51. Cutting out the middleman?

 Why should we assume a priori that any score prepared some centuries ago by one musician for another musician to play from cannot be used by a musician today without the intervention of someone (often a non-musician) with a Ph.D. in musicology?  Why does the score first have to be entirely rewritten in some other form and accompanied by a lengthy critical report that can only be understood by someone else with a Ph.D. in musicology?

Alexander Silbiger:  “In Defense of Facsimiles”

Historical Performance 7/2 (Fall 1994), p. 103

SOURCE:  screenshoot of Facebook group post (June 26, 2022); poster name and profile picture redacted 

Today this blog turns six years old.  My original intent was to get to about fifty posts, posting twice a month for two years.  But it has slowed to a trickle.  With a change of jobs and then the more crowded schedule imposed by a global pandemic, it has taken me six years to reach the intended scope, and yet the list of topics that I scrawled in the summer of 2016 is not yet exhausted.  Today’s topic is from that list; it came to mind again this past February with the death of composer George Crumb (1929-2022).

There’s nothing quite like examining one of Crumb’s scores—especially those which require an oversize format to be legible at all.  His published scores very often bear this notice:

SOURCE:  found in a considerable number of Crumb scores; this is scanned from the title page of Mundis Canis (Edition Peters, 2000).

With good reason:  his notation is meticulously clean, and his innovative and idiosyncratic graphic notation would be challenging for a professional copyist/engraver to reproduce (which would add substantially to the cost of publishing the music).  If you haven’t seen any of it, just google george crumb graphic notation; sure, you’ll also get some non-Crumb images, but you will quickly get a sense of what he accomplished.  (Here’s a tribute that appeared in the Washington Post and has remained on my desk for several months to remind me to write this post.)  Crumb reminded me of this long-delayed post because he effectively bypassed an editor:  he cut out the middleman, with his carefully prepared fair-copy scores becoming the unmediated text for the performers’ eyes.

That sounds like a good idea:  no interfering editor meddling with the text.  And I have sometimes taken editors to task for overreach; even for making things too easy (in a way, like subverting the teaching of cursive by putting everything in print).   But I’m not in the least ready to dispense with the what a good editor offers us.

I remember very clearly a particular moment on an afternoon in the autumn of 1997.  It was in my first semester in graduate school, and my classmates and I were in a seminar about Haydn’s string quartets.  The professor was at the time finishing up a volume of some of the quartets for JHW—Joseph Haydn Werke, the (still ongoing) complete critical edition of Haydn's works.  One very important thing I learned from that seminar was to never assume I knew “what the composer wrote,” based on whatever edition I was using.  The JHW is actually among the most scrupulous of the 20th-century critical edition projects to give a pretty clear indication within the musical text of what the principal source really shows.  The edition uses a complicated series of brackets:  < > angle brackets to indicate material which the editor has filled in following Haydn's shorthand instructions in the autograph score; ( ) rounded brackets to indicate something that is lacking in the principal source but is supplied from a source that is right below it in the chain of textual authority, like performance parts prepared by one of Haydn’s copyists; and [ ] square brackets to indicate something that the editor has added by analogy with some other passage or in order to spell out something musically necessary.  Taking an example almost at random, here is the first system of the musical text of Symphony no. 46 as it appears in JHW, although I have highlighted these distinctions in different colors:

SOURCE:  scan (detail) of the opening of Haydn's Symphony no. 46, as it appears in JHW series I, volume 6 ("Sinfonien 1767-1772"), ed. C-G. Stellan Mörner, p. 104; I have added color highlighting.

The orange passages (indicated in the edition by angle brackets) are realizations of Haydn’s abbreviated instructions for the second violins to double the firsts; the markings highlighted in blue (indicated by the rounded brackets) do not appear in Haydn’s autograph at all but are taken from the most authoritative set of parts.  (Notice that the initial f dynamic is assumed in the autograph, but made explicit in the parts—except for the horns, who maybe never have to be told to play forte.)  The green markings (indicated by the square brackets) are not found in the first or second tier of sources, but were deemed by the editor to be intended or otherwise necessary.  All of this is clearly marked on the published score itself, without even having to turn to the critical report (which gives further details of Haydn’s corrections and second thoughts, as well as other variant readings in the principal sources).

I forget what the particular issue was that afternoon in the Haydn quartet seminar, but when the class was reminded that we needed to know what Haydn wrote (not just what was presented in whatever edition we might have to hand), one of my classmates posited that the ideal music library would comprise high-resolution scans of every source:  then we could all see exactly what was on the page—not just of a composer’s manuscript, but of all the copyists’ scores and parts, and even all the early editions.  Then we would not have to take anyone else’s word for it (as even a critical commentary gives one only the information an editor bothers to relay, and sometimes it may be in error).  

My reaction at the time: Wouldn’t this just compel us all to reinvent the wheel every time?  

