ARGUMENT: STRUCTURE AND FALLACIES
Most people think they already know how to argue, because most people have argued. But the arguing that wins a text-message fight and the arguing that earns a B-plus on a college paper are different skills. Academic argument is slow, visible, and inspectable: you say what you think, you say why, you show what makes the why believable, and you name the assumption that connects them. Then — and this is the part almost everyone skips — you go looking for the strongest reason you might be wrong and you deal with it in the open. This chapter takes an argument apart into nameable pieces, shows you how to arrange those pieces on the page, catalogs the ways arguments break, and hands you a stress test to run on your own draft before an instructor, a charge nurse, or a stranger on the internet runs it for you.
By the end of this chapter you can
- Distinguish an arguable claim from a statement of fact, a matter of taste, and a topic.
- Break any argument into claim, reason, and evidence, and test the link with the word because.
- Identify the unstated warrant beneath a reason, supply backing for it, and decide when to state it outright.
- Add a qualifier and a rebuttal condition so a claim survives contact with a skeptical reader.
- Arrange an argument using classical arrangement or a Rogerian opening, and justify the order of your reasons.
- Write a genuine concession and a targeted refutation using standard sentence moves.
- Name and repair eighteen common fallacies, including straw man, false dilemma, post hoc, and begging the question.
- Explain why fallacy labels are context-dependent and why calling “fallacy” is not itself a refutation.
- Run a reverse outline and an objection audit on your own draft and revise from the results.
Section 1What an Argument Is Made Of
Claim, reason, evidence: the smallest working unit
An academic argument is a claim supported by reasons, and reasons supported by evidence. That is the whole machine. A claim is the sentence you want the reader to accept. A reason answers the question why should I accept that? Evidence answers the next question, and how do you know? Everything else in a persuasive essay — the hook, the transitions, the graceful last line — is packaging around those three parts.
Watch the machine assemble itself. Claim: Hospitals should be required to cap the number of patients assigned to a medical-surgical nurse. Reason: because heavier assignments are associated with worse patient outcomes. Evidence: a widely cited 2002 study in the Journal of the American Medical Association, led by Linda Aiken, found that each additional surgical patient added to a nurse’s workload was associated with roughly a seven percent increase in the odds of death within thirty days of admission. Three sentences, and a reader now knows what you think, why, and what you are standing on.
Beginning writers usually have the claim and the evidence and are missing the middle. They write a thesis, then drop in a statistic, then move on, as if the number argued for itself. Numbers never argue for themselves. The reason is the sentence that tells your reader what the number is doing there.
The because test
Here is a two-minute diagnostic you can run on any paragraph. Write your claim, add the word because, and finish the sentence in your own voice. If the finished sentence sounds obviously true, your reason is too weak to need arguing. If it sounds like a restatement of the claim, you have gone in a circle. If it sounds like something a reasonable person could deny, you have found a real reason and you now know what the paragraph has to prove.
| Attempt | Diagnosis |
|---|---|
| Nurse staffing ratios should be capped because understaffing is bad. | Circular. “Understaffing is bad” only restates the claim in gloomier words. |
| Nurse staffing ratios should be capped because nurses are tired. | True but weak. Fatigue alone does not establish that a legal cap is the right instrument. |
| Nurse staffing ratios should be capped because voluntary staffing policies have not produced stable ratios where they have been tried. | Arguable and checkable. A reader could deny it, which means it can be supported. |
Sorting what is arguable
Not every sentence can anchor an argument. Verifiable facts are settled by looking them up, not by essays. Pure preferences cannot be adjudicated at all. Topics are not claims: “social media and teenagers” is a subject area, not a position. The sweet spot is a sentence that a reasonable, informed person could disagree with, and that evidence could move.
