TOEFL Vocabulary: What the 2026 Test Actually Asks

The January 2026 redesign changed how TOEFL tests vocabulary rather than which words it favours. One reading task now makes you type missing letters instead of picking from options, and it is scored on spelling as well as meaning. Recognising a word is no longer enough to earn the point.

What actually changed in January 2026

The updated test took effect on 21 January 2026. The Reading section was rebuilt around three task types: Complete the Words, Read in Daily Life, and Read an Academic Passage. The section is adaptive, shorter than it used to be, and reported on a 1 to 6 band scale rather than the old points.

For vocabulary specifically, the important change is the first of those tasks. Everything else on the test still checks whether you understand words. Complete the Words checks whether you can produce them.

Complete the Words is a production task

Here is the shape of it. You get a short academic paragraph with exactly ten incomplete words in it. Each one shows its opening letters and hides the rest, and you type what is missing. A test contains several of these tasks.

Machine scoring makes the standard unusually strict. The word has to be right in meaning, right in grammatical form, and right in spelling. Get the sense but write the wrong ending and you lose it. Get the word and misspell it and you lose it.

That is a real break from how most people study for this test. Multiple choice rewards recognition, and recognition is the cheap half of knowing a word. Typing rewards production, which is the expensive half, and the two do not come as a set. Anyone who has read comfortably in a language for years and then tried to write in it knows the gap.

There is a sensible order for attacking one of these blanks, and it is not left to right. Read the whole paragraph first, because a blank near the start is often only decidable from a sentence further down. Then work out what job the missing word does: is it carrying the content of the clause, or joining two clauses together? Those are different searches. Only then use the visible letters, which are a constraint rather than a hint, and finally check the ending against the grammar of the sentence around it. The candidates that survive all four filters are usually down to one.

The habit worth building is treating the stub as arithmetic rather than a memory prompt. Four visible letters plus a known grammatical slot plus a known logical relation rarely leaves room for two answers. Guessing wildly at a word that starts with those letters is the slow way round and it tends to produce something that fits the spelling and not the sentence.

The surprise is what those blanks are made of

This is the part I did not expect, and it is the single most useful thing to know about the task.

Reporting on the new format notes that a large share of the blanks are function and linking words, the ones that depend on grammar rather than subject knowledge: however, because, although. Not the specialist vocabulary that TOEFL word lists are built from.

Think about what that does to the standard preparation. Memorise three hundred academic nouns and you have prepared for a fraction of the blanks. The rest are decided by whether you can read the logic of a sentence and tell that this clause contradicts the last one rather than following from it. That is grammar and discourse sense wearing a vocabulary costume.

It also explains why fluent readers sometimes lose points here for reasons that feel unfair. The academic content is comfortable, the connectives are automatic in comprehension, and then the task asks you to produce one cold, from four letters, with the spelling counted.

The practical response is to stop treating connectives as a vague category and sort them by the relation they signal, because that relation is what the paragraph gives you. Roughly four jobs cover most of them. Contrast reverses the direction of the argument: however, whereas, conversely. Concession admits a point before overriding it: although, despite, nevertheless. Cause and consequence run the logic forward: because, therefore, consequently, hence. Addition extends in the same direction: moreover, furthermore, likewise.

Sorted that way, the task becomes tractable. You read the two clauses, decide which of the four relations holds between them, and that narrows twenty candidate words to about three before the visible letters have done any work at all. Learning fifteen connectives by relation is a smaller job than learning three hundred nouns, and on this task it pays considerably better.

How many words you actually need

Figures circulate widely, and they are worth knowing as a rough map rather than a target.

Target score	Active vocabulary commonly cited
80 and above	roughly 5,000–6,000 words
100 and above	roughly 8,000–10,000 words
A working academic core	200–400 high-frequency words

Read the first two rows as a description of a person, not a to-do list. Nobody assembles ten thousand words deliberately. Most of that total arrives from years of reading and listening, and the deliberate part of your study sits in the third row. Two hundred words you genuinely own will do more for you than five hundred you half recognise.

