13. The best and worst of amateurism

...which is the third installment of the Settling Scores

Having discussed descriptive and prescriptive notation in my last post, I find an opportunity to explore an example (and which is it supposed to be?) relating to Bach.  Here is a surprising recent publication:

SOURCE:  Anthony Tommasini, "Glenn Gould's Every Detail.  But Why?" nyt.com (June 1, 2016)

This curious edition presents two texts of the “Goldberg” Variations on facing pages throughout.  The left-hand [verso] pages are labelled “Original Version,” while the right-hand [recto] pages are labelled “Gould’s 1981 Version.”  I reproduce an extract of the beginning of Var. 29 below, although with the twin texts arranged here vertically rather than horizontally just to be more legible in the blog medium:

SOURCE:  cropped scans of pp. 146-147 of Hopkins, ed., Glenn Gould's Goldberg Variations.

From what I can tell, Hopkins’s transcription of Gould’s 1981 performance is impressive, even if it leaves me with some questions.  (In the example above, what distinction does Hopkins intend between Bach’s triplet notation and his transcription of Gould’s performances as sextuplets? I found no explanation of this idiosyncracy, and the only thing I can imagine from it is that Gould’s performance downplays the half-beat.)   This project was clearly a labor of love for Hopkins, and the right-hand pages thus reveal the very best of amateurism—doing something for the love of it, going far beyond what might be asked or imagined.

It is the left-hand pages that leave me disturbed.  Even just the use of the word version for both the “original” (and I’m coming back to that word in a minute, too) and the performance troubles me.  I would have favored different words—maybe “text” for what Gould read, and “interpretation” for what he played.  Granted, the interpretation is filtered through Hopkins’s interpretation into score.  Whatever.  There would be great value in comparing what Gould had studied with what he played.  Of the text Gould used, Hopkins asserts that it was the 1938 G. Schirmer edition prepared by Ralph Kirkpatrick:

“We know that Gould used the Kirkpatrick edition, and only this edition, because three copies of this edition are presently housed in the Glenn Gould Archive (the official repository for Gould’s archives) in the National Library of Canada (NLC).  On the basis of Gould’s editorial markings in these scores (or lack of markings), each copy was seemingly used by him at various points in his career for various purposes.” [p. 10]

Hopkins suggests that one of these three copies (“almost entirely free of markings, such as fingerings, articulations, dynamics and tempos”) was likely the one that Gould used as he learned the work prior to the 1955 recording, and informs us that some pages have gone missing; a second copy seems to be a “reference score during the post-production process for the 1981 recording.”  (I wondered if it is a more recent printing; Hopkins doesn’t give any such information, still less any shelf-mark or locating information.)  The third copy, also incomplete but with very neat and comprehensive fingerings added for the aria and the first eight variations, belonged to Gould’s girlfriend during his conservatory years.  Hopkins concludes

“The relevance of the three copies of this edition is that they show Gould had little concern with the quality of editions that he used over the course of his career.  There is no evidence that he ever researched or consulted other editions for the purpose of critical analysis.” [p. 10]

And yet a page later Hopkins quotes Kevin Bazzana discussing films made immediately after the 1981 recording saying “in [NLC] videotape no. 50A, [Gould] can be seen with the 1979 [recte 1978?] Henle edition of the score.”  Well, what was he doing with that?  Is that not evidence—at least circumstantial, if not an actual smoking gun—of Gould “consulting” another edition?  And I noticed an instance (Var. 26, b. 14, 2nd beat, middle voice) where the note Gould plays in 1981 (D) is in the Henle text and NOT in the Kirkpatrick text (which has E; in 1955 he had played E).  As Hopkins uses these tapes to determine Gould’s fingering as best he can (and an impressive job it seems to be), clearly he deems the videos relevant to the 1981 audio recording.

If a facing-page edition is going to have some value, the facing pages need to relate to each other.  I can imagine two ways that this might have been done:  1) on the left-hand pages, provide a transcription of Gould’s 1955 recording, so that the two performances might be compared, or 2) on the left-hand pages, reproduce the Kirkpatrick text that Hopkins claims Gould used.  (Probably there would be copyright issues with that.  Did Hopkins ever pitch this project to Schirmer, who presumably holds that copyright?)

