Farming maize in ice age Michigan, predicting the future climate of cities, and our host takes a quiz on the sounds of science
43:05 Share

This podcast is supported by the Icahn School of Medicine at Mount Sinai, one of America's leading research medical schools.

The school is the academic arm of the Mount Sinai Health System in New York City. It's consistently among the top recipients of NIH funding. Researchers at the Icahn School of Medicine in Mount Sinai have made breakthrough discoveries in many fields vital to advancing the health of patients, including cardiology, cancer, immunology, neuroscience, and artificial intelligence. The Icahn School of Medicine at Mount Sinai. We find a way.

This is the Science Podcast for June 5th, 2025. I'm Sarah Crespi. First up, staff writer Paul Vussen talks about the tricky problem of regional climate prediction. Although global climate change models have held up for the most part, predicting what will happen at smaller scales, like the level of a city, is proving a stubborn challenge.

Next on the show, evidence in the upper part of Michigan that indigenous peoples once cultivated maize for 600 years, even during an ice age. Researcher Madeline McLeister talks about using LIDAR to search among the trees and the anthropological conundrums raised by such extensive agriculture without nearby urban centers.

Finally, producer Kevin McLean quizzes me on some mysterious sounds that have appeared on the site as part of news stories. I am not going to give any clues so you can play along. Climate change is happening and scientists are working tirelessly to model the global changes in temperatures, effects on polar ice, sea levels, all these things.

planet scale effects. But these models don't contain the detail needed for a lot of local decisions that have to be made. How much will this river in this city rise or fall? Will we have more droughts in this particular county? This kind of information helps if you need to decide on what infrastructure to build. This week in science, staff writer Paul Vussen takes us through the challenges of regional climate prediction. Hi, Paul. Welcome back.

Hi, thanks. Sure. One thing that really struck me about your story was this we're here aspect that kept coming out. That so much projecting of climate change has been done, you know, what the effects are going to be modeling into the future. But now the researchers that you're interviewing are saying we can take data and refine the models. We're here. We're in the global climate change period. Yeah, we are in the future climate period.

It was never easy to do climate modeling, but you used to have nothing to compare it to. You know, now with CO2 levels building up enough and warming to 1.5 degrees-ish, not really quite that, but around that, it's just showing up enough that you can see this regionally, all these effects that have been well predicted by the models and some not so well predicted by them. What were some of the surprises? This is kind of a sidetrack, but I do think it's really interesting.

Kind of the classic one is the Eastern Pacific Ocean has been cooling a little bit or not warming, depending on your interpretation, for decades now when it's supposed to warm a lot. That's a big driver of weather. And it's going the opposite direction of what the models predict. But we've also seen things like rainfall predictions that we've had rain happening in places where we thought it wouldn't be increasing or you've had big heat waves in Europe beyond what the models are.

or have predicted. Little pockets of this stuff pops up all over the world. And we'll say little pockets because on the whole, globally, the predictions are there. This is happening. But once you scale down into these smaller regions, things get less clear about what's going to happen. So let's...

Let's take it back up to the planet scale modeling. How does that work? And then we'll talk about how the smaller scale works. Yeah, so your classic global climate model, GCM as they call it, essentially you divide the world into say like 100 kilometer scale boxes and the flows of heat and weather that you pass from one box to another using fluid mechanics. But everything that's smaller than that box, which is

a lot of stuff has to be handled with more ad hoc parameterizations, they're called, where they're using our physical understanding of clouds or land surface processes.

how things flow over a mountain to approximate what would be going on. Add this together and you get your kind of model prediction. And you could see how trying to model a desert, mountain, river, all that stuff, it adds a lot of complexity to this modeling. So onto the main point of your piece. A lot of climate-related decisions need to be made by governments, cities, companies, families, and the regional prediction is a lot harder to do. So what

So let's talk about the example you kind of used to walk through this problem in your story. This is Austin, Texas. So jealous that you got to visit. Although you did go to a water treatment plant, so maybe not. Not as smelly as you might think. So can you talk about the situation there, like what they're contending with? Austin right now is...

in a drought. But from 2007 to 2016, I think it was, they had this truly massive drought, kind of the worst in modern history in Central Texas. And that really prompted them to use climate modeling as part of this long-term 100-year plan for their water system called WAP.

water forward. So they're very driven by a river that comes through there and they're drinking water. It's rain fed. They're not extracting stuff from groundwater like some other cities do. One input into this was you couldn't just use what the past climate was. There are good reasons to think that there could be more drought in the future, more dry days, rainfall getting concentrated into fewer days with bigger storms as the warmer atmosphere can hold more water.

