Eat Street Food China is an English-language project about Chinese regional foods, everyday snack traditions, culinary terms and histories. This page sets out how the research is done, what kinds of evidence the articles rest on, and what we do when the evidence runs out.
Chinese food history is rarely a single, well-documented story. Recipes change, names travel, local traditions overlap, and origin stories are often repeated far more confidently than the record supports. Our working assumption is that a dish’s past is usually less certain than the internet makes it sound, and that saying so is more useful than filling the gap.
What We Research
Our articles focus on questions such as:
- What does the dish’s Chinese name mean, character by character?
- Which city, province or regional tradition is it associated with?
- What ingredients and techniques define it?
- How does it differ from similarly named foods?
- Which parts of its history can be traced to a text we can read?
- Which stories are local tradition, legend, or a later commercial claim?
- How is the dish commonly served and understood today?
The Texts We Work From
We work from Chinese-language sources, in the printed editions we actually have in front of us. The material falls into three groups.
Early and pre-modern texts. Among those quoted on this site: the Shiming 释名 (Eastern Han), the Qimin yaoshu 齐民要术 (sixth century), the Shishuo xinyu 世说新语, the Nan Qi shu 南齐书 and Nan shi 南史, the Taiping yulan 太平御览 (which preserves Shu Xi’s 束皙 Bing fu 饼赋), the Zizhi tongjian 资治通鉴, the Song city memoirs Dongjing meng Hua lu 东京梦华录, Meng liang lu 梦梁录 and Wulin jiushi 武林旧事, the Qingyi lu 清异录, and the Yuan court manual Yinshan zhengyao 饮膳正要.
Late-imperial and Republican surveys of city life. The Dijing suishi jisheng 帝京岁时纪胜 (1758), the Yanjing suishi ji 燕京岁时记 (1906), the Dumen jilue 都门纪略 (1864), the Chengdu tonglan 成都通览 (1909), the Jiading xian xuzhi 嘉定县续志 (1930), the Ganquan xian xuzhi 甘泉县续志, and the Dazhonghua Jingzhao dilizhi 大中华京兆地理志 (1919).
Twentieth-century recollections. Essays and memoirs by writers who grew up in the cities they describe: Tang Lusun 唐鲁孙, Liang Shiqiu 梁实秋, Chen Hongnian 陈鸿年, Wang Dunhuang 王敦煌, Deng Yunxiang 邓云乡.
Every article lists the specific sources it uses, with chapter or section, at the foot of the page. We also keep a single dated list of the earliest record we have been able to check for each dish, with the original line in each case.
How a Quotation Is Checked
This is what a Chinese passage normally goes through before it appears here:
- We look for it in a printed edition we hold, rather than in a later book’s quotation of it. A passage cited by a later book is a lead to the original. Where that is as far as we have got, we say what we are relying on and describe what it records, instead of presenting it as the original wording.
- We transcribe it exactly as that edition prints it, including its characters. We do not convert between simplified and traditional forms to make a page look consistent, and we do not tidy punctuation.
- We name the edition. Where a text survives only in a modern collated or annotated edition, we say so. A modern Chinese rendering of a classical passage is a translation, not the text, and we describe what it records rather than putting it in quotation marks as if it were the original wording.
- Where our copy is a scan, we read the passage against the page image rather than relying on machine-read text, which we treat as a lead.
“We Have Not Found It”
When the record stops, we say that it stops. A sentence such as “the earliest text we have found that names this dish was published in 1919” is a statement about our search, not about the dish. It leaves room for an earlier text to exist, and for a reader to find it.
We also apply this to claims made elsewhere. Where a widely repeated origin story has no source we can check, we say we have not found the record it rests on. That is not the same as saying it is false.
Records and Recollections
These two kinds of evidence are not interchangeable, and the articles keep them apart.
A record is a text written close to what it describes: a gazetteer, a city survey, a manual. A recollection is written decades later, from memory, usually after the writer has left the place. The twentieth-century writers above are recollections, and they are marked as such wherever they are used. They are excellent on how a dish was eaten and argued about in one household; they are not evidence for when it began.
Documented History, Tradition, and Legend
Food stories often combine several kinds of evidence, and we try to label them:
Documented information is supported by a text or other identifiable evidence we can point at.
Attributed tradition is a story associated with a community, restaurant, family or region, but not established as fact.
Legend is presented as a story people tell, not as history.
Editorial interpretation connects established information to explain a pattern. Where an explanation is our inference rather than something a source states, the wording says so.
We would rather leave a famous story unresolved than turn it into a false certainty. A story can matter culturally even when its history cannot be confirmed.
Names, Translation, and Chinese Terms
Chinese food names do not always have a clean English equivalent. We normally give the common romanized name, the characters, and what the name literally says, because the name is often the best evidence about the dish. We distinguish between spelling variants, dialect forms, menu translations and genuinely different dishes.
Regional Differences Matter
China does not have one uniform street-food tradition. We avoid presenting one preparation as the only authentic version where households and regions plainly differ, and we try to say whose version we are describing. Migration and contact between regions are discussed where the evidence supports it.
First-Person Passages
A few articles include a short passage in the first person: something the editor ate, somewhere, at a particular time. These are marked as personal recollection, they are the editor’s own, and they are there because a reader is entitled to know when a description comes from experience rather than from a text. We do not invent them, and we do not use them as evidence for anything historical.
What We Do Not Invent
We do not claim restaurant visits, interviews, conversations, prices or fieldwork that did not happen. We do not present a modern retelling as an old text, attribute a quotation to a book we have not read, or convert a plausible guess into a date.
How This Site Is Made
The site is written and edited by one person, working from Chinese-language printed editions. Searching, draft translation and draft prose are done with AI tools. The editor reads the sources, decides what goes in, and stands behind the editorial choices on these pages.
What we do with Chinese quotations is set out above: we quote from the edition we have, transcribe it as that edition prints it, and name the edition so a reader can go and look. Where we are working from a scan, or from a later book’s citation rather than the original, the article says so.
We say all of this because the site is built on a reader being able to check it, and describing how the work is done is part of that.
Corrections
If something here is wrong, we would like to fix it, and we would particularly like to hear about an earlier source than the one we have cited. Corrections are made on the page itself.
For Researchers and Journalists
If you are writing about one of these dishes and want to verify a passage, we can supply the edition details and, where we hold it, an image of the page. Write to us through the contact page.