Inventing tradition: Webern's Piano Variations in early recordings

Webern's Piano Variations Op. 27 were first performed two years before the outbreak of Second World War, but it was only after the war—and in very different aesthetic and ideological circumstances—that a performance tradition developed. Through a combination of computer-assisted close reading and contextual analysis of the first movement, this article traces the competing stylistic and ideological directions of early performances—one deriving from the pre-war tradition of the Schoenberg circle, the other from the European avant-garde associated with Darmstadt and die Reihe —and the processes of negotiation through which a more or less consensus interpretation developed by the 1970s. My aims are to explore ways in which empirical approaches can be deployed for musicological purposes, to relate recorded performances to their historical contexts, and to link the story of Op. 27 to the development of music-theoretical approaches that see performance style as a function of compositional structure. Key recordings on which the article is based are presented online as Sonic Visualiser session files coordinated with the text.

There is an affinity between the situation of Op. 27 after the war and historically informed performance: in each case notations from one historical context are interpreted in another. But there is an obvious difference. Leading performers from the Schoenberg circle, including Rudolf Kolisch and Eduard Steuermann, attended the Darmstadt Ferienkurse (annual summer schools), which were initiated in 1946 and became the principal anvil on which post-war musical modernism was forged; Stadlenwho had now settled in London-attended from 1948 to 1951, and it was there, on 31 July 1948, that he gave the German premiere of Op. 27. Fortunately a private recording was made, and though it is of poor quality-and was commercially released only in 2006-it reveals a performance bearing all the marks of the pre-war style. 1 As Christopher Fox (2007a: 11) remarks, however, 'it was Webern on the page which became the great touchstone for younger composers', and in this context a new performance style developed that ostensibly owed little to pre-war traditions. Stadlen was vocal in his opposition to this new style, insisting in a number of publications (the best known of which is 'Serialism reconsidered', published in 1958) that Webern's intentions had been quite different. All this made the Piano Variations an important arena within which post-war modernist values were negotiated. It also resulted in an unusually extensive recorded heritage: as of 2012 there were 65 commercial releases, falling into two groups separated by a six-year break (1948-77 and 1983-2012). I shall suggest that by the 1970s Op. 27 was not only firmly ensconced within the modernist canon but had also acquired a well established performance tradition, and so it is the first group-consisting of 30 recordings-with which I am concerned in this article.
This article is an output of the project "Shadows of meaning: Webern's Piano Variations on record," funded by the UK Arts and Humanities Research Council, whose assistance is gratefully acknowledged. The recordings on which it is based were collected jointly with Miriam Quick, and more collectors helped us than can be named individually. I am also grateful to Daniel Leech-Wilkinson for his constructive reading of the draft article.
They are listed in Appendix 1. 2 Building on Timothy Day's (2000: 178-85) brief but perceptive account of how performances of Webern's music changed over the years, Miriam Quick (2010: 103) writes in a doctoral thesis, which includes a chapter on Op. 27, that there were 'in the late 1940s, 50s and 60s … not one but two Webern performance styles: the Viennese tradition and the avant-garde "Darmstadt" practice that rejected-or more accurately, perhaps, ignored-this tradition'. In this article, which focusses on the first movement, I explore recordings of Op. 27 that represent both these traditions, relying on recently developed techniques of computer-assisted close listening. 3 I also consider relationships between recordings and the contexts from which they emerged, including other recordings, critical and analytical writings, and the aesthetic ideologies that informed them. Woven into the story is the development of theoretical approaches that have understood performance style to be a function of compositional structure-what, borrowing a term from Susan Melrose (1994: 215), I call page-to-stage theory. If, as Nicholas Mathew (2011: 61) writes, Stadlen's annotated score represented 'a serious blow to the Darmstadt image of Webern', it also problematises some still prevalent assumptions about the relationship between scores and performances.

The presence of the past
In an interview with Gunther Schuller (1964: 28), the leading pianist of the Schoenberg circle, Eduard Steuermann-to whom Webern dedicated his Piano Variationsremarked that 'Webern himself was the freest interpreter of his own music that could be imagined', adding (with specific reference to the Concerto, Op. 24) that 'He played so freely that I hardly could follow the music, but it was extraordinary'. And in this context 'free' does not mean unconsidered: that becomes evident in a letter that Webern sent to the conductor Hermann Scherchen in 1938, in which he described how the fugal subject should be played in his arrangement of the Ricecar a 6 from Bach's The Musical Offering: 'from g via f# to f faster, then holding back a little on the e♭(accent given by the harp) and again rubato on the trombone progression (including the tied e♭ of the horn where the trombone has a crotchet rest in bar 6) '. 4 The intricate performance style current in pre-war Vienna is fairly represented in Stadlen's 1948 recording, which I shall describe at some length. But in order to do so I need to sketch out the compositional framework represented by Webern's original score of 1937, as well as the annotations in Stadlen's score. Because the small rhythmic values used in the movement-particularly its central section-make the usual bar:beat citations unwieldy, I often refer to particular points by event number (preceded by E). These correspond not only to the score at Appendix 2, but also to the Sonic Visualiser session files through which the majority of the recordings in the public domain may be accessed (see Appendix 3); fortunately this includes most of those I discuss in detail. The session files include a piano-roll representation that further simplifies cross-reference between sound and score, together with tempo graphs aligned with playback. These graphs are more perceptually salient than is often the case in post-tonal music, because so much of this movement is isochronous, and they will make my observations easier to follow. There are also dynamic graphs, although their usefulness is more limited. 5 In short, my analyses will make more sense-and be much more interesting-if read in conjunction with playback on Sonic Visualiser.
It is hardly necessary to point out that the first movement of the Piano Variations falls into three symmetrically disposed sections, which I shall call A1, B, and A2; the outer sections exactly match one another, apart from the serial derivation (which means the pitch characteristics and patterns of hand crossing are quite different), dynamics, and an alteration to the final two attacks that gives A2 a stronger cadential quality. Each of the A sections consists of a series of rhythmically defined gestures-what I shall call groups-that are made up of of isochronous sixteenth notes; in each section the groups are organised into two clearly marked parts, the second longer than the first. The first part of A1 comprises four groups each of four attacks notated in the form of enclosed pairs of notes and separated by sixteenth-note rests; it is an exact palindrome, with the mirror axis (what Webern scholars call Spiegelbild) in bar 4. Webern's original score marks it pp, but Stadlen's annotated score adds a few further markings in Webern's hand. The melodic and expressive thread is brought out through diagonal lines and arrows between the notes to be connected, together with tenuto signs and tiny-though not literally playable-pairs of hairpins; there is also a 'tpo' marking on the last beat of bar 4 (E9), implying a preceding ritenuto. The other annotations are added by Stadlen, based on his recollections (which may or may not have been accurate, but at least represent his impression of how the music should go), and fill out this picture. Stadlen writes 'subdued plaint' at the opening (putting it in inverted commas, which should mean it is a verbatim quotation from Webern); 6 rings the principal melody notes, which are in each case the highest note at that point; adds a pp marking at the beginning of the last of the four groups (E13), so implying some kind of dynamic arch; and at the same point also adds 'Echo', again in inverted commas.
That last instruction might be thought redundant, but in his recording Stadlen observes it to the extent that for a moment the music becomes inaudible. Otherwise the main feature of his playing of this passage is the irregular and curiously-perhaps plaintively-halting way he plays the sixteenth notes. The first sixteenth lasts fully 82% longer than the second; the unexpected prolongation of E7 (73% longer than the previous sixteenth) creates an oddly limping effect that turns into a kind of lilt at E9-12.
The effect is less of projecting structure than of searching for meaning, and in this way it projects agency. You have the sense that Stadlen is trying to say something, even if you don't quite know what it is. At the same time a traditional scheme is in operation: he plays the opening seven bars as an antecedent, a question, to which the rest of the A1 section-the consequent-provides a series of answers in the form of more sustained gestural groups. A three-bar group begins at bar 8 (E17), where Stadlen not only writes in 'new vigour' but also replaces the printed p by mf; all this underlines the stretto composed into the music through the overlapping of the paired attacks. A climactic group of four bars follows at bar 11 (E26), where Stadlen adds 'more!' (this time in inverted commas), together with huge hairpins centred on the Spiegelbild (bar 13, E31-2). A wavy line over bar 14, as Stadlen explains in his preface, indicates 'a modest holding back', which in combination with a caesura added in bar 14 turns E35-7 into a closing gesture within this four-bar group. Two shorter groups (E38-42 and E43-45) conclude the A1 section: added 'tpo' markings in bars 15-17 imply unnotated ritenutos in the preceding groups, matching the notated one in the final group. The annotations, then, generally flesh out the implications of the dynamic markings in Webern's original score-an overall rise and fall of intensity through bars 8-18, matching the group lengths.
As might be expected, Stadlen's recording of these passages is more or less in line with the annotations, though not completely so, and of course any performance contains detail that is not captured by verbal or graphic annotations. (John McCabe [2003: 17], who made his own recording of Op. 27 in 1969, writes that 'all those little interpretative nuances that simply cannot be written down, are what brings the music to life'.) At the beginning of the three-bar group, in bar 8, Stadlen hurries through E17-19 and then makes E20 dramatically longer (the sixteenth note at E20 is 88% longer than that at E19), creating the same kind of halting effect as earlier. By contrast he abbreviates the rest before the next four-bar group (bar 11, E26), almost knitting the two groups together and in this way creating a clear linear continuity from the a 2 at E24 to the b 2 at E26 (you can imagine him thinking of those ringed notes). He accelerates from E27 to E29, to which he gives a dynamic emphasis, and from where there is an overall decrease of both tempo and dynamics to the end of the group, though it is complicated by the prolongation of E31 (of which more shortly). Perhaps it is to avoid an excessively disjointed effect that Stadlen completely ignores his own caesura before E35, instead playing through it with the legato articulation that characterises his performance of this section in general. Where-as we shall see-later recordings shape these groups through the principles of temporal and dynamic phrase arching that spread rapidly in the aftermath of the war (Cook 2013: chapter 6), Stadlen in this way shapes them more in the form of a decrease from fast to slow, loud to soft, intense to serene. And he follows the same general pattern for the rest of the A1 section, though in a less exaggerated manner.
Also contributing to the overall expressive trajectory is Stadlen's use of what was already in 1948 the old fashioned practice of hand breaking or dislocation (spreading the notes of a chordal attack), often employed as a signifier of intense feeling. Stadlen's purpose however seems to be different. He employs it at E18, 21, and 24, and again at E39, which is to say at all the points where the pattern of interlocking pairs of notes changes. The piece has begun with one-and two-note attacks, but the first three-note attack is at E18 and the first four-note one at E21; at E39 the rests between the pairs of notes disappear, representing a further tightening of the stretto. In textural terms, then, the group at E38-42 is the most intense point in A1, but at the same time it is enclosed within a process of dynamic winding down from the climax of the previous phrase, and the result is a kind of expressive bottling up. Yet the hand breaking at E39 has no obvious part in this process: rather it belongs with the previous hand breaking, the intention presumably being to mark an aspect of textural design that Stadlen thought it important to bring out. At such points his playing both expresses and induces in the listener (or analyst) a kind of hermeneutic approach-as if meaning is to be found beneath the immediately perceptible surface of the music-which is perhaps not surprising in an interpretation shaped through intense coaching by a composer who, as Julian Johnson (1999: 186, 202-3) says, 'hardly ever wrote a piece which did not have, for him, significant extra-musical associations'. (Op. 27 is no exception.) The 'for him' is crucial: the associations were private, hidden from view. And here a link might be made with a curious ambivalence in Stadlen's relationship to the Piano Variations.
Stadlen claimed a position of authority in post-war debates about musical modernism through his insider knowledge of Webern's conception of this serial composition. But at the same time he used his position to further a polemic against serialism, speaking in his 1958 article of the 'fundamental errors and contradictions' of Schoenberg's system (Stadlen 1958: 22). He squared this circle by insisting that, in the coaching sessions, Webern 'acted as if he himself were not aware of the serial aspect of his work' (Stadlen 1958: 16), in other words as if-like the extra-musical associations-it was a private matter. Instead Stadlen emphasised what in an interview with Hansjörg Pauli he called the 'pre-classical shapes' of the music: 7 the palindromic patterns into which Webern fashioned the music, patterns that are afforded but not mandated by the serial organisation. In the preface to his edition Stadlen calls the palindromes 'autark' structures, presumably meaning that they appear self-sufficient, and his recording suggests an almost obsessive concern with them. The most obvious example in the A1 section is at E31, which he prolongs in such a way as to emphasise the Spiegelbild (E31-2)-an effect that, as I put it, complicates the decrease of tempo and dynamics at this point. More obviously than E39, this is-as Naomi Schor (1987) would call it-a 'bad' detail, one that is not synthesised or sublated within the larger context but rather intrudes upon the listener's experience. It is this intrusiveness that turns it into a hermeneutic marker, an invitation to interpretation akin to Lawrence Kramer's (1990: 12) illocutionary 'breaking points': once more you have the sense that Stadlen is trying to say something. And while this point in the annotated score is marked 'molto espressivo, particularly the repeated notes', there are other, unmarked points where Stadlen's playing is apparently shaped by the palindromes. Indeed one of them is the matching dyads at E7 and E10, the prolongation of which creates the oddly limping effect I described as searching for meaning, and which comprise the same pitches as E31-2 (c 1 and d 2 ). It is as if Stadlen wanted to bring out anything except the serial structure.
