> A QUINTET OF GOLDBERGS - for the pianoforte [CH]: Classical CD Reviews- Nov 2002 MusicWeb(UK)

A QUINTET OF GOLDBERGS - for the pianoforte

Johann Sebastian BACH (1685-1750)
Goldberg Variations BWV 988 (1742)

(pianoforte), also includes Partita no. 5 BWV 829
Recorded live, Philharmonie, Berlin, 25th April 1999 (Goldbergs)
Meany Theatre, University of Washington, Seattle, 26th November 1996 (Partita)
AT 00-00341
[tt: 64’ 15"; Goldbergs: 43’ 03"]
For information and sales contact:
Annette Tangermann
Friedenstrasse 16
D-14109 Berlin

Glenn GOULD (pianoforte)
Two performances, both recorded at Columbia 30th Street Studio, New York City
1: 10th, 14th-16th June 1955
2: 22nd-25th April, 15th, 19th, 29th May 1981
Plus interview of Gould on the Goldbergs by Tim Page, rec. 22nd August 1982, Toronto,
and studio out-takes from the 1955 recording
SONY CLASSICAL LEGACY SM3K 87703 [3 CDs: 38’ 26", 51’ 14", 63’ 27"]
Crotchet   AmazonUK   AmazonUS

Murray PERAHIA (pianoforte)
Recorded at Musica-Théâtre, La Salle de Musique, La Chaux-de-Fonds, Switzerland
9th-14th July 2000
SONY CLASSICAL SK 89243 [73’ 29"]
Crotchet   AmazonUK   AmazonUS

Wilhelm Kempff
Recorded in Beethovensaal, Hanover, July 1969
DG GALLERIA 439 978-2 [63’ 05"]
Crotchet   AmazonUK   AmazonUS

The stimulus for this article was the submission for review of Craig Sheppard’s recording of the Goldberg Variations. Since a recording of such a massive work seems to call for something far more detailed than would normally fall within the scope of a straightforward record review, I have made a variation-by-variation study of five CDs including, obviously, that by Sheppard.

I want to stress that this is not intended to be a complete survey of Goldbergs on record, or even pianistic Goldbergs on record (apart from the many harpsichord versions, at least one recording has been made on the organ); even a reasonably complete survey would have to include at the very least, in addition to the following, the piano versions by Rosalyn Tureck, Charles Rosen and the recent and much-praised Angela Hewitt. My intention has been to study, through a representative sample of noteworthy performances, some of the interpretative problems which this magnum opus presents. The versions chosen, then, are those by Gould, whose 1955 recording has become one of the icons of our time and who returned to the piece at the end of his life in 1981, Perahia, who "came out" at the same time as Sheppard (at the 1972 Leeds Festival) and so presents an obvious point of comparison, and Kempff, as a representative of the older, pre-Gould generation whose roots go right back to Busoni.


Alone of these, Sheppard’s is an unedited live performance. A priori, you might feel that this of all works could benefit from the performer’s having all the time necessary to sort out the problems it poses on his own. I can only report that Craig Sheppard is able to present even such a large-scale work as this in concert with total assurance. Just here and there a variation ends with a rallentando which seems a little excessive and I wondered if he always does this or whether it was a spur of the moment decision. But on a purely technical level there are none of the slips or hesitations which might be forgivable in a live performance over the span of such a mighty work, and there is also a sense of ready communication which is not so easily attained in this piece, live or not. Audience noise is small and the piano quality compares favourably with that of Perahia’s recent studio recording, as well as being richer than the elderly Kempff, which is nonetheless excellent for its age.

The Gould recordings are a special case since Gould intervened strongly on the technical side and saw to it that the piano and the recording equipment between them matched his vision of an instrument which maintained far more of the dry, pinging quality of the harpsichord than is normally realisable on a modern grand. Even so, the 1955 recording does sound rather close and airless – it is getting on for half a century old, after all – and even threatens distortion at times. The 1981 version presents a more attractive picture of Gould’s pianistic ideals, and in fact he has posthumously not had it all his own way. In 1981 digital technology was in its infancy but Gould insisted on having it just the same. To be on the safe side the engineers had a traditional analogue tape going as a backup. In the event it was not needed – or so it seemed, for the original issue was based on the digital recording. The reissue producer, Louise de la Fuente, explains how she and her colleagues were listening to some of the session tapes in preparation for this special edition and were using the analogue reels for easier handling, when it struck them that the analogue recording was actually "more musical and natural" than the digital one. Of course, all the editing prior to the original issue had been done with the digital tapes, so to prepare for the present reissue the analogue tapes had to be edited again from scratch, using Gould’s own annotated score and session notes, which had fortunately been preserved. This will be music to the ears of those of us who were convinced all along that the finest analogue recordings had a lifelike quality and above all a warmth which subsequent digital technology has scarcely matched.

It should be pointed out to anyone who wonders if the two Gould performances are sufficiently different to justify buying them back-to-back that this is a very different matter from those small adjustments of detail you would find in a hypothetical issue of, say, Karajan’s first and last recordings of Beethoven’s 9th. At certain points the differences between the two Goulds are greater than those between one of the Goulds and any of the other pianists. And yet the child remains father to the man. So yes, it most definitely is worth it, especially when you get a third CD with a lengthy interview – Gould’s last – which raises a number of fascinating and provocative points, and some out-takes from the 1955 sessions.



