Engineering and Music
"Human Supervision and Control in Engineering and Music"

Orchestra Concert
Ensemble Concert
About us

Guerino Mazzola 

Music Performance and Interpretation

Performance is an substantial component of the musical work, adding the physical reality to the "fictitious" score symbols and the power of sensorial evidence to the interpretative reflection of analysis. We discuss theory and software for interpretation and its performance. 
What is Performance? 
Performance is the physical execution of a work of art. This work may be a written poem, a musical score or a sculpture (a sculpture is performed while you walk around it and interact with its perspectives and its presence in space). This terminology subtends that performance adds an essential part to the somewhat abstract or symbolic "text". To put it with Paul Valéry: "C'est l'execution du poème qui est le poème." The point of performance is that its added value depends on the way of executing the symbolic data, performance is a rhetoric category. In performance, we communicate the work's contents, we make clear what could, to our mind, be the signification of the symbolic score. 

So why is performance essential? Couldn't we just stick to the abstract analysis of a work and try to understand it on the level of reflection? There are at least two reasons why this is not sufficient. First, the existential level of physical execution is different from the mental level of score symbols: Reading a recipe never replaces its cooking and consumption. Second, the compact sensual presentation of a fact has a degree of evidence which cannot be paralleled by intellectual meditation. After all, geometric visualization is an essential extension of the abstract mathematical concept framework and enjoys the well-known power of evidence. In a first approach, performance is to the fine arts what is geometric visualization to mathematics.

Performance is a very complex, global phenomenon [Mazzola 2002]. It involves four globalization factors: (1) instrumental differentiation, (2) splitting the score into voices, periods, and similar groupings, (3) specifying the roles of the musical parameters (pitch, duration, onset, loudness, etc.) into basic parameters, such as onset time, vs. dependent parameters, such as duration (depending on onset data),  and (4) unfolding the level of detail and sophistication of a performance as a function of its rehearsal history in the spirit of a genealogical tree.

Rhetorically speaking, each performance is a perspective view upon a complex object, and its understanding results from a sum of varied perspectives. If you perform a sculpture, you walk around it and integrate the various perspectives. This is the most common variant of the famous Yoneda lemma in mathematical category theory [Mac Lane1998]: Classification means integration of morphisms. This also suggests that the famous unicorn of hermeneutics: the ideal performance,  does not exist. On the contrary, a work needs an infinity of performances to comprise its complete understanding. Summarizing, a concrete performance is one of an infinity of expressions of the hidden and ambiguous contents of a given work. 

Rationales of Performance
Performance theory splits into two main concerns: structure theory and semantic theory. Whereas the former deals with the precise description of a given performance, the latter tries to understand the rationales of a given performances, i.e., why a performance is produced in the way it appears, for what reason an artist shapes his/her execution in one way and not in another one.

There are three main rationales for performance: emotion, gesture, and ratio. The first one has strongly been preconized by Alf Gabrielsson. He maintains that "we may consider emotion, motion and music as being isomorphic" [Gabrielsson1995].  While this conjecture may please psychologists, it is completely useless to scientific investigation. In fact, such an isomorphism is  a piece of poetic literature as long as the components: emotions, gestures, and music, are not described in a way to make this claim verifiable. Presently, there is no hope for a realistic and exhaustive description of emotions. Same for gestures, and as to music, the mathematical categories of local and global musical objects are so incredibly complicated that the mere claim sounds like a cynical joke. For example, the number of isomorphism classes of 72-element motives in pitch and onset (modulo octave and onset period) is 2.23.10^36 [Fripertinger1993]. How could the claimed isomorphism fit in this virtually infinite arsenal?

Gestural categories as a rationale for performance have been advanced in approaches [Kronmann1987] which maintain that musical retards, for example, share a structure of Newtonian mechanics. Such approaches cannot, however, explain the agogic phenomena within a motivic movement, or the dynamical differentiation within a chord, for example. Moreover, the gestural motivation for a determined instance of performance is extremely complex: How could one deduce Glenn Gould's performane when knowing his beautiful dance of fingers, arms, and body?

This is why we shall stick to rational semantics in performance, it is the easiest and most explicit rationale. This means that we have to investigate the score text by means of metrical, rhythmical, motivic, harmonic, contrapuntal etc. analyses and to correlate these findings to the expressive shaping of performance. This is also a traditional and important requirement of rhetorics: to convey the text's meaning, and not personal emotions or gestures. Theodor W. Adorno has strongly recommended such an analytical performance approach [Adorno1963]. It is an interesting question, whether traditional performances have much to do with analytical performance, and if not, how such a performance would sound like! We shall give an example of such a performance in this talk.

