Franz Hünten (aka François Hünten) published a Premier Quadrille de contredanses variées, suivi d'un galop as his Opus 63. The date is about 1834; a facsimile is available on IMSLP, as a digital copy of the facsimile published in an anthology by Garland about twenty years ago.

The work contains the usual five numbers of the quadrille as they were standardized in the early 19th century, plus a specific genre dance appended, as was the custom in published quadrilles at least through mid-century. In this case it's a galop; it might also have been a waltz or, after 1840, a polka.

The interesting feature of Hünten's quadrille is that the alternate strains in all five numbers are varied when repeated but the refrain (first strain) is not. The design of each number is ABACABACA. This provides 72 bars of music for the dance, or 8 for the promenade and 64 for the figures. See the graphic below for incipits of the refrain, B, C, and the variants of B & C.

I suspect that the composer is duplicating a common practice of performers (most often pianists) in house balls and similar dancing occasions (as with the post-Schubertiade dancing of Schubert's friends, for example). The unembellished refrain provides a stable guidepost for the design -- and therefore for the dancing -- while the variations on the alternate strains not only provide additional contrast with the refrain but also reflect what a creative (or bored) pianist very probably did in the course of dancing that could go on for a half hour to an hour at a time.

more to Czerny, op. 300

Among the 121 preludes in Carl Czerny's op. 300 are several that play very directly to rising gestures and follow through into the design of the whole. No. 30 uses the same broad schema as no. 15 (discussed in the preceding post)--a line from ^3 to ^8--but clothes it in rapid scale and octave figures. The first half takes F# up to D but does not close the cadence; instead it starts over with broken octaves and then concludes with a stereotyped short cadenza that carries ^5, ^6, and ^7 up to close on ^8. This "freeing of the ^6" is one of the most characteristic and distinctive figures in 19th century music, especially in the first half of the century.

Czerny, op. 300

Carl Czerny's The Art of Preluding, op. 300, may be the ultimate compendium of this aspect of sophisticated concert practice in the first half of the 19th century. Unlike The Art of Improvisation, op. 200, which is a treatise with examples, The Art of Preluding is a collection of preludes of many types in all keys. All are "ossified" (that is, published) versions of the kinds of preludes that pianists would apply to any and all of the music they performed in concert, including chamber and ensemble concerts, not just solo recitals.

A surprisingly large number of the 121 preludes have prominent rising gestures, but only a few follow through into the cadence. The most obvious of them elaborates a simple V7-I cadence by expanding the V through an octave's worth of a harmonized chromatic scale: see no. 31 below.

 An earlier Prelude in Ab major takes a rising line very gradually upward from ^3 to ^8:

Handout for SMT presentation

I presented "Schubert's 'Riemannian Hand': An Archaeology of Improvisation for Social Dancing" during last week's joint meeting of AMS/SMT in Indianapolis. I have posted a pdf file with the handout here. [NB: This link was updated on 3 July 2016]

Here is the PROPOSAL as it was sent to the program committee last spring:

Proposal for the annual meeting of the Society for Music Theory, Indianapolis, 2010

Title: Schubert's "Riemannian Hand": An Archaeology of Improvisation for Social Dancing

Research on improvisation practices in the nineteenth century has looked primarily at the fantasia. The reminiscences of Schubert's friends provide a window into another, equally important, practice: playing music at the piano for dancing. The celebrated Schubertiades varied greatly in their formats but often consisted in musical performances followed by dancing (Litschauer 245). For the latter, Schubert was obliged to string his waltzes into "endless cotillons" (Deutsch 230; also, Litschauer and Deutsch 101). A close relative of the contredanse, the cotillon required frequent repetition of strains, particularly the principal one, which acted as a kind of rondo theme. The number of repetitions depended on the dance caller (some published instructions stretch 24 bars of music across 80 or more bars of dancing).

Using the three-layer texture of the waltz (as played on a piano) and "endless cotillons" as the design, I will demonstrate with examples and through performance (1) how strict small forms, repetition, and variation can reveal pairings and groupings among Schubert's surviving waltzes, suggesting relationships that may have arisen through varied repetition in performance; and (2) how the chordal offbeats can effect transformations with parsimonious voice leading by simply moving thumb, middle finger, or little finger, thus anchoring the more distant modulations that Schubert attempted in improvisation.

