A source dossier lets you return to a piece of evidence without starting the investigation again. It preserves the identity of the item, how you encountered it, the claims it can support and the questions that remain open. A useful dossier is more than a bookmark and more disciplined than a collection of interesting notes.
This guide builds a completed teaching dossier around the historical print Singapore Scene in Kampong Bahru. It then shows how to revise a mistaken note, organise corroboration and use a reusable form with another source. The central skill is keeping the original item, its digital reproduction, catalogue information and your interpretation distinguishable.
You do not need specialist archival training to begin. You need one clearly bounded question, access to an identifiable source and a way to record what you actually inspected. Unknown information belongs in the dossier when it matters; an empty field should never pressure you into inventing a photographer, date or explanation.
This shared guide belongs to Guided Reading and Inquiry. The companion Reading Historical Photographs and Prints: Singapore’s Kampong Bahru supplies the extended visual reading and image. Here we concentrate on making the evidence record reusable and correctable.
What the dossier must preserve
Imagine finding an attractive historical image, writing a paragraph about it and returning six weeks later. The original link now opens a different image size. Your saved file is called old-singapore.jpg. The caption in your notes says 1890, but you cannot remember whether that was an exposure date, a publication date or a label supplied by a later uploader.
The problem is not simply poor filing. Your interpretation may depend on distinctions you no longer possess. If you describe the image as a direct colour photograph from an exact year, the medium and date have become part of the historical claim. Losing their source makes that claim harder to assess or correct.
A dossier preserves the smallest sufficient record for the question. It should tell another reader which item you mean, which version or reproduction you used, where the relevant evidence appears and how you moved from evidence to conclusion. It should also say what the item cannot establish on its own.
This does not require copying an entire website. A concise record with reliable identifiers, exact locators and clear limitations can be more useful than hundreds of downloaded pages without an explanation of their purpose. Save enough context to prevent distortion, then make the important relationships easy to see.
Define the question before collecting fields
Our teaching question is: “What can this print support about the representation of a Singapore waterfront, and which details require additional evidence?” This keeps the inquiry close to what an image and its catalogue record can provide. It does not ask the image to reveal every resident’s occupation, income, feelings or daily routine.
Different questions would require different dossiers. A study of printing techniques would need more production information. A study of the publisher’s market would need catalogues, distribution records and other editions. A study of the depicted location would need maps and additional views. The same item can enter several investigations without becoming sufficient evidence for all of them.
Write the question near the top of the dossier. It explains why a particular feature, passage or metadata field matters. Without it, every detail can seem equally important, and the record gradually becomes an unsorted biography of the object rather than support for a specific inquiry.
A good question can change as you learn. Record that change. If the investigation shifts from “where was this taken?” to “how was this waterfront presented to viewers?”, the new purpose may alter the evidence you prioritise. The old work remains useful, but its role should be reconsidered.
Identify the exact item
The starting record is the Library of Congress item Singapore Scene in Kampong Bahru, 2017657641. Its indexed MARC publication field gives an approximate 1890–1910 publication range and names Photoglob Company as publisher. The digital image identifier is ppmsca.41482. These identifiers distinguish the item from other Singapore scenes and similarly named reproductions.
The Wikimedia Commons reproduction record describes a colour photochrom print and identifies its Library of Congress source. It also displays a simplified 1890 date. That simplified display should not silently replace the qualified publication range or establish the date of the underlying photographic exposure.
Treat the item identifier as an anchor for research, not as a substitute for reading the record. An identifier can help distinguish two entries while saying little about their completeness or accuracy. If a source has no formal identifier, record a combination of title, creator, repository, collection and location sufficient to find it again.
Avoid relying on a local filename alone. A descriptive filename is useful for your own work, but other people may have renamed the same image. Keep the source’s identifier alongside the filename so that the relationship survives moving the file to another folder or sharing a copy.
Separate the object from the reproduction
The historical print and the digital file are related objects with different properties. The print has a physical medium, dimensions and publication history. The file has a pixel size, file format and digital history. A crop or resized copy introduces another version that may show less than the full reproduction.
Record which object each statement describes. “The source is a colour print” concerns the historical item. “I inspected a JPEG” concerns your access copy. “The right edge is absent” might concern a crop rather than the original object. A dossier that merges these statements can make a reproduction choice look like historical evidence.
For a visual source, keep a note of whether you examined the full frame or only a detail. A small preview may be sufficient to identify broad composition but inadequate for reading a tiny inscription. Enlargement cannot recover detail that the chosen file does not contain, and a sharpened image should not be mistaken for new evidence.
The same distinction applies to text. A scanned page, an optical character recognition transcript and a modern edition are different routes to a document. Each can be useful, but a transcription error or editorial change should not be attributed automatically to the original author.
