The Journal of Seventeenth-Century Music

Volume 10 (2004) No. 1

A Schenkerian Look at Lully

Gregory Proctor*

Abstract

Although Schenker himself did not include music as early as Lully’s in his canon, a strict Schenkerian perspective is applied to Act I, scene 4 of Persée, following the lead of Rosow’s description of large-scale structural patterning in Lully. Lully’s music is shown to be classically tonal in the strictest Schenkerian sense, while the analysis highlights harmonic-contrapuntal stylistic features favored by Lully. Long-distance connections are also made through the registral disposition of final cadences, symmetrical disposition of recitatives and airs about a central axis, and the extension of Schenkerian structure to distinct recitatives that otherwise form a dramatic unit.

1. Introduction

2. Issues in Schenkerian Systematics

3. Stylistic Vexations

4. Lully as Contrapuntal-Structuralist

5. Arch Form and Symmetry

6. The Challenge of the Schenkerian Approach

References

Musical Examples

Audio Examples

Figures

Tables

Appendix

1. Introduction

1.1 Lois Rosow has recently provided tools for demonstrating just how Lully “pensoit en grand.”¹ Let us set the stage with an extended quote:

As a result of the supple structure of Quinault’s poetry, along with the fluid relationship between recitative and air (from both structural and functional points of view), Lully had both the latitude to shape his scenes in a variety of ways and a special responsibility to help his audience perceive the shape of each scene as it unfolded. Thanks to the subtle means used to differentiate structural elements, by both poet and composer, the simplest expedients—repeating an introductory phrase or changing mode, for instance—could make the difference between articulating a salient moment and allowing that moment to be subsumed into relative continuity. Through clever placement of cadences of varying strength, closing formulas, clear beginnings and ambiguous beginnings, and recurring elements, Lully ensured that his listeners experienced a hierarchy of points of articulation, one that closely complemented the drama.²

Rosow’s words “hierarchy of points of articulation” will attract the attention of those trained in Schenkerian theory. Schenkerians always count on hierarchy of some sort. Although the concept of hierarchy applies in the world at large and in musical analysis in many different and overlapping ways, Schenkerians rely on generative hierarchy as opposed to, say, the status hierarchy of Riemannian function theory. Clearly, Rosow’s “hierarchy of points of articulation” is not directly generative in the normal Schenkerian sense. Yet it invites investigation in Schenkerian terms to see how close the correlation can be made between Schenkerian theory and Lully’s music, which in principle lies outside Schenker’s own time-boundaries of applicability. This investigation thus provides an opportunity to reflect on some basic issues in Schenkerian studies.

1.2 Schenker’s theory constitutes a remarkable definition of classical diatonic tonality, especially as confined to the period to which he himself consigns it: Bach through Brahms. Even within his time frame Schenker’s range of applicability is selective. That is to say, Brahms is in but Bruckner is out; Chopin is in but Liszt is out. Many are called but few are chosen. Although most Schenkerians do not consider themselves bound by these stylistic strictures, most of the analytic work done by Schenkerians nevertheless concentrates on music within Schenker’s canonic time span and canon of musical literature. It should not be surprising to discover that the application of Schenkerian theory at the edges of this canon will raise issues of what constitutes normative structure for those “edges,” and also confront the analyst with problems less likely to be encountered toward the center of the Schenkerian frame. It may be that Schenker’s own bias against this music was confirmed by its apparent ambiguity. We can therefore reject or ignore a style that might raise ambiguity issues, or we can seize the opportunity to make finer discriminations among styles. Although the Schenkerian ground has been shifting over the years toward a more comprehensive approach to different musical styles, especially with respect to post-tonal music (on the other side of classical tonality from the seventeenth century), Schenkerians—with few exceptions—continue to follow Schenker’s canon when dealing with the tonal repertoire. Geoffrey Chew is an outstanding exception.³

1.3 Once a theory is devised that satisfactorily defines the structure of a style of music and produces coherent analyses, it is not especially interesting or informative to repeat the analytic process indefinitely, especially after a composer’s work has been shown to fulfill the theory’s predictions about its structure. Schenkerians, therefore, having invested time and effort in the analysis of some Schenker-legitimized music and having discovered (to no one’s surprise) that the music fits the theory, publish their analyses anyway with the reader’s attention drawn to some characteristic of the piece incidental to the analysis but which the analysis helps demonstrate. Although tedious to read, this at last comes close to a different and more useful practice: the application of an analytic system to a literature whose membership in the category covered by that system is in doubt. The degree to which the analysis succeeds and fails precisely constitutes its relation to that category (always providing that the mechanism has been applied faithfully). Pure Schenkerian theory applied at the edges of classical diatonic tonality thus helps to define how some music is not like Haydn’s, say, in contrapuntal-structural terms.

1.4 A sixteenth-century motet, for example, is only remotely connected to Schenker-tonality. It holds in common with classical tonality the triad as primary referent structure; the well-formed line; the diatonic field within which the lines and triads operate; and the guarantee of clear cadences of several types at regular time intervals and construed to correlate with the musical form, which is an image of the form of the text. Yet it is difficult to show that any string of contiguous sections of the motet are governed by an abstract contrapuntal structure whose realizations at successive levels have led to this string on the surface, which is what Schenkerian theory demands. Add the complexities contingent upon modal characteristics (what would a pure unmodified Schenkerian linear close be in a real Phrygian mode?) and we can easily grant Schenker his observation that such music lies outside the range of his theory. This is not to say that others have not continued to build theories to deal with such music, for that is exactly what they do and should be doing; such theories are often presented as adaptations or extensions of Schenker’s theory, and whether or not they can be reconciled easily with his theory as adaptations, they are not his theory in full bloom. They may seem to answer, or actually answer, problems of outlying music, but at the same time they virtually abandon the core of his contribution—that tonal music is structurally thick.

