Rats (as well as mice) have a rich language. Or at least something like a language. However, we humans can’t hear them. Human ears can hear to about 20kHz, at least young humans. Can you hear a 20kHz tone? Or hopefully you can hear a 15kHz tone? As humans age and lose hearing, the higher frequency tones are the first to go. However, for rats, 20kHz is a very low frequency. This is because rats “talk” from 22kHz all the way up to 100,000 Hz or more. We call these “ultrasonic” frequencies because they are frequencies humans cannot hear. And for the centuries we humans have been living with, or at least along side rats, we had no idea. It wasn’t until Professor John Anderson from Cornell recorded rats talking in 1954. John figured that bats can make ultrasonic calls as well as guinea pigs , why couldn’t rats? By 1954 we did know rats could make some ultrasonic sounds, however, all that had been recorded at that point were rats making pure tones around 20kHz (see below video at 37 seconds). These would sound a lot like the sine waves linked above; just solid continuous calls with no real substance (tonality) to them. However, with a special amplifier, John Anderson was able to record rats making “a whole series of sounds” from 20 to 80kHz with shifting frequencies; much more like calls or “speaking” than just pure tones (see below video at 40 seconds). Further work has shown rats had a rich repertoire of calls, they even seem to laugh when tickled!

Rats laughing

With modern advances in technology, the study of rodent ultrasonic vocalizations have grown substantially

Pubmed results for ultrasonic vocalization papers

Microphones that can record into ultrasonic frequencies cost only a few hundred dollars. Software has lagged behind, with some commercial products available for hefty price tags. However, there’s been great deal of open source software for recording and analyzing ultrasonic sounds. Audacity can record (and edit) at any frequency as long as your microphone can pick it up. Until recently, analyzing bioacoustic data was very laborious, with poor students scrolling through hours and hours of recordings looking for ultrasonic calls. However, utilizing tools like machine learning and neural networks, open source bioacoustic software has become available. One of these software package is called “DeepSqueak” made by Dr. Kevin Coffery with Russel Marx in the lab of Professor John F Neumaier at the University of Washington. DeepSqueak is a library built with Matlab that can automatically detect rat calls in an audio recording. Rather than spending weeks analyzing rat calls, an hour of audio can be analyzed in a few minutes (depending on your computer). Further still, DeepSqueak will conduct analysis of the calls it detects. For example DeepSqueak will classify calls; basically trying to find calls that are similar and grouping them together so we can better understand what rats are talking about.

DeepSqueak

Rat conversations

Here are some recordings I made of my rats while they carried out various tasks. I use a Dodotronic Ultramic 250k microphone with Audacity to record my rats. To analyze their calls I use Audacity and DeepSqueak. As covered at the top of this page, we humans can’t hear rats. All their calls here have been slowed down by ~10 – 20 times so we can hear them. Keep in mind while listening these calls go on in real time for just a few hundred milliseconds, many are shorter than the time it takes to blink.

Three trills

Two Trills

Two rats greeting each other

Two rats talking

A rat conversation

More conversing

Rats making up after having a little fight

A rat saying bye to his friend