The proverb is about trades, and extending it to languages requires an argument rather than an assumption. The evidence backs it on hours and thresholds, and contradicts it on the additive effect of bilingualism. The real dividing line turns out to be the distance between the two languages.
Better one trade mastered than nine half-learned — the Vietnamese proverb một nghề cho chín còn hơn chín nghề — is a saying about trades. Extending it to languages requires an argument, not an assumption, because languages have a property trades do not: they transfer to each other.
The honest answer is that it holds in part, and the part it holds in is quite specific. This piece locates the line.
Start where the proverb is hardest to argue with.
You have a certain number of hours a week. Learning two languages at once does not create hours; it halves the ones you have.
Five hours a week all directed at one language is 250 hours a year. Split between two, each gets 125. The guided-learning estimates that accompany the European framework put an early CEFR level at roughly 180–200 hours.
Which means: concentrated, you clear a level a year; split, you clear no level in either.
This is not an argument about focus. It is division.
But division is only half the problem. The other half is that the return curve is not linear — and this is where the proverb has real weight.
A language’s usefulness has a threshold. Below it, you can do almost nothing with it.
Hu and Nation (2000) put comfortable reading at roughly 98% known words on the page. Nation (2006) puts everyday conversation at two to three thousand word families and reading a novel at eight to nine thousand.
Look at that shape. Someone at A2 in two languages can greet, order food, ask directions. They cannot read an article, follow a meeting, or say anything worth saying — in either language.
Someone at B2 in one language can do all of that, in one language.
Twice nothing is still nothing. That is the real reason the proverb holds here — not because specialisation is a virtue, but because the payoff sits above a threshold, and the threshold is high.
If the article stopped there it would be a proverb in a lab coat. There is a body of evidence pointing the other way.
Cenoz (2003) reviewed third-language acquisition research and described what she called the additive effect of bilingualism: people who already know two languages often acquire a third better than monolinguals do, with other factors controlled.
This directly contradicts a crude reading of the proverb. If learning several languages dilutes competence, bilinguals should do worse. They do not.
De Angelis (2007) surveys the field and explains why: someone who has already learned a foreign language brings language-learning experience — they know how to attend to form, they know some things do not translate, they can tolerate ambiguity.
This fits Cummins’s interdependence hypothesis (1979) and the long-term work of Sparks and colleagues (2009), who followed students for a decade and found first-language skills predicting foreign-language outcomes years later.
So: languages are not separate trades. They share a foundation.
If the proverb is partly right and partly wrong, where does the line fall?
The most useful answer in this piece comes from Håkan Ringbom (2007), who spent a career on cross-linguistic similarity.
His central conclusion: positive transfer is far larger between closely related languages, and largest in comprehension. Someone who knows Spanish can read a good deal of Italian immediately, because thousands of words look alike and the structures run parallel.
Ringbom and Jarvis (2009) set this out systematically: what governs transfer is not objective similarity but perceived similarity. When learners see two languages as alike, they transfer — sometimes correctly, sometimes not.
This yields a concrete rule, and it is not the rule the proverb gives:
Two close languages at once bring a lot of help and a lot of interference. Two distant languages at once bring neither.
The interference half deserves detail, because it is where learners get ambushed.
Rothman (2011) proposed the typological primacy model: in third-language acquisition, learners do not transfer selectively but wholesale from whichever language is typologically closest — even where that transfer produces errors.
Bardel and Falk (2007) identified a complementary phenomenon they call the L2 status factor: a previously learned foreign language often blocks first-language influence in the early stages of the next one. Your English will insert itself between your Vietnamese and the German you are starting.
Lindqvist (2010) measured cross-linguistic lexical influence in advanced learners of French and found it rising as the languages grew typologically closer.
For a Vietnamese learner taking Chinese and Japanese together this is very concrete: the shared characters help enormously with meaning and mislead constantly with sound, because the same character reads completely differently in the two languages. The advantage and the trap come from exactly the same source.
There is one further cost, measurable in the laboratory.
Meuter and Allport (1999) had bilinguals name digits, switching between languages on cue. They found switch costs — and something surprising: the costs are asymmetric. Switching back into the stronger language costs more than switching into the weaker one.
The accepted explanation is that to speak the weaker language the brain must suppress the stronger one harder — and that suppression has to be undone on the way back.
Green and Abutalebi (2013) place this in a wider frame: language control adapts to the context of use, and different contexts impose different control demands.
The scope should be stated honestly: these are millisecond switches, not Japanese in the morning and German at night. Extrapolating too far would be dishonest. But it does show that holding two systems live has a cost, and that the cost lands somewhere unexpected.
There is one level at which the proverb is not partly wrong but entirely wrong: inside a single language.
Master one trade suggests picking one skill and going deep. For languages that is bad advice.
As the piece on this site about being good at a language argues, the skills separate but are not independent. Receptive vocabulary feeds productive vocabulary. Listening improves speaking. Reading improves writing. Dropping a skill does not concentrate effort on the others — it starves them of input.
Schmitt (2010) describes word knowledge as a multidimensional continuum developing unevenly. You cannot choose to grow one dimension and abandon the rest; they pull each other along.
And Skehan (1998) adds another layer: complexity, accuracy and fluency trade off within any single utterance, but over the long run all three need training. Specialising in accuracy and abandoning fluency produces someone who writes correctly and cannot speak.
Within one language, nine trades is the right answer.
The pro-specialisation side is oversold too.
