The Socratic Review Method
20 honest minutes with 1 question beats 2 polite hours with 50.
The table was my parents’ kitchen table, and the time was somewhere past 11.
In front of me: 1 Logical Reasoning question, a legal pad, and a deal I’d made with myself. I wasn’t standing up until the question made sense. Not “I see why B is right” sense. I’d had that twice already. This question had beaten me twice, weeks apart, and both times I’d read the explanation, nodded the nod you met in chapter 5, and moved on. Though I didn’t know it yet, this was the 9th time the same skeleton had taken my money. 9 invoices. That night I finally opened one.
Out of options, I did something no book had told me to do. I stopped asking why the right answer was right, and I started interrogating the only witness who’d actually been there. Me.
At the top of the legal pad I wrote the argument in my own words, badly. Under it I wrote what I had believed at the moment I picked my answer, which felt embarrassing and oddly difficult, like describing your own handwriting. Then I asked the question I’d never once asked in a year of prep: why did that feel safe? I wrote the answer. It was flimsy, so I asked why again, and again. 4 whys down, I hit bedrock, a belief so obvious I’d never said it out loud: if one thing happens after another, the first thing probably caused it. There it was. Not a careless moment. A policy. I’d been running it on every causal argument for a year, and the test had been billing me for it, 9 times, in 9 different costumes. Somewhere in there the question went quiet. Not solved.
Quiet. I could see to the bottom of it.
And then the proof it wasn’t luck. I spent the next 2 weeks going back through months of old practice tests, running the same interrogation on every troubled question I could find, and the misses started collapsing into a handful of skeletons, just as chapter 4 promised. My next practice test came back 5 points higher. Within a month I’d crossed 160 for the first time. The year ended where chapter 1 told you it ended.
The name came later. I’d been reading about what law school is like, and I kept running into the Socratic method: the professor who answers a student’s answer with a question, then questions that answer, until the reasoning underneath is standing in the open with nowhere to hide. That’s the kitchen table, moved into a classroom. Nobody was going to do that for my LSAT prep. So I’d done it to myself, on paper. Years later I learned Toyota had built an empire on the same move; when something breaks on the line, they ask why 5 times, because the first 2 answers are always symptoms and the cause lives at the bottom. Same tool in 3 different buildings: a law school classroom, a factory floor, and my parents’ kitchen.
Drilling measures you. Review changes you. That table is where the sentence comes from.
That’s the story. Now the method itself.
The protocol: 1 troubled question, 20 minutes, in writing
Step 1: Take the deposition (2 minutes). Before you touch the answer key, an explanation, or anything else, write down your answer and 1 honest sentence about what you were thinking when you picked it. This must come first, because the instant you reread with the truth in your pocket, hindsight repaints the scene and your real reasoning is gone forever. You’re deposing a witness who’s about to start lying.
Step 2: Rebuild the argument in your own words (3 minutes). 2 lines on paper: what’s the evidence, what’s the conclusion. Do a bad job. I mean that as instruction, not consolation; a clumsy rebuild that’s honestly yours beats a polished one copied from the page, and you have my permission to be terrible at this for a month. If you can’t separate evidence from conclusion at all, you’ve already found something better than an answer. You’ve found the actual problem.
Step 3: Mark where the point died (2 minutes). Every miss dies in 1 of 3 places. The stimulus: you misread or never saw the structure. The stem: you answered a different question than the one asked. The choices: you understood everything and got seduced anyway. Circle 1. Over weeks, the circles alone become a map of you.
Step 4: Run the why-chain (8 minutes). The engine. I planted the opening question in chapter 5, and this is where it goes to work: what did I believe, at the moment I picked my answer, that made it feel safe? Write the belief. Then ask why you believed that. Then why again. 3 whys is the minimum; the first answer is always a symptom (“I rushed”), and rushing is never the bottom. You’ve hit root cause when 2 things are true: the answer is a belief about how arguments work, not a fact about this question, and it would explain other misses too. That second test is how you know you’ve found a skeleton and not a costume.