In terms of organisation-and despite the curiously opaque appearance of Webern's rhythmic notation-the B section (bars 19-36, E46-132) looks more straightforward in the score than it sounds in Stadlen's recording. Successively longer groups of isochronous thirty-second notes (E46-51, 58-66, 73-84) alternate with groups of six slower notes of varying lengths centred on the Spiegelbilder: Stadlen marks the first of these 'particularly intense'. These groups of six notes involve prominent melodic tritones in contrasted registers, while the hand crossings that are scripted in the scoreand that during the A sections create an effect of visual and kinesthetic interweavinghere take on an overtly dramatic quality. (That Webern was aiming at such a quality is demonstrated by the annotation that Stadlen-supposedly quoting Webern-adds at bars 20-21: 'right hand replacing left not until the last moment, almost too late'.) The hand crossings are perhaps the most immediately striking feature of Op. 27 in performance, and the reason I do not focus on them in this article is that I have done so in another (Cook forthcoming): it is based on two televised performances by Glenn Gould, and addresses the issue of how physical action can be encoded in recorded sound.
All these groups-both the isochronous and non-isochronous ones-are marked 'rit. . . . tempo' in Webern's original score, and for each pair of groups the same major seventh dyad acts as a terminus, with the 'tempo' marking following it (for example f# 1 /f♮ 2 at E52 and 57). This parsing is made explicit by the brackets Webern added in the annotated score, which suggests that Stadlen had trouble with it. And while in the original score the second and third groups of thirty-second notes are given a dynamic arch from p to f and back to p, the annotated score adds a crescendo to give the same shaping to the first group, which is how Stadlen plays it: a flurry of notes slowing from E48 to target E50 before coming to a stop on E52, with dynamics closely following tempo (as may be seen in the Sonic Visualiser session file). Of the 'tempo' marked in Webern's score at E53 (or at E68 or 86), however, Stadlen's recording carries no trace. At the note-to-note level the extent of flexibility-the thirty-second note at E50 is 160% longer than at E47-is in line with the annotation at the beginning of the B section ('freely improvisatory', in inverted commas). But the following six-note group is taken slower, averaging around dotted eighth = 20, that is at half the notated tempo; Stadlen accidentally foregrounds the hand-crossing by mis-hitting E53. The next two pairs of groups are played in much the same way, perhaps the only detail that calls for comment being the separation of E73-5, which results in a gesture that is rather convoluted but precisely corresponds to the annotation Stadlen ascribes to Webern at this point: 'in parenthesis'.
The increasing length of the three thirty-second-note groups gives them a directional force that culminates in the much more extended passage of thirty-seconds from E91; Stadlen's only intervention during these three groups is to slightly increase the tempo of his initial flurries of notes from one to the next. It is the remainder of the B section-E91-121, and after a short rest a briefer passage of similar material from E122-32-that prompted my comment about the music sounding less simple in Stadlen's recording than it looks on the printed page. It is again not hard to parse the passage. There is first a group of eleven thirty-seconds (E91-101), modelled on the previous thirty-second-note groups and again marked as a dynamic arch, with the last note overlapping a group of three tritone-related notes (E101-03) that coincide with the Spiegelbild and are distinguished from the rest of the section by being marked staccato. This in turn overlaps with group of twelve thirty-seconds (E103-114), now reaching ff and overlapping with a further group of three staccato notes (E114-16); then comes a group of six thirty-seconds (E116-21), again reaching ff but torn off mid-phrase. Finally there are five thirty-seconds (f diminishing to p), overlapping with the third and last of the groups of three staccato notes (E126-28)-which in turn overlaps with another five thirty-seconds (diminuendo to pp) that peter out into a pause before the A2 section. The rhetoric of building to a climax, tearing it off, completing what was left hanging but in a fragmentary or desultory way, and concluding with a pregnant pause is familiar from any amount of nineteenth-century piano music conceived within a broadly narrative or dramatic paradigm. As Robert Wason says (1987: 99), 'the musical meaning of this stormy passage is clear'.
To my ears Stadlen's recording confuses this basically simple and legible design for reasons that are indicated in his edition. In these thickly annotated bars Webern adjusts or adds dynamics, adds sforzandi at E112 and 117 together with an accent at E120 (left hand), and most importantly specifies caesuras before E100, 113, and 125, that is before each of the groups of three staccato notes. It is at the first of these caesuras that we see perhaps the most famous of all these annotations, in Stadlen's writing but apparently quoted verbatim: 'left hand like a mysterious drum'. The annotation continues, now in Stadlen's words, 'ditto bars 33 and 36. Sharply set off against what comes before and after'. That is what Stadlen does, but the result is problematic. It is as if, at E100, Stadlen suddenly turns aside to deal with other business in the middle of the gesture before resuming it at E105 (think of a cat breaking off to wash in the middle of playing). And though the effect is less extreme from E113, here there is a further problem: after the interruption there is not enough time to develop sufficient momentum for the end of the phrase at E121 to sound like a tearing off-or for E122-32 to sound like a completion of what was left hanging. Finally, the breaking up of the sustained passage indicated in the original score means it no longer sounds like the climactic block towards which the previous, increasingly long thirty-second-note groups have been directed. A certain effect of splashy rhetoric does not altogether make up for the lack of interpretive direction, and though later pianists also struggled with this section, it is hard not to wonder whether it was in part Webern's coaching that confused the issue.
After a gigantic pause (Webern added a pause sign in Stadlen's score), Stadlen plays the A2 section in a manner that does not exactly replicate the nuances of A1 but embodies the same general features; that is in line with the annotations, which indicate (perhaps redundantly) a 'tone of voice as at the start' in bar 37, 'onward' in bars 44 and 47, and 'very warm and heartfelt' in bars 48-9-all of which come to much the same as the annotations at the corresponding points of A1. To leave it at that, however, would be to ignore a massive discrepancy between the score-either score-and almost all the recordings of Op. 27, Stadlen's included. Other than the frequent 'rit' and 'tempo' signs, the only indication of tempo in Webern's score is at the beginning of the movement (dotted eighth = ca 40), and Stadlen's score adds nothing-though in his preface he says that Webern's metronome marks in this piece 'seem to me considerably faster than the tempi he wanted', and instead suggests sixteenth note = 96 (i.e. dotted eighth = 32). In practice he plays still slower, with the average tempo of the first section being dotted eighth = 28, and the movement as a whole 26. 8 Such figures have to be interpreted with caution: they impose a spurious regularity upon Stadlen's highly mobile tempos. I have cited occasional percentage deviations to make the point that Stadlen gives nominally the same note values wildly different interpretations at different times, in other words, that he plays with constant rubato. But the massive discrepancy to which I referred is something else.
Stadlen uses tempos that may not be constant but are obviously different in order to characterise different materials, such as his tempo of around 20 for the groups of nonisochronous notes in the first part of the B section: this is the principle behind the multitempo plans adopted by early twentieth-century pianists for movements built out of strongly contrasted materials (see e.g. Cook 2013: 115-16). Despite this his average tempo for the B section is just a little faster than for A1 (29 versus 28). By contrast he plays A2 much slower, at an average tempo of 24. Other than the serial derivation and its consequences, as I said, the two A sections are close to identical-except in terms of affective expression: played slower, A2 takes on a quality of serenity (or resignation, or subjectivity, depending on your preferred scene) that makes it quite different from A1. And it's not just Stadlen who does this. Playing A2 significantly slower than A1 is standard practice: there are just three out of the 65 recordings in which it is played faster, one from 1977 and the others from the 1990s. In the absence of any notational or other documentary basis, this is a purely aural/oral tradition, and I know of no reason to suppose that Webern envisaged it. Yet it is a basic determinant of Op. 27 as we know it. Symmetrical on the printed page, the form of this movement as heard traverses an emotional trajectory that is the creation of its performers.

Continuing tradition
We know from Neil Boynton's research (2002) that Webern also coached Else Cross in the performance of Op. 27. Like Stadlen she emigrated in the late 1930s, and subsequently taught at the Royal Academy of Music, 9 where a prize for performing contemporary piano music is named after her. While she did not record the work, however, there is a recording from 1954 by the French pianist Jeanne Manchon-Theis, another contemporary music specialist who is supposed to have been coached by the composer. Hard evidence of this appears to be limited to the statement on the record sleeve that the pianist 'gives us an authoritative interpretation approved by the composers themselves' (the record also includes works by Schoenberg and Berg), but, in an at times acerbic correspondence with the composer William Bolcom, Anthony Stadlen-Peter's nephew-referred to it as established fact: in a letter dated 14 February 2012, he recounted how moved Peter had been on hearing Manchon-Theis's playing of Op. 27, 'because he knew immediately that she must have studied them with Webern'. 10 And in her thesis, Quick (2010: 198-9) pursues a comparison between Stadlen's and Manchon-Theis's recordings, writing that 'Their approaches are quite similar: both are very expressive, with very wide fluctuations in tempo and lots of sustaining pedal'. Other shared features include a generally lyrical tonal conception with a consistently legato articulation, a focus on detail, and some indications of the multitempo plans adopted by the previous generation of pianists.
There is, in short, enough similarity in basic approach for the differences to be telling. For instance, where in the A1 section Stadlen begins the groups at E17-25 and E26-37 at a faster tempo and then slows down, Manchon-Theis shapes both as tempo and dynamic arches (though they are beginning-rather than end-weighted). And where in the B section Stadlen begins E46-52, E58-67, and 73-85 with a flurry of short notes, Manchon-Theis smoothens their profile into something closer to arches, again mirrored by dynamics, and creating a more measured effect reminiscent of recitative: her playing is less impulsive than Stadlen, more classical. (Like Stadlen, Manchon-Theis takes no notice of Webern's 'tempo' markings in bars 20, 24, and 28.) And while the isochronous and non-isochronous groups are treated without any obvious hierarchy, Manchon-Theis creates a certain sense of direction throughout this passage by building on something Stadlen did, but now more systematically: where each of his flurries of notes peaked a little faster than the previous one, Manchon-Theis increases the tempo of her archshaped groups in line with their increasing number of notes, so that the sustained thirtysecond-note passage from 91 to 121 becomes the fastest of all. Although the first part of the B section has never settled down to a consistent tempo-itself a characteristic of early twentieth-century pianism-there is a clear sense that a quite different tempo, around 40, has been established during this climatic passage.
Manchon-Theis handles this second part of the B section differently from Stadlen, and perhaps more successfully. Where Stadlen breaks up the continuity, in line with the caesura marking Webern pencilled in after E99, Manchon-Theis continues at full speed; only at the second of Webern's pencilled caesuras-after E112-does she momentarily hesitate, before resuming motion and breaking off more strongly at the notated rests after E121. In this way she avoids Stadlen's premature loss of motion at E99, and throws extra weight onto the points that, in his pencilled annotations, Webern marked sff: the matching B♭/D♭ dyads at E112 and 118. (I know of no reason to think Manchon-Theis had seen Stadlen's score, so this conceivably offers some confirmation of the story that Webern coached her.) She takes the average tempo of the A2 section down to 23, and flattens the tempo and dynamic profiles in a way that Stadlen does not. I described Stadlen's playing of the A2 section as serene, but Manchon-Theis's is more so. Johnson (1999: 233) describes serialism, in Webern's hands, as 'the means of a music that does not trace a narrative of the subject', and in Manchon-Theis's playing of this section there is a hint of the inorganic or objective quality that modernist critics attributed to Webern's music through the use of terms such as 'geometrical' and 'crystalline'. All the same, her historically conditioned playing can hardly be said to embody the same values as the sleeve note, by the composer Alexandre Spitzmueler: in Op. 27, he writes, serialism is 'pushed to the extreme point of contrapuntal rigour. Every ornament is excluded, and the composition is governed entirely by logic'. One wonders whether Spitzmueler ever listened to the recording.
