Śrī Śaṅkuka was probably a Kashmir scholar writing around 850 CE, a generation after Bhaṭṭa Lollaṭa and a generation before Ānandavardhana. The chronicle Rājataraṅgiṇī identifies a poet of the same name as author of a court poem about a battle in 850/851 — almost certainly the same person. He was also a dramatist; a 12th-century work refers to him as "a minister who was also a dramatist" and remembers a romantic comedy he wrote called Citrotpalāvalambitaka (The Earring of the Many-Colored Lotus).1
Pollock raises a question that subsequent scholarship has not resolved: was Śaṅkuka a Buddhist? The evidence is circumstantial but accumulates: he quotes Dharmakīrti (the great seventh-century Buddhist logician); the honorific "Śrī" sometimes suggests Buddhist affiliation; Bhaṭṭa Toṭa critiques one of his positions on grounds that would force him to accept universals, which only non-Buddhists accepted; he treats karuṇa (the tragic) as general-compassion-rather-than-grief-for-loved-one, paralleling Mahāyāna developments; and he accords central place to inference, one of only two means of valid knowledge Buddhists accept.
His commentary on the Treatise on Drama is lost. What survives are fragments preserved by Abhinavagupta — who quoted Śaṅkuka primarily to refute him, but whose reconstruction is detailed enough to recover the framework.
Śaṅkuka's central thesis: rasa is an imitation (anukaraṇa) in the actor of the stable emotion in the character.1 The cognitive event by which the spectator apprehends this imitation is inference (anumāna) — hence the name of his position, anumiti-vāda.
The framework departs from Lollaṭa at exactly the point where Lollaṭa's silence is loudest: Lollaṭa had said rasa is in the character, but had not specified what the spectator does. Śaṅkuka specifies. The spectator INFERS the character's stable emotion from inferential signs: the foundational factors (a beloved person), the stimulant factors (a garden, the moon), the reactions (sidelong glances, horripilation), the transitory emotions (longing, agitation). These signs are factitious — they have been produced by the actor's training — but the spectator does not register their factitiousness. The spectator apprehends them as inferential signs for the character's stable emotion, and infers the emotion as PRESENT IN THE ACTOR (insofar as the actor has identified with the character) but BELONGING TO THE CHARACTER.
The inferred stable emotion in the actor is what we call rasa.
Śaṅkuka's most famous example: when we see a painted horse, our apprehension does not take any of the four forms ordinary cognition takes:1
Instead, we have a fifth kind of apprehension that says simply: "This is a horse." The painted-ness is not registered; the horse-ness is not held with reservation; we simply see the horse.
The same applies to the actor playing Rāma. The viewer does not think:
We simply apprehend: "Here is Rāma."
The cognitive event is a fifth thing, distinct from all four ordinary cognitive types. Aesthetic apprehension is its own category.
Śaṅkuka's other illustration, drawn from Dharmakīrti: two men run toward two sparkling lights — one a gem, one a lamp — both thinking the lights are gems. "Although there is no difference regarding their false knowledge, there is a difference regarding their real effects."1
The example shows that real effects can arise from cognitively-false content. The men have false cognition (the lamp is not a gem), yet their actions and outcomes are real. Similarly with rasa: the spectator's cognition of "Rāma's emotion in the actor" is in some sense false (Rāma is not really present; the emotion is fabricated by the actor's training), yet the rasa-experience is real.
This is what differentiates aesthetic apprehension from ordinary error-cognition: in ordinary error, the false content fails to produce its proper effects when the error is corrected. In aesthetic apprehension, the false-by-ordinary-standards content produces real aesthetic effects continuously throughout the performance, never breaking down.
Śaṅkuka's most consequential conceptual innovation is the distinction between referential and expressive language.1 Words that name emotions (desire, grief, anger) operate by reference — they point to internal states. But these words alone cannot communicate emotion to the spectator. The naming of an emotion is not the production of an emotional effect.
What is required is expressive language — language that does not just refer to but enacts the emotion. The actor's speech, accompanied by appropriate physical reactions, produces an inferential sign that the spectator processes as evidence for the character's interior state. This is fundamentally different from naming the state.
The referential-vs-expressive distinction is the conceptual seed of what Ānandavardhana would develop into dhvani (implicature) — the doctrine that rasa cannot be denoted but only suggested. Pollock notes: "Śaṅkuka's critique a few generations before Ānanda was based on a powerful and convincing distinction he was the first to draw, between mere referential language and the expressive language required for rasa."1 Ānanda would build the dhvani doctrine on this foundation, often without acknowledging Śaṅkuka's priority.
Bhaṭṭa Toṭa (Abhinavagupta's teacher, c. 975) launched a devastating critique of Śaṅkuka's framework. The critique operates at three levels:
Against the inferential view from the spectator's perspective: The actor's body, costume, gestures, etc., are material things perceived through ordinary senses. They are radically different from the immaterial mental state of desire. How can material reactions inferentially produce knowledge of an immaterial emotion? More importantly: we have no idea what Rāma actually felt — none of us has ever met him — so the inference cannot ground itself in any prior knowledge of "Rāma's actual desire."
