The social lives of rats

The lab I am in mainly studies effort behaviors in rats and humans (and recently pigeons with my collaborators). We want to know what goes on in the brain when animals work at a difficult task. But knowing that a task is challenging to rats or pigeons is difficult. With humans it’s relatively easy to find out how hard they thought a task was; you can just ask them. However, with the rats I work with, I’ve yet to figure out how to ask to them. But this doesn’t mean they couldn’t answer me if I did ask. There’s been a lot of recent work into rodent vocalizations and rats seem to “talk”, or at least they make some fairly complex calls (press play to hear some rat “trills”) that sort of look and sound like talking.

I wanted to see if rats were making any sort of call when they were working hard at a task. We had some money for supplies so I purchased a specialized ultrasonic microphone that could pick up the high frequency calls rats make. It was very simple to setup with just a USB cable and some amplification settings. I installed some free open-source audio software (Audacity; note 5) capable of taking in the massive amounts of data the microphone was picking up from the high frequency ranges. And that was it, I was ready to go!

But this all seemed too easy and nothing in science is easy… When I hit record for the first time I saw a lot of noise being picked up by the microphone. Worse, the noise was right in a frequency range which rats are known to make calls (~20kHz)! I wondered if it was the room I was in so I installed Audacity on a tablet and began wandering around my building with the tablet in one hand and the microphone in another searching for a silent spot. Yet, there were none. The noise was constant and everywhere (Note 2). I even went into a special sound proof booth that another lab had and this loud noise was there too.

Well maybe the noise was just something out of my control; alien signals, the earth vibrating, who knows (Note 2). I figured I would try to record my rats talking anyways and see if their calls would be distinguishable from the noise (Notes 3 & 4). The noise too only went to about 40kHz and rats are known to make vocalizations around 50kHz, in addition to their 20kHz calls, so I figured I should still be able to capture these higher frequency ones. So I grabbed my favorite rat, put him in the testing apparatus, and hit record.

Aaaand there was nothing. Not a single call or shout or muttering or vocalization of any kind. The software I used to find the rat calls, DeepSqueak (Note 6), just could not find anything that looked like a rat call. I think the software started to get desperate and began to pick out other noises, like the sound of a pump or running motors, as rat calls. But it was very clear they were just noises and not rat calls.

I repeated this for a couple days as any good scientist would do and to my somewhat joy I did indeed capture a rat call. But it was one single call over quite a few attempts. The papers I was reading on rat vocalizations were capturing dozens or even hundreds of calls over short sessions. And I had one over hours.

I tried different recording software. I tried different analysis software. I moved the mic around. Most other experiments made sound proof boxes but I couldn’t do this due to my experiment. I reached out to #scienceTwitter for help. They (esp Dr. Jennifer Honeycutt) had some great tips but none resulted in me finding more rat calls. Further, everyone else seemed to be able to just record tons of calls so easily. Defeated, I tucked the mic safely away on a shelf and that was that. It seemed rats just don’t talk a lot during my task so how could I ever listen in to how hard they thought it was.

But a few days later a thought boomed from somewhere deep in my brain; almost all the previous experiments on rat vocalizations are done in a social context! Many are done with a mom and her pups or with males and females. Meanwhile I just had my one rat working by himself. If rat vocalizations are some form of social communication, of course one rat alone won’t be “talking”. He has no one to talk to. With the motivation only a spark of insight can give I plugged the mic back in and grabbed two of my rats who are cagemates. First, I put one rat in the experimental apparatus and recorded him alone for about 3 minutes. Then I put his cagemate in with him. After 10 minutes of them hangin’ out together I removed his cagemate so the original rat was alone again. All the while I have no idea if they are calling or not. Rat calls are well out of the human hearing range so I just have my fingers crossed hoping they’re talking.

I put the rats back in their home and loaded up the audio recording in DeepSqueak. I selected the short rat calls neural network and pressed “Detect calls”. My computer whirled up in search of rat conversations.

Huh, 1,231 calls? That’s more than usual but nothing too special. Remember the software was picking up other noises as “calls” and it wasn’t uncommon to get a few hundred false positives from motors and pumps. But then I started to look through the events that DeepSqueak thought were calls. And oh boy were there calls! Hundreds of them! Perfect clean calls of all types! Short ones and long ones and wavy ones and ones that rose in pitch and others that fell. It was just like the figures in papers that other scientists reported!

But it got even cooler. The first part of the session where the rat was in the box alone, no calls. Not a single one was detected in those solitary three minutes. However, just seconds after his buddy went into the apparatus, a cacophony of calls. Almost all the calls that were detected were from the time they were together. Once I took out his friend and he was alone again he made a few more calls but quickly grew quiet again. So rats do seem to talk after all. It’s just they tend to talk to their friends but don’t mutter to themselves (Note 1).

So now I just need to figure out how to speak rat so they think they’re not alone and will tell me how hard they thought the task was.

---

Notes: Here’s some things I found while setting all this up.

1. Science actually already knew rats don’t really make any calls when they are alone and in an environment they are familiar with (Wohr et al., 2008). So I could have avoided all of this but it was a fun discovery to make on my own.
2. The sensor in Dodotronic’s 250k microphone is the cause of some of the noise from the 20-40kHz, per a Dodotronic rep I emailed with. So of course I could never seem to escape it. DeepSqueak does a good job at finding squeaks in noise but this microphone noise really isn’t ideal for 22kHz long calls from rats. Dodotronic said their 384k mic does not have this noise.
3. Neuralynx Saturn commutators, and I imagine many other motorized commutators, put out a lot of high frequency sounds. Saturn commutators have a particularly loud 38kHz tone. I turned it off if I could to get better audio recordings.
4. Disk drives and other computer parts also seem to put out a lot of high frequency noise. Keeping the mic far from the computer and pointing the mic away from the computer helps cut down on noise.
5. Audacity is really great for recording and was better than any other (free) software I could find. Just make sure you set the sample rate to the appropriate rate. I did use Raven Lite a lot which I think has a better spectrogram than Audacity when I was looking for noise and calls in recorded files. However, I could only ever get Raven Lite to sample at 96k which limited the frequency range to 48kHz which is not good for ultrasonic vocalizations.
6. DeepSqueak is very easy to use and I highly recommend it.