I recall reading that the average person can only effectively compare 7 things at once. Has anyone heard of this and can reference any studies/research done on object comparison?

To give you some context for why I am asking: I'm designing a bar chart and wondering how many bars should be displayed. We are dealing with relatively large sets of x-axis members (100+) so displaying all the members is not an option, the plan is to sort the members, display the top x, and group the remainder into an 'others' bar (which would show the avg. value of all the grouped members) - I'm wondering what "x" should be.

  • Could you please provide a bit more context still: E.g. what is the user goal in relation to deciphering the bar chart?
    – agib
    May 13, 2011 at 8:50

4 Answers 4


First some clarification of the working memory theory.

Miller never said that people can only keep seven things (+/- 2) in their short term memory. His paper was about working memory (an entirely different concept). There are various numbers (note the plural) representing the chunks of info we can work with that are affected by context and type of information. As Miller and numerous others have found, they are generally lower than 7, with some types of info being as low as 1.

I hate citing wikipedia for something like this, but theirs is the neatest layman's write-up I have seen: http://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_or_Minus_Two

With specific regards the bar chart - users will typically only compare two bars at a time (by compare I mean not just look for the blunt affect of higher and lower, but actually check the numbers represented). They will fix one bar as the standard, and then scan the other bars measuring each against the standard, one at a time. The entire graph can be scanned for the trend over time - erikrojo's highs and lows - but that's a different form of comparison.

In short - I'd recommend you use as many as you need to provide meaning to the numbers in your context. If it's too many on the x-axis (requiring a horizontal scroll), consider providing a means for the user to freeze a section of the graph or select the number to display per jonshariat's answer.


This is covered on UX Myths and is false.

The original study done by Miller says that people cannot keep more than 7 (give or take 2) items in their short-term memory at one time. However, that is not the case with the web and any GUI. In fact, people can do well with lots of information and links on one page.

If you can let them choose how many are displayed. You could go with a default of 10 displayed at once.


If you are talking about a regular bar chart (I have no clue what x-axis members are) you can show as many bars as fitting to a screen or paper. They could even be only a pixel wide. As a viewer you will focus on peaks and lows. If you have descriptions (or images) to each bar than the width of the text will limit the amout of bars next (or top) to each other. You could add a line on top of each bar on mouse over to compare this specific bar to other bars. Additionally you can have subtle lines (with steady gaps inbetween) to help comparing bars that are far from each other. A flexible sorting feature can also be a further help. Highest -> Lowest. A - Z. Popular -> Unpopular. etc. You can add further information into the bars by color or showing two or more bars for each entry. Instead of grouping some bars into an "others" bar you could show all bars and display an "average"-line. And also label that line, to make clear what it's for.


There's some talk towards a cognitive limit at the point where people stop subitizing things and start to actually count them (subitize = unconciously build groups and quickly infer their number out of them).

According to Cowan, for the majority of people, this number is five, i.e. four items can still be subitized.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.