Resources · Plain-English explanations

Glossary

What the words on the research pages mean, from mythemes and traditions to trees, bootstrap numbers and Mantel tests. Used mainly by the Workbench and the Phylogenetic Analysis.

Find a word

Stories and the data about them

The building blocks: what is being compared, and how it is recorded.

Motif: A story idea that turns up in many places, such as “the raven steals the sun”. The motifs on this site come from Yuri Berezkin’s Analytical Catalogue, and each one has a reference code such as A13a.; Used in: Workbench · Phylogenetic Analysis
Telling also: narrative, version: One version of a story, as written down from one storyteller or source. A single motif can have dozens of tellings from all over the world.; Used in: Workbench · Phylogenetic Analysis
Tradition: A people or group whose stories are recorded together, such as the Ainu of Japan or the Navajo of North America. The site uses the traditions in Berezkin’s catalogue, and each one has a place on the map.; Used in: Workbench · Phylogenetic Analysis
Mytheme: One small building block of a story: a single event or idea, such as “the trickster hides the sun in a box”. Splitting tellings into mythemes lets us compare them fairly, a bit like comparing recipes ingredient by ingredient. The Fielder-White Mytheme Catalogue (FWMC) lists the mythemes for each motif.; Used in: Workbench · Phylogenetic Analysis
Mytheme list also: vocabulary: The set of mythemes worked out for one motif. Every telling of that motif is checked against the same list. The list itself is checked automatically, and if it fails those checks, no verdict is given.; Used in: Workbench · Phylogenetic Analysis
Coding also: coded: Going through a telling and recording, for each mytheme, whether it is present, absent or unknown. On this site the mythemes were recorded with the help of a computer model, and they are analysed again whenever they change.; Used in: Workbench · Phylogenetic Analysis
Present, absent and unknown: Present means the telling includes the mytheme. Absent means someone checked and it is not there. Unknown means nobody could tell, for example because the text is short or damaged. Unknown is never counted as absent: missing information is not the same as a “no”.; Used in: Workbench · Phylogenetic Analysis
Mytheme that differs also: informative mytheme: A mytheme that some traditions have and others do not. A mytheme that every tradition has (or that none has) cannot tell traditions apart, so it does not help to build a tree.; Used in: Phylogenetic Analysis
Run: One complete pass of recording the mythemes for a motif. When a motif is recorded again it gets a new run number, and results from the older run are marked out of date.; Used in: Workbench · Phylogenetic Analysis
Data fingerprint: A short code worked out from the exact data used. If anything in the data changes, the code changes too, so you can tell whether two results came from the same data.; Used in: Phylogenetic Analysis
Approximate location: Some traditions cover a wide area, so their dot on the map is a rough central point rather than an exact place. These dots are drawn with a dashed or double ring, and some distance checks leave them out.; Used in: Workbench · Phylogenetic Analysis
Great-circle distance: The shortest distance between two places across the curved surface of the Earth, the way a plane flies, rather than a straight line on a flat map.; Used in: Workbench · Phylogenetic Analysis
Documentation bias also: sampling bias: Some traditions have many recorded tellings and others only one, mostly because of where collectors happened to work. Traditions with more tellings can look more alike just because more is known about them. The analysis checks for this.; Used in: Phylogenetic Analysis
Evening out also: levelling, rarefaction: A check against documentation bias. Every tradition is cut down to the same small number of tellings, picked at random, and the test is run again many times. If the pattern survives, it is not just coming from the best-recorded traditions.; Used in: Phylogenetic Analysis
Sample: A smaller set picked from all the data, for example one telling from each tradition. Repeating a test on many random samples shows how much the result depends on which tellings happened to be included.; Used in: Workbench · Phylogenetic Analysis

Comparing tellings

Ways of measuring how alike two tellings are, and the diagrams that show it. Most of these are used in the Workbench.

