Several hypotheses have been put forward to explain the evolution of inaccurate mimicry. Here we investigated the novel hypothesis that inaccurate mimicry (in color and shape) is maintained by opposing selective pressures from a suite of different predators: model-aversive visually oriented predators and model- and mimic-specialized predators indifferent to mimetic cues. We hypothesize that spiders resembling ants in color and shape escape predators that typically avoid ants but fall prey to ant-eating predators. We tested whether inaccurate myrmecomorphic spiders are perceived as their models by two types of predators and whether they can escape from these predators. We found that model-specialized (ant-eating) predators captured mimics significantly less frequently than their ant models, because mimics changed their behavior by fleeing predatory attacks. The fastest escape was found in less accurate mimics, indicating a negative association between visual resemblance and effectiveness of defenses. In trials with spider-eating predators, mimics were not captured more frequently than their models. The quality of defensive mechanisms appears to result from opposing selection forces exerted by the predator complex: mimics are more accurate (in color and shape) in microhabitats dominated by model-aversive predators and less accurate in microhabitats with model- and mimic-specialized predators. Copy; 2011 by The University of Chicago.