Manuscripts and Statistical Anomalies
The first indication that Schubert’s Kosegarten settings might form a unique collection of songs emerged quite by accident. In the course of conducting a broad survey of Schubert’s song manuscripts and working methods I decided to put together a statistical overview of the sources to all of Schubert’s surviving Lieder, some 625 in all. The method of cataloging relied on the three fundamental types of song manuscripts widely used in Schubert studies: sketches, rough drafts, and fair copies. Sketches (Entwürfe) are incomplete renderings of songs, either in the extent of melodic material or accompaniment that Schubert put to paper or, though they appear to be fully scored songs, only a portion survives. This latter type is more properly called a “torso”; approximately 8 percent of Schubert’s songs have a manuscript in this category. Rough drafts (erste Niederschriften) are completely rendered songs that are performable but have a hasty appearance and often contain corrections; this is by far the largest group, with about 54 percent of his songs represented. Fair copies (Reinschriften) are carefully prepared presentation copies of songs that Schubert wrote out in a clean, careful hand; this is a much smaller group, about 23 percent. There is also a fourth category that is often overlooked: songs for which we have no manuscripts in Schubert’s hand; this accounts for a surprisingly large number of works, 142 in all, or 23 percent of his songs.2
I was specifically looking for groupings within these three main types of manuscript, especially among the fair copies, relationships that might reflect unknown larger-scale organizations of songs. The documentation of Schubert’s life is so incomplete that it did not seem implausible to speculate that even today some of his musical projects may have been left unnoticed in the historical record. The information at first seemed haphazard, but the survey started suggesting some patterns when I began matching the percentages given above with poets whose works Schubert had set. During the course of his short life the composer turned to texts by 98 authors. The vast majority of these writers (83) were set fewer than ten times, and only seven were set twenty or more times, as can be seen in Table 1. The large number of works set to Goethe, Mayrhofer, and Schiller reflects an ongoing engagement with these poets’ works throughout Schubert’s composing career, as can be seen in the dates of composition. For Müller the concentration centered on his last years; Matthisson, Hölty, and Kosegarten held Schubert’s interest only in his highly productive early period.
Table 1. Schubert’s Most Frequently Set Authors of Lied Texts
These numbers speak for Schubert’s sometimes shifting literary taste, but—as already intimated—there is yet another way to measure their importance in his output: by examining the number of fair copies he produced for each. Fair copies required more time to prepare than rough drafts and were usually put together for a particular reason. These were the patterns I was looking for, and I could largely explain the results. In 1816 Schubert planned to bring out a series of volumes of German songs for distribution and publication. The first two volumes contained a total of 24 songs by Goethe but was met with rejection by the great poet and the project was called off.3 Fair copies of Mayrhofer settings over the years make sense, as he was a close friend of the composer. The Müller fair copies were compiled in the preparation of Winterreise.4 But what was the motivation for the abundance of Kosegarten in fair copy, second only to Goethe? This anomaly was even more glaring considering that Schubert copied out virtually every Kosegarten setting he ever wrote (95 percent, last column). No poet—not even Goethe—came close to this type of treatment. A glance at the numbers, then, suggests a short but intense engagement with this writer that may have led to the production of a collection of fair copies. An odd conclusion, given that the research on Schubert’s Kosegarten songs suggested no such large-scale project.
Piqued by this observation, my thoughts eventually moved from statistics to a consideration of what might explain the anomaly. Curious, but with no expectations, I started examining the evidence. The more I looked, the more I found, and as one clue led to the next it became increasingly obvious that something important was emerging from the data, a conclusion that seemed too significant to have remained hidden all this time: that the Kosegarten songs constituted a cycle. In the months that followed it was necessary to test the veracity of such a possibility, but the more I shook the foundations of this notion, the more the pieces fell into place.
