Some matters on taste
Why Kant and Sontag Cannot Speak Otherwise
Preface
There is a familiar reaction I have to Kant’s account of taste and to Sontag’s sensibility of Camp: admiration, followed by irritation. Admiration, because both are plainly tracking something real—an experienced lawfulness of judgment that is neither a private whim nor an application of a checklist. Irritation, because when pressed for “grounds,” both appear to retreat into a kind of semantic juggling: universal without proof, logic without system, communicable without codification. The standard conclusion is that we have reached the limits of theory and must settle for evocative description.
RGA’s claim—and the reason I find it useful here—is that this conclusion is mistaken. The “quasi-paradox” in Kant and Sontag is not a terminal mystery; it is a symptom of a well-known structural predicament: competence under poverty of stimulus (PoS). Once I state that predicament explicitly, the semantic juggling is no longer an embarrassment; it becomes exactly what one should expect from anyone trying to describe a lawful internal capacity while refusing to falsify it by turning it into an external rubric.
My task, then, is not to accuse Kant and Sontag of lacking RGA. They could not have had it. My task is to show, that RGA’s PoS + minimal universal grammar of aesthetics (UG(A)) predicts why their formulations must take the form they do—and to show how the apparent paradox dissolves once I replace the wrong alternatives with the right one.
I need a brief reminder of the framework, because otherwise several of the terms I will use (“basis,” “I-state,” “lawfulness,” “codification”) will read like metaphors rather than commitments. RGA distinguishes an internal, intensional system (I-aesthetics) from its externalization (E-aesthetics). Internally, aesthetic competence is stratified: C-layer (compositional core) generates structured candidates under ranked, violable constraints; A-layer (affective appraisal) evaluates these candidates by a graded resonance procedure; P-layer (pragmatic-intentional stance/contract) gates what counts as licensed, serious, playful, devotional, satirical, and so on; H-layer (historical-narrative priors) supplies registers, genealogies, and learned expectations; J-layer (institutional-juridical designation/curation/adjudication) stabilizes exemplars and enforceable categories; and EPI-layer (ethical-epistemic claims) constrains cases where harm, truth, or justice are at stake. Externally, the same competence is encountered through interfaces: MF-layer (material form), IF-layer (interpretive form: attentional arcs and felt affordances), and SF-layer (social form: status/use, policy, law). SF-layer fixes what RGA calls the “basis” of uptake—the operative frame that selects among admissible parses. Given an I-state (the current configuration across C/A/P/H/J/EPI), an MF input, and an SF basis, resonance collapses lawfully; apparent indeterminacy is often nothing more than multiple admissible parses awaiting a basis.
I
Kant’s distinction (as I have been using it) can be put schematically.
Determinative judgment: the rule is given; the particular is subsumed; correctness is in principle demonstrable.
Reflective judgment: the particular is given; the rule is not; nevertheless the judgment presents itself as universally claimable—“you ought to agree”—while refusing the sort of proof that determinative judgment permits.
The peculiarity of taste is precisely this combination: a demand of universality without criteria. If I insist that universality requires explicit rules, Kant starts to look like he is saying something close to nonsense: a public claim for which no public grounds can be supplied.
Sontag’s remark about sensibility has the same logical shape, though the vocabulary differs. Here I want her sentence in full, because it is already a compressed diagnosis:
“Taste has no system and no proofs, but there is something like a logic of taste, the consistent sensibility which underlies and give rise to certain taste. A sensibility is almost but not quite ineffable. Any sensibility which can be crammed into the mold of a system, or handled with rough tools of proof, is no longer sensibility at all. It has hardened into idea.”
If I assume that a “logic” must be an explicit set of criteria, then this sounds like a self-protecting doctrine: there is a law, but it cannot be stated; there is an essence, but it must not become an idea. Yet Sontag is not merely posturing. She is trying to name a third category: something stable and recognizable—hence “logic”—yet not propositionalizable as a rubric without losing what it is.
In both cases, the trouble is not the phenomena. The trouble is the background assumption that leaves me only two permissible kinds of explanation:
(1) an explicit rule (rubric, criteria, system), or
(2) a private or merely sociological fact (agreeableness, fashion, clique code).
Kant and Sontag are both trying to describe a third thing: a lawful competence that is not stored or transmitted as a rubric. Without an explicit theoretical category for that third thing, their prose has to oscillate between the vocabulary of lawfulness (“universal,” “logic”) and the vocabulary that blocks rubricization (“no proofs,” “hardens into idea,” “betrayal”). It is this oscillation that later readers experience as paradox, or as shifting goalposts.
