Reconstruction and charity: the strongest version first

Someone you know posts an argument you think is wrong. You read it once, fast, and start typing. Twenty minutes later they reply that you've answered something they never said, and they're right. You had a good objection to a bad argument. It wasn't their argument.

This happens to careful people all the time, and it happens for a reason. An argument in the wild is a compressed thing. The premises are half-stated, the general rule that carries the step is left out, and the words around the reasons ("obviously", "probably", "although") are doing work that is easy to miss. Before you can test an argument you have to rebuild it, and the rebuilding is where most evaluations go wrong.

This lesson is about rebuilding well. It fills in step 5 of the six-step check you've been carrying since lesson 1, and with it the method is complete: find the conclusion, find the premises, supply what's missing, test the link, test the premises, and only then consult your opinion of the conclusion. Everything in the lesson serves one rule: evaluate the strongest version of what the other person actually holds, because that's the only version worth refuting.

Read the small words first

Before you rewrite an argument, read it once for the words around the reasons. Walter Sinnott-Armstrong and Robert Fogelin call this close analysis, and it comes before the deep analysis of putting things in standard form.⁵ Matthew Van Cleave, whose open textbook follows the same method, sorts the words into three kinds in section 1.10.⁴

Assuring terms tell you that reasons exist without giving them. "Clearly", "obviously", "everyone agrees", "studies show", "I can assure you". Van Cleave's definition: assuring is "informing someone that there are further reasons although one is not giving them now."⁴ One of his exercise sentences shows the shape: "Most pediatricians agree that the single most important factor in childhood obesity is eating sugary, processed foods."⁴ Whether or not that's true, notice what the sentence gives you: not a study, not a number, but a report that agreement exists. Van Cleave's advice is that assurances aren't necessarily illegitimate, "but a critical thinker will always regard them as somewhat suspect."⁴ The practical version is sharper. Writers often reach for an assurance exactly where they have no evidence to hand. So the premise under "obviously" is the one to check first.

Guarding terms weaken a claim so it's harder to refute. "Probably", "in most cases", "some", "it may be that". Van Cleave: guarding is "weakening one's claims so that it is harder to show that the claims are false."⁴ He gives a ladder: "All U.S. Presidents were monogamous" is a strong claim and easy to test; "Some U.S. Presidents were monogamous" is weak and nearly impossible to deny.⁴ The thing to notice is that a guarded claim is usually the honest one. A writer who says "probably" is telling you exactly how far they'll go. Your job is to hold them to the guarded claim, not to the bold one you might have preferred to attack, and also not to let the conclusion quietly outrun the guard. "Probably stale" does not license "throw it away without looking."

Discounting terms anticipate an objection and wave it through. "Although", "even though", "but", "nevertheless", "still". Van Cleave: discounting is "anticipating objections that might be raised to one's claim or argument as a way of dismissing those objections."⁴ His example is "The worker was inefficient, but honest."⁴ The point after "but" is the one the writer wants you to keep; the point before it is the one they've decided you'll set aside. Sometimes that's fair. Sometimes the objection they've discounted is the strongest thing in the passage, and the "although" is the only sign it was ever there.

None of this is evaluation yet. It's a map of where the writer feels strong and where they feel weak, drawn by the writer. Assurances mark the weak spots. Guards mark the claims they've measured. Discounts mark the objections they know about. Read the map, then rebuild.

Linked or convergent: does the argument survive a lost premise?

Take two everyday arguments, both plainly made up.

"The flat is ten minutes from her office, it's under budget, and it has a garden. She should take it."

"The last train leaves at 22:10 and the trip takes fifty minutes. She'll be home by 23:00."

Now suppose in each case that the second premise is false. The flat is over budget. The trip takes ninety minutes, not fifty.

