Skip to content Skip to footer

The Generation Poem: Reading the Verse That Places You in a Lineage

A thread-bound Wuyi lineage genealogy open at two facing pages, the printed centre-fold strip visible down the gutter

A lineage book is mostly a list. Names in columns, sons under fathers, a generation number in the margin. One page in the book is not a list.

It is a verse. Twenty to thirty-odd characters, four or five to a line, printed on a leaf of its own — usually after the house rules and before the descent register, though the order of leaves and the length vary between editions and branches. One character belongs to one generation, in order. The verse is the generation poem (班派, bānpài); the leaf is the generation verse page (派语, pàiyǔ).

The chart cannot hold it: the 五世一图 convention that the guide to reading a Chinese genealogy sets out restarts every five generations, so no chart carries twenty characters, let alone thirty. The poem is therefore the index to the register printed behind it, and the centre-fold strip (版心, bǎnxīn) down its gutter — title, volume, branch, page — tells you whose index you hold.

What the index does is seat a person. A boy born into the branch (房, fáng) was given the generation character (字辈, zìbèi) inside his two-character genealogical name (谱名, pǔmíng). For life it told every literate man in the county where he stood: which row at the spring and autumn sacrifices, whose seniority he answered to, whether his name was called for sacrificial meat. The poem looks like scripture and reads like a seating plan.

The page has neighbours, and they are half the evidence: the house rules (凡例, fánlì) that declare which slot the character occupies, the anchor number fixing where the sequence begins, and the strip that names the branch. The poem places but does not date: it will tell you a man was the twenty-third of his line, not the year he was born.

The promise is narrow: which name carries the character, which leaf declares the rule, how an anchor turns verse into a person, and five moves — four of them indoors. Two refusals, stated early. There is no “Taishan Chen generation poem” to look up, because poems belong to branches and households, not to surnames; and the poem is not a birth certificate.

One man, four names and one slot

A Taishan man carried a milk name (乳名, rǔmíng) from infancy, a school name (书名, shūmíng) from the schoolmaster, and a genealogical name from the register. If he emigrated he acquired a fourth, the overseas name (侨名, qiáomíng) — the name on a remittance envelope, not the register’s (what those letters are). One man, four words; a household entry, a naturalisation file and a descent row may each quote a different one.

Only the genealogical name is slotted. It takes the character, and through it a generation ordinal (世次, shìcì) — the number the book is arranged by. Place a man from his milk name, or from the courtesy name on a gravestone, and you are reading a field that was never indexed.

Then the slot. The house rules say whether the character sits first or second; a few branches alternate by generation. If the rules are silent — in reprints they often are, or the clause survives only in an older layer — the rows will say it: read three consecutive generations and watch which position moves.

A framed 世系表 descent-table entry in a woodblock-printed Chinese genealogy, one man's row and its fields
A 世系表 entry in the 璜皌余氏宗譜, from the Shanghai Library and National Library of China digitisation: parentage, birth years, the wife’s natal surname, three sons, a 幼殤 note, burial. The book is a 余氏 register from outside Wuyi; the printed strip at the leaf edge carries the section title and the hall name.

So the first act is not reading the verse but writing down which field the document quotes, and whether the character sits first or second.

One village, three poems

What is the Chen generation poem? At the level of the surname the question has no answer, which is exactly why a poem can place a reader precisely.

The proof is smaller than a county. In Lingchong — 250 households, about 1,150 people, one ancestor — the county government’s surname entry records three sequences in use: the Yan Shi branch’s, the Bo Liang branch’s, and that of the twenty households which arrived last. The third is not a corruption of the first two. It belongs to a group that came later and kept its own.

Poems travel with registration groups, not bloodlines, so a shared poem does not prove shared descent. Two readings are live: one ancestor, with the poem recording it; or a weaker branch absorbed into a stronger one through lineage merger (联宗, liánzōng). The page cannot settle it, and several poems in one village should be expected rather than suspected.

The Chens divide by ancestor, as the page states: Shuinan and Chongpan from Mo Gong, Chongjin from Ying Gong, Wencun from Kai Gong, the six villages of Duntou from Xuan Gong. Shuinan opens 圣文万世昌, Chongpan 籍凤毓家秀, Chongjin 先德开基, Wencun 贻谋燕翼.

