Translation Is Literally Impossible
Translation is literally impossible. No matter what you do, you can't translate from one language into another. All you can do is give a general approximation, and as a translator, your job is to get as close as you can.
It's said that the baseline for translation is for readers to receive "the same experience as readers of the original, as if the original creators had written it natively in both languages". It's easy to argue that this isn't possible, no matter what, because a translation will never have exactly the same patterns of words as the original work. But if you give up on exactness, and treat this baseline as an ideal instead of a target, you can start using it.
It's important to think about translation this way. Translation isn't the activity of replacing words, sentences, or conversations in one language with ones in another language. It's about replicating the original work, for people who can't consume the "original" untranslated work.
By changing it at all, it's no longer exactly the same. But consider this. Everyone experiences media differently in the first place. A native English speaker will see a song with engrish in it much differently than a native Japanese speaker would. Someone who grew up as a devout Christian will see Christian symbolism much differently than someone from an agnostic hippie family.
When you adapt a work to a new language, the target audience is inherently different, and would have taken different things out of the work even if it was originally meant for them.
In a sense, you just have to make the differences caused by translation be lesser than the differences between different readers.
What constitutes the experience of consuming a work? There are countless elements, but it's easy to list off a bunch of important ones where translation will have an obvious impact. I'm going to start off with ideas that don't necessarily have to do with translation, then show how they make translation "impossible".
It was merely an act
Statements are actions, not a jumble of words. Phrases that are used in specific situations, like "you're welcome" and "thank you", are the most obvious example. It would be asinine to translate "thank you" into a phrase that isn't commonly used to express gratitude, no matter how literal such a translation might be.
But this extends beyond common phrases. You can't draw a line somewhere and say, "stuff like this has to be translated based on the action it's doing, everything else has to be translated by literal meaning".
There's no such thing as literal meaning. There are a million ways to interpret the intent and nuance behind a given statement. Skilled authors are capable of manipulating the reader into the right state of mind to interpret everything correctly, but if the reader is too far from the work's target demographic, they're not going to get it.
The purpose of communication is to convey information in a particular way. This has to deal with the assumed knowledge of the person receiving the communication. By making a statement, you are performing the action of conveying information in a particular way to someone with a particular background knowledge. Statements are always actions, and the typical word-by-word translation does not perform the same action as the original statement.
You can think about this as "linguistic" (having to do with language itself) vs "non-linguistic" information. A statement is encoded with linguistic information, but really, what's being communicated isn't linguistic information, the communication itself is non-linguistic information that cares about mood, expectation, knowledge, and so on.
Right here, I'd like to kill some old concepts and introduce new ones.
"Literal" translations are ones where words or phrases are replaced with equivalents in another language, and anything that can't be translated is adapted in a way that makes sense. "Liberal" translations are ones where words, phrases, and sometimes even entire interactions are changed so that they have the least friction with a given audience. You might think about this like translation versus localization, or something like that. But that's all wrong, because this distinction is assuming that meaning and linguistic information are the same thing. Here's a better idea.
There are "spartan" translations, that focus on representing the original linguistic information, so named because of their indifference to comfort. And there are "astute" translations, that focus on representing the original non-linguistic information, so named because of their focus on sense. Spartan translations respect "wording", and violate messages. Astute translations respect "senses", and violate words.
The typical spartan translation from Japanese to English might use the phrase "it can't be helped" a few times. This isn't a literal translation. The phrase it's translating, しかたがない or しょうがない or similar, literally means "there is no method" or "there is no way of doing". Let's call this Phrase X. There's nothing about helping here. But this is a stock phrase, with a meaning about giving up on something, and "it can't be helped" is the same way. While this isn't a literal translation, it's definitely a spartan one.
Most of the places where Phrase X is used are places where no native English speaker would use "it can't be helped". Maybe an English speaker would use "oh well" instead. That's radically different than "it can't be helped". This is what makes it spartan.
