Before 1870, Your Ancestor Is in Someone Else's Paperwork

You have found the family in the 1870 census, and it is a good page. A whole household under one surname, with ages that roughly agree with what the family has always said, in a county you can find on a map. Every person in it is named, which is the first time that has ever happened to this family in a federal census.

So you go back ten years to the 1860 census, and the household is not there. It is not misspelled, not living next door, not in the neighbouring county. It is not there at all.

This is where an enormous amount of African American family research stops. It sounds like a records problem, as though the documents were lost. It is not. It is a paperwork problem, and once you see how the paperwork was designed you can work round it.

The short answer. Before 1870, your ancestor is rarely recorded in a document about people. They are recorded in documents about property, written by the person who claimed to own them, and those documents carry names far more often than their reputation suggests: probate inventories, estate divisions, deeds of sale, mortgages and tax lists. Three further record groups, created between 1865 and 1874, name people directly and in family groups. All three exist because a government needed to administer something, not because anyone set out to document a family. The Freedman's Savings and Trust Company's depositor registers, filmed by the National Archives as Microfilm Publication M816, record a depositor's parents, brothers, sisters, spouse and children, and in some volumes the name of the former enslaver and the plantation. Virginia's cohabitation registers, created by an act of 27 February 1866, turned a marriage begun in slavery into a dated civil record. The Bureau of Refugees, Freedmen, and Abandoned Lands, established by act of Congress on 3 March 1865, left labour contracts, marriage records and complaint files in Record Group 105. And this research became dramatically easier in 2024, when FamilySearch's Full-Text Search made unindexed deed and probate books searchable by name for the first time.

The 1850 and 1860 census schedules were designed not to name your ancestor

Start with why the wall is there, because the shape of the wall tells you where to go around it.

The 1850 census was the first to enumerate enslaved people on a separate schedule. The Missouri Secretary of State's office, which publishes those schedules county by county through its State Archives, sets out exactly what the schedules carry: the name of the slave owner, how many slaves they owned, and for each enslaved person an age, a colour and a sex, together with whether that person was a fugitive from the state, whether they had been manumitted, and whether they were deaf, dumb, blind, insane or idiotic. Then the sentence that defines the problem: slave names were not listed in 1850.

The 1860 schedule adds a column for the number of slave houses on the property, and it adds one exception that occasionally rewards a look. Enumerators were instructed to record the names of enslaved people aged 100 years or older. The same Missouri page is candid about how well that was followed: only about half of those over 100 were actually named.

So the two censuses immediately before the wall are not failed records of people. They are competent records of property, built to a specification with no field for a name. Read as a census they look useless. Read as an inventory they do one genuinely useful thing: they place a named enslaver in a named county, holding a countable group of people of stated ages and sexes.

The practical move. Take the 1870 household you already have. Note the county, and note the stated age of every adult in it. Then look at the 1860 slave schedule for that county for enslavers holding people of matching ages and sexes, and holding enough of them to contain the family group you can see in 1870. That proves nothing by itself. It gives you a shortlist of candidate enslavers, and the records below are how you test each one. The same discipline of reading a census as the document its enumerator was actually filling in, rather than the document you wish existed, applies across the whole series and is covered in how to read old US census records.

The records that name your ancestor are the ones that treated them as property

This is the uncomfortable fact that reorganises the whole problem, and it is the most useful thing in this article.

An enslaved person was legally property, and American law is exhaustive about property. Property is inventoried when its owner dies. It is divided among heirs. It is sold, and the sale is entered in a deed book because the buyer wants the title on record. It is mortgaged, and the mortgage is recorded for the same reason. It is taxed, and the tax lists survive because counties kept them. Every one of those transactions left a document, and unlike the census schedules, those documents carry names, because the entire purpose of an inventory or a bill of sale is to identify exactly which property is being described.

Four record types to look for, in rough order of how often they name people:

Probate inventories and estate divisions. When an enslaver died, the county court had the estate appraised. Enslaved people appear in those inventories individually, by name and with an appraised value, because a valuation of "seven negroes" is not a valuation anyone could act on. The division of an estate among heirs is often better still, because it has to allocate specific named people to specific named heirs, which means a family group can be split across the page in a way that records who was living together immediately before the split. This is the same class of document, and the same reading problem, as a pre-1858 English will: a court was settling obligations, not describing a family, and what it happens to record about relationships is a by-product.