Step 5: Write the rule and the forecast (3 minutes). 1 sentence, your words, written for future-you: the rule that would have saved this point. Then 1 line forecasting where this skeleton will show up next in a different costume. The forecast matters more than it looks; it converts the review from a memory into an ambush you’ve set.
Step 6: Book the rematch (2 minutes). Date, 1 week out, written in the notebook: re-attempt this question cold. And log the skeleton in your journal (more on that ledger in a minute). A review without a rematch is a theory. The rematch is the experiment. 20 minutes. That’s the price chapter 3 quoted you, and it buys what the napkin in chapter 4 said it buys.
Why it has to be in writing
Because a thought in your head is allowed to stay vague forever. The sentence “I sort of confused the cause thing” can float around your skull for a year feeling like insight. Write it down and it dies of exposure in 10 seconds, and you’re forced to replace it with something that would survive on paper. The pen isn’t recording your thinking; it’s doing it. Speaking works too, which is why a sharp study partner or a coach asking you the whys out loud hits even harder; you’re adding a second interrogator. (Engaging more senses helps in general. I haven’t found a use for smell yet. If you crack it, email me.)
Watch it run: 1 question, both endings
Your turn to be the witness. Untimed. Pen ready, pick an answer for real, because the next page only works if you’ve committed.
After the city installed bright LED lighting in Hartwell Park last spring, evening visits to the park rose 40 percent over the summer. Clearly, if the city wants to increase evening visits to Riverside Park, it should install the same lighting there.
Which one of the following, if true, most weakens the argument?
(A) Riverside Park currently draws more evening visitors than Hartwell Park did before the lighting was installed.
(B) The Hartwell lighting project cost more than the city’s entire annual parks budget.
(C) The summer after the installation was unusually warm and dry, and evening visits rose at every park in the city, lit or unlit.
(D) Several residents living near Hartwell Park have complained that the new lights shine into their windows.
(E) A neighboring city installed similar lighting in one of its parks 2 years ago.
Committed? Good.
The answer is (C). And (B) is the trap on this question, engineered for one specific reader: the one who judges the plan instead of the bridge.
One honesty note first: I’m about to write a deposition for you, and yours may read differently. If it does, trust yours. The printed chain is a demonstration; the method only ever runs on the reasoning that actually happened.
If (B) got your vote, your rescue follows, and it’s the full protocol compressed. Deposition: “B makes the plan look terrible, so it felt like it weakened things.” Rebuild: evidence, visits rose after the lights went in at 1 park; conclusion, the same lights will raise visits at another park. Where the point died: the choices; you understood the argument fine. Now the why-chain. Why did B feel safe? Because if the city can’t afford it, the plan is bad. Why does a bad plan feel like a weak argument? Because I was judging whether I’d vote for the idea. Why were you judging the idea? And there’s bedrock: I evaluate arguments by whether the conclusion sounds wise, not by whether the evidence actually supports it. That’s a skeleton, and it will cost you points on weaken, strengthen, and flaw questions for months if it goes unwritten, because the test loves stocking choices that make plans sound expensive, unpopular, or annoying without laying a finger on the logic. The rule, in 1 line: attack the bridge, not the destination. A plan can be unaffordable, hated, and doomed, and the argument for it can still be airtight; your only target is whether the evidence carries the conclusion. Forecast: expect this skeleton dressed as cost in 1 question and as angry neighbors in another. (Which is (D)’s whole job, by the way: a side effect, logic untouched. (A) is a baseline difference that does nothing to the causal claim, and (E) is a fact with no verdict attached.) And (C)? Look closely, because you’ve met it before. If visits rose everywhere that summer, lit or unlit, then the warm weather explains Hartwell, and the lights explain nothing. This happened after that, therefore because of that, dismantled by an alternative cause. The skeleton from my kitchen table, wearing its 10th costume. Chapter 4 told you they’re not allowed to change the bones. Now you’ve watched it.
