Chopin’s Nocturne in D flat Op.27 No.2:

Towards an Historically Informed Performance


My slightly unusual recording of this famous nocturne was inspired after the discovery of a remarkable book over the summer of 2014:  Neal Peres da Costa’s Off the Record: Performing Practices in Romantic Piano Playing.  The basic premise of the book, argued painstakingly and meticulously throughout, is that early recordings (those by pianists born in the nineteenth century) provide a vital and often overlooked window onto 19th-century piano playing.  Far from being the mannerist distortions that later schools of pianism dismissed in favour of so-called fidelity to the score, Peres Da Costa argues that these recordings embody a performance tradition that is quite probably quite close to the playing of Schumann, Brahms and Chopin.  He also charts some of the changes in playing style that occurred during the course of the twentieth century, during which some of the older expressive approaches were deliberately derided and discarded. He refers to a wealth of recordings, pedagogical texts and contemporary accounts of piano playing in support of his argument, and focuses on Chopin’s D flat Nocturne amongst other works.

The musician who spends time listening to the many audio examples Peres Da Costa provides (via an associated website) as well as to full length recordings available elsewhere (see discography below), will be at the very least intrigued by this remarkably different pianistic universe, though to begin with the sound quality and some of the expressive habits of particular pianists can be annoying or puzzling.  Above all it is an emotional, improvisatory, sometimes wildly spontaneous world, though perhaps for that very reason unsuited to the definitive act of recording in the modern sense, in which a masterpiece is perhaps interpreted in an idealised way that will stand the test of being listened to many times.

There are four general areas of early recorded piano playing that differ markedly from the today’s style of playing, discussed in depth through individual chapters of Peres Da Costa’s book, and I have crafted my interpretation with each of these in mind.  These are: dislocation (playing one hand after the other at moments where they are written at the same time in the score), un-notated arpeggiation, rubato (and other forms of rhythmic alteration), and tempo modification.  I have tried to avoid a merely academic application of each of these expressive devices but rather feel my way towards the style intuitively.  To this end I selected my own preferred pool of early recordings based on a number of key figures in Chopin interpretation in the early twentieth century, which I listened to intensively over a period of several months, taking notes on certain aspects whilst also allowing the complete sound to wash over me.

Above all these early recordings show us elements of style that can only be approximated by, or perhaps cannot be clearly indicated by, the written score itself.  The extent to which these elements are relevant to a piece written in the first half of the nineteenth century can always be debated as there are obviously no recordings from that time.  However, it is worth noting that two of the pianists I have allowed myself to be influenced by were alive at the time of Chopin – the notably eccentric Vladimir Da Pachmann (b. 1848) and iconic Louis Diemer (b. 1843).   The other three pianists whose recordings played a role in shaping my own were all born in the nineteenth century and have important links with the musical traditions of Chopin.  Moriz Rosenthal (b.1862) was a student of Liszt and of Chopin’s student  and disciple, Karl Mikuli, and a friend of Brahms.  My favourite recording, perhaps because I became rather attached to it before researching the others, and perhaps because he recorded it on Chopin’s Pleyel,  is that by Raoul Koczalski (b. 1885), another pupil of Mikuli.  Finally, two recordings made by Artur Rubinstein both early and late in his career (1936 and 1965)  make a fascinating comparison, with rather less of the early style present in the later recording as Rubinstein no doubt adapted to modern trends.  In the minds of many pianists and connoisseurs of the nocturnes, this later recording is from a definitive Chopin nocturne set, and I must admit  it is staggeringly fine.  I include a few short audio extracts from some of  these recordings where relevant.  (All recordings are currently commercially available: see discography for details.)

To give an idea of how these recordings may have influenced me, I have included annotated snippets from the score in the video clip below, highlighting details that are shared with the recordings on the discography – possible influences, so to speak – so this is probably a good point at which to listen (and watch) this version.  You’ll need to stop the other version first if it’s still running.  If possible set to watch in HD mode.  This is followed by detailed discussion.


