Over some four decades, one man read a quarter of a million Latvian folksong texts, one at a time, each on a slip of paper three centimetres by eleven, and decided where every one of them belonged. The instrument he built to hold them — a cabinet made in Moscow in 1880, seventy-three drawers, 268,815 slips — is the Dainu skapis, on the UNESCO Memory of the World register since 2001. The edition that came out of it, Latvju dainas, ran to six volumes between 1894 and 1915 and holds 217,996 songs.
Barons was clear that the sorting, not the collecting, was his contribution. It is not a book that happens to be organised. The organisation is the work.
Latvian folkloristics has read the dainas overwhelmingly through close reading: quote a song, read outward from it. That method has produced nearly everything worth knowing about them. What it cannot do is look at forty thousand songs at once.
This page does that, on the slice of the collection that is machine-readable. Four questions, in order of how much each leans on the one before it: where the reusable language sits, what a song does at its start, where variants of the same song come apart, and — the one that turned out to be about Barons rather than about the songs — whether the words agree with how he filed them.
What is being counted
The corpus is published by the Institute of Mathematics and Computer Science at the University of Latvia at korpuss.lv: 46,587 song texts, 921,508 words, recorded 1770–1903, lemmatised, and — the thing that makes any of this possible — each song still carrying Barons’ own hierarchical classification, with his Latvian descriptions intact: 2.4.4.2.1. Meita tura godu, sargā vaiņagu, netura godu, zaudē vaiņagu.
It is one volume of the six, not the cabinet. Specifically it is Volume 2, dated 1903 — the youth songs. Every classified song carries a code beginning 2.4: adornment, the wreath, the ring, learning to work, the dowry, reputation, girls’ and boys’ songs, night visits, old maids and bachelors. Childhood, weddings, married life, death and the mythological songs are in the other five volumes. Everything below is true of the youth songs and is not claimed of the rest.
One more gap, invisible in the data: the printed volume holds 47,781 texts and this corpus has 46,576, because the Latgalian-language songs were excluded when it was built. So it is not quite complete even for its own volume, and the exclusion removes a dialect region — worth knowing before reading anything here as a statement about Latvian folksong at large.
AWhat travels
The formula inventory
Oral song is built from reusable pieces — that has been the working assumption since Parry and Lord went looking for Homer in Yugoslavia. The question a corpus can settle is which pieces, and how far they get.
“Reusable” needs care here, because this corpus is variant-rich: the same song was often written down several times from different singers with a word changed. A run of words appearing in “two songs” may just be one song recorded twice. So the test is stricter: a four-word run counts as travelling only if it turns up under two different themes of Barons’.
By that measure, almost nothing travels.
| A four-word run appears in… | Runs | Share |
|---|---|---|
| one theme only | 266,402 | 98.3% |
| two themes | 4,051 | 1.5% |
| three themes | 387 | 0.14% |
| four or more | 133 | 0.05% |
Most four-word runs occur exactly once, and something that occurs once cannot cross a boundary, so that table is partly a statement about rarity. Controlling for it changes the number and not the conclusion: of the 6,060 runs common enough to appear in ten or more songs, 81.8% still never leave a single theme, and 1.7% reach four or more.
The stock a singer draws on is not one reservoir. It is a set of reservoirs, one per subject, and they barely leak into each other.
And the ones that do get out
| Themes | Songs | Sequence | |
|---|---|---|---|
| 10 | 99 | kad es biju jauna | when I was young |
| 10 | 96 | es biju jauna meita | I was a young girl |
| 10 | 23 | ai dieviņi ai dieviņi | oh God, oh God |
| 9 | 43 | vai dieviņi ko darīšu | oh God, what shall I do |
| 8 | 17 | kad es iešu tautiņās | when I go to the suitors’ house |
| 7 | 65 | es mātei viena meita | I am my mother’s only daughter |
| 7 | 47 | visi saka visi saka | everyone says, everyone says |
| 7 | 12 | visi ļaudis tā sacīja | all the people said so |
| 6 | 49 | es bagāta mātes meita | I am a rich mother’s daughter |
| 6 | 45 | drebi drebi apšu lapa | tremble, tremble, aspen leaf |
| 6 | 32 | vai māmiņa vai māmiņa | oh mother, oh mother |
Read the list and the shape of the system is visible without any further statistics. What travels is frames, not images: self-introductions, addresses, laments, the “everyone says” that sets up a report of gossip. This is the machinery for getting a line started and pointed at a subject.