Manchon-Theis was born in 1902, which makes her the earliest born performer to have recorded Op 27. Born in 1910, Stadlen was the fourth earliest. The second and third earliest were both American pianists born in 1908, and both known for their playing of Beethoven and Schubert rather than as contemporary music specialists: Webster Aitken, who recorded Op 27 in 1961, and Beveridge Webster, whose recording dates from 1967 (by which time he was 59, the oldest pianist to record Op. 27 during the period covered by this article). Both, as it happens, trained with Artur Schnabel, while Aitken also studied with Emil von Sauer-himself born as long ago as 1862. Given the extent to which performers are generally believed to internalise stylistic practices around their late teens or early twenties (Leech-Wilkinson 2009: 250), one might expect to hear aspects of pre-war performance in their recordings, despite their significantly later date, and this proves to be the case. Both share obvious features with Stadlen or Manchon-Theis, ranging from legato articulation and use of pedal to hand breaking: Aitken uses the latter sparingly, for example to bring out the Spiegelbilder, but Webster uses it so frequently that it becomes an audible expression of Webern's manual choreography. Both recordings exhibit the unsettled tempo in the first half of the A sections that, in Stadlen's performance, I described as searching for meaning. And in the second half of the A sections both follow Manchon-Theis in consistently shaping the groups as beginning-weighted arches.
But as usual it is the B section that offers the greatest interpretive contrasts. Sapp's similarity analyser detects an extremely close correlation in timing between Stadlen's and Aitken's playing of E46-53 and E58-67-but the aural effect is quite different, and this presumably reflects the perceptual salience of absolute tempo. 11 Aitken's average tempo during the B section is an exceptionally slow 22, which means that Stadlen's flurries of initial notes turn into vocalistic declamation; the effect is that much stronger in the corresponding group E73-85, where the tenor melody intensifies the recitativelike quality I observed in Manchon-Theis's playing of this passage. In contrast, Webster adopts a strategy different from any of the other pianists I have discussed so far: he consistently arches all the groups from E46 to E90-not only the isochronous but also the non-isochronous ones (though in the latter case the effect is less perceptually salient). He also parses the f# 1 /f♮ 2 dyad at E57 differently from the other pianists. I referred to the dyads at E52 and 57, E67 and 72, and E85 and 90 as the terminus points of their respective groups, as indicated by Webern's 'tempo' markings (and confirmed by his brackets in the annotated score): Stadlen, Manchon-Theis, and Aitken all treat them this way, reducing speed and volume as they approach them. By contrast, Webster creates a perceptible break between E56 and E57, leaving the latter oddly hanging. Perhaps it was because Stadlen was doing the same that Webern parsed this passage in his score.
Equally variable is the treatment of the climactic group of the B section, from E91 to E121. Aitken builds up speed at the beginning of the passage, like everyone else, but slows perceptibly to E99, almost as if he had seen Webern's annotation at this point. He picks up speed and slows again at E105-6 (this time contrary to Webern's annotation). Then he attacks E107 and maintains a tempo of close to 40-the highest sustained tempo of the performance-up to E116, at which point he slows down drastically to the end of the group: this and the following group (E122-32) in effect become a transition to the tempo of the A2 section. But whereas in Aitken's performance, like Manchon-Theis's, there are at least patches of sustained tempo, Webster's tempo during this passage is continuously mobile. The same might be said of Stadlen, who rendered this passage as a series of discrete gestures, and the effect in each case is the same splashy rhetoric. Yet the mobility of Webster's tempo is different. There is a sense of rapid motion underlying it-of the climactic block I missed in Stadlen's performance-but it is articulated around a series of events that require particular attention and are given extra time: the passage from E101-03 that Webern linked to a mysterious drum, and the successive climaxes at E107 and 112.
It makes sense to compare Webster's recording with that of the other three earliestborn pianists to record Op. 27 not, then, because they share particular interpretations of the score but because of similarities in their handling of time, in their essentially dramatic conception of the music's unfolding, and in their lyrical and melody-driven approach to musical expression-in short, because of similarities in style. But there are two respects in which Webster's recording combines aspects of performance practices from pre-and post-war modernism. One is the systematic use of phrase arching to which I referred, a feature that became widespread in the performance of commonpractice repertory: while tempo or dynamic arching is found in recordings from before the Second World War, only after 1945 did the coordination of these parameters with one another, and with composed phrasing, become the systematic phenomenon that researchers such as Neil Todd have documented in present-day performance. Manchon-Theis, Aitken, and Webster retain the very high level of tempo and dynamic inflection found in Stadlen's performance, but in every successive recording the tempo and dynamic arching becomes increasingly rationalised, with the arches coinciding with the gestural groups.
The second and more crucial respect concerns absolute tempo. By comparison with Stadlen's and Manchon-Theis's recordings, Aitken's features the fastest average tempo for the A1 section (31 as against the others' 28), but the slowest for the B section (22 as against Stadlen's 29 and Manchon-Theis's 27); he speeds up only slightly for A2 (24, close to the others). Webster retains the distinction between A1 on the one hand and B and A2 on the other, but he plays everything much faster (35,31,30). This affects the expressive character-and the physical choreography-of the performance, and has other interpretive consequences too: for example, the lack of interpretive definition over the dyad at E57 might be more problematic in a slower performance, where coherence would be more of an issue, while the strong effect of tearing off that Webster creates at E121 is more easily made effective at his tempo than at Aitken's (and Aitken handles this point quite differently). In short, perhaps unsurprisingly considering its late date, Webster's recording represents something of a stylistic hybrid.
But you would never guess that from the record itself, which like Manchon-Theis's couples Op. 27 with works by Schoenberg and Berg, and comes with an insert containing analytical notes by the American critic and composer Eric Salzman. Here, even more than in Spitzmueler's sleeve notes for Manchon-Theis, we have the Op. 27 of post-war modernism: Salzman speaks of the work's 'symmetrical quality-the musical objects have almost the character of tangible, geometrical objects'. Critical language lends itself to such simplifying characterisations: that is how it can be so responsive to fashion. By contrast, the grounding of performance in tacit knowledge and overlearned physical schemata makes it more resistant to change, and we shall see in the next section that the much younger pianists who came to maturity during the immediate post-war years struggled to express in their playing the qualities that tripped so easily off the critics' tongues.

Zero Hour
Performance is evanescent, and as a result the narratives that cluster around it easily become entrenched. One such narrative turns on the impermeable division that Stadlen sought to draw between those musicians-such as himself-who had first-hand experience of the Schoenberg circle, and the post-war avant-garde whose enthusiasm for Webern he saw as matched only by their failure to understand him. It was as if, on the strength of his coaching sessions, he was licensed to speak with the delegated authority of the composer. This becomes particularly clear in an article originally published in 1972, and pointedly called 'The pointillist misunderstanding', in which Stadlen (2004) dismissed interpretations of Op. 27 by a range of analysts from Friedhelm Döhl to Peter Westegaard by the simple strategem of demonstrating that they did not conform with the annotations in his score-which on its publication seven years later turned him into 'the composer's representative on our music stands' (Mathew 2011: 66). In essence Stadlen accused post-war performers such as Leonard Stein (of whom more shortly) of ignorance: in the letter from which I have already quoted, Anthony Stadlen-quoting from Peter-writes that 'Stein "could not be expected to divine" the nuances which could only be learned from "direct, detailed tradition"'.
But even when one allows for the cultural fault-line of the war, it is hard to imagine that pre-war performance practices had been forgotten so completely as that would imply. After all, Stein made his recording of Op. 27 in 1954, just seventeen years after Stadlen premiered it. 12 And as I said, performers of the Schoenberg circle-including Stadlenattended the early Ferienkurse, forming part of a representation of pre-war modernist music that, as Fox (2007a: 11) observes, was more extensive and significant than is commonly recognised. Equally significant was the presence of Theodor W. Adorno, as a result of which critical theory became a lasting component of avant-garde musical thinking in Europe. On the other hand there was strong opposition to what Fox (2007b: 2) calls 'Adorno's commitment to the continuity of development in the European art music tradition': this is illustrated, for example, by the analytical lectures he gave at the 1955 Ferienkurs, which emphasised the roots of Schoenberg's twelve-tone thinking in late Beethoven (Borio 2007: 64-5). This commitment to continuity was shared by Schoenberg circle performers such as Kolisch and Stadlen-and by Webern himself, whose The Path to the New Music 13 sets out a teleological progression through tonality to atonality and serialism. Quick (2010: 103) refers to 'an ideological and stylistic clash between pre-war and postwar strands of modernism', and that is the context within which Webern was received at Darmstadt. The point is neatly made by juxtaposing the comparison between Schoenberg and Webern that Adorno drew in his Philosophy of Modern Music, first published (in German) in 1949, with another that appeared six years later in die Reihe, widely if inaccurately seen as the house journal of the Darmstadt avant-garde. For Adorno (2006: 85-6), the originator of serialism subordinated compositional technique to artistic expression: 'Schoenberg assaults the row. He composes twelve-tone music as if twelve-tone technique did not exist'. By contrast, Adorno continues, 'Webern brings twelve-tone technique into reality and no longer composes….
[T]he material is itself vested with the capacity to posit musical meaning'. And as usual, Op. 27 (along with the String Quartet Op. 28) is in the front line: 'They feature nothing more than monotonously symmetrical presentations of serial marvels that, in pieces such as the first movement of the Piano Variations, come close to a parody of a Brahms intermezzo' (Adorno 2006: 86-7).
Writing in the second (1955) issue of die Reihe, which was dedicated to Webern, Herbert Eimert -six years older than Adorno but arguably the leading ideologue of postwar serialism-says the same, but draws the opposite conclusion: 'Schönberg thematises his series, Webern de-thematises it; Schoenberg uses it for that "composition" from which Webern frees it…. The thinner, the leaner Webern's music becomes, through its compulsion to extreme refinement, the more structure is manifest in it' (Eimert 1958: 33, 35). Henri Pousseur (1958: 52, 60) makes a similar point in his contribution to the same issue, but the argument had been most notoriously made three years earlier in Pierre Boulez's article 'Schoenberg is dead'-and in his own contribution to the issue Boulez summarised it: 'Whereas Schönberg and Berg ally themselves to the decadence of the great German romantic tradition, [Webern] reacts violently against all inherited rhetoric, in order to rehabilitate the powers of sound' (Boulez 1958: 40). In short, what Adorno criticised in Webern was precisely what attracted the Darmstadt avant-garde. When he spoke of twelve-tone technique being 'infinite in its ahistorical stasis', Adorno (2006: 53) intended it as criticism, but for the Darmstadt avant-garde that was exactly the point. The brief period of Webern obsession memorialised in the pages of die Reihe was based on seeing his work as a new start. Returning to Stadlen and Op. 27, it was not then a matter of simple ignorance but rather a highly selective, scriptist, so to speak fundamentalist appropriation of Webern's music (recall Fox's comment about Webern on the page being the touchstone for post-war composers). The aim was not to continue but to repudiate a history perceived as hopelessly compromised.
That is what lies behind Eimert's (1958: 32) strikingly positivist claim that 'Webern uses only the proved and the provable; everything else is rigorously excluded, and rightly, since it has no firm foundation'. It lies behind Armin Klammer's (1958: 87) claim-in yet another article from the Webern issue of die Reihe, on the third movement of Op. 27that 'traditional relationships are eliminated when tonality is replaced as the basis of the composition by the serial principle. We are thus compelled to seek fundamentally new possibilities which must arise from the series itself'. It resonates with such diverse cultural and intellectual movements of the time as existentialism, with its emphasis on individual freedom and choice, and structuralism, with its insistence that human thought and action are not determined by nature but culturally constructed. And when Ernst Krenek (1966: 10) writes that, in Op. 27, 'the composer … withdraws more and 13 A series of lectures delivered in 1932-3 and published in 1960 (English translation Webern 1963). more into the detached, cool, miraculous, and exciting world of the musical patterns, where the abstract spirit of music seems to have its own enigmatic life, sufficient unto itself', it is not hard to detect an echo of the controversies over the 'inner emigrants'those who remained in Germany but withdrew from public politics into a private, nonpolitical world-that erupted as the programme of denazification began. 14 It is this cool Webern, the Webern of the late instrumental music, that dominated anglophone musicology and theory until overturned in the 1990s by the combined efforts of Kathryn Bailey, Anne Shreffler and Julian Johnson. In this context there is considerable irony in Eimert's (1958: 34) claim-in the same issue of die Reihe-that the Webern constructed by the post-war avant-garde was 'the "real" Webern'. It is not so hard to understand Stadlen's exasperation.