Deeds of sale and mortgages. County deed books record transfers of title, and a deed conveying a named person from one enslaver to another is a dated, located, name-bearing record of a forced move. A mortgage that pledges named people as security is the same thing in a different legal wrapper, and these are common because enslaved people were the most liquid asset many enslavers held.

Tax lists. Counties assessed enslaved people as taxable property, usually by counting those within a taxable age band. Tax lists more often count than name, so treat them as a way of tracking an enslaver's holdings year by year, which tells you when a group grew, shrank or moved.

Court and chancery files. A disputed estate produces depositions, and a deposition is a narrative. Suits over the division of an estate are disproportionately likely to name people, state ages, and describe who belonged to which family unit, because those were the facts in dispute.

The obvious objection to all of this is that county deed books and probate packets are enormous, unindexed by the names of enslaved people, and were until recently searchable only by reading them page by page. That objection was correct until 2024.

Full-Text Search changed this in 2024, and most advice online predates it

FamilySearch first released Full-Text Search in FamilySearch Labs at RootsTech 2024. It moved into the standard search tools the following year, announced on the FamilySearch blog on 30 August 2025.

That post describes what the tool does plainly. It lets you search nearly 2 billion images using transcripts generated by artificial intelligence. Unlike a traditional indexed search, it reads the whole transcript of a record, so a match can turn up anywhere in the document rather than only in a field somebody chose to index. FamilySearch is careful to add that Full-Text Search does not yet cover all of its image collections, and that more are being added regularly.

Set that next to the previous section and the consequence is immediate. The deed books and probate packets that name enslaved people were never indexed by those names, because nobody indexing land records was indexing for genealogy. Searching the transcripts of those images by keyword is the first method that can find a name inside a document that was never catalogued by name.

One caveat, stated honestly. Those transcripts are machine-generated from handwriting, and the coverage is partial and still growing. So a null result from Full-Text Search does not mean the name is absent from the record. It means the name is absent from the transcripts of the images covered so far, which is a much weaker claim. That is precisely why a failed search is worth writing down, the argument made at length in negative evidence in genealogy.

A law passed on 27 February 1866 turned a marriage into a record

Some of the richest documents of this period exist because a state legislature had to solve a legal problem, and the solution required writing down families.

Virginia is the clearest case and the best documented. The historian Brent Tarter, writing in Encyclopedia Virginia, sets out the sequence. Governor Francis H. Pierpont had been pressing the legislature since 1864 to recognise marriages that enslaved people had entered into, and in December 1865 he spelled out the mechanism he wanted. Every county and corporation court clerk should keep a book. A man and woman who had been living as husband and wife could jointly ask the clerk to enter their names and the date of their marriage in it. That entry, Pierpont argued, should then count as a lawful marriage and make the couple's children legitimate.

On 27 February 1866 the General Assembly passed "An Act to Amend and Re-Enact the 14th Section of Chapter 108 of the Code of Virginia for 1860, in Regard to Registers of Marriage; and to Legalize the Marriages of Colored Persons now Cohabiting as Husband and Wife." It passed both houses on voice votes, and Tarter notes that the newspapers reported nothing of the debates, so it is not known whether anyone opposed it.

The act produced a register in each county, compiled over the following weeks and months, listing couples who had been married in slavery and were now putting it on record. Tarter judges it very likely that almost every Virginia city and county clerk compiled one, though not all of them survive. The Library of Virginia's account of the Augusta County register, published on its UncommonWealth blog in July 2015, adds a detail about how these documents physically worked: clerks often filled out large sheets and then submitted them to the Freedmen's Bureau, which means a register can survive in a federal record group even where the county copy is gone.

That same Library of Virginia account turns up a second, much rarer document type in the Augusta papers, titled Register of Children of Colored Persons in Augusta County, State of Virginia, whose Parents had ceased to cohabit. The Library says only about three of these are known to exist in Virginia. If you find one, it is a list of children attached to parents who were separated, which is precisely the relationship that the sale records in the previous section created.

What the register proves, and what it does not. A cohabitation entry proves that two named people presented themselves to a county clerk on a datable day and asserted a marriage and, usually, a set of children. It does not prove when that marriage began, because the couple were reporting it rather than documenting it, and the year given is memory. Treat the date in it the way you would treat any family story: good evidence of the relationship, weak evidence of the date.

The Freedmen's Bureau wrote down whatever it had to administer

The Bureau of Refugees, Freedmen, and Abandoned Lands, which the National Archives holds as Record Group 105, was established in the War Department by act of Congress on 3 March 1865. The Archives notes that although the Bureau was not abolished until 1872, the bulk of its work was done between June 1865 and December 1868.