There are many instances in my own interpretation where the use of dislocation on these recordings may have influenced me – almost too many to document  each and every one; some were certainly planned, others arose more spontaneously.  It is usually the left hand that precedes the right, especially on downbeats where in this piece there is always a lower bass note in the left hand.  The extent to which these left hand notes are technically before the beat (or on it) is really quite subjective in a situation where the tempo is constantly fluctuating due to the feeling of rubato.  However, the right hand never seems to sound unpleasantly late in the historical recordings.  An example is bars 2-4 of the main theme as shown in the annotated recording (also as Fig. 1 below). Koczalski and Rosenthal are similar in this regard (listen to the beginnings of Ex 1 and Ex 2).  With some of the right hand notes unhinged from the left hand we hear the attack points of the two hands distinctly, one no longer masked by the other, lending a sort of hyper clarity to the singing line whilst also giving an independent voice to the bass notes.  One very important use for this technique is to allow the melody to sing.  Recording my own version on a rebuilt 1930s Bechstein was in some sense better than playing on a modern Steinway from this point of view.  The treble of this beautiful instrument is not as powerful or sustaining than that of a large modern Steinway, and the dislocation helped to give an added presence to the melody without having to force the tone.

Koczalski’s 1948 recording on Chopin’s Pleyel creates an even more striking sense of a singing line, and the displacements no doubt add significantly to this impression on an instrument which similarly has a narrower dynamic range than that of a modern grand piano (Ex 1).  It is not clear from my annotations here – because they are limited to commenting on aspects of my own interpretation – that in Koczalski’s recording there is only a single note that is synchronised with the left hand in the entirety of bars 2 and 3!  Whereas some modern players (Biret and Pires, for example) have used occasional moments of dislocation to heighten expression it is clear that in some of these older recordings it is used far more liberally and never to this extent.

Fig. 1: bars 2 -4

Nocturne bar 2

Ex 1 Koczalski (1948) opening melody


Rosenthal’s 1936 recording also has quite a lot in common with this approach (Ex 2) but the time-lag in the dislocation is often shorter so less noticeable.

Ex 2 Rosenthal (1936)


Indeed, each pianist has their own style of dislocation, deliberately avoiding it at times.  Pachmann, for instance, who has a habit of pronouncing his bass notes very powerfully at times and often very much before the beat, is very sparing with dislocation at the beginning of the Nocturne, preferring a simple presentation of the main theme, at least to begin with (Ex 3).  The exception is the way he uses it to highlight the arrival on the high Bb of bar 3, followed by the low A in bar 4.  Idil Biret chooses exactly the same moments for slight dislocation in her modern recording of the piece (see post-1950s discography).

Ex 3 Pachmann (1927)


A second kind of dislocation appears to result from the use of dotted rhythms in the right hand.  Chopin writes quite a few of these into the melody (dotted semiquaver followed by demisemiquaver) and they often fall at the end of a bar, energising the upbeat into the next bar.  By their nature we will hear the left hand semiquaver accompaniment as separate from the final demisemiquaver in such instances.  However, listen again to Ex 2 and Ex 3 and you will hear that both Rosenthal and Pachmann introduce this effect at the end of the bar 6 where it is not written (Fig. 2 below).  In this case though is it that the rhythm is actually dotted in the right hand, or just that the left hand is placed first giving the impression of this by falling a fraction before the right?  It is almost impossible to answer this without recourse to some elaborate system of measurement beyond my present scope.  All I can say is that again there is the sense of a melody floating freely above accompaniment.  I have tried to allow melodic groups like this to float in a slightly unanchored way throughout my performance.

Fig. 2: bars 5-6

Nocturne bar 5

Related to this rhythmic shaping is the lengthening of an already dotted rhythm, which in an Baroque style would be known as ‘overdotting’.  Referring back to audio examples 1 and 2 and Fig. 1 this can be heard clearly in bar 2, the final semiquaver being held back from its expected position.  This does not mean it is always played as a very short demisemiquaver as would be the case in Baroque music, however (although it is in Rosenthal’s version and mine).  The flexibility of the tempo means it can still have a subjective length to it, even if played late.  Rubinstein (1936) synchronises this delayed semiquaver with the left hand.  Rosenthal and Koczaski let it dislocate after the left hand accompaniment, as I do.  This note is treated similarly by all the artists when the main theme returns later on in the piece.