The logical place to begin an investigation was with the actual autograph scores. This in itself presented some difficulties, since making sense of the fair copies also required understanding the rough drafts on which they were based. This doubled the number of manuscripts that had to be considered from 11 to 22. A further complication was that, of these 22 documents, 14 were held in public archives in Austria, Germany, the UK, and the United States, 6 were in private collections in unknown locations, and 2 had disappeared entirely.5 It would be no small task simply gathering the information. Nevertheless, work progressed and a number of important observations came to light. The first concerned the dating of the manuscripts. If the fair copies had been compiled as a set, it would make most sense if this process had been undertaken for all of the songs at about the same time and after all of the rough drafts were finished. The dating of the rough drafts was largely unproblematic, as in most cases Schubert himself provided this information in marginal annotations. A look at the dates of these drafts (Table 2) reveals the remarkable intensity of the output of these songs, most produced in conjunction with other Kosegarten settings over only a day or two. After producing thirteen such songs over the course of roughly the month of July (including five Kosegarten songs on 7–8 July and five on 25–27 July), Schubert stopped until three months later, when, on a single day, 19 October, he wrote the remaining seven.
Table 2. The Rough Drafts to Schubert’s 1815 Kosegarten Settings, by Date
Dates are noted by Schubert on the manuscript except for those given in brackets or parentheses; those in square brackets are taken from the Witteczek-Spaun Collection (A-Wgm) as given in the Deutsch Catalogue, and the date in parentheses is inferred from a dated composition on the same leaf. The sigla shown in the Rough Draft column stem from RISM, the international index for music sources.
Figure 1. Cutting and folding of paper into a bifolio.
Establishing the dates of the fair copies proved far more elusive. Schubert merely indicated the work’s title on these documents. Solving this dilemma required turning to an arcane branch of musical research: paper types. Understanding this method of dating requires some knowledge of paper manufacturing in Schubert’s day. The process involved dissolving cotton fiber in a chemical mixture that broke down the bonds between the strands of the material into a gooey broth. The papermaker then needed a large porous screen made of metal wire arranged in a crisscross pattern and held in place by a wooden frame around the perimeter. The screen was first dipped into the mixture and then, while being held flat, carefully lifted out of the vat. As the liquid ran through the screen and over the edges, the cotton would immediately start to re-bond in a flat layer on top of the screen. The screen with the material forming on it would be pressed to free up excess water, the wet paper lifted out and laid on racks to dry. The large sheets were eventually cut, often along one axis, A, and then folded along another axis, B. The result was a “bifolio,” consisting of a folded sheet forming two pages and four sides (see Figure 1). If the paper was to be used for music notation, a device called a rastrum was employed to create the lines of the staves across the width of the page.
Fortunately for historians, paper production was a process that led to slight variations in the finished product. Paper mills commonly left their imprint on the paper by hand-weaving letters or symbols with copper wire into the dip screen; when the paper was pressed, water would form along the edges of these shapes and leave behind a gray shadow inside the paper that would become permanent. This “watermark” functioned as a type of stamp of origin that could be trace
d back to the manufacturer. What is more, as screens had to be replaced after extended use, the watermark wiring would usually differ ever so slightly from its predecessors, a natural consequence of the manual process involved. These slight discrepancies can be very important in isolating unique batches of paper. Other differences can be noted as well. The process of cutting the paper rendered the bifolios (and single leaves) oblong or upright in orientation. Different formats could be used for adding staff lines (usually 12 or 16 to a page), which in turn determined the dimensions on the page—a measurement known as the “total span,” from the top of the top staff to the bottom of the bottom staff. Thus Schubert’s manuscripts can be examined and categorized according to their physical characteristics. Since the composer tended to use up piles of paper in sequence, we can plausibly group works with the same traits to the same period of notation; and if several of these manuscripts are dated within a specific time frame, it is reasonable to assign undated works on the same paper type to the same period. Even if exact dates are not possible, these groupings at least allow us with certainty to exclude the possibility of manuscripts being written before their respective types appeared. 6
Applying this method to the Kosegarten manuscripts revealed three distinct types of paper (see Table 3). Type 1 appears in the Kosegarten songs set in June and July, Type 2 in the October songs, and Type 3 for all of the fair copies. By identifying dated manuscripts from each of these types, we can isolate the time period in which they were notated. For the dated rough drafts this is clear; for the undated fair copies, related documents suggest a time frame around April 1816.7 In other words, Schubert returned to all of the Kosegarten songs a number of months after the last of them had been composed in order to copy them out again in a clean hand. He did so on the same paper and likely within a short period of time.
