Duchamp's “Fountain” vanished almost immediately; perhaps thrown away, perhaps smashed to pieces by the outraged committee at the Society of Independent Artists, which had opened an exhibition to all comers only to find Duchamp calling its bluff. Yet as we know, the story does not end there.

“Fountain” was copied: 15 replicas adorn art galleries, each endorsed by Duchamp. But more interesting are the transformations of “Fountain” that produce something new - for instance, Andy Warhol's “Campbell's Soup Cans” or Pablo Picasso's “Bull's Head”, to say nothing of the idea that a work of art could be just that: an idea.

It has become fashionable to assert - as the writer and director Kirkby Ferguson has done in his films - that “everything is a remix”. All creative acts, he says, copy, combine and transform earlier ideas. It's a convincing thesis. Gutenberg's printing press was inspired by a wine press, while Apple's Macintosh borrowed from Xerox's Alto, Nirvana's “Smells Like Teen Spirit” transforms a riff from Boston's “More Than A Feeling” and George Lucas's Star Wars owes a debt to Akira Kurosawa's The Hidden Fortress.

But what is it about an idea that makes it remixable? It's worth distinguishing between an idea that provokes lots of derivative work, and one that inspires something new and exciting. “Fountain” did both but perhaps there's a trade-off between fecundity and the ability to inspire original successors. “Apache” by the Incredible Bongo Band has been much sampled but hardly inspired a generation in the way that the Velvet Underground did.

Andres Monroy-Hernandez, now a researcher at Microsoft, and Benjamin Mako Hill, a hacker and researcher at the University of Washington, have conjectured that there may be a “remixing dilemma“: the elements that encourage people to appropriate and adapt a previous work are, alas, not conducive to originality in the new work.

Mako Hill and Monroy-Hernandez suggest that an idea that is fairly simple, that comes from an already-famous creator, and that is itself a remix, will tend to spawn many imitators. But these are precisely the qualities - moderate simplicity, notoriety and being part of a chain of remixes - that might reduce the originality of further derivative work.

Mako Hill and Monroy-Hernandez have tested their hypotheses in one particular setting, Scratch, a child-friendly programming language with a strong community. Scratch programmers are encouraged to share their programs and to use those of others as a basis for further work. A rich dataset is available, allowing the researchers to compare the complexity, popularity and (to some extent) the originality of programs shared on the site.

In Scratch, there does seem to be a remixing trade-off: more famous community members find their work remixed a lot - often in trivial ways; the same is true for already-remixed projects. But the trade-off is less apparent on the important metric of complexity. More complex programs are remixed more often, yet also with more originality, than simple ones. This is surprising given the conventional wisdom in open-source software that it is best to release simple, early versions to encourage the community to improve on them.

Then again, Duchamp's idea could hardly have been simpler - and nobody could suggest that subsequent artists have ignored it.