You say your sentence and the native speaker frowns for half a second. They reply in English. Or they praise your fluency straight after a simple greeting. All three reactions have clear mechanisms behind them, and almost none of them is the judgement you assume.
You say a Chinese sentence you have practised many times. The person opposite frowns for about half a second, then replies — in English.
Nearly every learner has had this moment, and nearly all of them read it the same way: I must have sounded terrible.
That reading is almost always wrong. Not because you sounded good, but because most of what is happening in the listener's head is not evaluation — it is processing. And the two produce exactly the same facial expression.
Comprehension is not passive. A listener's brain is continually predicting — the next sound, the next word, where the sentence is going. Prediction is what makes it possible to follow speech at normal speed without exhaustion.
When the speaker is a learner, prediction works much less well. Stress lands in the wrong place, the rhythm is not what was expected, a sound is replaced by a near neighbour. Each failed prediction forces the listener back onto a slower, more expensive route.
Put plainly: a listener's patience is a finite resource and you are spending it. Not because they are unkind, but because listening to you genuinely costs more cognitive energy than listening to a compatriot.
This explains a lot of behaviour that reads as rude. They finish your sentence — that is the prediction system running ahead and completing. They look away — many people cut visual load to free resources for listening. They say sorry? — prediction failed outright.
Here is the most important finding in this area, and it is a startling one.
Shiri Lev-Ari and Boaz Keysar (2010) had people listen to trivia statements — the sort of thing that a giraffe can go without water longer than a camel — read aloud by native and accented speakers. Listeners rated the statements read in an accent as less likely to be true.
The crucial part came next: even when participants were told explicitly that the speaker was reading someone else's words — so the accent could have no bearing on the accuracy of the fact — the effect persisted.
The mechanism is not conscious prejudice. It is processing fluency. The brain treats the feeling of ease as a quiet signal for truth. When speech is hard to process, that difficulty gets misattributed to the content rather than to the delivery.
Dragojevic and Giles (2016) traced the same mechanism behind language attitudes generally: listeners rate hard-to-understand speakers less favourably, and much of the effect runs through fluency rather than bias.
This is bad news and good news at once. Bad: how easy you are to process affects how much people believe you, in situations as real as an interview or a negotiation. Good: the target is not losing your accent — it is being easy to process. Those are different things, and the second is achievable.
Murray Munro and Tracey Derwing separated three things learners usually merge:
Accentedness — how different you sound from a native. Comprehensibility — how much effort the listener feels. And intelligibility — whether they actually understood.
The key finding is that these come apart. A speaker with a strong accent can be almost entirely understood. A speaker with a light accent can still make a listener strain because the rhythm is off.
The consequence is liberating: you do not need to erase your accent. That is not a realistic goal for anyone who started as an adult, and it is not a necessary one. What matters is reducing the listener's processing cost — and Derwing and Munro found that rhythm, stress and phrasing do far more for this than perfecting individual vowels.
This is the part that stings most, so it deserves a fair dissection.
Communication accommodation theory, from Howard Giles, describes how people automatically adjust their speech to move closer to or further from the person they are talking to. Switching language is one such adjustment, and it can come from several quite different motives:
Efficiency. If they believe their English is better than your Chinese, switching is the fastest way to finish the transaction. In a busy café this is pure logistics.
Kindness. Many genuinely think they are sparing you effort.
Their own practice. The native speaker is also a language learner, and you are a free English speaker standing in front of them.
And sometimes, social sorting. In some places language is tightly bound to group membership, and switching is a way — usually unconscious — of marking you as outside it.
Notice that only the last of these has anything to do with how well you spoke, and it is not usually the most common. The first three happen even when you are doing well.
The practical response is to keep speaking the language. Language choice is mutual — if you do not switch, most people drift back. And choose your contexts: the shopkeeper with a queue will switch; the neighbour with time will not.
Learners in East Asia notice this paradox early: you manage a greeting and receive warm praise. You study for three years, become genuinely fluent — and the praise disappears.
The compliment is largely ritual, not assessment. It reflects surprise and goodwill, and it is calibrated against expectation rather than ability. Expectations of foreigners are low, so clearing them is easy.
When the praise stops, that is usually a good sign: you have crossed the line where your speaking is no longer surprising. The silence is the real compliment.
So far this article has leaned on costs. The other half is real too, and also has evidence behind it.
Making the effort to speak someone's language reads as a costly signal. It shows you spent time, accepted the risk of embarrassment, and valued their world enough to step into it. Few social acts communicate that as quickly.
