Build a Source Log Another Reader Can Follow

A source log is a running list, one entry per document you consult, where each entry records what the document is, where you found it, and when you looked. If a claim about a soldier sits in your notes without an entry like that, the claim is a memory, not a finding. The whole point is that a stranger, or you in three years, can pick up the file and land on the same page you did.

What fields does every source log entry need?

At minimum, seven. Write them once at the top of your log as column headers so you never have to invent them mid-search.

1. Soldier name as written. Copy it exactly, including misspellings. Spelling variants belong in the log, not silently corrected.
2. Document type. Muster roll, compiled service record card, pension file, regimental history, index entry, cemetery list, or something else.
3. Repository and collection. The archive, the record group, the microfilm publication number, or the database name.
4. Identifier. The box, folder, page, card number, image number, or URL path that puts a reader back on the exact item.
5. Date of the document. Not the date you found it. The date on the record itself, or its stated period.
6. Date accessed. The day you read it. This matters most for anything online, because pages get revised or disappear.
7. What it states, in one sentence. One neutral sentence of what the document actually says, not what you think it means.

That last field is where most logs quietly fail. "Pension file proves he was at Shiloh" is an inference wearing a fact's coat. "Pension deposition states he was at Shiloh" is what the document says. Keep the two apart, and every later reader knows exactly where your evidence ends and your argument starts. If you want a fuller treatment of that split, see Separating Document Facts from Inferences.

Why log the date you accessed a document?

Because the same nominal source can change under you. The National Park Service's Civil War Soldiers and Sailors System explains that its soldier data was built from the General Index Cards of the Compiled Military Service Records held at the National Archives, and that the project has grown in phases over the years, with records added and features expanded over time. A database that gains records is a database whose search results on a given day cannot be assumed for another day. Logging the access date is the cheap insurance that says: this is what the system returned when I checked.

Should I log the document date separately from the access date?

Yes, and never merge them. A muster roll from 1863 and your Tuesday afternoon search are different facts about the world. The document date tells a reader what moment in the soldier's service the record describes. The access date tells a reader which version of a catalog, index, or website you were reading. Mixing the two produces entries that look precise and mean nothing.

One practical wrinkle: the same name often appears in more than one regiment, under more than one spelling. The Park Service notes that roughly 3.5 million soldiers actually served, yet 6.3 million General Index Cards exist, partly because of men serving in more than one regiment or under variant names. Your log should expect duplicates rather than treat them as errors. Two cards for two spellings are two entries.

What belongs in the notes field, and what should stay out?

Use the notes field for short, mechanical observations: the handwriting was faint, the image was cut off at the right margin, the card listed rank in and rank out. Keep out anything you cannot point to on the page. If you find yourself writing a paragraph of interpretation, that paragraph is a separate working note, and it should cite the log entry number rather than absorb it.

The test is simple. Could a skeptical reader open the same item and see the same thing you wrote? If not, rewrite the note until they could, or move the sentence to a clearly labelled speculation file.

If you are tempted to write	Write this instead	Why
"He was at Gettysburg"	"Regimental history places the unit at Gettysburg, July 1863"	Unit presence is not personal presence
"His name was John H. Carter"	"Index card reads 'Jno. H. Carter'; second card reads 'John Carter'"	Preserves the variants a later search needs
"Age 24 at enlistment"	"Enlistment record states age 24; 1860 census states age 22"	Keeps the conflict visible
"Found on Ancestry"	"NARA microfilm publication [number], image [number], accessed [date]"	Points to the record, not the viewing screen

Row two deserves a pause. Abbreviations and spelling variants are not noise to be tidied away; they are search keys. If you normalize them out of your log, the next researcher, including future you, will fail to find the card. There is a fuller discussion of that problem in Name Spelling Variants on Index Cards.

How do I cite an image I found online, rather than in a reading room?

Treat the image as a document and the website as a delivery method. Your entry names the document, the collection it belongs to, the archive that holds the collection, the identifier of the item, and the date you viewed the digital copy. If the site supplies a persistent link, record it whole. If it does not, record the search terms and filters you used, because that is the only route back.

That habit connects directly to Citing a Muster Roll Image You Found Online, and it matters even more when the underlying index is a secondary finding aid. The Compiled Military Service Records are themselves a nineteenth century clerical product: the Park Service describes how, beginning in the 1880s, General Ainsworth's staff indexed these records originally to determine pension eligibility, and that War Department clerks transposed the information by hand to an estimated 140 million cards. The Park Service also notes that the approximately 140 million cards include 5.4 million General Index Cards, each containing a soldier's name, and that these General Index Cards form the basis for the soldier names in its system. Your log should say plainly whether you are reading the card or reading someone's later index of it.

How do I handle a source that contradicts an earlier entry?

Do not overwrite the earlier entry. Add a new one and cross-reference both. Contradiction is data. If two documents disagree on a birth year or a birthplace, that disagreement is a finding about the record-keeping, and it may be the most interesting thing in your file. The mechanics of reconciling those discrepancies get their own treatment in Reconciling Conflicting Ages and Birthplaces.

Resist the urge to resolve the conflict by picking the tidier number. Log it, flag it, move on, and let the next document vote.

When is a source log finished?

Never entirely, but it can be complete enough to hand over. An entry is complete when a reader can move from your sentence to the document without asking you anything. The log is complete enough when it contains every claim you are prepared to make, each attached to at least one entry, and when every unresolved question is visibly marked as unresolved rather than smoothed over.

There is a discipline in stopping. When a gap cannot be closed with the records available to you, say so in the log, in plain language, and date the note. A flagged open question is more useful to a reader than a confident guess, and it costs nothing to record. That practice is the subject of When to Stop and Flag an Open Question.

A source log is not bureaucracy. It is the difference between a stack of interesting pages and research someone else can use. Keep the fields fixed, keep the dates honest, keep the inferences outside the entry, and the file becomes portable. Anyone who follows it should land where you landed.