Roger Reynolds’ Thoughts, Places, Dreams:
genetic and sonic approaches1

Philippe Lalitte

Thoughts, Places, Dreams for cello and chamber orchestra (2012-2013) is a particularly fascinating work not only because of its virtuosi writing for the cello – one of the most demanding pieces of the contemporary repertoire for this instrument – but also because of the complex interrelations between the soloist and the ensemble. The composer imagined a formal design where different blocks of textures played by the ensemble contextualize the soloist’s positings and ruminations. The work is based on two short pieces – imagE/cello and imAge/cello2 – composed for the cellist Alexis Descharmes. Although it is a familiar practice in contemporary music to “reuse” and to develop the material of a soloist piece in a new piece for solo instrument and ensemble (one of the most famous example being Berio’s Chemins series3), Thoughts, Places, Dreams is not a mere orchestration around a soloist part. Reynolds composed a specific design, which grew up from the materials of imagE/cello and imAge/cello pieces, but also took advantage of their expressive potentials in the writing of the orchestral parts.

The goal of this paper is not to give an exhaustive analysis of the work, which would require long developments, but to approach the work both from a genetic and a sonic perspective. First, I will show how the concept of “image” becomes a key-concept in imagE/cello, imAge/cello, and Thoughts, Places, Dreams, as in numerous recent works of the composer. Secondly, I will examine the process of composition using the sketches and notes available in the composer’s notebooks. In particular, I will focus on the elaboration of the temporal proportions and the organization of the layered events. Finally, I will investigate the sonic shape of the work through several analyses based on the recorded performance by Alexis Descharmes (cello) and the Ensemble Court-Circuit conducted by Jean Deroyer4. After dividing the score into 23 audio segments, the duration, the tempo, the texture and the “tension-release” curve will be analysed.

1. From the “image” concept to Thoughts, Places, Dreams

Process and Passion for violin, cello, computer processed sound, and real-time computer spatialization (2002) represented a turn in the way Reynolds composes. Since then, he has been exploring a more flexible approach based on “sonic images” (in the composer’s own words). The work was built on twenty-one musical images both in the sense of sonic images which acted as motivating resources for the whole and “dramaturgic” images themselves characterizing some emotional manifestations such as evil, blood, fear, black, heart, death, weight, and love. The compositional strategy experimented in Process and Passion was carried further with The Image Machine (2005), a real-time interactive computer music, in which some sound material of 22 (2004-2005) provides seeds for the seven computer generated sonic images.

At around the same time, his London agent Graham Hayter suggested he writes small-scale pieces that performers could more easily program. Then, the composer started thinking about the possibility of composing a set of solos lasting 4 to 7 minutes each”5. imagE/cello and imAge/cello (2007) where the first pair of the solo piece series (still in progress), each being dedicated to a performer. Reynolds then composed 6 other pairs: imagE/pianoimAge/piano (Eric Huebner, 2007-2008), imagE/contrabassimAge/contrabass (Marc Dresser, 2008-2010), imagE/guitarimAge/guitar (Pablo Gomez Camo, 2009) imagE/fluteimAge/flute (Rachel Beetz, 2012-2014), imagE/violaimAge/viola (John Pickford Richards, 2012-2014), imagE/violinimAge/violin (Irvin Arditti, 2015).

According to Reynolds (2015): “The imagE/… and imAge/… series explores – respectively – evocative and articulate ideals. […] The evocative studies are gentle, atmospheric and symmetrical while those of articulate bent are sectional, forceful and asymmetrical. Each pair explores characteristic aspects of its instrumental medium in a fashion that may be reserved or acrobatic”. Thus, the imAgE concept concerns both opposed expressive characteristics such as gentle, atmospheric, relaxed, reserved vs tense, forceful, assertive, acrobatic and opposed types of writing such as progressive, symmetric, linear, ornamental vs sectional, asymmetric, harmonic, dense. Speaking of the pair of imagE/pianoimAge/piano, the pianist Yuji Takahashi said (2009): “The composing technique is based on montage and editing. It may be too soon to summarize the implications of this project, but we might say it aims to represent a refracted double image: the done and doing”.

We will take the example imAgE/cello in order to see how these ideals take shape in the writing for cello. These musical characteristics will have refractions in the composition of Thoughts, Places, Dreams. imagE/cello (circa 4’) is composed around several axial pitches (with long duration or repetition) on which some ornamentals take place (grace notes and harmonics). High pitches alternate with low pitches in order to create a virtual polyphony or the illusion of a violin/cello duet. The global articulation is legato. The piece is stable both in terms of meter (5/4 meter all along) and tempo (54/60 beats per minute) avoiding ritardando or accelerando. All these characteristics give the piece a sense of lightness and softness, only tarnished at the end when the cello explores darker sonorities. On the opposite, imAge/cello (circa 3’) is constructed around forcefully strummed chords alternately played arco and pizzicato. They are interrupted by silences, microtonal wails, harmonic on held notes, or fast sequences of notes often played staccatissimo. Numerous meter change mark out the piece, while some fermata, ritardando or accelerando intervenes punctually to break the flow of the basic tempo (80 beats per minute). Some intervals (e.g. G/C, E/E-flat, B/E-flat), and some chords (e.g. B-flat/F/B/A-flat, G-flat/D/E-flat/E) reappear but in an unpredictable way. Here, while listening to the piece, neither regularity nor stability in terms of harmony, rhythm, dynamics and timbre can be detected. imAge/cello could depict the dark, irascible, undecided side of the same character. Lastly, note that both the two pieces are extremely virtuosi, summoning all the resources of the cello in terms of tessitura (C2-G6) and extended techniques6.

Thoughts, Places, Dreams finds its roots as well in imagE/cello and imAge/cello, as in several collaborations with the Ensemble Court-Circuit under the artistic Directorship of composer Philippe Hurel7. Alexis Descharmes, member of the Ensemble, asked the composer to contribute to his own project 30 years / 30 creations. Reynolds wrote for him a pair of pieces – imagE/cello and imAge/cello – premiered by Descharmes on March 3, 2007 at the Cité de la Musique (Paris), then during the Rencontres d’Ensembles de Violoncelles (Beauvais) on May 14, 2007. The cellist gave the American premiere at La Maison Française (Washington, DC) on March 7, 2010.

