How many pianists are there at any given moment in time who are capable of playing any piece in the vast repertoire of music written for the piano?  Of course, no one plays everything.  It takes time to prepare a piece for performance and not every composer or every work will be of interest to any particular musician.  But how many concert pianists who forego learning, say, Beethoven’s “Hammerklavier” Sonata, do so because the piece doesn’t speak to them rather than because they find it impossible to play?  I would guess that the number of such individuals is in the low triple digits, perhaps 250 at any moment in time.  That’s a pure guess and it may be high.

A surprisingly large number of those gifted pianists have turned to Bach’s Goldberg Variations in recent decades.  Any number of gifted harpsichordists have done the same.  I have read that there are 193 recordings of the work.  That article was written a few years ago; some recent recordings were not included.  There may be 200 recordings by now.  I am giving serious consideration to a project of listening to all of them.  I have made a start.

On occasion, I will obsess over a piece of music, listening to it repeatedly until the desire to hear it dies from overfeeding after a day or two.  I have been listening to Bach’s Goldberg Variations (BWV 988)[1] and virtually nothing else for over two months, so in this case obsession has graduated into mania.  I thought I would summarize what I have learned about the Goldberg Variations so far.  I may continue exploring the piece, so I reserve the right to supplement this note with additional comments.

Except for the comments on individual recordings, none of what follows will be news to anyone familiar with the work, but perhaps the phrasing will be different.  This is a long post on a modestly obscure topic.  I hope anyone reading it finds it interesting.  The length of the post is a measure of the intensity of the therapy required to move me to a different piece of music.

Form of the Work

The work begins with a melody, titled an “Aria”, followed by thirty variations.  Bach concludes the work by repeating the Aria.  Thus, the piece consists of 32 individual compositions.  The Aria itself is 32 bars long, which is not a coincidence.

The Aria is a long, slow, highly decorated, wandering affair.  The 30 variations display a staggering range of styles, emotional content, and technical challenges.  They include dance movements, fughettas, canons, toccatas, as well as others that don’t fall into any particular classification.

Contrasts between pieces, or between different performances of the same piece, can be startling.  Variation 7 is a pleasant dance movement (a “gigue”) followed by a piece that, as played by some musicians, sounds like we have entered the Bach Pump House where rows of perfectly tuned machines are running at full speed.  Yet there is one performer (Gustav Leonhardt) who can make the same piece sound almost like a lullaby.  Variation 13 is one of the most beautiful pieces of music I have ever heard, but some performers seem intent on draining any beauty out of it.  The next variation, number 14, is a drunken riot where the performer’s fingers tumble up and down the keyboard.  Number 25 is the longest and slowest of all the variations (nicknamed the “Black Pearl” by Wanda Landowska) and is followed by four pieces calling for extreme dexterity and virtuosity.

After the 30^th variation, the Aria returns.  One of the many oddities of the piece is that, when the Aria makes its second appearance, we realize (as the pianist David Jalbert points out) that the melody has disappeared for more than an hour (much less if you listen to a Glenn Gould performance of the work).  The Goldberg Variations were not built by varying a theme, the way the “Theme and Variations” compositions of Mozart, Beethoven, Schubert, Schumann, or Brahms are constructed.  Rather, the variations are built on the Aria’s bass line and the harmonies it implies.

That harmonic structure is not complicated.  After presenting the first part of the Aria in eight bars of G Major, Bach moves the next eight bars into the dominant, D Major.  A repeat sends us back to the beginning to hear those sixteen bars again.  Some performers ignore the repeat.  The second half of the Aria gives us eight bars in the key of E Minor (the relative minor, that is, the minor key with the same signature as G Major), after which the Aria finds its way back to G Major and closes with a few bars of sixteenth notes, the fastest-moving part of the Aria.  Again, the performer is directed to repeat the second sixteen bars, although many don’t take the repeat.

Of the 30 variations, 27 follow this harmonic pattern.  The three exceptions, numbers 15, 21, and 25, are written in G minor.  Jeremy Denk calls these “three oases of sadness in a desert of joy”.  All of these give us eight bars in G minor followed by eight bars in D minor.  Two of them, numbers 15 and 21, begin the second half with eight bars in E-flat major, to give us a little ray of sunshine through the minor key gloom, before returning to G minor to take us home.  The third of the minor key variations, number 25, the “Black Pearl”, stays in minor keys all the way through and ends on a note of desolation.

Bach was a happy, optimistic person, so it is not surprising that the overwhelming tone of the piece is one of joy, playfulness, and spontaneity.

Some History

The piece was published in (or about) 1742, when Bach would have been 56 or 57.  Bach did not have a lot of his music published, so the publication of the Goldbergs indicates the importance he attached to the work.  The Variations were composed sometime between 1738 and 1741.  Bach’s first biographer, writing in the first decade of the nineteenth century, explained how the piece came to be written.  The story has become the stuff of legends.  The Russian ambassador to the court of Dresden, a man named Hermann Carl, Count von Keyserlingk, suffered from insomnia.  His house musician, one Johann Gottlieb Goldberg, was a former student of Bach.  The Count was a music lover and a friend of Bach.  He commissioned Bach to write some pieces that Goldberg could play at night when the Count was having difficulty getting to sleep.  The piece we call the Goldberg Variations was the result of this commission.  In payment, the Count gave Bach a golden goblet filled with 100 gold coins (“louis d’or”).

