The same desire to engage all musicians resulted in the decision to compose most of the final polyphony of the concert, the music for the communion. The concert was performed by seven musicians, two of which had not participated in workshops on improvisation. One of them sang the tenor line and did not need to be able to improvise. The other sang one of the top lines, which required either reading or making a counterpoint, the last of which was not an option. I briefly considered composing a top line and letting all other musicians improvise around it, but found the idea unsatisfactory. With the skills and time at our disposal, it would bind the other voices, without the attentive, searching quality we wanted, and without offering an approach we had not tried before, such as in the offertory described above. I also wanted to experiment with a composition with five voices. There are no examples of that in Ars Nova, but the idea of creating a sound richer than four-voice polyphony was tempting to me. I therefore stacked four voices on top of the tenor, all following the rules of two-part counterpoint. The score I handed out was drafted quickly and without much detail. This allowed for an active shaping of the music by the whole ensemble, through small diminutions, timbre, dynamics, text underlay, and other parameters apart from the general melodic lines. I took the central, opening motif from Holters improvised part at an earlier performance. I also asked Danilevskaïa to make her own part, somewhere between the lower voices. With this, I intended to leave the responsibility and ownership of the music with the ensemble rather than taking it on myself. The score is to the right and can be opened and downloaded.


This part of the exposition will focus on our development of improvisation within styles of the late fourteenth and early fifteenth centuries, and show some of the results as presented in the concert Mouvance I. Mass. It contains several short audio excerpts from the concert, some musical notation, and a video documentation of the full concert available at the end of the page.

Improvisation within historical styles, often referred to as "historical improvisation", is an activity receiving more and more attention within the field of Historically Informed Performance (HIP). Some early music departments offer courses in historical improvisation (e.g. Hochschule für Musik und Theater "Felix Mendelssohn-Bartholdy" in Leipzig, the Royal Conservatoire The Hague and the Schola Cantorum Basiliensis), and the number of ensembles and performers who in some way or other incorporate improvisation into their performances of early music is by now so high that a comprehensive survey would be futile. Concentrating on the improvisation of polyphony in the styles of the late Middle Ages and Renaissance, and more precisely the cantare super librum where the polyphony is improvised over a pre-existing melody, the picture becomes less profuse. I will nevertheless not attempt a comprehensive account here, but limit myself to the musicians and researchers who had a more direct influence on our work. (For a more thorough and relatively recent overview over improvised polyphony in artistic practice and musicology, as well as a discussion on different notions of the term improvisation, see Berentsen.)1

We have worked closely with prof. Barnabé Janin (Conservatoire National Supérieur Musique et Danse de Lyon), whose experience in improvisation of polyphony in a number of historical styles with the ensembles Le Chant sur le Livre, Obsidienne, and Ensemble Coclico was of very high value to us. In developing this leg of our project, we were also inspired by workshops on improvisation held by his colleagues in Le Chant sur le Livre, Jean-Yves Haymoz, Pierre Funck and Raphaël Picazos. Finally, I would like to mention and thank Dr Niels Berentsen, who kindly shared with me his PhD treatise "New Approaches to Vocal Polyphonic Improvisation 1300–1470" prior to its publication.2 His project is naturally highly relevant to our undertaking.


Goals and research questions

As I have argued in an essay available on Research Catalogue, historical probability does not function as an aesthetic goal.3 We see HIP as a methodological field, an artistic process where part of the decision-making is based on interpretations of a selection of historical sources. The study of historical material serves to question our practices and habits, helps us imagine and search for other solutions or other ways of thinking and creating music. With our project, we wanted to question our notions of music and our typical compartmentalisation of music-making into composers and compositions, performers and performances, and the juxtaposition of composition and improvisation. One of the reasons for choosing late medieval sources was that the terms "composer", "composition", and "improvisation" were not in common use in this period.4 We wanted to explore this kind of music-making, less dependent on performances of stable pieces and more dependent on the capacity of the participants to spontaneously create a common expression, in a flat ensemble structure and with common authorship.

The goal was not to uncover, restore, or re-enact with any kind of accuracy any historical practice – an impossible task for a number of obvious reasons. Leaving aside hermeneutical discussions on the shifting understandings of the past, what we currently believe to understand of medieval music culture cannot replace our culture in the course of a three-year project. Knowing that medieval musical culture involved concepts of visual, tactile, and auditive organisation and memorisation tremendously different from our own (from general theory of the modes, to the Guidonian hand, solmisation of hexachords, musica recta/ficta, and down to the number of lines and the types of clefs used in staff notation), does not replace or overwrite our basic musical training, culture, and instinct. What one can do, is to introduce a limited number of historical concepts, and see how these interact with the musicianship of the participants.