If you picked (C), don’t skip this paragraph. You don’t get the night off; you get the other half of the method, because right answers hide gaps too. Run the confidence audit, in writing: 1 sentence on why (C) wins, and 1 sentence on why (B) is wrong, precise enough that the version of you from 3 weeks ago would be convinced. If the second sentence comes out fuzzy (“B just isn’t about the logic, you know?”), you’ve located a soft spot that a meaner question will find under time pressure. That’s a troubled question by chapter 4’s definition, and it earns the full 20 minutes.
Run it again: a different bone, both endings
One demonstration can be luck. So before the template, run the machine on a second skeleton and watch it produce a different root cause with the same 6 steps. Same rules: cover, commit, pen.
In a survey of the 80 customers who attended Brightleaf Garden Center’s annual members-only sale, 9 in 10 said they would recommend the store to a friend. Clearly, the store enjoys this same enthusiasm among its customers generally.
The reasoning is most vulnerable to criticism on the grounds that it
(A) generalizes about the store’s customers from a sample that is unlikely to be representative of them
(B) presumes, without justification, that customers who say they would recommend the store have actually done so
(C) relies on customers’ opinions rather than on any objective measure of the store’s quality
(D) treats evidence that most surveyed customers are enthusiastic as proof that every one of the store’s customers is
(E) overlooks the possibility that other garden centers enjoy even greater customer enthusiasm
The answer is (A). And (C) is this question’s trap, engineered for a different reader than (B) was: the one who grades the ingredients instead of the bridge. Same honesty note as before; if your deposition reads differently, run yours.
If (C) got your vote, the compressed rescue. Deposition: “opinions aren’t real evidence, so leaning on them is the flaw.” Rebuild: evidence, 90 percent of the members at a members-only sale would recommend the store; conclusion, customers in general feel the same. Where the point died: the choices. Why did (C) feel safe? Because survey opinions feel flimsy. Why does flimsy-feeling evidence read as the argument’s crime? Because I audit the quality of the ingredients instead of the fit between the ingredients and the claim. And bedrock: I downgrade arguments for using soft evidence even when the conclusion is about soft things. That’s the skeleton. The conclusion here is literally about enthusiasm, which is an opinion, so opinion evidence is the right species. The actual crime is the guest list: people at a members-only sale are the store’s superfans, a sample that selected itself. The rule, future tense: match the evidence’s species to the conclusion’s species before judging it, then attack how the sample was chosen, not how soft it feels. Journal handle: grading the ingredients. Forecast: flaw and weaken questions where 1 choice sneers at surveys, anecdotes, or opinions on principle.
The supporting cast, fast: (B) accuses the argument of assuming recommendations happened, which it never needs. (D) is a crime that didn’t happen; “generally” is not “every one.” (E) compares stores when the conclusion only ever ranked this store’s own customers. And (A) names the real gap: a handful, chosen badly, became everybody. You’ll meet that bone formally in the next chapter; it has a file.
If you picked (A): confidence audit. One sentence on why (A) wins, and 1 precise sentence on why (C) loses. If your second sentence is some version of “(C) describes the evidence accurately but describing evidence isn’t naming a flaw, and opinion evidence fits an opinion conclusion,” you’re done in 2 minutes. If it’s “(C) just felt off,” the question earns its 20.
2 questions, 2 different bedrocks, 1 machine. That’s the point of running it twice: the method was never an answer key for causal arguments. It interrogates, and it finds whatever’s down there.
Your turn, with a net
You’ve watched the machine run twice. Before you get the blank template, run it once with guide rails, on a question I’ll hand you but won’t solve. Cover nothing this time; just write at each prompt. The point isn’t the right answer. The point is feeling each step happen in your own hand.
A consumer group reports that the 200 households it surveyed spent an average of 12 percent less on groceries after switching to the store’s new loyalty app. The group concludes that the app saves the average household money.
