What 10,000 Real Cognitive Test Scores Look Like

1. The textbook number is not your number

Published research puts median simple reaction time around 250–273 ms. Our pool's median is 370 ms across 2,688 attempts. Neither number is wrong — they measure different situations. Lab figures come from dedicated hardware, practiced participants, and controlled conditions. Ours come from laptop trackpads, phone screens, background tabs, and first tries. The gap between the two is mostly hardware and context, not people. If you have ever scored 350 ms and wondered why you are “slower than average,” you probably are not — you are mid-pack among people actually taking the test in a browser.

2. Phones cost you ~54 ms on reaction time — and nothing on clicking

The same test, split by device, tells two different stories. On reaction time, desktop's median is 352 ms against mobile's 406 ms — and at the fast end the gap widens (top-10% desktop: 255 ms; mobile: 329 ms). Touch screens add sampling and processing latency to every tap, and it shows up in the data as a flat tax.

Click speed shows no such tax: desktop's median is 6.7 CPS, mobile's is 6.5. Tapping a screen is as fast as clicking a mouse when the score depends on repeat cadence rather than response latency. Practical takeaway: compare reaction-time scores only against the same device class, and treat click-speed scores as roughly device-neutral.

3. The practice effect is ~110 ms, and almost everyone retries

Of the people who took the reaction-time test, 83% tried more than once. Their first attempts have a median of 402 ms. By the fifth attempt and beyond, the median attempt is 367 ms — and looking at each returning person's best score, the median is 293 ms, more than 100 ms faster than a typical first try. Familiarity with the test — knowing the rhythm, where the screen changes, how your own device feels — is worth more than most people expect. It is also why comparing your first attempt against anyone's posted best is unfair to you by about a tenth of a second.

The full reaction-time distribution

Two ends of this table deserve honesty. The “fastest 1%” at 140 ms is at the edge of what a human can do in response to a visual stimulus — some of those are genuine, and some are lucky anticipations that happened to land inside our plausibility bounds. And the 99th percentile at 1,657 ms is not slow reflexes; it is someone who looked away mid-test. Real data has both tails, and we would rather show them labeled than trim them quietly.

Click speed, for comparison

The click-speed pool is tighter than folklore suggests. Forum wisdom says “average is 6–7 CPS and pros hit 15+”; our median is indeed 6.7, but only 1% of real attempts exceed 12.7 CPS. Scores far above that on other sites usually involve techniques like drag clicking — or autoclickers, which our submission pipeline is built to reject.

Every test at a glance

Median and top-10% score for every test with at least 100 submissions as of September 3, 2026. Three tests (Spatial Reasoning, Color Blind, Paired Memory) have not crossed that bar yet and are excluded rather than shown with unstable numbers. Accuracy-scored tests (Attention Span, Task Switching, Stroop, Peripheral Vision) cluster near the ceiling by design — most people who finish them score very high, and the interesting signal is in speed and consistency rather than the percentage itself.

A typing median of 46 WPM is worth a note: the commonly-quoted “average typing speed is 40 WPM” dates to studies of general office workers. People who voluntarily take a typing test skew faster — self-selection is part of every number on this page.

Where these numbers come from

Every score in this analysis was submitted anonymously from the test pages on this site. Submissions are validated server-side against per-test plausibility bounds, rate-limited per person and per network, and issued through single-use sessions so scripts can't stuff the pool. A trust-scoring system additionally weights which rows feed the percentiles shown on test pages. No accounts exist, no names are attached, and anyone can opt out — the full description is in our privacy policy.

What we deliberately do not do: publish a leaderboard. Public leaderboards on tests like these fill up with scripted scores within weeks, and every number above would become garbage. The trade we chose is percentiles over rankings — you find out where you stand in a real distribution, not whether you beat an autoclicker.

The numbers on this page are a snapshot as of September 3, 2026. Each test page shows the live version of its own distribution, refreshed nightly, in its “Live scores on MindBenchmark” table.