However this black-and-white interpretation of traditionalists versus radicals, or prewar versus post-war modernists, has been both contextualised and challenged by Morag Grant, and masks a more complex picture. As Grant explains, the idea of Zero Hour (Stunde Null)-a rebuilding of culture from the ground up, analogous to what happened in heavily bombed cities such as Darmstadt (Fox 2007a: 9)-is less credible than it once seemed. Even if those who attended the Ferienkurse believed they were creating something quite new, continuities between the pre-war and post-war periods-in which the 'inner emigrants' played a pivotal role-make 1930-60, as Grant (2001: 17) puts it, 'a much more significant periodisation'. Referring specifically to Webern, Gianmario Borio (2005: 87) agrees: 'behind the diffuse notion of a radical rupture there emerges a fine line of historical continuity'. A symptom of this is that, as Bailey (1995: 647-8) pointed out, the representation of Webern in die Reihe by no means conformed to the Zero Hour image of Webern. In the 1990s Shreffler complained that Webern's solo music had been ignored and sought to rehabilitate 'Webern as champion of the lyric genre' (Shreffler 1994: 4), but the Webern issue from forty years earlier began with a note by Hildegard Jone, whose poems furnished the texts for so many of Webern's songs, while Eimert's article-perhaps the most comprehensive in its reconstruction of the composer as post-war modernist-emphasised not only Webern's songs but also the 'scanty, undecorated, unrhetorical lyricism' of his music in general (Eimert 1958: 31). Then again, by 1955, when the Webern issue of die Reihe appeared, the critique of Schoenberg advanced by Boulez, Pousseur, and Eimert was beginning to be applied to Webern himself: Klammer (1958: 92) writes that even in Webern's work 'serial thought is contradicted by the remnants of thematic thinking'. In short, there were shifting and competing currents within the ostensibly monolithic modernism of Darmstadt and die Reihe. And it is this more complex picture that emerges from the recordings of Op. 27.
As I mentioned, the first commercially issued recording of the Piano Variations dates from 1951, and it is frequently cited as epitomising what Nicholas Mathew (2011: 58), quoting Taruskin, describes as the 'text-centred, hence literalistic' performance ideologies within which Op. 27 became entangled. It was made by the French pianist and conductor Jacques-Louis Monod, and appeared on an all-Webern LP released by the New York-based Dial label. Monod was aged 24 at the time of the recording, and attended a number of the early Ferienkurse (according to Wikipedia he was at the 1948 one, in which case he could have heard Stadlen's performance of Op. 27). 15 But although 14 As detailed in e.g. Garland and Garland 1997: 415, this term was coined in 1945 by Frank Thiess by way of response to Thomas Mann's writings on German guilt for the war: the 'inner emigrants' who remained in Germany, Thiess argued, had shown more courage than those who had left and now claimed to speak for the German people. This sparked off an enduring controversy that overshadowed critical responses to the cultural productions of the war years. 15 Entry on Monod, accessed 12 January 2014; this claim has since been removed. he was in this way very much part of the Zero Hour generation, there were strong connections with the past. As documented by David Hoek (2013), the recording project was masterminded by the Polish-French composer, conductor, and writer René Leibowitz, who had studied with Webern and whose book Schoenberg and his School, published in French in 1947 and in English translation two years later, was the first comprehensive introduction to atonality and serialism. Leibowitz also played a significant role in transmitting the performance practices of the Second Viennese School to the post-war world, and the performances he directed on the same Dial disc-of the Symphony Op. 21, the Quartet Op. 22, and the Concerto Op. 24-illustrate his concern to extract the maximum expressive charge from these spare scores, if only in the very nonpointillist shaping of individual notes. The more traditional textures of Schoenberg's First Chamber Symphony, which he recorded for radio in 1962, 16 allow a better insight into his performance values, which-despite the date-are much of a piece with Stadlen's recording of Op. 27.
Leibowitz's sleeve notes draw on his book and conform to the Zero Hour conception of Webern in a way his performance practices do not. Taking his cue from Leibowitz's claim that the Piano Variations dispense with orchestral colour and so 'crown Webern's evolution in the direction of musical "nudity"', Mathew (2011: 59) summarises that conception as 'Naked music; colorless music; pure music'. And that is what Mathew hears in Monod's recording. He discusses it in the context of a persuasive critique of Stadlen's assumptions concerning the authority of the text. Stadlen is just as much of a literalist as Monod, Mathew says; it is just that for Stadlen the published score of Op. 27 is defective, an inadequate embodiment of Webern's intentions. By contrast, Stadlen treats the annotated edition he published in 1979 as the comprehensive prescription of what is to be done. As Mathew explains, this way of thinking about scores leaves no room for the historically situated knowledge that enables performers to make sense of the notation by supplying what is implied but not stated. Stadlen speaks in the preface to his edition of the 'perplexing sparseness of expression marks in [Webern's] late instrumental works' (p. V)-which this is hardly true of Op. 27, the original score of which is as full of dynamic markings and instructions to slow down or speed up as late Beethoven or Liszt: as Fox (2007a: 11) says, 'Webern marked all those rallentandi … because he came from a musical tradition in which rhythmic suppleness was a hallmark of expressive sophistication'. However the underlying point is that scores do not attempt to spell out everything that is to be done, but rather reference a performance tradition in which their users are assumed to be enculturated. Stadlen's perplexity, then, is the result of a literalist mindset that is not just unhistorical but simply unrealistic.
And it is this mindset that Mathew sees as leading to performances like Monod's, in which the text is literally translated into sound. Monod decodes Webern's score, Mathew (2011: 60) says, in the manner indicated by the General Foreword to Karlheinz Stockhausen's Klavierstück V: 'depress the key for the duration indicated'. And the result is 'unmediated, un-interpreted Webern', a 'style of self-negating anti-pianism'. The problem with this is that Monod's recording is by no means literal, or better put, there is nothing literal about its literalism. An illustration of this in miniature is the opening bars, where Mathew (2011: 59) observes that Monod 'observes the rests with the kind of punctiliousness that can only lead to abrupt musical punctuation'. There is more to this than Monod's brisk tempo (his average tempo for the A1 section is 39, 40% faster than Stadlen's and almost exactly what Webern indicated). But it is not quite what Mathew describes. As the tempo graph shows, rather than playing the rests literally, Monod consistently shortens them: each group begins just before you expect it to. As I said, tempo is a salient parameter in this isochronous music, and if you listen analytically you can hear what is happening-but in normal listening the effect is precisely the abruptness that Mathew refers to. Rather than being punctiliously metronomic, then, Monod is deviating from literal execution in order to create the abrupt and inflexible quality that Mathew calls literalism.
Mathew also complains of the 'meticulous metrical rigidity' of Monod's playing: 'there is little sense of the give and take that might follow from local articulation or a sense of phrasing'. That may be fair comment up to and including the beginning of bar 15 (E37): it is as if Monod goes out of his way not to give the end of the phrase the quality of expressive inwardness that Stadlen drew from the music at this point. But from E38 there is a very perceptible softening of the profile of Monod's playing as he homes in on the end of the section, slowing down in a manner that is conventionally 'musical', though of course understated by comparison with Stadlen. And in the A2 section the change of quality is that much more obvious. Though-contrary to the score but like almost everyone else-Monod takes A2 at a slower tempo (average 35) than A1, he again plays its first part more or less metronomically (E133-48). But from E149 to the end he plays with increased flexibility and range in both tempo and dynamics: indeed there is clear, if again understated, tempo and dynamic arching in the groups E149-57, E158-69, and E170-74, while the final notes (E175-77) are subsumed within another conventional rallentando. And when he wants to give E169 the same inexpressive quality that he gave E37, he does so through a staccato articulation that is nowhere to be seen in the score. While he is not, obviously, in the business of milking the emotion out of expressive moments in the manner of Stadlen, he plays with greater or less flexibility according to how he wishes to shape different parts of Webern's score. In Stravinsky's terminology, he is not executing but interpreting it.
As usual it is the B section that best reveals Monod's interpretive strategies. Actually they are rather straightfoward and not so different from those we have seen in Stadlen, Manchon-Theis, Aitken, and Webster. Monod has clearly taken careful note of Webern's 'rit. … tempo' markings and incorporated them within his interpretation: Quick's (2010: 115) observation about Webern performance in general-that 'The Darmstadt avantgarde could not have been unaware of the notated tempo changes' but 'chose to ignore them for aesthetic reasons'-does not apply in this instance. Monod clearly conveys the parsing implied by Webern's 'rit' and 'tempo' markings, not through literal execution but by incorporating them within tempo and dynamic arches that correspond to each of the groups, slowing down to the cadential dyads. In the terms of Mathew's critique of Stadlen, Monod interprets Webern's markings as referencing established pianistic style and gesture: he does not share the literalist mindset that Mathew identified in Stadlen. As for the sustained group from E91, Monod creates some light and shade through careful attention to Webern's p markings at E100 and 113, and by keeping the speed up almost until E121 and not prolonging the rest, he not only conveys the torn-off effect we have already seen, but also creates the impression that E91-132 is a single block with a slight hesitation near the end. Beyond that there is not much to say, because at this tempo (36 averaged over the B section) there is not much to do, other than play the notes.
I do not deny that, compared to Stadlen's and many more recent recordings, Monod's creates a certain effect of literalness, to which the business-like quality of his tempos, the abruptness of his silences, and the understating of potentially expressive moments like E37 and E169 all contribute. My point is that it is just that, an effect of literalness, artfully constructed to play on expectations engendered by received performance practices. Pursuing a comparison between the 'consistent, and consistently hard, timbre' of Monod's playing and brutalist architecture, Mathew (2011: 60)-who has more faith in the tonal fidelity of mid-century recordings than I do 17 -cites Rose Subotnik's suggestion that 'Stravinsky's professed formalism … evokes "not the conceptual […] but the stylistic attributes of objectivity"'. Much the same might be said of Monod's literalness. And it follows from this that Monod's playing sits more squarely within the performance practices of his time than either Mathew's description or the recording's general reputation might suggest. At some points Monod may be playing against tradition, but most of the time he conforms to it, though in an understated way. It is rather like what happens when pianists are asked to play deadpan: they play as usual, only less so (London 2004: 176). As the tempo and dynamic graphs how, Monod shapes gestures, articulates groups, and slows down at the end of sections in an quite conventional manner, only less so. And his understatement is like his avoidance of the hand breaking that by the 1950s was becoming a signifier of old-school performance: a matter of aesthetic fashion, in other words-Subotnik's words-of style rather than conception. In short the traditional codes of expressive performance persist in Monod's playing. It may be cool, but it is not a new beginning.

Darmstadt pianism
An informative comparison can be made with a recording which-like Monod's-is generally seen as epitomising Darmstadt literalism, and which I have already mentioned: Leonard Stein's, recorded in 1954 (the same year as Manchon-Theis's and probably one by Glenn Gould, of which more shortly). There are some obvious similarities to Monod's recording, perhaps because Stein had heard it, and Stadlen (2004: 37) described Stein's playing of the A sections rather as Mathew describes Monod's: 'neutral and therefore distorting'. (In his radio review of the box set, Stadlen said Stein played Op. 27 in a way that 'Webern would have regarded as a mere spelling out of the notes'.) At an average of dotted eighth = 41 across the A1 section, Stein's tempo is even more business-like than Monod's-and, like Monod's, far closer to what Webern indicated than Stadlen's. Stein plays the opening bars without obvious inflection, and plays even more metronomically in the first half of the A2 section, where the still fast average tempo-38, only a little slower than A1-robs the performance of the serene, emotionally washed out effect of Manchon-Theis's relatively uninflected playing at the same point. Nevertheless Stein introduces a degree of tempo and dynamic arching from E17 and E149, like Monod creating the sense that each section thaws a little as it continues; more puzzlingly, he also abbreviates E19, creating the kind of apparently arbitrary nuance familiar from Stadlen's recording, and does the same at the corresponding point of A2 (E151). There are conventional rallentandi at the end of A1 and A2.
Stein takes the B section substantially slower than Monod (31 as against 36), and this means he faces more pressing problems of coherence. Like Monod he arches the groups of thirty-second notes-or at least he plays them faster than the surrounding notesand, again like Monod, he gives careful attention to Webern's 'rit. … tempo' markings. He too decelerates as he approaches the terminal dyads of each of the groups, though unlike Monod, to whose handling of these dyads I shall return, Stein gives them all equal weight: this parses the groups clearly enough, but offers little interpretation of the relationships between them. Once more like Monod, he creates the sense that E91-132 is a single group with an interruption at E121. From E91 to 121 the only significant inflection is the prolongation of E99, coinciding as it happens with Webern's annotation of Stadlen's score; this is one case where the apparent mobility indicated by the tempo graph does not translate into listening experience-probably through lack of coordination with other parameters such as dynamics and articulation-and at this tempo the thirty-seconds acquire a chugging quality they do not have in Monod's recording. Finally, like Monod, Stein embraces E122-32 within a sustained rallentando, with a caesura before the A2 section that is much more modest than Stadlen's gargantuan silence at this point, but still unmistakably rhetorical. Literalism, then, is relative.
But it is less through its inherent properties than through its background that Stein's recording adds to the emerging picture of Op. 27 in early performance. As we have already seen, Stadlen said that Stein 'could not be expected to divine' the nuances that could only be learned from 'direct, detailed tradition'. But while Stein did not study with Webern, he was studying with Webern's teacher at the very time when Op. 27 was being composed: from 1935 he was taught by Schoenberg at the University of Southern California and subsequently at UCLA, and from 1939 until Schoenberg's death acted as his assistant. And like Stadlen, Stein traded on his first-hand experience of the Second Viennese School. He wrote in the proceedings of a 1995 conference that 'My own insights into Schoenberg's music came from direct contact with the master himself' (Stein 2002: 62), and went out of his way to mention occasions on which he had played for Schoenberg. More than that, he developed an argument reminiscent of Stadlen's claims concerning the inadequacy of scores to communicate composers' intentions in the absence of a complementary oral tradition. That is the context of the remark about his own insights, but he set out the argument more fully in an earlier article: 'In twentieth-century composition', Stein (1963: 63) wrote, 'no ... norm of common notational practice seems to exist that might lead to an immediate recognition of the composer's intention'. The value of such arguments for purposes of career development is self-evident.
