To repeat…..or not to repeat? Thoughts on Schubert’s D960

Whether or not to meticulously observe the exposition repeat in Schubert’s final sonata, the D960 in B-flat major, is a question which continues to trouble pianists, musicologists and listeners alike. The debate concerns aspects such as authenticity, personal taste, prevailing musical fashion, and timing. It has cropped up the press this autumn as British pianist Paul Lewis completes his cycle of Schubert’s sonatas, and has exercised myself, colleagues and other musical friends in discussion.

The opening bars of the first movement of Schubert's Sonata in B-flat Major D960
The opening bars of the first movement of Schubert’s Sonata in B-flat Major D960

When I first started learning the D960, the day my new piano arrived from Chappells in January 2007, I did not doubt the correctness of adhering to the score and repeating the exposition. Throughout my musical studies, I had been taught to trust the composer’s intentions, that composers know what they are doing, and that repeats are there for a reason.

In Baroque repertoire, sectional repeats are commonplace, often used to reinforce material and to offer soloist and/or ensemble the opportunity for some interesting extemporisation or ornamentation in the repeated material. At a recent harpsichord masterclass I attended at Handel’s House with some of my students and those of a teaching colleague, the harpsichordist, Claire Williams, encouraged students to experiment with different effects in repeats, such as using the upper manual of the harpsichord, or employing a lighter touch. Bach and his contemporaries would have expected and encouraged it.

In the piano sonatas of Mozart, for example, a repeat of the exposition in the first movement is often a reinforcing device, a reminder that this is a sonata, in Sonata Form (Exposition, Development, Recapitulation). And Mozart would have expected his keyboard player to offer some extemporisation – changes in dynamics, ornamentation and so forth – in a repeat. As musicologist, pianist and noted Mozart expert Robert Levin states:

“If you take a repeat, for instance, heaven forefend you should play exactly the same way you did before. It’s like telling somebody, ‘Look, if you could go back five years and relive your life would you change anything?’

Some people might say, ‘Nah, I think I basically did what I wanted to do,’ but a lot of people would say, ‘Oh boy. I can name you a dozen things that I would done differently.’ A piece of music is an almost cinematic opportunity to revisit a situation and reinterpret it.”

In a repeat of the exposition, the musical experience does not stay the same – there can never be a “straight repeat” because we are human and do not seek to replicate MIDI recordings – but the repeat of the same material helps the listener to assimilate the musical ideas expressed in the composition, and makes it easier to comprehend the structure of the piece.

In Schubert’s last sonatas, the repeat sign is written for an exceedingly long exposition, while the material of the exposition is repeated a third time in the recapitulation with little alteration. This has led some pianists to omit the exposition repeat in performance. But in the last Sonata (and, indeed in the penultimate Sonata, D959 in A), the first ending of the exposition contains unique material, leading the music back to the movement’s opening. If the music is performed without the repeat, this material is missed out completely as it does not appear in the second ending of the exposition (if one does not observe the repeat, one goes straight to the second time bar and thence to the development section). These bars contain striking material, which does not appear anywhere else in the piece, and is significantly different from the second ending. When the exposition ends a second time, Schubert introduces an extraordinary bridging section, three ethereal chords a single bar, which seem to come from somewhere else completely, and lead the music into darker, minor-key territory.

The first and second time bars at the end of the exposition
The first and second time bars at the end of the exposition

British-Hungarian pianist Andras Schiff feels that the omission of the exposition repeat is as “the amputation of a limb” (Schiff, “Schubert’s Piano Sonatas”), while Alfred Brendel declares the exposition repeats in the final two sonatas to be “unimportant”, that the transitional bars in the D960 are too unconnected to the rest of the movement, and that their omission actually contributes to the coherence of the piece. Brendel also states that “repeat marks must not be taken as orders to be automatically obeyed, as if the repeatable section were written out by the composer.” (“Schubert’s Last Sonatas: An Exchange”; Frisch and Brendel). But the last two sonatas have more than mere “repeat marks”: the bars of music specifically written out by the composer suggest that Schubert requires the repeat to be observed.

