The Generalised Kaohsiung Technique

The Generalised Kaohsiung Technique is a particular set of rules that one can use to turn one name into another. It involves using the interplay between multiple languages and the friction between translating between them to create names that are easily explainable if needed, but otherwise create results that are completely unrecognisable given just the original.

The technique can be used if one wants to parody, skewer, obliquely refer to, riff off of or otherwise hint at something without referring to its name directly, and this was the original idea behind doing so. However, it can be used for other purposes, and its origin is not at all based on the idea of generating parody names.

There are two forms of the Technique: the original version, which is tightly defined, and the generalised form, which allows for many variations to be made. In this article, I will describe the technique, starting from why it got its name to how I like to use it.

Origin of the name

The Kaohsiung Technique is called that because of a city in Taiwan that had its name changed several times throughout its history.

At the time when indigenous Taiwanese, there once was a settlement that was called "Takau" in Siraya, which was spoken then. Back then writing was not prevalent in Taiwan at all, and the exact etymology is not known well, so this is not a settled truth, but it works well enough as a starting point for our adventure.

We'll skip over the time when the Dutch took over the island for a brief stint, but we will note that they notated the name as "Tancoia".

Soon after when the Chinese set up outposts on the eastern side of the island, they introduced writing to the island, if only just to tell where things are and how to get there. As with all Chinese writing, all names are written in characters, which inevitably bring in unwanted meaning to words that don't have them. The history is complicated now, but back then they just used whatever characters sounded the closest without caring about what those characters mean. So they chose to use the characters "打狗", which in the settlers' tongue is pronounced "Takau" at the time (with 打 being the part that is pronounced "Ta", and similar), to represent the place name.

Fast forward a few more years and now we have the Japanese taking over the island. They saw the characters "打狗" and they thought, "wow, that means 'beat dogs', that is so ugly and bad." And so, they respelt the name "高雄", which in Japanese is pronounced almost the same (Takaō). That it was also more Japanese is helpful! (And also that I was reminded that at the time they did not have the Western sensibilities with regard to the treatment of dogs make the justification lean more to the latter than the former, but surely both had some impact with each other, so whatever.)

And finally, the Chinese are back again. But unlike last time, when they had to make up words, the Japanese have helpfully left behind a lot of words ready-made for them. They can just take the characters and pronounce them in the way that they have always done. And so, the spelling 高雄 was re-understood as being Chinese and then re-pronounced "Kaohsiung". And that is what it is now pronounced today.

The technique

So if you have figured it out by now, the key behind the technique is to exploit the interplay between the written language and spoken language, and using one to pun on the other, and then iterate until unrecognisable.

The original Kaohsiung Technique involves using a word in the original language (usually English), and also Japanese and Chinese. The latter can be any "Chinese" that you have on hand, and in fact should be more accurately termed as any language that uses a lot of Chinese characters a lot. Even Korean can work at a pinch if that's really needed!

Here's how it goes:

Step 1.

You take your original word, written in whatever language you like.

Step 2.

You mould that word so that it conforms to Japanese phonology. Japanese phonology is very restrictive – (C)V(n)|Q – so this should be easy to validate as well as being essentially impossible to faithfully duplicate the original. This is not a fault; in fact, it is key: this allows us to take a first step toward a distortion. You can use the standardised transcription language from the source language to Japanese. A few hours looking at foreign words transcribed into katakana should give you enough intuition to invent your own.

Step 3.

Here we introduce the secret sauce: ateji. "Ateji" is the term for writing words that are not normally written in kanji (characters) in characters. The assignment from sounds to characters is arbitrary and highly flexible, but it isn't anarchic; in general, you should observe these rules:

  • Use only existing associations between characters and pronunciations; do not invent your own. However, advanced users can insert a pun here like it is one of those "kira-kira names".
  • You can ignore or insert long vowels if it helps you.
  • The distinction between small and full-sized kana may be ignored if it helps you.
  • The distinction between un-, dakuten- or handakuten-marked kana may be ignored if it helps you.
  • You may use a kanji with a reading that normally has an okurigana applied to it without actually writing that okurigana.
  • Prioritise native readings (kun·yomi) over Chinese-derived readings, for reasons which we will see later.
  • You can bring in Ainu and the Okinawan languages into the equation by allowing their direct translations to be used. The classical example is to associate 幌 (< Ainu poro) with 大 and allow their readings to interchange with each other.
  • Above all, observe or ignore these rules all in service of creating maximum distortion across all steps.
  • At this point you should now have a whole bunch of Chinese characters. Which is where we activate the second part of the secret sauce, the hardener to the resin in a two-part epoxy glue:

    Step 4.

    Take our Chinese of choice and then read all the characters in that language.

    That's all you need to do. You should, if you adjusted all your parameters correctly, now have a name that's pronounced greatly differently from the original, with only vague hints to its previous identity, if any remain.

