There's a Sanskrit category called mahāpuruṣa-lakṣaṇa — the characteristic marks of a great person. Thirty-two of them, by tradition. Long arms, broad shoulders, a particular shape of palm, the wheel-imprint on the foot. The marks do not make the person great; they are the signs by which a complete instance is recognized. A great person who lacks them is incomplete in the diagnostic sense. The marks are the inventory by which the tradition checks whether the body it is looking at is what it claims to be.
Bharata at NS XVII.1-41 does the same move with a dramatic composition.1 He names thirty-six lakṣaṇas — characteristic marks of a complete dramatic body of speech — and Ghosh's commentary spells out what's actually at stake: lakṣaṇa is not alaṃkāra (figure of speech) and not guṇa (excellence of composition). Lakṣaṇas are the diagnostic marks of the work as such.2 A play that has figures of speech, excellences, and Sentiments but lacks the lakṣaṇas is not a worse-styled play. It is an incomplete instance of the dramatic body. The lakṣaṇas are how the tradition checks whether what it is looking at is what it claims to be.
What's interesting is the range of the thirty-six. They are not all rhetorical figures. Bharata's inventory contains rhetorical moves (Ornateness, Compactness, Brilliance), reasoning forms (Causation, Discovery, Inference, Deliberation, Convincing Explanation), social pragmatics (Mediation, Series of Offers, Clever Manners, Compliment, Censure), psychological tells (Slip of Tongue, Indirect Expression of Desire, Concealment, Inversion), persuasion moves (Persuasion, Excellence, Distinction, Presumption), and aesthetic ornaments (Description, Pointed Utterance, Enumeration of Merits). The lakṣaṇas are a taxonomy of the full range of moves that a complete body of human speech makes — and Bharata's claim is that a dramatic composition must contain them all because human speech contains them all, and the play exists to recompress human speech into watchable form.
Ghosh's first footnote at NS XVII.1 is load-bearing. Three categories operate in the chapter, and conflating them collapses the system:
Alaṃkāras (figures of speech) — the four canonical ones at NS XVII.43-89 are Simile (upamā), Metaphor (rūpaka), Condensed Expression (dīpaka), and Yamaka (phonetic repetition). These are decorations of expression. They make a sentence more beautiful but do not change what kind of utterance it is.
Guṇas (excellences of composition) — the ten at NS XVII.95-106 are Synthesis, Perspicuity, Smoothness, Concentration, Sweetness, Grandeur, Agreeableness, Directness, Exaltedness, Loveliness. These are qualities the composition achieves. They are properties of the work, not actions in it.
Lakṣaṇas (characteristic marks) — the thirty-six at NS XVII.1-41 are the discourse-types the work performs. They are functional acts of speech — to argue from precedent, to mediate, to make an indirect request, to slip in revealing tongue, to compliment, to dispute. They are what the speech does, not how it is decorated and not what quality it achieves.
A complete Nāṭaka deploys all three categories at once. Alaṃkāra makes the diction beautiful; guṇa makes the composition excellent; lakṣaṇa makes the body of speech complete. Modern poetics, having inherited the alaṃkāra and guṇa categories more easily, often collapses lakṣaṇa back into "rhetorical figures," which is exactly what the system warns against. The 36 are not 36 more rhetorical figures next to Simile and Metaphor. They are a different kind of object — discourse-acts, not embellishments.
Sorted by function rather than by Bharata's enumeration order, the thirty-six fall into six bands. Each band names a kind of discourse-act the complete dramatic body must contain.
Rhetorical-aesthetic — Ornateness, Compactness, Brilliance, Description.3 How the work adorns its meaning and renders specific images.
Reasoning and epistemic — Causation, Discovery, Inference from Similitude, Deliberation, Convincing Explanation, Favourable Precedent, Unfavourable Precedent, Pointed Utterance.4 How the work conducts argument, draws conclusions, and supports claims with evidence.
Persuasion and influence — Persuasion, Excellence, Distinction, Presumption, Series of Offers, Celebrity, Indirect Expression of Desire, Wit, Semi-uttered Expression.5 How the work moves the listener toward an act or recognition without direct command.
Social pragmatics — Mediation, Clever Manners, Compliment, Censure, Accusation of Virtues, Identity, Interrogation, Enumeration of Merits.6 How the work manages the live relational field between speakers.