Notice one honest complication. The line between fact and claim shifts with audience and time. “The earth’s climate is warming” functions as settled fact in a science course and as a contested claim in a 1985 op-ed. Part of analyzing your rhetorical situation, as Chapter 1 argued, is knowing which sentences your particular readers will accept without a fight.
| Sentence | What it is | Fix or verdict |
|---|---|---|
| Community colleges enroll millions of American undergraduates. | Verifiable fact | Useful as evidence, not as a thesis. Cite it and build on it. |
| Night shift is the worst shift. | Preference | Convert to something checkable: Rotating night shifts impose costs that scheduling policy should compensate for. |
| Textbook prices in higher education. | Topic | Take a position: Course-material costs should be listed at registration, before a student commits to a section. |
| Mandatory clinical hours should count toward the credit load used to calculate financial aid. | Arguable claim | Keep it. Reasonable people differ, and evidence about student workload bears on it. |
| Everyone knows group projects are unfair. | Claim in disguise, propped up by “everyone knows” | Drop the crowd. State it plainly and support it: Group grades reward coordination skill more than subject mastery. |
Every argument reduces to a claim, the reasons that support it, and the evidence that supports those reasons — and the missing piece in most student drafts is the reason in the middle.
Section 2Warrants: The Assumption Doing the Real Work
Toulmin’s six parts
The philosopher Stephen Toulmin, writing in 1958, noticed that real arguments almost never look like the tidy syllogisms of formal logic. They look like law. So he built a vocabulary from how practical arguments actually behave, and composition courses have used it ever since because it is diagnostic: it tells you which part of your paragraph is missing.
Toulmin’s parts are the claim (what you want accepted), the grounds (the data or evidence you offer), the warrant (the general principle that makes the grounds relevant to the claim), the backing (support for the warrant itself, needed when readers doubt it), the qualifier (the word that limits how strongly you assert — usually, in most urban districts, probably), and the rebuttal (the conditions under which your claim would not hold).
| Part | Question it answers | Worked example |
|---|---|---|
| Claim | What are you asking me to accept? | Our campus should move the writing center’s hours to include Saturday mornings. |
| Grounds | What have you got? | Of the students who requested appointments last term, sixty-one percent listed a weekday work conflict. |
| Warrant | Why does that count as support? | Support services should be scheduled when the students who need them are able to attend. |
| Backing | Why should I accept the warrant? | The center’s own mission statement commits it to access for working and commuting students. |
| Qualifier | How strongly do you assert this? | At least during midterm and finals weeks … |
| Rebuttal | When would this not hold? | … unless Saturday usage data from the pilot fall below the weekday average. |
Finding the warrant you did not know you had
Warrants hide. They hide because they feel like common sense to the writer, and common sense is invisible from the inside. The trick is to write your reason, write your evidence, and then ask: what would a person have to already believe for that evidence to matter? The answer is your warrant, and it is very often the only place a hostile reader will attack.
Take a simple argument: The college should drop the placement essay, because seventy percent of students placed by the essay ended up in the same course their high school GPA predicted. The grounds are a number. The warrant is that a placement instrument that mostly duplicates cheaper information is not worth its cost. That warrant is defensible — but it is doing all the work, and a reader who values the remaining thirty percent will go straight for it. Say it out loud in your paragraph and you control the fight. Leave it buried and your reader controls it.
Different audiences accept different warrants. An argument to a hospital’s finance committee that rests on the warrant patient dignity outweighs efficiency may be true and still fail, because that committee has been assembled to weigh efficiency. This is not cynicism; it is audience analysis. You may need backing that translates your warrant into terms the room already accepts, such as readmission costs or turnover.
Making the warrant visible: before and after
Compare two versions of the same paragraph: a competent first draft, and the same argument with the warrant surfaced and qualified.
| Draft | What is happening |
|---|---|
| Before. Employers should stop requiring bachelor’s degrees for administrative roles. A recent analysis found that many job postings for these roles list a degree requirement even though most people currently doing the work do not hold one. | Claim plus grounds, no warrant. The reader has to guess why a mismatch between postings and incumbents matters. |
| After. Employers should stop requiring bachelor’s degrees for most administrative roles. A recent analysis found that many postings for these roles list a degree requirement even though most people currently doing the work do not hold one. The mismatch matters because a credential can only justify itself as a screen if it predicts something the job actually needs — and here the incumbents are the evidence that it does not. That reasoning holds unless the degree is a legal or accreditation requirement, in which case the screen is not the employer’s to remove. | Claim, grounds, stated warrant, and a rebuttal condition. The argument is now harder to knock over because it has already leaned on itself. |
Evidence supports a claim only through a warrant — the assumption that makes it relevant — so find yours, decide whether your audience shares it, and state it whenever they might not.