It helps to be precise about why these records exist, because that governs what you will find in them. The Bureau was not a documentation project. It issued rations and clothing, ran hospitals and refugee camps, supervised labour contracts between planters and freedpeople, handled apprenticeship disputes and complaints, and helped benevolent societies establish schools. The National Archives describes the resulting records as rich with names and personal information about individuals, including marriage certificates, schooling information, labour contracts, hospital records, complaints, relief rolls, land applications, requests for legal aid and protection, and trial summaries.

Three of those categories repay attention first.

Labour contracts name a freedperson, name the planter, state the terms, and carry a date and a place. If a contract from 1866 names the same planter as the enslaver on your 1860 shortlist, that is a lead worth chasing hard, though it is not proof on its own and you should not assume a freedperson stayed where they had been held.

Complaint and trial files are narratives, and a narrative names the people who were in the room. These are the Bureau records most likely to state a relationship outright rather than leave you to infer it.

Marriage records are a separate series of their own. The Archives describes marriage records of newly liberated African Americans collected from 1861 through 1869, first by the Union Army and then by the Bureau in its field offices across the Southern states and the District of Columbia.

A further route out of this period runs through military service. The service records of Black soldiers and sailors, and the pension files that followed them, contain sworn statements about marriage, children and identity, given by people who knew the applicant personally. Those files are large and follow a known order, and the question of which pages in one actually carry the genealogy is covered separately in which pages of a pension file matter.

A bank account application is the richest single document of the period

If you take one record type away from this article, take this one, because of what a depositor was asked when they opened an account.

Reginald Washington, writing in the National Archives' Prologue magazine in the summer of 1997, describes the records of the Freedman's Savings and Trust Company. The bank was chartered by "An Act to Incorporate the Freedman's Savings and Trust Company", signed into law by President Abraham Lincoln on 3 March 1865, the same day the Freedmen's Bureau was established. It collapsed in 1874.

Fifty-five volumes survive, covering twenty-nine branches, filmed as National Archives Microfilm Publication M816, Registers of Signatures of Depositors in Branches of the Freedman's Savings and Trust Company, 1865-1874. Washington lists what a signature record generally shows, and the list is worth reading slowly: the name of the depositor; account number; age; complexion; date of application; place of birth; place raised; occupation; spouse; children; names of parents, brothers, and sisters; remarks; and signature. He adds that some of the earlier volumes contain the names of former owners or mistresses and the plantations where depositors resided, and that some entries include copies of death certificates.

Read that as a research instrument rather than as a bank form. A single account application can give you three generations, a birthplace, a place of upbringing distinct from the birthplace, a named former enslaver, and a named plantation, all recorded from the depositor's own statement.

Washington's own example makes the point better than any summary could. Dilla Warren opened an account at the New Bern, North Carolina branch on 2 November 1869. Her signature records list eleven of her fifteen children, including the ones who had died and the ones who had been sold away. They name her husband, Pompey Nixon, sold away seventeen years before the Civil War. They name her father, Ned Clark, sold thirty years before the war, and her mother, Harriet Nixon, sold with two of Warren's brothers seventeen years before it. They name two sisters, one of them sold thirty-five years before Warren opened the account.

That is one document, and it holds a family, a chronology of forced separations, and a list of people to go and look for.

Now the practical warning, and Washington states it flatly. The registers of signatures are not indexed. A separate microfilm publication, M817, Indexes to Deposit Ledgers in Branches of the Freedman's Savings and Trust Company, 1865-1874, covers forty-six volumes of indexes for twenty-six branch offices. Because those index entries carry account numbers, they work as a rough finding aid to the registers.

Washington's cautions about the indexes are specific and they are worth taking seriously. More than one index exists for some offices. Some are not in strict alphabetical order. Some omit depositors whose surnames do appear in the registers. Many account numbers are missing, and in some cases the number in the index does not match the one on the signature card. Where that happens, he says, you have to search entire rolls of signature cards for the offices where an ancestor lived.

Three figures from the same article show how many families this touched. When the bank was liquidated, five dividend payments were declared between 1875 and 1883, amounting to 62 per cent of what was actually owed to depositors. Of the 61,131 depositors eligible for those dividends, only 29,996 sent in their passbooks. Some 31,000 small depositors never claimed their money at all.

What none of these records prove

Every strength above comes with a matching limit. Read these as the operating manual, not as a disclaimer.

