Part 1: Words 1–25 — Knowledge Claims, Evidence & Uncertainty

These twenty-five words give Secondary 3 students a vocabulary for asking one of the most important questions in serious reading: “How do we know?” The goal is not to turn English lessons into philosophy or research methodology. The goal is to stop students treating every confident sentence as the same kind of claim.

High-Value Fences for Deck 1

Fence	Boundary that matters
epistemic / factual	Factual describes content claimed to be true; epistemic concerns the status, basis, limits or justification of knowledge itself.
proposition / hypothesis / conjecture	A proposition is a claim capable of being considered true or false; a hypothesis is a testable explanatory proposal; a conjecture is a reasoned but insufficiently established proposal.
empirical / anecdotal / testimonial	Empirical evidence comes from systematic observation or measurement; anecdotal evidence comes from isolated cases; testimonial evidence comes from what a person reports or attests.
inferential / direct	Direct evidence bears immediately on the fact; inferential reasoning moves from evidence to a conclusion not directly observed.
deductive / inductive / abductive	Deduction applies premises to a necessary conclusion; induction generalises from cases; abduction selects the best available explanation for observations.
probabilistic / deterministic	Probabilistic claims allow likelihood and variation; deterministic claims specify that given conditions produce a fixed outcome in the model.
provisional / speculative	Provisional conclusions are temporary but evidence-grounded; speculative claims go further beyond currently secured evidence.
definitive / indeterminate	Definitive evidence settles the particular question to the required standard; indeterminate evidence does not support a determinate conclusion.
falsifiable / verifiable	Falsifiable claims expose themselves to possible disproof; verifiable claims can be checked or confirmed against evidence or procedure.
replicable / reproducible	Both concern whether findings can be obtained again; usage varies by discipline, so always state the operational meaning being used.
provenance / traceability	Provenance concerns origin and history; traceability concerns whether the chain from source through transformations can be followed.
triangulation / repetition	Triangulation combines genuinely different evidence sources, methods or perspectives; repeating one source does not create independent support.
uncertainty / ignorance	Uncertainty can be bounded and quantified or described; ignorance means relevant knowledge may simply be absent.

1. Epistemic

Meaning: relating to knowledge—how a claim is known, justified, limited or assigned confidence
Fence: “factual” concerns what the claim says; “epistemic” concerns our basis for accepting it
Word family: epistemic, epistemically, epistemology  ·  Collocations: epistemic uncertainty / epistemic status
Knowledge job: makes the strength and basis of knowledge visible
Example: “The meeting occurred at 3 p.m.” is a factual claim; whether we know that from a timestamp, memory or hearsay is an epistemic question.
Misuse check: Epistemic does not mean “intelligent” or “academic”.
Secondary transfer: Ask two separate questions: “What is being claimed?” and “What justifies believing it?”

2. Proposition

Meaning: a statement or claim that can be considered true or false within an argument or inquiry
Fence: a topic is something discussed; a proposition is something asserted about that topic
Word family: propose, proposition, propositional  ·  Collocations: central proposition / evaluate a proposition
Knowledge job: isolates the exact statement that evidence must support
Example: “Later start times improve attendance” is a proposition. “School schedules” is merely a topic.
Misuse check: Questions and commands are not propositions in the same straightforward sense because they do not assert truth conditions.
Secondary transfer: Rewrite a vague debate topic as one clear proposition before arguing about it.

3. Hypothesis

Meaning: a proposed explanation or prediction that can be examined against evidence
Fence: a proposition can simply state a claim; a hypothesis is framed for testing, comparison or investigation
Word family: hypothesis, hypothesise, hypothetical  ·  Collocations: test a hypothesis / competing hypothesis
Knowledge job: connects explanation with a possible evidence test
Example: “Notifications reduce comprehension because each alert interrupts working attention” is a hypothesis with a proposed mechanism.
Misuse check: A hypothesis is not “just a guess”; good hypotheses are informed and testable.
Secondary transfer: State what evidence would increase confidence and what evidence would weaken the hypothesis.

4. Conjecture

Meaning: a reasoned proposal or belief that remains insufficiently established by evidence
Fence: speculation may be loose; conjecture is often more structured but still unproven
Word family: conjecture, conjectural  ·  Collocations: reasonable conjecture / remain conjectural
Knowledge job: labels an idea honestly before evidence warrants stronger language
Example: From one unusual ending, a reader may conjecture that the narrator is unreliable, but the interpretation needs evidence from elsewhere in the text.
Misuse check: A well-tested explanation should not continue to be dismissed as “mere conjecture”.
Secondary transfer: Use conjecture when the idea is worth exploring but the evidence state remains thin.

5. Empirical

Meaning: based on observation, experience or measurement of the world rather than reasoning alone
Fence: theoretical reasoning develops an account conceptually; empirical evidence tests or describes what is observed
Word family: empirical, empirically  ·  Collocations: empirical evidence / empirical study
Knowledge job: identifies claims grounded in observed data
Example: Attendance records collected across a term provide empirical evidence about how many students actually entered the library.
Misuse check: Empirical does not automatically mean high quality; poor measurement is still empirical data.
Secondary transfer: Ask who observed what, how it was measured and under which conditions.

6. Anecdotal

Meaning: based on individual stories or isolated experiences rather than systematic evidence
Fence: anecdotal evidence can reveal possibilities and lived experience; it is usually weak for estimating frequency across a population
Word family: anecdote, anecdotal, anecdotally  ·  Collocations: anecdotal evidence / anecdotal report
Knowledge job: prevents vivid examples from silently becoming population claims
Example: One student saying a new app improved revision is anecdotal evidence of that student’s experience, not proof that the app improves outcomes for everyone.
Misuse check: Anecdotal does not mean false or worthless.
Secondary transfer: Ask whether the claim concerns possibility, experience or general frequency.

7. Testimonial

Meaning: based on a person’s report, statement or attestation about what they experienced, observed or believe
Fence: testimony concerns what a person says; an anecdote concerns an individual story and may be told by anyone
Word family: testify, testimony, testimonial  ·  Collocations: testimonial evidence / witness testimony
Knowledge job: makes source position and memory part of evaluation
Example: A witness describing what they saw provides testimonial evidence that may be important but still needs assessment for perspective, memory and corroboration.
Misuse check: Testimonial evidence can be strong for some questions and weak for others; evaluate fit to the claim.
Secondary transfer: Distinguish first-hand testimony, second-hand report and repeated hearsay.

8. Inferential

Meaning: relating to a conclusion reached from evidence or premises rather than directly observed or stated
Fence: direct evidence bears immediately on the claim; inferential evidence requires a reasoning step
Word family: infer, inference, inferential  ·  Collocations: inferential step / inferential reasoning
Knowledge job: makes the bridge from evidence to conclusion inspectable
Example: Seeing wet umbrellas is direct evidence that umbrellas are wet; inferring that it recently rained is an inferential move.
Misuse check: Inference is not guesswork when the reasoning is evidence-based.
Secondary transfer: Write the missing bridge explicitly: “Because ___, it is reasonable to infer ___.”

