Guides
Dating timeline claims that need a separate check: order, gaps, durations
It explains why derived claims in a dating timeline, such as order, intervals, durations, superlatives and completeness, need a separate checking pass.
Every row in a chronology can be individually sourced and individually true, and the chronology can still assert half a dozen things that nobody ever checked, because they are not written in any row.
Order is one. Duration is another. So is "first", so is the length of every gap, and so is the implication that the sequence is complete. These are derived claims: they emerge from the arrangement rather than from the evidence.
What to take away
- A derived claim is only as good as the weakest date it was derived from, and the display usually hides which date that was.
- Check derived claims separately, as their own pass, because the row-by-row pass cannot see them.
Five claims that live in the layout, not the rows
Order. Two rows appear in a sequence, so one happened before the other. Order is the one thing chronology is supposed to deliver, and it is only established if both dates are precise enough to be compared. Two entries dated to a month, in the same month, are not ordered by anything except the order you typed them in.
Five claims nobody sourced
- Ordersequence only if dates are comparable
- Intervalgap inherits both endpoints' error
- Durationsubtraction on announcement dates
- Completenesseleven rows means eleven reached record
- Superlativesclaims about your archive, not world
Interval. The visual distance between rows, or the arithmetic a reader does from the dates, produces a gap. If either endpoint is imprecise, the gap has an error range that the display does not show.
Duration. "Together for six years" is subtraction performed on two dates, at least one of which is usually the date something was announced rather than the date anything began. The result inherits every weakness of both endpoints and presents itself as a single confident number.
Completeness. A timeline with eleven rows implies that eleven things are what there is. It cannot mean that. It means eleven things reached this record.
Superlatives. First, last, longest, most recent. Each is a claim about the whole set, and the whole set is your archive rather than the world.
Order is only as good as your worst precision
Sorting is a mechanical operation and it will happily order things it has no business ordering. Give a sorting routine a date accurate to the year and one accurate to the day and it will place them, silently and confidently.
Can these two items be ordered?
Do their date precisions overlap?
show as unordered, do not pick one
order is defensible; if display must choose, say position is not evidence
The workable rule: two items are ordered only when their precisions do not overlap. An item dated to a year and one dated to a day inside that year cannot be ordered at all. The page should show them as unordered, not pick one.
A worked example with two real dates. One entry reads 2023. Another reads September 24, 2023, the day Taylor Swift first appeared at a Kansas City Chiefs game. The year contains the day, so the two entries are unordered, and any sequence between them is a typing choice rather than a fact.
The same pair shows what subtraction does. Compare 2023 with February 11, 2024, the Super Bowl Swift attended. The interval is a range, not a number: somewhere between 42 days and 406 days, depending on where in 2023 the first entry sits.
Where the display must choose a position, the entry text must say that position is not evidence.
This matters more than it sounds, because order is where causation gets manufactured. A reader who sees one row above another reads the first as leading to the second, and if the sequence was an artifact of your sorting rule, the causal story is an artifact too.
Intervals, and the arithmetic that hides a guess
The moment two dates are subtracted, their uncertainty vanishes into the answer. Two dates each accurate to a month yield a duration that could be nearly two months off in either direction, and the number that gets published looks exactly as solid as one derived from two certificates.
Three habits keep this honest.
Three habits for honest intervals
- Carry the precision throughmonths in, months out
- State a range, never pick a midpoint
- Never compute across two different date types
The five different things a date can be, and why they must not share a column, are set out in the notes on relationship dates guide, filed at /relationship-dates. Intervals are where mixing them does the most damage, because the mixing disappears into a single number.
"First" is a claim about your archive
Superlatives feel like observations and are actually statements about the boundaries of a collection.
Superlative vs qualified wording
Unqualified claim
- First public appearance
- The first
- Longest gap
- The longest gap
- Most recent
- Most recent
Qualified wording
- First public appearance
- Earliest entry in this record
- Longest gap
- Longest stretch with nothing recorded
- Most recent
- Most recently added to the record
"First public appearance together" means the earliest instance in your record, a fact about photographers and publication, not about a first.
"Longest gap" means the longest stretch with nothing recorded, which may be the stretch with the least attention.
"Most recent" means most recently added, which depends on when you last did a pass.
The fix is to write the qualifier into the sentence rather than into a footnote. "The earliest entry in this record" is accurate, survives being quoted out of context, and costs three words. "The first" is not accurate and cannot be defended by anything you hold.
The same reasoning applies to counting entries per period, which is why a year-grouped view needs a sentence explaining what its numbers count.
A pass for derived claims
Row checking and sequence checking are different jobs and should be different passes. The second one takes a fraction of the time and almost nobody does it.
| What to check | What you are looking for |
|---|---|
| Order | Do the two items' precisions overlap? If they do, the page must show the pair as unordered. |
| Intervals | The earliest and the latest the gap can be, given both endpoints. Publish the range, not the midpoint. |
| Durations | Whether both endpoints are the same kind of date, and whether the sentence names which ones it used. |
| Superlatives | The words that tie the claim to the record, such as earliest in this record or longest recorded gap. |
| Counts | Whether the sentence says what the numbers count, and which period they cover. |
Derived-claims checking pass
- Every adjacent pairare dates comparable?
- Every stated durationwhich two dates, same type?
- Every superlativequalified to the record?
- Every visible gappage says which kind?
- Any total or countsentence says what it counts?
Run it after any addition. One inserted row changes the order, the intervals on both sides, and possibly a superlative. That property separates derived claims from row claims: adding a correctly sourced row can make an existing sentence false.
The guide to corrections, filed at /public-statement-corrections, treats that as a correction to be recorded, not a tidy-up.
Common questions
Is it acceptable to publish a duration at all?
Yes, when both endpoints are the same kind of date, at a precision that supports the arithmetic, and the sentence says which endpoints were used. Most published durations fail the third test alone.
A reader has computed a figure from the page that we did not intend.
Then the page implied it. Readers subtract dates; that is what a timeline invites. If the subtraction gives a wrong answer, the fix is in how the dates are displayed rather than in a disclaimer.
Does an unordered pair look like an error?
It looks like precision, once the page has explained the convention. The alternative is an order you invented, which looks like knowledge you do not have. The ladder of what each source type can support is in the notes on facts guide, filed at /current-couples-facts.