Section 3Arranging the Argument on the Page
Two default shapes
Classical arrangement comes from Roman rhetorical handbooks and still fits most college argument papers. In modern terms: open with something that earns attention, supply the background a reader needs, state your thesis and preview your reasons, develop the reasons with evidence, handle the strongest objection, and close with the stakes. Roman teachers called these moves the exordium, narratio, partitio, confirmatio, refutatio, and peroratio. You do not need the Latin. You do need the sequence, because it front-loads what the reader needs in order to follow you.
Rogerian argument — adapted for writing classrooms in the late 1960s and 1970s from psychologist Carl Rogers’s work on listening — inverts the opening. Instead of declaring a thesis and defending it, you first restate the opposing position so accurately that its holders would say yes, that is what I think. Then you identify the conditions under which their view makes sense, and only then introduce your own, framed as an extension rather than a defeat. It is slower and it withholds the thesis, which some instructors dislike. It is also the only shape that reliably works when your audience already distrusts you — a letter to a manager, a policy memo to skeptical colleagues, an essay on an issue readers have already fought about.
A third question has no settled answer: should your strongest reason come first or last? Classical advice and much modern practice put a strong reason first to establish credibility and the strongest last for emphasis, burying the weakest in the middle. Persuasion research on primacy and recency effects is genuinely mixed, and results vary with how motivated and how tired the audience is. Reasonable instructors differ. What is not in dispute: do not open with your weakest reason, and do not end on one.
Concession and refutation as moves, not decorations
A concession grants that something in the opposing view is true or reasonable. A refutation shows why the opposing view still fails. Together they are the single highest-yield revision available to most student arguments, and the most commonly faked. A fake concession names a weak version of the objection and dismisses it in half a sentence. A real one states the objection at its strongest — sometimes stronger than an opponent bothered to — and then answers that version.
You can refute at three different depths, and knowing which one you are doing keeps the paragraph focused. You can challenge the evidence (the data are outdated, unrepresentative, or measuring something else). You can challenge the warrant (the evidence is fine but does not support the conclusion drawn from it). Or you can challenge the relevance (the point is true and irrelevant to the decision at hand). Attacking the warrant is usually the most powerful move and the least used.
| Move | When to use it | Sentence template |
|---|---|---|
| Strong concession | The objection has real force and you can still win | The most serious objection is not X but Y, and Y deserves an answer: … |
| Concede and pivot | The point is true but does not decide the question | Granted, ___. But that settles a different question than the one on the table, which is ___. |
| Refute the evidence | The data do not say what they are said to say | The figure is accurate, though it comes from ___, where conditions differ in ___. |
| Refute the warrant | Evidence is fine, inference is not | Even accepting those numbers, the conclusion follows only if we assume ___ — and that assumption fails when ___. |
| Qualify | Your claim is right in a narrower range than you first wrote | The case is strongest for ___; for ___ the evidence is thinner and my claim is correspondingly weaker. |
| Reframe the stakes | Both sides are arguing over a secondary cost | Framed as a question of cost, the objection wins. Framed as a question of ___, it does not. |
Before and after: a real refutation paragraph
Before. “Some people say that four-day school weeks hurt students. But this is not true. Studies show that students do fine, and parents like the schedule. Therefore districts should adopt it.”
After. “The strongest objection to the four-day week is not that students dislike it — they generally do not — but that the missing day shifts childcare costs onto families who can least absorb them, and that this cost is invisible in district budgets because it never appears there. That objection is correct as far as it goes. But it argues against adopting the schedule without a supervised-care option, not against the schedule itself: a district that funds Friday care removes the very cost the objection is built on. A four-day week is defensible, then, only as a package — and a district that cannot fund the care has not made its case for the calendar.”
Notice what changed. The after version names a specific objection instead of “some people,” concedes that it is correct, refutes its scope rather than its truth, and ends with a claim that is narrower and far harder to dismiss. It also lost the word “therefore,” which in student drafts is usually a promise the sentence does not keep.
Choose classical arrangement when your reader will follow a thesis and Rogerian order when your reader distrusts you — and in either shape, state the opposing view at its strongest before you answer it.
