Prelude and Trouble at Troldhaugen


‘The devil is in the details’, as the saying goes. Although our approach to research is fundamentally a more positive one, we found ourselves strangely near the more literal interpretation of this saying in the qualifying rounds of our project.

As people with real devotion, Tony and I can sometimes be rather fanatical about the details of what we are doing, and we greatly value the potential of these smaller details even in shaping a larger entity. I suppose we all have had unforgettable memories of great performances, moments that consciously or not put an entire performance in perspective – moments that stay with us for days and weeks, or sometimes forever.

These unforgettable moments of a performance are important in themselves, as moments of sheer beauty or strong emotion. But what makes them unforgettable is often the way in which these moments say something about the totality – the way in which they illuminate the architectonic or expressive concept of a work. To me, Alfred Cortot illuminates the duality of Chopin Sonata b-flat minor in an unforgettable moment: the transition to the second theme of the recapitulation in the first movement:

We are talking about a very brief moment here. It is however a section where Cortot is capable of displaying the full expressive range of the movement in a truly remarkable way. At a moment of great vision, he unifies the haunted despair of the first theme with the poignancy of the second.

To both of us, this fascination with certain momentary events and their large-scale effect is coupled with an urge to comprehend. We wanted to investigate some of these unforgettable moments of ours and started our collaboration experimenting with another radical idea: prior to the performance recreation, we called it analytical or ‘dissective editing’. This is in principle very much what Tony is doing as an editor when investigating the possibilities of the recorded material in a normal recording, but it was in our case taken to the extreme – extreme because of our limitations.

We chose a couple of performances that were very special to us and fed certain sections of these recordings into our digital editing system. We of course had no additional takes to play with, which is always the case in a normal editing situation – so our task was rather difficult. We wanted to find out whether it would be possible to ‘normalise’ or neutralise the unforgettable moments of great performances through minute changes of timing in the existing material. Would we be able to destroy Cortot or Rachmaninoff and learn from it? Our criteria were to make the edits comply with a professional standard, a realisation unnoticeable to someone not knowing the original recording. We wanted to keep all the other elements, such as the overall tempo and dynamics of the performance, unchanged.

To most people this sounds like a totally absurd venture, and we certainly felt like two nerds having lost contact with reality. We did however get some very interesting results when presenting our initial tests to a mixed group of artists. There were musicians among them but they were not in the majority.

In our double blind test we played back the original version and our manipulated version, awaiting the listeners´ reactions before revealing the results. We finally discussed the two versions. Initially, our ‘guinea pigs’ were a little confused because most of them were expecting much greater differences between the original and our manipulated version. After ‘tuning in’, everybody could however tell there was a difference between the two versions, which of course was a great relief to us. The really interesting tendency was that the untrained – the non-musicians – responded more negatively to our ‘normalised’ or ‘neutralised’ version than the trained musicians did. The non-musicians appreciated more the characteristics of the ‘real’ Cortot and Rachmaninoff. We also found a striking ability among some of the musically untrained listeners to describe verbally the effect of the changes we had made. I should emphasise that this was nothing close to being a scientifically valid test, but the tendency of the reactions was certainly interesting enough.

I have included one example of our experiments, in two versions. The performance is by Alfred Cortot, and the cut is from the opening of Chopin´s Piano Sonata in b-flat minor. In one of these versions we have manipulated the timing in the transision to the second theme – in the section from around ten to twenty seconds into the take. The result of the ‘quiz’ is found at the end of this article….

Version x:

Version y:

The answer to the question above was however clear and not very surprising: We were not – given our extreme limitations – able to destroy the greatness of Cortot or Rachmaninoff by minute subtractions of time only.

There are at least two reasons why the results of our dissective editing were not as convincing as we wanted them to be:

  • The first is the fact that we had no additional material to play with; there were no additional takes of Cortot available. In reality we were only able to take away from, and not add to the original recording. There are in fact ways of subtly stretching (adding) time in digital editing, which Tony introduced me to early on. This can sometimes be done to great effect when there is enough material to play with. With our extremely limited material however, only very subtle changes could be made to the original recording without making the changes appear ridiculously obvious.
  • The second reason is that we were strictly dealing with timing issues in this process. (Timing, with all its facets, is where performances have changed the most over the last century, and this is where our main interest lies.) Whatever importance the timing issues have, it soon became very clear that the minute changes of timing that we were able to make (because of our limitations) did not alone destroy the greatness of Cortot´s performance. Not surprisingly, we found his concept of performance far too complex.