Psychological tells — Slip of Tongue, Indirect Expression of Desire (again — it crosses bands), Concealment, Inversion, Hesitation.7 How the work registers what the speaker did not mean to reveal.
Compositional-structural — Parallelism, Multiplex Predication, Fancy, Compliment.8 How the work weaves repeated meaning across the dramatic body.
The taxonomy is wider than the modern category of "rhetorical devices." It includes things rhetoric never quite handles — the slip, the indirect, the concealment, the body-language of pragmatic courtesy. Bharata is naming the full discourse repertoire of human social being and claiming that a dramatic composition must contain it all.
The system lands when you watch specific lakṣaṇas operate.
Brilliance (śobhā, XVII.8): "If a charming and novel meaning arises when a less-known object is referred to by likening it to a well-known one, and a wonderful sense is expressed through double entendre."9 This is not Simile. Simile says A is like B for clarity; Brilliance says A is like B and the double-meaning sparks. The aesthetic event is the spark, not the comparison.
Slip of Tongue (bhraṃśa, XVII.27): "Manifold deviation of proud and similar other persons from the intended words to something else."10 Bharata is naming, in 200 CE Sanskrit, what Freud will eventually call Fehlleistung — the parapraxis that reveals what the speaker meant to suppress. The lakṣaṇa is a named compositional move: the writer is expected to script slips of tongue into the dramatic body as functional discourse-acts that disclose subterranean material.
Indirect Expression of Desire (manoratha, XVII.36): "Expressing one's secret desire of the heart by a pretence of referring to somebody else's condition."11 The character speaks of someone else's love-longing while in fact disclosing their own. This is not "subtext" as a modern craft notion — it is a named, definable, deployable move that Bharata expects the writer to be able to execute on command.
Three lakṣaṇas, three different bands. The scale of the system becomes visible: thirty-six named, defined, deployable discourse-acts that the writer holds available as a working repertoire. Modern scriptcraft has perhaps ten of these as half-named conventions ("subtext," "callback," "rhetorical question"). Bharata's repertoire is roughly three times richer.
NS XVII.42 closes the chapter with a single rule that prevents the lakṣaṇas from becoming a generic toolkit: "These thirty-six characteristic marks of a dramatic composition conducing to the object in view will beautify a play and [hence they] should be properly used according to the Sentiment [introduced in it]."12 The lakṣaṇas are not interchangeable building blocks. Each must be deployed in tune with the prevailing rasa. A Slip of Tongue lands differently in a Pathetic scene than in a Comic one; an Indirect Expression of Desire is the soul of an Erotic moment and an off-note in a Furious one. The writer's craft includes not just knowing the thirty-six lakṣaṇas but knowing which ones the present rasa makes available. This makes the lakṣaṇa system rasa-tuned at the discourse level — what the vṛtti does at the tonal-register level (Chapter XXII), the lakṣaṇas do at the sentence-and-exchange level.
The lakṣaṇa system has three modern uses.
As a vocabulary audit for scriptwriting. When a scene's dialogue feels flat, run it against the six bands. Are there reasoning-acts? Persuasion-acts? Social-pragmatic acts? Psychological tells? Or is the scene operating in only one band — usually persuasion or reasoning — and starving the others? A scene built only on argument lacks the social-relational lakṣaṇas that make characters feel like people in a room. A scene built only on subtext lacks the rhetorical-aesthetic lakṣaṇas that give the language texture.
As a diagnostic for "characters all sound the same." If every character deploys the same lakṣaṇas in the same proportions, they are functionally one speaker in different costumes. The fix is to vary lakṣaṇa-distribution by character — one speaker leads with Pointed Utterance and Distinction; another with Indirect Expression of Desire and Concealment; another with Censure and Clever Manners. The texture of difference is built at the lakṣaṇa level.
As rasa-tuned discourse selector. Before drafting a high-emotion scene, name the rasa and then ask which lakṣaṇas the rasa makes available. The Erotic rasa wants Indirect Expression of Desire, Compliment, Semi-uttered Expression, Wit. The Furious rasa wants Censure, Accusation, Distinction, Interrogation. The Pathetic rasa wants Enumeration of Merits, Indirect Expression, Recollection-style Description, Slip of Tongue. Deploying off-rasa lakṣaṇas — a Compliment in a Furious scene — produces tonal incoherence the audience feels but cannot name.
