Back to blog

GRE Vocabulary: Cards That Survive the Question Format

GRE verbal puts the word inside a sentence instead of asking for a definition. How to build vocabulary cards in that shape, and how to card near-synonym pairs.

If you are memorizing GRE vocabulary with the word on the front and the definition on the back, you are drilling a move the test never asks you to make. The vocabulary questions on GRE verbal put the word inside a sentence. You either fill a blank so the sentence holds together, or you pick the choices that complete it so the finished versions come out meaning close to the same thing. Nothing asks for a definition. What you get instead is a handful of words whose dictionary entries are nearly interchangeable, and a sentence that only tolerates one or two of them.

So the front of the card should be a sentence with the word missing, or the word sitting in a sentence with its meaning hidden. That is the argument. The rest is how to build those cards without writing sentences all month, and how to tell when the deck has started teaching you sentences instead of words.

A little about me: I am a third year engineering science student at the University of Toronto, machine intelligence. I also work on Nebulearn, a flashcard app, so treat that as a conflict of interest when a tool comes up.

What the test asks you to do with a word

The card shape falls out of the question format, so start there. Both vocabulary-driven question types hand you a sentence or short passage with something missing, plus a set of candidate words. One asks you to fill the blanks so the passage makes sense. The other asks which choices complete the sentence so the completed versions come out alike in meaning. What is scored both times is compatibility: does the word fit the logic of the sentence, and does it fit the attitude.

No question type hands you a bare word and asks what it means, or asks for its opposite. That gap is the whole problem with the standard vocabulary deck. A word-to-definition card is a self-contained unit and the test never gives you one, so the skill you need is: given a sentence, produce or verify the word. Why the other direction does not transfer is visible as soon as you look at the answer choices.

Six words, one dictionary definition

Take the family of words for saying very little: laconic, taciturn, reticent, terse, concise, succinct. Write six cards with those on the front and every back ends up saying some version of "brief, using few words." You will learn all six backs, and you will have erased the only information that could have helped you choose between them.

Here is what separates them in use:

  • taciturn is a temperament: quiet by disposition, across topics.
  • reticent is about a subject: holding back on this thing in particular.
  • laconic is brevity that reads as cool, dry or faintly withholding.
  • terse is brevity that reads as curt, often a little rude.
  • concise and succinct are compliments, and describe expression rather than mood.

Now a sentence that forces the pick: "He would happily talk about the weather all afternoon; only about his years in the army was he ______." The answer is reticent. Taciturn is not wrong because the dictionary says something different. It is wrong because the sentence already told you the man talks plenty, so a claim about his disposition contradicts the first half. The sentence did the discriminating, and a definition card never showed you a sentence.

Money words behave the same way: frugal and thrifty are neutral to approving, parsimonious and miserly are accusations, and a gloss of "sparing with money" covers all four.

Two card fronts that match the question

If the test scores sentence fit, only two fronts are worth building, and they are mirror images. The first puts the sentence on the front with the word missing, and the back is the word plus one line on why the nearest neighbour fails:

Front: Her cooking was ______ rather than parsimonious: simple ingredients, nothing wasted, and nobody left the table hungry.

Back: frugal. Sparing without the stinginess. "Rather than parsimonious" is the tell, so the blank has to be the approving version.

The second puts the word inside a sentence and hides the meaning. You answer with a paraphrase of the sentence, not a synonym:

Front: The reviewer's fulsome praise made the author uncomfortable. What is fulsome doing here?

Back: Excessive, cloying, laid on so thick it reads as insincere. Not a fancy word for generous, which is why the author squirms instead of preening.

Both keep one distinction per card, the same atomic encoding rule that applies everywhere else. The temptation with vocabulary is to stack four near-synonyms onto one back because they arrived from the same list. Resist it: a card with four words on the back cannot be graded honestly, and a card you cannot grade honestly stops feeding your scheduler real information. One addition does earn its keep on both templates. Tag the back with compliment, insult or neutral. One word of typing, and it handles a good share of the equivalence questions, which turn on attitude more often than denotation.

Card the pair, not the word

Once you are writing sentences, stop treating the single word as the unit. On the questions asking for choices that produce equivalent sentences, the decision is rarely between a word you know and a word you do not, but between two words you half know whose dictionary entries look identical.

So build the pair card directly. Front: "parsimonious or frugal: which is a criticism, and what in the sentence tips it?" Back: parsimonious, because it carries excess, and the tell is usually a consequence, someone going without something they needed. Same shape for ingenuous against disingenuous, where one is unguarded and the other is a performance of innocence. Same for prosaic, banal and insipid: banal means it has been said too many times, insipid means it has no flavour at all, prosaic means plain rather than lyrical. "The lyrics were not so much banal as insipid" only works as a sentence if you hold all three apart, and it beats three separate definitions.

