I · The dial
A tale of 686 words and one of 17,792 keep the same time
Folk tales are not the same length, and they were not written down in the same place. The shortest in this corpus, Zapisul lui Adam, runs to 686 words; the longest, Creangă’s Povestea lui Harap-Alb, to 17,792. To lay them on top of one another, each was stretched across the same dial: first word at zero, last word at one hundred. A word at 51% of a three-thousand-word tale and a word at 51% of a ten-thousand-word one end up in the same place.
The dial works with words, not with scenes tagged by hand. “The zmeu” means zmeu, zmeoaica, balaur and their other forms, counted across all 138 tales at once. Where the occurrences pile up, that is where the rung happens.
The order that comes out was not imposed from outside. Nobody tagged scenes, functions or motifs here: words and their positions were counted, and the rungs sorted themselves.
II · The ladder
Poverty at a quarter, the wedding at three quarters
Cut open and laid flat, the dial becomes a ladder. Every row below is one rung, and the ridge shows how many of its occurrences fall in each twentieth of a tale. The dashed line is the midpoint.
Poverty and widowhood crowd into the first two twentieths and never come back: the tale needs a lack in order to start, and gets rid of it quickly. The travelling bag, the provisions and the saddlebags arrive at 36%, the forest at 39%. The fight and the dragon sit on top of each other in the middle. The ring, the token and the verb to recognise climb to 64%, because everything hidden in the first half has to be proved in the second. The wedding stands at 77%, with a spike pressed against the right-hand edge.
The last rung is the closing formula: saddle, I mounted, lie. Those words fall at 96% of the tale and almost nowhere else. They are the only words in the corpus whose hour is fixed to the minute.
III · Tale by tale
The ends are nailed down, the middle is free
Averages always hide something, so each tale is worth looking at on its own. Below, every tick is one tale: its position is the midpoint of that rung inside that particular tale. The pale band covers the middle half of the tales, and the black line is the midpoint of the whole corpus.
The order holds very well at the ends. Among the tales that have both poverty and a wedding, poverty comes first in 55 out of 61. The closing formula comes after poverty in 52 out of 55, and after the forest in 53 out of 60. The end of a folk tale is not up for negotiation.
In the middle the arithmetic falls apart. The dragon shows up before the fight in 26 tales out of 50, which is what a tossed coin would give you. The forest comes before the dragon in 31 out of 52. The teller needs a lack at the start and a wedding at the finish; in between, the forest, the dragon, the crone and the three trials can be dealt in whatever order comes to hand, because an audience remembers the ends, not the order.
IV · The cast
God turns up in more tales than the emperor
Asked which character is hardest to keep out of a Romanian folk tale, most people would guess Făt-Frumos, the Prince Charming of the genre. His name appears in 29 tales out of 138. The dragon, in 59. The emperor, in 107. And God, in 124.
God rarely walks into the story as a character, the way he does in the tales about his godson, where he sits down to dinner and takes to the road. Most of the time he is a turn of phrase: God grant it, God keep us, as much as God gave. These tales were collected from people for whom heaven was a daily presence, and their speech stayed in the text even when the story is about dragons and enchanted horses.
The horse appears in 88 tales, almost thirty more than the dragon, and keeps the same hour as the emperor: around the middle. The fates, by contrast, appear in only ten tales, and when they do they come early, at a fifth, since their job is to speak before anything else happens.
V · The frame
The tale tells you itself that it is a lie
103 of the 138 tales open with a formula of time, and 41 of those with a fost odată ca niciodată, “once upon a time as never before”. The formula dates nothing: it says only that nobody knows when, and that it makes no difference. Some tellers start negotiating with the listener straight away.
How you get in
“Story, story, word of a story, tonight it reaches us on a stalk of rye, to keep you in talk this evening.”
Vasilie-finul-lui-Dumnezeu“Story, story, but I am not from the days of stories. I went one day to my mother-in-law and found a sack of stories, and coming home I dropped it and the sack split open.”
Finul-lui-Dumnezeu“Somewhere near the start of the ages, further than the mind’s sling can throw, it is told, as if made up, that man was cut from other cloth.”
Palatul de cleștar · Barbu Ștefănescu-DelavranceaHow you get out
“And I climbed on a nail, so as not to tell it again; I climbed on a rusty blade, and perhaps the story was somewhat of a lie.”
Din fată, fecior · Ion Pop-Reteganul“And then I mounted a strawberry and told you, good people, one great fat swollen lie!”
Capra cu trei iezi · Ion Creangă“Let whoever knows the rest tell it; but somebody give me a pipe of tobacco and then tell me to be quiet!”
Tata florilor · Ion Pop-ReteganulThe tale closes with a retreat. The teller rides out of his own story, asks to be paid in tobacco and blames the rusty saddle. In Ispirescu the formula is so well known that it no longer needs writing out in full: 31 of his 37 tales end with it, and in 21 he stops halfway, with “And I mounted a saddle etc.” The reader finishes it alone.
Between the two formulas the tale behaves as if everything were true: emperors with names, journeys counted in days, prices, food, brothers quarrelling. The frame is what holds the lie at arm’s length, at 0% and at 96%.
VI · Your own clock
Put any word on the dial
Type a Romanian word. If it occurs at least three times in the corpus you get its curve: at which moment of a tale the tellers use it, how many tales it appears in, and three real sentences it was written into. Nearby forms are gathered together and listed under the curve, so you can see what was counted.
Everyday words come out flat: casă (house), drum (road) and bani (money) spread evenly, because the tale uses them all the time. The ones with a sharp peak are the ones with a job to do at a particular moment: tată (father) early, inel (ring) late, murit (died) in the last twentieth, where it mostly sits inside “if they have not died, they are living still”.
VII · How it was made
The method, briefly
The texts come from the Romanian Wikisource archive, downloaded on 17 August 2026. The pages kept are the ones marked there as folk tales, with at least 600 words and with diacritics: 138 tales, 517,844 words, 26,942 distinct forms. Most come from the collections of Ion Pop-Reteganul in Transylvania and Petre Ispirescu in Wallachia; the rest are from Creangă, Slavici, Delavrancea, Filimon, Stăncescu and Eminescu. All of them left copyright long ago.
Each tale was cut into words, and every word was given its position in the text, from 0 to 1. The curves show how many occurrences fall in each twentieth, and the figure beside each rung is the midpoint of those occurrences. Old spelling was modernised only far enough to keep one word from splitting into two: î inside a word read as â, and cedilla ş, ţ brought to the comma-below forms.
The rungs are hand-picked word lists, printed in full under the dial. Choosing them is the only place on this page where human judgement enters, and its limits are visible: the fight also collects a tăia, “to cut”, which in these tales cuts firewood as well as dragons, and the zmeu cannot cover what happens when the enemy is a crone or the hero’s own father. In 79 tales out of 138 there is no zmeu and no balaur at all.
In the explorer, words are grouped by a simple Romanian suffix stripper, which does make mistakes: măr (apple) ends up in the same family as mării (of the sea). That is why every gathered form is written under the curve with its count, and can be taken one at a time.
The texts are under the Creative Commons Attribution-ShareAlike 4.0 licence, like all of Wikisource.
Marius Comper · 17 August 2026
↑ back to top