To some extent we were still encouraged by the results of our ‘dissective’ editing. After all, we were not the only ones who had responded to this subtle manipulation of timing. And our attempts did highlight a few interesting elements in various performances. It certainly had increased my sensitivity to the importance of timing issues even further, and it had been a valuable preparation for what was to come.

We had realised that our ‘dissective’ editing would not take us where we really wanted to go. It told us something about the power of micro-changes in timing. But it didn´t tell us much about where to move ahead. This knowledge wasn’t really applicable to us, nor to anybody else, in getting a more comprehensive understanding of a specific performance style.


Eventually Tony came up with the idea of recreating a historic performance as truthfully as possible to see what we could learn from it. Even though the various aspects of timing still seemed to us the most important element in the understanding of an early 20th century performance, this would force or hopefully enable us to capture all the elements of a great performance and capture them at the same time. We would experience first hand how the various elements interact in time. This is the only way to fully understand a performance.

Only later did we realise that the idea of recreating a historic performance in a somewhat related way had been developed by both the Russian pianist Alexander Melnikov and at Zenph Sound Innovation. Both have important points of contact with our project but in rather different ways. The most high tech project is clearly the re-performance from Zenph. Using modern computer technology, John Q. Walker and his team have recreated performances by Sergei Rachmaninoff, Glenn Gould, and others. The recorded sound is computer analysed, and through a complex process, the computer data then generates an automated ‘live’ recreated performance.

Judging on their own presentation on the web, the major concern with Zenph seems to be that of technological innovation and business development rather than artistic research, which takes Zenph in a radically different direction from us. I am not able to evaluate their technological achievements, which seem astonishing, but the proof is in the pudding.

Below are three versions of the same piece performed by Rachmaninoff. Version one is a ‘normal’ playback of an Ampico piano roll from 1919. This is the piano roll with the most advanced playback system today, reproduced on a modern Bösendorfer, realised by Wayne Stahnke:

Next is the Zenph recreation of an electrical recording from 1928 (RCA Victor):

Finally, we have the original electrical recording from 1928, on which the Zenph re-performance is based:

To my ears, the two first examples are basically interchangeable. The big difference is found between these and the original electrical recording. Listening to the original against these, we suddenly hear that a human is playing. The micro-dynamic detailing of the original is absent in both the other versions. This information is essential to the connection between the notes, to fill the space between them. In the original the phrasing breathes, and I am kept on the edge of my chair listening. Not in order to hear better, but spellbound by the magic of his playing. The musical argumentation is so fantastic that it survives in all of these versions; but when compared with the original, the Zenph recreation drops dead to the ground. Clarity is what´s left.

I am much more fascinated by Alexander Melnikov´s recreated performances from 2006, which are based on Scriabin´s piano rolls from 1908 and 1910 and Pavel Lobanov´s deciphering of them. The piano rolls clearly reveal far less information about the performance than acoustic recordings like Grieg´s. Nevertheless, there is in Melnikov´s performances an assimilation of this information, and we hear Melnikov interacting with Scriabin the performer: He says: “I have used those materials extensively in the preparation for the recording as regards rubato and other stylistic matters.” As an artistic project I find this much more tantalising, and it opens new perspectives to the original roll recordings.

Scriabin piano roll, Poéme op 31 no 2:


There are even interesting attempts at recreating a performance style by performing together with pianists on reproducing pianos. Joshua Bell performed the slow movement from Grieg´s Violin Sonata c minor with Rachmaninoff (replacing Fritz Kreisler) in 2009, and Rolf Gupta made a very interesting recording conducting the Grieg Piano Concerto with Kristiansand Symphony Orchestra, and with Percy Grainger as a piano roll soloist.