Table 3. Kosegarten Settings, Paper Types, June 1815-April 1816
Four leaves of rough drafts, D221/233, 227, 230/231 and 313, have not been examined; the watermark to D236/237/238 is illegible; and D241 is notated on older paper. Fair copy leaves not examined: D236/237, 238, and 315/316. “Kiesling” and “Weilhartiz” refer to two paper manufacturers whose names appear on the respective watermarks. “Type” is used for illustrative purposes. For a more detailed treatment, see the critical edition of Franz Schubert’s works that uses a different classification: Types 1 and 2 listed here would be included in “Type II” and Type 3 in “Type I.”
It was beginning to look like the fair copies constituted a set of some sort, but the next question loomed even larger: Why were they compiled? The historical record provides no indication of a special occasion or a dedication that would explain such a major project or singular event. My investigation might have ended there, were it not for a curious development. In collating what I could of the fair copies, I noticed numbers in the lower left-hand corner on all of them, numbers that suggested a sequence from 1 to 20 that had nothing to do with the order of composition. Perhaps this could explain the existence of the fair copies. If so, the questions had shifted: Who had numbered them and for what purpose?
Cracking the Code
To answer these questions I needed to return to the manuscripts and take a closer look. In doing so, I found that, for my purposes, the most important information regarding the documents lay not in the composer’s own notations but rather in the markings left behind by others. Figure 2 shows the autograph of the rough draft of Schwangesang and divides the inscriptions found on the page into eleven elements: 1–6 stemming from Schubert and 7–11 attributable to later hands.
Schubert’s inscriptions:
1. The score is quite legible and represents a complete rendering of the work. The noteheads are mostly round and clearly positioned; the text underlay of the first verse is complete; dynamic markings, phrasings, and accents are also inserted. Corrections are infrequent, though revealing. In measure 4 the voice line half note on the third beat is changed to a fermata quarter to allow for the anacrusis to the next measure. Note that the piano right hand shows no such correction (quarter note fermata following long quarter rest), implying that the voice line was notated first. A similar change in the vocal line in measure 14 (last measure in system 3) demonstrates the same characteristic. A cross-out in measure 15 shows the composer simply finding a better voice-leading solution for the descending seventh in the piano left hand.
2. Schubert carefully wrote out and underlined the title toward the center of the page.
3. Schubert notes the exact day, month, and year of the manuscript, a common procedure during this period for this type of autograph, though it was a habit he would later largely abandon. Schubert wrote out eight Lieder that day, seven of them set to texts by Kosegarten.8
4, 5. Instrumentation (“S[ingstimme]” and “P[iano]F[orte]”) and clefs, tempo marking, time signature, and key signature are all provided at the beginning of the work with the key signature and clef repeated in each system. This is hardly an unusual procedure in a musical composition, yet not always the case in Schubert’s less polished song scores.
6. Schubert identifies the author of the text, “Koseg[arten],” at the end of the score and makes a note of how many additional verses not written out here should be appended to the strophic song (“6 Str[ophen]”).
Later inscriptions (not in Schubert’s hand):
7. A reference to a copy of the song “in Franz Schubert’s hand” listed as “[Number] 17” of Series “Abth[eilung]” IV.
8. A listing of songs from the “Repertorium” with similar titles, giving text, author, publishing information, publisher or plate number (“3523”) and location or catalogue number of these related manuscripts. Idens Schwanenlied, for instance, is listed as “No. 32, Series II.”
