Aesthetic Taste Becomes a Score Only After Its Judgment Is Narrowed

Immanuel Kant begins with a person saying that a painting is beautiful, and immediately makes the sentence much more difficult than it looked.

The pleasure is disinterested. It does not want to eat the painting, own it, or recruit it to a moral campaign.

Yet its speaker still asks for agreement.

The verdict comes from feeling, not a determinate concept, and it behaves as though others ought to find the work beautiful too. Taste has arrived carrying a demand and no proof.

Kant gives this peculiar confidence four moments. Beauty is separated from appetite, given subjective universal validity, grounded in purposiveness without a purpose, and finally attached to exemplary necessity and a sensus communis.

Then he lets the thing generate its own administrative emergency. Taste rests on no concept, and yet a dispute about taste must have some basis.

The antinomy is not a decorative crack in the furniture. It is the furniture admitting that it creaks.

Hume had already made the predicament less cozy.

Sentiment has no external standard in the way a measurement does. Nevertheless his qualified critic needs strong sense, delicate sentiment, practice, comparison, and freedom from prejudice.

Neither Hume nor Kant mistakes I like it for the end of the matter. They merely decline to pretend that agreement can be manufactured by a ruler.

Psychometrics enters later with a ruler, a questionnaire, a pilot sample, and a justifiable suspicion of anyone who calls a score self-explanatory.

Spearman’s work on correlation and attenuation gives classical test theory its sober little equation. An observed score contains a hypothetical true score and random error.

Reliability is the proportion of observed-score variance attributed to the true score rather than the error.

This is a real achievement. It is a thoroughly different achievement from deciding whether the person has noticed grace, balance, garishness, or only a familiar face.

There is one perfectly reasonable wish behind a taste scale. Make private-seeming responses comparable, without treating them as weather reports from separate planets.

Ask about artworks, styles, reactions, or discriminations. Sum answers. Look for a stable pattern. Then give it a name.

The name is usually where the little cosmic fraud puts on a lab coat.

Cronbach and Meehl saw the danger.

A test can correlate respectably with something and still not have earned the name it has been given.

A construct requires a nomological network: linked theoretical claims, other constructs, observable indicators, and correspondence rules.

Reliability is an achievement, not an answer.

The construction sequence makes the rebuke procedural.

Begin with a written construct definition, then an item pool, expert review, pilot testing, item analysis, factor analysis, and reliability and validity testing.

A purported one-factor taste scale that yields three factors has not discovered complexity with a tiny trumpet. It has encountered a question its definition did not answer.

So the reconciliation seems to be waiting on the table.

Kant asks how a felt judgment can claim common standing. Psychometrics makes individual answers public numbers. Surely a reliable scale of taste has finally measured aesthetic value, with the sensus communis upgraded from philosophical furniture to a factor score.

It has not.

A scale can measure a narrower object exquisitely. Consistency in stated preference. Sensitivity to a specified formal relation. Familiarity with a tradition. Convergence with a critic panel. Willingness to endorse an item written by somebody having a very productive afternoon.

These are not equivalent objects. The reduction occurs before the reliability coefficient arrives to certify its paperwork.

The aesthetic theories make this impossible to ignore.

Tolstoy defines art through genuine transmission of felt emotion. Collingwood makes expression the clarification of an emotion not initially clear even to its maker. Bell finds aesthetic value in significant form.

Each provides a different candidate for what a taste response might be responding to.

A single factor cannot settle their argument any more than a thermometer can decide which poem was the fever.

Nor may a scale bypass interpretation by asking what the maker intended.

Wimsatt and Beardsley’s intentional fallacy does not say authorial intention is irrelevant. It denies that private intention is the final court for a work’s success.

Internal evidence from language and convention is legitimate. Private letters, diaries, and remembered conversation are external evidence that cannot settle the matter alone.

Hirsch disputes the conclusion. The verdict remains argued, which is annoying for a spreadsheet and healthy for a work of art.

Expression and interpretation therefore need not share a column.

Tolstoy asks whether a feeling passes to an audience. Collingwood asks whether making clarifies a feeling. Interpretation asks what a work means and what evidence bears on that meaning.

A preference inventory may record reactions to all three. It cannot, merely by returning the same number twice, announce that they are one attribute.

The ordinary psychometric answer is not despair. It is more exactness.

Item response theory shifts attention from one test-wide precision estimate to the probability of each response, given an item’s properties and a proposed trait level.

Simpler models use difficulty alone. Later ones add discrimination, then guessing.

The machinery becomes finer. Its metaphysics does not become automatic.

And a taste scale carries a particular hazard, because taste is where a tradition is doing much of the work.

An item written from inside one canon can measure familiarity with that canon and return a stable number anyway. Stability is not evidence that the number means the same thing to everyone answering it.

There may be a scale whose latent factor is identical with aesthetic capacity. That is speculation.

Factor analysis finds patterned answers. It does not discover a little faculty of taste sitting behind the eyes, waiting to be issued a badge.

What both traditions share is not that measurement has solved aesthetics, nor that aesthetic judgment is too delicate for measurement.

Both refuse private whim as the final court.

Kant asks why a feeling can demand company. Cronbach and Meehl ask what warrants attaching an attribute-name to a pattern of answers.

One asks a verdict to answer to a shared world. The other asks a score to answer to the construct it claims.

The first task makes criticism possible. The second makes comparison answerable.

Neither becomes the other merely because both can be made to sign a form.