The limitations of Sexual Selection

First, you are assuming that predators rely on the same perceptual apparatus (color vision) used by the female guppies to detect fitness in the male. Many marine predators instead use other senses: echolocation (dolphins), for example, or a bio-electrical field disturbance mechanism (sharks). Indeed, the rainbow panoply of marine life around coral reefs suggests that bright colors are not an insurmountable hazard. Also of note, many terrestrial creatures warn of their toxic character with bright colors.

Second, you posit that natural selection must limit sexual selection, but then assert sexual selection has explanatory power "only in the complete absence of predators"--you have no grounds to suggest the limiting effect of natural selection is absolute.

Finally, if there are color-targeting predators present, perhaps the brightly colored males display their fitness by surviving despite their attention-grabbing garb. The logic of sexual selection dictates that the selected trait denotes fitness, and perhaps sufficient speed and wariness to escape predators in this context is that fitness.

A female who prefers the most brightly colored mate in the vicinity isn't going to be subject to adverse selection as a result. And she's always going to pick some mate, therefore her genes are always going to be passed on, assuming she lives long enough.

So where this preference is widespread, males will equilibriate at the greatest, boldest coloration they can maintain in the face of predation. Shouldn't they?

The female preference for colorful mates is spurred by the fact that such displays are signs of health and strength. Obvious evolutionary advantages for her offspring.

It is fairly easy to see why conspicuous secondary sexual characteristics persist once they exist. They are part of an ESS --- an evolutionarily stable strategy --- which may be defined as a strategy which is optimal to follow if everyone else is following it.Consider the conspicuous plumage of the peacock. A male with drabber plumage would be more likely to escape predation, and would be favored by natural selection (in the narrow sense, i.e. excluding sexual selection), but because females make sexual choices on the basis of plumage, he would be disfavored by sexual selection.But why, you might ask, can't the species evolve so that peahens prefer drabber plumage? Well, this is where the notion of an ESS comes in. Such a preference would have to start with changes to the genes of one peahen. She would mate with the peacock with the drabbest plumage. Her sons would be more likely to survive, but they would be selected against by sexual selection because all the other peahens would still be selecting for conspicuous plumage.We'd therefore expect the balance of conspicuousness to camouflage to be around an equilibrium such that any more or less conspicuousness would be penalized by, respectively, the decreased chances of survival outweighing the increased chances of reproductive success, and the decreased chance of reproductive success outweighing the increased chances of survival. Natural selection pulls one way, sexual selection pulls the other way, and "the complete absence of predators" is not (as you suggested) necessary for sexual selection to make the males of a species somewhat conspicuous, any more that the complete absence of female choice would be necessary for natural selection to make them somewhat drab.

The only way natural selection could win out completely is if the predatory pressures were so great that only drab peacocks would survive. If this situation came about suddenly then it is more likely that the species would be driven to extinction.So generally speaking once a species has conspicuous secondary sexual characteristics it's stuck with them.

I think that answers your first question. The second question is, I think, more interesting: how would such a state of affairs arise in the first place? Note that what follows is my own hypothesis --- I don't mean that it definitely hasn't also been developed by others (and I would suspect that it has, because scientists aren't all idiots) but I do mean that I have no positive knowledge that it has been submitted to the scrutiny and criticism of the wider scientific community. I feel, however, that it is so luminously obvious that you might be willing to overlook this.I think the key to understanding the evolution of conspicuous secondary sexual characteristics is the concept of a supernormal stimulus. You will probably be familiar with the classic investigation into juvenile herring gulls begging their mothers for food. They identify the maternal beak by very simple cues; it is possible to construct a fake beak which presents these cues and elicits this behavior even more strongly than a real beak, so that presented with a fake beak and a real female herring gull waiting to feed them the chicks will concentrate all their attention on the fake beak and slowly starve to death.Now, similar things happen with mate selection. For example, consider the long-tailed widowbird. The female has a ~7cm tail, the male ~50cm. When researchers artificially extend the tail by sticking feathers to it, so that the tail is much larger than that found in nature, birds so treated become even more attractive to females, rather than being disdained as freaks. (Andersson: Female choice selects for extreme tail length in a widowbird, Nature, 1992.)I read of a fascinating study for which I am temporarily unable to find a reference. A computer simulation was made of the visual cortex of a bird (again, of a species with a female preference for long tails) to see what was going on in there. What I found interesting was that there was not one bit of the brain to detect male birds and another to find how sexy they were; instead, long-tailed birds looked more like male birds; they stimulated the "male bird detector" more strongly; and an image of a supernormally long-tailed bird looks more like a male bird than an actual male bird does.Now all this suggests how extravagant displays of secondary sexual characteristics get started. Suppose that a female bird can recognize males of her species by a wing pattern as in pattern A in the diagram below, which might be part of the male's camouflage plumage. Then she will probably be even more strongly attracted to plumage patterns B through E, which look even more like a male bird's plumage than the actual male bird's plumage does by virtue of being crisper and brighter and more colorful and more contrasting.So long as the reproductive rewards of being conspicuously male outweigh the hazards associated with being conspicuous, the net effect of selection will be to drive the species towards the more conspicuous state.Now of course natural selection can't look ahead and say: "My gosh, this mechanism for females to recognize males can be subverted in a way that will ultimately be bad for the species as a whole, since it will place a premium on being conspicuous to predators, so I'd better come up with another recognition mechanism which can't be subverted in this way"; natural selection never looks ahead. Given only the small brain of a bird or a fish or the smaller brain of a butterfly in which to instantiate a mechanism for identifying potential mates, it would not surprising that this subversion of mate recognition systems should happen quite often.(I believe that a similar mechanism applies to aposematism. I have noted that the markings on aposematic species often strongly resemble in shape the camouflage markings of closely related non-toxic species. However, I should admit that (a) my researches into this phenomenon have not been terribly systematic and so may be affected by confirmation bias and (b) this might be the result of developmental constraints. This said, a similar mechanism is at least plausible: once a bird has learned to avoid a beetle with certain markings originally evolved for the purposes of camouflage, it would be still more inclined to avoid beetles in which markings of the same form were clearer and more distinct.)