Darwin's On the Origin of Species included all of the essential elements of our current understanding of natural selection.   Natural selection is a mechanism of evolutionary adaptation that results from a combination of heritable variation among individuals and differences in their survival or reproduction correlated with this variation.
On the one hand, is natural selection so simple that it reduces to a tautology and explains nothing?   On the other hand, is there enough complexity to explain the emergence of cooperation, culture, language?   Is it even specifically a biological process?
Natural selection is a change in the frequencies of alleles in a population as a result of differences in the survival and reproduction of individuals that carry those alleles.   Natural selection, along with mutation, migration, and drift (randomness), produces evolution.   It is a matter of arithmetic:   in any population, genetic variants spread when they leave more copies in successive generations.
The mechanisms of heredity and variation have become progressively clearer and their complexities better understood...   The genome, we now know, is not the only way that parents can transmit their features to their progeny.
Methylation, often called "epigenesis," meaning "beyond genetics," expands the possibilities for natural selection.   With its susceptibility to environmental influences and its progressive loss, it provides a mechanism for heredity more flexible and less stable than other ways to regulate DNA...
Still less stable influences on development, but nevertheless hereditary, can result from direct responses to an environmental feature sustained across generations.   Learned habits and customs are examples that can propagate in families and populations of interacting individuals.
Direct environmental influences have often not been accepted as natural selection...   Yet environmental influences on development, including learning, require physiological mechanisms and predispositions to respond to features of the environment and thus are subject to adaptation by natural selection.
The result can take different forms.   Development might vary continuously with some environmental feature. Alternatively, developmental switches might produce several alternatives in response to particular environmental features.   In other cases, development might be especially sensitive to an individual's genome, rather than to its environment ...
The pattern of an organism's development is embedded in the pattern of a population's evolution.   Natural selection, an interaction between genomes and environment directing the evolution of a population, results from interactions of genome and environment directing the development of each organism.
Variation in Heritable Features
In any particular case, the flexibility of development depends on the nature of both genetic variation and environmental variation.
The stability or plasticity of development or evolution depends on how natural selection responds to different degrees of environmental periodicity.
Environmental variation with periods much shorter than an individual's life is best accommodated by direct influences of the environment on an individual's development.
Environmental variation over periods of one or a few generations is often better accommodated by a few alternative subplans for development.
Variation over intervals of many generations is handled most efficiently by revisions of the basic plan.
For biological organisms, these three alternatives correspond respectively to learning, developmental switches, and genomic encoding.
Each of these developmental alternatives results from an interaction of environment and genome, with progressively decreasing reliance on environmental flexibility and increasing reliance on genomic stability...
Natural selection occurs at all periodicities of environmental variation.
Cooperation and Competition
The evolution of cooperation has created another challenge for natural selection.   Darwin#39s initial summary of natural selection emphasized a "struggle for existence"...   In his discussion of honeybees, Darwin acknowledged the challenge that such competition presents for the evolution of cooperation...
The spectrum of possibilities for the evolution of cooperation by natural selection involves increasing behavioral complexity.   Some of the options thus might apply only to humans.
For instance, although cooperative interactions with kin are widespread among nonhuman animals, only a few nonhuman primates have enough complexity of individual recognition to support the formation of reputations.   Evidence for policing by animals, even primates, is also sparse.
On the other hand, neither theory nor fieldwork has yet plumbed the complexities of helping and cheating, either animal or human.
Constraints on Natural Selection
Fisher#39s Fundamental Theorem suggested that natural selection would move populations in a particular environment toward ever greater adaptation provided a source of genetic variation, such as mutation, persisted.   Sewall Wright argued, on the other hand, that natural selection usually moved populations toward a local optimum in an adaptive landscape with multiple optima ...
Interactions among alleles, as Wright argued, make multiple local optima for genotypes nearly inevitable...   Such interactions are frequent in genomes and populations...   Furthermore, social interactions can involve traits with advantages for one individual but disadvantages for another, or traits with advantages only when present in both individuals concurrently.
Alleles associated with such traits often do not spread in a population when rare...   For instance, alleles associated with producing a signal cannot spread when alleles for responding to the signal are too infrequent, even if a response would benefit a signaler.   Vice versa, alleles for responding cannot spread when alleles for signaling are rare.   Only in a population with enough of both sorts of alleles can they both spread...
Interactions between heritable variants, whether pleiotropic or epistatic, place constraints on evolution by natural selection.   The results are thresholds and isolated adapted optima...
These interactions prolong, even when they do not prevent, evolution toward global optima by natural selection...
Optimizing structure by means of heritable variation and selection applies equally to evolution, epigenetics, and learning.   In recent decades, it has also been applied to computation and molecular synthesis.
An example is the use of computed neural networks to discriminate sets of inputs...   Randomly adjusting the interactions of nodes in each iteration of a neural network and then selecting those variants that improve discrimination between different sets of inputs eventually yields the best performance possible.   In a similar way, pharmacologists search for optimal molecular structures by progressively altering the components of complex molecules.
Evolutionary computing or adaptive synthesis occurs in a multidimensional adaptive landscape just as biological evolution does...   The adaptive landscape is the performance of an algorithm or synthetic molecule as a function of the hyperspace of possible structures (nodes and parameters or types and positions of chemical functional groups).   Any solution encounters two widely discussed problems: too much precision to capture an entire adaptive peak; and too little accuracy to capture a global peak.
These problems in evolutionary computing have parallels in biological evolution.
Humans making decisions with the assistance of evolutionary computing have learned to extend the principles of natural selection.   Making decisions based on trial and error, whether by brains alone or by brains assisted by machines, results in optimal responses to each of numerous inputs. It is the basic process of human behavior.
Indeed all animals, not just humans, learn to match responses to inputs.   Decisions occur whenever an organism discriminates between alternative inputs when choosing what to eat or where to go or whom to associate with or to imitate.
They do so because decisions in response to unpredictable inputs allow greater specificity in adaptations.   An organism's capabilities and predispositions are specified by a stable plan, the organism's genome.   Such a plan is the basis for all forms of learning and culture.   This plan then develops in conjunction with its immediate context, the organism's environment.
Evolutionary computing is thus itself a result of evolution by natural selection.
[back to CONCLUSIONS (.pdf file)]