Two recensions, seventeen names in common: Ghosh's footnote at XVII.1 records that NS manuscripts split into two recensions of the thirty-six lakṣaṇas, with "no more than seventeen names in common, and among these, definitions of eight only are similar in both recensions."2 The older recension (used by Viśvanātha and Śiṅgabhūpāla) is the basis of Ghosh's translation. The later recension (used by Kīrtidhara, Abhinavagupta, Dhanañjaya) diverges substantially. The number thirty-six is preserved across both; the contents are not. The page reads this as a doctrinal commitment to the count combined with a working tradition that revised its contents — the same pattern as the sixty-four limbs.
Definitions partly lost: Ghosh's footnote at XVII.6 admits "the exact meaning of some of the terms at least relating to this subject has been to some extent lost, and various explanations have been partly based on guess."14 The system survived in name and structure but parts of its specific deployment have been reconstructed rather than transmitted. This is a real limit on the modern recovery of the framework.
Lakṣaṇa vs alaṃkāra confusion in later poetics: Ghosh notes that the lakṣaṇa category later became confused with figures of speech in the Sanskrit alaṃkāra-śāstra tradition. The page argues that the distinction at NS XVII.1 should be restored: lakṣaṇas are discourse-acts, not figures.
Order vs band-organization: Bharata enumerates the thirty-six in an order that does not group them by band-of-function. The page reorganizes them into six functional bands; this is an interpretive move not present in the source. The source order has its own logic (one Ghosh footnote suggests it follows traditional mnemonic conventions) but does not aid comprehension for a modern reader.
Bharata is the single primary source, and the productive tension is between his explicit definitions and the later Sanskrit tradition's struggle to maintain them. The Ghosh footnote at XVII.6 is unusually candid: Bharata gives definitions that "are mostly not at all clear without examples," and Abhinavagupta's commentary supplies the examples while sometimes shifting the definitions toward variants. Across the centuries, the thirty-six lakṣaṇas behave like the sixty-four Limbs of Chapter XXI — a sacred count whose actual contents drift under the pressure of continued use. The tradition treats Bharata's framework as authoritative; the practice treats Bharata's framework as a working vocabulary to be refined.
What this reveals about the lakṣaṇas: they are not handed down as final truth and they are not simply lost. They are a vocabulary the practitioner is expected to inhabit and inflect. Bharata's commitment is to the thirty-six-band structure — that there are roughly that many irreducible discourse-acts a complete dramatic composition must contain. The exact names and definitions are negotiable in the same way the sixty-four Limbs are negotiable. This reframes the lakṣaṇa system not as a fossil of classical Sanskrit poetics but as the earliest working specification of a craft vocabulary the modern dramatist still lives inside without knowing the names. When the modern writer asks themselves "what is this character doing in this line of dialogue," they are reaching for a lakṣaṇa Bharata named in 200 CE. The recovery move is not to memorize the thirty-six. It is to remember that the body of human speech has thirty-some named irreducible acts and that the writer's job is to deploy them in calibrated combinations.
The thirty-six lakṣaṇa system is a creative-practice page, but the structural move — a granular, named, defined inventory of discourse-acts that exhausts the field of a complete domain — recurs across adjacent vault domains. Two pages hold the same shape from different angles.
behavioral-mechanics: The Eleven Propaganda Ploys — Comprehensive Tactical Taxonomy — propaganda theory's enumerated catalog of named tactical ploys is structurally identical to Bharata's lakṣaṇa system, applied to a different field of speech. Both systems insist that effective discourse-craft is granular: it does not deploy "rhetoric" or "propaganda" generically; it deploys named, defined acts in calibrated combinations. The eleven propaganda ploys (Name Calling, Glittering Generalities, Transfer, Testimonial, Plain Folks, Card Stacking, Bandwagon, etc.) are the modern operator's working repertoire; the thirty-six lakṣaṇas are Bharata's. The structural identity goes further than catalog-shape: both systems make rasa-tuning or audience-tuning explicit. NS XVII.42 says the lakṣaṇas must be deployed according to the Sentiment; propaganda theory says the ploys must be deployed according to the audience's existing affective state. Neither system treats its named acts as context-free building blocks. The insight neither domain alone produces: the working craft of moving people through speech, whether toward aesthetic experience in drama or toward action in influence operations, has the same structural requirements — a granular named inventory of acts, plus a state-tuning rule that selects which acts the present audience-state makes effective. Bharata is doing influence theory and Aristotle is doing dramatic theory in pieces; together their move is the same. The 200-CE Indian and 20th-century American systems converge because the requirement is structural — moving humans through speech is not a context-free craft, and granularity plus state-tuning is what the receiving nervous system requires.
