What composers say about their scores

ShiroKuro

In @TulipAcid 's pedaling thread, there was a bit of discussion with @pianoloverus about Einaudi's score, and whether the pedaling directives in the I Giorni score where his directives, or were coming from an editor etc. I still haven't found that kind of info about Einaudi, but I just got ahold of the score for Madison (for cello+piano) by Ola Gjeilo, and inside the front cover is the following:

About Madison

At the recording session, only the cello part was written out - I played my piano part partly from memory, partly improvised. It was later transcribed from the recording by Thomas Barber and then edited by me.

I always recommend listening to the original recording before playing the work, as it conveys some elements that were challenging to notate, given the improvisatory, sometimes almost impressionistic nature of the piano part. You can hear the recording at my website, olagjeilo.com.

That's really interesting! I wish more composers included that kind of info about their scores.

Also, it's interesting to me because I don't feel like it sounds improvisatory, nor does the score look like a written-out improvisation. Although maybe that's because the cello part was written down and so the structure of the piece itself was set. (And of course he mentions "playing partly from memory").

The next part is interesting too -- basically "listen to the recording because there are elements not notated in the score."

BTW here's the piece. If my cellist friend is willing to do more music after we play together tomorrow, I'm going to suggest this one!

TulipAcid

ShiroKuro

Agree with Ola Gjello, in this era of technological marvels, we actually CAN hear the work performed by the composer (or conducted by, in the case of Gjello’s choral works…) and model our own performances towards the composer’s intent.

Gives future musicologists little to speculate and opine about, although I’m certain they’ll find something!

pianoloverus

TulipAcid Agree with Ola Gjello, in this era of technological marvels, we actually CAN hear the work performed by the composer (or conducted by, in the case of Gjello’s choral works…) and model our own performances towards the composer’s intent.

Gives future musicologists little to speculate and opine about, although I’m certain they’ll find something!

This is certainly reasonable but I would add some caveats:

The composer's performance could differ from the actual score approved by the composer.
The composer's performance is only his interpretation on that particular day. It could change a little or a lot the next day or a year later. I think, especially for a great work, many different interpretations are valid.
The composer could think that many interpretations are possible. I recently watched a video where Stephen Hough discussed a requited piece he wrote for some major piano competition. Since the piece was required in the first round, he was afraid that the judges and audience would become utterly bored after hearing it so many times. But he said this didn't happen because the performances were quite different but still very good.
The composer might think other pianists could give a better or at least equally good but different interpretation. I think Rachmaninov said that about Horowitz playing his third concerto.
In classical music, one can hear the actual performances of Rachmaninov playing some of his concerti. But although great contemporary pianists can hear those and probably listen to them, they usually don't choose to try to just copy them because they think they have their own
ideas that are equally valid.
Performance styles change over time. Some of Debussy's recordings of his own music sound so bizarre/distorted that one might conclude he couldn't count or maybe the recording mechanism was faulty. I haven't heard Paderewski's playing of his own works, but if they are like his playing of works by other composers, I doubt many of today's pianists would use his playing of his own pieces as model performances.

So IMO using a composer's recording as a very basic reference is reasonable but copying it may be less desirable.

TulipAcid

pianoloverus

That’s good news! While I’m interested in how to get and working towards “that sound” or “that feel” to a certain extent, as I am in no world or lifetime a performer of note, so I can, within reason, do my own thing.

pseudonym58

pianoloverus A slightly tangential topic, but in terms of historical recordings before ca. 1920:
Sometimes what is sold as a recording of a composer or pianist from the era is actually a piano roll, often a Welte-Mignon or Duo-Art roll designed to be played on a reproducing piano. Whether or not these reproductions actually represent anything close to what the artist "recorded" or would have played in live performance is speculation at this point. There are for instance early recordings of Mahler playing his transcription of the final movement of his Fourth Symphony as a piano roll, and it's blazingly fast. Recordings made by George Gershwin of Rhapsody in Blue are also extremely rapid. Those of Ravel playing various of his own works don't sound like anything that a man known as a perfectionist would have wanted released; they are awkward and uneven.
To what extent are these less than inspiring recordings inaccurate representations of the original players (technical limitations of either the recording or playback technology)? Who knows?

