Reading Notes: The Canaan Hymns — 2,246 Songs by Someone the System Never Certified
July 31, 2026
Original: https://ygpgsgl.org/posts/2026-07-31-canaan-hymns-notes
Published: July 31, 2026
Reuse policy: https://ygpgsgl.org/about#reuse-ai
What Triggered This
A question that turned out to be unanswerable: why is Protestant density highest in Henan, and specifically around Nanyang?
It is unanswerable in the ordinary sense — the county-level distribution is not public, and the field considers it open. Huang Jianbo names it the problem of spatiality: why did the growth happen in the countryside rather than in the cities, where the missionary presence had been heaviest? Where the missions concentrated is precisely not where the growth occurred.
Three days of reading did not produce an answer to that. What it produced was a pile of checked and unchecked claims, one corpus of primary material nobody seems to have processed, and a fairly consistent experience of finding that things in the secondary literature do not close when you add them up.
Three days is worth stating rather than blurring, because it bounds everything below. Nothing here rests on sustained familiarity with the field. It rests on reading a specific set of documents attentively and doing arithmetic on them, which is a narrow competence — it catches tables that do not add up and claims with no comparison group, and it does not catch what a specialist would know to look for and I did not.
These are the notes. They are organised by what I actually did, not by an argument.
The rest of the same three days went into What Grows Where the State Has Been, which reads the secondary literature on Chinese Protestantism and on what registration does to a congregation. These notes are the part that could not go in there: everything below rests on counting a body of text directly rather than on reporting what other people found, and putting it inside a literature digest would have given one section a different evidentiary footing from the rest of the page.
The Corpus
Lü Xiaomin (吕小敏), born 1970 into a farming family in Fangcheng county, Nanyang, Henan. One year of junior middle school, then back to farming. She cannot read musical notation and has had no musical training. According to a hostile critic, she began composing "without even having read much of the Bible."
The collection is 《迦南诗选》, the Canaan Hymns. The notated edition contains 2,246 songs. The index I was able to obtain covers 2,245 — number 825 is missing. They have been sung in unregistered congregations across China for about thirty years.
What makes the corpus worth the time is not its size. It is that no institution issued anything to its author. She has no theological degree, no ordination, no clergy registration, and appears on no roster. The state's licensing apparatus for religious personnel — qualification requirements, seminary coursework, filing and certification, and, where it applies, periodic assessment — acts entirely on clergy. None of it touches her.
That is not a claim about persecution. It is a claim about what kind of object this is: a body of religious material with no institutional address at any point in its production.
Four numbers, and why they differ. The notated edition holds 2,246 songs; the index I obtained lists 2,245; title comparison runs over the 1,976 covered by both indexes; lexical counts run over the 1,962 whose full lyrics I could retrieve. The denominators differ because comparing titles requires two overlapping indexes while counting words requires an obtainable full lyric. Each figure below is reported against the denominator it was actually computed on.
How the counting works, stated plainly enough to discount. Every proportion here is a song-level binary match against a hand-written word list — does this lyric contain any term on this list, yes or no. It measures lexical presence, not meaning. A song can be saturated with a theme and score zero because it uses none of my words, and the section below contains a live instance of exactly that. Treat the figures as orders of magnitude with a known instrument, not as semantic classification.
A caution about counting by title. Two indexes covering the same 1,976 songs agree on only 82.1 per cent of titles once punctuation is normalised. Most differences are function words (足迹/脚迹, 愿做/愿作, 祂/他); some are substantive — no. 89 appears both as 「我多想见到你远方亲人」 and 「送给远方的亲人」, no. 90 as 「心愿」 and 「心声」. This is a rate of title variation, not an error rate: a differing title may still contain the same keyword, and how often variation actually changes a keyword count is something I did not measure. It is a reason to work from lyrics rather than titles, which is what the figures below do.
Testing the Nationalism Reading — and Where the Test Was Blind
There is a standing characterisation of this corpus — most forcefully by a writer publishing as 崔以撒 (Cui Yisa) — as nationalist, worldly, verbose, and musically derived from Henan opera. I checked his evidence carefully (below, it holds up) and set out to test the reading.
