If a writer describes a dog as saying woof, the word woof is onomatopoeia: the word not only references the sound a dog makes but also reproduces it. Onomatopoeia is one of the earliest figures of speech children learn, and is only one of many ways to use language's musical aspect to evoke things and even ideas. The back of my old copy of Smyth's Greek Grammar has an extensive catalog, including repetition, rhyme, rhythm, assonance, alliteration, and many more. Figures of speech are what distinguish artful use of language from merely referential use of it. What do you call it, though, when a thing or idea in the real world is reproduced in music? Strictly speaking, perhaps the answer should be "language," but that completes the circle without answering my question. It may be easier to explain what I mean with examples.
The most straightforward way to represent something in music is to dub it in. In the age of sampling, a musician can easily insert the sound of, say, a telephone ring or a clock alarm. "The Big Picture," by Bright Eyes, for example, starts with an extended recording of a car trip, including jingling keys, the reassuring clump of a car door slamming shut, and an ignition turning over.
At the next level of artistic transformation, the phenomenon that I'm talking about closely resembles onomatopoeia. At the end of The Clash's "Somebody Got Murdered," a single drum stroke seems to be meant to sound like a gunshot.
In Radiohead's "Videotape," the limited progression of piano chords repeated throughout the song suggest the rotation of a videotape, and the rhythm that the drum track settles into by the end sounds, at least to me, like the slap and flutter of the loose end of a tape circling on an old-fashioned reel-to-reel tape recorder (or of the loose end of a reel of film turning on a projector)—so that the song manages to evoke the recording technologies of two different eras.
Some of these "figures of song" represent the real world more abstractly. Regina Spektor's "On the Radio" is busy with words, and one of the song's pleasures is the way the words tumble over and into one another. When Spektor describes hearing the song "November Rain" twice during a single radio broadcast, however, she breaks her own talky rhythm by stretching out the word "sleep" in the short line "because the DJ was asleep" so as to reference the DJ's snooze. Note that she doesn't repeat herself to reference snoozing; her allusion, more abstractly , is to the way a sleeper's oatmeal boils over, or the way a sleeper's pencil flatlines if she drifts off while taking notes.
The most sophisticated example I can think of is in the refrain of Belle & Sebastian's "The Blues Are Still Blue," when Stuart Murdoch sings that when he meets up again with the girlfriend he has left in the launderette, "the black will be white and the white will be black, but the blues are still blue." The melody meanders up and down during the first part of the refrain, reproducing the circling motion of a laundry machine. But when Murdoch sings the words "The blues are still blue," the vocalists accompanying him sing all the words at a single pitch, reinforcing the concept of blue's identity with itself.
To call something like this "onomatopoeia" seems a little Procrustean. The sound of laundry isn't being mimicked; the melody is tracing the shape the laundry makes, and then the accompanying voices are tracing that of identity. It seems like another terminology is called for; maybe a musicologist will steer me to it?