A name match is not an identification. Surnames adopted after emancipation were often, though not always, the enslaver's, and the same surname then recurs across every family that enslaver had held. A shortlist built from a slave schedule and backed by nothing but a shared surname has not been confirmed.

An age and a sex do not identify a person. The 1850 and 1860 schedules give you a demographic profile, and a profile fits many people. Two enslaved women of the same stated age in the same county cannot be told apart until a document names one of them.

A date reported after the fact is testimony. The marriage year in a cohabitation register, the age in a bank register, the birthplace a depositor gave a clerk in 1869: all of these are a person's statement about their own past, given decades later, under conditions that made exact dates hard to keep. They are excellent evidence of relationships and weak evidence of years.

The absence of a record is not the absence of a person. Bureau field office records survive unevenly, cohabitation registers do not all survive, and Full-Text Search covers a growing subset of images rather than all of them. Each of those is a reason to write down what you searched rather than to conclude anything.

And the hardest one. These records were created by institutions that treated your ancestor as an asset, a labour problem or a liability, and the information in them is shaped by what those institutions wanted to know. A probate inventory records a valuation because a court needed a number. That the same page preserves a name is incidental to everyone who made it, and it is the only reason the name survived.

What KleioBase does with records like these, and what it does not

Disclosure, since we sell a tool.

The particular difficulty of this period is that one person has to be assembled out of documents that never mention each other: a line in an 1860 inventory, a deed from 1863, a labour contract from 1866, a cohabitation entry, a bank register, an 1870 census household.

KleioBase is a research companion that turns exactly that kind of scattered material into one connected archive. You upload the page. Its people, dates, places and relationships are extracted, and the original wording is kept beside them, so a phrase like "Dilla, aged about fifty, and her four children" stays readable as a sentence instead of collapsing into disconnected fields.

Each record is held as a fixed snapshot of what that one document said. When the bank register's age and the census age disagree, both claims stay visible against their own source rather than one quietly overwriting the other. You can then put a question to the Research Companion about the whole archive, which is how you ask what six separate documents together say about one candidate enslaver's holdings.

Two smaller things matter more here than usual. Extraction goes through a Review step where every field is editable before anything is committed, which counts for a lot when the handwriting belongs to a nineteenth-century county clerk. And when a search comes up empty, the negative evidence feature records what you searched and over what range, and writes it into a GEDCOM 7 export if you ever leave.

Now the honest limits. Three of them.

KleioBase holds none of these images. They live at the National Archives, on FamilySearch, in state archives and in county courthouses, and finding the page is work you do yourself before anything is uploaded.

Nothing in the software understands the legal structure described above. It will pull a name and a value off a probate inventory without knowing that the value is the reason the name is there, and it will not tell you that a deed dated between two censuses explains why a family vanishes from one county and appears in another.

And a relationship you reasoned your way to, rather than read off a page, has no structured place in the tree yet. It has to sit as a note until you are satisfied it is settled, which in this period covers most of the relationships you will establish. Deciding what a document proves is still the researcher's job, and here that judgement is not part of the work. It is the work.

Where to start with the 1870 entry you already have

Work in this order, because each step narrows the next.

Write down the 1870 household in full, with every name, every stated age and the exact county and township. Then pull the 1860 slave schedule for that county and list every enslaver holding a group whose ages and sexes could contain the adults in your 1870 household. That is your candidate list and it is allowed to be long.

Next, take each candidate enslaver's name, not your ancestor's, into FamilySearch's Full-Text Search, and look for probate inventories, estate divisions and deed entries in that county in the 1840s and 1850s. You are searching for the enslaver because that is the name the clerk indexed. Read the documents that come back for first names and ages that match your 1870 adults.

Then check whether a cohabitation register survives for that county, and check the Freedmen's Bureau field office records for the same county for labour contracts in 1865 and 1866. Finally, search the Freedman's Bank registers for the nearest branch, and if the index fails you, read the roll.

And write down every one of those searches that found nothing, naming the county, the years and the collection. The next person to work this family should not have to repeat them, and that person may well be you in two years. Recording the failures alongside the finds is the discipline argued for in the evidence layer your research log is probably missing.

The clerk who wrote your ancestor's name into an estate inventory was valuing an asset for a court that wanted a total. The enumerator who left that same name off the 1860 schedule was following a printed form with no column for it. Neither man thought he was recording a family, and neither was.

The name survived anyway, and only because somebody one day might need to prove in court exactly which person had been sold.