↓ Skip to main content
Charted Roots
  1. Guides/
  2. Research guides/

I want to handle conflicting evidence between two sources

1238 words·6 mins
Table of Contents

Use this when two sources tell you different things about the same fact — two census records giving different birth years, a tombstone contradicting a death certificate, a family Bible against a baptism record. Per the Genealogical Proof Standard, conflicts must be resolved, not ignored. This guide walks through Mills’ three-axis framework for doing that, and how to capture the resolution as a proof summary that holds up to review.

At a glance: five patterns for reconciling conflicts
#

  • Direct beats indirect — a birth certificate outweighs a census age implying a year
  • Primary beats secondary — the person who was there outweighs the one reporting later
  • Original beats derivative — the original document outweighs a transcription of it
  • Contemporary beats retrospective — created at the time beats created from memory
  • Negative evidence narrows the field — a meaningful absence is real evidence too

The rest of this guide walks through how to identify which axis applies to a specific conflict and how to capture the resolution.

What you’ll need
#

  • Charted Roots v0.22.17 or later.
  • The conflict in front of you — at minimum, two source notes covering the same fact about the same person.
  • Some patience. Conflict resolution isn’t a 5-minute task; doing it right means recording why one piece of evidence outweighs another.

Steps
#

1. State the conflict precisely
#

Vague conflicts don’t resolve. Write the question down in one sentence:

“The 1870 census suggests William Anderson was born ~1818; the 1880 census suggests ~1825. Which is correct?”

That sentence is the research question for the proof summary you’ll write at the end. It scopes everything that follows.

2. Classify each conflicting source on three axes
#

Mills’ Evidence Explained methodology treats source quality as three independent classifications, not one. Set the relevant frontmatter properties on each source note:

source_classification: original | derivative | authored_narrative
information_classification: primary | secondary | undetermined
evidence_classification: direct | indirect | negative

What each axis answers:

AxisQuestion it answersExample values
SourceWhat kind of document is this?A census image is original. A FamilySearch transcription of that census is derivative. A published genealogy citing the census is authored_narrative.
InformationWho provided this information, and how close were they to the event?A census respondent reporting their own age provides primary information. The same respondent reporting a parent’s birthplace provides secondary information. An unsigned record where you can’t tell who informed it is undetermined.
EvidenceHow does this information relate to your specific question?A birth certificate proving the date is direct evidence. Age on a census implying a birth year is indirect. A person’s absence from a census where they should appear is negative — and meaningful.

The two census records in the William Anderson conflict are both original sources but contain secondary information (the household informant likely didn’t witness William’s birth). That changes how heavily either record weighs against, say, a contemporary family Bible entry recording the birth date in original ink.

3. Look for negative evidence
#

Negative findings are first-class. If the conflict is “where was William born,” and you’ve searched a county’s church baptism registers covering the relevant years and found nothing, that’s evidence_classification: negative on the absence — and it has to be recorded, not silently dropped. Document it in your research log:

## Research Log

- **2026-04-30** — [[Lancaster County Baptism Register 1815-1820]] — Searched all surnames Anderson → negative. No baptism record located for William.

The charted-roots-negative-findings block on the person note will surface entries like this when the time comes to write the proof summary.

4. Reconcile the conflict
#

With the three-axis classifications in front of you, the conflict usually stops being symmetrical. Common patterns:

  • Direct beats indirect. A birth certificate (direct evidence of the date) outweighs a census age (indirect evidence implying a year).
  • Primary beats secondary. The person reporting their own marriage date outweighs a death-certificate informant reporting the deceased’s marriage date decades later.
  • Original beats derivative. The original handwritten census schedule outweighs a database transcription of it — especially when the transcription is what you’re getting at first glance.
  • Contemporary beats retrospective. A document created at the time of the event beats one created from memory years later, all else equal.
  • Negative evidence narrows the field. If a baptism register exists for the right place and years and doesn’t list the person, that’s a real signal — possibly that the family wasn’t there, possibly that the family wasn’t of that denomination.

Sometimes the conflict doesn’t resolve cleanly. That’s also a finding. Capture it as confidence: possible rather than forcing a conclusion.

5. Capture the resolution as a proof summary
#

Open Control Center → Data Quality tab → Source conflicts card → Create proof summary, or go to the person directly via People tab → New proof. Either entry point opens the same modal.

In the modal:

  • Add each conflicting source as an evidence item.
  • For each, set Supports to Conflicts with for the contradicting source(s) and Strongly / Moderately / Weakly for the source(s) you favor.
  • Write the Conclusion as the answer to step 1’s research question.
  • Set Status to Conflicted if you couldn’t reconcile, or Complete if you did. Set Confidence to match — Proven, Probable, Possible, or Disproven.

In the body of the saved note, write the analysis: which axis tipped the decision, why the negative evidence supports the conclusion, what would change your mind. See the proof-summary guide for the full structure.

6. Verify the conflict surfaces (or doesn’t) in Data Quality
#

Control Center → Data Quality tab → Source conflicts card displays a count of unresolved conflicts (proof summaries with status Conflicted and at least one evidence item marked Conflicts with). When you flip a status from Conflicted to Complete, the count decrements. That’s how you confirm the resolution registered.

Variations
#

  • Three or more sources in conflict. Same workflow, more evidence items in the same proof summary. The three-axis classification matters more, not less, when the field gets crowded.
  • Conflict you can’t resolve yet. Leave status as Conflicted and confidence as Possible. The proof summary becomes a placeholder that surfaces in Data Quality every time you open the Control Center — a nudge to revisit when more evidence surfaces.
  • Conflict turns out to be two different people. This happens. Resolve by splitting the person record (or running duplicate detection in reverse — the dismissals are also persistent). The proof summary’s conclusion becomes “Sources A and B are evidence about distinct individuals, not the same person.”
  • Quality classification you’re unsure about. The defaults inferred from source_type (e.g., census defaults to primary source quality) are usually right for the document but say nothing about information quality. Override information_classification explicitly when you can — it’s the axis that most often shifts how a conflict resolves.

Related guides#

Reference
#

Found something wrong or unclear? Suggest an edit — opens a pre-filled issue with the guides label.