II
RGA’s PoS claim is, for my purposes, the claim that the “rule-or-whim” dichotomy is not a neutral starting point but a theoretical error. It is the aesthetic analogue of a familiar mistake in empiricist speculation about language: to suppose that because there is learning and convergence, there must have been explicit instruction or rich external evidence sufficient to determine the internal system.
RGA states the relevant facts in a form that is hard for me to evade.
First, poverty of instruction: the crucial operators and tradeoffs are typically not taught as rules, but discovered as necessities when attempts fail; the negative knowledge arrives as disappointment, not doctrine. Accordingly, the mind must already be equipped with something like a search grammar over derivations, else learning would be blind wandering through an intractable space.
Second, poverty of evidence: the artist receives almost no systematic negative evidence distinguishing admissible from inadmissible derivations; nevertheless the artist rejects vast regions of their own production as “dead” or “false,” often before any audience speaks. This presupposes the capacity to supply negative evidence internally.
Third, one-shot generalization and transfer: competence is amodal; people grasp what kind of structured move is in play and project it across media; this is inexplicable on a picture that treats competence as a mere inventory of rewarded tokens.
From these considerations RGA draws the conclusion that linguistics was forced to draw: I must distinguish competence from performance and posit an internal structure rich enough to make the observed performance possible. But now I need the refinement that has been hovering in the background.
UG(A), in a serious formulation, need not be a baroque innate catalog of aesthetic categories. If I follow the minimalist turn in linguistics, the species-specific core can be treated as minimal recursive structure-building plus interface constraints—the ability, in C, to generate hierarchical candidates, and to map them to the sensorimotor and conceptual-affective interfaces in a way that is usable at all. The familiar operator inventory (grouping, repetition-with-variation, focality, contrast, identity-under-transform, and so on) can then be treated as stable descriptive families of effects of that minimal engine under particular interface pressures, priors (H), stances (P), and bases (SF). RGA itself states the compositional core in operator-and-constraint terms and evaluates by graded resonance; but nothing in the PoS logic requires that every named operator be an innate substantive “feature.” What PoS requires is that the internal system be structured and restrictive enough to guide search, and to yield lawful internal negative evidence.
This is the crucial point for me: PoS forces internal structure; it does not force rubric-like explicitness.
III
With that in view, I can say directly how PoS + minimal UG(A) anticipates the Kant/Sontag quasi-paradox.
If aesthetic judgment is the output of a structured internal competence operating under severe underdetermination, then two things follow immediately.
(1) The judgments can be lawful and convergent (hence “logic,” hence “universal voice”), because shared internal structure and shared interface constraints restrict the space of admissible parses. In RGA’s terms: C generates candidates; A evaluates them; and across agents there is enough common architecture for nontrivial convergence, especially when P, H, and SF are aligned.
(2) The judgments cannot, in general, be justified by a finite set of public criteria, because the competence is procedural: it consists in a search over derivations and a graded evaluation, routed through stance and priors, not in a list of necessary-and-sufficient conditions.
Lawfulness without codifiability is exactly what PoS predicts.
Kant and Sontag did not possess this argument in explicit form. But they were describing its surface phenomenology from within aesthetic life. Their “semantic juggling” is the natural rhetoric of someone who refuses to reduce competence to a rubric while refusing, at the same time, to collapse it into private liking. They are trying to name the third category without having the third-category machinery.
IV
The resolution of Kant
On the rubric model, “universal validity” can only mean one of two things:
(a) everyone who applies the criteria will reach the same verdict (perfection), or
(b) many people happen to agree (a sociological generalization).
Kant denies both: taste is not demonstrable by criteria, and it is not mere agreement by coincidence. Hence the familiar appearance of paradox.
RGA resolves this not by producing a hidden rubric, but by replacing the wrong question with the right one.
The right question is: what must the internal system be like, such that—given sparse instruction and little negative evidence—agents can nonetheless converge on parseable structures and experience a felt necessity in appraisal?
RGA’s answer is: the internal system is a generative search-and-evaluation procedure distributed across layers. The “lawfulness” is not a set of public criteria but the stability of the computation given an I-state and a basis. Under fixed conditions, resonance is lawful; alter the basis (i.e., alter SF), and the outcome shifts systematically rather than randomly.
Now Kant’s “reflective judgment” stops looking like mystical privilege and starts looking like a perfectly ordinary condition of competence under PoS:
I do not have a rule in hand, because what I have is not a rule-list but an internal generator (C) and evaluator (A), routed by stance (P) and priors (H).
I nonetheless expect agreement, because the generator/evaluator is not idiosyncratic whim; it is a shared capacity operating under shared interface constraints, and (Kant wagers) often under sufficiently aligned priors and basis to support a public claim.
