Accuracy to the score (Beethoven, Schumann, Bruckner, et al.)

[(poco) Sforzando]

I'll have to sit down at the piano sometime today and see how fast I take the slow movements from 22 and 10/3 myself. (Just to try to hear it with the metronome in my head may be inexact.) But here's Rosen:

on 22: "The Adagio is often taken too slowly: the basic rhythmic element is not the eight note but the dotted quarter."

on 10/3: "The extraordinary pathos presents a temptation to play too slowly, and pianists today often choose a tempo that would be unjustifiable with the weaker sustaining power of Beethoven's keyboards."

[both quotations slightly abridged to save keystrokes]

Not the last word of course, and Rosen's own sense of Allegretto causes him (in the "free companion CD") to take a tempo for the second movement of 54 that feels absurdly slow to these ears. Another controversy attends the first movement of the Moonlight: a very slow Adagio in 4, or a relatively flowing and faster one in 2? No specific comment from Rosen, but this is one piece where the pedals found on the fortepianos of B's day are needed, and where modern instruments fall short.

[zamyrabyrd]

I just listened to Gulda playing Op. 22. "Adagio" is also a mood and I don't believe he conveyed that in the slow movement. The pace is too quick to convey the "molt'espressione".

I would like to call attention to another "Adagio (ma non troppo)", much later of course, also in triple time, the Klagender Gesang of Op. 110. The units are 16ths with 12 in a measure. It is difficult to decide on a slow enough tempo if it is to be calibrated to the 6/8 fugue in between, marked "Allegro" but also "ma non troppo". In short, how does one play both at the same tempo and convey the differences in feeling?The next sonata has 9/16 in the arietta.

It's interesting of course that one of his first major expressive slow movements was the D minor "Largo e Mesto", Op. 10 No. 3 in 6/8. For the quick 32's not to sound over fast, the tempo at the outset must be quite slow. Comparing his other works in triple time might give a clue how to play them.

[(poco) Sforzando]

I just listened to Gulda playing Op. 22. "Adagio" is also a mood and I don't believe he conveyed that in the slow movement. The pace is too quick to convey the "molt'espressione".

I would like to call attention to another "Adagio (ma non troppo)", much later of course, also in triple time, the Klagender Gesang of Op. 110. The units are 16ths with 12 in a measure. It is difficult to decide on a slow enough tempo if it is to be calibrated to the 6/8 fugue in between, marked "Allegro" but also "ma non troppo". In short, how does one play both at the same tempo and convey the differences in feeling?The next sonata has 9/16 in the arietta.

It's interesting of course that one of his first major expressive slow movements was the D minor "Largo e Mesto", Op. 10 No. 3 in 6/8. For the quick 32's not to sound over fast, the tempo at the outset must be quite slow. Comparing his other works in triple time might give a clue how to play them.

I have the Gulda set and haven't listened to 22 for a while, but I often think he's rushing things and pushing the music too hard. The "molt'espressione" seems to call for an execution that is quasi-operatic in style, regardless of actual pace. As for 110, I think the Adagio and fugue ought to be proceed at the same pulse (see the arpeggios before the G minor version of the Adagio), whatever pace that might be.

[amw]

I believe the "standard tempo" suggestion is contentious because there are statements from Beethoven that "tempi ordinarii" would be a thing of the past and for "modern" music he was happy to have the metronome. But op.22 is still a fairly early piece, so the "standard tempo" might well apply.

So you would conclude that most of these movements are usually played too slow?

My view is that Beethoven's adagios and andantes are almost always taken too slowly, yes. So are most of the allegros. (e.g. dotted minim = 60 in the first movement of Eroica is rarely observed even by modern "HIP" performers—a metronomic reading would be a little over 14 minutes, though Beethoven of course meant for the tempo to speed up and slow down with the character of the music. The first movement of 106 should not be much over 8:30 with the repeat. The last movement of 59/3 should be around five minutes. etc.) Part of that is just to make it sound more "sublime" (I won't blame modern instruments, perhaps the increased resonance of modern pianos plays a role but string instruments don't have a decay in the first place and we still get 20 minute Heiliger Dankgesangs from the Busch Quartet/Quartetto Italiano/Belcea Quartet and so on.) but it serves to decrease coherence. When Beethoven marks "con gran' espressione" in 106 he means to heighten the urgency—you don't really get any urgency in an 18+ minute performance, though there are a few long-liners who know what they're doing, Solomon probably being the best.