The performer’s first decision is whether to make both, one or none of the repeats (and, presumably, follow his chosen pattern right through the work), and what to do about the ornaments.

Kempff immediately distances himself from the others by ignoring the ornaments altogether; all the others play them and are fairly consistent with each other in interpreting them (the ornaments are not as ambiguous as these things can be elsewhere). My only query is the arpeggio in b. 11 which is played downwards by Gould and Sheppard instead of upwards as usual.

The effect of Kempff’s omission of the ornaments is that he seems to be playing a different piece – the more so when, with fewer notes to deal with, he opts for a more flowing tempo and his performance, including the first repeat, is barely any longer (1’ 58") than the Gould 1955 (1’ 53") or the Sheppard (1’ 51"), which make no repeats at all. Even if you can get used to the barer outlines of the unadorned theme I still find it perplexing since, apart from the actual ornaments as such, Bach also often breaks into more florid writing (as in bars 4 and 7) which is, in effect, written-out ornaments. Kempff obviously plays this as written with the result that the listener unencumbered by scores or technical knowledge of what all this is about will nonetheless feel a certain lack of equilibrium as the music veers between the "straight and narrow" and the decorative.

But before we get too superior about the old fuddy-duddy ("they didn’t know any better in his days"), we should remember that ornaments are supposed to be decorations applied to a basic structure, and while electing to play the ornaments today, we might hope not to lose sight of the basic simplicity as revealed by Kempff. In this respect Gould 1955 is the most successful, followed pretty closely by Sheppard. Perahia is alone in playing both repeats (the performance lasts 3’ 58") and does not entirely avoid the suspicion that his ornamental trees are obscuring his simple wood. When you see that Gould 1981 takes 3’ 05", you might suppose that he has taken at least one repeat this time but no, it’s just very, very slow (and accompanied by groaning vocalises). Such is his mastery of tonal gradation that he does hold the ear, more or less, but it’s weird.

A further decision to be taken comes towards the end, when Bach moves into more flowing semiquaver writing. How staccato to make the bass? Sheppard is the most staccato, and I find this draws a little too much attention to itself. Gould and Perahia are more natural (Kempff, as you might imagine, is completely legato).

So the stage is set with Gould 1955 in the lead, followed fairly closely by Sheppard and a little less closely by Perahia, with Kempff and Gould 1981 seriously compromised from the start.

Variation 1

The performance I loved was Sheppard’s. His light detached touch and crisp accents express a rhythmic joi de vivre which I don’t find in any of the others. Perahia is not that much different, but when that little amounts to a slightly fuller tone and marginally slower tempo the result is noticeably more heavy and serious. Perahia adds a few ornaments in the repeats (please assume henceforth, unless I point out any anomaly, that each artist pursues throughout the work the policy over repeats that he established in the Aria itself). Gould pitches in at a bristling tempo. It bowled people over in 1955 and no doubt still would, but for repeated hearing it is surely a little too manic. In 1981 Gould presented in a certain sense the same performance (strong-as-steel articulation at an unremitting forte) vastly slowed down (from 0’ 45" to 1’ 10"). In a way the nearest to Sheppard’s sheer likeableness is Kempff; although he has a tendency to put the pedal down (all the others keep their feet well out of harm’s way) whenever the writing makes this possible without actual dissonance, he also demonstrates that when the touch and the pedalling itself are properly light the result need not be either muddy or cloudy.

There is, however, one aspect of Sheppard’s performance over which some listeners might disagree. When the right-hand melody of the first four bars is passed to the left hand, he has no hesitation in letting it dominate over the right hand. On the instruments available to Bach (if we ignore the fact that he was asked to try out a few fortepianos at the end of his life) this could only be done with a double manual harpsichord or organ. Some of the variations are marked for 2 keyboards (though the object was basically to allow the hands to cross freely) but this variation is marked for one. None of the others points out to the listener Bach’s contrapuntal dialogue to this extent and it is Kempff who most completely lets both hands remain at the same volume, leaving the counterpoint to speak for itself. Interestingly, of the four pianists, Kempff was the only one (as far as I know) who also trained as an organist, so he would have been used to equal balancing of the parts from the beginning. Personally I rejoice in the lively dialogue between the voices which Sheppard creates and I don’t feel it goes actually beyond the baroque aesthetic.

Variation 2

This has an obvious continuity of mood with the preceding variation so it is not surprising if the performers yield similar results: Sheppard the most sheerly delightful with his light detached playing, Perahia in similar vein but a mite heavier, Kempff more legato but by no means heavy, Gould 1955 brilliant but a little manic. Gould 1981 convinces more here since he lets us appreciate the two-part writing in the upper voices, and a detail like the suddenly singing legato bass-line after the double bar (the bass has been scrupulously staccato till now) show how deeply he had re-thought his interpretation, whatever we think of the results.

