Writing Detective & Noir Fiction
Week 2 of 12

Hard-Boiled Scene Craft

Clean Prose, Sharp Dialogue, and Visible Behavior

Phase 1 — The American Engine
Craft Target Scene as Observable Action
Key Authors Hammett, Leonard
AI Lab Diagnostic Mirror
Portfolio Case Bible updated
Homework 6–8 hours
01

Lecture — Hard-Boiled Scene Craft

From Week 1: You defined your story's contract — the promise you make to the reader about what kind of question the story asks and what kind of fairness you owe. You wrote a contract paragraph and a Voice + Code scene. This week, you learn to build scenes that earn that contract line by line. Your case bible, started last week, grows with every scene you draft.

Last week you chose a contract. This week you learn to build the scenes that honor it. If contract is the architecture, scene craft is the bricklaying — and in hard-boiled fiction, there is a specific way to lay brick. The practical mechanism is blunt: behavior first. Hard-boiled scenes are constructed from observable action and audible speech. Inner commentary is not forbidden, but it is rationed, deployed the way a sniper deploys ammunition — one shot, placed for maximum effect, and then silence. Everything else must be visible, hearable, or physically felt.

Hammett remains the essential model because his prose achieves something that sounds simple and is ferociously difficult: it removes the narrator's interpretive apparatus and forces the reader to do the interpreting. The NEA's teaching materials on The Maltese Falcon attribute the novel's readability and momentum to exactly this quality — clean prose and precise dialogue driving the story forward without authorial explanation. Britannica reinforces the point, noting that the novel's sustained tension comes from vivid scenes and what it calls spareness of style. But spareness is not thinness. Hammett's scenes are dense with information. They are just ruthlessly selective about what kind of information is allowed on the page. What you see, hear, and can touch: yes. What the narrator thinks you should feel about it: almost never.

The test is crude but reliable: write scenes that can be filmed. If a camera could not capture it — if the information exists only in the narrator's unspoken thoughts — you should question whether it belongs in a hard-boiled scene. This does not mean interiority is banned from crime fiction. It means interiority must earn its place. When a hard-boiled narrator pauses to reflect, the pause itself is an event. The reader notices because it is rare.

Scenes in detective and noir fiction generally perform one of three jobs. An intake scene is where someone brings a problem — a client walks in, a body appears, a phone rings with bad news. A pursuit scene is where the investigator chases truth through a hostile system — knocking on doors, running records, asking questions that make people uncomfortable. A squeeze scene is where pressure is applied until someone breaks — an interrogation, a confrontation, a moment when the investigator forces a choice. All three types share a governing principle that you must internalize now and carry through the rest of this course: each scene must change the investigation's state. A state change can be a new clue discovered, a false lead confirmed, a relationship damaged, a new threat introduced, or a code compromise forced upon the protagonist. If you finish writing a scene and cannot articulate the state change in a single sentence, the scene is probably decorative. In a detective or noir story, decoration is dead weight.

If you finish writing a scene and cannot articulate the state change in a single sentence, the scene is probably decorative. In a detective or noir story, decoration is dead weight.

Now we reach dialogue, and this is where most emerging crime writers go wrong. Hard-boiled dialogue does not "sound cool." It does work. In hard-boiled fiction, dialogue is a weapon used to establish status, test for weakness, and deny information. Every line of dialogue your characters speak should perform at least one of these actions: threaten, evade, bait, confess, lie, bargain, insult, seduce, or recruit. If a line of dialogue is merely informational — if it exists only to convey facts to the reader — you have two choices: move the information into subtext (let the reader infer it from what is not said) or turn the exchange into a conflict (make the information contested, reluctant, or dangerous to share).

The practical tool for learning this is the dialogue ledger. Take any exchange you have written and, beneath each line, write two things: what the speaker wants and what the speaker is hiding. This is especially important in noir because Chandler's world-model — the system of corruption you learned about in Week 1 — implies that people do not speak truth for free. Testimony is dangerous. Witnesses have reasons to lie. Officials have reasons to redirect. Even allies have information they will withhold to protect themselves. When your protagonist extracts truth from a reluctant mouth, that extraction is not a conversation. It is a plot engine.