The Kirkpatrick edition is apparently still in print, and it is also very widely available in libraries (and, I imagine, in piano benches here and there)—indeed, more available in libraries than this Gould transcription will ever be—so interested individuals should have no trouble getting their hands on a copy to make the comparison.  (At a glance, WorldCat lists over 450 library copies of various printings of the Kirkpatrick edition, with just over 30 of the Hopkins/Gould score.)  Comparing the two is particularly interesting because Kirkpatrick often resorts to extra staves to realize Bach’s ornamentation or (yet more significant in this context) to modify the part-crossing to facilitate performance on a piano.  Here are two examples—the aria, where Gould’s ornamentation is somewhat slower than Kirkpatrick’s instructions (32nd-notes rather than 64th-notes), and an example of adapting the music for a single keyboard:

SOURCES:  first sixteen bars of the Aria, in scans of Kirkpatrick p. 3 and Hopkins p. 49.

SOURCES: the end of Var. VIII in marked-up cropped scans of Kirkpatrick p. 23 and Hopkins p. 75.

Instead of either of these strategies, on the left side Hopkins provides an entirely new text, which he explains thus in his introduction:

“Each of the variations [and the Aria] is presented in its original form on verso pages, accompanied by Gould's realization on recto pages, thereby allowing for ease of comparative analysis.  The original forms, labelled ‘Original Version,’ were produced from the Bach-Gesellschaft Ausgabe edition (BGA) of 1853, the Hans Bischoff edition of 1883 and the Ralph Kirkpatrick edition of 1938.  The Handexemplar, Bach’s personal copy of the first engraved edition, was likewise used for this purpose, yet the editions produced by the Neue Bach-Ausgabe (NBA) of 1977 and Henle of 1979 could only be consulted, due to copyright restrictions.  Discrepancies amongst these editions are noted in the Critical Notes on pp. 45-47.”  (p. 9)

Bach’s Handexemplar, corrected in many places by the composer, was rediscovered in 1975.  It was naturally the most important source for Christoph Wolff’s NBA text and its discovery also prompted Henle to issue a revised version of Rudolf Steglich’s 1973 edition (with the revisions undertaken in 1978 by Paul Badura-Skoda).  What does Hopkins mean by “could only be consulted”?  All of the sources he lists were presumably “consulted,” and as his critical notes list variants in each of these texts, it is unclear how “copyright restrictions” have impinged on his task at all.  No, what we have here appears to me to be a reinventing of the wheel:  a new edition that presumes to be scholarly, but executed, in my opinion, in an haphazard way.

Before giving some examples of this, I should note that the NBA text of the Goldbergs has not been universally acclaimed.  In particular, in a 1990 article in Performance Practice Review, Erich Schwandt took Wolff’s edition to task for a number of perceived deficiencies, concluding

“Something must be broken in the mechanism when musicological ‘overkill’ produces 27 pages of Critical Apparatus (roughly two-thirds of a page of words per page of music) and then gets the notes wrong.  I believe that the Neue Bach Ausgabe should seriously consider withdrawing Christoph Wolff’s edition of the ‘Goldberg Variations.’” (p. 69) 

I have examined Schwandt’s critique closey, and I see no warrant for such a charge.  Wolff has not gotten “the notes wrong.”  An edition is an interpretation—a performance, if you like—and although I would like to see every alteration to a source text logged in the critical report, I find only four instances where this not is the case, each having to do with an added appoggiaturas that are not clearly marked in the score as editorial additions.  Schwandt may disagree with Wolff’s and the NBA’s editorial principles (over the ornamentation symbols particularly), but in the critical report Wolff articulates those principles and the decisions that proceed from them.

