"Telephone game" effect when translating between languages

Question

Let's say there are three people, Bob, John and Jane. There are also three languages, L1, L2 and L3.

John is a fluent speaker of L1 and L2. Jane is a fluent speaker of L2 and L3.

Bob only knows L1, but wants to learn L3.

John, to help his friend Bob, translates a list of vocabulary from L1 to L2. Then he passes it on to Jane, who translates the words from L2 to L3. Thus, Bob has a dictionary of words between L1 and L3, as a tool for learning L3.

The question is, given any set of three languages, is there likely to be significant "telephone game" or "chinese whispers" effect by doing this? Does Bob risk learning inaccurate or plain wrong definitions of words by working this way? Are there any good examples of this? Would this depend any on the language families of L1, L2 and L3? Would this be more, or less, complicated with full sentences, than with loose words?

Answer 1

Does Bob risk learning inaccurate or plain wrong definitions of words by working this way?

Yes, he does, in most cases.

First, languages are often ambiguous. You may grasp the meaning by the context, but if there's no context, you don't know, for example, what is the meaning of refrain, is it a verb or a noun? If it's a verb, it can be to repeat or to stop yourself of doing something.

Some other paradoxes may also occur:

"time flies like an arrow" → "each of a type of flying insect, "time-flies," individually enjoys a different arrow (similar comparison applies)" (Wikipedia)

Note, no translation yet, it's just a result of machine-aided syntax analysis.

Would this depend any on the language families of L1, L2 and L3?

Ambiguity is not an attribute of translation, it's a property of an individual language. Translating between two languages belonging to the same language family, you preserve the original ambiguity, hence the translated word works fine in the same contexts as it did in original language.
However, if the languages are too far away, they naturally have different sets of homonyms, different rules of morphology, idiomatic constructs, phraseology, and so on. In this case, during translation, you should choose a single meaning that may appear to be wrong in another context.
Obviously, the more translations in chain, the higher the risk of a mistake.

Finally,

Would this be more, or less, complicated with full sentences, than with loose words?

The more context, the smaller the inaccuracy is because the translator knows which context to pick.

Answer 2

The answers by @bytebuster is certainly accurate but I would like to elaborate on certain aspects I would consider important.

1. If you were to translate individual words from L1 > L2 > L3 you are almost guarantee a game of telephone. In fact, all you have to do is to try to do this between L1 and L2. When you then go try to use the words in a sentence, you'll get almost immediate nonsense. Word to Word translation without context will only work on the most concrete and universal words such as 'orange' or 'automobile'. Even words like 'bread' or 'beer' may get you into trouble depending on the languages. And you can forget it with verbs 'go', 'see', etc. Sentences may still give you some issues but when you get to complete texts (assuming competent translators), you actually get relatively good results.

2. This is not at all a hypothetical situation. Explorers and conquistadors often used native translators who only knew one language who would find somebody who knew another language spoken by the locals and translate this way. I think Cortez is reported to have done so but I think you'd find reports going to antiquity. I'm sure any such translation would have had to rely on a lot of negotiation and clarification. Also, throughout the ages many texts (most notably the Bible) would be translated from Latin rather than the original sources. Today, you would mostly find original translations of published books but newspaper reports are often translated from third languages (most often English). There are even some academic disciplines where at least some data is collected in this way. Some aspects of Holocaust studies come to mind. Survivor memoirs have been written in so many languages (Czech, German, Italian,...) that no one scholar is likely to master all of them and many rely on English translations. Comparative religion is also likely to rely on at least some translation, as is much of anthropology, etc. What you will find in all of these examples is that nothing like a 'game of telephone' develops. In fact, these third translations are pretty faithful. Of course, they lose all nuance and any semblance of language play but they get the gist over. There will be many local mistranslations and misunderstandings but fewer than one might expect. As always, the more shared context, the easier this will be.

3. One area where this type of translation is quite routine is poetry translation from less commonly taught languages. It is quite unlikely that a there is a Hungarian poet who is also very good at Urdu. So such a person might collaborate with someone who speaks Urdu and Hungarian who will convey the meaning and point out any language play when they want to translate Urdu poetry. Of course, here, we're not looking for accuracy of meaning but rather keeping the gist and impact. My favorite book is 12 Czech translations of Poe's Raven - of which at least half of the 'translators' (all well-known Czech poets) did not know much if any English. There is a lot of diversity of expression but in all 12 instances you feel like you're reading Poe.