Find a Source Again: Search Terms, Item Identifiers and Broken-Link Recovery
eduKateSingapore · Guided Reading and Inquiry
Choose your reading route
1. Decide what must remain the same
2. Build an identity packet from what you already have
3. Search the institution before searching the whole web
4. Recover the Kampong Bahru record by identity
5. Recover the correct Bedroom painting
6. Compare candidates with a small decision record
7. Handle broken, blocked and restricted access differently
8. Recover the claim, not merely the page
9. Practise with three recovery problems
10. Write a recovery receipt that can be trusted
Source routes and the next inquiry
A broken link does not necessarily mean a source has disappeared. Its institution may have changed the catalogue, moved a record, replaced an image service or retired an old interface. The practical job is to recover the same work, item or version you used, then establish what you can still inspect. A visually similar image or a matching title may be helpful, but it must not silently replace the original source.
This guide teaches a bounded recovery method: reconstruct the identity, search the responsible institution, compare candidate records, follow the record to the relevant material, and record what was recovered. It uses two real identity examples already represented in the eduKate learning collection: the Library of Congress record for Singapore Scene in Kampong Bahru and the Art Institute of Chicago's version of Vincent van Gogh's The Bedroom. Practice failures and invented candidate records are clearly labelled as exercises.
Access matters. During preparation on 30 September 2026, direct text-fetch attempts returned access errors. The Library of Congress item record, its rights panel and its linked image viewer subsequently opened in the research browser. The Art Institute of Chicago's official public API supplied the painting's record and public-domain indicator, while its ordinary artwork page presented a security-verification screen. This guide distinguishes those successful recovery routes from the material that remained uninspected.
1. Decide what must remain the same
Begin by asking what you are trying to recover. A work can exist in several editions, translations, reproductions or digital files. A book's argument may be unchanged across two printings while its page numbers differ. A painting may be represented by several photographs. A historical photograph may survive as a print, a copy negative and multiple scans. Your task determines which distinctions matter.
If you cited a sentence on a numbered page, recovering the title alone is insufficient. You need the edition or a transparent explanation of how the passage corresponds in another edition. If you analysed a cropped image, a full image of the same object may help, but you should record that the view changed. If the claim concerned the crop itself, substituting the full image would change the evidence under discussion.
Write a short identity statement before searching: “I need the catalogue record and digital reproduction associated with this institution's item identifier.” Or: “I need the 1889 Chicago painting, not either of the other Bedroom paintings.” This statement prevents your search from drifting towards the most convenient result. It also helps another researcher understand why a plausible candidate was rejected.
Separate essential identity fields from useful clues. An accession number and holding institution can be strong identifying information. An old thumbnail filename may be only a clue. A handwritten title might contain a spelling error. Record all of them, but do not treat them as equally decisive. A robust match combines several compatible fields and explains any disagreement rather than depending entirely on one attractive resemblance.
The recovery target can change, but the change should be deliberate. If the original edition is unavailable and you decide a later edition is adequate, say what claim survives the substitution and what must be rechecked. That is a new research choice, not proof that the original source has been recovered unchanged.
2. Build an identity packet from what you already have
An identity packet is a small set of facts that can travel independently of a URL. Include the title, creator or responsible organisation, holding institution, item identifier, date or date range, format, edition or version, and the precise part you used. Preserve the old URL as one locator among these fields. If you recorded an access date, retain it, while remembering that it describes your earlier consultation rather than the source's creation.
You may already have pieces of this packet in a bibliography, download filename, screenshot caption or research note. Copy them without silently correcting uncertain details. Mark your own additions separately from text transcribed from a source. If a date appears in brackets or includes an approximation, preserve that qualification. Removing it makes the identity look more precise than the underlying description was.
For a picture, distinguish the physical object from the digital representation. The institution's item number may identify the catalogue record, while a reproduction number identifies a particular digital file or reproduction. Both can help you search, but they serve different roles. Record which number came from which field so that you can compare like with like when a new interface labels them differently.
