Think like a test writer

One great way to start smashing test questions is to start thinking like a test writer, instead of a test taker.  You’d better believe that the test writers spend tons of time studying how you think,* so don’t you think you should return the favor?

If you were assigned to write questions for your choice of test, what would your objective be?  Well, this is a standardized test, and the standardization means that the people who take the test this year have to have basically the same experience as last year and the year before, only the questions have to be completely different.  Not only does it have to feel the same, but the same person (who hasn’t done any test smashing) should get the same score every time she takes the test, within some narrow band.  And in order for that to happen, you have to know as precisely as possible, how many people — and of what “sort” — will get your question correct or incorrect.  Of those who get the question incorrect, you should have a pretty good idea of the distribution of which wrong answer they will choose.  The hopeless test takers will go for one wrong answer, and the almost-there test takers will, perhaps, fall for a different trap.

Every time you take an official test, you are contributing your lab rat skills on those experimental sections.  (Does that make you feel good?  You are contributing to science!  For free!)  That is where they test and calibrate their questions.  But you’d better believe that they aren’t writing random questions that they then learn about on the experimental section: they already have a pretty good idea what is going to happen by the time the question shows up on your test.

So, if you are the manager of a bunch of question writers, what are you going to do?  Well, there have to be templates that writers share, so that Sarah’s questions look the same as Karim’s questions.  And, dear TestSmasher, if you can figure out the templates, it is like having a master key to the test — or a sledge hammer!

This could work for any question type, but I think arguments are particularly well suited for this exercise, which is to sit down and write (at least) 3 questions in the style of your test:

  1. Pick one stem (maybe a question type you hate?) and use it for all three.
  2. Find a couple of official questions of the same type.
  3. Analyze the best you can the argument structure: number of sentences, the role they serve, etc.
  4. Choose 3 totally unrelated topics
  5. Craft an argument on each topic that conforms to the structure
  6. Write at least three answer choices (five might get tedious)

Most of the art of question writing lies in writing the answer choices, so let’s drill down into that one.

The correct answer has to be:

  1. Indisputably correct.  (Think about aggrieved test takers and their lawyers.)
  2. Not obvious.

How do you do both at once?  The way I would go about it is write an obvious and correct answer choice and then revise by picking synonyms or other phrases that mean the same thing but aren’t the way you would think to characterize the correct answer if you were a normal person instead of a test writer.

The wrong answers have to be:

  1. Indisputably incorrect.  (Think lawyers again!)
  2. Attractive.

And how do you do that?  Well, you say things that are correct but not answering the question, of course!  Or things that are in response to the question but are slightly, subtly, but fatally wrong.   Write at least one of each.

Once you have done this exercise, first, post your fake questions in the comments — I’d love to see what you come up with.  Then, go do a set of real questions with the same stem.  Do you approach the questions differently?  Identify the wrong (or right) answers more consistently?  Do you catch yourself before you fall into a trap that used to trip you up?

*There is an entire scientific field, psychometrics, that studies how you think and respond to questions designed to measure something about your cognition, psychology, or character.

Advertisement

A note on question types

In my last post, I was a bit snarky about the tendency of test prep companies to develop complicated taxonomies of, say, argument question types.  I’d like to be a bit more nuanced about that.  I like simple taxonomies, ergo my DESTROY, SUPPORT, DESCRIBE for arguments, and I do think that all of the arguments fit into one of those three broad categories.

HOWEVER.  In the course of your studying, if you start to notice that it helps you to think about “inference” questions differently than “conclusion” questions, that is fantastic.  If it identifies for you a difference that is meaningful to you in what approach you take/traps to look out for, then you know you are studying and learning well.  My only quarrel is when you spend a ton of time (and money) learning someone else’s taxonomy, with nuances that don’t mean a whole lot to you.  Then you stress out a bunch over your inability to see any difference between an inference or a conclusion question, when at the end of the day, there isn’t a difference, and it shouldn’t cause you any stress, because if you weren’t so fucking stressed out, you’d get the right answer.

This applies to other question types as well (RC, LSAT games, math questions on the various math-y tests).

A note on arguments on the GMAT

In recent years, the GMAT has been trying hard to become more “relevant” to business school.  So, a decade ago, its arguments were just slightly dumbed down versions of the LSAT arguments.  Now, some of their arguments (but by no means all, see the first worked example for a classic one) have a more “business decision” feel to them.  These questions fit less cleanly into the basic structures of the question type, though at root, the rules are the same.  However, to smash these questions, you need to have a stronger intuition about how these questions work so that you can apply the rules more flexibly.

