Writing in both scripts

Lesson opening

This is the first part of the lesson. The rest, the exercises and the reviews are in the app.

Two scripts in one text: the conventions of switching

Where each script lives you know from “Domains of use”; reading both, from “Digraphia in practice”; how Serbian remakes foreign names, from “Transcribing foreign names”. What remains is the grown-up part: what happens to a whole text when it changes script. A text does not change its clothes all at once, and knowing how to carry each part across is an editor’s job; in Serbia everyone does it who writes an application, runs a website or hands in a contract.

The conversion is not symmetrical

The first thing anyone who converts texts learns: the two directions are not equal. From Cyrillic to Latin a text goes mechanically: thirty Cyrillic letters map onto twenty-seven single Latin letters plus three letters that Latin writes with two signs, the digraphs lj, nj, dž, and a table is enough. The way back is harder, and those three digraphs are why.

Here is the catch. Latin nadživeti “to outlive” is built from the prefix nad- “over” and the verb živeti “to live”: here d and ž are neighbours, not a pair. In Cyrillic it is надживети, with two letters, дж. Yet exactly the same string of Latin letters would come from a Cyrillic “наџивети”, if such a word existed, because the single letter џ also becomes dž. So the Latin spelling is poorer than the Cyrillic one: it has lost the border between prefix and root, and only someone who knows the word can put it back. English has the same trap in hothouse: you read t + h, not the th of the, because you know the word. A converter does not.

Hence a conclusion worth learning whole: Cyrillic to Latin can be done by a machine; Latin to Cyrillic only by someone who understands the text. So if a text is needed in both scripts, keep the master copy in Cyrillic: Latin comes out of it in a second, while the way back always loses something.

Where a prefix meets a root: the place automation breaks

Words where d and ž (or n and j) stand side by side but belong to different parts of the word are few, and they are exactly the ones that spoil machine conversion. They are built alike: a prefix ends in a consonant, the root begins with the second letter of the pair.

The test is simple: mentally take the prefix off. If a living word remains (živeti, jezički), you are looking at a seam, and Cyrillic writes two letters: дж rather than џ, нј rather than њ. If nothing remains (džep “pocket”, njiva “field”), it is a true digraph and a single Cyrillic letter. For the Latin loans the check is the international word (in-jection, con-junction, con-jugation); Tanjug you simply learn. For the pair l + j, such seams practically never occur in Serbian words: it is almost always a single letter, ljudi “people” is људи.

A keyboard trap too: without a Serbian layout it is easy to slip a Latin a, e, o or j into a Cyrillic word. On screen Cyrillic ја and Latin ja look the same, but they are different characters, so search, spell-check and the converter all stumble. Set up a Serbian Cyrillic layout (“Digraphia in practice”).

Cheat sheet

The cheat sheet — everything essential from the lesson on one page. In the app it opens at any time.

Two scripts in one text: cheat sheet

Пресловљавање (preslovljavanje, transliteracija) changes the script, not the language. Three layers: Serbian words (letter for letter), foreign names (by sound), machine labels (hands off).

Prefix seam = two letters:

Test: take the prefix off. A word remains (živeti, jezički) → дж, нј. Nothing remains (džep “pocket”, njiva “field”) → џ, њ. l + j at a seam: practically never (ljudi is људи).

Foreign names: by sound, not by letters:

Latin allows both (prilagođeno “adapted” and izvorno pisanje “original spelling”), though newspapers usually spell by sound; Cyrillic only the adapted one (a rare name may add the original in brackets). Back: Шекспир → Šekspir, Њујорк → Njujork, never the original.

Latin even inside Cyrillic:

• brands, models in documents, on goods, usually: Google, Coca-Cola, BMW X5, iPhone 15 (newspapers: also Гугл);
• web/email addresses, files, always: www.politika.rs, ugovor-2026.pdf;
• tech and formats, usually: IT, PDF, USB, HTML, GPS, Wi-Fi;
• units, formulas, codes: kg, km/h, °C, H2O, EUR, RSD;
• Roman numerals: XXI век, IV издање;
• but everyday derivatives in the text’s script, small letter: гуглати, мејл, јутјубер.

Numbers and typography never change (not 1,250.00); only a month written as a word does. Abbreviations change with their endings; IT, PDF usually stay.

Working order: typography → mechanical conversion → prefix seams → names by sound → Latin islands → abbreviations with endings (МУП-а) → read heading and names aloud.