Reviews
Relationship Dates History: A Step-by-Step Procedure for Tracing Citation Chains
Relationship dates history for timeline editors: an eight-step procedure for tracing a date back to its origin, a worked celebrity example, how to store the chain, and what to do when the trail ends.
In celebrity relationship timelines, an editor may ask where a date came from and find no document—only a page that cites another page, which cites a third, which cites nothing. Relationship dates history includes this publication history, but a chain of references is easy to confuse with a source.
This guide shows how to trace and store that chain for celebrity relationship timeline entries. It explains why being first in the chain is not evidence by itself; it does not establish a date for a specific couple.
What to take away
- Store the citation chain, not just the citation. What a page cites is more informative than what it says.
- The earliest published version of a date is often the least checked one, because everything after it inherited rather than verified.
How a date acquires a history
The sequence is always roughly the same and it takes very little time.
How a date gains authority
- Origin: one approximate assertion
- First pickup: approximation dropped
- Reference layer: citation to pickup
- Consolidation: later pages cite compilation
- Fossilization: known, author unknown
| Stage | What happens | What the date has gained |
|---|---|---|
| Origin | Somebody states a date, often approximately, often in passing | Nothing; it is one assertion |
| First pickup | An outlet repeats it, dropping the approximation | Apparent precision |
| Reference layer | A compilation records it with a citation to the pickup | An apparent source |
| Consolidation | Later pages cite the compilation | Apparent corroboration |
| Fossilization | The date is now "known" and nobody can say by whom | Authority, from nowhere |
Illustrative example, invented to show the shape: a fan post in 2013 says two people met "around 2014." A gossip site repeats it as 2014. A timeline page records 2014 with a citation to the gossip site. A magazine then cites the timeline. The date is now bare: 2014.
Nothing was verified at any stage. Each step added confidence and removed a qualifier. That is the exact shape described under circular reporting: a claim returning to its own origin through intermediaries, and being counted as independent support on the way back.
Storing the chain
A citation records where you got something. A chain records how it got there, and it is the part that lets a later reader assess a date without redoing the work.
What a citation chain records
- Source used, its date, its wording
- What that source cited, and if reached
- Where the trace stopped, and why
- Earliest reachable item and its qualifiers
- Whether any two sources are independent
| Field | What it holds | Example | Uncertainty |
|---|---|---|---|
| chain_id | Short label for the chain | couple-start-2014 | No |
| claim_text | The date as first stated | "around 2014" | Yes |
| origin_source | Where the claim first appeared | Fan forum post, 2013 | Yes |
| pickup_source | First outlet that repeated it | Gossip site, 2013 | Yes |
| compilation_source | Page that recorded it with citation | Timeline page, 2014 | No |
| fossil_source | Later page that cites compilation | Magazine, 2016 | No |
| event_date_claimed | Date your row will display | 2014 | No |
| date_uncertainty | Qualifier you keep | circa 2014 | Yes |
| trace_date | When you last checked the chain | 2024-06-01 | No |
| trace_note | Where the trace stopped | Origin post deleted | Yes |
Tracing the chain is the expensive step and the one worth taking on a date you intend to keep. The procedure is in the next section, and how to anchor and check dates covers the anchoring side. The distinction between a claim that traveled and a claim that was corroborated is the ladder of source types: documented, first-party, reported without a source, unresolved.
The tracing procedure, step by step
Work backwards from the date in your row to its origin.
- Save the page that carries the date, with its URL, its publication date and your retrieval date. Archive a copy.
- Copy the exact wording of the date, including any qualifier such as "around" or "reportedly".
- Read the page for what it cites or attributes, such as "according to" or "first reported by", and list each one.
- Follow each citation to its source. If a link is dead, open the Internet Archive Wayback Machine for the capture nearest the page's date.
- Repeat steps 2 to 4 at every hop, adding a line to the chain record for each page.
- Search the exact phrase with a date filter, and use the siteoperator, to find earlier pages that carry the same date.
