Study for genealogy certification by practicing judgment, not recall. For every record you already own, write what the source is, who supplied the information, and what it proves alone or in combination. Then resolve one real conflict in writing each week.
Why source, information, and evidence are three different layers
A source is the artifact you consult, information is what a person states on it, and evidence is what you conclude from that information after judging its quality. Keeping the layers separate stops you from quoting a record for a claim it cannot support.
Conflation happens because all three layers arrive at once: you open a marriage register (source), read that the bride was 'of full age' (information), and start treating that as proof she was over twenty-one (evidence). Each step involves a separate judgment. The source question asks about the record's creation; the information question asks who supplied the statement and how they would know; the evidence question asks whether the statement answers your specific research question at all.
Apply the layers with a three-line annotation on every record you study. For a baptism entry, write: source — parish register, original; information — birth date reported by the clerk from the family, probably a parent; evidence — direct for the baptism date, indirect at best for the birth date. The annotation takes two minutes, and it exposes the gap between what a record states and what you want it to prove.
Direct versus indirect evidence: making each item carry its weight
Direct evidence answers your research question by itself, such as a birth date stated on a registration. Indirect evidence suggests an answer only in combination with other items, like household composition pointing to a mother's identity. Classify before you weigh.
Classification depends on the question, not the record. A census entry is direct evidence of who lived in a household on census night and indirect evidence of a marriage, a birthplace pattern, or a death window. This relabeling is where self-study pays off: take ten records from your own family file and, for one fixed question such as 'when did John die?', mark each item D or I. Expect to find that most of your file is indirect for most questions, which tells you where correlation work is needed.
Add negative evidence to the same exercise: the meaningful absence of a person from a record series where they should appear. A man missing from every tax list after 1888, in a district with unbroken survival of those lists, is evidence of departure or death even though no document says so. Distinguish this from indirect evidence, which is present but partial. Writing one sentence per classification keeps the distinction honest.
Original versus derivative sources: where errors multiply
Original sources are the first recording of information; derivatives are transcripts, abstracts, indexes, and copies made later. Each derivative step can introduce transcription slips and dropped context, so track which version you actually used and what it descends from.
Derivatives are not disqualifying; they are risks to manage. An index entry that omits a middle name, a transcription that reads one surname for another, or an abstract that silently drops a condition clause each changes what the record seems to say. The study skill is recording the lineage: note in your citation that you consulted a typed transcript held in a county archive, and state what the informant or archivist says about the original's availability. Then you can decide whether to seek the original or to use the derivative with stated limits.
Build a comparison habit rather than a blanket rule. When the original survives and is reachable, compare it against the derivative you used first and log every difference you find; those differences become your personal calibration for how much to trust that record class. When the original is lost, the derivative may be your best evidence, and your proof argument should say so plainly instead of implying you saw the original register.
| Record version | What it is | Typical risk | Study action |
|---|---|---|---|
| Original register or filing | First recording of the information | Informant error at creation | Weigh the informant's knowledge |
| Certified or photographic copy | Reproduction of the original | Usually low; legibility limits | Confirm it reproduces the whole entry |
| Transcript or abstract | Human re-typing or summarizing | Misreadings, silent omissions | Compare against the original where possible |
| Index | Finding aid, often entry-level | Names and dates reduced or normalized | Use to locate records, not as proof itself |
The Genealogical Proof Standard as a checklist you run in order
The Genealogical Proof Standard is a five-part discipline: reasonably exhaustive research, complete and accurate citation, analysis and correlation, resolution of conflicting evidence, and a soundly reasoned written conclusion. Treat it as a sequence to run, not a slogan to cite.
The order matters because each step feeds the next. Citations are only meaningful once you have searched broadly enough to have something worth citing; correlation is only fair once every item is correctly described; conflict resolution only works after correlation shows which items genuinely disagree. A useful self-audit is to pick one concluded relationship in your file and ask, in sequence, which record classes you searched, which you did not, and why. Gaps in the first element surface immediately under this questioning.
Practice the written element deliberately, because it is the element that reveals whether your analysis actually holds together. A soundly reasoned conclusion states the question, lists the relevant items, explains how they combine, addresses contrary items, and ends with a stated degree of confidence; drafting one exposes any gap in the earlier steps at once. Draft such a paragraph for a question you have already answered informally, then check it against the five elements and revise. The revision you make on paper is the skill you carry into any assessed task.
