Predation Experiment
For the predation trials, I deployed four different treatments of artificial larvae in a 2 x 2 factorial design: eyespotted in leaf rolls, eyespotted on open leaves, non-eyespotted in leaf rolls, and non-eyespotted on open leaves. I affixed prey to individual host plants, interspersed by both treatment and host plant species (S. albidumor L. tulipifera ). I generated unique treatment assignments for each trial. During deployment, I selected the artificial prey’s location on the plant haphazardly and measured its height (in cm) from the ground. Prey were placed 11.0-281.0 cm high on plants (mean: 83.5, SD: 44.1cm), comparable to the heights of live P. troilus prey I observed in the field (25-164cm, mean: 83.8, SD: 40.5cm; supplemental table 1).
For the open leaf treatment group, I attached artificial prey to the adaxial side of fully expanded host plant leaves by poking the loose wire-ends of each prey through the leaf, then twisting them tightly around the midrib. For the prey in leaf rolls, I attached them to the leaf in the same way, then folded the leaf over the artificial prey and secured it shut with a strip of Scotch double-sided tape (Fig. 1e). I positioned all prey with the “head” pointed up towards the leaf petiole, which reflects this species’ typical resting position (Fig. 1a, 1f). During Trial 2, I also included a fifth treatment group of eyespotted prey in leaf rolls, oriented down away from the petiole, to test the effect of eyespot orientation on predator perception. However, as I was unable to confidently determine whether predators perceived this treatment group as eyespotted or not, I excluded data on these prey from the final analysis. Before the start of the trial, I took a photo of each artificial caterpillar in place.
I collected artificial prey after approximately 5 days of exposure (mean: 121 hours, SD: 8 hours). At the end of each trial, I visually inspected prey for evidence of predation, using the pre-trial photographs as a baseline for non-attacked prey. Avian and mammalian attacks are clearly distinguishable by the shape of the bite-marks in the clay (see supplemental figure 3). I recorded any missing prey items (that could not be found after carefully scouring a 1m2 area around the original location) as attacked by an unknown predator. I photographed all recoverable prey with visible attack marks. In a few cases (n = 9), either the artificial caterpillar or the entire leaf roll fell from the plant with no sign of predator damage; these prey were excluded from analysis. I also excluded one artificial caterpillar in which a live P. troilus caterpillar had crawled into a leaf roll, and three artificial prey where the plant could not be found (and thus the clay caterpillar could not be recovered). In total I deployed 809 artificial caterpillars. Of those, data from 659 artificial caterpillars are included in the analysis, excluding prey attacked by non-avian predators. Because birds are common visual predators of insects (Nyffeler et al. 2018), avian predation patterns are the most informative (relative to mammalian or unknown predators) for the goals of this study (Hossie and Sherratt 2012, 2013). Additionally, as expected, mammalian and unknown predators showed little variation in predation between experimental treatments (see supplemental figure 4). As it is difficult to detect evidence of arthropod attacks using clay caterpillars, these types of predators were not considered in this experiment.