Cognitive accessibility audit (2026-10-22)
Audit of the portfolio (core/web, theme, admin delete flow) against the Likwidmack design system, WCAG 2.2 and the W3C COGA objectives. It sits on top of the existing WCAG 2.2 AA work (contrast roles, focus rings, 44px targets, live regions), which already passes on most criteria.
Method: code and design-system review (design-a11y-neurodivergent Mode A). No rendered pages, screen readers or real users were tested.
What already works
- Text pairs are AA in both themes, with one focus ring everywhere and 44px controls (
control-height,type-min). - Body text is 16px or more,
pline height is 1.5, reading text stops atprose-max(about 70 characters), and headings grow with browser text size. - Status messages are live regions (“Applied: Dark mode”, result counts, “Deleted X.” with Undo).
- Labels say what happens, icons sit beside words, and status always pairs a glyph with a word.
- Reduced motion stops the depth field, tile previews and video autoplay. The site Motion setting (
html[data-motion]) does the same. - Filters live in the URL, Resume exists, and Personalize applies changes instantly and remembers them.
Friction profiles
Ratings: Low = a tired first-time visitor would not notice; Medium = slows people down; High = can block the task or push people away.
Every page except /gallery (depth-field background on)
- Sensory and attention load: High. The default
particlesbackground animates forever behind the text (portfolio-haze-pulse12–13s,portfolio-particle-approach,grid-scroll4.5s, allinfinite), and it moves with the pointer. Only an OS reduced-motion setting or Style Studio’s Motion setting stops it. Personalize can switch to Custom, but nothing is labelled “stop the animation” (WCAG 2.2.2). - Visual crowding: Medium. Every
phas a two-layertext-shadow(_texts.scss), so letter edges blur. Blockquotes andciteare italic. - Layout instability, interaction ambiguity, time pressure: Low.
Case study
- Memory demand: Medium. Resume opens the page at the top, not at the saved position, and no page passes
progressFor, so no percentage shows (design-system open question). - Visual crowding: Low to Medium (the
ptext-shadow again). - Everything else: Low (PageNav, StoryPager, reading progress).
Gallery grid
- Sensory and attention load: Medium. On touch devices with a fast connection, every video tile half in view plays a looping preview without being asked, so several can move at once.
- Interaction ambiguity: Medium. On touch the tile title is hidden; only the tags show (design-system follow-up).
- Everything else: Low.
Gallery full view (feed)
- Sensory and attention load: High. The shown video autoplays, muted and looping, and the bar has Sound but no Pause (WCAG 2.2.2: moving content over 5 seconds needs pause, stop or hide). Only reduced motion gives a Play button.
- Interaction ambiguity: Low. Every gesture has a button and a key.
- Predictability: Medium. Tab can leave the dialog for the browser chrome (no focus trap; known gap).
Personalize and Style Studio
- Interaction ambiguity: Medium. Motion lives only in Style Studio, so the one setting that stops the background is two steps away from the panel people actually open.
- User control: there is no text size, spacing or reading-focus option. The browser settings still work, but the site doesn’t offer its own (Consensus; COGA objective 8).
Admin media delete (DeleteWithUndo)
- Time and pace pressure: Medium. Undo lasts 8 seconds. Check that the timer pauses while the notice has focus or the pointer is over it (WCAG 2.2.1).
