Research Guides
Verify a Research Filter When Forced Colors Remove Its Shadow
A filter's selected state can disappear even when its text remains perfectly readable. If a research interface uses a shadow as its only selected marker, the browser may remove that shadow in forced colors mode. Check whether readers can still tell which evidence the filter includes.
This is a worksheet for a fictional research interface, not a finding about Colquhoun.shop or a complete accessibility audit.
1. Test the mode, not a similar-looking theme
The W3C CSS Color Adjustment specification describes forced colors as a mode in which the browser uses a constrained palette chosen by the user. Under the default adjustment, effects including box and text shadows are removed. A site's own dark theme is a different condition and cannot substitute for this test.
Record how the mode was enabled. Browser emulation is useful for an initial pass; label it as emulation rather than claiming a physical device or operating-system configuration was tested.
2. Start with one distinction the reader needs
Imagine a website audit dashboard with “All sources” and “Verified sources.” The latter is selected, but the buttons look identical except for an inset shadow on the active one. The selected filter genuinely narrows the dataset and changes the report's figures.
With forced colors enabled, the shadow disappears. The reviewer can still read both labels and press both controls, but cannot tell which filter produced the numbers. Record that as a lost state cue, not a broken filtering operation.
For the fictional repair, expose the selection through the appropriate control semantics and add a visible cue that survives the mode, such as explicit selected text and an outline that renders with the user's palette. Check the unselected state too. A persistent border on every option does not identify the active one.
3. Test three things separately
- Focus: where the keyboard will act next.
- Selection: which filter is currently applied.
- Result: whether the displayed rows and totals reflect that filter.
Move focus to the inactive option without activating it. The reviewer should still distinguish the focused control from the selected filter. Then activate it and verify the evidence changes to the expected set. A border that appears only while a button has focus is not a persistent indication of selection.
4. Keep a comparison record for each sampled view
- Setup: page, browser, viewport and forced colors configuration.
- Baseline: current selection and the visual cue in the default mode.
- Forced colors: cue retained, removed or replaced, and exact reproduction steps.
- State change: exposed selection, observed focus and resulting data.
- Retest: behavior after the repair and any untested combinations.
Prefer a local repair that cooperates with the user's palette. The specification describes an opt-out for cases where the author supplies appropriate adjustments, but disabling forced colors throughout a dashboard to keep its styling can defeat the user's preference. Do not use brand consistency as a substitute for an operable state cue.
5. Report the verified task, not a blanket pass
A useful finding reads: “In the tested forced-colors setup, the selected marker stayed visible when focus moved, and switching filters updated the report to the expected records. The result was checked in one browser configuration.” That establishes whether the reader could identify and change the evidence scope in the sampled view. It does not demonstrate conformance across all assistive technologies or pages.