The meta-analysis by Macnamara, Hambrick and Oswald (2014) across domains found deliberate practice explaining only about 12% of performance variance. Hours poured into one thing matter, but they are not the whole story, and there is no magic threshold that converts a person into an expert.
On the other side, Erard (2012) shows the language counts people display are almost always inflated. Tracing the records, he found that genuinely extraordinary multilinguals typically have a small core of strong languages and a long tail at reading level.
Which is, interestingly, exactly the shape the proverb describes — but reached sequentially, not by choosing only one.
Pulling it together, here is the practical conclusion the evidence supports. It is neither only ever learn one nor learn as many as you like.
Get one language above the usable threshold before adding the next.
Usable threshold means something specific: you can read what you want to read without constant lookups, and you can follow a real conversation. For most people that is around high B1 or B2.
The reason is technical rather than moral. Above that line a language begins to feed itself: you can maintain it by reading and watching what you wanted to read and watch anyway, rather than through dedicated study hours. It converts from an expense into something close to free.
Below that line, every language demands dedicated hours. Two languages below the line demand double, and that is where things collapse.
On maintenance: Cepeda and colleagues (2006) show spaced review works by returning to the same material. A language already above the threshold needs a few hours a month not to decay — which a language at A2 cannot manage.
There are circumstances where parallel study is reasonable — work requires it, or you already have one language above the line. If so, the evidence suggests four conditions.
One: do not pick two close languages. Spanish and Italian together invites precisely the interference Rothman and Lindqvist describe. Spanish and Japanese is far safer.
Two: keep them at different levels. One language in maintenance and one in construction is much easier than two in construction.
Three: separate them by context, not by clock. Green and Abutalebi show language control adapts to context of use. One language for work and one for entertainment separates itself; two studied at the same desk does not.
Four: do not split evenly. Seventy-thirty means at least one will cross the threshold. Fifty-fifty may mean neither does.
The proverb is right about hours and about thresholds. Twenty-five per cent of two languages really is worth less than fifty per cent of one, because the payoff sits above a high threshold and linear is not the shape of the curve.
The proverb is wrong about what languages are. Trades are separate; languages share a foundation, and someone who has learned one foreign language learns the next more easily — Cenoz’s finding, and it cuts against the trade metaphor.
And the proverb is flatly wrong applied inside a language, where specialising in one skill starves the others.
A more accurate formulation for language learning might be: master one trade properly, then take up the second — and inside each trade, do all nine jobs.
It will never become a proverb. But it describes the shape of the problem correctly.
The same article, told in plain words — for younger readers, or for anyone who wants the point quickly.
There is a proverb about it being better to master one trade than to dabble in nine. Does it hold for languages?
Partly. And where it fails, it fails in an interesting way.
The payoff from a language does not rise in a straight line. It sits above a threshold, and the threshold is high.
Below it you cannot read what you want to read, or follow a real conversation — so the language gives you almost nothing back and needs constant feeding. Above it, the language begins to feed itself.
Twice nothing is still nothing. That is the real reason the proverb holds here — not because specialising is a virtue, but because of where the payoff sits.
If the story stopped there it would be a proverb in a lab coat.
People who already know two languages often acquire a third better than people who know one, with other factors controlled. That directly contradicts a crude reading of the proverb — if learning several languages diluted competence, bilinguals should do worse. They do not.
The reason: someone who has already learned a foreign language brings the experience of learning one. They know how to attend to form, they know some things do not translate, and they can tolerate not understanding.
Here is the concrete rule, and it is not the one the proverb gives:
Two close languages at once bring a lot of help and a lot of interference. Two distant languages at once bring neither.
Learners do not borrow selectively; they borrow wholesale from whichever language feels closest — even where that produces errors. And a previously learned foreign language often pushes in front of your mother tongue: your English will insert itself between your Vietnamese and the German you are starting.
For a Vietnamese learner taking Chinese and Japanese together this is very concrete. The shared characters help enormously with meaning and mislead constantly with sound, because the same character reads completely differently in the two languages. The advantage and the trap come from exactly the same source.
When bilinguals switch languages on cue, there is a measurable cost. And it is asymmetric: switching back into the stronger language costs more than switching into the weaker one.
The explanation: to speak the weaker language, the brain must suppress the stronger one harder — and that suppression has to be undone on the way back.
To be honest about scope: these are millisecond switches, not Japanese in the morning and German at night. But it does show that holding two systems live has a cost.
At one level it is not partly wrong but entirely wrong: inside a single language.
Listening, speaking, reading and writing do not substitute for each other. Specialising in accuracy and abandoning fluency produces someone who writes correctly and cannot speak.
Within one language, nine trades is the right answer.
Across many fields, deliberate practice explains only about 12% of the differences in performance. Hours poured into one thing matter, but they are not the whole story, and there is no magic threshold that converts a person into an expert.
On the other side, the language counts people display are almost always inflated. Traced carefully, genuinely extraordinary multilinguals typically have a small core of strong languages and a long tail at reading level.
Which is, interestingly, exactly the shape the proverb describes — but reached sequentially, not by choosing only one.
Neither only ever learn one nor learn as many as you like:
Get one language above the usable threshold before adding the next.
Usable means something specific: you can read what you want without constant lookups, and follow a real conversation. For most people that is around high B1 or B2. Above that line, a language needs a few hours a month not to decay — which a language at A2 cannot manage.
A more accurate formulation might be: master one trade properly, then take up the second — and inside each trade, do all nine jobs.
It will never become a proverb. But it describes the shape of the problem correctly.
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,...