In 20128, the Ensemble Court-Circuit commissioned a new work for cello and chamber orchestra. Then the composer decided that imagE/cello and imAge/cello would serve as genetic source for a new piece entitled Thoughts, Places, Dreams. The composer (Reynolds, 2015) invented “a range of new, but derived musical behaviors and arrayed them in irregular reappearances across a 24-minute span”. This new material, called by Reynolds “Thoughts”, interspersed with quotations from imagE/cello and imAge/cello, constituted the almost uninterrupted cello part.

During the Summer 2013, Descharmes make a demo recording of the entire cello part of Thoughts, Places, Dreams that he sent to the composer. After listening to the recording, Reynolds noted, in his notebook (21/08/2013, p. 62), some remarks showing that he took into account some of the particularities of the cellist’ playing style when writing the score of the ensemble. Few changes occurred in the final cello part of Thoughts, Places, Dreams. Some of them are listed below:

  • Dynamics adjustment: most often by increasing the level (less ppp than in imAgE pieces); bar 20: a decrescendo instead of a crescendo; bar 352: a crescendo instead of a decrescendo
  • A few textual indications supressed (e.g. bar 81 “microtonal wail”)
  • Pitch changes: bar 8: B-flat replaced by F; bar 198: A replaced by B
  • Inserting notes: bars 82-87: group of notes (from “Thoughts”, bars 48-58) instead of silences; bars 197-200: “echoes” added (repeated notes)
  • Meter changes: bars 197-210: the quote of imAge/cello is written in 3/4 instead of 4/4; bars 201-202: 2 bars added to extend the “microtonal wail”
  • Tempo changes: bar 20: “ritardando” added; bars 300-315: 72 BPM instead of 80; bar 360: “senza tempo” added

Some of these changes appear to be the result of Descharmes’ collaboration with Reynolds. The change of tempo bars 300-315 for the quote of imAge/cello (72 BPM instead of 80) is probably the consequence of the difficulty to respect the tempo with such thirty-note sequences. The change of dynamics bar 352 (crescendo instead od decrescendo) was noted Descharmes in his own score of imagE/cello in 2013 when working with the composer (Fig. 1)9.

Then Reynolds composed the ensemble parts – the “Places” and the “Dreams” –, which we will analyze later on. According to the composer (2015), the work is “a sonic mosaic, a successive of distinctive though sometimes overlapping domains”: the soloist’s (almost) uninterrupted performative journey, the “Places”, and the “Dreams” both provided by the ensemble. Clearly the piece gives a predominant role to the soloist. But, its omnipresence pulls the work towards a monologue rather than a concerto (in the traditional sense of a dialogue between a soloist and an orchestra). The cello part evokes an inner monologue – a stream of consciousness – the content of which being influenced by mental images, feelings and emotions represented metaphorically by the “Places” and the “Dreams”. The few passages where the cello plays alone10 never function as a typical concerto cadenza. While they are almost exclusively composed by the quotes of imAgE/cello (except bars 316-319), their function is comparable to a refocusing of the discourse on the original statements after some “ruminations” (the “Thoughts”). They bring back the listener to the primal evocative vs assertive opposition of the imAgE project, and thus lead to recalls in memory what was previously heard.

Alexis Descharmes and the Ensemble Court-Circuit, conducted by Jean Deroyer, premiered Thoughts, Places, Dreams in the Teatro Piccolo Arsenale, for the Venice Biennale (57th yearly contemporary music festival altra voce, altro spazio), on October 13, 2013. The same performers recorded the piece at the Auditorium Marcel Landowski, Conservatoire à rayonnement régional de Paris, on November 15,2013.

Figure 1. Excerpt of Alexis Descharmes’ own score of imagE/cello.

2. Genetic approach: the compositional process

Since the 1970’s, Reynolds has progressively adopted a specific musical approach with intensive control over all aspects of composition. Reynolds’ compositional process is methodical and distinctive (Boyd, 2008; Lalitte, 2005; Reynolds, 1987, 2002). For each work, he follows a highly formalized procedure that moves from a global to a local perspective. Nevertheless, the different methods used have a significant flexibility, giving every piece a sense of individuality while allowing the composer to explore new ideas and concepts, creating a body of music simultaneously diverse and cohesive. This compositional process can be observed in the numerous documents – notebooks and sketches – left by Reynolds when composing a piece. The composer also described it in several texts (Reynolds, 1987, 2002, 2007).

However, for more than a decade,Reynolds has express the need for exploring more flexible approaches based on “sonic images” rather than core elements11. Process and Passion (2002) marked a significant methodological turn for him (Reynolds, 2015):

So I sought a starting point that was less specific in its quantitative detail, more emotionally, timbrally centered. I wrote out rather schematic “sonic dramaturgies” and worked with my then San Diego-based collaborators (violinist Mark Menzies and cellist Hugh Livingston) as a film or theater director might, urging their realizations in one direction or another – their common touchstone was found in the notation’s harmonies, gestures, and sequences – as events in the moment suggested.

When the Ensemble Court-Circuit commissioned a new work for cello and ensemble to Roger Reynolds, the composer imagined that he could recycle the pair of imAgE/cello. The compositional process began on January 3, 2012. The composer wrote in the notebook of Thoughts, Places, Dream: “Thought is to build on image/cello works written for Alexis. Somewhat in the manner of MARKED MUSIC & Dream Mirror with the ensemble playing the role of the real time computer materials”12. Early in the process, the instrumentation of the ensemble was decided: Cello, Flute, Bass Clarinet, Horn in F, Trombone, Percussion, Piano, and Violin. The fundamental idea/strategy was specified on January 8:

Whole is a montage of (sometimes) overlapping, staged ‘trends/processes’ that proportionally allow for OR are interrupted by

IDEALIZATIONS of very strong character (normally) sharply contrasting with surroundings.

