Fully/Partially Improvised Works

I’m wondering how best to represent work level credits for work that’s fully improvised or partially improvised.

For example, Made to Break, a jazz group, has compositions by Ken Vandermark, but the entire group is also credited as ‘music improvised by’. My guess is that these compositions are fairly loose, potentially grooves, or various ideas that can be interpreted on the spot to create the piece.

This is slightly different than pieces that are entirely improvised. see, for example, pieces featured on this release: https://musicbrainz.org/release/a8ce6de3-0ae8-4229-8e02-1aa45df8c4c0 In this case the pieces are fully improvised, so I gave each member of the group a composer credit, as it seems like they were composing on the spot. But this seems a strange solution.

In the case of Made to Break (which I’m in the process of cataloging), some of Vandermark’s compositions are played on other live releases, so it would make sense to link them. One solution would be to create a new work that represents Vandermark’s score, and have the various improvisations of it as spontaneous arrangements, but, again that relationship seems misapplied.



I think one of the essential qualities of a Work is that it can be performed again. There is a grey area with partial improvisation.
But “fully improvised”, even with a common theme/focus/name, would say to me that the performances would not be the same Work.

1 Like

Well, no, I wouldn’t think fully improvised works would have much potential to be performed again (though they could, like the various arrangements of Metal Machine Music). But I still think it’s important to reflect the creative role the various improvisers have. I’m just not sure if we have anything better than ‘composer’ though.

1 Like

“Performer” in improvisation at least, and if I read Glenn Gould correctly all the way across the performance of scored works, trounces “Composer” thoroughly (IMO).

I’m probably going too far but GG seems to view the composition as something that the performer is using as an aid to their creativity - an voluntary limitation to spark higher performer creativity.

I see something similar in composer = [traditional] - variations by the performer are expected and usually non-remarkable (as in usually not claimed as Arrangements).

These are very much my views and I have no idea what wider MB community views are on this. I’d be interested to hear more views.


That’s a good point about Gould, but I think free improvisation usually comes with the expectation that there’s a greater level of contribution to the work, or at least a different level of contribution.

At any rate, would something that’s fully improvised just not have a ‘work’ entity?

I would generally not create a work at all for a full improvisation.

1 Like

If the so called work is only recorded once, in particular. → no works.

If the work is recorded a second time, then it’s no longer a full improvisation (except if it only shares the title but there are no common musics inside), it can have a MB work, I guess.

1 Like

Maybe an improvisation attribute to subtypes of the Artist-Recording relationship type Performance would allow to credit improvisers, just like there is a solo attribute for the Performer to credit soloists.

1 Like

Maybe, though one could argue that that would be implied for any jazz work, or even as @mmirG pointed out, any performance.

Personally, I would rather have that attribute attached to ‘composer’ at the work level. This seems a controversial opinion, but I don’t think the definition of a work should hinge on its ability to be recreated. I realize that that seems like the most sensible definition from a database perspective (why have works that necessarily only correspond to one performance), but it doesn’t make sense to me from an ontological perspective.

It’s also problematic, in some cases, from a publishing perspective. For example, the box set listed above (https://musicbrainz.org/release/a8ce6de3-0ae8-4229-8e02-1aa45df8c4c0), all of the pieces are published and copyrighted by the improvisers. I didn’t reflect this in the records for the works, because I got a bit lazy, but despite the fact that it’s unlikely that the works will be reproduced at any other time, they are still viewed as a creative work subject to copyright.


I very much hesitated to reopen this 6y-old topic but I ran into this problem after adding https://musicbrainz.org/release/386da5a7-fa67-4ff5-bf14-7c55cc5ae3ec
a 3CD release where the recordings of CD2 & 3 are improvisations.
I did not add works yet, but they all have ISWC entries referencing the 4 group members with ‘C’.

by a label as seen on cover art
https://ia801208.us.archive.org/12/items/mbid-386da5a7-fa67-4ff5-bf14-7c55cc5ae3ec/mbid-386da5a7-fa67-4ff5-bf14-7c55cc5ae3ec-37495453599.jpg beneath the lineup and
https://ia601208.us.archive.org/12/items/mbid-386da5a7-fa67-4ff5-bf14-7c55cc5ae3ec/mbid-386da5a7-fa67-4ff5-bf14-7c55cc5ae3ec-37495450190.jpg on lower left.

So add works or not? :thinking:
Edit: Maybe with disamb ‘Improvisation’.
Any help welcome.

okay, since it’s been opened, I’ll take this as a chance to add my two eurocents.

