Everyone - quite reasonably - lauds the Richter account of this profound and groundbreaking work, but it turns out Schubert had some ideas of his own about how this sonata to be played, and this breathtaking account by Kovacevich comes as close to definitive as I can imagine. He plays a genuine pianissimo and molto moderato, and his interpretation contains a sense of unease and urgency that is hugely compelling. There is no slouch toward profundity, no gesture of unnecessary sophistication. Instead there is a consistent variety of articulation -- warmth [23:22], sadness, violence, tranquility -- all as the work demands it.
Schubert’s final sonata is famed for its sense of inner stillness, but often that’s just a function of tempo - there’s a lot of compulsion and vigour in it, as Kovacevich’s recording shows. What is truly extraordinary about this sonata, however, is its harmonic life. The D.960 is full of harmonic shifts of vast structural violence which nonetheless sound perfectly serene - “magical” modulations, as they are often called, which receive little (obvious) preparation. It’s worth noting that there’s even such a “magical” shift writ large on the structure of the entire piece, since the second movement is in the key of C# minor, making it the most harmonically isolated of all the movements in Schubert’s work.
You encounter the first such magical modulation at 0:54, when the tonality shifts from Bb major to Gb major without any apparent preparation. At 1:57 there is another less sudden shift to F# minor, which means that we get a three-key exposition. In fact both shifts have been anticipated by the Gb trill which trembles minutely and ominously in the midst of the exposition’s first theme.
The next magical modulation appears when we get a sudden shift from an F major to a C# minor chord at the beginning of the development section [9:45]. This leads into a whole series of incredibly colourful modulations, which apparently culminate in Db major [11:01], except that we then wander via increasingly dramatic dissonances into D minor, where we get the climax of the movement in which the first theme of the exposition is repeated [11:50] in barely recognisable form. (The modulation from D minor back into Bb major is fairly noteworthy too, as the D minor dies away in the most achingly gradual manner possible, in little stepwise chromatic shifts, over about 90 seconds of music.)
The second movement contains what is perhaps the most famously moving modulation in Schubert’s work, a sudden shift from C# minor to the remote terrain of C major [26:22] right after a half-cadence on the dominant, although that is followed by an equally beautiful shift from C to E [25:56].
There are many other fun things going on in this sonata apart from just striking modulation, however. There is some unexpected playing on your expectations of how a melody is phrased [30:55, where the melody contains a false beginning that isn’t actually a false beginning, if you look at the phrase marks - it’s just that it’s grown by an extra bar], and some more harmonic cheekiness - the finale begins in apparent C minor. The climax of the finale [37:18] is, remarkably, in C-flat major, from which the bass descends in chromatic modulation eventually to G to return to the main theme.
The interrelations between the movements are also ingeniously set up: in the finale's coda, the octave on G [40:24] descends through Gb [40:30] to F, which reaches all the way back to that Gb trill in the opening of the first movement which resolved downward to F so inaudibly: a lovely instance of cyclic form. (In fact, the G-Gb-F resolution is central to the whole finale -- its how the tonality in the beginning of the movement moves from C minor to Bb major). And there’s the fact that the second theme of the finale [37:24] has the same melodic contour (5-8-7-6-6-5-(5-4-4-3)) as the remarkable C major modulation in the second movement.
And it must be mentioned that in the first movement the end of the exposition (when you play it the first time) contains 9 bars of weird, stuttering, totally new material [4:42], and the only instance when that Gb trill is played ff (this makes omitting the repeat kind of obscene, since this material totally disappears from the piece if you do.)
00:00 - Mvt 1, Molto moderato
19:58 - Mvt 2, Andante sostenuto
29:16 - Mvt 3, Scherzo
33:15 - Mvt 4, Allegro ma non troppo