Buying boxed sets of central repertory is often a lazy way of collecting, fraught with danger for the discriminating listener. But there are exceptional cases, and the case of the 32 Beethoven piano sonatas is certainly one of them. The sonatas make up a wonderfully diverse yet utterly cogent and self-contained body of work. There isn't a single sonata that it is safe to ignore (even the Op. 49 sonatas have their merits) yet try to collect them separately and you will have the devil's own job over couplings and availability. Having the complete opus played by a single great artist is also surprisingly satisfying: the equivalent of the Goldberg Variations writ large.

But which great artist? Of those listed above I would contemplate only the Brendel on Philips with a reasonable degree of equanimity. Ashkenazy's Decca cycle, made over many years, lacks the kind of consistent insight and acumen the music demands, though the playing is never less than beautiful and the recordings are always expert. As for Barenboim, here I would prefer this, the earlier of his two cycles, made in the late 1960s for EMI, but even this is flawed by too many unduly protracted—and so musically dumb—slow movements. In fact, the gramophone has given us great sonata cycles by three pianists: Schnabel, Arrau and Kempff. Some would say that Schnabel's is the greatest of all, despite some erratic execution and the very obviously dated 1930s sound (to be released by EMI in due course, this cycle is presently available on Nuova Era/New Note (CD) 90034/41). Nowadays, though, it must be seen as an historical set, a specialists' choice, however phenomenal. Arrau, mirabile dictu, is still recording Beethoven in his late eighties but his 1960s Philips cycle, yet to be transferred to CD, will probably remain as the great representative example of truly searching Beethoven playing. It is an epic journey (every repeat is honoured) that one approaches in the kind of frame of mind one might otherwise reserve for a re-reading of the works of Tolstoy or Dostoevsky. And then there is Kempff: Arrau's anti-type, and, it has always been said, Schnabel's nominated heir; the pianist Schnabel recommended should replace him if he failed to complete his own integral recording of the sonatas. In fact, Kempff recorded the 32 sonatas three times; once, before the Second World War, on 78rpm discs; for mono LPs in 1951–2; and again, this time for stereo LPs, between January 1964 and January 1965. It is this last, best recorded, and in many ways greatest cycle, that has now been reissued on nine CDs.

It is a wonderful set, one of the finest things the gramophone has ever given us, not simply because it offers great music and great music-making, but because it is, over its ten-hour span, a thing of such unquenchable vitality. When the set first came out at Christmas 1966 I remember playing it through over a single glorious weekend. And since then I have met others who acquired the records and were similarly spellbound by them. For all Beethoven's rages, moods and temporary glooms, he remains one of humanity's great torch-bearers and it is this torch-bearing quality—a transcendent vitality of mind, imagination, and spirit—that Kempff, like Schnabel before him, so unerringly and upliftingly conveys.

Kempff (now in his 95th year but long since retired) probably owed much as a Beethoven interpreter of genius to his early grounding in the music of Bach (Kempff's father was a Lutheran pastor) but he was also an interpreter blessed with a rare sense of fantasy. He was also, even in the recording studio, a wonderfully spontaneous musician which is why there is throughout this set a sense of Beethoven's music being liberated at its well-source only to be further transcribed and transformed by the dazzle of an especially 'fine' poetic imagination. It is, for the most part, light-filled playing—the rising arpeggio which opens the finale of the early A major Sonata assumes an almost symbolic quality here—that chimes well with the essentially transcendentalist temper of Beethoven's own mind in these works. It is also technically fine. We don't think of Kempff as a virtuoso after the manner of Horowitz, but behind that spare and resilient keyboard style there is always great virtuosity. Listen to the last movement of the first sonata, a superb stormscape that is evoked by Kempff with a thrilling mastery of tone and dynamics, the sound burnished and alive; but then note, at the arrival of Beethoven's strangely angular lyric cantilena, how exquisitely the melody is floated and sustained.

