Try it on yourself.

Twelve short experiments from the research behind our work. Do one, see your own result, then see how everyone else did and what it means for design.

Results so far
–
No pooled results yet
Experiments
12
Done by you, this visit
0

Your own result

Five effects you can feel in under a minute.

Each compares you with yourself: one condition against another, on the same device, in the same minute. That is why the result can be trusted on any phone or laptop.

When two cues disagree

Say what colour the word is printed in. Easy, until the word disagrees.

50 seconds

A colour word appears. Tap the colour of its ink, not what it says. Twenty-four words, as fast as you can without mistakes.

The theoryRobust
What it is
Naming the ink of a colour word is slower when the word names a different colour. Reading is automatic, so it competes with the answer you are trying to give.
Where it comes from
Stroop (1935). One of the most repeated findings in psychology: an online replication measured 859 ms when word and ink agreed against 1,152 ms when they clashed (Crump, McDonnell & Gureckis, 2013).
What it means for design
When a label, colour or icon contradicts what a control does, such as a red “Confirm”, people slow down and make more mistakes. Make every cue on a control say the same thing.
What it does not show
It is not a measure of your attention or self-control, and we found no published test of the effect on real buttons: the design reading is a reasonable extension, not a finding.

Everyone’s result

One question. Two versions. You get one.

Half of visitors see one wording and half see another. Answer, then see which you had, what the other half saw, and how the two groups differ.

The first number you see

One fact, then one guess. The fact is true for everybody. It is not the same for everybody.

15 seconds

Read the fact, then give your best estimate.

The theoryRobust
What it is
A relevant number seen just before an estimate pulls the estimate towards it, even when the number is only a true outer limit.
Where it comes from
Jacowitz & Kahneman (1995). In the Many Labs 1 replication (Klein et al., 2014) the effect appeared in all 36 samples: people told Chicago had more than a low figure guessed 1.03 million on average, and people told it had fewer than a high figure guessed 3.00 million (among answers inside the stated limits).
What it means for design
The first explicit, relevant number someone sees becomes their reference point. Show the comparison price or the higher tier on purpose, not by accident.
What it does not show
Numbers that are merely nearby do nothing: in a 6,826-person test, a product called “P97” instead of “P17” did not move estimates.

Design “laws” on trial

Two rules of thumb, tested on you.

Much of what is quoted as a law of design was never measured. These two can be.

Is a tenth of a second fast enough?

Two buttons. One answers a little late. Can you tell which?

45 seconds

Press both buttons, then choose the one that answered late. The delay shrinks each time you are right.

The theoryModerate
What it is
The well-known limits of 0.1, 1 and 10 seconds for response time are reasoned bands, not measured thresholds. This measures the smallest added delay you notice.
Where it comes from
The limits come from Miller (1968) and Card, Robertson & Mackinlay (1991), made popular by Nielsen. A later touch-screen study found people notice added delays of about 11 ms when dragging and about 69 ms when tapping (Deber et al., 2015).
What it means for design
A tenth of a second is too generous for anything a finger moves directly. The right budget depends on whether someone is dragging, tapping or waiting for a page.
What it does not show
Your device has its own delay underneath, so this is the added delay you noticed on this device, not an absolute threshold.

Which wait felt shorter?

Two progress bars. Watch both. Which finished sooner?

15 seconds

Watch the first bar, then the second, then say which felt faster.

The theoryModerate
What it is
Two waits of the same length can feel different. A bar that speeds up towards the end tends to feel faster than one that moves steadily.
Where it comes from
Harrison, Amento, Kuznetsov & Bell (2007) and Harrison, Yeo & Hudson (2010): in paired comparisons, bars that accelerated felt faster, and one animated style cut the wait people perceived by about 11%.
What it means for design
How a wait feels can be shaped by how progress is shown. It is cheaper than making the wait shorter, and no substitute for it.
What it does not show
The 11% is one design from one lab on short waits. And a progress bar does not by itself make people finish a form: in web surveys, steady bars did not reduce drop-out.

The myths file

Things we will not tell you.

Design writing is full of numbers that sound like research. Tap one to see where it comes from.

“Respond within 400 ms: the Doherty threshold.”Myth as attributed

What the source really is

The 1982 IBM report it is credited to does not contain the figure. It argues for sub-second response from the company’s own internal productivity studies. The name appears to come from a 2014 television drama, Halt and Catch Fire, and was picked up in a 2015 blog post.

“Attention spans are now 8 seconds, shorter than a goldfish’s.”Myth

What the source really is

A 2015 marketing report citing a statistics website. No research has been found behind either number.

“Users read only 20% of a page.”An estimate, not a measurement

What the source really is

An estimate by Jakob Nielsen (2008) from the browser logs of 25 people: how long pages were open, set against how long the words would take to read. Nobody’s reading was observed.

“53% of mobile visitors leave if a page takes over 3 seconds.”Observational

What the source really is

Google’s own wording (2016) is “likely to be abandoned”, from analytics data on sites that opted in. It is an observation, not an experiment.

“Unfinished tasks are remembered better: the Zeigarnik effect.”Fails meta-analysis

What the source really is

A 2025 meta-analysis (Ghibellini & Meier) found no memory advantage for unfinished tasks. What does hold is a tendency to go back and finish them.

“Hard-to-read fonts make people think harder and remember more.”Myth

What the source really is

Seventeen studies with over 7,000 people found no benefit to reasoning (Meyer et al., 2015), and a typeface designed for the purpose gave equal or worse memory (Taylor et al., 2020).

“Fewer options always convert better: the jam study.”Mixed

What the source really is

Across 50 experiments with 5,036 people, the average effect of offering more choice was close to zero (Scheibehenne, Greifeneder & Todd, 2010). Sometimes more choice helps, sometimes it hurts.

“Add a decoy tier to push people to the premium plan.”Mixed

What the source really is

With real, pictured products the decoy effect did not appear, and sometimes reversed (Frederick, Lee & Baskin, 2014).

“Nudges reliably change behaviour.”Overstated

What the source really is

After correcting for the studies that never got published, the overall evidence is undecided (Maier et al., 2022). Some nudges may work; the average claim does not hold.

“Subtle cues prime behaviour, willpower runs out, and power poses work.”Failed to replicate

What the source really is

Large multi-lab replications found priming with flags and money near zero, and the willpower effect near zero across 23 and 36 labs; seven pre-registered studies found no effect of posing on behaviour.

What happens to what you do here
  • Nothing identifies you. There is no sign-up and nothing is stored on your device.
  • When you finish an activity, one small record is added to the totals: which experiment, which version you saw, and which band your answer fell in. For the pointing tasks, also whether you used a finger or a mouse.
  • Because nothing links a record to you, it cannot be taken back out once it is added. If you would rather not add to the totals, do not start an activity.
  • Where two versions of a question exist, you are shown one at random and told afterwards which, and what the other was.
  • These are demonstrations and pooled counts from whoever visits, not a controlled study. People who choose to visit a design studio’s lab are not everyone.