Back to the on-screen lesson ·
evaluate what makes testimony more or less reliable
Paper packet. Every task here also exists on screen, where it is checked automatically; answers written on paper are not assessed by Nydus. When you are back at a device, enter your answers there.
You will evaluate what makes testimony more or less reliable, explaining the inference and its limits in a supplied case.
A report and its content are different claims. Evaluate both the source's route to the information and the hearer's reasons for relying on it.
| Term | What it means |
|---|---|
| Testimony | An assertion through which a speaker informs an audience. |
| Competence | Ability to judge accurately within the relevant domain. |
| Independence | Absence of the specified shared informational dependence. |
| Provenance | The origin and transmission history of a report. |
Much of what we know comes through other people. A student learns the date of an eclipse from a historian, a traveler learns a platform number from an announcement, and a researcher learns a measurement from a colleague's report. Testimony is not restricted to courtroom speech. It includes many ordinary assertions intended to inform an audience. The central question is how another person's statement can supply a reason for belief when the hearer did not directly observe the event.
Begin by distinguishing the report from its content. 'Rae says the bridge is open' describes an assertion. 'The bridge is open' is the proposition the assertion communicates. Accurate evidence of the first does not automatically guarantee the second: Rae could be mistaken. Yet requiring every hearer to inspect every bridge personally would prevent most ordinary coordination. The philosophical task is to explain warranted dependence, not to choose between universal credulity and complete isolation.
Two broad approaches frame a longstanding debate. Reductionist approaches seek the warrant for testimony in other kinds of evidence, such as a speaker's observed accuracy and the hearer's background knowledge. Non-reductionist approaches allow an initial entitlement to accept appropriate testimony without first proving reliability in each case, while still recognizing defeaters. Both descriptions cover diverse theories. Neither means that every report should be accepted uncritically, and neither makes testimony useless merely because it is mediated by a person.
In the exercises, reconstruct the support supplied by a reporting network. Some reports are independent observations; others are copies of earlier reports. Identify origins, paths, and possible failures before counting agreement. This is a bounded evidential task. You do not need to adopt one complete philosophical theory of testimonial justification to show that three copied messages contain less independent information than three separately produced observations.
Another way: Competence, opportunity and sincerity
A speaker can be honest but mistaken. Sincerity concerns whether the speaker reports what they take to be true. Competence concerns whether they are able to judge the relevant matter accurately. Opportunity concerns whether they were in a position to acquire the information in this case. These features can come apart. An excellent mechanic speaking honestly about an engine they have never inspected may lack the relevant opportunity, while a novice who directly saw a particular switch move may accurately report that limited event.
Expertise has a domain. A skilled botanist's judgment about plant identification does not automatically carry the same authority about railway engineering. Even within a domain, a specialist may need access to the relevant records or specimens. To assess a testimonial reason, state the precise question first and then match the speaker's competence to it. 'An expert said so' is too vague when the disputed issue is whether the expertise fits this claim.
Distinguish possible incentives from established distortion. A financial interest can give us a reason to seek independent verification or inspect methods. It does not by itself prove that a particular statement is false. Likewise, lacking an obvious motive to deceive does not guarantee accuracy. A fair analysis identifies the specific concern and explains how it bears on the report. It avoids turning a biographical detail into an all-purpose verdict on everything the person says.
The hearer's own limits matter too. A highly technical report may be accurate even if a beginner cannot reproduce its calculation. Responsible dependence can involve checking credentials, methods, replication, and institutional correction practices. It need not require pretending to possess the speaker's expertise. Conversely, prestige cannot replace access to what was actually claimed. A famous name attached to an unattributed summary gives less inspectable support than a clear report with an identifiable method and scope.
Another way: Draw the reporting network
Represent each original observation as a source node. Draw an arrow from a source to a person who repeats its information. Suppose Ada directly observes that a gate is locked. Ben hears Ada's report, and Cy reads Ben's message. Three people now assert that the gate is locked, but the chain has one observational origin. If Ada misidentified the gate, the error can propagate through all three reports. Agreement at the end of the chain is therefore not three independent checks of the lock.
