I was reading this 2003 paper by Philip Van Peer and colleagues, which is a quick introduction to the site 8-B-11, Sai Island, Sudan. I wanted to make a note of their description of an apparent sandstone mortar for pigment grinding:
In this Lower Sangoan level [date unclear from text], we observed the presence of a dense concentration of red and yellow ochre lumps, some with ground surfaces. An extraordinary object (Fig. 3) was found in trench 02/C, near the southern edge of the Middle Pleistocene gully (Fig. 2). The overall shape of this 10 cm thick Nubian sandstone slab is the result of human action. Steep flake negatives are present on its sides, except for the lower right part of the perimeter where oblique flake scars occur. Two of these flakes were found near the slab and refitted onto its right side. Apparently, these scars represent posterior reworking.
Except for small areas in the centre and to the right, the upper face of the slab is made perfectly flat by pecking. A large depression, of which the sides have su ronment, then natural selection must be caused by genes plus the environment. </p>
This view clearly encompasses Lewontin's chemostat. Why does one strain do better than the other in the same environment? Because, at least in that environment, its genes are better at reproducing. We don't need to know how the strains reproduce or survive in other environments, we don't need them to interact with each other; the only competition necessary is the obvious one: one reproduces more than the other.
It also deals with the "different causes" problem. Consider sexual selection: It is not a struggle for existence, but for mates. In its effects on allele frequencies, it is the same as natural selection. Geneticists generally define sexual selection as a special case of natural selection. But naturalists often treat them as entirely different phenomena. I read this as one of the reasons why the "natural" is lost from "natural selection" when geneticists start talking about it; there is no reason for them to commit to a cause, and to the extent that cause matters it can be described as genetic causation.
In other words, we don't need another word for every causal agent, be it mating, intrinsic advantages of alleles, epistasis, or whatever. Selection is as selection does.
Reaction the fifth
Instead of redefining agency, we may choose to redefine environment -- or at least, define it sufficiently broadly to cause every instance of natural selection. The environment includes not only Darwin's "hostile forces of nature" but also the social environment, the cognitive environment, the intracellular environment, and the proteomic environment. Everything that isn't genetic information is in this conception part of the environment.
This has the advantage of casting molecular interactions among genes as external rather than intrinsic. The fact that an allele's fitness may depend on the genetic background is just a recognition that an allele's function is properly viewed as differential against the action of other genes. Thus, epistasis is environmental. Sexual selection is caused by the social environment. The fast-reproducing strain in the chemostat is enabled by its environment in a way that the slow-reproducing strain isn't.
This the point of view that I have come to favor, because I think there is a point to viewing the genes themselves as information units rather than functional units. If we maintain that distinction, then the non-informational content of cells must be part of the environment in which genetic information is embedded.
It also makes sense to describe many instances of genic selection as a consequence of the cellular environment. For instance, codon bias is thought to often be a result of selection resulting from the relative numbers of tRNA species, amino-acid binding affinities, or nucleotide abundance. This selection has to do with factors that are intrinsic to cells, and have been for a long, long time. The selection in question is incredibly weak, and may not be strong enough to affect species with populations as small as most mammals. But if the two bacterial strains differ in their codon usage, with one highly optimized to intracellular tRNA abundances and the other highly disoptimized, and one reproduces faster than the other, then it is clear why the genetic information of one is favored by natural selection imposed by intracellular environments.
I think that this broad view of environment deals with Skipper's objection. Unless we really construe natural selection to literally require a "struggle", all of the deterministic outcomes possible between populations of organisms are caused by the environment. This leaves only the non-deterministic, stochastic outcomes, which are the basis of genetic drift. So under this view, natural selection encompasses all cases in which alleles are causally correlated with reproduction.
Selection or culture?
One may ask, why does any of this matter? There is a reason that is near to my heart: recent human evolution. During the last 10,000 years, human evolution has involved substantial genetic change in concert with great population growth. Some human populations maintained high intrinsic growth rates for a long time, and today their descendants make up a much higher proportion of the human gene pool than their ancestors did during the Late Pleistocene.
One 800-year-old Y chromosome lineage is found in 14 million Asian men. The haplotype has clearly been correlated with reproductive success over that time period. But what is the cause of that correlation? Is it a case of "Y chromosome hitchhiking" on some cultural advantage? That would be the Genghis Khan hypothesis -- one man's reproductive advantage for extreme cultural reasons was passed to many of his descendants through cultural mechanisms, and it has had an accidental effect on the Y chromosome haplotype frequencies of Asia.
Or maybe some random (i.e., non-Genghis-related) man had a Y chromosome variant with a functional advantage. For whatever reason, it makes its bearers have more sons.
The latter case is natural selection. The former case, presumably, isn't. It's an allele hitchhiking on a cultural advantage. In other words, the hypotheses are distinct because they involve different causes. But they also entail different predictions: for example, the selection hypothesis may predict that bearers of the haplotype have a reproductive advantage regardless of their current social status, while the culture hypothesis clearly requires that the advantage comes from high social status.
In other words, these different causes have different mathematical consequences, that can in principle be tested (although in this case they haven't yet been). I like my causes that way: different cause, different effect. So I like the broad-environment definition.