Research Guides
Sort a Comparison Without Separating Claims from Sources
A comparison table can look convincing while quietly attaching the wrong citation to a result after sorting. Reviewers should test the relationship between a row's label, values, caveats and sources, not only whether the largest number appears first.
This is a focused editorial and functional QA workflow for research tables. It does not certify accessibility, validate an entire dataset or describe an automated Colquhoun.shop feature.
1. Build a small deliberate fixture
Use an isolated test copy with invented rows. Give each row a stable identifier, a name, a numeric value, an observation date, a qualification and a distinct source label. Include values such as 2, 10 and 100, two equal values, a zero and a missing value. Keep the real content unchanged while testing.
Write the expected result first. For an ascending numeric column, 2 belongs before 10. Decide explicitly where missing values belong; do not silently turn an unknown into zero. When values tie, specify a deterministic secondary order or explain that order is not meaningful.
2. Track whole rows through each change
Before sorting, record the row identifier with its value, date, qualification and citation. Sort ascending, descending and then by another column. Compare the same identifiers again. A changed position is expected; a changed association is not.
For example, invented row R-02 might contain 10 observed pages, a note that only public pages were counted and source B. If the number moves but the note or source stays in the old position, the interface now tells a different story. Capture the sequence of actions rather than only the final screenshot.
3. Check what the control communicates
W3C's sortable-table example uses native table structure and buttons in sortable headers. Its aria-sort state identifies the currently sorted column and direction; the example moves that state when the sorted column changes. W3C warns that this is illustrative code, not production-ready code, and that assistive-technology testing is essential. W3C: Sortable Table Example
Check the actual implementation rather than copying attributes without behavior. Can a reader tell which columns sort, what the present ordering is and where keyboard focus went after activation? Document the browser and assistive technology combinations actually checked. A visual arrow alone does not establish that these checks passed.
4. Test the boundaries around sorting
Apply a filter, sort its results and clear it. If pagination exists, establish whether sorting applies to the whole result set or only the visible page, then check that the interface explains the scope. Inspect an exported file separately: it may deliberately use a different order, but the row associations must remain intact.
Do not infer that the first row is the best provider. Ordering by price, count or date does not settle suitability. For the underlying meaning of each cell, pair this test with the comparison-table evidence checklist.
5. Save a reproducible acceptance record
- Fixture: row IDs, values, missing-value rule and tie rule.
- Actions: exact sort, filter and pagination sequence.
- Expected: order and unchanged value-note-source associations.
- Observed: result, focus behavior and communicated sort state.
- Coverage: browser, viewport, assistive technology and export checks performed.
- Retest: affected component, repair and repeated sequence.
Keep an inconclusive check distinct from a pass. The useful outcome is a comparison whose evidence survives interaction, not merely a table that rearranges itself neatly.