I have written over twenty English script revisions for European computer games being localized to the United States market. Invariably these revisions were undertaken and completed prior to a game being shipped, during either the development of the original game or the development of the version specific to the U.S. Recently, however, I was asked for the first time to revise the translations of a game that had already been released in the U.S., the in-game texts of which were being severely savaged by American reviewers. Though already committed to releasing the revisions in a patch, even at that point the European publisher was still unsure how poor the translation of their game, Destiny*, had actually been. (*Name changed; see endnote  for more.)
Inspection of Destiny’s in-game texts indeed showed them to be woefully inadequate by almost any measure, and revision of the bulk of the 100,000-word translation was necessary simply to prevent the user from being constantly annoyed or confused during play. As the revision process went on and I came to better understand the original story, the main characters, and the often humorous dialogue, I also came to realize just how badly the translation had damaged one of the strengths of the original product, which was its whimsical tone. [For specific examples, see Appendix A: Translation Horrors.]
When the revisions were completed and the patch posted I spent some time thinking about how the myriad mistakes I had witnessed could have made it to store shelves. Having previously worked on script revisions that had gone about as well as they possibly could, I already knew that translation was simply one step in a successful localization process, not the only step. While the competence of the translators in this case was clearly less than it should have been, the publisher’s failure to recognize the difference between translation and localization also contributed to the hostile critical reception their product received in the U.S.
Recognizing the Requirements
While the translation of, say, a computer manual, only requires that the translator have sufficient knowledge of the languages and terms involved in order to communicate facts, translators of narrative works must also possess storytelling skills if they are to have any hope of conveying emotions between two languages. Moreover, because the craft of storytelling often requires knowledge of, and even reference to, the latest cultural norms, there is the added requirement that a localizer be able to translate a narrative from one culture to another.
It is in these steps beyond linguistic translation that Destiny also suffered, because there is always damage done to the connective tissue and emotional context of a narrative during translation. Only by making sure that someone with local knowledge is available to repair damage at the story level after a translation has been completed can a publisher or developer hope to convey the richness of the original texts. This means the ideal candidate for localizing (as opposed to simply translating) narrative in a work of interactive entertainment should have the following minimum qualifications:
- Q1 Fluency in the language of origin.
- Q2 Fluency in the language of destination.
- Q3 Professional storytelling ability.
- Q4 Current familiarization with the destination culture.
Obviously, the likelihood of finding a single individual with these qualifications for any given pair of languages is fairly small, and only becomes smaller when other desirable qualities – including punctuality, reliability, scheduling flexibility, good interpersonal skills, and a working familiarity with the jargon and development process of the interactive entertainment industry – are added to the requirements.
In instances where one person cannot be found, it stands to reason that these same skills will need to be distributed across two or more individuals. Recognizing the practical difficulty of finding translators in any country of origin who have both storytelling skills and cultural knowledge applicable to a particular destination country, it’s fairly obvious that the simplest division of labor involves initial translation by an individual qualified for that particular task, followed by localization of the translated texts by a second individual. This would divide the four previous qualifications as follows:
- Q1 Fluency in the language of origin.
- Q2 Fluency in the language of destination.
- Q3 Professional storytelling ability.
- Q4 Current familiarization with the destination culture.
With this division of labor the odds of actually finding qualified people who can do the work go up considerably, and the job of translator now accurately reflects the description of the people and companies routinely offering such services. Additionally, because the position of translator is usually factored into the development budget, only one additional hire will need to be made in order to ensure that all qualifications for localization are met for any given destination country.
Protecting the Product
While all of this analysis has helped us identify the individuals and skills required for the job, that’s not the primary goal. The primary goal is to create the least-convoluted effective path between the person who created the original content and the person who will be trying to resuscitate that content after it has been translated. This will not only ensure that the entire process is as cost-efficient and rapid as possible, but it will also provide the greatest assurance that the localization will preserve the entertainment value of the original content.
Expressed as a rule, the translation and localization of narrative elements will be more successful as the number of people handling the content decreases, which is specifically why I have referred to the translator and localizer as individuals. Effectiveness in translation or localization is inevitably diminished if more than one person works on these tasks precisely because multiple individuals never fully assimilate the elements of a work as a single individual must. While that’s not a problem with the translation of a computer manual, where one person can write the set-up instructions and another can describe the motherboard features, it is a very real problem when translating fiction. The reason for this is that what a character says or does at the beginning of a story can and should have a profound effect on what that character will say or do at the end of the story, or what other characters will say or do in response to that character’s choices. Translators and localizers who do not have visibility to an entire story by definition have no hope of being able to replicate these causal connections during the localization process.