At the time of Tony´s suggestion, I had been asked by Simax to find a repertoire, which would compliment my previously recorded piano concerto by Edvard Grieg in a planned expanded re-issue. This request came just at the right time, and gave birth to the idea of recreating some of Grieg´s performances. It would of course fit well with the commissioned recording from Simax, but there were a number of other, more important reasons why Grieg especially would make an excellent first candidate for recreation (other than being a Norwegian composer of course):

  • Grieg´s performances are certainly coloured by his individual style, but being trained in Leipzig: he was also a representative of the German school. Grieg was even the second German-trained pianist in history to make a recording. (Brahms was the first, and his recordings are of limited value for quantity reasons and technical quality, as can be heard in section ‘Tempo Modification and the Musical Narrative’) As such, his style of playing has wide implications.
  • We are in Grieg´s case not only investigating an outstanding historic performer, but a performer who is playing his own music. Recreating an important composer-performer would add weight to the results of our research.
  • Grieg performed his own music in a way that is radically different from what we hear today, even if many pianists have actually listened to his performances. His own way of performing his music really challenges the criticism that he is often exposed to as a composer.
  • Grieg recorded a total of nine pieces, which at the time looked like a manageable task to recreate as a whole set. We saw the possibility of investigating an important and complete recorded oevre. The pieces have very different character, and together they provide an intriguing clue to Grieg as a performer of his own music.
  • Finally, Grieg´s recordings are of such low technical quality that the recreated performances – if done to the best possible level – could even help to make Grieg´s performance style accessible to an audience who would otherwise never find an interest in listening to something as noisy as these old discs.

The ideal venue for a project like this would of course be Troldhaugen, playing in Grieg´s own living room, on his own piano. We know that Grieg himself never recorded there, but the instrument is of the right period, and doing this at Troldhaugen would certainly give the project a greater authenticity. We also had to work at a venue where people would be prepared to have us around for the extended period of time we needed, and who would see the value in it. Troldhaugen and everybody there turned out to be all that we could ever dream of.

Troldhaugen with Edvard and Nina Grieg in the doorway





Bergen Public Library, The Edvard Grieg Archives


I had a very clear view of what I wanted to do with this as a recording project: I wanted to recreate Grieg performing all the nine pieces, and consequently let this influence my own, recorded interpretations of the same pieces. I had a strong wish to keep my own voice and integrity, even where these nine pieces were concerned. I wanted to get deeply involved with Grieg but still keep him at a certain distance. A rather important part of the plan was to record the nine pieces again on a ‘proper’ instrument, meaning a modern Steinway D.

I tried to hold on to this idea for quite a long time, but eventually I had to realise that it was not going to be. After meticulously recreating Grieg, in fact being brain-washed by him, I simply found myself incapable of adding anything of value to Grieg´s performances of these pieces. In retrospect, I find that these pieces are ruined for me for a time to come, as I am neither capable of performing them with the level of detail in the recreation as I once did. Nor can I really distance myself from Grieg´s performances. The recreation process has however had a fundamental impact on me as a musician, which is far more valuable than playing these particular nine piano pieces.

We were both very clear that our working method had to be developed as we progressed, at least to a certain extent. But we had another fixed idea, which eventually failed as well: we assumed we would get to the end of this, working our way through the pieces bit by bit, and let the editing do the rest. Considering the vast amount of performance-details in the span of even a three-minute piece, we thought this would be the only realistic approach to the recreation process.

From the moment I met Tony, which was at the time of my first recording, I found him obsessed by the importance of recording complete takes – at least longer sections of music at the time. From my own experience, I have also later realised the great value of this. In recreating Grieg´s performances, none of us would dream of trying to record complete takes.

Comrades in Arms at Troldhaugen





Photo: Øystein Fyxe

Butterfly was the first piece we looked at. This was partly because we already had done a very small pre-test of this particular piece at my home, and partly because we wanted to go straight to one of the most difficult pieces of the set to recreate. We started out by working our way through Butterfly bar by bar -sometimes note by note. Keeping Grieg´s coherence in the musical gestures was of overall importance, and we tried to let this determine the minimum length of our recorded takes. Still, the takes often ended up shorter than we wanted, simply because Grieg´s performances are so incredibly rich in detail and unpredictable to our modern ears.

We soon realised that we had gravely underestimated our endeavour. The initial takes were not very promising, and we had to record certain small sections over and over again for hours, trying to capture the essence in Grieg´s shaping of the first phrases. The moment we had captured an important detail of his – and I had managed to realise something similar in my own performance – at that moment something else had slipped away. I constantly found myself disillusioned by the vast difference between what I meant to do – and thought I did – and what actually came out when playing back. The problems we had hearing properly, due to all the noise from the old discs, only added to the difficulty.

