Scientists document remarkable sperm whale phonetic alphabet
May 23, 2024 5:57 AM Subscribe
Scientists document remarkable sperm whale phonetic alphabet. A new analysis of years of vocalizations by sperm whales in the eastern Caribbean has found that their system of communication is more sophisticated than previously known.
So what I'm hearing is that we understand the sounds but not the language and that we'd be responding in gibberish, right?
posted by RonButNotStupid at 6:49 AM on May 23
posted by RonButNotStupid at 6:49 AM on May 23
Breaking news: scientists say the other people on this planet are also people.
posted by The River Ivel at 7:04 AM on May 23 [14 favorites]
posted by The River Ivel at 7:04 AM on May 23 [14 favorites]
Here's the original paper; I gave it a pretty close reading last week.
They're essentially trying to figure out the 'bitrate' of sperm whale vocal communication: more discrete units means more bitrate. They find that there is more variation in the vocalizations than previously known, though one of those variations (rubato) seems a bit tenuous to me.
There's also a discretization of tempos into five buckets; two of those buckets seem very well motivated by the data (clear spikes in the histogram), but the other three seem like arbitrary division of a big blob. (Figure 2a) My guess is that if the long tempo case is sematic feature, it may be a discrete feature with a continuous component, rather than three discrete tempos.
We also have continuous components in our speech; think of the degree of 'up sweep' we use for indicating a question, which can range from very subtle to quite exaggerated. In this case, the upsweep itself indicates a question, and the amount of upsweep can indicate amount of uncertainty, perhaps.
For sperm whales, we don't really know what this means or how it functions. Perhaps it's directional information, and the length of the phrase helps indicate how far away something is? No idea.
The really tricky part, though, is that gathering observational data to try to ground communications with the physical world is very difficult. They're whales, after all. It'll be a slow process, in all likelihood.
posted by kaibutsu at 7:06 AM on May 23 [31 favorites]
They're essentially trying to figure out the 'bitrate' of sperm whale vocal communication: more discrete units means more bitrate. They find that there is more variation in the vocalizations than previously known, though one of those variations (rubato) seems a bit tenuous to me.
There's also a discretization of tempos into five buckets; two of those buckets seem very well motivated by the data (clear spikes in the histogram), but the other three seem like arbitrary division of a big blob. (Figure 2a) My guess is that if the long tempo case is sematic feature, it may be a discrete feature with a continuous component, rather than three discrete tempos.
We also have continuous components in our speech; think of the degree of 'up sweep' we use for indicating a question, which can range from very subtle to quite exaggerated. In this case, the upsweep itself indicates a question, and the amount of upsweep can indicate amount of uncertainty, perhaps.
For sperm whales, we don't really know what this means or how it functions. Perhaps it's directional information, and the length of the phrase helps indicate how far away something is? No idea.
The really tricky part, though, is that gathering observational data to try to ground communications with the physical world is very difficult. They're whales, after all. It'll be a slow process, in all likelihood.
posted by kaibutsu at 7:06 AM on May 23 [31 favorites]
(pedant mode) Alphabets are writing systems, i.e. systems for representing sounds; what they discovered is a phonetic system, as in the sounds themselves are used in a systematic way. Which to be honest I thought we already knew about whalesong, and (I thought) about birdsong.
posted by trig at 7:10 AM on May 23 [10 favorites]
posted by trig at 7:10 AM on May 23 [10 favorites]
Thank you for the link and details, kaibutsu. You know this but reinforcing it for the general audience.. A key part of human phonology is what we call discreteness. Phonemes in human speech are distinct things. There's the sound m and the sound n and while you can imagine sounds halfway between the two in English it's always produced and heard as one or the other. And while it's true English speakers do use continuous tones when asking questions, that's exactly why tone isn't a phoneme in English. In tonal languages like Mandarin there are exactly 5 tones; again the voice can produce something inbetween but it's not part of the language. (One fun fact: signed human languages also have discreteness, just like spoken languages.)