    For even better results, and perhaps to aid pronunciation by your target audience as well as being able to better cover your tracks, transliterate the pronunciation back to the origin language. Most of the languages that you use in Step 3 would be fairly restrictive, so some level of creativity might be needed to disguise the things somewhat. If your goal is obscurity (which if you are doing this technique it probably is) then that's so much the better.

    If your target language is English, as it usually is, then you have an easier time as there is a rather poor fit to Chinese phonotactics and you can use this as an opportunity to slip the pronunciations a little bit further. There is also a number of Romanisation rules you can use to further complicate matters if this is necessary.

    And finally, in some cases, you can take the result and put it back around the circle again to get a second level of obscurity. This does not normally yield as dramatic a change as the first time around but it could help immensely if you make some changes in this step so that the result looks very different from the result from step 3.

    Worked examples

    Let's have a look at some examples of how this turns out. We'll only go once around the circle in order get a feel of the relationship between the start and the end.

    A game

    Step 1. Let's have a look at how the word "Warhammer" gets mangled.

    Step 2. The formal way to transliterate "Warhammer" in Japanese is ウォーハンマー (wōhanmā). However, we will alter this a little bit in the next step in order to make it a bit easier to mould.

    Step 3.

  • We first remove the long vowels, yielding ウォハンマ.
  • Then we make the small o large, making it ウオハンマ.
  • We then remove the gemination on the /m/, giving ウオハマ.
  • At this point, it is easy enough to identify ウオ as "fish" 魚 and ハマ as "shore" 浜, and so we can just write it as 魚浜.

    Step 4. We translate it back into Cantonese to complete the cycle. After making a basic transformation into the traditional Chinese characters, we have 魚濱, which we can transliterate using an informal transliteration scheme as "Yue Pan".

    And we're done. The final score is: "Warhammer" → "Yue Pan". Not bad, isn't it?

    Note that at the last step we could have went with "Yew Bun" as an equally legal transliteration scheme, and we can then snowclone the name in English to form a much more natural-sounding "Treebread"!

    A car

    Step 1. Now we'll play around with the word "Mustang" and see what happens.

    Step 2. That's マスタング in Japanese.

    Step 3. We can use this directly without any change at all:

  • マス identifies with 増す "increase"
  • タン can be written as 舌 (as in 牛タン "gyūtan", "beef tongue", even though it's not written like that in that phrase)
  • and グ is identified with 具 "tool".
  • String the characters together and get 増舌具.

    Step 4. For no reason whatsoever, let's move it over to Korean and get "Jeungseolgu".

    This is much more straightforward other than the punning part, but as you can see the result is nevertheless a near-total annihilation of the original value with very little of it remaining, though if you look closely signs of it still remain. The result however does make it a little bit difficult to have it run through the system a second time.


    If we go back and look at the description of the technique, it relies on a number of specific facts about the source language, Japanese and Chinese:

  • The three languages have very little to do with each other.
  • Japanese and Chinese have a shared history of writing with the same characters, even though they are pronounced entirely differently.
  • Japanese has a limited syllable inventory, but it makes extensive use of that inventory.
  • The Latin script allows for additional friction between translation making more sparks fly, depending on the choice of transliteration.
  • If the source language is English, there is the additional benefit of a vast amount of choice as to handling foreign languages.
  • Japanese in particular has an established and extensive tradition of using its orthographic nightmare to create lots of wordplay opportunities.
  • And yes, it is Japanese that makes the whole thing work properly, because its history allows it to fulfil all these roles.

    Because both writing and speaking are involved, and new scripts nowadays tend not to use the organically-grown, fatally knotted mess that sinograms are, it's not very easy to replicate this pattern anywhere else. However, the formula remains fairly easy to understand, and you can make this setup work in its classical form if you can find languages that fit these criteria:

    Obviously, you can't control the source or target languages, so we can't talk much about that.

    The other two languages should have:

  • A shared writing system, enough so that you can easily transfer words and pronunciations between one and the other
  • Be mostly unrelated
  • Have different phonological behaviours
  • Be able to accept almost all the possible syllables (or equivalent) in its configuration
  • And with regard to the source and target language, it should:

  • Be mostly unrelated
  • Have a highly restrictive phonology, so to enhance the distortion effort
  • Readily accept loanwords
  • Be able to "make native" that loanword, if needed
  • Obviously, I'm planning to make something of the sort in my own conworlds, and I think I have enough languages to make it work multiple ways, so there's going to be that in mind. With the freedom of a constructed world, we can add on any number of techniques to make it even more obnoxious, including punning based on the shared script, a fourth language and more.

    Finally, it should be noted that it's not necessary to follow the strict four-step programme for every word. For additional chaos, you can always alter it the order the languages are being used, add more languages into the mix, have it go back and forth between languages, or even introduce single-language puns, though of course you'll need extra expertise (~fluency) to actually handle all these new changes. Ultimately, it's still artistry. You can do a lot of things with artistry, and the rules are only there to reach your goal. At least, since the goal is to introduce chaos and be obscure in a clever way, there isn't too much of a risk to having it be too much!