Scholars now doubt the truth of this tale, but it was accepted as late as 25 years ago.  The story is repeated uncritically in the notes to recordings from 1980 (Trevor Pinnock, harpsichord) and 1988 (Rosalyn Tureck, piano).  Both sets of notes were written by the respective performers, each of whom is a reliable Bach scholar.  There are numerous reasons to doubt this romantic tale.  First, the story appears in a biography written in 1802, more than sixty years after the piece was composed.  The story was told by two of Bach’s sons, neither of whom was present in the Bach household when these events are said to have occurred.  Second, the title page of the work does not contain a dedication to the Count, which would have been customary then and now to honor the patron of a work.  Third, no one ever mentioned seeing the gold goblet and it was not listed among the effects in the Bach estate when he died in 1750.  Fourth, the estimable Herr Goldberg would have been about ten years old when Bach began work on the Variations.  He would have been a mere Kindsklavierspieler[2]  It is not likely that he was in the employ of an ambassador at such a young age.  It would be interesting to know if anyone referred to this magnificent piece of music as the “Goldberg Variations” prior to the publication of the story in 1802.  Finally, if you were looking for music to help you sleep, there is little in the Variations that would help.

The truth is probably that at some point in the 1740s, young Goldberg, who had become the house musician for Count Keyserlingk, asked Bach, his former teacher, for some music to keep his patron entertained.  Bach drew a printed copy of the Variations out of inventory and gave it to his former pupil, someone, Bach knew, who could actually perform the piece.  The 1802 version is a story that is too good to check, but the likelihood is that things didn’t happen that way.

As to the Aria itself, it used to be an accepted fact that it was written no later than 1725 and possibly not by Bach.  The reason for dating the Aria to such an early point in time is that it was found in one of the notebooks of Anna Magdalena Bach (J.S.’s second wife), a notebook that bears the date 1725.  It appears to have been copied into the notebook by Anna Magdalena.  The original notebook survives and investigators have determined that the Aria was copied into the notebook in the late 1730s.  Apparently, a couple of pages were left blank when the notebook was first used in 1725 and those pages were available to receive the Aria when it was copied around the time the Goldbergs were composed.  Chemical analysis of the ink seems to establish the later date to the satisfaction of the leading scholars in the field.

The Larger Structure of the Work

The work has a mathematical structure, or rather several overlapping mathematical structures.  As I mentioned, the Aria is 32 bars long, while the entire composition contains 32 individual pieces (the Aria, 30 variations, and the return of the Aria).  The first 16 bars of the Aria have a different character from the second 16.  The first 16 have a lighter mood than the second.  The first half is written in major keys, starting in G major, while the second half starts off in a minor key before using a series of sixteenth notes to take us back to G major.

Variation 16, labeled an “Overture” by Bach, sitting at the half-way point of the piece, has two distinctly different halves.  The opening half with block chords and dotted rhythms, can be thought of as summing up the first half of the piece.  The second half of Variation 16, a lighter, quicker, fugue-like affair, points us to the second half that still lies before us.  Murray Perahia thinks there is more to it than that.  He argues that if you divide the 32 compositions into four groups of eight, just as the Aria has four distinct eight-bar sections, each group in its mood and tone corresponds to the four respective elements of the Aria.  Some parts of the work fit this pattern more closely than others.  For example, the first eight pieces (Aria and Variations 1 through 7), can be likened to a Bach dance suite, ending in a “gigue,” a practice that Bach often followed.  (I would point out that Variation 7 does not sound like a piece to conclude with, but that’s a subjective impression.)

The last eight pieces can be mapped more comfortably onto the last eight bars of the Aria.  Measure 25 of the Aria is a pivotal moment, when the melody slowly moves from E minor and starts the meander toward G major.  Variation 25 is a long, slow, highly chromatic piece that performs a similar function for the piece as a whole.  The next four variations are quick moving numbers, the last two of which approach the limits of what can be sounded on a keyboard instrument.  These four correspond roughly to Measures 26 through 29 of the Aria, when a series of sixteenth notes carry us close to the Aria’s conclusion.  However, I think one can push this kind of mapping too far.  I think the best that can be said is that there is a rough correspondence between the form of the Aria and the structure of the 30 variations.

Trevor Pinnock notes another structural element.  There are two particularly monumental fugue-like variations, numbers 10 and 22, that each stand exactly six numbers away from the central Overture, acting as pillars holding up the rest of the structure.  His insight suggests a three-part division of the work, possible further evidence of the attention that Bach lavished on the piece.