ENTITIES are brief ‘components’ of trends/processes used by soloist.

FIELDS are extensive areas of strongly characterized textures/behaviors by the ensemble’s components elements.

These three ideas mark, frame, support the whole, but they do not constitutes the whole: they are features of/in the musical continuity13

Reynolds began by composing the entire cello part first. He segmented imagE/-cello in three globally equal sections (a: 1-20, b: 21-35, c: 35-52), the frontier between the first and the second being delineated by an ascending sequence of notes (b. 19-20), while those between the second and the third being separated by a descending sequence of notes (bars 35-37). imAge/cello was divided in four unequal parts, the frontiers being marked each time by a strummed chord (a: 1-17, b: 18-42, c: 43-58, d: 58-66). Theses sections constituted the seven quotations of imAgE/cello which served as seeds for the work. However, the composer did not feel that only segments borrowed from the two source solos could support a work of the length he had planned. Thus, on January 19, he began to elaborate nine “Entities” (later on called “Thoughts”): “1.Disjunct gestures”, “2. Multi-segment glisses”, “3. Converging (lines)”, “4. Diverging (lines)”, “5. Axis-straddling”, “6. Trend anchor”, “7. Diad succession”, “8. Common-tone succession”, and “9. State changes”. Before composing them, each had been specified in terms of number of segments, register profile, dynamic profile, trends (none, compression, expansion, cyclical), texture (staccato, sustained, irregular/intermittent, alternation, slurred), axes (stable, stopped, reset), and continuities (conjunct, disjunct, interrupted)14. It would take too long to examine in details all theses entities, so I would just choose the third and the fourth as examples (Fig. 2).

“Converging” draws an envelop from a wide range to narrow intervals in the low register around an axis which define the point of convergence over time. Pizz. and arco alternate in a soft dynamic (sotto voce). “Diverging” begins in the low register with a narrow distribution around an axis, and then progressively expands toward a maximum range, ending in the high register. The trajectory is also led by the move from glissandi in high dynamics to soft stable tones and harmonics. In fact, all “Entities” borrow and develop one or more specificities of the imAgE/cello seeds. Reynolds planned to use them in the work as reoccurring and varied elements along the cello part, providing both a sense of familiarity and novelty. Thus, the line formed by imAgE/cello quotes and “thoughts” acts as the thread of a monologue constantly inflected by the imaginary landscapes crossed.

Figure 2. Sketches of two “Entities” : “Converging” (top) and “Diverging” (bottom).

In order to contextualize the soloist’s ‘positings’ and ‘thoughts’, Reynolds specified in his notebooks two types of sonic textures – “Fields” (later called “Places”) and “Idealizations” (later called “Dreams”) – intended to be played by the ensemble. Each of them was thought with a distinct identity in terms of harmonic content, register, rhythm, texture, temporal shape, and duration. However, the composer did not give them the same function in their relationship to the soloist. In the composer’s mind, the seven “Fields” – “Pulsing”, “Diad slurs”, “Brief inflections”, “Sustains with breakouts”, “Rhythmic bounce”, “Interjecting rips”, “Cross-fade harmonies” – should simulate the kind of processing operated by some of the four computer algorithms he used in his recent music15. For example, in “Pulsing”, the writing simulates the action of MATRIX algorithm, which generates complex rhythmic patterns based on a temporal fracturing and patterned and expanded replication of source material, though it manifests the original morphology of the source sample. The sketch (Fig. 3) shows that the texture is composed of nine layers rhythmically independent constituted by repeated patterns on one pitch dispatched on the seven instruments (the piano and the percussion each have two layers). The number of repetitions and the duration of the tones are specified for each layer. For example, the left hand pattern of the piano occurs three times with a 182-unities duration (the unity corresponds to a sixteenth note), while the right hand pattern occurs 34 times with a 21-unities duration. Harmonically, “Pulsing” is structured by three chords, the first one (D, E-flat, B, G, B-flat, E, F, G-flat, D-flat) with a 240-unities duration (30 s), the second one (F-sharp, E-flat, G, C-sharp, A-flat, C, A, B-flat, E) with a 248-unities duration (31 s), the last one, with a 248-unities duration (30 s), and keeping 5 pitches of the first (G, B-flat, E, F, G-flat).

Figure 3. Sketch of “Field” 1: “Pulsing”.

The four “Idealizations” (later called “Dreams”) – “Freeze”, “Argument”, “Intercutting”, and “Alignment” were thought as areas of particularity musical behaviours. A static harmony, trills, tremolos and patterns repeated freely characterized “Freeze”. “Argument” was made of chaotic gestures with irregular spacing, and violent dynamics. Several gestures distributed alternatingly to the seven instruments spread during “Intercutting”. “Alignment” included synchronised chords and fast gestures. Figure 4 shows an early sketch of “Freeze”. The whole sequence is composed of a static harmony, a nine-pitch chord (F-sharp/E-flat/G/C-sharp/A-flat/C/A/A-sharp/E). In fact, this is the same chord as the second one in “Pulsing”, showing the inner link between both passages. The pitches are distributed on the seven instruments by shorts cells16, which are permuted in time through to four layers (Horn/Tbn., Vl./Fl., Pno., Perc.) having their own temporality (represented at the bottom of the page).

Figure 4. Early sketch of “Freeze” from the composer’s notebook (p. 67).

We will now look at how Reynolds assembles these elements to form the entire work. Generally, Reynolds begins by drawing lines on a semi-log paper for generating series of numbers that serve as a source for the temporal proportions17. For Thoughts, Places, Dream, he reused the original log series of imAgE/cello in order to specify two sets of numbers. The first one for the solo part consists of 7 terms (imagE: 1, 2, 51; imAge: 5, 9, 17, 30) and the second one for the ensemble part consists of 11 terms (“Places”: 1, 2, 7, 19, 29, 40, 51; “Dreams”: 4, 11, 24, 34). Three numbers are common to the two sets: 1, 2, and 51. Figure 5 shows that the duration of the whole piece was planned to be about 20 minutes.

