There's an important distinction in logic between inductive and deductive reasoning. Deductive reasoning proceeds from law-like premises to sound and valid conclusions along the sure path that: if the premises were all true, then the conclusion would follow necessarily.
Despite the image of science as deducing sure conclusions, most science is inductive. Induction starts from observations and assumptions that are not necessarily law-like and ends with conclusions that are only probable.
The best example is the statistician's urn analogy. There are no laws about the contents of urns, but only (supposed) observations about the marbles drawn from these urns and then conclusions about the probability of their marbles being white, red or a mixture with some specific frequency.
That is, even if all assumptions were true, the conclusion would not follow necessarily. It might still be false. It is my hunch that a lot of the debate about evolutionary psychology is due to a mistake of the logic categories of induction with deduction.
Take the recent publication on cunnilingus being a sperm retention strategy. As a piece of deductive reasoning it would be valid but not necessarily sound. As a piece of inductive reasoning, however, it is based on many assumptions, with only weak observational support for some of them.
For example: "Men mated to more attractive women also are at greater recurrent risk of sperm competition because rival men more frequently approach, and successfully lure away, more attractive women." (Pham et al. 2013. 'Is cunnilingus-assisted orgasm a male sperm-retention strategy? Evolutionary Psychology 11(2):405-414) How is that? The rival men are all equally attractive to the gals and therefore even the ugliest men rival for them and never for the lesser females?
Leaving the assumptions about the function of the female orgasm aside, the authors assumed that they reliably measured the risk of the sperm competition that a male faces by asking them in a questionnaire: how (1.) physically and (2.) sexually attractive they deemed their mates as well as their opinion about how (3.) physically and (4.) sexually attractive they thought other males would find them.
What if their assumed connection between attractiveness of the women and their propensity to cheating does not hold in the first place? The measure would be useless.
They also asked the males how satisfied they were with their partners (1.) sexually, (2.) emotionally and (3.) overall as well as how (4.) committed they were to them. This set of questions about satisfaction was then used to control statistically against a bias of the sample in this respect.
Wouldn't it have been more reliable to ask females (or even to ask both sides of each pair) whether or not they have extra-pair sex, whether or not they perform cunnilingus and to look for a correlations in the data?
Given their data, one can conclude that the degree of satisfaction in a relationship does not correlate with the perceived beauty of the female.
As the results were interpreted on a Darwinian theory, however, it is irritating to me that some economists are in the habit of interpreting very similar data very differently, because their theory is not evolutionary but economic.
Tim Harford, for example, published a book called The Logic of Life (Random House, 2008), in which you can read about oral sex being a strategy to evade the risk of unwanted pregnancy.
There are differences in context and sample, of course. Harford muses about teenagers or other people not in a committed relationship and about oral sex that is not being followed by coitus with ejaculation. It nevertheless strikes me as odd that a bit of context (relationship vs. not) in combination with a differences in the underlying theory (economics vs evolution) can completely reverse the conclusion (maximizing fertilization probability vs. minimizing it).
What's next? Fellatio being a cost imposed by females according to Zahavi's handicap theory?