5:30 sounds about right—the Gulda recording I checked out was 6:27, so I guess I was right that it's on the slow side. (Presumably you meant the one from the stereo cycle which isn't on Qobuz. I've heard some of it on Youtube and elsewhere, but seem to prefer the earlier set in every work I've heard both in, for some reason...) I can't imagine why anyone would mistake Op. 22 for a "sublime" work. Apparently pianists disagree though!

[both quotations slightly abridged to save keystrokes]
Not the last word of course, and Rosen's own sense of Allegretto causes him (in the "free companion CD") to take a tempo for the second movement of 54 that feels absurdly slow to these ears. Another controversy attends the first movement of the Moonlight: a very slow Adagio in 4, or a relatively flowing and faster one in 2? No specific comment from Rosen, but this is one piece where the pedals found on the fortepianos of B's day are needed, and where modern instruments fall short.

The first movement of the Moonlight is marked cut time, so the beat is the minim. "Adagio sostenuto" seems to be about 30-36 beats per minute as a standard tempo, but I've heard successful faster performances (all on modern pianos though, so who knows)

I would like to call attention to another "Adagio (ma non troppo)", much later of course, also in triple time, the Klagender Gesang of Op. 110. The units are 16ths with 12 in a measure. It is difficult to decide on a slow enough tempo if it is to be calibrated to the 6/8 fugue in between, marked "Allegro" but also "ma non troppo". In short, how does one play both at the same tempo and convey the differences in feeling?The next sonata has 9/16 in the arietta.

"Adagio ma non troppo" is meant to be exactly half the speed of "Allegro ma non troppo", as far as I know, though I've heard a handful of disorienting performances that don't do so. 72 as a basic tempo for the movement (sans opening recitative) works well.

The Arietta is in three (basic unit dotted quaver) and is also an Adagio without qualifications (molto cantabile e semplice are expression marks rather than tempo indications), I guess somewhere around 40-45 would work. I've gotten somewhat attached to the modern style of playing the theme so slowly that any sense of beat or rhythm is completely destroyed (and so quietly it's almost lost in the background noise of the hall), then gradually speeding up to allow variations 1-3 to display their amabilità, then slowing down a little for variation 4 & the coda. But that's obviously very inaccurate and mostly works by acting on ~feels~ rather than any sense of musical architecture.

[Jo498]

If amw is right (and the Rosen quotations point in a similar direction) and we take seriously that all (or most, I'd be probably easier convinced to make exceptions for extraordinary late movements like the op.111 arietta) such movements should, despite slow tempo, still "go" in dotted crotchets (or dotted quavers in 9/16 movements) most actual interpretations are way too slow or at the very least on the slow side of what seems reasonable.
I just listened to Lucchesini in op.22. He takes quavers at about 82-84 which does not sound bad per se but is already too slow for me to keep a "feeling" of dotted crotchets at a beat of 28 or slower. And his is a mainstream tempo for that piece, the slow ones like Arrau are at quaver ca. 66.
As I said, even if we find Gulda (amadeo 1967, I have not heard any others, this one is sonically and overall his best cycle I think) too fast here, there is hardly anyone (except Brendel) around the Czerny/Moscheles suggestion of 100-116 or even the 92 one could infer from op.106.

The largo e mesto from op.10/3 suggests a considerably slower tempo (melody in quavers, 64ths figurations). But even this one is marked by Czerny at quaver=72 which is both too slow for a feeling of dotted crotchets (at corresponding 24) and still much faster than it's usually played!
Gulda is again one of the fastest in op.10/3,ii at 66-69 (but I have to listen to it, I only guesstimated this from playing time, although he usually keeps a steady tempo, so this should not be too far off) more typical is 56-60.

I don't think we should be pedants about those tempi. Beethoven was (almost) deaf and the editions by Czerny and Moscheles are several decades later. So I can often be happy with a broader range of tempo. As longs as most tempi are "in the ballpark", e.g. most play the first mvmt. of op.22 at crotchet 140-150, the minuetto might be more often at 112 than 126 (Czerny) and the finale a tad slower than 72-76 as well, that's no big deal. Same for an Eroica first movement at 55 instead 60 and lasting 15-16 min (incl. repeat) with some flexibility.