Figure 5. Log series (notebook, p. 29).

The two sets of numbers were used to define both the duration and the position of the events. A first draft of the plan, drawn by the composer (notebook p. 42), shows that the planned total duration (1186 unities)18, the number of blocks (7 “Solos”, 7 “Places”, 4 “Dreams”), their name, duration and position (Figure 6). The proportions indicated inside the boxes of the solos (102, 70, 80, 33, 48, 90) signal that the order of the quotes drawn from the pair of cello pieces was decided at this state of the compositional process. The order of the quotes does not follow the original order. Thus, imagE/Cello-c (solo 7) is separated from the two other imagE/Cello (solos 1 & 2), and displaced at the end of the work. imAge/cello-c (solo 6) and imAge/cello-d (solo 5) are inverted, the first one ending the succession of imAges. The principle that no overlap exists between the layers “Idealizations” (latter called “Dreams”) and the other blocks was also established at this stage of the compositional process.

Figure 6. Early draft of the plan.

After several adjustments, the final plan was established, probably mid January 2012 (Fig. 7). The diagram in the bottom shows the process which led to elaborate the durations and positions of the blocks. Seventeen points were determined by adding some of the numbers (or addition of these numbers) of the set (4, 9, 29, 40, 51, 80, 91, 120): 4, 51, 102, 142, 193, 273, 353, 433, 484, 524, 644, 739, 826, 946, 1066, 1186. These numbers correspond to most of the beginning or ending of the blocks. The other numbers not belonging to this set (212, 375, 484, 554, 677, 994, 1095, 1157) come from the durations or sub-divisions of some blocks. Below the numbers series, at the bottom of the final plan, incremented lines and their denominations under a synthesized representation of the three layers represent the occurrences of the nine “Thoughts” having their own durations.

Figure 7. Final plan of Thoughts, Places, Dreams.

In the final plan, most of durations and positions were adjusted. Concerning the positions, it seems that some blocks were slightly pushed forward so that the concentration of events would be less important at the beginning of the piece and therefore better distributed. It is worth noting that the composer also decided that the solos would always begin without overlapping with other blocks in order to make their appearance clearly perceptible (which was not the case for solos 2 and 7 in the early plans). The duration of the blocks was also adjusted. The composer expanded “Fields” 3, 5, and 6 (respectively from 40, 51 and 51 to 80, 58 and 120) and “Idealizations” 2, (from 40 to 51), while he shrank imAge-a and b (respectively from 80 and 91 to 51 and 70) and “Idealizations 3 and 4 (respectively from 120 to 91 and 72). Therefore, it seems clear that Reynolds wished to give more time to the “Fields” to the detriment of the “Idealizations”. So, the total duration of “Fields” goes from 524 unities to 640, while those of “Idealizations” goes from 331 to 265. This change in duration probably affects the formal design, reinforcing the roles dedicated respectively to the “Idealizations” and the “Fields” as previously mentioned.

The occurrences of the “Thoughts” are drawn in the bottom of the final plan. They are represented by seven segmented lines the first appearance being indicated by a number: “Diad succession” (104), “Disjunct gestures” (212), “Multi-segment glisses” (233), “Common tone succession” (418), “Converging” (575), “Axis-straddling” (588), “Diverging” (614), “State changes” (664) and “Trend anchor” (759). The short vertical stripes give the number of occurrences for each “Thought” and, show their approximate position. Some “Thoughts” have only few occurrences (i.e. 3 occurrences for “Diad succession”: bars 22-24, 26-28, 30-32), while some others reappear up to 9 times on a long duration (i.e. “Trend anchor”: bars 238-243, 268-271, 295-298, 316-318, 324-327, 332-333, 338-339, 344-346, 347-350). Some “Thoughts” occur at equal time-intervals, while others have time-intervals increasing or decreasing. Their appearances are often varied and truncated. For example, “Trend anchor” appears in its full and original version only in its fourth occurrence. It is worth noting that in the plan several “thoughts” have their lines superimposed. However, in the score they are intermingled with each other, sometimes very quickly such as in the third “Dream” (“Intercutting”) where no less than 6 different “Thoughts” follow: 6 (b. 238-240), 5 (b. 2402-43), 4 (b. 243-246), 9 (247-248), 4 (b. 249-254, 254-257), 5 (b. 257-260). That leads to continuous reappearances of fragments already heard, but more or less transformed, continuous rotation of extended cello techniques, and continuous expressive changes in the cellist’s part.

3. Sonic approach: duration, tempo, and tension/release curve

The different methodological approaches to controlling musical events over time used by Reynolds are not an end in themselves, but a means. It is not only a means to achieve a formal coherence, but also to create continuities and contrasts, tension and release in the musical form. In order to analyse the piece from a sonic approach, we have chosen to divide the work into 23 audio segments (named from a to w) each of them having their own acoustic, and perceived particularities. The boundaries of each segment correspond to the beginnings and ends of the blocks designed by Reynolds. For example, the first part of the block imagE-a (bars 1-9) constitutes the a-segment. The first part of the block “Pulsing”, superimposed to the second part of imagE-a (bars 9-20), constitutes the b-segment, the second part of the block “Pulsing” (bars 21-32) constitutes the c-segment, the first part of imagE-bconstitutes the d-segment, and so on. So, we have four types of segment distributed in the work (Fig. 8-11):

  1. “Solos-imAgE” (IS): 3 segments unaccompanied imagE (a, d, w) and 4 segments unaccompanied imAge (h, l, o, s)
  2. “Places+Image” (PI): 5 segments superimposing “Places” with imagE or imAge (b “Pulsing”, e “Diad slurs”, i “Inflections”, m “Sustains”, v “Crossfade”)
  3. “Places+Thoughts” (PT): 7 segments superimposing “Places” with “Thoughts” (c “Pulsing”, f “Diad slurs”, j “Inflections”, n “Sustains”, p “Bounce”, r “Rips”, u “Crossfade”)
  4. “Dreams+Thoughts” (DT): 4 segments superimposing “Dreams” with “Thoughts” (g “Freeze”, k “Argument”, q “Intercutting”, t “Alignement”)
Figure 8. Excerpt of a-segment (Solo-imagE), score p. 1 (Peters Edition, EP 68443).
Figure 9. Excerpt of m-segment (Places+ImAge: “Sustains”), score p. 25 (Peters Edition, EP 68443).
Figure 10. Excerpt of c-segment (Places+Thoughts: “Pulsing”), score p.5 (Peters Edition, EP 68443).
Figure 11. Excerpt of k-segment (Dreams+Thoughts: “Argument”), score p. 22) Peters Edition, EP 68443).

