By Alice Lipscombe-Southwell

Published: Thursday, 04 November 2021 at 12:00 am


Even people who aren’t fans of spiders can appreciate the intricate beauty of their webs. It’s even more fascinating when you consider the fact that the arachnids have tiny brains, yet somehow can build these geometrically precise creations.

Now, scientists at Johns Hopkins University have used artificial intelligence and night vision to establish how exactly spiders build their webs.

“I first got interested in this topic while I was out birding with my son,” said senior author Dr Andrew Gordus, a Johns Hopkins behavioural biologist.

“After seeing a spectacular web I thought, ‘if you went to a zoo and saw a chimpanzee building this you’d think that’s one amazing and impressive chimpanzee’. Well, this is even more amazing because a spider’s brain is so tiny and I was frustrated that we didn’t know more about how this remarkable behaviour occurs. Now we’ve defined the entire choreography for web-building, which has never been done for any animal architecture at this fine of a resolution.”

Read more about spiders:

First, the scientists had to systematically document and analyse the behaviours and motor skills involved.

They took six hackled orb weaver spiders, which are small, nocturnal spiders native to the western United States. They selected this spider species as they do not need humid conditions, and can happily co-exist with each other.

In the lab, each spider was placed on a plexiglass box, under an infrared light. Each night, the spiders were recorded using a camera that operated at a fast frame rate, to capture all of their tiny movements as they built their webs.

The researchers then tracked the millions of individual leg actions with an algorithm designed specifically to detect limb movement.

“Even if you video record it, that’s a lot of legs to track, over a long time, across many individuals,” said lead author Abel Corver, a graduate student studying web-making and neurophysiology. “It’s just too much to go through every frame and annotate the leg points by hand, so we trained machine vision software to detect the posture of the spider, frame by frame, so we could document everything the legs do to build an entire web.”