In later life Stein returned to USC, completing a DMA in 1965 (at the age of 49) with a thesis entitled 'The performance of twelve-tone and serial music for the piano', and this provides a further perspective from which to approach his recording (though we must remember that it was by that time eleven years old). In a chapter on the Piano Variations, Stein (1965: 82) insists that through analysis it is possible for the performer to 'arrive at a convincing presentation of the contents and shape of the composition, rather than to allow the notes to fall where they may indiscriminately'. But he provides little guidance on how exactly this is to be done. For example he emphasises the 'structural significance' of textural density in Webern's music, explains that simultaneities represent 'verticalizations of row-cells', and concludes that they demand 'special attention'-but he does not suggest what the performer might actually do about these things, and neither does his recording. Despite the title of his thesis, Stein's emphasis is overwhelmingly on analysis of the printed notes, and he is thinking more like a theorist than a performer when-following the Boulez party line-he writes that 'Webern eliminated the expressive rhetoric of Romantic music in his search for an essential underlying structure' (Stein 1965: 72): you only have to compare Stadlen's and Stein's recordings of Op. 27 to appreciate how far such rhetoric is the work of performers rather than inherent in the notation. Perhaps unsurprisingly then, there are times when there seems to be a disconnect between what Stein says and what he does, between his explicit and his tacit knowledge. For example, when in his thesis he parses the first part of the B section, one of his groups is a 'phrase b consisting of a four-note syncopated mirror'. 18 This is E53-6, which as we saw Webern parsed-implicitly in his original score, and explicitly in Stadlen's-as the first four notes of a group that terminates with E57. Stein's parsing reflects a mathematical conception that is expressed through visual symmetry, whereas Webern's is based on a more traditional sense of motion towards a cadential point. And the point is that, in his recording, Stein plays it Webern's way. I imagine he did the same when he played Op. 27 in his 1963 DMA recital (Stein 1965: 195).
What is most revealing about Stein's thesis, however, is the evidence it provides thatmore than a decade after his recording-he was still thinking of Op. 27 in terms of prewar analytical categories. He divides movements into sections, periods, phrases and groups, which he relates to the forms and processes of classical music, in much the same way as Leibowitz or-more particularly-the composer, theorist, and Webern pupil Leopold Spinner (whose very traditional analytical contribution to the Webern issue of die Reihe rubs shoulders awkwardly with the contributions from Boulez, Stockhausen, and Pousseur). Like Spinner, Stein calls A2 the recapitulation, and though he simply refers to the central section of the movement as the B section (Spinner calls it the elaboration), Stein describes the build-up to the climax towards the end of that section as 'a familiar developmental device' (Stein 1965: 85, 83, 85;Spinner 1960: 10). Here then is direct evidence of what I deduced from Monod's playing: the extent to which performances within what Quick called 'the avant-garde "Darmstadt" practice' were based on traditional thinking, with their literalism being a matter of style rather than concept. As a pianist Stein retained pre-war habits of thought even as he took on board the products of post-war modernism: he attended the Ferienkurse, 19 and in his DMA recitals played music by the aforementioned Boulez, Stockhausen and Pousseur. In short, like Monod, Stein illustrates the continuity between pre-and post-war culture to which Grant referred. And in this way, the most ostensibly literalistic of all recordings of Op 27 represents a more complex and perhaps conflicted historical, aesthetic, and ideological position than could ever be deduced just from listening to it.
There are two later recordings that might equally be seen as exemplary of what Mathew calls 'Darmstadt pianism': one by the American, Juilliard-trained Paul Jacobs, recorded live at the 1956 Cité Radieuse Festival in Marseille, and the other a studio recording by Yvonne Loriod, who trained at the Paris Conservatoire. Both pianists were more centrally involved in the European new music establishment than either Monod or Stein-Jacobs through a close association with Boulez (though he also worked with Leibowitz), and Loriod through her husband from 1961, Olivier Messiaen-and both performed regularly at the Ferienkurse. Jacobs's recording might be seen as the first clear sign of a crossing over between the hitherto distinct styles initiated in the recorded repertory by Stadlen and Monod. He plays the outer sections with the same kind of careful understatement as Monod, the care extending to the interpretation of Webern's notated slurs (most obviously at E26-37); again like Monod's, his playing thaws as both A1 and A2 proceed, with increasingly well formed tempo and dynamic phrase arches. The element of crossing over comes in the tempos of the A sections, which at respective averages of 29 and 25 are close to Stadlen's (28 and 24). At that tempo what would be relatively imperceptible nuances when played faster generate the effect of cool expressivity-the expressivity of the pre-war style coupled with the coolness of the post-war avant-garde-at which many later performers aimed.
By contrast, Jacobs's average tempo for the B section (34, the same as it happens as Loriod's) is only slightly slower than Monod's (36), and distinctly faster than Stein's: as a result, while eight out of the 30 recordings made up to 1977 take B faster on average than A1, the differential is more marked in Jacobs's case than any other (apart from an altogether eccentric recording by Jean-Rodolphe Kars, of which more later). Jacobs's performance of the B section resembles Monod's-which I imagine he listened to-not only in its tempo and overall shaping but also in a tell-tale detail, the handling of the terminal dyads of the paired groups, at E52 and 57, E67 and 72, and E85 and 90. We have seen that, in his coaching sessions with Stadlen, Webern clarified the parsingwith the dyads terminating each group-and that Stein also played the music this way, even though the diagram in his thesis parsed it differently. But parsing is not all that is at issue. Stein expresses the parsing by prolonging the terminal dyads: as I said, his playing is clear but the groups remain disconnected. Monod, by contrast, has a more subtle approach. He consistently prolongs the note that precedes each dyad, but shortens the dyad itself: this makes the parsing equally clear but means that each group presses on into the next. Jacobs's strategy is identical to Monod's, and the effect is that much more articulate at his slightly slower tempo.
As for Loriod's recording, it is in some ways a compendium of Darmstadt pianism. Her no-nonsense tempos are close to Monod's-a little faster in A1, a little slower in A2while once again there are front-loaded tempo and dynamic arches in the second half of the A sections. And on Sonic Visualiser the tempo and dynamic profiles of the B section exhibit the familiar arching for each group from E46 to 90, the same sustained tempo from E91 to 121 relieved by the piano passages from E100-104 and E113-17, the same slight tearing off at E121, and the same decelerando and diminuendo to the section break after E132. But all this is put into the shade by the clipped articulation, lack of pedal, and aggressively dry recording acoustic that perhaps led Borio (2005: 95) to describe her playing as 'virtuoso', and Peter Hill as 'completely unemotional and even rebarbative, you know, really fearsome and forbidding' (Siepmann 2003: 24). At the beginning of both A1 and A2 Loriod consistently plays the third and fourth notes of each group staccato, breaking up the residual sense of more or less legato melody that is present in almost all the other recordings I have discussed: in doing this she is of course disregarding Webern's slurs. In the B section, the light, dry touch of her performance of the isochronous passages perhaps draws on the jeu perlé style associated with the Paris Conservatoire-she studied with one of its leading exponents, Isidor Philipp-while she plays the non-isochronous groups (E53-7, E68-72, E86-90) as a series of stabs. The lack of flow is enhanced by the fact that-like Stein rather than Monod or Jacobs-she prolongs the terminal dyads rather than the notes preceding them. In short, Loriod intensifies the effect of literalism that Mathew complained of in Monod, and like Monod achieves it through deviating from literal execution. The result is the kind of pointillism that Stadlen condemned in the playing of the younger generation-even as in terms of tempo and dynamics Loriod adopts the same, coherence-building strategies as other representatives of her generation.
For Hill and others, the dry, apparently disengaged pointillism of Loriod's recording makes it the paradigm of Darmstadt pianism. The odd thing about this is the date, given on the record label as May 1961. (The sleeve is as ascetic as Loriod's playing, with no image, the same sans-serif lettering that Universal Editions used at this time, and a note by the Bulgarian-French composer André Boucourechliev that stresses the rigorous and athematic nature of Webern's serialism.) This, after all, is long after the moment of what Bailey (1995: 648) called 'Darmstadt madness'-which she dates to 1951-2-while for composers pointillism was passé after around 1955. And this chronological disjunction prompts the question of what it really means to speak of 'Darmstadt pianism' in relation to recordings of Op. 27. If the term is not clearly focussed in terms of time, it is hardly more focussed in terms of place. Darmstadt is in central Germany but, as the international nature of the Ferienkurse might lead one to expect, recordings of Op. 27 were not centred there or even on the German-speaking countries more generally. The one recording actually made in Darmstadt was Stadlen's-of all recordings the most conspicuously opposed to Darmstadt pianism-while of the three recordings from the 1950s that are generally associated with it (Monod, Stein, Jacobs), one was made by a French pianist in America, one by an American pianist in America, and one by an American pianist in France. A fourth, made for radio in Canada and probably dating from 1954, 20 might have been considered equally emblematic of Darmstadt pianism had it been released at the time, and had there been any documented link between the pianist and Darmstadt: Kevin Bazzana (2003: 100) remarks that Glenn Gould was isolated from the contemporary music scene even in Canada. And though Gould could always have heard Monod's recording, there are no particularly compelling links (for example, his treatment of the tell-tale dyads in the B section is not like Monod's). It probably make more sense to look for the sources of Gould's particular take on literalism-which is no more literal than Monod's-in the performance practices he had developed in relation to the sixteenth-and eighteenth-century music that was already in his repertory before he started playing that of the Second Viennese School.
And this pattern of geographical dispersal continued. During the period up to 1977 there is a total of six recordings by performers from France, six from Canada (all Gould), five from the US, three from Italy, two from Japan, and one each from Austria (that's Stadlen), Poland, Russia, Turkey, Spain, and the UK (McCabe). It is remarkable that Germany comes in with just two: Franzpeter Goebels (1964), whose playing carries echoes of Stadlen's, and Christian Zacharias (1973). If then the idea of Darmstadt pianism is to stand for anything, it must be stylistic features or the aesthetic ideologies that informed them. Some common features between the recordings of Monod, Stein, Jacob, and Loriod can be identified, most of which are also found in Gould's. Tempos tend to be fast (though that is not true of Stein's performance of the B section, or Jacobs's of the A sections-while the fastest come from Gould, especially the live recording he made in 1957 at the Moscow Conservatory). There is a general quality of understatement, though even at its most extreme it does not preclude traditional rhetorical features such as caesurae or tempo modification, particularly in the B section: the effect ranges from the cool expressivity of Jacobs's performance to the alienating dryness of Loriod's. But the most distinctive feature is a negative one: all these pianists distanced themselves from the overtly expressive playing inherited from the pre-war years, as represented by Stadlen, Manchon-Theis (whose recording of Op. 27 dates from the same year as Stein's), or Aitken (the same year as Loriod's).
It may then be more productive to identify key features of the Darmstadt aesthetic and map them onto the recordings. In her book Serial Music, Serial Aesthetics, Grant locates a fundamental characteristic of Darmstadt serialism in the swerve away from thematicism that is exemplified by Klammer (who as we saw detected the vestiges of thematic thinking in Webern) and Boucourechliev (who insisted that his music was free from any such taint). An early advocate was Leibowitz (1949: 241), who argued that in Op. 27 'everything is theme', which is to say that nothing is more thematic than anything elsefrom which it follows in an apparent paradox that 'Webern's piano variations are … a basic contribution to the "athematic" method of composition'. This idea was taken up by Eimert, Boulez, and Pousseur, so becoming part of Darmstadt orthodoxy. And on this basis Grant (2001: 163) makes a key distinction between traditional thematic hearing, which is based on a kind of causality that results in specific expectations, and a 'serial hearing' oriented to processes of 'continual change and continual non-directed expectation'. She also sees as essential characteristics of serial aesthetics a 'conscious concentration on the internal structure and character of individual events' (she calls this 'microaesthetics'), and the use of silence for 'indicating the boundaries between groups and textures, … throwing elements into focus and then, contradictorily, leaving them hanging in space, as if to question their significance' (Grant 2001: 161, 164).