As a consequence, I actually recommend to GMAT studiers that you go buy some LSAT tests — they sell individual tests, which might be enough, or books of 10 tests, which, if this stuff is hard for you right now, might be a good investment.  These questions are good for building your foundation for two reasons: they have very transparent, predictable structures, and the types of logic they use can, on harder questions, be more sophisticated than what you will see on the GMAT.  This builds extra capacity and good intuition.  Then you can go work on the GMAT questions with a much stronger foundation.

Arguments: Working an example II

This argument is Q26 in the Critical Reasoning Practice Questions in the 13th Ed. of the GMAT Official Guide.  Let’s work through it in the steps I discuss in this post.

1.  The stem reads: “Which of the following conclusions can most properly be drawn from the information above?” This may sound like it doesn’t fit neatly into my DESTROY, SUPPORT, DESCRIBE framework, but it actually does.  When you are asked to draw a conclusion from an argument, you may not make any assumptions.  This means that the correct answer is an amalgamation of what is in the argument already, and therefore is a DESCRIBE question.

2.  The argument reads:

When a polygraph test is judged inconclusive, this is no reflection on the examinee.  Rather, such a judgment means that the test has failed to show whether the examinee was truthful or untruthful.  Nevertheless, employers will sometimes refuse to hire a job applicant because of an inconclusive polygraph test result.

This is actually all premise (you get to come to a conclusion), though the facts have some discrepancies.

3.  I do not worry about how I happen to know that polygraph testing in general is highly unreliable with lots of false positives.  This is not relevant.

4.  Okay, I don’t need to look for assumptions, I just need to describe.  But it is still useful to simplify the argument to build my description:

(1) Inconclusive = meaningless re applicant.  (2) Inconclusive = bad test.  (3) Inconclusive –> employers turn down applicants.

Which answer choice combines these 3 statements accurately?  Let’s do this by process of elimination.

  • (A) Most examinees with inconclusive polygraph tests are in fact untruthful.  This says that Inconclusive = usually untruthful re applicant.  Contradicts (1).
  • (B) Polygraph tests should not be used by employers in the consideration of job applicants.  This says that Inconclusive + Conclusive results = useless.  We know nothing about Conclusive results.
  • (C) An inconclusive polygraph test result is sometimes unfairly held against the examinee.  We know that Inconclusive = meaningless re applicant (1), and Inconclusive –> bad outcome for applicant.  Since suffering a bad outcome for something that you did not control can be described as unfair, this answer choice is not totally sucky.
  • (D) A polygraph test indicating that an examinee is untruthful can sometimes be mistaken.  The argument tells us nothing about conclusive results. Note Step 3!!!
  • (E) Some employers have refused to consider the results of polygraph tests when evaluating job applicant.  The argument only tells us about how employers use polygraph results, nothing about how they don’t use results.

You should be hyper wary of (C) because of its use of a value-laden word “unfairly.”  Almost always, these answer choices are traps.  But here is an exception worth noting, and you can be confident that it is an exception because the other choices are so badly wrong.

Arguments: Working an example I

This argument is Q2 in the Critical Reasoning Practice Questions in the 13th Ed. of the GMAT Official Guide.  Let’s work through it in the steps I discuss in this post.

1.  The question stem is: “The argument is flawed primarily because the author”.  So, I write down DESTROY on my paper.

2.  The argument is:

Homeowners aged 40 to 50 are more likely to purchase ice cream and are more likely to purchase it in larger amounts than are members of any other demographic group.  The popular belief that teenagers eat more ice cream than adults must, therefore, be false.

The first statement fits in the premise slot, the second in the conclusion slot.  How do I know?  The first sentence is stated as a fact.  The second sentence uses the telltale “therefore” to indicate that it is the conclusion.

3.  I do not indulge in thoughts about how 30-something single females are inclined to eat large quantities of ice cream after romantic missteps!

4.  Since I didn’t write down DESCRIBE, my work is not done.

5, 6.  A shortened, focused summary of the argument:

H buy most X.  Therefore, T do not eat most X.

What is the important shift between premise and conclusion?  buy –> eat.  So I need to find an answer choice that phrases that in a way that DESTROYS the argument.

7.  Looking at the answer choices, A pops out:

fails to distinguish between purchasing and consuming

Since purchasing = buying and consuming = eating, we have a perfect match.

8.  I read again the stem, yes, I am supposed to DESTROY.  A is phrased in a way, “fails to distinguish,” that DESTROYS.

Smash!!!

Arguments

These are the Critical Reasoning questions on the GMAT and the Logical Reasoning questions on the LSAT.  Indeed, the approach to the short reading comprehension passages on the GRE should be approached blending the Arguments and Reading Comprehension question approaches.  There is no equivalent question type on the SAT.