Variation 3

The first of the canons. Here Perahia turns the tables for in this case it is Sheppard’s insistence on very staccato bass semiquavers which seems a shade heavy and dogmatic. Perahia is very slightly faster with light detached articulation, creating that sense of sheer joy which, in the first two variations, I had missed in his playing and found in that of Sheppard. Also, while all four are good at keeping the lines clear, Perahia’s separation of the lower line towards the end of the first half is quite miraculously clear. Of course, if you maintain that the lines should be balanced equally, you will not like this, miraculous or not, and in that case your choice will likely be Gould 1955. The brittle articulation comes as near as imaginable to an actual harpsichord performance. There is élan but this time it is not so hard-driven as to seem manic and this was my favourite after Perahia. The coolly poised Gould 1981 (which makes the first repeat) has its attractions since the music is more than eventful enough to fill the slower tempo. Kempff stands apart, treating the piece as a songful pastorale, with the bass-line a seamless legato; it could have been dull and heavy but such is his transparency of texture that it manages not to be. He repeats both halves in all the canons.

Variation 4

All four are agreed that this has a sturdy Handelian character – even Kempff is not wholly legato. He differs from the others, however, in not maintaining an even forte throughout but grading the dynamics so as to avoid sounding too insistent, a danger not entirely avoided by Gould 1955. Sheppard shows that it is possible to make the variation effective at a steady forte (and he repeats both halves here) by keeping a spring to the rhythm. Gould 1981 (with the first repeat) is in similar vein, just a shade more solid. Perahia softens the dynamics for his repeats, as well as adding ornamentation. I find each equally effective here, according to his own lights. Sheppard is also notable for the way in which he gives the bass in the first part the sonority of an organ pedal.

Variation 5

This variation can easily seem a mere technical study. Sheppard gives us a further taste of his delightful light fingerwork, leaving Perahia just a mite serious and heavy in comparison. Gould 1955 seems to want to show us how fast he can play (very fast indeed), yet the curious thing is that Gould 1981 comes in with the same timing – 0’ 37" – but to totally different effect. It has both playfulness and grace and is seemingly unhurried. This is the one that gets my delighted vote.

Oddly enough, the most valid alternative to a very fast tempo would appear to be a pretty slow one. Once you’ve adjusted to Kempff’s leisurely pace you can appreciate the transparency of his part-playing and the translucency of his touch.

Variation 6

The second canon. For a change the slowest, most legato performance comes from Perahia. Far from sounding heavy this flowing, songlike performance gives us the time to hear what is happening. Sheppard, as in the first canon, essays a very staccato bass which comes to sound mannered, and points up the chromatic notes in the first half less then any of the others. He gives both repeats. Kempff (also with both repeats) is not merely swifter than Perahia but finds a different kind of rhythmic impetus, due also to the fact that he gives a sharp, bell-like accent to the first note of each canonic entry, creating a "hidden melody" (beginning G-A-B-C-D-E). This is, in a sense, "pianistic Bach" of the old school, a type of pianism which might seem more suited to Schumann. However, while it is true that neither the harpsichord nor the clavichord (nor even the first fortepianos which Bach possibly knew) could do this, it is also true that if you play a passage like this on the organ the "hidden melody" will come out willy-nilly, so the concept is not inherently foreign to Bach’s thought. Gould 1955 has some odd rubatos (very slightly echoed by Sheppard) which he mercifully abandoned in 1981 to give a flowing performance at a tempo similar to Kempff’s (and with the first repeat) but with a drier, more "baroque" timbre. Perahia, Kempff and Gould 1981 seem to me equally successful here, each in his different way.

Variation 7

"Al tempo di giga", says Bach, before loading it with ornaments which can easily clog the dance movement. The problem with an ornament on the first beat of most bars is that the musical accent can get shifted from the first beat itself onto the last note of the ornament, thereby compromising the even rhythmic movement. This risk is not wholly avoided by Sheppard. He is also the only one of the four to bring out strongly the theme in the left hand at bar 8, not something I feel strongly about either way.

The performance for me is Perahia’s. By making the first note of each dotted group staccato (why did no one think of this before?) he obtains a dance-like lilt which eludes all the others. Not that Kempff or Gould even seem to want to try. Though Kempff, by eliminating the ornaments (save, inconsistently, that in b. 28) might seem to have avoided one problem, he plays the piece as a gentle pastorale. So does Gould, even if in 1955 he had a certain elegance which could be thought of as a gigue in the French style. Except that, by calling it a "giga" in Italian rather than a "gigue", Bach would seem to have made his intentions clear. Kempff, as befits his generation, is evidently using an edition which kindly "corrects" Bach’s dissonance in b. 25.

Variation 8

A variety of approaches here. For Perahia the striding quavers are the thing and the semiquavers a gently murmuring background. However, in the repeats he varies the balance between the parts and finds some hidden melodies in the left hand. A strikingly imaginative interpretation.

For Kempff it is the semiquavers which count. At his calmer tempo they are gently melodic, with the quavers a very light and delicate staccato. These two contrasting views seem to me to be those that find the most in this variation. Sheppard is more straightforward and his light staccato touch has much vitality. Brilliant fingerwork is to be heard from Gould 1955; hearing the same interpretation slowed down in 1981 reveals it to be not especially insightful.

Variation 9

The third canon. Sheppard, Perahia and Kempff all agree that this is a flowing piece in the manner of a cappella choral writing. Sheppard, however, opts for staccato treatment of the semiquavers. Having commented that he hardly made the unexpected harmonies register in the first part of variation 6, I should say that the surprise F at bar 13 is pointed out in no uncertain manner.

Perahia and Kempff go in for a serenely flowing treatment, but Perahia is a shade sticky (the repeat of the first half moves a little more and sounds better; was this another take?) and it is Kempff (with both repeats) who persuades us that the are four beats in the bar not eight. Lovely, limpid playing.