9. Deductive

Meaning: reasoning in which a conclusion follows necessarily if the premises are true and the logical form is valid
Fence: deduction moves from premises to a logically required conclusion; induction supports a likely generalisation rather than necessity
Word family: deduce, deduction, deductive  ·  Collocations: deductive reasoning / deductive argument
Knowledge job: separates logical validity from empirical truth of the premises
Example: If all entries submitted after midnight are late, and this entry was submitted at 12:10 a.m., then deductively it is late under that rule.
Misuse check: A valid deduction can still produce a false real-world conclusion if a premise is false.
Secondary transfer: Test both the logical form and the premises.

10. Inductive

Meaning: reasoning that moves from observations or cases towards a broader pattern, generalisation or probability
Fence: induction supports rather than guarantees its conclusion
Word family: induce, induction, inductive  ·  Collocations: inductive inference / inductive reasoning
Knowledge job: helps calibrate generalisations from samples
Example: If many comparable classes show the same improvement after an intervention, an inductive argument may support the proposition that the intervention often helps under similar conditions.
Misuse check: One dramatic case is weak induction for a broad claim.
Secondary transfer: Ask about sample size, representativeness, consistency and exceptions.

11. Abductive

Meaning: reasoning that selects the most plausible explanation for an observed pattern from competing possibilities
Fence: deduction asks what must follow; induction asks what pattern is supported; abduction asks which explanation best accounts for what we observe
Word family: abduction, abductive  ·  Collocations: abductive inference / inference to the best explanation
Knowledge job: supports diagnosis and interpretation under incomplete information
Example: A laptop shuts down only during demanding tasks. Overheating may be an abductive explanation if it accounts for the timing better than a random battery fault.
Misuse check: “Best current explanation” is not the same as proven cause.
Secondary transfer: Compare at least two explanations for fit, simplicity, mechanism and counterevidence.

12. Probabilistic

Meaning: expressed in terms of likelihood, chance or varying probability rather than certainty
Fence: probabilistic relationships allow exceptions and distributions; deterministic claims specify a fixed outcome under defined conditions
Word family: probability, probable, probabilistic, probabilistically  ·  Collocations: probabilistic model / probabilistic claim
Knowledge job: prevents tendencies from becoming absolutes
Example: “Sleep deprivation increases the probability of attention lapses” is probabilistic; it does not claim every sleep-deprived student will fail to concentrate.
Misuse check: Probabilistic does not mean arbitrary or unknowable.
Secondary transfer: Use frequency or likelihood language that matches the evidence.

13. Deterministic

Meaning: describing a system or model in which specified conditions determine a particular outcome
Fence: deterministic is stronger than “likely”; it assumes the relevant conditions fix the result within the model
Word family: determine, deterministic, deterministically  ·  Collocations: deterministic model / deterministic relationship
Knowledge job: exposes overstrong cause language in human or social contexts
Example: A simple computer rule may be deterministic: if the password is wrong three times, the account locks. Human behaviour is often less deterministic.
Misuse check: Strong correlation does not create determinism.
Secondary transfer: Ask whether exceptions are logically impossible, merely rare, or simply unobserved.

14. Provisional

Meaning: accepted or used temporarily while remaining open to revision when stronger evidence appears
Fence: provisional conclusions are evidence-grounded; speculative claims may extend beyond secured evidence
Word family: provision, provisional, provisionally  ·  Collocations: provisional conclusion / provisionally accept
Knowledge job: allows action without pretending uncertainty has disappeared
Example: The team provisionally concludes that server load caused the slowdown while waiting for complete logs.
Misuse check: Provisional does not mean careless or weak.
Secondary transfer: State what new evidence would trigger revision.

15. Speculative

Meaning: based on possibility or conjecture beyond what current evidence can strongly establish
Fence: a provisional claim is the best temporary conclusion from evidence; a speculative claim reaches further into uncertainty
Word family: speculate, speculation, speculative  ·  Collocations: speculative explanation / highly speculative
Knowledge job: marks imaginative possibility without disguising it as finding
Example: Claiming that a small attendance rise will transform long-term achievement would be speculative without longitudinal evidence.
Misuse check: Speculative ideas can be useful for generating hypotheses.
Secondary transfer: Preserve the idea but downgrade the certainty language.

16. Definitive

Meaning: sufficiently authoritative or conclusive to settle a particular question for the relevant purpose
Fence: strong evidence increases confidence; definitive evidence closes the specific question to a much higher standard
Word family: define, definitive, definitively  ·  Collocations: definitive evidence / definitive answer
Knowledge job: demands a high threshold before certainty language is used
Example: A timestamped submission record may be definitive for whether a file reached the system before the deadline, though not for why it was late.
Misuse check: “Definitive” should not be used merely because one source sounds authoritative.
Secondary transfer: Define exactly which question has been settled—and which have not.

17. Indeterminate

Meaning: not capable of being determined reliably from the available evidence or current method
Fence: uncertain means confidence is limited; indeterminate means the evidence does not permit a determinate answer at all
Word family: determine, determinate, indeterminate  ·  Collocations: indeterminate result / remain indeterminate
Knowledge job: legitimises “the evidence cannot decide” when that is the accurate conclusion
Example: If two equally plausible explanations fit the limited evidence and no discriminating data exist, the cause may remain indeterminate.
Misuse check: Indeterminate is not a synonym for “I have not thought enough”.
Secondary transfer: State which missing observation would make the question decidable.

18. Falsifiable

Meaning: capable in principle of being shown false by some possible observation or test
Fence: a claim that explains every possible outcome cannot be meaningfully falsified by evidence
Word family: falsify, falsification, falsifiable  ·  Collocations: falsifiable claim / potentially falsifiable
Knowledge job: forces a claim to expose what would count against it
Example: “This method improves every student’s score” is falsifiable because one properly measured counterexample can defeat the universal claim.
Misuse check: Falsifiable does not mean false; it means testable against possible disproof.
Secondary transfer: Ask, “What result would make me abandon this claim?”

19. Verifiable

Meaning: capable of being checked against evidence, records, observation or a defined procedure
Fence: verifiable means checkable; verified means the checking has actually occurred successfully
Word family: verify, verification, verifiable  ·  Collocations: independently verifiable / verifiable record
Knowledge job: separates claims with checkable anchors from claims that rely only on trust
Example: “The announcement was published at 10:14 a.m.” is verifiable if an archived timestamp exists.
Misuse check: A verifiable claim may still turn out false when checked.
Secondary transfer: Name the record, observation or procedure that could verify the claim.

