Denotation is the foundational mode of symbolic reference in Goodman's framework—the relation by which a symbol picks out, labels, or depicts its subject. A portrait denotes the person portrayed; the word 'dog' denotes dogs; a diagram denotes the system it represents. Denotation operates through conventions that establish which symbols refer to which subjects—pictorial conventions for painting, linguistic conventions for words, diagrammatic conventions for scientific representation. The conventions are not arbitrary (they are constrained by purposes and by fit with other accepted versions), but they are conventional—there is no natural, pre-conventional connection between a symbol and what it denotes. Goodman's analysis of denotation displaced the assumption that representations work by resemblance—that a portrait denotes its subject by looking like it. Resemblance, Goodman demonstrated, is symmetric and reflexive (if A resembles B, then B resembles A, and A resembles itself), while denotation is neither. A portrait denotes its subject; the subject does not denote the portrait. The connection is established by convention, not by similarity. Understanding what a symbol denotes requires knowing which conventions are in play, and the knowledge is acquired through immersion in the symbol system's practices.
Goodman's account of denotation applied with equal rigor to pictures and to words, rejecting the traditional distinction that treated pictorial representation as natural and linguistic representation as conventional. Both are conventional, he argued—the conventions differ, but both require learning, both operate through established systems of reference, and both can be understood only by someone who has internalized the relevant conventions. A person who has never seen Western pictorial art does not immediately perceive that a two-dimensional array of pigment on canvas denotes a three-dimensional landscape—the perception requires familiarity with the conventions of perspective, foreshortening, and atmospheric depth that Western painting employs. The conventions are so thoroughly internalized by people raised in picture-saturated cultures that the conventionality becomes invisible, producing the illusion that pictures are natural copies of what they depict.
In the age of AI, denotation operates straightforwardly: AI-generated images denote their subjects through the same pictorial conventions human-made images employ. A diffusion-model landscape denotes a landscape; a language-model description denotes the object described. The denotational function is preserved across the transition from human to machine production, because denotation depends on conventional symbolic reference, not on the producer's consciousness or intentions. What changes is not whether the symbols denote but what the denotation achieves in the context of a worldmaking project—and this change is consequential in ways that the surface preservation of denotational function conceals.
The limit of denotation is that it is the thinnest form of reference—the symbol points to its subject but does not, by denotation alone, tell you what to attend to about the subject, what properties matter, what understanding to extract. A portrait that merely denotes—that picks out the sitter without exemplifying visual properties or expressing qualities—is deficient as art, because art's cognitive contribution operates through the thicker referential modes of exemplification and expression. AI-generated images reliably achieve denotation. Whether they achieve exemplification and expression—whether they possess and highlight properties in ways that yield aesthetic understanding—is the question Goodman's framework reveals as the critical one for evaluating AI art.
Goodman's analysis of denotation developed across multiple works, beginning with The Structure of Appearance and reaching full articulation in Languages of Art (1968), Chapter I. The argument built on the work of C.S. Peirce (whose icon/index/symbol taxonomy Goodman rejected as insufficiently rigorous) and of Rudolf Carnap (whose extensional approach to semantics Goodman adopted and refined). The key move was the demonstration that resemblance cannot ground pictorial reference—a move that separated Goodman from the phenomenological tradition (which treated perception as immediate and pre-conventional) and aligned him with the structuralist insight that all reference operates through conventional systems.
Denotation is conventional. Symbols denote their subjects through established conventions of reference, not through natural resemblance—pictorial denotation requires learned perceptual habits as much as linguistic denotation does.
Denotation is asymmetric. A portrait denotes its subject; the subject does not denote the portrait—the directedness of the referential relation distinguishes denotation from resemblance.
AI preserves denotational function. Machine-generated images and texts denote their subjects through the same conventions human-made works employ—the surface referential function is maintained.
Denotation is not sufficient for art. The cognitive contribution of art operates through exemplification and expression, not through denotation alone—pointing to a subject is not the same as constructing understanding about it.