A screenshot can preserve useful evidence about what you saw, but it has limits. It may omit the date, creator, rights panel or surrounding catalogue context. It may also capture an incomplete loading state. Treat it as a retained observation rather than a complete substitute for the record. If the screenshot is your only surviving trace, explain that constraint in the recovery note.
Do not add personal information or credentials to the packet. A source identity rarely requires an account password, session token or private sharing link. If your old URL contains authentication details, use the institution's public record route where available and keep sensitive account material out of a bibliography. Recovering a source should not expose an unrelated person's information or create a new access risk.
3. Search the institution before searching the whole web
Start with the organisation responsible for the record. An institutional catalogue search may understand identifiers, creator names or collection titles that a general search engine handles poorly. Search the exact identifier first when you have one. Then try a distinctive title phrase and a creator or collection name. Change one part of the query at a time so that you can tell which clue helped.
A general search engine can still be useful, especially when an institution's navigation has changed. Restricting a query to the institution's domain can reduce unrelated results. A query containing an identifier, title phrase and institution is often more discriminating than a broad topic search. Nevertheless, the search result is a lead. Open the actual record when access permits and check the fields that matter to your claim.
Avoid treating ranking as evidence of identity. The first result may be a shop reproduction, exhibition page, educational essay or another version of the work. It can contain a familiar image and still be the wrong object. A less prominent catalogue entry with the correct identifier may be the better route. The decisive question is whether the record matches the identity packet, not how confidently the search engine presents it.
Spelling variation can help explain an empty result. Historical place names, transliteration and punctuation may differ between your note and the catalogue. Keep a record of plausible variants, but use them to find candidates rather than to erase distinctions. Two names that resemble each other can still identify different places or people. Return to the full record to resolve the match.
Set a bounded search purpose. You are trying to identify a record or recover a particular passage, not map every mention on the web. After a few targeted routes, review what changed. If none produced a stronger candidate, inspect the missing identity field or ask the repository through an authorised channel. Repeating increasingly vague searches can create a large pile of uncertain results without improving the match.
4. Recover the Kampong Bahru record by identity
The Library of Congress record identifies Singapore Scene in Kampong Bahru with control number 2017657641, in the Photochrom Print Collection, and reproduction number LC-DIG-ppmsca-41482. The verified record route is https://www.loc.gov/item/2017657641/. It names Photoglob Co. as publisher and gives the qualified publication statement [Zürich] : [Photoglob Company], [ca. 1890-1910]. The object is a colour photochrom print, with call number LOT 14192, no. 338 [P&P] and printed number 20203. P.Z. These are complementary identifiers, not interchangeable file names.
Suppose your saved image link no longer works. Begin with the item number rather than guessing a replacement image filename. Search it within the Library of Congress domain or catalogue, then compare the result with the title and collection. If a record points to the same item but presents new download sizes, you may have recovered the catalogue identity even though the file-delivery route changed.
Next separate the record from the reproduction. A small thumbnail, a larger viewing image and a high-resolution file can represent the same catalogued item while differing in detail available for inspection. If your argument concerns a tiny inscription, a thumbnail may be inadequate. If your claim concerns the broad composition, a viewable larger image may be sufficient. State which representation you actually used rather than treating every file as interchangeable evidence.
Do not infer an original photographer or exact exposure date from the title. The Library of Congress's collection guidance explains that catalogue dates describe the creation or publication of the catalogued image and may be approximate. The relationship between a photographic source, a printed object and a later digital reproduction can require separate investigation. Preserve uncertainty instead of turning a recovery success into a new unsupported historical claim.
In this preparation session, the item page opened in the research browser after the initial text-fetch error. Following its View Enlarged Image link opened https://www.loc.gov/resource/ppmsca.41482/, where the digital reproduction was visibly displayed. The item-specific rights advisory states “No known restrictions on publication.” That is the Library's advisory, not a new licence granted by this guide. Its rights panel also explains that researchers must assess applicable restrictions. The recovery therefore reached the record, relevant identity fields, viewing image and rights statement; a high-resolution TIFF download was not tested.
