Skip to content

Ravel’s Bolero

Ravel’s Bolero in images

Basic rythm of Bolero

Ravel’s Bolero has a fairly readable structure built around a central theme (A) and a counter-theme (B), both in 16 bars, each repeated 9 times and a final section. The whole piece is punctuated by regular beats that give it a repetitive, and even obsessive character. There is an overall crescendo which leads to the final collapse in the last section. This structure lends the Bolero surely more easily than other works to a visual representation. Many interpretations of Ravel’s work have been given : the wellknown ballets, paintings, sculptures. This section compares some of them with what give more quantitative methods of analysis and particularly with a musical strip that I realized by a mapping with a set of abstract paintings.

Béjart

The ballet of Béjart is perhaps the best known interpretation of the Bolero. I give here some images taken from a video of the ballet of the 20th century (https://www.youtube.com/watch?v=tub81Sb-n4w). The dance is organized around a central figure – Jorge Donn. His quasi mechanical moves, light and elegant, follow the melody. The movements gradually become very ample. Jorge Donn is surrounded by a circle of 40 dancers who rise in turn and become more and more present as the music rises. The ballet ends with a collapse of all the dancers on the central circle. The relationship with the structure of the piece is obvious and easy to read.

The Bolero by the ballet of the 20th century by Maurice Béjart
The Bolero by Maurice Béjart’s ballet of the 20th century

Anne Adams

In 1994, Anne Adams, a Canadian biologist, gave a visual representation of the Bolero, called “Unraveling Bolero”. In her painting, the musical piece is represented by 17 lines of 20 rectangles corresponding to its 340 measures. Height corresponds to volume, color to pitch. The transformation of the musical structure into visual ones is clear. The general crescendo becomes an increase in the size of the rectangles, the final collapse a brutal break in the design of the figures.

Unravelling Bolero by Ann Adams

Unraveling  Bolero by Anne Adams


One of the interests of this painting is to make explicit the overall structure of the work, something more difficult to perceive with the ears. Space replaces time in a way.

Catherine Lhoir

Le Bolero by Catherine Lhoir

Le Boléro by Catherine Lhoir

Catherine Lhoir paintings and artistic creations are highly influenced by music and this since the early 90s; her artistic production is wide ranging from enamels to design, through painting and sculpture.  Generally, her paintings and sculptures reflect a general impression created by music without the will to stick closely to the internal structure of a particular piece, except for her Bolero. Catherine Lhoir here – like Anne Adams – relies heavily on the insistent repetition of the theme to give rhythm to her work. In the silhouettes we find the elegance of the music and in their alignment the recall of the repetitive theme of Bolero. The crescendo is very apparent in the increasingly sophisticated character of the characters. The final fall is represented by the move of the trumpet of the penultimate figure, falling with the last silhouette. Catherine even suggested to partially bury the last characters to better create this feeling of final collapse. The inspiration is obviously totally different from that of Anne Adams and the link with the score is slightly less strict.

Mireille Berrard

Mireille Berrard was a French painter (1930 -2005). From the 1980s, Mireille Berrard created works that translate into abstract paintings musical pieces. Olivier Messiaen, a composer also interested in the correspondence between sounds and images, said  that she produced  “graphical equivalences between sounds and colors”. Her Ravel bolero is reproduced in the figure below.

Le Bolero by Mireille Berrard
Le Boléro by Mireille Berrard

What is there in common with previous works? At first glance, relatively little. Circular patterns where Anne Adams and Catherine Lhoir use more linear representations. A large red spot on top of the painting that breaks the progression of the little colorful and elegant twisted patterns, while the other two artists kept the same increasing motif to represent the repetitive theme of the Bolero. On the other hand, similarities are also apparent: a modular design consisting of a succession of small red and brown patterns, a crescendo movement rendered by the magnification of the motifs and the final fall with the explosion of the red at the top of the painting.