Similarity score: A number from 0 to 1 (or 0% to 100%) showing how alike two tellings are on one measure. 0 means nothing in common and 1 means identical. A high score shows that two tellings are alike; it does not prove that one came from the other.; Used in: Workbench
Jaccard also: presence similarity, shared mythemes: Compares which mythemes two tellings have. Only mythemes that were checked in both tellings count. The score is the number of mythemes both have, divided by the number that either has. If two tellings share 3 mythemes and have 6 between them, the score is 3 ÷ 6 = 0.5. A mytheme missing from both tellings adds nothing: two tellings are not called alike just because they both lack something. On the tree page, “difference in mythemes” is 1 minus this score.; Used in: Workbench · Phylogenetic Analysis
Order concordance also: chronology: Looks only at the mythemes two tellings share, and asks whether they happen in the same order. It checks every pair of shared events, such as “the sun is stolen” and “the sun is set free”: does the first come before the second in both tellings? The score is the share of pairs that agree, so 1 means the same order throughout and 0 means the opposite order every time.; Used in: Workbench
Sequence overlap also: LCS Dice: Lines up the full list of events in two tellings and finds the longest chain of events that appear in the same order in both, even with other events in between (the “longest common subsequence”). The score is twice the length of that chain, divided by the number of events in both tellings together. 1 means the two tellings have exactly the same events in the same order.; Used in: Workbench
Event pair also: comparable event pairs: Two shared events whose order is known in both tellings. Order scores only mean something when there are enough of these: a perfect score from a single event pair tells you very little.; Used in: Workbench
Joint coverage also: mythemes assessed in both: The share of the motif’s mythemes that were actually checked (present or absent, not unknown) in both tellings, for example 15 of 17. Low coverage means a score rests on very little evidence.; Used in: Workbench
Ranking-eligible: A pair of tellings with enough evidence to be listed in the rankings: normally at least 6 mythemes checked in both tellings and at least 3 that both have. It is a minimum amount of evidence, not a sign that the pair is unusually alike.; Used in: Workbench
Similarity matrix: A grid with every telling along the top and down the side. Each square shows how alike that pair is, with darker colours for more alike.; Used in: Workbench
Network: A diagram in which each dot is a telling and lines join pairs that are alike. Thicker or darker lines mean more alike. Where a dot sits on the screen has no meaning on its own: only the lines matter.; Used in: Workbench
Weighted modularity also: how groups are found: The method the Workbench uses to find groups in the network. Each telling starts in a group of its own. The two groups whose links are stronger than you would expect (given how strongly each links to everyone else) are joined, and this repeats until no join makes the groups any clearer.; Used in: Workbench
Ungrouped: A telling that is not in any group: in the network, because its links are too weak, or in the dendrogram, because it does not join anything above the cutoff. Ungrouped tellings are shown as grey dots.; Used in: Workbench
Groups also: communities: Sets of tellings in the network that are more strongly linked to each other than you would expect. A computer method finds them automatically, and they can change when you change the settings.; Used in: Workbench
Dendrogram also: hierarchical clustering: A tree-shaped diagram that joins the most alike tellings first, then joins those groups, and so on. Cutting it at a chosen level splits the tellings into groups.; Used in: Workbench
Average linkage: The rule the dendrogram uses to join groups. Two groups join at the average score of every pair of tellings between them, and only if every one of those pairs has a score.; Used in: Workbench
Cutoff also: minimum similarity: A threshold you choose. Only pairs that score above it are linked or grouped. A low cutoff gives bigger, looser groups; a high one gives smaller, tighter groups.; Used in: Workbench
Co-occurrence: How often two mythemes turn up in the same telling. Mythemes that keep appearing together may form a fixed episode of the story.; Used in: Workbench
Sequence transitions: Which mytheme tends to come straight after which, counted across all the tellings. It shows the usual flow of the story.; Used in: Workbench
Median position: Where a mytheme usually falls in a telling, from the start (0%) to the end (100%). Line up its position in every telling and take the middle one: half the tellings have it earlier and half later. Unusual tellings throw a median off less than they would an average.; Used in: Workbench
Percentage points also: pp: The difference between two percentages, found by taking one from the other. If a mytheme is in 60% of one group’s tellings and 40% of another’s, the difference is 20 percentage points.; Used in: Workbench
Model confidence: How sure the computer model that recorded the mythemes was about one particular coding, shown as a percentage beside the quotation it used. It is the model’s own estimate, not a check by a person.; Used in: Workbench
Sensitivity check: Running an analysis again with one setting changed, or with a different sample, to see whether the result stays the same. Results that change a lot should be treated with care.; Used in: Workbench

Trees

The branching diagrams on the Phylogenetic Analysis page.

