Egyptian fruit bats argue about food, sleeping spots, and other bats getting too close. To us, the calls sound much the same. The bats may be telling apart sounds we can’t hear.
In a 2016 study of Egyptian fruit bat calls, Tel Aviv University researchers Yosef Prat, Mor Taub, and Yossi Yovel recorded bats for 75 days. They matched the calls to video so they knew who was calling, who the call was directed at, and what was happening. Of the more than 160,000 calls they recorded, they analyzed nearly 15,000 for which they had enough detail.
Then came the machine learning. The researchers measured patterns in the sounds and trained a model to classify four kinds of squabble: food, mating attempts, perching, and sleeping. On calls left out of training, it identified the context with 61% balanced accuracy, compared with 25% by chance.
This was not a bat translator. But a call that sounds to us like a generic complaint carried clues about who was complaining and what had happened. Somewhere in that colony, a bat was trying to sleep while another bat got too close. Science has now classified the complaint.
Have a good weekend.