There is a deeper reason. Kinzler, Dupoux and Spelke (2007) showed that infants already prefer people who speak with a familiar accent — and that for young children, accent is a stronger cue to social grouping than appearance is. Voice is one of the oldest in-group markers our species has. When you speak someone's language, imperfectly or not, you are touching that system directly.
That is why the reaction is often warm out of all proportion to the quality of the sentence. People are not responding to your grammar. They are responding to the fact that you tried.
Charles Ferguson named this in 1975: foreigner talk. Native speakers automatically slow down, simplify vocabulary, repeat more and articulate more clearly.
Michael Long (1983) showed that the valuable part is not the simplification itself but the negotiation that happens when understanding breaks down: asking again, rephrasing, confirming. That negotiation is what produces input that is both comprehensible and pitched at the right level — the best kind for learning.
Which means the awkward moments you find embarrassing are the most valuable part of the conversation. A perfectly smooth exchange teaches you less than one that had to stop three times to be clarified.
Re-read the puzzled expression correctly. It is almost always processing load, not judgement. The right response is to repeat more slowly or differently — not to switch to English and blame yourself.
Invest in rhythm before individual sounds. If the goal is reducing the listener's cost, stress and phrasing buy far more than polishing a vowel. This is exactly what shadowing trains.
Choose the situation rather than summoning courage. Whether someone has patience for you depends heavily on circumstances: whether they are rushed, whether it is noisy, whether something has to get done. Picking the moment matters more than being braver.
Let them negotiate with you. When you do not understand, say so. Please say that again more slowly is not a sign of failure; it is precisely the mechanism the research shows produces the most learning.
And do not use compliments as a measure. Early praise measures low expectations; later silence measures competence. If nobody has complimented your Chinese lately, you may well have just improved.
The same article, told in plain words — for younger readers, or for anyone who wants the point quickly.
You say a sentence you have practised many times. The person opposite frowns for about half a second, then replies — in English.
Nearly every learner has had this moment, and nearly all of them read it the same way: I must have sounded terrible.
That reading is almost always wrong. Not because you sounded good, but because most of what is happening in the listener's head is not evaluation — it is processing. And the two produce exactly the same facial expression.
Understanding speech is not passive. A listener's brain is continually predicting — the next sound, the next word, where the sentence is going. Prediction is what makes it possible to follow speech at normal speed without exhaustion.
When the speaker is a learner, prediction works much less well. Stress lands in the wrong place, the rhythm is not what was expected, a sound is replaced by a near neighbour. Each failed prediction forces the listener onto a slower, more expensive route.
Put plainly: a listener's patience is a finite resource, and you are spending it. Not because they are unkind, but because listening to you genuinely costs more energy than listening to a compatriot.
This explains a lot of behaviour that reads as rude. They finish your sentence — that is the prediction system running ahead. They look away — many people cut visual load to free resources for listening. They say sorry? — prediction failed outright.
Here is the most striking finding in this area.
People were played trivia statements — the sort of thing that a giraffe can go without water longer than a camel — read aloud by native and by accented speakers. Listeners rated the statements read in an accent as less likely to be true.
The crucial part came next: even when participants were told explicitly that the speaker was reading someone else's words — so the accent could have no bearing on whether the fact was correct — the effect persisted.
The mechanism is not conscious prejudice. The brain treats the feeling of ease as a quiet signal for truth. When speech is hard to process, that difficulty gets misattributed to the content rather than to the delivery.
This is bad news and good news at once. Bad: how easy you are to understand affects how you are judged, unfairly. Good: it points at exactly what to fix, and it is not vocabulary.
When a conversation with a learner breaks down, something useful happens: asking again, rephrasing, confirming. That negotiation is what produces input that is both understandable and pitched at the right level — the best kind for learning.
Which means the moments you find embarrassing are the most valuable part of the conversation. A perfectly smooth exchange teaches you less than one that had to stop three times to be clarified.
And do not use compliments as a measure. Early praise measures low expectations; later silence measures competence. If nobody has complimented your Chinese lately, you may well have just improved.
These articles summarize well-established research in learning science and linguistics. Key sources and further reading:
The question assumes focus is a muscle you train. The strongest evidence says the training barely carries over, and the...
No time, no talent, too old — those are explanations, not causes. What can be measured is duller: one sentence with a ti...
Not one event but four, running on clocks a billion-fold apart: a wave at 400 milliseconds, a word locked in overnight,...
,_by_Aurelio_Tiratelli.jpg){kind=link}