Edgard Varèse's Poème Électronique: From the Sketches to the Sound Spatialization Free

In the late 1950s, during a transformative period for electronic music, Edgard Varèse's Poème électronique (often referred to hereafter as PÉ) (1958) emerged as a remarkable exception that defied the prevailing trends. At the time, Parisian musique concrète and production in the Cologne studio dominated the music landscape. In line with his other works, Varèse did not provide interpretative guidance for PÉ, nor did he explicate his compositional methods through theoretical writings. This composition continues to challenge scholars seeking to position it within the context of late 1950s electronic music, even two-thirds of a century after its creation. A significant factor contributing to this difficulty is the presumed scarcity of written sources related to the work.

PÉ was created for a multimedia performance at the Philips Pavilion during Expo 58 in Brussels, the first world's fair following World War II. The pavilion, for which Le Corbusier was commissioned, had a futuristic structure designed by Iannis Xenakis, who at that time was his assistant. Le Corbusier also curated the visual components of the performance, consisting of colored lights and a film projection. From September 1957 to April 1958, Varèse worked in a dedicated facility provided by Philips, specifically, the garage at the Strijp III complex in Eindhoven, near the Philips Research Laboratories. As Kees Tazelaar (2013) has meticulously documented, Varèse had access to cutting-edge equipment for electronic music production, supported by expert Philips engineers such as Willem Tak, Jan de Bruyn, and Anton Buczynski.

Varèse recorded PÉ using three monophonic tapes, which were played simultaneously on three tape recorders. An approximate synchronization was achieved via a lead-in tape containing sync and count markers. During the performance of PÉ Philips engineers used a specialized three-channel perforated 35-mm tape, commonly used in the film industry. The perforations ensured synchronization with the projected films and an additional perforated tape that contained control signals for the fully automated performance. A total of 325 loudspeakers were mounted on the walls of the pavilion, supplemented by an additional 25 speakers positioned around the perimeter for low frequencies. This arrangement allowed sound to emanate from all directions, either revolving around the audience or descending from a height of approximately 20 meters at the pavilion's apex. Unfortunately, it is no longer possible to experience PÉ in its original form. The pavilion was demolished shortly after the Expo's conclusion, and, with it, the only equipment capable of reproducing this multimedia spectacle was lost. Since the pavilion's demolition, PÉ has primarily been regarded as a standalone musical composition. This is largely due to a version released on the album Music of Edgar Varèse (Columbia 1960, MS 6146), which reduces the original multidirectional sound spatialization to a stereophonic format.

The sound material made available for public performance followed a different path. In the 1960s, Frits Weiland produced a four-track version based on the original Varèse tapes, redistributing the sound events in quadraphonic space. This tape, preserved at the Studio voor Electronische Muziek Rijksuniversiteit in Utrecht, also served as the basis for the Decca recording edition (Varèse: The Complete Works, Decca 460 208–2, 1998) and the stereophonic materials rented by Ricordi for concert performances. Recently, two reconstruction projects have aimed to recreate the original spatialized listening experience in the Philips Pavilion. In 2009, the Virtual Electronic Poem (VEP) project (https://www.cirma.unito.it/vep) utilized the original preparatory tapes, now preserved at the Institute of Sonology of the Conservatory of The Hague, and developed dedicated software for spatialized sound reproduction. Through this system, PÉ could be experienced in a virtual simulation of the original acoustic space, based on a 3-D model of the building's interior (Lombardo et al. 2009). In 2015, Kees Tazelaar performed a spatialized (5.0) film version of the work in the concert hall of Zürcher Hochschule der Künste, based on further research into the visual projections during the multimedia show inside the Philips Pavilion. Both the VEP project and Tazelaar's reconstructions of the sound distribution were based on an excerpt of Philips technical control score containing the information regarding the sound routing. A short excerpt of this diagram, published in Philips Technical Review in 1958, displays the information about sound distribution for a 30-second section (De Bruin 1958/59). In the absence of any other data beyond this 30-second segment, the spatialization performances by the VEP project and Tazelaar were based on the curators’ interpretation of the piece. In 2021, Luca Cossettini and Alessandro Olto, working in the Mirage laboratory at the University of Udine, created a restoration of Weiland's quadraphonic tape for Ricordi, resulting in a DVD intended for spatialized concert performances (Cossettini and Olto 2021).

Following initial studies on PÉ, which were based on the limited documentary sources available in the 1990s (Lukes 1996; Treib 1996), recent scholarly research has predominantly focused on its tapes (Cossettini 2017), the audiovisual system setup (Lombardo et al. 2009; Tazelaar 2013), and Le Corbusier's scenario for the light and film projections (De Heer and Tazelaar 2017).