Two things about the sample, before anything rests on it. The close reading covers twenty-eight songs on national themes, and they are not a random draw: I selected them from the ones Cui names plus a spread across the numbering, which is a purposive sample built to find the thing under discussion. Separately, and later, I obtained plain-text lyrics for 1,962 of the 2,246 songs, which is what the frequency figures further down rest on. Where a claim below is about proportions, it comes from the 1,962. Where it is about grammar, it comes from the 28, and it cannot be generalised to the corpus by counting.
Within those twenty-eight, "the nation as an end in itself" does not appear once. Where 中国 or 中华 occupies the subject position, the predicate is always a religious task. More often China is the grammatical object.
Because this turns on grammar and not on sense, the Chinese is necessary.
Song 664《中国大有希望》:
神的仆人哪你要刚强……有你迫切的祷告，中国大有希望 圣灵的水流气势磅礴……向东流啊流向中国
Servant of God, be strong… with your urgent prayer, China has great hope / The stream of the Spirit surges mightily… flowing east, flowing toward China.
The agents are the Spirit and the person praying. China is what the verb is done to — prayed for, flowed toward.
Set beside a standard nationalist construction, the difference is structural rather than tonal. My working test assumed that nationalism makes the nation the agent and God a resource, and that these songs do the reverse — God the agent, the nation the object, prayed over, watered, used, warned. Hold that assumption loosely; it is the thing that fails later on.
The actual content is a route, not a sentiment
The "national" material is not diffuse patriotic imagery. It is a specific, named geography.
Song 1491《耶路撒冷的天空和中国的天空》:
犹太的弟兄和我们的弟兄／我们今天今天在喀什相逢／接过神圣的使命 沿着前辈的脚踪／一定要把福音传回耶路撒冷
Jewish brothers and our brothers / today, today we meet in Kashgar / taking up the holy commission, following in the footsteps of those before us / we must carry the gospel back to Jerusalem.
Kashgar is the westernmost city in Xinjiang, the Silk Road's exit point. The thread runs through roughly ten songs — the Silk Road, "He will push open China's door, to send missionaries to the world," "China owes you too great a debt of gospel," "the Lord will bless China, and China will bless the world," all nations flowing to the holy mountain, two songs about Xinjiang peoples.
China is assigned a position in a relay: the gospel travelled east, and this generation returns it west. The nation is a segment of transmission, not a destination. That much I would still defend.
Where this reading fails, and how I found out
I wrote, at this point in the notes, that the songs therefore are not nationalist. That conclusion does not hold, and the way it came apart is the most useful thing in these notes.
"Carry the gospel back to Jerusalem" is not a neutral errand. It belongs to a specific movement with a documented organisational history, and that movement's own theology assigns China the final leg of salvation history — the one people who will complete what no one else completed. Two critics inside the Christian world say so plainly. Kim-kwong Chan (2013) objects to its "ethnocentric claim of China's role in bringing the gospel on its final leg," noting that some regard the movement as a hoax. Ezra Jin, of a Beijing church, objects that "a stronger biblical theology of mission does not encompass a Zionist theology."
So the two levels come apart:
Grammatically, China is the object — sent, used, blessed, pushed through the door. Theologically, China is the elect — the only one who can finish the route.
Being the grammatical object is, in this idiom, the highest position the text has to offer. My test asked who occupies the subject slot, and a text can nominate itself for a unique role in world history while sitting in the object slot the whole time. The test was structurally blind to the form of nationalism actually present.
Two consequences I would rather state than bury.
Cui Yisa was half right, and more right than his argument. I had scored his reading as wrong because the predicates are missionary. The predicates are missionary; that is not the point. He read these songs as nationalist and, at the theological level, there is a case for it — just not the case he made, which rested on tone and vocabulary. His conclusion stands up better than his reasoning does.
Having the whole corpus did not help. I had 1,962 full lyrics and thirty-four score pages when I wrote the paragraph that was wrong. No amount of additional text would have corrected it, because the error was in the question, not the sample. What corrected it was reading two theologians who work on the movement. A complete sample does not guarantee you can see the structure — it only guarantees you can count what you already thought to look for.