During the presentation, I will provide a handout charting and comparing the dances that make use of third relations, and I will perform a "music-stretching" cotillon that will gradually transform one Schubert waltz (as the first iteration of the principal strain) into another (as the final iteration).

An instance of the pairings that emerge from this study is given in Example 1, which shows that the A Major Waltz D. 779 no. 13 (familiar to music theorists from analyses by Schachter and by Lerdahl and Jackendoff) might easily have arisen as an improvised variation on D. 365 no. 6. [ Example 1: At the top, Schubert, D. 365 no. 6, opening; at the bottom, D. 779 no. 13; in the middle, underlying voice leading pattern -- see the presentation handout for this.]

The three-layer texture is associated with the most common ensemble in tavern or small restaurant settings in Vienna about 1800: two violins and bass. The layers are clearly differentiated in Example 2 -- the second violin's double stops would simply become the offbeats in waltzes by Lanner and Strauss. [Example 2: Beethoven, Sechs ländlerische Tänze, WoO15, no. 1, first strain. -- see the presentation handout for this.] The texture of this ensemble could be appropriated for domestic settings when the piano became popular as a replacement for the traditional violin as accompaniment for dancing. For those pieces that used the Ländler-derived "oom-pah" rhythms, the result was three functionally differentiated layers, two in the left hand, one in the right.

I focus attention on patterns of the middle layer (left hand chords) in relation to upper voice figures, particularly on those that generate third-related key areas in the second strain. See Example 3, where the left hand executes an LP transformation. [Example 3: Schubert, D. 779 no. 13, move from the first to second strain, and from A major to C# major, as an LP transformation in the left hand -- see the presentation handout for this.]

By doing multiple comparisons among dances, I try to reconstruct some sense of how Schubert, during improvised performance, may have been—in Kofi Agawu's terms—"thinking in music about music."

Works cited

Agawu, Kofi. "How We Got Out of Analysis, and How to Get Back in Again." Music Analysis 23/ii-iii (2004): 267-86.

Deutsch, Otto. Rosamond Ley and John Nowell, trans. Schubert: Memoirs by His Friends. London: A. & C. Black, 1958.

Lerdahl, Fred, and Ray Jackendoff. A Generative Theory of Tonal Music. Cambridge, MA: MIT Press, 1983.

Litschauer, Walburga. "Unbekannte Dokumente zum Tanz in Schuberts Freundeskreis." Studien zur Musikwissenschaft 42 (1993): 243-249.

Litschauer, Walburga, and Walter Deutsch. Schubert und das Tanzvergnügen. Vienna: Holzhausen, 1997.

Notley, Margaret. "Schubert's Social Music: The 'Forgotten Genres'." In Christopher H. Gibbs, The Cambridge Companion to Schubert, 138-54. Cambridge/New York: Cambridge University Press, 1997.

Schachter, Carl. "Rhythm and Linear Analysis: Durational Reduction." Music Forum 5 (1980): 197-232.

Archaeology of Improvisation

In November, I will be giving a paper-presentation during a conference session on improvisation. The place is Indianapolis; the occasion is the joint national meeting of the American Musicological Society and the Society for Music Theory.

The title is "Schubert's 'Riemannian Hand': An Archaeology of Improvisation." Here's the abstract:

Schubert was said to string his waltzes into "endless cotillons" for dancing. A close relative of the contredanse, the cotillon required frequent repetition of strains, particularly the principal one. Using the three-layer texture of the waltz (as played on a piano) and "endless cotillons" as the design, I will demonstrate (1) how strict small forms, repetition, and variation can reveal pairings and groupings among Schubert's surviving waltzes, suggesting relationships that may have arisen through varied repetition in performance; and (2) how the chordal offbeats can effect transformations with parsimonious voice leading by simply moving thumb, middle finger, or little finger, thus anchoring the more distant modulations that Schubert attempted in improvisation. By doing multiple comparisons among dances, I try to reconstruct some sense of how Schubert, during improvised performance, may have been—in Kofi Agawu's terms—"thinking in music about music."

[note added 5-19-10: Legler and Kubik reproduce instructions from a dance manual by Edward David Helmke (1830), one of which is "A waltz may last no more than 15 minutes and a cotillon no more than 45 minutes" (95).]

