Persona 5: Phantoms of Translation
How Atlus fails fans of a landmark JRPG

Cutting to the Chase

Persona 5 is out in English, and it's arrived with a shower of praise from critics and fans alike. This should come as little surprise to anyone who played the game in Japanese. It's brilliant and innovative, and very fun.

But, there's a problem: the translation is riddled with errors.

That's not to say it's unreadable, or completely without merit. But for a massive RPG from a highly successful franchise, it falls incredibly short of the standard it should be held to. A video game is a professional work, no different from any other form of media. Yet no other form of media would ever get away with the number of errors found in Persona 5's English script.

This isn't the 1990s, when localization was still young and errors were endemic. Nor is Persona 5 a small game, or an insignificant one.

Toggle theme:

How Bad is Bad?

The baseline for any translation is this: readers of the translation should receive the same experience as readers of the original, as if the original creators had written it natively in both languages.

Persona 5's English localization does not meet this standard. Readers are receiving an inferior work containing awkward language, mischaracterization, grammatical mistakes, and outright translation errors.

ASK

Question Zone

What is translation?

#1

I'll start with what translation is, and what translation is not.

Translation is:

  • Adapting text from one language to another
  • Ensuring that the adapted text is faithful to the creator's intent
  • Conveying foreign concepts gracefully and without distortion
  • Allowing readers to enjoy the same experience in both languages

Translation is not:

  • A word-by-word conversion of language A to language B
  • Removing content that is difficult or controversial to readers
  • Leaving content untranslated because it is challenging to convey
  • Changing content to suit the translator's subjective preference

Localization is a broader process, and involves adapting content so it's more easily understood in different regions. This generally involves translation, and may involve other adjustments—such as converting between regional measurement systems or exchanging obscure pop culture references for ones appropriate for the target audience.1 However, localization cannot “fix” bad translation.

Localization can be a controversial term because of how video game translation and consumer awareness of it have evolved over time. Despite misconceptions to the contrary, good localization will never violate the principles of translation listed above.

You can find more information on the topic under the ASK item “What does good translation look like?”

1: The trick is to ensure that the reference is equally obscure and has the same effect both before and after, relative to the given audience. A common British reference might be literally impossible to understand for a Japanese reader, so a localizer will attempt to find a common ground that clicks with the target audience without culturally undermining the text. Localization is hard.

What does good look like?

#2

Translation can be a murky concept, so first I'll define a standard to measure against: imagine if translation weren't necessary at all.

In other words, imagine that the original writer is perfectly fluent in both languages and writes both versions personally. That creation—a perfect reflection of creator intent—is the standard to which translators must strive.

A good translator, accompanied by their editor and QA staff, gets as close to that vision as possible. By the time they're finished, the only flaws in their work should belong to the creator, not themselves. Occasionally, a translator may even correct minor errors that the creator did not intend to be present in their creation, such as typos and inconsistencies.

By extension, adding errors is a cardinal sin and a fundamental failure on the part of the translation team.

Translators must be skilled writers, ideally even more skilled than the creators of the works they translate—though this is not always possible. An excellent translator will see deeply into the creator's work and draw out the nuance and fine detail dwelling there.

Above all, translation is never a word-by-word conversion.

This is one of the greatest translation myths, and one of the most harmful. Translators do not convert words from one language to another: they convert ideas. This is why it's so important for them to understand creator intent. A word-by-word translation will always be a stiff, inferior product that betrays the original work and its readers.

How did this happen?

#3

Unfortunately, there's no way to know for sure without Atlus telling us themselves or one of the translation staff coming forward anonymously. However, there are some hints at what the underlying reasons may have been.

#1: The translation team was massive.

Yu Namba of Atlus has stated that Persona 5 “boasted the most number of translators and editors on a team”. 1 That sounds impressive, but it rings instant alarm bells for anyone in the localization industry: it's like marketing your hit novel as being “written by the greatest number of authors.”

And Namba's claim checks out: Persona 5's credits list a whopping six translators and eight editors as having worked on the project's localization.

Those eight editors are especially alarming: their job is to unify the translation, and that's hard enough under normal circumstances. But with more editors than translators, they're actually liable to worsen the stylistic inconsistencies.

#2: The release was delayed for localization reasons.

In November 2016 Atlus released a statement announcing that it would be pushing back the release of Persona 5 from February to April 2017:

“[T]he Japanese release of Persona 5 smashed all our expectations, and as a company, we decided that we owed our fans the very best effort to make Persona 5 our gold standard in localization.

“Practically, this means redoubling our QA and localization efforts, even returning to the studio to record previously unvoiced lines. We don’t want to rush this game and in this case, it meant we needed to move the release back for the last time.”

Of special note is the mention that they would be "returning" to the studio. This implies that they had already finished recording voiced lines back in November 2016, which necessarily required those lines to have been translated and edited by that point in time.

#3: Even voiced lines contain outright errors.

Normally, a script will be adjusted during recording to better reflect character voice and fix lingering mistakes. Not so with Persona 5.

