Batesian mimicry is characterized by phenotypic convergence between an unpalatable model and a palatable mimic. However, because convergent evolution may arise via alternative evolutionary mechanisms, putative examples of Batesian mimicry must be rigorously tested. Here we used artificial butterfly facsimiles (N=4000) to test the prediction that 1) palatable Limenitis lorquini butterflies should experience reduced predation when in sympatry with their putative model, Adelpha californica, 2) protection from predation on L. lorquini should erode outside of the geographical range of the model, and 3) mimetic color pattern traits are more variable in allopatry, consistent with relaxed selection for mimicry. We find support for these predictions, implying that this convergence is the result of selection for Batesian mimicry. Additionally, we conducted mark-recapture studies to examine the effect of mimicry and found that mimics survive significantly longer at sites where the model is abundant. Finally, in contrast to theoretical predictions, we found evidence that the Batesian model (A. californica) is protected from predation outside of its geographic range. We discuss these results considering the ongoing hybridization between L. lorquini and its sister species, L. weidemeyerii, and growing evidence that selection for mimicry predictably leads to a reduction in gene flow between nascent species.