It must be stressed that in a critical edition, the printed score and the critical report are equally necessary for an understanding of the textual situation.  This seems to have escaped Nicholas Hopkins as he prepared his new “original version” for his Glenn Gould project.  His basic editorial principle seems to be that the Handexemplar, bearing corrections in Bach’s hand, establishes a final text for the Goldbergs once and for all.  His “original version” thus aims to be essentially a resetting (following modern notational practices) of the corrected reading of the Handexemplar, and his critical notes detail differences between his new text and the other five published texts he cites above.  In fact, the vast majority of his critical notes list places where an accidental is missing in the Handexemplar but appears in the other editions.  There is a good reason for this:  the first edition (of which the Handexemplar is a copy, of course) is a product of different notational conventions than ours today:  an accidental was regarded as affecting only the note to which it was affixed.  As an example, in Var. XXVIII b. 23, the left-hand part is crowded with repeated sharps for the recurring note:

SOURCE: cropped scan of the Handexemplar, p. 30 (my mark-up)

This would be too many accidentals by our current standards; the following bar has one too many, while also lacking two that would now be needed:

SOURCE:  ditto

The second natural sign seems superfluous by today’s standards, but we would now expect a natural sign for the last note in both hands—no longer D-sharp but D-natural.  As Hopkins seems unaware of the older practice, he documents it as if it is news.  More than half (at a rough count, 37 out of about 70 total) of his notes deal with this, an item not even worth noting.  Several of Hopkins’s notes record details such as “a dotted tie is notated ... in the NBA.  This tie is found in no other sources.”  Exactly:  the tie is dotted because it is an editorial emendation.

A more serious problem is that—as he gives no indication that he has seen the NBA critical report—he is apparently utterly unaware that of the seventeen other copies of the original print which Wolff examined, six have corrections that Wolff is able to attribute to Bach, and no two of these copies have exactly the same corrections:

SOURCE:  cropped scan of Kritisher Bericht for NBA Ser. V, Bd. 2 (1981), p. 93.

In Wolff’s table, source A1 is the Handexemplar, but not all of the corrections made it into the Handexemplar.  (This table only lists corrections found in the other six copies, noting when they are or are not duplicated in the Handexemplar; there are quite a number of corrections unique to the Handexemplar which Wolff documents elsewhere.)  Evidently not knowing of this, Hopkins re-introduces errors into this text which Bach had corrected:

SOURCE: Var. XVI, from marked-up cropped scan of Hopkins, p. 102.

As the table above indicates, in five of the seven copies with corrections traceable to Bach, this E has been changed to F-sharp.  The note is given as F-sharp already in the BG edition (1853), as the copy Wolff identifies as A5 was the source for the edition and has the correction.  Kirkpatrick was working from a different copy, lacking the correction, and so gives E.

Another instance concerns a missing accidental:

SOURCE: marked-up cropped scan of Hopkins, p. 130

The circled A is given as A-flat in all of the editions that Hopkins consulted save the original print; it had been changed to A-flat in only two copies of the first edition (Wolff’s A6 and A8)—neither of which was used as a source for editions prior to the NBA, although it did show up in an early manuscript copy (Wolff's C1) which Bischhoff and Steglich had used.  All of these editors recognized the musical sense of A-flat; even in the intensely chromatic vocabulary of Var. XXV, the A natural is jarring.  Try it for yourself.

Hopkins did catch a genuine lapse in all of the editions he used, and apparently hitherto unnoticed.  It again concerns a missing accidental in this same movement, and it seems clear how it was missed for so long:

Composite of Var. XXV, b. 10
SOURCES:  marked-up cropped scans of  (top) Handexemplar p. 25; (lower left) NBA Ser. V Bd. 2 p. 104; (lower right) Hopkins "original version" p. 130.

The natural sign (missing in the NBA and all the other editions) would not have been expected in the original print, given its conventions regarding accidentals; but subsequent editors may have missed it because 1) in the original print there is a system break before beat 2 of this measure, so that the preceding D-sharp is out of sight and out of mind; and 2) it is so manifestly obvious musically that it was implied even when not on the page.  Neither is a satisfactory excuse.  But has anyone ever played a D-sharp here?  I wonder.  Hopkins is the first of these editors to publish the work as a computer-set (rather than engraved) score, and this is the sort of detail that computer-setting captures more readily than the weary eye.