One mathematical pattern in the piece is much less speculative because the variations in question are labeled by the composer to point it out.  Every third variation is a canon, a piece where two different voices play the same theme, starting at different times.  (“Row, row, row your boat” is an example.)  All of the canons are in two voices and all but the last have a free, non-canonic, third voice in the bass.  The canons have a particular pattern.  In the first canon, Variation 3, the leading voice and the following voice start on the same note; the canon is at the “unison”.  In the second canon, Variation 6, the following voice starts one note higher than the leading voice, so the canon is at the “second.”  This continues right up to Variation 27, which is a canon at the ninth.  Variation 27 differs from the other canons in that there is no free voice in the bass.  The two canonic lines roll up and down the keyboard nine notes apart.  Ninths don’t sound all that pleasant together, so it is a remarkable achievement to send the two voices all over the keyboard with no harmonic cover or diversion of any kind.

There are a few Bach tricks in the canons.  A couple of them are in contrary motion – the following voice does not mimic the leading voice in the usual way.  Instead, when the lead voice, say, goes up a third, the following voice goes down a third.  The rhythmic pattern is the same in both voices but every motion by the lead voice is met by the opposite motion from the following voice.  In a couple of variations, the lead voice and following voice change places here and there.

None of this mathematical wizardry is part of anyone’s conscious enjoyment of the music.  I doubt that anyone who hadn’t looked at the score or read a mathematical analysis of the music has ever listened to the piece and thought, “Ah, yes, the canons keep rising, one tone every three variations.”  It’s the kind of thing you would discover only if you were reading the score.  However, the continued rising of the tones separating the voices is undoubtedly felt by the listener as the piece progresses, even if we can’t say what we are feeling or why.  We can feel the structure, as we feel the structure of a great architectural masterpiece, even if we do not understand the elements that produce it, or even know what they are.

In addition, because we know that this kind of in-depth planning went into the construction of the music, we know that the composer saw the work as something more than a gold cup of coins’ worth of sleeping pills.  He didn’t build the piece on such a complex foundation so that we would have the pleasure of decoding a puzzle over the course of centuries.  He produced the Variations with patterns of such complexity in order to enhance the enjoyment of his music by those who would play it and hear it, to “refresh their spirits” as he put it.

When we come to the final variation, number 30, the previous pattern has led us to expect a canon at the tenth.  Instead, we get a piece labeled “Quodlibet” – “What pleases you”.  It was the custom at Bach family gatherings for the guests to entertain themselves after a heavy meal by blending disparate melodies together into a workable counterpoint.  These efforts must have produced some remarkable impromptu compositions while the Madeira and Schnapps were circulating round the table.  Bach had 19 children, some of whom became important composers in their own right.  In addition, there were older and collateral members of the Bach family who had been composers and court musicians for generations.  Doubtless many family friends were musical.  So when this group decided to blend a couple of melodies, they got blended.  And, if somehow the group got stuck due to a particularly obdurate passing tone, they had history’s greatest composer of counterpoint available to bail them out.  It would be like having Richard Sherman playing defense on your touch football team.

In the same spirit of fun and good fellowship, Bach fitted two popular songs of the day[3] to the harmony of the Aria and offers the result as the final transformation of the bass pattern that inspired the previous 29 variations.  The counterpoint is so seamless that it is not always obvious that more than one melody is in play.  The composer is having some fun, digging an elbow into the sides of the performer and the listener, and invites us to do the same back.  A good performance of the Quodlibet leaves us smiling, and it’s easy to imagine the composer smiling back at us across nearly three centuries.

After all that, the Aria returns, just as quiet and contemplative as before.  Even though the notes are identical to the ones that opened the piece, a good performance leaves us feeling that the Aria has gone through the fire and has been transformed.

Thoughts on Recordings

In general, my favorite recording is the one I’m listening to at the moment.  There are only three that I would label “Avoid” and one of those is very much a specialty item.  Here are some impressions of the recordings in roughly chronological order of recording date.

Rudolf Serkin 1928.  Before high fidelity recording was ever thought of, the Aeolian Company was using a sophisticated method to record piano performances using piano rolls.  The company developed a “Duo-Art” reproducing piano and persuaded Steinway to manufacture a few pianos that were fitted with the Duo-Art mechanism.  From 1916 to the late 1920s, many famous pianists of the early twentieth century recorded this way.  During the stereo era, it was possible to play these rolls through the old Steinways and record the result onto then-modern tape.  Everest records produced several of these recordings.  I have an old LP of Jan Paderewski and another of Vladimir de Pachmann[4], both performing a variety of pieces by Chopin.  Rudolf Serkin used this method to lay down a record of a Goldberg performance in 1928.  It’s a survivor in the twentieth century of what we understand performance practice of Bach’s works to have been in the nineteenth century.  The idea was to use then-modern instruments to create a massive wall of sound, through which the searing genius of Bach would somehow emerge.  The 1928 recording fits right in.  There is little in the way of line or definition.  The notes run together in a way that makes the music unrecognizable.  Serkin omits Variations 6 through 10.  I would not plan to return to this recording (which can be heard on YouTube).

