UPDATE - Another paper by Stuart-Fox et al. that appeared shortly after the work on social color change shows that chameleons do indeed change color in response to predators to increase their crypsis. Specifically, this paper shows that the dwarf chameleon (Bradypodion taeniabronchum) exhibits different responses to snake and bird predators, and does so in a manner that likely optimizes its camouflage in each case. As much as I hate to admit it, chameleons are obviously much better at this whole color change thing than anoles... [Thanks to commenter Michael Meadon for pointing this out]
Ask anybody why a chameleon changes its colors and you're sure to hear something about it trying to blend in with it's background. It may come as a surprise to many of you, then, that this simply isn't true. Chameleons, and most other reptiles that are into color change (e.g., Anolis carolinensis, a.k.a. "the American Chameleon"), do so almost exclusively for social reasons. Stuart-Fox and Moussali drove the nail in the coffin of this myth with a remarkable series of studies a few months back (one in American Naturalist and another in PLoS Biology). By combining data about chameleon color, the light environment in which they're displaying, and the visual systems of both chameleons and their potential predators, they were able to show that the conspicuous color-changes observed in the South African dwarf chameleon are specifically designed to stand out against their background for the purpose of social interaction. Got it? Color change in chameleons didn't evolve so that they could blend in with their background, but so that they could stand out against it! Of course, this doesn't change the fact that chameleons are still pretty damned good at being cryptic, just that the color change isn't used for this purpose.
PS - I couldn't agree more with their assertion "that quantifying signal conspicuousness to different receivers can be used to gain insights into the evolution of signal diversity in animals." In lizards, this type of work has been made possible to the remarkably thorough studies of Leo Fleishman and colleages, who have painstakingly measuring the properties of lizard visual system. Although the models remain imperfect, they're a hell of a start. Thanks Leo!