A few words need that treatment against themselves. Sanction can mean approve or penalize, so give it two sentences on one card and name which sense each is running. Where the sentences come from then matters more than the template, because a card is only as good as the sentence on it.

Where the sentences should come from

Do not write them yourself. I tried that first, and it fails for a simple reason: you write a sentence that fits the definition already in your head, so the card confirms your gloss instead of testing it. You need sentences from people who were not trying to make your card easy.

Two sources are worth the trouble. The best is the explanation text in practice material, because an explanation names the exact nuance the question turned on, and that line belongs on the back of your card. The second is real prose in the register the test uses: essays, book reviews, academic abstracts, long-form journalism. Fiction is less useful, since novels reach for these words for voice rather than argument.

A word list's own example sentence sits in between, fine as a scaffold and rarely worth keeping, because list sentences are written to make the definition obvious. Which is the pattern under all of this: the sentences worth carding arrive while you are doing something else, and the weak ones arrive when you sit down specifically to make cards. That makes this a capture problem, not a writing problem.

Capture is the one place I will name a feature, conflict of interest already stated. Nebulearn has a Chrome extension that snips anything on the page into a card, and a highlight plus a keybind drops the sentence in front of you onto a card front without leaving the page. The free tier holds 1,000 saved questions, room for a word list plus pair cards. The point is not speed for its own sake. It is that the sentence you kept was written by somebody else, in the wild, which is what makes it worth reviewing at all.

The failure case: your deck teaches the sentence

Two weeks in, this method has a failure mode that looks like success. You hit your own card, fill the blank instantly, feel good, then stall on a practice question using the same word. You learned the shape of your sentence. Your brain found the cheaper pattern, as it always does, and matched on rhythm and neighbouring words instead of retrieving a meaning.

That is the trap I wrote about in cloze deletion versus question and answer encoding, where hiding a word in a sentence trains pattern matching rather than understanding. Vocabulary is the honest exception, since here the hole is the thing being tested, but it holds only if the sentence is not the only thing you keep. Four fixes, in the order I would try them:

  1. Two sentences per word, from different sources. The word becomes the only thing the two cards share, which is what you wanted to learn.
  2. Flip the direction. Turn the blank card into the hidden-meaning card for a cycle. Same word, different retrieval path.
  3. Promote the pair card. If the blank has gone stale, the discrimination question is usually still hard.
  4. Never card a sentence from a practice question you plan to retake. You will answer it off your own card.

Order matters too. If your vocabulary cards always arrive in the same run you start leaning on the run, which is the context dependency problem doing quiet damage, the same thing that leaves people with ghost words from premade language lists: passable in a session, invisible in a sentence. And if a pair keeps failing after many reviews, do not grind it harder. That card's sentence probably does not force the distinction it claims to test, so rewrite it before reaching for the leech protocol.

When a plain definition card is the right call

The gloss card is not useless, it just has a job that lasts about two days. You cannot judge whether a word fits a sentence with no approximation of what it means, so for a word you have never seen, build the definition card, run it until the word stops being noise, then replace it with a sentence card. The mistake is keeping five hundred scaffolds and calling that a vocabulary deck.

Some tools also beat building your own cards outright. If you want a thousand GRE words in your queue tonight with nothing to build, a maintained community Anki deck is genuinely the fastest route, and Anki's shared-deck ecosystem is unmatched. If you and a friend are quizzing each other off one list on a Friday, Quizlet has less friction than anything else for that hour. If your vocabulary notes are already threads inside a larger set of notes, RemNote wins, because the links are the point.

Whichever you land on, pick it before you build eight hundred cards: review history does not move between flashcard apps, so switching in month three means the words come with you and the schedule starts over. Premade decks earn their keep when you have no material of your own, and less here, since sentence quality is the thing you wanted to control.

What a month of this looks like

Keep the volume honest and none of this is heavy. My own rule of thumb, not a study: a handful of new pair cards a day beats a big list-clearing session, since pair cards are the expensive ones to review properly and the ones that pay off on the equivalence questions.

PhaseWhat I addWhat I review
First two weeksGloss scaffolds, disposableEverything, one short sitting
Middle stretchHarvested sentences, then pair cardsWhatever is due, no list clearing
Final weeksNothing newPair cards and recent misses

The last row is the one people get wrong. Late prep is not the time to add words, it is the time to let the scheduler surface the ones you are still missing, which is what FSRS is for and what a high-stakes exam schedule is built around.

The split on this page was never more words against fewer. It is definition recall against sentence fit. Definition recall is the easy deck to build, easy to review, and the one the question format does not score. Sentence fit is harder, because every card needs a real sentence from outside your own head, which makes the method stand or fall on capture. That is the part I built Nebulearn's extension around: highlight the sentence where you found it, put it straight onto a card front, and let the scheduler handle when that pair comes back.

GREVocabularyFlashcardsExam Prep