- Stop at a document (a clerk's record or a court docket), a first-party statement that carries the date, or a page that cites nothing.
- Grade the datedocumented, first-party, reported without a source, or unresolved. Keep the qualifier that matches the grade.
A chain that ends at a first-party source
The Jennifer Lopez and Ben Affleck marriage shows a short chain. News coverage of the Las Vegas wedding pointed back to Lopez's own newsletter, On The JLo, published July 17, 2022, which described a ceremony on July 16, 2022. That is a first-party statement carrying the date, so step 7 stops there, and the row can be graded first-party. The independent check is the Clark County marriage record, which has to be pulled and dated separately. Until it is, the row stays at first-party, not documented.
Why earliest is not best
There is an intuition that the first published version is closest to the truth. In this subject it is often the opposite.
Earliest item vs. real document
Earliest published item
- Written
- Quickly, one conversation
- Checking
- None possible
- Qualifiers
- Hedges, later dropped
- Effect
- Sets inherited value
Document or first-party statement
- Written
- At the time, with the date
- Checking
- Verifiable
- Qualifiers
- Stated plainly
- Effect
- Improves the date
The first item was usually written quickly, on a single conversation, with hedges that later disappeared. It was not checked, because there was nothing to check it against. What it did was set the value that everything afterwards inherited, which is a different property from being reliable.
What actually improves a date is a document, or a first-party statement made with the date in it, arriving at any point in the chain. Those are worth more than any amount of early publication, and they are worth flagging as such in the row.
The difference between a source made at the time and an account written later is the ordinary distinction between a primary source and a secondary one. And in this subject the primary layer is often missing entirely.
Qualifiers disappear as a claim travels
Watch what happens to the words around a date rather than to the date itself.
Qualifiers dropping away
Origin wording
- Certainty
- Approximate, reported
- Attribution
- Given by one person
- Cause
- Hedged claim
Fourth appearance
- Certainty
- Bare date
- Attribution
- None
- Cause
- Formatting artifact
An origin item says a date is approximate, or reported, or given by one person. The pickup drops one qualifier for length. The compilation drops another for format, because its table has a date column and no uncertainty column. By the fourth appearance the date is bare, and the bareness is a formatting artifact rather than a finding.
The defense is a schema that can hold uncertainty, so your own table does not do the same thing. A row might read "circa March 2015 (reported)" instead of "March 2015". Precision you display has to match precision you hold, and the ways to render an uncertain date without inventing certainty are set out in the sourced index of relationship dates.
What to do when the chain ends at nothing
It usually does. The honest outcomes are limited and none of them involves keeping the date at face value.
When the chain ends at nothing
Can the entry support this date?
attribute to earliest reachable item, restore its qualifiers
record that it circulates with no findable origin, or drop it
When the chain ends at nothing, the steps are:
- Mark the date as unresolved and keep the original wording.
- Record what you searched and where the trace stopped.
- Use the Internet Archive Wayback Machine, Wikipedia's page history and What links here, and Google's site: operator to look for earlier pages.
- Replace the bare date with an uncertain rendering, or drop the date if the claim depends on it.
An old entry should carry the reasoning that produced it, so this work is done once. When you revisit an old entry, repeat the trace and attach the new reasoning to the record so a later editor can retrace it.
Common questions
If three sources agree, is that not corroboration?
Only if they are independent, which means separate underlying evidence rather than separate mastheads. Three pages citing one item are one source with three faces.
Should the row show the whole chain to readers?
Show the source you relied on, its wording, and a note if the trace ended at an item citing nobody. The full chain is internal, but the fact that it ended nowhere belongs on the page.
What if the earliest item is unreachable?
Say so, and record what you could reach. An honest note that the origin is behind a paywall or gone is more useful than a citation to the nearest live page, which implies you saw something you did not.
Does a first-party statement close the question?
It settles what the person said, on a date, which is usually the strongest thing available. It does not settle the underlying date if they were recalling it, and a recalled date is still worth marking as one.
