Inference Questions: The Black Sheep of the GMAT Critical Reasoning Family
A quick note: this is a pretty deep dive into a single GMAT Critical Reasoning question type. If you’re just beginning to learn CR strategy, check out The GMAT Critical Reasoning Mindset or How to Master Every GMAT Critical Reasoning Question Type
Inference questions are not super common on GMAT Critical Reasoning, usually only accounting for 1 of your 10 CR questions. However, it tends to be a question type that students miss more frequently, in both CR and Reading Comprehension. Some of this stems from the inherent difficulty, but much of it can result from students’ possessing an incorrect or incomplete sense of what they’re supposed to be doing on these problems.
There are seven main question types in Critical Reasoning: Explain Discrepancy, Role of Bold, Strengthen, Weaken, Evaluate, Assumption, and Inference.
Inference question types are pretty unique. Unlike Assumption, Evaluate, Strengthen, Weaken, and Role of Bold, Inference questions are not based on arguments. And unlike Assumption, Evaluate, Strengthen, Weaken, and Explain Discrepancy, Inference questions are not usually about presenting your brain with some form of cognitive dissonance.
RELATED: Top 10 Tips for GMAT Critical Reasoning
How to Learn GMAT Critical Reasoning
In addition to learning the patterns surrounding analyzing Plans, Predictions, and Causal Explanations, we should also be learning a little index card’s worth of technique for each of these seven main question types.
For each question type, we’re trying to memorize the following:
- What keywords in the question stem tell me it’s this type of question?
- What am I reading for in the paragraph?
- How, and to what extent, should I pre-phrase a potential correct answer?
- Are there any tendencies relating to the paragraph or the answer choices?
So let’s make sure everyone has a great index card for Inference questions.
What Keywords in the Question Stem Tell Me It’s an Inference Question?
Here are a few examples of Inference question stem wording:
– If the statements above are true, which of the following must be true?
– If the information above is correct, which of the following is most strongly supported?
– The claims above most strongly support which of the following assertions?
– Which of the following can be properly inferred from the passage?
One of the most salient features is the noun being used to describe the paragraph we read. Remember, when we’re doing Strengthen, Weaken, Evaluate, and Assumption, we’re going to almost always see the paragraph described as one of the following: argument, plan, prediction, or hypothesis. Those nouns connote the idea that there will be an opinion within the paragraph:
Arguments have an opinionated conclusion, based on some (untouchable) evidence.
Plans have the opinion that if we follow this plan, we will achieve the stated goal.
Predictions contain a conclusion that is in the future tense, thus an opinion.
Hypothesis means that our author will be opining some ‘causal explanation’ for a curious fact.
With Inference, you see that the nouns being used connote that we’re just reading some facts: statements, information, claims, or passage
Inference may be asked in the “must be true / properly inferred” style (indicating 100% provability), or they may be asked in the “most strongly supported” style (indicating that the correct answer is the most provable claim, even if not 100% provable).
Students often confuse “most strongly support” Inference question stems with a Strengthen question stem.
INFERENCE: The statements above most strongly support which of the following conclusions?
STRENGTHEN: Which of the following, if true, most strongly supports the argument?
The two big differences are that Inference deals with statements/information, while Strengthen deals with arguments/plans/hypotheses, and that in Inference questions the paragraph provides support for the correct answer choice, while in Strengthen questions the correct answer choice provides support for the paragraph.
Now you tell me, how do you tell the difference between an alligator and a crocodile? Or better yet, one of my favorite SNL premises ever, how do you tell the difference between Dylan McDermott and Dermot Mulroney?
What am I Reading For on an Inference Question?
Here’s the real reason for the season. I wanted to write this blog because any time I’ve asked students this question, I’ve always gotten blank faces or wrong answers. People are tempted to say something like, “The conclusion and premise?”, “the assumption?”, “the gap?”
These answers indicate that students are still in default Critical Reasoning mode, thinking that they’re about to read another argument or plan. Remember, Inference is specifically not giving us arguments and plans. We’re just getting two or more facts and being asked which answer is derivable from those facts.
