Skip to content
View in the app

A better way to browse. Learn more.

Learn more

Stochastic Thinking: Xenakis, ST/4, and the New Opusmodus Constraints System

I have been interested in the music and ideas of Iannis Xenakis since my earliest encounters with computer music in the early 1980s. His work remains a source of fascination, not simply because of the mathematical procedures that inform it, but because of the extraordinary musical imagination that those procedures made possible. I have always found the music itself more revealing than the theories considered in isolation. The scores, and especially the experience of listening, disclose a world of sound in which complexity, violence, delicacy, and unexpected coherence coexist.

Among the works that have particularly interested me are ST/4 and ST/10, compositions from the early 1960s associated with Xenakis's pioneering experiments in computer-assisted stochastic composition. I have returned to these pieces many times over the years, attracted by their peculiar combination of rigorous organization and seemingly spontaneous musical behavior. There is something fascinating about music in which individual events can appear almost arbitrary, yet the larger texture possesses such a compelling and unmistakable identity. I am especially interested in how these works move between states of relative transparency and extraordinary complexity, and how musical coherence can arise from the behavior of an ensemble rather than from conventional thematic development.

My own engagement with stochastic processes reaches back to my earliest work with computer music, first in the world of UNIX, C, and software synthesis, and subsequently through various systems for algorithmic composition. I have never regarded these procedures simply as ways of generating notes. Rather, I have been interested in how they might enlarge the composer's imagination, suggesting relationships, textures, and temporal structures that would be difficult to conceive through more conventional means. The attraction is not randomness for its own sake, but the possibility of shaping musical circumstances in which something unexpected, and perhaps musically revealing, can happen.

Recently, I decided to return to the ST works in a more practical way, by developing my own stochastic string quartet model in Opusmodus. This has become a particularly absorbing project, partly because it allows me to revisit questions that interested me more than forty years ago, and partly because I can now approach them with a programming environment that makes experimentation unusually immediate. I am not attempting to reconstruct Xenakis's original computer programs or to compose another ST/4. My intention is to investigate some of the principles that make these works so interesting to me and to discover what they might suggest for my own compositional practice.

I began with a relatively straightforward question: what happens when four instrumental voices are allowed to develop according to independent stochastic temporal processes, while remaining participants in a common musical situation? From there, the program has gradually acquired a more elaborate architecture. Each instrument generates its own sequence of events, drawing upon changing distributions of pitch, duration, and temporal density. At the same time, the ensemble is understood as a collective entity whose overall activity, registral disposition, and patterns of silence can be examined and shaped. I am interested in the tension between these two levels of organization: the independence of the individual instrumental parts and the emergent behavior of the quartet as a whole.

As the project has developed, I have become increasingly concerned with what I think of as the morphology of the musical texture. I want the quartet to do more than alternate between statistically different sections. I want it to breathe, to gather energy, to become concentrated or dispersed, and sometimes to withdraw almost entirely into silence. These are familiar compositional concerns, of course, but the challenge here is to approach them through statistical relationships rather than through the direct specification of individual musical events. I find myself listening not only to the notes that emerge, but to the changing character of the ensemble, and asking whether the processes I have designed are producing something musically persuasive.

This work has coincided with my exploration of the new Opusmodus Constraints System, which I regard as a major advancement in the development of the environment. For me, its importance lies in the possibility of moving beyond the generation of musical material toward a more comprehensive consideration of the relationships that material establishes. Constraints allow certain conditions to be expressed explicitly, and possible realizations can be examined in relation to those conditions. In the quartet project, I am combining the generation and statistical evaluation of complete ensemble textures with native constraint-based validation. This opens a particularly interesting avenue for further research.

What appeals to me is that the use of constraints need not eliminate unpredictability. Quite the contrary, it can provide a framework within which unpredictability becomes more musically useful. I can establish a preference for particular kinds of ensemble behavior without determining precisely how those behaviors will be realized. The resulting music is neither entirely prescribed nor simply left to chance. It emerges from an interaction between the statistical processes, the conditions I establish, and the compositional judgments I make while listening to the results.

There is also a personal pleasure in the programming itself. I have spent much of my creative life working with computer music systems, and I continue to find the development of compositional software an extension of composition rather than a separate technical activity. Working in Lisp and Opusmodus has given me a renewed sense of freedom in this regard. Ideas can be formulated, tested, revised, and sometimes abandoned with considerable immediacy. The quartet model has already passed through numerous revisions, each suggesting new questions and exposing limitations that become apparent only when the music is heard. I rather enjoy this process of discovery, in which the program and the composition seem to develop together.

