Study also finds repeating whoops can improve identification

As dusk begins to mask the grasslands of the Maasai Mara of southwestern Kenya, a spotted hyena creeps under the woody umbrella that is acacia.

The carnivore pauses, its rounded ears sticking forward as a faint sound sails by, an airborne missive traveling three miles at 767 miles an hour. Again, then again. Whhhhooo-OOOppp! There it is…the call of another spotted hyena, repeated quickly enough to deserve attention. A warning of lions in the area, perhaps, or one clan of hyenas encroaching on another’s territory.

To help or not to help? With so much ground to cover and so many potential perils beyond, the answer might depend on who, exactly, is on the other end of the long distance call. For spotted hyenas, identification is no joke. But it’s a whooping cough, says a new study by Kenna Lehmann and colleagues at the University of Nebraska-Lincoln.

By applying machine learning to audio clips collected in the field, the team concluded that hyena calls exhibit signatures unique to individuals – a form of caller ID distinct enough that hyenas are likely to be able to call them. distinguish from each other. For the first time, the researchers were also able to quantify how repeating a call, as spotted hyenas do, could improve the chances of being identified.

The fact that spotted hyena clans are built on hierarchies of social rank, but consist of several families that meet regularly and disperse across the savannah, makes individual identity particularly important.

“Hyenas don’t treat all clan members the same, so if they decide to show up and help someone, they want to know who they’re showing up to help,” said Nebraska postdoctoral researcher Lehmann. . .

In their search for voice signatures, the team turned to what is known as a random forest model. The researchers first trained the model by feeding it the identities of each hyena they had recorded, along with a considerable number of acoustic features extracted from each of its calls.

From there, the model used a series, or episode, of randomly selected hyena whoops to generate decision trees. Each branch of a tree represented a binary choice within an acoustic trait of an also randomly selected batch. The model might start by dividing hyena calls into higher and lower frequencies, for example, and then further dividing those groupings into, say, longer or shorter calls, and so on. Ultimately, the tip of each branch represented a vote in favor of a particular hyena.

After assembling 500 of these decision trees at random – a random forest – the model predicted the identity of a given whoop based on which hyena received the most votes from those 500 trees. The team put their trained model to the test by asking them to identify which of 13 hyenas produced a randomly selected bout of whoops, then repeated that test 999 times.

The model correctly paired a whoop fight with her hyena about 54% of the time, about six times more often than you would expect by chance. This success rate suggests that there is enough variation in the calls of different hyenas, and enough consistency in the calls of a single hyena, for the model to reasonably distinguish them. And if the model can discern those differences, Lehmann said, it’s reasonable to assume that hyenas can too.

Three traits of the whoops seemed particularly instructive: the duration of a call, the highest frequency of the call, and the average frequency during the portion of the call with the most consistent pitch. The greater the disparity between these traits, the more likely the model – and potentially the hyenas themselves – would be to distinguish between the sources of the respective whoops.

Still, 54% is well below 100%, even before taking into account the inherent challenges of communicating with another hyena in the Maasai Mara. For one thing, spotted hyena clans can grow to over 125 members, a number that seems to strain even the bulkiest and most airtight memories. There’s also the possibility of acoustic nuances getting lost in the transmission, especially when these signals travel many miles before reaching rounded ears. Wind, rain, and other animal calls, meanwhile, can introduce noise into the signal.

“It’s understood that one of the ways to get your message across is to repeat it,” Lehmann said, “especially if you’re in a noisy environment or communicating over long distances.”

Previous research has shown that penguins, for example, repeat their cries more often when the wind picks up. And other studies have found evidence that various animal species promote repetition under similar noisy circumstances. But as far as Lehmann and his colleagues could tell, none had quantified how much repeating an animal call could actually improve information transmission.

So the team again resorted to their random forest model. When the model guessed a hyena’s identity based on a single whoop, it correctly identified that identity about half as often as when it received three whoops. This accuracy increased further with additional calls, peaking at seven whoops.

“It’s like getting a little bit more information (every time),” said Lehmann, who has previously studied vocalizations in killer whales. “The first time you hear it, you might notice: Oh, that was definitely a male or female voice. Then the next scream, you might be able to narrow the range further.”

Lehmann and his colleagues knew that the calls of some animal species also contain signatures that differentiate the groups to which they belong from other groups of the same species they might encounter – somewhat akin to human accents or dialects. She recalled that some researchers studying killer whales had become so familiar with pod signatures that they could instinctively tell them apart. (One researcher claimed that calls from a certain pod were “more nasal” than others.)

Given the size of spotted hyena clans, Lehmann thought their whoops might also use a group-specific signature.

“Obviously if you just have to remember what your band sounds like and you don’t have to remember each of the 100+ individual voices, that would be a lot easier to do,” he said. she declared.

When the researchers went to look for a group signature in the random forest, however, they couldn’t find any. One potential explanation: the apparent ability to memorize so many individual signatures may have rendered a clan signature useless or, at best, not useful enough to bother growing.

“If you know who the individual is, you know what group they’re in,” Lehmann said. “Animals are quite good at associating this information.

“So if they need individual signatures for other reasons, it may never have been necessary to develop a group signature as well, which this finding suggests. They should be able to track all the individual voices and to be able to distinguish: if it is individual X, he is part of my group. I can choose to help him according to his membership of the group, but there may be other decisions to make about whether this is a party mate I actually want to help.”

“A Million Different Stars That Must Align”

All of the team’s findings – the presence of individual signatures, the absence of clan signatures, the usefulness of repetition – ultimately come not from a random forest but from the savannah of Kenya’s Maasai Mara National Reserve. . There, Kay Holekamp and her colleagues at Michigan State University have been researching spotted hyenas since the late 1980s.

Lehmann herself spent a year in the Maasai Mara, which takes its name from the Maasai people who have long inhabited it. From 2014 to 2015, the then PhD student and several colleagues made regular trips west of Kenya’s capital, Nairobi, to a field site in the reserve.

“The first time I went there…I was like, ‘Oh, I’m going to sleep on the floor for 10 months, in a sleeping bag,'” said Lehmann, who soon learned that a a large canvas tent and a soft bed awaited him. “But we were quite spoiled there, to be perfectly honest.”

While the accommodations were more comfortable than expected, the data collection turned out to be anything but. From their vantage point in a Toyota Land Cruiser, Lehmann and his colleagues pointed a directional microphone out the window and turned on an audio recorder. Unfortunately, the team was very subject to the vagaries of Murphy’s Law.

“You must not drive. And the car must be turned off,” she said, noting that her engine drowned out the sounds of the Maasai Mara. “And the hyena has to scream. And you have to be able to… see who it is. They can’t be in a bush. And they have to be close enough that you can get a good recording. And the “other hyenas need to be quiet at the same time. There’s just, like, a million different stars that need to line up to get a good recording that you can then use in an analysis like this.”

Under these circumstances, Lehmann said, patience was more than a virtue. It was a necessity.

“With this portable recording equipment, we were recording opportunistically, constantly and just hoping they would yell at us,” she said with a laugh.

During these months of hope and expectation, the researchers remained busy observing and reporting behaviors that would inform further studies. In doing so, they glimpsed the individuality that their analyzes of hyena calls would, years later, affirm.

“You definitely learn that different people have different personalities or may react in different ways in different situations,” Lehmann said. “So it’s always fun to get to know hyenas and their little interactions and the dramas that might happen in their lives.”


Source link

Comments are closed.