The flat argument has convergent premises: each one is a separate reason for the conclusion, and each would count for something on its own. The train argument has linked premises: they support the conclusion only together, and neither does anything alone. The Stanford Encyclopedia's entry on informal logic makes the distinction the basic move in argument diagramming: convergent premises are drawn as separate arrows into the conclusion, and linked premises are joined with a plus sign to show "that they support the argument's conclusion (only) when combined."² The drawing below does exactly that for the two arguments you just read. You've met the pure convergent case before, too: the conductive arguments of lesson 5, where independent reasons are weighed against counter-considerations, are convergent support with the other side of the ledger written in.

The reason to tell them apart is that the whole point of evaluating premises one at a time (which is what step 5 does) is to know what happens to the argument when one of them fails. If the premises are linked, a single false premise sinks the conclusion. If they're convergent, a single false premise costs the argument one reason and leaves the rest. So a diagram is a prediction about damage. It tells you, before you've tested anything, which premises are load-bearing and which are extra weight.

Most real arguments mix the two. The flat argument, written out fully, would probably have a linked pair hiding inside it: "it's under budget" only counts because "[she can't afford anything over budget]", and those two go together. You'll see a mixed case in the worked example below.

Toulmin's layout: the argument seen from the arguer's side

In lesson 1 you met Harry. "Harry was born in Bermuda, so Harry is a British subject." You found the missing premise, a general rule about people born in Bermuda, and you saw that writing it down made its exceptions visible. Now here's the full machine that example came from.

Stephen Toulmin, in The Uses of Argument (1958), thought the premise-and-conclusion layout hid too much, and proposed six slots instead.¹ Laid out on Harry:

• Claim: Harry is a British subject.
• Data (Toulmin's later textbook with Rieke and Janik says grounds): Harry was born in Bermuda.
• Warrant: "A man born in Bermuda will generally be a British subject." This is the rule that licenses the step from the data to the claim. "Propositions of this kind I shall call warrants", Toulmin writes; the later textbook's gloss is "a general, step-authorizing statement."¹³
• Backing: the statutes and legal provisions on British nationality, which are why the warrant holds.
• Qualifier: "presumably". The warrant says "generally", so the claim can only be made with that strength.
• Rebuttal: unless both his parents were aliens, or unless he has since become a naturalised American. The circumstances under which the warrant doesn't apply.¹

Here are the six slots drawn out. Lesson 1's drawing showed four of them; this is the whole machine.

Now write the same argument in the standard form you've been using since lesson 1:

1. Harry was born in Bermuda.
2. [A man born in Bermuda will generally be a British subject.]
---------------------------------------------------------------
C: Harry is (presumably) a British subject.

Look at what happened. The warrant is the bracketed premise. The two layouts are one argument viewed twice, so the second view has to earn its place.

David Hitchcock, in a careful paper on Toulmin's warrants, gives the answer. "A warrant is an inference-licensing rule, not a premiss": a rule the arguer is relying on, not a fact they're asserting.³ Seeing it as a rule does two things a premise list doesn't. First, a rule has a scope, so it invites the question "does this rule apply here?", which is the rebuttal slot. Second, rules can belong to fields. Toulmin held that warrants are field-dependent: the rule that licenses a step in law ("born in Bermuda, so a British subject") is nothing like the rule that licenses a step in medicine ("this test is positive, so probably this disease"), and a step that's fine in one field can be worthless in another.³ When someone argues from a courtroom rule to a medical conclusion, the layout shows the mismatch at once. Hitchcock adds a caution that Toulmin's picture needs: plenty of everyday warrants belong to no organised field at all. They're rough generalisations from common experience, like the fifteen-year rule you're about to meet, and for those there's no professional body to check the rule against; there's only the work of writing it down and asking how far it reaches.³

This is the reason the layout finds weaknesses a premise list hides. The rebuttal slot has to be filled. Writing "unless" and then stopping to think is how you find that Harry's parents might have been diplomats, that the bread might have been in a sealed bag, that the doctor might be a dermatologist being asked about a heart.

Worked example with a wrinkle: the boiler

Here's a message of the kind a person might send to the family group chat. It's made up, so nothing in it needs to be true, and that's the point: you're testing the argument, not the boiler.

