Choose Between Two Strong Answers

When two answers both look right

By this point the basic question types are familiar. A main idea can be separated from a detail. An inference has to stay inside the evidence. A semicolon cannot join a fragment. None of that disappears when a question gets harder. What changes is the amount of room left for imprecision.

The frustrating version arrives after the easy work is already done. Two choices are gone. Two remain. Both use the right topic. Both contain something the passage really says. Reading the passage again from the beginning rarely helps, because most of both answers was never actually in dispute.

When two answers look right, stop asking why each one is right. Ask where they stop agreeing.

Put the two survivors next to each other. Ignore the words they share. Find the smallest meaningful difference between them: a verb, a qualifier, a speaker, a time phrase, a comparison, a causal link, a grammatical relationship. Turn that difference into one question the passage can settle, and go back to the passage for that question only.

The method in five moves

1. Get to two. Eliminate what is clearly wrong first.
2. Compare the survivors directly against each other, not against the passage.
3. Mark the exact word or relationship where they diverge.
4. Turn that divergence into a question.
5. Return to the passage only for the evidence that settles that one question, and choose the answer that stays inside what the passage actually establishes.

A worked example

A research team compares two explanations for a coastal marsh losing plant cover: rising salinity, or grazing by an introduced crab. In fenced plots that excluded crabs, plant cover recovered even though salinity stayed high. The researchers conclude that grazing contributed substantially, while noting that salinity could still affect plant growth. The question asks what the fenced-plot result does to the salinity explanation.

Suppose the last two choices are "It rejects the possibility that salinity affects marsh plants" and "It shows that grazing can account for part of the loss without ruling out a role for salinity." Both agree that the fenced plots matter. They part ways on a single question: does the result leave any room for salinity at all? The researchers' own last sentence says yes, and the second choice is correct. Once the difference is visible, the passage has only one small job left to do.

Ten ways survivors diverge

Ten categories cover almost every way two survivors actually diverge. Learn the category, not just the individual example, and a new pair of close answers stops feeling unfamiliar.

The point is not to memorize a list of paired words. The point is to get fast at finding the part of two close answers that actually needs deciding.

Round 1: compare the exact wording

Each of these patterns comes up again and again. The practice questions that follow drill them.

• Strong evidence with remaining uncertainty. A randomized, one-summer experiment earns a cause-and-effect claim inside the test it ran, and nothing beyond it. An answer that adds "proves" and "under all weather conditions" turns one finished experiment into a second, untested one.
• Sample evidence and population scope. Eighty-four letters from three ports establish something real about those letters. "Merchants throughout the Mediterranean" quietly makes a much larger claim than the sample can carry.
• Possibility and prediction. An estimate is not a finished plant. "May reduce" and "will reduce" differ because the test that would turn may into will has not happened.
• Challenge versus rejection. New evidence can narrow a claim without collapsing it. If the author of the account keeps arguing part of it, the evidence "questions" the claim; it does not "reject" it.
• Exception size and claim scope. "Generally shifted" is a majority claim. A handful of holdouts do not defeat a claim that never promised zero exceptions.
• One contributor among several. When a report names two plausible contributors and says it cannot separate them, the safe answer credits one factor as a possible contributor, not the sole cause.
• Observed timing and historical priority. "Oldest recovered" is not "first ever made," especially when the passage says older layers were destroyed.
• The required comparison. Every number can be correct and still answer the wrong question. A claim about improvement needs the change (final minus starting score), not the largest final score.
• Percentages hidden by raw counts. The library with more damaged books can have the lower damage rate once the size of each pool is accounted for.
• Topic match versus mechanism match. Say the claim as "X happened because Y." Evidence about the same topic but a different mechanism (dye shortage instead of vanishing buyers) is a rival, not support.
• Shared data, contested interpretation. In cross-text questions, find whether the second author disputes the measurements or only what they add up to. Do not invent a disagreement the second text never makes.
• Observed behavior versus inferred motive. Behavior is visible on the page; the reason for it usually is not. Stay with hesitation, not with a specific feeling.
• True detail versus whole-passage scope. A true detail can be too small to be a main idea. Ask what the detail leaves out.
• Attribution across speakers. Keep a running note of which claim belongs to which name, especially when a passage starts to sound as if its voices agree.
• Method versus finding. A passage can describe its procedure in enough detail to feel like a result. If the stem asks what was found, procedure is the wrong answer.