What continues to excite me about Opusmodus is its capacity to sustain this kind of inquiry. Its combination of Lisp, musical data structures, generative procedures, and now a sophisticated Constraints System offers a remarkably fertile environment for compositional research. I see the new system as especially significant because it encourages us to think not only about the production of musical events, but about their relationships, their collective behavior, and the conditions under which a musical idea might become coherent.

I do not yet know where this quartet project will lead, and that is part of its attraction. It may eventually become the basis for a finished composition, or it may remain an evolving laboratory for exploring statistical musical behavior. Either outcome would be worthwhile to me. The questions themselves are compelling, and I find considerable satisfaction in returning to ideas that first occupied my imagination decades ago, now equipped with new tools and a somewhat different understanding of what I am listening for.

Perhaps that is the enduring lesson I take from Xenakis: mathematical and computational procedures are not substitutes for musical imagination. They are means of extending it. Ultimately, the value of any such system must be discovered in the music it enables us to make, and, above all, in what we hear.

12 Comments

Recommended Comments

AM

Members

The code would be the most interesting part — maybe you could share it here?

RST

Members

Yes, I do plan to create an application based on the current work. This will eventually be available at my Substack.

AM

Members

Just here, as simple as possible. Thx

RST

Members

8 minutes ago, AM said:

Just here, as simple as possible. Thx


I am sorry to say that my algorithm is not all that simple. It is based on the program that Xenakis developed called ST for what he termed "free stochastic music."

The original program ST was coded by M. F. Génuys and M. J. Barraud in FORTRAN IV on an IBM 7090 . In his book Formalized Music, Xenakis discusses some of his applications of the program to his work generally.

The Fortran code is available online here:
https://github.com/ThemosTsikas/XenakisFreeStochasticMusicFortran/blob/main/XenFSM.f

The technique (and to an extent the ST program itself) is also fairly well documented elsewhere: for example,
https://books.openbookpublishers.com/10.11647/obp.0390/ch6.xhtml
https://www.iannis-xenakis.org/en/stochastic-synthesis/
https://www.iannis-xenakis.org/en/stochastic-music/
https://ccrma.stanford.edu/~dkeller/pdf/KellerFerneyhough2004.pdf
https://www.jstor.org/stable/40926355

The Keller and Ferneyhough discussion (and their work in Patchwork) is particularly informative, but I am not sure where that program is located or how well documented it is. In my view, Patchwork and OpenMusic which followed it, are programs that are sometimes difficult to read and follow, though others may like the graphic approach.

It is possible that someone has continued working with ST-like ideas in OpenMusic.
https://openmusic-project.github.io/openmusic/

AM

Members

I know the Xenakis things. I thought you had programmed your track/example in Opusmodus, so the code would be easy to share.

RST

Members

4 minutes ago, AM said:

I know the Xenakis things. I thought you had programmed your track/example in Opusmodus, so the code would be easy to share.


Yes, the audio example was created using Opusmodus and the current program version is v1.8.1 - basically it creates a continuous stochastic morphology and uses some of the new constraints functions that have recently been added to Opusmodus. I used the VSTi Opus, a sampled string quartet, and recorded the example with Nuendo.

I do plan to share this program when I am finished working on it.

The Opusmodus program and reference materials will uploaded to my Substack.

AM

Members

I think it’s much better for the forum community and engagement if you post it here directly rather than referring people to your website. In my opinion it’s more direct and helps the forum community thrive ☺️

RST

Members

6 minutes ago, AM said:

I think it’s much better for the forum community and engagement if you post it here directly rather than referring people to your website. In my opinion it’s more direct and helps the forum community thrive ☺️


Oh yes, I see.

Thank you for your thoughts. I certainly understand your point, and I agree that a thriving forum community depends upon the generosity of its members and their willingness to exchange ideas, experiences, and knowledge. I have always valued that kind of collegial exchange, particularly in the world of computer music, where so much of our collective progress has depended upon people sharing their discoveries.