Two limits on the correction itself, so it does not travel further than it should. Chan and Jin are criticising the movement's theology, not these songs; I have found no study that analyses this group of fourteen songs directly. And the earlier observation survives intact at its own level: the vocabulary of national feeling is genuinely conscripted, and the sentences it is conscripted into genuinely aim at mission. That is compatible with the destination itself carrying an exceptionalist claim. Two layers, both real.
The mechanism is role substitution, not "adding Chinese elements"
The commoner explanation of such corpora is cultural blending — indigenous imagery layered onto imported doctrine. What I found on the scores is different and more specific: scriptural passages rewritten with the Chinese church inserted into the grammatical position of the original subject.
| song | source text | substitution |
|---|---|---|
| 728 | 2 Cor 4:8–9, rewritten line by line | subject replaced with 中国教会 |
| 823 | Gen 12:2, the Abrahamic covenant | "主要祝福中国，中国要祝福世界" |
| 1446 | Rom 1:14 + Isa 6:8 | "中国欠下你太多福音的债" + "我们在这里" |
| 706 | Luke 15 | "浪子都要回家" |
| 519 | Rev 5:5 | "犹大支派的狮子要来" |
| 1100 | John 12:24 | "一粒种撒向神州田，深埋在土下" |
This is not decoration. It is a congregation writing itself into the subject position of a canonical text. Whether that is more or less "nationalist" than the usual reading is a question I would rather leave open than answer badly — but it is certainly a different operation from adding local colour.
Two counter-cases, which I am recording rather than explaining away
Song 184《今日要象猛狮苏醒》 uses the sleeping-lion trope — a standard motif of modern Chinese nationalism — and the song contains no missionary predicate at all. Meanwhile song 519 uses the same lion image for Christ. The same vocabulary runs in opposite directions inside one corpus, which is hard to reconcile with a unified ideological product, and equally hard to reconcile with my own reading above. I have no account of it.
Song 551 contains an overseas and cross-strait perspective — "how many children of the nation abroad pray and weep for you, standing on the far shore of the ocean… hearing a mother's call across the strait." That does not sit comfortably with the standard account of a Henan farm woman composing alone. I have not verified that song's attribution and cannot.
One song attaches itself to Hudson Taylor
Song 1561《祖国啊，祖国，我要回来》 opens:
主啊愿祢感动戴德生的灵，今日也来感动我；愿祢燃烧前辈们传福音的火焰，今日也来燃烧我
Hudson Taylor founded the China Inland Mission. This does not show institutional continuity — the mission structures did not survive — but it shows that the name remained an invocable authority in congregational hymns written after the 1990s. Memory persisted where organisation did not.
The Henan Opera Attribution: A Question I Could Not Decide
Cui Yisa traces the melodic idiom to Henan opera. This is the part of the notes I am least satisfied with, because the honest result is a null one.
I read six scores. None is pentatonic. Four close on gong, two on yu, none on zhi.
The sourced reference for the opera modes (科普中国, 2018; corroborated in part by Zhang Guangtao's 2022 article on modal classification in Henan opera): the eastern Henan style closes predominantly on 5, a seven-tone zhi mode; the western Henan style closes on 1, a seven-tone gong mode.
Seven-tone, closing on gong — that matches western Henan style.
And it equally matches the default form of the Western hymn.
The feature yields identical predictions under both hypotheses. It therefore cannot discriminate between them. It is therefore not evidence, however good the match looks.
This is the most useful thing I learned in the whole exercise and it applies well beyond hymnody: a matching feature is only evidence if the competing hypothesis predicts something different.
And the claim has no source
Tracing "Henan opera influence" upstream: the sentence in the Chinese Wikipedia article on 《迦南诗选》 carries no footnote, and the references section is empty.
So the situation is: one unsourced encyclopaedia sentence, plus one hostile critic's ear, corroborating each other into an apparent consensus in search results — and they may well not be two independent witnesses.