We wanted to find out what kind of music we would be able to make when we went beyond reading medieval scores to creating the polyphony together. We wanted to see what kind of medieval sounding polyphony we would we be able to improvise, how the process of developing the necessary skills could influenced our performance both as individuals and as an ensemble, and whether anything of this process would be of value to the other legs of the project.



As stylistic framework we chose polyphonic music of the Italian Ars Nova, also known as Trecento, lasting from about 1320 to 1420. We focussed on the later part of the period, excluding the style known as Ars Subtilior, but allowing some space for experimentation in an early Renaissance style. Our goal was to be able to make our own liturgical polyphony, which meant we would work on cantus super librum, a practice where one or more of the musicians sing a plainchant whilst others create counterparts. We studied in particular compositions by Antonio Zacara da Teramo (ca1365–1416), Paolo da Firenze (ca1355–1436), and Guillaume Dufay (1397–1474), as well as a number of anonymous pieces. For instrumental diminutions, our main source was the famous Codex Faenza,5 a manuscript with virtuoso diminutions to liturgical and secular music. (Diminution is a term used to describe the technique or practice of ornamenting or elaborating one or more of the parts in a counterpoint with faster-moving note values. It can be applied by vocalists as well as instrumentalists.) We also studied Paola da Firenzes treatise on counterpoint, one of very few late medieval treatises that offers not only the general rules but also a rudimentary training method for improvising counterpoint.6 We were interested in the relation between the simple outline of these rules, or rather recommendations, and the richness and openness of the late medieval contrapuntal styles.

Since the goal was to find out what music could emerge when the musicians co-created the music, the most important component was obviously the participants: one recorder player, a male and a female singer, and two medieval fiddle players participated directly and continuously in the studies and performances of improvised medieval music. (Other project participants were present to varying degrees in workshops, some of them participating in one or two public performances.) The musicians were connected through their previous work with the ensemble Currentes. Nevertheless, their backgrounds were quite dissimilar, from the musician mainly based within the field of medieval music, to the ones who might have felt more like guests in the early music field. The experience with improvisation was highly varied as well, and very few had anything but a modest experience in improvisation in medieval styles. In the discussion, we will come back to how the varied experiences and areas of expertise became visible and influenced the directions of the project as a whole.

Basic theory, methods, and examples of results

Our basic method of work was, firstly, to study and practice in workshops Paolos basic rules for simple counterpoint. To the right are images of three of the twenty examples of rudimentary contrapuntal motions provided by Paolo. The lower line is a hypothetical tenor, the higher line the recommended movement of the counterpoint. Note how the simple counterpoint always moves between octaves and fifths.

Secondly, we studied repertoire, which we analysed according to a few simple categories of medieval counterpoint, the principalis, finalis, and penultima, i.e. the first, last, and penultimate sonorities of each phrase. (See examples below.) We looked for other improvisable structures as well, and practiced our findings in workshops. I do not have space here to describe the richness of approaches we developed. They will be offered in a separate compendium on at a later point. The interested reader can in the meantime refer to the above-mentioned PhD thesis by Berentsen, which offers a solid overview of improvisatory strategies for the styles of Ars Nova and the early Renaissance. For Renaissance styles, one can also consult the highly practical and instructional publication by Barnabé Janin, "Chanter sur le livre".7 I will here offer a few key points from our work:

Our improvisations were always based on a cantus firmus, i.e. a pre-existing melody that we had in front of us in musical notation. This particular technique of improvisation is frequently referred to as cantare super librum or "singing upon the book", since it involves reading a melody from a book (usually of plainchants) and improvising polyphony around it.


In Ars Nova, the counterpoint rules are based on two voices: a tenor, which in our case would be the cantus firmus, and a cantus, a top voice. The treatises of the time do not prescribe how a third voice, a contratenor, should relate to these two. The contratenor follows the same rules as the cantus, but seeks a line that is different from it. Looking at the repertoire, the practice clearly allows for collisions, such as intervals of a second or a seventh, or also unison movements between the two counterpoints. But at cadences, the voices will usually be distributed in one of the following manners:


Jostein Gundersen

The Italian Ars Nova does not have other basic cadence movements. Adding a fourth part, a second contratenor or a second cantus, therefore implies doubling the cantus or the contratenor at the cadence, either at the unison or at the octave. Music of Ars Nova never has more than four voices.