Obviously the boiler needs replacing. It's fourteen years old, and every engineer will tell you they don't last past fifteen. It's been serviced twice this year already, which is probably more than a new one would need in five. And although a new one costs about £2,500, the savings on gas alone would pay that back within a few years. So we should replace it this autumn rather than wait for it to go.

Close analysis first. "Obviously" is an assuring term on the conclusion. "Every engineer will tell you" is an assuring term on the fifteen-year rule, and that rule is the premise with no evidence attached. "Probably", "about", and "within a few years" are guards, and they look honest: the writer is measuring what they're claiming about the servicing and the payback. "Although a new one costs about £2,500" is a discount: the cost is mentioned so it can be set aside.

Now the reconstruction. The conclusion is the last sentence. The stated premises are the age, the fifteen-year rule, the servicing, and the payback. Put them in standard form:

1. The boiler is fourteen years old.
2. Boilers don't last past fifteen years.
3. It has been serviced twice this year, which is probably more than a new one would need in five.
4. A new boiler costs about £2,500 and would pay that back in gas savings within a few years.
--------------------------------------------------------------------------------------------------
C: We should replace the boiler this autumn rather than wait for it to fail.

Premises 1 and 2 are linked. The age means nothing without the rule and the rule means nothing without the age; together they support an intermediate conclusion, "the boiler will fail within about a year". Premises 3 and 4 are convergent: each is a separate reason for replacing, and each would still count if the other fell.

Now the gap. Something is missing between "it'll fail within a year" and "replace it this autumn rather than wait." Write down what the writer needs to be true that they haven't said. Do it before you open the box.

So the full reconstruction adds:

5. [Replacing a boiler before it fails is cheaper, or less disruptive, than replacing it after.]

with 1, 2 and 5 linked, and 3 and 4 convergent alongside them. Here's the diagram, and notice what it predicts before any premise has been tested: the chain on the left fails as one piece, while each arrow on the right stands or falls alone.

Now the evaluation, which is the next section's subject, done here in preview. The link: this is an inductive argument (you can tell from the guards), and given the premises it's reasonably strong. The premises: 1 is a matter of record. 3 and 4 are guarded and plausible; a reasonable person would accept them provisionally and might ask to see the gas bills. Premise 2 is the problem. It's the one under "every engineer will tell you", and as stated it is almost certainly false: boilers don't have a fixed lifespan, and "don't last past fifteen" is the kind of round number people quote rather than measure. Weaken it to what an engineer might actually say ("older boilers fail more often and cost more to repair") and it still supports the intermediate conclusion, but only weakly, and premise 5 now carries more weight than the writer noticed.

The verdict is not "wrong". The argument is alive on its convergent legs (3 and 4) even if the linked chain (1, 2, 5) is weaker than it looked. A fair reply is "the fifteen-year rule isn't a rule, but the servicing and the savings might be enough on their own; can we see the bills?" That reply refutes the version the writer actually holds, keeps the parts that work, and moves the conversation forward. The reply "boilers don't have expiry dates, so no" refutes one premise and pretends the other three don't exist.

Charity: reconstruct the version they'd sign

The principle of charity is old and plain. When a passage can be read more than one way, read it the way the arguer would accept. When you supply a missing premise, supply the one they'd agree to, not the one that's easiest to knock down and not the one that makes the argument valid at the price of being absurd.

That last case is the one people get wrong most. Take the café: "It's closed on Mondays, so let's go Tuesday." You could make that valid by adding "[Tuesday is the only day it opens]". Nobody believes that. You could add "[Tuesday is the next day that suits us both]", which makes the argument strong rather than valid and is exactly what the speaker meant. Charity says take the second. Lesson 1 gave you the rule as "the most plausible thing the arguer would accept"; here's the reason behind it. If you reconstruct an argument nobody holds, your refutation refutes nobody. You've done the work and nothing has changed, because the argument you demolished was never in the room.

