# How Accurate Are Family Stories, Really?

Family stories are not lies, and they are not transcripts either. Psychologists and historians who study how memory travels between generations have found a consistent pattern: some parts of a story survive almost untouched, and other parts change in predictable ways. Here is which is which, and how to test the difference.


Every family has one. A castle somebody's ancestors supposedly owned. A name changed at Ellis Island. Three brothers who came over together and scattered to three different states. A grandmother who hid someone during the war.

The instinct when you start real research is to treat these stories as either true or false, and then to feel a little foolish when the paper trail contradicts the version you grew up with. That instinct is aimed at the wrong question. Put the work of memory psychologists, historians of wartime memory and genealogists who test legends against records side by side, and something more useful appears: family stories do not fail randomly. They distort in predictable places, and they preserve other details with real fidelity. Learning which is which turns a story you cannot verify into one you can.

**The short answer.** No study has measured what percentage of family stories are accurate, and treat any number that claims to be that figure with suspicion. Put the memory research and the published genealogical case studies side by side, though, and the distortion has a consistent shape: concrete props, settings and the kind of event tend to survive retelling; names, exact numbers, ages and who played which role tend to drift, and where the family has a stake, the drift runs in the family's favour. A story is rarely invented from nothing. It is usually a real event with the wrong person's face on it, or a true relationship with an inflated timeline. Test each detail separately rather than accepting or rejecting the whole story at once.

## Nobody has measured a truth rate, and that gap is itself informative

Start with what does not exist. There is no published study that takes a large sample of family stories, checks each one against documentary or genetic evidence, and reports a percentage that turned out to be true. Researchers have measured how *far* a specific fact drifts, and how *often* a specific distortion pattern appears, but nobody has produced the single number a curious reader actually wants.

The closest anyone has come is not really about truth at all. A widely used family-history questionnaire developed by psychologists, called the "Do You Know" scale, asks children twenty yes-or-no questions about whether they know things like how their parents met or where they grew up. In a 2008 brief report by Duke, Lazarus and Fivush in the journal *Psychotherapy*, higher scores went with higher self-esteem, lower anxiety and better family functioning. But the scale measures whether a child has *heard* the family's stories, not whether those stories are accurate. It suggests knowing family stories goes with doing well. It says nothing about whether the stories are true, and it should never be cited as if it did.

## Living memory has an edge, and it moves

There is a real limit to how far back a family's own memory reaches, and it is worth knowing where that edge sits before you decide a story is either confirmed or refuted by "everyone always said so."

The Egyptologist and cultural theorist Jan Assmann calls the everyday, person-to-person memory of a family or community "communicative memory," and he puts its horizon at roughly eighty to a hundred years, or three to four generations, writing in *New German Critique* in 1995. Beyond that horizon, what a group remembers is a different kind of memory: origin stories and founding myths, fixed and formal rather than conversational. The horizon is not measured, and he does not claim it is. It is his summary of what oral history studies suggest.

A related idea came earlier, from the historian Jan Vansina, who studied oral tradition in Central African societies for decades. Vansina called it the floating gap: recent generations are richly remembered, the deep past is covered by a handful of origin traditions, and in between sits a stretch of relative silence that moves forward through time as the oldest living generation ages and dies. He found that genealogies covering the last two or three generations before a community's oldest members could be genuinely reliable, because they carry incidental, unpatterned detail rather than a tidy origin story. One correction is worth making here, because it gets repeated incorrectly online: the "eighty to a hundred years" figure belongs to Assmann, not to Vansina. Vansina's own writing gives no year figure at all, only "two or three generations," and he was describing oral societies with no writing, which makes the comparison to a modern family with photographs and letters an analogy rather than a direct transfer.

How thin living memory already is at three generations shows up in a 2007 Ipsos MORI survey for the National Trust: 57 per cent of UK adults could not name a single one of their eight great-grandparents in full, and 2 per cent could name all eight.

The practical version of both ideas: a story about your grandparents' own lives is operating inside a family's living memory and is worth taking seriously as a lead. A story about "the old country" four or five generations back has usually crossed into the territory both writers describe, where the specific facts have mostly fallen away and what is left is closer to a founding legend than a report.