- Reasonably exhaustive research: list the record classes relevant to the question and note which you searched and which remain.
- Complete and accurate citations: enough detail for another researcher to find the same item, including derivative lineage.
- Analysis and correlation: compare items item-by-item for agreements, disagreements, and patterns across sources.
- Conflict resolution: explain why you prefer one item, using source and informant quality, not convenience.
- Soundly reasoned written conclusion: a coherently argued statement of what the combined evidence supports.
Scenario one: a birth year that disagrees across three records
A death certificate says the deceased was born in 1874; the 1901 census implies 1876; a baptism register gives only a baptism date. The tempting move is to trust the most official-looking record. The better move is to classify, then correlate.
The plausible mistake is treating the death certificate as the winner because it looks formal. Classify it instead: an original source carrying secondhand information, since the informant was likely an adult child reporting a birth decades earlier. The census is an original source with household information, probably supplied by a parent who lived through the birth, but rounded or approximated ages are common in such schedules. The baptism register is an original with contemporaneous information, yet it dates the baptism, not the birth. Each record is strongest for something other than the question asked.
The better decision is a written correlation: use the baptism date as an anchor, treat the parent-supplied census age as the stronger birth-year indicator, and treat the death certificate's year as the weakest for this question while noting it may be the strongest for the death itself. State the residual conflict rather than burying it, and give the conclusion a confidence level. This matters because an unexplained conflict in a proof argument signals that the analysis step was skipped, whichever record finally gets cited.
Scenario two: two same-name men and one identity decision
Two men with the same common name live in one county in the same decade, and their records blur into one person in a family file. The plausible mistake is using name similarity as identity proof. The better decision is testing each record against distinguishing criteria.
Work the scenario concretely. One man appears in land records with a wife named Ellen; another appears in militia lists with occupation recorded as cooper. A researcher merges both into one person because the name, county, and period match. The mistake is structural: no single shared attribute distinguishes identity, and merging pollutes every later conclusion built on the merged profile, including census household reconstruction and parentage claims. Once merged, the error is hard to see because the combined profile looks richer and more convincing than either thread alone.
The better decision is to keep the threads separate and correlate distinguishing criteria: occupations, associates, signatures or marks on land documents, church membership, and neighborhood of residence across several years. Where an associate network — witnesses, neighbors, employers — overlaps between threads, note that as linkage evidence; where it stays disjoint, note that too. Only after this comparison does a merge or a split become a reasoned conclusion. Practicing the split-first discipline on your own file is the fastest way to make it automatic.
A practice loop, a rubric, and an adaptable preparation sequence
Practice on records you already own rather than chasing new searches: annotate, classify, correlate, and write. A repeatable weekly loop builds the judgment that certificate-level study demands faster than accumulating more documents does.
Run the loop on one document per week. Step one: annotate source, information, and evidence in three lines. Step two: mark each item direct or indirect for one fixed question. Step three: correlate it with one other record and write a five-sentence proof paragraph. Expected observations after a month: your annotations get shorter because classification becomes immediate, and your proof paragraphs start naming conflicts unprompted. If your paragraphs still read as narrative summaries rather than arguments, slow down at the correlation step and compare items attribute by attribute.
An adaptable sequence: spend the first weeks on annotation and classification with one record type at a time; the middle weeks on correlation and one written conflict resolution per week; the final stretch on full proof arguments under a time limit, using unfamiliar records from a free practice set. Readiness checks before any assessed task: you can classify a new record in under three minutes, you can state why you prefer one conflicting item over another in one sentence, and your proof paragraphs address contrary evidence without being reminded.
- Self-check rubric, scored 0 to 2 per line; these are learning milestones, not passing predictions:
- Layer separation: are source, information, and evidence stated as three distinct lines?
- Classification: is each item labeled direct or indirect for the stated question?
- Citation lineage: does the citation show what version was used and what it derives from?
- Conflict handling: is the disagreement described and a preference justified, or ignored?
- Written conclusion: does the paragraph argue, not narrate, and state its confidence level?
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.