But if hardly anyone plays that adagio in a tempo where it's plausible to hear it in dotted crotchets, and those who do (Brendel, Gulda) are considered way too fast (and are in Gulda's case statistical outliers, about a minute faster than the next one (Brendel, other Gulda performances) who are again almost two minutes faster than median (around 8:30)) there is either a very wrong tradition in place. Or we are wrong about the dotted crotchet as tempo reference.

[zamyrabyrd]

"Adagio ma non troppo" is meant to be exactly half the speed of "Allegro ma non troppo", as far as I know, though I've heard a handful of disorienting performances that don't do so. 72 as a basic tempo for the movement (sans opening recitative) works well.

I would have a hard time with speeding up the fugue in Op. 110 to twice that of the Aria. Usually it is assumed the dotted quarter of the fugue corresponds to the dotted 8th of the aria but that makes the fugue too plodding, not allegro at all.
Actually I just had another thought after all these years of breaking my head over this piece, what if every 8th of the fugue were equal to 2 16ths and not 3!!! This might actually work! 

[Cato]

I have read through this topic with interest, and noted some of the considerations, e.g. "Beethoven was deaf," "...given the sustaining power of keyboards in Beethoven's day..." and

Yes, the scherzo. My point has been that Bruckner went out of his way to notate the motif as always double-dotted quarter/sixteenth/quarter, when it would have been clearly much easier on him and the players to go with single dotting. That being the case, it frustrates me no end that I have yet to hear a performance, live or recorded, that even seems to try to get the rhythms as written.

Amen to Sforzando here!

When I was composing, I was always able to hear the music at various tempos, even imagining it in very slow motion as I inked the notes on the music paper, although such slowness is not what I wanted for a performance.

I will not pretend to know what Beethoven heard as he composed, wrote down notes, revised, etc.  As the deafness progressed, I do wonder how much of a consideration the "sustaining power" of keyboard instruments mattered to him.  One can make the case both ways, I suppose: that it mattered more precisely because he could no longer hear them, and wanted to be very careful, and that it mattered less, because then he heard the music on a level of imagined perfection divorced from practicality.

Recently, for the first time in decades, I returned to composing: a 9-voice choral work.  It is an Adagio-Largo work in 9/8, and in general I was imagining the work with a quarter note at 40 or so, slower for the Largo.  And so there are a good number of 16ths and 32nds.  (I suppose one could make a case for 9/4, but that just did not seem right.)   It is this subdivision which gave me an impression of approaching the heart of certain themes, that the 32nd was an indivisible packet of energy, like an elementary quantum particle.  And of course, mentally, I was able to slow these down, or speed them up as I wished.

So for Beethoven, I have wondered whether he mentally heard things quickly, and liked it that way!  Certainly the Czerny metronome markings, although not consistent throughout the various editions, are faster than what we usually hear today.  And in his later years, Ignaz Moscheles complained that conductors were slowing down not just Beethoven's works, but Haydn's and Mozart's as well.

[(poco) Sforzando]

Amen to Sforzando here!

Hey, I got an amen! (Someone else seemed to imply that since none of the conductors did it right, it can't really be wrong.)

[Jo498]

The interesting thing is that while one often reads that the original markings by Beethoven were insanely fast (because he was deaf and one hears stuff faster in the head than when actually playing) the markings by Czerny and Moscheles (which are taken from different editions, with some differences in between) are mostly quite consistent with Beethoven's own and are also very fast. Some in slow movements might have to do with the pianofortes of that time but this does not apply in general.

It's also odd that there are a few movements that are "traditionally" played in the tempo suggested by Beethoven's markings or sometimes even faster, namely most of the scherzi and also the finales of the 5th and 7th symphony. So if Beethoven was deaf or his metronome was broken why did he suggest a sensible tempo for the Eroica scherzo but too fast tempi for the first two movements?

As I said, one can be pedantic here and it also seems an open question how flexible performers were. (Beethoven said both that the tempo marking should only hold for the beginning of a piece, that feeling could not be captured by a mechanism, that the tempo should change a hundred times within a piece, but so slightly that only a trained listener would realize etc.