A few years after this conversation came sites like Bach Digital:  such portals have the potential of fulfilling my classmate’s wish, with high-resolutions scans not just of autographs, but often a range of derivative sources too—and with no limit to how many sources might be uploaded in the future.  Or take the Online Chopin Variorum Edition—a misnomer, since there is really no editorial authority producing an edited text.  These resources encourage a see-for-yourself approach, even maybe a do-it-yourself approach, and I heartily approve making such resources available beyond the walls of the libraries and other holdings in which they are found.  But do the users of such sites know how to read them?  We are making the materials of editing available without any instruction of what it means.  

In an earlier post, I strongly criticized a particular “urtext” edition of Bach’s “Goldberg” Variations because of its amateurish assumption that it was presenting the Holy Grail (i.e., Bach’s intended text), while the editor had unknowingly taken a shortcut that undermined his claim.  And the unknowingly is the issue:  when we don’t know what we don’t know, we don’t really know what we’re doing.

Take this example, from a scan of Beethoven’s autograph of his Septet, op. 20, very near the end of the first movement.  The top stave is the clarinet, and below that the horn.

SOURCE:  (cropped) scan of Beethoven's autograph of the Septet, op. 20 (held in the Biblioteka Jagiellońska in Krakow), available on the ISMLP as #774464 and #774465.  This detail  (showing bb. 271–273) comes from the upper right corner of p. 25 (which appears as scan p 15 of $774464).

The last note in the clarinet staff is clearly marked c-natural.  The natural is ostensibly superfluous, as the key signature for a b-flat clarinet in a movement in e-flat major would be just one flat.  Thus it is no surprise that this accidental is eliminated in the new Beethoven Werke edition (ed. Egon Voss, 2008). 

SOURCE: detail of BW IV/I:  Kammermusik mit Blasintrumenten, p. 25.

The critical commentary does not even mention that there was ever a natural sign there—nor does it comment on b. 273 at all.  The commentary does list as a supplemental source Beethoven’s later arrangement of the work as a piano trio, op. 38 (with the option for a clarinet to substitute for the violin), and one would expect that substantial textual differences in otherwise analogous materials would be mentioned.  But when we look up the corresponding moment in op. 38 we may well be surprised:

SOURCE:  detail of p. 95 of Beethoven: Klavier Trios III, ed Friedhelm Klugmann (Henle, 1968), showing bb. 271–273 of the first movement of op. 38.

The top staff here is the violin, which may be replaced by the clarinet on the second staff.  Note that the last note in the clarinet here is c-sharp ( = the b-natural of the violin).  Voss doesn’t mention any of this in his critical commentary to the Septet. 

For the do-it-yourself editor, this discrepancy may seem like an inconsequential difference between the two texts, and that Beethoven clearly must have meant c-natural ( = concert b-flat) in the Septet at this moment, as he actually notated the natural sign.  That is the reading transmitted in virtually every edition.  (I see that someone involved in producing the French edition (Pleyel) in 1828 caught the mistake and emended it, but that emendation didn’t catch on.)  But, looking across a much wider range of manuscripts and with decades of experience, editor Jonathan Del Mar has spotted a pattern of errors which very neatly explains this problem.  This pattern manifests in erroneous accidentals in transposing parts, and indicates that Beethoven used the “clef substitution” method of transposing.  That is, instead of thinking that, for example, each note of a b-flat clarinet part sounds a whole tone lower than notated, he merely thought of this staff as being in the tenor clef, on which he could then note that sounding pitch (albeit sounding an octave higher).  

Beethoven indisputably notated c-natural; Del Mar makes the case that he meant concert b-natural (and thus c-sharp, when transposed up a tone).  He points to exactly the same mistake in a place in the second movement of the Eroica where there really cannot be any doubt of the error.  The autograph is lost, but as the wrong accidental occurs in the closest copyist score, the surviving parts used in the 1804 first performance, and the first edition of 1809 (shown below), the mistake surely descends from Beethoven himself.

SOURCE:  detail of p. 60 of the first edition of Beethoven, Symphony no. 3 (Cianchettini and Sperati, 1809), taken from scan available at https://www.beethoven.de/en/media/view/4716012494651392/scan/61; this shows bb. 139-144 of the second movement.

Surely the clarinets and the violas should be in unison; the viola b-natural would need to be transposed as c-sharp for the clarinets, and yet the same natural persists.  Del Mar has observed the same pattern with horns, for example the dubious f-flat that occurs in the fourth movement of the Septet.  Using the clef-substitution method for e-flat horn, Beethoven would have imagined a bass clef on this stave, and the flat is thus coming from the intended sounding a-flat.  Again the argument is that what Beethoven meant is not what he wrote (which should have been f-natural), but we need an experienced editor to recognize that; a facsimile won’t clue us in.

SOURCE:  (cropped) scan of Beethoven's autograph of the Septet, op. 20; This detail  (showing bb. 100–102 of the fourth movement) comes from the upper right corner of p. [52] (which appears as scan p 7 of IMSLP #774465).