One place where dislocation is noticeably reduced or absent in Rosenthal’s and Koczalski’s recordings is the moment where the theme returns after the tumultuous middle section: at the moment where the triplet cross-rhythms, which are by their very nature partly dislocated over the left hand, resolve into the first note of the main theme on the beat (bar 46), the hands are synchronised, resolving the tension (the hands being absolutely together in the case of Koczalski and only a fraction dislocated in Rosenthal).  And this despite the fact that both artists dislocate the equivalent note at the beginning of the piece.  Pachmann misses this trick (dislocating heavily the RH), as he does at another structural point where the Rosenthal and Koczalski seem deliberately synchronised: the climactic arrival on the dominant (bar 60) before the pivotal perfect cadence to the tonic.  This resynchronisation seems to add an extra sense of resolution, resolving the bass dislocations which occurred in the bar before (Fig 3).  My own version copies these seemingly deliberate moments of ‘resychronisation’.

Fig. 3

Nocturne bar 59


The reader will note I have not yet mentioned Rubinstein, for the simple reason that he seems, even in 1936, to have eschewed dislocation as a general rule.  However, he can be caught making very slight separation of the hands here and there in 1936, though only in one place by 1965.  For example, listen to  Ex 6 where at the end of bar 6, similar to the Rosenthal and Pachmann recordings, a very slight split between hands can be detected at this moment in Rubinstein’s 1936 recording (but not in the 1965), but it is only a whisker.  Louis Diemer’s very early recording of 1904 is rife with dislocations, but they are harder to hear clearly due to the muffled recording.  I have noted them, where similar to my own, in the annotated video.

Tempo and tempo fluctuation

Another interesting assertion by Peres Da Costa is that tempos were faster than today, even in slower types of music.  Indeed, of the recordings I surveyed, Pachmann managed a total duration of only 4’24’’ with several others clocking in around the 5 minute mark.  Compare this with many later 20th-century performances (for example, Barenboim at 6’28’’ and Pires at 6’37’’).  Of course, there were limitations on the duration of recordings in the early years of the twentieth century but these varied with the decade and recording procedure (the recording cited by Louis Diemer is the earliest, and has to cut out the coda!).  Telling from this point of view is Raoul Koczalski’s 1948 recording which is still only 5’13’’ in spite of advances in recording.  At this faster tempo the piece becomes more impassioned than in the meditative slower approach and thus invites a completely different approach to phrasing and rubato.   One could argue that from a notational point of view the piece is not in two lots of three-in-a-bar but just two beats in a bar and therefore should be played at sufficient speed for this two-in-a-bar to be felt.

Two of the most exquisite and justly famous post-1950s renditions, those by Rubinstein and Maria Joao Pires, manage to combine elements of both fast and slow approaches, choosing significantly different tempi for different sections and thus combining the best of both worlds. Whereas Pires gets faster for the episodes which divide the varied statements of the slower main theme, Rubinstein inverts this relationship, yet both recordings are similar in overall length (Rubinstein at 6’20’’, Pires at 6’37’’).   My initial trial recordings always seemed to be around the five minute mark like Kolczalski and Rosenthal, but in the end I decided to take statements of the main theme a little slower and also relax the tempo through the coda section to bring the work to a peaceful close, inspired by the sectional approach to tempi of these two great artists, and adding a further minute to the overall recording length.  The idea of performing different sections at different tempi is to some extent vindicated by another observation from Peres Da Costa, that musical terminology was often interpreted differently today such that markings we find in this work, such as ‘espressivo‘ (bar 10), ‘dolce’ (bar 34) and ‘calando‘ (bar 70), could have suggested a change of tempo.  Indeed, since these indications coincide with the onset of new sections or phrases in this particular piece, it is at these moments that tempo changes are felt.  For example,  ‘calando‘, for the last section of the coda, encourages Rubinstein and Koczalski to slow down considerably, which I have emulated.

Another surprising aspect of early recordings that clashes with modern taste is the rather volatile shifts of tempo which feature in some of the recordings.  In earlier studies to which Peres Da Costa refers (namely Robert Philip’s Early Recordings and Musical Style: Changing Tastes in Instrumental Performance 1900-1950) one very noticeable feature of early performance style that comes to light is the unfettered use of accelerando.  Whereas modern performers as a rule tend to make tempo modification in the form of slight slowing, early interpretations tend to use both slowing down and speeding up, and both to a greater degree which modern artists may find excessive or irritating.  This is not quite true of the D flat nocturne, in the sense that even modern performers find it fairly irresistible to push forward in the tempestuous middle section of the piece to some extent (bars 37-45).  I have deliberately tried to do something excessive here from the point of view of modern taste, but I actually stand by this as a historically valid aesthetic choice, and it is not that dissimilar to the approach of a great modern interpreter, Maria Joao Pires.  Louis Diemer’s recording contains a number of tempo surges during this section and is without doubt more excessive than my own in the way it hurries forward sporadically, uneven and out of control from the point of view of modern standards (Audio Ex. 4) but creating a fascinating feverishness.  He was professor of piano at the Paris Conservatory, and taught Alfred Cortot amongst others!