Two mechanisms produce that. A branch that split early and took its own hall and chartered estate composed its own verse — Shuinan, where Wenqi moved out of Chongpan. And a household (户, hù) carved out of a crowded registration block (图甲, tújiǎ) could take a new name and a new poem.

A village hall on a stone platform in Shuinan, Taishan, firecracker paper on the road in front
水南, in Taishan, roughly six kilometres west of the county town: a hall on its stone platform, firecracker paper still on the road, a wooded hill behind. The village’s 陳 branch keeps two poem layers of its own.

The text is plural too. Between the government page and a clan transcription of the same poems, five characters differ: 家 and 嘉, 铭 and 名, 起 and 喜, 尊 and 俊, 绩 and 博. The first three are homophones in Cantonese, the kind of drift a reciter produces; the last two are sense-substitutions, pointing at a compositor. The same collapse runs up a level: 李氏 and 黎氏 fuse in Siyi-dialect transcription, as the Li (黎) surname article sets out. This article quotes the government page and says so.

The surname articles call the poem a portable identity card, and it is — the Tan and Lee pieces argue it with poems of their own. The refinement is that the card is portable but slotted and anchored, and portability without the anchor is what produces wrong placements.

Which leaves the search string: surname plus village or branch, not surname alone. A descendant who knows his branch’s poem can place his generation from the character alone, because the unit is the household and the branch.

The number that anchors the verse

Sequences are extended, never invented whole. An extension is written as an instruction: from such-and-such a generation onward, the sequence is revised (自第某世起, zì dì mǒu shì qǐ). Scholarship states that as the norm; no Wuyi source retrieved here prints it verbatim.

The more common device is an instruction to add. The Taicheng Tan page prints a thirty-character poem and says how to use it: take Tianlin as the first generation, add your own position in the poem, and you have your generation. Er is the first character and Er Tian stands at the twenty-first generation, so the offset is twenty. Si the fifth: Si Bing and Si Huan at the twenty-fifth. Jie the sixth: Quan Jie and Jie Ji at the twenty-sixth. Kai the seventh: Kai Ta and Kai Zhen at the twenty-seventh. Wen the eighth: the page states that the twenty-eighth is the Wen generation.

Five men, five alignments, one offset. Character position plus twenty equals generation. The verse is a list; the anchor is what makes it a ruler.

The third device is a table: a Xinhui Lingcun Chen transcription keys every character to a generation column, and in the copy consulted here the headers run three generations ahead of the page’s prose. Only the two-line sum finds such a slip — one identified error in one transcription, not a verdict on clan transcriptions — and a reader’s photocopy may carry it.

One Wuyi case has a complete record, and it is a Li story. The Zida Gong branch (子达公房) of Taishan’s 李氏 had no sequence for twelve generations after its first ancestor. The lineage’s account places the commission at the thirteenth generation, in the Kangxi gengzi year (1720), when Xingru Gong — a graduate of that year — asked Chen Yun, then Guangdong’s education commissioner and a Hanlin reviser (翰林院检讨, hànlínyuán jiǎntǎo), for an upper line to serve as the branch’s sequence. Three limits travel with it: this is a 李氏 story, never transferable to the Chens; it is the lineage’s own account, 族谱称 rather than a preface; and the year is printed because the county page documents it, not because it is the origin of Wuyi poems generally.

The Chen layers carry no such sentence.

Why the poem is younger than the pedigree

The commonly cited institutional floor is 1536, when Xia Yan’s memorial let commoner lineages build a family shrine and sacrifice to a founding ancestor. A fixed sequence presupposes what the reform first made legitimate: a traceable founder and a line continuous enough to number.

The county is younger than the poem claims. Xinning was carved out of Xinhui in 1499 — the county gazetteer’s 沿革表 puts the act in the twelfth year of Hongzhi — and was renamed Taishan only in 1914; Kaiping did not exist until 1649. A preface with a Song ancestor “moving into Xinning” commits an anachronism in its first line.