So charity isn't politeness. It's accuracy about the object you're evaluating. A reconstruction that the arguer wouldn't sign is a wrong answer to the question "what is this person's argument?", in the same way that a wrong sum is a wrong answer, and the fact that it was uncharitable is just the way it went wrong.

The name has a history worth a paragraph, because it says how deep the idea goes. "Principle of charity" was coined in 1959 by the philosopher Neil Wilson, and it entered philosophy not as advice about arguing but as a claim about understanding language at all. W. V. O. Quine picked it up for his theory of translation, and Donald Davidson made it central: on his account you cannot interpret another speaker in the first place except by reading them as mostly rational and mostly right, because a translation that makes someone spout nonsense is better evidence of a bad translation than of a nonsensical speaker.⁷ The informal-logic tradition borrowed the name for the humbler job in this lesson, and the borrowing is apt. The uncharitable reconstruction fails the same way the uncharitable translation does: it tells you about the interpreter, not the speaker.

Internet usage has a word for the charitable version: the steelman, by contrast with the straw man. It's a useful word with no canonical source; it appeared in the early 2010s and the older academic name is the principle of charity. Use either. But hold on to what it means. A steelman is the version the best defender of the view holds. It is not a version nobody holds. If you build a magnificent argument for a position and no one who holds the position would recognise it, you haven't steelmanned them; you've built your own argument and put their name on it, which is a straw man with better manners.

Rapoport's rules

The best-known statement of how to do this comes from the philosopher Daniel Dennett, in Intuition Pumps and Other Tools for Thinking (2013). He credits the social psychologist and game theorist Anatol Rapoport, and he presents the rules as the antidote to the habit of caricaturing an opponent. Under the heading "How to compose a successful critical commentary," Dennett writes:⁶

1. "You should attempt to re-express your target's position so clearly, vividly, and fairly that your target says, 'Thanks, I wish I'd thought of putting it that way.'"
2. "You should list any points of agreement (especially if they are not matters of general or widespread agreement)."
3. "You should mention anything you have learned from your target."
4. "Only then are you permitted to say so much as a word of rebuttal or criticism."

Dennett's stated reason is practical: following the rules turns your opponent into a more receptive audience for your criticism.⁶ There's a second reason that matters more for this course. You can't do rule 1 without doing the reconstruction properly. To restate someone's position so well that they'd thank you, you have to have found their conclusion, their premises, the rule they're relying on, and the strength they're claiming. Rule 1 is the whole of this lesson, tested by the one judge who knows the answer.

Let me model it on one of the disagreements from the list below, the family dog. Suppose I'm against getting one and my sister is for it. Her case, restated by me under rule 1:

A dog would get the children outdoors every day, which nothing else we've tried has managed. It would give the youngest, who's anxious, something to care for that depends on her. We've both said we want the house to feel more like a home than a timetable, and a dog does that in a way a plan can't. The costs are real, but they're known and steady, which is the kind of cost we handle well. And if we don't do it while the children are this age, we won't do it at all.

Rule 2: I agree that the children are indoors too much and that our previous fixes have failed. Rule 3: I hadn't thought about the youngest and responsibility; that's a real point and I learned it from her. Rule 4, only now: I think the daily walking will land on one adult, not on the children, and the case depends on the children doing it. That objection is aimed at the version she actually holds, and she'll know it.

Compare what I'd have said without the rules. "Dogs are a lot of work and the kids will lose interest in a week." True, possibly, and useless, because she never claimed the kids wouldn't lose interest; her case rests on something else.