20. Replicable

Meaning: capable of being tested again through a sufficiently similar procedure to see whether a comparable result is obtained
Fence: replication asks whether the finding survives another test rather than depending on one original occurrence
Word family: replicate, replication, replicable  ·  Collocations: replicable finding / replication attempt
Knowledge job: checks whether an observed effect is stable beyond the first study or trial
Example: If another class using the same intervention under comparable conditions shows the same pattern, the finding gains replication support.
Misuse check: Disciplines differ in how they distinguish replication from reproduction; always state the operational meaning being used.
Secondary transfer: Ask what must remain similar and what should be independently re-collected.

21. Reproducible

Meaning: capable of being obtained again when the stated data, procedure or analysis is repeated closely enough for checking
Fence: a useful teaching distinction is that reproducibility often checks whether the same materials or data produce the same result, while replication often tests the finding with new observations; terminology varies across fields
Word family: reproduce, reproducibility, reproducible  ·  Collocations: reproducible analysis / reproducible result
Knowledge job: checks transparency of process as well as outcome
Example: Another analyst using the same dataset and documented calculation should reproduce the reported average.
Misuse check: Reproducibility of calculation does not prove the original data were representative.
Secondary transfer: Separate “Can I reproduce the analysis?” from “Would new data replicate the finding?”

22. Provenance

Meaning: the origin, ownership or documented history of information, evidence, an object or record
Fence: source names where something is encountered; provenance asks where it originally came from and what happened to it along the way
Word family: provenance  ·  Collocations: data provenance / establish provenance
Knowledge job: protects against treating reposted information as independent evidence
Example: Three articles may all trace back to one press release; their apparent plurality disappears when provenance is checked.
Misuse check: A prestigious current host does not erase uncertain original provenance.
Secondary transfer: Trace the earliest identifiable origin before counting sources.

23. Traceability

Meaning: the ability to follow information, evidence or a decision through a documented chain from origin to current use
Fence: provenance identifies origin and history; traceability emphasises whether each transformation, handoff or decision can be followed
Word family: trace, traceable, traceability  ·  Collocations: source traceability / maintain traceability
Knowledge job: makes evidence chains inspectable
Example: A graph has good traceability if the reader can move from chart to dataset, from dataset to collection method, and from method to original records.
Misuse check: Traceability does not guarantee quality; it makes quality easier to evaluate.
Secondary transfer: Ask whether every important claim has an inspectable path back to evidence.

24. Triangulation

Meaning: the use of multiple methods, sources, perspectives or evidence types to examine the same question from different angles
Fence: repetition gives more of the same evidence; triangulation gains strength from partially independent ways of checking
Word family: triangulate, triangulation  ·  Collocations: triangulate evidence / methodological triangulation
Knowledge job: tests whether a conclusion survives different evidence routes
Example: Attendance records, interviews and room-usage observations may triangulate a claim about why a study space is heavily used.
Misuse check: Three websites copying one dataset are not meaningful triangulation.
Secondary transfer: State what each evidence line can reveal that the others cannot.

25. Uncertainty

Meaning: the state in which a claim, measurement or outcome is not known with complete confidence
Fence: uncertainty can be measured, bounded or explained; ignorance concerns information that may simply be missing
Word family: uncertain, uncertainty, uncertainly  ·  Collocations: residual uncertainty / uncertainty about
Knowledge job: keeps conclusions proportional to evidence
Example: Even when several sources converge, uncertainty may remain about the exact size of an effect or whether it will transfer to a new context.
Misuse check: Uncertainty is not the same as “anything could be true”.
Secondary transfer: Name the source of uncertainty: measurement, sampling, model, missing variable, ambiguous interpretation or future conditions.

Pressure Test 1: Proposition → Hypothesis → Conjecture

Statement	Best label	Why
“Students using the new room stayed longer.”	proposition	a claim that can be checked
“The quieter design increases stay duration by reducing interruptions.”	hypothesis	testable explanation with mechanism
“Perhaps the room feels more psychologically secure.”	conjecture	interesting but currently weakly evidenced explanation

Pressure Test 2: Empirical, Anecdotal or Testimonial?

Evidence	Type	What it can support
six months of entry logs	empirical	observed usage pattern
one student’s story about studying better there	anecdotal	possibility and individual experience
a librarian reports repeated complaints heard directly	testimonial	reported observations requiring source evaluation

Evidence type is not a simple quality ranking. The correct question is fit to claim. A first-hand testimonial may be excellent evidence for what one witness heard and poor evidence for how common the event is across a population.

Pressure Test 3: Deduction, Induction or Abduction?

• Deductive: All locked accounts reject new logins. This account is locked. Therefore it will reject the login, assuming the premises and rule are correct.
• Inductive: Across many comparable weeks, attendance rises before examinations; we infer a recurring tendency.
• Abductive: Attendance rose sharply this week; approaching examinations are one candidate explanation, but rain and room closures compete with it.

Pressure Test 4: Probabilistic Is Not Weak

“Students who sleep fewer than six hours are more likely to report concentration problems” may be a strong probabilistic conclusion if the evidence is good. Replacing it with “sleep deprivation always causes concentration failure” makes the sentence sound stronger while making the reasoning worse.

Pressure Test 5: Provisional, Speculative, Definitive or Indeterminate?

Evidence state	Calibration
best current explanation, but more records are pending	provisional
possible explanation with little direct support	speculative
question settled by an authoritative direct record for that exact issue	definitive
available evidence cannot distinguish two explanations	indeterminate

Pressure Test 6: Falsifiable vs Verifiable

Claim A: “Every submission before 11:59 p.m. receives an on-time timestamp.” This can be falsified by one valid counterexample and verified against system records for particular submissions.

Claim B: “The system always behaves correctly, even when records appear to show otherwise.” If every possible contradiction is reinterpreted as proof of correctness, the claim becomes difficult to falsify meaningfully.

Pressure Test 7: Replicable vs Reproducible

For this learning page, use a practical distinction while remembering that disciplines vary in terminology:

• Reproduce the analysis: another analyst uses the same data and documented method and obtains the same result.
• Replicate the finding: another trial or sample collects new observations under comparable conditions and finds a similar pattern.

A result can be reproducible yet fail to replicate if the original dataset captured an unstable or context-specific pattern.

Pressure Test 8: Provenance, Traceability and Triangulation

Four news pages report the same figure. Two cite each other. All ultimately trace to one undated infographic whose source data are unavailable.

1. Provenance: the figure’s original source is weakly established.
2. Traceability: the path can be partly followed, but it ends before raw evidence is reached.
3. Triangulation: four pages do not equal four independent evidence lines because they share one origin.
4. Uncertainty: confidence should remain limited until an independent or primary source is found.