Attribution is a status signal, and the least datable sentence in a book. Feng Erkang names a Ming genre, the scrambling of famous men into one’s ancestry; Segawa Masahisa documents the prestigious preface as a routine later insertion, Chen Baisha among his examples. The mechanism matters more than the names: a prestigious founder needs unbroken descent down to the branch, so the sequence is back-dated to fit the pedigree, its starting generation a product of the register rather than evidence for it. The genre is documented; individual poems may be retrospective.

Shuinan’s two layers are the worked case. The earlier is twenty characters and opens on 圣文万世昌, quoted here as a first line and a count; the later is given whole:

仁厚承恩泽,书香启俊良;
邦家常有耀,绩学自名扬。

The government page prints 绩学 in the last line; a clan transcription prints 博学. The lineage credits the earlier sequence to Chen Baisha (陈献章, 1428–1500), and that is all that can be said of its date: no preface, colophon, fee or paying branch has been located. The government page states the attribution as authorship — the verb 撰, “composed” — an assertion stronger than the evidence carries, and itself a datum about the lineage’s strategy rather than a fact about the 1440s. The later quatrain the lineage credits to Chen Botao, a Dongguan man who took third place in the 1892 palace examination; the sentence names an act, an honorific, an origin and a product, and no year of commission, no fee, no compilation-office scene.

One piece of arithmetic makes retrospection unavoidable. Mo Gong died in 1242; Chen Xianzhang was born in 1428, a hundred and eighty-six years later. Such a man could not have named the founder’s sons. So either the poem begins later, or the attribution is retrospective — probable rather than certain, because it rides on the founder’s dates as the government page gives them.

Everything above dates a composition, and measures how much shorter a poem’s antiquity is than the descent it claims. The poem gives order and never dates.

Four numerals, three tests

The inflation trap has names. In the Chongpan Chen register the first generation is four brothers: Wen Shen, styled Feng Shiyi Lang; Wen Zhen, Feng Shier Lang; Wen Long, Feng Shisan Lang; Wen Qi, Feng Shisi Lang. Four men, four numerals, one generation. These are birth-order names (郎名, lángmíng) and generation ranks (行辈, hǎngbèi) — serials within a generation, not generation characters. Read Feng Shisan Lang as a generation character and a first-generation man becomes a thirteenth-generation one. The numerals occupy the first generations; generation characters begin only after the branch’s move.

Three tests follow, all arithmetic.

The first is the band. Twenty-five to thirty years per generation is a working figure, and two Wuyi clans’ own numbers land inside it. One Xinhui branch states eight hundred years and thirty generations — 26.7 years each. One Taishan 李氏 page counts living members between the twenty-fifth and thirty-second generations — at the top edge, nearer thirty-three years each, worth saying rather than smoothing. These are depth counts, not start claims: the band describes depth as the lineage counts it, and what it dates is a compilation. Qian Hang holds that a sequence gives order and never dates; the arithmetic school dates compilations with the same numbers. This article uses the arithmetic only on compilations.

The second is length. One character maps to one generation in principle, and a poem covers as many generations as it has characters. A Lingcun Chen sequence of thirty-two characters spans thirty-two recorded generations; Shuinan’s twenty ran out and a second twenty were composed. So a poem shorter than the pedigree is not a scandal: it is proof that an extension happened, and an extension is dated, authorised and paid. Verify the anchor before applying the count.

The third is density. Entries thicken from roughly the fifteenth generation, where contemporary record-keeping begins. The cleanest datum is a branch’s own sentence: twelve generations after the first ancestor with no sequence, the first poem fixed at the thirteenth. That is the one branch-start figure this article prints, and it is lineage-derived.

None of these tests dates a birth. They date compilations and expose impossible pedigrees.

What the words were doing, and why they rhyme

Rhyme and parallelism let a member who could not read hold twenty generations in his mouth. The form is a mnemonic, not a literary achievement.

