59. The latent textual value of a derivative score

A few weeks ago, my colleague Justin Vickers posted a very generous profile of my blog.  Speaking with Justin a few months ago, I gave him an example of material that has been on my idea list for posts even before I posted the very first one in the summer of 2016.  These ideas sometimes linger while I wait to find the right examples to illustrate the issue at hand.  One of those ideas—the second on the list, actually—was about the importance of vocal scores to the textual record.  (A vocal score is an edition of a vocal or choral work in which the orchestral accompaniment has been “reduced” so that is reasonably playable by one person at a piano; as opposed to a full score, in which all the orchestral lines are spelled out—but there are varieties of full score too numerous to discuss in this post).  Knowing that Justin is a Britten scholar, the example I shared was Peter Grimes.  But now is a good chance to finally write this post (not as post number 2, but as number 59), and thus to back up and look more into the Peter Grimes instance.  And, as so often in this blog, there is a long backstory here.  If you want to cut to the chase, skip the indented paragraphs and I’ll see you on the flip side.

Decades ago, during my graduate studies, the music seminar room doubled as the 

“locked press” of the music library.  In my very first blogpost, I described my favorite space in the Cornell music library before the building was renovated:  it held the M3s (collected editions) and the ML134s (work catalogues).  The locked press inspired a certain awe, as it held the stuff that could be accessed only with special oversight—the holdings that would be only rarely consulted and which would be impossible or expensive to replace.  I remember the bewildering array of facsimiles (in those days before digital scans were so readily available online), and sometimes during a class my eyes would wander up onto the shelves to behold these special volumes.  I remember one catching my eye, noteworthy because it wasn't a facsimile of a medieval or renaissance source, nor even of eighteenth- or nineteenth-century manuscript, although it was as impressively lavish in its presentation.  Two big volumes:  The Making of Peter Grimes:  I. Facsimile of Benjamin Britten’s composition draft and II.  Notes and Commentaries.  I knew very little of Britten’s music at the time; I knew the name of the opera, but I had never seen or heard it.  I knew it was an important work, but I was still surprised to see that something so recent—hardly fifty years old at the time—would merit a place in the holy of holies in the music library.

You should keep in mind that I was a snob of the first order at that time.  I had very strong opinions about what was worth looking at and what was worth listening to.  When in the late 1980s I came across the Dover reprints of full scores, what interested me most was that they were full scores (i.e., what I assumed to be “everything the composer wrote”), not really anything about what editions were being reprinted.  With my interest in Gilbert and Sullivan, it was a little frustrating at the time to be pretty much limited to vocal scores, but I still generally held vocal scores in contempt—at least at one remove from the real thing, at best.

I occasionally requested something from locked press, and those seemed like particularly momentous times.  In fact, almost soon asI arrived on campus campus I called up the facsimile of The Mikado, spending several hours at my carrel poring over it.  I perused several other locked press items over there years, but never Peter Grimes.  That wasn't until many years later, by which time I had come to know the opera pretty well, and was devoting much of my energies to British music generally.  So, in the spring of 2013, when I had some faculty development funds to spend, I splurged on several facsimiles:  Bach’s Orgelbüchlein, the composition score of Messiah, and a few others.  But also, seeing a chance flyer from Boydell & Brewer, I ordered The Making of Peter Grimes.  I spent a lot of time with it over a few weeks, looking at the score with the full score at hand, and reading the commentaries.  But then it went up on the shelf and didn't often come back down.

Fast forward a few years, when I was considering using something from Peter Grimes for one of my classes to analyze.  Finally over my full-score-only snobbishness, I had recognized that it would sometimes be easier for my students to see what I wanted them to see if they could use a reduced score of some sort (like Ravel’s two-piano transcription of La Valse... and, I thought to myself, at least the transcription is by the composer...).  So I went down the hall and borrowed the vocal score of Peter Grimes from one of my applied voice colleagues.

