7/28/2019 Fitness Landscape of Music (2012) - Christoph Adami

1/2

Adaptive walks on the fitness landscape of musicChristoph Adami1

Department of Microbiology and Molecular Genetics and BEACON Center for the Study of Evolution in Action, Michigan StateUniversity, East Lansing, MI 48824

T

he evolution of musical abilitiesin man was an area of intenseinterest for Charles Darwin: inhis bookThe Descent of Man, and

Selection in Relation to Sex (1) he discussedthe emergence of voice and musical pow-ers in terms of sexual selection and madethe case that the vocal organs were pri-marily used and perfected in relation tothe propagation of the species (p. 315).

Although music has been performed bypeople at least as far back as the Paleo-lithic (2), musical styles have changeddramatically throughout the ages and dif-fer significantly across geographic regions.We could therefore ask whether musical

preferences change according to laws thatare similar to those of biological evolution,namely via inheritance, selection, and

variations of tunes that are induced bychance. In PNAS, MacCallum et al. (3)present an experimental approach to studyhow music evolves, by encoding soundsinto programs that can evolve via randommutation and recombination. Using com-puter programs to study evolution is ofcourse not new (4), but in this case selec-tion was at the hands of people, who lis-tened to the musical phenotype of theseprograms over the Internet and scoredthem according to their personal prefer-ence. Within a population ofloopingpolyphonic sound sequencesor loops,for shortthose sequences that generatedhigher ratings by users were given moreoffspring, and their genetic materials re-combined. In this manner, the prevailingmusic in the population changed overmany generations. However, did the musicget any better? To assess this, the authorssample loops from the entire evolutionarytrajectory and ask a different set of con-sumers (not those who selected the loopsin the original experiment) to rate thequality of the music. This average rating

the musical appeal M

increased signifi-cantly over the first 600 generations, butthen seemed to flatten out. Over the re-maining 2,100 generations of the experi-ment, the appeal of the loops remainedessentiallyflat. What happened?

In evolutionary biology, we are used topatterns of adaptation that are consistent

with rapid change followed by stasis (5).A handful of different causes are usuallyadvanced to explain the slowdown. Whena population adapts to a new niche, it canbe thought of as climbing a fitness peak,using the metaphor of the fitness land-scape introduced by the mathematical ge-

neticist Sewall Wright. If there is onlya single peak (or if higher peaks are un-reachable from that location), then adap-tation must stop when the peak is reached:every step upward leads to diminishingreturns (6, 7). However, there are otherreasons why adaptation may slow downor stop. One of them is particularly in-teresting from the point of view of geneticsand may be at the root of the stasis ob-served in this experiment. Although re-combination of genetic material is a great

way to create novelty, it is an equallygood way to destroy sets of genes that arefinely tuned to work with each other, byripping them apart in the recombinationprocess. The loops do not have well-defined sets of genes, but individual traitscan be examined. For example, the au-thors use two traits that are commonlyused to classify music genresthe clarityof the chords CL and the complexity of therhythmic signature Rto assess the fitnessof each loop. Both traits can be measuredobjectively (and without error) for eachloop, and even though they explain onlya small percentage of the overall musical

appeal, they correlate with it. Moreover,both CL and R increase within the first 600generations, when the appeal is also in-creasing. Importantly, the authors showthat, in control experiments in which loopsevolved in the absence of selection, bothmeasures remain at low levels. These twotraits now can be used to parameterize thefitness landscape of music in this experi-ment. Interestingly, the traits are not in-dependent: rather, the topology of thelandscape implies that they interact syn-ergistically; that is, loops that have a high

value in one trait are especiallyfit if theyalso have a high value of the other. In

genetics, such interacting traits are calledepistatic, and they are at the very heartof everything that makes evolution inter-esting. In particular, it has been hypothe-sized that synergistic epistasis betweengenes is necessary for the evolution ofsexual recombination to begin with (8).Indeed, according to that theory, thesynergy between deleterious mutationsensures that they do not accumulate inthe genome, and therefore keep themutational load of a population low.However, a low mutational load implies

a reduced genetic variance, which leadsto reduced adaptive potential viaFishers Fundamental Theorem ofevolution (9). Thus, sexual recombinationmay be detrimental to the future adaptivepotential if the synergistic epistasis istoo strong.

So, are the loops evolving toward anepistatic conundrum between traits, ordid they simply run out of beneficial mu-tations at the top of a peak? To test this,the authors use a description of evolu-tionary processes that is more general thanFishers Fundamental Theorem, namely

Prices equation (10), and that is able todisentangle the effects of variation, and

the strength of inheritance, on evolution.An analysis of Prices equation shows thatthe slowdown of the furious rate of adap-tation during the early 600 generations isdue mainly to the increased productionof inferior types via recombination, an ef-fect that is expected if traits interact

Fig. 1. An adaptive walk on a rugged landscape defined by the two traits CL and R, beginning at the

blue circle and ending at the black circle. The height of the surface is given by the replication rate of a

loop with a given set of traits (the landscape depicted is an idealization of the landscape in ref. 3).