Ex 4 Diemer (1904) bars 38-47


The listener will also note how these surges in tempo are balanced by equally bold ritardandi, corresponding to one definition of rubato that is much debated, that of robbed time wherein time removed from one part of a phrase is restored in another.  Since this aspect is, in its more subtle details, virtually impervious to verbal description, I am not going to attempt to chart it through the various recordings in a detailed manner.  I am convinced that every time the piece is played such rubato will come out slightly differently, let alone when it is played by such a mixed bag of pianists as we have here.  I will confine my observations in the final part of the discussion to a few significant points on the matter.  However, I will mention here the variances in approach to the reprises of the main theme, only one of which is marked to be preceded by a rit. by the composer (bar 25).  We have just heard Diemer make a pronounced rit. before the next reprise in bar 45 where this is not marked, and so to do all the other recordings surveyed here, though to different degrees.  What goes up must come down, and to the extent that the section before the return of the main theme develops momentum, it must return to the relative calm of the theme by slowing in some way.  The modern recordings by Idil Biret and Maria Joao Pires are less pronounced than most of the the earlier recordings  in their respective rits in bar 45.  I tumble forward through the triplets, reserving the rit. for the end of the bar, though I have experimented with lots of different ways of playing this bar.  The most remarkable version I have come across is Rosenthal’s, where he makes a sudden reduction in tempo during this bar (before the main theme) which is almost visionary, surely something unrehearsed? Here it is:

Ex 5 Rosenthal (1936) bars 42-47



I have not heard a recording which matches exactly the description of ‘metrical rubato’ found in some accounts of Chopin’s playing, in which the left hand keeps exact time whilst the right floats above it in its own rubato time (although Rosenthal is close at times).  Nevertheless, there is a sense in which many times the dislocation between hands is a sort of side effect of the right hand living and breathing, as it were, in its own time, suggesting something of this nature.  A curious aspect of several of these recordings is the unsteadiness of the left hand accompaniment, especially compared with modern performances.  Listen again to Ex 1 and 2, and also Ex 6 (below) from Rubinstein’s 1936 recording, with attention on the left hand (refer back to Fig. 1 for the score).  Despite the noticeably extraordinary variance of pulse if one tries to tap quavers along with these examples, we may feel there is a sense of rightness about this way of playing, but why?   I would describe it that  the left hand sometimes rushes on a little where the right hand melody comes to rest on a longer note, but where there is more movement in the melody time expands a little, effectively slowing down (in some instances quite a lot) to, as it were, savour each note of the melody (noted in Fig. 1 at the end of bar 4).   Both of Rubinstein’s recordings embody this approach, at least with regard to the main theme, so that although he does not share the dislocation elements of the others, he does share a similar concept of shaping melody and accompaniment in terms of tempo rubato.  Having attempted to copy this by slowing down to accommodate groups of right hand semiquavers (for example,in  the second half of bar 4), I couldn’t convince myself to rush forward in the same way in bar 3.  Audio Example 6 is from the opening of Rubinstein’s 1936 account, which is a little more extreme with these shifts than in his 1965 account, especially in the way bar 3 plunges forward.  One can imagine a teacher asking him – ‘what, you’re bored with this bar and have to play it faster just because the melody isn’t doing anything?’  But there is that unquantifiable magic to the whole, an over all balance to the speed,  lingering and forward moving in equal measure.  Whatever the precariousness of the left hand in all these versions the right hand sounds very evenly proportioned, almost the reverse of metrical rubato in terms of melody and accompaniment since the right hand sounds steady and the left more improvisatory.

Ex. 6 Rubinstein, 1936, opening.


