TOEFL Complete the Words Practice

What Complete the Words actually tests

You get an academic paragraph with the back half of certain words removed and a stub of letters left in place, and you type the rest. It looks like a spelling test and it is not. The letters you are given rule out most of the vocabulary you know, and the sentence around the blank rules out most of what survives that, so the real skill is holding the paragraph's meaning in your head while working at the level of single words.

The task sits at CEFR B1 to C1 and measures three things at once: word ordering rules, lexical knowledge and reading comprehension. A confident speaker can still lose most of these items, because speaking fluently never requires you to produce the exact suffix, and this task does.

Where students lose marks on this task

The most common mistake is working blank to blank without reading the paragraph first. The topic sentence does most of the disambiguation for you, so anyone who starts typing at the first gap is guessing with a fraction of the information available. Read the whole paragraph at normal speed, fill the blanks with strong collocational pull first, then go back for the rest.

After that it is suffix families and agreement. The -ence and -ance pair, -tion against -sion, -able against -ible, and the choice between a noun and its adjective when the slot only takes one. Then plurals and tense, because the sentence often forces a form the isolated word does not suggest. Last is the blank left empty. There is no penalty for a wrong guess, and a guess assembled from the letters you were actually given is right far more often than students expect.

How the 20 levels drill it

Level 1 gives you four blanks in a paragraph of 55 to 70 words. By level 15 it is seven blanks in 90 to 110 words, at the length and density the real section uses. Levels 16 to 20 push to nine blanks in 110 to 130 words with vocabulary above the exam's ceiling, so real exam paragraphs read as short and plain by comparison.

Each paragraph is one of the level's five questions and it only counts as correct if every blank in it is right, which is the same all-or-nothing rule the real task uses. The grader does accept British spellings, so metre, colour and analyse are all marked correct even though the site writes American English. After you check, each blank shows the full word next to what you typed, with an explanation of the pattern behind it.

Count the letters before you guess the word

A blank on this task is not an empty space, and the number of missing letters is free information almost nobody uses. Every gap in these paragraphs draws one dash per letter that was cut, and the box will not accept more than those letters plus two, so the shape of the answer is on screen before you have thought about meaning at all. Used properly, that turns a recall problem into an elimination problem.

Take opti followed by four dashes, from the level 8 paragraph on eye evolution. Optimal and optimum need three letters, so the count rules both out before you read the sentence, which leaves optimize and optimism, and the does not in front of the blank forces the verb. Two checks, no vocabulary search. The same move settles sile with three dashes, where silent needs two and silently needs four, so silence is the only word the slot can hold. It settles melt with five, which can be neither melted nor melting and turns out to be meltwater. It separates compensated from compensation, because the noun needs six letters and the blank offers five. Across the hundred paragraphs here the missing tails run from two letters up to ten, and most are three, four or five, which is exactly the range where a wrong guess feels plausible.

The habit worth building is to read the stub and the count as one unit. Say compen, five to yourself before you look at the rest of the sentence, and the sentence is then choosing between two or three survivors rather than everything you know. The count also catches the error the box was sized against: if you are running out of room, you are typing the whole word instead of the half that was removed. Once that is automatic, take it into a full Reading section, where these paragraphs share one clock with everything else.

The blanks where no ending rule can save you

Close to a quarter of the blanks in this bank are not suffixes at all. Nectar, retina, seabed, canopy, threshold, masonry, dashboard, limestone, sugarcane and windward carry no ending you can derive, so the suffix families that decide most of this task are silent on them. Grammar tells you a noun belongs there and then stops. These are the blanks that quietly decide paragraphs, because a student who has drilled endings and nothing else has no procedure for them and starts inventing letters.

There is a procedure. Read the sentence, name the thing in your own words first, then check your word against the stub rather than the other way round. In the level 8 paragraph on the ocean conveyor, the sentence says that cold salty water sinks and begins a slow return south along the bottom of the sea. You do not have to recognize seab. You have to say bottom of the sea, notice that seabed is the single word English keeps for it, and let the stub confirm you. The same move handles the compounds, and there are plenty: meltwater, overflow, daylight, countryside, cardboard, sugarcane. The second half is an ordinary word and not an ending, so stop hunting for a suffix and ask which plain noun the sentence has already described.

This split is also the most useful thing to look at when you lose a paragraph. A blank lost on an ending is a rules problem, and a week of deliberate drilling moves it. A blank lost on a content word is a reading problem: the sentence named the thing and you were reading letters instead of meaning. The second kind is why subject range matters here. These paragraphs run from honeybee dances to karst caves, jazz harmony, radiocarbon dating and Gothic masonry, and the academic passage track covers the same ground at greater length if that is the half you keep dropping.