Here is a result that should change how you practice TOEFL Listen and Repeat. In the sentence "Carry-on liquids must be placed in clear plastic bags," I looked at how often 184 test takers dropped each word when they repeated it. They dropped "plastic" 15% of the time. They dropped "clear" 58% of the time. Same sentence. Same position. Both are adjectives in front of the same noun. One gets dropped almost four times as often as the other.
That gap is not random, and it is not about memory capacity or sentence length. It comes down to one thing: whether the word belongs to a phrase your brain already stores as a single unit. This post explains the mechanism, shows the pattern across the test, and gives you the fix.
Quick answer
In Listen and Repeat, test takers drop the words that do not fit a predictable phrase. "Plastic bags" is a collocation, a word pair English speakers store and recall as one chunk, so it survives. "Clear plastic bags" is not a stored chunk, so "clear" falls out. The deeper cause is that most people repeat the meaning of a sentence, not its sound. When you rebuild a sentence from meaning, you keep the predictable frame and lose the specific, low-predictability words, which are exactly the words Repeat Accuracy scores. The fix is to hold the sound, not the meaning.
The proof: two words, same spot, very different fate
"Carry-on liquids must be placed in clear plastic bags" is one of seven sentences in a standardized airport set, each repeated by 184 test takers. Here is what happened to the content words.
| Word | Dropped by | Part of a stored phrase? |
|---|---|---|
| clear | 58% | No. "Clear plastic" is not a fixed phrase, so the word stands alone. |
| liquids | 40% | No. A specific, low-frequency noun that has to be held on its own. |
| placed | 35% | No. Predictable in meaning, so it gets reconstructed loosely or skipped. |
| bags | 29% | Partly. The head noun of the collocation "plastic bags." |
| plastic | 15% | Yes. "Plastic bags" is a strong collocation and comes back as one unit. |
Look at "plastic" against "clear." They sit in the same place, modifying the same noun. The only real difference is that "plastic bags" is a phrase English speakers hear constantly, so it is stored and recalled as a single chunk. "Clear" is the word that breaks the expected phrase, so it is the word that drops.
Why it happens: you repeat meaning, not sound
Here is what is actually going on in your head during Listen and Repeat. You hear the sentence, you understand it, and then you rebuild it from that understanding. This is meaning-based repetition, and it is the default for almost everyone. The problem is that when you rebuild from meaning, your brain fills the predictable slots automatically and drops the words it could not predict.
Collocations are the clearest case. "Boarding passes." "Plastic bags." "Mobile app." Your brain treats each of these as one chunk, so they come back whole. But a word that does not fit the chunk, an unexpected modifier or a specific detail, has to be stored separately. Under time pressure, that separate word is the first thing to go. That creates a penalty, because Repeat Accuracy scores the exact words, including the one you dropped.
The same pattern shows up across the test
Once you see it, it is everywhere in the data. The words that drop most are the low-predictability content words, the ones you cannot guess from the rest of the sentence.
- In "You will find restrooms near every main gate," "restrooms" dropped 51% but "gate" only 15%. A gate is predictable in an airport. The specific destination is not.
- In "If your flight is delayed, you may receive updates through the airline's mobile app," the specifier chain "through the airline's" dropped between 61% and 74%, while the predictable anchors "flight" and "app" survived.
This is the information-load idea at the level of a single word. The words that carry the most new, unpredictable information cost the most to hold, so they fall first.
The fix: hold the sound, not the meaning
If meaning-based repetition is the problem, the fix is to switch to sound-based repetition. Hold the surface form of the sentence, the actual string of sounds, instead of understanding it and rephrasing. Here is how to train that.
- Shadow the exact words. Repeat sentences immediately after you hear them, matching the sounds, before you have time to translate the meaning. You are training your ear to hold form, not gist.
- Chunk into two or three sound groups. "Carry-on liquids" / "must be placed" / "in clear plastic bags." Holding three sound groups is easier than holding nine separate words, and it keeps the off-collocation word inside a group.
- Protect the word that breaks the phrase. When a sentence has a word that does not fit the expected collocation, like "clear" before "plastic bags," mark it in your mind and say it deliberately. That is the word you are most likely to lose.
- Drill the dense sentences on purpose. The longer, information-heavy lines are where low-predictability words pile up. Train past the length you expect so a normal sentence feels easy.
- Check your transcript word by word. After each attempt, compare what you said to the prompt. The missing word is almost always a specific, low-predictability content word. Seeing it is how you stop dropping it.
See where your words go
If you want to know which words you drop, and whether you are repeating sound or meaning, take a free Listen and Repeat test on My Speaking Score and read the data. You will see your Repeat Accuracy and the exact words where your version drifted from the prompt. That gives you a much clearer path forward than guessing which part of your speaking to fix next.
Frequently asked questions
What is a collocation, and why does it matter for Listen and Repeat?
A collocation is a word pair or phrase that occurs together so often that your brain stores it as one unit, like "plastic bags" or "boarding passes." In Listen and Repeat, collocations come back whole because they are recalled as a single chunk. The words that do not belong to a collocation have to be held separately, which is why they drop more often.
Why do I drop "clear" but not "plastic" in the same sentence?
Because "plastic bags" is a stored phrase and "clear plastic bags" is not. When you rebuild the sentence from meaning, the collocation "plastic bags" returns automatically, but "clear" is an extra word that breaks the expected phrase, so it is the one that gets lost. In the data, "clear" dropped 58% of the time and "plastic" only 15%.
Is Listen and Repeat a memory test?
Not in the way most people think. It is an encoding test. It rewards people who hold the sound of the sentence and penalizes people who process the meaning and rephrase. Two people with the same memory span can score very differently depending on how they encode the sentence.
What does it mean to repeat the sound instead of the meaning?
It means holding the actual string of words you heard, rather than understanding the sentence and reconstructing it in your own words. Sound-based repetition preserves the specific, low-predictability words that Repeat Accuracy scores. Meaning-based repetition tends to lose them.
Which words am I most likely to drop?
The specific, low-predictability content words: the exact noun, the unexpected modifier, the prepositional detail. In the airport set these were words like "clear," "restrooms," "updates," and the phrase "through the airline's." Predictable anchor words and strong collocations tend to survive.
How do I practice holding the sound?
Shadow short sentences immediately after hearing them, before translating, and chunk each sentence into two or three sound groups. Mark any word that breaks an expected phrase and say it deliberately. Then check your transcript against the prompt word by word to catch the words you lose.
Final takeaway
Listen and Repeat is not testing whether you understood the sentence. It is testing whether you can hold it. The clearest proof is one sentence: test takers keep "plastic bags" but drop "clear," because one is a stored phrase and the other is not. Stop rebuilding sentences from meaning. Hold the sound, chunk it, and protect the word that breaks the phrase. That is how you keep the words the score is actually counting.