Loriod comes closest to this in the way she breaks up melodic continuity in the A sections, and disrupts musical causality in the non-isochronous groups of the B section (what I described as a series of stabs): it would not be unreasonable to say that through her use of silence she leaves these notes hanging in space, as if to question their significance. It is only Loriod's recording to which this applies, however, and as I have said, such effects-created primarily by articulation-run alongside a quite traditional concern for parsing, pacing, the balancing of elements and the creation of direction. And to these ends Loriod incorporates strategies and rhetorical features comparable to the other 'Darmstadt' recordings I have analysed. But perhaps the best demonstration of these concerns and strategies comes from the series of live recordings that Gould made from the 1950s to the 1970s. In the article I have already mentioned, I trace the stages through which his interpretation developed by focussing on the tell-tale dyads in the B section. In 1954 Gould played them like Stein, parsing the groups but offering no interpretation of the relationships between them. They sound much the same in his 1964 film, but here there is an additional element: he can be seen to create a sense of overall direction though increasingly energetic physical gestures.
In his audio recording of 1970, by contrast, Gould made each pair of dyads shorter than the previous one, maintaining the same overall tempo but in this way creating an audible analogue of the motion towards the extended, climactic group at E91 that he had previously created through visual means. 21 This illustrates precisely the musical causality that Grant associates with thematic rather than serial hearing, the sense of continuous temporal entailment that Heinrich Schenker (1987: I, 291) vividly evoked when he wrote that 'harmonies appear to be linked more intimately and with seemingly greater necessity the more drastically and obtrusively a tone of one harmony hooks into the flesh of the following one'. And it is by transforming 'before' and 'after' into 'because and 'therefore', and so creating a sense of logical sequence independent of the time of performance, that such causality gives rise to musical narration. In this way the narrative paradigms inherited from nineteenth-century music may be more deeply buried in the playing of pianists like Monod, Loriod, and Gould than in Stadlen or Manchon-Theis, but they are still there. The conclusion is obvious. In his article on composers at the Ferienkurse, Fox (2007a: 13-14) questions 'the extent to which there ever was a "Darmstadt School"', and concludes that it was 'really an amalgamation of a number of quite different projects'. Much the same might be said of Darmstadt pianism.
Or maybe it would be better to say that, when it came to pianism, there was no project. In the Webern issue of die Reihe, Stockhausen (1958: 38) spoke of the value of analysis to performers, so advancing the idea that there is-or should be-a commonality of purpose across the domains of performance, composition, and discourse. But this claim is hardly substantiated by Klammer's analysis, later in the same issue, of the final movement from Op. 27. It is not simply that, in his quest for 'fundamentally new possibilities which must arise from the series itself', Klammer begins by dismissing such performable dimensions of the music as thematicism as irrelevant, and instead-in Grant's (2001: 119) words-'dissects every possible angle of the piece into a series of neat but mostly meaningless numbers'. Whereas 'in traditional music determining applies mainly in a forward direction', Klammer (1958: 92) says (he gives the example of antecedent and consequent), 'with Webern we find it working retrospectively as well': the structure depends on different groups having different intervallic attributes, but their significance can be known only when all have been heard. Musical logic, in other words, reverses the flow of time. And Borio makes a similar point about Dieter Schnebel's analysis of Op. 27 (which dates from 1952 but was not published until 1984). When he speaks about motivic transformations, Borio (2005: 95) observes, Schnebel does not talk about their chronological development but rather their formal derivation. The contrast is with the Schoenbergian analytical tradition in terms of which Stein conceived Op 27, and which continued after the war in the publications of Leibowitz, Spinner, Walter Kolneder (1961) and Friedrich Wildgans (1967): designed to emphasise the continuity of serial music with the classical tradition, such analyses adopt a narrative, diachronic stance that allows at least a degree of mapping from analysis to performance. But in analyses like Klammer's or Schnebel's the narrative axis is no longer that of time: the music is rationalised in terms of what Iannis Xenakis (1971) called 'outside-time structure'. Such synchronic analysis is more or less untranslatable into the quintessentially temporal practices of performance or, indeed, listening.
But there is more to it. Shreffler (1994: 238) speaks of Webern's vetoing performances because they could never be good enough, and of 'a kind of "cult of rehearsal", which was actually used to limit outside performances'. Again, Schoenberg's notorious quip that the performer was 'totally unnecessary except as his interpretations make the music understandable to an audience unfortunate enough not to be able to read it in print' (Newlin 1980: 164) is echoed in Kolisch's insistence as late as the 1970s that he 'no longer believed that performance was important or useful at all' (Satz 2002: 207). In short, there was within the Second Viennese School a tension between composition and performance, and this continued into the post-war avant-garde-most obviously through the translation into practice of Schoenberg's (1975: 326) pre-war fantasies about mechanical instruments making it possible to do away with performers. 22 But at Darmstadt a new element was added to the mix: a celebration of the contradiction between the composition and its actualisation. Pousseur (1966: 95) wrote that 'the divergence between serial procedures and the perceptible result is sought after and exists effectively'. And citing an analysis of Stockhausen's Klavierstück III by Christoph von Blümroder, Grant (2001: 66) concurs: 'the tension between the predisposed order and the individual's interaction with it … can indeed be seen as its raison d'être'.
The result of this way of thinking was to leave performers outside the charmed circle of 'mystical idealism' (a term that Quick [2010: 39] applies to Webern but is even more applicable to the post-war avant-garde). Because abstractly conceived new music relied so heavily on translation into viable performance practice-a point to which I shall return at the end of this article-those who actually played it occupied an increasingly vital role in the production and dissemination of new music. Yet at the same they were increasingly elbowed out of the official discourses of musical modernism. In the long term this gave rise to the untheorised culture of what Brian Ferneyhough calls 'the "gig" musician . . . who, in a couple of rehearsals, is justly proud of producing a "professional" realization of just about anything' (Ferneyhough and Boros 1990: 6-7). Theory was the preserve of composers.

New mainstreams
Even if 'Darmstadt pianism' designates less a distinctive style than a reaction against overt expressivism, the Piano Variations bear out Quick's claim about there being two broad approaches to Webern performance after the war. From the 1960s up to the six-year break in recording after 1977 and the publication in 1979 of Stadlen's annotated score-which effectively closed the first chapter in the history of Op. 27 on record-the broad narrative is one of convergence. I shall provide an overview so brief that it might best be seen as a listening guide. There is on the one hand the tradition that stems from Stadlen and Manchon-Theis, the echoes of which are to be heard in the recordings not only of Aitken and Webster-who belonged to the same generation-but also of younger pianists such as Carlo Pestalozza (1961) andFranzpeter Goebels (1964): in Pestalozza's, for example, there is the same searching for meaning that I identified in Stadlen's and Manchon-Theis's recordings, while there is enough similarity between Pestalozza's and Goebels's recordings-in the slow tempos, the rather disjointed B section (echoing the Stadlen tradition), and the noticeably more tranquil playing of the A2 section-to suggest that Pestalozza's recording formed part of Goebels's preparation. 23 Both pianists were born in 1920 but made their recordings before Webster made his: that prompts questions of how far their recordings (or indeed Webster's) should be regarded as the continuation of a pre-war tradition or as a revival-and of the criteria on which the distinction might be made. I shall not attempt to answer these questions but will simply refer to these six recordings as representing the first wave in performances of Op. 27. On the other hand there is the tradition of Monod, Stein, Jacobs, and Loriod, supplemented by Gould's six recordings from 1954 to 1974, which span almost the entire period covered by this article. I refer to these ten recordings as the second wave in performances of Op. 27, while recognising the extent to which the two waves overlapped.
By the 1960s these two approaches had developed their own momentum as ways of playing the piece. Whereas in the 1950s to play Op. 27 one way or the other amounted to a statement of aesthetic or ideological affiliation, by now they simply represented alternative stylistic options from which performers could choose. As I suggested, it is perhaps Jacobs's 1956 recording, with its combination of first-wave expressity and second-wave cool, that most tellingly anticipates the accommodation that was eventually reached between the two styles. In essence Jacobs maintained the stylistic features of second-wave performance while adopting first-wave tempos; nine years Goebels did something similar, combining first-wave tempos with something akin to Loriod's staccato articulation on the third and fourth beats of the groups in the opening section (though he artfully sustains certain notes behind the staccato ones, creating a kind of halo around the sound that results in a much less aggressive effect). And as we saw, with Webster it is the other way round: like Goebels, I placed him in the first wave because of a preponderance of backward-looking stylistic features-the most obvious being his generous use of hand breaking-but he combines these with tempos closer to those of the second wave. These hybrid interpretations might suggest a looseness of entailment between available stylistic features, and hence the viability of a rather free kind of mix-and-match approach.
In practice, however, that is not what happened. There was instead a convergence on a kind of standard or default interpretation, in which the overtly gestural playing characteristic of the first wave was accommodated to a greater or lesser degree within a framework defined by the relative-though not literal-compliance with the score characteristic of the second. A typical performance in the third-wave or 'consensus' style would feature tempos somewhere between those of the first and second waves; the opening A section would begin fairly metronomically but become more expressive as it continued; there would be strong tempo arching of at least the isochronous groups in the middle section, with continuity across them perhaps being achieved through the same handling of the terminal notes as in Monod's and Jacob's recordings; E91-121 would be projected as a single group enlivened through attention to Webern's piano and staccato markings at E100-104 and 113-17, with something of a torn-off quality at E121 and a major rallentando and caesura at the end of the section; and the final section would be taken at a significantly slower tempo than the first one, though in other respects a variety of approaches could be adopted (it might be handled similarly to the opening section, be played more expressively throughout, or become increasingly sublimated as it continued). After the decades of verbal and musical controversy, a tradition had been invented.
It is not, of course, that all recordings within the consensus style are the same, but that they are easily classified in relation to this hypothetical norm. For example the pattern of beginning the A section metronomically but becoming increasingly expressive is found in the recordings of Claude Helffer (1968), Marie-Françoise Bucquet (1969), Charles Rosen (1969), Alexei Lubimov (ca 1971), and Bruno Mezzena (1973, while the Monod-Jacobs strategy for the terminal notes in the B section is adopted by Helffer, Lubimov, and Yuji Takahashi (1977). On the other hand there are idiosyncratic elements. In a singularly accident-prone live performance from the Fourth Cliburn Competition, Christian Zacharias (1973) repeatedly abbreviates the first notes of the four-event groups in the A sections, while Andrzej Dutkiewicz (1975) prolongs the third notes; again, Zacharias accelerates rapidly around E109, rips off E121 like a strip of cloth, and lunges viciously to an unnotated fortissimo at 124-5, whereas-equally contrary to prevailing norms-Dutkiewicz breaks up E91-121 into a series of discrete gestures (rather like Stadlen, though the breaks fall in different places). Both the degree of consensus and the occasional striking deviations from it suggest that pianists were by now increasingly playing in relation to one another, encouraged no doubt by the increasing circulation of recordings. One might say it was at this time that Op. 27 ceased to be a World War 2 orphan subject to contested interpretations and mainly the preserve of contemporary music specialists. As in all invented traditions, its performance was naturalised, as if the way it was now played was simply the way the music goes (Hobsbawm and Ranger 1983). And in this form it increasingly entered the repertory of generalist pianists. It became part of the modernist mainstream-a development that might be dated to Eric Salzman's claim, in a December 1973 review of the Turkish pianist Idil Biret's recording, that 'a sensitive performance of the by-now classic Webern Variations completes an impressive American recording debut '. 24 Of no less interest, however, are the outliers, four recordings that cannot be sensibly classified. One is the recording that prompted Salzman's claim, which has something of the hybrid quality of Jacobs's, Goebels's, and Webster's. Biret's tempos, especially in the A section, are slow even by first-wave standards. Yet she plays the B section with a striking pointillism that sounds so anomalous in this context as to suggest a deliberate intertextual reference to-again-Loriod's recording of twelve years earlier, for many the paradigm of Darmstadt pianism. (At Biret's much slower tempo any link with jeu perlé disappears.) Biret also slows down gradually during the second half of the final section, so creating a particularly intense version of the sublimated narrative I first identified in Manchon-Theis's recording. I spoke of pianists increasingly playing in relation to one another, but Biret's recording sounds like something more than that: a personal commentary on a now established discographic repertory. A second outlier is Maurizio Pollini's recording, from 1976, which is individualistic in several ways. In the first seven bars of the A1 section, for example, he creates a single arch encompassing E5-16, using a combination of tempo, dynamics, and articulation to weld the separate groups into a single gesture, while at the opening of A2 he plays almost metronomically but uses dynamics to shape the grouping pattern. There is a kind of simplicity to his playing, as if the answers to the interpretive problems pianists had been grappling with over the previous quarter century were really obvious all along-which is perhaps why this recording became one of the poles in relation to which interpretations were formed and evaluated during the second chapter of Webern's Piano Variations on record.