Step 5: the acceptability, relevance, and grounds conditions

You now have an argument reconstructed in the version its author would sign, diagrammed, with its warrant found. Time to test it. Lessons 2 through 6 gave you the tests for the link. The informal-logic tradition, following Ralph Johnson and Anthony Blair, adds the tests for the premises and puts the whole thing in one frame.² This course spells the three conditions ARG, for acceptable, relevant, grounds, and calls an argument cogent when its premises are:

(A) Acceptable. A reasonable person in the arguer's audience could accept them without further argument; the Stanford entry's compressed version is that premises are "acceptable when they are true or acceptable in some other way."² In practice, ask where the premise comes from. It can stand without more argument when it's common knowledge, when it rests on testimony you have no reason to doubt, or when it comes from an authority who is actually an authority on the point, and lesson 8 gives you six questions for that last case.⁸ Note what acceptability is not. It isn't "I accept it", and it isn't "it's true". A premise can be true and unacceptable (it needs support the arguer hasn't given), and a premise can be acceptable and turn out false. The condition is about whether the arguer has earned the premise for this audience.

(R) Relevant. Each premise gives some support to the conclusion. The Stanford entry's wording is that premises are "relevant when they provide some (i.e. any) support for the conclusion of the argument."² That sounds like a low bar, and it is one, but it's the condition that real arguments fail most often and the hardest to show. When a premise is irrelevant, it usually feels relevant, because it's true and it's about the same subject. "He's been late three times this month" is relevant to whether he should keep his job and irrelevant to whether his budget figures are right, and the sentence carries no marker telling you which question is on the table.

(G) Adequate grounds. The premises, taken together, give the conclusion enough support: the link is valid, if a deductive link was claimed, or strong enough, if an inductive one was. This is the test you already own. Lessons 2 to 4 are the grounds condition for deduction; lessons 5 and 6 are the grounds condition for induction, with the likelihood ratio as one way to say how much a premise supports a conclusion.

Johnson and Blair's own name for the third condition is sufficiency, so you'll also see the letters ARS; a family of textbooks says "adequate grounds" instead, and this course keeps the G because it points back at the link tests you already own.² The letters don't matter. What matters is that the three conditions are separate, so that an argument can fail exactly one and you can say which.

Now you have the whole six-step check. Find the conclusion. Find the premises. Supply what's missing, charitably. Test the link (grounds). Test the premises (acceptability, and relevance to this conclusion). Only then consult your opinion of the conclusion. Lesson 8 will hand you a list of fallacy names, and every one of them is a shortcut to a failure of one of these conditions. You'll be able to say which, which is more than the name tells you.

What people get wrong

"Reconstruction means putting it in my words." It means putting it in theirs, made explicit. Your words are fine for the bracketed premises, as long as the arguer would sign them. The moment the reconstruction contains something they'd deny, it's no longer a reconstruction of their argument.

"Charity means agreeing." It means evaluating the right object. You can restate someone's case so well that they thank you and then show, on the acceptability condition, that its central premise has no support. That's not a contradiction; it's the order Dennett's fourth rule requires.

"A steelman is the strongest argument I can imagine for the position." It's the strongest argument the position's best defenders actually make. If nobody holds your improved version, you're arguing with yourself. Ask the question the first rule asks: would they say "I wish I'd put it that way"? If they'd say "that's not what I think", start again.

"Relevance is obvious." It's the condition most real arguments fail, and showing the failure takes work, because an irrelevant premise is usually a true one on the same subject. The test is to state the conclusion precisely and then ask what this premise adds to it. If the honest answer is "nothing, but it's true", you've found the failure.

"Acceptable means I accept it." It means a reasonable member of the audience the arguer is addressing could accept it without more argument. Your acceptance is neither necessary nor sufficient. This is also where your opinion of the conclusion sneaks back in early: you'll find yourself calling premises unacceptable because you dislike where they lead. Step 6 exists to stop that; keep it in sixth place.

"Guarding is weaselling." A guard is a measurement. "Probably" says how strong the claim is. Hold the arguer to the guarded claim and don't let them cash it later as the bold one, but don't punish the honest writer for telling you the strength. The suspect words are the assurances, not the guards.

Practice

Now close the page and write down, from memory: the three kinds of terms close analysis looks for; the difference between linked and convergent support and why it matters; Toulmin's six slots; Rapoport's four rules; and the three ARG conditions, with one sentence on what each asks. Then check. Reread only what you missed.

