The working databases behind the Counter-Archive — the source reads, name recoveries, and research leads — published as downloadable files. This is a research snapshot, not a finished archive: it mixes verified findings with working hypotheses, leads, and bounded negatives, and it says which is which.
Current snapshot: 2026-10-11. The 2026-10-09, 2026-10-07 and 2026-09-28 snapshot cards are kept below for comparison.
5,089 rows × 23 columns. One row per source read, finding, or research object: what the source says, why it matters, the citation, hypothesis assessments (H1–H8), and the evidence-confidence rating. The backbone of the project.
9,805 rows × 7 columns. One row per person believed to have been enslaved: name as recorded, age, race descriptor, sex, the source evidence, and working notes including do-not-merge warnings. Blank cells mean the source gave no data — never fill them by inference.
333 rows × 10 columns. The to-search list: names that might belong in the enslaved-names table once verified, with the record that would confirm or kill each lead. The overwhelming majority are unverified — this tab is a finding aid, not a finding.
3,884 rows × 11 columns. The open-question log: every research question and hypothesis with its evidence basis, status, confidence, next test, and the archive that could answer it. Do-not-merge firewalls are load-bearing; critical-fabulation narratives are clearly labeled as interpretation, not fact.
4,039 rows × 23 columns. Previous snapshot, kept for comparison.
5,499 rows × 7 columns. Previous snapshot, kept for comparison.
327 rows × 10 columns. Previous snapshot, kept for comparison.
1,485 rows × 11 columns. Previous snapshot, kept for comparison.
3,446 rows × 23 columns. Previous snapshot, kept for comparison.
5,004 rows × 7 columns. Previous snapshot, kept for comparison.
328 rows × 10 columns. Previous snapshot, kept for comparison.
1,305 rows × 11 columns. Previous snapshot, kept for comparison.
6 files · CC BY 4.0. The same snapshot, preserved off-site: all three redacted CSVs plus the data dictionary, reading guide, and redaction explanation. Citable as Bendickson, Alexis. Counter-Archive Research Databases — Redacted Public Release (Snapshot 2026-09-28).
1,853 rows × 23 columns. One row per source read, finding, or research object: what the source says, why it matters, the citation, hypothesis assessments (H1–H8), and the evidence-confidence rating. The backbone of the project.
1,450 rows × 7 columns. One row per person believed to have been enslaved: name as recorded, age, race descriptor, sex, the source evidence, and working notes including do-not-merge warnings. Blank cells mean the source gave no data — never fill them by inference.
169 rows × 10 columns. The to-search list: names that might belong in the enslaved-names table once verified, with the record that would confirm or kill each lead. The overwhelming majority are unverified — this tab is a finding aid, not a finding.
100 edges × 84 nodes. The friends–associates–neighbors network behind the research: who witnessed whose deeds, stood surety, posted bonds, and where enslaved people's lives crossed white capital. Every edge carries a confidence level — confirmed_primary (read from the manuscript) is citable; lead and unresolved are not. Identity safeguards are built in: same-name people are distinct nodes (Turner Anderson ≠ Turner Moore ≠ Turner Jones; Nelson A. Thompson ≠ Nelson Anderson).
13 search entries + 3 interpretive entries. A record of archival silences: who and what was searched for and not found — ten name negatives across the Freedmen's Bureau Alabama reels, the missing NARA Roll 15, the absent Clift slaveholder, the cold maternal line — each with its bound. A bound is not a conclusion: "no hit in these three reels" never means "never existed." Absence of evidence is not evidence of absence beyond what each entry states.
Download the full data dictionary (Markdown) — every column defined, hypothesis codes, and the complete data-quality caveats. Download the FAN graph schema (Markdown) — node and edge types, the confidence vocabulary, and the do-not-merge discipline. Download the Silence Ledger (Markdown) — the project's negative-evidence log: what was searched, what came back empty, and the bound on each silence.