So what this article is showing is evidence for something like discrete phonemes in whale sounds. Specifically rhythm, tempo, and ornamentation are three features that they claim has evidence of discreteness. Ie: there are 5 tempos in figure 2a and an utterance either takes 0.3s, or 0.5s, or 0.8s, or 1.1s, or 1.3s. (At least that's the claim, as kaibutsu says the histogram looks a little less clear than that but I'll trust the authors and peer reviewers here.) There's also a fourth feature, rubato, that may be discrete or continuous but seems part of the communication.
Discreteness is believed to be an important part of how we process language. It lets the brain turn complex waveforms into strings of small symbols which can then be processed at a higher level. If whales do have language it's interesting and possibly inevitable that their language would share certain structural similarities to human language, specifically in the way its set of phonemes are structured.
In figure 3 the paper goes on to lay out a table of all observed phonemes in their whale vocal samples.
posted by Nelson at 7:28 AM on May 23 [15 favorites]
So what this article is showing is evidence for something like discrete phonemes in whale sounds. Specifically rhythm, tempo, and ornamentation are three features that they claim has evidence of discreteness. Ie: there are 5 tempos in figure 2a and an utterance either takes 0.3s, or 0.5s, or 0.8s, or 1.1s, or 1.3s. (At least that's the claim, as kaibutsu says the histogram looks a little less clear than that but I'll trust the authors and peer reviewers here.) There's also a fourth feature, rubato, that may be discrete or continuous but seems part of the communication.
Discreteness is believed to be an important part of how we process language. It lets the brain turn complex waveforms into strings of small symbols which can then be processed at a higher level. If whales do have language it's interesting and possibly inevitable that their language would share certain structural similarities to human language, specifically in the way its set of phonemes are structured.
In figure 3 the paper goes on to lay out a table of all observed phonemes in their whale vocal samples.
posted by Nelson at 7:28 AM on May 23 [15 favorites]
yooooooouuuuuuuuure nooooooooo straaaaaaaaaaaaaangeeeer toooooooo looooooooooooove...
posted by GCU Sweet and Full of Grace at 7:38 AM on May 23 [15 favorites]
posted by GCU Sweet and Full of Grace at 7:38 AM on May 23 [15 favorites]
Here's the original paper
codas, hmm
"For example, the sentence 'Let's meet next to the statue of Claude Shannon...'"
*groans*
posted by HearHere at 8:30 AM on May 23
codas, hmm
"For example, the sentence 'Let's meet next to the statue of Claude Shannon...'"
*groans*
posted by HearHere at 8:30 AM on May 23
According to wikipedia, we know sperm whales express clan identity. I suppose they do not sing similar songs simultaneously like humpback whales do though.
As an aside, Bulk Food by Peter Watts is a cute sci-fi short story about communications with killer whales.
posted by jeffburdges at 8:39 AM on May 23 [2 favorites]
As an aside, Bulk Food by Peter Watts is a cute sci-fi short story about communications with killer whales.
posted by jeffburdges at 8:39 AM on May 23 [2 favorites]
I can't wait to hear more about this on season 2 of Extraordinary Attorney Woo!
posted by bigendian at 9:02 AM on May 23 [4 favorites]
posted by bigendian at 9:02 AM on May 23 [4 favorites]
Awesome. I just went down a little youtube rabbit hole about how wrong-headed out attempts were to try and converse with dolphins. Carl Sagan was the voice of reason then: don't try to teach them English, dummy. They can kinda communicate with their blow holes but they're ALREADY communicating with sonar clicks. We can make machines to bridge the gap and they cannot. It behooves us to meet them on their terms. This is so cool!
*links selected for comedic value
posted by es_de_bah at 9:15 AM on May 23 [1 favorite]
*links selected for comedic value
posted by es_de_bah at 9:15 AM on May 23 [1 favorite]
wrong-headed out attempts were to try and converse with dolphins. Carl Sagan was the voice of reason then: don't try to teach them English, dummy.
Fa... LOVE... Pa.
posted by hanov3r at 10:05 AM on May 23
Fa... LOVE... Pa.
posted by hanov3r at 10:05 AM on May 23
cute sci-fi short story about communications with killer whales
I think we have different definitions of cute.
posted by lizjohn at 10:07 AM on May 23 [1 favorite]
I think we have different definitions of cute.
posted by lizjohn at 10:07 AM on May 23 [1 favorite]
(pedant mode) Alphabets are writing systems, i.e. systems for representing sounds; what they discovered is a phonetic system, as in the sounds themselves are used in a systematic way. Which to be honest I thought we already knew about whalesong, and (I thought) about birdsong.