However, two critical aspects of the work remain largely unexplored: Varèse's compositional process and the sound spatialization used during the performance. This article addresses these overlooked elements by examining previously neglected written sources, including Varèse's musical manuscripts and the complete control score, which displays information for the spatialization of the entire piece. The primary aim of this study is to propose a philological approach for analyzing the sources related to PÉ, from the initial sketches to the final sound projection plan. Because the analysis of all the sources related to PÉ is beyond the scope of this paper, the investigation focuses on a short section of the composition, rather than the entirety of the work.

Varèse's manuscripts became available to scholars in 2003, as part of the Edgard Varèse Collection, which had previously been curated by his student, composer Chou Wen-chung. Now housed in the Paul Sacher Stiftung (Foundation) in Basel, the collection is hereafter referred to as EVC-PSS. Many sketches related to PÉ were overlooked for years, dispersed among numerous unidentified documents. The extreme heterogeneity of these materials likely hindered their identification. Indeed, this collection displays a broad range of notational styles, including multicolored plans for electronic montage, symbolic annotations, diagrams on graph paper, draft scores for electronic pitches in standard staff notation, sketches of brief rhythmic elements, and the use of Cartesian axes to notate frequency or sound envelope.

The diversity of these documents mirrors the characteristics of PÉ itself, which, compared with much of the electronic music of its time, stands out for its wide array of sound materials. However, the variety within the manuscripts also reveals the multiplicity of working methods Varèse used in the creation of PÉ. Through a systematic comparison of various sources, primarily preserved at the Paul Sacher Stiftung, I was able to identify over 50 manuscripts related to PÉ. Among these, I recognized a small group of pre-definitive manuscripts, or drafts, which Varèse likely used as reference points for the final realization and assembly of the tape.

In two sections of the composition (3:38–4:39 and 6:44–7:06), electronic sounds are almost entirely absent, replaced by two acoustic episodes—the first for alto voice, male choir, percussion, and piano, and the second for soprano and male choir. To record these segments, Varèse prepared a specific score by reworking a piece from 1947, Étude pour Espace. This score represented Varèse's final attempt to realize an ambitious project he had pursued since the mid 1920s. Throughout this period, Varèse envisioned a challenging stage work, which he referred to by various titles, including The One-All-Alone, The Astronomer, and Sirius (MacDonald 2003; see also Jostkleigrewe 2006). During these years, Varèse engaged with intellectuals and writers from diverse backgrounds, including his wife Louise Varèse, the Cuban writer Alejo Carpentier, and the French playwright and theater director Antonin Artaud, collecting several proposed scenarios and libretto drafts. Despite the various options he explored, the subject of the work remained anchored to a few central visionary images. On one hand, there is the figure of an astronomer, isolated in his observation tower in the middle of an American city, and capable of establishing a mysterious communication with Sirius. On the other hand, the star is transforming into a supernova of destructive power, and the forewarning of an impending apocalypse drives the city population into a state of irrational panic.

Among the main characters described in all these scenarios, the Astronomer faces a recurring, evil antagonist. In an early draft from the mid 1920s, Louise Varèse describes this character as a Native American “Arrow-Maker,” and a “Sorcerer” (Varèse 1972, p. 260; Macdonald 2003, p. 218), a term that reappears years after in Varèse's manuscripts for PÉ. In 1931 Carpentier wrote an account about his collaboration with Varèse (Carpentier 1931), partially translated from Spanish by Malcolm Macdonald (2003), full of details regarding this character, now called “the Visionary” (el Iluminado). The lines describing the Visionary and his fellow, referred to as “the negro,” reveal striking similarities with the characteristics of vocal passages in PÉ. In Carpentier's synopsis, “the Visionary announces the end of the world in long, slow, heavy phrases, while the choir, directed by the negro, howls voodoo monosyllables mixed with ñáñigas, Indian words and magic incantations—a verbal counterpoint,” where the word “ñáñigas” refers to an Afro-Cuban ritual dance (Carpentier 1931, translated in English in Macdonald 2003, p. 230).

After abandoning the project of a stage work, Varèse attempted in the 1930s to adapt these ideas for a new piece: an equally grandiose choral symphony entitled Espace. Certain aspects of the proposed staging reveal Varèse's visionary approach. Initially, he conceived the use of an amplification system to create the effect of sound moving around the audience. According to Fernand Ouellette's account, Varèse later envisioned vocal parts being performed simultaneously in multiple capitals across the globe, with the various choirs coordinated through a system of radio links (Ouellette 1981, p. 132). For the literary component, Varèse initially sought contributions from trusted authors such as André Malraux and Henry Miller. However, as with several of his vocal works, he ultimately opted to assemble his own text in a collage format. This involved selecting verses from pre-existing poetry collections—The Journal of Albion Moonlight by Kenneth Patchen (1941), Temblor de Cielo by Vicente Huidobro (1931), and the 16-century work Noche oscura by San Juan de la Cruz—and interspersing these with a unique form of glossolalia, based on phoneme clusters devoid of literal meaning.