Connections

This lesson completes the six-step check. Steps 1 to 3 came from lesson 1, and the missing premise you learned to bracket there is what Toulmin calls a warrant. Step 4 is the grounds condition, which is lessons 2 to 4 for deductive links and lessons 5 and 6 for inductive ones; the likelihood ratio from lesson 6 is one way to put a number on how much a premise supports a conclusion. Step 5 is the acceptability and relevance conditions you met today. Step 6 is where belief bias, from lesson 2, waits for you, and the only defence is to keep it sixth.

Lesson 8 gives you the fallacy names. It comes after this one on purpose. A fallacy name is a hypothesis that an argument fails one of the ARG conditions, and you now have the tests that decide the hypothesis. Lesson 9 turns the method on the news, where the assuring terms come from sources and the missing premises come from editors. Lesson 10 turns Rapoport's rules round: you'll steelman your own opposition before you write, so that your argument has already met its best objection by the time a reader does.

One thing this lesson didn't settle. The whole course evaluates arguments as products: a set of premises, a conclusion, a link, judged on their merits. That's the logical approach, and it isn't the only one. The Stanford entry on informal logic describes two others. Rhetoric "understands arguing as an attempt to persuade an audience", and dialectics "understands an argument as an attempt to make a move in an exchange between interlocutors with opposing points of view."² Each approach sees things the others miss: rhetoric explains why a valid argument can fail to move anyone, and dialectics explains why the same argument can be fair in one exchange and a foul in another. People who study argument for a living disagree about which approach is basic, and some try to combine them. This course picks the logical one because it's the one a single reader can run alone on a page. If you go further, you'll meet the other two, and the "Go deeper" list points at the map.

Go deeper

• Think Again I: How to Understand Arguments (Duke, Coursera, free to audit), Walter Sinnott-Armstrong and Ram Neta. The video companion to this lesson: close analysis and deep analysis, taught by the people who named them, with graded practice passages.
• Daniel Dennett, Intuition Pumps and Other Tools for Thinking (Norton, 2013), chapter 3, "Rapoport's Rules". Three pages, and the surrounding chapters are short essays on other thinking tools, several of which belong in this course.
• Matthew Van Cleave, Introduction to Logic and Critical Thinking, 2nd ed. (open textbook, CC BY 4.0), chapter 1, sections 1.9 to 1.12 on LibreTexts. Missing premises, the assuring/guarding/discounting toolkit, and a full reconstruction of a real speech, longer than anything here.
• David Hitchcock, "Toulmin's Warrants", in Anyone Who Has a View (Kluwer, 2003; from the ISSA 2002 proceedings). The paper this lesson's account of the warrant comes from; it's the clearest statement of why a warrant is a licence and not a premise, and it says where Toulmin's field idea holds and where it doesn't.
• The Stanford Encyclopedia of Philosophy, "Informal Logic" (Leo Groarke), sections 3 to 5. Diagramming, the ARS criteria, and the map of the logical, dialectical, and rhetorical approaches, for readers who want to know what this course chose not to do.

