The Baldwin Effect stands kind of half-way between Darwinian evolution and Lamarckian evolution. It suggests how the characteristics acquired by individuals during their lifetimes can affect evolutionary changes in the genetic base to make the course of evolution more efficient. This idea is currently getting as much attention in the field of evolutionary computing and artificial intelligence as it is in biology. It promises to provide a very important functional link between artificial neural network learning systems and genetic algorithm approaches.
The Baldwin effect is not a recent idea. It was first proposed by James Mark Baldwin in 1896, at a time when the idea of Darwinian evolution was young and vigorous and the idea of Lamarckian evolution was perhaps mortally wounded but not yet dead.
Baldwin's proposal was that an advantageous characteristic that is acquired by individuals through some physiological change in response to the environment or through a learned behavior will tend to be replaced by a gene-based trait that offers the same advantage. In technical terms, a successful result of phenotype plasticity comes to be reflected in the genotype. While that sounds like simple Lamarckism, there is a critical difference between Baldwin's idea and Lamarckism. Whereas Lamarck proposed that the new characteristic was directly and immediately reflected in the genotype, there is no direct or immediate genetic change in the Baldwin effect.
In the Baldwin proposal, the actual change in the genotype happens in the Darwinian way, through mutation and selection, but that change is facilitated, which is to say expedited, by the acquired characteristic. This effect works in two ways. One is that the success of the acquired characteristic gives a leg-up to partially successful mutations that may not have otherwise provided sufficient advantage by themselves to survive the selection process; the other is that individual development in response to the environment serves a kind of scouting role, searching the terrain of the solution space and pointing the way to success for the much slower process of random mutation and selection. The genetically 'hard-wired' trait, once it appears, tends to displace the acquired characteristic because the new trait is a short-cut; the process of individual learning or adaptation is costly in time and resources, but the genetic solution comes 'free,' so there is selective pressure in favor of the new genotype and against the phenotype adaptation.
Dennett places major importance on the Baldwin effect in his description of the evolution of consciousness (Consciousness Explained, pp. 184-187)
"A New Factor in Evolution," (Baldwin's original paper) is at http://www.santafe.edu/sfi/publications/Bookinforev/baldwin.html and many other sites
A review of "A New Factor in Evolution" by a Baldwin contemporary can be read at http://paradigm.soci.brocku.ca/~lward/sup/CATT_001.HTML
Evolutionary Computation, Volume 4, Number 3 (Special Issue on the Baldwin Effect).
| 