More obviously calculated to controvert existing norms is Carles Santos's recording from the following year, the basic strategy of which is to do everything differently from everyone else. He begins very slow, with exaggeratedly metronomic and dynamically even playing, but arching the group from E17, and again at E26-this time so violently that it sounds like a sudden change of tempo. In the first part of the B section he consistently plays the repeated notes at the palindrome points (E55, 70, and 88, all of which Webern marked sf) much more softly than the preceding note, so controverting the idea of symmetry. In the same section he adopts Monod's and Jacobs's strategy of prolonging the notes before the dyads at E52, 67, and 85, but at E57, 72, and 90 it is the dyads that he prolongs, and in this way he pairs the groups-a quite different but eminently logical approach. He breaks up E91-121, like Stadlen and Dutkiewicz, but at points that do not fit with Webern's indications. And in the A2 section he breaks the first rule in the book by playing the section faster than A1: Santos's is the earliest of the three recordings in which the average tempo for A2 is faster than for A1. 25 Perhaps even more idiosyncratic is the way Santos plays individual notes of a chord staccato while the other(s) are held. A kind of reverse version of the hand breaking that persisted up to the 1970s in the recordings of Rosen and Zacharias, this technique had already been tried by Goebels, but Santos uses it in a systematic manner that borders on the perverse. The first clear example is E20-21, where Santos sustains the C# and G; as E22 is a single C#, the result is to underline a palindrome axis that Webern's asymmetrical texture camouflages, and Santos does the same at E40-42 and elsewhere, including in A2. Once again the strategy is transparent: where Webern makes the symmetry obvious, Santos disguises it, while where Webern disguises it, Santos makes it obvious. In short, Santos handles the piece more in the manner you might expect of a composer than a performer-which makes sense, as within a few years he had moved on from his initial career as a pianist to become a composer, film-maker, and sculptor. There is one final example, as it happens by a protegé of Manchon-Theis, 26 that is equally subversive of standard performance practice. Jean-Rodolphe Kars's recording, from 1969, stands out by virtue of its tempos: his A sections, in particular, are not just slower but far slower than anyone else's. The average tempo for A2 is 17, and during E133-48, where Kars consistently prolongs the third attack of each group, the sense of musical motion practically disappears. The result is something that might also be said of the reverse hand breaking in Santos's recording: the focus of listening is directed away from musical causality and onto the quality or microstructure of individual sounds. The same might be said, if to a less extreme degree, of the Takahashi siblings's broadly similar recordings, 27 which are the slowest within my 'consensus' category-and it is not irrelevant that Aki's recording was part of the famous 'Piano Space' album, which coupled works by composers of the European and Japanese avant-garde, together with Cage. Morton Feldman (2000: 55) wrote of her performance of his Triadic Memories that 25 The others are Pollini, in a live recording from Berlin (1990), andChristoph Eschenbach (1996). 26 As stated on the sleeve of Kars's recording. 27 The similarity analyser shows these to be closely related in both tempo and dynamics across the movement as a whole, a finding consistent with Quick (2010: 197).
'Takahashi appears to be absolutely still. Undisturbed, unperturbed, as if in a concentrated prayer…. The effect of her playing to me is that I feel privileged to be invited to a very religious ritual'. That resonates with Kars, who credits Messiaen for his conversion from Judaism to Catholicism, and gave up public performance in 1981, becoming a priest five years later. 28 And this spiritual, even New Age conception of the Piano Variations has its own roots in Darmstadt. The focus away from causality and onto the microstructure of sound better illustrates Grant's 'microaesthetics', and more generally her definition of serial hearing, than any of the second-wave recordings. But it also reflects an idea of space that resonates with the synchronic quality of Schnebel's and Klammer's analyses. Takahashi's album title echoes the Darmstadt idea of musical space that Ligeti specifically ascribed to Webern, whose music- Ligeti (1965: 16) says-'brought about the projection of the time-flow into an imaginary space by means of the interchangeability of the temporal directions, provoked by the constant reciprocity of motivic shapes and their retrogrades…. Webern's structures seem to circle continuously in their illusory space'. And in his contribution to the Webern issue of die Reihe Stockhausen (1958: 65) says that when in music 'we realise … that we have "lost all sense of time", then we have in fact been experiencing time most strongly'-an idea that gave rise to his later conception of forms that 'cut across horizontal time experience into the timelessness I call eternity '. 29 Representing what Fox (2007a: 14) describes as the Darmstadt composers' 'desire to make other-worldly music', Stimmung was published the year before Kars's recording of Op. 27 was released. That recording, and those with which I have associated it, suggest how it might after all have been possible to move beyond traditional narrative paradigms towards a performance practice more deeply imbued with Grant's serial aesthetic. It is striking, however, that in the first movement of Op. 27 Kars's basic strategem for achieving this is to perform the music at an average tempo of just half Webern's metronome marking, and the final section even slower. As in the case of Santos's recording, this takes non-literalist interpretation to a level that begins to challenge the very idea of performance as conventionally understood within the Western classical tradition.

The shadow of meaning
If the Piano Variations became mainstreamed during the 1960s and 70s, the same might be said of music analysis and theory, at least in North America. As Grant (2001: 2) has explained, dissatisfaction with the European music journals' neglect of American developments was one of the factors behind the establishment in 1962 of Perspectives of New Music, and the first volume included a scathing attack on die Reihe's pseudoscientific mystification of new music. As usual Op. 27 was in the front line, although it was the 22-bar second movement on which attention focussed. Christian Wolff had made this the topic of his contribution to the Webern issue of die Reihe-the only article by an anglophone writer-but in the 1960s there was a spate of writing on it to which several of the leading figures in American music theory contributed: authors included Milton Babbitt (1960), David Lewin (1962), Wilbur Ogden (1962), Peter Westegaard (1963, Roy Travis (1966), and James Rives Jones (1968). The aim was to move away from the European analysts' numerological obsession with imperceptible underlying systems and towards a more pragmatic focus on contextually defined and perceptible structures: several of the American writers focussed on the intricate relationships between the serial structure and the contrapuntal design of the movement. But there was still a general process of abstraction from surface to system, which is to say from 28 Information from his Wikipedia entry. 29 Translated (from volume 1 of Stockhausen's Texte, 1963) in Hasty 1997 temporal unfolding to atemporal structure. Hence the problems of translation from analysis to performance that I identified in Stein: his DMA thesis cites the writings of Lewin, Westegaard, and Ogden, among others, but they appear in footnotes rather than being thought into the main argument.
Where the European modernists of the immediate post-war years were content to celebrate the contradiction between idea and actualisation, however, their American counterparts (who were more securely embedded in the higher education system) made a determined attempt to bring performance within the academic orbit-a development aided and indeed mandated by the requirement that teachers of performance in American universities should possess doctorates. (Hence Stein's 1965 DMA.) The particular form this took reflected something that American theorists shared with their Zero Hour predecessors: a social-constructionist distrust of the traditional idea that the tonal system is given in nature and hence uniquely privileged. This idea is nowadays mainly associated with twentieth-century conservatives from Schenker to Pfitzner, but it was equally an article of faith for Webern (1963: 12-13), for whom the historical modes and scales were 'a manifestation of the overtone series….
[A]s a material it accords completely with nature', while for Stadlen (1958: 19) tonality was not a 'phase or a style, but … the nature of music'. Like the post-war avant-garde in Europe, American theorists, who were often also composers, were determined to reconstruct musicwhether through composition or analysis-on the basis of explicit, artifactual principles.
Whereas performers had ended up outside the magic circle of Darmstadt serialism, in America the constructionist ideology was pursued to its logical conclusion. In effect performance was thought of as an extension and supplement to composition. And if that composer-centric view echoes Schoenberg's disparagement of the self-importance of performers, there was also a parallel to his distinction between style and idea, between presentation and the essential content that was presented-which in 1960s-70s America meant structure. Within this context, as Westegaard (1974: 71) wrote, 'style' became a dirty word, admitting an unacceptable element of historical contingency. In this way the page-to-stage approach, as it developed through key texts by Edward T. Cone (1968) and Wallace Berry (1989)-both composers as well as performers-and became part of the music-theoretical mainstream, was at heart radically anti-historical. It sought to rationalise and discipline performance, giving it a subordinate place within a value system centred on composition. Its hegemonic nature has been widely criticised (Mathew [2011: 58] refers to 'Edward T. Cone's authoritarian notion that "analysis is a direction for a performance"'), not least for leaving little room for performance as a creative practice.
Although it came from the other side of the Atlantic, a study by Christopher Wintle (1982) of the second movement from the Concerto Op. 24-the main title of which is 'Analysis and performance'-provides a good illustration of what is at issue. The article explores the way in which Webern organises his music hierarchically-that is, creates relationships between musical surface and underlying structure-without drawing on traditional tonal formations. Despite the title, no discernible theoretical model for the relationship between compositional structure and performance is in evidence. There is simply a slippage from the one to the other via the notion of interpretation: after explaining his atonal background structures, Wintle (1982: 91) asserts that 'An awareness of these articulations is essential to a properly balanced interpretation of the movement'. In other words, the musical structure-if correctly interpreted-entails the manner of its performance. That puts the analyst in charge, and by the end of the article-though he has made no reference to any particular performance of the work- Wintle (1982: 98) is in a position to conclude: it is striking that in the available commercial recordings of the Concerto, so little comprehension of structure is evinced. Dynamics are ignored, phrasing is under-articulated, tempo gradations are over-ridden, and the whole deprived of the sense of directed motion that alone can bring this music to life.
Bald value judgements of that kind are rarely encountered nowadays, but the underlying thinking is still to be found within the music-theoretical mainstream. And as I said at the beginning, such thinking was problematised by Stadlen's evidence of how Webern expected Op. 27 to be performed. The resulting confrontation was staged in an article by Robert Wason, published in 1987. The trouble starts with Webern's insistence, which Stadlen records in both his edition (p. V) and his 1958 article (16), that Webern did not consider it necessary for performers to understand the serial structure: Wason (1987: 62) suggests that Webern was unthinkingly parroting Schoenberg's statements to this effect, and adds, 'we cannot but wonder whether his advice is to be taken at face value'. Then there is the matter of the melody lines that-as Stadlen (1958: 12) records-Webern said must be 'as telling as a spoken sentence', and which Stadlen dutifully ringed in his score. The problem is that these notes cut across the serial structure, which leads Wason (1987: 95) to observe that 'The derivation of these "melody" notes is itself an interesting phenomenon for further study, although … it has so far eluded systematic explanation'. The implication is that analysis is still a direction for performance, it's just that in this case we don't yet have the right analysis.
But the central standoff concerns the nature of the relationship between musical content and its realisation in performance. Stadlen addressed this in his interview with Pauli: after referring to the 'very artistic mirror forms and retrogrades' composed into the first movement of Op. 27, he said that in performance Webern wanted 'not only something additional … but something that occasionally even contradicts the construction, something that destroys it, so that … the listener … must-or should-be misled now and again'. 30 This dialectical conception, familiar from Adorno's writings on performance, is reduced in Wason's (1987: 102-3) article to the idea that the performer may on occasion play against structure, 'although obviously one must have a clear understanding of just what one is "playing against"'. And a few lines earlier, Wason speaks-with specific reference to the B section-of the possibility of playing 'against the structural segmentations, while certainly assuming their existence'. It's a lawyer's argument: if there is no sign of the structure in performance, that only proves how deeply embedded it really is. Mathew (2011: 61) refers to theorists such as Wason as 'so entangled in the poietic fallacy that they proceeded from the premise that Webern's performance instructions must somehow be linked to the serial structure of the Variations'. For theorists who think this way, the assumption that phenomenal surfaces-what we play and hear-are generated by underlying structures is too deeply embedded to be questioned, and that gives rise to the further assumption that explanation must mean derivation-as in what Wason says about the 'melody' notes, or for that matter Schnebel about motivic transformations.
At the same time, Wason (1987: 64) does recognise that 'twelve-tone "structure" is not the only structure going on' in Op. 27. He cites the formal plans of all three movements as 'examples of traditional thinking that Webern transports into the twelve-tone domain', and adds that 'such notions as "motivic structure" … reveal significant structural aspects of Op. 27, despite the fact that these "structures" are not per se twelve-tonal'. At these points, as also in his reference to the 'stormy passage' at the end of the B section-where, he says, traditional climax-building techniques such as phrase compression, rapid dynamic alterations, and registral expansion clarify the meaning-Wason is circling around something that almost breaks the surface when he mentions Webern's own comparison of the B section to a Brahms intermezzo: 31 the extent to which this movement draws on the familiar rhetoric of tonal music. Wason's reluctance to articulate this may reflect the social-constructionist ideology to which I referred, but it also has to do with the valorisation of structure rather than style. Here again Wintle's 1982 article illustrates what is at issue. Wintle builds on Webern's (1963: 54) remarks in The Path to the New Music about how effects of tonal excursion and return can be created by purely serial means. In Op. 24, Wintle argues, Webern creates pitch hierarchies that are in essence structurally isomorphic with those of tonal music, yet do not entail the stylistic practices of tonal music: 'Schenkerian tonal operations', Wintle (1982: 98) says, need not be invoked. In this way the historical dimension is acknowledged, but on terms that hold nature at bay.