Reference.

Legler, Margit, und Reinhold Kubik. "Anmutige Verschlingungen. Tänze des Vormärz: Quellen – Notation – Ausführung." In Boisits, Barbara, and Klaus Hubmann. Tanz im Biedermeier: Ausdruck des Lebensgefühls einer Epoche, 89-131. Proceedings from the Symposium Musizierpraxis im Biedermeier: Tanzmusik im ländlichen und städtischen Bereich, Graz, Austria, 26.-27. März 2004. Series: Neue Beiträge zur Aufführungspraxis, vol. 6. Vienna : Mille Tre Verlag Robert Schächter, 2006.

More to hexatonic cycles and the Riemannian Hand

I've written in an earlier post about Richard Cohn's hexatonic cycles in relation to the C# move in D779n13. I probably should have labeled today's entry another "regret post." As with Eco and Westergaard, Cohn's article on Schubert's treatment of tonality has many points of resonance with posts on this blog, probably because it has been in my mind several times already when I make reference to Neo-Riemannian historical narratives.

I will write more about the article at another time. In thinking about it this morning, however, I went back to the post with the waltz that has a close parallel to D779n13 in a direct A-C# move from first to second strain (then back again): D971n2: see the score in this post. The move is interesting in itself, of course, since D971n2 might well be another trace of a Schubertian improvisational formula, but the piece as recorded is curious in the absolute regularity of its left-hand chord shifts: in every case, as shown below, two pitches change by a half step while the third remains stationary. A rotation is required between each key area, as the register shifts of E3 up to E#4 and back reveal.

All this seems the more remarkable in light of Cohn's observation that

Theorists recognized voice-leading efficiency as an alternative basis for harmonic relations already during Schubert's lifetime and throughout the nineteenth century, but were reluctant to explore its systematic implications, primarily because to do so would have required them to relinquish their deep prior commitment to acoustic theory and tonal centricity, in favor of a less hierarchic, more networked conception of harmonic relations. (214)

I would extend that early knowledge of voice-leading efficiency to the afterbeats played by Schubert's left hand. Pragmatically, as any pianist who has played waltzes, ragtime, or stride piano knows well, anchoring the after-beats on the keyboard serves as a way of making the (frequently blind) leap down to the downbeat bass note more secure. Paradoxically then, for the improvising pianist it is the afterbeats -- and not the accented bass -- that provide a stable voice leading thread and can influence the direction of improvisation as well as harmonic formulae do.

[2-23-10: corrected text and graphic] In that light, consider the figure below. At (a) are the three Riemannian transformations R, L, and P. These are all the 1-note step-changes that will produce major or minor triads. At (b) are the diatonic 2-note changes; at (c), the chromatic -note changes. At (d) are the diatonic and chromatic triads generated through 3-note changes. In all cases except familiar inversions of V7 and ii7 or forms of viiø7, no non-triadic results are shown, and those moves that generated parallel fifths were also excluded.

It should probably not be remarkable that Schubert usually maintains close voice-leading patterns in the after-beats of his dances, given the habits of his training and the practical uses mentioned above. That such patterns are sometimes maintained even in the face of unusual or awkward chord progressions, however, is notable, an indication perhaps of how the "middle" -- always literally close to the player's eye, after all -- can guide the "edges" of the texture. In D779n7 (below), we would expect the stationary F3 in the second strain, given the T-D Ländler figures, but look at how tight are the after-beats through the odd (and not altogether elegant) harmonic meanderings of the first strain. (As usual, click on the thumbnail to see the graphic at original size.)

Reference:

Cohn, Richard. "As Wonderful as Star Clusters: Instruments for Gazing at Tonality in Schubert." Nineteenth Century Music 22/3 (1999): 213-32.

Fauxbourdon

Was the C# major section a creative (and perhaps spontaneous?) response to the problem of parallel fifths in the first strain of D779n13 and the realization of an underlying fauxbourdon figure?