The failure to complete these two major projects—a multimedia stage work and a choral symphony—coincided with a prolonged period of creative and existential crisis for Varèse. This challenging phase ultimately led to a significant reduction of the original project. Varèse chose to distill the musical ideas he had developed into a new work: Étude pour Espace, a study for two pianos, choir, and percussion lasting approximately 10 minutes. The piece was performed only once, in New York in 1947, with Varèse conducting the New Music Society choir and recording the performance using portable equipment. Dissatisfied with the outcome, he abandoned the choral symphony project.

These drafts correspond to various discrete sections of the composition, each ranging in duration from 14 seconds to 1 minute 30 seconds. A comparison with the tape reveals a significant correspondence between the notated drafts and the recorded sounds. The sequence of the manuscripts is confirmed by a final mark: a capital letter in red pencil, likely added to nearly every draft upon completion. Some manuscripts bear two alphabetical markings and, accordingly, constitute two adjacent sections. In the absence of a comprehensive structural plan for the entire work, these red letters, spanning from A to I, serve as the only indicators of the sequence of electroacoustic plans, and thus the arrangement of the sections in the piece. The last three manuscripts lack this red mark but were included in the group of final drafts by deduction, as they are the only existing detailed plans corresponding to the final part of PÉ. Therefore, the letters “J,” “K,” and “L” have been added by the author to complete the alphabetic sequence. By arranging the sketches and the fragments derived from Étude pour Espace in the correct sequence, one can discern a written plan for nearly the entire piece.

The manuscript has been known since 1958, when Jean Petit reproduced it in the volume Le Poème électronique: Le Corbusier (Petit 1958, p. 196). Subsequently, it has been featured in several publications, including the cover of Edgard Varèse und das ‘Poème électronique’: Eine Dokumentation, edited by D. A. Nanz (2006), and Peter Wever's Inside Le Corbusier's Philips Pavilion: a Multimedial Space at the 1958 Brussels World's Fair (Wever 2015). However, its content and exact relationship with the tape remained unclear owing to the difficulty in understanding it without the other missing pieces, until a new research project was developed in 2015 (Izzo 2015). Among the early studies devoted to Poème électronique, Roberta D. Lukes's dissertation has been the only attempt to establish a connection between Varèse's early sketches and the tape (Lukes 1996). However, in 1996, she had access only to the few manuscripts held in the Philips Company Archive in Eindhoven. The abundance of Varèse's sketches and drafts identified in the EVC-PSS attests to his efforts during the composition process of PÉ. At the age of 74, with no prior experience in electronic music production, Varèse was attempting to handle new instruments and confront new technical and musical challenges. From September 1957 to April 1958, Varèse worked in Eindhoven on the creation of three mono tapes—producing electronic tones; collecting sounds from various sources (such as a church bell, a church organ, and a jet fighter); and recording excerpts of his music for a percussion ensemble, a female singer, and a choir. As Tazelaar (2013) has shown, when the tape was finally completed, Buczynski extracted some excerpts from the mono tapes and inserted them into a stereophonic tape. He also added reverberation effects and distributed the sound on the left and right channels, operating with a panorama control. The two channels were then respectively mixed into channels two and three of the final three-channel master tape. Buczynski's new mixing added complexity to the final spatialization. Presumably, Philips engineers began working on the spatialization of sounds shortly after the tape was completed.

Although Varèse's music manuscripts provide detailed information about sound events, they contain minimal annotations regarding sound spatialization. Upon his arrival in Eindhoven, Varèse documented the spatial possibilities of sound projection in his notebook; his music drafts for PÉ, however, include very few annotations concerning the evolving sound distribution during the piece. The only written plan for sound spatialization was likely prepared by Philips engineers after the completion of the three tapes. Furthermore, in his recollection, engineer Willem Tak (1958/59) claimed that Varèse did not participate in the planning of sound spatialization within the pavilion, with these decisions being made by Philips engineers.

This drawing is accompanied by an explanatory legend. The speakers were arranged in two configurations: Some groups, referred to as “clusters,” served as static sources, whereas others, referred to as “routes du son,” simulated the movement of sound along a line. The clusters A, B, and C were located on the three peaks of the building, at 13, 18.5, and 20.5 meters, respectively. The clusters J and U were positioned at the entrance and exit as antiphonal sources. Additionally, other sources for the lowest sounds surrounded the audience at floor level along the perimeter of the building. Tazelaar (2013) conducted extensive research on the sound projection system inside the Philips Pavilion and almost entirely reconstructed the placement of the speakers on the walls.

