A lexical distance heatmap turns a table of language-to-language scores into a coloured grid you can read at a glance. Each cell shows how far apart two languages are in vocabulary: the lower the number, the more words they share. The chart below covers 19 languages, labelled in the data with two-letter codes such as En, Du and Da, and it is a good example of when a heatmap beats a table.
What does a lexical distance heatmap show?
A lexical distance heatmap shows a score for every pair of languages in a square grid, with a colour for each score. Lexical distance measures how different two languages' vocabularies are, so a small score means closely related words and a large score means little overlap. It is a measure of words, not of grammar or of how well speakers understand each other.
The idea is not mainstream linguistics, and the scores depend on the method that produced them. For background on lexical distance maps of Europe's languages, see Big Think's Strange Maps article and this discussion of linguistic distance data.
How to read the grid
Four things make a matrix like this easy to read once you know them.
- The diagonal is zero. Every language is at distance 0 from itself, so the diagonal is the one stripe that never changes. It also tells you where to start reading.
- The grid is a mirror image. The distance from A to B equals the distance from B to A. In this data all 171 pairs match both ways, so you only ever need to read one half.
- Colour carries the score. Look for the light and dark extremes first, then for blocks where neighbouring cells share a shade. Blocks are groups of languages that sit close together.
- Row and column order matter. A heatmap only shows groups if related languages are next to each other, so sort the rows before you judge the pattern.
What the numbers say
The scores here run from 0 to 51. Across the 171 distinct pairs the median is 36 and the average is 33.8, so a typical pair sits in the middle of the scale. Only 12 pairs score below 20, while 40 score 40 or higher.
| Pair | Distance | What it means |
|---|---|---|
| Da and Nb | 4 | The closest pair in the whole grid |
| Af and Du | 7 | The second closest pair |
| Fa and Ic | 10 | The third closest pair |
| En and Sc | 13 | The fourth closest pair, well below the median of 36 |
| Lu and Ic | 51 | The largest score in the grid |
| Lu and Sr | 51 | Tied for the largest |
Two readings stand out. First, a few pairs are very close and most are not: the closest pair scores 4 and the median is 36, nine times higher. Second, some languages have no close relative in this set. Sr's nearest neighbour is at 36, and Yi, Lu and Wf have nothing closer than 30, 27 and 26. A table would hide that. In the heatmap those languages show up as rows with no light cells.
When a heatmap is the right chart
Use a heatmap when your data is a grid of two categories with a number in each cell: languages by languages, products by regions, hours by weekdays. It works best when:
- you have somewhere between about 10 and 40 categories on each side, enough to see patterns but few enough to label;
- the cells are all measured on the same scale;
- you care about clusters and outliers more than exact values.
If you need exact values, add a table or turn on value labels. If you only want to rank a handful of items, a bar chart is clearer. If the story is flows between categories rather than distances, a chord diagram can fit better.
How to build a distance matrix heatmap
Here is the short version. The full walkthrough is in our guide to creating a heatmap.
- Put your data in long format, with one row per cell: the row label, the column label and the value. This chart's sheet has 361 rows, one for each of the 19 × 19 cells.
- Upload the CSV and bind the row, column and value fields.
- Sort the labels so related items sit together.
- Choose a sequential colour scale from light to dark, with the diagonal included, and add the source in the footer.
Colour choices matter more than they seem. A scale that changes too quickly in the middle makes ordinary differences look dramatic, which is covered in the hidden bias in colour scales. The heatmap chart maker lets you set both the scale and the missing-value style.
Frequently asked questions
What is lexical distance?
Lexical distance is a measure of how different the vocabularies of two languages are. It counts differences in words, so it says nothing directly about grammar, pronunciation or mutual understanding.
Why is the diagonal of the heatmap zero?
Each cell on the diagonal compares a language with itself, and a language has no vocabulary difference from itself. The zero diagonal is also a useful anchor for judging the scale.
Why is the heatmap symmetric?
The distance from language A to language B is the same as from B to A, so the grid is a mirror image across the diagonal. In this chart every one of the 171 pairs matches both ways.
How many languages does this chart cover?
The chart covers 19 languages, which gives 361 cells and 171 distinct pairs. It is a subset, not a map of every language in Europe.
Can I make a heatmap from a spreadsheet?
Yes. Export your sheet from Excel or Google Sheets as a CSV, upload it to PlotSet and bind the row, column and value fields. See how to create a heatmap for each step.