I cannot “prove” the judgment in the determinative sense, because proof requires criteria externalizable as a checklist—i.e., the sort of stabilization typically carried by J and enforced by SF—whereas reflective judgment concerns competence prior to, or in excess of, that stabilization.
In RGA terms, determinative judgment corresponds to what happens when the J-layer has stabilized criteria and exemplars within an SF basis that makes them operative; reflective judgment corresponds to the case where C/A/P/H are doing the decisive work before J has fixed a portable rule, or where no such rule can capture the competence without distortion. The “gap between the private and the provable” is, in my terms, precisely the gap between lawful competence and public rubric.
V
The resolution of Sontag
Sontag’s sensibility claim becomes transparent under the same move. Her phrasing already tells me what must be true if it is to be more than rhetoric: “something like a logic of taste” is affirmed, while “system” and “proofs” are denied; and then comes the diagnostic—what happens when one tries anyway: the sensibility “hardens into idea.”
On the rubric model, “logic” must be explicitly stateable. On the private-feeling model, whatever cannot be stateable must be ineffable whim. Sontag denies both. RGA dissolves the tension by making explicit what “sensibility” is in the only form that makes her sentences coherent rather than magical: a stable parameterization of internal competence across layers, guided by basis.
The reason systematization “betrays” is not that language is forbidden; it is that public articulation typically functions as basis engineering. The moment I externalize a sensibility as a set of teachable markers, I shift uptake from competence-guided reflection to determinative application under a new SF basis. In RGA terms, I push labor upward into J: designation becomes doctrine; doctrine becomes enforcement; and what had been a living competence distributed across C/A/P/H is replaced by a portable projection.
So “betrayal” is not romanticism. It is a predictable externalization effect:
A sensibility lives as a way of searching and weighting—what derivations I entertain (C), what feels resonant (A), what stance is licensed (P), what priors make the move legible (H), and what bases sustain uptake (SF).
A “system” is what results when that living competence is projected into J as a portable doctrine.
The projection is necessarily lossy: it captures surface correlates in MF/IF terms, not the internal routing that made them function.
Once portable, the projection becomes an enforcement device—often unintentionally—and the sensibility hardens into “idea.”
Under PoS this is exactly what I should expect: since the competence is not learned or stored as explicit doctrine, any attempt to turn it into doctrine will alter the social conditions of learning and uptake—SF and J—and therefore alter the phenomenon.
VI
Why their prose must look the way it does
I can now say, without euphemism, why Kant and Sontag sound as if they are moving goalposts.
They are trying to defend a lawful phenomenon against two bad reductions:
reduction to rubric (determinative criteria stabilized by J and enforced under some SF), and
reduction to whim (agreeableness) or to brute sociology (mere clique labeling without competence).
But without PoS and without an explicit postulate of internal generative competence distributed across layers, there is no stable vocabulary for “lawful but non-codifiable.” So their texts perform a balancing act:
They invoke lawfulness (“universal,” “logic”) to block the slide into mere liking.
They invoke non-codifiability (“no proofs,” “hardens into idea”) to block the slide into rubric.
RGA’s PoS + UG(A) does not merely reconcile these slogans after the fact. It explains why any honest description of competence under underdetermination will have to take on this apparently paradoxical form until the underlying structure is made explicit.
VII
What changes once the structure is explicit
Once I accept PoS and posit a minimal UG(A), the “paradox” is not solved by adding a hidden rulebook. It is solved by withdrawing a mistaken assumption: that public normativity requires public criteria.
The resulting picture is straightforward.
There is a minimal generative core (recursive structure-building in C plus interface constraints that make it usable).
There is a layered internal state (C/A/P/H/J/EPI) that weights derivations, evaluates them by graded resonance, and is steered by priors, contracts, and basis.
Publics and institutions do not create the competence; they route it and stabilize some projections of it as standards through SF and J.
And then the Kant/Sontag theses become ordinary consequences:
Kant: reflective taste is the experience of lawful competence (C/A routed by P/H) operating where no stabilized J/SF rubric yet exists; it therefore feels universal and yet resists proof.
Sontag: sensibility is the stable parameterization of that competence under a particular basis (SF) and set of priors (H) and stances (P); it is shareable and yet collapses into cliché when externalized as a system, because J/SF codification converts competence into determinative application.
This is not relativism. It is the relativity RGA insists upon: given an I-state and a basis (SF), resonance collapses lawfully; change the basis, and the same material form (MF) can be computed differently without any appeal to whim.
That is the resolution. The rest is commentary.