Against the inferential view from the actor's perspective: The actor does not think "I am imitating Rāma." If "imitation" means "making similar to," the actor cannot make himself similar to someone he has never seen.
Against the very category of "imitation": There is no actually-existing desire-of-Rāma for the actor to imitate. The framework breaks down at its foundation.
Bhaṭṭa Toṭa's critique was so thorough that Abhinavagupta took it over wholesale into his commentary on Bharata. Within two generations of Śaṅkuka's death, his framework was consigned to oblivion. We know it primarily because Abhinava preserved it as the position he was refuting.
Śaṅkuka and Bhaṭṭa Lollaṭa, separated by 25 years, share the conviction that rasa pertains to the character, but they specify the framework completely differently. Lollaṭa: rasa is the stable emotion intensified by aesthetic elements, located in the character primarily and in the actor through complete identification. Śaṅkuka: rasa is the actor's IMITATION of the character's stable emotion, inferentially apprehended by the spectator. The convergence is rasa-in-character; the tension is whether what we apprehend is the emotion itself (Lollaṭa) or an imitation of it (Śaṅkuka). The split reveals the philosophical sophistication Śaṅkuka brought to the discipline: Lollaṭa treated audience-apprehension as automatic; Śaṅkuka demanded a specific cognitive account of HOW the apprehension worked. The demand was philosophically right; the answer (imitation + inference) was the wrong answer, but the demand itself was foundational.
Śaṅkuka and Ānandavardhana, separated by 25 years, are connected through Ānanda's silent appropriation of Śaṅkuka's referential-vs-expressive distinction. Ānanda built the dhvani doctrine on this foundation. The convergence is that rasa-communication requires something different from ordinary linguistic operation. The tension is what that something is: for Śaṅkuka, inference from imitative signs; for Ānanda, manifestation via implicature. Both positions agree that ordinary reference + ordinary metaphor cannot produce rasa. They diverge on the alternative mechanism. Pollock argues this is one of the clearest cases in Indian aesthetic history of unacknowledged conceptual debt: Ānanda took the foundational insight from Śaṅkuka, transformed it into a different theoretical framework, and never credited the source.
Śaṅkuka's framework operates at the intersection of philosophy of mind, theory of inference, and aesthetic phenomenology.
Behavioral-Mechanics: Manipulation and Influence Hub — Śaṅkuka's referential-vs-expressive distinction has a precise structural parallel in the practitioner-influence tradition: the difference between TELLING SOMEONE TO FEEL SOMETHING and PRODUCING THE FEELING IN THEM. Hughes (Six-Minute X-Ray), Cialdini, Greene all teach that the words for emotional states are weak influence tools; the production of the actual states via environmental cues and behavioral elicitation is far stronger. This is Śaṅkuka's distinction restated for an applied-influence context. The handshake reveals: the most rigorous theorization of how language produces (rather than merely names) emotional effects in another person comes from an 850-CE Sanskrit-tradition theorist of aesthetic experience, and modern influence-mechanics is a slow rediscovery of the same conceptual structure.
Creative-Practice: Non-Fiction Writing Craft Hub — "Show don't tell" is the contemporary form of Śaṅkuka's referential-vs-expressive distinction. To "tell" is to use referential language (naming the emotion). To "show" is to use expressive language (producing the inferential signs from which the reader infers the emotion). The handshake clarifies what "show don't tell" is doing at the cognitive level: triggering the same inferential apparatus the spectator deploys in a theatrical performance. The principle works because human cognition is structured to infer emotional states from behavioral signs, and writers exploit this. The Śaṅkuka framework provides the underlying theory that "show don't tell" pedagogy mostly leaves implicit.
AI-Collaboration: AI-Assisted Creator Hub — LLM-generated prose tends to OVERUSE referential language and UNDERPRODUCE expressive language. The systems are trained to name emotional states (the character "felt sad", "experienced joy") rather than produce the inferential signs from which the reader would infer those states. This is what makes much LLM prose feel "flat" — it operates in the referential mode Śaṅkuka identified as inadequate for rasa-production. The handshake produces a diagnostic for AI-assisted writing: revise referential constructions into expressive ones; replace named-emotions with inferential-signs. The Śaṅkuka framework predates LLM systems by 1,200 years but identifies precisely the failure mode that makes contemporary AI prose feel artificial.
The Sharpest Implication
Śaṅkuka's framework was the LOSING side in Sanskrit aesthetic history: Bhaṭṭa Toṭa demolished it, Abhinava preserved it only as the position-to-refute, and within two generations the imitation-and-inference view was considered settled-against. Yet contemporary empirical research on reader cognition increasingly supports something closer to Śaṅkuka than to his refuters. Reader-response research, Theory-of-Mind studies, and computational text analysis all suggest that readers DO infer characters' mental states from behavioral signs, do operate the inferential apparatus Śaṅkuka described, do not in fact undergo any of the four ordinary cognitive types in their apprehension of fictional characters. The discomfort: the Sanskrit tradition rejected a framework that may have been substantially correct, on philosophical grounds (the problem of the unseen original) that empirical research has not validated. Śaṅkuka may have been more right than his successors knew.
Generative Questions