In order to see how the temporal proportions designed by Reynolds are embodied in the performance, will analyzed the duration of each segment calculated from the Descharmes, Deroyer and Ensemble Court-Circuit’s recording (Fig. 12). Overall, the segment durations (indicated inside the histogram bars) correspond to the temporal proportions of the plan. There are some deviations in duration from the plan, which we will examine further on with the tempo. The segment durations show a slight tendency in increasing, the longest being placed in the second half of the piece. The total duration of the performance is 1430 s. The segment durations extends from 23.46 s to 140.06 s, with a mean of 62.17 s (SD = 23.56). Globally, the shortest segments are those where the quotations of imAgE/cello occur (M = 50.05 s, SD = 17.24), while the average durations of “Places” and “Dreams” are almost equal (respectively M = 67.01 s, SD = 35.15 and M = 66.87 s, SD = 11.29). Yet, the longest average duration comes from the “Thoughts” (M = 75.40 s, SD = 29.27). The major part of the “Thoughts” appears in the second half of the piece (first half: M = 60.35 s, SD = 10.44; second half: M = 87.94 s, SD = 34.79). It means that the “Thoughts” derived from the imAgE/cello pieces invade more and more the cello part, and particularly during the “Interjecting rips” segment.

Figure 12. Duration of the 23 segments calculated from Descharmes, Deroyer and Ensemble Court-Circuit’s recording.

It is interesting to compare the theoretical tempo (indicated in the score) to Descharmes, Deroyer and Ensemble Court-Circuit’s recorded performance. Figure 13 shows the average tempo for each segment and the theoretical tempo (the curve)19. Globally, the performers’ tempi are under those of the score (respectively M = 62.19 and 68.26). Such a difference is not unusual, with such technically difficult piece. In particular, when the performers play knowing that they are being recorded (in the studio or live), they tend to take fewer risks and therefore choose slightly slower tempi in order to ensure the cleanest possible technical realization. Furthermore, the rhythmic notation of the score leaves a certain freedom of tempo to the performers. Alexis Descharmes confirmed this to us: “my performance in the recording is not metronomic. Jean Deroyer followed me and managed to understand my tempo fluctuations”20.

Nevertheless, this difference in tempo between the score and the performance is not dispatched uniformly through the 23 segments. The tempo tends to be slower when the soloist plays unaccompanied than when he plays with the ensemble. It is not surprising because, without the chamber orchestra, the cellist has no longer the constraint of the synchronization. He gets more timing freedom, hence generally taking more time for some particular events. This is particularly noticeable for the imAge- (h-segment), imAge-d (o-segment), and imagE-c (w-segment). A clear differential of tempo score/performance appears also between imagE and imAge solos, the last one having a slower tempo (respectively M differential = 4.18, SD = 8.18; M differential = 13. 03, SD 12.42). Surprisingly, the cellist takes more time with the assertive images (when he plays without the ensemble) even so the contrary might have been expected. One particularly striking example is the tempo adopted by the cellist in the first part of imAge-a (h-segment), which is well under the tempo of the score (57 BPM instead of 80). Sometimes, the decrease of tempo is due to an event such as the wailings in bar 201 which the performer significantly stretches out for an expressive goal. Now, if we compare the tempi of the “imAgEs” (with and without the ensemble), to those of the “Places” and the “Dreams”, it appears that the latter ones has almost always a tempo slightly quicker than the score (respectively M = 7.82, SD = 9.70; M = 6.04, SD = 6.28; M = -0.73, SD 4.77). Thus, it seems that the writing of the “Dreams” (principally k-segment “Argument” and t-segment “Alignment”) tends to lead the performers to increase the tempo (at least during some passages) possibly to reinforce the expressive effect.

Figure 13. Average tempo for each segments compared to the theoretical tempi (score).

The succession of the 23 segments also varies in texture density because some of them are filled by a single instrumental part (when the cello plays unaccompanied: a, d, h, l, o, s, w), while other contain multiple instrumental parts. Obviously, they also vary in pitch and rhythm content, dynamics, timbre, etc. In order to see how these changes affect the acoustic content of the segments, and thus presumably the sonic shape and its perception, we performed an analysis of the acoustic features from the performance audio recording with MIR Toolbox (Lartillot, et al., 2008). Fifteen audio features were used21 covering a large panel of musical dimensions: dynamics (Amplitude RMS22), timbre (Spectral Brightness23, Spectral Rolloff24, Spectral Flux25, Spectral Flatness26 Spectral Spread27, Inharmonicity28The Inharmonicity represents the divergence of the signal spectral components from a purely harmonic signal.[/efn_note], ZCR29, Attack Slope30, sensory dissonance (Roughness31), harmonic and spectral changes (HCDF32, Novelty33), rhythm (Pulse Clarity34, Event Density35), and duration.

We ran a Discriminant Analysis36 in order to test if the variables (audio features plus duration) allow discriminating the four types of segments (Solos-imAgE, “Places+imAgE”, “Places+Thoughts”, “Dreams+ Thoughts”). A Wilks’ Lambda test showed that the difference between the means vectors of the groups is significant F(4.52), p = 0.001. Table 1 shows the eigenvalues and the corresponding percentage of variance.

