The GRE Verbal section is mostly vocabulary in disguise. Text Completion and Sentence Equivalence both reduce, in practice, to “do you know what this word means in context.” Reading Comprehension passages come weighted with the same academic vocabulary, and the harder questions hinge on noticing whether a word is being used in its primary sense or a secondary one.
Magoosh’s 1000-word list is the canonical study deck and it works. But a thousand words is not where you start. A thousand words is where you finish. The problem with starting on a thousand-word list is that the first 200 cards consume all your attention before the algorithm has had a chance to build any spaced-repetition leverage. By card 700, you’ve forgotten cards 1-100.
This deck is the 50-word starter that gives the larger list something to anchor onto.
How the 50 were chosen
Etymological roots and cluster density. If you learn truculent, belligerent, and pugnacious together, you’re learning one concept (warlike). If you learn truculent on Monday from one list and belligerent on Friday from another, you’re learning two unrelated cards.
The 50 cards in this deck represent 50 high-frequency GRE words that anchor 12 conceptual clusters. Knowing the 50 makes the next 200 cheaper because each new word slots into a cluster you already have.
The 12 clusters
Praise (5). laudatory, encomiastic, panegyric, eulogistic, lauded.
Censure (5). vituperative, excoriating, castigating, scathing, vitriolic.
Hostile (4). truculent, belligerent, pugnacious, bellicose.
Reserved (4). taciturn, laconic, reticent, reserved.
Effusive (4). voluble, garrulous, loquacious, prolix.
Skeptical (4). dubious, incredulous, skeptical, sanguine.
Genuine vs. fake (4). ingenuous, disingenuous, candid, sincere.
Stubborn (4). obdurate, intransigent, recalcitrant, refractory.
Ephemeral (4). ephemeral, transient, evanescent, fugitive.
Deferential (4). obsequious, sycophantic, fawning, unctuous.
Dispassionate (4). dispassionate, equanimous, phlegmatic, sangfroid.
Vehemence (4). vehement, fervent, ardent, zealous.
That’s 50 words, 12 clusters, each cluster averaging 4 words. The cluster size is deliberate — small enough that you can hold the cluster in working memory, large enough that one-of-four spaced-repetition sampling becomes useful.
What the cards look like — cloze with etymology hints
Every card in the deck is a cloze deletion, not Q&A. The reason: GRE vocabulary lives or dies on subtle differences inside a cluster, and Q&A format can only test “what does this word mean” — it can’t test “why is truculent the right word here when belligerent would also fit.”
The cloze format with the ::hint syntax solves this. Each card embeds the word in a context sentence and pairs it with an etymology hint:
Someone described as "{{c1::truculent::Latin trux, fierce — note the
'truc-' root}}" is aggressively assertive and verbally combative,
typically in a way that invites further conflict.
The hint shows only on the card front. Student tries to produce truculent; if they got “belligerent,” the hint’s etymology pointer (“trux, fierce”) nudges them to truc- without giving the answer. SuperMemo’s writing on hints calls this “guided retrieval” — the hint is supposed to disambiguate without trivializing.
For tight clusters where the words are near-synonyms, the overlapping cloze pattern fits perfectly:
The "praise" cluster: {{c1::laudatory}}, {{c2::encomiastic}},
{{c3::panegyric}}, and {{c4::eulogistic}}. The "censure" cluster
mirrors them: {{c5::vituperative}}, {{c6::excoriating}},
{{c7::castigating}}, and {{c8::scathing}}.
One note generates 8 cards. Each card hides one word and uses the other 7 as context. The praise/censure pairing is itself reinforced — every review of the praise cluster shows the censure words as ambient context, and vice versa, so the dichotomy gets practiced 8× per cycle.
For etymology-heavy words, the hint carries the root:
The opposite of "{{c1::voluble::Latin volvere, to roll — flowing speech}}"
(talkative, fluent) is "{{c2::taciturn::Latin tacere, to be silent}}"
(reserved, sparing of words).
The roots become hooks. Six months after the test, a student might forget the exact answer card but remember “starts with vol-, has to do with rolling speech” — which is enough to pick the right answer on a Sentence Equivalence question.
The math behind the 50
Spaced repetition’s effect compounds with the number of items, but only up to a point. Cognitive science research suggests cluster sizes of 3-7 items are optimal for retention; above 7, items in the same cluster start interfering with each other. This deck stays in that zone deliberately.
Avorio’s FSRS-5 implementation tracks each card’s stability (how long until you’d forget it) and difficulty (how often you struggle on this specific card). For vocabulary, difficulty correlates with semantic distance from words you already know. That means truculent will spend longer at short intervals than vehement (which has a more familiar root). The algorithm handles this without you having to think about it.
How to use this deck
Plan for an 8-week ramp before your test date. Two weeks to learn — every card 4-5 times each. Four weeks of consolidation — the easier clusters (praise, censure) stretch out, the harder ones (deferential, dispassionate) keep coming back. Two weeks before exam day, the algorithm will surface the cards you’re least sure on. Don’t fight that — let the scheduler do its job.
After this 50-card deck is solid, Magoosh’s 1000-word list gives you 950 more words that mostly extend clusters you already know. That’s when a thousand-word list works.
Try the first 10 below
The praise + censure clusters — 10 words that test whether you can detect strong positive vs. strong negative valence. If they’re already easy, you have your foundation; this deck won’t add much. If 3+ feel uncertain, install Avorio and load the 50.