That said, Hopkins also introduces what is, in my opinion, a shocking error:

SOURCE:  cropped scan of Hopkins, p. 60.

The first two notes in the right hand are a third too high.  That might be regarded as a typo, but the error is compounded on the facing-page transcription:

SOURCE:  cropped scan of Hopkins, p. 61.

These are not the notes that Gould plays—he plays B-G, exactly as notated in every other edition.  And although this could be a simple copy-and-paste error if Hopkins used his own “original version” as the base text for his “Gould’s 1981 version,” it is further compounded by a footnote at the bottom of the same page, explaining the asterisk and even naming the notes (incorrectly) “D and B.”

SOURCE:  ditto

This is not a typo, but rather an editorial blunder.  That there could be an error of this magnitude—even redoubled in a footnote—suggests to me the possibility that there was no editorial oversight whatever to this production.  (Hopkins apparently set this edition himself.)  Lack of editorial oversight would be par for the course for a product that appears under a Creative Commons license on the IMSLP; it is astounding in a publication from a house of the reputation of Carl Fischer.  Granted, there is no backlist of “urtexts” at Carl Fischer, but now Hopkins’s “original” version (that is, the left-hand pages) has just been issued as a new urtext of the Goldberg Variations.  I have not examined that publication, but I hope that—at the very least—this error has been fixed.  Even so, given the rest of the concerns voiced above, Hopkins’s “original version” presumes too much.  As followers of the blog will know, I am all in favor of the proliferation of editions offering valid texts of all sorts, but I think the market has no need of this “urtext” edition of the Goldberg Variations.

The right-hand pages show the best of amateur devotion (although I think the introduction sometimes veers over into hagiography); the left-hand pages appear, in my view, to manifest ad hoc amateur naïveté presented as serious scholarship.  Drink deep, or taste not the Pierian spring.

11. Dart’s Brandenburgs

...which is the second installment of the Settling Scores

I am never in Manhattan over New Year’s, but years of following the concert listings in The New Yorker and The New York Times indicate that in NYC there is a holiday tradition of programming all six of Bach's Brandenburg Concertos.  (Okay, so that’s a nineteenth century name for these works, but I think we’re stuck with it, so I’ll dispense with the scare quotes.)  Somebody—the Chamber Music Society of of Lincoln Center, the NY Phil, or St Luke’s, or a visiting group at Bargemusic, Symphony Space or somewhere—presents these six exquisite pieces as an essential secular holiday event.  With that in the back of my mind, this seems like as good a time as any to look at textual issues relating to these works.

Bach’s 1721 fair-copy dedication manuscript survives to this day, and one might assume that it should settle any textual questions about the music.  It served as the sole source for Wilhelm Rust’s 1871 Bach Gesellschaft edition, except for the fifth concerto, for which Rust also consulted an autograph set of parts.  Rust’s edition is readily available in inexpensive reprints, and that is a good thing:  with really just one source, the edition is quite faithful to that dedicatory manuscript.  In one important respect, the BG edition is closer to Bach’s notation than that of the Neue Bach Ausgabe (ed. Heinrich Besseler in 1956):  the BG preserves the transposing notation, not converting everything to concert pitch as the NBA does (although determining “concert pitch” in Bach is an issue I must return to later this year).  This affects not only the brass:  importantly, it affects the violino piccolo (conc. no. 1), whose triple-stops appear completely unidiomatic when rendered in concert pitch.  (Granted, the BG changes the clef of the recorders (if that’s what they are, in nos. 2 & 4), which the NBA does as well.  The NBA goes further, and changes the clef of the violas da gamba in no. 6 from tenor to alto.)  It’s nice when the cheap and convenient editions are also good ones.  The Dover reprint even includes an English summary of Rust's preface, which at least makes users aware that there are textual issues to consider.