But I do not think it is pedantic to think that there is a difference whether a movement goes in (slow) dotted crotchets resulting in comparably flowing quavers or in fairly slow quavers without any sense of the "larger beat" left. If interpretations were all over the place in case of the op.22 adagio, fine. But they aren't.
There is basically only one on the fast end of a plausible tempo (Gulda), a few more that are flowing enough to sense the dotted crotchet (Brendel, Goode, Backhaus...), lots so slow that this is impossible while quavers are still somewhat flowing and a few by very famous pianists (Gilels and Arrau) that are almost half the tempo suggested by sources and analoguous movements.

[Jo498]

I agree that the "espressione" here is the expression of an opera aria (which he carried to self-parody in 31/1 later), not a dramatic scena (18/1) or whatever, and is essentially static—which is why you need to characterise the music a lot to make the emotion convincing.

Unfortunately, I don't care enough for the piece to seek out more interpretations but it would be interesting if Brendel or Goode (with comparably fluent tempi) are more emotional than Gulda.

I did a quick check of the parodistic op.31/1,ii. It seems plausible that this one should be roughly in the same tempo as op.22,ii.
Pianists beg to differ. Arrau, who played op.22,ii at glacial 65 or so for the quaver, sounds fairly stiff in op.31/1,ii but he starts the movement with 90 for quavers. Gilels who was even slower in op.22 is around 104. Still on the slow end of the speed for that type of movements if one wants to hear the dotted crotchets but completely different from op.22.

[Jo498]

Gulda (amadeo) is a little faster in op.31/1,ii than in op.22,ii, but quite close, ca. 45-46 in the former, ca. 42 in the latter.
But most of the (way too) slow ones in op.22,ii go for a comparably flowing tempo, if still fairly slow  "in 3" (34-36 for the dotted crotchet) in op.31/1ii. Fairly glacial are Arrau and Schnabel at 30 or a little less (but note that with this tempo they would be among the faster ones in op.22,ii!)
I am only guesstimating because I do not have these recordings but Brendel in his 1970s recording seems to take almost the same tempo in both op.22,ii and op.31/1ii (ca. 35) or maybe he is even a little bit slower in the latter.

Quite strange; the similarity of those movements should be obvious. The "grazioso" in the op.31 might hint at a slightly faster tempo but certainly not by so much that this one should go "in 3" and the op.22 "in 9". Kolisch gives the same range (dotted crotchet: 44-46), Czerny very similar (quaver: 116 vs. 126). And one could play op.22ii in, say 35 for the dotted crotchet, op.31/1ii in 42 which would be a noticeable difference but both clearly "in 3".

[PerfectWagnerite]

Another question I have is if the composer is performing or conducting his own music does his/her rendition supersede the written score?

[(poco) Sforzando]

I'll have to sit down at the piano sometime today and see how fast I take the slow movements from 22 and 10/3 myself. (Just to try to hear it with the metronome in my head may be inexact.) But here's Rosen:

on 22: "The Adagio is often taken too slowly: the basic rhythmic element is not the eight note but the dotted quarter."

on 10/3: "The extraordinary pathos presents a temptation to play too slowly, and pianists today often choose a tempo that would be unjustifiable with the weaker sustaining power of Beethoven's keyboards."

[both quotations slightly abridged to save keystrokes]

Not the last word of course, and Rosen's own sense of Allegretto causes him (in the "free companion CD") to take a tempo for the second movement of 54 that feels absurdly slow to these ears. Another controversy attends the first movement of the Moonlight: a very slow Adagio in 4, or a relatively flowing and faster one in 2? No specific comment from Rosen, but this is one piece where the pedals found on the fortepianos of B's day are needed, and where modern instruments fall short.

