TOEFL Build a Sentence: which grammar spots cost points?

What does the Build a Sentence task ask you to do?

In Build a Sentence, you read one sentence the other person says, then move scrambled words and phrases into place to build a response sentence that fits it. Page 13 of ETS's "Updated TOEFL iBT Test Overview" (the Test Overview, no publication date, checked 2026-10-30) says a complete sentence is followed by a sentence whose words are in the wrong order, and that the two sentences form an exchange between speakers. The response must be grammatical and appropriate to the context. In the one sample item on the same page, the first and last words of the response sentence are already in place, and one of the given chunks is left unused. The leftover chunk was a different form of the same verb used in the correct answer.

ETS says the task measures whether you understand English sentence structure (ETS "TOEFL iBT Writing Section", no publication date, checked 2026-10-30). The Specifications Document gives the skill the task measures as "reassembling a range of sentence structures," the context of the language used as everyday communication between people (Social Interpersonal), and the difficulty range as CEFR (Common European Framework of Reference for Languages) A1 to C2.

The figures for the three Writing tasks are below. Item counts and scores come from the Specifications Document and Table 2 on page 30 of ETS's "TOEFL iBT Technical Manual" (ETS Research Report RR-25-12, October 2025, hereafter the Technical Manual), scoring comes from Technical Manual pages 34, 35 and 39, and timing comes from Test Overview pages 4, 14 and 15. All page numbers in this article follow PDF page order (page 1 = cover).

Page 34 of the Technical Manual says a Build a Sentence item earns 1 point if correct and 0 points if wrong. There is no partial credit within an item. Page 39 of the same document says the task is scored by matching against the answer key, using neither human scoring nor AI scoring, and the Specifications Document classifies the same task as machine scored. Email and Discussion use ETS's AI scoring engine together with human scoring, and when the AI scoring has low confidence in a response, a human rater gives the score (Technical Manual pages 35 and 39). Writing as a whole has 12 items, a raw score of 0 to 20 and a standard time of about 23 minutes (Test Overview pages 4 and 21). Of that, 10 points belong to Build a Sentence. In this article, "losing points" means an item scoring 0, so you do not get that 1 point (raw score). Write an Email and Write for an Academic Discussion are covered in How do I prepare for the three new TOEFL Writing tasks?.

Where does word order in questions go wrong?

Word order in questions turns on the order of the auxiliary and the subject. In a direct question, the auxiliary moves in front of the subject; in an indirect question (a question placed inside another sentence), the subject comes first, as in a statement. The sets below are examples the Unlock Prep (unlockprep.com) editorial team wrote for illustration. They follow the format the Test Overview describes (one sentence from the other speaker + a response sentence, words already in place, one leftover chunk), and they do not copy any ETS item. Blanks are shown as ___.

Set 1: Indirect question

• Partner says: Can I help you find something?
• Response: Could you ___ ___ ___ ___?
• Chunks: tell me / where / is / the bus stop / does
• Wrong order: Could you tell me where is the bus stop?
• Right order: Could you tell me where the bus stop is? (Leftover chunk: does)
• Why: What follows where is an indirect question serving as the object of tell me. In an indirect question, the verb (is) comes after the subject (the bus stop). The question mark at the end is there because the opening Could you makes the sentence a question.
• Why the wrong order looks plausible: The word order of the direct question "Where is the bus stop?" gets carried over as is.

Set 2: Direct question and does

• Partner says: I'm going to the library this Sunday.
• Response: What time ___ ___ ___ on Sundays?
• Chunks: does / the library / open / opens
• Wrong order: What time does the library opens on Sundays?
• Right order: What time does the library open on Sundays? (Leftover chunk: opens)
• Why: In a direct question, the question phrase (what time) is followed by the auxiliary (does), and then the subject (the library) and the base verb (open). When does is in the sentence, the verb after it is a base verb.
• Why the wrong order looks plausible: Because the subject the library is third person singular, you reach for opens with -s. If you put the same content after Do you know, does drops out and the sentence becomes "Do you know what time the library opens on Sundays?"

Where do the auxiliaries go when there are two or more?

In a question, only the first auxiliary moves in front of the subject. The other auxiliaries and the main verb stay after the subject. The set below is another example we wrote for illustration in the same format.

Set 3: Two auxiliaries (should + have)

• Partner says: The deadline for the form was yesterday.
• Response: ___ ___ ___ ___ the form earlier?
• Chunks: should / I / have / sent / send
• Wrong order: Should have I sent the form earlier?
• Right order: Should I have sent the form earlier? (Leftover chunk: send)
• Why: Only the first auxiliary, should, moves in front of the subject (I). The second auxiliary, have, and the past participle sent stay after the subject.
• Why the wrong order looks plausible: If you have memorized should have as one chunk, you end up moving both words to the front together.

The negative word not follows the same position rule. In "I have not received the email yet.", for example, not sits right after the first auxiliary.

What has ETS not disclosed?

What we confirmed in ETS documents covers the task's format and description, the number of items, the score of 1 or 0 per item, the scoring method and the Writing raw score range. Page 23 of the Test Overview says a scoring guide is used for each of the two Writing tasks in which you write your own response, Write an Email and Write for an Academic Discussion, and it does not include a scoring guide for Build a Sentence. ETS has explained how raw scores are converted to bands but has not published the conversion table. Page 35 of the Technical Manual says the Writing raw score total is converted to a scale score through weighted equipercentile linking (a statistical procedure that adjusts for differences in difficulty between test forms). What we could not find in the Test Overview, the Specifications Document, the Technical Manual or ETS information pages is listed below.

• Whether every item has words already in place and a leftover chunk (the Test Overview shows only one sample item)
• Time set aside for Build a Sentence on its own (ETS states about 23 minutes for Writing, 7 minutes for Email and 10 minutes for Discussion, but gives no separate time for Build a Sentence)
• A conversion table showing which score on the 1 to 6 band each Writing raw score becomes

We did not find any place where ETS documents disagree on Build a Sentence figures. We did not compare articles from test prep companies or blogs. The Specifications Document still contains wording that minor changes may be made before launch. How to read band scores is covered in How do TOEFL 1 to 6 scores compare with 0 to 120 scores?.

What should you check when you practice?

In our view, if you decide the sentence type and the auxiliary position before placing any chunks, you can avoid the wrong turns shown in the three sets above.

1. Look at the words already in place and the question mark at the end. If a phrase like Could you or Do you know comes first and a question word like where, what or if follows it, what comes after is an indirect question, so the subject comes first.
2. Find the auxiliaries (do, does, did, can, should, have). In a direct question, put only the first auxiliary in front of the subject.
3. If do, does or did is in the sentence, check that the verb after it is a base verb.
4. If the leftover chunk is a different form of the same verb, check again that the form you chose fits the subject and the auxiliary.