1.5 Now if the Renaissance is out of the frame and Bach is in, then where in the Baroque does Schenker’s system begin to operate consistently? This is the same as inquiring as to where in music history must one adapt or alter Schenkerian theory. It seems plausible to guess that Schenker-tonality might well be established for most Western European music by the middle of the seventeenth century, and in that case Lully’s music would likely be part of it. I propose, then, to engage the musical intersection between Schenker (unmodified) and Lully. I make special reference to Act I, scene 4 of Persée, with some invocation of Act V, scene 1 of Armide.⁴ A word of caution must be inserted here. Since I approach this music from a Schenkerian perspective, a reader may expect substantial graphic analysis of the composition. Aside from reservations already expressed about the tedium of published analysis altogether, my aim here is different. I mean to engage issues of structure raised by Rosow that have elicited responses in me that I associate with Schenkerian theory. Table 1 lists those issues in Rosow’s chapter as I read them. The spinning out of the discussion as the issues arise does not necessarily follow a completely Schenkerian course, but it is the Schenkerian orientation that leads me to the discussion.

1.6 Two terms in Table 1 that require comment are “air” and “extended binary.” Much discussion of this music depends on the differentiation between “air” and “recitative.” The original scores had no such designations; they are our labels, not Lully’s.⁵ That said, musicians tend to agree on which of the two labels they might use in any particular case. In general, we judge a passage to be “air-like” if it has relatively few differing note values and shows parallel phrase rhythms.⁶ It is an “air” if it further displays a closed small form. The term “extended binary” refers to a small binary form (AB) in which the second group of poetic lines (the “B” part) is repeated with different music. It can be symbolized as ABB', but I follow Rosow’s convention of writing ABb to clarify that the “B” sections have the same words but different music.⁷

2. Issues in Schenkerian Systematics

2.1 Schenkerian theory is subject to a variety of interpretations, some of them seriously at odds with others. Fifty years ago, Schenkerians considered themselves engaged in a war with “conventional analysis.” The war was more imaginary than real, and indifference to Schenkerian assertions was taken to be hostility to the whole Schenkerian enterprise. Under those conditions, the field of Schenker studies appeared to be more of a piece than it was. Some of what concerns me in Lully will be mystifying to some who consider themselves Schenkerians, who will not be able to conceive that there is an issue when I raise one. This is because most American Schenkerians are in fact Salzerians.⁸ To the disinterested—as well as uninterested—observer, a glance through Free Composition and Structural Hearing will not raise immediate suspicions that they are at odds with one another.⁹ The similar format (two volumes, one of text and the other of musical and graphic examples), analytic notation (distorted ordinary musical notation without metric value), and concern with large stretches of music as forming single trajectories would seem to consign them to one and the same camp. Yet there are foundational differences. Salzer understood the entire corpus of Western music as subject to his theory, quite the contrary of Schenker’s Bach-through-Brahms constraint. To get this breadth of application to work, Salzer had to undermine Schenker’s theory at its core. Schenker’s theory envisages a musical composition as a layered series of conceptual states, of slices of musical time, where each layer is the structure for the subsequent layer, and each subsequent layer is one of an infinite number of musical realizations of the prior layer. Salzer’s theory posits points of “structure” separated by webs of “prolongation.” A Schenkerian seeks to determine how each layer expresses the contrapuntal/harmonic facts of its adjacent layers (much in the way that variations relate to their theme and to the other variations).¹⁰ A Salzerian eliminates or adds notes to the facts of the adjacent layer, thereby either bringing the points of structure literally closer together (in the direction of structure) or spreading them farther apart (in the direction of prolongation). All textbooks in Schenkerian theory and most examples in the scholarly literature are Salzerian. In general, Schenker makes more analytic decisions than do Salzerians, and this is often reflected in the relatively large number of what Daniel Harrison refers to as “floating note heads” in Salzerian graphs.¹¹

2.2 The point of view in this article is strictly Schenkerian.

2.3 When I refer to Schenker’s theory proper, I additionally have in mind two slightly different theories, one earlier and one later. The assumption that Schenker's theory was a single entity that developed gradually over time is so ingrained in the Schenkerian culture that many theorists are unaware of the discontinuity, with the result that the difference between the two theories is typically ignored by Schenkerians. They tend to see the first theory as an immature version of the second, when they see it at all. The second theory then becomes, of course, the “mature” theory. The preference is thus to follow the assumptions and notational patterns of the later theory. Common to both theories is the assertion that no sense can be made of a detail except with reference to its context—there is no figure without a ground. Nevertheless, there is a substantial change in viewpoint from one to the other theory, and there are occasions when the earlier viewpoint seems to address more directly issues of interest to the analyst in the piece at hand.