Personally, I have never doubted the inclusion of the exposition repeat, and it irritates me when I hear performances, either live or on disc, which omit it.

In reading reviews of live performances of Schubert’s last three sonatas, it strikes me that many pianists omit the exposition repeat/s simply to save time. Most concerts last around 90 minutes (plus interval); any longer and the audience starts to get restless, worrying about last trains etc. Critics, who praise the omission of the exposition repeat, are similarly impatient, presumably keen to get out of the Wigmore and into the pub before last orders! Played in its entirety, with all repeats intact (including in the third movement Scherzo), the D960 comes in at around 40 minutes.

Another issue which relates to this is the tempo of the first movement. It is marked Molto Moderato, and in some pianists’ hands (Richter), this can verge on Adagio! Moderato means “not rushing or dragging”, and Schubert also used the German term mässig, implying the calm flow of a measured allegro. A quick glance through Spotify of recordings of this work reveals that most versions which include the repeat come in at around 20 minutes; Richter’s is 25 minutes, which is verging on self-indulgent, while Pollini’s is 18 minutes. At this length, the opening movement of the D960 is as long as an entire sonata by Haydn or Beethoven.

It is customary to programme Schubert’s last three sonatas together, just as Beethoven’s last three are, to stress the interrelations between the three. The D958 and D959 are played in the first half of a concert, and the D960 after the interval. It can feel like a long haul for the audience, and one of the solutions to this problem is to shorten the programme by omitting repeats, mainly those of the opening movements’ expositions. However, it irks me when I read reviews of pianists such as Paul Lewis (a protégé of Brendel), who is completing his survey of Schubert’s piano and lieder music with performances of the final three sonatas, in which critics praise the omission of the repeat without outlining at least a little of the background to this ongoing debate. I feel that without the exposition repeat the audience is cheated of the opportunity to experience Schubert’s compositional intent, and the drama, which comes from the tension between the contrasting harmonies in the exposition, and the transition between exposition and development, is lost.

To my mind, a well-played opening movement of the D960, with repeat, need not feel unnecessarily long. The music has an spaciousness (its key and scale always calls to mind, for me, a great river charting its final course towards the sea) which is offset by a considered interpretation of the Molto Moderato marking. I particularly like Pires and Uchida in this work. I have enjoyed all three sonatas in a single concert, or the D960 on its own, preceded by one of the D899 Impromptus in a lunchtime recital at the Wigmore (Andreas Haefliger – repeat omitted!). Both programmes work equally well.

Here is Maria Joao Pires

Maria João Pires – Schubert : Piano Sonata No.21 in B flat major D960 : I Molto moderato


  1. It is with some trepidation that I contribute to this, as I do not have a formal musical education, and can barely read music: I can only go by the subjective impressions I get when I listen.
    I have heard this piece in live performances only twice – once played by Elisabeth Leonskaja, and, more recently, by Maurizio Pollini. They both observed the repeat. On recordings, I have performances by Rudolf Serkin and Wilhelm Kempff (both with repeats), and Clifford Curzon and Alfred Brendel (both without). I must confess that, to my ears at least, that ff trill seems out of place in Kempff’s recording, and – I realise this may sound like blasphemy to you – wonder whether it would have been better omitted. However, that same trill in Serkin’s recording makes perfect sense. I don’t, to be frank, miss those at the end of the exposition when they are omitted. Serkin is the only pianist I have heard in this piece who has completely convinced ma about that trill.

    Beyond that, I will leave this debate to those more learned in musical matters than myself. I do take your point that if Schubert wrote it in, we should respect his intentions; and that if our interpretation can’t make sense of that ff trill, then it’s our interpretation that must be wrong rather than the trill itself.

    • Thank you for a thoughtful post, Himadri. As pianist Alisdair Hogarth said to me on Twitter earlier, in response to this post, it’s really about how convincing/involving the performance is. So I suppose in the end it becomes very subjective & a matter of personal taste (as is the way with all music!)

Comments are closed.