Performers, however, did not share such qualms. Kolisch and other performers of the Schoenberg circle believed that the same principles applied to the performance of contemporary and earlier music. Lowell Creitz (2002: 162-3), who joined Kolisch's Pro Arte Quartet in 1955, writes that 'The procedures we used were common for all music, not just that of the 20th century', and cites Webern saying his Quartet Op. 28 should be interpreted in the same way as 'many of the piano sonatas by Beethoven in three movements'. Creitz (2002: 166) goes on to describe how the Pro Arte Quartet developed its interpretations, saying 'The goal at all levels of structure is to discover the gesture of the motif, the theme, the section and the movement'. Given the Schoenberg connection, it is not surprising that the approach resonates with Stein's thesis, and this is the context within which Stein makes reference to the tonal tradition. It is not only his use of terms such as recapitulation and elaboration, or his reference to the building of the climax in Wason's 'stormy passage' as 'a familiar developmental device'. It is also, for example, when he says that the periodic construction of Webern's themes and in particular his use of mirror forms-Stein cites the opening of Op. 27 as one of several examples-'suggest harmonic movement of the order, I-V, V-I' (Stein 1965: 73). While this illustrates how Stein invoked tonal models as he sought out the musical gestures, however, an especially vivid sense of the shaping of serial materials in tonal terms can be gained from Stadlen's (1958: 18) account of how he interpreted the series: you can imagine him thinking with his hands when he says that 'we instinctively group the notes of a pitch-set according to chords that are, however vaguely, suggested by its outline and according to possible metric analogies whose rhythmic emphases again allocate harmonic priorities'. On the same page Stadlen speaks of atonality 'reminding us of our shady past', and questions whether atonality has any real existence other than through these historical references. And Adorno (2006: 87) extends the argument to performance when, following his complaint that Op. 27 offers 'monotonously symmetrical presentations of serial marvels', he adds that 'for its performance to give the monotonous tone groups even the shadow of meaning, it must distance itself infinitely far from the rigid notation, especially of its rhythm'.
But the best evidence of the persistence of tonal ways of thinking within the performance of Op. 27 is, of course, provided by the recordings. I have hardly mentioned tonality in my descriptions of them, but really I have been talking about the practices of tonal performance all the time. It is particularly in the B section that performances of Op. 27 embody modes of temporal and dynamic shaping-the kind of gesture Creitz was talking about-that are conditioned by the nineteenth century. All the recordings I have described in detail are consistent in shaping isochronous groups into distinct gestures, usually involving some degree of tempo and dynamic arching; articulating relationships within non-isochronous groups by means of agogic and dynamic accentuation, though in very different ways; using tempo and dynamics to build the second half of the section towards climax and collapse; and separating off the section as a whole through more or less extended silences. The opening bars of the outer sections, too, form a storehouse of ways in which different patterns of emphasis can be created through the interaction of duration, dynamics, and articulation. These are the standard techniques of expressive performance as analysed, for example, by Eric Clarke (1988), and-to follow up Webern's, Stadlen's, and Adorno's hints-could be illustrated equally readily by recordings of any Brahms intermezzo. It is as if the crystalline Piano Variations of the lecture hall and the shadowy Piano Variations of the concert hall-the serially and the tonally conceived pieces known by the same name-existed in parallel universes. That explains why Webern thought it unnecessary for performers to know about serial design, and said his music should be played in the same way as Beethoven's. It also explains a sense of disconnect between performance and the institutionalised music theory of American academia that is particularly obvious in the case of serial music but hardly unknown in relation to common-practice repertory: only a handful of professional performers-Murray Perahia is the outstanding exception-have seen deep structures such as the Schenkerian background as offering an indispensable insight into the music they play. If academics and performers have in recent years moved towards a more productive alignment of interests, the reason is not that theorists have discovered the right analyses to direct performance, as Wason envisaged, but rather that they have become more pragmatic, more open minded, and more sensitive to the different purposes that analysis can serve. It is also worth pointing out that such moves have tended to occur in countries, such as the UK, where theory has never been as fully institutionalised as in the US, and where the divisions between theorists and historical musicologists are more permeable-and moreover where collaborations between performers and academics have been incentivised by research funding regimes.
And if performers of Op. 27 have played Webern's serial music as if it was tonal music, then their performances have been received as if they were tonal music. That might seem an unfounded claim given a critical emphasis on the geometrical, crystalline-in short objective-aspects of Op. 27 that has still not disappeared from view: in a 2007 review of Mitsuko Uchida's recording, Steve Schwartz remarks that 'She doesn't play Webern like a mathematical proof, but in the same way she plays Schubert and Mozart-clearly, simply, with an adult's point of view and experience', and then adds, 'Webern purists may hate this approach, excoriating her for a "non-objective" view of the score'. But I can offer a negative demonstration of my claim. As we saw, the same objective qualities have been repeatedly attributed to the modernist performances of the 1950s-60s. I cited Mathew's characterisation of Monod's recording as 'unmediated, un-interpreted Webern', and Wason (1987: 101) broadens the critique to 'the mechanical rhythm and flat, emotionless dynamics heard all too often in performances of Op. 27', while Malcolm Hayes (1995: 223) extends it to Webern's music in general ('when it came to performing Webern, a dispiriting species of metronomic soulfulness was the order of the day'). Yet recordings of Op. 27-even second-wave recordingsactually embody a degree of expressive nuance that, in simple quantitative terms, is quite comparable to what might typically be found in performances of a Brahms intermezzo: the point can be made by comparing performances of Op. 27 and Brahms's Op. 116 No. 5 by the same pianists, though I am consigning the details to a footnote. 32 And this contradiction between fact and perception need not be put down solely to the stubborn persistence of aesthetic ideologies, or to musicologists' failure to listen. Even recordings like Monod's and Loriod's draw on deeply internalised, historically sedimented codes of expressive performance, and it is because these codes are also internalised by listeners-because of what Bruno Repp (1999) calls listeners' 'timing expectations'-that they are not perceived for what they are. The result is that the objective qualities attributed to the composition are incorrectly ascribed to the playing as well. And the reason this happens is that performances of Op. 27 are being heard on the model of tonal music.
-o -o -o -Seeking to circumscribe but not wholly eliminate the role of the natural, Fred Lerdahl (1989Lerdahl ( , 2001 has argued that atonal music lacks the acoustically grounded stability conditions of tonal music, which means that what would in tonal music be understood as purely surface features play a key role in generating perceptual hierarchies. The point can be made through the specific example of cadences. In an article that applies his idea of the structural downbeat to the second movement of Op. 27, Cone (1962: 45) succinctly defined the nature of a cadence in tonal music: it is 'the point in the phrase at which rhythmic emphasis and harmonic function coincide'. In atonal music one half of this equation is absent, since without stability conditions there can be no harmonic function. Consequently other means have to be found to create cadential effects. There is an example at the end of the A1 section of Op. 27's first movement: both Wason (1987: 97) and Bailey (1991: 192) comment on the introduction of various irregularities into the palindrome (octave transposition creates a registral descent, rhythmic adjustment isolates the end of the group), the purpose being-in Bailey's words-'to serve as a cadence'. Viewed in terms of tonal theory, the point is that a cadence like this does not emerge from the underlying structure. It is all surface and no background, an exampleas Quick (2010: 205) puts it-of 'the goal-directed, harmonically-driven rhetoric of tonal music-even in the absence of such an organisational system'.
In an article designed to debunk the European avant-garde's attempt to portray Op. 27 as a precursor of total serialism, Westegaard (1963: 30) argued of the second movement that the patterns in its various 'nonpitch' parameters-by which he meant rhythm, articulation, and dynamics-make no sense in themselves: they 'are still playing their traditional role of differentiation. They interact with one another and with pitch to clarify pitch relationships, sorting out for the ear those pitch relationships which are to shape the movement'. That applies also to the first movement, where the gestural shaping to which I referred often involves the coordination of pitch, rhythm, and dynamics, and it turns on the distinction made by Leonard Meyer (1989) between on YouTube [http://www.youtube.com/watch?v=23pwn2__fzA]). Standard deviations weighted against tempo are used as an admittedly crude measure of rubato, and on this measure, the pianists on average employ the same amount of rubato in Op. 116 No. 5 and Op. 27 overall, though in the central sections there is 18% more rubato in Op. 116 No. 5 (with the repeats averaged). That however masks large individual differences between performers and recordings. Biret plays with considerably more rubato in Op. 27 (31% more overall and 35% more in the central sections); so does Richter, for whom the corresponding figures are 17% and 62% (the latter figure averaged across his four recordings, between which there is considerable variation). Pollini does the opposite: on average he plays with more rubato in Op. 116 No. 5 than in his three recordings of Op. 27 (22% more overall and 23% more in the central sections), but here the variation between recordings is massive: whereas his performance of Op. 116 No. 5 has 47% more rubato overall than his 1976 recording of Op. 27, it has 3% less than the Berlin recording of Op. 27 from 1990. 'primary' and 'secondary' parameters and by Schenkerians between 'structure' and 'design': the difference is that whereas in tonal theory it is the conjunction of pitch and rhythm that defines the musical content, with the other parameters serving an ancillary role, in serial theory it is the sequence of pitch classes alone. The result is the gap beween serial theory and broader musical significance that leads Christopher Hasty (1988: 285) to complain that 'pitch-class relations per se offer us little insight into the totality of musical organization. Meticulously crafted details of duration, accent, contour, and instrumentation can rarely be rationalized by the serial structure, and when they are treated by analysts, they are generally relegated to the "musical surface"-a surface curiously detached from the serial "background."' The conclusion Hasty draws from this is that serial organisation, in itself, is a much weaker determinant of what he calls 'the true structure of the work' than its tonal equivalent: it relies much more heavily on the interactions between pitch classes and and other parameters, and for this reason Hasty urges analysts to devote much more attention to the role of nonpitch parameters in 'the creation of musical gesture and form'. But of course it is not just analysts who devote attention to the role of nonpitch parameters in the creation of musical gesture and form. That is basically the pianist's job description. It is then telling that early performers from both camps agreed in emphasising the formative role of nonpitch parameters in the performance of Op. 27. Stadlen (1958: 14) wrote that for Webern 'the notes had become almost incidental and were only regarded as carriers of expression', while Stein (1965: 71) spelled out more precisely what this meant when in his thesis he spoke of 'duration, dynamic, touch, register … acquiring equal importance with the pitches themselves'. It was also Stein (1965: 173) who spelled out what this meant for the performer's role. Because such music 'is usually bereft of the form-building tendencies of harmony and meter', he argued, 'the performer must begin from scratch and gradually build up the various levels of formal relationships which are characteristic of each new composition'. (That is an example of what I meant about performers playing an increasingly vital role in new music.) Pousseur (1964: 82) made a similar observation, again making the comparison with common-practice repertory: in the works of Webern's late period, he wrote, 'you feel (hear) that the performers are there (much more than in tonal music, which seems to perform itself), that they produce tones and meanings which they define'. And so we end up with a paradox. As we have seen, it was within a context largely conditioned by serialism that North American music theorists developed the page-to-stage approachan approach that sought to embrace performance within a value system centred on composition, and so left little room for thinking of performance as a creative practice. Yet it was precisely in serial music that the detachment of 'musical surface' from serial 'background' to which Hasty refers expanded performers' opportunities to co-create the music that listeners heard. The agency denied in theory could be asserted all the more strongly in practice.
The reception at Darmstadt of Webern's serial compositions represented in its most extreme form the resistance to history, and the search for meaning in the text, that have underpinned the disciplinary identity of music theory. In this article I have attempted not only to trace the history of Webern's Piano Variations in early recordings, but also to show how the recordings are themselves permeated by history. Perhaps all I have really done is unpack and illustrate Mathew's (2011: 67) pithy claim that 'Webern's music is made of the kind of themes, topoi, and gestures that we are familiar with-things that we know how to sing and play'. That historicising claim reflects and contributes to the critical reconstruction of Webern that has taken place over the years since Zero Hourfrom arch structuralist to arch expressionist, from composer of extreme autonomy to composer of extreme referentiality, from the symbol of a new beginning to a figure deeply entangled in history. It has taken musicologists and theorists half a century to accomplish this 180-degree reversal in critical interpretation. But that only illustrates

Abstract
Webern's Piano Variations Op. 27 were first performed two years before the outbreak of Second World War, but it was only after the war-and in very different aesthetic and ideological circumstances-that a performance tradition developed. Through a combination of computer-assisted close reading and contextual analysis of the first movement, this article traces the competing stylistic and ideological directions of early performances-one deriving from the pre-war tradition of the Schoenberg circle, the other from the European avant-garde associated with Darmstadt and die Reihe-and the processes of negotiation through which a more or less consensus interpretation developed by the 1970s. My aims are to explore ways in which empirical approaches can be deployed for musicological purposes, to relate recorded performances to their historical contexts, and to link the story of Op. 27 to the development of musictheoretical approaches that see performance style as a function of compositional structure. Key recordings on which the article is based are presented online as Sonic Visualiser session files coordinated with the text.