For many years one of the best games in town as far as language maps went was the Language Atlas of China (中国语言地图集) published by the Chinese Academy of Social Science (中国社会科学院). The original 1987 edition was not without its problems, and there was a fair amount that was just left unclear due to insufficient data at the time (the same sort of lack of data that poked a few then-unknown holes in Jerry Norman's 1988 Chinese
). But, generally, it was a good atlas representing the combined efforts of a great many scholars.
For the last couple years we've been talking about picking up the second edition, published in 2012. You can find an article from 2008 floating around online that shows the new classification system of the second edition. From this article I already knew that I wasn't the biggest fan of the new system, but at least in some areas there was huge improvement. It was enough to give me hope for the second edition. Last week I finally picked up a copy. I hate to say it, but I'm really quite disappointed.
The way the Academy gathers resources — as told to me from their own mouths on a couple visits to their offices this past year — is by essentially outsourcing the data. This makes sense, since there's so much to be addressed. The Academy contacts various researchers throughout China and Taiwan and requests their data to go into the Atlas. This sounds like it would get the best data from the people who know the areas the best. But in talking to a number of regional scholars, I'm by far not the only one upset with the latest version.
The Wú section
I don't know where to start with the Wú map. I see so much wrong with it that I think it needs a separate post. Unfortunately Hangzhou is still listed as Wú, but that's not terribly surprising. Changzhou is now grouped with Wuxi — despite some superficial similarities — as well as with Suzhou. That's a bit odd. So I'm actually going to leave Wú out of this post for now beyond what I've already said, and then more fully address it at a later time.
Taiwan & the Hakka section
The Taiwan map is pretty atrocious. Look at a city like Hsinchu 新竹. You have Hoiliuk 海陆 dialect/accent spoken in the city and Siyen 四县 dialect/accent spoken to both the north and the south. If you ask any Hakka speaker in the area, scholar or layperson, they will tell you that Hsinchu is clearly a Hoiliuk city.
In the Language Atlas, Hsinchu isn't shown as Hakka speaking at all. In fact there are 5 distinct Hakka dialects spoken in Taiwan, but according to the Atlas we're only four, with Rhaoping 饶平 dialect absent. Speaking of Rhaoping, in the mainland city after which the dialect is named, there's also not any Hakka being represented in the Atlas.
Meanwhile Hoiluk is over-represented outside of Hsinchu as well, but completely absent from Hualien County where it's one of the dominant dialects. In fact no Hakka is shown in Hualien at all.
Just so that it's not all negative, I will say there's been some good improvement in Manchuria. In the 1987 edition, there were huge gaps where it showed no Mandarin speakers at all, despite there being speakers with their own non-Putonghua dialects. This has been remedied, with towns like New Barag Right Banner (新巴尔虎右旗) now properly represented as speaking a 黑松片 dialect. Big improvement there.
Sampling methodology issues
This is a big issue for me in how a lot of dialect work is done in China. It's something we try very
hard to avoid with Phonemica. The idea is that you go in and ask around and find the most representative speaker of the dialect in the town. You then talk to him for hours and hours and hours and take him as representative of the whole town. This ignores peripheral dialects. This doesn't show how large the area is of that one person's dialect. And it also obviously ignores ongoing changes in the dialect or small variations. I'm pretty sure that's what's going on with this second edition in most cases.
On the one hand, dialects are now drawn following county lines rather than random smooth lines throughout the blank map space. Cool, except dialects don't always follow county lines. But if you're only talking to one dude in the middle of the administrative division, then that's what you're going to see. You say "Zhang Dou is representative of Minhang" so all of Minhang gets marked as that dialect, right up to the political boundaries of Minhang. That doesn't work when you have Southern Minhang sounding very different from Northern Minhang (which it does). Grab a guy from Wujing and you're going to get pieces of both North and South dialects, ultimately giving a clean representation of neither.
It's obvious why it's done this way. It is already time consuming and expensive to do the fieldwork. But now that the work is being done (for example 汪平's multiple speakers from Suzhou for 江苏方言研究丛书 or 徐越's mutlple towns in North Zhejiang, and now that the data is out there
, it'd be great to see it being used for such major project as this.
In biology you have what's called the type species. It's the species that best represents the defining qualities of the group. This happens in dialectology as well. However I think in some cases people are just a little too focused on finding the type species, that they miss all of the other incredibly valuable information on the periphery. This is a bigger issue when less data is used or available. So for that I can only hope that the third edition, if there ever is one, will be a significant improvement over the second.
It would also be quite nice to see them include more than just the PRC's political claims, since that reflects neither jurisdiction nor linguistic distribution.