So, evolutionary psychology is frequently criticized as reductionist, and many of its critics complain about the employment of evolutionary accounts for human social behavior. In my dissertation, presently going through final revisions, I argue in favor of reductionism, and I extensively and usually approvingly cite Ruth Millikan's Language, Thought, and Other Biological Categories, which advances an evolutionary account of thought and language. So it seems that I ought to be predisposed in favor of evolutionary psychology. For that matter, when I was studying at UCSB, I did one year have a roommate who was an evolutionary psychology grad student, and he was a nice enough guy. However, I'm not actually finding myself all that impressed so far.
I'll explain one important difference between Millikan and the evolutionary psychologists I've read so far. Millikan recognizes that there are many ways in which things can be reproduced, and central to her view is the point that selection can operate on things which are reproduced by any means. So, for example, tools can easily be seen as produced via a complicated evolutionary process; in modern times, people buy tools because they perform various functions more or less well, and tools are manufactured because people are willing to buy them. Tools that perform their function better sell better, so when manufacturers come up with a new and better tool, they start producing copies of that instead, driving the old, less good tool out of the market. Tools obviously don't have genes, though they have a variety of constraints; a tool which performs better but requires much more expensive materials will not be selected for, and a tool may gain an advantage in the marketplace because machines specialized for more quickly and efficiently manufacturing that tool are already available, and alternate designs which couldn't be built using the same machines may be at a disadvantage despite performing their functions better. And there are endless further complications (for example, legal issues or aesthetic issues may affect which tools can be sold and which are preferred by buyers; a particular tool's manufacturer may have deals with certain efficient distributors which give the tool advantages over tools made by other manufacturers independent of the merits of the tool).
Millikan thinks we should interpret beliefs on a similar pattern; we have evolved some machinery for generating beliefs, and one of the mechanisms of belief formation involves the ability to transfer beliefs to others and acquire beliefs by transfer from others through communication (indeed, Millikan thinks communicability is one of the criteria distinguishing beliefs from other kinds of states, though communicability with one's future self via memory may be sufficient for something to be belief-like, so animals are not necessarily excluded from having any beliefs whatever). Selection pressures obviously function on the beliefs, and the exact details of the mechanisms surely play a big role in how easily various kinds of beliefs are developed and transferred, but there are plenty of other selection pressures.*
For example, and this is critical for one of my big worries about the evolutionary psychologists I've read, if a particular belief often leads to a quick death, its ability to spread to others is greatly reduced, as its advocates will have little time to pass it on. Further, it is highly likely that observing the tendency of a belief to lead to death will produce a certain resistance to the acceptance of that belief, even if the exact psychological mechanisms by which people accept or reject beliefs presented to them in communication are not fully understood. Thus, it is to be expected that beliefs which lead to swift death would be rare due to selection pressures on beliefs. Beliefs which lead to swift death might also be rare due to selection patterns on belief-forming mechanisms; a belief-forming mechanism which can never form a rapidly self-destructive belief seems to have some advantage, and so it is possible that people should have genes which code for belief-forming mechanisms hostile to a particular self-destructive belief. However, the fact that the self-destructive belief faces powerful selection pressures anyway means that the difference between having such genes and not having such genes is small (even those without genes specifically hostile to a particular self-destructive belief are unlikely to develop that belief because of the general features of belief formation and the selection pressures on beliefs, so those without the genes are at only a very slight, perhaps negligible, disadvantage).
Thus, it is an empirical question where the forces of selection have exerted the most influence in bringing about a particular behavior; selection pressures don't have to work on genes. They certainly don't in the case of tools, which don't have genes, and beliefs undergo sufficiently rapid change that it is implausible that genetic changes could account for changes in belief frequencies. Indeed, behavior in general is quite variable between times and cultures, in ways which seem plausibly responsive to selection pressures, but which change too quickly for the selection pressures to have operated on genes.
However, note that while great variability does tend to rule out the possibility of gene selection, less variability does not necessarily show gene selection; if selection forces are particularly strong, then a behavior transmitted through explicit instruction or imitation would be expected to be very persistent and to spread quickly and thoroughly, and if such a behavior first appeared sufficiently far in the past, it would not be surprising for the behavior to be universally or nearly universally taught and imitated by humans everywhere.
Evolutionary psychologists all, to varying degrees, concede that behavior results from a complex interplay of genes and environment, but it appears that in practice they focus almost all of their attention on genes. So far, I have not been able to dig up much evidence in favor of this procedure. Admittedly, we are presumably interested in all of the causes of human behavior, so evaluating selection for genes which influence behavior is certainly an interesting and worthy research project. However, in some evolutionary psychology writings (such as A Natural History of Rape, the example I'm currently investigating), the evolutionary psychologists insist that genes are the most important factor, and that concentrating on other factors invariably leads to worthless results. This does not appear to be the case, even on the basis of the arguments they themselves present.
* There are obviously connections between Millikan's account and the trendy notion of "memes," first proposed by Dawkins somewhat earlier than Millikan's book. Millikan's account is much more detailed than most discussions of meme theory, and to my knowledge she has never adopted the meme terminology, but she is clearly in sympathy with the views of Daniel Dennett (and vice versa), and Dennett has recently jumped on the meme bandwagon. I've avoided it because of the misleading implication of atomism, though arguably "gene," on which the term is based, is equally misunderstood when interpreted atomistically, so perhaps the lesson is that people should be wary of atomism rather than of a particular terminology.