An absent name in the register may mean no casualty was reported. It may equally mean no copy of that day’s paper could be found. The difference is recorded here for every date in the study.
2,192dates in the study
4issues located
18pages held
0pages machine-readable
The database carries one row for every one of the 2,192 days between 1 January 1860 and 31 December 1865, each saying whether the paper published that day and whether a copy has been found. That is deliberate. A database that holds only what was found cannot tell you what it is missing, and a register of casualties that cannot tell you what it is missing is a misleading document.
The Mercury stopped with the Union occupation of Charleston in February 1865 and did not resume until November 1866, after this study ends. The month of the suspension is certain; the date of the last wartime issue is not established here, so every date in February 1865 is left open rather than guessed at. Sundays are marked as days on which no issue is expected, that being the general practice of Charleston dailies, though it has not been confirmed date by date.
Month by month
issue locatedpublished, no copy foundceased during the monthnot publishing
1860
JanFebMarAprMayJunJulAugSepOctNovDec
1861
JanFebMarAprMayJunJulAugSepOctNovDec
1862
JanFebMarAprMayJunJulAugSepOctNovDec
1863
JanFebMarAprMayJunJulAugSepOctNovDec
1864
JanFebMarAprMayJunJulAugSepOctNovDec
1865
JanFebMarAprMayJunJulAugSepOctNovDec
The issues that were found
Legibility is not judged by eye. Each page is put through a real OCR pass and scored on how many of the long words it returns are real words — long words, because garbled text throws off short real words by chance. Clean English prose scores 83 per cent. Anything under 50 per cent is noise carrying the occasional real word.
1860-12-27 Daily
Internet Archive (Lincoln Financial Foundation Collection) · charlestondai18601227char · 10 pages · Complete Issue
Seven days after South Carolina seceded, and by a wide margin the best-preserved wartime Charleston newspaper in public hands. Pages 2-9 are two-page spreads scanned together. A local Tesseract pass recovers 1,200-2,000 long words a page, but only 19-39 per cent of them are real words, against 83 per cent for clean English prose. Real phrases do surface — 'the present attitude of South Carolina', 'FATIGUE MILITARY CAPS' — but the columns splice, so a name on one line may belong to a story in the next column. Readable enough to tell what the paper is about; not readable enough to attach a name to a casualty.
Page
Scan
Resolution
Verdict
Why
1
2887×3743
~144 dpi
Unusable
969 long words recovered, 21% of them real (clean English scores 83%)
2
4047×3032
~101 dpi
Unusable
1837 long words recovered, 36% of them real (clean English scores 83%)
3
4047×3032
~101 dpi
Unusable
1891 long words recovered, 36% of them real (clean English scores 83%)
4
4047×3032
~101 dpi
Unusable
1690 long words recovered, 38% of them real (clean English scores 83%)
5
4047×3032
~101 dpi
Unusable
1917 long words recovered, 39% of them real (clean English scores 83%)
6
4047×3032
~101 dpi
Unusable
1413 long words recovered, 39% of them real (clean English scores 83%)
7
4047×3032
~101 dpi
Unusable
1494 long words recovered, 32% of them real (clean English scores 83%)
8
4047×3032
~101 dpi
Unusable
2039 long words recovered, 31% of them real (clean English scores 83%)
9
4047×3032
~101 dpi
Unusable
1175 long words recovered, 19% of them real (clean English scores 83%)
10
2839×3743
~142 dpi
Unusable
617 long words recovered, 22% of them real (clean English scores 83%)
1862-03-19 Daily
Internet Archive (Lincoln Financial Foundation Collection) · charlestonmer18620319char · 4 pages · Complete Issue
Repository OCR reads straight across four columns and splices unrelated stories into single lines. A local Tesseract 5.5.3 pass at 400 dpi recovers the masthead and essentially nothing else: 3 long words on page 1, 22 on page 3. The cause is resolution, not software.