9. A number in the lower-left-hand corner in black ink (“24”).
10. A temporary plate number (“16755”), though not yet engraved (“noch nicht gest[ochen]”).
11. Various markings in various colors to the left of the title.
Though these marginalia at first seemed like little more than cryptic scribblings, curiosity drove me to wonder what they meant and who had gone to the trouble of adding this information. The references to plate numbers in items 8 and 10 indicated the work of someone in the publishing industry in the nineteenth century, but item 9 turned out to be far more puzzling. This number in the lower-left-hand corner turned up in dozens of manuscripts, yet clearly did not indicate plate numbers. Intriguing connections seemed to emerge between the seven Kosegarten rough drafts written on 19 October, with 23, 24, 26, 29, 30, 31, and 32 appearing in this position, whereas only two of the five songs from 7–8 July were even numbered (8 and 9), and the numbers from 25–27 July read 12, 13, 14, 1, and 2.
Figure 2. Inscriptions on rough draft of Schwangesang (D318).
The key to understanding the mysterious numberings turned out to be not the numbers themselves but the references to them provided in item 7. In the case of this manuscript of Schwangesang, the annotator provides the following information, as mentioned above: “A copy (or fair copy [Reinschrift]) in Franz Schubert’s hand see 17 Series IV.” There is in fact a fair copy of this song, Schwangesang, in Schubert’s hand; see Figure 3. Fair copies such as this resemble the rough drafts, but a few differing features stand out. The musical notation is more carefully prepared, with rounder note heads; there are more musical performance indications, and usually no corrections. At the top of the page, Schubert merely indicated the title of the song; the author attribution (“Kosegarten”) and date (“comp[osed] 19/10 [1]815”) were added by the annotator mentioned above, as was most of the rest of the marginalia.
Figure 3. Fair copy of Schwangesang (D318).
The connection between the marginalia numberings on the rough draft and the fair copy of the same song now became obvious. In the lower-left corner of the fair copy stood the number “17,” as indicated in the annotation on the rough draft. What is more, this fair copy provided a cross-reference to the r
ough draft in a perfectly complementary fashion; it referred back to the other manuscript as follows: “Das ursprüngliche Manuscript vide Nr 24 Schubert’s Lieder mit Namensfertigung und Datum Abth. I” (The original manuscript see No. 24 of Schubert’s Lieder with signature and date Series I). Compare the numbers in the lower-left corners of both manuscripts along with the references in the marginalia along the bottom edge in Figures 2 and 3. Whoever made these markings was concerned about pointing out manuscripts of the same song. Oddly, the rough drafts and fair copies were assigned not the same but separate “series” designations. What is more, the number in the lower-left corner, though associated with a specific series, never appeared with its Roman numeral designation; this could only be derived from the autograph that referred to it. Thus the “17” on the fair copy appears with no Roman numeral next to it; the assignation of the fair copy to “Series IV” could only be determined by consulting its rough draft, where it states: “A copy (or fair copy [Reinschrift]) in Franz Schubert’s hand see 17 Series IV.” Likewise, the fair copy explains the “24” in the lower-left-hand corner of the rough draft: “The original manuscript see No. 24 of Schubert’s Lieder with signature and date Series I.”9
Having unlocked the secret of the numbers in the lower-left corner, it was now possible to organize the manuscripts by what looked like someone’s attempt to put the songs in order. The sequence that emerged on the rough drafts of the Kosegarten songs led to no significant patterns, but the case with the fair copies proved highly suggestive (see Table 4). By referring to the rough drafts of the 1815 Kosegarten settings, it was possible to determine that all twenty of the fair copies of these songs belonged to Series IV. In fact, Series IV would turn out to include only these twenty works. What is more, the twenty formed a set numbered consecutively from 1 to 20 that did not reflect the order of composition or sequences in any other series. But what criteria had determined these markings? Were these numberings arbitrary or did they reveal the sequence of songs as they had been grouped together as a set? It was necessary to identify who had entered these marginalia and the purpose for doing so.
Franz Schubert and His World Page 17
