Roth, when asked whether all characters need to be likable, answers fast: No. I think you should potentially understand why they are what what makes them tick even if if you may not always understand it.1 Then he names the case he can't solve. You wonder for instance about those... two boys at Columbine who had seemingly a pretty normal life and decent parents. They worked at a pizza parlor and this and that. Why did they decide to go shoot all these people? And I don't know the answer.2
This is the villain's edge. The writer's job is not to make the villain comprehensible — fully comprehensible villains are domesticated villains, often disappointing on the page — but to bring the audience close enough to the inside of the villain that the question what makes them tick feels live. The villain Roth wants is intelligently evil like Silence of the Lamb. Hannibal Lecter you understand enough to bond with. You do not understand him fully. The unreached interior is what makes him grip.
The understandable villain is not the sympathetic villain. Roth is explicit: villains make great movies. If you can find a great villain, you're in good shape.3 He's not asking the writer to redeem the villain or to give them a sob story that explains the violence. He's asking the writer to make the villain's internal logic legible without making the violence reasonable.
The contrast he reaches for is the traumatized lash-out villain — the character who hits people because their hand was put in an oven. That, he says, is complicated: there's a causal chain you can follow. The harder villain, the better villain, is the one whose causal chain you can't follow but whose interior you can taste. The two Columbine boys, in Roth's framing, are interesting precisely because the available facts don't add up to the outcome. The writer who can render that gap on the page has a great villain.
Roth distinguishes implicit modes:
The traumatized villain. Backstory explains the present action. The audience understands the violence as response to inflicted harm. Empathy is the writer's tool, and the villain becomes a lashing victim. This works for many stories but produces a domesticated antagonist. The villain's actions are predictable once the backstory is known.
The opaque-interior villain. Backstory is insufficient to explain the present action. The audience experiences the gap between what's known about the character and what they did. The villain feels like a real psychological problem rather than a plot mechanism. Hannibal Lecter. The Columbine boys, if a writer could reach them on the page. The two-line speech where the villain reveals something but not enough.
The opaque-interior villain is harder to write because the writer must know the interior — must have done the imagining — and then withhold it from the audience in a way that produces felt presence rather than confusion. The writer who only withholds without first knowing produces a vague antagonist. The writer who reveals everything produces a domesticated one.
The understandable-not-likable threshold gives the vault's antagonist material a governing distinction it has been missing. The vault has multiple pages about manipulator profiling, dark-triad clinical literature, Greene's strategic operators, and Hughes' field-deployed influence tactics. None of those pages told the writer what the audience-side of a successful villain feels like. This page supplies that side: the audience must arrive at understanding without endorsement. The criterion is now installable across the vault's manipulator pages — a manipulator-on-the-page is craft-eligible when readers can name the tactic without rooting for the tactician.
The page also gives character as inherited determinism its complementary protagonist/antagonist reading. Roth's Sicilian-mafia-1860 villains are fated — they do what their position requires. Roth's protagonists are usually bewildered by the same mechanism. The split-by-position is what makes the inherited-determinism frame dramatic rather than philosophical. The villain is the inheritance-machine running cleanly; the protagonist is the inheritance-machine running with a glitch the protagonist cannot name. Villain-craft and protagonist-craft are the same engine in two states.
Third — the page gives the vault's dark triad page a craft-application. Until now the dark-triad page sat as a clinical-psychology reference. With this page it becomes a writer's instrument: each dark-triad subtype maps onto one of the three villain modes, and the writer can choose a mode by choosing a clinical profile. The clinical knowledge becomes structural craft.
Take the scene in Silence of the Lambs where Clarice first meets Lecter — the long approach down the corridor, the speech across the glass, the moth-detail offered as bait. The scene is roughly seven minutes. By the end of it Lecter has become the most memorable villain in modern film, and the audience can articulate why without endorsing him.
The Thomas Harris–Ted Tally writing achieves this through a specific application of Roth's understandable-not-likable principle. Three moves: first, Lecter is given a tactic the audience can name — psychological probing through Clarice's vulnerabilities, framed as offered help. The audience reads the tactic clearly. They don't endorse it; they recognize it. Second, Lecter's interior is shown as partly legible — he is articulate, cultured, attentive, and the audience can imagine the texture of his cognition without ever reaching its origin. Why he kills the people he kills is never explained. The withholding is precise. Third, the villain is given one detail the audience cannot square — Lecter's intervention on Clarice's behalf, late in the film, that the audience reads as neither sympathetic interest nor manipulation, and that the writers refuse to resolve. The unresolved residue is what makes Lecter durable.
Compare to a counter-case: a generic serial-killer antagonist whose backstory is fully explained (childhood trauma, head injury, neglected by mother). The audience now understands the violence too completely. The character can be filed and forgotten. Lecter cannot be filed. The unreached interior is the durability. Roth's I don't know the answer5 about the Columbine boys is the same epistemic posture applied to villain-construction.
The case study reveals that opaque-interior villain craft is a negative space discipline. The writer's job is not to render the interior but to make the audience feel the shape of an interior they cannot enter. Lecter's interior is glass behind glass. The audience can see the writer pointing at it. They cannot reach it. They cannot stop trying to.