Wanda Landowska 1933.  Landowska trained as a pianist and then decided around the turn of twentieth century that the harpsichord was her instrument.  The critic/composer Virgil Thompson said that “Landowska plays the harpsichord better than anyone else plays anything.”  She was a great collector of old keyboard instruments but decided that her needs could be met only by a modern instrument built to her specifications.  She hired the venerable Paris firm of Pleyel, the firm that built Chopin’s pianos, to construct a harpsichord along the lines of a Sherman tank.  She had the strings laid on a steel harp, just as they would be for a modern grand piano.  The bass register included strings that were sixteen feet long.  One problem with the massive construction of the central core of the instrument is that the floor and walls of the instrument had to be equally massive to support the weight of the steel.  That meant that the instrument’s sound would be affected by the thickness of the wood.  To overcome this problem, she had Pleyel install steel plectrums (the part that plucks the string of the harpsichord).  This alteration helped, but evidently the sound engineers decided during recording sessions that it was necessary to bring the microphones closer to the instrument than would have been ideal.  The result is an instrument that has a unique sound, not necessarily the sound that Bach would have recognized as native to the harpsichord.  It reminds me somewhat of an acoustic steel guitar.

Yet, when the great cellist Pablo Casals suggested to Landowska that her performances might not be authentic, she drew herself up to her full height of four foot eight inches and said, “You play Bach your way, and I’ll play him his way.”

In the early decades of the twentieth century, she began to present the work to a public that was largely unfamiliar with it.  Evidently, she was quite a show-woman and people flocked to her concert performances of the piece to watch (she said) the dramatic crossing of hands necessary to bring off the effects demanded by Bach.  In 1933, she made the first ever recording of the full work, also available on YouTube.  Her intense commitment to the music is evident from the first note, and one quickly gets used to the fact that the sounds she produces from her harpsichord have to be marked with an asterisk.  She uses rubato where few modern performers would and she adds romantic touches that are, one feels, her own invention.  But when you listen to her play this music, none of that matters.  Her intense commitment, virtuosity, and obvious joy in sharing her discoveries ultimately carry the day.  This would not be anyone’s first choice of a recording of this work, but it is well worth listening to for its historic importance and for the power of the performer.

Glenn Gould 1955 and 1981.  Landowska opened the door to the Goldbergs in 1933 and Gould kicked the door down in 1955.  A few recordings were made between those dates, but the 1955 recording is a landmark.  Gould had given his first U.S. concert in New York in 1955 and the head of CBS Masterworks immediately signed him.  For some reason, they decided that Gould’s first recording would be the Goldberg Variations, a piece that was not a household term in 1955.

Gould is often described as an eccentric, but if we give him that word we have to find another weaker term to describe the people who merely cover their windows with aluminum foil or who stand on street corners talking into cell phones that have not been activated.  Once his recording career provided him with sufficient security that he could give full vent to his eccentricities, they began to pile up and became the stuff of legends.  To name just a few: By the early 1960s he stopped giving concerts.  All of his public performances were via studio recordings.  On top of that, he remained secluded in an apartment in a Toronto hotel.  When he went out, he wore an overcoat, scarf and heavy gloves, even in summer.  When he practiced the piano, he played numerous radios, each tuned to a different station.  When cable television came in, he replaced the radios with TVs, each tuned to a different cable station.  He would arrive at a recording studio with several towels because he soaked his hands and forearms in hot water for about twenty minutes prior to a performance in order to loosen his muscles and tendons.  Observers reported that his skin turned red from the heat.  He was notorious for humming along when he recorded.  The best efforts of the engineers could not remove all of it and his hums can be heard still.

That is just a sampling of what Gould would later become.  In 1955, all of that was still in the future but the seeds were there.  You can hear it in the performance, and not just because he hummed.  His tempos are lightning quick.  He takes no repeats, so that we have barely begun to digest Bach’s latest idea in time to hear Mr. Gould race to the final bars of the variation, take a breath and start the next one.

Throughout his career, Gould made negative comments about famous composers or notable works.  He at one point said that Bach was a better mathematician than he was a composer.  He wrote about how Mozart became a “bad” composer, and indicated that Beethoven’s Appassionata Sonata was, as far as he was concerned, among the least important of that composer’s works.  These comments and others like them seem to be more of an adolescent desire to shock the grown-ups than sincerely held opinions.  That trend is already evident in the liner notes to the 1955 recording, which have a decidedly snarky tone.

Here’s a small conundrum.  Gould plays the piece faster than anyone on record that I am aware of.  His timing for the 1955 recording was just over 38 minutes.  But he takes no repeats.  If he had taken all of the repeats, his timing would have been about twice as long, or one hour, sixteen minutes.  That makes his pace slightly slower than Rosalyn Tureck, the “High Priestess of Bach,” who is generally regarded as rather slow and deliberative in this music.  Her 1988 recording of the Goldberg Variations runs one hour fifteen minutes.  So, Gould the speed demon plays the work in slightly more than half the time than a slow performer requires to perform all the repeats.

The answer to the riddle is that Gould takes everything at blazing speed except for the three minor key variations, which he plays at a very slow pace.  To take only the Black Pearl as an example, and not to belabor the point overmuch, Gould takes about six and a half minutes on Variation 25 without taking a repeat.  Tureck, noted for her deliberate pace, takes 4:11, omitting the repeats in that variation.  Angela Hewitt, who takes a very measured hour and nineteen minutes for the entire work, devotes 7:54 to Variation 25, taking both repeats.