Record creator roles precisely
A historical source may involve an author, photographer, illustrator, printer, publisher, translator, editor, collector and digitising institution. One named organisation should not fill every role merely because it appears prominently on the record. A publisher credit does not identify the person who made the underlying photograph.
In the teaching dossier, the publisher role is known from the cited record, while the underlying photographer is not established by the material used here. That is a meaningful distinction. It prevents a polished caption from giving a false impression that the whole production chain has been recovered.
Use role-specific fields rather than a single “creator” box when the distinctions affect the question. If the record names only an editor, preserve that limitation. If authorship is attributed rather than certain, retain the qualification. A question mark or phrase such as “attributed to” can carry important information that should survive later summarising.
Also distinguish the custodian from the rights holder and from the creator. An archive can preserve an item it did not produce. A website can display a reproduction without having created the original. Those relationships should be documented separately before you write a credit or a reuse note.
Give each date a job
A source can have a creation date, an event date, a publication date, an acquisition date, a digitisation date and the date on which you consulted it. These are not competing answers to one question. They describe different events in the source’s history.
For a photograph-derived print, the photographic exposure and the publication of the print need not occur on the same date. A modern file timestamp concerns a later digital process. If you place all three on one timeline without labels, the source can appear to have been created repeatedly or to contain impossible chronology.
Preserve the precision of the evidence. A year range remains a range. “About” remains approximate. A month does not become the first day of that month for narrative convenience. Some software requires a single sorting value; if you use a technical placeholder internally, it must not become the public historical date.
You can often order events without knowing their exact dates. Publication must follow the production steps required to make that published object, but this does not tell you the length of the interval. A useful dossier records what is ordered, what is dated and what remains only broadly bounded.
Locate the evidence within the source
A URL takes someone to a source, but a locator takes them to the evidence relevant to a claim. For a book, this might be a page and paragraph. For a recording, it might be a timestamp and speaker. For an image, use a clear positional description tied to the full frame.
In the waterfront print, a note about foreground figures should identify the foreground area rather than simply saying “in the image.” A note about distant vessels should distinguish them from small craft nearer the shore. You can make those descriptions useful without asserting an exact coordinate or a measurement that you have not established.
The locator should remain intelligible if a preview is resized. “Upper-right area of the full image” is often more portable than “at pixel 842,” unless your task genuinely depends on a particular file and its dimensions. If a coordinate is necessary, name that file version and explain the reference system.
For catalogue information, name the field or section. “Publication statement in the MARC record” is a better locator than “the website says.” This helps distinguish a repository’s structured metadata from a later explanatory caption or a reader’s own interpretation.
Separate observations from interpretations
An observation describes something inspectable in the chosen source. An interpretation explains what it might mean. Both belong in a dossier, but they should not be written as if they carry the same kind of support. A visible boat can be recorded; its owner’s destination may remain unknown.
One useful entry has four parts: the feature, its location, the possible interpretation and the limit. For example: “Small craft appear near the waterfront; this may support discussion of the water’s role in movement; the image alone does not identify particular journeys or frequency of use.” The limit prevents a modest visual clue from becoming a complete economic history.
A strong interpretation also considers an alternative. A quiet-looking scene might reflect the selected viewpoint, moment or presentation rather than the absence of activity in the wider area. The dossier should explain which additional source could help distinguish those possibilities, instead of adding “perhaps” to every sentence and leaving the investigation directionless.
Use How Research Methods and Source Evaluation Work when the difficulty concerns evidence quality or inference. The dossier is the record that keeps those judgements attached to the exact material you used.
Record reuse information without enlarging it
The Commons file page used here reports no known publication restrictions and displays public-domain markings. Record that statement with its source and consultation date. It describes the information attached to this reproduction; it should not be rewritten as a blanket guarantee about every image on the same website or every possible use.
Access and reuse answer different questions. Being able to view a scan does not necessarily establish permission to republish it. A catalogue description, an image and a modern commentary may also have different terms. Keep those distinctions visible when deciding whether to quote, reproduce or link.
For this dossier, the image is available through the linked visual-reading companion and reproduction record. A reader preparing a different publication should inspect the relevant item and intended use again. The record should support that decision by preserving the exact rights statement, not by replacing it with an undocumented “free to use” label.
If rights information is missing or conflicting, keep the uncertainty and choose an appropriate next step, such as linking to the record while seeking clarification. A completed research note can honestly contain an unresolved reuse field. Completion means the current state is accurately documented, not that every question has a convenient answer.
Trace repeated descriptions to their origin
Finding the same caption on three websites does not necessarily give you three independent confirmations. The pages may copy a repository record, mirror one another or draw from the same upload. The repeated statement can be useful for discovery while adding little independent evidence for its accuracy.
