DEVIL13 / READER REFERENCE / EDITION 1.0
Evidence Glossary
Useful words for following a claim from its source to a defensible conclusion. Eighteen working definitions, each with an example, to help assess what the evidence can support.
These are plain-language reading aids, not a complete philosophy of science or a substitute for a study’s own definitions. Examples labelled below are invented illustrations, not reports of real events or detector tests.
Browse by task: Claims and explanations · Source records · Testing again · Patterns and uncertainty.
Find a term
Authenticity · Base rate · Causation · Claim · Correlation · Corroboration · Evidence · False positive · Hypothesis · Independence of sources · Inference · Observation · Primary source · Provenance · Pseudepigraphy · Replication · Reproducibility · Uncertainty and confidence
Claims and explanations
Claim
A statement presented for someone to accept, reject or investigate. Make it specific enough to compare with a record or observation. A statement about an image’s date is a different claim from a statement about how that image was made.
Hypothetical example: “This photograph was taken at the harbour yesterday” contains a place and a date that can be checked.
Evidence
Material used to support or challenge a particular claim: a record, observation, measurement or other relevant information. Its value depends on how it was produced and how well it bears on the question. Explain the connection instead of treating a source link as the conclusion.
Hypothetical example: An earlier archived copy challenges “taken yesterday”; it does not by itself identify the camera or editing software.
Observation
What you notice or record through a stated method. Keep the description close to what was available to inspect, and note limitations of viewpoint, instruments or resolution. The Library of Congress distinguishes observation from inference in its source-analysis approach.
Hypothetical example: “The page has three red marks” describes visible features; “the marks warn of danger” adds an interpretation.
Inference
A conclusion drawn from observations and other information. It can be well supported, weak or mistaken. State the reasoning that connects the evidence to the conclusion, along with alternatives that could explain the same observation.
Hypothetical example: Matching tears and page numbers suggest two leaves belonged together; further material evidence could strengthen that inference.
Hypothesis
A proposed explanation that can be investigated. For an empirical question, specify observations that would count for or against it. An explanation that accommodates every possible result gives the investigator little way to distinguish it from competing explanations.
Hypothetical example: If reflections caused the repeated lights, changing the camera’s angle should change the pattern in a predictable way.
Apply these distinctions with the Evidence Engine → · Back to terms ↑
Following the source
Primary source
An original or firsthand record relevant to the research question. The Library of Congress notes that the category varies by discipline. A source can be primary evidence of what someone reported while leaving the reported event uncertain. “Primary” is not a guarantee of accuracy.
Hypothetical example: A dated diary directly records its writer’s account; its story about a distant event still needs assessment.
Provenance
The documented history of an object or record: where it came from and how it passed between people or collections. Getty explains the term through ownership history. For a digital file, trace available publication and transfer history too; identify gaps rather than inventing a continuous chain.
Hypothetical example: A catalogue, sale record and collection inventory connect a book to three documented owners, with an earlier gap.
Authenticity
Whether something is what it is represented to be in a specified respect. Ask “authentic as what?” A genuine historical letter can contain a mistake, exaggeration or deliberate falsehood. Establishing the document’s identity and assessing the truth of its contents are separate tasks.
Hypothetical example: Confirming that a letter is genuinely from 1890 does not establish that the creature described in it existed.
Corroboration
Additional evidence that supports a claim or a specific part of it. Compare what the sources actually agree about, including their differences. The Library of Congress recommends seeking support and contradiction across sources instead of relying on one fragment of history.
Hypothetical example: A diary and an independently kept harbour log agree that a ship arrived, but only the diary mentions lights.
Independence of sources
The extent to which sources arose separately rather than copying, sharing or depending on the same information. Here the term concerns reporting origins, not a formal calculation of statistical independence. Count the underlying records, not only the websites repeating them.
Hypothetical example: Ten articles quoting one press release provide ten publications, but only one underlying account of the event.
Read the guide: Ten Websites, One Source →
Pseudepigraphy
Attribution of a work to someone who was not its actual author. Such an attribution is part of the text’s history and requires examination. Queen’s University Belfast uses this term for the Key of Solomon; the name in its title is not verified authorship.
Hypothetical example: A later manuscript credits an ancient sage. Study the attribution, the text tradition and this copy’s date separately.
Open the Source Desk → · Read the grimoire guide → · Back to terms ↑
Testing again
Reproducibility
Under the National Academies’ 2019 convention, computational reproducibility means obtaining consistent results from the same input data, code, computational steps and analysis conditions. Other fields may use the word differently. Matching an output checks a computation; it does not guarantee the inputs or interpretation are correct.
Hypothetical example: A second analyst runs the shared script on the shared dataset and obtains the same reported totals.
Replication
In that same National Academies convention, a replication study collects its own data to address the same scientific question. Consistent findings support replicability, assessed with the relevant uncertainty. A replication can be informative even when its result differs; investigate methods, context and possible errors.
Hypothetical example: A new set of observations tests whether the reported light pattern recurs under the stated conditions.
Practice with a fictional translation claim → · Back to terms ↑
Patterns and uncertainty
Base rate
How common an event or condition is in the relevant group before considering a particular new clue. Specify the group and where the rate came from. Penn State’s probability lesson explains why a prior probability and a probability after new evidence answer different questions.
Hypothetical example: Before interpreting an alarm, ask how often the target event occurs among the things this system examines.
False positive
A positive flag when the condition being sought is absent. The precise meaning depends on the task: NIST’s security-testing glossary applies it to an alert for a vulnerability that is not present. Identify what “positive” means before interpreting a detector result.
Hypothetical example: A fictional detector labels an ordinary camera image “generated.” That is a false positive for its generation claim.
Correlation
A measure of how variables vary together. State the measure and inspect the pattern rather than relying on a number alone. An association may suggest a question worth testing; it does not identify its cause. NIST’s scatter-plot guide explains this limit.
Hypothetical example: Reports of strange lights rise on nights with more visitors. The two counts move together.
Causation
A relationship in which one factor helps produce a change in another under specified conditions. Ask about timing, plausible processes and alternative causes. A controlled comparison may help, but the relevant design depends on the question; a correlation alone cannot establish the causal account.
Hypothetical example: More visitors might produce more reports through increased observation, without changing how often the lights occur.
Uncertainty and confidence
Uncertainty describes what remains unresolved, such as a missing record, measurement limits or competing explanations. Confidence should reflect the strength of the grounds for a particular conclusion. Here these are qualitative judgments, not calculated statistical confidence levels. Explain what would change the assessment.
Hypothetical example: The archive establishes that an image existed earlier; its place of creation and original author remain uncertain.
Investigate coincidences → · Read the image-verification guide → · Back to terms ↑
Sources and scope
Checked November 18, 2026. Definitions are DEVIL13’s explanatory synthesis. Institutional links support the distinctions identified in the text. Terms can have specialized meanings in different disciplines; the National Academies convention above is explicitly identified for that reason.
- Library of Congress: Claim, Evidence and Reasoning
- Library of Congress: What is a primary source?
- Library of Congress: Evaluating Sources and Using Evidence
- Getty: What Is Provenance?
- Queen’s University Belfast: Clavicula Salomonis
- National Academies: Understanding Reproducibility and Replicability
- Penn State STAT 414: Bayes’ Theorem
- NIST SP 800-115: Glossary, Appendix F
- NIST: Scatter Plot
Put the vocabulary to work in the Research Toolkit, or try a clearly fictional investigation in The Inquiry Room. Report a substantive error through Contact & Corrections.