Gould 1955 is clear but brittle; he had rethought his approach totally by 1981. Here his quavers are a delicate staccato; his mastery of line is such as to demonstrate that this need not militate against a horizontal, rather than vertical, approach. His staccato semiquavers seem much more natural in this context than do Sheppard’s. Though Kempff and Gould 1981 represent opposite extremes, these are the two I found most rewarding.

Variation 10

The "Fughetta". Sheppard manages to be sturdy and forthright without heaviness. His dynamic gradations in the second part help and his staccato quavers, which in some variations I find mannered, are an aid to buoyancy here.

Perahia is even more sturdy and forthright but, while at the outset I thought it super, in the end he does not entirely avoid heaviness and over-insistence.

Kempff concedes us at least the trill on the second note of the theme and also some springy, detached playing. He grades the dynamics and builds the variation up purposefully towards a full organ climax. Nowadays they tell us we shouldn’t do this sort of thing with Bach, but I found it very effective. The forceful and energetic Gould 1955 narrowly succeeds where Perahia narrowly failed (but since he doesn’t play the repeats the risk of over-insistence is smaller). Gould 1981 finds a variety of touches and dynamic gradations. He makes the first repeat and, having played the theme staccato first time round, now he plays it legato. He builds up to a resounding climax and concludes with a fuller chord than Bach’s written one.

This is another case where the two extremes – Gould 1981 and Kempff – prove the most rewarding.

Variation 11

Sheppard’s detached touch produces delightful results here – it has sparkle even at a fairly moderate tempo and the interplay of the hands is very clear. So is it with Perahia, who is more legato and rather faster, a more mellifluous interpretation. Either is preferable to Kempff who is also mellifluously fluent, but pedals whenever the semiquaver triplets revolve around a triad, with unsuitably impressionistic results. And perhaps also to Gould 1955, with its rather empty-headed clarity, but maybe not to Gould 1981 which in spite of being only a second shorter (0’ 54" against 0’ 55"; Sheppard takes 1’ 03") seems to have all the time it needs to savour the contrapuntal exchanges, and to combine mellifluousness with sparkle.

Variation 12

The fourth canon. For Sheppard the canonic writing in the two upper parts is almost an accompaniment to the forthrightly striding bass. Marvellous rhythmic buoyancy here. Perahia, on the other hand, lets the bass remain a harmonic support to the canonic parts which wrap around each other in a much gentler manner. However, if you like this, you will probably like Kempff even more (with both repeats), who brings a vocal quality to the flowing semiquavers and has the bass line tolling like a bell. I found this profoundly beautiful and can hardly believe that it is almost twice as slow as Perahia (4’ 03" compared with 2’ 17"). Gould 1955 is a burst of energy, just slightly tamed in 1981. For Gould as for Sheppard the striding basses carry the variation forward and the difference between Sheppard and Gould 1981 is basically one of timbre. I haven’t dwelt on this matter so far but Gould, as is well-known, had his pianos specially prepared for baroque music whereas Sheppard is evidently using a normal concert grand. So Gould has a hard-hitting timbre which seems a half-way stage between a piano and a harpsichord, while Sheppard, in the context of a basically mellower timbre, uses a staccato touch to obtain brilliance. You may have preferences; I find I enjoy Sheppard and Gould 1981 equally, as I do the totally different Kempff. A pattern which is emerging is that the most effective performance of each variation is that which risks most in one direction or another. The Goldbergs don’t seem to respond to the "middle way".

Variation 13

The florid song-like writing of this variation marks it out as different from any of the others up to this point. It is somewhat surprising to find Perahia applying a considerable degree of rubato, frequent de-synchronisation between the hands and some fussy bringing out of the inner parts. Kempff also essays an all-legato approach, with an easily flowing tempo and a simplicity of expression which makes even the most decorative passages sound sublimely natural. Although Kempff worked from an aesthetic viewpoint that in many places seems dated to us today, there are many moments, and this is a supreme one, when his profound musicianship provides illumination which time cannot so easily obscure.

The remaining performances adopt a more daintily staccato approach to the accompanying quavers. Sheppard and Gould 1955 are both attractive from their own standpoint. Gould 1981 has much of the right-hand melody staccato as well, but also a feeling for line which is the equal, in its totally different way, of Kempff’s. So here again, I respond most to the two performances which have the courage to take their particular viewpoint to its logical extreme.

Variation 14

How hard it is for the mordent on the bass G at the beginning of this variation not to sound like a scrunching wrong note. Perahia avoids this risk in the repeat, so what a pity he didn’t retake the beginning. A generally clear performance. Sheppard starts well enough, but when the music breaks into demi-semiquavers the rhythm is not at all clear and the effect is messy.

Kempff, by omitting the mordent, avoids one problem, only to create worse ones by some heavy pedalling. It is a good, vital example of his Beethoven style, except that he happens to be playing Bach.

Gould 1955 has clarity, but quite frankly only the "steady as she goes" Gould 1981 succeeds in making sense of this bizarre little capriccio.

Variation 15

As well as being the fifth canon, this variation is an important stage along the way in other respects. It is the first in the minor key, and it marks the numerical half-way point. The fact that the following variation is in the style of a "French Overture", and thus a prelude, certainly suggests that, aside from mere arithmetic, Bach wanted us to sit up and feel "this is part two beginning". Since the work was allegedly written for Bach’s patron Count Hermann Carl von Keyserlingk to while away his sleepless nights (he suffered from insomnia), no doubt it amused Bach to think that if by any chance he had nodded off during this gentle minor-key piece, the crashing start of the Overture would wake him up again.

