Punctuation in a Right-to-Left Text

Modern Hebrew punctuation is borrowed. The full stop, comma, question mark, exclamation mark, colon, brackets and quotation marks all entered the language with the nineteenth-century press and the Haskalah; classical Hebrew had a different system entirely, in which a verse ended with a sof pasuq ׃ and the phrasing was carried by cantillation marks rather than by commas.

So the glyphs will look familiar. What is unfamiliar is that they now live in a line that runs the other way, and a few of them change shape because of it.

Paired marks mirror — but you still type the opening one first

Put a parenthesis in a Hebrew sentence and the bracket that opens the parenthetical is drawn on the right, because that is where the parenthetical starts. The same goes for square and curly brackets and for angle brackets.

Here is the mechanical fact that nobody tells you, and it saves a great deal of grief: the character you type is still the opening one. Unicode calls U+0028 LEFT PARENTHESIS, but that name describes the shape it takes in a left-to-right context. In a right-to-left run the renderer mirrors it and draws the other shape. The stored character does not change; only the picture does.

The consequence is a rule to follow blindly: type the parenthetical in the order you say it — open, content, close — and let the display sort itself out. If you "correct" it by typing the closing bracket first because that is the shape you see on the right, the string is now genuinely wrong. It will look fine in that one app and fall apart the moment the text is copied into a different one, or read by anything that parses brackets.

Quotation marks, and the mark that is not one

Everyday Hebrew quotes with a straight double mark at both ends: היא אמרה "אני עסוקה". Typographic pairs exist and turn up in books and careful design, but usage is inconsistent enough that you should be prepared for either.

The trap is that this mark has the same shape as the gershayim ״ that sits inside an acronym. In plain text a quoted word and an acronym can look confusingly alike. One rule separates them cleanly:

One convention differs from American English and is worth taking on deliberately: Hebrew puts the sentence-final full stop outside the closing quotation mark, as British practice does.

The sentence that ends in a number

This is the case that makes learners think their keyboard is broken, and it deserves the actual mechanism rather than reassurance.

Every character in a bidirectional text belongs to one of three classes. Hebrew letters are strong right-to-left. Latin letters are strong left-to-right. Digits are weak — they run left to right internally, and when the algorithm decides what to do with a neutral beside a number it treats the number as though it were right-to-left. And punctuation is neutral: a full stop, comma, colon or bracket has no direction of its own and takes one from its neighbours.

A neutral character sandwiched between two runs of the same direction joins them, which is why a comma in the middle of a Hebrew sentence never misbehaves. The problem is a neutral at the edge. The period after that 4 has a weak digit on one side and the end of the paragraph on the other, so there is nothing to copy — and in that case the algorithm falls back to the paragraph direction. In a properly Hebrew paragraph that is right-to-left, and the period is drawn at the far left, where the sentence ends.

Change the paragraph direction and the same stored string draws differently. A chat box, a spreadsheet cell or a form field that has not been told it is Hebrew usually guesses from the first strong character it meets: a message beginning with a Hebrew word gets an RTL paragraph, and a message beginning with a Latin brand name gets an LTR one — and then the trailing period appears next to the number instead. Nothing about your text changed. The container's direction did.

There is also the solution Israelis actually use most of the time, which is simply not to end the sentence on a number:

A Hebrew word after the digits gives the period a strong right-to-left neighbour and the ambiguity never arises. Watch for it and you will notice how often Israeli writing puts a unit, a weekday or a qualifier after a figure rather than letting the figure sit at the end.

Where Hebrew punctuation is genuinely looser

The marks are the same; the rules for using them are not as tight as English ones. The Academy of the Hebrew Language publishes כללי הפיסוק, the rules of punctuation, but they are advisory and widely ignored, and Israeli editing tolerates far more variation than an English style guide would.

• The comma before ש is optional. English requires nothing before that, and Hebrew likewise leaves it out in ordinary writing — but you will also see it written with a comma, and neither is an error.
• No comma between subject and predicate, even when the subject is long.
• Lists take commas, and the last item is joined with ו and no comma before it — there is no serial comma in Hebrew.
• The colon introduces a list or a quotation, exactly as in English.
• The dash — a long mark with spaces around it — introduces a clause or an afterthought. It is a completely different character from the maqaf, which glues words together.
• Question and exclamation marks behave normally, and Hebrew has no inverted opening marks.

Common Mistakes

1. Typing the closing bracket first because that is the shape you can see. The display is mirroring the character; the character itself is unchanged.

2. Putting the full stop inside the quotation marks. American English does this; Hebrew does not.

3. Using the maqaf as a dash. The joiner and the clause-separating dash are different characters with opposite jobs.

4. Deleting a mark that "looks wrong" and retyping it. This is the expensive one. The rendering is not the storage: the period after a number is already in the right place in the string, and moving it to where it looks right puts it in the wrong place logically. The text will then be broken everywhere except in the one window you fixed it in. If a line genuinely renders wrong, insert an RLM instead of rearranging characters.

5. Concluding that pasted Hebrew is corrupted. Paste a Hebrew sentence into a field or a document that is set to left-to-right and the punctuation jumps, the brackets flip and any trailing number moves. The characters are intact; the container is telling the algorithm to lay them out as a left-to-right paragraph. Set the field's direction — or start the line with a Hebrew word, which is what a direction-guessing field looks at.

Key Takeaways

• Hebrew uses the European punctuation set, adopted in the nineteenth century; classical Hebrew ended a verse with sof pasuq ׃ instead.
• Paired marks mirror on screen but not in storage. Type the opening bracket first, always; U+0028 renders as the right-hand shape in a Hebrew line by itself.
• Gershayim sits inside a word; quotation marks flank it. Hebrew letters on both sides of the mark means it is spelling, not punctuation.
• The full stop goes outside the closing quotation mark.
• Punctuation is direction-neutral: at the edge of a line it takes the paragraph direction, which is why a trailing period lands on the left — and why the same string renders differently in a container that guessed the wrong direction.
• RLM (U+200F) is the deliberate fix. Never delete and retype a mark that merely looks misplaced.
• Hebrew comma practice is looser than English: the comma before ש is optional, there is no serial comma, and the Academy's rules are advisory.