You're writing the antagonist for a new project. Before drafting any scenes, you take ninety minutes alone. You write three pages of the villain's interior monologue in their own voice — what they think about while doing the thing the protagonist will eventually confront them about. These pages will not go into the script. They are the writer's private working document. You write them anyway because the audience will feel the difference between a withheld interior the writer has and a withheld interior the writer never bothered to build.
Now you draft the first scene the villain appears in. The villain says four sentences. The first sentence is operational — they say what the scene needs them to say. The second sentence is adjacent to their interior — a phrase or detail that hints at the unwritten three pages without revealing them. The third sentence is operational again. The fourth sentence — and this is the move that turns understandable-villain into great-villain — is contradictory to the interior. A small detail that does not fit the profile the audience has just begun to assemble. The contradiction is the durable residue. It is what the audience will remember and try to solve.
The protagonist responds. The villain leaves the scene. You re-read what you've written. You ask yourself: can the audience name the villain's tactic right now? If yes, the legibility is achieved. Can the audience endorse what the villain just did? If no, the moral distance is held. Is there one detail in the four sentences that the audience cannot place? If yes, the opaque-interior is functioning. All three conditions present = the threshold is crossed. If any one is missing, you rewrite the four sentences with the missing condition in mind.
You've collapsed the understandable-not-likable threshold when audiences quote the villain affectionately. There is a kind of fan reception in which the villain becomes a meme-favorite, charm-detached from menace, and the writer reads this as success. It isn't. The audience has migrated from understanding to endorsement. The villain has become a domesticated mascot. Roth's threshold required understanding without endorsement; affectionate quoting is endorsement. The repair is to write the next villain a level less charismatic and a level more disquieting.
The second sign — and the more common one — is the villain's backstory now explains the violence. You will know this has happened when, asked why the villain does what they do, you can give a three-sentence causal answer. A great villain resists the three-sentence explanation. Lecter does. The Columbine boys do. The moment you can fully explain the villain, the villain has become a complicated victim, which is a different and lesser character.
The third — the one Roth implicitly warns against — is the villain's behavior follows plot requirements rather than internal logic. You'll catch this by asking: would the villain make the choices they're making if the plot didn't need them to? If the answer is "the plot needs them to," the villain has become a mechanism, and the audience will detect this even if they can't name it. The repair is to throw out the plotted villain-scene and ask what the villain — given their interior — would actually do in this moment, then build the next scene around their choice rather than the plot's need.
The fourth, only visible in retrospect: the villain has no contradictions. Cartoon villains and bad-realistic villains both share this. Fanciful villains contradict themselves in small specific ways that the audience cannot reconcile. The unreconcilable contradiction is the marker of an interior the writer has imagined fully and withheld precisely.
The opaque-villain mode runs the risk of glamorization. Hannibal Lecter is fun, charismatic, articulate. Audiences love him. The writer who builds opaque-interior villains may inadvertently produce villains audiences root for. Roth doesn't address this risk directly. The implicit safeguard is that the villain remains evil in the audience's moral judgment even as their interior remains mysterious. The audience can be fascinated without being aligned.
A second tension: the understandable not likable criterion makes villain-craft less stable in television than in feature. Across 60 hours of a Breaking Bad arc, the audience's relationship to Walter White shifts. He starts understandable-but-not-likable and ends, for many viewers, likable-but-no-longer-understandable. The form's duration warps the villain. Roth's two-hour feature is the natural home for understandable-not-likable. Long-form villain craft is a different discipline.
Roth's villain craft converges with Lieberman's mind-reader's logic of intent inference4 from the existing vault. Lieberman's diagnostic question — what would they have to be thinking to do that? — is precisely the move the audience makes with an understandable villain. The writer who builds a Lieberman-eligible villain has built one the audience can run intent-inference on. The writer who builds an inscrutable-without-interior villain has built one Lieberman's mechanism cannot grip.
The split between Lieberman and Roth is direction. Lieberman is teaching the reader (real life) to infer intent. Roth is teaching the writer to construct intent-inference-eligible villains. Same mechanism, opposite ends of the apparatus. Reading both surfaces the polymath insight: villain craft is the literary application of a real-world social-cognition skill. Writers who can't infer intent in real life cannot construct villains amenable to intent-inference on the page.
Greene's 48 Laws of Power villains — historical figures whose strategic ruthlessness Greene catalogs — are a third reference point. Greene's villains are fully legible in their own logic. The strategic mind is laid bare. This is a different mode again: the transparently evil villain whose evil is interesting because the strategy is precise. The vault now has three distinct villain modes: traumatized (causal-chain), opaque-interior (gap-rendering), and transparent-strategic (Greene). A writer's choice among them is a major craft decision.
The compound insight: villain craft is the application of clinical psychology + manipulator profiling to the construction of intent-inference-eligible characters. The polymath's villain reads as more real because the polymath has structural maps the literary writer has not encountered. Greene's villains are real because Greene read the historical strategists. Roth's villains can be real to the degree he reads the psychological literature on opacity.
The Sharpest Implication
The standard advice to make your villain three-dimensional by giving them a sympathetic backstory may be producing the wrong villains. Sympathetic backstory produces traumatized-lash-out villains, which work but are predictable once the backstory is known. The more durable villain is the one whose interior the writer has imagined fully and whose interior the audience cannot reach. Withholding requires having something to withhold. The work is in the imagining the audience never sees.
Generative Questions