Tagless dialogue for the rest of us—why fixing attribution is fixing the wrong problem.
Most dialogue advice boils down to “use ‘said’ because it’s invisible” or “try ‘whispered’ and ‘exclaimed’ for variety.”
I’m about a month in on my first book. Here’s my take:
The real play is making tags unnecessary in the first place.
I’ve been stress-testing how far I can push tagless dialogue before a reader loses the thread.
Here’s what I’ve figured out.
Voice Is a Fingerprint
If you cover up the dialogue tags and can’t tell who’s speaking, you have a voice problem. I’m not saying you can NEVER use “he said,” but…
Every character needs a verbal signature so distinct that the reader hears them before they see the name. In my book, those signatures look like this:
L speaks in fragments. Clipped. Profane. Single-word commands and questions she already knows the answer to. “That sucks.” “Run.” “A little.”
M (the POV character) spirals. Medium sentences that get longer under stress. Anxious cataloguing of details. He doesn’t say “I’m scared” — he inventories every reason he should be.
D Exclamation marks! Yes, even though some famous people said not to use them. Whatever. Cheerful non sequiturs during life-threatening situations. “Yahoo! I’m super happy to hear that!” — while someone is actively bleeding.
R speaks in measured, longer sentences oscillating between warmth and menace. Dry humor and dad jokes wrapped in polite suburban language. The kind of person who says “It’s time for you to go home” and you’re not sure if that’s hospitality or a threat.
The test is simple: if two characters could swap a line and it still works, one of them may not have a voice yet. If you can swap ANY of their lines and it still works, you definitely have a voice problem.
Sentence Length Is Identity
This one isn’t immediately obvious, but your reader’s “ear” picks it up immediately.
Different characters think in different lengths. Fragments. Spiraling run-ons. Enthusiastic bursts with extra words bolted on. Deliberate, complete compound sentences.
The reader learns each character’s rhythm and starts identifying them by cadence before content. They feel who’s talking before they process what’s being said.
This means you can drop a line of dialogue with zero attribution, and if the sentence length matches the character’s established pattern, the reader already knows.
Action Beats Kill Two Birds
The simplest tagless technique: replace “said” with a physical action on the same line.
Instead of:
“What floor?” L said.
You write:
L spewed cigarette smoke toward the overhead speaker and crunched another button on the control panel. “What floor?”
The action tells you who’s speaking AND shows their emotional state. Two jobs, zero tags.
This works for every character if you give them distinct physical habits. L smokes and breaks things. R leans against walls and fiddles with keys in his pocket.
The action IS the tag.
Vocabulary as a Fence
Each character should own certain words that nobody else uses.
L curses. D says “yahoo” and “haha.” M uses mundane terms to describe otherworldly phenomena. R says “neighborly” and uses domestic language to downplay danger.
Every character should have verbal tics you can use where attribution would otherwise be murky.
This creates invisible fences in your dialogue. When the reader sees “yahoo” in a line of dialogue, they’re already with D before their eyes move to the next sentence. When they see “shit,” they’re with L.
Response Patterns Matter More Than Content
Here’s one I didn’t see very often in craft books:
HOW a character answers a question matters more than WHAT they say.
L deflects with counter-questions, without answering the original question. M answers and immediately asks his own question. D answers literally and completely, then adds something irrelevant. R answers with a statement that closes the conversation — “It’s time for you to go home.”
Once these patterns are established, you can drop a three-line exchange with zero tags and the reader tracks every speaker by their response pattern alone.
The Two-Person Rule
In a two-person dialogue, you almost never need tags after the first exchange.
Establish who speaks first with one tag or action beat. The alternating rhythm carries the rest. Only re-tag when a long action beat or internal monologue interrupts the volley and the reader might lose track.
Two characters talking is a tennis match. The reader’s head turns back and forth automatically. Don’t interrupt that with unnecessary attribution.
The Three-Plus Problem
Three or more speakers is where things get tricky.
Start with action tags for every speaker change. Once the reader has the rhythm, drop some. When a new character enters the conversation, tag them immediately. When a character who’s been silent suddenly speaks, tag them.
The character who speaks least needs tags most, because their voice is least established in the scene. The loud, distinctive character can go pages without attribution. The quiet one needs more tags.
Silence Is a Tag
A character NOT responding is attribution for the next character who does.
“L said nothing” followed by a line of dialogue tells the reader someone else is filling that silence. The reader tracks who is present and who hasn’t spoken.
And silence from certain characters carries more weight than speech from others. L refusing to answer a question tells you more than M talking for three paragraphs.
When You Actually Need Tags
I’m not saying never use them. I use them a lot. Here’s when you should:
- The first line of a scene where the reader doesn’t know who’s present.
- When more than two characters are rapid-firing and the rhythm could confuse.
- When a character speaks in a way that contradicts their established voice and the reader needs confirmation.
- When a long passage separates two lines of dialogue from the same conversation.
And when you do use one, try a character’s name in an action beat rather than “he said.” It does more work.
The Golden Rule
Write the dialogue. Read it aloud. If you can’t tell who’s speaking without the tags, the problem isn’t missing tags.
The problem is the voices aren’t distinct enough.