Listeners’ Box
Rolls: On the Record
I read David Bratman’s article “History’s Mysteries” about early recordings and the impact on modern performance practice with a great deal of interest. I’m not at all an expert in this field, but I’ve been informed by those with more knowledge than I have that piano roll recordings are in no way accurate representations of the performances of these artists.
As I understand it, the performance of piano rolls can be manipulated by the reproducer, both in terms of tempo and dynamics, and are therefore extremely unreliable. To prove this you only have to compare Rachmaninov’s recordings on piano rolls compared with those recorded acoustically. Depending on which “reproduction” of the piano rolls you are listening to, the piano roll recordings sound uncontrolled and rhythmically extremely unsteady, whereas Rachmaninov’s acoustic recordings do not sound like this at all. In fact, his sense of rhythm is very “modern” and quite strict. Refreshingly so, I might add. His sense of rubato is not overdone to the point of sentimentality that one often hears today.
So, I am amazed that any knowledgeable musicologist would take his piano roll recordings as being in any way a true representation of his playing. Similarly, a comparison of the piano roll recordings with acoustic recordings by Robert Casadesus reveal a very different pianist. In fact, they are too different to be trusted as the accurate representation of his playing at different times in his career. I would like to know if this aspect of piano rolls was discussed at the forum? I certainly would hope so.
I wouldn’t go as far as dismissing piano rolls, but I certainly don’t listen to them with any degree of trust in their authenticity. Perhaps we are going too far in our acknowledgment of them as valued documents of historical performance practice?
There should be no doubt that interpretation of a score in light of historical information concerning performance practices that were generally followed at the time of its composition is an appropriate procedure. However, this is not a license to exercise mere personal idiosyncrasy. My piano teacher, Olga Samaroff, strongly urged her pupils to be guided by historical performance traditions, but demanded that each deviation from the strict notation of the score be validated by the possibility of finding some reasonable requirement — whether intellectual, structural, emotional, historical, or all of the above — for the deviation. To merely say, “I want to play it this way because I like it this way” was unacceptable.
There are two people who, David Bratman might want to contact: Dick Wahlberg and Sandra Sonderlund.Dick has a collection of original, vintage recordings dating back to Edison cylinders. His collection would make any university library jealous. He also has vintage victrolas to play his records. Listening to vintage recordings on vintage victrolas is an astonishing experience. It is not the same as listening to remastered recordings on CD. Some years ago, I was visiting Dick and he played a vintage 78 on a refurbished, high-end victrola for me. He asked me to guess who the singer was. I didn’t know but, more to the point, it sounded like the singer was there, singing live, standing right next to me. I was nonplussed by the quality of the recording. Turns out, it was Billie Holliday. I was amazed.Sandra is arguably the world’s leading authority on early keyboard practice. If you want to know how Bach, Beethoven, Chopin, and so on, performed and taught, she is the one to ask. Her new book on the subject is a definitive discussion of the subject.
Musicians all seem to know the same thing: The score is not the music. Unfortunately, the music police, and especially the early music police, sometimes get bogged down, forgetting that music is not on the printed page. Scholarship can only go so far. We don’t know what music sounded like prior to the invention of recordings. We can guess but we don’t know. We weren’t there. Nor will urtext scores always help. The purpose of urtext is not to dictate performance practice. All urtext does, and it’s a valuable tool, is to show us an unedited score. Urtext shows us what the composer wrote originally, not how the music was performed.
Scholarship, however good, can only take us so far. After study and practice, musicians create music, not scholars, however valuable their contribution. When I play, no two performances are ever the same. I look at the score, sometimes I listen to how other musicians performed the same piece, but I always interpret the music as the moment dictates. Musicians study, they practice, and then they let the music sing.
A Response from David Bratman
The first letter-writer, Daniel Glover, is disturbed that piano rolls are being used as evidence of performers’ actual practice. You would not wish to take an 1800-word journalistic summary as a complete description of a complex musicological problem. In fact, the issues he raises were discussed at the symposium, especially by Donald Manildi, but also by others. Piano rolls are virtually useless for determining dynamics, and can serve only as indirect evidence for overall tempo. But practices such as chord rolling, free phrasing of melodic lines, and relative tempo variation within the piece are accurately recorded in piano rolls. Someone asked if the tempo variations, sometimes so surprising to hear, could have been caused by flaws in the recording or reproducing process. The speakers thought not: The tempo variations are too consistent and make too much sense musically to be processing errors.
The speakers stressed the desirability of playing rolls on pianos similar to the ones on which the rolls were made, and also of choosing appropriate tempos to do so. They roundly criticized many modern commercial CDs of piano rolls which fail to do this. They noted that editing piano rolls was easy, much easier than editing sound recordings would be until a considerably later date. They questioned whether players making piano rolls played the same way as in concert, but they also raised the same question about early sound recordings, which were not made under concert conditions either.
Also, George Barth and Anatole Leikin emphasized that no piano roll, nor any sound recording for that matter, is a blueprint. Players should use these to increase their understanding of the work and develop their own performing style.
This last point should allay the concern of letter-writer Victor Wolfram, who does not want to see performers use historical practice as merely a license for personal idiosyncrasy. In fact, the advice Will Crutchfield gave was exactly the opposite of this fallacy. Rather than using historical style to justify your pre-existing preferences, he says you should use your preferences to justify your choice to use historical style. He cautions players to use historical style only if they have come to understand it and to absorb its ethos. Without this, it’s just affectation — the same arbitrary, unvalidated playing that Mr. Wolfram’s piano teacher warned him against.
Lastly, letter-writer Phil Robbins says that vintage recordings sound best on vintage victrolas. They certainly do. One session at the symposium that was not mentioned in my article featured Stanford music librarian Jerry McBride demonstrating early recordings on early machines. Not only is the sound better on the victrola the recording was designed for, but even the choice of needle makes a difference. We heard the same recording with both steel and cactus-spine needles. The steel needle gave a crisper, more “present” sound, but with much more background noise.
This ties back in with the advice on choosing the right player piano for piano rolls, and shows that the variables in studying historical recordings are many, and the certainties few. This theme, along with many others, ran through the entire conference.