2018-04-23

mindstalk: (Enki)
_If a Lion Could Talk_, Stephen Budiansky, 1998.

...then probably we could understand it, but it wouldn't be a lion anymore.

I saw this book mentioned a lot in grad school, when I was doing my own reading about animal cognition and ape language experiments, but I never read it before. It's much more skeptical than most of what I read then, but fairly convincing in its own right. Independent of that is a lot of animal cognition and communication stuff I didn't know.

This isn't a comprehensive review, just a sweep through of stuff that stood out to me.

**

People report anecdotes of animal intelligence, but rarely animal stupidity under similar situations. People seek to prove animals are some defective version of humans, rather than attending to the impressive things animals can do, in effect denigrating what animals are actually like.

"Smart animals" often means the ones who do what we want; a "stupid horse" learns that the right behavior will avoid being made to work.

Importance of 'peripherals': it's easy to confuse a difference in sensory ability with a difference in intelligence. Rats learn smell associations better than visual ones. Then again, animals learn things that tie into their natural behaviors: dogs can easily learn a tone means food, but not which speaker means food.

If goldfish had hands to play with, would they seem any dumber than monkeys?

At a basic non-verbal general cognition level, humans aren't much different from many animals: we can learn lists, keep 7 things in memory, subitize 4 objects. One researcher claims pretty much all vertebrates have the same general intelligence. This is too strong -- apes do seem better than monkeys who are better than dogs -- but as an order of magnitude thing, maybe kind of right.

Humans special because of our own specialized learning system, language -- which just happens to be capable of unbounded levels of abstraction and reference, creating a real discontinuity of capability, of thought about thoughts, intentions about intentions about intentions.

"Unthinking intelligence" -- it takes hard work to find animal behavior that can't be explained as associative learning. It's possible, but hard.

Lots of 'intelligent', adaptive, behavior -- stalking is brilliant for predators. But puppies stalk bugs at a few weeks of age. Instinct.

Evolution is trial and error; at a faster timescale, so is learning.

Utility of intentional stance, or mock anthropomorphism: assuming a behavior or organ has intention or purpose is often productive and predictive, but that doesn't mean a behavior actually is intentional, any more than a heart is.

Many animal calls classified as "food call" or "alarm call" may at root be attention, "here I am", calls. Or for pets, "summon human". Semantics provided by context of use, not the call itself.

Chomsky: "If you want to find out about an organism you study what it's good at. If you want to study humans you study language. If you want to study pigeons you study homing instinct."

Old approach: mentalism, trying to find human thought, even syllogisms, in animals. Later approach: behaviorism, denying anything other than stimulus-response. Even later approach, inspired by computers and cybernetics: cognitivism, looking for mental representations, whatever those might be.

'animats', very simple models that can produce impressive behavior by interacting with information in the environment. "Follow a gradient, or tumble randomly until you find a gradient" for bacteria. Model frog with 3 interacting simple loops, producing complex prey-grabbing behavior. Model cricket that emulated mate-finding behavior with 100 lines of code, plus an 'ear' responsive to the right frequencies.

Horses and chimps can learn to learn, eventually getting faster at learning new discrimination tests. Chimps can transfer 'sameness' to new match-to-sample tests; pigeons can't.

Timing tests on sequential learning tasks, like ABCD, then being asked about pairs like AC or BD. Monkeys can learn 6 items, pigeons 5 and slower; monkeys also give evidence that they've learned a list they mentally run down, pigeons had learned simpler and less complete rules: A comes first, D comes last; BC baffles it.

Counting vs. rhythmic memory: people can easily repeat "Deck the Halls" 'fa la la la la la la la' without conscious knowledge of how many 'las' there are.

Ground squirrels do make different "hawk alarm" and "mammal predator" alarms. They're not semantically random, but have different acoustic properties: the hawk alarm is a high tone hard to locate, or that is even deceptive. Call benefits both receiver (there's a hawk, run) and sender (lots of running squirrels make the caller not stand out.) The mammal call is part of "look, I see you, your ambush failed, go away."

Animal esperanto: across species, high whines convey fear or appeasement, nonthreat; deep growls convey aggression and threat. Exploitation of big things making low sounds.

Key question: not what an animal is trying to 'say' but what it is trying to accomplish. The whole point is manipulation; often that means 'honest' signals, but not always.

Bird song varied because sender and receiver -- and each male bird is both -- have different goals. Receiver wants to judge how far away someone is, and respond if they're nearby. Sender has no reason to be honest, it'd be ideal to make rivals run around responding to fake threats. Songbirds can only judge distance of a song they know, creating an arms race of multiple songs per bird -- if you use a song your neighbor doesn't know, they have to worry you're intruding. Birds in less competitive areas (more resources per territorial bird) have more stable regional dialects.

Acoustically a bark is between a whine and growl, rising and falling in pitch. Many species 'bark' in a general sense, including bird chirps. Content neutral, "I'm here, now what?" Because they mean nothing, they can mean anything. "Follow me", "stranger approaches", "feed me", "let me in", "let me out". Human vowels and consonants are kind of like whines and growls, making words different kinds of barks.

Humans children use words as names, not as requests, much more obviously than any trained animal. Likewise lots of mutual attention games, with pointing and gaze, that even ape mother-infant pairs don't show.

Bunch of undermining of mirror tests of self-awareness I don't want to summarize.

Animal social intelligence often greater than nonsocial, e.g. easily learning a dominance hierarchy but not other sequential relations, or in-group membership but not arbitrary categories.

Training of child hunting behavior by adult predators is impressive in many cases: fairly good matching to the child's abilities. But it mostly seems to be setting up a learning oppotunity for the child; actual imitation is vanishingly rare. Even the famous snow macaques learning to wash potatoes seems to have been serial re-invention; for example, it didn't spread any faster even as the number of washing monkeys grew.

Infant chimps stick twigs in holes, and follow adult chimps; those combined is enough for them to learn that fishing in some holes yields termites. "stimulus enhancement": animals are drawn to places where conspecifics are finding food, putting them in the same situation in which to learn how to get food.

"The things an animal is good at generally do not require three decades of ambiguous experiments to uncover." -- an indictment of "do apes think like humans" research.

Profile

mindstalk: (Default)
mindstalk

July 2025

S M T W T F S
  12345
6789101112
13141516171819
20212223242526
272829 3031  

Page Summary

Most Popular Tags

Expand Cut Tags

No cut tags

Style Credit

Page generated 2025-08-18 03:43
Powered by Dreamwidth Studios