Page
Scan
Resolution
Verdict
Why
1
2735×3634
~137 dpi
Unusable
only 3 long words recovered — too little to judge or to use; median text line 1 px against the ~30 px a recogniser needs
2
2735×3634
~137 dpi
Unusable
only 2 long words recovered — too little to judge or to use
3
2735×3634
~137 dpi
Unusable
only 22 long words recovered — too little to judge or to use; median text line 1 px against the ~30 px a recogniser needs
4
2735×3634
~137 dpi
Unusable
only 0 long words recovered — too little to judge or to use; median text line 3 px against the ~30 px a recogniser needs
1863-05-14 Tri-Weekly
Internet Archive (Lincoln Financial Foundation Collection) · triweeklymerc18630514char · 2 pages · Complete Issue
The repository's own OCR file is 170 bytes for the entire issue. A local Tesseract pass recovers around 230 long words a page at 17-23 per cent real, which is noise carrying the occasional real word.
Page
Scan
Resolution
Verdict
Why
1
2254×3204
~113 dpi
Unusable
227 long words recovered, 17% of them real (clean English scores 83%)
2
2254×3204
~113 dpi
Unusable
240 long words recovered, 23% of them real (clean English scores 83%)
1864-02-09 Daily
Internet Archive (Lincoln Financial Foundation Collection) · charlestondai18640209char · 2 pages · Complete Issue
Mid-siege. Page 1 yields 21 long words at 10 per cent real; page 2 yields 87 at 25 per cent. Far too little, and far too uncertain, to extract from.
Page
Scan
Resolution
Verdict
Why
1
2499×3410
~125 dpi
Unusable
only 21 long words recovered — too little to judge or to use; median text line 1 px against the ~30 px a recogniser needs
2
2499×3410
~125 dpi
Unusable
87 long words recovered, 25% of them real (clean English scores 83%); median text line 1 px against the ~30 px a recogniser needs
Where the papers were looked for
The search was not confined to the Mercury. Newspapers.com’s public catalogue lists eight titles for Charleston, South Carolina, of which four cover any part of 1860–1865. Each was then looked for in the free archives.
Charleston title
Wartime span
Newspapers.com (subscription)
Free and public
The Charleston Daily Courier
published throughout
84,234 pages, 1803–1873
2 issues on the Internet Archive. Nothing in Chronicling America.
The Charleston Mercury
to February 1865
52,214 pages, 1822–1868
2 issues on the Internet Archive. Nothing in Chronicling America.
The Charleston Tri-Weekly Courier
1861–1864
36 pages in all
Nothing.
The Charleston Daily News
from 1865 only
11,474 pages, 1865–1873
In Chronicling America, but it begins after the city fell.
And in the repositories themselves
Repository
What it holds for 1860–1865
Usable here
Internet Archive
4 issues in all: Mercury 19 Mar 1862 and 14 May 1863; Courier 27 Dec 1860 and 9 Feb 1864 — all Lincoln Financial Foundation Collection, all public domain
All four downloaded and measured. None readable.
Chronicling America (Library of Congress)
Nothing. 0 pages of the Mercury (LCCN sn83045168 and sn84026951) and 0 of the Courier
Checked and settled, 30 August 2026. Nothing to take.
Newspapers.com
Both papers in full, by its own catalogue
Subscription. Catalogue read; nothing signed into.
Accessible Archives
Mercury 1861–1865 in the Civil War Newspapers Collection
Subscription. Not opened.
HathiTrust
No serial record returned
Nothing.
This machine
No run of either paper
Not a source for casualties.
The Library of Congress question was settled on 30 August 2026 and the answer is a clean negative. Chronicling America holds 7,793 digitised pages of Charleston newspapers and not one of them is the Mercury or the Courier. The same query returns 383 pages for a control title, so the nought is a real nought and not a broken search.
Worse for the wider project, Chronicling America has no Charleston paper of any name covering 1860 to 1864. The earliest wartime-window titles it holds — the South Carolina Leader and the Charleston Daily News — both begin in 1865, after the city fell. For the years this register is about, the free route is closed, and not only for the Mercury.
That leaves the subscription archives and the microfilm. Which is not a technical decision: it needs an account, and somebody to read what that account’s terms permit by way of systematic download. Nothing was signed into, and no paywall, access control or bot check was worked around in building this.