We would almost always start by making a cantus in simple counterpoint over the cantus firmus. (Simple counterpoint means one note in the cantus for each note in the tenor.) We would then add a contatenor, eventually a fourth voice, and slowly add more notes, often called diminutions, in the main mensurations to the counterpoints.


We initially thought of the simple counterpoint as an exercise, a sort of necessary first step in developing a richer polyphony, but soon discovered that the simple polyphony grants a possibility which disappears once a more rhythmically complex counterpart part is added: A simple counterpoint is rhythmically as flexible as the cantus firmus. This opens for a free, almost recitational style that was very attractive to us, particularly in longer sections with much text in the liturgy, such as the introit. In the following audio example, we hear Ingvill Holters simple counterpoint to the introit "Benedicta sit". The plainchant is sung by Tyler Ray and Kjetil Almenning. The plainchant is given to the right, in square notation and next to it in modern notation.

One problem with the method of always starting with a simple counterpoint, is that diminutions quickly take on a secondary character, as if they would be a kind of ornamentation to a more important structure. This is the way musicology tends to describe diminutions, as ornamentation of a Gerüst or foundational contrapuntal structure. I have shown elsewhere how a composition can utilise the same diminution in a number of different contrapuntal situations to create structure and unity, and that ideas of a fundamental contrapuntal structure or Gerüst might fail to recognise the structural importance of diminutions.8 Reducing a composition to a simple counterpoint would in most cases mean neutralising its character and structure. Working the other way around would after a while tend to be equally uninspiring: All counterpoints would sound similar and yield few ideas for a continuation. We therefore found it useful to sometimes start with diminutions: We would search in the repertoire for typical figures and employ them as often as possible in our own settings. In the following audio example, Ingvill Holter used motifs from Paolo da Firenzes setting of the introit "Gaudeamus omnes"9 in her counterpoint to the introit "Benedicta sit, sancta Trinitas". Some of the motifs she used are notated to the right. The plainchant is the same as in the above example.

As can be heard already at the end of the first audio example, and throughout this second example, tenor Kjetil Almenning is adding a simple contratenor to the tenor (sung by Tyler Ray) and Ingvill Holters diminished counterpoint. His choice of notes is largely determined by register and interval: His register is always between cantus and tenor, and he usually sings thirds or fifths above the tenor. The expressive quality of the three voices as a whole is particular in that it moves between sometimes clearly directed lines, sometimes more hesitant and tentative searching. It is a kind of expression that differs from reading a score, and results from the complex process of listening attentively, remembering solutions that might have worked in rehearsals as well as adapting to unexpected developments.


The method of starting with motifs rather than with simple counterpoint became even more important when we added faster moving counterpoints in the style of Codex Faenza. In the following audio example, I chose motifs from Codex Faenza, organised them according to contrapuntal situation (principalis, finalis, penultima) and intervallic movement, and applied them as often as possible in a diminished counterpoint to the offertory "Benedictus sit, Deus pater". We moved the tenor up an octave, and then added a few notes and rhythms in order to create directions towards cadences. Adding notes and changing the rhythms of plainchants is a practice we learnt from Codex Faenza. I sketched a simple contratenor for short sections of the plainchant. In the recording, soprano Sunniva Eliassen sings the tenor, Ingvill Holter the contratenor, while I play the cantus on a recorder. The fiddle player Hans Lub doubles the lower soprano whilst Anna Danilevskaïa created her own melodic lines between that of the recorder and the contratenor. Below the audio example is a small sample of motifs we selected from Codex Faenza. To the right is the plainchant in square notation from Liber usalis and our arrangement of it.

Planning one voice dictates to some degree the movements of another as one tries to avoid doubling each other in counterpoint, especially in cadences. This results in an expression that sounds more planned and studied than the previous example. The fast diminutions add to this quality: As a general rule, the faster movements we wanted to play, the more we needed to rehearse. The more we rehearsed, the more we remembered in the concert of the rehearsed solutions. Whilst gaining in richness and virtuosity, the expression lost some of the at times hesitant and searching quality that was quite attractive to us. Danilevskaïas part, the least audible here, is maybe the one that retains most of this quality.