But editors should tread lightly on this thin ice, seeking what is meant in preference to what is actually on the page.  (Query:  What did the framers of the Constitution intend?)  A notorious example of this is Jacques-Louis Monod’s recomposed edition of Arnold Schoenberg’s A Survivor from Warsaw, in which Monod altered Schoenberg’s score to accord with Schoenberg's theories... or at least with Monod’s analysis.  As James Grier cautions, "Let the editor not be accused of printing the piece the composer would have written had he or she known as much as the editor" (p. 136).  (I think what my client was meaning to say...)

Many of my posts have dealt with overly intrusive editing—which I think often comes from the temptation to prefer a variant reading just for the sake of an audible difference, maybe in some sense justifying the new edition.  (Here’s a recent post on precisely that topic.)  But I find myself torn between the two quotations with which I started this post.  Silbiger (who himself earned two PhDs) is right to champion the great value of experiencing the sources closest to the original context of the music; and I looooove facsimiles.  I am interested in seeing sources that allow me a glimpse of the composer at work—composing scores, or revisions happening on the page of what otherwise seems to be an attempt at a fair copy.  Here, for example, is Bach revising the text as he transforms the E major violin concerto (BWV 1042) into the D major harpsichord concerto (BWV 1054) in D-B Mus. ms. Bach P 234, a collective score of all seven of the extant concertos for (one) harpsichord, plus a fragment of another.  He had already copied out and transposed the solo line (which at this point features some double stops in the violin version) when he decided to rework the figuration to be more idiomatic on the keyboard.  Having run out of space, he opted for keyboard tablature below the system:

SOURCE:  cropped scan from Bach Digital of P 234, p. 48 (detail), showing BWV 1054/iii, bb. 77-87.

There are many examples of similar re-workings throughout these seven concertos—sometimes squeezing in passing tones, at other times inserting much more florid ornamentation.  There is a lot to learn from it. Werner Breig, the editor of the NBA volume of these works carefully documents the corrections and changes to the manuscript in his critical report.  I, for one, am glad to be guided by an editor who has the experience and expertise to discern the patterns behind the markings on the page as it stands; and I would certainly not attempt to perform from a scan of P 234, even the beautifully-produced Bärenreiter facsimile of it.

That is not to say, of course, that Bach intended anyone (other than himself?) to perform directly from this score.  But what of a text from the composer’s hand that is intended to be the performance copy?  This brings up the Facebook group post above:  “Why don’t composers just do their own engraving [recte computer-setting] and page layout?”  Cut out the middleman—like George Crumb, or (for that matter) Wagner.

I have learned a lot from the facebook “Music Engraving Tips” group:  there are all sorts of issues of spacing, orthography, and best-practices in preparing and printing parts, a lot of which had never crossed my mind.  But I’ve also learned that many people who prepare editions of music (whether professional or not) have very little education about music history.  I see the most elementary questions about the meaning of some mark in a manuscript or an old edition; thankfully, these queries are generally answered gracefully and without contempt.  And the group’s members are genuinely concerned with presenting the text in the way that is most immediately accessible to the user—that is, the performer.

Although an individual may interact with a musical text in many different ways, we should conceive of these as distinct roles.  There are surely very few people truly suited to fulfill all of them.  The primary roles might be delineated as: 

1. Composing:  originating and finalizing musical ideas;
2. Notating:  getting those ideas into a graphic form comprehensible by others;
3. Editing:  encompassing a wide variety of activities, but essentially amounting to tweaking the notated text (ideally in dialogue with the composer) to ensure that it presents what is meant; Ralph Vaughan Williams referred to this as “washing its face,” and his sometime assistant Roy Douglas has written an illuminating account of such activity;
4. Setting:  actually getting this text into a form that can be mechanically printed and/or distributed; and
5. Reading/Interpreting/Performing:  the “end user”(?) is presumably a performer, scholar, or listener.

Of course we might interpolate other such functions:  1a might be Orchestrating.  That in itself is a good reminder that an individual may be prodigiously gifted in one of these tasks while having no appreciable ability in another.  Certainly we expect roles 1 and 2 to go together, but we do not necessarily expect a composer to be the best conductor of their own music, though.  (William Walton, a composer I’ve done a lot of work on in my research, is a good example of someone who was generally better off leaving the conducting to the professionals.)  The Facebook post I reproduce above suggests that role 2 is essentially the same as role 4, and that role 3 might be bypassed altogether.

But at what cost?  The value of the middleman editor is to push back:  Is this what you mean?  (Or, for the dead or otherwise unavailable composer, Is this what was meant, given what we know about this composer’s usual practice?)  From my own experience with this blog, I know that I need someone else’s eyes to look over my prose.  I never publish a post without several pairs of eyes reviewing it—and always at least one of the reviewers is neither a musicologist nor a musician.  The comments and suggested revisions I get from that person are particularly helpful because they show where I have made a complete pig’s breakfast of my explanations.  I don’t always adopt the suggestions for revision—as sometimes they have completely misunderstood what I was trying to convey—but I am sure to do something to fix the problems that led them to misread it.  Here I am composing verbal text; musical text it is no different, and it can only be improved by an editorial review of some sort.