The reasoning is most vulnerable to the criticism that it
(A) overlooks the possibility that households that chose to switch to the app already shopped more frugally than average
(B) presumes that 12 percent is a large enough saving to matter to most households
(C) fails to consider that the store could discontinue the app at any time
(D) ignores the possibility that grocery prices rose during the period studied
(E) takes for granted that the app is the only loyalty program the households use
Pick 1, for real, then run the steps on yourself, writing a line at each:
Deposition: what were you thinking when you picked? 1 honest sentence, before you read another word.
Rebuild: evidence on 1 line, conclusion on the next, in your own clumsy words. What’s doing the supporting, and what’s leaning?
Where did it die: stimulus, stem, or choices? Circle the place, even if you got it right; a right answer that died in the choices and got rescued by luck still teaches you something.
Why-chain: if you picked a wrong answer, why did it feel safe? Then why again, twice more, until your last line is a belief about how arguments work and not a fact about groceries. If you picked the credited answer, run the confidence audit instead: 1 sentence on why it wins, 1 precise sentence on why your closest runner-up loses.
The rule: write the 1 sentence that would save future-you on a question like this, in your words, future tense.
Forecast and rematch: name the costume this bone wears next, and put a date 1 week out in the margin.
All 6 steps, walked with a net. The skeleton in this one, if it helps you check your work, is the same self-selecting sample you met at the garden center: people who choose to adopt an app may already be the budget-watchers, so the savings might belong to the shopper, not the software. But notice what mattered more than knowing that: you just ran all 6 steps in your own handwriting. The net comes off on the next one.
The template
This is the page you’ll fill out, in full or in the 5-minute rerun version chapter 10 will teach you, roughly 200 times between now and test day. Copy it into your notebook now, or print the pad version from the resources page.
•••
Free companion #6: that pad version. The template above as a printable, 1 autopsy per sheet, every field sized for a real pen. The pad includes 1 full autopsy worked end to end, deposition to rematch date. Free at unpluggedprep.com/books.
•••
SOCRATIC REVIEW: 1 QUESTION, 20 MINUTES
Date / question ID: My answer / correct answer: Deposition. What I was thinking when I picked it: Rebuild. Evidence: / Conclusion: Where the point died (circle 1): stimulus / stem / choices Why-chain (3 whys minimum, last line must be a belief about arguments, not about this question): Root cause: / The rule (1 sentence, my words, future tense): Costume forecast (where this skeleton shows up next): Rematch date: / Skeleton logged?
Y / N
•••
What 4 months of templates look like, honestly
Real entries, lightly cleaned up, from a student who gave me permission to embarrass her. Watch the same pen across 4 months, because this gradient is what you should expect from yours.
Month 1:
“Picked D because it sounded strong?? Argument is about fish. I think I rushed. Rule: slow down on the last 2 choices.”
That’s a 2 out of 10, and it’s what yours should look like in month 1, because she did it in writing and booked the rematch, which puts her ahead of the entire nodding public.
Month 2:
“Picked (B) because it used the same words as the stimulus and felt safe. Died in the choices. Why safe? Familiar = confirmed. Root: I treat recognition as proof. Rule: a strengthen answer has to add a new plank, not repeat an old one in the same wood. Forecast: any choice that’s just the stimulus in a costume.” Structure has arrived. The whys still take her 10 minutes, and the root cause is real.
Month 3:
“after = because, costume: plant openings and county wages. Caught the conclusion verb (boosted) and predicted the missing suspect before the choices. Still lost 40 seconds to (D) because it named a real-world worry. Rule: a sentence can be true in life and a stranger to this argument. Tally: 6.”
Now the journal and the prediction are doing half the work before the template even opens.
Month 4:
“Picked the trap because it restated the conclusion in louder words, and I rewarded volume instead of new support. Root: I treat emphasis as evidence. Rule: a strengthen answer must add a fact the conclusion needs, not repeat the conclusion with feeling. Forecast: watch for this on strengthen questions where 1 choice is basically the stimulus wearing a megaphone.”
Same pen. The difference is 90 days of reps, and that month-4 entry is what a 12-point jump looks like in handwriting.