Claim Calibration Ladder

Language	Evidence state	Example
possible	compatible with evidence	may reflect
plausible	fits evidence reasonably well	plausibly explained by
provisional	best current conclusion but revisable	provisionally conclude
probable	evidence favours the claim over alternatives	is likely to
well-supported	multiple good evidence lines converge	the evidence strongly supports
definitive for the question	relevant uncertainty has been sufficiently closed	the record establishes

This is a calibration aid, not a universal numerical scale. The correct phrase depends on the kind of question and the quality of the evidence.

Unseen Source Analysis: How Much Do We Actually Know?

“A student survey conducted after a new study programme found that most respondents felt more productive. One teacher said the class appeared more focused. The school newsletter therefore concluded that the programme has definitively improved academic performance.”

1. Which evidence is empirical?
2. Which evidence is testimonial?
3. Is self-reported productivity the same construct as academic performance?
4. What inferential steps are required to reach the newsletter’s conclusion?
5. Would “definitively” be justified? What calibration is safer?
6. What alternative hypothesis could explain increased reported productivity?
7. What evidence would make the academic-performance proposition more verifiable?
8. What evidence would falsify the strongest version of the claim?
9. How could the school triangulate the conclusion?

Writing Transfer: Upgrade the Epistemic Precision

Weak sentence	Secondary 3 upgrade
“This proves students learn better online.”	State what was measured, identify whether the evidence is empirical, and calibrate the conclusion to the observed scope.
“My friend improved, so the method works.”	Label the case as anecdotal and avoid generalising frequency or causation from one example.
“Experts say it is true.”	Identify who the experts are, their relevant expertise, the evidence they use and the provenance of the claim.
“Nobody can disprove it.”	Ask whether the claim is actually falsifiable or merely protected from counterevidence.
“Five websites agree.”	Check whether the sources are independent before treating agreement as triangulation.

Morphology & Epistemic Network

Core word	Network	Useful question
proposition	propose, propositional	What exact statement is being evaluated?
hypothesis	hypothesise, hypothetical	What observation would discriminate this explanation from a rival?
conjecture	conjectural	What evidence is still missing?
empirical	empirically	What was actually observed or measured?
testimony	testify, testimonial	What did the witness directly experience?
inference	infer, inferential	What reasoning bridge connects evidence to conclusion?
deduction	deduce, deductive	Does the conclusion necessarily follow if the premises are true?
induction	induce, inductive	How representative are the observed cases?
probability	probable, probabilistic	Is the claim about likelihood rather than certainty?
provisional	provisionally	What evidence could revise the current conclusion?
speculation	speculate, speculative	How far beyond the evidence is the claim reaching?
falsifiable	falsify, falsification	What possible result would count against the claim?
verifiable	verify, verification	What record or observation can check it?
replicable	replicate, replication	Would a new comparable trial show the same pattern?
reproducible	reproduce, reproducibility	Can another analyst repeat the stated process?
triangulation	triangulate	Are the evidence routes genuinely different?

Misuse Clinic: Knowledge Claims & Evidence

1. Epistemic does not mean sophisticated; it concerns the basis and limits of knowledge.
2. A topic is not a proposition until a claim is stated.
3. A hypothesis should be testable or discriminable against alternatives.
4. Do not dismiss all unproven ideas as mere conjecture when evidence already exists.
5. Empirical evidence can still be poorly measured.
6. Anecdotal evidence can show possibility but rarely population frequency.
7. Testimonial evidence should be evaluated for source position, memory and independence.
8. Inference is not guessing when the evidence-to-conclusion bridge is explicit.
9. Deductive validity does not guarantee true premises.
10. Induction does not provide logical certainty.
11. Abduction identifies the best current explanation, not final proof.
12. Probabilistic does not mean weak or random.
13. Deterministic claims are unusually strong and often inappropriate for complex human behaviour.
14. Provisional conclusions should still be evidence-grounded.
15. Speculation should be labelled rather than written as fact.
16. Do not call evidence definitive unless the exact question is genuinely settled.
17. Indeterminate is a legitimate conclusion when evidence cannot discriminate.
18. Falsifiable does not mean false.
19. Verifiable means checkable, not already verified.
20. Replication terminology varies by field; define the intended meaning.
21. Reproducibility of an analysis does not prove generalisability.
22. Current hosting location is not always original provenance.
23. Traceability exposes a chain; it does not guarantee every link is good.
24. Multiple copies of one source are not triangulation.
25. Uncertainty should be described, not used as an excuse to say “we know nothing”.

Retrieval Checkpoint 1 — No Notes

1. Explain an epistemic question about an ordinary factual claim.
2. Turn one discussion topic into a clear proposition.
3. Write a testable hypothesis with a proposed mechanism.
4. Write a conjecture that is reasonable but not yet well supported.
5. Give one empirical observation and state how it was measured.
6. Give one anecdote and explain what it cannot establish.
7. Give one testimonial claim and identify a source-evaluation question.
8. Separate direct evidence from an inferential conclusion.
9. Write a valid deductive argument and identify its premises.
10. Write an inductive generalisation and state what would strengthen it.
11. Compare two explanations abductively and choose the better current fit.
12. Rewrite a deterministic claim as a probabilistic one.
13. Give a truly deterministic rule under defined conditions.
14. State a provisional conclusion and what could revise it.
15. Turn an overconfident statement into an honestly speculative one.
16. Give one question for which a record could be definitive and one it could not settle.
17. Create an indeterminate scenario with two equally supported explanations.
18. Write a falsifiable claim and the observation that would falsify it.
19. Write a verifiable claim and name the checking procedure.
20. Describe a replication attempt using new observations.
21. Describe a reproducibility check using the same data or method.
22. Trace the provenance of a fictional statistic through three reposts.
23. Build a traceability chain from conclusion back to raw evidence.
24. Design a three-source triangulation plan using genuinely different evidence types.
25. Select five Words 1–25 and use them to evaluate an unfamiliar claim while explicitly calibrating uncertainty.

What Comes Next

Part 2 will teach Words 26–50 — Models, Measures & Counterfactual Explanation. The next deck moves from “How strong is the claim?” to “How was the world represented and measured?” through model, paradigm, schema, representation, proxy, indicator, metric, benchmark, baseline, parameter, confounder, mediator, moderator, counterfactual, hypothetical, precondition, necessity, dependency, asymmetry, nonlinearity, sensitivity, saturation, latent, manifest and attribution.

For progression, revisit the Secondary 2 Hero Vocabulary system, compare the separate Secondary 3 Intermediate bridge, review Vocabulary Learning with the Fencing Method, or return to the eduKate English Vocabulary Lists hub.

Part 2: Words 26–50 — Models, Measures & Counterfactual Explanation

Secondary 3 students now move from judging how strong a claim is to judging how the world has been represented, measured and explained. A model is not reality. A metric is not the thing being measured. A proxy can be useful without being equivalent to the construct. A correlation can remain after a confounder has created the appearance of a causal relationship. These distinctions matter whenever a text uses data, comparisons, rankings, surveys, diagrams, models or causal explanations.

