Research Guides
Verify Every Responsive Version of a Research Graphic
You corrected the chart, refreshed the article and saw the new numbers. That does not establish that every reader receives the corrected graphic. A responsive page may reference several image files, and an overlooked small variant can preserve an old claim.
This review is about consistency across an image family, not another alternative-text checklist. Use it for research graphics on a site you are authorized to inspect. It describes a manual workflow, not an automated Colquhoun.shop feature.
1. Inventory the family, not just the visible file
Choose one graphic that carries evidence, such as a comparison chart or annotated screenshot. Record the page address, surrounding caption, approved source revision and every declared candidate you can identify in its image markup. Keep the fallback file in the record too.
Google's image guidance explains that responsive images can use srcset or picture and recommends retaining a fallback src. It also explains that images can be discovered through the img element's src. Google: Image SEO best practices
Do not assume similarly named files share the same content revision. A filename ending in “small” describes a naming decision, not evidence that its numbers were regenerated with the larger image.
2. Record what the browser selected
For each tested layout, keep the browser, viewport dimensions, pixel density where available and the selected image URL. MDN documents currentSrc as the URL selected by the browser; importantly, that value does not establish that loading succeeded. MDN: HTMLImageElement.currentSrc
Keep three columns separate: declared candidate, selected file and observed result. A file can appear in the markup without being selected in your test. A selected address can still fail to display. Record a successful visible rendering separately from the address observation.
3. Compare the meaning, not only the dimensions
Imagine a fictional research chart whose revised dataset changes June enquiries from 42 to 36. The large image is regenerated, but the smaller file still says 42. Both addresses return successfully. A narrow-screen reader can therefore see a different result from a wide-screen reviewer even though neither file is technically missing.
The acceptance check should compare both images with the approved data and caption. Confirm period labels, units, legend entries and notes as well as the headline number. A smaller crop that removes a necessary qualification is also an editorial problem, even if the remaining values are correct.
4. Treat an update as a family release
Before publishing a correction, identify which candidates need replacement, which page references need updating and who will verify the public result. Preserve the approved source graphic and its data privately so the next reviewer can reproduce the comparison.
After the update, repeat the recorded test cases and inspect every intended candidate directly where permitted. Direct inspection checks the asset; viewing it in the article checks its context. Neither step alone substitutes for the other. If a candidate cannot be inspected, leave it unresolved rather than inferring a pass from its sibling.
5. Keep a reusable asset-family record
- Evidence: approved data revision and explanatory caption.
- Family: fallback URL and other declared candidates.
- Selection: test conditions and observed selected URL.
- Content: values, labels, cropping and qualifications checked.
- Release: replacements, reference changes and owner.
- Verification: public observations and unresolved variants.
This process can reveal inconsistent evidence without promising image indexing, rankings or coverage of every device. Report exactly which assets and layouts were checked. The goal is one supported research claim across the versions readers can encounter, not merely one attractive screenshot.