Evolutionary games on the computer

Originally published 14 April 1986

Those who oppose evo­lu­tion often claim that the the­o­ry is not “sci­en­tif­ic.” They say that no hypoth­e­sis can qual­i­fy as sci­ence unless it can be test­ed by a con­trolled experiment.

Evo­lu­tion­ary hap­pen­ings, say the objec­tors, are unique, unre­peat­able, and irre­versible. Evo­lu­tion­ists may claim that fish­es turned into amphib­ians and rep­tiles into birds, but there is no way the the­o­ry can be put to the exper­i­men­tal test.

As Stephen Jay Gould has point­ed out, the objec­tion aris­es from a wrong notion of sci­en­tif­ic method. There is no such thing as “the sci­en­tif­ic method.” There are many meth­ods by which sci­ence seeks truth, and one of them is the his­tor­i­cal method.

Like the the­o­ry of con­ti­nen­tal drift or the the­o­ry of star for­ma­tion, evo­lu­tion is a his­tor­i­cal sci­ence. It describes events that occurred in the past, on a time scale that is long com­pared to human expe­ri­ence. Gould believes that Dar­win’s chief con­tri­bu­tion to sci­ence was not a the­o­ry of evo­lu­tion, but a con­vinc­ing demon­stra­tion of the use­ful­ness of his­tor­i­cal meth­ods in science.

The data is in the rocks

We can have con­fi­dence in evo­lu­tion for the same rea­son we know that Colum­bus sailed to Amer­i­ca in 1492. It is a mat­ter of rig­or­ous infer­ence from his­tor­i­cal data. In the case of evo­lu­tion, the data are the fos­sils in the rocks.

There is plen­ty of room with­in the his­tor­i­cal method for test­ing hypothe­ses. One way his­tor­i­cal hypothe­ses can be test­ed is by cre­at­ing com­put­er mod­els of his­tor­i­cal events. A series of events that requires mil­lions of years in the real world can be read­i­ly sim­u­lat­ed with computers.

Karl J. Niklas and his col­leagues at Cor­nell Uni­ver­si­ty recent­ly report­ed a fas­ci­nat­ing exer­cise in com­put­er mod­el­ing of bio­log­i­cal evo­lu­tion. They used a com­put­er to study the ear­ly evo­lu­tion of land plants. Mil­lions of years of vari­a­tion and nat­ur­al selec­tion unfold in min­utes on the screen of their machine. Plants stand up, spread branch­es, scat­ter spores, and reach for the sun — all electronically.

First, the Cor­nell researchers for­mu­lat­ed hypothe­ses about what fac­tors had the great­est effect on plant evo­lu­tion. They select­ed the abil­i­ty of the plants to gath­er sun­light for pho­to­syn­the­sis, to sup­port ver­ti­cal branch­ing struc­tures, and to dis­sem­i­nate seeds or spores. The effi­cien­cy of all of these things were defined geometrically.

The plants in the com­put­er mod­el includ­ed only sim­ple “stick-fig­ure” branch­ing struc­tures. The prob­a­bil­i­ty of branch­ing, the branch­ing angles, and the rota­tion of the branch­es about the trunks or stems were con­trolled by let­ting the com­put­er sim­u­late ran­dom “genet­ic” muta­tions of indi­vid­u­als with­in a species. The pro­gram main­tained a degree of “genet­ic” con­ti­nu­ity between ances­tors and descendants.

Survival games

If nature is “red in tooth and claw,” so was the com­put­er. Niklas and his col­leagues let the com­put­er play war games. On the field of bat­tle at the start were sim­ple ground-hug­ging plants, like those that appear ear­li­est in the fos­sil record. The com­put­er cranked out suc­ces­sive gen­er­a­tions of plants, allow­ing for muta­tions, and scored each new species on its abil­i­ty to gath­er sun­light, dis­sem­i­nate spores, and avoid the shade of its neigh­bors. The win­ners con­tin­ued to the next gen­er­a­tion. The losers were eliminated.

The plants that won the war games look aston­ish­ing­ly like mod­ern trees. Niklas claims that the inter­me­di­ate com­put­er plants are con­sis­tent with the fos­sil record. He does not claim that the com­put­er games “prove” the hypothe­ses, only that they raise our con­fi­dence in the use­ful­ness of the hypothe­ses to explain cer­tain his­toric events that occurred hun­dreds of mil­lions of years ago.

Oppo­nents of evo­lu­tion will object that the Cor­nell com­put­er mod­el is an over­sim­pli­fi­ca­tion of a very com­plex set of cir­cum­stances. Niklas read­i­ly admits this. But even a glance at the love­ly plants unfold­ing on the screen of his com­put­er impress­es one with the pow­er of the method.

Every day we use maps of the world that incor­po­rate only a few fea­tures of the ter­rain they describe. Nev­er­the­less, the maps are enor­mous­ly help­ful, and it is hard to imag­ine how we could get along with­out them.

Sci­ence is a map of real­i­ty. A map is use­ful if it gives us a sense that it incor­po­rates some essen­tial ele­ment of the real.

Evo­lu­tion is map of the his­tor­i­cal past. As a map, it is par­tial and incom­plete, but more use­ful by far than any alter­na­tive map that has yet been devised for explain­ing the evi­dence that is doc­u­ment­ed in stone.