The diction is where a page earns its keep. The earlier Shuinan layer asks for literary standing, an auspicious realm, virtue outrunning the forebears, law-abiding endurance: 文, 世昌, 瑞国祥, 德业, 遵法. The 1892 layer asks for something else — 仁厚 producing 恩泽, 书香 producing 俊良, 邦家 producing 耀, 绩学 producing 名扬. The nouns move from the state and the law to the household and the book, and the shift dates the two layers better than a colophon would. A poem can carry a route as well: the Taishan 李氏 sequence turns the Zhuji Lane migration charter into a couplet, 迺祖南来,长发其祥, and names Zeng Shen as the family’s model in the line before.

The queue, not the plaque

The poem was not a monument. Ancestral tablets are ordered by generation and by shrine alternation (昭穆, zhāomù), not by the characters in anyone’s name. The hall’s memory is finite: after five generations a tablet moves back, and when the mourning bond runs out it is retired. The register drops no generation; the hall drops most of them. Shrine position and generation seniority are two orders, and they do not coincide.

What the poem allocated was bodily and distributive. Which row a man stood in at the spring and autumn sacrifices. Whether he took a first share at the winter distribution of sacrificial meat (分胙, fēnzuò). The arithmetic became a queue.

Its edge shows in who is missing. Boat-dwelling households on the Tan River estuary and the coast stood outside the registration system: no register, no sequence, and the blank where a poem should be is an exclusion record rather than a lost leaf. Military households at Guanghai, registered to the guard garrison whose walled town went up in 1394, held land, a hall and a poem without appearing in the registration rolls — the clearest evidence that the poem’s operative boundary follows registration, which may or may not coincide with a single village. Women enter as a surname plus “née”, with no generation character. Live-in sons-in-law and adopted sons of other surnames were barred by the house rules, and formal adoption entries existed to smooth the gap. Scholarship identifies these categories; no Wuyi house-rules leaf was retrieved to show how any one branch wrote the rule down.

One sentence: the generation character proves a man was registered as a branch’s descendant, not that he biologically was one. That is not the same as saying the poems are fiction. A register that seats some men and leaves others off is a real register, doing real work, for the people inside it.

The object in the drawer

A genealogy prints once a generation, and revision destroyed the old copies, so the poem a family produces today is often a 1990s typeset artefact: remittance-funded, reset in simplified characters.

Variants cluster by branch for a reason. Each branch held one numbered set of the printed edition, so a copying error propagates down that branch’s copies and nowhere else. In 1901, at the Taiqiu Academy in Xinhui county town, local gentry settled an extension to a Chen sequence; the Lingcun branch went home and changed one character, 后 to 广. One graph, one branch, permanently.

Two other forces work on the characters. Taboo avoidance forces substitutions when a character collides with an ancestor’s name, and stroke omission is its polite form. And the house rules often pre-draft twenty or thirty future characters, so the tail of a long poem was never used by anyone. A clan transcription still online in the 2020s ends its table with a promise to decide the rest later: citing that tail as evidence of antiquity inverts the document.

Emigrant periodicals ran revision notices, births, deaths, generation characters and donor lists to overseas readers; the oldest of them was founded in 1909, and the longer emigration story is told elsewhere. From the 1980s, families are described as asking village kin to supply a character for a grandchild born abroad — attributed practice, no letter or dated instance retrieved, so no family is put in the sentence.

One outward glance, because the poem was administrative and therefore testable: a receiving-port inspector could check a man’s father, brothers and sons against the family’s characters, so a forged pedigree had to forge a poem too. The paper-son files carry that history; this article does not.

And the ranking, stated once so a reader can weigh his own family’s version: a printed genealogy outranks a preface, a preface the house rules, the house rules a photograph of a page, a photograph a clan transcription, a transcription a recitation, a recitation a website list.

Five moves, four of them indoors

  1. The newest preface. Read for the revision year — often the only date if the colophon is missing — and for two names, the sponsor-editor (主修, zhǔxiū) and the actual compiler (纂修, zuǎnxiū). The first is often the donor, the second the writer. Look for any sentence asking a named outsider to compose a sequence, and for its date. Write down: the year, and both names.

  2. The house rules. Read for the register’s cut-off, the slot rule, taboo substitutions, and exclusions for adopted-out sons and sons-in-law. Write down: the slot rule, verbatim.

