Chinese characters written with water

Chinese characters written on the pavement with brush and water. I have seen this a few times in Beijing, once in the Houhai park (see picture) and once on Wangfujing opposite of the Kaogu bookstore, near the Fine Arts Museum. The scenario is this: an old man dips a huge brush into water and writes large Chinese characters on the ground, which then slowly fade away. So the first characters are gone before he has finished writing the last ones.

Now where does this art of hydro-calligraphy come from? Or what is the significance of it? To be honest, it does grab you as you see beautifully written characters fade away under the rays of the sun. Something about temporality. And you can never tell when they are really gone, in one moment you still seem to see them but then maybe you don't anymore. There must be a second in time when they are actually gone but you still register them as being there.

Kanji variants in Japanese temple names

The post Kanji used to ID Honnōji at No-sword describes how a specific kanji variant used on a tile was used to tracing it to the Honnoji temple in Kyoto. Apparently, the same variant has been used exclusively in the name of Honnoji so the excavation of the same form showed the connection of the object with the temple.


Here is the name of the Honnoji temple with the variant for the character 能. You can see that the right side of the character is different from the modern form.


Here is the variant in question on a roof tile.

Actually, this variant was quite common in medieval China, it appears on Dunhuang manuscripts more often than the form used today as the standard. I assume it was the same in Japan during the Nara period and possibly even after that.


The kanji variant in a Chinese manuscript of the Mahaparanirvana sutra from Dunhuang, dated 506. This manuscript is currently in the collection of the British Library, shelfmark Or.8210/S.81.


The character 能 was one of the most varied characters in China. These examples are from the dictionary called Guang beibiezi 廣碑別字 by Qin Gong and Liu Daxin, showing variants on stone inscriptions from the Han through the Tang dynasties.

Anyway, I think it is a fascinating phenomenon that a variant would continued to be used in a trademark fashion for a particular institution. As another example, I noticed that the name of the Honkokuji temple 本圀寺 is also using a special kanji variant, instead of the standard way which would be 本国寺. The character 圀 is an Empress Wu character form, called in Japan sokuten moji 則天文字. These were used during the fifteen-year reign of Empress Wu Zetian (690-715) and have been used by scholars for dating manuscripts and inscriptions. If an Empress Wu character occurs in a manuscript, it is usually automatically assumed that it dates to 690-715. This is why it is rather surprising to see it in Japan as a standard way of writing a name of a temple. But it also occurs in Japanese surnames.

There are probably plenty of similar examples in Japan but these are the only ones I have noticed. If you know of others, I would love to hear about them.


The name of Honkokuji temple with the Empress Wu character variant. The character 圀 is actually a compound structure comprising the characters 四方 or 八方, depending on how you look at it. They say that the initial form consisted of 四方 but because of the inauspicious sound of 四, the Empress had it changed to 八方.

I am always amused by Chinese signs that mix traditional and simplified characters together. Obviously, in modern China they should never be mixed, writing should be standardized. There is even a committee to watch out for character irregularities in public spaces. Nevertheless, there are plenty of such examples, I am listing a few of them below.


This is a street sign in Dali, Yunnan province, advertising a family business which is a hotel, a barbershop and a bath. There are number of interesting things on this board, including the characters 館 and 照 both of which are today non-standard character variants. Most of the characters are traditional but 間 and 準 are simplified. In fact, 間 appears in the same row twice, once as fantizi and once as jiantizi.


This is a sign advertising mobile phone cards. You can see that the characters 動, 眾 and 終 are simplified, whereas 國, 無 and 費.


A good luck sign, written in gold on a red background. 進 is simplified but 寶 and 財 are not.


This is a warning that you should use the safety rail on the cable car.


This is just a good wish board.



A store sign for alcohol and tobacco. 煙 and 昇 are traditional, 東 is simplified.


Watch out for hygiene!


A hotel sign mixing fantizi and jiantizi.


A no parking sign.


This is the sign of the celebrated Donglai Shun restaurant in Beijing, famous for its dumplings. I am a little ambivalent about including it here because the simplified characters could simply be considered a calligraphic feature. What do you think?


This is a parking lot sign. Pay attention also to the variant forms of the simplified 點, consisting of a 占 and 大.

Chinese character structure: Tracing it back

Let's look at the structure of variant Chinese characters for a second. (This post is triggered by Amida's comment to the blog entry below but I thought it deserves to be in a main entry.)

Does the structure of a particular character form always have to go back to a standard or a historical form? Or can you look at the structure of a form in its own right, without automatically tracing it back to earlier or more standard forms? I have always felt suspicious about the practice of analyzing Tang character forms by digging up earlier and earlier forms of the same character and eventually tracing it back to jiaguwen. It is true that even the Tang character forms originate from the oracle-bone script, I am not trying to deny that. At the same time the relevance of such a genealogy is highly debatable.

Would it make sense to look at running hand character forms in their own right? Instead of saying, "Ah, this is the abbreviation of this" or, "This should actually be written like this," just look at the way it is actually written. Take, for example, the character 盈 in the Santi Qianziwen page on the picture below. The caoshu form in the left-most column is clearly not of the same structure as its kaishu form. You cannot say that it has the 皿 component at the bottom, even though this particular character has that component in other styles, such as the kaishu and xingshu styles on its right. On the other hand, the components in this caoshu forms could also be analyzed and categorized. The little thing that replaces the component 皿 probably occurs in other caoshu forms, replacing the same or even other elements.

Now there might be some cases when you do not have a kaishu form to rely on, simply because it has not survived. Imagine a scenario where you only have caoshu characters and you have to make sense of those alone. You would not be able to take each character form and say, well, this actually should be written like this. Instead, you would have to work with what you actually see.

The same goes for tracing Han or Tang character forms back to oracle-bone inscriptions. Yes, there is a lineage but you should be able to look at the evidence at hand in its own light, shouldn't you? Plus then the oracle-bone inscriptions should be traced back to something as well, only we do not have anything to trace it back to.

I am really curious about what others have to say about this.


This is the beginning of the Santi Qianziwen, or the Tri-style Thousand Character Classic, which is essentially the same text written in three different calligraphic styles: kai (standard), xing (running hand) and cao (drafting).