Readings from ^3 -- including Carl Schachter's -- inevitably include a string of parallel sixths underneath the principal line (see this post). In a recent session that included both improvising on right- and left-hand 6/3 passages and playing through a number of Schubert waltzes, especially those I know least well (that is, those written down before 1819), I realized (a) that following a tight fauxbourdon figure in the right hand frequently led to rather dull results; and (b) that, given their obvious utility in structuring the physical path of a dance improvisation, Schubert frequently uses strings of parallels in the right hand but is surprisingly reticent about the line of sixths down from ^3. When he does use the latter, he will find ways to vary it -- as with the truncation in the first strain of D365n9 (below) or the upward extension in D365n5.

I should have made an obvious stylistic point much earlier in this blog: ii6 -- or for that matter any S-type or predominant chord -- is a marker of the German dance, not the Ländler. The latter, as Litschauer documents, is characterized by alternations of tonic and dominant or prolongations of one or the other (see my Music Analysis article, 214-15). The German dance makes use of a range of progressions, including many taken from menuets. On these terms, D779n13 is a perfect marriage of the two dance types: it announces itself as a German dance immediately but is zärtlich like a Ländler -- indeed, one might speculate that the rare expression indication was meant to alert a contemporary player that, although this looks like a deutscher, it should be played more slowly and sweetly. (Virtually the same, btw, can be said of D365n2, the Trauerwalzer.)

In posts sometime next month, I will write about another right-hand figure that is quite common in Schubert's waltzes: the diatonic wedge, or ^7 to ^1 below with ^6 to ^5 above.

Postscript: One of the most direct treatments of parallel sixths in the ^1-^3 space is in the first of the two schottisches that follow the German dances of D783:

Schottisches in the first decades of the nineteenth century are often surprisingly direct, even crude (Beethoven's are good examples). With this one, the minor-key topos and pedal point announce zingarese exoticism but the second strain suddenly turns the motive into a loud march, and the dance ends with horn calls. All this in the space of 16 2/4 bars. The result is a comic portrait of Hungarian soldiers, perhaps an invitation to a parlor game rather than dancing.

Citations for recomposition posts

Here are two citations (with comment) to literature relevant to the recomposition posts (see "recomposition" in the Label sidebar). Both concern that intersection between arrangement (in this case) and analysis to which Kofi Agawu argues (post).

Nicholas Cook,"Arrangement as Analysis," Journal of Music Theory Pedagogy 1/1 (1987): 77-90. Writing when the Society for Music Theory was barely a decade old, Cook laments the growing gap between the abstract nature of music analysis and the practical activities linked to composition and performance. He argues that for certain kinds of basic analytical insights, at least, arrangement is as good a tool as analytical notation: the arranger is required to make analytical decisions in order to decide among textural and timbral options.

Cook illustrates this particularly well by comparing Beethoven's string quartet version of the Piano Sonata, Op. 14n1. In one case, the quartet medium allows Beethoven to add instruments one by one and thus "write out" a crescendo missing in the piano version. In another case, recasting a pianistic arpeggio passage gives a good sense of what elements Beethoven regarded as "basic" and which could be varied:

The two versions share the same bass, and consequently the same harmonic outline; also the downbeats of the melody usually coincide, though the register is not always the same. But that is all. This represents Beethoven's conception of the work's underlying structure. . . . The piano and quartet versions are different elaborations of the same underlying structure--elaborations appropriate to the specific medium in which they are realized (84).

If Cook says, later, that "one might reasonably describe arrangement as analysis made audible," one can also say that improvisation under the constrained circumstances of playing for social dancing also fosters a kind of audible analysis -- indeed, requires it for the practice of varied repetition that was still a firmly embedded element in performance practice in the 18th and 19th centuries.

Thomas Christensen, "Four-Hand Piano Transcription and Geographies of Nineteenth-Century Musical Reception," Journal of the American Musicological Society 52/2 (1999): 255 -298. This comes at arrangement from the opposite direction: Christensen is mainly concerned with transcriptions of symphonic compositions for piano four-hands.

The reduction of the trio texture to the solo piano is related, to be sure, but the effect was very different in the 1820s from those that Christensen describes for much of the rest of the century. One example: the four-hand versions of symphonies played in the home resulted in a "dislocation of musical genre from performing geography" (290) and "is implicated in the process of etherialization that became such a cornerstone of Romantic formalist music aesthetics." "Dislocation" is much too strong for the difference between the tavern or restaurant and the domestic setting of eating, drinking, music making, and dancing. Indeed, the two venues were very closely related to one another. I would also argue that for Schubert's friends, the piano was an alternative timbral resource as rich in its own possibilities as the trio (or larger group) for which it substituted. This would especially have been true where original music was being created on the spot -- and where expressive keys of many sharps or flats, and expressive key relations such as the mediants, were being explored.