Discrimination (%)81,2017,061,74
Cumulative %81,2098,26100,00
Table 1. Discriminant Analysis: eigenvalues and the corresponding % of variance.

Two factors explain 98% of the variance: the first one (F1) representing 81%, and the second one (F2) 17%. F1 is mainly correlated with ZCR (.45), Spectral Flatness (.36), Spectral Spread (.35), Spectral Rolloff (.25), and negatively with Duration (-.26) and Inharmonicity (-.24). F2 is mainly correlated with Spectral Flux (.84), Roughness (.79) and Amplitude RMS (.79), and negatively with Spectral Spread (-.60), ZCR (-.54) and Spectral Flatness (-.58).

Figure 14 represents how the 23 segments are grouped and dispatched on the factor axes. We observe that “Solos-imAgE” (SI) and “Dreams+Thoughts” (DT) segments are clearly separated from “Places+imAgE” (PI) and “Places+Thoughts” (PT) segments on the F1 axis. It means that SI and DT segments have globally a brighter and sometimes noisy sound, a larger spectrum, more abrupt changes and a clearer pulse than the other types of segments. These two pairs of segments are strongly opposed from an acoustic point of view and thus offer expressive contrasts. F2 axis is less discriminant, explaining only 17% of the variance. Nevertheless, it allows differentiating SI from DT segments. DT segments are louder, more dissonant, with sharper attacks, and contrasted changes in harmony. PI and PT segments are grouped together on F2 axis meaning that they have close acoustic and temporal features. When followed, these two segments offer a form of textural continuity in the unfolding of the work. The Discriminant Analysis confirms the specificity of the unaccompanied solos (SI) and the “Dreams” segments (DT) previously observed when analysing the tempo. The former are characterized by a freedom of tempo and a great diversity of sonorities while the latter have a more rigid tempo and a more monolithic sound.

Figure 14. The four types of segments dispatched on the factor axes.

The last analysis concerns the variations of tension and release during the unfolding of the piece. Tension-release schemas in tonal music have been approached both from music theory (Berry, 1976; Lerdahl, 2001; Lerdhal & Jackendoff, 1983; Meyer, 1956; Schenker, 1935) and psychology of expectation (Bigand et al., Huron, 2006; Margulis, 2005; Narmour, 1990). Few studies have been done with atonal music (Dibben, 1999; Lalitte et al., 2009; McAdams et al., 2004; Pressnitzer et al., 2000). Theories of musical tension have generally been based on the harmonic dimension, however other dimensions are likely to contribute such as metrics, rhythm, dynamics, timbre, texture, and spatialization. Hjortkjær (2011) proposed a parametric model in which tension is predicted as a combination of low-level acoustic features including measures of intensity variation and distribution of spectral energy, as well as higher-level features related to changes in pitch class distribution and in tonality.

In order to analyse tension and release in Thoughts, Places, Dreams, we chose seven audio features among the fifteen used the Discriminant Analysis, those with a significant p-value37: Spectral Flux (p < 0.0001), Amplitude RMS, and Roughness (p = 0.000), ZCR (p = 0.006), Spectral Spread (p = 0.008), Pulse Clarity (p = 0.019) and Attack Slope (p = 0.032). The real values of the descriptors were transformed in a 0-1-scale in order to make it comparable and representable. The stacked bar graph (Fig. 15) shows the contribution of the seven audio features for each segment. The curve give the mean multiplied by 7 for each segment. Globally, four of the seven descriptors (HCDF, Spectral Spread, Attack slope, and RMS) contribute the most to the tension and relaxation curve. Musically speaking, this means that the curve represents variations in harmonic changes, pitch/timbre range, rhythm activity, and sound intensity. Looking at the diagram, we observe that the piece is acoustically organized around four peaks of “energy” (higher mean values) which globally correspond to the “Dreams+Thoughts” segments (or around these segments). Dynamically, there is a long progression from b-segment to g-segment (“Freeze”) followed by a short release of tension (from h-segment to j-segment). Then, a sudden contrast between j-segment and k-segment, with the higher tension of the piece (“Argument”), is followed by a tense group of segments (from l-segment to n-segment), and a brief release. Then, a progression in two steps (from o-segment to t-segment) leads to the last climax (“Alignment”) and the final release (from u-segment to w-segment). The most important contributors for “Freeze” (g-segment) are Spectral Spread (.90), and HCDF (.79), being the produce of a large pitch range and a dense harmony. “Argument” (k-segment) shows the biggest peak, with all the descriptors at the higher value (1), except for ZCR (.16) and Spectral Spread (.00) which indicate a maximum of energy, density and dissonances in the medium-low register. The main contributors for “Rhythmic Bounce” (p-segment) and “Intercutting” (q-segment) are RMS Energy (respectively .76 and .72), Attack Slope (respectively .62 and .68), and Spectral Flux (respectively .50 and .68), and HCDF (respectively .52 and .82). Here, the musical factors are more complex to determine because different types of writing follow, but strong dynamics, large pitch range and dense harmony are also present. The last major peak comes with “Alignment” (t-segment), the main contributors being Roughness (1), RMS Amplitude (.94), HCDF (.79), Attack slope (.67), and Spectral Flux (.65). All that perfectly matches the writing of “Alignment”, with its stroke dissonant chords, rhythmic complexity and strong dynamics. As the descriptors show, it appears that the “Dreams” are not “sweet dreams”. But, they well fit with the composer’s aesthetic intention to provide concentrated and intense moments. Moreover, this latter analysis confirms the special role of the “Dreams” segments in the formal design of Thoughts, Places, Dreams. Finally, we observe that the “imagE” segments (with and without orchestra) get lesser values (M = 0.26; SD = 0.02) than the “imAge” (M = 0.38; SD = 0.05), in accordance with their evocative vs assertive expressive meaning.

Figure 15. Tension and release curve calculated from seven audio features.

