What a progress bar's own behaviour communicates about an unfolding wait, and what Harrison et al.'s controlled comparison actually measured when they varied that behaviour under a fixed duration.
A file is copying. A bar fills from left to right. A percentage climbs: 40, 50, 60… This is a hypothetical, not a specific observation — but it’s one you’ll recognise, because almost every digital wait now comes with some piece of interface trying to represent how it’s going while it goes.
This week’s claim, plainly#
How an interface represents an ongoing wait does more than disclose that progress exists: its changing behaviour gives the waiter a particular picture of how the process is unfolding. In Harrison et al.’s controlled comparison, different progress-bar behaviours changed which bar was perceived as faster, even though the actual duration behind every bar stayed fixed. That’s the empirical backbone this week rests on — a measured change in perceived speed and preference under a fixed, controlled duration. It is not a measured change in what someone expected, and this lecture doesn’t claim otherwise.
From the log to the system#
Three mechanisms have built our vocabulary for reading a wait: what we attend to, what we expected, and what information arrived. Now we turn that vocabulary onto the system doing the waiting with us.
Redrawing Week 4’s line#
Week 4 established a first distinction: whether explicit information was available during the wait — arriving only after it had already begun, and kept separate from Week 3’s expectation, which is fixed before a wait starts. This week’s question follows that distinction: given that a system represents the wait’s progress in some way, what does the specific behaviour of that representation communicate about the process as it unfolds? Two interfaces can both satisfy Week 4’s distinction — both disclose something, during the wait — and still differ completely on this week’s question, because one behaves as though steadily converging on completion and the other behaves as stalled, erratic, or offers no positional claim at all.
This course’s own vocabulary for that behaviour — representation, trajectory, shape, implied completion — is ours, not the research literature’s. Keep that apart from what the cited studies themselves measured, which is narrower: perceived speed and preference, under a fixed duration.
What Harrison et al. actually did#
Harrison, Amento, Kuznetsov and Bell built nine different progress-bar behaviours — linear, ones that paused early or late, ones that accelerated or decelerated, ones that moved in visible steps — and had 22 participants compare pairs of them, judging which of the two seemed to finish faster, while every bar in every trial ran for the same fixed 5.5 seconds (Harrison, Amento, Kuznetsov & Bell, 2007). That’s a tightly controlled, paired-comparison task on a short, fixed, identical duration — not a study of how a real, longer wait gets judged in minutes, and not a measure of expectation. Read their findings at that scope:
- Progress bars with a pause near the end of their run were reliably perceived as slower, even though the underlying duration hadn’t changed.
- Accelerating bars — ones that moved fastest right before finishing — were reliably preferred and perceived as faster.
- Participants also showed an order bias, tending to rate whichever bar they saw first as the faster one, independent of its actual behaviour.
Harrison et al. don’t claim any of this generalises to how long a wait feels in general, and they didn’t measure whether a bar’s behaviour changes what someone expects would happen next, or a free-form judged-duration report of the kind this course’s log collects. What their result licenses is narrower: within a short, controlled, paired comparison, a progress bar’s own behaviour near its end changed which one people perceived as faster, and which one they preferred.
Reading Myers alongside Harrison#
Myers ran an earlier experiment on a related but different question: whether having a percent-done indicator at all changes anything (Myers, 1985). His work establishes that a percent-done indicator gives a user information about a task’s progress and completion — something to estimate how much has been done, or roughly when a task might finish — and that a formal experiment found a significant preference for systems that included a progress indicator over systems that didn’t. Read that as evidence about whether an indicator is present, not about which specific behaviour a bar should show. That’s why this week reads Myers and Harrison together rather than treating either as sufficient alone: Myers speaks to presence, closer to Week 4’s territory; Harrison speaks to behaviour, which is this week’s new question.
What this doesn’t establish#
None of this licenses “progress bars make waits feel shorter” as a general claim, and it doesn’t license “an accelerating bar is always the better design.” Harrison et al.’s result is about perceived speed and preference in a short, fixed, paired-comparison task; it says nothing about longer real-world waits, about this course’s free-form “Judged duration” field, about what someone expected going in, or about whether a misleading bar backfires once someone notices the trick. That’s a question this course is raising, not one either paper answers.
Why this matters#
Weeks 1 to 4 built one log and reread it through three mechanisms that live in what a person attends to, expects, and is told. From this week, the course turns outward, to how waits get built — starting with the most literal act of representing a wait’s state that exists, a progress bar or a spinner. Week 6 asks the same question of a physical room instead of a screen: what a waiting environment, rather than an interface, communicates about how a wait will go. Together they’re the first matched pair in this course’s analyse block, one designed and digital, one physical and social, both asking what a wait’s own design communicates about itself.
What you’re doing this week#
Keep the wait-observation log’s method — a structured entry, no manipulation, honest about what you actually saw — but extend it with one new, clearly labelled field. The session has you log what a digital wait’s progress or status representation actually did, not just whether one existed, and think through what its behaviour communicated about the process as it unfolded.
Teaching team
- Marisol Quaye — convenor