So they asked folks to project out what the regional climate signal might be. What are some of the choices then if they are trying to model the future of this one specific location? What can they do with a global climate model to make it particular to where they live? One researcher said to me,

So global climate modeling is kind of like your classical music. The effort to get these local signals is more like jazz. There's a lot of bebopping and scatting that goes into this. So there's this thing, it's called bias correction and downscaling. It's kind of like a police procedural where they have a

a crime scene photo and they see something and one little part of it and they zoom in and say enhance and it just magically enhances the information density increases somehow to actually support decisions. Scientists have tried to figure out the best way to enhance while staying true to what both we know of the past and what we could know of the future. And this is very tricky. There's no real perfect way to do this. Either you could be biasing what you get

to the past or to climate models that have biases themselves and may not be perfectly rendering what the local atmosphere looks like. And this is the give and take that there's no perfect answer to. And this is what Austin had to work through itself and what local managers are working through themselves all across the world.

How much do I trust these results? How do I get to what I need to know? Downscaling and bias correction are kind of the two main things that they're looking into. So what's bias correction? Essentially, you might get a kind of totally different number for rainfall than what it seems like you should get from the climate model for what you know has been happening, right? And so you use these

existing historical records of rainfall to modify what you get from the climate model to bend it towards the ground truth. But the problem is reality of the present does not necessarily capture reality of the future. And so these bias corrections, if they're kind of only tweaking, if you're being very careful about how you do it, it's like only modifying it a little bit. It's okay. Sometimes they can truly change the signal you're getting from the climate model where

It doesn't matter if you're just inputting random noise into it and you get the same thing as what the climate model gives you. So, you know, you're not getting any future insight from the climate model. Yeah. So you're overpowering any kind of input from the modeling with your bias correction. And then downscaling is what just being able, this is like looking at a smaller trunk in greater detail. Yeah. So essentially going from this coarse scale to the kind of more local landscape. There are lots of different ways to do this, but...

You can use the weather stations as well to go back to it. You can also downscale using higher resolution modeling. Could you just make your boxes smaller and model more finely across the whole globe? It's super computer intensive and not everyone can do this. Austin couldn't do this.

But it is a best practice. But you still have the problems where this is a one way process. So all the stuff from the global climate model is being fed into the regional climate model. But the stuff that comes out of the regional climate model is not going back to the global climate model. So you're still inheriting biases. So what did Austin end up doing? What were kind of some of the choices they had to make in order to model their future?

So they've done it twice now. So the first time they kind of went with best practices at the time, which meant using the output from every known climate model, essentially only using one scenario, very extreme temperature rise to inform their projections. They got drought signal, but there was kind of shock at some points for like the 2100

They had these huge rainfall events, far more than they would have thought that even canceled out the drought signal. Then they were like, you know, these could be a problem. Is this something we should be worrying about? Should we worry about flooding? You know, they have like flood control systems because flooding used to be a huge problem there. But this just caused a lot of uncertainty and concern. And so.

Some of these bias correction models have been shown to cause things to get wetter, or is this the model that's causing this? Anyway, they went through a whole second round that they just finished this past year, kind of using researchers at UT Austin, University of Texas.

They were much more careful in just using climate models that seemed to capture the weather system around Texas better, using a few of those that required a lot less bias correction. When they did it this time, they saw that drought signal. They still had more rain in kind of individual days because that's just what you expect, but didn't see these huge flooding events. So it seemed more reasonable. How much this is really needed, that's still an open question. Could they just do a...

a more shorthand assumption for this and get near the same result? Maybe. What are some common pitfalls for people who are trying to find out what might happen in their region? What could they do that maybe is just generally not going to work? So there are risks of just believing too much. One thing is the federal government has not, and certainly is not in the near future, gone into climate services in this way or making these kind of

regional, local projections, partially because of these issues. But now there have been a lot of people or companies are being pushed to disclose climate risk and there are purveyors with like black box models of the future that claim to capture down to like a block, a city block of kind of... Your rainfall for the next 50 years on your block. You talk to a climate scientist about these and they're just like, oh my God, they're just, they can't even. So...

