Accessibility
Accessibility statement
What this site conforms to, how that is checked, and — because a claim of accessibility is worth exactly as much as the evidence behind it — what has not been checked at all.
- Conformance target
- WCAG 2.2 Level AA
- Current status
- Partially conformant — self-assessed
- How assessed
- Automated tests in CI, plus manual review
The standard, and where we stand against it
This site aims at WCAG 2.2 Level AA. Every page listed in the sitemap currently passes the automated portion of that standard in both English and Polish, at a 1280px desktop viewport and a 375px phone viewport. That is a real result and a narrow one, and the next section is about how narrow.
What the tests actually prove
Automated checking finds somewhere around a third of WCAG failures. It is good at the mechanical questions — contrast ratios, missing names, broken ARIA, unlabelled controls — and blind to every question that needs a reader. It cannot tell whether a heading describes the section beneath it, whether a link makes sense read out of context, or whether the reveal at the end of an experiment lands as an explanation or as a jump-scare. So a passing suite means the floor is solid. It does not mean the site is pleasant to use with a screen reader, and we are not going to claim it does.
Not yet done
This site has not been tested by disabled users, and it has not had a third-party audit. Both would tell us things the tooling cannot, and both are on the list. Until then the status above says self-assessed, which is the accurate word.
What has been built in
- Every section heading is a real heading element, so the page can be skimmed by heading rather than only by eye.
- Text colours are checked against the surface they sit on. The lowest ratio anywhere on the site is 4.76:1, above the 4.5:1 the standard asks for.
- Everything interactive works from the keyboard, including the reading slips in the blind horoscope test and the sideways-scrolling lineage diagrams.
- The focus outline is never removed — it is drawn in brass, offset from the control so it stays legible against a dark surface.
- Results that appear without moving focus — a drawn card, a hit or a miss, a failed request — are announced to assistive technology rather than only rendered.
- Controls are at least 24 by 24 pixels, and the language switcher marks each language name in that language so it is pronounced correctly.
- Animation is limited to a single pulsing indicator, and it stops entirely when the system asks for reduced motion.
- Nothing depends on colour alone: evidence labels carry words, register states carry text, and the charts label their own lines.
Known problems
- 01The lineage diagrams are wide drawings that scroll sideways. The container is focusable and named, and every claim in the diagram is also written out in the article's labelled claim list — but the drawing itself conveys its structure visually, and the text is the accessible route rather than an equivalent one.
- 02The twelve reading slips in the blind horoscope test are a listbox that moves with Tab rather than the arrow keys most listboxes use. Everything is reachable and operable; it is simply not the convention.
- 03The statistical charts are labelled images with the figures repeated in an adjacent table. A reader using a screen reader gets the numbers, not the shape of the distribution.
How it is built
Static HTML with progressive enhancement. Nothing on this site requires JavaScript to read; the participation flows need it to submit an entry, and say so when it is unavailable. The site is usable at 400% zoom without horizontal scrolling, and respects the text size and spacing your browser is set to.
Tell us what is broken
If something here is unusable, awkward, or merely irritating with the assistive technology you use, write and say so — including which page, what you were trying to do, and what you use. This is the same address as everything else, and the same rule applies: we would rather be told than not know. An accessibility problem is treated as a defect, not a feature request.
We aim to reply within a few days and to say plainly whether a fix is coming, when, or why not.