I spent an hour today at the piano, just turning on Sound Recorder from my phone and playing through some of the Beethoven movements we've discussed as well as a few more for good measure. Given the pretty wretched quality of the pianism, the piano, and the sound, I won't inflict the results on you, but here are my average tempos:

op. 7 q=40
op. 2/1 q=48
op. 10/3 dq=40 (8=120) [ETA: I'm an idiot, mathematically speaking. My 10/3 is actually about dg=23 / 8=68.]
op. 22 dq=26 (8=78)
27/2 h=44
31/1 dq=34 (8=102)
110 Arioso in Ab minor: d8=36, Fugue in Ab: dq=52, Arioso in G minor: d8=44, Fugue in G: dq=52

Some of these results surprised me (I especially didn't realize I took the 110 fugue that much faster, and the G minor arioso as it seems more agitated than the Ab minor version). And I wasn't metronomic by any means. Also how slow I was in 22 and how fast in 10/3. But the notational system provides no means for a composer to indicate whether a 6/8 should be felt in 6 or 2, or a 9/8 movement in 9 or 3, unless perhaps the metronome mark indicates a dotted quarter or an eighth.

[amw]

Lmao I did that too (before you made your post), and just got around to looking at the results.

This is just me playing at the tempi that felt most natural to me. I can't speak to historical accuracy.

10/3/ii - 7:49 (approximately eighth = 66 / dotted quarter = 22)
22/ii - 4:46 (approximately eighth = 145 / dotted quarter = 48)
27/2/i - 4:46 (approximately half = 29 / quarter = 58)
106/iii - 14:23 (approximately eighth = 78 / dotted quarter = 26)
110/iii
Klagender Gesang - 1:25 (approximately dotted eighth = 54)
Fuga - 2:28 (approximately dotted quarter = 71)
Ermattert klagend - 1:38 (approximately dotted eighth = 54)
L'inversione della Fuga - 0:54 [until Meno allegro] (approximately dotted quarter = 71) or 2:09 [to the end] (approximately dotted quarter = 74)
111/ii - 17:09 (approximately dotted eighth = 42)

Listening to myself afterwards reveals I have absolutely no sense of rhythm, and should invest in a metronome app for future practice. That said, it's sort of interesting that my tempi in 110 remained consistently wrong. (The fugue should have been dotted quarter = 54, and I always started it at that tempo, but within 10-20 seconds was back at my "natural" tempo of dotted quarter = 71.)

I'm not sure this serves any purpose but it's interesting to see how me and Sforzando basically reverse the tempi of 22/ii and 10/3/ii. I treat the former "in three" and the latter "in six", I think.

I can't actually imagine an 11 minute 22/ii. Will queue up some Gilels for listening later.

[Jo498]

Poco sforzando is pretty close to mainstream in both the actual tempi and their relation in op.22,ii and op.31/1ii. Do you have any idea why the for the former movement a much slower felt appropiate when playing despite having considerably less fioritures etc.

I have trouble imagining op.10/3ii at ca. double the usual speed. Maybe the beginning, but the later passages with 64ths must be a blur at that tempo...

[(poco) Sforzando]

Poco sforzando is pretty close to mainstream in both the actual tempi and their relation in op.22,ii and op.31/1ii. Do you have any idea why the for the former movement a much slower felt appropiate when playing despite having considerably less fioritures etc.

I have trouble imagining op.10/3ii at ca. double the usual speed. Maybe the beginning, but the later passages with 64ths must be a blur at that tempo...

I wonder if my counting was wrong especially on 10/3. I should check my samples again to be sure.

ETA: I'm an idiot, mathematically speaking. My 10/3 is actually about dg=23 / 8=68. Actually I accelerate in the passages with 64ths and slow down again for the ending. Interesting how one does those things unconsciously. Interesting too how all my tempos for Beethoven adagios tend to fall in a similar range.

I can't explain my tempo decisions for 22 and 31/1 other than that they feel right to me. If I took 22 any faster, it would be more like a barcarolle in Andante rather than a true Adagio. Perhaps the parodistic nature of 31/1 explains the faster tempo, though in some cases one simply has to hold up the downbeat in order to fit in some of the more egregious fioriture. By the middle of the movement, when the bass is moving in constant sixteenths, I can't feel it any slower. (It would be a mistake, I think, to believe that slow movements do not often have quite a bit of fast music within them.)

The ending of 110 is another problem altogether. It's a complicated notational problem where a Meno mosso is marked with the movement in 16th notes but the listener may well feel the music has gotten faster, and then towards the end we accelerate back to the original fugue tempo. Rosen has a fascinating discussion of this. As I recall, Penelope Crawford gets this just right except for her very last chord where she takes a slight retard; this seems to me so wrongheaded in an otherwise admirable reading that I am tempted to rip the audio and do a little editing to keep her strictly in tempo.