You have to be skeptical and work with people who know the region well. There are some promising new research solutions that are coming down the pike that could help this as well.

So these are the ones that we were talking about before where they could just increase the resolution of the modeling? You can take that to the global level, right? Instead of having these hundred kilometer scale global models, now like in Europe, they're running it at five kilometers for the whole globe from 2020 to 2040. It's called destination earth. We've

covered this before in science, then you can just skip the whole process. You get a globally consistent, already local, accurate kind of model. Sounds good. Is it expensive? Energy intensive? Yes, yes. But, you know, if you do this, then maybe you can use AI to learn from it, build more ensembles or this information is out there.

It's one model. It's not a suite of models. They haven't really validated all this yet, but that's one big push. Using AI to do this downscaling is another big push of like the type of AI that's now being used for weather forecasting, right? It's also been shown, it's learned how to do that with the weather that can also do downscaling to a certain extent. I was going to say-

When does climate prediction meet weather prediction? You know, are we going to be able to close that gap and say 50 years it'll be like this and in three weeks it'll be like this? Yeah, well...

They definitely, I mean, it's all connected to each other, right? Weather models or climate models or weather models, but the timescale you're trying to get at matters a lot. But ECMWF, the European weather forecaster, they also do climate and they're behind that European effort. And they're using AI as well to do downscaling for local climate. So there's potential kind of solutions there. Can we touch on the other story that you've been working on lately? And this is about weather prediction, get

getting past what is seen as the limit right now. It's been a truism for decades that essentially there's a limit to the predictability of weather around two weeks. This is often ascribed to Lorenz and the famous butterfly effect of the 60s. Lorenz, I think it's been shown that he didn't suggest a two-week limit. That's more derived from early climate models run by like Charney and others. The basic idea is

The true idea of the butterfly effect is that errors in the small scale of your model of the weather, those propagate, increase much faster than the large scale. So eventually those errors at the small scale overwhelm your system. So even if you know perfect state of the atmosphere today, you will still not be able to predict the weather past a couple of weeks or so because of this chaos that comes from these small scale kind of cloud formation patterns.

Now, some reachers have used one of the new AI models developed by Google. We know the weather that's happened in the past. And so they said, hey, model, OK, starting from this kind of weather, we already know we're going to work back through you to figure out what the perfect initial conditions is, what they're called perfect initial conditions for you would be. And with those perfect initial conditions, how far out can you predict the weather? And that's when it shows accuracy.

oh, it could predict 33 days, potentially more. That's when they ran out of memory in their kind of computer system. That's pretty amazing. So how much of this have they done? How close is this to the clinic? Yeah, it's, I don't know. So, I mean, they did this for 2020, all of 2020. It wasn't this one-off thing, but.

But the big issue here is they already know what the weather is. They need to do it for an actual prediction. Like the theoretical limit maybe is not there, but it also...

What we don't know for sure is this kind of perfect initial condition that they derived. Was that actually what the reality or was that just what they formed to then give them the present day weather? That's still uncertain. But the tweaks the model seemed to make all seem reasonable. You know, one way this could be really helpful is we have these things called reanalyses that essentially it's like capturing the weather over time using weather records. And you can run these

this system on these reanalyses and perhaps improve those a lot. These reanalyses are what these AI models then train on. And so then perhaps these AI models get better. So it's not close to kind of happening, but it also suggests that maybe, maybe we'll see for sure in the future, because right now we're pushing at that 14 day limit. The weather researchers would say this is not a time to divest from basic research and weather

weather, which is, you know, a imminent threat in the United States as well. That's true. So someday we'll be able to pick where we want to live, depending on climate, 100 years out and book our vacation on a not rainy weekend, six months out. That's what I'm hoping for. The perfect future. I don't know if you're going to get six months, but there probably still is an intrinsic limit. It's just what that exactly might be. We're not sure.

