Stop Trying to 'Speak Up More' in English Meetings. Your Real Problem Is Listening.

Search for advice on surviving meetings in English when English isn't your first language and you get the same list every time. Speak up in the first five minutes. Prepare your key phrases the night before. Learn the thirty idioms native speakers use. Practise your point out loud. Work on your accent. Be confident. Every single tip is about output — what comes out of your mouth. Which is strange, because if you ask people who actually sit in these meetings what goes wrong, the answer is almost never "I couldn't say my sentence." It's "I lost the thread around minute twenty, and by the time I found it again, they'd decided something."

This isn't a niche problem. There are roughly 1.5 billion non-native English speakers in the world against about 370 million native ones, and in a 2024 survey more than 62% of employees said they work with colleagues whose native language differs from their own. The English-language meeting is the default operating mode of global work, and in most of those meetings, most of the people in the room are working in their second (or third) language. If the standard advice were right, this would be a speaking problem at planetary scale. It isn't. It's a listening problem, and the research on that is uncomfortably clear.

A meeting isn't a speaking test. It's a listening marathon.

In 2018, researchers at University College London put native and non-native English listeners in front of sentences played in varying amounts of background noise and measured their pupils while they listened. Pupil dilation is one of the more honest measures of cognitive effort — you can't fake it, and you can't suppress it. The result: non-native listeners showed significantly greater pupil dilation than native listeners even when their comprehension scores were the same. They were getting to the same answer, but paying far more to get there. In harder conditions, the native listeners could throw more effort at the problem and keep up; the non-native listeners appeared to hit an effort ceiling and comprehension dropped.

Now stack a real meeting on top of that. Add video compression that chops consonants. Add four speakers with four different accents, one of whom has a bad microphone. Add crosstalk. Add the fact that you are simultaneously supposed to be evaluating what's being said, forming an opinion, deciding whether to voice it, and writing something down. Listening effort research consistently links high effort to faster fatigue and a reduced ability to do a second task at the same time — and a meeting is nothing but second tasks. This is why the ninety-minute Thursday planning call leaves you wrung out in a way your native-speaking colleagues don't seem to understand. You didn't do more work. You did the same work on hard mode.

Here's the part the "speak up early" advice misses entirely. When you decide to contribute, you start assembling the sentence. Vocabulary, grammar, the tone that makes it sound like a suggestion rather than an objection. In your first language this happens in the background. In your second language it takes the foreground, and while it's there, you stop listening. Call it the composition tax. The thing you were preparing to say is usually fine. The thing you missed while preparing it is the decision at minute 38 that changes your team's quarter.

Why 'sound more native' is chasing the wrong prize

The bias is real; let's not pretend otherwise. In a well-known 2010 experiment, Shiri Lev-Ari and Boaz Keysar had native English speakers rate trivia statements ("Ants don't sleep") read aloud by native and non-native speakers. The same sentences were judged less true when spoken with a foreign accent. The crucial detail is why: the effect was driven by the listener's processing difficulty, not by any judgement about the speaker's intelligence. A 2026 study in the Journal of Language and Social Psychology found a related paradox — professionals with lower English fluency were rated as more likeable but less competent.

So yes, an accent costs you something. But accent reduction is a multi-year project with modest returns, and it attacks the least fixable part of the problem. Look at what the competence judgement is actually made of in a working team: did this person understand what was decided, did they act on it correctly, did they remember the constraint the client mentioned in passing. Those are all listening outcomes. Being the person in the room who can say "actually, Priya said Thursday, not Friday, and she said it depended on legal signing off" is a competence signal you can start sending next Tuesday. No phonetics coach required.

What actually helps (and it isn't a phrasebook)

1. Move as much as possible from listening to reading. For most second-language users, reading comprehension runs well ahead of listening comprehension — text doesn't have accents, doesn't go by at 160 words per minute, and can be re-read. Ask for the agenda in writing, ask for slides before the call, and ask for decisions to be typed into the chat as they're made. Every sentence you can read instead of decode by ear is bandwidth you get back.
2. Stop doing capture and comprehension at the same time. Writing notes while listening in a second language is the worst possible combination: two working-memory-heavy tasks fighting for the same resource. We've written about why note-taking during meetings backfires for everyone; for non-native listeners the cost is doubled. Get a verbatim record from somewhere else and spend the meeting purely understanding.
3. Ask for clarification of decisions, not words. "Sorry, could you repeat that?" is fine, but it asks a native speaker to repeat the sound. "So we're saying the launch moves to the 14th and marketing owns the announcement — is that right?" asks them to confirm the meaning, which is what you actually needed, and it reads as leadership rather than struggle.
4. Cut the acoustic tax at the source. Headphones, not laptop speakers. Cameras on, because lip movement measurably improves speech comprehension in noise. Politely ask the person on the airport Wi-Fi to type their point. These sound trivial; the pupillometry says they're the difference between coping and hitting the ceiling.
5. Review within the hour. Whatever you did retain is on a steep forgetting curve — most of it is gone by tomorrow, and for effortful second-language listening the encoding was shallower to begin with. Ten minutes with a transcript right after the call beats an hour of reconstruction the next morning.

Notice that none of these make you sound more native. All of them make you understand more per unit of effort, which is the actual scarce resource. And notice that the second one — get the verbatim record from somewhere other than your own overloaded head — is doing most of the work. Which raises the obvious question.

The dignity problem with 'just use an AI notetaker'

The standard AI-notetaker fix comes with a cost that native speakers never notice. When you add a meeting bot, a line appears in everyone's video window: "Sofia's Notetaker has joined the meeting." In a team where you're already fighting a competence-versus-likeability paradox, that line reads as "Sofia needs help." Some people are fine with that. Many are not, especially in cultures where asking for accommodation in front of colleagues carries a real social price. And that's before the practical problems: the recording goes to someone's cloud, your colleagues in Germany or France quite reasonably want to know whose, and consent for a bot in a client meeting is a negotiation in itself.

The quieter version is a phone on the table, or a phone next to the laptop, doing transcription locally. It works for in-person meetings where bots can't go at all. It doesn't announce itself. And because the model runs on the device, the transcript never leaves it — which turns the privacy conversation from "where does this go?" into "nowhere." One honest caveat: speech recognition has its own accent problem, and heavily accented English in a noisy room will produce imperfect text from any tool. The point isn't a flawless transcript. The point is that a 95% transcript you can read at your own pace beats a 70% memory you formed under load.

A different script for your next English meeting

Before: get the agenda and any slides in writing, and read them — you're front-loading comprehension into the channel where you're strongest. Decide the one thing you want to say, so the composition tax is paid in advance rather than mid-meeting. During: start the recording, close the notes app, and give the meeting your full listening budget. When you speak, confirm meaning, not words. When you lose the thread — you will, everyone does — let it go instead of chasing it; the transcript has it. After: within the hour, read the summary, scan the transcript for the decisions and who owns them, and send the two-line recap to the group. Do that three times and watch what happens to the way the room treats you. You didn't get more fluent. You became the person who always knows what was decided.

The advice industry sells non-native speakers a speaking problem because speaking problems have products attached: courses, coaches, accent apps. The listening problem has a much cheaper fix, and it's the one that actually moves the competence needle. Protect your ears. Read what you couldn't hear. Stop performing fluency and start being the most reliable memory in the room.

English meetings as a non-native speaker: common questions