Knowledge and evidence: follow a reason to its source
Agreement can transmit one reason without adding another. Ask which person observed the fact, what a learner actually knows about the source and what changes when a defeater arrives.
Study route and exercise: University Reading List editorial project. How we curate · Sources checked:
On this page
Scope and responsibility
The archive, people and reports below are invented. The exercise separates truth, belief and evidential support; it does not stipulate one universally accepted definition of knowledge.
Choose your starting point
Begin free with the questions
Use Rebus to compare analysis of knowledge, justification and social epistemology.
Study sources systematically
Use Audi for perception, memory, testimony and inferential support.
Challenge a reductive definition
Use Williamson after epistemology and elementary logic to examine knowledge-first arguments.
Compare the three books
Introduction to Philosophy: Epistemology
Brian C. Barnett (book editor); Christina Hendricks (series editor) · 2021
Free conceptual route
- Choose it when
- You need introductory arguments with social and feminist epistemology alongside traditional questions.
- Use another route when
- You need an advanced monograph or only a probability manual.
- Selected reading
- Read chapters 1–2 on knowledge/justification and chapter 7 on social epistemology.
- Limits and trade-offs
- A survey presents competing approaches, not a verdict machine for every testimony case.
Try before choosing
List the truth, belief and support claims in the archive packet separately.
Check your work: The analyst’s supplied truth is not silently added to the learner’s evidence.
Epistemology: A Contemporary Introduction to the Theory of Knowledge
Robert Audi · Edition: 3 · 2011
Systematic sources and structure
- Choose it when
- You want a sustained account of how testimony and inference depend on sources.
- Use another route when
- You only need a short first definition or knowledge-first specialist argument.
- Selected reading
- Read the selected third edition’s testimony, inference and structure-of-knowledge discussions.
- Limits and trade-offs
- A source taxonomy does not decide every reliability dispute without facts about the particular chain.
Try before choosing
Draw S→D→E and identify every original observation.
Check your work: There is one observation at S; faithful copying adds transmissions, not two eyewitnesses.
Knowledge and Its Limits
Timothy Williamson · 2002
Knowledge-first theory
- Choose it when
- You know standard knowledge analyses and can track a dense philosophical argument.
- Use another route when
- You need a first introduction to reasons or elementary logic.
- Selected reading
- Read the introduction and the discussions of knowledge as a mental state and evidence in the selected copy.
- Limits and trade-offs
- Do not reduce a theoretical disagreement to a simple number-of-sources formula.
Try before choosing
Ask where knowledge or adequate support first enters the testimony chain.
Check your work: Three consistent assertions cannot manufacture an original evidential basis if none is supplied.
Check your preparation
Whose evidence is being assessed?
The learner’s. The analyst also receives the actual location so that truth can be checked; that extra fact is not a new message to the learner.
Does dependence make a report false?
No. It changes how much independent corroboration agreement supplies, not its truth value by itself.
Is a defeater always a disproof?
An undercutting defeater can remove confidence in the source-to-claim link without establishing the opposite claim.
A six-step study plan
Steps completed: 0 / 6
These are your own study marks, not an assessment. They are not sent to this site.
-
Separate perspectives
Selected reading: Rebus: knowledge and justification.
What to do: Make analyst and learner evidence columns.
Check your work: The actual box location appears only in the analyst column.
-
Trace provenance
Selected reading: Audi: testimony.
What to do: Draw the complete report graph.
Check your work: Every arrow has a supplied copying relationship.
-
Count observations
Selected reading: Audi: inference.
What to do: Distinguish one observational ground from three agreeing texts.
Check your work: No independence is inferred from different signatures.
-
State the warrant
Selected reading: Rebus: social epistemology.
What to do: Explain why the original source is credible in the main packet.
Check your work: Use stipulated observation/reliability, not mere repetition.
-
Apply the defeater
Selected reading: Audi alongside follow-up.
What to do: Identify the exact premise the audit removes.
Check your work: It defeats observation-based trust, not p by logical negation.
-
Compare conceptual claims
Selected reading: Williamson with the introductory routes.
What to do: Write truth, justified reliance and knowledge questions separately.