Going back to climate, the longer look instead of weather, it sounds like there are companies that are trying to sell black box models because there's a demand for this information from shareholders saying you need to future proof your supply chain or in towns trying to decide on infrastructure for decades to come.

So can you make a prediction for us on how this field's going to go? I think it's going to see a lot more kind of investment from climate scientists or as kind of possible solutions crop up, maybe there'll be ways forward. But there are also simpler ways of thinking about this, where depending on the severity of the signal that you're worried about and the infrastructure you're dealing with, like

Sea walls are one thing, and there's a whole different breed of uncertainty with ice sheets and stuff like that. But the climate models on temperature do pretty well. Even locally, there's some uncertainty. But for more heat, yeah, that's pretty easy. Yeah, there are certain things you don't have to worry about. What about my tornadoes? We've had 40 tornadoes already in Indiana this year. The climate signal to tornadoes is not very clear. There are signs of tornado alleys pushing east, but...

but you just jumped from the most sure to the other end of it. So definitely hotter, maybe not so clear on the tornado effects. Okay. Yeah, or just rain. All right, Paul, thank you so much for talking with me. Yeah, my pleasure. Paul Voussin is a staff news writer for Science. You can find a link to the story we discussed at science.org slash podcast. Stay tuned for a conversation about evidence of intensive indigenous farming in Michigan's Upper Peninsula.

What was farming or agriculture like in the Americas when Indigenous people had the run of the land before colonists arrived? It's actually really difficult to tell because of the way the land has been turned over.

by people who came later, both for farming and also just for urbanization. This week in science, Madeline McLeister and colleagues write about using LIDAR, which is laser imaging of the ground from the air, and excavation to uncover evidence for intensive agriculture by indigenous peoples in the upper peninsula of Michigan during the Little Ice Age. This is very far north for what they are growing.

the subtropical plant maize, and at much larger scales than expected. Hi, Madeline. Welcome to the Science Podcast. Hi, thanks so much for having me. Sure. So the Upper Peninsula of Michigan, I've actually never been up there, but it's very far north. It's tucked between Lake Superior and Lake Michigan, right under Canada. Not a place that you might expect to be covered

covered in cornfields. What made you decide to look there for this? This area that we're working at, the site is called 60 Islands, and it's one of several sites associated with Ana M. Omet, which is on the National Register of Historic Places. It's a cluster of archaeological sites that are ancestral menominee, and these are

very important to the contemporary Menominee Indian tribe of Wisconsin. And they were keen for

for us to go record some of these features before they might be destroyed by things like climate change or also an impending mine that might be coming in the area. Okay, so this was part of a collaboration where there was kind of a scientific interest, but then also an interest from the people whose ancestors might have created this agriculture, who've done this agriculture. That's right. Yeah. So these were agricultural fields created by their ancestors. Can you

Can you refresh us on how LiDAR lets you see features? That is so funny. In the paper, there's like the LiDAR and then there's the ground view and it's just greenery. It's just green ground cover and trees. So you can't see this necessarily by taking pictures from airplanes. You have to do something else. What does LiDAR get you? So LiDAR just sends lasers down to the ground and this can penetrate the tree canopy to a certain extent. And then the

they kind of hit the ground and go back up. What that does is it creates a map of the topographic surface of the earth because the forests are so dense in the upper peninsula of Michigan. We conducted our LIDAR during May, which is leaf off. So they're not just hitting a bunch of leaves. It's

And then you have to filter the point cloud. And this is detailed in the paper. And then you get a really nice picture of the ground cover. They are very difficult to see on the ground. These features have a pretty low relief, anywhere from like 10, sometimes to 25 centimeters off the ground. The second feature

the vegetation starts growing. Even just when you're standing on the ground, you can't see them. Even if the vegetation isn't growing, they're hard to see. So they weren't identified by archaeologists until the mid-1990s. So yeah, from the LIDAR imagery in your figures, this kind of looks like, I guess it would be a

patchwork of striped squares. They're kind of like if you have a quilt and have squares and each square has stripes in it. So what exactly are we looking at when we talk about these ridges with this low profile of elevation? These were constructed garlands

