Fitness landscapes
The fitness landscape is the standard mental picture for evolution and for search: a surface where each point is a possible organism or solution, height is how well it does, and a population is a cloud of points climbing uphill. It is a genuinely useful picture, and most of what makes it useful is knowing where it breaks.
The picture
Sewall Wright introduced the metaphor in 1932. Lay out every possible genome so that similar genomes are near each other, give each one a height equal to its fitness, and evolution becomes movement on that surface. Mutation is a small step in a random direction. Selection is a bias toward uphill steps. The vocabulary that follows is familiar: peaks, valleys, ridges, plateaus.
It explains several things well. A population on a local peak can be stuck, because every single step available makes things worse even though a much higher peak exists elsewhere. A rugged landscape with many peaks is harder to search than a smooth one with a single summit. A plateau gives selection nothing to work with, so a population drifts across it at random until it finds an edge.
Motoo Kimura's neutral theory adds an important correction to the naive picture. Many mutations change fitness not at all, so genomes of equal fitness form connected paths through the space. A population can wander a long way along such a path without ever going downhill, and arrive somewhere from which a new peak is reachable. Drift is not just noise on top of selection: it is part of how a search escapes a local peak.
Where the metaphor breaks
Three assumptions are buried in the picture, and none of them holds in a real ecology.
- The landscape is fixed. It is not. When the fitness of one organism depends on what the rest of the population is doing, the surface deforms as the population moves across it. Climbing changes the hill. This is the normal case in biology, not an exception.
- Fitness is a single number. It is not. Surviving longer, reproducing sooner and producing more offspring trade against each other, and combining them into one height is a choice made by the modeller rather than a fact about the organism.
- Nearby genomes have similar fitness. Often false. A single change in a regulatory gene can restructure an organism, so the space has cliffs that a smooth surface does not suggest.
Why this project avoids the term
There is no fitness landscape in Emergent Life, because there is no fitness function to define heights with. Nothing scores an organism. Organisms spend energy, find more or die, and reproduce as an action they take, and what survives is whatever survived, described afterwards rather than measured during. Reasons for that choice, and what it costs, are on genetic algorithms.
The environment is also not fixed. Resources deplete where organisms feed and regrow over time, and terrain material can be moved and later recovers, which is set out in reshaping the world. So even the deformed-surface version of the metaphor understates it: the ground moves for reasons that include the population's own past actions, which is the subject of niche construction.
What can still be measured are the things that do not need a landscape: which lineages persist, how the distribution of each gene shifts over time, and when a lineage has diverged enough to count as its own under a stated rule. Those are on genome, inheritance and mutation.
References and further reading
Primary sources for the claims on this page. Where this project's own results are referred to, they link to the page carrying the evidence instead.
- Wright, S. (1932). The roles of mutation, inbreeding, crossbreeding and selection in evolution. Proceedings of the Sixth International Congress of Genetics, 1, 356-366.
- Kimura, M. (1968). Evolutionary rate at the molecular level. Nature, 217, 624-626. doi:10.1038/217624a0
- Kauffman, S., & Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. Journal of Theoretical Biology, 128(1), 11-45. doi:10.1016/S0022-5193(87)80029-2