Variation 15 is also one of only two which has a tempo marking – "andante". Contrary to romantic ideas, "andante" is the present participle of the Italian verb "andare" which means "to go". How often do we hear an "andante" which seems to mean "stopping", not "going"! Furthermore, in the baroque era it was not uncommon – in some of John Stanley’s Organ Voluntaries, for example – to give an indication such as "Andante allegro" or even "Andante vivace". These might be translated as "With a lively (or vivacious) movement". If we add the fact that this variation has two crotchets (fourth-notes) to the bar rather than four quavers (eighth-notes), the indication turns out to be not so much a request for a slow tempo as a warning against one. Sublimely (perhaps literally) unaware of this, Gould 1981 actually seems to be giving us eight semiquavers (sixteenth-notes) to the bar. With the first half repeated he takes longer than either Kempff or Perahia who give both repeats. Frequently staccato into the bargain, I found this, at one and the same time, to have a hypnotic fascination which reminded me of W. B. Yeats’s line about "peace comes dripping slow", and to be utterly interminable.

The other performances are fairly well agreed over tempo and style though Gould 1955 has a tendency to broaden out, by an emphasis on the individual semiquavers, which suggests that presages of the 1981 version were already forming in his mind. But it is not only a question of tempo, as I found when, having appreciated Kempff for his greater mobility, I discovered that he is actually slower than Perahia (4’ 38" against 4’ 19"). Over and above mechanical/temporal consideration, there is that intangible thing called "humanity" which shines through Kempff’s performance and places it before the similarly conceived versions of Sheppard and Perahia.

Variation 16

The "French Overture", with a typical first part in dotted rhythm and a lively fughetta to follow. Perahia unfolds the panoply of the full orchestra, but avoids heaviness through his lively double dotting. His fughetta is also robust; Sheppard has slightly less profile in the first part but his lightly-played fughetta is the best of the lot and I wish he had given us the repeat.

Kempff has no time for double-dotting but has a certain spring to his step even so. Surprisingly, Gould has no time for double-dotting either. His 1955 performance has a vitality which had turned to heaviness by 1981.

Variation 17

This is a variation where Sheppard’s light detached touch produces delightful results. He also draws attention to the voice leading with various accents and a constantly changing balance between the hands which most listeners will find helpful. If you don’t agree, then Perahia, at a faster tempo, has everything beautifully neat and as laid-back as a Czerny study. He is well aware, obviously, that the music is inherently far more interesting than a Czerny study and evidently feels it is best left to tell its own tale. Kempff vindicates a slower tempo with a light staccato touch and much voice-leading. It seems to me that everything he does here is done equally well done by Sheppard, whose fleetness of foot is an added attraction. Gould is swift and light both times (0’ 53" in 1955, 0’ 54" in 1981!) but seems to have more time to breathe in the later one. At 0’ 59" Sheppard has just that fraction more of space around the notes and remains my favourite.

Variation 18

The sixth canon finds all four pianists remarkably in agreement that it is to be played with a light staccato bass and with the upper canonic parts wrapping coolly around each other. Sheppard here achieves a tripping delicacy which just eludes the others. I found it captivating.

Variation 19

Here all five performances are totally different, and each is effective in its way.

Sheppard, in line with his general approach, is light and dance-like. He highlights details such as the C natural in b. 4 in a more specifically "pianistic" way than any of the others and he evidently has a particular affection for this variation since he plays the first repeat. Perahia is considerably slower, beginning with a flowing legato. However, on the repeat he makes the semiquavers staccato, and thereafter he alternates legato and staccato to make a remarkably detailed interpretation. Kempff is fairly swift, applying a gently flowing legato throughout. Once again that indefinable quality, humanity, shines through his performance. Gould 1955 has brittle staccato semiquavers which become legato towards the end. Outwardly not far different from Sheppard, he gives the variation an iron strength which Sheppard does not attempt. Gould 1981 is at his most provocative; very slow and very staccato, the effect is that of a musical box whose clockwork mechanism is almost run down. My initial reaction was that this was just ludicrous, but gradually I came to admit that it works on its own terms. As in 1955, the staccato subsides to a legato towards the end.

Variation 20

One of the most virtuosic of all. Apart from the hand-crossing, a recurrent problem in all those variations which Bach intended for a double-manual harpsichord, it breaks into triplet semiquavers at several points. So if you pitch in at what might seem a plausible tempo, you risk finding at bar 9 that you just can’t cope. One solution is to pitch in anyway and just trust that you’ve got superhuman fingers. And if you happen to be Glenn Gould in 1955, that’s precisely what you have got, but can a listener without superhuman ears follow it? Gould 1981 relented just a little (0’ 50" against 0’ 48"); somehow it sounds more relaxed than those mere two seconds would lead you to think, although his bashing out of the repeated quavers towards the end suggests a certain lack of sympathy with this variation..

Another solution is to start by finding a tempo in which the triplets flow easily and then let the rest fit in. This, I take it, is how Kempff arrived at his much slower tempo. He also indulges in much pedalling wherever the quavers revolve around a chord but apart from this, he only succeeds in demonstrating that a certain feeling of virtuosity stretched to the limits is an essential element of this variation; its various scales and arpeggios are not inherently especially interesting and at a relatively comfortable pace it sounds rather dull.