When Gould came to record the Goldberg Variations a second time, in 1981, he said that he had made the Black Pearl sound like a Chopin Nocturne.  Even so, he played the variation the second time in 6:03, again without repeats.  The 1981 recording is marked by slower tempos and a gentler handling of the piece.  One of the raging debates in this music seems to be whether one prefers the 1955 to the 1981 performance.  I know that there are many who revere Gould as a great musical genius and an important interpreter of the music of Bach in general and this piece in particular.  There are enough recordings of this work for every taste to be indulged.  For myself, I believe there are so many more compelling performances that I don’t think I will return very often to either of the Goulds.

I should note a few other points about the 1981 recording.  Gould recorded it in April and May 1981.  He did an audio interview on the work in August of that year.  In late September he suffered a stroke and died on October 4, 1981.  The 1981 recording was released after his death.  It was the last recording made at CBS Masterworks studio in New York.  Gould made one other recording, in September 1981 for Decca, but it is nevertheless remarkable that his controversial recording career began and (almost) ended with the Goldbergs.

In an interview during his prime recording years, Gould had indicated that his only pianistic idol was Rosalyn Tureck.  This was surprising to many, because their styles are so different.  When this comment was passed on to Ms. Tureck when she was interviewed some time later, she appeared not to be terribly pleased to have been identified as the mentor of someone whose approach to Bach she disagreed with.  When he died of a stroke, she is supposed to have said that it was no wonder given how fast he played.  When she herself died a number of years later, a Gould loyalist posted the comment that the world-wide average speed of Bach performances would now increase by ten percent.

The 1981 recording was made at the very dawn of the age of digital recording.  The theory behind the technology appeared to be sound, but the equipment in use at the time sometimes produced harsh results.  Engineers were enamored of their newfound ability to produce a broad dynamic range and sometimes forgot that the objective was to reproduce music, not “perfect sound forever,” which was the motto adopted at the time by the pro-digital forces.  The Gould 1981 digital recording is perhaps not the highest example of the recording engineer’s art.

Then, about two decades later, someone discovered – or remembered – that it had been the custom in the early days of digital recording to run an analog tape as a back-up, just in case.  In the first years of the new century, the producer and sound engineer dug out the 1981 analog master tapes and compared them note for note against the digital recording.  They decided that the analog recording was superior.  In 2002, Sony, who had purchased the CBS Masterworks library, released a three disk set consisting of the 1955 recording, the 1981 recording remastered using the original analog tapes, and a third disk containing interviews with Gould from 1981 and out-takes from the 1955 recording.  The three are released together in a set titled “Glenn Gould – A State of Wonder”.  It is well worth owning for anyone who is an admirer of Gould or who has an interest in how the performance of this piece has developed over time.

Rosalyn Tureck 1957 and 1988.  I have listened to Tureck’s 1957 recording on YouTube.  I have a CD of her 1988 recording.  A mere two years after Glenn Gould provided us with his groundbreaking, high-speed, high energy performance of the Goldbergs, Tureck gives us the anti-Gould performance.  The 1957 recording is possibly the longest one ever made.  Where Gould gets through the piece in 38 minutes with no repeats, Tureck takes more than 90 minutes, taking every repeat.  It’s really hard to believe that Gould viewed Tureck as his musical model.  The remarkable thing about the 1957 performance is that, once you get used to the idea that this is an extremely slow performance, it makes perfect sense.  She is absolutely committed to her approach.  The slow tempos allow her to emphasize rhythmic elements that might be lost in a brisker reading.  In addition, the lax speeds necessarily prevent the creation of what I think of as a “wall of sound” so that the intersecting planes of the music stand on their own more clearly than they might in a rapid-fire performance such as Gould’s 1955 recording.  The 1988 recording is decidedly brisker; she gets through the piece in 73+ minutes, but still has the same commitment to rhythm and clear articulation.

Gustav Leonhardt 1965 (harpsichord).  Leonhardt recorded the piece three times that I am aware of: 1953 (ahead of Gould); 1965; 1976.  I listened to the 1965 version and did not compare it to the earlier or later renditions.  Leonhardt plays at an extremely slow pace.  The 1965 version as heard on YouTube is 48 minutes long without repeats, implying a performance time of one hour 36 minutes if he had taken all repeats.  That makes even Tureck’s 1957 recording look speedy.  Frankly, if you are planning to conduct a study of 20^th century harpsichord performance practices, you need to listen to this disk, but otherwise you don’t.

Trevor Pinnock 1980 (harpsichord).  Trevor Pinnock was a stalwart of the “authentic performance” movement that was the rage in the 1980s.  The idea was to perform the music of the Baroque and Classical periods on instruments that were authentic to the period, using orchestral forces and tempos that were proper to the time and place of composition.  He and his colleagues Christopher Hogwood, Nikolaus Harnoncourt, Roger Norrington, John Eliot Gardiner, and a number of others created an atmosphere in which audiophiles and music lovers, myself very much included, would race around to find the latest recording on ancient instruments that could then be played through the most modern stereo playback equipment available.  It never occurred to me to insist that we could have a more authentic experience if we used wax cylinders and wind-up Victrolas to listen to these recordings.

