Study also finds repeated yelling may improve identification – ScienceDaily

As dusk begins to obscure the Masai Mara grasslands in southwestern Kenya, a spotted hyena slips under the wooden canopy that is the acacia tree.

The carnivore pauses, its round ears point forward as a faint sound of sailing, an airborne message travels three miles at 767 miles per hour. again, then again. Whhhhhooo-OOOppp! Here it is … the call of a fellow spotted hyena, repeated quickly enough to attract attention. Warning of lions in the area, perhaps, or of one of the tribes of hyenas encroaching on other lands.

To help or not to help? With so much ground to cover, and so many potential dangers behind, the answer can depend on who exactly is on the other end of the long-distance call. For spotted hyenas, identification is no laughing matter. It’s Dickey’s problem, says a new study by Kina Lyman of the University of Nebraska-Lincoln and colleagues.

By applying machine learning to audio clips collected from the field, the team concluded that hyenas hitting individuals have unique signatures — a form of caller ID distinct enough that hyenas are likely to distinguish one from the other. For the first time, the researchers were also able to quantify the frequency of a call, as spotted hyenas do, which may improve their odds of being identified.

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

“Hyenas don’t treat everyone in the clan the same way, so if they decide to show up and help someone, they want to know who they’re going to help with,” said Lyman, a postdoctoral researcher in Nebraska. .

In its search for acoustic signatures, the team turned to what is known as a random forest model. The researchers first trained the model by providing it with the identities of each hyena they recorded, along with a massive number of vocal traits extracted from each of its cries.

From there, the model used a randomly selected series or spell of chanting from a single hyena to generate decision trees. Each branch of the tree represents a binary choice in a phoneme from a set that was also randomly selected. The model might start by dividing the hyena that shouts at higher frequencies against lower frequencies, for example, and then splitting these groups into longer calls against shorter calls, for example, and so on. In the end, the head of each branch represents a vote for a particular hyena.

After collecting 500 of those random decision trees – a random forest – the model predicted a particular noisy identity based on which hyena got the most votes out of those 500 trees. The team put their trained model to the test by asking them to decide which of the thirteen hyenas produced A random bout of shouting, then repeat this test 999 times.

The model correctly paired a squawk with her hyena roughly 54% of the time, or about six times more than would be expected by chance. This success rate indicates that there is sufficient variation in the sounds of the different hyenas, and sufficient consistency within a single hyena’s screams, for the model to be able to reasonably distinguish between them. If the model can discern those differences, Lyman said, it’s reasonable to assume that hyenas can, too.

Three attributes of shouting seemed particularly useful: call duration, highest call frequency, and average frequency during the most consistent part of the call on the pitch. And the greater the disparity in those traits, the more likely the model—and perhaps the hyenas themselves—distinguished the sources of the chant involved.

However, 54% is well below 100%, even before accounting for the challenges inherent in communicating with a fellow Maasai Mara hyena. For example, spotted hyena clans can swell to over 125 members, a number that seems to strain even the most capricious and tight-knit memories. There is also the potential for audio differences in transmission to be lost, especially when those signals travel several miles before they reach round ears. Meanwhile, calls from wind, rain and other animals can make signal noise.

“There is an understanding that one way to get your message across is to repeat it, especially if you are in a noisy environment or if you communicate over long distances,” Lyman said.

Previous research has shown that penguins, for example, repeat their calls more often when the wind picks up. Other studies have found evidence that different types of animals prefer repetition under similar noisy conditions. But as far as Lyman and her colleagues were able to tell, none of them quantified the extent to which repeating an animal call could improve information transmission.

So the team turned again to the random forest model. When the model guessed the hyena’s identity based on just one shout, it correctly correlated that identity only about half as much as it did when supplied with three sounds. This accuracy increased further with additional calls, peaking at seven beeps.

“It’s like getting more information (every time),” said Lyman, who previously studied vocalization in orcas. “The first time you hear it, you might notice: Oh, that was definitely a male or female voice. Then, the next shout-out, you might be able to narrow it down even more.”

Lyman and her colleagues knew that the calls of some animal species also contain signatures that distinguish the groups to which they belong from other groups of the same species they may encounter – somewhat closer to human dialects or dialects. She noted that some researchers who have studied orcas have become familiar with signatures of the horns that researchers can instinctively discern. (One researcher claimed that calls of a particular pod were “more nasal” than calls of others.)

Due to the size of the spotted hyena clans, Lyman thought that their shouts might also use a particular group’s signature.

“Obviously, if you just have to remember what your group sounds like, and you don’t have to remember each of the 100-plus individual votes, that would be a lot easier,” she said.

When the researchers went looking for a collective signature in the random forest, they were unable to find one. One possible explanation: the apparent ability to save a lot of individual signatures may have made the clan signature either useless or, at best, unhelpful enough to bother development.

“If you know who an individual is, you know what group they belong to,” Lyman said. “Animals are very good at correlating this information.

“So if they needed individual signatures for other reasons, perhaps there was also no need to develop a collective signature, which this finding suggests. They should be able to keep track of all individual sounds and be able to distinguish: If this is individual X, they In my group. I can choose to help them based on their being a member of the group, but there are probably more decisions to be made about whether they are a fellow group member that I really want to help.”

A million different stars have to line up.

All of the team’s findings — the presence of individual signatures, the absence of a clan signature, the utility of repetition — ultimately originated not from a random forest but from the savannah in Kenya’s Masai Mara National Reserve. There, Kay Holkamp and colleagues from Michigan State University have been conducting research on the spotted hyena since the late 1980s.

Liman herself spent a year in the Maasai Mara, whose name is derived from the Maasai people who inhabited it long ago. From 2014 to 2015, a doctoral student at the time and several colleagues traveled regularly west from the Kenyan capital, Nairobi, to a field site in the reserve.

Lyman said, who soon learned that there was a large canvas tent and a soft bed waiting for her. “But we were spoiled there, to be perfectly honest.”

If the facilities were more accessible than expected, the data collection proved anything but. From their vantage point in a Toyota Land Cruiser, Lyman and her colleagues were pointing a directional microphone out of the window and flipping over the voice recorder. Unfortunately, the team was largely subject to the vagaries of Murphy’s Law.

“There is no need to drive. The car must be stopped,” she said, noting that her engine was drowned out by the sounds of the Masai Mara. “And the hyena has to squawk. And you have to actually be able to… see who it is. They can’t be in the bush. And they have to be close enough so you can get a good score. And the other the hyenas have to be quiet at the same time. Time. There are just, like, a million different stars that have to align to get a good score that you can then use in an analysis like this.”

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

“With this portable recording device, we were opportunists, recording constantly and just hoping they would yell for us,” she said with a laugh.

During those months of hope and waiting, the researchers kept busy observing and chronicling behaviors that would benefit other studies. As they did, they glimpsed the individuality that their analyzes of the hyenas’ cries would confirm, years later.

“You certainly know that different individuals have different personalities or may react a certain way in different situations,” Lyman said. “So it’s always fun to learn about hyenas, their little interactions, and the drama that might happen in their lives.”

Source link

Related Posts