It was such congenial environments that were displaced later in the century by the sacralization of the concert.

more on Schumann and the F-major experiments

I originally ended yesterday's post with the following:

One has to wonder if Schumann ever played for dancing, because, despite the party-game context of his review, he had a skewed notion of what Schubert's dances, even in their published forms, represented. It was a small step from mistaking these F-major deutscher for a closed text to the monumentalizing priorities of the Bach Gesamtausgabe. And it really seems unnecessary because the record of Schumann's early music, finished and unfinished, certainly confirms that he had improvisational skills.

In the meantime, I found this in Andreas Boyde's reconstruction of the unfinished set of variations on D365n2:

In 1827 Schumann began "revelling" in the music of Franz Schubert. He heard his songs for the first time, was introduced by Agnes Carus to the four-hand works for piano and fell in love both with "Schubert's Waltzes and her". His diary on 2nd March 1829 mentions a "fruitful improvisation on the Waltz of Longing"; eight months later he requested the music for the complete Schubert Waltzes from Friedrich Wieck and according to Friedrich Täglichsbeck played them "beautifully and whenever he had the chance".

The question of playing for dancing, however, remains open.

Reference:

Robert Schumann. Variations on a Theme by Schubert - Sehnsuchtswalzervariationen,

reconstructed from the manuscripts by Andreas Boyde. Hofheim: F. Hofmeister, c2000.

Imaginary continuo (after Rothstein)

Is it possible to retain a sense of Schubert's risky improvisations (see posts (1), (2)) in a Schenkerian graph? Only if the graph can be made to reflect the human and fallible, not just the unerring creative instinct of an idealized notion of genius. To attempt this shift, I will start with the continuo grid. William Rothstein (296) uses the construct of the "chordal scale" (derived from Lerdahl) as the repository of tones for the imaginary continuo, his term for what I will call the continuo grid.

The beginning of the first graphic below shows the chordal scale for D779n13 with respect to its tonic triad: all members of the A major triad between the lowest and highest notes struck. Out of this scale emerge the four voices of a standard part-writing progression representing measures 1-10, or the introduction and the first eight measure phrase. The analogy to a continuo keyboard part would be better if the right hand held three voices and the left just one, but I resist that because I want to preserve the textures of this waltz as much as I can.

Resisting the idealization of textures is in fact the basic strategy of this rewriting. Note that one result is the prominent set of parallel fifths. A respect for textures suggests that the pedagogically simplified, chorale-style texture of (a) is not the best solution for this passage, which after all contains several additional voices in the left hand: these are inserted in (b), the end result looking more like the textures in orchestral settings of dances in the late eighteenth and early nineteenth centuries. Note the octave doublings that have now been added to the parallel fifths. The fullness of this texture contrasts with the three-part reduction that keeps only the two imitative upper voices and the bass (c). It is this latter that Carl Schachter uses in his durational reduction of the Waltz, but Schachter corrects the parallels by texture inversion of the two melodic voices.

Instead of collapsing the texture as Schachter does, I have removed the "non-harmonic" I 6/4 chord in (a) of the next graphic below, leaving a simple ii6-V7-I harmonic progression. Level (b) expands (a) with the closing cadence.

The contents of (b) are reproduced in the lower system of the final graphic (below) and renotated to correspond more closely to my preferred Schenkerian reading, the rising line from ^5 to ^8. The alignment is not meant to suggest an equivalence, but a replacement: if maintaining textures is thematic here, then the lower system takes the place of the upper one.

The continuo grid has a quality of concreteness about it, as if the improviser is looking at the piano keyboard and making choices about register (A major, melody in the fifth octave, accompanying voices in "standard" positions for a waltz in this key). Once the fateful decision is made (at bar 3) to model this waltz on D365n6, the parallel fifths emerge very quickly, and Schubert's decision to maintain the register despite the parallels opens the door for the sublime counter-move to the downward pull of the ancient suspension figures: the sudden lift over V in the final cadence. The graphic suggests further that, having transmuted an error into the sublime, so to speak (the parallel fifths into the rising cadence gesture), Schubert adopts the rising itself as a design in the second strain: tonicizing C# major (rather than treating it as V of vi to start a cycle of fifths sequence), then the peculiar chord progression it leads to, which facilitates another expressive moment associated with upward shifts of register (in measures 31-32--not included in the example).