garden beds. They can be up to a meter wide. The ones at 60 Islands are a little bit skinnier than that, probably about 50 centimeters wide, but they really vary, especially when you're talking about how extensive they are. And this patchwork, what it is on the ground is these are individual beds where the ancestral Menominee farmers would have planted harvests

probably maize, beans and squash, along with a suite of things for their, you know, based on personal preference. There's evidence of melon and sunflower at the site as well. You were saying before that there are one, one and a half meters wide. How long are they when we're looking at these squares from

from the air. So they really vary, but they're typically about 20 to 30 meters and sometimes up to 50 meters in length. If you think about fields, they're the rows where everything was planted. They're just elevated a little bit. How much was this compared to previous finds? Like how much more area do we add to this? So we're

we've added about 10 times the area and we haven't even mapped it all yet. And it's not just the area. So it was mapped in the early to mid 1990s by a woman named Marla Buckmaster. And I will say she did a fantastic job mapping very, very difficult to see features without the technology that we have today. But what LIDAR gives you is this opportunity to cover just so

such a bigger area as well relatively quickly. So our survey, we covered 135 hectares. And within that 135 hectares, 95 of those hectares were densely clustered agricultural garden beds. So within that nine and a half hectares that was previously known, we found

even more ridges in that nine and a half hectares. And then beyond that, we found ridges far, far exceeding what we had previously known. So we're just seeing a total reworking of this landscape for agriculture. All right. We're going to get into the implications of that a little bit later.

But one of the things you when you kind of identified the scale of this and where all these different ridges were, then you went and sampled them. So you went in and looked at the content of the ridges and you did radiocarbon dating and other analyses. What did you find looking deeper into these agricultural fields? The neatest thing that we found through the cross sections was that we're seeing expectations.

episodes of rebuilding in the stratigraphy. So not only are you seeing that this huge area is being dedicated to agriculture, they're using the same spaces over and over again. The radiocarbon dates give us about a 600 year window when this landscape was in use by the ancestral Menominee community. So that would have been about

1000 CE going up to 1600 CE. And then we're also seeing lots of small finds in terms of artifacts, things that are really small, things that you would drop and go missing and lots of charcoal. And so the way that we interpret that is we're seeing that they're using compost basically from their household.

to mix in with these soils, as well as using wetland soils to mix, to build probably structure as well as fertilizer. So it's not just the topographic reworking of the surface. It's also a reworking of the soils themselves.

And there's no village in the immediate vicinity. We actually don't know where the village is for all these agricultural fields. So stay tuned. Yeah. Did you find any other evidence of habitation besides the fields? Absolutely. So this is deeply enmeshed in a really important landscape. I mentioned this Anaem Omid broader district. And what that's composed of is a range of burial mounds,

dance rings. And some of those are in the paper as well. So the other great thing about LIDAR is, you know, it's not looking just for the fields, right? Anything with a topographic relief, it'll pick up. Even a fur trader shack. That's exactly right. Yeah. So you're talking about composting and remediating the soil, huge scale, like so intensive as in there's a lot of work going on, but also there's a lot of crops. And it's

cold. We haven't really emphasized this. This is kind of the upper limit for where corn can be happy. What does this all mean put together about what agriculture was like at this time? We were just not anticipating the scale of this till we got back home. For archaeology, when we think about intensive agriculture and the impetus for intensive agriculture, we're typically thinking about

hierarchical societies, increasing inequality. We're also thinking about things like population pressure and everything we know about ancestral Menominee communities. They're largely egalitarian. They're semi-sedentary, meaning they move on the landscape as opposed to staying in one place. And we're not seeing...