While creating a third voice to a two-part counterpoint was a more complex task for us than making the initial top part, making a fourth voice was not very different from creating the third. The process for the contratenor and the cantus secundus is more or less the same: listen to where the cantus goes, and avoid doubling its line. The contratenor does this below the cantus, the second cantus does the same above, the latter usually a bit more agile than the lower voices. The contratenor and second cantus do not try to avoid each others lines. This results in frequent doublings at the octave as well as small, dissonant clashes, and belongs to the style. We also took advantage of the difference in sound quality and volume between fiddles and voices, and let one of the fiddle players invent a contratenor parallel to that of the vocal contratenor. This is not a practice we know from history, but an experiment we initiated wishing to engage both fiddle players: One already doubled the tenor. The other could not double any voice as such, since none of them were notated. The solution was to do more or less the same as the contratenor. Their lines coincide to a large degree, but sometimes deviate or have different diminutions that cause small collisions. In our opinion, the collisions do not disturb, neither at the octave nor in the unison, but because of the different timbres rather contribute to a richness in sound.

The audio example below is a four-part counterpoint to the section "misericordiam suam" of the offertory "Benedicta sit", given above in square notation. The modern notation for the section is to the right.

Discussion of results

The borders between improvisation and composition are fleeting, not least because there are no clear and stable definitions of what constitutes either of the two. I believe it is safe to assume that most people associate composition with music that is prepared, stable, and lasting, and with a visible authorship, whilst improvisation represents something spontaneous and ephemeral, performed rather than authored. Many are the practices, historical as well as contemporary, that in different ways challenge the categories: compositions made through improvisation, improvisations within compositions, notated improvisations, spontaneous compositions, memorised improvisations, or cultures such as the late medieval where the terms composer and composition are not in use. The term comprovisation has been suggested by several researchers and musicians, as denoting some middle ground between composition and improvisation.10


I do not wish to enter into etymological discussions or propose new definitions. Through what might be labelled a re-enacting of historical practice, we have tried to investigate artistically how we understand the practices of composing, performing, and improvising. We have explored what kind of music and music-making such re-enactment could lead to. The definition of the verb "improvise" in the Oxford Dictionary of English proposes that improvisation is without preparation.11 The music we have made through the project was never without preparation. On the contrary, in order to obtain possibility of choice, some flexibility of movement, we have had to prepare more than ever. Through this process of preparation, we have found solutions that we have liked, kept, and repeated, sometimes to such an extent that the audible result is almost as stable as the compositions we read from scores. (This was also noted in a review for the Norwegian journal Ballade of the concert series Mouvance.12) We have also developed skills and a vocabulary to create polyphony with increasingly shorter preparation time, with increasing flexibility, and an ability to find solutions even when all memory of solutions fails.


It is difficult to know when we cross the line between improvisation, memorisation, and composition, not only for the performers but also for the audience. Starting with the latter, I suspect the main reason they know we are improvising is that we tell them so. The performative effect of verbalising and sharing our creative strategies with an audience can hardly be overestimated. If we communicate to the audience that we improvise, their experience of the music will undoubtedly be coloured by their understanding of what it means to improvise, an understanding which probably is at some distance from that of the musicians. The aesthetic and ethical ramifications of this performative action of sharing processes with the audiences are very interesting to me, and deserve to be discussed in more detail some other time.


The common denominator for the music we make through this project is that we make it, together. Nobody has a complete overview of the activity of the other voices. As a rule, we dont put our solutions into notation. There is no score, only the parts of each of the performers, who all have different strategies for visualisation and memorisation of performance possibilities: Some read from c-clef, some from g-clef or f-clef, some write no annotations, some write notes and others write numbers for their own voices, some jot down key cadences, etc.. If a performer leaves, nobody can perform his or her part, so the music changes. Even for the above-mentioned composed communion, the only music we have notated in detail, no-one has a score with the full version of the piece: Diminutions, dynamics, breathing, one of the fiddle parts, and more are lacking from the score. For the performance, text underlay was also lacking. I added it after the concert based on the singers solutions.


This practice of making the music together is both empowering and limiting. Starting with the latter, it means we cannot easily exchange musicians, especially as we get close to a performance. Musicians with the same skills are hard to come by, so we need each other to perform our music. This mutual dependency is also empowering: we have a common ownership of the music. Rather than an ensemble performing compositions under the direction of a leader, everyone makes his or her part. The lack of scores, and hence the lack of overview, produces a risk in the performance different from that of notation-based musicianship, a risk that I believe is perceptible and forms part of the experience also for the audience. (Again, see the review in Ballade referred to above.) Be it through inspiration and initiative, lack of memory or concentration, mistakes or invasive acoustics: There is always a chance that the music will take an unexpected direction which everyone needs to react to. And if the music sounds boring, bad or uninspired, we have no composers to blame: The process of re-enacting an improvisatory approach to music-making means that the music is ours. The ownership and the risk are shared.

Wheels within Wheels final performance Mouvance I: Mass in St. Marys church, Bergen.