Round 2: compare grammar and structure

The same move works on grammar questions. Find the small decision hidden inside a long sentence.

• Find the small decision in a long sentence. Strip everything between the subject and the blank. Plural nouns inside phrases never control the verb.
• Track the clause until it closes. If a comma opened a supplement, the supplement must close before the main verb arrives, even after a long list inside it.
• One subject with two predicates. When one subject does two things, the join needs only the word "and." Length alone never creates a second subject.
• Restore normal order in inverted sentences. Turn the sentence back into subject-verb order, then check the number of the true subject, including compound subjects.
• Finite verb versus participle. An opening phrase is not a clause. The sentence still needs its own finite verb.
• Pronoun antecedents. A pronoun can agree grammatically with two nearby nouns and still be unclear. When the logic works only one way, naming the noun removes the ambiguity.
• Let the final words set the relationship. Finish the sentence before choosing a transition. "Similarly" looks fine until you read that the second group's improvement was smaller.
• Sequence versus consequence. A program starting next to a larger count does not mean the program caused the count. "By then" tracks time; "as a result" asserts cause.

Synthesis: two gates, not one

On notes-based questions, a choice must pass two tests: every fact must be true to the notes, and the sentence must do exactly what the goal says. A sentence built from accurate notes can still fail the assignment if it breaks the one rule the assignment stated, such as "do not discuss awards." Combining true fragments is not the same as producing a true sentence: each fragment must keep its exact status. A proposed design is not an approved one, and a scale-model test is not a finished observatory.

Quick pairs to practice in your head

For each pair, name the exact difference before choosing. A few examples of the kind of reasoning that settles them:

• Vocabulary. "The union's position yielded to a compromise." The preposition decides: "yielded to" means gave way to, not produced.
• Same direction, different degree. A revision that changes several claims but leaves the argument intact is "restrained," not "minor."
• Association is not meaning. A room with bare walls and straight-backed chairs is "austere," not "cruel."
• Function. A correct rhetorical label attached to the wrong sentence is still the wrong answer. Qualification belongs to the sentence that qualifies.
• Percentage points versus percent change. A rise from 20% to 30% is 10 percentage points but a 50 percent increase.
• Rate versus risk. "4 incidents per 1,000 device-hours" is not the same as one device's 4-in-1,000 chance of an incident.
• A conditional does not run backward. "If the catalyst is contaminated, the reaction slows" does not mean a slowed reaction proves contamination.
• No measurement is not no effect. A sensor that failed below -20°C tells you about the study's equipment, not about zero flexibility.
• Joint possession. One shared proposal by two editors puts the apostrophe on the second name only.
• Long choices. If two answers share twenty identical words, compare only the phrases that differ.

Check whether a true statement answers the question

Every comparison in this unit tests the same idea: a choice must clear two gates. It must be factually accurate, and it must actually answer what was asked. A choice that clears only the first gate is not a weaker version of a correct answer. It is still wrong. This failure shows up in six familiar forms:

1. A true detail is too small for a main idea.
2. Evidence is related to a claim without supporting its specific mechanism.
3. A plausible inference adds a motive the passage never gave.
4. A synthesis answer is entirely true but serves the wrong goal.
5. A cross-text answer invents a disagreement the second text never makes.
6. A quantitative answer uses real numbers to answer the wrong comparison.

The fix is identical every time: name the exact target the question asks for before judging whether a true-sounding choice actually hits it.

What to carry forward

Two answers can both be partly right. That is not a reason to reread everything. It is a reason to make the disagreement smaller. Keep population, time, certainty, cause, comparison, attribution, and grammatical structure exactly where the text puts them.

When the choices still feel close, run one more check: for this answer to be true, what else would have to be assumed? The problem, when there is one, is usually an unsupported bridge: a motive the text never gives, a cause the design never isolates, a population the study never reaches, a future test that has not been run, or a speaker who never made the claim.

You do not need to become suspicious of every strong word. Sometimes the passage earns a strong answer outright. The goal is to make the strength of the answer match the strength of the evidence behind it.

A useful error log sounds like this: changed may to must; used the final value instead of the change; used the right topic but the wrong mechanism; attached a claim to the wrong speaker; turned behavior into motive; let a nearby noun control the verb; reread without knowing what was being looked for; changed an answer without finding new evidence. That level of diagnosis is the point. The question may still be dense and the two choices may still look annoyingly close, but the disagreement is smaller now, and small disagreements can be checked.