No row may be cited as fact on its own authority. The verification level of each row is encoded in the table itself — read it before you quote anything:
Thin-evidence convention: where evidence is thin, cells are left blank. A blank cell means "no data" — never "no," never "unknown person."
| # | Column | What it holds |
|---|---|---|
| 1 | Document Title | Working title of the source, finding, or research object. |
| 2 | Document Type | Compiler-assigned label for the kind of thing the row is (e.g. Primary source, Census transcription, Finding Aid / Index Pointer (Lead)). The parenthetical qualifier is part of the meaning. |
| 3 | Source | The archive, publication, collection, or informant the row derives from. |
| 4 | Evidence | What the source says, in the researcher's words or as transcribed. |
| 5 | Why It Is Important | The researcher's assessment of significance. Interpretive — not a source claim. |
| 6 | Connections | Related people, places, records, and project threads. Used for FAN-cluster linkage. |
| 7 | Conclusions | What the researcher concludes. Check Evidence Confidence before citing. |
| 8 | Questions | Open questions the row raises. Research notes, not findings. |
| 9 | Citation | Source citation in a loose APA-like format. |
| 10 | File | Filenames the row draws on — project-library filenames, not hyperlinks. |
| 11–18 | H1–H8 | Hypothesis assessments (see codes below). |
| 19 | Evidence Type | Kind of evidence (e.g. Chancery manuscript — firsthand read of page images). The dash-qualifier carries the verification claim. |
| 20 | Thesis Category | Mixed-use: status-like values (Supports, Verified findings, Unresolved, Lead, Context) or topic labels. Read each value literally. |
| 21 | Notes | Working notes: transcription caveats, do-not-merge warnings, read-level notes. |
| 22 | Evidence Confidence | The row's verification level in the researcher's own words. The most important column for judging whether a row is citable. |
| 23 | Source Fingerprint | Researcher-assigned provenance slug for the source object. |
| Code | Hypothesis | What counts |
|---|---|---|
| H1 | Institutional knowledge / erasure | Institutional notice, memorialization, omission, disturbance, or erasure |
| H2 | Legal-archival erasure | Law, property, probate, bureaucracy, archival naming, archival silences |
| H3 | Black Jones lineage | Identifies, distinguishes, or connects Black Jones individuals and families |
| H4 | Shandy parentage | Bears on Shandy's identity, mother, age, father, or descendants |
| H5 | APJ children hypothesis | Bears on whether Alexander P. Jones fathered children with an enslaved woman |
| H6 | Plantation kinship / economy | Jones-Donnell-Haywood-Perkins kinship, landholding, plantation commerce, transatlantic business |
| H7 | Black agency / post-emancipation | Enslaved or freedpeople's action, family formation, claims, work, politics, education, community life |
| H8 | Patrol / KKK / racial violence | Slave patrols, KKK, racial terror, coercion, organized white violence |
Cell values: Supports (affirmative evidence; strength depends on source quality) · Refutes (contradicts a proposed connection) · Mixed · Context (background, not proof) · Unresolved (bears on the question, doesn't decide it). Blank means not assessed. A few cells carry legacy Yes/TRUE values whose exact meaning is unconfirmed — do not treat them as Supports without checking.
| Column | What it holds |
|---|---|
| First Name / Last Name | As recorded in the source. Most enslaved people are recorded without surnames; blank is normal. 84 rows have no name at all — these are method rows, negative controls, or name fragments; read Evidence/Notes first. |
| Age | As recorded (22, 24-35, under 10). Transcribed as found, not normalized. |
| Race | Mix of source language and researcher annotation. Quoted period terms preserve what the record said; they are not the researcher's language. |
| Sex | M/F, sometimes with a basis note. Never inferred beyond what the value states. |
| Evidence | Compact pipe-separated: finding | record description | citation. |
| Notes | Working notes including do-not-merge warnings and identity caveats. |
Name · Name Variant · Associated Enslaver or Property · Date or Range · Location · Source Lead · Record Needed (what would confirm or kill the lead) · Verification Status · Source Citation or URL · Source-Only Notes.
This release is redacted. The project's working databases name living people — descendants who gave interviews, family correspondents, local historians, and staff who answered research requests. Interview participation is not publication consent, so until each person opts in, they appear here as [REDACTED — living person].
Also removed: one personal email address, five staff email addresses used in research correspondence, and the researcher's own university email address. Public reference-desk addresses (archives, special collections) are kept. Historical figures — including the 1950s Willie Jones family at the center of the land-reclamation research, who appear in public court records and news reporting — are named as the sources name them.
Where a redaction would have removed a historically significant public record, the name is kept: redaction targets living private individuals, not the public record. If you are named in this data and want your name restored or removed, use the contact link below.
Bendickson, Alexis. The Counter-Archive Research Databases, snapshot 2026-10-09 (redacted public release). The Counter-Archive, Huntsville, Alabama. https://www.beyondavalon.app/data.html