5. Recover the correct Bedroom painting
The title The Bedroom is a useful starting clue but an insufficient identity by itself. The Art Institute of Chicago's institutional exhibition material distinguishes three paintings of the bedroom: the first in Amsterdam, a second in Chicago, and a smaller third in Paris. For this exercise, the target is the Chicago work associated with accession number 1926.417 and artwork record 28560.
The institution's ordinary artwork route is https://www.artic.edu/artworks/28560/the-bedroom. Its public API record at https://api.artic.edu/api/v1/artworks/28560 opened successfully and connects record 28560 with accession 1926.417, Vincent van Gogh and 1889. Its description identifies this as the second version. These compatible fields distinguish the object from a shop product or another version. Search the accession number together with the institution or artist rather than using the bare number alone.
Why add the institution? Identifiers operate within namespaces. A sequence of digits and punctuation can occur in unrelated systems. In a general search, an accession number can coincide with a regulation number, product code or other record. The match becomes meaningful through the responsible collection and compatible descriptive fields. “1926.417” is not globally self-explanatory without context.
Now imagine a candidate image labelled “Bedroom in Arles, 1888, Amsterdam”. It shares the artist, subject and much of the composition with the Chicago target. It is still a different object. Reject it as a replacement for an analysis specifically tied to the Chicago painting, while noting that it may be useful for a comparison if your task expands. Similarity explains why the error is tempting; it does not remove the error.
In this preparation session, the ordinary artwork page presented security verification, so that browser route was not pursued. The successfully retrieved official API record marked the work as public domain. This is item-specific evidence, but the painting's viewing image and download interface were not freshly inspected. Keep that distinction in the recovery note, and do not turn a verified catalogue match into an assertion that every image-use condition or downloadable file was checked.
6. Compare candidates with a small decision record
When several results look plausible, compare them explicitly. Create fields for institution, identifier, title, creator, date, format, edition or version, and relevant content. Mark each as match, mismatch, unknown or not applicable. This is a practical reasoning aid rather than a mathematical scoring system. One decisive mismatch can matter more than several superficial matches.
Consider three invented recovery candidates for a fictional map. Candidate one matches the title and place but has a different edition date. Candidate two matches the item identifier and edition but offers only a partial scan. Candidate three matches the visual appearance but gives no institution or record number. For a claim tied to the original edition, candidate two is the strongest identity match, yet its partial scan may still be inadequate for inspecting the relevant area.
Separate the identity decision from the usability decision. “This is the same item” does not mean “this representation lets me check the claim.” Conversely, a different edition may be usable for a new, clearly revised question. These decisions should appear as two sentences in your notes. Keeping them apart prevents convenient access from deciding identity without your noticing.
Do not resolve unknown fields by copying them from another candidate. If a mirror omits the creator, you cannot simply insert the creator from a similar-looking record and then count that as confirmation. Instead, trace the mirror's stated source or retain the field as unknown. A recovery packet should preserve evidence, not manufacture agreement through editorial tidying.
A useful decision note reads: “Candidate B matches institution, identifier and edition. Only pages one to three are viewable, while the cited diagram is on page five. Identity is strongly supported; the relevant content remains uninspected.” That note is more informative than a single tick beside “found”. It tells the next person exactly what has been achieved and what the unresolved task requires.
7. Handle broken, blocked and restricted access differently
An access failure is an observation, not a complete diagnosis. A missing-page response, a temporary server error, an authentication prompt and a permission restriction imply different next steps. Record what you encountered in ordinary language. Avoid declaring a source deleted, a website down or a collection closed when the evidence only shows that one request failed in your environment.
If an old route is missing, try the institution's current catalogue and search for the identifier. If the site reports a temporary problem, return later or use an authorised alternative interface provided by the same institution. If a record is visible but the item requires a reading-room visit, treat that as an access condition rather than a broken link. You may have recovered the record successfully while still lacking the material.
