Writing · 10 items per section
TOEFL Build a Sentence Practice
Build a Sentence is 10 of the 12 items in TOEFL Writing, and it is pure pattern knowledge: relative clauses, embedded questions, inversion, participle phrases. That makes it the fastest thing on the whole test to improve, and most students spend their writing preparation on the two essays instead. This ladder is 100 items in 20 levels of 5. Level 1 is free.
- 100 questions
- 20 levels of 5
- CEFR A2–C1
- Level 1 free
How a level works
Each level is 5 questions: 5 independent sentences to assemble. You get feedback the moment you answer, and at the end you can review all 5, with the correct answer, why it is correct, why the option you picked was not, and what the question was training. 3 of 5 opens the next level, and any level can be replayed for a better star count.
The difficulty ladder
Foundations
Levels 1–5 · CEFR A2Everyday language and short texts. The point here is to get the mechanics of the task automatic before the content gets hard.
Building
Levels 6–10 · CEFR B1Longer material and less familiar topics. This is where the task starts to cost you attention.
Test Ready
Levels 11–15 · CEFR B2Real exam length, real exam difficulty. Clear this band and the task itself will not surprise you on test day.
Beyond Test
Levels 16–20 · CEFR C2Past what the exam uses, on purpose. Denser text, faster audio, closer distractors. Test day should feel like a step down from here.
Levels 1 to 15 stay inside real 2026 TOEFL parameters for this task. Levels 16 to 20 are deliberately harder than the exam. The first 1 levels are free and need no signup.
What Build a Sentence actually tests
You are given the chunks of a sentence out of order, plus one extra chunk that does not belong in it, and you assemble the sentence. The answer has to be both grammatical and natural, which is a higher bar than grammatical alone: several orders will parse and only one reads like something a native writer would produce.
There is no essay skill involved and no opinion to have. It is syntax, and the stock of structures it draws on is small enough to learn: relative clauses and where they attach, embedded questions and their word order, inversion after negative adverbials, participle phrases, comparative structures, and where adverbs sit.
Where students lose marks on this task
The first mistake is trying to use every chunk. One of them is a distractor and it is chosen to look plausible, so a sentence that only works if you force in the extra piece is the wrong sentence. Build the core clause first and see what is left over.
The second is embedded question word order, which almost every learner gets wrong under time pressure. You ask where the library is, not where is the library, once the question sits inside another clause. The third is producing something grammatical but not idiomatic, usually by putting an adverb in a position English tolerates but does not choose. Read your assembled sentence back to yourself before checking it, because your ear will catch what your grammar knowledge does not.
How the 20 levels drill it
Level 1 gives you five chunks and one distractor. Levels 11 to 15 give you seven chunks and one distractor, at exam difficulty. Levels 16 to 20 give you eight chunks and two distractors, which is harder than anything the test will show you.
Where a sentence genuinely has more than one natural order, the alternative is credited rather than marked wrong. After you check, the explanation names the pattern the item was built on, which is the part worth remembering: the same dozen or so patterns come back all the way up the ladder and all the way through the test.
What the given opening already decides for you
Every item hands you more than a bag of chunks. It also hands you a fixed opening, the punctuation that must end the sentence, and a line from the other speaker that you are answering. All 100 items here are built that way, and so is the task inside the 23 minute Writing section. The opening is not decoration. It is a grammatical commitment somebody has already made on your behalf, and it kills most of the orders you would otherwise compare.
Watch what a stem does. Start an item with "No sooner" and an auxiliary has to come before the subject, so every order beginning with a plain subject is gone before you weigh it. "What surprised me" can only be followed by "was", because the opening is the front half of a cleft and nothing else joins it to the rest. "There is" pulls you toward an existential quantifier phrase, and once "hardly" appears in the pool, "any" is the only quantifier that fits. "If the shuttle had run on time," commits the result clause to "would have", settling a choice between two near identical chunks before you think about conditionals.
The other speaker's line constrains the meaning the same way. When the question asks which of two campus cafes is better, your reply compares two things, so a superlative would be wrong even if the chunks allowed one. Closing punctuation counts too. One item in this bank ends in a question mark, and that one character tells you the finished sentence is a request, not a statement.
So read the stem, the punctuation and the speaker's line before you touch a chunk, decide what word class has to come next, and place that chunk first. Most items collapse from dozens of orders to two or three at that moment, and what follows is checking rather than searching.
The leftover chunk names the rule being tested
One chunk stays unused, and the useful question is what it is a version of. Across all 100 items here, the extra chunk is almost never an unrelated word. It is a rival form of a chunk that does belong, put there to find out whether you know which of the two the structure needs. Spot the pair and the scramble becomes a two way choice, which is a much smaller problem.
"checking" sits beside "checked", and the point is what has to follow "should have". "collecting" sits beside "collected", and the point is whether a reduced clause describing the samples is active or passive. "make" sits beside "makes", and the point is that the verb agrees with "neither" rather than the plural noun in between. "if" sits beside "whether", and the point is that only "whether" can sit in front of a to-infinitive. The same design runs the whole length of the ladder: "more cheap" against "cheaper" at level 1, "the more easily" against "the easier" in the middle, "turn out" against "turns out" near the top.
This changes what is worth writing down when you miss one. Do not copy out the sentence. Record the pair and the one reason the loser lost, because that reason is the entire item and it will come back in a different sentence. Levels 16 to 20 hand you two extra chunks instead of one, so there are two pairs to settle. Sometimes both belong to one contrast, as when "starts" and "started" compete with "starting", and sometimes they test two separate points.
A list of pairs you have lost stops growing after a few weeks. That is the clearest sign this task is finite in a way the two essays are not, and once the list is short, put the skill under a section clock in the writing practice tests.