When a glimpse is enough: Partial mimicry of jumping spiders by insects

Area of Science:

• Evolutionary Biology
• Animal Behavior
• Computer Science

Background:

• Many insects, including flies and moths, mimic the appearance of jumping spiders.
• This mimicry, termed 'partial mimicry,' differs from Batesian mimicry as it involves spider-like patterns on only parts of the body.
• The effectiveness of this mimicry may depend on the viewing angle, suggesting predators target mimics from specific perspectives.

Purpose of the Study:

• To investigate the hypothesis that partial mimicry is angle-dependent.
• To utilize Deep Convolutional Neural Networks (DCNNs) as a tool to analyze insect mimicry patterns.
• To understand the role of predator visual perception and cognitive shortcuts in the evolution of partial mimicry.

Main Methods:

• Training a DCNN on images of jumping spiders, focusing on key features like eyes and legs.
• Testing the DCNN's ability to classify images of mimicking flies and moths as jumping spiders.
• Analyzing classification probabilities across different species and viewing angles, considering predator visual acuity.

Main Results:

• The DCNN successfully misidentified mimicking insects as jumping spiders, with varying probabilities depending on the species.
• Classification accuracy was influenced by the angle from which mimic images were presented.
• The study demonstrated that mimicry effectiveness is linked to signaling angle and predator cognitive processes.

Conclusions:

• Deep Convolutional Neural Networks are effective tools for testing evolutionary hypotheses in mimicry.
• Partial mimicry likely evolves due to the interplay of signaling angle, mimic orientation, and predator reliance on cognitive shortcuts.
• Future research should incorporate predator visual system properties, like ultraviolet vision, for a comprehensive understanding.