Christine Huffard sent me a note alerting me to the publication of her latest paper, and she thought I might be interested because I “seem to like cephalopods”. Hah. Well. I’ve noticed that Dr Huffard seems to have some small affection for the tentacled beasties herself.
The paper follows on an old tradition and an old problem. While people have no problem distinguishing human individuals, we have a tough time telling one individual animal from another. This perceptual difficulty complicates problems of studying variations in behavior or physiology, or monitoring numbers and behavior, in natural populations. One solution is tagging or marking the animals in some way, but that always has the risk of changing or harming the disturbed animals — non-invasive procedures are much preferred. This is an especially difficult problem with small animals, like zebrafish or small octopus; I’ve struggled myself with trying to track individual fish in experiments.
I came up with one solution, and Huffard et al. have come up with something similar: humans can be trained to recognize distinctive individual variations, and learn to identify single animals. In this paper, they describe a pattern of white pigmented regions that are consistent within single animals of the species Wunderpus photogenicus…and as you might guess, that is a great excuse to put together a collection of photographs of these aptly named animals.
You might think that animals with a reputation for dynamic camouflage, who can change their skin color and texture at will, might lack any patterns that remain constant. Not so! Cephalopod coloration is not infinitely plastic, so there are always shadows of the underlying biology that can be detected. This is especially true of animals that use distinctive markings, such as warning coloration, like the blue-ringed octopus.
The wunderpus has distinctive markings as well — they look like they belong in the 1960s, either as a tie-dyed t-shirt or better yet, a psychedelic concert poster — and as it turns out, they have characteristic individual white markings that can be easily distinguished. With a little practice, people can get good at recognizing individuals by spotting the spots.
Wunderpus is among the most sought after subjects for underwater photography, so this means that the tourist trade can be tapped into to provide data on the populations — all those tourist snaps can be used to record the existence of individuals over time. This is a great idea for monitoring populations. It also opens up another useful possibility: wunderpus is in high demand for private aquaria (at over $700 each!), so with a complete catalog of natural populations, it might well be possible to recognize illegally acquired animals. Look for wunderpus photos on milk cartons someday!
Meanwhile, you can browse the online database of Wunderpus photos, and if you’re diving in Indonesia, you can contribute to it, as well.
Huffard CL, Caldwell RL, DeLoach N, Gentry DW, Humann P, MacDonald B, Moore B, Ross R, Uno T, Wong S (2008) Individually Unique Body Color Patterns in Octopus (Wunderpus photogenicus) Allow for Photoidentification. PLoS One 3(11):e3732.
Cat of Many Faces says
Huh, that really does make sense. I mean, if I understand it correctly, chromatophores are just little expanding and contracting color sacks right? If so, I guess they wouldn’t move around on the body, and thus we would see consistent patterns.
Very cool :)
Brock says
Cool. This seems pretty similar to the way blue whales are tracked though — by the patters on their tails. Still, I guess it’s good to know more elusive creatures can be tracked without tags.
Christie says
Wow, that’s really awesome. I still think I’ll have trouble telling them apart, though…
Glen Davidson says
One assumes that they can tell each other apart, and, being quite visual animals, one would expect they could do so by vision.
It’s probably easy for them, while they probably would have trouble differentiating humans
Glen D
http://tinyurl.com/2kxyc7
Matt Heath says
Is the name “wunderpus” really a portmanteau of “octopus” and the German “Wunder” (meaning “ZOMFSM!AWESOME!!!1”)? If so it’s highly appropriate.
James F says
Wunderpus…what is the secret of your power?
Wunderpus…won’t you take me far away from the mucky-muck now?
JSug says
Why not adapt computer software used to identify human features so that it could be used to identify animal individuals?
Blake Stacey says
Ah, yes, the standard solution technique employed across the sciences: Let the grad students do it!
Last Hussar says
If you like cephalopods (identify ’em, I can’t even spell ’em) that much why don’t you publish a photo of a different one each week. Course if you don’t know the markings it could be the same one each week and we’d never know.
Oh hang on.