I had never looked at this vocal score before, but flipping through it, I kept thinking I’ve seen this before.  And I knew exactly where I had seen it.  It looked a lot like the “composition draft” that was in the facsimile on my shelf.  No, it’s not note-for-note the same—hardly that.  But what the vocal score transmits is derived not from the full score but from the composition draft.  The draft is written in “short score,” meaning lots of stuff crammed in a few staves, but generally not really conceived as a piano accompaniment.  Maybe because Britten was himself a formidable pianist, the notation of the short score betrays to my eye a sort of “pianistic thinking” which transferred nicely to the clean vocal score text prepared by Erwin Stein.  Looking at it now, it is actually hard to find a really good example to demonstrate this.  The section that had first attracted my attention was the interlude that opens Act III; the example below is from the storm interlude in the middle of Act I (the beginning of which has always struck me as prefiguring some of the “rumble” in West Side Story; Bernstein had conducted the US premiere of Peter Grimes in 1946.  Just sayin’...).  These are some of the most fascinating pages in the composition draft:  they reveal a complicated gestation, as Britten was asked to lengthen it, adding and revising material in the middle of the movement.  I think this is sufficient example, though, of how the compositional draft is taken over into the piano reduction with little more than cosmetic retouching (as well as spelling out everything that is otherwise indicated with ditto slashes).

 

SOURCE: Britten: Peter Grimes, Interlude II (Act 1); composite of (l.) p. 110 of the published vocal score; and (r.) detail of f. 35r from the composition draft, scanned from the Boydell facsimile.  The material of the composition draft begins in the second bar of the vocal score page (indicated by the red bracket).

There are even times when Britten brackets the upper two staves as “I” and the lower two as “II”—which certainly suggests thinking of it as two pianists.  Stein's reduction even sometimes asks for two pianists, but it interesting to note that these do not really match up with the times Britten writes that way.  The great Passacaglia in the middle of Act II, for example, appears in the composition draft on no more than three staves (though Stein expands it to full piano duet).

What this suggests to me is that the heart of Britten’s ideas of the piece are more easily seen in the notation of the vocal score than of the full score.  Moreover, since this was Britten’s regular working habit—short score before full score—maybe the vocal scores of his other works would similarly convey the essence of the work in ways that might be obscured in the full score, especially such complicated and overwhelming scores as War Requiem.  When Imogen Holst prepared the War Requiem vocal score, she relied on the compositional draft, and Philip Reed has made this point before (and doubtless others have, too):

“As with the vast majority of Britten’s output, the War Requiem took shape on the manscript paper as a through-composed short score draft written in pencil throughout, with the orchestral texture reduced onto two, three, or occasionally four staves and the vocal lines occupying their own staves.  At first glance the draft resembles something approaching a vocal score and was certainly used as a guide by Imogen Holst when she prepared the vocal score (under Britten’s supervision) for Boosey & Hawkes.  The instrumentation is indicated by verbal abbreviations—‘str’, ‘trbn’, ‘ww’—at the time of composition, ready for instant retrieval when the moment came for the full score to be made.  This simple technique was effective in allowing Britten to press on to the end of a work before making the full score, safe in the knowledge that the piece was in effect written; in the case of complex works, such as the War Requiem, it also allowed an assistant to follow behind the composer using the draft as a basis for the all-important vocal score from which the soloists and chorus would learn their music” (37).

And to add another nuance:  my guess is that greater fluency with the published reductions in the vocal scores of Britten’s works would yield similar fluency with reading his compositional drafts.  I think that it would give a greater facility to perceiving aspects of his compositional process than the rest of us can readily discern.

