Representation always involves a certain degree of abstraction—that is, the taking away of one characteristic or more of the original. Even a fairly realistic painting of a person, for example, lacks some features that characterize actual persons: a painting is two-dimensional, whereas every actual person is three-dimensional; the surface of a painting is paint, but not so the person; actual persons have very numerous pores and other marks on their faces that are lacking (in whole or in part) in a painting, and so on. The depiction of a person in a painting is usually sufficient to enable human viewers to recognize the figure as a person—though it is apparently not sufficient for most animals, who see only a coloured canvas where people see on the coloured canvas a representation. When the degree of abstraction is so great that it is no longer possible to recognize this shape as a human shape or as the shape of any identifiable object, the painting is then spoken of as non-representational. (In popular parlance such paintings are called abstract, but this is misleading, for abstraction is a matter of degree, and, as has just been shown, all depictions are necessarily abstract—that is, abstracted from reality to some degree.) The actual object with all its millions of qualities is at one end of the spectrum, and the painting so abstracted that a depiction subject is unrecognizable is at the other end; between the two extremes lie all the possible degrees of abstraction.

Literature can be representational but not in the same way as visual art. It is quite natural to say that in a novel or drama a number of characters and actions are represented. The representation is, of course, not a visual one; it is representation through language. The painter portrays Napoleon by making a portrait of him, while the writer does so by describing him in words. The writer, unlike the painter, can also depict action. Not all literature, of course, is representational in this way: a sonnet may contain no characters at all and no action, consisting solely of an expression of feeling by some unspecified speaker.

Any of the mixed arts that include words as part of their medium, such as drama or film, can be, like literature, representational. Indeed, they have a further advantage: they can depict action not only through words but also by showing the characters and exhibiting the action before the spectator. These arts are visual as well as verbal, and since they are not limited to one moment in time, as painting and sculpture are, they are temporal arts as well as spatial. These mixed arts, then, can be doubly representational.

Is it possible for music, too, to be representational? Music cannot visually show characters or objects, nor can it describe them in words. Can it “depict them in tones”? Program notes at concerts usually assume without question that it can. The audience is told about the tone poem Don Quixote, by the Austrian composer Richard Strauss, “The composer has given us a musical representation of the Don’s adventures. The 17th-century Spanish writer Miguel de Cervantes has described them in words, and Strauss has done so in tones.” But the claim to representation in music is, to say the least, quite dubious. Without the title, with the music alone, would there be any clue that the music was supposed to be “about” the adventures of Don Quixote? True, there is a passage that resembles the bleating of sheep sufficiently for that much to be guessed, but even to conjecture that this passage is a representation of sheep bleating is a far cry from being able to reconstruct the entire story. Suppose that Strauss had left every note in the score just as it was but changed the title. Would the piece then have been a representation of something else? The very fact that this question can be asked shows how different music is from visual art. If a painter has drawn a house but indicated in the title that it was supposed to be a tree, viewers could still say, on the basis of what they saw in the picture, that it was not a tree but a house. But in music the listener is never in this situation: if a listener says that this series of tones represents the adventures of Don Quixote, then that is because of the title Strauss used. If the composer had given it no programmatic title, one listener might think of one represented subject, another a different one, and a third none at all, and there would be no way of showing who was right or even whose opinion was to be preferred. The conclusion seems to be that music by itself—without title, without words, without depicted action (as in a combination of music and drama such as opera)—is incapable of representing anything. There is simply a series of musical tones that may suggest differing associations, programmatic or otherwise, but the musical tones by themselves cannot be said to represent anything at all.

This might be objected to as an overstatement. If a picture can represent a house by looking more like a house than anything else, cannot a work of music represent the sea by sounding more like the sea than any alternative? And is this not the case in, for example, the French composer Claude Debussy’s tone poem La Mer? Even this, however, is highly questionable. Almost no one guesses the title to Debussy’s tone poem without first knowing what it is; it may seem obvious enough after the composer has channelled the listener’s response by means of a title but not beforehand. And surely this is because the sounds in the tone poem do not sound more like the sea than like anything else: the tone poem consists, after all, of a series of complex musical tones, emitted by violins, cellos, clarinets, flutes, trumpets, and so on, and it would be difficult indeed for these sounds, which are musical tones, to sound very much like the sea, whose sounds consist after all of a series of complex noises. There is no great similarity between any one series of musical tones and any one series of nature’s noises. Hence, the first cannot be said to constitute a representation of the second.

The matter is even more obvious in the case of those numerous programmatic titles in which the supposedly represented subject contains no sounds at all. Debussy’s Reflets dans l’eau (Reflections in the Water) is taken by some as a musical representation of reflections in the water. But reflections in the water emit no sounds at all, not even noises. No one, then, could say that the sounds in Debussy’s piano composition resemble the sounds of reflections in water. The resemblance, if there is any, is much more remote: it may be that the feeling obtained when Debussy’s composition is heard is somewhat like the feeling that arises when reflections in water are seen. This is highly improbable without knowledge of the title, but at most it would provide a mood resemblance, which is far removed from a representation by music of things in the world. The conclusion seems inescapable that music is not to be classified as a representational art, at least not in the same straightforward meaning of “representation” that applies to the other arts.

So much, then, for the capacities of the various arts as far as representation is concerned. But the question remains: in those arts that are properly called representational, what should be the nature of the representation?

That art should be an outright duplication (incorrectly called “imitation”) of reality is a view that was put forward by the French novelist Émile Zola in his book Le Roman expérimental (1880; The Experimental Novel) and has been occasionally held (though not practiced) by painters reacting against Romanticism, such as the 19th-century French artist Gustave Courbet. Zola advocated a novel that resembled a scientific investigation into reality. Plot was to be of no importance, rather an aspect of reality was to be examined searchingly, and from this the story would unfold without imaginative effort. Persons or groups of persons would be depicted, and from them the action would evolve.

It would be impossible, of course, to carry out such an ideal of art as “report” and undesirable even if it were possible. First, the author or painter must select a subject and, within the subject, must select which details to treat, for they cannot all be described even in a hundred lifetimes: since every object and event has an indefinitely large array of qualities, there is no point at which a description of it would be completed. Besides, the very language used (no matter how neutral a description is attempted) will colour the account. Even if the words were colourless, the mode of putting them together would yield a style, which would colour the account once again. Indeed, should such an ideal be achieved (as in the verbatim transcription of an actual trial) it would be the deadliest possible bore.

Art, even representational art, is not a reproduction of reality; it is a transformation of reality. How, specifically, is reality transformed in being represented in art? There is probably no general satisfactory answer to this question. Each art, each style of art, and each work of art transforms reality in its own way—the 19th-century French painters Paul Cézanne in one way, Pierre-Auguste Renoir in another; the 19th-century English writers Mary Shelley in one way, Emily Brontë in another. No set of rules can lead to predictions as to what transformation of reality will be conceived in the mind of the next creative artist. Reality is the common base, but each artist deals with it in a unique way.