References:

Rothstein, William. "On Implied Tones." Music Analysis 10/3 (1991): 289-328.

Schachter, Carl. "Rhythm and Linear Analysis: Durational Reduction." Music Forum 5 (1980): 197-232.

Postscript: More to Improvisation

As a postscript to yesterday’s improvisation history, here is an alternate solution that Schubert might easily have come to: a simple sixteen-bar form that resolves the hypermetric peculiarities of its model: see graphic below (NB: this is a thumbnail; click on it for the original). The roughness of its harmonic figures in the second strain is not evidence against it, as there are many precedents in the waltz repertoire for such sudden harmonic turns between the four-bar components of an eight-bar strain.

The key to the changes lies at the beginning: the two-beat (four-eighth-note) pickup has been altered to a clichéd single quarter-beat (as in D365n6), and the harmony begins "in progress," as it were, with the 7-6 suspension over ii6. The fact that this opening is plausible (and closely resembles an existing piece) tends to invalidate Carl Schachter’s claim that "to omit the first two bars [of the A Major Waltz] would be to suppress the opening tonic altogether; [this] would make the whole piece pointless and nonsensical" (72).

The prosaic clarity of the recomposed first strain serves as well as any preceding analysis to highlight the strangeness of Schubert's original, perhaps the last bit of evidence we require in order to affirm that the A Major Waltz is a poor piece of social-dance music and is misplaced in D. 779--perhaps it would have been more successful had it been expanded a bit to act as the trio to a minuet or scherzo.

My crude rewriting barely masks the most obvious metric problems, however. The A Major Waltz, with its repeated second strain, consists of 29 two-bar groups, or the two-bar introduction plus 14 four-bar phrases, or 7 eight-bar strains: a waltz in an 8+8 design has four eight-bar groups, and a waltz in an 8+16 design has six groups.

The fact of an introduction itself is unproblematic; they are not common in Schubert's own dances, but brief introductory figures or flourishes had become familiar to dancers nearly a decade earlier through their use in the "extraordinarily popular" Linzertänze by Michael Pamer, principal predecessor of Lanner (Reeser, 47)--both Strauss, sr., and Lanner had played in Pamer's orchestra. Three other waltzes in the familiar sets by Schubert include two-bar introductions: D146n10; D365n34; and D734n15. (In a performance setting for dancing, such introductory "vamping" was undoubtedly commonplace, as I suggested in yesterday’s post.)

And uneven hypermetric groups do happen, also: three of the Valses nobles, D. 969, have them. Numbers 9 and 12 have an extra four-bar group in the second strain: the former adds up to seven eight-bar groups with the repeat of the second strain, the latter to nine such groups. Number 3, however, wins the prize as the oddest design in the major waltz sets: a four-bar introduction is followed by a repeated strain of 8+8; the second strain consists of four eight-bar groups plus one six-bar group (as 4+2). Thus including the repeat of the second strain, the total is 4 + (8 x 13) + 4, or 14 eight-bar groups altogether.
According to Litschauer and Deutsch (111), D. 969 was very probably meant as a concert cycle, not a functional dance collection, and one has to wonder whether the A Major Waltz would not be better placed as a trio in that set, rather than in D. 779, whose members are otherwise all functional dances not far removed from Schubert's first published collection, D. 365.

References:

Schachter, Carl. "Rhythm and Linear Analysis: Durational Reduction." Music Forum 5 (1980): 197-232. Reprinted as "Durational Reduction" in Schachter 1999a, 54-78. (Schachter, Carl. Joseph Straus, ed. Unfoldings: Essays in Schenkerian Theory and Analysis. New York/London: Oxford University Press, 1999.)

Reeser, Eduard. W. A. G. Doyle-Davidson, tr. The History of the Waltz. Stockholm: Continental Book Company, 1949.

Litschauer, Walburga, and Walter Deutsch. Schubert und das Tanzvergnügen. Vienna: Holzhausen, 1997.