After two complete days fighting our way through Butterfly, we really felt we had achieved something! Tony was editing as we progressed, and we were increasingly happy with the level of detail in our recreation of this piece. The play back of the complete edited version thus turned out to be somewhat disappointing to both of us. We tried to hold on to the fact that we had managed to capture a whole lot of Grieg´s subtle performance details, but there was no getting away from the fact that it all sounded like nice patchwork, like a ‘Frankenstein’ of modern music technology. The details were mostly there, but there was no coherence in the performance; the performance was not a living organism. Grieg´s beautiful and extraordinary line in Butterfly was gone – or rather, it had never been there in our version.

We realised that we would have to capture the details of Grieg´s playing in the span of whole takes – or at least longer stretches – if we were to capture anything but the lovely details. This is what our first version of Butterfly sounded like:

This totally changed our approach to the working process, and of course made everything twice as difficult. We now had to work our way through each piece in detail, and try to preserve as much of this detail as possible in the consequent complete takes. In the end this set-back really turned out to be a blessing in disguise, as we later found many of Grieg´s most important performance features appearing clearly only with this ‘bird´s eye’ perspective. When comparing the two versions, many people will probably find them sounding almost identical – which they to some level are. And I do believe many researchers would have left it at this stage. When comparing the first version with Grieg´s performance it was however clear to us that there were fundamental but less visable elements of his playing that we had failed to capture. We had not managed to recognise and capture the most important things happening at the root level of his playing: his wonderful swinging line and his tantalising overlapping phrasing. The process of lifting Butterfly to this next level proved to be the most rewarding part, where these less instantly recognisable features of his playing emerged.

The final version:








View from Grieg´s window 2010 and 1903




My very first plan was to do the performance recreations at Troldhaugen, and consequently to record my own interpretations of the same pieces, the Piano Sonata, and finally Ballade on a modern instrument. I left the idea of doing a second version of the nine pieces relatively early, but doing Ballade and the Piano Sonata on a modern instrument was still fixed in my mind. I had always been a little apprehensive towards period instruments, and more concerned with the limitations of many early instruments than their apparent qualities. I wanted to bring Grieg and his performance style into the 21st century – fulfilling all the aural expectations to a modern recording – other than the performance style itself. In a ‘scientific’ manner, I wanted to isolate the effect of the original performance-style by using the best of modern recording technology, coupled with a new concert instrument.

I was even slightly concerned that using anything other than a modern Steinway would make the recording an acquired taste, and not very attractive to a modern audience. Surely, our point was to make Grieg´s performance-style accessible to a much larger audience, rather than limiting our target group and make our release a library issue. And most importantly: I loved my relatively new Steinway and all it is capable of delivering. I still do.

After fighting our way through the recreations, everything was now set for the next big battle. This was taking place in Oslo, on a brand new concert grand as planned. The Steinway was brought in for the occasion and was beautifully prepared by the piano technician Thron Irby. Tony had made the journey to Oslo and had set everything up with our recording engineer. We were trying out a recording venue that was new to all of us, but already a proven venue for string recording: Ris church.

Ris was clearly not ideal for solo piano. The acoustics were far too rich, and I had difficulty hearing the details of my own playing. Strangely enough, the piano and room was still recording rather well. Geoff Miles had done a fabulous job adjusting the recording setup, and when listening to the tests, I was happy with the sound of everything.

It still didn´t take me long to realise that this new instrument wouldn´t give me what I now wanted. Working over such an extended period on Grieg´s 1892 Steinway B (even if it has been modified), had made me very much aware of the differences between these two instruments, which in principle are quite similar. Grieg´s instrument is a relatively small one, and has got limited resources with regards to volume and richness of sound. It is also a rather fragile instrument, which had to be worked on continuously by Troldhaugen´s piano technician Richard Brekne to keep it in the wonderful shape he had brought it to. But more importantly: I now found myself missing the ‘spoken’ quality of the old instrument. I missed the old instrument´s immediate response to my most subtle impulses in a way which only the voice is capable of.

Nina and Edvard at the piano



Bergen Public Library, The Edvard Grieg Archives

A modern Steinway concert grand is an amazing instrument. It is capable of projecting sound in the most fantastic way even in the largest concert halls. There is an even beauty of sound, which in fact makes it hard to produce anything sounding harsh. It is fully ‘pianist proof’. To me, this wonderful quality sometimes becomes a limitation when trying to communicate the subtlest narrative content in the music, and in the expressive range which goes beyond beauty.

