What is a great musician?

wiebe skole iets lichterIn The remains of the day by Kazuo Ishiguro, the butler, Mr. Stevens, reflects upon a fundamental question. 

"But let me return to the question that is of genuine interest, this question we so enjoyed debating when our evenings were not spoiled by chatter from those who lacked any fundamental understanding of the profession; that is to say, the question "What is a great butler?"

We can easily replace "great butler" for "great musician" or "musical performance" and then we have questions that can be debated in a similar manner for many many evenings.

What is a musical performance?

One of the things we regularly do in my lessons (contextual studies, music philosophy and aesthetics, ...) is listening to and comparing performances.

This is really fun, but also difficult if you want to go beyond quick judgement and prejudice.... It proofs to be  difficult to postpone a judgement, and just stick to factual observations, and use these in a later stage to arrive at an evaluation. Such a "socratic" approach is often really rewarding, even if it takes some time. 

In the case of "factual observations" we are talking about things like: comparing tempi, fluctuations in tempo, rhythmical freedom, use of dynamics, articulation, transparency, details, phrasing, articulation of form, proportionality in the use of musical parameters, and so on.

Let me give an example. In Episode 001 I mentioned three performances of the Chopin Etude Op. 10 nr. 3.

This is the Spotify playlist with performances by Murray Perahia, Amir Katz and Lang Lang. Just take some time to listen to all three, put your smartphone away (you will survive 11 minutes without a smartphone, I promise).

Just to get some cues I made a project in Audacity to be able to compare the performances: three tracks, three performances. In label tracks I put the beginning of all sections.

Tristesse comparison waveforms and proportions

Some general observations about length (in seconds) of each performance:

  Total  A' 
 Perahia  232 79  96 67 
 Katz 184 58 76  50 
 Lang 286 106  83 97





The first impression of the starting tempo: Katz is considerable faster than Perahia, Lang is (much) slower than Perahia. 

The tempo of the central section seems to be closer in all three performances, but Perahia is (even) slower than Lang, and Katz and Lang differ just 7 seconds, while in the first section Lang takes almost the double amount of time in comparison to Katz...

Also the total length of the performances differs considerably. Between the shortest and longest 102 seconds (!), which is substantial if we take the total length of the composition into account. 

As mentioned before, if we compare the central sections, the lengths are clearly much closer to each other than in the outer sections. It seems obvious that  the performers differ in their views with regard to proportion and balance in the form and form sections.

And the interesting question is of course: What kind of arguments can be found to underpin a performance of this etude? There is of course the nickname "Tristesse" which has influenced  the choice of many performers and publishers (!) for a relatively slow tempo. The "vivace" in the manuscript should have made them think...

There is the question of "rubato", of course: at certain moments in Langs performance it's difficult or even impossible to hear a time-signature...

These are just a few observations. 

Of course we can observe the fluctuations in tempo, and put them into words, but wouldn't it be nice to be able to measure them? And the same for dynamics and other characteristics? Of course that would be nice.  But is it possible? Yes, it is. Daniel Levitin conducted really interesting research on this topic.

Measuring musicality (or: how musicians deal with musical parameters)

We all know that musical sound has a limited amount of parameters: pitch (linear/vertical), dynamics, metre/rhythm, articulation and timbre/colour. We can use these to describe what we hear. (See my working model "Describing musical sound").

Take a look at two short videos of Daniel Levitin searching for an answer on the question:

What makes a performance "musical"? (The title of the video is a little bit misleading).

He explains his research by conducting experiments with controlled manipulation of musical parameters in quality perfomances with the help of special adapted Yamaha pianos.

Levitin video screenshot

video 1: https://youtu.be/CJMwWX8WX3o

video 2: https://youtu.be/k4--Pq0bci4

Nowadays we have a heritage of more than a century of recorded music. Wouldn't it be interesting to look for trends and patterns in performing history? Yes, but it's a huge amount of recordings. So we would need computers and smart software to help us. Tadaa! This is where Music Information Retrieval enters the stage.

The next level: Music Information Retrieval (MIR)

It's possible to measure characteristics of recorded performances of music. And that on a big scale. Take a look on the following site:



Browse their website and Research Projects on:

Take a look at the tool Sonic Visualiser.

And the site of Queen Mary University of London

That's it for now.