  3. The verse page. Copy the characters, the strip and the anchor. Photograph the leaf with its centre fold and, if present, the facing rules page: the strip carries title, volume, branch and page, and it is the only per-page branch identifier. Then run the two-line sum. Write down: first character, character count, anchor number.

  4. The descent table. Find a known grandfather’s genealogical name, take his generation ordinal, and note his taboo name and courtesy name, his wife’s natal surname and his sons. Walk up to the branch founder and across to the first migrant: the Shuinan chain runs Wenqi, Mo Gong (d. 1242), Chen Hui, Chen Wen. If the village name is all you have, the ancestral-village guide is the next step, and for a neighbouring county’s line the Heshan clan genealogy guide lists its surname groups and where the registers are held. One trap: the chart carries names only, the table names plus five fields, so counting boxes in the chart counts graphics. Write down: the chain as one line.

  5. A datable stone or a contract. This works only if the inscription records the taboo name rather than a courtesy name, and only if birth and death are given. Qian Hang holds that a generation character cannot be recovered from a stone; David Faure treats the gravestone as the lineage’s second archive. Both are right, conditionally. No Wuyi stone meeting the condition was retrieved for this article, so the method is stated without an example. Write down: the taboo name and both dates.

Only the fifth move requires leaving the house. An elder’s recitation is the cheapest check on the third, and the channel through which homophones corrupt a sequence: the poem gave you order, and only the stone gives you a date.

When there is no poem, and what the character is for

Some will do all five moves and find nothing. The line may descend from boat households or military households never in the rolls. The book may have been burned in the 1960s or 1970s, or scattered generations earlier when the coast was cleared. The surname may have changed at emigration and never changed back. Or the “poem” may exist only as a cousin’s phone transcription, two characters wrong, with no way to tell which two.

What survives is not the page but the practice: characters still standing in living names, taboo names cut into stones, whatever an elder can recite. A sequence cannot be rebuilt from two names, and this article will not pretend otherwise.

The character in a living name is a seat number inherited from people who wanted their descendants countable — and countability is a form of care. Not nobility, not continuity, not five thousand years: a man who wanted his grandsons distinguishable, and placeable, bought a verse and had it printed.

This month, ask the eldest living person one question: what was your genealogical name, and who gave it to you. At the next Qingming, photograph one page — the verse leaf with its centre fold, or, if there is no book, the taboo name and the two dates on the stone.

A position is not a proof. A position is still something a person can stand in.

Continue Reading

Stay Connected

Follow RootsOfChina on Facebook and Instagram for new article updates, behind-the-scenes from Kaiping, and stories from the Chinese diaspora.

Sources

  • 台山市人民政府, 台山姓氏 · 陈氏 (2025) — 水南 and 冲泮 poems, both attributions, the 郎名 run — cnts.gov.cn
  • 台山市人民政府, 台山姓氏 · 李氏 (2017) — the 子达公房 commission, 康熙庚子, 陈筠 — cnts.gov.cn
  • 台山市人民政府, 台山姓氏 · 谭氏 (2017) — 凌冲’s three sequences, the 台城 counting rule — cnts.gov.cn
  • 錢杭, 《中国宗族史研究入门》 (复旦大学出版社, 2009) — 谱牒 and 宗族 structure; a sequence gives order, never dates — iahs.fudan.edu.cn
  • 何福海, 《新寧縣志》, 光緒十七年 (1891), 卷二 沿革表 — the 弘治十二年 (1499) carving of 新寧 from 新會 — ctext.org
  • 陳氏宗親網, 凌村支譜 — the 32-character table, its header slip, the 1901 太丘書院 续派, 后→广, the 博學 variant — chens.org.cn
  • 湖北省图书馆, 寻霖 on the components of a 族谱 and the scope of 凡例 — library.hb.cn

Suggested Reading

Is this your family’s story?

Roots Of China is built from readers’ family histories. If a story here sounds like your own — a surname, a village, a journey — leave a comment below, or share your family’s story with us and we may be able to help you find your ancestral village.

Share your story →

Leave a comment

0/100