The skeleton journal
Last piece, 60 seconds per review. Keep a running page in the back of your notebook, 1 line per skeleton, in your own words (“after = because,” “emphasis = evidence,” “judging the plan, not the bridge,” “grading the ingredients”), with a tally mark for each new appearance. The journal is your rerun detector, the thing row 7 of the check-in was asking about. The rule of 3: any skeleton that collects 3 tallies stops being a note and becomes your top priority, because the test just told you, in writing, where your next several points are buried.
When the chain stalls: the 5 dead ends
Fair warning, from watching people learn this since 2005: step 4 is the step that fights back. Smart people stall in the same 5 places, so I’m giving you the map and the way out for each. Tape this list inside the notebook until you stop needing it.
Stall 1: “I just misread it.” A misread is an event, not a cause. Eyes don’t slip randomly; they slip in the direction of something. The way out: what was I hunting for while that line went by? Usually the answer is “the conclusion I’d already decided on,” which means the real bedrock is I read to confirm, not to find, and that’s a skeleton worth a tally.
Stall 2: “I rushed.” Rushing is a decision, and decisions have reasons. The way out: what did I believe about this question’s price? Common bedrock: I believe early questions don’t deserve full attention, or I believe time spent rereading is time wasted. Both are policies. Both are billable.
Stall 3: “I didn’t know the word.” Sometimes legitimate, and in month 1, vocabulary really can be the floor; if “sufficient” or “presumes” was genuinely foreign, the rule is a definition and the chain can stop. But ask 1 more why anyway: why did I answer instead of flagging? If the answer is because guessing felt better than admitting I didn’t know, you’ve found a more expensive belief than the vocabulary.
Stall 4: the shame spiral. The chain turns into name-calling: “because I’m careless,” “because I’m bad at these.” Insults aren’t beliefs about arguments, so the chain has left the rails. The way out: what specifically did careless-me believe at the moment of the pick? Convert the adjective into a policy, then interrogate the policy. The method has no use for your character, only your reasoning.
Stall 5: the infinite regress. The opposite failure: the chain blows past bedrock into philosophy. “Why do I trust evidence at all? Why does anyone believe anything?” Stop. Reread step 4’s stop condition: a belief about how arguments work that would explain other misses. If your last line names a pattern you could forecast on next week’s questions, you’re done. If it could be carved on a monument, you went 2 whys too far.
If you stall somewhere that isn’t on this list, you’ve found dead end number 6, and I’d genuinely like to hear about it. The list got to 5 by students mailing me their stalls.
•••
Your turn. Tonight. 20 minutes.
You walked in here holding 1 troubled question; chapter 5 made sure of it. This is the experiment chapter 1 promised, the one that asks for no faith. Set a 20-minute timer, open the template, and run all 6 steps on your question, doing a proudly bad job of step 2. When the timer dies, look at what’s on the page. If there’s a root cause and a rule in your handwriting that no answer key could have given you, the method just paid you for the first time. Date the page. That’s your first invoice opened.
•••
Stuck at step 4? Read this box.
The dead-end map above handles most stalls. If you’re still stuck, you’re not failing, you’re at the exact spot where a second voice pays for itself. That’s what the free tutoring lesson is for. Book it at unpluggedprep.com/start.
•••
KEEP THESE 3
The protocol: deposition, rebuild, mark where the point died, why-chain to a belief about arguments, rule plus forecast, rematch. 1 question, 20 minutes, always in writing, because the pen is what’s doing the thinking.
The why-chain’s stop condition: you’ve hit root cause when the answer is a policy, not an accident, and it would explain other misses too. “I rushed” is never the bottom, and the 5 dead ends all have a way out.
Right answers get audited too. Can you say, in 1 precise sentence, why the trap is wrong? If not, that question is still wearing its costume, and it earns the full 20 minutes.
•••
You own the method, and you’ve now watched it find 2 different bedrocks in 2 different readers. Next chapter aims it at the biggest target on the board: Logical Reasoning, 2 sections, 2/3 of your score, and the 6 skeletons that show up so often they’ve earned police files. Bring the journal. We’re going hunting.