D779n13 originates in improvisation

The Opus 9 (D. 365) collection was published in 1821; as most of the waltzes of D. 779 were probably composed in 1822-23, it is entirely plausible that Schubert might have played a version of D365n6 at some point not long after the publication of Opus 9 and decided to use its distinctive opening formula to improvise a new waltz. The likelihood is increased by the fact that D365n6 was apparently among Schubert's favorites: as Litschauer reports, it is among only six Schubert waltzes that appear three times in different manuscripts and one of only two found in manuscript copies made by one of his friends (Litschauer 1995, 4). The fact that some of these are variants is all the more intriguing. The first of the three manuscript versions (in 9 Deutsche (1819)) is identical to the published version, but the second (in 2 Deutsche (1821)) is striking in that it gives a fully realized Bb minor 6/3 chord in the left hand of bars 1-2, whereas the third (in 4 Deutsche (undated)) combines the first strain of D365n6 with the second strain of No. 7 from the same set (see Litschauer 1989).

Thus, we can imagine the new waltz starting as depicted at (a) in the graphic below: as necessary, vamping to establish the waltz meter and tempo for the dancers (or to gather one's thoughts), then open with a slight rhythmic variant of the figure in D365n6. The voice leading formula (second line in (a)) is varied too from the moment that F# becomes the upper voice (by the end of the first full bar of melody), but that is also where the trouble begins (third line in (a)), also realized as an initial attempt at a complete phrase in (b), which has a rather flat ending that allows the suspensions to drag the upper-voice melody down through ^4 and ^3).

Level (c) shows this ending again with the lead-in to the repetition of the phrase. Here is the first inspired moment in what has so far been a rather dismal effort: the hemiola rhythm of the initial gesture has taken control by now and the "turn" in four eighth notes gracefully (and convincingly) gathers energy that is directed toward the F#. Finishing out this varied repetition of the first strain, Schubert brings the upper voice down from its perch on ^6 and ^5 with a conventional ländler cadence that strikes ^1 but leaves the register of ^3 open. This cadence will also close the second strain of an 8 + 8 recomposition in Example 4.34.) Like the descent to ^3 in the first statement, this cadence seems weak.

At this point, the decision has to be made whether to set the new waltz in a two or three-part form, but for our purpose here, Schubert's choice makes little difference. I will assume that he composes a contrasting second strain (in the manner of D365n5). The dancing continues, and I imagine that Schubert plays a trio (perhaps one of the Ab major waltzes from D. 365 transposed to A major), but the awkwardness of the newly invented waltz bothers him, and he returns to it. This time around, the first phrase and most of the second are as we know them in the published version, but the cover-tone-qua-melody, now much more insistent than in the first attempt, brings about the second inspired moment, the cadence that lifts the upper register to ^8 (see the lower line in (d)).

Having reached the register of ^8 (A5), Schubert decides to stay there, perhaps expecting to use G#5 to initiate a cycle of fifths sequence as in (e'). For reasons unknown (which might include a simple lapse of attention during a long evening of playing), he decides to ground the contrasting middle of a small ternary form in a tonicization of the C# major supporting that G#5. There are a few examples of this design, where the contrasting middle closes in the tonicized key, and the reprise returns to the main key without modulation, but in this case the design will not work because the theme starts on a non-tonic position one half-step above C#.The third and last inspired moment, then, is to take advantage of yet another motion upward to carry the melody into the pianoforte's thin-toned and ethereal upper octave where the reprise begins in the final version.

Schubert chose to keep the A Major Waltz and eventually found it a place in the D. 779 collection, but I fancy it was not because of its odd combination of counterpoint and dance topoi or even its dramatic tonal contrast (something Schubert is known to have liked), but instead for the sake of its charming two-tiered melody. I imagine one of his musically skilled friends, perhaps Josef von Spaun, coming over after hearing several repetitions of the Waltz and whispering, "Schön, zärtlich," while the dance continued.

References:

Litschauer, Walburga. "Franz Schuberts Tänze: Zwischen Improvisation und Werk." Musiktheorie10/1 (1995): 3-9.

Litschauer, Walburga, ed. Franz Schubert. Neue Ausgabe sämtlicher Werke, Series 7, part 2: Werke für Klavier zu zwei Händen, Band 6: Tänze I. Kassel: Bärenreiter, 1989.