Apart from all of that, Mr. Pinnock is a superb musician and has an obvious love for this music.  His recording is my favorite rendition on the harpsichord.  He performs the work on an instrument that appears from the date on the soundboard to have been constructed in Antwerp in 1646.  There is a record of numerous modifications made to the instrument.  It was updated in 1756, 1770, and 1780, then completely restored in 1881 and again in 1968.  It has a marvelous sound, if you like the sound of harpsichord.  (Sir Adrian Boult said that a harpsichord sounds like two skeletons copulating on a metal roof.)

Pinnock omits about half of the repeats (including in both occurrences of the Aria).  His tempos are nicely judged and the entire performance has a pleasant, natural feel.  He does not rush anything, nor does he turn the slow variations into extended studies in anguish.  This is the first performance I ever heard on harpsichord and it remains my favorite on that instrument.

Grigory Sokolov 1982.  This is a recording made in performance, in February 1982 in Leningrad (as it was then called).  I understand that Sokolov records no other way.  He records the work as it comes to him on that particular evening, with no attempt made at editing.  Sokolov has been called the world’s greatest living pianist.  Listening to this performance, I can believe it.  He draws effects from the piano that are breathtaking.  At the same time, the performance is highly idiosyncratic.  What you get when you listen to this performance is Sokolov’s impression of the music on this particular night.

His advice to a younger pianist was to practice the right way so you can say what you want to say.  It was an off-hand comment, but it suggests that when he performs, it is he rather than the composer who is speaking.  Sokolov is said to spend his entire waking life practicing when he is not performing.  So, he is one of the small band of brothers and sisters I referred to at the beginning who can do anything on the piano.  He certainly proves it in this performance.  The question I am left with is whether I have just heard brilliant pianism or a profound exploration of one of Bach’s masterpieces.  Every listener will have to decide for him/herself, but I lean to the former.  After I have succeeded in getting away from this music for a couple of months, I will plan to return to this performance to see if I have a different opinion. 

Tatiana Nikolaeva 1985.  Nikolaeva recorded the work five times: 1970, 1983, 1985, 1987, and 1993.  The last of these was recorded by Hyperion (that is, in the West) and is available on Amazon.  Amazon has one other version also available, but it is not clear which one it is.  The first two and the fourth are Russian pressings.  The one that I heard is the 1985 recording, recorded in performance by the BBC.  There is wild, massive applause at the end of the recording, but it is a measure of the respect that the audience had for the performer and the power of her performance, that they waited for the last quiet note to die down before erupting.

Nikolaeva won fame in 1950 when she placed first at a competition held in Leipzig (then in East Germany) to commemorate the 200^th anniversary of Bach’s death.  The idea was that each competitor would come to the stage and play a prelude and fugue from The Well-Tempered Clavier that the performer had prepared for the judges.  When Nikolaeva’s turn came, she went to the stage and said to the judges, in effect, “Which one do you want me to play?”  I don’t know which one they asked for, but she played the prelude and fugue from memory and won the competition.  Dmitri Shostakovich was one of the judges and was so impressed that he composed his own set of 24 preludes and fugues specifically for Nikolaeva to perform.

This is the most lyrical performance of the Goldberg Variations that I have heard.  The music sings under her fingers.  She produces drama, pathos, joy as she feels them.  She also produces a wide range of pianistic effects.  She clearly loves Bach and loves the piano.  I thoroughly enjoyed listening to this performance, but I am not sure I would return to it.  Her approach to many of the pieces is highly romanticized and emotionally charged.  She is obviously a brilliant performer, but if my objective is to discover what Bach is doing in this music, rather than what the performer is feeling, I think I need to look elsewhere.

Zhu Xiao-Mei 1990.  This is one of the finest performances I have heard.  Zhu was born in Shanghai in 1949 and was giving concerts on Chinese radio and television by the age of 8.  At the age of ten she entered China’s national music school for gifted children.  Unfortunately, she ran afoul of the Cultural Revolution and spent five years in a labor camp in Inner Mongolia.  She was forbidden to play the piano, but somehow her fellow prisoners found one for her to use, so she was able to keep practicing during her imprisonment.  She left China in 1979, lived in the U.S. for a while and then moved to France, where she now lives.

Her performance is clear, sensitive, beautifully performed.  Her tempos are sensible and well-judged.  She has a unique solution to the problem of whether or not to take repeats.  She avoids the school that takes them all, the school that takes none, and the school that takes some.  She takes first repeats (with the exception of number 16) and she avoids second repeats (with the exception of number 30).  It’s a superb performance, but ultimately her eccentric decision regarding repeats puts me off.  The individual variations sound unbalanced because what should, to my ear, be something made up of twos or fours is made up of indivisible threes.

Angela Hewitt 1999.  Angela Hewitt is an artist of great subtlety and intelligence.  She definitely belongs to the “slow” school of Goldberg performance.  Her performance runs more than one hour eighteen minutes.  Every minute is enjoyable.