Author contributions: C.A. wrote the paper.

The author declares no conflict of interest.

See companion article on page 12081.

1E-mail: adami@msu.edu.

1189811899 | PNAS | July 24, 2012 | vol. 109 | no. 30 www.pnas.org/cgi/doi/10.1073/pnas.1209301109

mailto:adami@msu.edumailto:adami@msu.edu

7/28/2019 Fitness Landscape of Music (2012) - Christoph Adami

2/2

synergistically. Over the whole evolu-tionary history, however, the authorsobserve a trend of reduced varianceinstead, which is a consequence of syn-ergistic epistasis under recombination.Indeed, a visualization of the adaptive

walk on the landscape spanned by thetraits (Fig. 1) shows that the evolutionaryprocess fails to locate the highest peak

in the landscape. The same dynamicsare observed in control experimentsperformed with undergraduate studentsas the selection force rather than over theInternet, suggesting that the observationis germane.

Although the fitness landscape meta-phor necessarily abstracts the full evolu-tionary dynamics by focusing on a few traitsonly (usually two so that the landscape canbe visualized), it is a valuable tool not onlyto gain a more intuitive understanding ofthe past unfolding of evolution but also tomake quantitative predictions for the fu-ture of evolution. Using tools similar tothose used here to reconstruct the fitnesslandscape from actual fitness data, re-searchers recently reconstructed the fitness

landscape for the evolution of drug resis-tance in the HIV and showed that epistasisbetween mutations on that landscape is

MacCallum et al. present

an experimental

approach to study how

music evolves.

strong and affects recombination betweenthe HIV genes (11). Recombination isa valuable tool that evolution uses whenthe landscape changes often, as is truefor certain viruses, notably HIV. Forthe musical loops, the environment isfairly constant, so perhaps an asexualmodel of replication would be bettersuited and might locate the peak ofthis landscape.

Do the experiments conducted byMacCallum et al. tell us anything abouthow music actually evolved in humansocieties? Certainly, music does not

emerge from a random process only:composers actively create music, albeitinfluenced by existing music within theirculture. Perhaps this influence can bethought of, remotely, as variation andrecombination of existing elements. Thesuccess of a tune, no doubt, dependson a process akin to selection by anaudience, and music spreads via a processsimilar to replication when performerslearn (and thus copy) a new piece.However, maybe the most importantrole of music in the history of our speciesis precisely the one that interestedDarwin the most: with music we areable to encode, and therefore transmit,complex and deep emotions. Forempathic animals such as ourselves,being able to compose or performmusic may indeed help in propagatingthose genes.

ACKNOWLEDGMENTS. I thank B. stman for

Fig. 1. This work was supported in part by theNational Science Foundations BEACON Center forthe Study of Evolution in Action, under Coopera-tive Agreement DBI-0939454.

1. Darwin C (1871) The Descent of Man, and Selection in

Relation to Sex (D. Appleton, New York), 1st Ed.

2. Montagu J (2004) How old is music? Galpin Soc J 57:

171182.

3. MacCallum RM, Mauch M, Burt A, Leroi AM (2012)

Evolution of music by public choice. Proc Natl Acad

Sci USA 109:1208112086.

4. Adami C (2006) Digital genetics: Unravelling the ge-

netic basis of evolution. Nat Rev Genet 7:109118.

5. Fitch WM, Ayala F, eds (2005) Tempo and Mode in

Evolution: Genetics and Paleontology 50 Years After

Simpson (National Academies, Washington, DC).

6. Khan AI, Dinh DM, Schneider D, Lenski RE, Cooper TF

(2011) Negative epistasis between beneficial mutations in

an evolving bacterial population. Science 332:11931196.

7. Chou HH, Chiu HC, Delaney NF, Segr D, Marx CJ (2011)

Diminishing returns epistasis among beneficial muta-

tions decelerates adaptation. Science 332:11901192.

8. Kondrashov AS (1994) The asexual ploidy cycle and the

origin of sex. Nature 370:213216.

9. Fisher RA (1930) The Genetical Theory of Natural Se-

lection (Clarendon, Oxford).

10. Price GR (1972) Fishers fundamental theorem made

clear. Ann Hum Genet 36:129140.

11. Hinkley T, et al. (2011) A systems analysis of mutational

effects in HIV-1 protease and reverse transcriptase.

Nat Genet 43:487489.

Adami PNAS | July 24, 2012 | vol. 109 | no. 30 | 11899

COMMENTARY