2.4 Musicians are most acquainted with the later theory as presented in Schenker’s posthumous publication, Free Composition. Here, abstract voice-leading forms are nested completely within other “more” abstract forms. Each form is contained in another, as a ramification, until the ultimate container—the Ursatz—is reached.¹² This is accomplished by treating the figure-ground relation as (at least partly) recursive. Once a ground that illuminates some figure is established, that ground itself is treated as a figure on its own level and its context is sought. Ultimately we come to a context that has no embracing context of its own: the fundamental structure (Ursatz).

2.5 The earlier theory, found in the series of periodicals Der Tonwille,¹³ is more flexible, and comes close philosophically to Schoenberg’s Grundgestalt concept. Here there are fewer distinct layers of abstraction. A triadic space of third, fifth, or octave is posited and brought to life through its being filled with descending scalar segments. (This filling of abstract space is in fact what the term Urlinie originally referred to.) Surface counterpoint is then interpreted in terms of those spaces. Some lines, read through the immediate surface, will present the ideal forms directly, while others do so indirectly. Lines filling intervallic spaces other than third, fifth, and octave, together with rising lines through any interval, are relatively complex and must be understood as linked in ways that ultimately fill the abstract space of the simpler forms.

2.6 The primary analytic-notational tool for this theory is the Urlinie-Tafel. The Urlinie-Tafel consists of a simplified version of the music (a “metric reduction”) overlaid with brackets that show the sequence of scalar patterns that directly or indirectly represent the simple Urlinien. Other standard analytic symbols, such as Roman numerals, are typically appended. Schenker’s preference was to see the entire movement in terms of one typical line, but this is not required in the earlier theory. The indispensable requirement is to reveal the ubiquity of the Urlinie as a motive of a higher order. Figure 1 is from the first issue of Der Tonwille. Observe that the brackets in the first theme-group cover descending seconds and one fourth; in the second theme group they cover only fourths. Figure 2, from the same publication, was apparently produced a little later and is more consistent, being thoroughly pervaded by the descending-third line. Sometimes these lines are conjunct, at other times disjunct. The reasons for the choices are not obvious, and one bracket near the end reveals an interesting departure from the emphasis of the later theory. Specifically, the line in mm. 32–6 (a'-flat–g'-flat–f') is preferred over the plausible g'-flat–f'–e'-flat of mm. 36–7. This excludes the cadence with its second scale degree from membership in the Urlinie, even though such membership is one of the core requirements of the later theory. Again, observe how the lines are not differentiated with respect to priority, a central feature of the later style.

2.7 Figure 3 shows how, in a short time, Schenker moved to the concept of nested lines, which, while extending the degree of abstraction through several more layers, concretizes each layer with specific membership and the responsibility for generating the next ramification in an orderly way. Figure 4 shows the stage leading up to Figure 3, revealing the orderly generation of every element within a level by means of an operation on the elements of the prior level. Note that the primary source is a descending line containing the perfect authentic cadence, and that the ascending line (or arpeggiation, or arpeggiation and line, depending on which level we attend to) is now dependent upon the subsequent descending line. This is virtually a pure instance of the second, “mature” theory.

2.8 Although the second theory seems more orderly, the constraints on the strict order of generation and on one overarching line can sometimes seem forced. Observe, for instance, Example 1. From the perspective of the later theory (and treating the passage as a complete structure for the sake of simplicity of notation), Example 1 can be read as based on a fundamental line of either a third (“^3-line”) or a fifth (“^5-line”). Figure 5 shows a ^3-line analysis and Figure 6 shows a ^5-line analysis.¹⁴ The ^3-line interpretation invokes the interruption transformation, in which a line is begun, proceeds partway, and then breaks off, whereupon it returns to the beginning and runs its full course. It is an exquisite conceptualization of parallel-period phrase structure with a half cadence ending the first phrase and a perfect authentic cadence ending the second phrase. With interruption, at remote structural levels the break occurs at the half-cadence point, ^2/V. On later, more surface levels, the break can be earlier in the line. In both cases, though, the return is to the beginning of the line. The ^3 as head tone of the line in Figure 5 falls to ^2, is regained and proceeds to its goal, ^1/I. The ^5-line interpretation of Figure 6 relies on the middle voice following the leading upper-voice line in (mostly) parallel thirds. Here a line is created to connect with the middle-voice a', whereupon it unfolds back to the c'' of the leading line.

2.9 Figure 5 captures the question-answer flavor of Example 1, while Figure 6 captures more directly the fifth-space in which the tune flows. Schenker’s earlier theory captures both, as shown in Figure 7. The relative simplicity of Figure 7 over Figures 5 and 6 is due to its more rapid retreat into abstraction, while the other Figures worry the details of transformational logic from one level to the next. They make the music look more complicated than it sounds. The earlier theory is often more attractive in music at the seams of tonality (and I normally entertain Lully’s music in those terms), but the later Schenker is capable of making finer distinctions. Through its insistence on a rule-bound logic of transformations, the later theory can force the analyst to consider the patterns of transformation themselves and aid in the discrimination of one style of tonality from another. Schenkerians, however, ought to be aware of the possibilities attendant upon use of the earlier theory. Unfortunately, they often doubt that examples such as Figure 7 even exist in Schenker’s writings, and to help them I offer Figure 8, another analysis by Schenker from the early 1920s that matches up quite nicely with Figure 7.