Interpretation of Interpretation
Rational performance is based upon music analysis which is susceptible of quantitative representation since, in the last instance, the artist has to hit the keys or pluck the strings in a numerically meaningful way. We have implemented a number of such analyses in the performance platform Rubato® (the dynamically loadable modules Metro-, Melo-, and HarmoRubette®) [Mazzola1994b]. The principle of these implementations has been to produce numerical weight functions which account for the analytical relevance of score events; a metrical weight function will be shown below.

The primary goal of rational performance is to express such analytical data by means of agogics, dynamics, articulation, and intonation. (In a more sophisticated stage of performance theory, the expression on the level of artistic performance gestures must be addressed.) Usually, the analytical information is too complex to be mapped in a one-to-one way onto the shaping of performance, which means that even for a fixed bunch of analyses, a number of performances has to be executed to allow a faithful representation of these analyses.

Adorno and Performance Fields
The core of performance structure theory is the description of performance transformations by means of performance vector fields. Such fields generalize the well-known tempo curves (which are one-dimensional fields [Mazzola1994a]) and arise from Jacobian matrices associated with diffeomorphisms from the symbolic parameter spaces to their physical counterparts. This machinery enables performance shaping of infinitesimal precision. This differential-geometric approach is by no means a mathematical overhead, it rightly fulfills Adorno's and Walter Benjamin's method of "micrological performance" which asks for "infinite interpolation" [Adorno1963]. This theory has been modeled and implemented in Rubato's PerformanceRubette.
Celibidache's Initial Magic
Performance fields can only produce performances if initial performance values are available. For the tempo field this means that the physical onset corresponding to the score's  initial onset must be defined.  Initial performance is a dramatic phenomenon since it specifies the point where the fictitious reality of the score is anchored in the physical reality of performance. After this anchorage, potentiality is suspended, and the performance field will take care of the physical coordinates associated with symbolic events. Semiotically speaking [Mazzola1998], initial performance is a shifter, it relies the fictitious symbolic work to its full-fledged realization by the sign's user: the artist, the conductor, the orchestra. This is why Sergiu Celebidache so violently insisted on the uniqueness of performance, which germinates on the magic of performance initialization [Schmidt1992].
Local Hierarchies
Not all music parameters play the same role in performance. For example, performance of durations is determined by performance of onsets. In the default setup, the performed duration is the difference between the performed onset of the event's end and the performed onset of the event's beginning. This type of dependencies entails hierarchies of parameters, and therefore, hierarchies of performance fields. Rational performance deals with the calculation of such field hierarchies by use of analytical weight functions.
Genealogical Trees of Rehearsal
The fourth global component of performance stems from the fact that performance is never a one-step process. Artists have to practice, to rehearse, to develop their valid performance through a cascade of intermediate performances, starting with the mechanical prima vista performance on the given score. This is formalized in the so-called stemma. This is a genealogical tree, starting from the mechanical prima vista performance, and successively splitting into performances of the score's parts (voices, periods, etc.) and generated from the previous performances by specific performance operators which are loaded with selected weight functions. For example, a child of a mother performance may be defined by the application of a metrical weight to the mother's tempo curve. It can be shown that the known shaping operators can all be deduced from a unified procedure involving the Lie-derivative of a weight along the mother's performance field. However, a general classification of shaping operators has not been settled to this date. 
Sundberg's Performance Grammars
The way how a given bunch of analytical weights acts on the successive refinement of performance on the stemma has been termed "performance grammar" by Johan Sundberg, one of the fathers of modern performance theory [Friberg1991]. The yoga of performance grammars is this: It could appear as if performance theory would just be interested in the production and simulation of "good" performances. But this is not very interesting. If we listen to a performance, we are not seeking to understand this particular result, but we want to recognize a possible system which leads to the present output. If a specific performance can be produces without any systematic insight, it is useless. Performance research is the search for systems which generate interesting performances. This is what a grammar should look like. It is a scientific requirement: The "experiment" should be reproducible, not just a matter of chance.
Experiments with Bach
In his habilitation [Stange1999], Joachim Stange-Elbe has investigated metrical and melodic weights for Bach's "Kunst der Fuge". He has studied the possibilities to generate interesting performances of this composition by use of performance operators which are implemented in Rubato's PerformanceRubette. Stange-Elbe [Stange1998] first tried the target-driven strategy: Use the given weights and try to produce a performance which should sound like what you expect from experience and tradition. This strategy was much less successful than the source-driven approach: try to just express the given weights without any predefined target. The latter has in fact led to very interesting performances which we will present in our lecture. Perhaps, this type of performance was one of the first in history which was uniquely driven by analytical considerations.
Inverse Performance
Control of performance need not restrict to its productive perspective, it is also important to control given performances, i.e., to understand the mechanisms which could lead to such a performance. This is the subject of inverse performance theory. To begin with, such a theory has to offer means for calculating the final performance fields of a supposed stemma (the fields at the stemma's leaves). In his ongoing PhD thesis, Stefan Müller has implemented the EspressoRubette which calculates all the performance fields of a MIDI-formatted recording with respect to the given score file (also in MIDI format) [Müller2002]. Given such an output of performance fields, we may consider all possible parameters which yield this output, i.e., all possible weights, stemma parameters and performance operator parameters.  In a special case of such a background parameter system, the locally linear grammars, Roberto Ferretti has calculated the algebraic varieties of parameters which yield a fixed given output. It turns out [Ferretti2002] that generically, such varieties are isomorphic, but under reasonable restrictions, the parameter varieties help distinguishing different performances on the level of algebro-geometric structures. For instance, this approach reveals more global coherence in Martha Argerich's performance of Schumann's famous "Träumerei" than in Vladimir Horowitz's performance [Mazzola1993]. A statistical approach to inverse performance theory is exposed in [Beran2000]. 

