Methodology

Background: Zener-style cards

Zener cards (circle, cross, waves, square, star) were used in mid‑20th‑century parapsychology research, notably by J. B. Rhine and colleagues at Duke University, to test claims of extrasensory perception under controlled guessing conditions. With five equally likely symbols, random guessing yields a 20% expected hit rate per trial.

This site uses the same five-symbol vocabulary for familiarity and simplicity. It is a modern, browser-based homage — not a reproduction of Rhine’s laboratory protocol. Useful background reading:

• Zener cards (Wikipedia)
• J. B. Rhine (Wikipedia)
• Parapsychology — historical context and mainstream skepticism (Wikipedia)

Trial procedure

1. When you tap Start tests, the site creates an anonymous session ID (and an internal resume secret for private continue links).
2. On each of 10 tests in a block, the server-side target symbol is chosen uniformly at random from the five Zener symbols using crypto.getRandomValues in your browser before your pick is recorded.
3. You select one of five face-up cards. Correctness is determined immediately.
4. Independently, the site flips a fair coin to decide whether you will see feedback (correct/incorrect reveal) or enter a short blind pause with no reveal.
5. Each trial row — target, pick, correctness, feedback visibility — is written to the database when storage is configured.

Blocks finalize every 10 logged picks. You may append more blocks later via a private continue URL; lifetime metrics always use every pick in that session.

Primary comparison: shown vs hidden feedback

The core within-session comparison splits your tests into two arms:

• Shown-feedback tests — you saw whether you were right or wrong before the next pick.
• Hidden-feedback tests — you did not see the outcome; you only advanced after a brief pause.

We report accuracy in each arm and the difference (shown minus hidden). If seeing results somehow correlated with better guessing — beyond what blind rounds show — that gap is what the home-page narrative explores. This is a descriptive split, not a claim that feedback “causes” precognition.

Why 20% is the chance baseline

Each trial draws one target uniformly from five symbols. If you guessed at random with no information, you would expect about one correct answer in five — 20%. Session reports compare your shown-feedback accuracy to that baseline and compute an exact one-sided binomial tail probability: the probability of getting your observed number of correct answers (or more) on shown-feedback tests alone, assuming independent 20% trials.

Important scope notes:

• The p-value is computed only on the shown-feedback subset, not on hidden rounds or the full session total.
• With only a handful of shown-feedback tests in a single 10-test block, the p-value is a coarse read — see the hints on your results panel.
• We do not apply multiple-comparison correction across sessions or participants in the live UI; treat exploratory hits as hypothesis-generating, not definitive proof.

Sample size and the 1,000,000-response milestone

A single 10-test block randomizes feedback visibility roughly fifty-fifty, so each arm often contains only about five tests. That is enough for personal curiosity, not for stable population inference.

Public, experiment-wide conclusions stay embargoed until 1,000,000 logged responses accumulate in the shared dataset. The milestone bar on the home and results pages tracks progress toward that threshold. The intent is to avoid over-interpreting noise from early, small samples while still letting individuals explore their own data immediately.

Limitations and honesty checklist

• Self-selected participants — visitors opt in; there is no random population sampling.
• Browser-only environment — distractions, tab switching, and device differences are uncontrolled.
• No preregistration — analysis choices are documented here but were not locked in a formal registry before launch.
• Resume sessions — motivated users can log many blocks; leaderboard ranking uses cumulative session stats, which may differ from strict independent-subject designs.
• Not medical or diagnostic — do not use this site for health, safety, or financial decisions.

We publish methods openly so you can judge for yourself. If you spot a bug in logging or metrics, please contact us.

Ready to try it?

Head to the home page to log your first block, or read about the project for publisher background and disclaimers.