In the present case, the Commons page identifies its Library of Congress source. Treat the repository record and the reproduction description as a related source family. Do not count their shared wording as two independent historical witnesses to the scene or its date.
A practical lineage note can be one sentence: “This reproduction and its description derive from the repository item linked above.” If another source contributes something genuinely different, state what it adds. A separately dated map might help locate a shoreline; a copied image caption would not serve the same purpose.
The distinction is especially valuable when an error spreads. If five later pages repeat one mistaken date, correcting the earliest accessible source may explain the whole pattern. Without lineage, the number of repetitions can make the error look more authoritative than a single qualified catalogue entry.
A completed identification record
The table below is a completed teaching record for the selected item. “Not established” is an intentional value. It tells the next reader where further evidence is needed and prevents unknowns from disappearing during reuse.
| Field | Teaching dossier entry |
|---|---|
| Research question | What can this print support about a represented Singapore waterfront? |
| Item | Singapore Scene in Kampong Bahru |
| Repository and identity | Library of Congress; item 2017657641; digital ID ppmsca.41482 |
| Medium | Colour photochrom print; digital reproduction used for viewing |
| Publisher | Photoglob Company |
| Photographer | Not established in the evidence used here |
| Publication date | Approximately 1890–1910; qualified catalogue range |
| Exposure date | Not established; distinct from publication |
| Access copy | Commons JPEG reproduction linked above; full-frame companion image |
| Date conflict | Simplified reproduction label retained as a discrepancy, not adopted as an exact exposure year |
| Reuse note | See the current file-page statement and markings; consultation 7 September 2026 |
| Dossier state | Teaching record, version 1; outstanding fields explicit |
The identification fields draw on the Library of Congress record and related reproduction record. The exact publication range was available in the repository’s indexed MARC field during this review; a full direct reopening of that record was unavailable. That access limit remains part of the teaching record rather than a claim that every primary field was inspected afresh.
A completed claim record
| Claim or question | Evidence and locator | Current judgement |
|---|---|---|
| The scene includes built structures close to water | Full-frame visual relationship between shoreline and structures | Supported as a description of the representation |
| The image can be discussed as a constructed printed view | Medium identified in the source records | Supported; colour should not be treated as an unmediated record of original hues |
| Every resident depended on fishing | No evidence about every resident or occupation in this item | Not established |
| The exact exposure year is known | No exposure date established in this dossier | Not established |
| The scene proves that the wider area was always quiet | One selected representation cannot establish all times and places | Unsupported extension |
| A map might help identify the viewpoint | Requires a suitable map and comparison of stable features | Proposed corroboration, not a completed finding |
The table deliberately distinguishes a supported description, an unsupported extension and a proposed next investigation. A planned map comparison is not evidence that the viewpoint has already been located. Writing it as a task keeps future readers from inheriting a conclusion that no one has demonstrated.
This claim record should travel with the identification record. Separating them into unrelated folders makes it easy to reuse a striking interpretation while losing its limits. The unit of useful evidence is the claim together with the source relationship that makes it assessable.
Plan corroboration around one uncertainty
Suppose the next question is where the image’s viewpoint lies. Begin by listing stable features that might be compared: shoreline shape, major structures, relative positions and distant landforms. Then identify a suitable map or another view with a known location and a relevant date range.
Do not begin by collecting every old Singapore image you can find. A large pile of vaguely related material can make the inquiry harder. Specify which feature would support one location over another and what result would count against your preferred interpretation.
A second source must also be evaluated in its own right. A map can generalise or omit structures. A later photograph can show a changed shoreline. A caption can be copied. Corroboration is not the act of placing two sources beside each other; it is an explained comparison of evidence relevant to the same uncertainty.
Record the outcome in a short note. “Candidate map found, date unsuitable for the current claim” is useful progress. It prevents the same dead end from being repeated and explains why the dossier still leaves the location unresolved. A negative result can be valuable without being turned into positive support.
Correct a note without hiding the change
Imagine that your first working caption said, “A colour photograph taken in 1890 by Photoglob.” After checking the source relationships, you see that the wording merges medium, date and creator roles. Quietly replacing the sentence improves the current caption but loses the reason for the change.
A correction note might read: “Version 2 replaces the earlier description of a direct colour photograph with a photochrom print, separates the qualified publication range from the unknown exposure date, and identifies the named company’s publisher role. The earlier caption should not be reused.” This records the correction without inventing a new attribution.
Keep the change proportional. You do not need a dramatic announcement for every punctuation edit. A correction affecting the source’s identity or a historical claim deserves a visible record in your working dossier. If the mistaken caption has been published elsewhere, identify those affected uses so that the correction can reach them.
Version labels should be simple and meaningful. Record the date, what changed and why. A file called final-final-new2 is less informative than a short change note attached to a stable dossier identity. The aim is to preserve the history of consequential reasoning, not to accumulate ceremonial version numbers.
