What a Variation Keeps

29 Aug, 2026

Variation and sampling both make new work from existing material. Variation transforms musical relationships, learned from a score or by ear. Sampling reuses recorded sound, which can carry those relationships along with the qualities of a particular performance.

Musicians have long made new work from material they did not invent. A tune can become a set of variations, and a recorded drum break can become the foundation of another song. Both invite us to hear something familiar differently. Grouping them together as borrowing hides the more interesting difference, which has less to do with how much a musician changed than with what was carried forward in the first place.

Variation offers the older way of working with repetition and difference. A musician retains some features of a theme and alters others, perhaps its rhythm, harmony, texture or melodic decoration. The theme need not belong to someone else. Nor must the process happen spontaneously. Variation describes a relationship between musical materials, and improvisation describes how a performer makes decisions in the moment.

Bach's Goldberg Variations, published in 1741, show how much can change around a continuing foundation. The aria's bass pattern and harmonic plan support 30 variations whose surfaces differ dramatically. Following the opening melody alone will not explain their relationship. The continuity lies deeper, allowing contrasting musical characters to belong to one design.

Mozart took a more immediately recognisable starting point in his 12 variations on "Ah, vous dirai je, Maman," written in Vienna in 1781 or 1782. The borrowed French tune gives listeners a reference against which to follow his alterations, so the thing being preserved sits right on the surface. Beethoven expanded the scale of such an encounter in the Diabelli Variations, completed in 1823, where a waltz by the publisher Anton Diabelli became the basis of 33 variations. He called them Veränderungen, transformations, the same German word Bach had used for the Goldbergs rather than the Italian-derived alternative. The starting material sets a task, and the composer discovers how much it can sustain.

Rachmaninoff shows how far that can run. Paganini's Caprice No. 24 had already drawn Liszt and Brahms before him, and in the Rhapsody on a Theme of Paganini, from 1934, the 18th variation inverts Paganini's theme so that rising intervals fall and falling intervals rise, moves it into D flat major and slows it into one of the most familiar melodies of its century. Listeners who do not know the relationship generally do not hear it.

The connection is demonstrable rather than obvious. Set Paganini's theme against Rachmaninoff's inversion and the correspondence can be traced interval by interval, though whether a listener then hears the 2 as the same idea is a separate question and not a settled one. What a variation keeps can sit in a familiar melody, in a harmonic pattern, or in a sequence of intervals turned upside down. Where it sits moves from work to work, on the surface in Mozart, beneath it in Bach, and well out of ordinary earshot in Rachmaninoff. Recognition depends on what a listener has been given to listen for.

Elgar carried the idea one step further, and possibly further than anyone can follow. His Variations on an Original Theme, first performed in London in June 1899, sketch 14 members of his circle in music, himself included. In a note for that premiere he also warned his audience that through and over the whole set "another and larger theme 'goes', but is not played." He never said what it was. A century of proposals has offered Rule Britannia, Auld Lang Syne, Ein feste Burg, the Pathétique sonata and the suggestion that the hidden theme is not a tune at all but an idea, and no proposed solution has won general acceptance. If Elgar was telling the truth, the set preserves something nobody has ever located, and it coheres anyway.

Beethoven and Elgar themes compared

The opening notes of a Beethoven theme (top) recurring in Elgar's theme (bottom), one of the resemblances advanced in the long argument over the Enigma. Diagram by Taxus2, 11 August 2017, licensed under CC BY-SA 4.0.

Recording introduced a different possibility, and the examples that follow establish points of comparison rather than a single line of descent. Studio experiments in postwar Paris and dance parties in the Bronx arose from different needs and communities, and classical variation confers no legitimacy on anything that came after it.

Pierre Schaeffer arranged recording sessions at the Batignolles railway depot and built Étude aux chemins de fer from the results, broadcasting it with 4 companion pieces in October 1948 among the first works of musique concrète. Repetition helped redirect attention from the source of a noise to its audible qualities.

What matters for the comparison is the starting material. Schaeffer began with recorded railway sounds rather than a composed theme, and those sounds already carried rhythms, pitches and textures he could select and reshape, which he did freely, transposing them by playing 78rpm discs at 33. Paganini's caprice exists apart from any performance of it, and a musician who learns it can produce a sound of their own. The Batignolles locomotives exist only as that recording. Recording had made a particular event available for repeated handling, which is a different inheritance from a theme, not a poorer one.

Hip hop developed its own practices of reuse in a different social setting again. In the Bronx during the 1970s, DJ Kool Herc worked with 2 copies of the same record on 2 turntables, cutting back and forth to extend the instrumental break and give dancers more time with the passage they wanted. What Herc inherited was a passage of a record, the same few bars arriving twice over, and he had no means of changing what was on the record, which makes it the event in close to its purest form. The turntable became an instrument for selecting and reorganising recorded performances, and the technique anticipated by years what samplers would later do electronically.

Digital samplers gave producers further ways to isolate, repeat and rearrange sound, and J Dilla built unusually expressive relationships among his samples and his drum hits. Rather than layering many sources, he tended to cut a single sample into fragments and rebuild them into a groove the original did not contain. He also worked without quantization at times, leaving off the function that pulls each drum hit onto the machine's rhythmic grid, so that the placement of his fingers on the pads survived into the finished track. Questlove compared the effect to drumming by someone drunk. Departing from the grid did not empty the music of structure, and substituted a looser organisation for a regular one.

The distinction now has a shape, and the practices overlap rather than divide. A sample may carry melodic and harmonic material, or it may carry nothing but noise or a single drum hit, and either way it brings the sound of one performance with it: a drummer's timing, a singer's grain, the acoustics of a room. A newly performed quotation recreates whatever relationships the borrower has grasped, by ear or from a page, and leaves that sound behind. Rachmaninoff turned Paganini's intervals upside down by understanding the relationship, which a musician can do without writing anything down. A producer who loops a break has taken an afternoon in a room along with it, and processing can bury that past recognition without altering where it came from.

Recognition can enrich the experience either way. A listener who discovers a sample's source may hear both recordings differently. Credit helps make that relationship visible, and the new setting can direct attention to a detail previously passed over. Borrowing can honour, question or simply exploit its source, and the technique alone settles none of those judgments.

Asking whether such a work is original turns out to be 2 questions wearing one phrase. Variation concerns how musical material changes. Sampling concerns the reuse of recorded sound, and a single work can do both at once, which is why the presence of borrowed material tells us where part of it began and very little else. What it does not tell us is which kind of borrowing is at work, and until we know that we do not yet know what to listen for.

My blog

What a Variation Keeps

Further Reading