<p>Researchers have long assumed that black widow coloration functions as a warning signal to avian predators. However adult female black widow coloration does not resemble typical warning coloration in two distinct ways. First, black widows are less colorful than most other documented aposematic species. Second, the hourglass shape of an adult female varies both between species and within a site. Here I examine the ultimate causes and consequences of North American black widow coloration. </p><p>In chapter two I present data that suggest that black widow coloration not only functions as an aposematic signal to avian predators, but has also been selected to be inconspicuous to insect prey. In choice experiments with wild birds, I found that the red-and-black coloration of black widows deters potential predators: wild birds were ~3 times less likely to attack a black widow model with a red hourglass than one without. Using visual-system appropriate models, I also found that a black widow's red-and-black color combo is more apparent to a typical bird than typical insect (Euclidean color distance ~2.2 times greater for birds than insects). Additionally, an ancestral reconstruction revealed that red dorsal coloration is ancestral in black widows and that at some point some North American black widows lost their red dorsal coloration (while maintaining the ventral hourglass). Behaviorally, differences in red dorsal coloration between two North American species are accompanied by differences in microhabitat that affects how often a bird will view a black widow's dorsal region. All observations are consistent with a cost-benefit tradeoff of being conspicuous to potential predators while being inconspicuous to prey. I suggest that avoiding detection by prey --- combined with Müllerian mimicry --- may help explain why red-and-black aposematic signals occur frequently in nature.</p><p>In chapter three, I examine the variation in hourglass shape. Classical aposematic theory predicts near uniformity in warning signal appearance because a uniform signal is easier to learn to avoid than a variable signal. However the shape of the hourglass of North American black widows appears to vary both within and between sites in ways that are inconsistent with classical aposematic theory. Using 133 black widows of three different species from nine sites across the United States, I quantified the variation in hourglass shape and examined how Müllerian mimicry, species type, and condition each influenced hourglass shape. A principle components analysis revealed that 84.5% of the variation in hourglass shape can be explained by principle components 1, 2, and 3, which corresponded to hourglass size (PC1), the separation between hourglass parts (PC2), and the slenderness of the hourglass (PC3). Both a black widow's condition and species significantly predicted hourglass shape; however I found no support for localized Müllerian mimicry within different geographical regions. My results suggest a relaxed role for selection on hourglass shape. I discuss several hypotheses that could explain the variation in hourglass morphology including that potential predators may avoid any red markings rather than an exact shape (categorical rather than continuous perception).</p><p>In chapter four I expand on my work from chapter two to examine the eavesdropper's perspective on private communication channels. Private communication may benefit signalers by reducing the costs imposed by potential eavesdroppers such as parasites, predators, prey, or rivals. It is likely that private communication channels are influenced by the evolution of signalers, intended receivers, and potential eavesdroppers, but most studies only examine how private communication benefits signalers. Here, I address this shortcoming by examining visual private communication from a potential eavesdropper's perspective. Specifically, I ask if a signaler would face fitness consequences if a potential eavesdropper could detect its signal more clearly. By integrating studies on private communication with those on the evolution of vision, I suggest that published studies find few taxon-based constraints that could keep potential eavesdroppers from detecting most hypothesized forms of visual private communication. However, I find that private signals may persist over evolutionary time if the benefits of detecting a particular signal do not outweigh the functional costs a potential eavesdropper would suffer from evolving the ability to detect it.</p> / Dissertation