3. Stylistic Vexations

3.1 Whatever composer we subject to the Schenkerian lens, there are likely to be peculiarities of that composer’s style that do not show up with the same frequency or distribution in other composers’ music. This makes the study of any composer’s work an exercise in style as much as it is an exercise in structure. Since Schenkerian structure is predetermined (there is a limited number of abstract musical plots), the discovery of stylistic traces is more informative in any case. The pleasure in analysis lies to a great extent in this region of idiomatics. We become accustomed to the ways in which a composer sets and solves problems; this gives us an advantage in the next piece we study by that composer, as we know better what we are likely to find. In the most general stylistic terms, we come to look for regularities. We are helped along when we find them and challenged when we do not. Given a Classical-era sonata’s first movement, for example, Schenkerians expect to find interruption at the first level of middleground, coincident with the second key area of the exposition if the piece is in major, and coincident with the end of the development section if the piece is in minor. We expect some proportional relations among sections of the sonata, and when they are subverted, but we know the piece is of the late eighteenth century, we suspect we are looking at Haydn rather than Mozart or Beethoven. Specialists become aware of numerous such peculiarities, and read through them subconsciously.

3.2 Analysis at the edges of the Schenkerian field intensifies this sense of idiomatics. One familiar with a style will have no difficulty in reading through the peculiarities, while newcomers will have to learn and become comfortable with those peculiarities (while bearing in mind that the style’s peculiarities are to be found in other Schenker-tonal music). Among these in Lully’s style are certain characteristics of melody building (e.g., a melodic peak attained by a leap upward, followed by one or two descending thirds, which may be filled by a coulé), the avoidance of sequences to propel the line, the lavish use of mediant and submediant harmonies in place of tonics and dominants, the linearity of the bass, and the relative weakness of the subdominant function in cadential formulas. What follows is a discussion of a few of these issues brought to the foreground, so to speak, in Lully.

3.3 The function of the bass part in Schenker-tonal music is threefold: to provide the elements of the structural bass arpeggiation and its transformations; to provide a root for a harmony when needed, as in the “leaping passing tone;”¹⁵ and to be, simply, one of the voices of the contrapuntal weave. In Lully’s operatic music, the third function is always near the surface, in distinct contrast to Italian music of the period. French preference for a singing bass line often results in the intertwining of the features of the separate functions in such a way as to seem to blend two distinct functional parts into one. Two voices, a line of the counterpoint and a structural bass arpeggiation, are combined into one part. Example 2, from Schumann’s Album for the Young, is one of the clearest examples of this phenomenon. It blends tenor and bass voices into one part. Schumann here evokes a child’s instrument with a limited range, such as a toy piano or music box. The compression of range causes part crossings, most explicitly presented in m. 4 (Example 3), where the bass voice’s cadential 6/4 ends up higher in pitch than the tenor’s passing seventh. The tenor voice is essentially assumed to be an inner voice and the bass an outer voice.

3.4 At least as well known, and highly typical of the issue as it appears most frequently in French Baroque music, is the G-major Menuet by Christian Petzold found in the Notebook for Anna Magdalena Bach (BWV Anh. 114). The minuet’s first period is shown as Example 4. Here the highest part (leading voice) ascends to the first-order neighbor and then proceeds down the scale to the second scale degree; the bass (following voice) supports the neighbor in the top voice and then proceeds downward in parallel tenths with the top voice. This lower-part accompanying scale is more typical of tenor than bass parts and accounts for the part assignments in the analysis shown in Figure 9.¹⁶ A somewhat similar instance from Lully’s Armide is shown in Example 5. The original is shown followed by a rewrite that restores the middle part to a registrally safe distance and disentangles the bass. Note that in this rewrite, unlike in the analysis of the Petzold minuet, the bass part is assumed to have maintained the subdominant harmony all the way through to the dominant, by virtue of rhythm and the suppression of the resolution of the neighbor to the fifth scale degree.

3.5 In general, I follow Edward Klonoski in accepting that upper-voice space and bass space are walled off from one another in Schenkerian structure.¹⁷ For those who do not accept this division of bass space from upper-voice spaces, there is no difficulty in the analysis of such places. The problem lies with those of us who do accept the division of the spaces from one another. Although transfer of register is normal in both spaces, transfer between the spaces is not. This means that a line begun in one space cannot end in the other. But an illusion of such a transfer is often given when middle voice and bass register become entangled, and the middle voice either crosses below the bass or joins it in a single surface part. The intertwining of voices for the sake of bass-part suppleness is by no means unique to Lully, but it is prominent in his music.

3.6 Prominent among the details of the seventeenth-century style but less to be found in later music are échappée-like figures where both preparation and échappée are harmonized. Example 6 shows the most common form, with the scale-degree succession ^3 to ^2 interrupted by the incomplete neighbor ^4. This would be of little concern in melodic analysis, except that the figure occurs near cadences and with the subdominant note in the bass pointing—in a cadence-signaling way—to the dominant. In later music this harmonic position is often associated with the II or II-6(/5) harmony, in which scale degree ^2 either is overtly above the fourth degree in the bass (as in m. 15 of Example 4), or is buried beneath ^4 (often literally in the bass part or as in Example 7 at the asterisk), thus displacing ^3 and continuing the line across the subdominant to the dominant harmony.¹⁸

3.7 In later styles such as Bach’s, it is easy to accept the implication of a II chord, a seventh chord, or an added sixth chord, but in seventeenth-century style, saturated as it is with triads and with few seventh chords, it is harder to make such assumptions. This is especially so if the purpose is solely to assure a continuous line to the cadence. Even in a later style, the surface may be so triadic that the II-6/5 choice cannot be made. Example 8 is an instance from Beethoven’s last piano sonata. The F chord marked with the asterisk cannot easily be assumed to have a d'' in it. Such an assumption flies in the face of the triadic surface.