The images stay home. The wreath, the oak, the dowry chest, the mill — the objects the songs are actually about — do not appear on this list at all. What crosses Barons’ boundaries is grammar under pressure of metre. What stays inside them is meaning.
BHow a song starts
The opening formula system
A companion analysis on the dissertation site found that the first tenth of a daina is 2.3× more formulaic than everything after it. That says the opening is a distinct functional slot rather than just the place a song happens to begin. So: what is in it?
Seven songs in ten start with a line some other song also starts with, and the 500 commonest openings — 2.4% of the inventory — account for better than a quarter of the whole corpus. There is a long tail behind that, but the head of it is small and heavily used: enough openings that a singer always has one ready, few enough that every one of them is familiar to the room.
And the slot is genuinely reserved. Of the 1,665 openings established enough to be used by five or more songs, only 11.8% ever appear anywhere but at the start of a song. These are not general-purpose lines that happen to turn up first. They are lines for starting with.
| Songs | Opening | |
|---|---|---|
| 177 | dod māmiņa kam dodama | give me away, mother, to whoever you will |
| 163 | ābelīte dievu lūdza | the little apple tree prayed to God |
| 160 | pieci brāļi viena māsa | five brothers, one sister |
| 121 | adu cimdus adu zeķes | I knit mittens, I knit stockings |
| 98 | kad es biju jauna | when I was young |
| 88 | kalniņā stāvēdama lejā laidu | standing on the hill, I let it down to the valley |
| 86 | jauni puiši mutes deva | the young lads gave kisses |
| 80 | ganos gāju kreklu šuvu | I went herding, I sewed a shirt |
| 76 | es meitiņa kā puķīte | I am a girl like a little flower |
| 73 | caur ābeļu birzi jāju | I rode through the apple grove |
| 70 | puķe puķe roze roze | flower, flower, rose, rose |
| 67 | protu protu redzu redzu | I understand, I understand, I see, I see |
Notice what these are. Almost every one establishes a speaker and a situation in a single line — who I am, what I am doing, what I am asking for. Two of them (puķe puķe roze roze, protu protu redzu redzu) do nothing but establish rhythm and a voice, saying essentially nothing at all. Those are warm-up lines in the most literal sense.
CWhere a song lets go
Variant drift
The archive records the same song many times over. That is usually treated as a nuisance — noise to be deduplicated before the real analysis starts. It is also a natural experiment: every family of variants is a record of what singers kept and what they changed.
Group every song by its first verse line, keep the families of five or more, and ask how far the family agrees, line by line. 1,702 opening lines qualify, covering 22,204 songs — nearly half the corpus.
The collapse is immediate, not gradual. By line two the family has already split in half, and after that it barely erodes further — lines three and four are only a little worse than line two. Whatever happens, happens at the seam between the first line and the second.
| Songs | 2nd lines | Shared opening |
|---|---|---|
| 175 | 66 | dod māmiņa kam dodama |
| 170 | 31 | ābelīte dievu lūdza |
| 160 | 46 | pieci brāļi viena māsa |
| 121 | 20 | adu cimdus adu zeķes |
| 113 | 16 | kalniņā stāvēdama |
| 111 | 23 | maza maza meitenīte |
| 95 | 44 | kad es biju jauna meita |
| 94 | 29 | nav neviena ozoliņa |
A daina variant family is not a song with small changes. It is a fixed opening with an open continuation.
Which is the same finding as the front-loading of formula, arrived at from the opposite direction. The opening is a shared, public, reusable object — a hundred and seventy-five singers reached for dod māmiņa kam dodama and then went sixty-six different ways. Everything after the first line is where somebody is actually singing.