Respect restrictions. Do not bypass a paywall, authentication requirement, blocked route or security warning to preserve a claim of complete recovery. A repository may offer a lawful public description, a request process or an alternative copy. Follow those routes within your authorisation, and keep the unavailable portion explicit. The research question can often be narrowed while access remains unresolved.
A web archive may preserve an earlier public page, but check what it actually captured. The text may survive while an image, interactive viewer or linked file does not. A saved page may also reflect an older description that was later corrected. If you use it, label it as an archived snapshot and retain its date. Do not present the snapshot as the institution's current statement without a current check.
For a consequential claim, incomplete access may be a reason to stop publication or seek stronger evidence. For a classroom demonstration of source identity, the same limitation can become part of the lesson if it is clearly stated. The appropriate action depends on the role the source plays and the consequence of being wrong, not on a universal rule that every failed link can be ignored or every failed link invalidates the whole project.
8. Recover the claim, not merely the page
Once a candidate opens, return to the claim that led you there. Find the exact passage, image detail, table, map area or record field you relied on. A recovered homepage does not recover the cited evidence. A record describing a collection does not necessarily contain the individual item. The final step is to reconnect the claim with the specific material that supports it.
Read the surrounding context. A sentence may be qualified by the next paragraph, a table may use a denominator explained in a note, and an image caption may distinguish a proposed date from an established date. If the new interface reveals context missing from your original notes, revise the claim accordingly. Recovery is an opportunity to correct an interpretation rather than merely restore a decorative citation.
Compare versions when something has changed. If a catalogue narrows an attribution or revises a date range, record the difference and ask whether your argument depends on it. A small metadata change can have a large effect on a chronology. Conversely, a corrected typographical error may leave the substantive conclusion intact. Explain the consequence instead of treating every alteration as either trivial or catastrophic.
Preserve a useful locator. For a document, that might be page and section; for a map, sheet and grid area; for a catalogue, the named field and item identifier. If a digital viewer uses unstable page numbers, add the printed page number or a distinctive heading where available. The point is to make another inspection possible without requiring the reader to guess what you saw.
Then write the claim again from the recovered evidence. Do not simply copy your old sentence out of habit. Ask whether it is still proportionate, whether the relevant qualification remains present, and whether an access limitation should be stated. A successful recovery may confirm the old account, narrow it, or show that it needs to be withdrawn. All three outcomes improve the integrity of the work.
9. Practise with three recovery problems
Problem one is an invented book case. Your note says “Harbour Stories, second edition, page 42”, but the available catalogue result describes the third edition. What can you claim? You have found a related edition, not yet the cited version. Search the edition details, inspect the relevant passage if available, and explain any substitution. Matching title and author alone does not establish that page 42 contains the same text.
Problem two uses the real painting identity from this guide. A search returns a gift-shop poster of The Bedroom and an institutional page naming the Chicago accession number. Which is the better record route? The institutional identity evidence is the stronger starting point for recovering the object. The shop page may offer a reproduction, but product availability and visual similarity do not replace the catalogue identity or the institution's conditions for image use.
Problem three is an invented historical photograph case. A mirror labels an image “1900”, while the responsible institution gives an approximate publication range spanning several years. What should your note preserve? Retain the institution's qualified date and record the mirror's simplified caption as a discrepancy, not an improvement in precision. Search for evidence if an exact date matters; do not select the narrower date merely because it fits a timeline neatly.
For each problem, write four lines: the intended identity, the candidate found, the unresolved difference, and the next legitimate action. A strong answer may end with “not yet recovered”. That is useful when it names the missing field or access step. A weak answer often ends with “found” while leaving the identity mismatch unexamined.
Now change the conditions. Suppose the edition or exact reproduction is unnecessary because your question concerns only the broad existence of a work. Some differences may no longer be decisive. Explain why the evidential requirement changed, and avoid carrying the looser identity standard back into a claim that depends on a specific passage or visual detail. The recovery method should follow the question rather than become a checklist applied without judgement.