Jeff Satterley says
JSug:
Most identification software is done using Machine Learning techniques, which require large training sets to tune the parameters for identification. So they would need a rather large collection of Wunderpus photos to do a reasonably accurate job, although this database sound like a good place to start training.
Also, animals don’t cooperate with photography as much as people do. When photo ID is done on photos of humans, we generally know where, and in what orientation, the useful facial features will be. I’m sure the photos of the Wunderpus will be at many different angles, rotations and appendage positions, which requires either human sorting and adjusting, or very intelligent software, which requires still more training data.
Sven DiMilo says
Whit Gibbons and crew at the Savannah River Ecology Lab got tired of individually marking the thousands of turtles they were trapping, so they started xeroxing their plastrons for natural ID.
Donnie B. says
Wunderpus: sounds like the name of the next Bond movie and/or girl.
The Chemist says
Hm, I guess I never considered tagging the animals to be an issue in research. Of course now that you mention it, it seems obvious.
Wasn’t there already a Bond movie called Octopussy. I guess you could call the next one Wunder– um on second thought that might sound too dirty.
Greg R. says
Is this PZ Myers as a child?
KC says
Once upon a time (okay, two years ago), when I was involved in mouse behavioural research in large, seminatural enclosures, it got to the point where I would recognize mice from about 5-8 metres away. The more time you spend near your subjects, the more you can start learning some of them. I knew some of them better than I knew some of the volunteers in the lab.
Brad D says
Wunderpus? I think we can do better than that!
My suggestion: Übërpüs
You can never have too many spurious umlauts!
Could be a band name too.
Noadi says
I can always be sucked in my photos of cephalopods, especially blue-ringed octopus.
Mozglubov says
@ #7:
Also, most face recognition software is not for particular individuals (or, if it is, it’s not excessively accurate). A lot of research is going into facial recognition (I have seen some papers that are trying to use online video databases to recognize famous actors), but the variability makes it extremely hard to do. Individual face recognition can be done with a passable degree of success pretty much only if it is done using pre-set orientation and lighting.
Jim Harrison says
The Nuer people of the Sudan know each of their cows individually and have developed a complicated system for keeping track of the color patterns that identify ’em. Maybe we could hire some of these guys as consultants on the octopus identification project.
ThatOtherGuy says
Hee hee. Wunderpus is such a great name :D
phisrow says
@Brad D: I think you mean “Yoü can nëvër have too many spürioüs ümlaüts!”
travc says
This reminds me (oddly) of a project years ago at Caltech to get more information out of MRI data. MRI data is actually at least 6 dimensional, so the typical images are just a projection throwing out lots of info. Of course, there are all sorts of cool machine learning techniques using the full data to classify tissue types ect, but this project took a different (eminently practical) approach.
Doctors (radiologists especially) are really good at pattern recognition problems. So just display more dimensions from the MRI in a visual form and have the doctor’s brains do the hard work.
What they came up with was displaying points as ‘brush strokes’ instead of just pixels. Brush strokes have a color, intensity, orientation, texture, ect… and each of those properties was mapped to a different dimension of the MRI data. It made for some really interesting ‘impressionist’ looking pictures of embryos, hands, and such… but importantly, information like how easily water diffuses was clearly visible and could be used to identify tissues or abnormalities.
Don’t think it took off. There was a working doctor on the project who was quite excited and got really good just glancing at the images and being able to extract all sorts of info, but I suppose the discipline as a whole is extremely conservative by nature. Maybe some day.
Anyway, figuring out what information to pay attention to when training a human (or any other system) to do a recognition problem may seem somewhat easy… but it is really important.
Tim H says
I don’t understand. Should the band be named “Übërpüs” or “The Spürioüs Ümlaüts”?
Dave Godfrey says
Tim H:
Both. Its Übërpüs änd thë Spürïøüs Ümläüts.
There’s supposed to be one over the “n” too, but unicode (nor anything else) supports it. :(
Interrobang says
Is there any correlation between prospagnosia and animal recognition? I can’t tell a lot of humans apart, but I can recognise individual squirrels in the neighbourhood. Squirrels and other animals are easy, but put four brown-eyed brown-haired white guys in identical outfits, and I’ll be calling each of them the others’ names forever. Humans, with their pesky nearly-monochrome skins and hair, and lack of fur, are simply not as distinctive as animals are…
Stripes, dammit! I demand that humans develop stripes!