Sources

1. Toulmin, S., The Uses of Argument (Cambridge University Press, 1958; updated ed. 2003), pp. 97 to 102 in the 1958 edition. The six-slot layout on the Harry example (claim, data, warrant, backing, qualifier, rebuttal); "Propositions of this kind I shall call warrants" (1958, p. 98, quoted by Hitchcock, source 3). The warrant and rebuttal wording here is the 2003 edition's, as read and recorded for lesson 1 (warrant p. 94 and p. 97; rebuttal p. 97). "Data" is the 1958 term; "grounds" is the term of Toulmin, Rieke & Janik, An Introduction to Reasoning (1984, p. 38, via Hitchcock), whose gloss "Such a general, step-authorizing statement is called a warrant" (1984, pp. 46 to 47, via Hitchcock) is quoted in the warrant bullet. That gloss is often misattributed to the 1958 book; Hitchcock's own quotation marks the difference, and this lesson follows him.
2. Groarke, L., "Informal Logic", Stanford Encyclopedia of Philosophy, sec. 3.2 (convergent premises as separate reasons; linked premises joined with a plus sign to show "that they support the argument's conclusion (only) when combined"), sec. 4.1 (following Johnson and Blair 1977, 1994, the ARS criteria: premises are "acceptable when they are true or acceptable in some other way", "relevant when they provide some (i.e. any) support for the conclusion of the argument", and sufficient when they provide enough to warrant acceptance), and sec. 5 (logic treats an argument as an attempt to justify a conclusion; rhetoric "understands arguing as an attempt to persuade an audience"; dialectics "understands an argument as an attempt to make a move in an exchange between interlocutors with opposing points of view"). Every quotation re-verified against the live entry at review time. The ARG lettering (grounds for the third condition) is a textbook variant of Johnson and Blair's ARS; the course uses it for the reason given in the text.
3. Hitchcock, D., "Toulmin's Warrants", in van Eemeren, Blair, Willard & Snoeck Henkemans (eds.), Anyone Who Has a View (Kluwer, 2003), pp. 69 to 82; read in full at review time. "A warrant is an inference-licensing rule, not a premiss"; the 1958 and 1984 page references for the Harry example; Toulmin's field-dependency thesis and Hitchcock's qualification of it ("Some warrants belong to specialized fields, but some are just generalizations, more or less rough-and-ready, based on common experience").
4. Van Cleave, M. J., Introduction to Logic and Critical Thinking, 2nd ed., CC BY 4.0, ch. 1 sec. 1.10 "Assuring, guarding, and discounting" (LibreTexts). Definitions: assuring is "informing someone that there are further reasons although one is not giving them now"; guarding is "weakening one's claims so that it is harder to show that the claims are false"; discounting is "anticipating objections that might be raised to one's claim or argument as a way of dismissing those objections". The presidents ladder of strong to weak claims; the example "The worker was inefficient, but honest"; the list of discounting terms; the remark that assurances "are not necessarily illegitimate ... but a critical thinker will always regard them as somewhat suspect"; exercise sentence 4 (pediatricians and childhood obesity). All quotations re-verified against the live page at review time.
5. Sinnott-Armstrong, W. & Fogelin, R., Understanding Arguments: An Introduction to Informal Logic, 9th ed. (Cengage, 2015), ch. 3 (assuring, guarding, discounting) and the chapter "The Art of Close Analysis". The origin of "close analysis" before "deep analysis" (reconstruction in standard form); used here through Van Cleave, who follows the same method, and the Duke Think Again course, not read directly.
6. Dennett, D. C., Intuition Pumps and Other Tools for Thinking (W. W. Norton, 2013), ch. 3 "Rapoport's Rules", p. 33. The heading "How to compose a successful critical commentary" and the four rules quoted verbatim; Dennett's description of Rapoport as a social psychologist and game theorist; the rules as an antidote to caricaturing one's opponent; and the stated reason that they make the opponent a more receptive audience. Wording verified against a published verbatim excerpt of the passage (The Marginalian, 28 March 2014), re-checked at review time; the printed page itself has not been seen, so the page number is from publisher and library records.
7. Malpas, J., "Donald Davidson", Stanford Encyclopedia of Philosophy, sec. 3.3, read at review time: in Davidson's work the principle of charity "counsels us to interpret speakers as holding true beliefs (true by our lights at least) wherever it is plausible to do", and "a version of which is also to be found in Quine". The coining of the phrase by N. L. Wilson ("Substances without Substrata", Review of Metaphysics, 1959) is the standard attribution, acknowledged in Quine's own footnotes in Word and Object; Wilson's paper itself was not read, and the lesson claims nothing from it beyond the name and date.
8. CSUN, PHIL 200 Critical Reasoning syllabus (Weimin Sun, Fall 2011). Lists evaluating sources of belief (media, experts, personal experience) as a core objective; the background for treating acceptability as a question about where a premise comes from.