On this practical matter of needing a piano reduction of the orchestral material for rehearsals, in other research before I started this blog I had a few examples where the composer would depart from his [sic] usual routine in order to produce a vocal score first--because it would be needed sooner than a full score.  When Charles Villiers Stanford wrote his Songs of the Sea in 1904, he seems to have first thought of it as an orchestral song cycle for one singer, and started work on the full score; only later did he decide to add a men’s chorus, and then set to work on the vocal score.  When, in 1910, he followed it up with another cycle, Songs of the Fleet, he knew from the first that it would feature a mixed chorus along with the baritone soloist, so he started with writing the vocal score (with a piano part) before then turning to the full score.  All four of his autograph manuscripts for these two pieces are preserved at the Royal College of Music, London.  (What do we make of the well-established Stanford having to prepare the piano reduction rather than it being hackwork for someone contracted by the publisher?  I’m not sure.)

Another example:  William Walton was not much of a pianist himself, and so generally composed in full score (but, if the sketches that remain for the Variations on a Theme of Hindemith (1963) are typical, writing in concert pitch before producing a fair copy with the correct transpositions); in the case of the Cello Concerto (1955-56), however, Walton produced a keyboard “reduction” first.  (Can you “reduce” something that doesn’t yet exist?)  In the case of his chamber opera The Bear (1966–67), Walton worked on the piano score in tandem with the vocal score, apparently using the discipline of working out a piano score as the opportunity to test the worth of what he had already written.

But perhaps the greatest musicological value of vocal scores must be the preservation of material subsequently cut and lost in other sources.  Among the treasures on the IMSLP (and there are many!) is a scan of the two early Choudens vocal scores of Bizet’s Carmen.  They prove to be from two different issues:  the Eastman copy includes the recitatives composed by Ernest Guiraud, documenting the early reworking of the opera—and thus an early exemplar of the performance tradition that dominated during the first century of Carmen performances; the Chapel Hill copy is a true first edition, without recitatives (although the scan missed a few pages—most frustrating!).  But what is so important about this edition is that it contains the “Scène et Pantomime” that initially followed the opening chorus but was cut during the first run.  The autograph for this portion was removed from the rest of the opera, although it survives now in a private collection.  Fortunately, it has since been published (as in the 1992 Schott/Eulenberg score), but for a long time the first edition vocal score was the only available source for it.  Of this score, as Winton Dean notes, the 1875 vocal score is the authoritative source in determining the provenance of the early changes to the score:

“No retouch to the autograph, whether it is in the hand of Bizet, Guiraud, Antony Choudens, General de Gaulle, or the Archbishop of Canterbury, can by any human possibility have got into the 1875 vocal score after Bizet’s death or (barring misprints) without his approval.  All such modifications to the autograph must have been made in conformity with Bizet's wishes as expressed in the vocal score.”  (p. 289)

(Note that the Eastman copy indicates at the end of the opening chorus to follow directly with “no. 2,” but there is no “no. 2” in that source.) If you're interested, Richard Langham Smith (pp. 161–89) argues persuasively for the importance of this pantomime scene in understanding the work, even if it does seem to me like a speed bump in a performance.

One other example of the vocal score as an important conduit of evidence:  this spring I was teaching an American Music course, and for the class session when were talking about late eighteenth- and early nineteenth-century American hymnody, I wanted them to look at a source I remembered seeing years ago (via a da Capo press reprint, but now of course more readily available as a scan online).

SOURCE: cropped screenshot of the title page of IMSLP 242918, the first edition (a copy at Eastman)

The reason I remembered looking at this source some thirty years ago was because of that line in the subtitle “together with many beautiful extras from the works of Haydn, Mozart, Beethoven, and other eminent modern composers never before published in this country.”  One such tune that had struck me as a particularly odd choice and thus memorable decades later.

SOURCE:  cropped screenshot of p. 180 of IMSLP 242918

I recognized this melody from Act I of Mozart’s opera La Clemenza di Tito, and it struck me as strange that it would show up here.  As was customary in such sources, the melody is presented in the staff directly above the bass.  (Lest this be thought the tenor—one of my pet peeves—please note that the first hymn in this collection explicitly labels the top staff as “TENOR” and the third as “TREBLE OR AIR.”  The fact that tenors may sing this melody is another issue altogether and has to do with the practical use of these tunebooks, not the initial intention they present.)  