10. Write a recovery receipt that can be trusted
A concise recovery receipt should state the target, earlier locator, search route, candidate record, identity evidence, material actually inspected, unresolved limitations and next action. Date the receipt. These fields need not become a complicated database. A short paragraph can be enough if it makes the distinctions visible and allows another person to continue without repeating the entire investigation.
For the Kampong Bahru example, a proportionate receipt for 30 September 2026 is: “Library of Congress control number 2017657641 and reproduction number LC-DIG-ppmsca-41482 match the target. After an initial text-fetch error, the institutional item page, item-specific rights advisory and linked digital-image viewer opened in the research browser. The displayed reproduction was inspected; the largest downloadable file was not tested.” This records the recovery achieved without claiming exhaustive access testing.
For the Chicago painting, the receipt should say that the official API record confirmed the artwork identifier and accession number, while the ordinary page remained behind security verification. It should separately record the API's public-domain indicator and the lack of fresh image inspection. Do not merge the Amsterdam, Chicago and Paris paintings because they belong to a related series. The receipt preserves the exact object and the actual access route.
Use status words that describe observations. “Identified” means you found a defensible record match. “Inspected” means you saw the relevant material. “Claim rechecked” means you compared the evidence with the statement you intend to make. “Rights checked” means you examined the applicable rights information for the proposed use. These are separate achievements, and none should be inferred automatically from the others.
End with a practical stopping point. If identity and relevant content are recovered and the claim has been rechecked, you can return to the original inquiry. If the object is identified but inaccessible, preserve the packet and choose whether to wait, seek authorised access or narrow the claim. If identity remains uncertain, do not let a tidy bibliography disguise the uncertainty. A clear unresolved record is a better foundation than a confident substitution.
Source routes and the next inquiry
The principal routes are the Library of Congress item page, https://www.loc.gov/item/2017657641/, and the Art Institute of Chicago artwork page, https://www.artic.edu/artworks/28560/the-bedroom. The Library's item page and linked viewer were successfully inspected in the research browser after initial fetch errors. The Art Institute's official API, https://api.artic.edu/api/v1/artworks/28560, supplied a verified alternative record route; the ordinary artwork page remained behind security verification. The successful API read does not imply that the image was inspected.
For collection-specific date and digital-identifier conventions, consult the Library of Congress Photochrom technical information: https://www.loc.gov/static/collections/photochrom-prints/about-this-collection/technical-information.html. The Art Institute's comparison of the three paintings is at https://archive.artic.edu/van-gogh-bedrooms/about-paintings, with its institutional timeline at https://archive.artic.edu/van-gogh-bedrooms/timeline. These sources support the bounded identity distinctions described here, not a claim that every current record field or download condition was inspected.
Continue with the worked primary-source dossier when you need to organise an object's history and evidential limits. Return to Guided Reading and Inquiry when the recovered source is one step within a larger question. The useful habit is simple: preserve identity, inspect the relevant material, recheck the claim, and leave an honest route back.
A final independent check
Use an ordinary source you already have permission to consult, such as a public report or a library book. Before reopening it, write the identity packet from your existing notes. Then hide the old link and try to locate the item through the responsible catalogue or publisher. Do not deliberately damage a record or remove a working reference; this is a rehearsal using your own notes.
After locating a candidate, compare it against your packet. Which fields were sufficient? Which were missing? Did the title lead to several editions? Did an identifier refer to the whole work, a particular volume or one digital file? Record one improvement to your original note that would have made the recovery more reliable. The improvement might be as small as adding an edition statement or a printed page number.
Finally, ask a second reader to use the revised note, or return to it yourself after a delay. Their task is to find the relevant passage, not simply reach the same website. If they stop at the wrong edition or cannot identify the claim's location, revise the locator and repeat the narrow check. This tests the practical quality of your reference without claiming that every source in a larger collection has been audited.
The result is a small, reusable capability: your notes can carry a source's identity across a change in interface. That capability supports reading, writing and correction, because another person can see both how you arrived at the evidence and where your knowledge still ends.