How to Attack an LSAT Logical Reasoning Question: Analyze, Predict, Eliminate

A common experience, a few practice sections in: you read the stimulus and it makes sense. You get to the answer choices and two of them sound pretty good. You go back and forth, pick one, move on. Later you check the answer, read the explanation, and think: yes, obviously.

The frustrating part is that you understood the argument. You could have explained it to someone. So why did the question beat you?

Because the point wasn't lost on the answer choices. It was lost on the stimulus. You finished reading without forming a position, so when you arrived at the choices you had nothing to compare them against. Four of those five were written specifically to sound reasonable to a reader in that state.

You can't fix that by reading harder. You fix it by doing the work in a different order.

The three steps

Every Logical Reasoning question yields to the same sequence:

1. Analyze — work out what the stimulus is, and react to it on its own terms.
2. Predict — read the question and say what a right answer will have to do.
3. Eliminate — cut everything that doesn't do it.

The order is the whole point. Almost everyone does some version of step one. Almost nobody does step two. And step three is nearly impossible to do well without it, because "eliminate" turns into "pick whichever one still sounds okay after a third read."

Step 1: Analyze

Read the stimulus and react to it. Stimuli come in three kinds, and each one asks for a different reaction:

• If the argument doesn't hold up, say what's wrong with it.
• If the argument does hold up, work out how it's built.
• If it's a set of facts with nothing being argued, work out what they commit you to.

Note that you are not sorting the stimulus into a bin and then going to work. The reaction is how you sort it. You see the jump from the evidence to the claim, and seeing it is what tells you this was a flawed argument. That order matters more than it looks, and the rest of this article is largely about why.

When the argument doesn't hold up

Most Logical Reasoning stimuli are arguments, and most of those arguments are defective. Find the conclusion, find the evidence, and name the jump between them.

The conclusion is the claim that everything else in the stimulus exists to support. It is not always the last sentence, and it is often not the sentence that sounds most important. The reliable test is to ask why. Take a candidate sentence and ask, "why should I believe that?" If the rest of the stimulus answers the question, you have the conclusion. If the rest of the stimulus doesn't answer it, you have a premise.

Two things to watch for. First, the LSAT frequently opens with a claim that the author is about to disagree with: someone else's position, quoted so it can be knocked down. Words like but, however, and yet usually mark the pivot from what someone else thinks to what the author thinks. Second, in dialogue-style stimuli with two speakers, be certain about whose argument you're working on.

Then be precise about what the evidence literally says, because the flaw almost always lives in the difference between what was shown and what was claimed. LSAT arguments are built to look valid while containing a specific, nameable jump, and your job is to say that jump out loud in your own words: the author is treating X as though it were Y.

A handful of jumps account for a large share of the test. Evidence of a correlation used to support a claim of causation. A sample used to support a claim about a whole population, without any reason to think the sample is representative. Something true of the parts asserted about the whole. A condition that is merely required for an outcome treated as though it were enough to produce it, or the reverse. A word that shifts meaning between the premises and the conclusion.

When the argument holds up

Not every argument on this test is broken. Some are valid, and some are defensible enough that there's nothing useful to attack. When that's what you're looking at, the question is almost certainly going to ask you to describe the thing rather than fault it: what job a particular sentence is doing, how the author gets from the evidence to the claim, which answer choice argues in the same shape.

So react by understanding the construction. What is concluded, what supports it, and how is the support meant to work? A case generalized into a rule. A rule applied to a case. A hypothesis rejected because something it predicts didn't happen. Two options presented and one ruled out.

Hunting for a flaw here wastes the time you should be spending on structure, and it also puts you in the market for an answer choice that names a defect this argument never had. Those choices are on the page, and they are written to be attractive to someone who arrived expecting to find something broken.

When there's no argument

Some stimuli argue nothing. They give you a set of statements and ask you to take them as true. There's no conclusion to locate and nothing to criticize, so the reaction is to ask what these statements, together, guarantee.