3.8 Example 6 has the more interesting case of the “missing” note (a') actually presented in the time span of the subdominant harmony, which encourages its assumption as part of the harmony. Yet it is also plausible as solely a surface anticipation of the fifth of the dominant harmony. This latter choice is motivically reinforced by a similar anticipation of the tonic harmony in the time span of the dominant that follows immediately. To summarize the issue with respect to Example 6: if a' is a chord tone above the c in the bass, then—in idealized voice-leading terms—b' goes both to c'' and to a', and we have a II-6/5. In that case a line is led downward from b' through a' to g' and the line has a true close. If a' is solely an anticipation, on the other hand, then b' goes only to c'', we have a IV chord, the a' derives from g', and the close is illusory in that the line is not completed in the g' (b' still hovers). This will not concern the Salzerian who is inclined to imagine the free insertion (and therefore the analytic removal) of full harmonies as pure ornaments into a stream of structural counterpoint. In that case one decides on other grounds whether or not the line has closed, and if the line is determined to have closed, discards the harmony on c and retrieves the continuation of the descent from b' in the a' of the D harmony.¹⁹ As one who sees Riemann’s functional harmony system as bearing a genuine relation to tonality, however, I resist treating subdominant-function harmonies in cadential environments as ornamental, especially given their prominence in signaling full close, both in Schenkerian and function theory. The surface counterpoint must be evaluated as a prime clue in the determination of linear completion, and the vexation is that surfaces differ sufficiently so that the choice is never automatic.

4. Lully as Contrapuntal-Structuralist

4.1 We shall make most frequent reference to Act I, scene 4 of Lully’s Persée (see the transcription and prose translation of the complete libretto). To help establish context and to aid the reader’s sense of continuity, a transcription of the scene is provided in an Appendix. A dance is added to the end of the scene in the source—dramatically part of the next scene, but since no speaking character has entered, only dancers, the scene change is not labeled until after the dance. We shall construe the scene as ending at the end of the trio before the entrance of the dancers. (There is some transitional recitative between the trio and dance; this is included in the transcription.)

4.2 Example 9 contains the opening of our scene. This opening consists of a ritournelle that sets the mood (and, as is common, summarizes the music that follows) and a duet (a quarrel) between Andromède and Phinée. The music of Example 9 is transcribed in a synchronic-diachronic form that allows comparison between the ritournelle and the duet.²⁰ The ritournelle is shorter than the duet but there are clear correspondences, indicated in the Example by the way the two passages are stacked one above the other.

4.3 The correspondences are obvious, but what further interests the Schenkerian is the relation between their structures: they are virtually identical through the second, and largely the third, level of middleground. Figure 10 is a graph of the analysis of the ritournelle, and Figure 11 is that of the duet.²¹ That the ritournelle and duet might form a complete unit together, or are even part of a much larger structural unit, will not yet be entertained (see section 5). Figures 10 and 11 reveal that the two are essentially the same at the stage of fundamental structure (Figure 10a and 11a) and the first level of middleground (Figure 10b and 11b). Further similarities evident on cursory inspection of the graphs in Figures 10c and 11c include the imitation of the head tone’s upper neighbor note as a lower neighbor on scale degree ^4 and as an upper neighbor on scale-degree ^2, unfolding in the course of the neighbor motion to the head tone, and the transfer of the inner voice to form cover tones which have their own upper neighbor motion.²² Of special interest is the subdominant function support for ^4 in the fundamental line in which IV becomes II by means of a 5-6 exchange. The treatment of the resulting A-minor harmony forms the core of the difference between the two passages.

4.4 Since the duet has more content, it can be viewed for our purposes as an expansion of the ritournelle. Expansion can be achieved either by the addition of segments relatively complete in themselves, or by the composing out of the prior material through new layers of countrapuntal elaboration. A ballet, for example, might be enlarged by the addition of segments, while composing out—the thickness of its levels—enlarges a symphony. The study of composing out is the central activity of Schenkerian theory. We are concerned, then, with how the duet is a composed-out expansion of the ritournelle.

4.5 Figure 10b shows an inner-voice ascent, g'–a'–b', which is joined through unfolding with the upper voice at the subsequent level (shown in the graph by the slanted beams). In the ritournelle, the goal tone of the inner voice line, b', continues into c'' while the line’s head tone, g', is regained (m. 4) to initiate the 5-6 exchange. The first expansion in the duet is at this point. Figure 11c shows how the duet’s b' is continued through c''-sharp to d'' (the c''-sharp accompanied by e'' as an additional neighbor to the d'') in a tonicization of D.²³ The second expansion begins at the C harmony that supports ^4 of the fundamental line. The 5-6 exchange, seen in both excerpts, generates an A-minor harmony, which is then tonicized. The prominent f-natural in the bass of both passages makes this modulation explicit. This is also the point at which the cover tone becomes prominent (Figures 10c and 11d). The A-minor content is far more elaborate in the duet than in the ritournelle. The modulation in the duet first reaches A minor in inversion, reinforcing the 5-6 exchange as the source of A minor, before moving on to the root position of the A-minor harmony. Figure 11d shows this elaboration.