Another curious aspect revealed at the beginning of this example is the ‘swing’ in the left hand semiquavers.  You can hear it especially in the left hand of bar 2 where the melody first enters.  That is, the left hand semiquavers are often in slightly unequal pairs tending towards a triplet relationship.  They are not just evenly spaced semiquavers.  Rosenthal does it too but it is not as constant, obscured by more wayward shifts in timing.  The swing element seems to have rubbed off on several modern performers.  Idil Biret uses it sometimes, especially when the music becomes more unstable in mood.  I do the same at bars 17 and 18, but must admit elsewhere that it is not yet sufficiently ‘in my blood’:

Ex. 7 Tebbs (2014) bars 17-18


At the corresponding point Rubinstein’s left hand swung quavers approach a more pronounced swing, approaching a definite triplet relationship:

Ex. 8 Rubinstein (1936) bars 17-18


Note that this swing does not involve dislocation in the right hand.  The right hand is simply pulled along with the shape of the left.  Such swing is ironed out of the same passage in his later 1965 recording, though it is in evidence elsewhere to a lesser degree than in this example.

Unwritten arpeggiation

The technique of playing chords in an arpeggiated fashion where not indicated in the score is of course associated with the harpsichord, yet the evidence points to the fact that this technique seems to have been handed down, as it were, to the piano.  According to Czerny, arpeggiation was commonplace among pianists, to the point where he reports that some pianists it seemed had forgotten how to play chords together.  Even later in the nineteenth century, Peres da Costa cites reports of Brahms having a penchant for spreading chords here, there and everywhere, though of course we cannot assume Chopin was the same.  There are few chords as such in this Nocturne, but arpeggiation of the two-note right hand chords that appear occasionally throughout the piece is indeed found in certain places by our historical pianists.  Rubinstein’s 1936 recording does so only once in bar 13 but he plays this chord with the notes together, as written, in his 1965 recording, updating his style again (Fig. 4) so that it reflects only what is actually written down in the score.

Fig. 4

Nocturne bar 13

This creates an effect  that Chopin himself notates with grace notes several times in the right hand part at climactic moments, for example bar 8, but did he expect pianists to use it elsewhere?  Whilst rubato and dislocation, it could be argued, are not possible to notate precisely and therefore beyond the bounds of what the composer can specify, simple arpeggiation is relatively easy to write down.  The point though is perhaps more that the composer makes certain choices and leaves others to the performer.  If arpeggiation was as commonplace in the nineteenth century as reports suggest then Chopin must have accepted that individual interpretations might apply the technique in other places as well as those specified, the notated instances merely being  places where the technique is absolutely necessary for  expressive effect.   In this fashion my own recording copies a number of arpeggiations similar to the one in Fig. 4, all of which can be found among  my chosen cohort of early recordings (in bars 11, 13, 33 and 61). I have also added an original instance in bar 73, after a similar one in the coda by Rosenthal.  But one could go further.  To give an idea how rife such impromptu arpeggiations could be, if the mood took one, have a listen to this section from Diemer’s recording where no arpeggiations are indicated in the score (I count eight here in the right hand!):

Ex. 9 Diemer (1904) bars 34-39


Another fascinating arpeggiation is introduced by Koczalski where he colours the left hand very subtly with an arpeggiated chord just before the beginning of bar 27, then following through with the written notes of the left hand accompaniment.  So subtle and soft is this that I only noticed it very recently.  He only uses such an effect twice in his recording and I have not attempted anything like it (compare Ex.10 and Fig.5 below).  This goes beyond merely arpeggiating written chords, effectively adding notes to the score that Chopin did not write.

Ex. 10 Koczalski (1948) bars 26-28

Fig. 5

Nocturne bar 25

Extended trills

One final brief topic mentioned in Peres Da Costa’s book is the unusual approach to trills which can be found in some early recordings and which is supported by various pedagogical texts.  To add extra brilliance to long trills they are played for slightly longer than the written note suggests, and in this piece the trill heralding the spectacular cadenza is indeed begun a little before the left hand reaches the middle of the bar in Diemer’s and Koczalski’s recordings, then prolonged a little more by slowing the left hand accompaniment underneath the trill.  I manage the early start to the trill,  but not the expansion of the left hand.  The audio example below is from Diemer’s recording.