Now add Dee, who independently checks the same gate at the same relevant time. There are two observational origins. Ada's and Dee's reports can provide corroboration because their routes do not simply duplicate one another. Independence is always relative to a possible error. If both observers relied on the same faulty sign rather than checking the lock, their apparent independence would not address the sign's error. A useful graph includes methods as well as names when the shared mechanism matters.
Count paths and origins separately. A message can travel through many intermediaries without acquiring new observational support. Intermediaries may still add useful verification, such as checking that a document has not been altered, but that is a different evidential contribution. Mark it explicitly. A person who authenticates a witness's signature has supported the report's provenance; they have not thereby independently observed the event described inside it.
Repetition can also hide changes in scope. An original observer says that one entrance was blocked at noon. A summary says that the building was inaccessible. A later headline says that it was closed all day. A source map should preserve the proposition at each stage. Otherwise you may correctly identify a common origin while missing the fact that the final report claims much more than its source supports. Trace both where information came from and how its content changed.
Another way: Dependence, disagreement and credibility
When reports conflict, do not immediately average them or choose the larger group. First check whether they concern the same object, time, and standard. Two observers can disagree about whether a line is 'long' while agreeing that it contains twelve people, because their comparison standards differ. They can report different gate states because they visited at different times. A contradiction exists only when the relevant propositions cannot both be true under the same conditions.
Next inspect the network. Five outlets repeating one mistaken bulletin do not automatically outweigh one direct correction with a verifiable record. Nor does a lone dissenter automatically deserve special trust. The evidential value depends on access, competence, methods, and responsiveness to correction. A minority report can be weak or strong for reasons that cannot be read from its minority status. The goal is to reconstruct support rather than reward agreement or disagreement as such.
Credibility judgments can also be distorted by prejudice. Dismissing a witness because of an irrelevant social identity can deprive the inquiry of relevant evidence and wrong the speaker in their role as a knower. The correction is not to declare every statement from that speaker infallible. It is to apply relevant standards fairly: what did the person observe, under what conditions, with what competence, and how can the report be checked? Fairness and accuracy support each other when both oppose irrelevant credibility shortcuts.
A practical response to uncertainty should fit the decision. If a report concerns an easily checked meeting location, a direct confirmation may resolve the issue. If it concerns a complex historical event, independent records and careful source criticism may be necessary. If no further evidence is available, it can be appropriate to suspend judgment or state a qualified conclusion. Acknowledging dependence and uncertainty is more informative than pretending that a message count settles every dispute.
Finish a source analysis with a compact explanation: identify the number of independent origins, name any shared route, and state one relevant check. For example, 'The four notices descend from one bulletin; the maintenance log would provide a different route.' This conclusion does not assert that the bulletin is false. It identifies what the current evidence can and cannot do. That bounded result is the philosophical skill being assessed.
Another way: Silence and missing reports
Absence of a report is evidence only when a relevant reporting process would probably have produced one. If an office reliably announces every cancellation through a monitored channel, silence there may count against a cancellation claim. If announcements are irregular, the same silence provides much less support. State this expectation rather than treating every missing message as proof that nothing happened. A source audit therefore records both the reports received and the procedures through which relevant reports were expected. Missing evidence becomes informative through an account of how it would have been generated and noticed.
A community hall receives six messages about whether a weekend event is canceled. Four are forwarded copies of the same neighborhood post. The fifth is a summary of that post on a local website. The sixth comes from the organizer, who says the event is still scheduled and attaches a current booking confirmation. A volunteer must decide what information to publish on the hall's noticeboard.
A head count gives five reports of cancellation against one report of continuation. A source map gives a different result: the first five reports all descend from one unidentified post, while the organizer's report has a distinct route and relevant access to the booking. The map does not magically prove the organizer correct. It explains why five-to-one is an inadequate summary of the evidence. The reports are not five independent observations of cancellation.
The volunteer checks that the booking confirmation identifies the same event and weekend, then calls the hall's scheduling office using its established contact details. The office independently confirms the current booking and says no cancellation has been recorded. There are now two relevant routes supporting continuation, provided the office did not merely repeat the organizer's message. The original post may concern a previous event with a similar name, which would explain the confusion without assuming deliberate deception.
