Source: https://restorebunny.com/guides/resolve-conflicting-photo-identifications/ # Resolve Conflicting Family Photo Identifications Without a Vote Record competing names, independent evidence and unresolved questions using a claim ledger instead of counting relatives who repeat one story. Family photo projects4 min readUpdated 11 October 2026 ## Try it on your own photo Repair a copy of your photo, then compare faces and details with the original. [Restore my photo]()1 credit per successful photo ยท Account required[See credit packs]() Quick answer ## The short version Record each identification as a claim with its source and reasoning. Ask people independently, compare evidence that can distinguish the candidates, and leave the name unresolved when the evidence cannot decide. Repeated agreement is weaker if everyone learned the same story from one person. ## Start with the disagreement, not a preferred name A family photograph may be labelled 'Aunt Leela' by one relative and 'her sister' by another. Do not overwrite one account as soon as a more confident speaker appears. Record both claims against the same photo ID, with the speaker, date, whether they knew the person and how they reached the answer. Confidence in a memory is useful context but is not a measurement of accuracy. Separate visible observations from interpretations. The Library of Congress analysis tools encourage observation, reflection and questioning. Apply that small distinction here: 'a ring is visible on the left hand' is an observation; 'therefore this is the married sister' is an interpretation requiring additional assumptions. The ledger should expose those assumptions rather than hiding them inside a final caption. ## Ask independently before sharing the alternatives Send an unlabelled reference copy or contact-sheet entry and ask what the person recognises. A leading question such as 'Is this Leela at her wedding?' may produce agreement with your suggested story. Ask about place, companions, clothing and how they know. Keep the original response before showing the competing account, provided the contributor agrees to this use of their comments. Then trace the origin of agreement. Three cousins may all be repeating an identification supplied by the same uncle. Record that relationship instead of treating it as three independent witnesses. Conversely, a dated invitation, another labelled photograph and a first-hand recollection can provide different kinds of support. They still need checking for whether they actually concern the same person and event. ## Worked example: an answer changes when the date range narrows In an editorial example, Ravi identifies the seated woman as his mother because he remembers the necklace. Nila thinks it is an aunt because the house was sold before his mother moved there. The photo itself shows the house but no clear date. Your ledger lists two candidates, the necklace observation, Nila's house-history account and the open question about the sale date. A dated family letter may later show that both sisters visited the house before the sale. That weakens Nila's exclusion argument without proving Ravi's identification. The correct result may still be 'possibly Ravi's mother; aunt remains an alternative'. This example illustrates reasoning, not an actual family case or a claim that jewellery uniquely identifies people. ## Use a claim ledger with six fields - Candidate: write the proposed name or event and the exact person position in the image. 'Woman second from left' is less ambiguous than 'the woman'. - Source: speaker or document, date recorded, and whether the account is first-hand, remembered from someone else or copied from a caption. - Reason: record the clue and the assumptions connecting it to the candidate. Keep a short direct quotation only when you have permission and an accurate record. - Contradiction: list evidence that might rule the candidate out, with its own source. Do not omit contradictory notes from the research file. - Next check: name a specific document, witness or related image that could distinguish the candidates. Avoid collecting more opinions that repeat the same basis. - Current status: confirmed on stated evidence, provisional, disputed or unresolved. Record who assessed it and when; this is an editorial judgement, not a statistical confidence score. ## Publish a caption the evidence can support For a shared version, choose language appropriate to the current record. 'Identification disputed: Leela or her sister' may be more honest and useful than a confident name. Keep the full ledger in the private archive where later researchers can inspect it. If a relative objects to public discussion, retain the disagreement privately and use a less specific public caption. Do not use an AI-restored face as new identification evidence. A reconstruction may make a face look familiar while altering the very feature being compared. Return to the capture master and other documented photographs. When new evidence changes the assessment, update the caption and keep the earlier claim as history. A correction should improve the record without making the disagreement disappear. ## FAQs Should the majority decide the name? No. Trace the evidence behind the answers. Several people can repeat one mistaken caption. Can face resemblance settle the disagreement? Use resemblance as a research clue. Age, pose, source quality and editing can affect it; seek independent corroboration. Is an unresolved caption a failure? No. A clear disputed or provisional status preserves useful evidence and leaves room for a later correction. ## Sources Preservation and technical guidance reviewed for this article. - Library of Congress: [Primary source analysis tools]() ## Related - [Write captions with explicit uncertainty]() - [Preserve the original interview account]() - [Make an identification contact sheet]() ## Ready to restore your photos? Upload an image, choose your settings, and preview results with a before/after slider. [Restore my photo]()[View Pricing]() RestoreBunny Team Photo workflow editors Guides follow our [editorial policy]() for sourcing, product claims, AI limitations, updates, and corrections.