The feature I thought would decide it, and what happened when I counted
Extended nonsense refrain. Song 1640《主要在中国掌权》 — the song Cui singles out — consists, apart from its final two lines, entirely of the syllables "HEI HEI HEI…", leaving as lexical content only "主要在中国掌权／美好的事情要发生." Song 1520 pairs "神州归神" with 锣鼓敲起来, drums and gongs.
Extended vocable refrain plus percussion imagery is characteristic of folk song-and-dance, work chant and yangge, and is uncommon in the imported hymn repertoire. I wrote here that it was the one feature the competing hypotheses predicted differently, and made counting it the first priority.
Then I counted. Strong vocable refrain occurs in 18 of 1,962 songs — 0.9 per cent.
0.9 per cent cannot carry "this corpus derives from local opera." I had taken one spectacular song and treated it as the character of a body of two thousand. That is the same error I had just spent a page attributing to other people, committed while the page was still open — which is worth recording, because it says something about how little protection naming an error gives you against making it. The name is not the safeguard; computing the denominator is.
The count did settle a smaller thing. Those 18 songs run from number 117 to number 2101, spread across the collection rather than clustered — so whatever the refrain is, it is not a house style local to one stretch of the numbering. And the first instance, 117《福音的勇士们》, carries the westward imagery and the vocables in the same song: crossing the Yangtze, crossing the Yellow River, out to the grasslands, out to the desert… with you hei-hei, with me hei-hei.
Two cautions I did not state the first time. Vocables of this kind are not specific to Henan opera — yangge, work chants, folk song and contemporary mass song all use them, so even a high rate would not have identified a source region. And gospel singing in English has its own non-lexical syllables and call-and-response. It was a feature with some discriminating potential, not a test.
Cui Yisa: Separating Evidence From Judgment
Because the critic is hostile and the conclusion I draw differs from his, his reliability had to be established independently. I checked his citations in ten places.
| check | result |
|---|---|
| Three numbered citations (664, 1591, 1652) | exact |
| Eight quoted lines against the printed score | verbatim |
| Textual variants he records (浪潮/水流, 自东流到/向东流向) | correct |
| Song 1640 described as almost entirely "HEI HEI HEI" | correct |
| Score page 664 printed title | matches |
Ten checks, no discrepancies. His evidence is reliable and I have used it.
His judgment — "full of worldly language and worldly style," "verbose and incoherent" — is an evaluation rather than a finding, and once the full lyrics were in hand it became a measurable one.
What the whole corpus does to two opposite verdicts
Two well-known characterisations of these songs point in opposite directions. Cui: full of worldly language, with named examples "having nothing to do with God at all." Duan Qi (2013): the songs are closely integrated with Confucian, Buddhist and Daoist tradition, with named examples — filial piety, repaying kindness, mother-in-law relations, neighbours. Counting across the 1,962 lyrics:
| claim | corpus | |
|---|---|---|
| Cui | saturated with worldly language | 96.9% contain religious vocabulary; those with none: 3.1% |
| Duan Qi | closely integrated with the three teachings | her four themes, combined: 1.9%; folk-religious content words (因果, 积德, 烧香, 风水, 命运, 菩萨, 祖宗 …): 1.1% |
Filial piety — 孝 — appears in one song out of 1,962. Mother-in-law, in two.
And both of them cited real examples. Cui's two songs are genuinely in the 3.1%: 521《家乡水，故乡山》 is about not having been home to harvest the sorghum or see one's parents in years; 1655《爸爸妈妈，听我说呀！》 is a left-behind child's letter to migrant-worker parents — you're away, I know you work so we can live like the children with money, I'd rather go without pocket money, I've woken up crying from dreaming of you. His description of them is accurate. Duan Qi's themes are presumably in there too.
Real examples, unsupported generalisations, in opposite directions, from the same corpus. That is what a striking case does when nobody computes the denominator — and I did it twice myself the same night, so this is a description of a common failure mode, not an accusation.