The story is that she spent several days recording the piece in studio during August of 1999.  After five days of recordings, she went back to the studio one evening with some friends to do a run-through.  The engineers were in the studio and recorded the performance, which they and the artist agreed was the one they had been looking for.  That is the performance that is presented on this disk, although they did a few patches and edits.

There is no heavy sentiment in her performance, but she wrings the emotional content from each piece.  Her judgment of tempos is perfect and she gives us a clear view of Bach’s counterpoint when it is in action.  The recording likewise is superb.  This would have to be on anyone’s short list of best Goldberg recordings.

Murray Perahia 2000.  This is another fine performance with well-judged tempos and superb musicianship.  It sounds like damning with faint praise, but Perahia is always reliable.  He never tarts up a piece of music to give you his impression.  Rather, he plays everything with great integrity and humility.  He does the same here.  If this were the only version of the piece available to me, I would be well contented to use this as my reference for what the Goldbergs are meant to sound like.  Because there is a much wider field, I have tended to gravitate to other performances as my favorites, but not because of any perceived flaws in Maestro Perahia’s presentation of the work.

Andras Schiff 2003.  This is said to be his third recording of the work, but the only other one I could identify was made in 1983.  The 1983 recording was made in the studio, the 2003 was recorded in performance.  I found a video performance on YouTube and listened to that.  The performance is lively and lyrical, which is what one expects from Schiff.  The audience was also active and could not quite synchronize their coughing, although they worked on it constantly throughout the performance.  It may be that the sound engineers were able to deal with audience noise when they mastered the CD, but it certainly detracts from the performance I heard.  

Sonia Dinnerstein 2007.  I am going to make some negative comments about this disk, but before I do, I want to acknowledge that anyone who can play this music has a relationship to it and with the composer that those of us who listen do not have.  I grant that anyone who can play the music has a right to his or her interpretation.  We listeners may comment, but I concede that the performer has a prior claim to understand the music.

Before I listened to this disk, I learned that Ms. Dinnerstein had been a guest on the Oprah Winfrey show and her performance of the Goldberg Variations had received the endorsement of Ms. Winfrey[5].  On top of that, Ms. Dinnerstein has an inspiring personal story.  I haven’t looked her up to find out what it is, but a classical pianist can hardly expect to earn air time with Oprah without one.  After receiving Oprah’s recommendation, her recording shot to the top of the Billboard classical charts.

Knowing that much, I was predisposed to dislike the recording.  However, as a fair-minded person, I held my prejudices in check as I listened.  There is nothing worse than snobbery and I worked as hard as I could to retain my objectivity and my sense of fair play.

After listening to the performance, I imagined two possible, but obviously fictional, scenarios that would explain how it came to be produced.  (1)  In a variation on the Mel Brooks script The Producers, the producer of the disk needed, say, 100 investors to fund the project.  He sold the project to 3,000 investors.  His plan was to produce a record so bad that the project would be a total loss.  He would tell each group of 100 that he had unfortunately lost their money, while he pocketed twenty-nine times the amount of the investment required to produce the disk.  It took months to find the right artist, but when Sonia turned up, he knew he had struck gold.  Then the artist had to volunteer to appear on Oprah, the record soared to the top of the charts, and the producer was ruined.  (2) Ms. Dinnerstein had read the history of the Goldberg Variations, believed the literal truth that the piece had been composed to help an ambassador fall asleep, and calibrated her performance to have the same effect on a modern audience.

Maybe the best summation is in a one-star review on Amazon.  This really should be subtitled “Bach for Lovers.”

Andreas Staier 2010 (harpsichord).  This is one of the most disappointing disks I have heard in a long time.  The piece was performed on an exact replica of a harpsichord built in 1734 by a builder with the delightful name of Hieronymus Albrecht Hass, who worked in Hamburg and was famous in his own time.  The Hass instrument was built with the idea that it could replicate many of the sounds that can be produced by an organ.  According to Staier, Hass’s instruments “are the biggest and most richly equipped with registers of any built before the twentieth century”.  Unfortunately, Herr Staier uses this wonderful instrument to create a “wall of sound” comparable to the 1928 piano roll performance that I complained about some pages back.  To my ear, we end up hearing all the wonderful features that the Hass harpsichord can produce, but we don’t end up hearing the Goldberg Variations in any recognizable form.

David Jalbert 2012.  This is yet another in a series of dazzling performances by a Canadian pianist.  Canada must have more virtuoso pianists per capita than any other country on earth.  Jalbert’s performance is thoroughly engaging, his tempos are beautifully judged, and his command of the piece is impressive.  In addition, the quality of the recording itself is exceptional.  This may be the finest recording of a solo piano I have ever heard.

Kiniko Ishizaka 2013.  This recording is part of the “Open Goldberg Variations” project.  There is more information at opengoldbergvariations.org.  At the site, you can listen for free to a recording of the piece by Kiniko Ishizaka, who is not only a fantastic pianist but has also been a member of Japan’s Olympic weight-lifting team.  At the website, you can download recordings with better audio quality and pay what you choose, or you can buy an audio CD (also available from Amazon).