Important:  If you are taking the LSAT, arguments are 50% of the test.  Games and Reading Comprehension are each 25%.  TestSmashing Arguments is the most valuable thing you can do to up your LSAT score.  Not to mention there is a lot of spillover between Arguments and RC.  Games are the strangest (and most fun) part of the LSAT, so they often suck up most of the attention of LSAT studiers.  This is a mistake.  If you can avoid that temptation, you are ahead of the game already.

Okay, ‘nough warm up.  How to smash arguments:

1. Read the question stem first.  A lot of prep companies spend a lot of time categorizing the question types in elaborate ways, and that isn’t a terrible thing to do, but I also think it isn’t the most important thing to worry about.  I would, however, write one of three things down on the test next to the question (or on scratch paper for the GMAT): DESTROY, SUPPORT, DESCRIBE.  Or symbols that are meaningful, like X, (check), ~.  (Spending time writing lots out is not the TestSmasher way, unless it is necessary.)

2.  Read the argument with one main purpose: which bit of the argument fits into which slot?  There are two slots: Premise and Conclusion.  They take the form, Premise: A is X.  Conclusion: Therefore, B must be Y.  In other words, there is a statement of fact (the premise) and then a statement about something else that must also be true (conclusion) because of the original fact.  The order in which these items come is not consistent.  Sometimes the conclusion comes first.  It is very important to know which is which.

3.  Do.  Not.  Ever.  Engage.  The substance of the argument is irrelevant.  The truth in the real world is irrelevant.  This is how people get snookered on arguments.

4.  If you wrote down DESCRIBE on your paper, you are done.  Look at the answer choices and, like in reading comp, prove that every word in your answer choice describes what is happening in whatever part of the argument you have been asked to describe.

5.  Otherwise, you have to figure out the mechanism by which the author of the argument got from Premise to Conclusion.  Unless the argument is:

A is X.  Therefore, A is X.

(and very, very occasionally, you will see an argument that is a tautology like the one above) there is an Assumption involved in the argument.

6.  The key to most argument questions is realizing that there is only a small set of structures to the arguments.  If you can see past the substance and the crappy writing style, you’ll learn that there are maybe 5 or 6 arguments on the tests.  Once you are able to see that, identify the class of argument, know how the correct answer is structured, you have smashed Arguments!!!

Some of the underlying structures are:

A is X.  Therefore, B is X.  Answer choice deals with the question: Is A = B?

A is X.  Therefore, A is Y.  Answer choice deals with the question: Does X imply Y?

There are a handful of these structures, and the prevalence and flavor are somewhat different between the LSAT and GMAT.  You should find a way to characterize the structures in a way that is meaningful to you, and your study time on arguments should be about noting and understanding the structure of arguments.

7.  Which word did you write down?  If you wrote DESTROY, and the argument is such that the answer choice deals with the question of, Is A=B?, then you look for an answer choice that says “B is not the same as A.”  If you wrote SUPPORT, the correct answer would be “B is indeed the same as A,” at least for the purposes of the argument.

8.  Double check that you executed the correct action on the argument.  The other way people get snookered is that they DESTROY instead of SUPPORT or vice versa.  This usually happens when they ask you to DESTROY a sensible argument or SUPPORT a really stupid one.  If you are not engaged with the substance of what you are reading, you won’t make this mistake as much.

Note: One frequent question type is “What assumption does the argument rely upon?”  This sounds DESCRIBE-y, but, since it deals with the assumption it isn’t.  Whether it is DESTROY or SUPPORT depends a bit on your temperament.  TestSmashing makes me snarky, so I like to think of identifying the assumptions as an act of destruction.  Some of you may feel more nurturing than me, and see identifying assumptions as a road to supporting the argument by making it stronger.  So interpret the question in the way that speaks to your perspective, but in the end the answer will always be “the argument assumes A = B” or “the argument assumes that X implies Y.”

How do you know you have smashed Arguments, once and for all?

Well, you’ll get them 100% correct, all the time.  I mean 100%.  Not 80% or 90%.  And, you will start to weird yourself out, since you’ll read the argument, have a pretty good idea of what the correct answer will say, and read the answer choices only to realize you predicted the exact wording of one of the answer choices.  That should scare you, because answer choices that sing to you are often traps, but if you are test smashing arguments in particular, you will know what the right answer should say almost to the word.  Of course, sometimes they will choose a bizarre route to the right answer.  Sort of like “X doesn’t imply Y when aliens invade,” but you should, by now, recognize that indeed it is true that X doesn’t imply Y when aliens invade, this is the form of answer I am looking for, therefore this must be right, bizarro though it may be.

I’ll do some worked examples in subsequent posts.