To conclude

This article has chosen to address the analysis of Thoughts, Places, Dreams through two complementary approaches: genetic (the generating ideas and the composer writing strategies) and sonic (the sound realization of the score in the Descharmes, Deroyer and Ensemble Court-Circuit recording). It allowed on the one hand to document the composition process through writings, sketches and plans left by the composer and on the other hand to show how a specific performance can embody and shape the compositional project in its own way. Of course, the audio features values will change to some extent with other performances, other performers, other recording contexts, but certainly not to the point of radically altering the obtained results. The acoustic characteristics of the blocks and the tension and release curve can thus be considered as reliable representations of the work. Our analysis confirmed the role of each “protagonist” imagined by Reynolds. Notably, the soloist assumes a crucial role in this work not only because he is the guiding thread in the unfolding of the work, but also because the cellist constantly manages the tempo and the expression, the conductor and the musicians of the ensemble adapting moment by moment to his playing style. The analysis of the tension/release curve confirmed that the performers further enhance the special role of “Dreams” in the formal design by increasing sound density, roughness, and loudness. Finally, the combined genetic and sonic approaches could lead to hypotheses about how the work might induce musical expectations and emotions in listeners, hypotheses that will need to be tested experimentally with subjects.


Berry, W. (1976). Structural Functions in Music. Dover: Mineola.

Bigand, E., Parncutt, R., and Lerdahl, F. (1996). “Perception of musical tension in short chord sequences: The influence of harmonic function, sensory dissonance, horizontal motion, and musical training”. Perception & Psychophysics, Vol. 58, No. 1, pp. 125–141.

Bithell, D. (2007). “Image, Engagement, Technological Resource: An Interview with Roger Reynolds”. Computer Music Journal, Vol. 31, No. 1, pp. 10–28.

Boyd, M. (2012). “The Evolution of Form in the Music of Roger Reynolds I & II”, Tempo, Vol. 66, No. 259 & 260, pp. 36-48 & 34-49.

Dibben, N. (1999). “The Perception of Structural Stability in Atonal Music: The Influence of Salience, Stability, Horizontal Motion, Pitch Commonality, and Dissonance”. Music Perception, Vol. 16, No. 3, pp. 265-294.

Hjortkjær, J. (2011). Toward a cognitive theory of musical tension (PhD thesis), Copenhagen University, Copenhagen, Denmark.

Hulse, S.H., Takeuchi, A.H. and Braaten, R.F. (1992). “Perceptual Invariances in the Comparative Psychology of Music”, Music Perception, Vol. 10, No. 2, pp. 151-184.

Huron, D. (2006). Sweet anticipation: Music and the psychology of expectation. Cambridge, MA: MIT Press.

Lalitte, P. (1998). “Archipelago de Roger Reynolds. La spécificité du projet dans son contexte esthétique et technique; son analyse en regard des sciences de la perception”. Mémoire de DEA, IRCAM/Université Paris-Sorbonne.

– (2005). “Compulsory Figures” in Stephen McAdams & Marc Battier (Eds.). Creation and perception of a contemporary musical work: The Angel of Death by Roger Reynolds, DVD, Paris: IRCAM-Centre Georges Pompidou.

Lalitte, P., Bigand, E., Kantor-Martyruska, J., and Delbé, C. (2009). “On listening to atonal variants of two Piano Sonatas by Beethoven”. Music Perception, Vol. 26, No. 3, pp. 223-234.

Lartillot, O., Toiviainen, P., Eerola, T. (2008). “A Matlab Toolbox for Music Information Retrieval”, in C. Preisach, H. Burkhardt, L. Schmidt-Thieme, R. Decker (Eds.), Data Analysis, Machine Learning and Applications, Studies in Classification, Data Analysis, and Knowledge Organization, Springer-Verlag.

Lerch, A. (2012). An introduction to audio content analysis, Hoboken (N. J.), Wiley.

Lerdhal, F. (2001). Tonal Pitch Space. Oxford: Oxford University Press.

Lerdhal, F. & Jackendoff, R. (1983). A Generative Theory of Tonal Music. Cambridge Massachussetts: The MIT Press.

Margulis, E. H. (2005). “A model of melodic expectation”. Music Perception, Vol. 22, No. 4, pp. 663–714. doi: 10.1525/mp.2005.22.4.663

McAdams, S., & Matzkin, D. (2003). “The roots of musical variation in perceptual similarity and invariance”. In I. Peretz & R. Zatorre (Eds.), The Cognitive Neuroscience of Music, Oxford: Oxford University Press. pp. 79-94.

McAdams, S., Vines B.W., Vieillard, S., Smith, B.K., & Reynold, R. (2004). “Influences of Large-Scale Form on Continuous Ratings in Response to a Contemporary Piece in a Live Concert Setting”. Music Perception, Vol. 22, No. 2, pp. 297-350.

Meyer, L.B. (1956). Emotion and Meaning in Music. Chicago and London: The University of Chicago Press.

Narmour, E. (1990). The Analysis and Cognition of Basic Melodic Structures: The Implication-Realization Model. Chicago, IL: University of Chicago Press.

Peeters, G. (2013). “A large set of audio features for sound description (similarity and classification) in the CUIDADO project”, Rapport technique Ircam (

Pressnitzer, D., McAdams, S., Winsberg, S., and Fineberg, J. (2000). “Perception of music tension for nontonal orchestral timbres and its relation to psychoacoustic roughness”. Perception and Psychophysics, Vol. 62, No. 1, pp. 66-80.

1 I would like to express my gratitude to Roger Reynolds, who has gently answered my questions and provided numerous helpful documents and to Alexis Descharmes for his helpful insights on the performance of the work. I also would like to thank Thomas Désarménien for his careful proofreading.

Reynolds, R. (1987). A Searcher’s Path, a composer’s Way, I.S.A.M., Monograps n° 25, New York.

– (2002). Form and Method: Composing Music, New York and London: Routledge.

– (2007). “Ideals and Realities: A Composer in America”, American Music, Vol. 25, No. 1, pp. 4-49.

– (2015). Booklet of the 2-CDs Roger Reynolds: Complete Cello Works – Alexis Descharmes.

Schenker, H. (1935). Der Freie Satz. Vienna, Universal Edition.