While localization will usually only involve one person because of the relatively rare professional storytelling requirement, the translation process may involve multiple translators simply because the company contracted with has a number of qualified personnel on staff. The tendency to use translation companies at all is partly reflective of the fact that they are easier to locate than qualified freelancers, but it also reflects the fact that the translation process often takes place under severe scheduling pressure. During crunch time, having multiple translators assigned on a rush basis may not only seem advantageous but essential, but this perception is short-sighted. Scheduling gains from accelerating the translation process by using more than one person will be at least partly, if not wholly forfeit at the localization stage, because the localizer will have to correct not only errors in translation but also disagreements between translators.
The final step to protecting the integrity of original content involves making sure that lines of communication with the original content creator(s) are maintained during the localization process. Both the translator and the localizer will need to ask any number of questions about the content as they each perform their work, and any delays in getting answers will inevitably slow the localization process down. If neither the localizer nor the original content provider speaks the other’s language the translator may also have to translate questions and answers between those two parties. If a translation company was hired to do the translation this may prove problematic, and particularly so in instances where more than one person was responsible for the translation.
Calculating the Cost
An implicit assumption driving this discussion has been that the economic cost of good localization will eventually be recovered through increased sales in the destination country. Unfortunately, if any cost/benefit analysis has been done concerning how the quality of localization impacts revenue generation, I’m not aware of it. While this means I cannot claim that improving the quality of localizations will necessarily result in increased sales sufficient to recoup the cost of attaining that quality, by the same token it’s impossible to state categorically that such an increase will not take place. Ultimately all sales-driven arguments for deciding whether or not to improve the localization of narrative elements in a game are moot, meaning we must appeal to other metrics on a case-by-case basis in order to determine whether it is cost-effective to invest in quality.
Unfortunately, data and standards by which we might actually assess various levels of localization quality are equally absent, but here at least we can describe a minimum threshold that we must reach. Assuming that the goal of localization is to communicate the positive qualities of a given game across a language barrier, the minimum standard of quality for the localization of narrative elements is the level at which no trace of the localization process is apparent to consumers in the destination country. At this level of quality users may not actually be entertained or moved by the narrative content, but at the very least they will not be confused or irritated by it. The most important benefit of reaching this minimum threshold is that all other aspects of the title, from gameplay to graphics, will still be accessible to the player.
In order to attain this minimum standard of quality it will in most cases be necessary, as outlined above, to hire two individuals: a translator and a localizer. While this cost will clearly put additional pressure on the development budget, it should be remembered that provisions for hiring a translator will probably already have been made, which means it is only in hiring a localizer in the destination country that we are actually adding to the project. A related problem, however, is that while this additional cost will look small during pre-production, by the time the need for these services actually arises the project may be behind schedule or facing considerable budgetary pressure. These factors, in turn, may tempt management to use the lack of direct evidence tying localization quality to increased revenues as an excuse for diverting localization funds to more pressing needs.
Mitigating this tendency is the fact that it doesn’t matter whether the developer or publisher is contractually obligated to conduct the localization, because the failure of the responsible party to adequately perform can materially damage the other party as well. For example, if the developer is responsible for the game’s localization, poor performance may mean the publisher’s marketing campaign will go to waste. If the publisher is responsible, poor performance can kill any chance of the developer seeing royalties as sales dry up. Because of this mutual reliance, not only should there be clear agreement beforehand as to who will perform the localization, but also which steps will be undertaken in the localization process, and who will perform that work.
Beyond preserving goodwill in the developer/publisher relationship, good localization can prevent other intangible problems from occurring. While putting an economic value on some of these benefits may be difficult, it is still important to recognize that gains can be derived from emphasizing quality during the localization process, particularly by limiting downside risk. For example, a single poorly localized game may be enough to tatter the reputation of a developer or publisher in the critic’s and consumer’s eyes, causing them to view subsequent titles as less worthy of interest, while simultaneously leading to increases in tech and customer support costs and returns from retailers. On the other hand, a reputation for producing good localizations may not only increase sales and enhance the developer or publisher’s reputation, it may also lead to additional business opportunities from other developers who would like to ally themselves with a publisher that has proven it can localize to foreign markets.