*    *    *

Now, it's difficult to say exactly what happened. Perhaps there were unreasonable time constraints. Perhaps the localization team was underqualified. Perhaps Atlus had internal issues between its American and Japanese branches—it's not unheard of for international teams to cooperate poorly, and that often hurts localization efforts. Perhaps it was all of these.

However, there's one thing we know for sure: this was 100% avoidable, and would have been avoided had the parties in charge cared enough. The developer always has full control over the work.

In short, whoever was responsible simply thought they could get away with it.

1: The use of “most number of” is indeed poor English—consider the oddity of “many number of” if it doesn't quite click for you—but we'll cut Namba some slack and blame the editors of Sony's blog.

Where's all the noise?

#4

There have been complaints, especially by industry professionals. However, you'd be hard pressed to find a major review site that actually makes note of the errors in Persona 5. And indeed, some people have said that the translation is fine, or even extremely good.

So what gives?

This is actually a chronic pattern for media translated from Japanese into English. Here's an abridged explanation of why that is:

#1: Japanese media is relatively niche, and started out even more so.

For a long time, the only reason fans were able to consume Japanese media was due to the dedicated efforts of fan translators. The modern economic viability of Japanese media in the West is largely rooted in their efforts to bring Japanese content into other languages.1

Fan translators rarely have professional experience or an expert understanding of Japanese, and fans were tolerant of mistakes. The following are hallmarks of the era:

  • Literal, word-by-word translations
  • Stilted writing that mimics Japanese sentence structure
  • Stock translations of common phrases
  • Leaving Japanese words untranslated even when they have English equivalents
  • Translation errors

#2: Early official translations couldn't be trusted.

If fan translations were often flawed, official translations were often downright incomprehensible.2 But that wasn't the root of the problem. The root was this:

The official translations sometimes... sounded good.

But while they sounded good, they were also often censored, altered to be more “friendly” in the West, or even completely rewritten.

A new belief was born within a significant segment of the fan base: translations that sounded good could not be trusted. Only fan translations were guaranteed to be faithful to the original creation—that they were often stilted and contained untranslated content was proof of this authenticity. And importantly: it was proof that could be identified without knowing Japanese.

All it took was a glance and anyone could tell.

#3: Japanese translation has a global audience.

If English-native fans felt starved for content translated into their native language, fans whose first language wasn't English were even more so. Many ended up consuming English translations in order to experience Japanese media in a language they had at least some knowledge in—and this still happens.

Skill levels vary, but speakers of English as a second language are understandably less likely to identify awkward English writing. In fact, flawed translations are often a blessing: the awkward wording, limited vocabulary, and stock lines that appeared in fan translations makes the content much easier to understand for some non-native speakers.

#4: Video games are more than just text.

This one's pretty simple: if a game is fun enough, it can distract from weak writing. When translated games are dubbed, talented voice actors are also able to supply intonation to awkward lines that makes them seem less so. However, this greatly limits how good their performance can be and results in an inferior dub.

*    *    *

The result of the above factors (and more) is that complaints about awkward, literal translation are often countered with accusations of wanting to remove all references to Japanese culture.

In Persona 5's case, there's one more factor: people are pretty upset by Atlus's streaming restrictions. Understandably, this has drawn attention away from the poor quality of the translation.

1: And it should not be underestimated just how much work this was, especially for video games. The translators have to get the translation into the game somehow, and Japanese programming practices are famously infuriating to deal with.

2: Consider the memetic line “All your base are belong to us,” which comes from the 1992 translation of Zero Wing.

Where are the problems?

#5

If you look below you'll see a selector that will allow you to browse a list of errors. Accompanying each entry will be an explanation of what's wrong. An effort has been made to keep excerpts spoiler-free.

Before you dig in, consider checking out the other questions in the ASK section. To view an answer, just click on the question in the list above. They'll help you understand exactly what to look for and how each example is judged.

The errors listed below are a limited sample from a much larger collection, and are confined primarily to the first 5-10 hours of the 100+ hour game.

Defenses of Bad Translation

#1

“I prefer literal translations because they're more faithful.”

This is a pretty common view! And it seems to make a lot of sense. However, there's a problem with it, and it stems from the following commonly held belief: that translation is about words.

Translation is really about ideas.

This can be a bit confusing to anyone who hasn't thought about the distinction before. Words are basically convenient common terms for certain concepts, which people put together as building blocks to express larger ideas in. Languages form when different groups don't agree on which words correspond to which concepts—but those concepts themselves are universal.1

To phrase it another way: words are just a way for a person to take their ideas and pass them on to other people. If there were no other people in the world, there wouldn't be any need for words, but there would still be a need for ideas.

So how does that factor into translation?

Literal translation converts the creator's words directly into translated words. Faithful translation converts the creator's words back into the creator's idea before producing a faithful, translated idea. Literal translation converts the creator's words directly into translated words. Faithful translation converts the creator's words back into the creator's idea before producing a faithful, translated idea.

The processes behind literal translation and faithful translation look fundamentally different. Namely, faithful translators need to understand the original creator's ideas and methodology before they can write anything, while literal translators will just skip to the word phase, causing a degradation of quality in the process.

