Research Guides
Keep a Methodology Panel’s Exposed State in Sync
A methodology panel opens visibly, but its control still reports that the panel is collapsed. A researcher using assistive technology can receive a different account of the page from someone looking at the screen. Review the exposed state through the interaction, not just the initial label.
This guide offers a state-consistency test for website research interfaces. It does not claim that Colquhoun.AI implements a particular disclosure widget or that one passing check certifies a whole website.
1. Define the state the control represents
W3C's Name, Role, Value guidance explains that interface components need programmatically available names and roles, with user-settable states and values exposed appropriately and changes available to assistive technologies. It specifically identifies expanded and collapsed states among the information that matters. Standard HTML controls used according to specification can supply these semantics; custom components require appropriate implementation.
For an audit, first identify the intended control pattern. A link to a separate methodology page is not the same interaction as a button that reveals a panel. Do not add state attributes simply because another component uses them.
2. Follow a fictional methodology disclosure
Imagine a fictional comparison page with a “Methodology” button. Initially, the associated panel is hidden and the control exposes a collapsed state. Activating it reveals the sampling notes, but the developer updates only a CSS class. The exposed state remains collapsed.
The finding is a disagreement between visible content and programmatic state. Changing the button text to “Methodology open” might alter its name, but that alone does not demonstrate that the component's appropriate state has been corrected.
The repair should make the rendered panel and the exposed state derive from the same intended interaction state. The reviewer then tests opening and closing, rather than accepting a single screenshot of the open panel as proof.
3. Check a complete state sequence
- Initial load: record whether the panel is visible and what state the control exposes.
- Open: activate the control and compare visible content with the programmatic state.
- Close: activate again and verify both representations change back.
- Repeat: confirm successive cycles do not leave stale values.
- Alternative path: where another action closes the panel, test that path too.
If the application restores a saved open panel after navigation, include that restored state in the test. A component can work after a direct click yet initialize incorrectly when the interface reconstructs a previous view.
4. Keep inspection and assistive-technology evidence distinct
Inspect the relevant accessibility information in a suitable browser tool, then test the interaction with the assistive technology and browser combination available to the reviewer. Record that combination explicitly. A source-code attribute is evidence about markup; it is not by itself evidence of the actual experience in every screen reader.
Use harmless sample methodology text and avoid placing private research notes in shared screenshots or logs. Capture the states necessary to reproduce the mismatch, not a complete customer workspace.
Also verify that the displayed sampling notes are the intended content. Correctly announcing an expanded panel does not establish that its research methods are accurate, complete or associated with the current comparison.
5. Write a state-specific acceptance note
A useful ticket names the control, its associated panel, the failing transition and the observed mismatch. Its completion note lists the tested open, close and restoration paths, along with any untested browser or assistive-technology coverage.
Keep related findings separate. Keyboard activation, focus placement, visible contrast and programmatic state each need their own evidence. This focused review helps ensure that a research interface does not describe one state to a reader while displaying another.