One more thing, because my own instrument was also faulty. Looking at those 60 songs with no religious vocabulary, most are not worldly at all: 85《良人和我》 is a paraphrase of the Song of Songs, which contains no divine name because its source does not; 127 and 145 are missionary songs. A keyword list that scores the Song of Songs as secular is not measuring what I said it measured — the 3.1% is an upper bound on "no religious vocabulary," not a count of worldly songs.
Note also what 1655 does to the argument. Cui uses it to show secularisation. It shows something else as well: who is in the room. It is a song about migrant labour and the children left behind — which is the same population the fieldwork literature describes filling these congregations. One text, two uses, neither false.
And he double-counts. Song 574《唱一首天上的歌》 opens with the line that gives 《生命的河》 its title; they are one song, listed as two representative works. Four examples are three. Not fatal to his case, but worth recording: one object with two names defeats counting by name, and it defeated both of us — see below.
Where My Own Counting Broke
Two of these are mine, and they belong in the notes for the same reason his double-count does.
The distribution claim I had to withdraw. I used keyword matching on song titles to chart where "national" material clusters, found it concentrated in the 1401–1800 numbering range, and — because a few dating anchors put that range after 2011 — was inclined to line it up with the discursive shift around 2012.
Then I obtained the missing numbering block, and my own data overturned it. The distribution is not monotonic: one peak plus scattered middling stretches, and among the last hundred-odd songs the national theme is entirely absent. Whatever produced those songs, it was not a trend rising to the present.
I had also predicted, before obtaining the earlier block, that the most famous songs were in it and so the national proportion there would be under-counted. Measured: 24/990 = 2.4 per cent in the first thousand, against 3.0 per cent after — the opposite direction. The famous song was there (《中国的早晨五点钟》, no. 268) but the ninety-nine songs around it contained no others. I had mistaken "where the famous ones are" for "what that stretch is like."
The count I made a priority before computing it. Set out above: one song consisting almost entirely of "HEI HEI HEI" led me to call extended refrain the decisive feature and rank counting it first. It is 0.9 per cent. Three times in one night — twice by me, once each by two published scholars — a vivid example stood in for a distribution. The instrument that catches it is boring and always the same: divide by the whole.
The inference I drew from title variation. I originally wrote that eighteen per cent title disagreement meant keyword counts carried eighteen per cent noise. That does not follow. A song can have two recorded titles and the same keyword in either; the disagreement rate is not a misclassification rate, and I never measured the misclassification rate. What the figure licenses is a decision rather than an error bar — work from lyrics, not titles — which is what the proportions here now do.
Materials for Future Work
If this ever becomes a real project
- Look at the eighteen vocable songs as music. The count is done and came to 0.9 per cent, which killed the feature as a test of origin. What the lyrics cannot show is whether those eighteen share a musical shape or only a notation habit — that needs the score pages.
- Read scores rather than lyrics. Plain-text lyrics are in hand for 1,962 of 2,246 songs; score pages for thirty-four. The national-theme subset alone is 64 songs, of which 52 have score pages available.
- Verify authorship of song 551. The overseas/cross-strait perspective is inconsistent with the sole-author account and nothing else in the corpus resembles it.
- Find fieldwork on Fangcheng. The county produced both this corpus and a well-documented case of a registered congregation being relocated to the edge of town, with the congregation dispersing into homes as a result. Both facts come from secondary sources.
- Locate the 1994 Nanyang survey (Li Liang) and the 2002 study by Wang Shenhong. These are the earliest local fieldwork on the region and I have neither.
- Read what has been written about the Back to Jerusalem movement's theology, properly rather than in the two critics quoted above. That is where my reading broke, and I have sampled the literature rather than worked through it.
Things I now think are not worth pursuing
- The county-level distribution. Not public. Would require gazetteer work or fieldwork, and the field regards the question as open for exactly this reason.
- The Henan opera attribution as currently posed. Not decidable by modal analysis, and the refrain count I was relying on to move it came in at 0.9 per cent. It would need a musicologist and a comparison repertoire, neither of which I have.
- Grammatical position as a test of ideology. It was the wrong instrument for this question, and it took two theologians to show me why. Where the highest role a text can offer is to be the one who is sent, counting subjects and objects measures nothing.