ShiroKuro

pseudonym58 I don’t know a lot about piano rolls, but I can remember reading about them in the past and just taking for granted that when someone said a roll was played by the composer, that it was the representative of the composer’s playing. So it’s interesting to learn how inaccurate that impression was.

On the other hand, you have these contemporary composers (post-classical of course, this wouldn’t hold as much for jazz), and certainly they make some changes in live performances, but their original recordings are often treated as the definitive version, and we do know that it’s them playing and that CD recoding is how they actually played it. And lots of sheet music and score books are marketed as “the version you hear on the CD.”

pianoloverus

pseudonym58 Your thinking about the inaccuracy is probably correct and I've seen this mentioned in other places. On another thread I mentioned that some of Debussy's recordings of his own pieces are so bizarre that I told myself I would never listen to them again. Some of his other recordings of his own pieces are much more normal sounding so it's possible that big difference is due to deficiencies of the recording method. Ravel, despite graduating from the Paris Conservatory, was not a great pianist so his recordings, especially of his more difficult works, will often not be very good even if the recording technique was OK.

MRC

When listening to a piano roll recording, it's important to understand what the recording technique was good at, and what it was not so good at. The big companies like Welte-Mignon and Ampico had developed very sophisticated machines that kept the roll going at a stable tempo and recorded with great precision the beginning and end of each note. This means that everything to do with timing - tempo, accelerandi, rallentandi, rubato, rhythmic subtleties, note lengths - was reproduced very faithfully.

It was much harder to record the dynamic variations of each note. Different techniques were tried, but they were not usually sufficient, and dynamics were often adjusted in post production. The same goes for subtleties of pedalling.

With that in mind, compare these two recordings Rachmaninoff made of his opus 39 no 6, first in 1921 on an Ampico piano roll:

and then an acoustic recording for RCA in 1925:

I'm very happy that both versions exist.

pianoloverus

MRC When listening to a piano roll recording, it's important to understand what the recording technique was good at, and what it was not so good at. The big companies like Welte-Mignon and Ampico had developed very sophisticated machines that kept the roll going at a stable tempo and recorded with great precision the beginning and end of each note. This means that everything to do with timing - tempo, accelerandi, rallentandi, rubato, rhythmic subtleties, note lengths - was reproduced very faithfully.

I have heard that piano rolls could be inconsistent with the tempo although I don't remember where I read that. Whether that applies to the type of rolls you mention or recordings by the great pianists I don't know. Here's what AI says although it's not so easy, at least for me, to draw a conclusion. I'd be interested in your comments about what it says. Do you think these comments agree or disagree with what you wrote or somwhere in between?

"The tempos on old piano roll recordings were moderately accurate for reproducing pianos, but heavily edited and standardized for basic player rolls. Key factors include human intervention during manufacturing, mechanical limitations of paper acceleration, and varying playback calibrations.Technical and Manufacturing InaccuraciesFactory Editing: Human editors regularly altered master rolls by hand, shifting hole placements and revising the length of notes to fix perceived mistakes or clean up a performance.Paper Speed Shifts: As paper wound onto the take-up spool, the roll's effective diameter increased, causing the paper speed to accelerate slightly over the tracker bar unless compensated for by the mechanism.Standardized Tempo Markings: Companies like Welte and Aeolian often stamped a generalized tempo guideline or a strict linear tempo track on the roll rather than capturing split-second micro-rubato from the original artist.Reproducing vs. Standard RollsHand-Played/Standard Rolls: These rolls featured printed red tempo lines that required the user sitting at the player piano to manually push or pull a tempo lever, meaning the tempo of the playback depended entirely on the amateur operator.Reproducing Systems (Ampico, Duo-Art, Welte-Mignon): Advanced reproducing systems punched dynamic and tempo perforations automatically during the artist's recording session. However, even these required studio technicians to fine-tune the master rolls post-session, frequently smoothing out extreme shifts in tempo or dynamic contrast.'

