let me know.

-
The answer was apparently b) but I completely understand your viewpoint. It is frustrating that the question is so ... vague and lacks context. – David M Jun 7 '13 at 10:38

This is still an open research question, and there are a number of different hypotheses. The issue is further confused by mounting evidence that short-term memory and working memory are different processes, with working memory being suggested to be, basically, short-term memory plus attentional and executive processes that help maintain and control the contents of memory. (The following paper provides a very clear demonstration: http://psycnet.apa.org/index.cfm?fa=search.displayrecord&uid=1999-11137-004) The addition of these control processes can drastically affect memory capacity, and there's not always a clear line between a working memory task or a short-term memory task.

I agree with the above poster - the question is too vague to answer. Working and short-term memory capacity can vary based on a wide variety of parameters, including type of stimulus (it's easier to remember a series of numbers than it is to remember a series of abstract geometrical shapes), concurrent tasks (do you just have to remember the stimuli, or do you have to maintain the stimuli in memory while simultaneously doing something else?), experience with the task, memory training...it's a long list.

You may find a recent book to be of interest: Cowan, N. (2005). Working memory capacity. Psychology Press; ISBN 184169097X. (Keep in mind, though, that Cowan has his own model of memory capacity, so does not take an entirely neutral position.)

-