I describe Rust’s edition as “good” because it is so close to the reading of the dedication score, although Besseler was right to question what value that source should have:  “The numerous errors show that Bach was careless.  As only some have been corrected, the dedication score does not have the value that a manuscript made for his own use would have.”  [NBA Krit. Ber. p. 12.]  This would mean that the dedication score is “authorized” (i.e., in Bach’s hand) while not really being “authoritative” (i.e., his intentional definitive presentation of the text)—an interesting distinction—although, as I discussed last month, “definitive” is not the most useful term in such contexts.  The search, then, is for the parent texts from which Bach (carelessly) prepared the dedication score.  That search led Besseler to some manuscript copies made by Christian Friedrich Penzel some ten years after Bach’s death.  Penzel was one of Bach’s very last students—not quite 13 when the old man died in 1750.  As the readings differ a bit from the dedication score (and, as that score would not have been available to Penzel in Leipzig anyway), these sources seem to be a useful point of departure in understanding a “first version”—even if, at best, Penzel’s source postdates Bach’s re-use of some of the material in cantatas (BWV 52, 207, and 207a), and it cannot be regarded as an uncontaminated earliest [i.e., Ur-] text.  Other Leipzig copyists transmit material we associate with the Brandenburg Concertos, but these too are disconnected from the dedication score, and so also likely transmit early readings—most famously, perhaps, the much shorter version of the harpsichord solo in concerto no. 5, transmitted in only one source.  Indeed, that source is eccentric in other ways, for example its description of the concerto as a “Concerto Quadruplo,” elevating the cello into the concertante group:

SOURCE:  Bach Digital, detail of title wrapper for parts of  Concerto no. 5 copied by J. C. Farlau.
To my eye, the word Quadruplo is clearly a later addition, as is the insertion Violoncello Concertato,
but even then it's not clear that both of those are written in the same hand:  look at the r in each.

(The cello part transmitted with that set—but in the hand of yet another copyist—is rather different than the traditional reading, adding it as continuo in the second movement, and doubling or adapting the left-hand harpsichord part in other places.  If you want to see something crazy, look at the first movement, bb. 95ff., where this cello is included with the six-bars of trills from the soloists.)

SOURCE: The Music Parlour blog

The most powerful way to become aware of musical textual variants is to hear them, and no one brought these audibly to our attention more than the short-lived but remarkable English musician and musicologist Thurston Dart.  Already in 1959 he released a recording of the Brandenburgs which was a veritable shot across the bow of conventional musical wisdom.  These recordings—available for download via this interesting post on The Music Parlour blog—seem  to borrow the rhetoric of the Sermon on the Mount: “You have heard it said..., but I say unto you....”  You think Bach’s corni da caccia are horns?  Dart says they should be Jagdhörner sounding an octave higher, so he does the first concerto using trumpets instead.  He inserts the slow movement of BWV 1021 to provide the Phrygian half-cadence between the two movements of the third concerto.  It is a provocative recording.

SOURCE:  discogs.com

Dart wasn’t finished with the Brandenburgs.  He prepared an version for the Academy of St. Martin-in-the-Fields for their 1971 recording, conducted by Neville Marriner.  (Although the Phillips release refers to “Dart’s edition” it was never published in score. Perhaps “version” or “reading” is a more accurate term?)  Here again the same I’ll-show-you spirit prevails in Dart’s interpretations of the evidence, and the album cover has the audacious claim “FIRST VERSION—FIRST RECORDING.”  He follows Penzel's text for the first concerto, so that the third movement is gone, as is the polonaise (both appear in an appendix) and the accompaniment to the horns in Trio II is a rather different unison line for strings rather than oboes.  In this recording the corni da caccia parts of the first concerto are restored to the horns in their usual octave, but the tromba of the second concerto is allocated to the horn rather than the trumpet, down an octave, no longer at the very top of the texture.  His evidence for this is a copyist’s part annotated (by whom?) “ô vero Corno di caccia”—and this alters the soundscape in a remarkable way.

Dart expended a lot of scholarly energy on Bach’s description Fiauti d’Echo in the fourth concerto:

SOURCE:  Bach Digital, cropped page scan from the dedication score, f. 38r. 

His conclusion was that these were really flageolets, sounding an octave higher than notated, so that on the 1971 Marriner recording these are performed on sopranino recorders.  You can’t miss them. Much more could be said here, and this issue has generated a literature of its own.