Like I said, we don't know where the village is. So there's no clear population pressure. And then just the scale of it in this unlikely location. So that tells us, one, some of our foundations, archaeological foundations of agriculture need to be rethought. There's so much debate about the role of agriculture

agriculture and people staying in one place and like kind of elaboration of culture. Did cities come because of agriculture? Because you could have food and have a lot of people in one place or, you know, all of the arrows are kind of in question here. Which way things go? Right. So there's no large urban center here. There's no kings. There's there's

Cool questions. And then just labor organization. I mean, this is not an easy project to reshape 95 hectares with changing the soil of it all. I mean, it's a massive undertaking. And so how that labor was organized...

raises these really critical questions. And then also what's interesting is that if we're seeing this level of reworking in the kind of most unlikely of places in Eastern North America, because this is the largest preserved raised bed field system in Eastern North America now, in the Upper Peninsula where we wouldn't expect it, what were the size of fields in places where we would expect

Right, where things are easier to grow. Where we have these large population centers. Further south. It really throws into question a lot of what we think about agriculture and the landscape and the efforts of...

ancestral Native American communities in eastern North America. Yeah, really interesting. Where else should people be looking for this? Like part of the problem here is that the Upper Peninsula does have kind of undisturbed land. I guess it's been forested, but still, where else can you look? If you came to Indiana, we have mounds, but we have a lot of cornfields that came much later too, right? Archaeologically, sometimes we do see these garden beds buried somewhere.

So, for example, there's been work done in Indiana at Angel Mounds where you can see the ridges were buried. That's one place we can look for them where they may have been plowed over and so lost in that regard. The other place that we can look for them is...

is using historical aerial photography. I saw your paper on this. This is so cool. That's something my colleague who's a co-author with me on this paper, Jesse Cassana, and I have done as well is we can't do LIDAR, you know, historically, but we can look at some of these aerial photographs of where we can find these fields again. This is so interesting because you have to have photography and airplanes, but you want to be as old as possible before the land was taken over. That's right.

That's right. So we're looking at some of these photos from the 1930s, which is about as far back as we can go reliably. But then I think there's other things that we can do to locate these fields. We need to start looking in less likely places because the good farmland, that has probably been plowed. We're going to not have that topographic relief. But how much have we looked at these forested spaces? Not really, because we just assumed it wasn't there.

So we need to start rethinking where we look for agriculture. And then also, I think, talking to Native American communities and continuing to collaborate with them. Some of their oral histories place agriculture in various locations. And archaeologically, we typically think of the size of agricultural fields going up to 121 hectares. But Blackhawk, who

was a SOC leader in Illinois, he, in his autobiography, mentions a tremendously larger area under cultivation. And so taking that kind of information into consideration, and our work supports the estimates that Blackhawk has, as opposed to the archaeological ones, where we're only seeing remnants. So fields were much larger than we can see archaeologically.

What are the Menominee collaborators that you worked with take away from this? What was their reaction to finding so much of this agriculture going on? So we were really fortunate that a number of representatives from the Menominee Indian Tribe of Wisconsin came out to observe the work.

and to have Dave Greenough, who is the Tribal Historic Preservation Officer, join us on this project. So I don't want to speak for what they think about it, but they, you know, certainly Dave is enthusiastic. One thing that happened during the survey is that we had the drone flying the LIDAR and we had two bald eagles who were nesting nearby watching this very, very closely. And

the representatives from the tribe that were with us told us that this was a good sign. And we were happy the eagles didn't take out the drone. But we...

But it turned out to be not just a good sign, quite a remarkable sign. That's great. All right. I'm going to wrap it up there. Thank you, Madeline. This has been wonderful. Thank you so much. Madeline McLeister is an assistant professor in the Department of Anthropology at Dartmouth College. You can find a link to the paper we discussed at science.org slash podcast. Up next, a Science Sounds pop quiz. You can listen to me do Just Okay and play along.

So often at Science, there are news stories or research papers that come with some sort of audio component, a recording of an experiment or a new organism or something. And those things don't always necessarily have a home somewhere on the podcast. We're not always covering them. But I've been collecting a few of them over the last few months. And Sarah, who listens to a lot of audio, I thought it'd be fun to test out her knowledge. So this is Sounds of Science. ♪

Okay, so Sarah, do you know what we're doing? Sounds of science. Things I've never heard before, but I'm going to guess from context clues. Yeah. Start out easy. Okay, you don't have to close your eyes. Okay, ready? There's sound one. I sense coral reefs munching, so I'm going to go with a marine mammal. Maybe like a, I'm going to just say a whale. Beluga whale.

