Week 6 Studio: Motion, and how long it felt

Three animations over identical twenty-second waits, ranked by the class, and a careful account of why that ranking is weak evidence.

1 April 2027

There is published work claiming that the shape of a progress animation changes how long the wait is perceived to be. Chris Harrison and colleagues at Carnegie Mellon ran the best-known set of these experiments; their summary page is in the reading below, and it reports that bars which accelerate toward the end were perceived as faster than linear bars of the same duration.

You are not going to take that on trust, and you are also not going to be able to replicate it properly in this room. Both halves of that sentence are the lesson, and the second half is the one that transfers.

By the end of this week#

You can:

• build motion treatments that differ only in easing, over waits of provably identical duration;
• run a simple ranking with a class and record the result without cleaning it up;
• name at least three specific, concrete reasons a result like this is weak — not “small sample” in the abstract, but what went wrong in this room;
• decide what to do with a weak result, which includes doing nothing.

The idea#

No lecture this week. This section and the 14:00 framing are the teaching.

What the published studies actually measured. Perceived duration, usually via forced-choice comparison between two animations of identical real duration, with participants who did not know the durations matched. Reported effects are in the range of a few per cent of perceived time. They are not claims about whether users trust the bar, understand it, or would rather have a number.

What a room of twenty classmates can and cannot reproduce. You can reproduce the procedure. You cannot reproduce the conditions, for reasons that are specific and worth naming out loud today:

• Not blind. You built the treatments. You know which one you expect to win.
• Order effects. Whoever sees the pulsed bar third has been waiting a full minute by then, and fatigue is not easing.
• Demand characteristics. Your participants are classmates who know this is a course about progress bars in a week about animation. They will produce the answer the exercise appears to want.
• No real stakes. Nobody in the room is waiting for a file they need. Perceived duration under no pressure is a different measurement.
• n ≈ 20, one room, one afternoon. Even a genuine effect of a few per cent is far below what twenty rankings can detect.

Recognising all five of those on your own work, in the moment, is the skill. Week 11 gives you a better-designed observation and you will need this instinct there too.

The measure. Forced ranking, three treatments, “which felt longest to shortest”. Ranking rather than seconds, because asking people for a duration estimate in seconds introduces a numeracy problem on top of everything else.

View this week’s slides — the 14:00 framing runs from these. They carry the worked example and the exercise brief. On a phone they are small — the deck is built for a projector, and this page carries the same material as prose.

In the room#

Thursday 1 April, 14:00–17:00. No lecture this week.

What you start with#

• Your prototype through week 5, with one linear animation and an easing hook that is wired up and unused.
• A 20-second simulated wait that all three treatments will run over.
• Printed ranking slips, and the Harrison summary page from the reading.

What you build#

1. Three motion treatments over a wait of identical duration — linear, decelerating, and pulsed. Same start, same end, same total, differing only in easing.
2. A duration assertion in code proving the three are equal. This is a real requirement: a treatment that is 400 ms longer invalidates the whole exercise and the difference is invisible by eye.
3. The raw ranking table, exactly as collected.
4. A weaknesses paragraph naming at least three specific reasons the result is weak, drawn from your own run rather than from the list above.

A worked example#

The three easings, and the assertion that makes the comparison valid:

const TREATMENTS = {
  linear:     t => t,
  decelerate: t => 1 - (1 - t) ** 2,
  pulsed:     t => t + 0.02 * Math.sin(t * 18 * Math.PI),
};

// A treatment 400 ms longer than another invalidates the whole exercise,
// and the difference is invisible by eye.
for (const ease of Object.values(TREATMENTS)) {
  console.assert(Math.abs(ease(1) - 1) < 1e-9, "must finish at exactly 1");
}

Easing changes the shape of the journey, never its length. If ease(1) !== 1 you are comparing durations, not motions.

Done when#

• Your duration assertion passes, and you have run it.
• The table is raw. If you were tempted to drop a response, write down the temptation instead — that sentence is worth more than the cleaned table.
• Your three weaknesses are about this run. “Small sample size” on its own does not count; “n≈20 against a published effect of a few per cent” does.

The mistake to expect#

Dropping a response.

Someone ranks the pulsed bar fastest “because they weren’t paying attention”, and excluding them is very tempting. If you are tempted, write the temptation down instead — that sentence is worth more than the cleaned table.

The second: writing “small sample size” as your weakness. “n ≈ 20 against a published effect of a few per cent” is a reason. The other is a phrase.

Before next week#

About two hours.

1. Decide whether you are shipping any of the three, and write one sentence saying what your decision rests on. “The ranking” is a weak basis and saying so is fine. “I prefer it” is also fine, if labelled as preference.
2. Week 7 audits the treatments you built today. Before Thursday, turn on your operating system’s reduce-motion setting and look at all three. Bring what you find.

If you are joining late#

git checkout stage-06 gives you the prototype through week 5 with a single linear animation and an easing hook that is wired up and unused. It is a floor, not an answer: the three treatments, the duration assertion and every decision after the ranking are yours.

What this week cannot tell you#

Almost anything about perceived duration, and that is the intended outcome.

You ran a procedure in conditions that cannot support its conclusion. The correct write-up says so, keeps the raw table, and does not convert twenty rankings into a design principle. If your treatments came out in the same order as the published work, that is not replication — with this design it is close to a coin landing the way you expected.

What the week does give you is calibration. You now know from the inside what a weak result feels like while you are collecting it, which is the only reliable defence against over-reading the slightly better evidence you get in week 11.