And yet, as stated above, I understand why publishers are glad to have camera-ready scores from their composers:  it saves money.  Again:  at what cost?  Here is a bit of the published score to Judith Weir's powerful one-unaccompanied-singer opera King Harald’s Saga.  In this scene, Harald’s two wives bid him farewell as he sets off to conquer England.  Weir wrote the libretto, too, and here it gives each wife the same 12-syllable text.  Weir duly sets this as a 12-tone row; presumably in order to throw the differences in characterization into sharp relief, the second wife presents the row in inversion.

SOURCE: detail of scan of  Judith Weir, King Harald's Saga (Novello, 1982), p. 6.

I have enjoyed teaching this work for many years now, and I’ve used it in classes for music majors as well as for general students who cannot even read music.  It is a fascinating work, particularly in the ways it pushes generic conventions to (maybe past) the breaking point.  And, as a meditation on the senselessness of violence and conquest, it seems always timely.  So I’ve spent a lot of time with this score.  Every time I look at this passage I am irked by the absence of a bar-line near the beginning of the second system:  the second wife should have “take / care,” shouldn’t she?  This is the only place in this section where the bar-line for the second wife is different than the first, and in every other place it seems to indicate which syllable is to be accented—namely, the one right after the bar-line.  I think this is a simple omission in Weir’s fair copy—one that may not matter, as in performance the singer may intuit what Weir was after despite any irregularity of the notation.  Or did Weir mean to break the pattern?  (I haven’t bothered to reach out to her and ask her.  It seems such a petty complaint.)  

Note that here the problem is what isn’t there:  it is not an erroneous marking (as in the Beethoven examples above), but a presumably missing marking.  I do not know in what ways the editorial staff at Novello were involved in the publication of this score, but I think someone should have caught that.  Sure, editors can and do introduce mistakes of their own.  Merely by imposing a house style, an edition can obscure a composer’s musical fingerprint (what in German is usefully called the Partiturbild).  John Eliot Gardiner’s Music in the Castle of Heaven posits that Bach's manuscript notation is expressive of performance gestures:  “There is also the bonus of his graceful and expressive orthography, which reveals the way he experienced his music and expected it to unfold—the shapes and gestures suggestion of his phrasing an motion” (p. 226).  Perhaps. For a much more nuanced argument (relating to Haydn), I recommend James Webster’s “The Triumph of Variability:  Haydn’s Articulation Markings in the Autograph of Sonata No. 49 in E Flat.”  (Webster was the professor directing the seminar discussed above, although I didn’t come across this article until years later.)

Clearly, there is no substitute for getting back into the sources as much as possible; but I don’t regard a musician as negligent if they don’t do that, nor do I regard the performer who plays from a facsimile as on a higher artistic plane than anyone else.  Maybe an analogy would be that a critical edition is like currency or a credit card in place of gold:  the edited text is a convenient substitute for the real thing that will suffice for most musicians in most situations.  But for anyone for whom the text itself is a supreme concern, a good edition is a useful guide and nothing more.

46. Look ere ye leape

For some ridiculous reason, to which, however, I’ve no desire to be disloyal,
Some person in authority, I don't know who, very likely the Astronomer Royal,
Has decided that, although for such a beastly month as February, 
twenty-eight days as a rule are plenty,
One year in every four his days shall be reckoned as nine-and-twenty.

And here we are on that leap day.  Perhaps I ought to be celebrating Rossini’s birthday while I have the chance, but this is a post I’ve had in mind almost from the beginning of this blog:  What about scores that seem to have one bar too many (or too few)?

An example that many pianists know is the extra bar that shows up in some editions of BWV 846/i, the first prelude of Bach’s Well-Tempered Clavier (Book I). 

SOURCE:  cropped from G. Schirmer (c. 1893) reprint of Czerny edition, from IMSLP #01005

This bar was introduced by Christian Friedrich Gottlieb Schwencke (1767–1822), I suppose as a remedy two faults he perceived in Bach’s text.  He seems to have wanted to make the prolonged Dominant pedal begin on a “strong” bar (the first of a group of 4 bars)—thus he needed that to fall on b. 25 instead of b. 24, and so introduced a new b. “23.”  He also took pains to eliminate the two false relations (here shown in color) in the voice-exchange.  The bracketed solid note-heads represent Schwencke’s interpolated bar.

(Incidentally, this doubly-chromatic voice-leading troubled analyst Heinrich Schencker too.  He made a big deal about the fact that in an autograph manuscript, Bach wrote stems for the bass F-sharp turned upward.  To Schencker this was conclusive proof that Bach thought of that note as nothing more than a foreground harmonization of the soprano E-flat.  I’m not convinced that the stemming was anything more than fortuitous.  Whatever.)  