Sheppard and Perahia seem to me to have the happy medium; particularly happy in the case of Sheppard who exudes more sense of enjoyment and whose clearer voice-leading will surely help the listener to hear two parts chasing each other around rather than a flood of brilliant sound. But if you think that any voice-leading at all is an extraneous, romantic concept you will prefer Perahia’s more laid-back approach.

Variation 21

The seventh canon and the second minor-key variation. All four agree that it is to be played legato and in a leisurely tempo. It seems to me that Sheppard is the pianist who best captures its gentle flow and finds an easy discourse between the different voices. Regrettable, I feel, is Gould 1955’s decision to play it as loudly as possible and, worse still, he allows it to become static. Gould 1981 tempers this only in that he takes the first repeat and does allow the volume to drop at that point, providing a welcome oasis of sensitivity. Also regrettable, it seems to me, is Perahia’s decision to split the hands so often; the effect is lumpy. Furthermore, since I’ve often praised Perahia for his way of varying the music in the repeats – whether literally, with added ornaments, or more subtly with changes of expression, there are also variations, including this and also the preceding one, where he performs each part so identically as to reduce the repeat itself to a mere slavish observance. No complaints about Kempff; I just find that Sheppard flows that little bit more naturally.

Variation 22

I love the way Sheppard starts almost hesitantly, gathering strength as he proceeds. There is a fair consensus of opinion that this piece has rich, sonorous long notes and detached quavers – yes, even from Kempff. Perahia is a shade more serious but he builds it up well – remember that, with both repeats, he has twice the span to cover that Sheppard has and he certainly succeeds. Kempff is a little more forward moving and builds towards a fine climax. Gould 1955 is forceful from the start and perhaps less interesting. In 1981 he has changed little, except that he takes the first repeat and does introduce some light and shade at that point.

Variation 23

If anyone still doubts that Bach had a sense of humour, this is the variation to hear, and especially in Sheppard’s performance which has all the puckish, darting lightness and the chuckling wit of a Mendelssohnian scherzo. Kempff is pretty puckish, too, in spite of a slower tempo (and I won’t spoil the joke by telling you what he does with the final chord). However, I do think that speed is an essential part of the bag of tricks, and this is where Kempff falls short of Sheppard.

Perahia and Gould 1955 take their delights a mite more seriously, not that this prevents Bach from speaking for himself. Gould 1981 is scarcely any slower (0’ 58" against 0’ 54"; Sheppard takes 1’ 01") but he is lighter on his toes and this is my favourite after Sheppard.

Variation 24

The eighth canon. It is a curious feature of this work that often in the "straightforward" variations the voices chase each other around like canons while the canonic variations proper frequently disguise themselves. In this case the effect can seem that of a sublimely simple pastorale. If you agree with this interpretation, you will want the most beautiful, serene and timeless (but still flowing) performance of all, and Kempff will be your man. If you basically agree, but want the music to press forward a little more, then you may like to try Perahia. He himself describes this variation in his notes as "a calming, pastoral canon", though for me the performance which illustrates his thesis is not his own but Kempff’s. If you think the whole pastoral idea too romantic, you may enjoy Sheppard, who keeps it lighter and adds some dainty staccato quavers. Or Gould 1955 who treats it as a robust gigue. As so often, though, it is the most extreme solution of all which turns out to be the most satisfying alternative to the "traditional" Kempff. Gould 1981 is outrageously provocative, with every note a light staccatissimo so short you’d think he was the proverbial cat and the keyboard a hot tin roof (though he takes the first repeat and allows himself a degree more legato there). I found it absolutely enchanting.

Variation 25

The third (and last) variation in the minor key, the second (of two) with a tempo indication – Adagio. This is the famous "black pearl" (Wanda Landowska’s words) and, oh, the shame of it, this sublime inspiration can seem amorphous and interminable. What’s the secret? I am reminded of a comment by Parry regarding another of Bach’s profoundest slow pieces – the E flat minor prelude from Book 1 of the "48": "The effect of coherence is attained by these chords being systematically grouped in threes … which serves as a unifying principle". In this case the unifying principle is the left-hand rhythm which always consists of three upbeat quavers and another two on the downbeat. At moments of heightened tension the two left-hand voices, instead of playing this rhythm together, play it in a kind of pseudo-canon. While this left-hand rhythm must never dominate, it must be heard as the backdrop against which and around which the right-hand melody flowers and expands. I’m not sure that we do hear this in Sheppard’s performance, which rather loses its way. Perahia manages better, and succeeds in holding the attention over his very long span (7’ 24" with both repeats). Kempff helps himself by choosing a more flowing tempo and nowhere in the entire performance is his profoundly simple yet deeply felt musicianship more apparent. The melodic line is apparently freed from the accompaniment which nonetheless proceeds with complete rhythmic rigour. At 4’ 52" with the first repeat he is only slightly longer than Sheppard’s repeatless performance (4’ 15") yet there is no sense of hurry. Gould’s two performances are also without repeats, yet last considerably longer than Kempff’s (6’ 28" in 1955, 6’ 03" in 1981). With Gould the unit of measure appears to be the semiquaver not the quaver. He does, however, succeed in keeping up a gently rocking movement which carries the music forward in spite of the very slow tempo. The right-hand melody is a little freer in expression in 1981; in 1955 the music risks stalling altogether towards the end.

