Music is an
ancient and important expression of human creativity, and is an intrinsic part
of culture. It can be argued that creative expression is what makes our species
unique among those on earth, and hence, musical appreciation is something which
is well worth studying.
What makes a series of sounds in a particular sequence more pleasing to the
human ear (and brain which interprets it) than another series of sounds, in a
different order? As any composer will know, only some tunes and melodies really
‘work’, but when they do, they can become so memorable and iconic that a piece
can outlast its composer in collective human memory and be appreciated through the
ages.
Was there something special about Mozart, Handel, and Bach that meant they were
able to write music that would, both figuratively and literally, ‘resonate’ with
their audiences? Does Lady Gaga, and others of her ilk, possess a special
ability to produce tunes that others will like, and that will stick in a person’s
mind?
These, are
questions that Prof. Armand Leroi, and his team at Imperial College, London,
hope to answer. They are taking music to task, and taking an evolutionary biologist's
approach to understanding this most unique form of human creativity. Essentially,
the team have set out to see if music can evolve, by a process of natural
selection- can it be ’composed’ without a ‘composer’?
Leroi and his
colleagues have designed a computer algorithm that uses the principles of
evolution by natural selection to create musical tunes from random ‘noise’
files. First, a computer generates short audio clips, which are essentially
random collections of sounds.
Then, using an interactive website, these ‘noise’
files are presented to listeners (members of the public), who score them for
how ‘good’ they are to listen to. Thousands of members of the public listen to
the files and score them, and the average scores attributed across all
listeners are taken forward. Of the thousands of noise files, only the ten
which score highest overall can ‘reproduce’ themselves in the next generation.
They are subject to random mutation, and can recombine with one another, just as
in a sexual system of genetic evolution.
The next generation are then scored by
the public, and again, only the highest scoring noise files reproduce. This
process repeats for thousands of generations, with listeners scoring each
generation of ‘tunes’.
If the
Darwinian selection process works, we should expect the ‘tunes’ to increase in ‘fitness’.
In this case they should become more
pleasing to listen to, and more melodic, with each generation. As we expect, if we have enough generations, the output of
the iterated process is music; music generated from random noise without
purposeful design. It really does seem to work- I have heard it for myself,
and, while after a 8,700 generations, it is still no Beethoven’s 5th (or even One Direction, for that matter) the audio files are tuneful. Don’t
believe me? You can see for yourself, and even score the tunes before the next
generation at www.darwintunes.org.
This is far
from surprising, really. Anthropologists and the like are very aware of the
phenomenon of ‘cultural evolution’- pleasing cultural artefacts have been
copied and passed down from human to human for centuries, and, undoubtedly, ‘mutation’/
changes are introduced when a mistake is made in the copying process. This
creates variation, and, perhaps only the ‘most successful’ (most appreciated?)
variants are chosen to be copied by others. Hence, painting styles, sculpture,
and, music for that matter, can, conceivably evolve, by a process of natural
selection, or, ‘audience selection’ if you like.
We might speculate that this could go on (at a
rapid rate) in composers’ brains when they write a piece, or, we may even
suggest that there is nothing ‘special’ about composers- natural selection produces
the same thing as they can. Is this ‘audience selection’ really the key to what
is driving the success of particular musicians and composers?
Of course,
when we consider any sort of change through time, we must consider the Price
Equation (a bugbear of my undergraduate study, but wonderful, nonetheless!):
Δ z’ = cov(w/w’,z) + E((w/w’)Δz) - I am in no way an expert on this, but
essentially, it splits evolutionary change of a character/ trait into two essential parts- the change due to
the pressure of selection, and the change due to transmission.
A caveat
with cultural evolution is that the change due to transmission (E((w/w’)Δz)
)- i.e. the change in a character (or piece of music, in this case) that is not
due to natural selection cannot be ignored. This component is just too large
here to usefully do so. Hence, we cannot say for certain how important ‘audience
selection’ might be in cultural evolution- music produced by different cultures
may have diverged as a result of simply being transmitted, as opposed to there
being active selection according to what sounds good to one group of people
compared with another.
What is clear is that music, can be ‘evolved’ under controlled
conditions. This provides a demonstration of natural selection working over a
short time, that even a non-expert can understand, and get involved in. While I
doubt Prof. Leroi will be getting a hit record any time soon, I have to say, I’ve
definitely heard worse!
