WORDS THAT TRAVELED
Arabic
Three languages. One traveling word.
Arabic · Turkish · Spanish
Words can travel much farther than the people who speak them.
Arabic, Turkish and Spanish belong to different language families. But centuries of trade, migration, conquest and cultural exchange have left traces of one language inside another.
Some words moved almost directly. Others crossed continents, changed pronunciation, picked up new spellings and passed through several languages before reaching the forms used today.
The similarities are not because these languages are siblings. They are evidence of contact.
One word, four stops
Coffee traveled before it became café.
The journey begins with the Arabic word قهوة, transliterated as qahwah.
The word entered Turkish as kahve.
From Turkish, it traveled into Italian as caffè.
Spanish eventually adopted café.
Arabic as a bridge
Arabic wasn't always where the journey began.
Some words arrived in Arabic after already traveling across languages. Arabic then carried them farther.
The story begins with Sanskrit nāraṅga.
The word moved into Persian as nārang.
Arabic carried the word west as نارنج, nāranj.
In Hispanic Arabic, the form became naranǧa.
Spanish inherited the word as naranja.
What changes in transit?
Some borrowed words barely changed. Others became harder to recognize.
Arabic loanwords in Turkish adapted to a different sound system and writing system. Some retained almost the same written form. Others changed much more.
Arabic maʿlūm entered Turkish as malum. Once transliteration marks are normalized, the written forms are almost identical.
kitāb became kitap. The structure is still easy to recognize.
waqt became vakit. The relationship is less obvious on the page.
And Arabic jawāb became Turkish cevap — one of the more visually transformed examples in this prototype.
A piece of grammar traveled too
Sometimes Spanish inherited more than the word.
In many Spanish words of Arabic origin, a form descended from the Arabic definite article al- became attached to the borrowed word.
Arabic al-maḵzan eventually became Spanish almacén.
Arabic al-qaṣr became alcázar.
Arabic al-quṭn became algodón.
But the pattern is not universal. Arabic as-sukkar became Spanish azúcar, where the article was assimilated rather than preserved as a visible al-.
Across the dataset
Not every borrowed word changed by the same amount.
Comparing normalized written forms shows a wide range: some Turkish loanwords remain close to their Arabic transliterations, while others look much less alike.
At first glance, these are simply Arabic-origin words used in modern Turkish.
But once we compare their normalized written forms, a pattern appears.
Some pairs remain very close: maʿlūm → malum scores 1.00 in this comparison.
Others diverge much more. jawāb → cevap scores 0.20.
The bigger picture
These words form a network of contact.
Some routes are direct. Others pass through several languages. Seen together, the dataset becomes less like a list of translations and more like a map of linguistic exchange.
Hover over a word or language to trace its connections.
Explore the words
Follow individual words across languages.
This prototype includes words connected through direct borrowing, longer transmission routes, and shared histories across Arabic, Turkish and Spanish.
Methodology
How these word histories were traced.
Word selection
The dataset was manually compiled for this project from word-level etymological research rather than downloaded from a single source.
The examples were selected to show different kinds of linguistic movement, including direct borrowing, multi-language transmission and changes in written form.
Routes
For each word, I recorded a source language and transmission route based on the etymological information collected during the research process.
A word appearing in Arabic, Turkish and Spanish does not necessarily mean it traveled directly from Arabic to Turkish to Spanish. The route field is therefore kept separate from the languages in which a form appears.
Written-form similarity
For words recorded with a direct Arabic-to-Turkish route, I compared the Arabic transliteration with the modern Turkish written form.
Selected transliteration marks and orthographic differences were normalized before calculating a string-similarity score from 0 to 1. A score of 1 means the normalized written forms match exactly.
The score is a comparison of written forms only. It does not measure pronunciation, meaning or overall linguistic distance.
Data and sources
For each word, the working dataset records the forms used in Arabic, Turkish and Spanish where relevant, along with source language, transmission route, relationship type, research notes and confidence information where available.
Etymological information and transmission routes were researched at the word level using dictionary and linguistic references recorded during the research process. Where an origin or route was uncertain, that uncertainty was retained in the notes rather than presented as definitive.
Geographic boundaries used in the interactive maps are drawn from World Atlas / TopoJSON.
Tools
Python and Jupyter were used to assemble and analyze the dataset. The interactive was built with HTML, CSS, JavaScript and D3.js, with World Atlas / TopoJSON used for geographic boundaries.
Every borrowed word carries a record of contact.