However, until recently, it was not possible to determine the sequence of the different spatialization scenarios during the performance. For a long time, the only source explaining the sound distribution during the piece had been an excerpt of the diagram created by Philips engineers and published in an article by engineer Simon Leo de Bruin (1958/59) in the journal Philips Technische Rundschau, with the caption “Schematische Darstellung der registrierten Klänge” (Schematic Representation of the Recorded Sounds). This diagram, also called the “control score,” illustrates how the control tape could synchronize tape recorders, lights, loudspeakers, and the film projectors through a sequence of multiple signals. The control tape contained all the recorded sound signals necessary to operate the complex system of sound distribution and lighting. The sound track of different tones activated various devices through a system of narrow bandpass filters and relays. The control score was used to plan and develop all this information. Lombardo et al. (2009) detailed the layout of the control score, which features a timeline at the top and two horizontal stripes. In the upper stripe, all the control signals are graphically represented as horizontal lines with technical labels referring to the devices to be activated and stopped. The lower stripe of the control score does not contain technical information but serves as a reference guide to the music. Three horizontal lines represent the three audio channels, where each short episode is symbolized by a horizontal segment that defines its duration. The most interesting data in this diagram pertains to spatialization. Each line (or tape segment) is associated with a letter or a number that mostly refers to the labels in Figure 3. Unfortunately, De Bruin's article included only a fair copy of an excerpt related to a short portion of the piece, from 2:05 to 2:35. Until recently, apart from these 30 seconds, there were no other clues about the sound projection during the rest of the performance. However, a newly discovered source, originally held by the composer Wen-chung and now in the Paul Sacher Stiftung, fills this gap. It consists of an approximately 10-meter-long scroll of paper, which is evidently a complete copy of the original technical diagram created by Philips engineers for the sound spatialization and later sent to Varèse. As previously mentioned, without any other autograph document, it is difficult to determine who made the decisions regarding sound spatialization. Nevertheless, this document is an invaluable source for understanding and reconstructing the performance of PÉ, including the sound projection, as it contains detailed information on the sound sources, the routes, and the timings of each sound event.

In addition to being a valuable source of technical information, the control score reveals the challenges faced by Philips engineers in naming and describing Varèse's sounds. In the lower part of the diagram, dedicated to the three mono channels, each sound or small group of sounds is described with a label, often using unorthodox terms unfamiliar to the common lexicon of electronic music laboratories. Table 1 summarizes the content of the control score related to sound distribution. Each name refers either to a single sound or a sequence of similar sounds. For each entry, the table reports the starting moment of the sound or series of sounds, as well as the distribution of each sound within the pavilion. The capital letters and Roman numerals mostly refer to the drawing “Les routes du son” (Figure 3). The control score contains additional indications not included in this drawing. Lacking a complete legend, three types of labels are still not entirely clear: (1) The indications “Lt I,” “Lt II,” and “Lt III” (“low-tone” or “lage toon” in Dutch), not listed in the drawing of Figure 3, probably refer to the 25 low-frequency speakers distributed at the bottom of the walls behind a concrete screen, dividing them into three sub-groups. (2) The letter “N” is associated with three simultaneous control signals in the control score and may indicate a group of speakers distributed in three different places inside the pavilion. (3) The label “14 O+III,” more difficult to decipher, likely refers to a line of speakers. The team of scholars that curated a 3-D model of the Philips Pavilion and studied its amplification system for the VEP project made a hypothesis about the position of the two groups of speakers “N” and “14 O+III” inside the building (Lombardo et al. 2009). Despite these uncertainties, the overall information about the sound distribution provided by the complete control score allows us to closely understand the characteristics of the spatialized performance inside the Philips Pavilion. Table 1 also contains some additional information not present in the control score, but useful to the reader:

1. The left column indicates the corresponding draft, reporting Varèse's marks (Figure 1).

2. The gray background indicates that a sound excerpt had been processed with reverberation and distributed, after a new mixing, on channels two and three (Tazelaar 2013). In two cases, sounds are distributed only in channel two (highlighted in gray), and channel three remains silent.

3. For some Dutch words an English translation is added.

4. A black rectangle within Section D indicates the portion of the control score that was published in the Philips Technische Rundschau.