## What a story loses each time it is told

In 1932 the psychologist Frederic Bartlett published the results of an experiment that still shapes how researchers think about memory in transmission. He had one person read a short, unfamiliar folk tale, "The War of the Ghosts," and write it down from memory. The next person read that version and wrote down their own, and so on down a chain: a written, slowed-down game of telephone.

The pattern that emerged was consistent enough that Bartlett named its parts. Details were dropped, a process he called omission. Gaps left by the missing details were filled with explanations that made the remaining story make more sense, which he called rationalisation. And unfamiliar elements were swapped for familiar ones: a canoe became a boat, paddling became rowing. By the end of a short chain, every trace of the supernatural had vanished from a story that was originally about ghosts, and the whole thing had become shorter and more internally consistent than the original, even though nobody in the chain was trying to shorten or simplify it.

One of Bartlett's specific findings matters more than any other for genealogy. He wrote that in serial retelling, "names and numbers rarely survive intact for more than a few reproductions," while the central, salient events of a story, the parts that carry its emotional weight, resist change far longer. Bartlett was also candid that his lab under-represented one thing: in real oral retelling, he wrote, much more elaboration takes place. Later work has held up. Bergman and Roediger replicated his related repeated-reproduction experiment, in which the same person retells a story over time, in *Memory & Cognition* in 1999, and found distortion growing with the delay. And in 2006 Mesoudi, Whiten and Dunbar found, in the *British Journal of Psychology*, that gossip about relationships between people was passed down transmission chains more accurately than equivalent non-social information.

Put those together and you get a working rule, which is our synthesis rather than a finding of any one study: that something happened, and the shape of the relationships involved, tends to survive. The precise name, the exact date and the number attached to it are the first things to drift.

## The moral of the story gets rewritten more than the story does

Bartlett's experiment used a folk tale nobody had a personal stake in. Real family stories carry something his ghosts did not: a reason to want a particular version to be true. The clearest demonstration of what that pressure does comes from a study of German families remembering the Nazi era.

The social psychologist Harald Welzer and his colleagues interviewed and recorded family discussions across forty German families, spanning grandparents who had lived through the period, their children, and their grandchildren, 182 interviews and family conversations in total. What they found was not simple denial. It was a specific, repeatable transformation Welzer called "cumulative heroization." Across two thirds of the families studied, stories about grandparents during the Nazi period grew more flattering as they passed from the generation that lived them to the generation hearing them secondhand, even when the grandparents' own accounts, on record in the same study, described something considerably less admirable. Roughly 15 per cent of all the stories told across the interviews cast a family member as a hero, and about 50 per cent cast them as a victim. When a grandparent did tell a story that implicated them, Welzer reports it went nowhere, as though the rest of the family had not heard it.

One case makes the mechanism vivid. In the grandmother's own account, she kept camp survivors out of her house. In her granddaughter's version, the grandmother hid an escapee. The granddaughter's story is built from real props: the road to Bergen-Belsen, a box, straw. But those props came from a different family story about a different woman, and the granddaughter had moved the rescue, and the credit for it, onto her own grandmother. The furniture of the story, the physical objects and setting, survived the retelling essentially unchanged. The person at its centre, and what she actually did, did not.

That is the pattern worth carrying into your own family's stories, especially any that involve moral judgement: a war, a feud, a scandal, an inheritance dispute. The scene tends to be reasonably reliable. The role each person plays in it is the part a family reworks, generation by generation, in whichever direction makes the story easier to live with.

## Four family legends, and what is actually going on inside each one

**"Our name was changed at Ellis Island."** This is one of the most widely repeated stories among American families with immigrant ancestors. The history office of US Citizenship and Immigration Services, the modern successor to the service that ran the station, says plainly that it is not how the process worked. The passenger list an immigrant's name was checked against was created abroad, close to home, when the ticket was purchased, not at the island itself. Inspectors worked from that list and were instructed not to alter a name unless the immigrant asked or the original was demonstrably wrong. The kernel of truth is real: names absolutely did get mangled, repeatedly, at every step of being copied from a ticket agent's handwriting to a ship's manifest to a clerk's transcription, and immigrants often did adopt a new spelling or an anglicised name within their first years in the country. What did not happen is a single dramatic moment at an inspection desk. USCIS offers the most likely explanation for how the story formed: recalling their arrival decades later, many immigrants referred to the entire experience of their first years as "Ellis Island." The name change was real. The place and the moment were compressed.