4.6 We have assumed the duet to be an expansion of the ritournelle, but if Lully had composed the duet first and then did judicious cutting, he cut in such a way as to preserve the integrity of the Schenkerian middleground. It is probable that he composed two formal units on the same plan with the same motivic material and knew as well as anyone how to create content at later levels of structure.

4.7 My summary of Rosow’s argument in Table 1 includes concepts such as “weak closure” (“Music,” item 4), and the effect of context on formal closure (“Music,” item 11). In connection with the present excerpt, we note that the duet ends with an ornamental turn to the third of the tonic harmony in place of the underlying cadential tonic.²⁴ For all that this may be a common way to end ensembles, it is also a simple way to force a cadence to remain “open” in some respect. This is a detail that presents the compositional surface as less closed than the analytic graphs claim it to be. The duet is reprised in mm. 183–92, starting with the material of m. 11, and in the reprise the cadence is overtly closed, using the formula of the end of the ritournelle, as shown in Example 10. This kind of simple yet subtle distinction between degrees of surface closure is reflected in another Lullian technique, the choice of register in which a cadence is executed.

4.8 The nature of the Schenkerian background leads to distinctions in voicing not made in any other harmonic-tonal theories. In Schenker’s background, the most representative voicing of the main opening tonic harmony has the third of the chord on top or in an inner voice. Regardless of where it originated, the third is necessarily voiced lower at the close of the structure. The concept of “tonic sonority” therefore becomes more refined than in its common use, in that with Schenkerian theory, the sonority (as opposed to the taxonomical category) differs according to its temporal location, as a function of where one is in the course of the background.²⁵ Further, the most representative voicing of the cadential dominant harmony is with its fifth (the line’s ^2) on top and its third immediately below that. Differently functioning triads thus have different sonority. As a by-product of this concept, there is difference of finality between the relative heights of melodic cadences. Given the essential range of a melody, a cadence at the top of that range is relatively open. A convincing formal close is conversely at the middle or bottom of the range. This echoes the tradition of medieval chant; with either an authentic or plagal final, the final is never the high note. In tonal music, a high cadence, however assertive, is in some sense not final. In Lully’s music high cadences are the typical cadences at the ends of recitatives, especially when introducing an air. More universally, internal cadences in Persée that mark a point of demarcation but also carry the drama forward are often high cadences.

4.9 For example, the duet of Example 9 is followed by the air sung by Mérope already referred to in Example 1. This air is in extended binary (ABb) form. The two endings (“B” and “b”) are shown in Example 11. The “B” section ends with a high cadence, softened by the hemiola. The “b” section ends with a low cadence, but the covering e'' moving to an implied d'' is prominent so as to suggest that there is more to come here as well. We observe this “more” in Mérope’s second air of the scene, over one hundred fifty measures later, in Example 12. Here the high cadence of Mérope’s first air becomes the low final cadence of her second air. The cover tone is still there at the end of her second air, but it is farther back in the melody. Together with the air, the back-and-forth dialogue of the scene is also brought to a close with this cadence. What follows this is the trio that acts as the effective finale to the scene. (The cover tone of her air that had been left in register is now picked up as the head tone of the finale.)

4.10 There are numerous such connections lying just beneath the surface, and some lying deep. It is clear that the duets are connected to Mérope’s ensuing airs as a formal unit. She is, after all, very much part of the interaction; she wants Phinée to succeed with Andromède, so Lully ties their segments together. The prominent pitches at the end of the opening duet form the foundation of her first melody. Just as the lovers’ recapitulated duet comes to a convincing close, so does Mérope’s second air. Even though the tunes of her airs as well as their forms differ, Mérope reaches back across the scene to turn an intermediate high cadence into a final low cadence. It will come as no surprise by now to learn that the structure of the primary strain of the final ensemble reflects many of the features of the duet, which is also to say, of the ritournelle that began it all. This is “thinking on a large scale.” Figure 12 depicts a middleground graph of this final ensemble from mm. 203–31 (Audio 1, Appendix). The ensuing segment is an “Amen” consisting of a subdominant prolongation followed by a transposition of that same material to the tonic harmony.

4.11 Of special interest in this trio is the recurrence of neighbor notes to the head tone of the line, whereas in the duet these neighbors were echoed on later tones of the fundamental line. The point of departure from the head tone is also propelled forward through its harmony’s transformation into a subdominant of II.

5. Arch Form and Symmetry

5.1 Rosow discusses Act V, scene 1 of Armide at length. One of this scene’s formal attributes is an arch form underlying almost the entire scene. Table 2 is drawn from Rosow’s summary.

This portion of her table reveals a well-defined arch whose shape is contributed to by both Quinault and Lully. The arch-form proper begins at the descriptor “4 individual lines,” but the earlier segment is included here to account for the rhyme scheme and to make sense of the text since the arch begins with a response. The scene as a whole is in C minor; the two central airs are in C major. The group of lines defined by the arch-form is also unified by the rhyme scheme, by patterns of line lengths, and by position of cadences.

5.2 We can discover a somewhat similar symmetry in our scene from Persée, but based on cruder standards than Rosow’s.²⁶ What stimulates this plan is the return of the opening duet near the end of the scene, reinforced by Buford Norman’s observation that the dramatic form of the scene is ABA', suggestive of ternary.²⁷ Figure 13 shows the scene divided into six subsections, grouped in pairs to reflect the ABA model. In ABA forms, the junction of the “A” and “B” sections is conventionally a major point of demarcation, the point at which repeat signs are found in dance forms. This point occurs at m. 55 of the transcription (Appendix), and coincides with a change of mode from G major to G minor. In contrast, the junction between “B” and “A'” is elided; the point of change is assigned to Andromède’s words “Juste ciel!” which pull the motivic material back to the ascending fourth/descending thirds so characteristic of the opening duet.