Section 4Fallacies: The Standard Ways Arguments Break
What a fallacy is, and what calling one is not
A fallacy is a pattern of reasoning that feels persuasive but does not actually support its conclusion. Formal fallacies break the structure of an inference. The ones that matter for writing are informal fallacies: failures of relevance, of sufficiency, or of honest framing. They persist because they work on tired readers.
Two cautions before the catalog, and they are not decorative. First, most fallacy labels are context-dependent. Argumentation scholars in the pragma-dialectical tradition and researchers such as Douglas Walton have argued persuasively that many so-called fallacies are legitimate moves gone wrong in a particular setting rather than forms that are always invalid. Citing an expert is not a fallacy; citing an unqualified expert, or treating one expert as settling a genuinely contested field, is. Warning that a policy will lead to further consequences is not a fallacy; asserting a chain of consequences with no support for any link is.
Second, naming a fallacy is not the same as refuting an argument. A conclusion can be true even when a particular argument for it is bad — treating a flawed argument as proof that its conclusion is false is itself a mistake, sometimes called the fallacy fallacy. In academic writing, the useful move is almost never “that is a straw man.” It is “that version of my position is not the one I hold; here is the one I do hold, and here is why the objection misses it.”
Failures of relevance and pressure
These fallacies substitute something emotionally live for something logically relevant. They are the easiest to spot in other people’s writing and the hardest to spot in your own, because in your own draft they feel like conviction.
| Fallacy | What it does | Weak version | Repaired |
|---|---|---|---|
| Ad hominem | Attacks the arguer instead of the argument | We can ignore the staffing proposal; the union rep who wrote it has never managed a budget. | The staffing proposal underestimates overtime costs by assuming a fixed census; here is what the numbers look like with seasonal variation. |
| Straw man | Refutes a distorted version of the opposing view | Advocates of open-book exams think students should never have to learn anything. | Advocates of open-book exams argue that recall under time pressure measures the wrong skill. That claim is worth taking seriously, but it understates how often clinicians must act without lookup time. |
| Appeal to popularity (ad populum) | Treats wide belief as evidence of truth | Most students in my cohort think the lab fee is unjustified, so it is. | The lab fee funds equipment that three of the four course sections no longer use, according to the department’s own inventory. |
| Misused appeal to authority | Cites an authority who is irrelevant, unqualified, or falsely presented as consensus | A famous physician says this supplement prevents dementia, so it works. | Two randomized trials of this supplement found no cognitive benefit; a widely quoted physician disagrees, but has not published trial data. |
| Red herring | Changes the subject to something easier | You ask about wait times, but the real issue is how underpaid our staff are. | Wait times and staff pay are linked: vacancies in triage account for most of the increase since March. |
| Appeal to fear | Uses dread in place of evidence | If we do not adopt this software now, patient safety will collapse. | The current system logs medication changes only at shift end, which delayed reconciliation in eleven documented cases last quarter. |
| Tu quoque | Deflects criticism by pointing at the critic’s conduct | You cannot criticize my late assignments; you returned midterms two weeks late. | The late penalty is steep relative to the assignment’s weight; here is a proportional alternative. |
| Genetic fallacy | Judges a claim by its origin rather than its merits | That statistic came from an industry group, so it is false. | That statistic came from an industry group and counts part-time roles as full positions, which inflates the total by about a third. |
Failures of evidence and framing
The second family is less about heat and more about structure: too little evidence, the wrong causal story, or a question rigged before it is asked. These are the fallacies that show up most often in research papers, where the writer has real data and draws more from it than it can carry.
