Evolution, self organization and neuroscience

Question

I have rudimentary knowledge of evolution, and biology in general, so bear with me if this question is a bit naive.

Let's say we have a particular trait, like highly sensitive peripheral vision. There are two ways to go about explaining how this came about:

One is to say that evolution selected for it over time, because animals with a poor sensitivity to peripheral activity could easily become prey. So there is an advantage to having sensitive peripheral vision, and such a trait is adaptive. The picture I have in my mind is of an assembly line where there are randomly wired visual cortices and the best ones get selected for automatically. This seems to be the view of the field of evolutionary algorithms, at least to my mind.

Now, if we focus on the random nature of the wiring, we might see that there is an explosion of possibilities in terms of how the visual system can be wired together. After all, evolution is blind, and it doesn't know beforehand which physical signals are of interest.

I'm trying to think of a more plausible mechanistic explanation. Suppose we assume that given certain reliable properties of the physical world, like permanence of objects for example, evolution has found the optimal way to wire the visual cortex so that peripheral vision (and a host of other good-to-have features) is a natural outcome, then it would give us more focus in what environmental features and algorithms to look for. This can also be thought of as the idea that evolution, over time, has come to assume that certain things in the physical world will be a given and exploits these. (For example, objects in the peripheral areas of vision will tend to have different perceived velocities, so perhaps normalizing velocities across the field of vision automatically results in a need for higher sensitivity in that region. It just so happens that this could also be useful in detecting predators).

My question then is, is my assumption in the above paragraph valid? (Of course, the question then gets shifted to "Why normalize velocities across the visual field in the first place?").

Answer 1

I don't fully understand your question. I hope the following will help a bit.

I'd like to pinpoint what seems to me to be one important mistake in your text. Reading you we have the feeling that selection acts on wires so to select the best combination. If this is what you meant, then it is obviously wrong. The thing is that the change that get inherited are alleles. An allele is a variant of a gene. For example: gene codes for eyes color. A bi-allelic gene has an allele which codes for blue eyes while the other codes for brown eyes. Assuming that there is genetic variance underlying the wire connection variance in the population, then yes selection will select for the best genetic variant. This does not necessarily mean that this best genetic variant is the most optimal possible to imagine solution. The most optimal allele might have never been created just because of the randomness nature of mutations.

EDIT:

To answer:

Do natural selection always bring a particular phenotypic (loosly speaking, phenotype = morphological) trait to its optimal state?

No! It does not.

1. The main reason is that the genetic variant coding for this optimal trait does not necessarily exist because of the randomness of mutational process. Also, because of the metabolic pathway bringing some variance to a trait, it is often very hard to even imagine a mutation that would allow creating a new, more optimal, phenotypic variant. So indeed developmental biology has to be taken in account to.

2. Because in absence of genetic drift (infinite population), selection can only bring genotypes uphill in the fitness landscape. If a very high peak exist somewhere in the fitness landscape but in order to reach such peak, 10 mutations are needed were all of them in any other genetic background are deleterious (correspond to valleys in fitness landscape), then the optimal high will probably never be reached (except if all 10 mutations appear suddenly in the same individual).

3. Because selection might be weaker than genetic drift. Also, different subpopulations might be in different environments were different traits correspond to optimal traits and because of migration (gene flow) between subpopulations, it is possible that no optimal phenotypic trait can be ever seen although all the genotypic variance exist.

4. The seemingly most optimal trait might actually not be really optimal. For example: Peripheral vision might be limited by the brain rather than by the eye because it would be an important energy cost for the brain to process more information, a cost that might outweight the benefit of having a better peripheral vision.

Answer 2

Your two arguments are actually the same

The import point you are missing about the algorithm approach of optimizing a genome is it optimizes for the environment it was cycled through in the past. the laws of physics have been constant through those generations so the system becomes optimized for the laws of physics. Well optimized for physics at the scale of life at least, we are not good at handling relativity for instance.

It is much the same reason the human brain favors a desire for sweetness that is counter productive in a modern industrial world, because in the environment of our ancestors getting problematically high concentrations of sugar is simply impossible, so there is no benefit to having evolved a limiting behavior. Evolution has only had a handful of generations to select for anything else. Advantage and disadvantage are always relative to the environment, so the environment of the ancestral line effects how evolution optimizes an organism. Consider being very good at conserving heat would be very beneficial in the artic and very detrimental in a hot jungle. Retaining water would be very helpful in a hypertonic environment and very deadly in a hypotonic environment. Things are adaptive or maladaptive depending on the environment so evolution favors or disfavors the passing on of genes differently depending on the environment. A brain the predicts motion well under earth gravity has a big advantage, on earth, but since earth life evolved on earth brains can be slowly optimized for predicting motion under earth gravity, note I say can be sometimes the brain is stead simply made plastic and learns to predict motion in earth gravity because it is only exposed to earth gravity. It can be very hard to tell which is occurring.

Now it is worth noting optimization is only local not ultimate, Remi covers this quite well in his discussion of fitness landscape so I will refer you to it, I will just say optimized is not the same thing as perfect, it is only the best given what is available and given a history of X.