Waiting and perceived duration
How long a wait feels depends on occupation, explanation and certainty more than seconds.
Perceived wait time diverges systematically from clock time. Unoccupied, unexplained and open-ended waits feel longer, as do waits before a process starts compared with waits inside it. Because satisfaction tracks the perception, a wait can be improved without being shortened, though shortening still helps.
Same clock time, different felt time
How it shows up in software
Loading is the most common waiting experience in software, and most of it is unoccupied and unexplained. Skeleton screens, streamed partial results and named processing stages all shift the wait from blank to occupied. The worst pattern is a determinate progress bar that lies, because it converts a wait into a broken promise.
Using it well
- Stream partial results where the architecture allows. First tokens or first rows reaching the screen converts a pre-process wait into an in-process one.
- Name the current stage in words when a process has real stages, and only show a percentage when you can compute an honest one.
- Occupy the wait with something related to the task, such as the document skeleton, rather than an unrelated animation.
- Give long jobs an expected completion and a way to leave. Letting someone close the tab and be notified removes the wait rather than dressing it.
Where it turns manipulative
- Artificial delays that stage an impression of computation, common in scoring, matching and AI-branded features.
- Progress bars that move on a timer rather than on actual progress, which trade a short trust gain for a larger loss at the stall.
- Filling a wait with upsell content, which converts idle time into a sales surface the person cannot decline.
Where you have seen it
ChatGPT
Streams output token by token, so the response begins before it is complete and the wait happens inside the result.
Facebook
Popularised skeleton screens, showing the shape of the content in place before the data arrives.
GitHub Actions
Long builds show named steps with per-step timings and live log output rather than a single overall bar.
What the research says
- Maister, 1985Mixed evidence
Set out propositions on waiting: occupied time feels shorter than unoccupied, pre-process waits feel longer than in-process, anxiety and uncertainty lengthen the felt wait.
An influential practitioner essay rather than an experimental paper. Several propositions have since been tested, with mixed support; the occupied-time and uncertainty ones hold up best.
- Hui and Tse, 1996Mixed evidence
Manipulated wait-duration and queue-position information in a service setting. Information improved evaluations for longer waits, with less benefit for short ones.
- Nah, 2004Mixed evidence
Measured tolerable waiting time for web page responses and found tolerance dropping sharply beyond a couple of seconds, with feedback extending tolerance.
Dated by modern connection speeds and expectations. Use it for the shape of the tolerance curve and the feedback effect, not for the thresholds.
- The Houston airport baggage-walk storyContested
Widely retold as an example of moving passengers further from the carousel to cut perceived wait.
Included here as a caution. It circulates without a citable primary source or measured result, and should not be used as evidence.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.