psychology: Defense Mechanism Inventory and Three Ego-Exoneration Methods — psychology offers its own named, defined catalog of discourse-acts and intrapsychic moves — the defense mechanism inventory (denial, projection, rationalization, displacement, sublimation, etc.) plus the three ego-exoneration methods. The structural parallel to NS XVII is exact in a subset of lakṣaṇas: Bharata's Slip of Tongue (XVII.27) is the parapraxis Freud will analyze 1700 years later; Indirect Expression of Desire (XVII.36) is displacement plus projection in dramatic form; Concealment (XVII.38) is the ego defense rendered as a named compositional act the writer deploys on purpose. What differs is the agency: psychology describes these moves as involuntary protections the psyche generates against threat. Bharata describes them as voluntary compositional moves the writer deploys to render the involuntary on stage. The insight neither domain alone produces: every defense mechanism the psyche performs involuntarily has a corresponding lakṣaṇa the writer deploys voluntarily. The dramatist's job is to render the psyche's involuntary moves as if the speaker did them on purpose, and the audience reads the rendering and recognizes itself. This is what makes drama therapeutic in a way it would not be if it merely depicted defenses externally — the lakṣaṇa structure makes the audience enact the defense in compressed form, in the watching, and that enactment is what allows recognition. Bharata's thirty-six lakṣaṇas are not just craft moves; they are a working catalog of how the psyche's involuntary repertoire becomes the writer's voluntary palette.
The Sharpest Implication
If you take the thirty-six lakṣaṇas seriously, the working vocabulary you use to talk about dialogue is impoverished by an order of magnitude. Modern writing craft has maybe ten named dialogue-moves — subtext, callback, status play, the line, the comeback. Bharata named thirty-six in 200 CE and the later Indian tradition revised but never reduced the count. The implication is not that you should memorize Sanskrit terms. The implication is that the work needs discourse-act granularity at roughly that resolution, and that modern writers compensate for the missing vocabulary either by overworking the ten moves they do have names for, or by writing on instinct without a working diagnostic. Bharata's claim is that instinct is not enough at this resolution — that the body of human speech has thirty-some irreducible acts and the writer who cannot name them produces work that lacks several entire bands of discourse. The flat-dialogue diagnosis is rarely about the line being weak; it is about the band being absent. Restore the band — add reasoning-acts to a scene built only on persuasion, or add psychological tells to a scene built only on social pragmatics — and the line that read as flat suddenly works because it is now operating in a complete discourse-field.
Generative Questions
The names of these — they sound kind of academic. They don't really matter. It's the formulas and the tactics that really make the difference. Indeed the ancient Greeks couldn't even agree on half of these technical names. They called them all sorts of different stuff.15 Mark Forsyth on the equivalent Greek rhetorical-figures tradition, recovered for contemporary English writers in The Elements of Eloquence.
The same applies to the 36 lakṣaṇas. Bharata enumerates them at NS XVII.1-5; later commentators add and re-categorize; some lakṣaṇa definitions are stable across the commentarial tradition while others (Ghosh's note at XVII.6) have suffered partial loss of meaning and require Abhinavagupta's commentary for reconstruction. The Sanskrit names are useful for navigating commentarial literature. They are not load-bearing for using the formulas. A writer who internalizes the formula for Brilliance (śobhā, NS XVII.8 — language with a particular striking effect on the listener) can deploy the formula without the name. A writer who memorizes names without internalizing formulas has decorative knowledge.
Forsyth's specific recommendation, applicable to anyone studying classical rhetorical-figures taxonomies — Greek or Sanskrit — is to build a personal-name cheat-sheet alongside the technical-name reference. I came up with little names for myself, and then I have a cheat sheet with a few examples and some bullet points. And then what I can do is if I'm trying to insert one, I can just look at the cheat sheet. The Sanskrit student who follows this approach deploys the lakṣaṇas more reliably than the student trying to operate from technical-name knowledge alone. The lakṣaṇa-tradition and the Greek-rhetoric tradition are sister catalogues — both valuable, both name-burdened, both best operated with formulas as the load-bearing unit. See Forsyth Rhetorical Figures Methodology and The Four Alaṃkāras.