Insight into the evolution of animal behaviour comes from an unexpected source – a computer program. In a report in the online journal Paleontologia Electronica, Oyvind Hammer of the Paleontological Museum, Oslo, Norway looks at how a simple program can generate patterns reminiscent of the 'trace' fossils – preserved tracks and trails – left by some of the earliest animals.

The program is an example of what computer scientists call 'artificial life' or 'AL', designed to simulate evolutionary processes such as mutation and natural selection. AL is currently a dynamic area of research in computer science: it is used to model everything from the behaviour of natural ecosystems, to the ability of computer viruses to penetrate computer security, and the ability of a computer program to solve a given problem.

Although Hammer cautions against drawing too close a parallel between computer-generated patterns and what happened in real evolution, he proposes that the experiment can provide useful insights of a more general kind: in particular, that complex patterns of behaviour that seem directed and 'intelligent' can be produced by a disarmingly simple set of subroutines. There is no correlation, then, between the complexity of a fossil trackway, and the creature that made it.

Fossils of complex animals appear quite suddenly in the fossil record around 550 million years ago. This transition, known as the 'Cambrian Explosion', appears to have been a real event. This is proven in the breach, by the existence of rare fossils of soft-bodied creatures in even older strata, and by the preserved traces of trails and burrows whose creators would have been large (that is, visible to the naked eye), and presumably of a degree of organization similar to that of a worm. Some of these trails are very old indeed: in a report in Science, Dolph Seilacher of the University of Tübingen in Germany and colleagues described possible worm burrows from India in rocks thought by some to be 1,000 million years old, though this evidence is contentious.

Hammer now uses AL to suggest what, if anything, such worm trails tell us about the behaviour of the unknown trackmakers. In his computer, he simulates a flat plane (imagine, if you will, an ancient sea-floor) over which 'food resources' are distributed in random patches. He then unleashes his synthetic trackmakers into this prehistoric worm heaven, and observes the tracks they make. Trackmakers that achieve success in finding food patches are allowed to make it to the next generation. They are allowed to reproduce, and mutation also takes place. After hundreds of generations, apparently complex patterns of behaviour emerge.

The synthetic trackmakers are, in reality, collections of subroutines, or 'modules', arranged into a network. The modules include analogues of sensors, which detect food (essentially, a calculation that reports output values in some proportion to the proximity of a food patch.) But Hammer's module list includes things that one would associate more with an electronic hobbyist's mail-order catalogue than with animal behaviour – summers, multipliers, sine-wave and sawtooth-wave oscillators, and so on. The modules are connected into networks in ways that can be changed by 'mutation', which can also affect the strength of the connections between each module.

Artificial evolution produced sets of routines progressively better at finding food patches, and staying within patches once they got there. One routine – a synthetic trackmaker – found itself with a sawtooth oscillator coupled to the direction parameter of a food sensor, "creating a sweeping sensor that could report the environmental state in all directions successively, much like radar."

"This may seem an obvious idea in hindsight,", says Hammer, "but nevertheless a clever solution that the programmer did not foresee." This routine produced a pattern of direct, straight lines between patches, and, once in patch, tended to meander within it. "All this must be regarded as rather elaborate behaviour given the small size of the control network," concludes Hammer.