Adorno, Th. W. (1976). Der getreue Korrepetitor (1963). Gesammelte Schriften, Bd. 15, Suhrkamp, Frankfurt am Main.

Beran, J., Mazzola. G. (2000). Timing Microstructure in Schumann's "Träumerei'' as an Expression of Harmony, Rhythm, and Motivic Structure in Music Performance. Computers and Mathematics with Applications, Vol. 39, Issue 5/6, pp. 99-130.

Ferretti, R., Mazzola, G.  (2002). Algebraic Varieties of Musical Performances. To appear in: Haluska, J. (ed.): Music and Mathematics. Bratislava: Tatra Mountains Mathematical Publications.

Friberg A. (1991). Generative Rules for Music Performance: A Formal Description of a Rule System. Computer Music Journal, Vol. 15, No. 2, 1991.

Fripertinger, H. (1993). Untersuchungen über die Anzahl verschiedener Intervalle, Akkorde, Tonreihen und anderer musikalischer Objekte in n-Ton Musik. Magisterarbeit, Hochschule für Musik und Darstellende Kunst, Graz.

Gabrielsson A. (1995): Expressive Intention and Performance. In: Steinberg, R. (ed.): Music and the Mind Machine. Springer, Berlin et al.

Kronomann, U., Sundberg J. (1987). Is the Musical Ritard an Allusion to Physical Motion? In: Gabrielsson, A. (ed.): Action and Perception in Rhythm and Music. Royal Swedish Academy of Music No.55.

Mac Lane, S. (1998). Categories for the Working Mathematician. Springer, New York et al.

Mazzola, G., Zahorka O. (1993-1995). Geometry and Logic of Musical Performance I, II, III. SNSF Research Reports (469 pp.), Universität Zürich.

Mazzola, G.,  Zahorka O. (1994a). Tempo Curves Revisited: Hierarchies of Performance Fields. Computer Music Journal 18, No. 1.

Mazzola, G., Zahorka O. (1994b). The RUBATO Performance Workstation on NeXTSTEP. In: ICMA (ed.): Proceedings of the ICMC 94, San Francisco.

Mazzola, G. (1998). Semiotic Aspects of Musicology: Semiotics of Music. In: Posner, R. et al. (eds.): A Handbook on the Sign-Theoretic Foundations of Nature and Culture. Walter de Gruyter, Berlin and New York.

Mazzola, G., et al. (2002). The Topos of Music—Geometric Logic of Concepts, Theory, and Performance.  To appear (1200 pp.), Birkhäuser, Basel/Boston.

Müller, S. (2002). The EspressoRubette: Visualization of Expressive Performance. To appear in the proceedings of the "Second International Seminar on Mathematical Music Theory", i2musics online publ., Berlin/Mexico City/Zürich.

Schmidt, J. (ed.) (1992). Celibicache. Book and Movie PARS, München.

Stange-Elbe, J., Mazzola, G. (1998). Cooking a Canon with Rubato—Performance Aspects of J.S. Bach's "Kunst der Fuge". In: ICMA (ed.): Proceedings of the ICMC 98,  pp.179-186, San Francisco.

Stange-Elbe, J. (1999). Analyse- und Interpretationsperspektiven zu Johann Sebastian Bachs "Kunst der Fuge" mit Werkzeugen der objektorientierten Informationstechnologie. Habilitation, Universität Osnabrück.