There is also a cheap way to check where you actually sit, and it is not a vocabulary size test. Open an academic article in a field you do not study and read a page. Count the words you could not produce if the page were taken away but you had to write about the same topic. That number is the gap between your reading vocabulary and your usable one, and it is the gap the new task charges you for. If the page is comfortable to read and the count is still high, you are exactly the candidate this redesign was built to catch.

The Academic Word List is still the right starting point

ETS publishes no official vocabulary list, which is exactly why every prep site can sell you a different one and none of them can be wrong.

The Academic Word List is the closest thing to a defensible substitute. It collects 570 word families that between them account for roughly ten percent of the words in academic texts, which is a large return for a finite list. It was not built for the TOEFL, and that is a point in its favour: it describes the register the test borrows from rather than guessing at the test’s own contents. We covered how to work through it in the AWL guide.

The word family part matters more than usual now. The AWL is organised by family precisely because analyse, analysis, analytical and analytically are one item of knowledge with four surfaces, and a task that scores grammatical form is a task that asks which surface you need.

Read in Daily Life changed the register

The second new task is easy to overlook because it looks trivial. Short everyday texts, emails and notices and the like, rather than academic prose.

The consequence for vocabulary is real though. A candidate who has prepared exclusively with academic word lists has prepared for one register and will meet another. The words in these texts are not hard, but they are a different set, and being fluent in the language of lecture notes does not automatically make you fluent in the language of a landlord’s email.

The specific vocabulary of everyday documents is oddly resistant to list-based study, because it is bound to situations rather than topics. Deposit, refund, deadline extension, itemised, in advance, subject to. None of that is advanced English, and most of it never appears in an academic passage, so a strong reader can hit it cold. Worth noting too that these tasks tend to arrive early, and on an adaptive section an early stumble shapes the difficulty of everything after it. That is an expensive place to meet an unfamiliar register.

What this changes about how you prepare

• Practise typing, not just recognising. If your entire method is flashcards you tap through, you are rehearsing the half of the skill the test stopped rewarding on its own.
• Study connectives deliberately. However, although, because, whereas, nevertheless. Learn what each one signals about the relationship between two clauses, because that is what tells you which one belongs in the blank.
• Check your spelling of words you already know. The words that cost points here are usually not exotic. They are ordinary words you have only ever read.
• Learn words as families. One item, four forms, and the test decides which form it wants.
• Read outside the academic register too. An email, a notice, a bill. Fifteen minutes of it is enough to stop the second task feeling alien.

Broader exam-vocabulary strategy, most of which is unchanged by the redesign, sits in how to study vocabulary for an exam.

The part a word list cannot fix

Two honest limits.

First, no list survives contact with a task that samples function words. You cannot memorise your way to knowing that whereas is the right connective in a particular sentence; you have to read the sentence.

Second, the score bands above describe a correlation, not a mechanism. Ten thousand words does not buy a 100. It is a description of the sort of person who scores 100, and the causal work is done by everything that produced the vocabulary in the first place, mostly a lot of reading and listening over a long time. Treat the numbers as a sanity check on your level, and be suspicious of anyone selling the list that closes the gap. If you want to think about the totals properly, how many words you need to be fluent goes into where those figures come from.

How Vocaby fits

The redesign rewards two things a review app can genuinely help with: knowing a word well enough to produce it, and knowing what it sounds like.

Vocaby uses FSRS spaced repetition, so each word comes back near the point where you would otherwise forget it, and every review is a retrieval attempt rather than a re-read. That is the same difficulty the new task exposes, met earlier and in private. Every entry carries IPA and real audio, which matters for the Listening section for the same reason spelling matters for Reading: a word stored only as a shape is a word you may not catch as a sound. With more than 29,000 words, the academic range is covered without a separate list. You can browse it in the word library.

The test moved from asking whether you have met a word to asking whether you can write it down. Study accordingly.

Download Vocaby on the App Store to learn words with audio, IPA, examples, and spaced repetition on every entry.