Perceived duration can diverge from clock duration

What it is

The clock moved four seconds; a person may say “that was instant” or “that was a long wait”. Perceived duration and clock duration are not the same quantity. What gets remembered, and what gets called “slow”, is the perceived one. Divergence goes both ways: an empty spinning wait stretches the interval; attention occupied by other processing shrinks the same interval. Perceived performance starts by admitting two rulers, not by treating milliseconds on a perf report as the user report.

Why it happens

In prospective timing, attention acts like a gate. While the gate faces time — watching progress, watching a spinner, watching a clock — pacemaker pulses are counted more faithfully and the interval is judged long. That is the hovering clock: the more you watch time, the slower it goes. While the gate is occupied by task-related work, pulses are missed and the interval is judged short. Fill also acts on its own: a filled duration packed with events is often judged longer than an empty one in a pure timing task, because the fill becomes stimulation to count. In a “waiting for a result” task, if the fill pulls attention off timing, the direction can reverse. Divergence is therefore not a spell that “fill always shortens”. It is the sum of timing-attention and event-fill.

Retrospective judgement (only asked afterwards how long that was) takes another path: the more changes remembered, the longer the interval is judged (a storage-size account). An interface that packs a wait with meaningless state flashes will look longer in hindsight even if no one felt they were watching a clock at the time.

Studying it

Hold objective delay fixed, vary whether people monitor time and whether the interval contains events, and collect judged duration. Zakay and Block’s attentional-gate model treats “monitoring time or not” as the switch; Thomas and Weaver treat filled-interval judgement; “a watched pot never boils” is the folk hovering clock. HCI injects delay and uses a post-hoc scale, or asks people to adjust a wait until it matches “the one just now”.

Independent variables: clock duration, visible clock or progress, event density in the interval, prospective versus retrospective query. Dependent variables: estimated duration, ratio to clock, reports of “I was watching the time”.

Lab participants know they are in a timing study, so the gate faces time more often and overestimation is inflated. In product waits the main task is “is it done”; timing is a side job. Absolute lab numbers should not become product SLOs. The direction should: make people watch time, and they will judge it longer.

Where it stops holding

Very short intervals (inside the instant window) have almost no room to stretch or shrink; what is coded is causation, not waiting. On very long intervals people switch to an external clock (the phone); the perceived ruler is taken over by a wall clock and fill tricks in the UI have limited reach. Age, load and mood move the gate’s default; the same person’s ratio changes across moments. Filled duration can reverse direction between timing experiments and waiting experiments; citations must say which task, not eat both.

Applying it

• Do not treat a single clock-millisecond number as “how slow it felt”. The same request, empty versus occupied, should collect subjective duration separately.
• Avoid training people to watch a clock during a wait: uninformative flashes and countable spinner turns pull the gate onto time.
• When comparing versions, log both clock duration and post-hoc estimates, and look at the ratio, not only mean milliseconds.
• How to check: the same 5 s delay, one group watching an empty spinner, one able to keep reading already-present content. Ask immediately “how many seconds was that”. If the estimates match, you are still using the clock as perception; if the empty group runs clearly longer, divergence has been moved.

Related

• Same group: I2.07.2 A wait with feedback is perceived as shorter · I2.07.3 Optimising perception does not replace optimising the clock
• Nearby: I2.01 Choosing skeleton versus spinner · I1.02 Continuity-of-thought threshold · I1.05 Latency jitter
• Search terms: perceived duration · hovering clock · filled duration