To draw a valid inference = to draw a valid conclusion. Validity means that we can derive an idea from the available evidence without speculation or exaggeration.
In Reading Comp, correct answers to Inference questions usually paraphrase something we know from the passage, but they use new and unexpected wording or some sort of inverted syntax. Pretend that in line 20 of an RC passage we are told that “George Washington wore a military cap to his Inauguration as the first President of the U.S.”
Can we infer that
(A) He wore the cap because it was a cold January day.
(B) All former generals have to wear military caps.
(C) He thought it flattered his deep blue eyes.
Of course not. None of those are provable claims. They are speculations or overstatements.
We can infer weird restatements such as these:
(A) Not all Presidents are sworn in hatless.
(B) On the first day of the American Presidency, at least some parts of the first President’s scalp were not visible.
(C) People seeing George Washington for the first time at his Inauguration could not have conclusively determined Washington’s level of baldness.
The personality of CR Inference questions is a little different: the correct answer almost always pulls together two or more facts provided.
Pretend we read a CR Inference paragraph that said, “George Washington wore a military cap to his Inauguration as the first President of the U.S. He wore this same cap during the Battle of Yorktown.”
Can we think of a safely worded claim that pulls information from both sentences?
A) Presidents do not always begin their terms wearing totally new garments.
“Presidents begin their terms” pulls info from the first sentence. “Not wearing totally new garments” comes from knowing that the cap he wore on Inauguration day had been worn before at the Battle of Yorktown.
So, in summary, what are we reading for when we read an Inference paragraph? We are reading for two or more facts that could be synthesized in order to derive a true claim.
Very often, this synthesis comes about because the two or more facts contain some overlapping piece of information (in this case, both facts referenced GW’s hat). There are four main types of Inferences, which I will detail a little later:
1. Math-y
2. Apply a Rule
3. Causal
4. Straddle the Pivot
How, and to What Extent, Should I Pre-phrase a Potential Correct Answer on Inference?
We are reading to see if we can combine two or more of the provided facts to derive some true claim (or incredibly likely claim, if we’re doing “most supported”). When we successfully find an available inference we can make from the paragraph, we should certainly anticipate that the correct answer will probably reinforce or reward that.
But we should stay very flexible. Ultimately, the only standard of right or wrong on Inference is, “Could you prove this answer choice, using only the information provided in the paragraph?”
So the correct answer is under no obligation to tie everything together or to present the ‘coolest’ takeaway. The correct answer just has to be the most provable claim. Keep any inference you discovered in your mind as your mantra of what the correct answer will probably sound like, but give each answer choice a fair hearing by asking, “Could I prove this, using the statements I just read?”
What will frequently happen with correct answers is that they will be a spin-off inference of what we inferred. For example, if we inferred “Spain outscored France in the first half”, the correct answer might say “Spain scored at least once in the first half”. If we inferred “the cost of paying for parking was more than the combined costs of taking the train and taking a Lyft from there”, the correct answer might say “The Lyft ride was cheaper than the cost of paying for parking”.
Those answers can feel annoying, because we’re like, “Yes, but I know even more than that!” It doesn’t matter. You can sign off on the truth of those answers, so they are correct. In general, expect that when you walk out of your real GMAT, you’ll reflect on all the stuff you studied that you never even saw on your test and feel like, “Yes, but I know even more than that!”
“Why did I spend a week working on Combinatorics, only to have ZERO combinatorics questions on my exam!”
“Sir, I’m just a janitor at the Pearson testing center. I’m not sure why you’re screaming at me.”
Are There Any Tendencies Relating to the Paragraph or the Answer Choices?
As I hinted before, there are four main types of Inferences, so we can learn to read the paragraph while seeing if we pick up on the scent of any of these.
- Math-y Inferences
- Spain beat France in regulation. However, France scored more goals in the 2nd half.
Infer: Spain scored more in the 1st half / Spain’s margin of victory in the 1st half was larger than France’s margin of victory in the 2nd half - My rent is scheduled to increase next month. Nonetheless, it will represent the same proportion of my income.
Infer: My income is also going to be higher next month (increasing by the same multiplier)
- Spain beat France in regulation. However, France scored more goals in the 2nd half.
