i

Accessibility statement

WCAG 2.2 AA · AAA where reached

This site publishes a collection about what the law requires and another about every WCAG success criterion. Holding itself to a lower standard of honesty than it applies to other people's interfaces is not available to it. So this page says what was measured, what was found, what has been fixed, and — the part most statements omit — what has never been checked at all.

The target

WCAG 2.2 Level AA across the site. Level AAA is met in specific places rather than claimed generally: the footer meets AAA contrast (7:1) throughout, and the enhanced-contrast criterion is met by the body text on every page. AAA is not claimed for the site as a whole, because several AAA criteria — sign language for media, a reading level below lower secondary — are not appropriate to or not met by a research site of this kind.

What was audited, and when

Two audits were run on 31 July 2026 against the built site, both restricted to Level A and AA, and both measured in the page rather than inferred from the CSS:

  • Keyboard and focus. Found two failures. The serious one: twenty horizontally scrolling table regions that no key could reach — 2.1.1. Every such region is now focusable, given a role and labelled. The last one was found and fixed on 14 August 2026, two weeks after the audit said they had been dealt with, because one page had been missed.
  • Pointer, touch and reflow. Found two failures. The serious one: thirty-four rooms of the building map whose names existed only in a hover tooltip — 1.4.13. The map now carries a native disclosure listing every room by name, operable by tap, keyboard and screen reader with no JavaScript, and Escape dismisses the tooltips without moving the pointer.

Both audits are now partly out of date, and this is the honest part. They were run over fifteen pages. The site is thirty-seven, because the Basement, Patterns and Whole systems were since split into a page per entry, and a contents rail and a citation builder were added. Those additions were built to the same rules and reviewed against them, but they have not been re-audited. Treat the audit findings as covering the site as it stood on 31 July.

What has never been verified

Both audits carry a "not verified" section, and its contents remain unverified:

  • No screen reader has been run against this site — not VoiceOver, not NVDA, not JAWS. The ARIA is written to specification and reasoned about; it has not been listened to.
  • No real touch device. Viewports were emulated. Target sizes are measured arithmetic, not thumbs.
  • No assistive-technology user has tested it. No cognitive-load, dyslexia or low-vision testing with actual readers has taken place.

An audit by one person with a keyboard and a browser is worth publishing and is not worth overstating.

What is deliberately built in

  • A skip link, and one <main> landmark per page.
  • A larger-text control in the header that does not depend on browser zoom.
  • Every interactive control reachable and visibly focused; :focus-visible outlines are never removed.
  • prefers-reduced-motion honoured on every transition.
  • No image carries meaning that is not also in text. There are no photographs at all except the director's portrait, which is decorative and marked as such.
  • The three pieces of JavaScript — the citation builder, the rail's scroll highlighting and the text-size toggle — are enhancements. Without them the pages are complete.

Known limits

  • The building map's slab cells rely on the target-size spacing exception below 1000px rather than clearing 24px on their own measurements.
  • Some collection tables are wide by nature. They scroll in a labelled, focusable region, which is a mitigation and not the same as fitting.
  • The search page depends on JavaScript. Without it, the building map and the floor indexes are the way through the site.

Reporting a barrier

If something here is unusable, that is a defect and it will be recorded rather than argued with. Write to the director and describe what you were trying to do, on which page, with what. A report naming a specific failure is worth more than a general audit, because it comes from someone actually using the thing.

This statement describes the site as of 14 August 2026.