I know how to trace a statement backward. A sentence points to a note, the note to a source, the source to an event or observation. Even when the chain is incomplete, its intended direction is clear.
An omission has no such visible handle.
Suppose three witnesses describe the same meeting and none mentions an argument that a fourth witness remembers vividly. The absence may matter. It may suggest that the argument was private, forgettable, deliberately suppressed, or invented later. It may also suggest nothing at all. People omit most of what happens to them. A record is made mostly of exclusion.
Yet once I notice the missing detail, I am tempted to treat the silence as evidence with no provenance.
That is dangerous because an omission does not begin in the archive. It begins in a comparison. Someone expected a thing to appear, looked for it in a particular field of records, and interpreted its absence against a model of what those records normally contain. Without that expectation, there is no omission, only a document saying what it says.
So the provenance of a silence must include the question that produced it.
Why did I expect this source to mention the event? Does the writer usually record disagreements? Was the document composed before the alleged event, or for an audience from whom it would be concealed? Is the surviving copy complete? Did I search the original language, the catalog description, or only a transcription whose editor silently regularized the text?
"The source does not mention it" sounds like a fact about the source. Often it is also a fact about my search, my theory of relevance, and the boundaries of the collection available to me.
This does not make negative evidence useless. Some silences are extraordinarily precise. A ledger that records every shipment except the disputed one tells us something. So does a diary that names every attendee but one, or a database whose documented inclusion rules should have captured a missing case. The force of the absence comes from the regularity surrounding it.
Perhaps negative evidence is strongest when I can describe the mechanism that should have produced a mark.
If the event occurred, this clerk was required to enter it. If the person attended, this roll was compiled to name them. If the phenomenon existed at the claimed scale, this instrument should have registered it. The missing mark then has a lineage: event, expected recording process, surviving record, inspected field. The chain can be challenged at every link.
Without that mechanism, silence is too easy to recruit. It will support almost any story that already expects it.
I want records of absence to carry more than the phrase "not found." They should say where the thing was expected, why it was expected there, what would count as finding it, and which failures of preservation or search could produce the same blank.
Provenance is usually imagined as the history of what survived. But truth also depends on tracing the shapes left by what did not. The difficulty is that a blank cannot testify to how it became blank.
The questioner has to testify too.