Ezra Pound once said that music rots when it gets too far from the dance, and there is no doubt that Kempff's vital projection of rhythm is one of his greatest assets. There is the wit and ebullience of the early Allegros, the superbly timed trills and gruppetti, and the proud festive character of something like the Presto of Op. 10 No. 3. His rhythmic detailing is both sophisticated and instinctive. He understands what is called in Plainsong, I believe, elan et repos, exertion and rest; though it should be said that the 1964-5 cycle is more direct, more rhythmically single-minded, than the slightly more Furtwanglerish 1951–2 mono cycle. In the opening paragraph of the C minor Sonata, Op. 10 No. 2, this sense of elan et repos is backed by superb phrasing to create special and significant patterns of movement: the angularities, the asperities, the sudden graciousness of the music. With many pianists—conductors, too—you get used to the rhythmic carriage of their Beethoven: vital, perhaps, but predictable and too little differentiated. With Kempff this is never the case. The dance rhythms of the first movement of Op. 31 No. 1 are quite distinct from the more dispassionate progress of the so-called Tempest Sonata which, in turn, is different again from the untiring energy and concentration of some of Beethoven's more driven toccata-like movements. And how finely differentiated are the various essays in fugal writing that so distinguish (in every sense of the word) the late sonatas.

Kempff never bores us with Beethoven, even if, on occasion, he may seem to go to another extreme. In the slow movement of Op. 10 No. 3 his tempo is astonishingly brisk at quaver=60. Compare that with Arrau at 48, Schnabel 42 and Barenboim 36. Schnabel's is perhaps the finest of the four performances but where Barenboim is patently far too slow, Kempff never seems unduly fast. Kempff can be something of an enigma: a devoted Beethovenian who may suddenly seem to be passing by the music too dispassionately. In this set I recall being shocked by the restless opening to Op. 101 and marginally disappointed, perhaps, by the final movement of Op. 111. But 25 years on I hear that flow and seeming detachment as something other: music-making that has won the right after this great journey to see the world from a loftier station and a further shore. (Kempff's view of the music chimes with what Beethoven eventually does on the last page of the Diabelli Variations.)

There are no obvious failures in the set, though it must be said that the 1964 account of the Appassionata sounds perfunctory beside its 1951 predecessor. Kempff's performances of all the great named sonatas—the Pathetique, the Moonlight, the Waldstein, Les adieux, the Hammerklavier—are triumphs of communicative music-making. It is here, though, that one has to ponder Kempff's view of the great repeats question. This is something we know that Beethoven also agonized over which tends to suggest that those we have should be observed. By and large, Kempff does but there are half a dozen sonatas where first movement expositions are not observed. Sometimes it is clearly a matter of intuition. Indeed, one can sense as the repeat sign approaches that Kempff will ignore it. This is certainly the case in the Pastoral Sonata in the Pathetique, and in Op. 31 No. 3. I was unprepared for the omission in Op. 10 No. 2 and the Waldstein, and still find the omission of the Hammerklavier's exposition repeat the most worrying. But this is the extent of the No Turning Back group. In general, Kempff relishes the chance to go over the ground one more time.

''Spare and resilient'' is how I described Kempff's playing. It is light-filled, too. As a result, it is possible to feel a certain initial surprise, disappointment even, with the actual sound-quality of these performances. As piano recordings go, these may strike you as being clear and brilliant but a shade brittle and lacking in bass. It was ever thus. The earlier mono set sounded much the same. In fact, the ear quickly adjusts to Kempff's style and one comes to relish the clarity of the sound and the superb balancing of the contrapuntal detailing. The LPs, with DG's famed silent surfaces, were very fine. On CD they come up superbly, with only minimal tape background; and on CD you know that there will never be any clicks or crackles to distract one's attention away from Kempff's playing.

Perusing the adverts in Gramophone, I see that the asking-price for this set is in the region of £50. Relative to a few tanks of petrol, a modest meal for two, or a couple of acts of a Puccini opera at Covent Garden, it is astonishing value. At about the time the set originally appeared I remember being told the story of an undergraduate at Trinity College, Cambridge who elected to buy the complete works of Tolkien with some academic prize-money. His tutor demurred: ''I don't believe, Mr ––––––, that this synthetic mythology will look well on your shelves in 15 years' time''.

Fifteen years, or 50: this set will always look well on your shelves, and in its new durable format it could go on giving pleasure and insight for generations to come.'