**"We have a Cherokee princess in the family."** The genealogist Elizabeth Shown Mills has published a fully documented case study of exactly this claim. A family memo from 1950 held that an ancestor lived to 108 and had a "half Indian" wife. By the dates in his own Revolutionary War pension papers, he was 91 when he died. And the Cherokee reservation records the family had attached to the story belonged to a different Thomas Jones, with a different wife. Mills's summary of the pattern is worth quoting directly: stories of Native ancestry "sometime prove to be rooted in fact, but typically many generations earlier than lore contends." The historian Gregory D. Smithers, writing in *Slate* in 2015, dates the pattern: in the 1840s and 1850s large numbers of white Southerners began claiming descent from a Cherokee great-grandmother, often a "princess", a title he notes the Cherokee never had.

**"We have a family coat of arms."** The College of Arms in London, the body that actually grants coats of arms in England, addresses this directly on its own website: there is no such thing as a coat of arms belonging to a surname. A coat of arms is granted to one person and descends to their descendants in the legitimate male line. Two people with the identical surname, and no traceable relationship to each other, can be entitled to completely different arms, or to none at all. A shop that will print anyone's surname onto a shield is selling decoration, not heraldry.

**"Three brothers came over, and one went north, one went south, and one went west."** This one has no institution to settle it and, as far as we could find, no study behind it either way. What exists is genealogists' long familiarity with it as a stock legend. The genealogist Sharon DeBartolo Carmack notes it is always three brothers, never two or four, and her advice is the right test: confirm the three brothers were not simply three men with the same surname. *Family Tree Magazine* adds that immigrant relatives were more likely to settle near each other than to scatter, though it cites no data for that either. Before accepting three-brothers-and-three-directions as an origin story, look for documentary evidence that the three men were related to each other at all, rather than assuming the shared surname settles it. Tracing the witnesses, neighbours and associates each man kept, the [FAN club method](/blog/breaking-genealogy-brick-walls-fan-club-research), is often the fastest way to tell a real brother from a coincidental namesake.

## Not every version of a story carries the same weight

Genealogists have a formal way of ranking testimony that applies just as well to a story told at Thanksgiving as it does to a document in an archive, and learning it changes how you listen to the next family story you hear.

The distinction that matters most is between primary and secondary information. Primary information comes from someone who was actually there. Your grandmother's own account of her own wedding day is primary information. Her account of what her grandfather went through on the boat, which she heard from her own mother, is secondary information, once or twice removed from anyone who actually witnessed it. Neither is worthless, but they are not the same kind of evidence, and a story's reliability tends to fall off sharply with each generation of hand-to-hand retelling between the event and the version you are hearing.

The Board for Certification of Genealogists writes the rule into its *Genealogy Standards*: Standard 39 "assigns preference to primary information provided by a reliable informant." A story told with confidence a hundred times is still secondary information if nobody in the chain was there. That single test, asking who this specific claim ultimately traces back to, does more to sort a good family story from a shaky one than any amount of debating whether the story "feels" true.

## A method for testing a story you cannot simply trust or dismiss

Elizabeth Shown Mills, whose Cherokee-princess case study appears above, has published a compact method for testing a family tradition that treats the story itself as a set of clues rather than as a fact to accept or reject outright.

Start by pulling every discrete, checkable detail out of the story: a name, an approximate date, a place, an occupation, a relationship. Then, for each detail separately, look for the earliest written record that could confirm or contradict it, deliberately working backward past every retelling to the original source, whether that is a vital record, a pension file, a ship's manifest, or a contemporary letter. Along the way, weed out any source that turns out to simply be repeating an earlier one rather than offering independent confirmation, and set aside pure hearsay that traces back to no identifiable original witness at all.

Mills points to two articles in the *National Genealogical Society Quarterly* as models of this kind of testing. In one, Louise F. Johnson tested in 1994 the popular lore that Marmaduke Swearingen was Chief Blue Jacket, and Mills describes that tradition as proving "radically wrong". In the other, Douglas S. Shipley's 2002 study of an enslaved family in Missouri, the core of the tradition held. Both outcomes are wins for the method: the point is not to vindicate the family story, it is to find out, one checkable detail at a time, which parts of it are load-bearing.