Ex. 11 Diemer (1904) trill in bars 50-53



The above discussions could go on a very long time in terms of both detail and debate, but it is enough to give the reader a feeling for the issues involved.  Most of the pianistic effects here might be regarded by the amateur pianist as very subtle, and indeed they are, but the sum of the parts amounts to a whole which is palpably different to our traditions of Chopin playing (and more importantly teaching).  I am sure, however, that some of the great modern pianists are very aware of some of these early recordings. We have already mentioned Idil Biret a number of times as there are moments where she plays with dislocation and swing, but this is done with careful moderation.  Pires’s very fine recording also seems uncannily aware of these early recordings in its flexibility of tempo and judicious use of  dislocation.  Then there is the influence that Rubinstein’s recordings have had as classic readings of the nocturnes.  It has been interesting to observe that most of his changes in interpretation between 1936 and 1965 correspond with trends in pianism also identified by Peres Da Costa, largely in a direction away from the early performance traditions.  There is less swing, virtually no dislocation or arpeggiation, for example in the later recording, and rubato has become more controlled.  I hope that my own recording, far though it may be from ticking all the boxes of the ‘Golden Age’ of pianism, may stimulate further debate about how to play Chopin in an historically informed way, and perhaps about whether this is possible or desirable for today’s pianist, and that this article might switch pianists on to the fantastic legacy of early Chopin recordings by these great artists.

To my mind my own recording does sound curiously different to modern recordings of Chopin interpretation, but actually does not go as far as some early twentieth century artists in the degree to which some un-notated elements are applied.  One example is so-called ‘dislocation’, which I reserve for certain moments during the three statements of the main theme, in the coda and in just a few other places, but which I do not use throughout, whereas several of these early recordings seem to play the two hands slightly out of time with each other at some point in almost every bar.  Such an approach existed in a completely different cultural milieu to today and it is hard to enter in to this convincingly having been exposed to modern pianism for so long in which this splitting of the hands – which can often happen unintentionally to the student pianist – may still be frowned upon by many teachers.  Having said this, setting the hands free from the modern norm of togetherness is an interesting exercise and begins to feel quite natural and expressive after a while.  Listening to the early recordings and trying out the technique, one learns that there can be many shades and uses for it.

What may convince us about the historical place for these techniques is perhaps the way they might express the ultimate role of subjectivity in the Romantic worldview, the exploration of a personal touch, a personal rubato and set of nuances.  At an extreme the solo piano work, and especially the nocturne, is an exploration of this inner world of nuance and personal affect.  Any attempt to project this state has to come from within rather than from an attempt to emulate this or that particular recording.  At least I can say, from the recordings I have made, that I have never played the piece the same way twice.  But I have felt my way to a different, possibly more authentic, interpretation by exposing myself to these fantastic recordings by pianists who very much had their feet  (and ears) planted in the same century as Chopin.

Discography  (pre-1950s)


Diemer, Louis  (1904). Piano Music – CHAMINADE, C. / MENDELSSOHN, F. / DIEMER, L. / GODARD, B. / CHOPIN, C. (Three French Pianists) (1901-1919), Symposium Records dur. 3’38’’, acoustic.

Koczalski, Raoul (1948). Chopin: Historical Live Recordings 1948, Fryderyk Chopin Institute, dur. 5’13’’, acoustic.

Pachmann, Vladimir Da (1927).  Chopin, F.: Piano Music (The Complete Issued Electrical Recordings) (Pachmann) (1925-1928), Radiex Music, dur. 4’25’’, acoustic.

Rosenthal, Moriz (1936). Moriz Rosenthal Volume II, Pearl dur. 4’55’’, acoustic.

Rubinstein, Artur (1936). Chopin: Nocturnes and Scherzi (Rubinstein) (1936-1937),  Naxos, dur. 6’14’’, acoustic.

Discography (post-1950s)


Biret, Idil (1991).  Fryderyck Chopin: Complete Piano Music, Vol.5, Naxos, dur. 6’16”.

Pires, Maria Joao (1996).  Chopin: Complete Edition. Deutsche Grammophon, dur. 6’37”.

Rubinstein, Artur (1965-66).  Rubinstein Chopin Nocturnes.  RCA Red Seal, dur. 6’20”.

Select Bibliography


Peres Da Costa, Neal (2012).  Off the Record: Performing Practices in Romantic Piano Playing.  OUP USA

Philip, Robert (2004). Early Recordings and Musical Style: Changing Tastes in Instrumental Performance 1900–1950, CUP.

Schonberg, Harold C (1987).  The Great Pianists from Mozart to the Present, Simon and Schuster.