The governing question for this deck is: “What representation or measurement stands between the world and the conclusion—and what would happen under a credible alternative?”

High-Value Fences for Deck 2

Fence	Boundary that matters
model / reality	A model is a selective representation built for a purpose; it should not be confused with the full system it represents.
paradigm / model	A model represents a specific system or relationship; a paradigm is a broader organising way of seeing what counts as a problem, method or explanation.
schema / framework	A schema is an organised mental or conceptual pattern; a framework is a more explicit structure for analysis or action.
representation / original	A representation stands for something else through words, images, symbols, data or models; it is not the original object or event.
proxy / indicator	A proxy substitutes for something difficult to measure directly; an indicator signals or reflects a condition without necessarily substituting for the whole construct.
metric / measure	A metric is a defined quantitative measure used for comparison or tracking; “measure” is the broader category.
benchmark / baseline	A baseline is the starting reference before change; a benchmark is an external or target comparison standard.
parameter / variable	A variable can vary across observations; a parameter defines or characterises a model, process or population and may be fixed within one analysis.
confounder / cause	A confounder is associated with both the proposed cause and outcome and can create or distort an apparent relationship.
mediator / moderator	A mediator helps explain how an effect travels; a moderator changes when, for whom or how strongly an effect occurs.
counterfactual / hypothetical	A hypothetical imagines a possible situation; a counterfactual specifically asks what would have happened if an actual condition or event had been different.
precondition / cause	A precondition must be in place for an outcome or process to occur but may not itself produce the outcome.
necessity / sufficiency	A necessary condition must be present; a sufficient condition is enough to produce the result under the stated model. The two are not the same.
dependency / correlation	Dependency means one state or process relies on another; correlation may exist without that functional reliance.
asymmetry / inequality	Asymmetry concerns directional or structural imbalance; simple numerical inequality does not necessarily imply an asymmetric relationship.
nonlinearity / irregularity	Nonlinearity means output does not change in a constant proportion to input; it can still follow a regular mathematical pattern.
sensitivity / instability	Sensitivity means outcomes respond strongly to changes in an input or assumption; instability means the system may fail to settle or remain controlled.
saturation / maximum	Saturation means additional input produces progressively little or no extra effect because capacity or response is approaching a limit.
latent / hidden	Latent refers to a variable or property not directly observed but inferred from indicators; “hidden” is broader and less technical.
manifest / latent	Manifest features are directly observable or expressed; latent features are inferred from what is observed.
attribution / association	Attribution assigns responsibility or causal contribution; association merely records that things occur together.

26. Model

Meaning: a simplified representation of a system, process or relationship used to describe, explain, predict or test
Fence: a model deliberately leaves things out; reality contains more detail than the representation
Word family: model, modelling, modelled  ·  Collocations: explanatory model / model of
Analytical job: makes simplification visible rather than treating diagrams or theories as the world itself
Example: A flowchart showing how information moves through a school is a model of communication, not a complete account of every conversation.
Countercase: A model can be useful even when it is not perfectly realistic if it preserves the relationships needed for the task.
Secondary transfer: Ask what the model includes, what it excludes and which question it is designed to answer.

27. Paradigm

Meaning: a broad organising pattern of assumptions, questions and methods that shapes how a field or problem is understood
Fence: a model answers within a chosen view; a paradigm influences which models and questions seem natural in the first place
Word family: paradigm, paradigmatic  ·  Collocations: dominant paradigm / paradigm shift
Analytical job: helps students see when disagreement is deeper than one disputed fact
Example: Treating learning mainly as information transmission and treating it mainly as active construction can produce different paradigms for teaching decisions.
Countercase: Every change in technique is not a paradigm shift.
Secondary transfer: Identify the assumptions that remain unchanged even when individual methods vary.

28. Schema

Meaning: an organised pattern of concepts or expectations that helps a person interpret and organise information
Fence: a schema can be an internal knowledge structure; a framework is usually a more explicit external structure for analysing a task
Word family: schema, schematic  ·  Collocations: conceptual schema / existing schema
Analytical job: explains how prior knowledge shapes interpretation
Example: A reader with a “mystery story” schema expects clues, suspects and later explanation, which changes what details become salient.
Countercase: Schemas help efficiency but can also cause misreading when a new case does not fit the familiar pattern.
Secondary transfer: Ask what prior expectation the reader is using and whether the text confirms or disrupts it.

29. Representation

Meaning: something that stands for, depicts or encodes another object, event, group, idea or relationship
Fence: representation is selective: a graph, quotation, photograph or summary presents some features and excludes others
Word family: represent, representation, representative  ·  Collocations: visual representation / representation of
Analytical job: keeps the student aware that evidence often arrives through a medium
Example: A bar chart represents attendance through counts and categories; it does not display the reasons students attended.
Countercase: Accurate representation does not require including every detail; it requires preserving what matters for the stated purpose.
Secondary transfer: Compare the representation with the underlying phenomenon and identify what information was compressed.

30. Proxy

Meaning: a measurable substitute used to stand in for something that is difficult, costly or impossible to observe directly
Fence: a proxy substitutes for a target construct; an indicator can simply signal that the construct may be present
Word family: proxy, proxies  ·  Collocations: proxy measure / use as a proxy
Analytical job: exposes the gap between what is measured and what the argument claims to know
Example: Number of books borrowed may be used as a proxy for reading engagement, but borrowing alone does not prove that the books were read carefully.
Countercase: A poor proxy can be measured perfectly and still represent the target badly.
Secondary transfer: Ask, “What is the real construct, and why is this substitute expected to track it?”

31. Indicator

Meaning: an observable sign or measure that provides information about a condition, state or trend
Fence: an indicator points towards a condition; it may be one component rather than a full substitute for the construct
Word family: indicate, indicator, indicative  ·  Collocations: key indicator / indicator of
Analytical job: allows multi-indicator reasoning instead of relying on one convenient measure
Example: Attendance, completion rate and voluntary participation can be separate indicators of engagement.
Countercase: One indicator can move for reasons unrelated to the underlying construct.
Secondary transfer: Combine several indicators and ask whether they converge before making the larger claim.

32. Metric

Meaning: a defined quantitative measure used to track, compare or evaluate performance or change
Fence: a metric is a formalised measure; not every observation has been converted into a metric
Word family: metric, metrics  ·  Collocations: performance metric / track a metric
Analytical job: forces the student to inspect how “success” has been operationalised
Example: Average response time is one metric for service speed, but it says nothing directly about accuracy or user satisfaction.
Countercase: Optimising one metric can worsen the larger system if the metric is mistaken for the whole goal.
Secondary transfer: Ask what behaviour the metric rewards and what important outcomes it ignores.