## How you ask the question shapes what you get back

If you are still in a position to interview the person who actually remembers something, the oral history profession's guidance is worth borrowing even though you are not a professional oral historian. The Oral History Society in the UK, the Smithsonian's folklife interviewing guide and the Library of Congress Veterans History Project agree on a few points.

**Ask open questions.** The Smithsonian guide notes that questions beginning "How", "What" or "Why" get fuller answers than ones beginning "Do" or "Did". "What was the crossing like?" also avoids handing the person your own assumption to confirm.

**Record it.** The Smithsonian guide prefers audio because it documents the stories completely and accurately, and the Oral History Society warns that a very cheap recorder is unlikely to produce audible recordings. Your own memory of the interview is subject to every distortion described above.

**Agree the purpose first.** The Oral History Society recommends discussing how the recording will be stored and used before you start, and the Smithsonian guide says never to record without the person knowing.

**Expect contradiction, and keep it.** The Oral History Society warns that relatives may treat their own version as the "true" story, and says the real value of oral history is that it can reveal varied and contrasting memories. Two relatives giving you different versions is not a problem to resolve before you write anything down. It is itself the finding. Record both versions, note who told you each one and their relationship to the events, and let the documentary evidence, not your own preference for one relative's account over another's, settle which one is closer to what happened. That habit of writing down what disagrees rather than quietly resolving it in your head is the same one behind [the evidence layer your research log is probably missing](/blog/genealogy-research-evidence-layer).

## What KleioBase does with a story like this, and what it does not

Disclosure, since we sell a tool.

A family story is not a document, but it behaves like one once you start testing it: it makes specific, checkable claims, and each claim needs its own paper trail. KleioBase keeps a record as a fixed snapshot of exactly what one document said, and when two records disagree, both stay visible rather than one silently overwriting the other. That disagreement is frequently the most useful clue you have. A pension file that gives an ancestor's age as ninety-one and a family memo that gives it as one hundred and eight are not a contradiction to quietly average away. They are two claims that need to be weighed against who was closer to the truth, exactly the primary-versus-secondary distinction described above.

You can record the family story on the person's profile as a note, and a fact wording you add by hand is stored with no source record, so it is visibly your own note rather than something read off a document. The Research Companion can then be asked about the whole archive, answering with clickable citations to the exact records behind each claim, which is a quick way to see which parts of a story have a document under them. If you go looking for a record the story implies should exist, a birth in the village the story names, an arrival in the year it gives, and you come up empty, KleioBase's negative evidence feature lets you record exactly what you searched and over what range, so the next person researching this family does not waste a weekend repeating a search that has already failed once, the same discipline covered in [negative evidence in genealogy](/blog/negative-evidence-in-genealogy).

Two honest limits. KleioBase does not accept audio. If you record an interview, you need to get it into text yourself before any of it can be attached to a profile or asked about by the Research Companion. And nothing in the software can tell you whether a story is true. It will show you where your documents agree, where they contradict each other, and where a claim has no document behind it at all. Deciding what that pattern means, the judgement a genealogist calls resolving the evidence, is the part that stays entirely yours.

## Where to start this weekend

Pick one family story you have never actually tested, ideally one with a specific name, date or place attached to it rather than a vague impression. Write down every checkable detail in it separately. Then chase just one of those details back to the earliest original document you can find, whether that is a census entry, a pension file, a ship's manifest, or a certificate. Do not try to prove or disprove the whole story in one sitting. Test one claim, see what the paper says, and let that result tell you how much weight the rest of the story can bear.

The people who first told these stories were not trying to deceive anyone. An immigrant folding their first hard years into the words "Ellis Island", a grandchild making a grandmother a little braver with each retelling, a family memo adding seventeen years to an old man's life: none of that is dishonesty. It is what memory does to a story it cares about. Your job is not to catch anyone lying. It is to find the real event still sitting underneath the version you were told.

Canonical: https://kleiobase.com/blog/how-accurate-are-family-stories