[Smt-talk] Aesthetics of Computer-Generated Music

Martin Braun nombraun at telia.com
Thu Apr 14 08:21:13 PDT 2011

Thanks to David Bashwiner for presenting the study by Steinbeis and Koelsch 

In order to understand the experiment one also needs to know that the music 
stimuli were presented to the subjects, while lying in the scanner, in 
blocks of five. That is they were told: "Now you will hear five pieces that 
were composed by a computer." or "Now you will hear five pieces that were 
composed by a composer."

Therefore it was clear, before anything was measured, that the subjects 
would be wondering about possible intentions of the composers only during 
the composed-by-a-composer blocks.

Generally, it is important to note that this study was not music specific at 
all. For example, a much stronger difference in activation of the putative 
"the-other-person's-mind" network could be expected by presenting two types 
of short emails: type A with a crystal clear intention of the writer (no 
wondering of the reader about the intention of the writer) and type B with 
an ambiguous intention of the writer (much wondering of the reader about the 
intention of the writer).

David Bashwiner generally commented:

"Koelsch's work is among the most musically sophisticated of that coming out
of the field of musical neuroscience."

For those who have no first-hand experience with this field one might add 
that Koelsch's work represents the cognitive faction of the field. Here the 
key terms are syntax, semantics, communication and understanding.

Interestingly, the first word in the title of the Steinbeis and Koelsch 
study "Understanding" is incorrect and should not have been permitted by the 
reviewers. A correct term would have been "Wondering about".

In support of the other side of the "field of musical neuroscience" one 
should add that for many, perhaps most, researchers music is as little 
cognitive as walking, skiing, dancing and whistling.


Martin Braun
Neuroscience of Music
S-671 95 Klässbol
email: nombraun at telia.com
web site: http://www.neuroscience-of-music.se/index.htm

----- Original Message ----- 
From: "David Bashwiner" <david.bashwiner at gmail.com>
To: <smt-talk at societymusictheory.org>
Sent: Wednesday, April 13, 2011 5:25 PM
Subject: Re: [Smt-talk] Aesthetics of Computer-Generated Music

On Sun, Apr 10, 2011 at 11:04 AM, Dmitri Tymoczko 
<dmitri at princeton.edu>wrote:

> It seems to me that it would be very easy to find pairs of pieces, one
> computer-generated and one human composed, where untrained listeners show 
> a
> strong if not overwhelming preference for the computer-generated piece. 
> Not
> sure what if anything to take form that ...

An article by neuroscientists Nikolaus Steinbeis and Stefan Koelsch may be
of interest here [Steinbeis, N; Koelsch, S (2009). *Understanding the
intentions behind man-made products elicits neural activity in areas
dedicated to mental state attribution.* Cerebral Cortex, 19(3):619-623.]. 
authors sought to demonstrate that musical stimuli are perceived differently
when they are believed to be communicative (i.e., composed by humans) than
when they are not. To that end, they presented listeners with short passages
(8–13 seconds in length) from dodecaphonic musical works, telling them that
half had been composed by human beings, the other half by computer. (The
excerpts had in fact all been composed by Schoenberg and Webern.) The
subjects were asked simply to rate the pleasantness of the pieces, and their
brains were scanned during the process (using fMRI). Pleasantness ratings
did not differ across the two conditions, but subjects did report, in a
post-imaging questionnaire, considering more strongly the intentions behind
the pieces they believed to have been composed by humans than those they
believed to be computer-generated. Brain activation differences
corresponding to this attribution of intention were found in a circuit of
areas demonstrated to be involved in the attribution of intentions generally
(not just to musical works): the anterior medial frontal cortex, the
superior temporal sulcus, and the temporal poles. Activation in the first of
these areas correlated the most strongly with reported strength of thinking
about intentions; the last of these, the authors believe, functions as a
storehouse for relevant information about composers and their intentions.

Koelsch's work is among the most musically sophisticated of that coming out
of the field of musical neuroscience. I highly recommend it to the
interested reader: http://www.stefan-koelsch.de/.

One other thought that comes to mind is J. J. Gibson's notion of "display" (
*The Ecological Approach to Visual Perception*, 1986, p. 42). For me to
communicate with you, I can use my hands and voice, etc., or I can write you
a note and leave it on your door. When you see the note (the "display"),
though it itself is inanimate, you recognize it to be communicative in
function, and you attribute intentions to its maker (or whoever you think
its maker is). Often, however, we interpret things that are not in fact
displays (not made by humans) as displays (e.g., signs from God, Cope's
Concerto). Studies like the one above suggest we can also believe true
displays (the music of Schoenberg and Webern) to be not displays if given
the prompt that they may have been composed by computer. What Dmitri appears
to to be wondering is whether believing a display to in fact not be one
might increase one's aesthetic appreciation of it. That's certainly
possible, and it could of course go the opposite way as well (though
Steinbeis and Koelsch found no differences in pleasantness ratings across
the two conditions). What the Steinbeis and Koelsch study gets at is that
the difference between believing something to be a display (i.e., to be
human-made) and believing it not to be one implicates social areas of the
brain, specifically those involved in the attribution of intentions

*David Bashwiner*
Assistant Professor of Music Theory
University of New Mexico
2103 Center for the Arts
(505) 277-4449 

More information about the Smt-talk mailing list