The modern standard of depth of touch is another side to the instrumental development that I find problematic, in some ways even detrimental to the playing. This became even more apparent to me when playing Grieg´s instrument for such a long time. Modern instruments are very slightly deeper in touch, which means that the keys need to travel a longer distance in order to make sound. We are only talking about a minute difference, a difference that non-professionals will have problems recognising.

Compared with instruments built a century ago, it is slightly more labour-intensive to make sound on a modern concert-grand, but it also gives more power when a big sound is required. This greater depth of touch comes with an unwanted side-effect: the slightly increased effort on each finger tends to add to the difficulty when trying to create coherent musical gestures in the most intricate pianistic structures. The differentiation between the important and the less important in complex textures becomes harder, and performing the most difficult passages with the ease sometimes required often feels impossible. Very frequently, the most difficult passages should be subordinate to – and only appear in the shadow of – a much simpler overall line. When hearing the great masters of early recording history, this apparent ease is often striking. Striking because it is rare to our modern ears. Some pianists and piano makers will certainly claim that the difference here is only a theoretical one. I do not agree. More than anything, I believe the instrumental development has had a neutralising effect on our attempts of realising convincing musical gestures.

I began to realise that the nature of our project made it impossible to move straight to a modern instrument after doing our research on a period instrument. I decided to call off the whole session at Ris church.

Our friends at Troldhaugen kindly let us return for another recording session – this time we requested a more normal one, which still would allow us a generous amount of time. We were both very happy about this choice, and felt reassured that our newly acquired knowledge could only properly be realised in this way.


Tony made a first edited version of my Ballade only a few weeks after recording it at Troldhaugen. The performance had many qualities to it, but in several important sections, essential features of Grieg´s performance style were absent. It could certainly be released the way it was, but none of us was entirely happy. The major weaknesses were: a certain stiffness and sectionalising of the theme, a lack of ‘schwung’ (cf. ‘Tempo Modification, Swing, and Structure’) in several variations, and a lack of a clear, natural, and dynamic overall line through certain variations. The unbroken, coherent line that is so characteristic of Grieg´s own playing was not there at all the points of my performance. Keeping this ‘thread’ throughout the span of a large-scale work is of course even more of a challenge than in a three-minute piece. Nevertheless, it is a feature of Grieg´s playing which is fundamental to his style, and preserving it on a larger scale would be crucial for the end result.

This is the theme from the first version:

The final version:

We knew that realising Grieg´s characteristic ‘schwung’ in my interpretation of the Piano Sonata and Ballade would be a major challenge, and one of overall importance. In my first version of Ballade I had captured his ‘schwung’ in certain parts, but in other important sections I had failed. As a result the phrase structure was stiffer, more sectionalised, and closed in important places:

Variation 2, first version:

The final version:

The central ‘funeral march’ of Ballade is often very problematic in performance. The repetitive meditation on the theme easily becomes too static, the overall line is lost and the structural function of the variation becomes unclear. The duality emerging from the meditative, static rhythmic pattern, and the overall dynamic but falling tendency of the variation, is highly important in my view. This overall falling tendency points towards the very end of the variation, which in my view is the point of greatest gravity in the work.

The realisation of this falling tendency requires an overall line, which is not only congruent with or restricted by the basic rhythmic pattern of the variation. The overall direction or shape of the variation is dependent on two elements:

The anxious, short breathed character of the previous variation must reside in the opening of the ‘funeral march’. It should not be done in a pedantic or obvious way, but the aftermath of the variation should be felt. The transition here is a point of two strongly overlapping directional tendencies (cf. “Ambiguity and Multi-layeredness”)

The repetitive, static rhythmic pattern must be also be disturbed slightly, in the right way, to allow a dynamic large-scale line to emerge. Finding the right kind of schwung is central to this. In the first version of Ballade these elements were missing, and this important variation became a closed, detached, and static episode:

Our unbelievably patient, but gradually tired friends at Troldhaugen kindly let us return again. This time we only allowed ourselves a short time to re-record Ballade, but we now knew where we were heading.

The final version:

*Version y is the original, the unmanipulated version of Cortot