Look for the pieces that combine. Two conditionals that chain into a third. A conditional you can run backward, which is available the moment you see one. A quantity that limits another. Anything that must be true given what you've been told is fair game, and that includes something the stimulus already said outright. Students throw away correct answers for "just repeating the passage," but a restatement of a premise is guaranteed by the premises, which is the whole standard.

The discipline here is staying inside what's written. The reading habits that earn points on a flawed argument, where you're actively looking for what the author hasn't established, will cost you points here, where nothing is being established and everything is simply given.

Step 2: Predict

Now read the question, and before you look at a single answer choice, say what a right answer will have to do.

The question is asking you to do something with what you just found. If you named a jump, the question might want you to describe it, widen it, close it, or state what the author needed to be true for the jump to be safe. If you worked out what a set of facts guarantees, the question wants something that is guaranteed. Answer the question in front of you using the material you already have. That's the entire step.

Notice that this is not a prediction of the answer's wording. You are almost never going to guess the sentence. You are writing a job description, and then hiring whichever choice fits it.

It is also not an attempt to name the one correct answer in advance. There is rarely a single right thing to say here, and treating prediction as target practice makes it feel harder than it is. What you want is the standard a choice has to meet. A wide, correct standard beats a narrow guess, because the standard is the thing you'll actually use in a moment.

Prediction is what converts the answer choices from five judgment calls into one matching task. Instead of asking "does this sound good?", which has no defensible answer, you ask "does this do the job?"

You may have noticed that none of this required you to know what the question is called. That's deliberate. Type names are useful for tracking your accuracy and for talking about the test, and there's a full breakdown of them in the question type guide. But a name you look up is a shortcut around the thinking, and the thinking is the part that transfers to the question you haven't seen yet. Work the stimulus, read the question, decide what an answer has to do. If a label falls out of that, fine.

Step 3: Eliminate

Now, and only now, read the choices. You are not looking for the best-sounding one. You are holding each choice up against the standard you just set and asking whether it meets it.

Be ruthless about this. A choice that doesn't do the job is trash, and it is trash immediately, not after you've thought about what it might mean, or found a reading of it that would work, or noticed that it's true. Plenty of wrong answers are perfectly true. Cross it off and keep moving. Most of the time lost in this step goes to rehabilitating choices that were already dead.

The same strictness cuts the other way. If a choice plainly does the job, take it and move on. You do not owe the test five evaluations, and the ritual of crossing off four before you're allowed to pick one costs real seconds on questions where the answer was the second thing you read.

Wrong answers on this test are not random. They come in recurring shapes, and learning to recognize them is most of what separates a fast eliminator from a slow one:

• Right idea, wrong direction. The choice engages the exact issue you predicted, then does the opposite of what the question asked, shoring the argument up when you were asked what would undermine it.
• Too strong. The choice would do the job, but claims far more than it needs to: only, never, all, impossible. Common when you were asked what the argument requires, since an argument usually requires much less than a confident-sounding choice asserts.
• Out of scope. True, plausible, sensible, and about something the argument never claimed.
• Answers a different question. The most dangerous shape. The choice addresses a nearby claim that the author didn't actually make, such as whether something is worth doing rather than whether it works.

If you get to the end and nothing met the standard, the standard was wrong. That is useful information and it has a specific fix: go back to the stimulus, not to the choices. Somewhere in step one you reacted to the wrong thing a premise mistaken for the conclusion, or a flaw hunted in an argument that didn't have one. Re-reading the five choices a third time will not tell you which, because the choices were written by someone who would very much like you to keep doing that.

And if two survive, resist the urge to read them against each other, because that comparison has no anchor. Go back to the stimulus instead. Usually one of the two is about a subtly different claim than the one the author made.

The three steps on a real question

Here is a question at the middle of the difficulty range. Try it before you read on, and try it the long way, with the three steps, even though that will feel slow. I've deliberately not told you what kind of question it is.

Analyze

Something is being argued here, and it doesn't hold up. That's the reaction, and everything else follows from it.

