The Boy Who Found Vermeer Boring

Taste is a posterior

A child in a viral reel calls Girl with a Pearl Earring boring and says The Scream gives him a feeling. The easy response is to call it innocence. It is more useful to treat it as a test case for a harder question: is there such a thing as good taste at all?

I. The reductio

Suppose there is not. Paul Graham follows that premise to its end: if no taste is better than another, no art is better than another, and so no artist is better than another. We could have popular artists but never good ones. Almost no one accepts that, so the premise fails. Taste exists, but as a matter of degree. Objectivity, Graham argues, depends on how much the observers have in common. It is not a binary.

II. The child

If taste is a spectrum, the child sits at its subjective end. His reasoning, unstated, runs: good paintings move me, this one does not, therefore it is not good.

This is the prosecutor's fallacy: he reverses the conditional probability, mistaking "good work, so it moves me" for "it moves me, so it is good."

His dataset is small. In it, emotion looks like a face doing something: Munch's open mouth, hands pressed to cheeks. Vermeer's girl shows no emotion in that sense. What she offers is a pearl built from almost no line, a glow that seems to come from the paint itself, and a technique that took a lifetime to earn. None of it registers, because none of it is in his training set. As Callum Flack puts it, taste is know-how compressed into feeling, and you can't feel what you haven't yet compressed. The child is not wrong so much as uncompressed. He is also using one yardstick, expressed feeling, and applying it to every work.

III. The same error in adults

The child's mistake is that he has fixed the criteria in advance. But the factors that make a work good are not constant. They vary from work to work, and often from genre to genre.

Consider Olivia Dean. If the factors for a great contemporary artist were rapping and beatboxing, she could not qualify. That is not a verdict on her, only on the measure. The same holds for artworks. If feeling were the only measure, Rococo would be a failure. If narrative were, abstract art would be unanswerable. Each work sets its own terms, and part of taste is knowing which terms to apply.

This is where the adult version of the fallacy appears. The crowd applies one factor, fame, to everything. Take the Mona Lisa and Géricault's Raft of the Medusa, hung in the same museum. Far more people photograph the first. Is it only what is painted that separates them? Part of the answer is curiosity fed by hype. After the theft the story drew people in; if it drew two, those two drew four to see what they had seen, and the growth compounded. That does not make the painting overhyped. It means the other reasons to admire it, the ones its own factors ask for, are undervalued beneath the noise.

Fame is real evidence, but it appears around works of every kind, so it cannot decide alone. Brie Wolfron adds that taste is not correctness and not richness. Knowing the right names and paying the right price are checklists anyone can complete without feeling anything, which is why they carry less weight as evidence.

IV. Why taste cannot be installed

Bayes' theorem explains why. A judgment is a prior updated by evidence, and the quality of the posterior depends on the quality of the evidence, which depends on the dataset. Good evidence is also specific: what counts for one work may be irrelevant to another, so the same dataset can mislead when carried unchanged from a soul record to a canvas.

Now suppose the dataset is supplied by someone else. Algorithms suggest, brands curate, social circles approve, and each narrows what you meet. If the evidence you update on has already been filtered by other people's preferences, your posterior reflects their priors, however personal it feels. That is the failure of the formula: the right clothes, the right art, the right music, without knowing what actually intrigues you. Suggestions shrink the training set until taste becomes a reflex.

A dataset, in this sense, is not raw data. It is accumulated experience, and it is personal. Taste tracks knowledge: the more you know of a subject, the more relationships you see between its parts, and those relationships let you predict outcomes, which intuition feels before reason can explain. Intelligence and intuition work together, and neither can be fed in overnight.

So the dataset has to be made, not received. Curiosity is the signal that is expensive to counterfeit, since nobody stands for an hour before a painting they do not care about. Correctness and richness belong in the calculation as evidence to be weighed, not as verdicts.

V. A flow state, not a destination

That is why taste is never fully achieved. Flack describes it as compression and acceleration through cultivated attention: the deeper you attend to a problem space, the more know-how turns into feeling, until judgment arrives before language does. Building it has the shape of a flow state. It is entered by absorbed looking, it cannot be rushed, and it does not finish.

It also develops unevenly, as Wolfron notes. Someone exact about painting may have no preference in coffee. There is no single faculty called taste, only problem spaces and a history of attention in each. Irony and satire are its frenemies, because they substitute posture for attention.

VI. Back to the child

The child is not wrong to be moved by The Scream. That is his first step in building taste. The remedy is not to tell him he should like Vermeer. It is to keep him looking, at the boring painting, the unfamiliar one, the one that resists, until his set of things to be curious about is large enough that the girl with the pearl earring finally has something to say. And he might still stand by his original point. That would be taste: a judgment held with a full dataset behind it.

Taste, then, is not a verdict handed down by the market or the rulebook. It is a posterior, built by attention, and always open to revision.