Celebrity family trees: how claim indexes keep context

Family trees are made of links. A list is made of entries. Flatten the first into the second and you keep the names and throw away the part that could be checked.

That trade is made constantly in coverage of famous families, usually for good reasons: a "who's who" of a celebrity family is easier to publish, easier to sort, easier to read on a phone. It is worth knowing exactly what is lost in the conversion, because the losses are silent and the result looks more authoritative than the diagram it came from.

What to take away

What the links were carrying

Each connection in a tree holds more than the fact that two names are associated. That is what a pedigree chart is built to carry, and a flat row discards it. It holds a direction, a type, a generation, and a state of confidence. A flat row carries those only if somebody has built the columns for them.

What each link carries, item by item. Direction: a tree line runs from parent to child; a list row has none. Type: a tree separates descent, marriage, adoption and step relationships; a list calls them all "family". Generation: a tree places each person on a level; a list puts everyone on one level. Confidence: a tree can mark a link as documented, inferred, reported or disputed; a list states every row with equal weight. Source: a tree can attach the record to the link; a list attaches it, at best, to the person. Absence: a tree can show two people side by side with no line between them; a list cannot show that they are unconnected.

The second row produces the most durable errors. A tree that distinguishes a marriage line from a descent line loses that distinction once it becomes a list of household members.

A child from an earlier relationship gets attached to a later spouse, and no one writes a false sentence. Blended celebrity families are where this happens most: a published roster of a star's "family" puts stepchildren, children from a first marriage and a current partner's children in one column, and readers fill in the relationships themselves. The Kardashian-Jenner family is the standard case: Kourtney, Kim, Khloé and Rob Kardashian are the children of Kris Jenner and Robert Kardashian, while Kendall and Kylie Jenner are the children of Kris and Caitlyn Jenner. A tree shows two descent lines meeting at one mother, so the six are half-siblings across the two groups. A flat "Kardashian family" list shows six siblings and loses the line that explains the two surnames. The destroyed distinction, and every other link type a list cannot hold, is set out in the guide to what a tree actually asserts.

Order becomes rank, adjacency becomes relationship

Two things happen in a list that never happened in the diagram, and neither is intended.

Order reads as significance. In a tree, position is structural: it means generation and descent. In a list, position means importance, because that is what lists mean everywhere else. A sort by any field at all will be read as a ranking of people unless the sort key is stated and dull.

Adjacency reads as connection. Two names next to each other in a list look related in a way names further apart do not, and readers construct the relationship themselves.

In a tree, two boxes side by side with no line between them are explicitly not connected, but a list cannot say that because it has no vocabulary for absence.

The confidence problem

A tree that marks its links honestly distinguishes what a record asserts, what several records together make unavoidable, what a family account reports, and what sources disagree about. Those four states carry the whole value of the research. As column values they read: documented, inferred, family account, conflicting.

Flatten it and they are gone. Every entry carries equal weight, so the weakest link looks as confident as the strongest. A reader cannot tell which is which.

Anyone copying from your list copies that false uniformity forward, and the copy gets cited back at you.

If a list is genuinely needed, the confidence state has to become a visible column, and it has to be filled for every row including the comfortable ones. A column that is only populated where there is doubt teaches readers to ignore it.

The list worth publishing

There is a list that works in this area, and it is not a list of people. It is an index of the work: one row per claim you hold, describing the claim's status rather than asserting it.

One filled row, with letters standing in for names:

Interchange formats for genealogical data, described under GEDCOM, keep exactly that separation between an assertion and the source behind it.

For a famous family, the same row format works with real names, because the parents have made the relationships public. Applied to the British royal family, an index row reads "Prince William is a son of King Charles III", type descent, confidence documented, source the royal household's own published biography. A second row for Prince Harry carries the same type and source. What the index does not do is turn those two rows into a ranked roster: the order of rows is the order of entry, and any line of succession is a separate, sourced claim, not something the list implies.

This version is publishable, useful, and honest. It also does something the roster version cannot: it can contain nothing about living people and still be complete.

Rows about living people stop at what they have made public, and rows about minors do not exist, even when the parents are famous and the children are photographed. The guide to why a list of children cannot be assembled sets out what consent would require.

Living branches are a different data type

The lower part of a tree is not a smaller version of the upper part. The evidence changes, and so does the question.

Most record types that make earlier generations tractable are closed for the living. State birth records are one: the public copy is commonly sealed for 75 to 100 years, the term depending on the state. What remains is what people have said about themselves.

A list of living relatives with dates and places is a compiled identity record, whatever the intention behind it. The profile standard this site works to sets the limits, because that compilation is what turns small facts into an identity record.

The living part of any published index carries only relationships that the people involved have acknowledged. It omits dates precise enough to place someone, locations, and children beyond a public parent relationship.

Common questions

Can a list and a tree be published together?

Yes, and it is the better arrangement, provided the list is generated from the same underlying entries rather than typed separately. Two hand-maintained representations of one dataset diverge, and the divergence is invisible until someone notices the two pages disagree.

Is an alphabetical list neutral?

More neutral than most, and still not free. State the sort key: family name at birth, then given name. That key files a woman who has used her married name for forty years under a name she no longer uses. And an ordered set looks complete because it looks orderly.

What about a list of relationships rather than of people?

That is a set of pairs, and it has its own behavior: symmetry, failed transitivity, and a second person in every row who may never have spoken. Those problems are worked through in the guide to compiling relationship pairs.

How do we show a household that does not fit one branch?

Not with a list, and not with a single enclosing box. Describe the relationships between people individually, name the type of each one, and give the ones that started and ended their period. The vocabulary for doing that without ranking anybody is set out in the guide to households formed from more than one family.