Leonard extends this principle to its sharpest point. His reputation as a dialogue master rests on a specific discipline: his characters reveal hierarchy, fear, and motive without the writer ever resorting to adverbs or dialogue tags that interpret. No one "said angrily" in a Leonard scene. Instead, anger is embedded in what is said, how it is structured, and what it leaves out. This is not stylistic minimalism for its own sake. It is the hard-boiled principle applied at the sentence level: let the reader do the work. If you have written your dialogue well, the reader does not need to be told the character is angry. They feel it.

Hard-boiled description follows the same logic. When a hard-boiled narrator describes a room, the details are chosen to signal class, vice, violence, or hypocrisy. But the best descriptions also indicate psychological state without announcing it. A narrator might note that the office is immaculate but the wastebasket is overflowing, or that the expensive painting hangs next to a patched wall. The reader learns the narrator's judgments alongside the physical facts. This is where setting becomes a character: the objects in a room reflect its moral economy. The important difference between noir atmosphere and noir observation is that observation has consequences. You will use those details later — as evidence, as motive, as thematic pressure. A detail that never pays off is a detail that never deserved the sentence it occupied.

To avoid parody, you need one more tool: tone calibration. Many emerging writers believe noir equals relentless snark — a narrator who wisecracks through every scene, who treats danger as material for one-liners. But snark flattens stakes. The noir voice can be humorous, but humor should usually be defensive: the character uses it to stay alive emotionally, to keep distance between themselves and a world that wants to swallow them. When a line becomes a joke the writer wants rather than a tactic the character needs, the scene loses authenticity. Chandler himself warned that simply having characters talk tough does not make a writer serious. You must separate surface markers — slang, smoke, cynicism — from structural realism: credible motives, plausible behavior under pressure, and institutions with actual power.

Finally, connect scene craft to your case bible. Every scene you write this week should be logged in your timeline: what changed, what clue was planted, what relationship shifted. This is how you transform what feels like "writing cool noir scenes" into the disciplined work of building a solvable, consequential investigation. The genre demands both pleasures simultaneously: you seduce the reader line by line with voice and texture, and you deliver an intelligible logic chain underneath. The seduction without the logic is decoration. The logic without the seduction is a police report. You are building neither. You are building fiction that works.

02

Readings

Reading 1 — Primary
Dashiell Hammett, excerpt packet (selected chapters from The Maltese Falcon and/or Red Harvest)
Craft reason: Demonstrates the "observable action only" method — scenes built from behavior and dialogue with almost no authorial interpretation.
Reading 2 — Contextual
NEA Big Read, Introduction to The Maltese Falcon
Craft reason: Teaching overview that frames clean prose and dialogue precision as tension mechanisms — not stylistic choices but structural ones.
Reading 3 — Primary
Elmore Leonard, excerpt packet (2–3 scenes)
Craft reason: Dialogue as action, wit as weapon, violence deployed with control — the modern heir to Hammett's behavioral method, with the adverb prohibition in full force.
Reading 4 — Optional
Library of America, description of Hammett's Black Mask stories
Craft reason: Historical framing of the hard-boiled voice as a literary event — the "jarring textures" and "revved-up cadences" that signaled a new kind of American prose.
03

Writing Assignments

Short Drill · 30–40 min

The Dialogue Ledger

Write 20 lines of dialogue between two characters in an interrogation, negotiation, or confrontation. No scene-setting, no narration — pure speech. Then annotate each line: beneath each speaker's words, write (a) what the speaker wants in this moment and (b) what the speaker is hiding. The annotations are not part of the story. They are your X-ray of subtext. If any line has no want or no hide, rewrite it until it does.
Longer Homework · 3–4 hours

Intake Scene + First Turn

Write a scene of 1,500–2,000 words in which your protagonist encounters the crime or receives the case. This is your intake scene. By the end of the scene, the investigation must have taken its first turn — a new piece of information, a lead, a contradiction, a door slammed shut. Include 2 clues planted invisibly: details the reader should be able to find on a careful reread but should not consciously register on a first pass. Use the behavioral method: build the scene from action and dialogue. If you catch yourself writing what a character thinks or feels, ask whether you could show that thought or feeling through what they do or say instead. Update your case bible with the scene's state change, the planted clues, and their intended function.
04