Dart died before the 1971 recordings were issued; he participated as a continuo player in only six movements (including all of Conc. no. 3, including the interpolated movement from BWV 1021).  Although Philips re-released the 1971 Marriner recording on CD at least as late as 1989, these recordings seem utterly unavailable today through newer releases or streaming/download services (other than Concerto no. 4 as part of a 2004 Marriner tribute collection).  I only ran across them because I married into a copy.  Dart’s approach—if not precisely his conclusions—has become much more widely known through the output of his former Cambridge student Christopher Hogwood, who was also always ready to perform a variant for the sheer excitement of hearing something new.  His 1985 recording of the Brandenburgs retains some of Dart’s ideas, as well as some new departures (particularly regarding the choice of instrument for the bass line).  The Dart attitude is clear in Hogwood’s liner notes:

[Bach’s] desire to impress the Margrave with variety above all is apparent, alarmingly in Concerto 1 where the revised version addas a new concertante third movement for the violino piccolo to a work that opens with a strongly ripieno movement; and in Concerto 5, where a harpsichord episode of nineteen bars is inflated out of all proportion to produce what is currently mistermed a ‘cadenza’ of sixty-five bars.  [Notes to Decca 414-187-1, p. 2]

SOURCE:  allmusic.com

How alarming!  How disproportionate!  How tasteless!  Add to this Hogwood’s characterization of the dedication score's “specious authority stemming more from its Dedication and calligraphy than from its value as source material,” and we may perceive that we are receiving Dart’s spirit through Hogwood’s hands.  At every turn he offers something different.  For the first concerto, Hogwood follows Penzel’s text strictly—no violino piccolo, no Allegro 6/8 and no Polonaise.  Marriner had conceded to the 1971 public with an appendix, but Hogwood’s public in 1985 was ready to be shocked by difference.    To differ from Dart he scraps the idea of interpolating a movement from BWV 1021 as a slow movement to Concerto no. 3; he restores alto recorders as the Fiauti d’Echo in no. 4; and the horns and trumpet are back in their usual places (nos. 1 and 2 respectively) at their usual octave.  With admirable restraint—or is it marketing savvy?—he eschews the lengthy harpsichord solo in no. 5 for the more abrupt version.  I can remember hearing this recording of no. 5 on the radio in about 1987, as I was about to shell out some cash for the Brandenburgs.  Hearing the short variant, I recall reacting “Aw, man! No!” and buying Trevor Pinnock’s recording instead.  And—much as I admire Hogwood and his work over the decades—I can’t say I’ve regretted that youthful investment.  As well-played as these recordings are, are they (and Dart’s) not just a little too eccentric to become standards?  As Richard Taruskin wrote thirty years ago, after quoting the same Hogwood passages I quote above,

In his recording, Mr Hogwood has rectified Bach’s lapse by reinstating the original nineteen-bar solo.  Let me suggest that this conglomeration of shallow fireworks and harmonic barbarities, however ‘in proportion’, and however it may conform to the performer’s idea of the stylistic norms of the day, is poor music by any standard, and that by replacing it Bach judged it so.  As a snapshot of Bach the improviser, it has human interest to be sure, but it is unfinished composition at best.  It is amusing to hear it as a once-only curio, but to offer it as a viable substitute for what Bach offered as representative of his best and most fully elaborated work is manifestly to devalue both that work and the critical sensibility that impelled its revision....  I see here the ultimate perversion of the idea of authenticity:  the elevation of what amounts to a rejected draft to the status of a viable alternative—and even a preferable one—because it is earlier, more in keeping with ex post facto historical generalizations, and less demanding on the listener.  [pp. 192f; later included in Text & Act, p. 139]

Indeed, as Taruskin goes on to insist, the label Brandenburg on the Hogwood (and Dart) recordings is false advertising, as that can only be used to describe the texts contained in the dedicatory score.  Certainly those Leipzig copyists would have been mystified if someone requested one of Bach’s Brandenburg works.  Was ist das?  The alternative texts really ought to be called something different.  “Dart’s Brandenburgs” don’t exist.