Wow. You're very close. Humpback whale. Oh, very close. Yeah, very close. This was a story that published a few weeks ago by our intern, Alexa, where they were looking at the...

uh language patterns of humpback whales and found that they're structured similar to human languages i do feel like the first as soon as they started getting good using ai to analyze sound they're like let's get those whales on i know right let's find out yeah yeah see what they're talking about yeah um okay all right i'm closing my eyes because that might help next one oh ah i'm gonna go on it oh do you have a guess i'm gonna say it's a frog

A frog? Yeah. Okay. You know what? Honestly, you're not in terribly wrong territory. It is an amphibian. Do you remember? Does the word Sicilian mean anything to you? Oh, it's one of those milk-giving worms? Yeah. Oh, no. So this one, yeah. This one was from a little while ago, but these are... These are...

These legless amphibians called Sicilians, they really look like a giant worm. But yeah, this was a story. This was actually a paper in Science. And yeah, these legless amphibians, the paper was reporting on their ability to provide the amphibian equivalent of milk to its young. Oh, boy. I'm so glad we're expanding out milk giving to different walks of life.

That's not noisy, right? What's this noise? So that sound is the babies. The babies make this little vocalization near the cloaca of the mother. And they think it's to sort of beg for the milk, I guess. They'll even bite her a little bit. That's pretty amazing. And I never would have guessed because I've actually watched those videos before. Yeah. And I did not have the sound on, obviously. I missed out on this baby.

Baby talk. Yeah. Okay. Should we move on to the next one? I'm ready. I know that one. You already know it. All right. Right out of a dolphin's blowhole. Right out of a dolphin's blowhole. No. Oh, I thought so. I honestly thought you were going to get this one right away anyway, because it was on the podcast. Yeah. Isn't that what it is? No, this one is the Counting Crows.

You want to replay it again? I know. I did it. That's why it was hard. Okay, go ahead. Let me hear it one more time. Yeah. Go back. Okay, you're going to have to... So you may have been responding to the...

Raspberry style noise. Yes, that initial sound is basically the cue that these crows are given. They're given auditory cues in order to give a certain number of vocalizations. I do remember this. This was one where this particular cue represents a number that the crows are meant to give that number of vocalizations for.

One, two, three, or one, two, three, four, like that. Exactly, exactly. I do remember it now, but I definitely didn't remember it in the moment. So yeah, you definitely got me there. Yeah, so this was one, this was actually also another science paper. That was, this was a cool one just because, you know, crows are, we already know crows are like so, so smart. But yeah, it was really interesting to hear how they train them and how they like get them to do all of these, these different things.

different tasks so all right i really hope the next one is not a dolphin because i'm going to just be like always guessing the same thing okay all right um so this is the last one we've got we're underwater again okay it's some kind of shells breaking that's what i'm going with it is what it sounds like um what is it i really don't know this one this is a sound that is made by a shark

And this is a news story. There's been reports of different elasmobranchs, that's like sharks and rays and stuff, that have been able to make sounds. One of our online news editors, Michael, he had mentioned that these attempts to record sounds made by sharks have gone back into the...

back to the 70s but it sounds like this is a sort of first of its kind from from a shark species that they've actually been able to confirm um but they think it's this like bony uh part of their mouth that they're clasping together wow i really had no idea that sharks made noise the ocean is like way noisier than we ever really but they're gonna have to redo that jaws theme it's gonna be like i

I know it's not the most ferocious. These sharks in particular are not super, super big. I think they're a type of dogfish. But yeah, I'm sure there are probably more noisy sharks out there. Wow. As we keep instrumenting the ocean, maybe it'll be something that we'll get to hear more about. Is that the last one? All right. Yeah, that's the last one. I think you did pretty well. No, that was fun though. I was very surprised by that last one especially. All right. Great. Well, thanks for playing, Sarah.

And that concludes this edition of the Science Podcast. If you have any comments or suggestions, write to us at sciencepodcast at aaaas.org.

To find us on podcasting apps, search for Science Magazine or listen on our website, science.org slash podcast. This show is edited by me, Sarah Crespi, and Kevin McLean. We had production help from Podigy. Our music is by Jeffrey Cook and Wenkui Wen. On behalf of Science and its publisher, AAAS, thanks for joining us.