An improvisation which is performed to a live audience and then lives on in their memories is hardly recreatable, however, a recorded improvisation can be analyzed and thus recreated, to a much higher degree than from memory. I don’t think the ability to be recreated should define a work, because the possibility to recreate a piece of music is not only a function of the complexity of the music itself, but also of the recreator’s cleverness, skill and, I suppose, free time.

Additionally, improvisation and composition can be reduced to the same core. Improv is just composition that happens very fast and without much thought, but rather with feel. I reckon that if improv can be viewed as a subset of composition, it shouldn’t need to be distinguished from it. I’m pretty sure there are many “composed” works which are really just polished improvs and unless the author tells us, we’ll be none the wiser. However, I can also see the benefit of having an improviser subtype of the writer, or even of the composer relationship.

Based on this, a work that is recorded only once shouldn’t be disqualified from having a work. Very many songs are recorded only once and then never again (because they’re not very good, perhaps). If one would agree that improvisation is composition, then there should be no difference between a work for an improvisation recorded only once, and a carefully constructed song recorded only once.

Apologies for the wall of text, I’d personally add works in any case, and shamelessly use the composer relationship. It’s the closest we have and it’s not even very far, in my opinion.


I agree with your wall of text. Some artists can’t write music. If they go into the studio and lay a new track down which then gets released on the album, this will get a work as it was released.

If they do the same at a concert, and that concert was being recorded for a live CD and is released, I don’t see what is different. They created something new in the same way.

If they keep messing around with that track in different ways at every gig then it still sounds like a work to me. I know the early Pink Floyd did a lot of this. New pieces would grow organically. It is one of the reasons why live bootlegs are so popular as you hear music grow into the final form.

A lot of music starts from “I was messing around on the guitar and I got this riff…


I think that’s perfectly correct. In jazz there are numerous improvisational works that receive official work credits on cover or in the ISWC database. And I know some that were actually recreated, e.g. Work “Köln, January 24, 1975, Part IIc” - MusicBrainz


dropping in with another jazz example, My Favorite Things by John Coltrane. it’s a nearly 14 minute improvisation based on the Sound of Music song of the same name, but bears little resemblance to the original, yet currently on MusicBrainz, it’s a recording of the original work

I don’t know if there’s any direct covers of the work, but there certainly are recordings of songs that are clearly referencing the Coltrane song and others which clearly don’t respectively

edit: maybe scratch that last point, at least for jazz versions of My Favorite Things, I apparently didn’t read the title for the Postmodern Jukebox song… lol

for more info, here’s Polyphonic’s video on How Coltrane Broke “My Favorite Things”


Meanwhile, Work “My Favourite Things” - MusicBrainz by Stephen Hough is listed as it’s own work based on My Favorite Things. I don’t see how the Coltrane one makes sense.
Maybe there should be a relationship to track “recording based on work” as opposed to “work based on work”.

1 Like

Here are a couple more examples.

Recording “Dies Irae” by Sky - MusicBrainz and Recording “Dies Irae” by Barbara Laronga - MusicBrainz are listed as recordings of Song “Dies Irae” - MusicBrainz. It is unclear to me whether they are improvisations, but not simple recordings or transcriptions of another work. There is new creativity involved.

That’s no free improvisation.
Barbara Laronga certainly improvises, but she improvises on the work “Dies Irae” and it’s the same work as that interpreted by Sky. See and listen on Wikipedia: Dies irae - Wikipedia

Of course, these could be arrangements, but we don’t usually use that for jazz (or pop). It would cause too many arrangements.

This recording would be better given its own work. Credits on release¹ say writer (Monkman). It has its own ISWC: T-010.476.458-0 TRAD. / MONKMAN FRANCIS

Therefore a new work based on the traditional.

¹) https://ia800507.us.archive.org/11/items/mbid-08d2ba1a-8788-43de-91c5-dfe36c008fd0/mbid-08d2ba1a-8788-43de-91c5-dfe36c008fd0-15290851679.jpg

EDIT: It already has its work: Work “Dies Irae” - MusicBrainz (I added ISWC and based on relationship)

1 Like

Sounds right to me - just voted yes on your edit.

1 Like

And we tend the other way for classical - a huge number of arrangements is the exception and we like to know which version of a piece is performed.
I was confused to find that MB’s only recordings of Dies Irae were clearly not the Gregorian plainchant until I linked Recording “Dies irae” by Aurora Surgit, Alessio Randon - MusicBrainz earlier today.

1 Like