Evolutionary puzzlements

[zeightyfiv]

Does phenotypic evolution operate in the thermodynamic limit?

What I mean by this is, does evolution always take a given prior state and environmental input and then produce the same, deterministic phenotypic changes, modulo a tad of noisiness? This is not to say that evolution would retrace its steps given equivalent environmental inputs, thermodynamic systems can be unstable and chaotic too. And, it certainly doesn't suggest it would retrace its steps without the same environmental input; that's patent nonsense, considering, for example, how important meteorite impacts have been in evolutionary history! But rather, given particular evolutionary circumstances, is the next step inevitable?

For example...

• If it's evolutionarily advantageous for an animal to have longer horns (an easily controlled trait), will they get longer?
• If it would be beneficial for species of a given lifestyle to develop sonar navigation, will one?
• If there is a large, empty niche for land animals, will amphibians inevitably evolve eggs like amniotes, tough against mechanical insults and permeable for respiration but not for dehydration? With a large yolk and the elimination of a distinct juvenile feeding morphology (tadpoles)? (I don't understand why the latter pair were coupled in.)

Clearly, there's some (or several) notions of scale, here. A tiny population can easily drift and screw itself over (just as a tiny cluster of atoms can bounce all over the place, not necessarily rolling "downhill"), but a larger population is more stable... and many questions are naturally posed across multiple species and families, asking whether perhaps one of them will follow the next logical step. A prerequisite for this "thermodynamic limit" notion is, to (ab)use the language of partial derivatives that thermodynamics is so fond of, that ∂Fitness / ∂t | environment  ≥ 0 -- that fitness is an entropy-like variable (but subject to the whims of environmental changes) which, given sufficient scale to drown out noise along with a constant environment, never decreases.

There's also some notion of scale in the traits themselves. It's one thing to ask if a specific gene or metabolic pathway will be duplicated or modified in a particular way. It's quite enough to ask if *some* gene (or genes) will arise to provide a certain function one way or another. The peculiarities of this choice may then affect evolutionary viability of further decisions downstream, but they don't immediately matter. (This is why I refer to "phenotypic evolution" rather than just evolution.)

Finally, some traits are just not particularly evolvable. Whether due to prior accidents of design or fundamental a priori difficulties, some changes or features are not easily to make -- there's a very high activation barrier, so to speak. This is not exactly a caveat, though. If some traits are insufficiently evolvable, they just won't happen, not repeatably, probably not ever. So, it's not so much a problem for determinism as it is a problem for determining just what is on the plate of options. It is, however, a potential caveat for some of the questions one might ask.

Now, here's a question I would ask, which has been bothering me today:

Why did placental development in land vertebrates evolve from amniotes? Why didn't it evolve directly from amphibians?

Placental development *seems* like an easier system to put together. Even ovoviviparity, which has evolved countless times in other places, seems like it would clear the way for land habitats and increasing size without needing protected freshwater environments to return to for breeding -- and without needing amniotic development as an intermediate step. So, is there a key constraint I'm missing? (something to do with lactation? extraembryonic structures for respiration? a need for endothermy?) Or did evolution just say "Ha, ha! We're going this way!"