The public correction should state the verified schedule and identify the date. It should explain that the circulating message concerned a different booking only if that origin has also been verified; otherwise, the reason for the original error remains open. It should not accuse every person who forwarded the post of lying. Those people may have been sincere while depending on a shared error. Recording the origin and the correction path also helps future volunteers recognize copied information quickly. The procedure improves inquiry by making dependence visible, not by demanding that nobody ever relies on another person.
Many sincere voices can repeat one error. Draw arrows back to original observations before treating agreement as corroboration. Conversely, identifying a shared source does not prove its report false; it limits what the repetition adds. Expertise must fit the proposition, and irrelevant identity stereotypes cannot substitute for examining access, method, and competence.
Identify the direct observer.
A checks the lock directly
The case gives A an observational route to the gate's state.
Trace the first repetition.
A → B
B reports only what A said and performs no separate check.
Trace the next repetition.
B → C
C's message inherits the same original information through B.
Count observational origins.
1
Only A independently acquires the relevant observation.
State the limit of agreement.
The three messages share A's possible error
Repeated testimony does not remove a defect at their common origin.
Map the first reporting chain.
A → B → C
All three messages in this chain originate with A's observation.
Locate the separate observer.
D independently checks the same lock
D does not derive the report from A, B, or C.
Check the proposition being observed.
Same gate and same relevant time
Reports of different objects or times would not directly corroborate each other.
Count the independent origins.
2
A and D provide separate observational routes.
Explain the additional support.
D can help check an error specific to A's observation
Independence matters because one route need not reproduce the other's mistake.
Locate the original measurement.
A measures the sample
A supplies the only new measurement in the stated network.
Trace the publisher's role.
B republishes A's report
Publication alone does not produce another measurement.
Trace the archivist's role.
C verifies A's signature
Authentication concerns authorship rather than the sample's measured property.
Count measurement origins.
1
Neither B nor C makes an independent observation of the sample.
Identify C's genuine contribution.
Support for the report's provenance
A contribution can matter without being a second measurement.
Specify a different evidential check.
A fresh measurement by D
Replication would test the result through a new route rather than only authenticate the document.
A observes and B repeats A.
A → B
B contributes a second report but no new observation.
C observes independently and D repeats C.
C → D
This chain starts from a different observational origin.
Count both kinds of node.
A observes a result. B repeats A. C repeats B. A, B and C each issue one report. Count independent observational origins and total reports.
| Response | |
|---|---|
| Origins | |
| Reports |
A observes; B repeats A; C independently observes; D repeats B; E repeats C. All five issue one report. Complete the source audit.
Trace the dependent reports.
B and D lead to A; E leads to C
Follow transmission arrows until a direct observation is reached.
Count the independent roots.
origins observational origins
Only the separately acquiring observers add origins.
Count the reporting people.
reports reports
A copied report remains a report even though it adds no observational root.
A and B independently observe the same result. C repeats A. D repeats B. All four issue one report. Count origins and reports.
Independent observation origins: origins. Total reports listed: reports.
A observes. B and C separately repeat A. D independently observes. E repeats D. Each of the five issues one report. Count origins and reports.
Independent observation origins: origins. Total reports listed: reports.
A and D independently observe. B directly repeats A; C directly repeats B; E directly repeats D. Construct only the three direct report-transmission arrows using repeats. Do not link the independent observers.
This task has no paper form; do it on a device.
A caretaker inspects a hall. Two volunteers each forward the caretaker's report. A scheduler independently checks the booking record, and an editor repeats that scheduler. Each named person issues one report. Treat the inspection and booking check as independent origins for the stated scheduling inquiry. Count origins and reports.
Independent observation origins: origins. Total reports listed: reports.
Lesson test: one question per skill, one attempt each, no hints. Your answers are checked when you submit.
A, D and F independently observe. B repeats A; C repeats B; E repeats D. Each of A through F issues one report. Count observational origins and reports.
Independent observation origins: origins. Total reports listed: reports.
Without rereading, explain how to evaluate what makes testimony more or less reliable. Give a fresh case and identify what would change your analysis.
10. Complete a two-origin map, step 3
2 origins and 4 reports
There are two independent acquisitions despite four reporting speakers.