The descriptions by Philips engineers are mostly in Dutch and English, and rarely in French. Some terms refer to the recorded sound sources, such as “Piano+koor” (“piano and choir”) or “Straaljager” (“jet fighter”). In certain instances, Philips engineers adopted Varèse's own private vocabulary for describing sounds. The exact meaning of some of Varèse's terms can be difficult to decipher. For example, “Bourdon” refers to the low frequency of a church bell recorded in Delft's Oude Kerk (Old Church) by Varèse himself. “Parameter” denotes a section of the piece where electronic sustained tones overlap to generate a chord, a process reminiscent of many passages in Varèse's instrumental music. In most cases, Philips engineers preferred to rely on onomatopoeic terms, attempting to imitate the auditory effect of the corresponding tape excerpt. This process of “translation” highlights the challenge of establishing a common language between the composer and the engineers. Interestingly, in some cases, Varèse adopted more precise terms than those chosen by the Philips engineers. For a sequence of short “staccato” impulses, the composer noted quintuplets of sixteenth notes on a staff with the annotation “square waves.” Conversely, in the control score, the same sequence of sounds is simply labeled “Tok-tok” by the engineers, a Dutch onomatopoeic word also used for the “cluck” of the chicken. Varèse's “Spiral” refers to a three-note motif achieved through a sustained sliding electronic tone. This term suggests a circular motion, and in one sketch, Varèse illustrated this movement with the drawing of an ascending spiral, an idea of sound movement shared with Le Corbusier at an early stage of development. The term used by the engineers, “Wauw,” loses any spatial connotations and simply refers to a sustained sound.

Varèse's seemingly enigmatic phrase “Atakaout transmuet” is another intriguing case. It refers to a passage noted in draft G and created by recording a speaking choir and then filtering some tape loops containing voices and percussion. The word “Atakaout” results from several phonetic mutations of the verse “There is a shout,” derived from the novel The Journal of Albion Moonlight by Kenneth Patchen (1941). Varèse used this process of phonetic transformation from lyrics to glossolalia in many of his vocal compositions, referring to the resulting text as “syllables of intensity.”

The discrepancy between Varèse's own terminology and the labels given by Philips technicians challenges the common belief that Varèse could not master electronic devices and their technical language. Interestingly, many of the onomatopoeic terms chosen by Philips engineers are related to ornithology, especially when describing electronic sliding tones. Terms such as “Neus-wiep,” “Oewoep,” “Tjang-wiet,” “Tsoewiet,” “Voep-voep,” and “Wiep-wiep” are sometimes used in Dutch spoken language to describe the singing of specific birds like the sparrow, the leaf warbler, or the owl. Table 2 compares the terms used by Philips engineers with those used by Varèse in his music manuscripts.

Varèse noted where to insert the alto recording with the phrase “1st Sorceress part” written in red pencil (Figure 5, upper right corner). The mention of a “Sorceress” echoes the character of the Sorcerer or Visionary in the several scenarios written for his abandoned stage project The One-All-Alone.

An additional annotation at the bottom of the same page specifies: “sporadic p[e]rc[ussion] electronic.” This note likely served as a personal reminder for Varèse— an indication of a passage to be developed later. In the same document, the pencil annotation “goutte” describes the overall effect of the percussive “opening cadenza” (Figure 7). The electronic diagram created for section F can thus be understood as the realization, in the form of a detailed draft, of the embryonic idea of the “sporadic percussion electronic” initially conceived in the revision of Étude pour Espace.

Both these annotations, and the phrase “1st Sorceress part” in diagram a serve as a guide for the tape editing of the two audio sources (the electronic percussion sounds and the vocal performance). For these reasons, both manuscripts—the representation of electronic sounds (diagram a) and the Étude pour Espace score—should be regarded as integral components of section F, despite their differing appearances and sonic outcomes. Like the other final drafts, diagram a bears an alphabetical marking. However, the “F” marked in red pencil is partially obscured by a strip of paper, though it can be faintly discerned in the upper left corner of Figure 5.

In diagram a, Varèse notated four distinct aspects of this electronic episode. The first pertains to the superimposition of simultaneous rhythmic aggregates based on “unrelated” metronomic tempos, a recurring feature in the sketches of PÉ. The percussive entries are structured around four concurrent metronome tempos: 60, 72, 80, and 92 bpm, each written on a different “part” or horizontal line. The episode begins with only the first “part,” corresponding to 60 bpm. After four beats, this part continues in the underlying “system,” with the addition of the other three “parts” (72, 80, and 92 bpm). The continuity between the two “systems” is clarified with a connecting arrow. This polytemporal framework served as the foundation for the preparation of the draft. The four lines are used to notate individual sounds, whose durations generally correspond to the tempo indicated by the metronome markings. The second aspect addresses the timbre and morphology of each sound, capturing their temporal evolution. In most cases, each “drop” consists of a pair of sounds: a sustained noise, represented by a wavy horizontal line, followed by a short pulse, depicted as a point. The episode features a subtle process of timbral differentiation among the elements of the electronic percussive sounds. To facilitate the selection of sound sources, Varèse added a red pencil number (ranging from 1 to 10) beneath each undulating line, likely referring to a reserve of sound materials already collected. Additionally, certain timbral qualities are described verbally using adjectives such as “thin,” “cutting,” “hissing,” “soft,” and “nasal.”