The retention of Schwencke’s extra bar was codified by Czerny’s edition (reprinted and reissued by a number of publishers, and probably in print continuously to this day.)  Gounod was likely working from Czerny’s edition when he created his superimposed melody—initially as an instrumental Méditation, but subsequently texted (and forever after known as) Ave Maria.  Gounod has thus transformed this “Schwenckenische Takt” into a load-bearing bar:  it is impossible to correct it, as the climax of Gounod’s melody depends on it.  I am reluctant to claim that it does much damage to the Bach original.  It matters if you know to listen for it; otherwise it passes unnoticed.  And, as Malcolm Boyd has observed,

“Schwencke wrote numerous compositions—oratorios, cantatas, concertos, sonatas, and songs—but his most frequently performed piece of work is without doubt bar 23 of the first prelude in Book 1 of Bach’s The Well-tempered Clavier....” [p. 444]

As it happens, Schwencke is a link to another work for which editions generally seem to have one bar more than the composer intended—although in this case the variant stems from the composer's hand.  It is Mozart’s celebrated thirteen–instrument serenade, K. 361, a work with a long history of textual problems, and to which I will return in future posts.  (The Schwencke connection:  he devised a charming version for piano quartet plus one woodwind.) The superfluous(?) bar is in the fifth movement, the Romance.  Here is the relevant page of the autograph manuscript:

SOURCE:  Mozart, K. 361, v (Romance) bb. 22-30; p. 54 of Mozart’s autograph; a downloaded digital scan on the Library of Congress website; a scan of the whole manuscript is available.

This movement is in a large ABA structure, but the A section is itself a binary form (aabb).  The page above shows the last three bars of the b and the first six bars of the B.  The return to A is indicated by the instruction da capo senza repliche a few pages later.  So the moment in question here is the third bar on the scan above:  it is to be played as b. 24 twice (as it is repeated); but is it to be played again as b. 111 on the third time through—the da capo without repeats?  

Mozart’s curved bracket above and below this bar was his usual indication of a first ending, but that wouldn't work in this case, as the transition from b. 23 to b. 25 is nonsensical.  Rather these seem to be an indication to skip b. “111” and go straight to the coda.  But—crucially—these curved brackets were both smudged while the ink was still wet.  Was Mozart changing his mind?  Or was the smudge accidental?  After all, if he made these markings after composing the B section, then all the rest of the ink on this page would have already dried.

Actually, I think this is exactly what happened.  Mozart was presumably impatient to start to work on the coda:  the previous page (folio 27v, numbered 53) of the manuscript shows a tell-tale mark where the still-wet slur near the top of this page (28r, p. 54) would have set-off when the page was put down too soon on top of the other folio.  I have marked the off-set smudge on p. 53 in red, as well as a space at the bottom of the page where we might expect to see a similar offset from the slur below the contrabass line:

So, is it a problem that the lower mark isn’t there?  Maybe.  But maybe not.  I speculate that the wet ink of the lower bracket of p. 54 could well have been smudged by the bottom edge of page 53, without leaving much on the page itself.  Here is my effort at representing the pages as they would have lain together to create the offset.  For this image you must imagine you are seeing through p. 54 (and so here its image has been reversed and made partially transparent).  The top brackets coincide when placed at such an angle—possible, as the pages were then unbound nested bi-folios:

The smudged top bracket and the mark on the previous page are uncannily similar.  The most suggestive detail to me is the “1” which, I suggest, explains the blob under the set-off bracket on p. 53.  Here’s a detail—p. 53 on the top, p. 54 (reversed, to align the offset) below, with the “1” and the set-off blob circled:

As I say, maybe.  I think editors are perfectly justified in omitting this bar on philological evidence (and speculation, it must be said) like that presented above.  One of the editors of the NMA volume including K. 361, Daniel N. Leeson, has written at length about this bar, voicing his regret that he did not fight more with the general editors in order to omit it from the NMA text.  (See, for example, his 2009 summary of his decades of research into this piece.  He seems not to have noticed the off-set on p. 53, however.)  Leeson, with his co-editor Neal Zaslaw, did at least manage to get a footnote in the NMA score to the effect that perhaps Mozart did not want this bar played, and directing the user to the critical report:

SOURCE:  detail of scan of NMA VII/17/2 (1979, ed. Leeson & Zaslaw), p. 191.

In fact, the critical report wasn’t issued until 2002, and was the work of a yet another hand—Dietrich Berke.  (See his comment here.)  When the NMA score was issued as a separate Bärenreiter offprint, the Leeson/Zaslaw footnote was modified:

SOURCE:  detail of scan of Bärenreiter TP 312 (otherwise an offprint of the above; this is from the 6th printing, 2006), p. 51.

The Henle edition (2005, ed. Henrik Wiese) at least puts the bar in brackets, with an explanatory footnote:

SOURCE: detail of scan of Henle 9809 (2005, ed. Henrik Wiese), p. 45.  The relevant comment cited here describes the notation and the smudging, remarking “It is impossible to determine with absolute certainty whether this ‘volta’ applies or not, and what it refers to.  Presumably it relates to the transition from the recapitulation to the coda in mm. 111–112, so that the chords from M 24 (= M 111) give way to the entrance of the coda” [p. 77].

