Jumping Spiders' Predator Mimicry Proves Perplexing Even to Computers

Travel restrictions caused by COVID-19 forced biologists from the University of Cincinnati to find alternative methods to study jumping spiders' mimicry of wasps. They presented digital portraits of jumping spiders, praying mantises, and wasps to a computer algorithm to determine if it could accurately identify them. The study, published in the journal Behavioral Ecology, found that even the computer was fooled about 20% of the time, with the algorithm incorrectly identifying the spiders as wasps.

Key Takeaways:

  • The study, led by University of Cincinnati student Olivia Harris, used computer vision techniques and machine learning algorithms to classify images of jumping spiders, wasps, praying mantises, and flies.
  • The algorithm accurately identified 13 species every time, but misidentified Maratus vespa and several other spiders as wasps over 20% of the time.
  • The researchers used the images to test their hypothesis that male jumping spiders use sensory exploitation to their advantage in attracting female mates.
  • The study found that the predator illusion works best at greater distances or in the female spider's periphery, where she relies on monochrome green vision.
  • UC Associate Professor Nathan Morehouse said that the females will break the illusion once the male gets closer, and it's beneficial for the males to break the illusion to make mate choices.

Statistics:

  • The algorithm was fooled about 20% of the time when trying to identify digital portraits of jumping spiders.
  • The algorithm correctly identified 13 species every time, but misidentified 49 species (including Maratus vespa) over 20% of the time.
  • There are 62 species in the study, and the algorithm misidentified 11.29% of them (7 species).
  • 20% of the time, the algorithm identified Maratus vespa and several other spiders as wasps.

Sources:

  • Behavioral Ecology, "Computer-based classification of jumping spiders and wasps" (study published in the journal Behavioral Ecology)
  • Olivia Harris, University of Cincinnati student and study lead author
  • Jürgen Otto, researcher and co-author of the study
  • David Knowles, researcher and co-author of the study
  • Nathan Morehouse, UC Associate Professor and study co-author