The third aspect relates to the distribution of sounds across the three channels (or, more accurately, mono tapes). Compared with other manuscripts, the draft for section F contains particularly detailed information regarding sound distribution among the channels. Although this draft, like other music manuscripts, does not provide specific information about the spatialization within the Philips Pavilion, the meticulous organization of the three channels suggests preparatory work for subsequent sound distribution. The assignment of sounds to the channels occurred in two stages. First, in diagram a, Varèse assigned each element to a channel, marking it with a large pencil number. Of the three sustained noises in Figure 8, the second and third are clearly allocated to channel 2, whereas for the first sustained noise there is not a clear indication of the channel.

In this illustration, the three spectrograms derive from preparatory mono tapes held in Paul Sacher Stiftung. These tapes were made during an intermediate stage of the composition, before the final mixing with the stereophonic tape made by Buczynski. This is evident in the “sorceress” part. In the spectrograms of Figure 12, the alto voice is recorded only in channel 3, whereas the control score shows that, after Buczynski's work, it was doubled in channels 2 and 3. It is worth noting that Buczynski added some changes in the distribution of this part between channels two and three (Tazelaar 2013). Doing this, he added a further subtle sound movement to the “sorceress” part, that, in this case, oscillates between groups of speakers I l. and N. As a result of the spatialization process, in addition to the “drop” effect, the alto singer's voice is projected from the left perimeter of the building (“I l.[eft]”), and the choir were projected from the low-frequency speakers (“L.t. I”) and from one of the building's peaks (“A”).

The lower portion of the diagram presents an interpretation of the piece's overall structure. This structure can be described as an ideal tripartite form, consisting of two similar sections at the extremities—Part I (letters A, B, C, and D) and Part III (letters J, K, and L)—and a contrasting central episode, Part II (letters F and G), separated by two shorter transitions: Tr. I (letter E) and Tr. II (letters H and I).

Part I (letters A, B, C, and D) is distinguished by heterogeneous sounds, rapid editing, and the recurrence of several structural elements, including the “bourdon,” the spiral motive, sliding sustained tones, and electronic percussion. Overall, Part I exhibits a high degree of timbral diversity, sharp contrasts between adjacent elements, and a fragmented discourse. However, it concludes with a sequence of sounds that mirror its opening—electronic percussion, sliding tones, and the “bourdon” sequence—where the “bourdon” sound functions as an articulation marker signaling the end of the first section.

The subsequent Transition I, coincident with letter E and named “parameter,” introduces a more “neutral” zone of stillness, characterized by timbral and dynamic continuity. As such, it plays a significant structural role, preparing the listener for something entirely different and unexpected: the entrance of a human voice.

Part II (letters F and G), as mentioned above, primarily incorporates recorded excerpts of an alto singer, a male speaking choir, and both electronic and acoustic percussion. Although section F features greater timbral discontinuity, the following section (G) is characterized by timbral consistency and smooth, fading “loops”—a term Varèse himself used—creating a multilayered mix of heavily manipulated voices and percussion.

Transition II (sections H and I) serves as a bridge between Parts II and III. This structural placement positions the two chordal sections named “parameter” around the central Part II, which is derived from Étude pour Espace, thereby isolating it from the rest of the composition and highlighting its distinctiveness.

Part III opens with a soprano solo and continues with the fragmentary editing of brief excerpts of musique concrète sounds and electronic sliding tones. Within a short span, this section appears to encapsulate many of the sonic characteristics found throughout the entire piece. The rapid editing in section K concludes with the repetition of the spiral motive, establishing a structural endpoint just before the final 30 seconds of thunderous electronic glissandos, which function as a coda.

As previously noted, during the eight months of working on PÉ, Varèse significantly changed his notational style depending on the specific sonic effects he sought to achieve. These differences in the layout of the drafts partially support the tripartite segmentation proposed here.

Drafts for Parts I and II, characterized by rapid editing of short, heterogeneous sounds from various sources, incorporate multiple types of notation. Episodes involving short electronic percussion or impulses are notated on handwritten staff segments for percussion made of one, three, or five lines. Sustained sounds—such as electronic sliding tones or musique concrète recorded noises—are represented using various graphic elements, including curves and straight lines of different shapes. Occasionally, Varèse indicated longer recorded sounds (such as “jet”) adding a name as a reference and without any other graphic sign. Additional descriptions and adjectives usually specify the timbre of individual sounds.