33. Benchmark

Meaning: a reference standard or comparison point used to judge performance, quality or progress
Fence: a benchmark is often an external target or comparator; a baseline is the starting state of the case being tracked
Word family: benchmark, benchmarking  ·  Collocations: benchmark against / performance benchmark
Analytical job: makes comparative judgement explicit
Example: A school may benchmark response times against a service standard of twenty-four hours.
Countercase: A prestigious benchmark can still be inappropriate if the compared contexts are not comparable.
Secondary transfer: Ask why this benchmark is relevant and whether its conditions match the current case.

34. Baseline

Meaning: an initial reference level used to judge later change
Fence: baseline answers “where did we start?”; benchmark answers “what are we comparing ourselves with?”
Word family: baseline  ·  Collocations: establish a baseline / baseline measure
Analytical job: prevents improvement claims without a credible before-state
Example: To evaluate whether reading time increased, the school first establishes a baseline from the previous term.
Countercase: If the baseline period was abnormal, later comparisons can mislead.
Secondary transfer: Check whether the baseline was measured under comparable conditions.

35. Parameter

Meaning: a quantity, condition or boundary that characterises a model, process or population and helps determine how it behaves
Fence: a variable changes across observations; a parameter may be fixed within one model while defining its behaviour or limits
Word family: parameter, parametric  ·  Collocations: model parameter / set the parameters
Analytical job: makes assumptions and constraints inside models visible
Example: In a simple queue model, service rate can be a parameter that strongly affects predicted waiting time.
Countercase: Everyday use sometimes treats “parameters” as general boundaries; in technical contexts the term may be more precise.
Secondary transfer: Ask which parameter values were assumed and how changing them alters the conclusion.

36. Confounder

Meaning: a factor related to both the proposed cause and the outcome that can create, hide or distort the apparent relationship between them
Fence: a third variable is any additional factor; a confounder specifically threatens interpretation of a causal relationship
Word family: confound, confounded, confounder  ·  Collocations: potential confounder / control for confounders
Analytical job: blocks premature causal attribution
Example: If students who attend extra lessons also tend to study more independently, independent study may confound the relationship between attendance and examination performance.
Countercase: A variable unrelated to either the proposed cause or outcome is not a meaningful confounder.
Secondary transfer: Ask what factor could influence both sides of the observed association.

37. Mediator

Meaning: a factor through which part of an effect or relationship operates
Fence: a mediator helps explain how X influences Y; it sits on the proposed pathway rather than merely confusing it
Word family: mediate, mediation, mediator  ·  Collocations: mediating factor / mediate the effect
Analytical job: converts “X affects Y” into a more complete mechanism
Example: If quieter study space improves task completion by reducing interruptions, interruption frequency may mediate part of the effect.
Countercase: A confounder is not a mediator when it precedes and independently influences both X and Y.
Secondary transfer: Complete the pathway: X → mediator → Y.

38. Moderator

Meaning: a factor that changes the strength, direction or conditions of a relationship between other variables
Fence: mediator explains how; moderator explains when, for whom or under which condition the relationship changes
Word family: moderate, moderation, moderator  ·  Collocations: moderating effect / moderator variable
Analytical job: prevents one-size-fits-all claims
Example: Task difficulty may moderate the effect of background noise: light noise may matter little for routine work but strongly affect complex reasoning.
Countercase: A moderator need not be caused by the original factor.
Secondary transfer: Ask whether the effect changes across groups, contexts or levels of another variable.

39. Counterfactual

Meaning: a reasoned comparison with what would have happened if an actual condition or event had been different
Fence: a hypothetical can imagine any possible scenario; a counterfactual changes something about the actual case to assess causal contribution
Word family: counterfactual, counterfactually  ·  Collocations: counterfactual scenario / counterfactual reasoning
Analytical job: is central to causal attribution
Example: To ask whether a revision programme caused improvement, consider the counterfactual: what would comparable students likely have achieved without the programme?
Countercase: Counterfactuals are not directly observable in the same case, so good design or comparison is needed to estimate them.
Secondary transfer: Write “If X had not occurred, what would we expect Y to have done?”

40. Hypothetical

Meaning: imagined or proposed for the purpose of reasoning rather than asserted as an actual event
Fence: hypothetical reasoning explores possibility; counterfactual reasoning modifies an actual or assumed factual situation
Word family: hypothesis, hypothesise, hypothetical, hypothetically  ·  Collocations: hypothetical scenario / hypothetically speaking
Analytical job: allows controlled thought experiments
Example: “Suppose a school had unlimited classroom space” is hypothetical even if no current school in the discussion has that condition.
Countercase: A hypothetical scenario can clarify principles but cannot substitute for empirical evidence about the real world.
Secondary transfer: State which assumption is intentionally fictional and why the scenario is useful.

41. Precondition

Meaning: a condition that must already be satisfied before a process, state or outcome can occur
Fence: a precondition enables possibility but may not be enough to produce the result
Word family: condition, precondition, preconditioned  ·  Collocations: necessary precondition / precondition for
Analytical job: separates enabling conditions from active causes
Example: Internet access is a precondition for joining a live online lesson, but access alone does not cause learning.
Countercase: If the outcome can occur without the condition, it is not genuinely a necessary precondition.
Secondary transfer: Ask whether the condition is required, helpful or merely common.

42. Necessity

Meaning: the state of being required for an outcome, rule or argument to hold
Fence: necessary does not mean sufficient: oxygen is necessary for ordinary human respiration but oxygen alone is not sufficient for healthy life
Word family: necessary, necessity, necessarily  ·  Collocations: necessity of / necessary condition
Analytical job: improves logical control of “must” claims
Example: Accurate evidence is necessary for a sound factual conclusion, but it may not be sufficient if the reasoning from evidence to conclusion is invalid.
Countercase: Something highly useful is not automatically necessary.
Secondary transfer: Test necessity by asking whether the outcome could still occur without the condition.

43. Dependency

Meaning: a relationship in which one state, process or outcome relies on another
Fence: correlation records co-variation; dependency implies some functional, logical or conditional reliance
Word family: depend, dependent, dependency  ·  Collocations: dependency on / dependency relationship
Analytical job: makes hidden reliance explicit
Example: A digital submission system has a dependency on network availability if submissions cannot be completed when the network fails.
Countercase: Two variables can be correlated without one depending on the other.
Secondary transfer: Ask what breaks if the supporting element is removed.

44. Asymmetry

Meaning: an imbalance in direction, power, information, effect or structure between related parts
Fence: inequality concerns unequal quantity; asymmetry concerns how the relationship itself differs across directions or sides
Word family: symmetry, asymmetric, asymmetry  ·  Collocations: information asymmetry / asymmetric relationship
Analytical job: exposes relationships that look reciprocal but are not equivalent
Example: A platform may know detailed user behaviour while users know little about how the platform ranks content; this creates information asymmetry.
Countercase: Different quantities do not automatically create important asymmetry if the relationship remains functionally balanced.
Secondary transfer: Compare what each side can know, control, risk or influence.