Trust as a Tool
One very real risk associated with hiring any prospective employee is that the applicant may not be who they claim to be. Every company makes a bad hire from time to time, discovering too late that a new team member requires constant management and motivation, or worse, that they are actually destroying morale and efficiency. This kind of personnel risk is only exacerbated when it comes to localization because of the inherent problem of getting to know people who may telecommute or speak a different language. This is true even when working with reputable translation companies, because it may be hard to gain visibility as to who exactly will be doing the work behind the corporate facade.
Unfortunately, it’s not only possible that the persons entrusted with the translation and localization of a game may be unmotivated, it’s also possible that they may be hostile to the industry or the content being localized, even to the point of sabotage. If neither the publisher or developer takes the time to check over the translation before the localized content is shipped – and this can be particularly problematic if the translator is the only person who speaks the language of destination – considerable damage can be done. For example, here is a line from another game I helped localize, as it was originally intended:
As [he] explained to me later, the satellite was a military artifact of the 21st century. With it, earthquakes and volcanoes could be triggered from space, destroying whole civilizations without waging war.
Now, here is the version that was passed to me by the translators of the original text:
As [he] explained to me later, the satellite was an American military device of the 21st century, whose function it was to destroy whole civilizations by triggering earthquakes and volcano eruptions without having to wage war.
Where the word ‘American’ came from is anybody’s guess, but it was clearly not in the original text. Did the translator simply make a mistake, or were they trying to inject their own political views into the game? Whatever the case, the important point is that monitoring a worker’s performance is hard enough in these situations without having to worry that they may actually be trying to damage the product. Because of the language barriers and potential distances involved, the localization process requires more trust that usual on the part of the employer, which means that requirement should be emphasized when making hires for the positions of translator and localizer.
In the case of Destiny, there seems little doubt that the translators gave assurances about the quality of their work that were factually false. Whether that was an innocent mistake or outright deception I don’t know, but the result was the same: the potential success of a product destined for the U.S. market was crippled because the quality of the localization was not what it was purported to be. In order to make sure that unproven translators and localizers are not the only people qualified to judge the quality of their work, the party responsible for overseeing that work should make arrangements to have a reliable third party review samples of the translations as soon as possible.
The long-term goal of any localization process should be identifying and maintaining relationships with individuals who have demonstrated their ability and commitment. Trust can not only speed the overall process of localization, but it will make subsequent projects flow more smoothly.
The localization of Destiny failed for multiple reasons, none of which was even remotely specific to the product itself. First, the translators were not qualified to do the work they were doing. Second, insufficient effort was made to double-check the work the translators were doing, which meant problems were not discovered until the game reached reviewers in the destination country. Third, multiple translators worked on the texts, leading to inevitable disagreements and mistakes between translators. Fourth, no one was hired to localize the translated text, effectively preventing the U.S. audience from enjoying the entertainment value of the original content.
If there is any overarching lesson to be learned from this it is that successful localizations can be made routine within the regular development process, provided attention is paid to the goal of protecting the original content. Given the ever-increasing complexity and sophistication of products vying for consumer dollars in the global marketplace, it is inevitable that the localization process must be improved if developers and publishers are to have any hope of consistently mining opportunities in foreign markets. This is particularly true for products being localized to markets in which there is already considerable domestic development.
— Mark Barrett
Appendix A: Translation Horrors
What follows is a sampling of the revisions I produced for Destiny. I have grouped these revisions into categories defined by the effect the original translation error had on the player, and listed the categories in ascending magnitude of the damage done. While these examples are in English, it should be emphasized that the same types of mistakes can and do take place when translating between any two languages. Note also that some revisions include changes not solely dictated by the original translation.
Grammatical mistakes include obvious errors in syntax, spelling, etc. These errors destroy suspension of disbelief and disrupt enjoyment and gameplay, but they do not actually confuse the player as to the original intent. Grammatical mistakes such as the next example indicate to the player that the game was rushed or badly produced.
Even if there is a grilled chicken for a fine meal, at the moment I’m not hungery.