In contrast, Varèse's drafts for the transitional “parameter” sections, such as E and I, are markedly different. Both are written as conventional scores, each on a single page of staff paper. These drafts are essentially scores for electronic tones, with each staff assigned to a different sound generator. The traditional layout of these drafts led to significant implications for subsequent research. Because Varèse was not a systematic archivist of his own documents, many unidentified manuscripts in his collection have yet to be fully analyzed. For a long time, drafts marked with letters E and I were overlooked because they resemble orchestral scores and their relation to electronic music is not evident.

Part II, situated at the heart of the composition, is derived from a combination of three distinct manuscripts: (1) a draft for a sequence of electronic percussive sounds (partially reproduced in Figure 5); (2) excerpts from the revised score of Étude pour Espace (Et.58, partially reproduced in Figure 7) for alto, choir, and percussion (both are parts of F section); and (3) a large editing diagram based on a rhythmic part for speaking choir (letter G). On the one hand, these three manuscripts had been written with entirely different notational styles and appearances. On the other hand, they were conceived as a coherent set, as they all originated from Varèse's long-standing project: his ambitious choral symphony, Espace, and its subsequent revisions.

As already illustrated, the music material for section F derives almost entirely from Étude pour Espace. The correlation between the following section (G) and Étude pour Espace is not evident simply by listening. The entirety of section G is based on a multilayered mixing, referred to as “procession,” which includes filtered speaking voices, hardly recognizable, played at different tape speeds and a persistent sound effect that imitates the noise of footsteps on gravel. Only through the philological study of the draft of section G is it possible to recognize that the vocal and percussion parts included in this section are derived from selected fragments of Étude pour Espace (Izzo 2015).

From these observations, it appears clear that the entirety of Part II is based on vocal and percussion parts drawn from Étude pour Espace, and that this central part plays a crucial role in the overall conception of the work. Situated between two transitional sections of electronic chords that evoke stillness, its centrality is further emphasized by a brief silence that Varèse inserted precisely at the center of the composition, at second 240. Varèse referred to this vocal section as “sorcerer,” “sorcier,” or “sorceress,” alluding to the Visionary character from The One-All-Alone. The choir's spoken text, “Atakaout,” even if derived from a phrase by Kenneth Patchen, echoes the incantatory words that the Visionary delivered to the masses in Carpentier's scenario.

In general terms, the evidence of sound distribution from the control score corroborates the structural articulation outlined in the analysis of the sketches. The control score clearly indicates that spatialization was used to accentuate contrasts between successive sections and to highlight similarities between recurring sounds or processes. Three primary principles govern the relationship between the tape content and its spatialization: (1) the consistency of motives; (2) the contrast between continuity and discontinuity; and (3) the enhancement of dynamic variations.

Motivic sounds were consistently played from static sources. For example, the “bourdon” sounds surround the audience through the low-frequency line of speakers (Lt I), concealed behind a concrete screen, whereas the spiral motive typically emanates from speaker clusters N and U, with additional sources introduced during the final statement of the motive, including the low-frequency speakers Lt II and cluster C, positioned at the highest point of the building. Similarly, electronic impulses or short sounds, referred to as “percussion,” move along the building's perimeter (route I). In contrast, sliding tones, which often reach high frequencies, are routed along the edges of the curved walls or overhead. For instance, the two sliding tones at 0:21 (“Oewoep” and “Tsoewiet”) follow distinct trajectories: route II (from the floor to the A peak) and route O (horizontally above the audience from peak B to point D, then vertically to peak C).

As a general principle, during sections featuring high and short sounds, discontinuity and fragmented editing, each sound emanates from a different point in the pavilion, creating rapidly shifting spatial scenarios. Conversely, sections marked by continuity and gradual transitions are played from static speaker groups. These complementary principles are particularly evident when comparing Part I to Transition I. In section D (2:00 to 2:40; see Figure 2 for a reproduction of the draft), the densely edited, morphologically diverse noises are spatialized using more than twenty changes of sound sources (trajectories and speaker clusters) within just 40 seconds. When this section was played in the Philips Pavilion, the listening experience for the audience must have been disorienting, with sounds coming from scattered and rapidly changing locations. In contrast, section E (2:40 to 3:25) features overlapping sustained electronic tones, with sound distributed only to the three peaks of the building (points A, B, and C) and the low-frequency speakers surrounding the audience (Lt I, II, and III). The resulting chordal structure changes smoothly, creating an immersive, enveloping experience.