45. Nonlinearity

Meaning: a relationship in which equal changes in input do not produce equal proportional changes in output
Fence: nonlinear does not mean random; the relationship may be highly regular but curved, thresholded or saturating
Word family: linear, nonlinear, nonlinearity  ·  Collocations: nonlinear relationship / nonlinear effect
Analytical job: challenges simplistic “twice the input means twice the effect” assumptions
Example: Adding one student to an empty room may have little effect on noise, while adding one student to an already crowded room may produce a much larger disruption.
Countercase: A straight proportional relationship is linear even if the numbers are large.
Secondary transfer: Plot or describe how marginal effects change as the input increases.

46. Sensitivity

Meaning: the degree to which an outcome, model or conclusion changes when an input, assumption or parameter changes
Fence: sensitivity is responsiveness; instability concerns inability to remain controlled or settle
Word family: sensitive, sensitivity  ·  Collocations: sensitivity to / sensitivity analysis
Analytical job: tests whether conclusions depend heavily on one assumption
Example: If a ranking changes completely when one criterion weight moves from 20% to 25%, the result is highly sensitive to that weighting.
Countercase: A sensitive measurement can be desirable if small real changes need to be detected.
Secondary transfer: Vary one assumption at a time and observe whether the conclusion remains robust.

47. Saturation

Meaning: a state in which additional input produces progressively smaller additional effect because capacity, response or available opportunity is approaching a limit
Fence: maximum is the final upper bound; saturation describes the flattening process as further input yields diminishing change
Word family: saturate, saturated, saturation  ·  Collocations: reach saturation / saturation point
Analytical job: explains diminishing returns
Example: Adding revision hours may initially help strongly, but after fatigue accumulates, learning gains can approach saturation.
Countercase: A plateau can result from measurement limits rather than true saturation of the underlying process.
Secondary transfer: Ask whether the response is flattening because of capacity, fatigue, scarcity or the measurement itself.

48. Latent

Meaning: present but not directly observed, often inferred from patterns in observable indicators
Fence: latent does not simply mean secret; it refers to an underlying construct or state inferred from manifestations
Word family: latent, latency  ·  Collocations: latent variable / latent factor
Analytical job: keeps constructs separate from the indicators used to estimate them
Example: “Motivation” can be treated as latent when inferred from attendance, effort, persistence and self-report rather than directly observed as one object.
Countercase: An unobserved variable should not be invented merely to save a weak explanation.
Secondary transfer: Name the observable indicators and explain why they are expected to reflect the latent construct.

49. Manifest

Meaning: directly observable, expressed or evident rather than inferred as an underlying state
Fence: manifest variables are measured or observed directly; latent variables are inferred from them
Word family: manifest, manifestation  ·  Collocations: manifest variable / manifest behaviour
Analytical job: separates what is seen from what is inferred
Example: Number of completed tasks is manifest; underlying persistence may be a latent construct inferred partly from that behaviour.
Countercase: Observable behaviour still needs interpretation before it can establish motive.
Secondary transfer: Make two columns: manifest evidence and latent interpretation.

50. Attribution

Meaning: the assignment of cause, responsibility, credit or explanation to a person, factor or process
Fence: association observes a relationship; attribution states that one factor deserves causal or explanatory weight
Word family: attribute, attribution, attributable  ·  Collocations: causal attribution / attribute to
Analytical job: is the final move that should come only after confounders, mechanisms and counterfactuals are considered
Example: Attributing improved scores entirely to a new programme is unsafe if students also received more teaching time and the assessment became easier.
Countercase: Attribution can be partial: several factors may each contribute.
Secondary transfer: State whether attribution is direct, partial, probable or indeterminate—and why.

Model Critique Ladder: Representation → Measurement → Explanation

Stage	Question	Failure to watch for
1 — Target	What real thing or construct are we trying to understand?	unclear construct
2 — Representation	How is it represented?	model mistaken for reality
3 — Measure	What proxy, indicator or metric is used?	metric substituted for goal
4 — Reference	What baseline or benchmark defines comparison?	unfair comparator
5 — Alternatives	What confounders or rival models exist?	single-cause attribution
6 — Mechanism	Is there a mediator or causal pathway?	association treated as explanation
7 — Conditions	What moderates the relationship?	one-size-fits-all claim
8 — Counterfactual	What would happen if the proposed cause were absent?	no comparison state
9 — Sensitivity	Does the conclusion survive altered assumptions?	fragile result
10 — Attribution	How much causal weight is warranted?	overclaiming certainty

Pressure Test 1: Model Is Not Reality

A school map shows every classroom as an equal-sized rectangle.

The map can be a useful model for navigation while being a poor representation of actual room size. A model should be evaluated against its purpose, not against an impossible demand to preserve every feature of reality.

Pressure Test 2: Proxy, Indicator or Metric?

Measure	Best concept	Why
books borrowed used to stand in for reading engagement	proxy	substitute for a harder-to-observe construct
voluntary attendance as one sign of engagement	indicator	signals one aspect
percentage of assignments submitted on time	metric	defined quantitative measure

Pressure Test 3: Benchmark vs Baseline

• Baseline: average response time was 42 hours before the intervention.
• Benchmark: the service target is 24 hours.
• Post-change result: average response time is 28 hours.

The system improved relative to its baseline but still falls short of the benchmark. These are different judgements.

Pressure Test 4: Confounder, Mediator or Moderator?

Students who join a revision programme tend to improve more.

• Confounder: students who join may already be more motivated, and motivation also predicts improvement.
• Mediator: the programme may increase retrieval practice, which then contributes to improvement.
• Moderator: the effect may be larger for students with weak prior knowledge than for students already near mastery.

Pressure Test 5: Hypothetical vs Counterfactual

Question	Type
“Suppose every student had unlimited study time. What might happen?”	hypothetical
“The programme ran last term. What would likely have happened to the same students without it?”	counterfactual

The counterfactual is harder because the alternative world cannot be directly observed for the same case. Good comparisons, controls or careful causal reasoning are therefore needed.

Pressure Test 6: Necessary Is Not Sufficient

A valid password may be necessary to enter an account, but it may not be sufficient if two-factor authentication is also required. Students should treat “required” and “enough” as separate logical jobs.

Pressure Test 7: Nonlinearity, Sensitivity and Saturation

Suppose revision time increases from one to two hours and scores rise substantially, but increases from six to seven hours produce almost no gain.

• The relationship may be nonlinear.
• The system may be approaching saturation at higher study durations.
• If the conclusion changes dramatically depending on whether one extreme student is included, the analysis may be highly sensitive to that case.

Pressure Test 8: Latent vs Manifest

Observation	Role
student completes optional practice	manifest behaviour
student asks for feedback voluntarily	manifest behaviour
underlying motivation inferred from several behaviours	latent construct

The crucial mistake is to write as though the latent construct were directly observed.