According to Leeson, the only edition yet to completely omit bar “111” is that edited by Roger Hellyer.  After explaining the situation in his Preface, Hellyer comments:  “If performers cannot accept my decision that I am here following Mozart's ill-expressed intentions, they are of course free to reinstate what has been played here at least since 1803, as in bar 24.”  Exactly so.

Unsurprisingly, performers have been more willing than editors to take a chance on omitting this bar.  Here, for example, is this moment in Christopher Hogwood’s recording with the Amadeus Winds.  The missing bar comes—or, rather, doesn’t come—at ten seconds into this clip, at the start of the coda:

I find this reading musically compelling, and certainly plausible as Mozart’s intention, with the suddenly reduced forces on the beginning of the coda (and the surprise dissonant harmony when we expect the full cadence).  Moreover, I think the textual evidence is suggestive enough to back it up, even if it is not conclusive.

Of course we may be deleting a bar Mozart intended, merely to satisfy our taste.  That seems to be what has sometimes happened near the end of the first movement of Beethoven’s fourth symphony.  To some ears this ending seems to have one too many bar.  Jonathan Del Mar reports that Schumann (1840) and Czerny (1853) were early advocates of deleting it, and in one early set of manuscript parts the bar has been deleted—but it is impossible to say when this alteration was made, and no other source close to Beethoven supports it.  Other critical editions have brought up this question, but the only one I have seen to delete a bar for the sake of metrical regularity is Peter Hauschild’s 1996 edition for Breitkopf [below on the right].  Hauschild’s astoundingly naive justification for relying so heavily on this single source as transmitting Beethoven’s supposed alterations:  “da es wohl ausgeschlossen ist, daß andere an Beethovens Symphonie herumkorrigiert haben”! [p. 84; “because it is surely out of the question that others would have corrected Beethoven’s symphony.”]  The irony is delicious.

SOURCE:  marked-up page scans of the last page of the first movement of Beethoven’s Symphony no. 4, op. 60:  (l.) Bäenreiter (1999, ed. Jonathan Del Mar—here from 2001 off-print); (r.) Breitkopf & Hartel (1996, ed. Peter Hauschild).

If you want to compare these in performance, compare John Eliot Gardiner (with the text on the left) and Daniel Barenboim omitting the bar, as on the right.  I suppose Barenboim and anyone else may do with the text as they see fit.  The composers themselves sometimes take such liberties.  Here is an extract from Saint-Saëns’s symphonic poem Danse macabre (1875) in the composer’s own transcription for violin and piano.  He added the bar marked in red when he produced this version; it does not correspond to anything in the orchestral version.

SOURCE:  cropped screen shot of 1877 Durand edition of violin/piano version, p. 11 (from IMSLP #33277); the extract begins at b. 340.

Liszt’s Mephisto Waltz no. 1 is a similar example, if even more complicated in Liszt’s piano version both adds and delete bars compared to the orchestral original—so that the two versions do not correspond.  Already in the first 150 bars each version contains a bar that the other lacks, and it is clear from the composer’s sketches for the piano version that he had second thoughts.

SOURCE:  marked up scan of first edition (Leipzig, c. 1862), bb. 1–32; b. 25 is new to the piano version.  Scan from IMSLP #13711.

SOURCE:  top, as above, bb. 120–136; the orchestral version has an extra bar after b. 134; below, detail of manuscript sketch of this passage, scan from the Morgan Library.

For an example where I believe a new scholarly edition is led by the early sources into an error that an older edition had set right, look at the last aria of BWV 52, “Falsche Welt, dir trau ich nicht.”  As often elsewhere, Bach did not write out the final ritornello, but indicated it merely with a da Capo instruction:

SOURCE: detail of scan of the autograph (D-B Mus.Ms. Bach P 85, f. 8v) available on Bach Digital.

SOURCE:  the same, f. 7r.

Because the final cadence is (or at least I would argue is) elided with the return of the ritornello, Bach has notated the first bar of the ritornello again before the instruction “DC.”  Consequently, he surely meant not really a return to the beginning, but rather to the second bar.  And, indeed, at the second bar we find the segno marking we would expect to see [at right].

The BG edition interprets it thus, assuming that Bach’s “DC” really meant “DS.”  The NBA, on the other hand, takes the DC literally—correctly pointing out that the earliest performing parts (which are all the work of copyists) have da Capo, following the autograph faithfully.  In fact, one of these early parts—Oboe III—has a segno at b. 2, which indicates that someone recognized the problem early on.  (The lack of a correction in the other parts does not mean that it wasn't corrected:  the mistake seems so obvious that the musicians could recognize and remember the error.)  Incredibly, there is no comment in the NBA critical report about any of this—nor about the different editorial decision that has been taken.  As it is the policy of the NBA to print in full such passages that Bach has abbreviated, a redundant bar is introduced into the text:

SOURCE:  marked-up composite page scan with detail from pp. 162 and 163 of NBA I/26 (1994, ed. Andreas Glöckner), here scanned from Bärenreiter offprint TP1290 (2007).