But this spring when I looked at it again, what stood out was that first triplet in the melody (second system, first downbeat).  I’ve never heard that before, I thought.  Where did they get that?  Out of curiosity, I looked up Mozart's autograph.

SOURCE:  cropped and marked-up scan of the autograph score, from IMSLP 807497 (p. 42); for reference, this is p. 74 as presented in the Bärenreiter facsimile of the autograph, which re-assembles portions held in different collections.

We are in a different key (A major) and in different clefs (the upper staff is the bassoon in bass clef, the lower staff is Annio, in soprano clef), but the reading shows two downward triadic arpeggios, a tonic and a supertonic chord.  None of that 8-7-8 neighbor-tone motion in the hymnbook.  Were the editors of the hymnbook trying to make it easier to sing?  (But if so, why include this tune at all?)

As the IMSLP had a scan of an early vocal score, I clicked on that.  Lo and behold!

SOURCE:  cropped marked-up scan of p. 25 of IMSLP 572308, a vocal score published by Simrock c. 1800

There was the Boston reading, clear as day.  So where did this come from?  The piano reduction in this source was credited (even on its title page) to Christan Gottlob Neefe, and it must have been one of the last things he did, since he died in 1798.  Was this figuration Neefe's idea?

I looked in the critical report of the Neue Mozart Ausgabe.  Crickets.  Not only was there no mention of the Simrock vocal score (and indeed very little reference to any of the early vocal scores), but there was nothing flagged about this passage.  So I pulled out the new edition of the Köchel Catalog.  Neefe's vocal score was not listed.  Instead, I found two 1795 editions of the vocal score:  one was published in Leipzig by Breitkopf, with a keyboard reduction by Siegfried Schmiedt; the other in Hamburg by Günther & Böhme, with a keyboard reduction by August Eberhard Müller.  I found a scan of the Schmiedt score right away; the only digital scan I could locate of the Müller reduction was from a subsequent Breitkopf score (c. 1803), but in any case both of them gave the same reading as the autograph:

SOURCE:  composite, marked-up, cropped scans of early vocal scores:  (top) p. 24 of Breitkopf (1795) available at archive.org; and p. 17 of Breitkopf (c. 1803), available at Harvard University Library.

I looked at other published vocal scores and manuscript copies; I couldn’t locate any other source that transmitted Neefe’s reading (assuming he originated it), but I was only looking at scores of the complete opera.  Maybe Neefe’s reading made it into some other excerpted anthology that I haven't seen.  Otherwise I can only conclude from all this at the moment that a copy of Neefe’s edition was the one that the Boston editor(s) referred to in preparing the Handel & Haydn Society Collection.  Not much, but it is gratifying to have at least something other than a wild goose chase.

ADDENDUM 5 July 2026

A fantastic recent example of the value of a derivative sources is Marc Shepherd's discovery of three cut numbers from Gilbert and Sullivan's Iolanthe, extant only in a manuscript copy of the full score; the score was under wraps in a D'Oyly Carte vaults for almost a century, and then languishing in the British Library until Shepherd paged it and saw what it contained.  It's a great story.

4. Moving targets (Episode #1)

As will be gathered from the title, I will be doing an occasional series of “moving targets” posts.  In these posts, my targets are not the musical works so much as the editions of them—editions which change while under cover; sometimes under the same physical cover without any notice from publisher of the nature of the changes.  Sometimes the ISBN or ISMN remains the same, so that the alterations might only be discovered by accident when they are not advertised as “corrected.”  But even when they are so advertised, it isn’t always immediately clear what we hold in our hands.

