Response Time and Display Rate

The bar above shows two important things about response time.

  1. The user starts to type before the activity that is being requested is initiated.
  2. The computer's response is neither instantaneous nor all at once.
Designers of user interface must take this into account. Thus when the requested process will take some time the system should react quickly with some indication that it is working. This is the origin of the Mac clock and the Windows hourglass.

Theoretical Foundations

Any moderately complex task involves both long term and short term memory. In a famous 1956 paper George Miller advanced the hypothesis that the capacity of human short term memory is seven chunks of information (plus or minus two). Thus solutions to problems that require the someone to remember more than seven pieces of information are likely to be more difficult to implement since some of the information must be stored in long term memory, which for most of us means writing it down.

(I am reminded of a description of the English University system that a contemporary of mine who was at Oxford gave me. He called it "the sponge theory of learning". You cram your head with everything about a certain topic and in the exam squeeze all the knowledge out leaving the examination room having forgotten everything you knew about the topic. Ironically it probably is very good training if you are going to be a trial lawyer or a cabinet minister.)

The challenge then is for the system to have the long term memory and present the short term needs "just in time" so that the user does not need to maintain information in long term memory. Long delays in system response can cause the user to forget her plan with a subsequent further slowdown. On the other hand if the user works too fast she is more prone to errors brought on by excessive haste. There is some balance between speed and accuracy that needs to be maintained.

Response Times: Expectations and Attitudes

Based upon the research in the field Schneiderman draws the following conclusions.

  1. People will work faster as they gain experience with a command so that it may be advantageous to allow people to set their own pace of interaction. (For a horrifying example of what can happen if the designers do not take this into account review the story of the Therac -25 disaster by Nancy Levinson and Clark Turner and originally published in IEEE Computer vol 26 no. 7 July 1993 pp 18-41. Or there is a shorter version of the history of the Therac-25).
  2. In the absence of such constraints as cost or technical feasibility users will force response times down to well below 1 second.
  3. Although people can adapt to slower response times they are generally dissatisfied with times longer than 2 seconds.

It should be noted that much of this is task dependent and that users can adapt to significant variations if they have to.


Last Changed: April 28th 1999.
Push Here To return to GUI home page.