MRC

To go back to the original subject of this thread, here are some thoughts.

A musical work usually goes through many changes before being published. The composer alters a melody here, a harmony there, decides to cut a section or add some more bars somewhere else... Even composers such as Mozart, who we were told imagined everything perfectly in his head before writing it down, in fact made quite radical changes, as shown by sketched-out first versions of some of his works. There's probably always something you might be able to improve on, but at some point you have to say that this is the version that will be published, either because you decide it, or because you have a deadline to meet.

That doesn't stop the work living on in the composer's head, and particularly in the case of composers who perform their own works the music may continue to evolve. Maybe a composer likes to play their pieces differently each time: in this case, there isn't a preferred, "best" version. Maybe they end up with a version they really prefer to the one they published. That doesn't necessarily mean they will go to all the bother of creating a revised score and insisting that their publisher releases the new version.

Take those two Rachmaninoff recordings I just posted. Listen to the first few bars of both. The RCA (acoustic) recording has understandably better dynamic shading than the Ampico (piano roll) one, but otherwise they are very similar. Rachmaninoff starts slow and menacing, making a noticeable rallentando and diminuendo in the third bar, then plays fast from the 6th bar onwards.

Now look at the score, which was published in 1917:

Just Allegro from the start. Since he recorded this piece twice, both times with the first bars very free in tempo and at a much slower speed than the printed Allegro, I think it's safe to say that this is the way he had come to prefer it.

Why didn't he make a revised edition? Bringing out a new edition takes time and money, and he had other worries anyway. Probably he didn't think the difference was that important, and of course people could listen to his recording.

A pianist who plays that piece now has the choice: copy (more or less) the way Rachmaninoff recorded the beginning, or play it all Allegro. I prefer the more dramatic version in the recordings, but nobody can say that the Allegro version is actually wrong.

MRC

pianoloverus I have heard that piano rolls could be inconsistent with the tempo although I don't remember where I read that. Whether that applies to the type of rolls you mention or recordings by the great pianists I don't know.

I know that the companies who recorded the famous pianists of the day had sophisticated systems that kept the recording roll at a steady speed. The high-end reproducing pianos also had stabilising mechanisms to keep the speed constant. Cheaper models didn't, and of course the hand-cranked ones gave you all freedom to change the tempo at will. That must have been great fun!

As far as I know, the recordings of great artists ("reproducing rolls") really did capture all the rubato and tempo changes. Standard rolls were produced without live recordings: a technician would create a roll directly from a musical score, like a MIDI file. These rolls could have tempo and dynamic markings, which the person operating the player piano could follow (or not!). There was also something in between, called "hand-played rolls". Here, a pianist would record the piece in real time while somebody noted the dynamics, which would be added to the finished roll as annotations to be followed.

Regarding editing, the piano roll system permitted an ease of correcting mistakes that wasn't seen again until digital recording techniques became available. If a pianist missed a note during the recording, it was very easy to correct, and this was certainly done.

Certainly the most faithful recordings we can get our hands on are the acoustical ones made from the first years of the 20th century until the early 1940s, before magnetic tape started being used. When a recording was done in one take, with no possibility of splicing bits of different takes together, you know that what you hear is what the performer played.

pseudonym58

pianoloverus I am reminded of the very cheap little open reel tape recorders (toys, really) produced in the early 1960s that lacked a capstan (revolving metal spindle against a rubber pinch roller) speed control. You had the same issue with the changing speed depending on the amount of tape on the spool doing the pulling! If you were to take a tape recorded on one of those cheap machines and play it back on another more sophisticated machine with a capstan, the effect was amusing. I believe all cassette tapes, beginning with the ones developed by Phillips in 1964 or thereabouts, included a capstan. Sorry, this topic has certainly wandered! I'll stop here.