| Fallacy | What it does | Weak version | Repaired |
|---|---|---|---|
| Hasty generalization | Generalizes from too few or unrepresentative cases | Both students I tutored improved after switching to paper notes, so handwriting beats typing. | Two tutees improved after switching to paper notes. Controlled studies comparing longhand and laptop note-taking report mixed results, so I offer this as a hypothesis, not a finding. |
| Post hoc (false cause) | Treats sequence as causation | Readmissions dropped after we launched the discharge app, so the app worked. | Readmissions dropped after the app launched, but the same quarter added two case managers; without separating those changes we cannot attribute the drop. |
| Slippery slope | Asserts an unsupported chain of consequences | If we allow revision on one paper, soon no deadline will mean anything. | Unlimited revision would strain grading capacity. A single revision window with a fixed deadline addresses that without removing deadlines. |
| False analogy | Compares cases that differ in the relevant respect | Running a hospital is like running a restaurant: satisfy the customer. | Both settings manage throughput, but patients cannot decline care the way diners decline a dish, so satisfaction is a weaker guide to quality. |
| False dilemma | Reduces a range of options to two | Either we cut the program or we raise tuition. | Cutting the program and raising tuition are two options; reallocating the facilities reserve is a third, and phased enrollment caps a fourth. |
| Begging the question | Assumes in the premise what it claims to prove | Required attendance is fair because students should be required to attend. | Required attendance is defensible because in-class work is graded and cannot be replicated asynchronously. |
| Equivocation | Shifts a key term’s meaning mid-argument | This remedy is natural, and what is natural is safe, so the remedy is safe. | “Natural” meaning plant-derived and “natural” meaning harmless are two different claims; digitalis is plant-derived and lethal at modest doses. |
| Loaded question | Smuggles an unproven assumption into the question | Why does the department keep wasting money on assessment? | What does the department’s assessment budget purchase, and what evidence is there that it changes instruction? |
| No true Scotsman | Redefines a category to dodge a counterexample | No serious student misses class — and anyone who does was never serious. | Attendance correlates with grades in this course, though the correlation is modest and confounded by work hours. |
| Moving the goalposts | Raises the evidentiary bar once it is met | Fine, the pilot worked, but it did not run for five years. | The one-year pilot met its stated benchmarks. A longer trial would test durability, which the pilot was not designed to measure. |
Fallacies are patterns where the support never actually reaches the conclusion — learn them to repair your own drafts, and remember that labeling one is a diagnosis, not a refutation.
Section 5Stress-Testing Your Own Argument
The argument reverse outline
Chapter 3 introduced the reverse outline, where each margin note records a paragraph’s topic and its job. The argument version records something narrower, and it is not a summary. Open your draft and, in the margin of each body paragraph, write two things and nothing else: the claim that paragraph makes, and the reason it gives. Then read only your margin notes, top to bottom. Three failures show up immediately. If two paragraphs carry the same margin note, they are one paragraph wearing two hats. If a paragraph’s claim does not connect to the thesis, it is a digression no transition can rescue. And if a margin note reads “background on X” three times in a row, your paper is a report that has not started arguing yet.
Do the same for evidence. Beside each piece of support, write one word: who, how many, when. If you cannot fill in all three, you have a sentence that sounds like evidence and functions as filler. Chapter 6 has more on what makes support count; this pass just finds the gaps.
The objection audit
Now argue against yourself on purpose. Write your thesis at the top of a blank page and generate objections in four specific directions rather than waiting for inspiration. Attack the evidence: where did these numbers come from, who was counted, what year, who was left out? Attack the warrant: what must a reader believe for my evidence to matter, and who does not believe it? Attack the scope: where is my claim weakest — which population, which setting, which edge case? Attack the cost: if I am right, what does acting on it break, and for whom?
Then triage. Objections you can answer become refutation paragraphs. Objections you cannot answer become qualifiers — you narrow the claim until it is true. Objections that would sink the whole thesis are the most valuable of all, because you found them and your grader has not yet. Changing your mind mid-draft is not failure; it is the process working. Chapter 3 made the same point about revision, and it is worth repeating here because argument papers are where students resist it most.
Hedging versus precision
Students are often told to sound confident, and then they overclaim. They are also told to be cautious, and then they hedge into mush. The distinction that resolves this: hedging weakens a sentence without adding information, while qualifying narrows a claim to the range the evidence actually covers. Qualified prose sounds more authoritative than absolute prose, not less, because a reader can tell you know where your evidence stops.
| Overclaimed | Hedged into mush | Properly qualified |
|---|---|---|
| Online courses do not work. | It seems that online courses may possibly not always work for some students in certain cases. | In the two large community-college studies I examined, fully online sections showed lower completion rates than in-person sections, with the gap largest among first-term students. |
| Everyone knows twelve-hour shifts cause errors. | Some research might suggest a potential link between shift length and errors. | Several observational studies associate shifts longer than twelve hours with increased self-reported errors; because the data are self-reported, the size of the effect is uncertain even though the direction is consistent. |
| This policy will fix the problem. | This policy could perhaps help somewhat. | This policy addresses the scheduling half of the problem. It does nothing about vacancy rates, which a second measure would have to cover. |
A rubric you can run before submitting
Grade your own draft against the criteria below before anyone else does. Be honest about the middle column; that is where grades are lost.