A reusable blank form
Copy this form into your own notes and fill it for one item. Use “not applicable,” “not found” or “not established” where those meanings are accurate. Do not leave them interchangeable: something can be relevant but still unresolved.
| Field | What to enter |
|---|---|
| Research question | One question the item may help answer |
| Exact item identity | Title, identifier and repository or source owner |
| Collection context | Collection, series, file or surrounding publication |
| Object type | What the original item is |
| Access copy | Scan, photograph, transcript, edition, crop or other reproduction |
| Creator roles | Named roles, with uncertain attributions preserved |
| Dates | Separate event, creation, publication and later processing dates |
| Relevant locator | Page, passage, timestamp, field or image area |
| Observation | What can be inspected directly |
| Interpretation | What you infer and why |
| Alternative | A serious competing explanation |
| Limits | What the item does not establish |
| Source lineage | Where copied descriptions or data originate |
| Reuse information | Exact statement, its location and consultation date |
| Corroboration needed | A specific next source and discriminating comparison |
| Access limitations | What was unavailable or only partly inspected |
| Version note | Date, consequential changes and affected uses |
For a short school task, a paragraph can combine several fields as long as their meanings remain clear. For a larger inquiry, separate tables may be easier to maintain. The form is a guide to preserving relationships, not a requirement that every source must fit an identical administrative shape.
Practice with a fictional letter
Consider this invented teaching packet: a scan shows a letter dated 14 May 1932. A catalogue calls it “Letter from A to B” and assigns the file to a collection donated in 1986. A typed transcript gives 17 May 1932. The letter mentions “the bridge work” but names no town. No rights statement accompanies the scan in the packet.
Write a short dossier answering: “What can this packet establish about the date and location of the bridge work?” Identify the source objects, record the date discrepancy and state which conclusion is too strong. Do not invent an author’s full name or a town because the form contains a field for one.
A good answer distinguishes the handwritten letter, the scan, the catalogue and the transcript. It notes that the donation year concerns collection history. It compares the visible date with the transcript rather than silently choosing the more convenient one. The mention of bridge work does not establish where the work occurred or whether it began on the letter date.
The missing reuse statement is a separate question from the letter’s evidential value. You can describe the packet’s limits in a classroom exercise without claiming permission to republish the scan. The next research action might be to inspect surrounding correspondence for location clues and the repository’s item-specific access and reuse information.
Check common dossier mistakes
A copied caption is treated as a second witness. Repair this by tracing its source and describing the relationship. Keep it as a discovery route if useful, but do not increase the count of independent evidence merely because the wording appears on another website.
A file timestamp becomes the historical date. Repair this by assigning each date to the event it actually records. If the historical creation date is unknown, retain that state. The existence of a precise modern timestamp does not reduce historical uncertainty.
An interpretation is stored as an observation. Repair this by separating the visible feature from the explanation. “Three figures are visible” and “the community was prosperous” require different evidence. A confident adjective cannot bridge that gap on its own.
A complete-looking form hides access limits. Repair this by stating what you could inspect and which fields came from a secondary description or indexed record. Another reader can then decide whether the evidence is sufficient for their intended use and where a stronger check is needed.
Change the source and preserve the method
Apply the form to a recorded interview. The exact item now includes the recording identity, participant roles, interview date and relevant timestamp. A transcript is a useful access copy, but a claim about tone may require listening to the recording. The interview date may be decades later than the event being remembered.
Apply it to a scientific chart. The important identity may include a dataset version, variable definition, units and reference period. A screenshot preserves appearance but may omit downloadable values or methodological notes. A claim about a trend needs the time range and comparable measures, not just the chart title.
The fields change in emphasis, while the central relationships remain: item, access copy, relevant evidence, interpretation and limit. Successful transfer means adapting the record to the new source type. It does not mean forcing an audio recording into a photograph’s date and creator categories.
For an example using numerical reports, continue to Source Comparison: NASA and NOAA’s 2023 Temperature Reports. Notice how baselines and publication periods perform the same protective role that medium and date qualifications perform in the print dossier.
Return later and test whether the dossier works
Put your source aside for a few days, then try to reconstruct one claim using only the dossier. Can you find the item? Can you locate the relevant evidence? Can you tell which part was observed and which part was inferred? Can you explain why an unresolved field remains unresolved?
If you must rely on memory for a crucial detail, improve the record at that point. If the dossier contains so much copied material that the relevant evidence is difficult to find, reduce or reorganise it. A dossier succeeds when it supports an accurate return, not when it reaches an impressive length.
Use How Archives Catalogues and Finding Aids Work to understand collection context, and How Citations, References and Scholarly Linking Work to make the final claim traceable. Keep the dossier as the working bridge between finding a source and using it responsibly.