The spatialization of Part II also emphasizes continuity and coherence, with voices and percussion primarily emanating from static sources. The whole section G, for instance, is made of two superimposed layers: On one level there is a complex mixing of lower-range voices, percussion, and “steps” noises. This is played from speakers around the perimeter of the building, enveloping the audience. On the top of this layer, occasional loops of higher, filtered voices are played from different sources (routes I and II and group of speakers A), with slow fading-in and fading-out.

In instances where Varèse indicates a high dynamic level, such as the fortissimo “Yell of the Choir” in section F, a single channel may be distributed across multiple speaker groups. In this case, the choir's voices are projected through clusters U and C, as well as along route O, further enhancing the spatial and dynamic impact.

In the near-total absence of documentary information, analytical studies dedicated to Poème électronique have traditionally relied primarily on listening. Despite the varied methodologies and conclusions regarding the structural interpretation of the piece, the analytical contributions converge on a common observation: One of the distinctive characteristics of Varèse's PÉ lies in the extreme differentiation of sound materials, often juxtaposed in opposition. Lawrence Ferrara described the structure of PÉ as a “nondiscursive form” (Ferrara 1984, p. 370), and Robert Cogan (1991) explained it through his idea of an “oppositional poetic.” According to Cogan, in PÉ this structural logic emerges through the juxtaposition of contrasting materials:

Francesco Giomi and Marco Ligabue, similarly to Cogan, described the music of PÉ with the term “poetica del contrasto” [“poetic of contrast”], but they saw this characteristic as a weak aspect, and, in their opinion, the composition has “elementi di relativa incoerenza” [“elements of relative incoherence”] (Giomi and Ligabue 1998, p. 25). More recently, Gary S. Kendall (2006), building on David Cope (1984) and Cogan (1991), described PÉ as a composition based on a “non-linear” and fragmentary discursive process. According to Kendall, the succession of unrelated elements is not perceived as contrast (an intensification within a unified narrative process) but rather produces a sense of fracture and discontinuity. Kendall defines this compositional logic as “juxtaposition and non-motion” and traces its origins to the modernist currents of the early twentieth century, in both the visual arts and music.

More recently, Kees Tazelaar published an interpretation of the form of PÉ considering the different sound sources available: the three monophonic tapes and the stereophonic working tape with reverberation effects. He acknowledges the central role of Étude pour Espace in the composition of PÉ and argues that this long citation dominates the center of the structure, with a “perceptible symmetry in the music's overall form” (Tazelaar 2017, p. 178). Nevertheless, Tazelaar notes that “one of the distinctive things about the sound material used in Poème is its tremendous variety” (Tazelaar 2017, p. 178). The resulting contrasts generate “individual sound objects, so that there is little musical continuity in the composition,” a characteristic it shares with Le Corbusier's scenario of images and lights.

Without disputing the principle of fragmentation, which most of these contributions identify as fundamental to PÉ, it is possible to discern an internal articulation in Varèse's piece when considering the documents. There are sections of PÉ where the fragmentation and heterogeneity of materials are more pronounced, and others where the composer appears to have followed a different logic. As explained, the central part of PÉ exhibits greater continuity, which can be fully appreciated by examining the manuscripts and the spatialization outlined in the control score. The analysis of the drafts reveals that, in preparing this part of the piece, Varèse drew upon ideas and musical materials from Étude pour Espace, which are, in turn, connected—albeit tenuously—to the initial conception of Espace, Varèse's early vision of spatialized sound.

The lack of information about PÉ has long been an impediment to its interpretation. However, documents from the Varèse archive reveal a complex, multifaceted, and intriguing development process, with extramusical references yet to be fully explored. The study of the work's genesis, through the examination of manuscripts and preparatory sounds, allows for a closer understanding of the composer's intentions by tracing the layers of revisions and variants. The absence of a systematic method in the piece's development can be disorienting, but it also reflects Varèse's perspective and, ultimately, constitutes one of the most fascinating aspects of his poetics. As he stated with his characteristic sharpness in a letter to Odile Vivier:

Investigating these documents to gain a better understanding of the composer's thought is perhaps the best way to avoid the risk of interpreting PÉ through “arbitrary systems” entirely foreign to the personality and poetics of the composer. The discovery of new sources has inaugurated a new phase in the study of Varèse's compositional thought. Furthermore, these sources offer the possibility of performing PÉ in its original form, grounding the performance practice of electronic music in philological research.

This article is an extended elaboration of a paper presented in April 2024 at the international conference TENOR Zurich 2024, where it was awarded with the Best Paper Award “for exceptional research contribution.” I am thankful to the TENOR Organizing Committee at the Institute for Computer and Sound Technology, and to the chair Prof. Germán Toro Pérez. I am also grateful to Paul Sacher Stiftung for the support in this long-term research and for making the reproductions of the documents available. I would also like to thank Kees Tazelar for his numerous suggestions during the revision of this article.