Takahashi, Y. (2009). Booklet of the 2-CDs Roger Reynolds Complete Piano Works – Epigram and Evolution, mode 212/213.


Roger Reynolds, Thoughts, Places, Dreams, Edition Peters, EP 68443.


Roger Reynolds: Complete Cello Works – Alexis Descharmes, mode 277-78 (2015).

Roger Reynolds the imagE-imAge set, CD NEUMA 450-114 (2015).

  1. I would like to express my gratitude to Roger Reynolds, who has gently answered my questions and provided numerous helpful documents and to Alexis Descharmes for his helpful insights on the performance of the work. I also would like to thank Thomas Désarménien for his careful proofreading.
  2. In the following article, we will keep Reynolds’s typography for the titles of these pieces: imagE for Evocative and imAge for Articulate.
  3. Chemins I (1965) for harp and orchestra on Sequenza I; Chemins II (1967) for viola and 9 instruments on Sequenza VI; Chemins III (1968-73) for viola and orchestra on Chemins II; Chemins IV for oboe and 11 strings (1975) on Sequenza VII; Chemins V (1992), for guitar and chamber orchestra on Sequenza XI.
  4. Roger Reynolds: Complete Cello Works – Alexis Descharmes, mode 277-78 (2015). Live recording, November 152013, at the Auditorium Marcel Landowski, Conservatoire à Rayonnement Régional de Paris, record engineer: Alice Legros.
  5. Reynolds’ email sent to the author on December 1, 2014.
  6. Among the extended techniques used there are: arco, pizz., sul ponticello, sul tasto, vibrato, non vibrato, harmonics, quasi flautando, glissandi, regular or irregular tremolos, staccatissimo, strummed chords, microtonal wail, jeté, over pressure, etc.
  7. Ensemble Court-Circuit and Jean-Marie Cottet (piano), conducted by Pierre-André Valade, premiered The Angel of Death, at the Pompidou Center during the Agora Festival (Ircam) on June 2001. Then the composer was again in interaction with the Ensemble in 2002, when the festival Why Note (Dijon) commissioned Process and Passion for violin, cello, computer processed sound, and real-time computer spatialization; Nicolas Miribel (violin) and Alexis Descharmes (cello) premiered the piece on November 15.
  8. Thoughts, Places, Dreams was commissioned by the Ensemble Court-Circuit with the assistance of the French Ministry of Culture and is dedicated to Alexis Descharmes.
  9. Descharmes notated in blue and red colours the changes suggested by Reynolds.
  10. Bars: 1-8, 7-43, 81-89, 133-144, 197-210, 300-319, 368-374. Occasionally, the cello is accompanied by slight interventions of a few instruments, but they are elements of a ghost polyphony designed to highlight the cello part.
  11. A core element is a kind of “genetic material” which encloses strong identities in terms of proportion, pitch content, rhythm activity, texture and morphology. It allows the composer to create “seed” structures and derivations that nevertheless preserve a high degree of perceptual invariance (Hulse, et al., 1992; McAdams & Matzkin, 2003). For more information concerning the notion of core elements in Reynolds’ music see Reynolds (1987, 2002) and Lalitte (1998, 2005).
  12. Reynolds’ notebook, p. 2.
  13. Reynolds’ notebook, p. 17.
  14. Reynolds’ notebook, pp. 35-41.
  15. These algorithms are PROLIF, SMEARZ, MATRIX and THINNR. They were used in live electronics pieces including SEASONS project, Dream Mirror, MARKed MUSIC, POSITINGS, george WASHINGTON, and FLiGHT, etc. For more information concerning these algorithms, see Bithell (2007).
  16. For example: Tbn/Horn a = C-sharp, b = G, c = E-flat, d = F-sharp.
  17. While he used other sources such as Fibonacci series, Henon attractor, or even armature texts, the logarithmic series remains the main source (Reynolds, 2002).
  18. The duration of the piece was planned to be 20 minutes, thus the unity here corresponds to 1 second.
  19. Values during extended pauses were removed in order to obtain a true average tempo for each segment.
  20. Interview of Alexis Descharmes by the author, on December 6, 2015.
  21. For more information about audio descriptors, see: Geoffroy Peeters (2013), and Alexander Lerch (2012).
  22. The Amplitude RMS measure the global energy of the signal computed by taking the root average of the square of the amplitude.
  23. The Brightness is the amount of energy above a cut-off frequency (here 1500 Hz).
  24. The Spectral Roll-off point is the frequency so that 95% of the signal energy is contained below this frequency.
  25. The Spectral Flux indicates indicate the temporal position of important contrast in the spectrogram. It is computed as the distance between the spectrum of each successive frames.
  26. The Spectral Flatness is a measure of the noisiness (flat, decorrelation) vs sinusoidality of the spectrum. It is computed by the ratio of the geometric mean to the arithmetic mean of the energy spectrum value.
  27. The Spectral Spread is a measure of the shape of the spectrum. It is defined as the spread of the spectrum around its mean value.
  28. The Attack slope is defined as the aver-age temporal slope of the energy during the attack segment.
  29. The Zero-Crossing Rate is a measure of the number of time the signal value cross the zero axe.
  30. The Attack slope is defined as the aver-age temporal slope of the energy during the attack segment.
  31. Sensory dissonance relates to the beating phenomenon when overtones are close in frequency.
  32. The Harmonic Change Detection Function (HCDF) is the flux of the tonal centroid.
  33. The Novelty curve represents the probability along time of the presence of transitions between successive states, indicated by peaks, as well as their relative importance, indicated by the peak heights.
  34. The Pulse clarity estimates the rhythmic clarity, indicating the strength of the beats estimated by the mirtempo function.
  35. The Event density estimates the average frequency of events, i.e., the number of note onsets per second.
  36. The discriminant analysis enables the researcher to examine whether significant differences exist among the categories, in terms of the predictor variables. It also evaluates the accuracy of the classification.
  37. “p-value” is a standard test of statistical significance, used throughout empirical sciences.