Admittedly that grilled chicken would make a fine meal, but I’m just not hungry right now.
One skill translators usually do not have is knowing when not to use formal or even correct grammar in order to preserve the voice of a character or narrator. The following example appeared after the line, “Oh! Then maybe you met him?”
The motivation for change in the following example was cultural and authorial, turning a flat and literal sentence into a part of the narrative context of the game.
That won’t be enough for the next winter.
That won’t be enough for the coming winter.
If an American reader knows any definition of the word ‘commode’ at all, it is that of a toilet. In Destiny a number of tables were identified as ‘commodes,’ which met a European definition of the term, but only provoked unintended hilarity in the U.S.
The following translation attempted to describe a ranger.
A sincerely hard-working nature-boy of nature.
A steward of nature.
The following weapon was intended to be something akin to an ice pick or dagger. The translators were obviously unaware that ‘prick’ has extremely limited use as a noun in the United States.
Prick of death
Spike of death
Even after viewing the following lines in the context of play I still had to ask what these dialogues meant. Mistakes of this kind, which actually confuse the player as to meaning, are worse than simple grammatical mistakes because they completely eject the player from the game until the player can determine (or guess) what the original intent was.
The first sentence below seemed to reference some earlier event that I could not remember. Had I missed it? No: it was simply another mangled translation.
I acted rashly. Maybe we’ll see each other again later…
Sorry, I’m in a hurry right now. Maybe we’ll see each other again later…
The following line was a sarcastic reply from a character whose neighbors had cowered indoors as monsters roamed their village. Notice how no sarcasm is detectable in the original translation.
It is noticeable that I’ve ended up in a village of heroes…
I suppose you noticed how many other heroes we have living here…
The following instructions pertained to a game of chance the player could play. I tried everything I could think of, but eventually had to ask for another translation from the developer.
With a card turned inside out you can turn the fate determined for you into the opposite. But be cautious, the unalterable can never be outwitted in such a simple way!
If you accept the symbol, leave the card face up. If you reject the symbol, leaving the card face down reverses the effect of the symbol. But be careful. It is not easy to intuit the meaning of the fates!
The following item was a beverage apparently familiar to European consumers. To U.S. audiences it sounded more like an insect, and made no sense in the game’s context.
This botched translation, complete with variables, made it almost impossible to figure out a basic game mechanic while playing the game. Translation of texts like these demands communication between the translator and developer.
Spell fails with %s
%s evades the spell!
Translation errors which confuse the player and prevent imaginative involvement with a game are bad enough. Errors which seem to make sense are even worse, because the player has no idea that they have just processed something incorrectly. When they later run into problems, they may have no idea how they got off on the wrong track.
Because of the use of the word ‘precious’ in the following translation I thought I needed to do something with the mirror. I tried numerous items, numerous actions, and came back to the mirror repeatedly, only to be told that it was purely decorative.
A precious mirror.
A flawless mirror.
The following use of the word ‘south’ implied a reference to the larger game map, not the smaller locale I was visiting. After heading off to search the southern reaches several times I finally found the character I was looking for only one room away from where I had been given this hint.
Mostly nowadays he spends his time in the south.
I think he’s resting up around here somewhere…
The following lines confused the word ‘incurs’ with ‘causes’, and led me to believe I was actually injuring my own party by performing a number of actions. Again, basic game mechanics were destroyed by the translation.
%s attacks %s with the cold hand and incurs %i damage.
%s attacks %s with the vampire hold and causes %i damage.
Here are some instructions for the game of chance noted above. I actually followed the instructions, which produced results, but at the time I didn’t know they were the wrong results.
Left mouse button: Click right mouse button: Move two cards around to the left then accept fate.
Left mouse button: Select and move cards. Right mouse button: Flip cards over.
Finally, here is a translation error which not only blows the player out of the game, it also rudely forces them to confront a hotly-debated aspect of the interactive entertainment medium itself. I call it a deal breaker because it actually punishes the player for having trusted the people who made the game.
A store for murder tools of all kinds.
A store selling weapons of all kinds.
One of the basic tenets implicit in any audience’s willingness to suspend disbelief and enter a fictional world is that the audience will not be punished for trusting an entertainer to take care of them. While not intentionally malicious, this mistake, like the others noted above, nevertheless laid waste to the delicate state of suspended disbelief that Destiny’s creators had originally achieved. As a result, critical and commercial reaction to Destiny was significantly less than it should have been in the United States.