In my opinion, this is just wrong—and the conclusive proof for it is an overlooked detail of the autograph score.  Where the other parts are given the da capo instruction, the singer—who will have nothing further to sing—is given exactly fifteen bars of rest (after the bar in which the cadence occurs):

SOURCE detail of scan of the autograph f. 8v again, just further enlarged.

 ...yet the NBA text calls for sixteen bars of rest to reach the end of the aria (as enumerated above).  Significantly, I haven’t found a single recording that includes this extra bar.  If they are using the NBA, the musicians are deleting it.  Sometimes an extra bar is just too much of a good thing.

34. So teach us to number our bars

Today’s post marks the second birthday of Settling Scores.  I have been having altogether too much fun with it, and I’ve met all sorts of interesting (and interested) people.  Some were names I knew professionally, but very many have been entirely new.  I am gratified by the response, even if I am sometimes completely in the dark on the reasons why some posts take off and others fall comparatively flat.

Although when I started this project I had a long list of issues I wanted to cover—and that list remains long—I never imagined I would spend a post on bar numbers.  What could there possibly to say?  The bars are numbered!  End of story!  But just a few weeks after I began blogging, I knew eventually this post would happen.  It was prompted by a post on the blog put out by the G. Henle Verlag.  Henle urtext editions have dominated the market (particularly for piano students) in the USA for as long as I can remember.  You’d know those slate blue covers anywhere, even if they have updated the look a bit over the years.  Their blog comes out every two weeks, written by their house editors in rotation.  It offers a fascinating behind-the-scenes glimpse at editorial work in progress.

The post that got me thinking concerned their new edition of Camille Saint-Saëns’s marvellous second piano concerto.  (To clarify:  the edition is a two-piano version, with a new reduction of the orchestral material.)

SOURCE:  cropped page scan of https://www.henle.com/en/detail/index.html?Title=Piano+Concerto+no.+2+in+g+minor+op.+22_1355, accessed 20 July 2018

Although in the preface editor Peter Jost goes to some pains to point out that the piano reduction published as the first edition in 1868 was not by the composer (but rather his pupil, Adam Laussel), the Henle blurb above gets this wrong.

Jost’s blog post concerns the arresting opening of this concerto—a free-flowing, unmeasured prelude at first, developing gradually into more conventional Romantic virtuoso piano figures covering the whole compass of the instrument.  Here are the first three pages as they appeared in first edition of the full score:

SOURCE:  scan of 1875 Durand edition from 1995 Dover reprint.

The Durand engravers have provided the conventional full score accolade on the first page, showing the complete resources required for the work.  In the autograph, however, the first page to be in full score is the third page, at the orchestral entrance, and the preceding two pages appear very much as a separate introduction, ending mid-page with a double bar and a clearly implied attacca across the page:

SOURCE:  scans of the autograph score, F-Pn Mus. MS-488, fully available here.  In this example, I have taken the images not from the Bibliotheque Nationale site, but rather from the Henle blogpost.  This has required cropping them to display them appropriately:  Henle inaccurately represents p. 2 abutting p. 1 (as if recto facing the preceding verso), although it really should abut p. 3, as above.

Jost points out that Saint-Saëns numbered the measures of this movement, starting with the orchestral entrance.  Thus the “prelude” is unnumbered—although it isn’t entirely unmetered, and even concludes with ruled bars.  Jost follows the composer on this, yielding a movement of a prelude plus 112 bars.

The first edition lacked measure numbers, but had rehearsal letters.  Sabina Teller Ratner’s thematic catalogue of Saint-Saëns works gives the total number of measures in each movement, and thus in this case numbers from the beginning, with the last bar as number 115.  (Her bar 11 below is Jost’s bar 8.)

SOURCE scan of Ratner catalogue (OUP, 2002) Vol. 1, p. 353

I do not understand the value of Mr. Jost’s “return” to the composer’s “original numbering.”  We don’t know enough to understand whether those numbers were intended to mean anything at all.  Was Saint-Saëns making a philosophical statement about the music (as Mr. Jost inevitably is—some music designated as “preceding” the real piece)?  Was there at that moment nothing written on the preceding pages, with the composer planning to improvise an introduction based on material that appears later in the movement—eventually codifying it as text?  I exchanged e-mails with Mr. Jost in the days following his post, but came away unsatisfied.

As I see it, bar numbers serve one principal and practical function:  orienting the user in a score.  A bar number is a coordinate used to locate something.  It need not be anything else. 

For any music requiring more than one player, numbered bars are useful in rehearsal (“We’ll start in bar 63”), where the system is more precise—and arguably less cumbersome—than rehearsal letters (“We’ll start six bars before F”).  In Jost's edition, taking it “from the top” is not the same as from bar #1, and that may lead to some confusion.

Measure numbers are essential, however, in critical editions (like Jost’s) so that the editor can cite a detail in the critical commentary and the user can locate it easily.  Compare, in this connection, how the new C. P. E. Bach edition deals with the unmeasured sections of the fantasies:

Source:  cropped scan of Carl Philipp Emanuel Bach:  The Complete Works, Ser. I, Vol. 3 (ed. David Schulenberg, 2005); detail of p. 34, the fantasy from Wq 63 no. 6.