Variation 26

This is yet another of the variations where Sheppard’s light fingerwork produces delightful results. Unusually for him, he plays both repeats. Perahia at virtually the same tempo (1’ 57" against Sheppard’s 1’ 58") has a more Handelian assertiveness. Kempff is like a slowed down version of Sheppard; the music bubbles along very nicely but lacks the exhilaration that a faster tempo can bring. Gould is faster still (0’ 52" both times, without repeats) but has more light and shade in 1981.

Variation 27

The last of the canons and another one that sounds remarkably little like a canon. All four pianists agree that this has the character of a gigue (which sounds like a good reason for not interpreting Variation 24 as a gigue). Sheppard’s staccato touch seems fussy here, especially in the left hand. Musing as to why this touch should sometimes prove delightful but at other times fussy, I think this is a pianistic touch which is effective up to a certain velocity (and that velocity is slower in the plummier middle-lower registers); beyond that velocity a detached, but not actually staccato, touch is more effective, and this is what Perahia and Gould give us. But to be fair, the actual speed at which staccato ceases to be effective is a rather subjective matter; it depends on the acoustics of the room and on the particular piano used, but also on the ears of the listener, so not everybody will agree with me anyway that Sheppard sounds fussy in this variation.

Having started by applauding the vitality and energy of Perahia’s touch and general presentation, I did find it rather unimaginative of him to play the whole variation (with repeats) at an unremitting forte. Kempff is much lighter even if his touch is more legato, and builds the variation up gradually (with both repeats). I could also have sworn he was faster than Perahia but the stopwatch says otherwise: Perahia 1’ 39", Kempff 1’ 43".

Gould 1955 is brittle and a little aggressive. Gould 1981 makes the first repeat and, as so often, takes this opportunity to find the light and shade that was missing before. He also plays the repeat with a more legato touch.

A small point; the end of this variation has a very unfinished feel and cries out to lead into the following variation. I find it strange that only Gould 1981 actually does this (as I shall discuss more fully in my final comments, he had given a lot of thought to the question of continuity in the years that separated the two interpretations). In all other cases the music apparently stops in mid-flow and the gap before the next variation is considerable. The work of the producer or of the pianist? Well, in the case of Sheppard’s live performance, this at least must surely document what he actually did.

Variation 28

Having got the canons – the "serious business" – out of the way, Bach seems to want to kick the traces, since he now writes two variations which are sheer virtuoso showing-off, leading to the Quodlibet on folk-themes, which must have seemed both unexpected and funny in the days when everybody knew the tunes he used, and finally a reprise of the original aria.

Played on the piano Variation 28, with its recurrent inner-voice trills, seems to have virtually invented the late Beethoven style, and it is Perahia who succeeds best in capturing this. Though not at all slow he manages an unruffled serenity – Sheppard is rather effortful here. Logically this should be Kempff’s preserve, but perhaps that is the problem since he seems to want to transform into late Beethoven, with the help of the pedal, even those passages of two-part writing which sound like normal Bach.

Of course, the late Beethoven analogy is only an a posteriori construction anyway and Gould has nothing to do with it, opting for much drier harpsichord sonorities. The lighter-touched 1981 version is particularly successful; this and the Perahia come out tops here.

Variation 29

In this second virtuosic variation the effect on the harpsichord of all the thick chordal writing alternating with triplets streaming down from top to bottom of the keyboard would be above all one of sheer noise. Mindful of this Gould 1955 goes at it hammer and tongs, rather like a mad genius bashing everything in sight. In 1981 he evidently repented of his youthful sins and offered a more controlled reading. I remain unrepentantly convinced that this variation is supposed to sound like a mad genius and therefore Gould 1955 realises it like no other.

Sheppard also goes for a very pianistic reading, to the extent of adding some octave doublings in the bass when repeating the first part. There may be nothing wrong with this in principle (some readers might not even concede this much) but it does come as a jolt considering that Sheppard has done nothing similar up till now. It is also a pity the piano bass notes were not better in tune.

Perahia is a little slower than Sheppard (let alone Gould 1955) and basically just sees that everything is vital and clear. Kempff is on the same lines but slower and with a certain affectionate sense of humour. I was reminded of the mock drama of Schumann’s hobby-horse in Kinderszenen.

Variation 30

The Quodlibet. If we go by its name this should be a "free for all", the different voices belting out their folk-song fragments like salesmen advertising their wares in the market-place. In the hands of Gould in 1955 this is pretty well how it sounds. Perahia, though more refined, also seems to subscribe to the theory of "the louder the better". Gould 1981, with the first repeat, is gentler than before, and Kempff plays it with a flowing legato and plenty of dynamic variation. Sheppard adopts a wholly pianistic approach. He makes both repeats and uses them to explore different voice leadings and varied dynamics. He has the second half fade away on its repeat, setting the stage for the return of the theme itself. A possibly anachronistic, but warm-hearted and imaginative solution.

Aria da capo

No one in their right mind is going to make this repeat of the Aria, after all that has come about in the meantime, a carbon copy of the original. All agree that it is to be played more slowly, with a certain leave-taking air. If the timings seem to suggest otherwise this is because those who included one or both repeats the first time round omit them now. The original edition simply stated "Aria da capo" without actually writing it out again. A note to the Henle Edition states that there is "no certainty with regard to the practice [of making repeats in da capos] prevailing in Bach’s day".