WCAG 2.2 cognitive criteria
| SC | Result |
|---|---|
| 1.4.1, 1.4.3, 1.4.4, 1.4.10 | Pass by design (theme roles, rem/clamp, reflow work). Glass marks at 2.4–3.0:1 are a known follow-up. |
| 1.4.12 Text Spacing | Not assessed live; the layout uses rem and wraps, so it is likely a pass. |
| 2.2.1 Timing Adjustable | Check: the Undo window (8 s). |
| 2.2.2 Pause, Stop, Hide | Fail risk: the default depth field and full-view video loop with no direct pause control. |
| 2.3.3 Animation from Interactions (AAA) | Partial: the background follows the pointer unless motion is reduced. |
| 2.4.7, 2.4.11, 2.5.8 | Pass by design. |
| 3.2.3, 3.2.4, 3.2.6 | Pass: the same header, labels and Personalize trigger on every page. |
| 3.3.1–3.3.3, 3.3.7, 3.3.8 | No public forms or logins in scope (Style Studio’s hex field has an inline error). |
Ranked fixes
| # | Fix | Needs | Effort | Impact | Confidence |
|---|---|---|---|---|---|
| 1 | Add a Pause animation toggle (aria-pressed, remembered) to Personalize, and to the footer or header so it is one tap away. It writes the existing Motion preference (reduced) or a new background-paused flag, and stops the depth field’s CSS animations and pointer tilt. |
Attention, sensory | S | High | Standard (2.2.2) |
| 2 | Add a Pause / Play button to the gallery full view’s bar (one label, aria-pressed), remembered for the open view the way Sound is. |
Attention, sensory | S | High | Standard (2.2.2) |
| 3 | Move Motion (System / Reduced / Playful) into Personalize as a third radio group; keep it in Style Studio too, using the same component. | Predictability, attention | S | Med | Consensus (COGA 8) |
| 4 | Remove the text-shadow from running text (p); keep it for display headings if wanted. Make blockquotes and cite upright, with emphasis from weight or a rule. |
Literacy | S | Med | Consensus (BDA style guide); Research for spacing over font shape |
| 5 | Add a Reading group to Personalize: text size (100 / 115 / 130 %), comfortable spacing (WCAG 1.4.12 values: 0.12em letter, 0.16em word, 2em paragraph), and an optional Focus mode that hides the background and side chrome on case studies and blog posts. Off by default and remembered. | Literacy, attention | M | High | Consensus (COGA 8); Research (spacing) |
| 6 | Make Resume scroll to the saved position and pass progressFor, so the percentage shows. |
Memory, executive function | M | Med | Consensus (COGA 6) |
| 7 | Limit touch auto-previews to the one tile nearest the centre of the screen. | Attention | S | Med | Heuristic |
| 8 | Pause the Undo timer while the notice has focus or the pointer is over it, and say so (“Undo is available for 8 seconds”). | Executive function | S | Med | Standard (2.2.1) |
| 9 | Trap focus in the gallery full view (useFocusTrap). This is already a known follow-up. |
Predictability | S | Med | Standard (2.4.3) |
| 10 | Show tile titles on touch, as a line under the tags. | Interaction ambiguity | S | Low | Consensus |
Done (2026-10-22)
Owner’s picks: fixes 1–3 and 7–9, upright blockquotes and cite from fix 4 (the paragraph text-shadow stays), Pause animation reuses the Motion setting, fix 5 goes to the TODO.
- Fix 1: a header Pause animation button (
useMotionPause). Found on the way: the depth field’s CSS animations kept running even under the OS reduced-motion setting (only the pointer tilt stopped). It now holds still whenever motion is reduced. - Fix 2: a Pause button in the gallery full view.
- Fix 3: Motion added to Personalize.
- Fix 7: one touch preview at a time.
- Fix 8: the Undo line names its window. The pause on hover and focus already existed.
- Fix 9: a focus trap in the full view.
Validation
- Friction profile: done for 6 screens and flows.
- WCAG 2.2: rows above; 1.4.12 and the Undo timer not assessed live.
- COGA: objectives 1, 2, 3 and 7 are met by the existing design. Gaps are in 4 (avoid data loss: Resume), 5 (help users focus: animation, autoplay, previews), 6 (don’t rely on memory: Resume position) and 8 (personalisation: motion, reading, focus mode).
- User control: nothing found that overrides OS settings. The Motion setting
playfuldeliberately overrides the OS, and only when the visitor picks it. - Non-diagnostic: no behaviour sensing anywhere; the camera background is opt-in, local and explained.
Scope: cognitive accessibility · non-diagnostic · functional needs: attention, memory, language and literacy, executive function, sensory regulation, predictability, numeracy, motor · validation: friction profile + WCAG 2.2 review of code and the Likwidmack design system, direct (Lead) — not independent · not tested: rendered pages, axe/Lighthouse, screen readers, 400% zoom and text-spacing bookmarklet live, touch devices, real neurodivergent users