Another feature of the site is that the score of the Goldberg Variations is available for download on a free license.  You can copy it, print it, send it to friends.  One of the charming aspects of the score is that it was crowd-sourced; those who contributed could “buy” a variation and post a message on that page of the score.  Some of the messages are quite touching.  Some are straightforward: “In memory of my mother, Lorraine Halse Vines, from Rose.  The thought of you is music in my heart.”  One of my favorites is in German: “Für Sylvia von Heiko.  Weil mit Dir alles schönsten ist.”  (I think: Because with you everything is most beautiful.)  A strange one is addressed to all Bach lovers, “now that this work is open and freely available to all of us.”  Up until the publication of the “open source” score, Schirmer would have parted with a copy for about ten bucks, and you still need an internet connection to get the free copy.  Anyway, I thought it was a charming conception, and it is very handy to have the score available on my hard drive.

As to the performance, Ms. Ishizaka’s style is energetic, focused, and a pleasure to listen to.

Incidentally, the “Open” concept has moved on to the Well-Tempered Clavier.  Ms. Ishizaka has recorded Book I.  I assume that Book II is in the works.

Jeremy Denk 2013.  I think this may be the ideal recording of the Goldberg Variations.  Tempos are forward-moving but not brisk.  He treats the bass voice as an equal partner in the eight canons that feature a free voice.  His counterpoint is impeccable.  He performs the Black Pearl with great emotional intensity, but he doesn’t rely on agonizingly slow tempos to achieve that result.  He finds insights in some of the variations (I am thinking of numbers 19 and 20 particularly) that no one else has pointed out.  The recording is first rate.  In addition, if you buy the CD (rather than the MP3), a DVD is included as a substitute for liner notes.  On the DVD, Denk illustrates various interesting features of the Goldbergs.  It’s a nice addition to an already superb recording.

Evaluating the Performances

Putting aside the harpsichord v. piano question and the “to repeat or not to repeat” question, after listening to this piece countless times, and after listening to some of these recordings as many as half a dozen times, I have focused on a few elements in evaluating them.

Variation 8 is a piece alive with energy.  I prefer performances that play it briskly, but with an emphasis on counterpoint.  Some performers over-emphasize the top line (Ishizaka).  A couple, surprisingly, emphasize the bottom (Perahia, Staier).  In my opinion, we should hear the two lines, bass and treble together (Denk, Hewitt, Jalbert).

Variation 13 is the beauty variation in my opinion.  Yet Gould and Tureck insist on excising that quality.  If you want to, you can make the piece sound like an old-fashioned typewriter or chickens pecking at corn.  But why would anyone do that?  At the same time, there is no need to schmaltz the piece up a la Landowska.  I think that Denk, Hewitt, Jalbert, Nikolaeva, Pinnock, Sokolov all do a lovely job with this variation.

The transition from Variation 25 to Variation 26 is a critical moment in this music, again in my opinion.  Number 25 ends on a note of dark desolation.  Number 26 is a lively virtuoso piece.  I appreciate it when the performer opens number 26 gently so that the spell of number 25 is still at work during the first few bars of number 26.  The standard treatment is to start banging on the keys right at the start of number 26.  Denk does a lovely job of opening number 26 gradually.

Variations 28 and 29 are both special cases.  The keyboard is barely big enough to contain the music.  It is as if the wizard has left the workshop for a few minutes to collect some items he is going to need for the Quodlibet that lies just ahead.  During his absence, his imp takes over and inserts these two maniacal gems into the masterpiece.  I prefer performances where the artist treats these variations as a climax that will be relieved by the Quodlibet and, ultimately, the quiet return of the Aria.  Denk provides the best example, followed closely by Hewitt and Jalbert.

In the final variation, the Quodlibet, I prefer performances that emphasize the counterpoint of the piece, which on many performances is not as clearly heard as it might be.

Desert Island Selection

If my desert island recording were to be selected at random, I could be happy as long as I could be guaranteed that the selection would exclude Serkin, Dinnerstein, and Staier.

If I could choose and could only choose one, and with the understanding that I might make a different choice tomorrow, I would take the recording by Jeremy Denk.  However, I would plan to hide the Pinnock, Hewitt, and Jalbert recordings in my luggage.

I feel much better now.  I think I may be able to move on to a different piece of music.

[1] Bach did not use the title “Goldberg Variations”.  That’s a nickname that was applied decades after Bach died.  The formal title, translated into English, is Keyboard Practice, consisting of an Aria with Diverse Variations for the Harpsichord with 2 Manuals. Composed for Music Lovers, to Refresh their Spirits.

[2] Child keyboard player.  It is possible that this is an actual German word, but it’s a coincidence if it is.  I couldn’t resist making it up.

[3] The titles: (1) “I have been away from you so long.  Come closer, come closer.” (2) “Cabbage and turnips drove me away.  If my mother made meat, I’d have opted to stay.”  People were more easily entertained in eastern Saxony in the 1730s than are their descendants.

[4] De Pachmann was a magician.  Search on YouTube for: “Vladimir de Pachmann plays Chopin Nocturnes (piano rolls)”.

[5] I assume it went something like: Best. Bach. Recording. EVER!!