Conclusion: adding professional landscaping to a house will increase the price it sells for. Apply the why test and it holds up, since the study is there to answer "why should I believe that?"

Evidence: in a study of recent sales, houses that had professional landscaping sold for about 10 percent more on average than houses that didn't.

The jump: the evidence establishes that two things went together. The conclusion claims that one of them produces the other. Stated without the stimulus's words: the author is treating the two groups of houses as though they were otherwise comparable, so that nothing except the landscaping could explain the price difference.

Predict

The question asks what would most undermine the argument. I have an argument that leans on the two groups being otherwise alike, so a right answer will make that harder to believe. Concretely: give me another explanation for the price gap, some other way the landscaped houses already differed from the unlandscaped ones.

That's the standard. Notice how little it depends on knowing what this question is called: I read the stimulus, saw what it was leaning on, read the question, and said what would have to happen. Now the choices are a matching exercise.

Eliminate

1. The landscaped houses were generally in more desirable neighborhoods. This is the prediction almost verbatim: a second difference between the groups that could produce the 10 percent premium on its own. If that's true, the sale figures are much weaker support for the columnist's claim, because the premium may be the neighborhoods rather than the landscaping. Keep.
2. Landscaped houses sold more quickly. Selling faster suggests buyers wanted them more, which points the same direction as the columnist's claim, not against it. Right idea, wrong direction. Cut.
3. Some homeowners landscape their back yards instead. Where people choose to landscape says nothing about whether front-yard landscaping raises a sale price. Out of scope. Cut.
4. Landscaping usually costs more than the price increase it produces. Cut, and this is the one worth slowing down for.
5. Buyers report noticing the front yard first. If anything this supports the columnist. Wrong direction. Cut.

The answer is A. For the record, this is what most people would call a Weaken question, and knowing that would not have saved you a second: the work was reacting to the stimulus and then answering the question that was asked. But the interesting choice is D, because D is where most of the missed points on this question go.

"I don't have time for three steps"

This is the first objection everyone raises, and it's worth taking seriously. A Logical Reasoning section gives you 35 minutes for around 25 questions, which is about 85 seconds each. Three steps sounds like a luxury.

But look at where the time actually goes when you skip prediction. You read five choices. Nothing is clearly right, so you read them again. You narrow to two, then read those two four or five more times, alternating, waiting for one to feel better. That process routinely burns two minutes and ends in a coin flip.

Analyze and Predict, on the example above, are quick. What they buy is a single clean pass through the choices with no second-guessing. The three steps aren't extra time. They move the thinking to the front, where it's cheap, instead of the back, where it's expensive and you're panicking.

They also compress with practice. At first, working a question this way takes several minutes and feels mechanical. Later, finding the conclusion is instant, you see what the argument is leaning on as you read it, and the prediction is ready by the time you reach the question. At that point it isn't really three steps anymore.

How to build the habit

Four things make the difference between knowing this method and actually using it under pressure:

• Practice untimed first. Speed on the LSAT is a byproduct of the method becoming automatic. It is not something you can add on top of a method you haven't internalized, and trying to do both at once usually means you get neither.
• Write the prediction down. For the first couple of weeks, physically write it before you look at the choices. Holding it in your head lets you quietly revise it once you see something attractive, which defeats the entire purpose.
• Cover the answers. With paper, with your hand, with a piece of card. If the choices are visible while you're analyzing, your eyes will drift and you will start evaluating before you have a position.
• When you miss one, find the step that failed. Did you read the stimulus as the wrong kind of thing, hunting for a flaw in an argument that didn't have one, or arguing with a set of facts? Misidentify the conclusion? Name the wrong jump? Answer a question other than the one asked? Set a good standard and then get talked out of it? Those are different problems with different fixes, and "I got it wrong" tells you nothing about which one you have.

That last point is worth more than it looks. Over a few weeks, the pattern of which step keeps breaking tells you what to do next in a way your practice test score never will. It's also why reviewing the questions you got right but felt shaky about pays: a lucky guess and a solid one look identical in your accuracy stats.