AI Lab

Phase 1 · Diagnostic Mirror

AI as Diagnostic Mirror

Guardrail: AI diagnoses. You decide. AI never writes your scenes. The diagnostic prompts below help you see what your scene is actually doing versus what you intended. Revision decisions remain yours.
Prompt 1 — Scene Anatomy
Here is my intake scene. Analyze it for three things: (a) What is the scene's state change — what does the investigation know or risk at the end that it did not at the beginning? (b) In each dialogue exchange, what subtext is operating — what do the speakers want and what are they hiding? (c) Flag any lines that feel like "writer performance" rather than character behavior — moments where the prose calls attention to itself rather than serving the scene.
Expected output: A scene map identifying the state change in one sentence, a subtext analysis for each major exchange, and 3–5 flagged lines with brief explanations of why they read as performance. Use the flagged lines as your primary revision targets.
Prompt 2 — Tension Rewrite
Select the two weakest dialogue exchanges in my scene — the moments where tension drops or information is delivered without conflict. For each, rewrite the exchange to increase tension while preserving the same factual information. Then explain what tension mechanism you used (status shift, information denial, implied threat, forced choice, or another technique).
Expected output: Two rewritten mini-exchanges plus a brief explanation of the tension mechanism in each. You do not paste these into your draft. You study the mechanism, then apply it yourself in your own voice. The rewrite is a demonstration, not a replacement.
Prompt 3 — Clue Visibility Audit
I planted 2 clues in this scene. Without me telling you what they are, read the scene and identify any details that might function as clues — physical objects, behavioral contradictions, environmental details, or statements that could carry more weight than they appear to on first read. List every candidate you find, then rate each on a scale of 1 (completely invisible) to 5 (too obvious). I want my planted clues to land around 2–3.
Expected output: A list of candidate clues with visibility ratings. If the AI finds your planted clues easily (rating 4–5), they are too visible and you should bury them deeper. If it misses them entirely, they may be too buried or too arbitrary. If it identifies details you did not intend as clues, consider whether those accidental clues are interesting enough to promote.
What you should learn from this exchange: The AI reads your scene as a first-time reader with no knowledge of your intentions. Where its reading diverges from your plan — where it misses clues, misreads subtext, or identifies performance you did not notice — that divergence is your revision map. The goal is not to make the AI "right." The goal is to use its misreadings as data about how your prose actually communicates.
05

Assessment Focus

Scene State Change
30%
Dialogue Subtext
30%
Descriptive Function
20%
Clarity & Pace
20%
06

Wow Element

Two-Pass Scene Surgery

This exercise teaches the hard-boiled principle by making you feel its constraints in your hands.

Pass One: Write a scene of 400–600 words using any method you like — interior monologue, omniscient narration, emotional exposition, whatever feels natural. Let the narrator think, feel, judge, and explain. Get the scene down.

Pass Two: Rewrite the same scene as if it must be performed on a bare stage with no voiceover. No internal narration. No "she felt" or "he thought." Only what an audience could see and hear: action, dialogue, physical gesture, environmental sound. Everything else is cut.

The Comparison: Place the two versions side by side. What was lost? What survived? What — and this is the surprise most students encounter — actually got better when the scaffolding came down? Often the emotion is more powerful without explanation, the tension sharper without interpretation, the character more vivid when the reader has to infer rather than receive.

This is not an argument that interiority is always wrong. It is proof that interiority must be earned. The hard-boiled method is the default. Deviations from it are deliberate acts — and the best ones land harder because of the restraint surrounding them.
07

Portfolio Tracker

Updated This Week
Case Bible
Add intake scene to timeline · Log state change · Record 2 planted clues and their intended function in reveal grid
Weekly
Craft Reflection Log
100–200 words: what did the behavioral method reveal about your writing habits? What did the dialogue ledger expose?
Week 4
Moral Ledger
Locked — introduced in Week 4
Week 5
System Map
Locked — introduced in Week 5
Week 6
Network Dossier
Locked — introduced in Week 6
Week 12
Final Portfolio Piece
Locked — 4,000–6,000 words + 750-word craft reflection
08

Estimated Homework Time

6–8
hours total
Reading: 2–3 hrs · Writing (drill + homework): 3–4 hrs · AI Lab + Craft Reflection: 1 hr
Next Week
Week 3: Fair-Play Engineering and the Unreliable Narrator
The engineering week. You will learn the fair-play doctrine as a craft technology, build a complete reveal grid, design solvable puzzles — and then confront the question of what happens when the narrator is the one hiding the truth. Van Dine sets the rules. Christie breaks them. Thompson makes you wonder if they were ever real.