Analysis of the Bolero by quantitative methods

A more literal – but fairly rough – visual transcription of the Bolero is also possible: spectograms elaborated from the MEL coefficients. To make it simple, let us say that the coefficients extracted from the different frames taken in each section of the Bolero are used to make an image. Here is the result obtained by this method using Segments 1, 4, 8, 12, 16 and 19 of Bolero (each segment corresponds to a section). We see in this visualization the progressive densification of the patterns with horizontal bars more and more marked which corresponds to the general crescendo of the piece. But this graph is essentially a representation of spect ra, the bands corresponding to frequencies (based on the initial frequencies of the sounds but not identical)  and the gray levels at intensities, as defined by the so-called Mel Frequency Cepstral Coefficients (MFCC).

Spectorgram of the Bolero elaboraed with MFCC
Representation of sections 1, 4, 8, 12, 16 et 19 of the Bolero

It is possible to make this representation a little less austere by giving it  a “style”. Software that uses artificial intelligence techniques (here deep neural networks) makes it possible  to combine content and style of images. In the figure below, I mixed the content of the “spectra” (rotated by 90 degrees and calculated with slightly different parameters) , with the  sky painted by Van Gogh in his famous “starry night. The application used is “Ostogram” developed from an algorithm proposed by Leon A. Gatys et al and implemented by Justin Johnson in 2015 (https://github.com/jcjohnson/neural-style).


Images derived from MFCC extracted from the Bolero and transformed by Ostogram
 Representation of sections 1, 4, 8, 12, 16 eand 19 of the Bolero with a Van Gogh (produced with Ostogram)

It is of course possible to go beyond these very syntactic representations. An initial analysis of the 19 sections of the Bolero provides the graph below. In order to obtain this figure, each section was characterized by 840 MFCCs, that is to say 840 diemnsions, and then projected into a plane by a statistical analysis called principal component analysis. This transformation, from high dimensional space to a plan, results of course, in a loss of information, here 53% of the general dispersion of the initial points representing the sections of the Bolero.

Projection in a plan of the 19 sections of Bolero
Representation of 19 sections (from Bo1 to Bo19) of the Bolero  in the space of the first two principal components


How to explain the outlying positions of the fourth and eighth sections ? This is due, on the one hand, to the high sensitivity to the timbre of the cepstral coefficients, but also to the deformations introduced by the projection in a plane. Section 4 is played by a small clarinet with a high tone and section 8 by a sopranino saxophone. The specificity of these two sections disappears when we consider the distances in the initial space with 840 dimensions. The interpretation that is analyzed here is given by the “Radio Bratislava Symphony Orchestra” directed by Ondrej Lenard. The projection of the points representing the sections clearly shows the general crescendo with a strong dispersion of the points along the first axis. The particular character of Section 19 is not highlighted by the analysis; it seems to be diluted in the general amplification introduced by the trumpets and the trombones in the last sections (Bo17 to Bo18).  


How to find visual equivalents of these points? Which images to choose? To make a parallel with the works of the three artists presente dabove,  I decided to start with 51 abstract paintings by Daniel Seret, already used to characterize the piece “Crescendo” in another section (“Experiment Musical Strips” under the section “Experiments”) of the site.

Paintings by Daniel Seret used to define a musical strip
Paintigs of Daniel Seret which have been used to define a visual strip for the Bolero



The optimal  representation as obtained with my mapping device is given below. Its quality – measured with a quality indicator I have defiend  here – is fairly high (95%). In the msucial strrip, some drawings have been rotated when necessary to keep the shape of the rectangles identical in the different sections.

Musical strip built for the Bolero
Musical strip elaborated from the 51 paintings by Daniel Seret to represent the 19 sections of Bolero

The general crescendo is visible in this representation of Bolero as the repetitive character of the theme. Notice that sections 4 and 8 have lost their marginal status, resulting, among other things, from a deformation introduced by the projection into a plane.