DWhat Barons knew
The taxonomy as an object of study
Barons had no theory of classification and no committee. He had slips of paper, a cabinet, and four decades. The obvious modern question is whether the categories he arrived at are real — whether they correspond to anything in the songs, or whether they are one man’s idiosyncratic filing.
The test: train a plain naive-Bayes classifier on the words of 80% of the classified songs, and see how often it can recover his filing decision on the other 20%. Forty-seven sections to choose between.
That is a high number, and it is the first thing worth saying. A taxonomy built by hand in the decades around 1900, by one man reading songs one at a time, is almost entirely recoverable from vocabulary alone a century later. Barons was consistent to a degree that would be unremarkable only if he had been following a written rule, and he was not.
The other 13%
The interest is in the failures, because the classifier does not fail randomly. It fails in exactly two ways.
It collapses sections that share a situation and differ only in who is speaking. Its commonest mistakes are girls’ songs, boys’ songs, and other people’s advice about courtship — three sections using the same words about the same events. What separates them is the voice, and voice is not in the vocabulary. Barons separated them anyway. He was tracking who sings and to what end — a judgement about the songs that the words cannot carry and that no clustering algorithm would ever have made.
And it fails on his finest distinctions. He divided the weaving songs into thirteen sub-sections by which garment is being made: shirts, mittens, stockings, shawls, skirts, sheets, belts, the bleaching, the fulling. Those are among the worst-classified sections in the corpus, because the songs are nearly identical apart from one noun.
| Recovered | Section |
|---|---|
| 94.6% | 2.4.7.1 Meitu dziesmas girls’ songs |
| 94.1% | 2.4.3.2 Vaiņags the wreath |
| 93.8% | 2.4.5.9 Maltuve the grinding room |
| 92.9% | 2.4.6.1 Ļaužu valodas… gossip and reputation |
| … | |
| 60.3% | 2.4.2.2 Seja, vaiga sārtums, acis… face, eyes, beauty |
| 59.3% | 2.4.5.8.8 Auž, raksta, pušķo villanes weaving shawls |
| 54.8% | 2.4.5.8.12 Audeklu balināšana bleaching cloth |
| 48.5% | 2.4.6.2 Ļauni ļaudis, skauģi, burvji… ill-wishers, envy, sorcerers |
That is not a flaw in his scheme. It is a record of how finely a nineteenth-century Latvian farm distinguished work that a lexical model — and most modern readers — would see as one activity. He kept a distinction the language itself barely marks, because the people he collected from kept it.
The cabinet is not a container for the songs. It is an argument about them, and most of it survives contact with the evidence.
Method, and a note on manners
Everything here runs on a single local snapshot of the corpus, harvested once and analysed offline. korpuss.lv is a public academic service run by a research institute, not a commercial API; firing a query per question at it would be poor manners, and analysing a fixed snapshot has the side benefit that every number above is reproducible. The code is a few hundred lines of Python with no dependencies.
Three things that had to be got right
Editorial parentheses are stripped. The printed edition records variant readings inline — Kam, meitiņa ( māsiņa), skaista augi. Those are the editor’s, not the singer’s, and leaving them in inflates apparent formulaicity exactly where an editor happened to notice variation.
Section-crossing, not song-crossing. In a variant-rich corpus, “shared between songs” and “written down twice” are the same measurement. Requiring a sequence to cross two themes is the difference between 27.2% of four-word runs looking formulaic and 1.7% of them actually being so.
Unpunctuated line breaks are recovered. The edition does not always set a comma at the end of a verse line, so splitting on punctuation alone silently fuses pairs of lines. The tell is a population of fifteen- and sixteen-syllable “lines”, which is twice a daina line and therefore impossible. Uncorrected, it distorts every count of song length and every variant family.
What this is not
It is one volume of Barons out of six, and roughly a fifth of the songs he collected. It is large enough for distributional claims and far too small to prove an absence: “not attested here” never means “does not exist.” Nothing above is a claim about Latvian folksong in general, and several of the results — particularly the theme-locality of formula — might look quite different across the whole cabinet. That would be the next thing to find out.