Counterfactual Attribution Drill

After a school launches a new revision platform, average scores rise from 62 to 70. The same cohort also receives an additional weekly lesson, the examination paper is slightly easier, and participation in the platform is voluntary.

1. What is the baseline?
2. What outcome metric is being used?
3. What construct is that metric intended to represent?
4. Name two potential confounders.
5. What counterfactual would be needed to estimate the platform’s causal effect?
6. Could self-selection into the platform create an asymmetry between users and non-users?
7. Suggest one plausible mediator.
8. Suggest one possible moderator.
9. What evidence would justify partial rather than total causal attribution?

Model Comparison Drill: Same Evidence, Different Explanations

A student’s marks improve sharply after joining a study group.

Model	Mechanism	Discriminating evidence
peer-explanation model	explaining concepts improves retrieval and organisation	measure explanation activity
motivation-selection model	already motivated students are more likely to join	compare prior motivation/baseline behaviour
time-on-task model	group simply adds more study time	control total study hours

Secondary 3 reasoning improves when the student asks which observation would discriminate between models instead of selecting the first plausible story.

Unseen Informational Text: Measure the Measure

“The school’s new well-being initiative is successful because visits to the counselling centre fell by 25%. The decline began two months after the programme started.”

1. What construct is “well-being”?
2. Are counselling visits a proxy, an indicator or the construct itself?
3. What alternative interpretation could explain fewer visits?
4. What baseline is needed?
5. What benchmark, if any, would be meaningful?
6. Name one potential confounder.
7. What mechanism would connect the initiative to lower counselling demand?
8. What counterfactual comparison would strengthen causal attribution?
9. Could reduced visits actually represent worse access? Explain the model dependence of the conclusion.

Writing Transfer: Replace Measurement Shortcuts with Precise Claims

Weak claim	Secondary 3 upgrade
“More likes prove the campaign was effective.”	Treat likes as an indicator or proxy and specify what kind of effectiveness is being measured.
“The new method caused the improvement.”	Identify confounders, mediator, counterfactual and degree of attribution.
“The graph shows reality.”	Describe the graph as a representation built from selected variables and scales.
“The score doubled, so the effect doubled.”	Check whether the relationship is linear.
“Motivation increased.”	Separate the latent construct from the manifest indicators used to infer it.

Morphology & Model Network

Core word	Network	Useful question
model	model, modelling	What is simplified or omitted?
paradigm	paradigmatic	What assumptions define the problem?
schema	schematic	What prior structure shapes interpretation?
representation	represent, representative	What stands in for the original?
indicator	indicate, indicative	What condition does the measure signal?
metric	metrics	What behaviour does the metric reward?
confounder	confound, confounded	What could influence both X and Y?
mediator	mediate, mediation	Through what pathway does the effect travel?
moderator	moderate, moderation	When or for whom does the effect change?
counterfactual	counterfactually	What would likely happen without X?
dependency	depend, dependent	What stops working if this element disappears?
asymmetry	asymmetric	How do the two directions differ?
nonlinearity	nonlinear	Does equal input change produce equal output change?
sensitivity	sensitive	Which assumption has disproportionate influence?
saturation	saturate, saturated	Where do additional inputs yield little gain?
latent	latency	Which observed indicators support the inferred construct?
manifest	manifestation	What is directly observable?
attribution	attribute, attributable	How much causal weight is actually warranted?

Misuse Clinic: Models, Measures & Counterfactuals

1. A model is a selective representation, not reality itself.
2. Do not call every methodological change a paradigm shift.
3. A schema can help interpretation and also bias it.
4. Representations can be accurate for one purpose and incomplete for another.
5. A perfectly measured proxy can still represent the wrong construct.
6. One indicator rarely captures a multidimensional construct completely.
7. Metrics can be gamed when they become targets.
8. Baseline and benchmark answer different comparison questions.
9. Parameter and variable are not always interchangeable.
10. A confounder threatens causal interpretation; it is not just any additional variable.
11. A mediator explains pathway; a moderator changes strength or conditions.
12. A hypothetical scenario is not automatically a counterfactual.
13. A precondition may be necessary without being sufficient.
14. Necessary does not mean enough.
15. Dependency implies reliance, not mere co-occurrence.
16. Asymmetry concerns the relationship, not merely unequal numbers.
17. Nonlinear does not mean random.
18. High sensitivity can make a conclusion fragile even when calculations are correct.
19. A plateau may reflect saturation or a measurement ceiling; distinguish them.
20. Latent constructs are inferred, not directly observed.
21. Manifest behaviour does not establish underlying motive by itself.
22. Attribution should follow—not precede—consideration of confounders and counterfactuals.
23. Partial causal attribution is often more accurate than total attribution.
24. Changing the representation can change what appears important without changing the underlying event.
25. Good models expose their assumptions and limits.

Retrieval Checkpoint 2 — No Notes

1. Build a simple model and state two things it deliberately omits.
2. Explain the difference between a model and a paradigm.
3. Give a schema that could help and then mislead a reader.
4. Turn one event into two different representations and compare what each foregrounds.
5. Choose a proxy for a difficult construct and state its weakness.
6. Give three indicators for the same latent construct.
7. Design one metric and explain how it could distort behaviour if overused.
8. Give a baseline and a separate benchmark for the same problem.
9. State one parameter that would materially change a simple model.
10. Create a causal claim and add a plausible confounder.
11. Create a mediator that explains the pathway from X to Y.
12. Create a moderator that changes the effect across contexts.
13. Write one hypothetical scenario and one genuine counterfactual about the same topic.
14. Give a precondition that is necessary but clearly not sufficient.
15. Give one necessary condition and one sufficient condition for a defined outcome.
16. Create a dependency and state what fails when the supporting element is removed.
17. Describe an asymmetric relationship in information, power or risk.
18. Create a nonlinear relationship and explain why it is not random.
19. Give a conclusion that is highly sensitive to one assumption.
20. Create a saturation pattern and state what resource or response is approaching its limit.
21. Name one latent construct and three manifest indicators.
22. Give one manifest observation that supports two competing latent interpretations.
23. Take a simple association and show why causal attribution remains uncertain.
24. Use a counterfactual to improve an attribution claim.
25. Select five Words 26–50 and use them to compare two competing explanations of an unfamiliar pattern.

What Comes Next

Part 3 will teach Words 51–75 — Argument Architecture, Discourse & Contestation. Students will move from evaluating models to examining the hidden structure of whole arguments through thesis, warrant, concession, reservation, refutation, fallacy, equivocation, circularity, syllogism, subtext, discourse, ideology, agenda, polemic, dogmatic, didactic, coercive, propagandistic, intertextuality, contextualise, problematise, interrogate, reframe, privilege and contestable.