At the same time, I am trying to develop something rather more substantial through my Substack publication. After more than four decades as a professional composer, researcher, and educator, including over thirty years teaching composition and music technology at the university level, I have accumulated a considerable body of knowledge, experience, and intellectual property. I am now devoting a significant amount of my time to documenting that experience, developing original compositional software and generative models, and writing about the ideas and techniques that inform my creative practice.

This is not simply a matter of sharing a few examples or fragments of code. Much of what I am publishing represents many hours, sometimes days or weeks, of research, experimentation, programming, testing, musical evaluation, and documentation. I regard this work as an extension of my professional creative practice, and I believe that both the intellectual effort and the time invested in it have value.

My Substack is intended to provide a sustainable way of continuing this work, with freely accessible writing alongside more substantial materials available to supporting subscribers. The modest subscription fees are a means of supporting the ongoing research and development, rather than an attempt to discourage the exchange of ideas.

I remain very happy to participate in the forum, share insights, discuss techniques, and contribute to the community. Indeed, I hope that what I am doing will ultimately enrich that community as well. But I also believe that composers, researchers, and educators should feel comfortable assigning value to their intellectual and creative labor, just as we recognize the value of the software and other tools we use, such as Opusmodus.

I hope you can appreciate the distinction. I see these two forms of engagement as complementary rather than competing, and I would very much like to continue contributing to both.

RST

Members

Where the algorithm is headed, generally speaking:

Compositional Form

Sequence, duration, proportion and recurrence of musical moments

to

Stochastic Field Definitions

Density, duration, tessitura, register, dynamics, silence and instrumental relationships

to

Field Relationships and Transitions

Juxtaposition, overlap, gradual morphing, interruption, recurrence

to

Stochastic Event Generation

Independent instrumental realizations governed by the active field

to

Constraints and Musical Evaluation

Local field identity, ensemble relationships and larger formal coherence

RST

Members

The field library for the current working version

;;; Each field: name, per-instrument arrival rate per 1/16,

;;; sounding durations in 1/16 units, register center/width

;;; (normalized within each instrument's pitch field),

;;; dynamics, and per-instrument activity multipliers.

;;; An activity multiplier of 0 silences that instrument.

(defparameter st4-field-library

'((:suspended

:rate 0.10 :durations (4 6 8 12 16)

:center 0.50 :width 0.46

:dynamics (pp p mp)

:roles (1.0 0.65 0.85 0.50))

(:agitated

:rate 0.75 :durations (1 1 2 2 3 4)

:center 0.65 :width 0.18

:dynamics (mp mf f ff)

:roles (1.15 1.0 0.95 0.75))

(:fragmented

:rate 0.26 :durations (1 2 3 5 8)

:center 0.50 :width 0.38

:dynamics (p mp mf f)

:roles (1.0 0.45 0.85 0.65))

(:saturated

:rate 0.85 :durations (2 3 4 6 8)

:center 0.50 :width 0.48

:dynamics (mp mf f ff)

:roles (1.0 1.0 1.0 1.0))

(:residual

:rate 0.08 :durations (1 1 2 3 4)

:center 0.80 :width 0.17

:dynamics (pp p)

:roles (0.70 0.35 0.45 0.20))))

;;; The composer determines the form explicitly.
;;; Each entry is (field-name number-of-4/4-measures).
;;; Unequal proportions and returns are intentional.
(defparameter *st4-field-form*
  '((:suspended 4)
    (:fragmented 5)
    (:agitated 5)
    (:suspended 3)
    (:saturated 6)
    (:fragmented 5)
    (:residual 4)))
;;; The composer determines the form explicitly.
;;; Each entry is (field-name number-of-4/4-measures).
;;; Unequal proportions and returns are intentional.
(defparameter *st4-field-form*
  '((:suspended 7)
    (:fragmented 1)
    (:agitated 1)
    (:suspended 3)
    (:saturated 2)
    (:fragmented 1)
    (:residual 8)))

Same seed, and all other parameters the same. Only change is the form definition above.


RST

Members

image.png

RST

Members

The Opusmodus Source Code and further documentation is now at Substack.

https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbbd3870a-497d-42b3-b0d4-872caca50dfa_1672x941.png

Search

Search

Configure browser push notifications

Chrome (Android)
  1. Tap the lock icon next to the address bar.
  2. Tap Permissions → Notifications.
  3. Adjust your preference.
Chrome (Desktop)
  1. Click the padlock icon in the address bar.
  2. Select Site settings.
  3. Find Notifications and adjust your preference.