You have to choose what to retain. Which action do you want your translation to keep? The action of using one of the most common stock phrases in the language? Or the action conveyed right then and there by using that phrase in that situation? Either choice is valid on some level, but it's a strict choice. You can't have your cake and eat it too.
Choosing "it can't be helped" introduces its own new meaning. It's no more "literal" than using a variant phrase like "oh well". This is a mirror reflection of what happens when you use "oh well". Choosing "oh well" means that you no longer make a statement about "something not being there, or not being possible". You have a pure speech act.
Phrases themselves are information, and that's true from both perspectives. A spartan translator wants to retain the identity of each phrase, because that phrase itself is part of the experience of the original work. An astute translator wants to retain the identity of each sentiment, because a given phrase will express a unique sentiment in different situations. It's rare that you have a perfect mapping where both languages use a unique phrase, of their own, in the same situations and nowhere else.
You might be thinking, of course. Different cultures act different ways. It's natural that you have to choose between adapting culture or representing it. Buddy, that's only the tip of the iceburg. This ship is going down.
The moment you start studying grammar or linguistics at a reasonable level, you start to wonder what kinds of similarities there are between languages. After all, the more similar two languages are, the easier it seems it should be to translate between them. It turns out that there are a lot. Languages are so similar, in fact, that someone who already knows a language can get stranded in any other society in the world -- absolutely any -- and become conversationally fluent in the language spoken there in a matter of months. They're going to miss a lot of nuance because they simply have not had the time for a lot of the cultural aspects of communication to sink in, but the linguistic parts adapt very quickly, much faster than learning a first language the first time.
Part of the explanation for this is an idea called "Universal Grammar". Like any important-sounding word that doesn't describe a specific theory, there are several ways to think about Universal Grammar, or UG. The hard version of UG, or the hard stance on it, is that all languages use exactly the same building blocks, and all humans have those building blocks in their head somewhere. The soft version is that every human has the same linguistic tendencies, and languages are never going to deviate from those tendencies too much.
UG is applicable to how translation should be done because, at some level, it gives you a list of things that are definitely linguistic details. But it gets more interesting than that. If something is universal, it means that a given language *must* have a way to deal with it. Every language has a way of dealing with certain things. Time. Embedded descriptions. Parts of speech. Topic versus comment. New and old information. Attitude. Hypotheticals. However, each language might deal with these things in radically different ways, so different that you don't even realize that they're the same thing.
The ways that Japanese and English deal with attitude are so radically different that it's hard to realize that you're losing information when you eliminate them during translation. When you translate いいんだ into something like "it's fine" or "it's alright", you lost information. Information that both languages can express. After all, that んだ means something radically different than what よ or です mean, but having those instead could translate into "it's fine" or "it's alright" just as easily.
In this panel, the んだよ in the box on the left is a clear followup on the attitude introduced by いいじゃないか in the box on the right. If you were translating this entire panel into English, establishing a similar connection would be trivial. But translating the presence of the んだよ itself, as a literal entity with the right meaning, is literally impossible.
This isn't a pathological case. The typical Japanese statement is, in some way, impossible to translate literally into English, when it comes to its attitude. This isn't limited to attitude, either. Anywhere where the two languages use sufficiently different ways of expressing things that have to be there, it's impossible to create a literal rendition of the original Japanese. Likewise, it's impossible to create a literal rendition of the original sentiment, because parts of that sentiment have tiny, miniscule amounts of nuance that are deeply tied to parts of Japanese grammar that cannot be adapted into English no matter how hard anyone tries.
This is without considering any differences in culture or available vocabulary, like honorifics, polite speech, and differences in social structure. This is only considering expression in the general sense, with things that are universal to all languages, regardless of culture.
Literal translation is impossible. Translation is literally impossible.
If you've been especially mindful while reading this rant, you might have noticed a lot of places where I suddenly stop explaining something to make a strong declarative statement. This is an example of statements as actions.