The constraint is space, not translation
A second language on a panel roughly doubles the text, and the second language is usually not the one most of your international visitors read. A third and fourth make the panel unreadable long before they make it useful. Most museums discover this at design stage and quietly cut the languages back to one, which is how a collection ends up monolingual by accident rather than by decision.
Translation length makes it worse in specific directions. German and Finnish expand against English. Chinese and Japanese contract in character count but need a larger type size to stay legible. Arabic and Hebrew run right to left, which is a layout change rather than a text swap. None of that is difficult on its own; all of it competes for the same square metre of wall.
Split the label from the interpretation
The way out is to stop treating the panel as the place where all the content lives. Keep the printed label short and factual, the things a visitor needs while looking: what it is, who made it, when, and one line on why it is here. Put the fuller interpretation somewhere that has no space limit, reached from a code on the label itself.
Once the depth lives off the wall, adding a language stops being a design problem. The panel does not change, the code does not change, and the printed exhibition stays as calm as the designer intended. This is also the only version of the problem where translating into six languages is realistic, because nothing physical scales with the number.
Printed label: object, maker, date, one line of why
Behind the code: the full interpretation, in every language you publish
One code per object, however many languages sit behind it
Published as text and audio together, so it also serves people who cannot read the panel
What to translate first
If you cannot translate everything at once, translate the things that determine whether a visitor can navigate the museum at all before you translate the things that make the collection interesting. Orientation first: how to get around, what is where, what is included, what the rules are. A visitor who understands the building can enjoy the objects; a visitor who has beautiful object labels but cannot find the toilets or work out whether they are allowed upstairs is having a worse visit.
Then the objects people already stop at. You know which ones those are, because staff answer the same questions about them every day. Translating twelve of those properly beats translating the whole collection thinly, and it gives you real usage data to decide what comes next.
Version control is where this quietly breaks
The problem arrives six months later. Someone corrects a date on the English panel, and now the six translations say something different. Nobody notices, because nobody on staff reads them.
The only reliable fix is to keep every language of a stop in one place, so changing the source flags the translations that no longer match. If your translations live in a folder of documents from the agency, that check does not happen and the guide drifts out of accuracy language by language.
It is also the argument for locking the source text before commissioning any translation. Every edit after translation begins is paid for again in every language, and the edits after launch are the ones that never get budgeted at all.
Frequently asked questions
How do museums translate exhibit labels?
Usually by keeping the printed label short and putting the translated interpretation behind a code on the label, because panel space runs out after one or two languages. The alternative, printing every language on the wall, makes panels unreadable and has to be redone whenever the text changes.
How many languages can you put on a label?
Two before it starts to hurt, three before it stops being readable at a normal viewing distance. Anything past that needs the depth to live off the panel.
What should be translated first?
Orientation before objects. Wayfinding, what is included, and the rules matter more to a visitor who cannot read English than any single object label does, and they are also the shortest texts to translate.
Does translating labels mean reprinting them?
Only if the translation goes on the panel. If the printed label stays short and the translated text sits behind a code, adding a language changes nothing physical and nothing gets reprinted.
How do you keep translations in step with the English?
Keep every language of a stop together so that editing the source shows you which translations no longer match. Translations kept in separate agency documents drift silently, because nobody on staff reads them well enough to notice.