Check your work: Do not claim the books unanimously endorse your preferred knowledge analysis.
Worked exercise
Three reports, one observation
Let p mean “archive item A is in box 4”. The analyst is told p is true and S’s check was correct. Learner L cannot access the item. L receives D saying p and E saying p, and a provenance report stating that S inspected the item and recorded p, D faithfully copied S and E faithfully copied D. L knows S is usually reliable but can make mistakes; L is not told that this particular check was correct. Compare truth, L’s support and independent corroboration. Construct another world compatible with L’s available information in which p is false.
Assumptions
The analyst’s actual-state facts are not separately revealed to L. L’s entire information is the reports, the inspection/copying provenance and S’s ordinary fallible reliability, with no supplied defeater. No numerical error rate or sufficient definition of knowledge is stipulated.
- The original observation occurs at S. D and E are two transmissions along S→D→E, not independent observations. Counting signatures would overstate corroboration.
- Under the supplied reliability and absence-of-defeat premises, L has a credible testimonial route to p through S. Dependence does not make the true reports false or automatically eliminate all support.
- A question about knowledge still concerns whether the relevant conditions for knowledge are met, not whether enough documents agree. The packet warrants a source-based support analysis without settling every theory of knowledge.
- In another world compatible with L’s information, the item is in box 5 and S makes a one-off label-reading mistake, recording box 4. D and E copy exactly as before; L receives the same reports and provenance. This differs from the analyst’s actual world, not from L’s evidence. Ordinary fallible reliability permits the mistake. Thus the available evidence does not logically guarantee p. If p is false, L cannot know p even if its belief remains reasonable; truth is necessary for knowledge, but a full sufficient analysis is not supplied.
Answer and interpretation
In the actual world p is true and L has a credible but fallible testimonial basis. Agreement supplies no independent second check, and the same available evidence is compatible with a false-p world. Truth, reasonable support and knowledge must therefore be assessed separately.
Test the result
Do not let L use the analyst’s knowledge that the actual check was correct. The false-p countermodel preserves L’s information, not every hidden fact of the actual world. Dependence alone does not erase S’s potential support.
Change an assumption
The original check was a guess
Change the underlying scenario: replace the main case’s correct inspection with an unchecked guess. Keep actual p true and L’s initial D/E reports and provenance claim unchanged. L then receives a trustworthy audit that S did not inspect the item and guessed box 4. D and E copied faithfully. L learns the audit but has no other location evidence; the analyst’s knowledge that p remains true is not revealed to L.
- The audit removes the supplied observational basis of S’s record. Faithful transmission preserves the content of the guess; it cannot restore the missing check.
- The reports remain true in the analyst’s actual-state packet. L’s former reason to trust this particular record is undercut, even though no evidence for “not in box 4” is supplied.
- L should withdraw the claim that the reports provide checked observational support and seek a new check. Withholding p is compatible with p’s being true; it is not asserting not-p.
Answer and interpretation
The case now illustrates a true repeated claim whose stated evidential route has been defeated. Repetition does not cure the missing basis, and the analyst’s truth does not repair L’s justification.
Test the result
Separate undercutting from rebutting and belief from its grounds. No numerical confidence or universal philosophical verdict is invented.
Common mistakes
Agreement becomes independence
Trace origin rather than counting documents.
Truth becomes justification
A true conclusion can outlast the defeat of the reason used for it.
Defeat becomes negation
Lacking support for p is not evidence that not-p is true.
Before moving on
- Choose readings by introductory questions, source structure and advanced theory.
- Trace testimonial dependence without erasing a credible original source.
- Explain how a defeater changes support without necessarily changing truth.
Sources and edition checks
Sources checked:
- Rebus: official open epistemology
The 2021 Rebus volume is CC BY 4.0 except credited material and includes traditional and expanded epistemology.
- Audi: selected third edition
Audi’s selected third edition is a systematic source-and-structure route, with named topics matched to that copy.
- Williamson: selected Oxford record
Williamson’s selected 2002 paperback is an advanced knowledge-first argument, not a simple evidence-counting method.
These are proposed study activities, not official course assignments or a claim that every book has been read cover to cover.