Poor translators worship words. Excellent translators respect words, but worship ideas.

1: Some languages make it harder to express certain concepts than others do (one language might require one word where another language might take five), but those concepts can always be explained. If this weren't true, it would be impossible to learn another language!

#2

“The language is a bit stiff, but everything is still accurate.”

Unfortunately, while it's possible for a translation to be stiff but understandable, stiff but accurate translations are pretty much a myth.

There are several reasons for this. The first is that writing ability is strongly linked to translation skill, and someone who produces stiff translations is therefore more likely to misunderstand the original text. Or they might understand the original, but make a mistake when conveying it in the destination language.

Another reason is the fundamental nature of language. Stiff translations generally suffer from being overly literal, and overly literal translations treat every language as 1:1 transformations of each other, as if every language can express the exact same concepts in the exact same way.

It's easy to prove that language doesn't work like this. Consider the Japanese ao, which is a color that encompasses both green and blue. English has separate words for both colors, meaning there's no easy 1:1 conversion between Japanese and English. The translator will always have to take context into account to decide whether to translate ao as blue or green.

If you're mathematically inclined, you might recognize this as a form of the pigeonhole principle. If you're not mathematically inclined, don't worry: it's barely math. The pigeonhole principle is just a fancy way of saying that if you have more items than buckets to put them in, some of the items will have to share a bucket.1

Coming across concepts that don't translate well into a particular language is common for translators. When this happens, they have to consider context and choose the option that best matches the original creator's intent.

Creator intent is also the third reason that stiff translations cannot be accurate. An accurate translation conveys more than just raw information: it conveys character voice, personality, and the charm of the author's style. A translation's accuracy is judged by how similar it is to what the creator would write if fluent in both languages. Unless the creator intended the text to be stiff, producing a stiff translation is inaccurate, and a failure.

Finally: while it's possible to detect overt errors, it's impossible to judge whether something is fully accurate just by the translation. Check out the first three error samples in the section above for a practical example of this.

1: So if you have to use your hands to carry three apples across the room, one of your two hands is going to have to carry an extra apple. Similarly, translation sometimes requires you to take a word from a single-bucket language and put it into a two-bucket language, or vice versa.

#3

“I like that I can feel the Japanese behind the translation.”

It's definitely great to get to experience the cultural aspect of a piece of foreign writing. However, that foreign nature should be expressed by the text's content, not by the text's awkwardness.

This goes back to creator intent. If the original creator were perfectly fluent in English, would they have made their writing intentionally awkward just so readers could feel how “foreign” it is?

Probably not. And thus, neither should the translator.

#4

“The game is massive, so mistakes were bound to happen.”

This sometimes also takes the form “You're just cherry-picking the bad parts.”

This argument is reasonable to an extent: translation is difficult, and mistakes do happen. But those mistakes should only be on the order of one misplaced comma per several thousand sentences, or a typo every hundred thousand characters. If the mistakes are frequent and noticeable, then they're unacceptable.

Consider—how would readers react if George R. R. Martin released his next book and every third sentence was awkward, with every fifth sentence containing an objective error? Writing is hard, and his novels are long, after all.

Yet I doubt anyone would claim those mistakes are fine.

#5

“There's nothing wrong.”

At the end of the day, this is the hardest defense to address. At the same time, it's also the easiest.

One reason someone might use this defense is that they genuinely don't see a problem, because to them those flaws aren't flaws. And that's valid, so long as they accept other people's right to believe otherwise.

Another reason stems from the fear people have to acknowledge mistakes in the things they like. After all, liking something flawed might mean one's opinion is bad or wrong. Or maybe acknowledging the mistakes will ruin the enjoyment entirely.

The default response is often defensiveness: those mistakes aren't serious, or they don't exist at all.

If this instinct ever grabs you, just remember: it's entirely possible to enjoy translated media and still demand that publishers do better. Because the work deserves it, and so do we.

Wrapping Up

I haven't listed every mistake in Persona 5, or even a substantial fraction of them. I've also been forced to focus on the translation aspect of localization, which means I haven't properly addressed other failings such as bad typography, untranslated images and video, and voiced lines that are unsubbed even when Japanese audio is enabled.1 Nor have I dedicated time to the sometimes strange handling of honorifics. For those aspects and more, I'll have to merely conclude by saying that there are few areas in which Persona 5's localization isn't lacking.

That I have to say that about such a great game is unfortunate.

If this bothers you even slightly, I encourage you to tweet at Atlus as well as discuss the problem with people you know. Fans deserve better, but publishers like Atlus will continue to produce sub-par localizations if nobody speaks up.

1: Keeping in mind that a lack of subtitles isn't a problem only for people who prefer Japanese voicing—anyone who has hearing loss is affected, regardless of their language settings. And roughly one in five people has hearing loss of some form.

Credits

This site was created and designed by Connor Krammer. You can get in touch with him on Twitter or via email at connorkrammer@gmail.com. He masquerades as an editor for a living.

Special thanks go out to BlackDragonHunt for hacking and translation help, as well as to the boatload of amazing people who contributed examples of poor localization and helped review the site before it went live.