Brad D says
“Übërpüs” was a new idea, but my original idea that was pretty much guessed at was “Spüriöus Ümläut”. Since my teenage daydreams of being in a band are long gone, and I’m really not a metal head anyway, I thought I might dabble in writing fiction, it’s just a starter idea:
http://www.goodreads.com/story/show/29146.Sp_ri_us_ml_ut
Brad D says
Random thought: Can you use a barcode scanner to identify individual members of a zebra herd?
David Marjanović, OM says
No, meaning “wonder”, as in “miracle”.
The one on über is not spurious…
Sure Unicode supports it. U+0308 is called CONNECTING CHARACTER DIAERESIS. Put this behind an n, et voilà: n̈. (Now let’s just hope it gets displayed correctly. Which would surprise me.)
David Marjanović, OM says
(Why am I not surprised.)
Owlmirror says
Why be sane?
Ü?ë??üs ?ñð ??ë ??ürïøüš Ü??äü??
Owlmirror says
hm.
a?n?d?
Owlmirror says
OK:
trying something different…
a?n?d?
Owlmirror says
OK, that looks better. I used font-family:Arial (because the typeface that has the most Unicode blocks is an Arial one, at least on this machine).
Owlmirror says
Which means that this should look more uniform:
Üƀëȓƥüs ǟñð ţħë Ŝƥürïøüš Üṁłäüťş
Brad D says
Nice :)
gillt says
Illegally acquired animals for the aquarium trade is no doubt problematic, but it is the hobbyists who teach the scientists how to breed these exotics in captivity.
fotomatt says
I know… I know… “we” all understand that saying “humans and animals” is the equivalent of saying “hammers and tools,” but a LOT of other people end up reading these comments, too.
So is it really too much to ask that we take an extra couple keystrokes to clearly differentiate human animals from non-human animals? (human and non-human animals; humans and other animals; etc.)
Otherwise, we may unwittingly reinforce the outrageous idea that humans aren’t animals. Yes, it’s hard to imagine that anyone could actually entertain such preposterous blather… but you might be surprised at some of the crazy things people come up with. ;-)
Peter Ashby says
yeah, how do you stop the octopus from smiling if it wants to?
Peter Ashby says
There’s a Grocer’s Apostrophe in there. Terrible, just Terrible.
Dave Godfrey says
David Marjanović, OM:
Sure Unicode supports it. U+0308 is called CONNECTING CHARACTER DIAERESIS. Put this behind an n, et voilà: n̈. (Now let’s just hope it gets displayed correctly. Which would surprise me.)
I was trying to get it exactly right. Oh and I forgot to remove the dot on the i too. But at least I know the difference between feet and inches. ;)
Brad D.:
Good story. Not that its ever happened to me. If I ever have children they’ll have to rebel against me by listening to the top 40.
Mike says
The Octopus vulgaris said to the Enteroctopus dofleini, “those Wunderpus all look the same to me”. The Enteroctopus blushed, chastised Octopus vulgaris for being so… vulgar, and then promptyly ate him.
I know, I know, this conversation could not have happened in nature!!!
pharynguphat says
THIS is where you excel, even if it is reviewing someone else’s work.
It’s in the REST of your writings that unmitigated wretchedness reigns supreme.
AdamK says
I’m not an animal! I’m a human being!
Owlmirror says
I know it looks that way. I thought it was funny, and left it.
(That’s actually a LATIN SMALL LETTER T WITH CARON)
Owlmirror says
Germane to the original post: I sometimes look closely at photos of animals that appear more than once in some image or set of images to see if I can spot distinguishing characteristics. Sometimes small details can be surprisingly persistent, if you pay attention.
andrea says
“While people have no problem distinguishing human individuals,”
Speak for yourself! I’m part of the estimated 2% of the population with prosopagnosia (faceblindness). Too dang many humans are indistinguishable to me. I’m not “into” tats or piercings personally, but they sure are handy for identifying a few of my students!
andrea