I seem to remember an exchange of letters between Rosen and Alfred Brendel in the NY Review of Books on the subject too; I myself was getting tired when I recorded, and I broke off the sonata before the end as my technique is no longer up to it.

[Cato]

Another question I have is if the composer is performing or conducting his own music does his/her rendition supersede the written score?

I was wondering about the same thing: it could be helpful to see whether e.g. Rachmaninov consistently followed his own metronome markings.  I suspect he did not.  On the other hand, I recall a story that, many years after the publication, he spotted an incorrectly printed metronome marking for one of his works.  The printed one was much too slow.  So printers' errors can also be part of the equation.

[amw]

Another question I have is if the composer is performing or conducting his own music does his/her rendition supersede the written score?

No but I would say it complements it. A work does not exist in a single inalienable form (obviously not since it's different every time you hear it), and I'd consider e.g. the score and the piano-roll performance of La cathédrale engloutie to be two different, equally valid versions of the piece, in the same way that e.g. the original and revised versions of Mahler's 6th are different and equally valid.

I can't actually imagine an 11 minute 22/ii. Will queue up some Gilels for listening later.

Gilels's performance was very beautiful if somewhat Romanticised. At the same time, I found it almost impossible to concentrate on it simply because the music lost all of its forward motion and most of its potential for drama (understated, obviously, but the semiquaver turns in the development section with the intermittent sigh figures above form an effectively soft climax in the hands of a good pianist, like Kovacevich or someone). It became soporific and in parts almost dirgelike. Gilels is highly regarded and a lot of it seems to be because of his slow tempi, so I don't know where I went wrong. :lol:

I experimented with playing the opening of 22/ii more slowly than seemed natural to me and found that the tempo would progressively accelerate whenever I wasn't paying attention.

The ending of 110 is another problem altogether. I seem to remember an exchange of letters between Rosen and Alfred Brendel in the NY Review of Books on the subject; I myself was getting tired when I recorded, and I broke off the sonata before the end as my technique is no longer up to it.

My technique has never been up to it and I still played to the end As far as tempo goes my instinctive feeling is that "poco a poco più Allegro" starting at bar whatever (during the meno Allegro) is valid until the end of the piece, so one should just keep accelerating until the final parabola of semiquavers is as fast as is playable. That said I didn't actually do that properly, because that would take practice and stuff, and also knowing how to play the piano

[(poco) Sforzando]

Another question I have is if the composer is performing or conducting his own music does his/her rendition supersede the written score?

Back in the Baroque era, music was always expected to be newly produced and the composer was ordinarily his own interpreter. Despite a few cases where Bach adapted concertos by Vivaldi, we do not usually think of independent performers going around performing the works of independent composers, especially composers of the past. Now the composing and performing functions are thought of as being largely separate, and even when the composer performs or conducts his own work, he is only one of many potential interpreters of the score he has written.

You might say in effect that today, once the composer has completed his score, his job as a composer is finished. He goes on to other things, and the work is now in the hands of whichever performers want to take it up. (Supposedly Brahms said of a surprising reading of one of his works, "So! it can be done that way too.") A composer's own interpretation of his work has obvious special interest, but I don't think we can say it's the only possible interpretation or even the definitive one. Britten for example has recorded excellent versions of his operas and other works, but so have other conductors. Boulez recorded Pli Selon Pli three times, and it's very different each way; the more violent early version on Sony (which I prefer) is quite different from the serene latest performance on DG, and these changes in temperament correspond to Boulez's evolving personality as a composer. On another thread I mentioned how Boulez, when supervising Charles Rosen's recording of the third sonata, asked Rosen to play a particular passage slower even though Rosen was already below Boulez's indicated metronome mark.

[(poco) Sforzando]

My technique has never been up to it and I still played to the end As far as tempo goes my instinctive feeling is that "poco a poco più Allegro" starting at bar whatever (during the meno Allegro) is valid until the end of the piece, so one should just keep accelerating until the final parabola of semiquavers is as fast as is playable.

And I see no need to accelerate past the point where the main fugue theme returns in the right hand in Ab (where the bass comes back in 16th notes); it has the effect of a recapitulation, and though I know there's no explicit marking, I feel a steady tempo from this point to the end seems appropriate.