5.3 The pairing of the six subsections is straightforward: The identification of “A” and “A'” (subsections I–II and V–VI) is overt with respect to musical material, and has been referred to in a variety of ways in the discussions of Examples 9–11 and Figures 10–12. The definition of the “B” section (subsections III–IV) as distinct from the others is explained in Norman’s discussion and reflected in the change of mode at the beginning of III, and negatively, by the separation of subsection IV from the recapitulatory material of subsection V. Reinforcing this is the evocative difference of treatment between the G major of the outer (“A”) sections and the G major within the “B” section (subsection IV): the outer sections have prolongations of A minor, while subsection IV does not have even an A-minor cadence.²⁸

5.4 Subsections II-V in Figure 13 are groupings of still smaller units. Considerations in the formation of these units were the assumptions that a change of character together with a change of texture creates a unit, a recitative sung by both Phinée and Andromède form a distinct unit, and a contiguous recitative and air—in either ordering—sung by a single character form a unit. Grouping the units into subsections relates to issues of texture, key change, and the boundaries formed by the large ABA pattern. Thus, subsection I is separated from II by the change of texture from orchestra to singers. Subsection II is separated from III by the change of mode (which correlates with the large formal division from “A” to “B”), and change of mode likewise accounts for the separation of subsection III from IV. Subsection IV is separated from V by the recapitulatory return of the large “A” section. Subsection V is separated from VI by the wholesale change of texture (the first vocal ensemble), which balances the instrumental opening of the scene (subsection I).

5.5 This balance between subsections I and VI contributes to the symmetrical formulation of Figure 13. It is justified by the use of these subsections as the scene frame,²⁹ and is reinforced by their musical-structural affiliation (as evident from the comparison of Figure 10 with Figure 12). While the ABA formulation is palindromic in only the most simplistic terms, as though the sections had no time-occupying material, Figure 13 also makes a stronger assertion of the reflexively symmetrical disposition of its component pieces. While Rosow’s analysis contains an arch, our formulation merely points to the symmetrical disposition of these individual pieces around a center point, coinciding with the center of the B section. This is reflected in the nested brackets at the top of Figure 13. The Figure thus asserts two disjunct formal analyses, Norman’s ternary and a palindrome inspired by Rosow’s arch in Armide. The reflexive symmetry of units within sections is subverted only in groups II and V by the treatment of the opening duet and the following, placating, air by Mérope as an ordered unit within both the first (“A”) section and its recapitulation (“A'”).

5.6 Calling a structure an “arch” implies formal characteristics which might press our Schenkerian orientation, while not, to be sure, disavowing the musical utility of such a structure otherwise, as Rosow discusses throughout her article. An arch implies a keystone without which the structure would collapse. It implies a high point without which the approach and dénouement are meaningless. Like our symmetrical disposition of Figure 13, it suggests a palindrome, a forward and then backward directionality inherently obscure if not impossible in Schenkerian structural terms. Schenkerian theory seems to have no ready place for the palindromic concept, especially translated as a moving toward a climax and then a fading. It has more use for the concept of a departure and return, but still under the control of a line that ultimately descends monotonically. Rosow’s arch and our symmetrical disposition may turn out to be, therefore, a prize instance of the tension between structure and design: Schenkerian structure and the design imposed by poetry, genre, and internal form.

5.7 This is not the place for an exhaustive discussion of the ways in which a Schenkerian sensibility might cast further light on symmetrical forms, but the possibilities are limited. There certainly are pieces informed by palindrome, mostly twentieth-century works. In tonal music it is harder to make such a thing work. Haydn's palindromic minuet³⁰ is impressive for its existence, not its beauty, and the harmonic structure is still heard from left to right, so to speak. Elizabeth Sayrs has pointed out, however, a straightforward tonal reversal embedded in Hugo Wolf’s Und steht Ihr früh am Morgen auf.³¹ If there is to be a Schenkerian departure-return kind of structure to an arch in tonal music, then the central component would have to be subsidiary to the framing motion. A typical arch-like ABA of the late eighteenth century might appear at the first level of middleground as in Figure 14. The ABA reveals a nesting of the “A1” and “B” sections under the umbrella of the line represented as the content of “A2”, which, although the same as “A1,” has higher status due to its location. Although the structural neighbor note of section “B” is literally a high point, it has lower status than have the “A1” and “A2” sections that frame it. This is to say that even the most simplistic symmetry of ternary form falls away under Schenkerian hierarchization.

5.8 Dramatic turning points are plausible in such symmetries, however, and we have one in our scene from Persée. The boundary between Andromède’s first air and Phinée’s second air is close to the exact center of the scene. At the axis point three things happen: the mode returns to G major; Andromède runs out of patience and begins to treat Phinée with sarcasm; and Phinée starts to blaspheme. (Parallel events to all of these also radiate about the keystone of the arch in the Armide scene.) As in Armide, the center point divides two airs, one for each of the two main characters; in our case, these are further surrounded by two more airs similarly apportioned. Whether or not this throws dramatic weight toward the center of the sequence of airs, the string of airs itself reflects a series of strong assertions, as is characteristic of airs in general.

