Spaces, Half-Spaces, and the Zero-Width Non-Joiner

Persian uses two different invisible boundary characters. An ordinary space separates words; a zero-width non-joiner (ZWNJ, U+200C), commonly called a “half-space,” prevents a cursive join inside one orthographic word. They may create similar-looking gaps, but they carry different structural information.

That distinction matters far beyond typography. Search, spell-checking, line breaking, copy-and-paste, and automated text processing all see the underlying characters. Standard Persian therefore treats boundaries as part of spelling: می‌روم mi-ravam “I go” contains one word with a ZWNJ, while کتاب خوب ketâb-e khub “a good book” contains two words separated by a space.

Three reasons a Persian line can appear to break

Before deciding that a visible gap is a word boundary, ask what produced it:

The third row is the trap. ا is one of the seven letters that cannot connect forward in Persian, so دانشگاه has a visual break after it even though the word contains neither a space nor a ZWNJ there. The seven non-connecting letters explain why glyph separation and word separation are not the same thing.

What a ZWNJ actually does

ZWNJ has no printed width of its own. It tells the shaping system not to connect two letters that could otherwise join. Unlike a space, it does not turn the material on either side into separate syntactic words. In می‌روم, for example, می‌ is an imperfective prefix attached to the verb روم. The half-space reveals the morpheme boundary while keeping the verb one word.

The same principle applies to many productive suffixes. Standard کتاب‌ها “books” and بزرگ‌تر “bigger” keep the stem recognizable and block an unwanted join at the morphological seam. The exact grammatical distribution of these affixes belongs on the pages about verb prefixes and plural and comparative spelling; the present page focuses on what the boundary character contributes.

Spaces separate independent words

An ordinary space belongs between independent words, including short grammatical words. The conjunction و va “and,” the preposition به be “to,” and the object marker را râ are written separately in the written standard. A ZWNJ is not a “smaller-looking space” that can replace every ordinary space.

Persian morphology does not line up perfectly with English spacing. English writes do not as two words but books as one solid word. Persian writes نمی‌روم with an internal ZWNJ and کتاب‌ها with another. The underlying logic is not “copy English spaces”; it is “mark Persian wordhood and productive affix boundaries according to Persian convention.”

Final he and a boundary you cannot hear as silence

After a word ending in silent ه, standard house style writes ezafe with ه‌ی, as in خانه‌ی ما “our house.” The ZWNJ prevents an inappropriate join before ی, but the boundary is not a pause: the linker is pronounced -ye.

The second example has نامه‌ای, where an ending follows نامه inside one word. A full space would incorrectly split that inflected form. This is one reason the ZWNJ is morphological rather than merely decorative.

Copy-and-paste integrity

Because ZWNJ is invisible, two strings can look nearly identical while differing in code points. Deleting, normalizing, or replacing it with a space can make a standard word nonstandard, change search results, or stop an exact match. Conversely, sprinkling ZWNJs into every visible gap damages text: the natural break after ا in دانشگاه needs no boundary character.

For reliable checking, place the cursor at the suspected seam and move or delete one character at a time, or use an editor that reveals invisible characters. Search separately for the correct word if copied text behaves oddly. Do not rely on visual width: fonts and line justification can make a full space, a ZWNJ-controlled join break, and a natural non-join look deceptively similar.

Register and real usage

ZWNJ is expected in edited (written-standard), (formal), and (academic) Persian. Newspapers and publishing houses may differ on a limited set of compounds, and older print can show spelling practices from before modern standardization. Those are genuine editorial variations, not permission to ignore ZWNJ in productive می‌/نمی‌, plural ‌ها, or ordinary comparative ‌تر forms taught here.

Informal messages often omit half-spaces because of keyboard friction (informal), and older software sometimes turns them into full spaces. That explains forms learners will encounter; it does not make those encodings preferable for careful writing. Colloquial Tehran pronunciation can accompany either careful or casual digital spelling, but pronunciation alone does not decide the Unicode boundary.

Common Mistakes

1. Replacing the half-space with an ordinary space. This incorrectly divides one verb into two graphic words.

2. Omitting the ZWNJ and allowing an unwanted join. The spelling hides a productive boundary.

3. Using a visible hyphen as a substitute. A Latin hyphen is a printed mark, not U+200C.

4. Inserting a space at every visible cursive break. The break after non-connecting ا is already supplied by the letter.

5. Importing Arabic letter code points while fixing spacing. Correct boundaries do not repair the wrong alphabetic encoding.

Key Takeaways

Use an ordinary space between independent words and U+200C inside a word where standard Persian blocks a join at a morphological boundary. Do not mistake a natural break after a non-connecting letter for either one. Persian spelling does not record every sound, but it does record word and morpheme structure systematically—and invisible characters are part of that system.