2.3. Transformations
The principle of applying transformations is standard in music. Typical examples
used in tradional music are shift (e.g. repetition, transposition), shear (e.g.
arpeggio), reflection (e.g. retrograde), dilatation (e.g. augmentation). In the 20th
century, other transformations such as exchange of coordinates and rotation were
also applied by some composers (e.g. Boulez, Kagel, Eimert, Stockhausen; also
Messiaen).
The transformations that have been used by composers (in western music)
throughout the centuries can be described in a unified manner by the mathematical
definition of affine transformations. These are defined by f(x) = g(x) + a (a is
a vector) with g linear, i.e. g(x + y) = g(x) + g(y) and g(bx) = bg(x) (b is a
scalar).
2.4. Hierarchy
Global and local structures are often shaped by starting with a global structure, then
refining local structures. This is the case for the compositional process as well as for the
performance (and rehearsal) of a composition. Mathematically, this can be formalized,
for instance, by defining different local coverings (subsets) of a global composition
(Mazzola 1990) and by hierarchical smoothing models (Beran and Mazzola 1999a,b,
Mazzola and Beran 1997).
3. Analysis: Musical analysis of historic compositions and performance
An important way of validating the relevance of mathematical models for music is to
analyze existing compositions and performances. For the theory mentioned above,
empirical studies are reported in Beran and Mazzola (1999a,b, 2000) and Mazzola and
Beran (1997). Scores by Bach, Schumann and Webern were analyzed using the RUBATO
software (Mazzola and Zahorka 1993–1995, 1996). RUBATO is based on an extension of
the mathematical music theory described in Mazzola (1990). The results revealed
interesting features of the compositions that were not obvious before the analysis.
Moreover, for one of the scores (Träumerei by Schumann), the numeric data obtained
from the analysis were compared to tempo curves from 28 historic performances by 25
pianists (Horowitz and Cortot were represented by three performances each). Using
so-called HISMOOTH models (hierarchical smoothing models), common features as
well stylistic differences and clusters of performers could be identified, and
explained by structures in the score. In particular, the following results were
obtained:
- A “Horowitz-Cluster” could be identified including Klien, Brendel and all three
performances by Horowitz. In spite of a time span of several decades, the three
performances by Horowitz were very similar to each other.
- A “Cortot-Cluster” could be identified including Argerich, Capova, Demus, Kubalek,
Shelley and all three Cortot performances. The same remark as above applies to
the three performances by Cortot.