| Criterion | Not yet | Strong |
|---|---|---|
| Claim | Announces a topic, or states something no one disputes | Takes an arguable position a specific reader could reject |
| Reasons | Evidence appears with no stated because | Each body paragraph opens with a reason, not a fact |
| Evidence | Unsourced, undated, or dropped without comment | Sourced, dated, and followed by a sentence explaining what it shows |
| Warrants | Assumed and invisible | Stated wherever the audience might not share them, with backing |
| Counterargument | “Some people say” plus a one-line dismissal | Named, steel-manned, and answered at the level of evidence, warrant, or relevance |
| Qualification | Absolute claims, or hedges everywhere | Claim narrowed to the range the evidence supports; limits stated |
| Arrangement | Paragraphs could be shuffled with no loss | Order is defensible; nothing depends on information the reader has not yet been given |
| Fallacy check | Loaded language, false dilemmas, or unsupported causal chains | Causal claims supported step by step; alternatives acknowledged |
Attack your own draft first — reverse-outline the claims, audit the objections in four directions, and convert the objections you cannot answer into qualifiers rather than pretending they do not exist.
The chapter in one breath
An argument is a claim, the reasons that support it, and the evidence that supports those reasons — joined by warrants, the assumptions that make evidence count. Start by making sure your claim is arguable at all: not a fact to be looked up, not a preference, not a topic. Use the because test to find a real reason, then ask what a reader would have to already believe for your evidence to matter, and state that warrant whenever your audience might not share it. Add a qualifier so your claim covers only the ground your evidence covers, and a rebuttal condition so a reader knows you have thought about limits. Arrange the whole thing classically when readers will follow a thesis, and in Rogerian order when they will not. Concede the strongest version of the opposing view, then refute it at the level of its evidence, its warrant, or its relevance. Learn fallacies to repair your own drafts rather than to score points, remembering that most labels depend on context and that naming one refutes nothing by itself. Then stress-test: reverse-outline your claims, audit objections in four directions, and let the ones you cannot answer narrow your claim instead of quietly disappearing.
Words worth knowing
| Term | What it means |
|---|---|
| Argument | A claim supported by reasons, which are in turn supported by evidence; in academic writing, an invitation to inspect your reasoning, not a fight. |
| Claim | The arguable sentence you want your reader to accept. A reasonable, informed person could deny it. |
| Reason | The because statement linking claim to evidence; the sentence most often missing from student paragraphs. |
| Grounds | Toulmin’s term for the data, examples, or testimony offered in support of a claim. |
| Warrant | The general principle, usually unstated, that makes the grounds count as support for the claim. |
| Backing | Support offered for a warrant itself, needed when the audience does not already accept the warrant. |
| Qualifier | A word or phrase limiting the strength or scope of a claim, such as usually, in most urban districts, or among first-term students. |
| Rebuttal | The stated condition under which your own claim would not hold. |
| Classical arrangement | The Roman sequence of introduction, background, thesis and preview, development, refutation, and conclusion. |
| Rogerian argument | An arrangement that opens by restating the opposing view fairly and locating common ground before advancing a position; useful with hostile audiences. |
| Concession | Granting that part of the opposing view is true or reasonable, without abandoning your claim. |
| Refutation | Showing why an opposing view fails, at the level of its evidence, its warrant, or its relevance. |
| Steel man | The strongest, most charitable version of an opposing argument — the version worth answering. |
| Fallacy | A pattern of reasoning that feels persuasive but does not actually support its conclusion. |
| Straw man | Refuting a distorted, weakened version of the opposing position instead of the real one. |
| False dilemma | Presenting two options as exhaustive when other options exist. |
| Post hoc | Treating “after this” as “because of this”; mistaking sequence for causation. |
| Begging the question | Assuming in the premise the very thing the argument claims to prove. |
| Fallacy fallacy | Concluding that a claim is false merely because one argument offered for it is flawed. |
| Reverse outline | A revision method in which you record something about each paragraph after drafting — its topic and job in general, or its claim and reason in an argument — then read only those notes to test structure. |