If you click through to the actual paper, as Nelson notes above they actually DO have a writing system for representing these sounds. It is unwieldy as all fuck, but it is trying to describe the sounds in an orderly way.
What is different here from previous whale song study is that, rather than seeing that they have these lengthy songs that seem to have structure that we generally only can detect with our human minds if we speed them up a lot... this study instead is actually trying to break the songs apart into individual sounds. And they've found a remarkable set of either phonemes or words or somethings which are standardized across the population and seems to be quite communicative in its purpose.
This is a huge breakthrough and is really quite fascinating. Thanks for posting!
posted by hippybear at 11:08 AM on May 23
If you click through to the actual paper, as Nelson notes above they actually DO have a writing system for representing these sounds. It is unwieldy as all fuck, but it is trying to describe the sounds in an orderly way.
What is different here from previous whale song study is that, rather than seeing that they have these lengthy songs that seem to have structure that we generally only can detect with our human minds if we speed them up a lot... this study instead is actually trying to break the songs apart into individual sounds. And they've found a remarkable set of either phonemes or words or somethings which are standardized across the population and seems to be quite communicative in its purpose.
This is a huge breakthrough and is really quite fascinating. Thanks for posting!
posted by hippybear at 11:08 AM on May 23
If we DO ever figure out how to speak with them - I sincerely hope we begin with an apology for all of the stupid, harmful things we've done to them and their environment for so long...
posted by caution live frogs at 11:08 AM on May 23
posted by caution live frogs at 11:08 AM on May 23
they actually DO have a writing system for representing these sounds. It is unwieldy as all fuck,
the Unicode Consortium is taking notes for Version 16
posted by AzraelBrown at 11:57 AM on May 23 [2 favorites]
the Unicode Consortium is taking notes for Version 16
posted by AzraelBrown at 11:57 AM on May 23 [2 favorites]
Excellent link, animals always exceed our preconceptions about them. I was looking to see an audio example of a chorus but couldn't seem to find any in the paper or on the DSWP website. I just think it would be neat to listen to a chorus while watching the corresponding exchange plot.
Is there an open database of whale codas I could find somewhere online?
posted by crossswords at 12:34 PM on May 23
Is there an open database of whale codas I could find somewhere online?
posted by crossswords at 12:34 PM on May 23
There's also a discretization of tempos into five buckets; two of those buckets seem very well motivated by the data (clear spikes in the histogram), but the other three seem like arbitrary division of a big blob. (Figure 2a) My guess is that if the long tempo case is sematic feature, it may be a discrete feature with a continuous component, rather than three discrete tempos.
I was interested in this too, and they justify their tempo clusters in section 3 of the supplemental info. Looks like they are satisfied with tempo clusters that also produce monomodal distributions of rubato, although their example is combining the much more clearly defined cluster 1 and 2. Groups 4 and 5 look all smushed out, I'd be interested if the histogram of those combined cluster is also multimodal.
posted by crossswords at 12:49 PM on May 23 [1 favorite]
I was interested in this too, and they justify their tempo clusters in section 3 of the supplemental info. Looks like they are satisfied with tempo clusters that also produce monomodal distributions of rubato, although their example is combining the much more clearly defined cluster 1 and 2. Groups 4 and 5 look all smushed out, I'd be interested if the histogram of those combined cluster is also multimodal.
posted by crossswords at 12:49 PM on May 23 [1 favorite]
This seems like a job for Amy Adams’s Linguist character in Arrival. I think she should stop her lecture on the history of Portuguese immediately and got on the case.
posted by umbú at 1:24 PM on May 23 [2 favorites]
posted by umbú at 1:24 PM on May 23 [2 favorites]
« Older Out Of The Shadows, Into The Spotlight [or so they... | Well, you know you can't take it with you Newer »
posted by BobTheScientist at 6:39 AM on May 23 [15 favorites]