Phylogenetic analysis: Methods first made for working out the family trees of living things from their features. Here they are used on stories: traditions are compared by their mythemes and arranged into a tree of similarity. A tree of stories shows which tellings are alike; it does not by itself prove a family history.; Used in: Phylogenetic Analysis
Pre-Filter: What this site’s phylogenetic analysis is for: a quick first pass over hundreds of motifs, to pick out the ones worth studying more closely. Its results are leads to follow up, not final findings.; Used in: Phylogenetic Analysis
Tree: A branching diagram, like a family tree drawn on its side. Traditions that are most alike are joined first, then those groups are joined, and so on, until everything is connected.; Used in: Phylogenetic Analysis
Tip also: leaf: The end of a line on the tree, marked with a dot. Each tip is one tradition. The number in brackets after a name is how many of its tellings were used.; Used in: Phylogenetic Analysis
Branch: A line in the tree. Branches meet where groups of traditions join. To see how closely two tips are related on the tree, follow their branches back to where they meet.; Used in: Phylogenetic Analysis
Branch length: How long a branch is. On this site, a longer branch means more difference in mythemes. Ticking “Equal branch lengths” draws every branch the same length, which makes the shape easier to read but hides how different the traditions are.; Used in: Phylogenetic Analysis
Group also: clade: A point where branches join, and everything that comes off it: a set of traditions that sit together on the tree.; Used in: Phylogenetic Analysis
Neighbour-joining: A quick and widely used way to build a tree. Again and again, it finds the two traditions (or groups) that are closest and joins them, until everything is connected. It is the main tree on this site.; Used in: Phylogenetic Analysis
Maximum likelihood: A slower way to build a tree. The computer tries many possible trees and keeps the one that best explains how the mythemes could have been gained or lost along the branches. Groups that appear in both this tree and the neighbour-joining tree are more trustworthy.; Used in: Phylogenetic Analysis
Bootstrap also: reliability number, support: A way to test how reliable each group in the tree is. The computer makes many new copies of the data by picking mythemes at random (some twice, some not at all), builds a tree from each copy, and counts how often each group appears again. That count, out of 100, is the small number where branches join. 70 or more is usually treated as fairly reliable; lower numbers are more of a guess.; Used in: Phylogenetic Analysis
Newick file: A plain-text way of writing down a tree with brackets, such as ((A,B),C). Tree programs such as FigTree or iTOL can open it.; Used in: Phylogenetic Analysis

Testing for patterns

How the pages decide whether a pattern is likely to be real or could just be chance.