- Apply a Rule Inferences
- For a grilled cheese to be delicious, the cheese must be melted into liquid form and the bread must be toasted but not burned. The cheese Patrick is using to make grilled cheeses for his wedding feast is so cold and dense that liquefying it will take at least five minutes of high heat, which is more than enough time to burn the bread he’s using.
Infer: At least one item at Patrick’s wedding feast will not be delicious.
(many different ways we could state the inference, but applying the rule to my specific situation tells us that these will not be delicious grilled cheeses)
- For a grilled cheese to be delicious, the cheese must be melted into liquid form and the bread must be toasted but not burned. The cheese Patrick is using to make grilled cheeses for his wedding feast is so cold and dense that liquefying it will take at least five minutes of high heat, which is more than enough time to burn the bread he’s using.
- Causal Inferences
- Patrick always buys his shampoo at CVS. Recently, CVS stopped offering Herbal Essences shampoo. As a direct result, the cost of Patrick’s hair care went up.
Infer: Patrick was previously buying Herbal Essences. He is now buying a different shampoo, and that new shampoo costs more.
- Patrick always buys his shampoo at CVS. Recently, CVS stopped offering Herbal Essences shampoo. As a direct result, the cost of Patrick’s hair care went up.
- Straddle the Pivot Inferences
- Patrick is a mean teacher. However, he gives his students candy.
Infer: Not all candy-giving teachers are nice. - People think that New York City is the most expensive city in the country. Yet, the cost per square foot of real estate in San Francisco is much higher than that in New York City.
Infer: New York City is not necessarily the most expensive city in the country.
- Patrick is a mean teacher. However, he gives his students candy.
Math-y and Rule-based inferences tend to feel more like Must Be True. Causal and Straddle the Pivot tend to go more with Most Strongly Supported.
What About Answer Choice Tendencies?
Since we are trying to find the most provable claim, this is a question type for which strong or new language in the answer choices should be a big red flag.
There are three big categories of strength of language:
CERTAIN: all, only, never, unless, requires, must
MORE THAN 50%: most, typically, generally, usually, likely, tends to, probably
AT LEAST ONE: some, sometimes, can, may, might, not all, not always, need not
When you’re doing Reading Comp, Assumption, or Inference, you should always consider the two stronger levels of language to be red flags; this doesn’t mean they’re automatically wrong, but it means you have to research in the passage whether you’re justified in saying something this strong.
Meanwhile, when you’re doing Strengthen, Weaken, or Explain Discrepancy, all of which begin with the words “Which of the following, if true, most …”, then the weakest level of language is a red flag. An answer choice isn’t going to have much impact if it’s only saying “at least in one case this is true”.
Inference Questions: Takeaways
In summary, if we want to improve at Inference questions on RC and CR, we need to remember that we’re not allowed to speculate or exaggerate. We’re only allowed to pick answers we feel like we can derive from the provided information. If nothing is 100% provable, then pick the most provable option. If you need to guess quickly, avoid strong language.
For CR, you can go one layer farther and proactively read the paragraph looking for facts that can be combined. In particular, if you see quantified wording, look for a math-y inference. If you see causal wording (e.g., “because of this”, “due to”, “this allows”, “this makes possible), look for a causal inference. If you see a Rule (e.g., “if/then”, “always”, “only”, “ensures”, “requires”, “guarantees”), look for an inference you can make by applying that rule to a specific situation. If you see the paragraph is divided up by a but / yet / however, think about what safely worded claim you could create that would integrate both sides of that pivot.
And if you’d like to submit an entry into a drawing contest I’m sponsoring, provide one drawing of a crocodile eating Dylan McDermott and another one of an alligator eating Dermot Mulroney.
You can attend the first session of any of our online or in-person GMAT courses absolutely free! We’re not kidding. Check out our upcoming courses here.
Patrick Tyrrell is a Manhattan Prep instructor based in Los Angeles, California. He has a B.A. in philosophy, a 780 on the GMAT, and relentless enthusiasm for his work. In addition to teaching test prep since 2006, he’s also an avid songwriter/musician. Check out Patrick’s upcoming GMAT courses here!