 The name of the game discussed in this article has been changed because I do not want to imply that the specifics of that title were uniquely responsible for the problems that were encountered. Botched translations are hardly new in interactive entertainment, and some, such as, “All your base are belong to us!” from Zero Wing (Sega Genesis/Megadrive), have even achieved notoriety.
 There will always be some publishers trying to make a quick buck by peddling garbage to unsuspecting customers, but I do not believe that to have been the case in this instance. The fact that the publisher was surprised by the negative critical reaction they received leads me to believe that they had every intention of meeting the threshold language requirements in the U.S., and the fact that they patched the texts after the game shipped supports that view.
 A foreign company that was unaware of the 9/11/01 attacks on the United States, for example, would run the risk of seriously misjudging America’s acceptance of terrorism as a motivation for gameplay. More commonly, cultural awareness affects subtle relationships such as how a character of one gender or social class might speak to another, or which words actually make a joke humorous in a particular context. Even if a translator also possesses storytelling skills, the absence of this kind of cultural knowledge can have a serious negative impact on the enjoyment of the intended audience.
 In some situations additional cost savings may be also possible. For games being localized to the U.S. market there is an increased likelihood that the original creator may be able to act as the translator of their own work, because a number of European and Asian countries include English in their core curricula. In instances where the destination languages are substantially similar, such as American English (AE) and British English (BE), a single translation to either variant may be sufficient in order to facilitate localization to both. The cost of those subsequent localizations could be further reduced by having a full localization done first in, say, AE, then a considerably less-involved localization done in BE. The AE localization would solve most of the translation and storytelling problems, leaving only grammatical and cultural differences to be resolved for BE.
 On Destiny there were multiple translators working on various parts of the texts, which meant that no single individual had custodial responsibility for the work as a whole. This led to instances in which, for example, the same character had a slightly different name in different parts of the story, which probably would not have happened if one person had been responsible for translating the whole work. There were also subtle differences present in how various translators handled tone and humor, which were apparent at the story and cultural levels.
 It has been my experience that if a developer has even minimal capacity in English it can be enough to allow me to ask and understand questions directly, without need of a translator. Although there might be occasions when I have to ask a question several times, altering my word choice, I have usually been able to bridge any language barrier when asking factual questions about the intent of a specific line. In the case of Destiny, the English of the developers was more than adequate to this task, and in many instances seemed superior to the English of the translators.
 By this measure the localization of Destiny clearly failed, because every reviewer specifically mentioned that the translations were abysmal. Indeed, many of the reviewers even expressed frustration at this, stating that Destiny was otherwise an entertaining game, but they couldn’t enjoy the good parts because of the badly mangled language. The most important thing to learn from this particular case is that even when all other aspects of a game are deemed enjoyable, poor localization can preclude the critical and commercial audience from enjoying a work.
 Localization efforts which go beyond the minimum level of competence and actually involve the user on an emotional basis simply add an additional facet of enjoyment for the user. Whether this additional level of quality needs to be present in order for the game to succeed in the marketplace will probably be determined by the degree to which that part of the game was central to the success of the title in the country of origin.
 I was not privy to the contractual agreement between the developer and publisher of Destiny, so I cannot comment on how that relationship may have impacted the resulting translation. It seems obvious, however, that if the agreement had specified that a single individual would be responsible for translation, and a single individual would be responsible for localization (anticipating the role I eventually assumed), most of the problems that did occur during translation would have been eliminated.
 A nod here to Steve McConnell, author of Code Complete (which every programmer should read) and Rapid Development (which anyone involved in software development should read). In Code Complete, Steve used examples he called “coding horrors” to show how difficult it could be to understand poorly organized or written code, and I have adapted that term for the same reason.
 The player’s usual reaction to a translation or localization mistake is to think that the people who made the game are stupid, even if other aspects of the game are outstanding. This tendency comes from the fact that while most people may not be particularly literate, their cultural sensitivity to the way their own language should read and sound becomes acute at a very young age. It is in fact probably the case that the younger the player the more likely they are to see a mistake as indicative of idiocy, precisely because even at their young age they already recognize the mistake themselves.
Copyright © 2002 by Mark Barrett