Correlation also: r, Pearson r: A number from −1 to +1 showing how closely two things go together. +1 means they rise together perfectly, 0 means no link, and −1 means one rises as the other falls. On these pages it usually compares distance with difference: a positive r means places further apart are less alike. Pearson r is the most common kind.; Used in: Workbench · Phylogenetic Analysis
p-value also: p: How likely you would be to get a result at least this strong by pure chance, if there were really no pattern. p = 0.03 means about a 3% chance. Smaller numbers are stronger evidence. A p-value is not the chance that the result is true, and it says nothing about how big the effect is.; Used in: Workbench · Phylogenetic Analysis
Significant: A result whose p-value is below a chosen line, usually 0.05 (5%). It means the result would be unusual if only chance were at work. It does not mean the result is large or important.; Used in: Workbench · Phylogenetic Analysis
Multiple testing also: uncorrected p-values: Run hundreds of tests and some will come out “significant” by luck alone: about 1 in 20 at the 0.05 line. The p-values on these pages are not adjusted for this, so check any single striking result carefully before relying on it.; Used in: Workbench · Phylogenetic Analysis
Permutation test also: shuffling, null model: A way of finding a p-value. The computer shuffles the data many times, for example swapping which location belongs to which tradition, and counts how often a shuffled version gives a result as strong as the real one. If that rarely happens, the real pattern is unlikely to be chance. The “null model” is the starting assumption that there is no pattern at all.; Used in: Workbench · Phylogenetic Analysis
Null model also: no-pattern assumption: The starting assumption that there is no real pattern, used to work out what chance alone would produce. In the Workbench’s distance test, it assumes the recorded locations could be swapped freely between traditions.; Used in: Workbench · Phylogenetic Analysis
One-sided test: A test that only looks for a pattern in one direction. The Workbench’s distance test only asks whether similarity goes down with distance, so it only counts shuffles where the drop is at least as steep as the real one.; Used in: Workbench
Exact test and random shuffles also: Monte Carlo test: Two ways of running a permutation test. With only a few traditions (six or fewer), every possible arrangement can be tried: an exact test. With more, there are far too many, so the computer tries 999 random shuffles instead (a Monte Carlo test) and estimates the p-value from those.; Used in: Workbench
Seed: A starting number for the computer’s random shuffles. Using the same seed repeats exactly the same shuffles, so anyone can get the same result again.; Used in: Workbench · Phylogenetic Analysis
Mantel test also: distance test: Compares two sets of distances between the same traditions: how far apart they are on the map, and how far apart they are on the tree. If traditions far apart on the map also tend to be far apart on the tree, the story seems to follow geography. Its p-value comes from a permutation test.; Used in: Phylogenetic Analysis
Partial Mantel test: The same test, but allowing for which continent each tradition is on. If the pattern disappears, the motif differs between continents rather than fading steadily with distance.; Used in: Phylogenetic Analysis
Continental contrast: When a motif is told differently on different continents, but inside each continent places further apart are no less alike.; Used in: Phylogenetic Analysis
Corridor also: route: A pair of continents joined by a possible route, such as Asia and North America across the Bering Strait. The test asks whether traditions on either side are more alike than their distance would predict. Some routes are marked uncertain because researchers disagree about them.; Used in: Phylogenetic Analysis
Verdict: The short summary of a motif’s analysis, such as “Follows geography” or “Too little data to judge”. It is decided by fixed rules from the test results, not by an AI, so the same data always gets the same verdict.; Used in: Phylogenetic Analysis

A–Z list

Every word on this page, in alphabetical order.

A Approximate location Average linkage

B Bootstrap Branch Branch length

C Co-occurrence Coding Continental contrast Correlation Corridor Cutoff

D Data fingerprint Dendrogram Documentation bias

E Evening out Event pair Exact test and random shuffles

G Great-circle distance Group Groups

J Jaccard Joint coverage

M Mantel test Maximum likelihood Median position Model confidence Motif Multiple testing Mytheme Mytheme list Mytheme that differs

N Neighbour-joining Network Newick file Null model

O One-sided test Order concordance

P p-value Partial Mantel test Percentage points Permutation test Phylogenetic analysis Pre-Filter Present, absent and unknown

R Ranking-eligible Run

S Sample Seed Sensitivity check Sequence overlap Sequence transitions Significant Similarity matrix Similarity score

T Telling Tip Tradition Tree

U Ungrouped

V Verdict

W Weighted modularity

If you wish to get access to data for projects and research, then please contact us, we do our best to help

How to cite: Fielder-White, Jonathan (2026). The Folklore Database. Citation version 1.261113.1 (13 November 2026). https://folkloredatabase.com.

0009-0006-1672-4451
The citation version records the access date and does not identify an archived database snapshot.
Please cite the relevant sources, catalogues, and classification systems, including the Folklore Database, Berezkin-Duvakin Catalogue, ATU, Thompson, alongside the page accessed and the last time accessed. Please respect copyright and licences.

Editorial annotations and database enhancements © Jonathan Fielder-White licensed under CC BY 4.0. Berezkin-Duvakin derived narrative and metadata licensed under CC BY-NC-SA 4.0. Full rights information can be found here.
Rights in original software, interface design and other protected material not expressly licensed for reuse are reserved to their respective rights holders.

Text and data mining: For protected material not offered under a Creative Commons or other reuse licence, the Folklore Database reserves the rights it holds in automated extraction, dataset creation and AI training, to the extent permitted by law. This reservation does not restrict uses permitted by the applicable Creative Commons or other licence, public-domain status, or statutory exceptions and limitations. Third-party material retains its own terms. Please contact us for help identifying the applicable sources, licences and attribution for bulk research access.

Designed and Developed by Jonathan Fielder-White.
© 2024–2026 Crecganford / Jonathan Fielder-White. Rights reserved except where licensed or otherwise stated.