For a while in the second half of the twentieth century, the concept of the Fassung letzter Hand held sway in scholarly editions of music, particularly those of 18th and 19th century music.  To a certain extent it still does.  It doesn’t really translate well:  “the version of the last hand” doesn’t convey much.  It really means something more like “the last authorized version.”  The concept has even been stretched to mean the version the composer indicated in some way that (s)he preferred, even if (s)he never documented it precisely.  (A good example of this is William Walton's opera Troilus and Cressida.  The version that appears in the William Walton Edition is not one the composer ever heard or even put down on paper, but it is consistent with the comments he made about the 1976 Covent Garden revision:  he preferred the cuts, but he didn't like the transpositions and other alterations to accommodate Janet Baker's lower tessitura.)  And of course the idea of “new and improved” is a good marketing tool for selling another copy of something to someone who already owns it.  So even if some musicologists have moved on to “process editions” which document a piece in various stages of its existence, commercial publishers love it when a composer’s final thoughts can boost sales in an already established work.

In the course of looking into something for quite a different post appearing months hence, I stumbled across an interesting example of this.  I knew already that Leonard Bernstein's Candide has a complex textual history.  The show has been different things at different times, not only with lots of music scrapped and then resurrected (sometimes with entirely new lyrics, assigned to a different character), but the whole book replaced and then later patched and re-patched.  A good sign of this textual confusion is the copyright page of the only published full (i.e., orchestra) score of any version of the show—billed on its title page as “”SCOTTISH OPERA EDITION OF THE OPERA-HOUSE VERSION / (1989).”

And of course this doesn’t account for the further vicissitudes this work has undergone since Bernstein’s death in 1990.  I don’t expect to live to see a comprehensive critical edition of Candide.  It would probably take at least four volumes:  the 1956 version, the 1973 version (in which Bernstein took no active part, but which had great consequences for the subsequent manifestations), the 1988 version, and a huge appendix of all the other material, including details of at least the 1971 and 1982 versions.  The only project I have run across that seeks such comprehensivity in a work with such textual complications is this edition of Bizet's Carmen, but I don't know how much of it has actually come to fruition, as all of the information detailing what it was supposed to encompass has disappeared from the web.  [Here are scans of a few pages I downloaded a few years ago.]

Regarding the 1994 full score, the Bernstein website acknowledges:

While this publication encompasses the complete score, it by no means reflects a final, frozen show. Like its hero, Candide is perhaps destined never to find its perfect form and function; in the final analysis, however, that may prove philosophically appropriate.

Fine.  But Boosey & Hawkes seems to want to have it another way.  In this self-same 1994 score, we find the following notice:

SOURCE:  detail of scan of p. [i] of 1994 score

And yet, when the overture was included in Leonard Bernstein Orchestral Anthology, vol. 2 (1998), we find a notice that at first glance would amount to pretty much the same thing:

SOURCE:  detail of scan of p. viii of 1998 score

These two scores cannot present identical readings, however, because the instrumentation is substantially different.  The 1998 score presents what must be called—although it isn’t anywhere that I can see—the “concert version” of the work, scored for full symphony orchestra (basically triple wind with three percussionists, in addition to timpani).  This scoring seems to be Bernstein's own, as here is the first page of his manuscript (according to the Bernstein website).  This orchestra is too big for a theater pit, and always was.  The original 1956 Broadway recording is for a smaller ensemble, and from my listening it seems to correspond closely to scoring of the “Scottish Opera” version.  (The two aren’t quite identical, at least to my ear:  the most prominent differences are in the percussion.)  The reduced scoring—and reducing also the expense of remunerating musicians—cuts and redistributes the music of one flute, one oboe, two clarinets, two bassoons, two horns, a trombone, and a percussionist.  That saves 10 players, not counting shrinking the string sections too.

Listening to the 1989 recording, one finds that Bernstein actually used the full symphonic scoring for the overture, and the reduced scoring for the rest of the show.  (Just listen to the percussion at the beginning.  If in doubt, watch the live performance given a few days before the recording was made.)  Thus, despite the apparent contradiction because of the different scoring, both statements reproduced above could be true, as both texts do relate to the 1989 recording.  But it seems more complicated than that.

