Appendix 1. Mesh enclosures placed in freshwater ponds can be invaded by early instars of invertebrate predators (e.g., larvae of beetles and dragonflies) small enough to pass through the mesh. The predators can then, in some cases, grow quickly enough to hunt and consume amphibian larvae placed in those enclosures. In the main text, we analyzed our data including all enclosures. Here we present an alternate analysis in which we excluded enclosures where we recovered invertebrate predators when the experiment was taken down and the contents were inspected. Among the 10 ponds that did not dry prematurely, invertebrate predators were recovered from 6 out of 20 enclosures. In one pond, both enclosures were invaded, eliminating it from consideration and leaving 5 Present ponds and 4 absent ponds.
In our re-analysis, performance of wood frogs differed between Present and Absent ponds (MANOVA: Wilks’ Lambda = 0.030; F3,3 = 32.022; P < 0.009). Univariate ANOVAs revealed no differences in survival (P = 0.264) but final size in Absent ponds was smaller (P < 0.001) and developmental stage of wood frogs collected from Absent ponds was more advanced (P < 0.001). Performance also differed among ponds within wood frog categories (MANOVA: Wilks’ Lambda = 0.001; F21,9 = 20.327; P < 0.001).
Eliminating invaded enclosures resulted in just one qualitative change in our results: the size difference between Present and Absent ponds changed from a nonsignificant trend to a significant difference. The pattern observed both here and in the original analysis may reflect the influence of temperature. It is likely that lower forest cover around many of the Absent ponds (Figure 1) contributes to higher temperatures in some of the absent ponds. A regression of final development stage against average water temperature shows a strong relationship (Linear Regression: N = 10, R2= 0.878, P < 0.001) in which 3 of the 4 warmest ponds were in the Absent category.