5.9 Ultimately we are faced with a question inherent in Lully but which can be invoked in a broader domain. When we encounter a scene united by an overall key scheme, rife with cross-references of musical material, with clearly open cadences contrasting with closed ones, and where repeated material may differ from its first statement only by the degree of melodic closure, can we posit an overall Schenkerian structure to the scene as a whole? If we can, the result would be more complex than, but on the order of, Figure 14. If we follow the more conventional view of operatic construction as composed of discrete numbers (similar to the discussion in paragraph 1.4 about the segments of the Renaissance motet), we would elevate each of the sections of Figure 14 to fundamental-structure status. They would remain separate and complete small compositions, perhaps linked together by extra-structural recitatives, and perhaps subordinated one to another by dramatic means outside harmonic-contrapuntal structural constraints.

6. The Challenge of the Schenkerian Approach

6.1 In his series of treatises on Wagner’s mature operas, Alfred Lorenz framed the issue in essentially the same way as is being entertained here: there may be numerous small forms, but they are segments of an overarching form extending to scenes, acts and even entire operas.³² Here we are limiting the question to scenes. Properly understood, Schenkerian theory forces the issue upon us. The differences between recitatives and airs in Schenkerian terms are illustrative.

6.2 Recitative is active; it develops the drama. The air is assertive; it reflects a character’s attitude, regardless of emotional content. This does not say that the drama cannot be advanced through airs, but that the orientation of the air is assertion as opposed to interchange and exposition. The distinction between air and recitative is not in the sphere of expression but in the character’s state of mind when the utterance is made. Whenever a character is sure of what to say and is making a grand declaration, he sings an air. Recitative involves dramatic interaction. In recitative, what the character says is not “canned”; it is (dramatically) spontaneous. This can easily lead to the assumption that a recitative that debouches in an air reflects a sense of conviction developing in the character’s mind. Near the end of our entire scene, recitative gives way to the recapitulation of the duet. The characters will develop no new relationship through interaction. They are back where they started, but with a worse taste in their mouths, and they sing at each other.

6.3 We expect airs to have complete Schenkerian structures. Their closed form and unity of key at least suggest that. We do not, however, typically expect recitatives to have such complete structures. Apart from its typical lack of unity of key,³³ the surface content of the recitative is normally jagged and difficult to subsume under the same constraints as “formal” music. This follows from its avowed purpose, to reinforce dramatic speech, rather than musical speech. Interestingly, in keeping with the lyric quality of the style, Lully’s recitative often reveals elegant recurring patterns, visible through at least a late-middleground lens.

6.4 Examples 13, 14, and 15 show patches of recitative in Armide; their analyses are in Figures 15, 16, and 17 . The pattern of counterpointed and subarticulated fourth and fifth spans appears in all of them, pulling together the recitatives and helping them toward formal unity. Although recitative is less likely than air to have a complete tonal structure, in Lully's hands it has considerable internal coherence. And in anyone’s hands, by virtue of its inability to stand for itself tonally, it is easily marked for subordination to a larger structure.

6.5 Again, this is not to say that recitative cannot have a complete formal structure. An example of such a complete structure can be found in Persée in the exchange between Phinée and Andromède at the first change to G minor (Audio 2). (See Figure 13, at the beginning of section “B”, and score, mm. 55–75, in the Appendix.) Figure 18 is an analysis of the recitative. Although the recitative passage proper ends on a half-cadence, the analysis shows the line as completed in the tonic harmony of the beginning of the next air (Phinée’s first air in Figure 13).

6.6 Less common are complete structures in recitative sung by a single character. But when one does arise, a most beguiling question arises with it: does the closed (therefore air-like) structure reflect the character’s state of mind? The use of recitative instead of air would suggest a subtle dramatic ploy in which the character seems to be participating in an exchange, but whose mind is made up.

6.7 We are left with the relative sureness that airs can be expected to have complete Schenker-tonal structures, and that some of the recitative can also claim such completeness. In deliberating as to whether or not entire scenes form such structures of a higher order, we have the obvious task of integrating the dependent recitative passages into a hierarchy interacting tonally with their surrounding material. More difficult, perhaps, is the distribution of priority with respect to closed structures, whether air or recitative. Again, this hierarchization of tonal structures is the kind of concern that Schenker asks us to consider, but it is not yet answered. We do know, though, that Lully’s music fits the Schenkerian model, at least for all the pieces examined. On the basis of these few instances we see, for example, that, in addition to stylistic features revealed throughout this article through the application of Schenkerian theory, Lully favors the fundamental melodic space of a fifth to such an extent that he preserves the fifth space through the course of the structural melodic descent by the use of cover tones; that he supports the fourth scale degree in the fundamental line with some form of the subdominant harmony; that although there are expected moves to the dominant at formal points of punctuation (as in Figure 18), there is a marked absence of broad dominant prolongations, so characteristic of eighteenth-century music; and that Lully’s sequences are subtle and brief, often consisting of 5–6 exchanges that might be clouded as such by leaping interpolations in the bass.

6.8 We see, then, that Lully’s sense of hierarchy, of large-scale distribution of functions, is revealed through the Schenkerian lens as well as through the lens of Rosow’s model, and that the two perspectives complement one another. It remains to be discerned if the large-scale coherence indicated in her model is entirely reflected in the tonal structure of scenes as well.

---