Although the 1994 publication is the only edition of the orchestral score of the complete show, the overture has been published in score three other times (not counting arrangements for band, etc.): in 1960 (G. Schirmer), reprinted with a few alterations in 1990 as “corrected edition” (Boosey & Hawkes), and newly computer-generated in 1998 as “corrected edition” (in the Boosey & Hawkes Anthology mentioned above).  I do not want to bog down this blog with all of the textual variants of these three editions, but the findings of my somewhat hasty collation of these sources are here, for the most indefatigable of readers.

A few of the most audible differences are worth mentioning.  One is the tempo:  the 1960 version is quite a bit faster (half-note = 152 at the beginning; whole-note equals 96 at the coda) than the subsequent editions (half-note = 132 at the beginning; half-note equals 152 at the coda), although the original metronome marks appear in the 1976 vocal score of the 1973 version.  This slowing down of the whole piece is consistent with Bernstein's own recordings.  His recording with the New York Phil made on Sept. 28, 1960 is at the tempi published in 1960; his 1989 London Symphony recording is at the tempi published in 1990.  In this respect at least the scores published in the 1990s can be said to reflect his performances in 1989.

This is not always the case.  The 1998 score (which claims to reflect changes made in 1989) includes an interesting change made at some later point.  Bernstein's performance 1960 recording, 1989 recording (and indeed the 1989 concert performance) all give this reading for the cymbals.  Note particularly the two clashes in b. 51

SOURCE:  detail of scan of 1960 G. Schirmer score, p. 10, bb. 47-53

The 1998 score—and only that score—prints this:

SOURCE:  detail of scan of 1998 Boosey & Hawkes anthology, p. 9, bb. 47-53

Sensible though this is (i.e., aligned with the brass pattern), it is nowhere to be heard on Bernstein's recordings, nor indeed on any recording I have located before the 1997 “New Broadway Cast” recording.  It is not in the 1994 score, which claims to be based also “on the manuscripts of Leonard Bernstein at the Library of Congress.”  One wonders on what authority this change was made.   I haven’t seen those manuscripts, but even if the 1998 reading accords with the original notation, do we know enough—does anyone know enough, that is—about how the 1960 reading came about to know it does not have an authority superceding the original?  Is it an accident that has been performed faithfully for decades (and recorded by the composer), or is it the correct reading?  Are Bernstein’s recordings of it evidence of at least tacet acceptance?  And if the 1998 score is going to interpose such a change, should there not at least be some indication of how and/or why?  (The battle scene in Act I of the 1994 score (no. 5f) has at one point cymbals matching the reading of the 1998 score of the overture, and that might have been the source of the change; but I would like to have been told.)

I think the most curious variant reading is one that was revised at some point before the 1989 performance and recording.  The 1960 reading of the horn line in the final canon before the coda was this:

SOURCE:  detail of scan of 1960 G. Schirmer score, p. 38, bb. 225-230

Note the lack of a syncopation at the end of b.227—and it is to be heard thus in the early recordings, even though this departs from the strict canon with the other upper parts.  For me the puzzle is why it was ever this way at all—but it upholds the old maxim of philology that the more “difficult” reading is more likely the original.  (Who would think to modify it to this reading?)

The three scores printed in the 1990s all give the syncopated figure, yielding a more conventional strict canon:

SOURCE:  detail of scan of 1998 Boosey & Hawkes anthology, p. 34, bb. 225-230

And, until now, that was the way I thought it had always gone.


I will return periodically to instances of publishers accumulating more confusion than clarity.  That works exist in multiple versions doesn’t trouble me in the least.  Indeed, I find such differences to be of consuming interest.  But I am irked when a publisher makes a text more difficult to access, and particularly so when the information is ambiguous or misleading.  Still, it does give me something to write about.