Thursday, September 03, 2009

Chupacabra = not very coyote-like coyote

We have another chupacabra candidate.

Jerry Ayer pretty much admits that not knowing what to call it is good enough for him. So he might as well call it a chupacabra. And why not? It'll help bring some attention to his taxidermy business.

But the issue of what we expect a chupacabra to be is harder to resolve. From an investigative position I have to say that longer front legs doesn't convince me this isn't a coyote. Some will argue that just looking at it should be enough evidence that it's not a coyote. It doesn't look like one and that's how we decide most animal classifications. This is an OK argument as far as we trust that seeing something provides us with enough information about it to include or exclude it in a class. If we take the time to identify what exactly doesn't look like a coyote, we come up with the same argument. Just more detailed. No fur = we trust it's not a coyote. But long legs? This is why just lookin' can be so easily countered. Because impressions need to survive investigation.

The old saying about looking sounding and... tasting like a duck is really based on an interplay of feature analysis and prototypes. Prototype theory suggests that we have an idea of an ideal duck in mind when we call something a duck. And we know when we see a good example of what we were thinking.1 Feature analysis works by proposing a checklist of those features common to a set. Proper features might be something like physical characteristics. Size. Shape. Skin/coat type. When distinguishing species, it's safe to say that something [+scales] is not the same species as something [-scales][+fur]. So for any category, feature analysis will suggest that there is a set of features that are necessary and sufficient for classification. So every member of a group is and must be [+F1] [+F2] [+F3]...

But features are tough to lock down. Features like quadriped or biped can be used when categorizing members of some classes. So among other things, a human is [+biped] and a dog is [+quadriped]. But a husky with two legs is still a dog. Even if it was born with two legs. So we can add a set of transitive features. A little mammal born of a dog is also a dog.2 Even if it doesn't have some of the features we use to identify dogs. But this is just passing the buck. Such transitive features rely on the premise that the class of one dog is already known. In some cases this is valid. Like the difference between a Ford and a Chrysler.3 But we haven't moved too far with the analysis. We find ourselves stuck with the ultimate feature being tautological. A dog is [+dog].

This question of the chupacabra works by subtraction. The argument is almost explicit that since it's "unlike anything native to Texas" it must be that mysterious chupacabra that we have never before seen. That's not a very strong argument. But we use some of the assumed features of the chupacabra in our analysis. I wrote about this exact topic a couple of years ago.

So now we see that Texas has produced two of what look like the same weird little sucker. And only 130 miles apart. I'll go out on a limb and say that if we find a few more of these creatures that look like bald coyotes with long front legs, a new category will be created. Some name will be suggested. But no one will accept that this is the little monster. The chupacabra will still be an elusive little wingless bat that must remain [+mysterious].

1. That's too simple to be very helpful. But we'll move on for the sake of space.
2. One big question here: are we sure it's a dog because when born of a dog it must be a dog, or because we trust that it simply will be a dog?
3. Let's avoid the arguments about quality usually heard in the vicinity of a set of truck nuts and bumper sticker of Calvin taking a leak.


  1. DNA analysis? -- for ruling out?

  2. yep. that's another one of those subtractive features. i mentioned that in the other post.


Thanks for reaching out.

You can also contact me at wishydig[at]gmail[d0t]com.