All performers manage to intensify the expression of their original performance. The only dubious offering seems to me to be Gould 1981 where an already woefully slow performance is made still more interminable.


So we’ve compared the performances variation by variation, and it would seem that all five come up tops in some variations, and also each has its share of "bottom choices". Readers might also have noticed that Perahia and Gould 1955 have a smaller share of "tops" than the others. But now how about putting the score away and listening to each one straight through. Does the overall impression confirm the sum of the details?

Yes and no. Sheppard’s tendency towards staccato where others prefer just very clear articulation falls into its place over the larger span; you accept it as a norm when you are not continually comparing it with performances which use other means. And also Kempff’s tendency to use the pedal – and I must emphasise that he never allows this to damage harmonic clarity – is less off-putting when the ear is not asked continually to adjust and re-adjust. On the other hand the close-up, airless nature of Gould’s chosen piano preparation and recording characteristics tends to become more, not less, overbearing when you’re chained to it over a long period of time. Having declared that the Goldbergs do not thrive on half-measures, I see I’m going to have to modify this a bit.

Listening to Kempff straight through I appreciated the beauty of his sound, the clarity even in swimming-pedalled textures – every note seems like a pebble thrown into a pool of water with the light glistening upon it – and the essential humanity of his art. I also found that the sum of his rhythmic flexibility and his failure to contrast the gentler variations with virtuoso ones which really pitch in and send the sparks flying made the total experience more of a drag than I would have expected. Perhaps the truth is that the world has moved on since then and while certain aspects of Kempff’s treatment of Bach are timeless, it is difficult now to relate to the performance as a whole.

Not least, of course, because of the impact which Gould’s 1955 recording made. Maybe Gould had to play it that way at that particular time. Here, too, the world has moved on. The overall effect is even more breathless and aggressive than the single parts. And, as we hear in the interview on the third disc, Gould himself distanced himself from it.

Despite the woefully slow presentation of the Aria in the 1981 performance, a lot of the individual variations came out tops in the separate comparisons, so what about the total effect? Well, I’m afraid that even if the aggressive aspects are tempered, they still add up as the work proceeds. And certain provocative performances, like the dead-slow no. 15 and the grindingly static no. 25 (but listen to what Gould has to say in the interview about his reasons for playing them like this) weigh on the forward flow like a ton of bricks. It’s unremittingly modern and I suppose that, just as there are people who really like living in open-plan houses adorned with steel girders and plastic upholstery, there will be people who relate to this. The trouble is you never meet those people, unless you’re a photographer for an architectural magazine, and I suspect they have cosy little hidy-holes somewhere else that they actually live in. And, while their show-houses will have a CD of the Gould Goldbergs strategically placed on the coffee-table, their secret pied-à-terres will harbour an old LP-player with Kempff not too far away.

Fortunately it doesn’t have to be "either/or". First of all, the sheer lightness of Sheppard’s touch makes his a supremely listener-friendly version. Here is a Bach who sings and dances, and even the more questionable variations seem to fall into place. This is a version with which I can stay the whole journey.

And so I can, rather to my surprise, with Perahia. I had half-expected that his rigour (with all repeats) would prove massively impressive, but there is something more, and I think it comes down to rhythm. Each variation has its own rhythmic life and somehow each new variation seems to take over the pulse from the last. I don’t know whether Perahia has made a conscious attempt to relate the variations to some overall underlying pulse. Gould reveals, in the most interesting part of his interview, that in 1981 he did aim to do this. But he also states that it is best for the performer to arrive at these tempo relationships subliminally. So whether Perahia – and also Sheppard – attained their very clear sense of continuity through occult arithmetical calculations or simply by playing the work through hundreds and hundreds of times until they started to "feel" how each variation grows out of the latter really makes no difference; the end result is that their performances cohere and add up to more than the sum of their parts. Also, I should add that Perahia’s non-interventionist approach, while it might mean that many single variations are shaped more strikingly elsewhere, means that nothing gets in the way of the cumulative impact.

All this is very heartening. I have often seemed to side with those critics for whom the only truly great pianists are long dead. However, Bach is a composer so inherently timeless that performances of him tend to relate to their own day, and then fall by the wayside. Certainly, the way of Kempff’s generation still has much to offer us, maybe selectively rather than as a whole. For a while, it seemed that Gould’s way was the one for our times, but time has moved on again. So perhaps it is not surprising that the two performances I find I can relate to overall are by two pianists still in the full flower of their careers: Sheppard, who is a little more friendly and communicative, and Perahia with his admirable but not oppressive rigour.

The addition to the Sheppard disc of the 5th Partita, by the way, proves to be neither here nor there. Apart from some uneven semiquavers near the beginning which betray the live origin, this is a more "interpreted" and more heavily pedalled Bach than we hear in the Goldbergs. Furthermore, his pedalling, even when light, is brought to our attention by a clanking of the mechanism. There was no actual need for a filler and I rather wish this performance had been left on one side.

But I wouldn’t let that influence me. It only remains to add that, since all these performances have imperishable insights, if you were to get all five and compare them closely, as I did, far from suffering a surfeit you’re quite likely to want to go out and get five more.

Christopher Howell


Return to Index