Here the first portion of the piece goes without a barline for several systems, so each system is given a letter:  bar 1a, bar 1b, bar 1c.  This illustration begins at bar 1h.  The first barline does not appear until after the 3/4 time signature, so that in this edition the bar marked Largo is still bar 1j, with the bar following it reckoned (finally) as bar 2.  (The critical commentary can thus cite a note in a specified portion of this extended bar 1.)  This method is necessarily idiosyncratic:  it works for this edition, but it would not be readily translated to another.  But it doesn't need to be:  the sole function of these bar numbers is to connect the critical commentary portion of the volume with the score, and this system works well enough.  (To be fair, Jost does employ a similar policy:  the opening systems of the Saint-Saëns are labeled with Roman numerals—like the front matter of a book—which inevitably suggests that we haven’t yet reached the “real thing.”)

It is a more honest method than, for example, Henle’s treatment of the Mozart “Modulating prelude” (K. Anh. C 15.11) which gets a new bar number for each system, despite no barlines:

SOURCE:  cropped scan of Mozart:  Klavierstücke (HN 22, ed Ullrich Scheideler, 2006), p. 66.

Glancing through their back catalogue, I see that Henle’s practice has been inconsistent.  Here is a page of K. 394 in their 1955 edition (no longer in print), and the circled bar numbers correspond with ruled bars rather than with systems:

SOURCE:  scan of p. 40 of Mozart:  Klavierstücke (ed. B. A. Wallner; Henle, 1955)

Incredibly, this same work—newly edited by Mr. Scheideler—appears in the same new volume as the modulating prelude (HN22) with the bar numbers allocated exactly the same way as in 1955, so the new volume itself is inconsistent.  The Neue Mozart-Ausgabe isn’t much better in this respect:  K. 394 is treated as above (although the Henle and NMA bar numbers do not correspond); other works in the volume, including the modulating prelude, use the a... b... c... system as in the C. P. E. Bach edition.  For a particularly interesting situation, see the NMA’s presentation of K. 284a [NMA IX/27/2, pp. 5–9]; bar “(25)” is my favorite.

Does any of this really matter?  It depends, of course, on whether a number is merely a milepost or whether it has any substantive meaning relating to the music.  Once you start disconnecting the numbers from the sequence of bars on the page you surely must mean something.  I looked to see what the Hallische Händel-Ausgabe does with those passages in the organ concertos in which they have interpolated Wolfgang Stockmeier’s “suggestions” of how to improvise in response to Handel's instruction “ad libitum.”  I, for one, don't think such interpolations belong in that sort of scholarly edition, but at least the editors had the good judgment to leave those bars unnumbered (and in small type):  Handel didn’t indicate how many bars to play, and neither should the HHA.

SOURCE:  cropped scan of a portion of the second movement Op. 7 no. 4 (HWV 309) as presented in HHA Ser. IV Bd. 8, p. 204

For comparison, here’s Handel's autograph for this section:

SOURCE:  page from Handel’s autograph for HWV 309 (Op. 7 no. 4), mvt. 2; British Library R.M.20.g.12, f. 66r

When I began work on my first editorial project—Walton’s Variations on a Theme by Hindemith for the William Walton Edition—I remember starting by numbering the bars and assuming that it would be a straightforward task (young and callow as I was).  The anxiety that awaited me!  I wanted to number the bars sequentially across all the variations.  In a way, this was a substantive statement:  it meant essentially the whole is greater than the sum of its parts.  But really there was a practical reason for this:  the critical commentary would be much harder to use if you had to keep track not only of the bar number but also of the variation number.  When I set to work, however, I found that Walton had paid no attention to the seams between the variations.  This might be because he would send off a completed variation to his publisher before starting another, but it is just as likely that he didn’t care if a complete final bar of one variation was followed by a pick-up bar of the next.  In many musical editions, “bar number 1” is the first complete bar rather than the first thing on the page—but I found I would have to count each of these incomplete tags at the beginnings and ends of variations as full bars if I wanted to have just a single numbering system for the whole piece.  It worked, but I still don’t like the look of it.

On the substantive (rather than practical) value of rehearsal marks, the words of Jonathan Del Mar are a useful reminder.  The following disclaimer can be found in the preface to each of the scores of his Bärenreiter editions of Beethoven symphonies (and a similar one for the concertos, etc.):

SOURCE:  cropped scan of p. V of Del Mar's edition of Symphony no. 9 (BA9009)

How orchestras survived for so long without rehearsal marks I can’t imagine, and at least those who attempt historically-informed-performance are not bound to historical rehearsal practices.  (The unions would never stand for it.)  I bristle against heavy-handed editing, when the editor goes out of the way to make a mountain out of a molehill.  Herr Jost’s treatment of the Saint-Saëns strikes me as just that.  Then again, this blog is made entirely out of molehills treated as if they were mountains, so I’m one to talk.