Atoms of shape

Discrete combinatorial systems underlie a lot of the variety in our universe and on our planet: atoms combine to make molecules, nucleotides to make genes, amino acids to make proteins, phonemes to make syllables, words to make phrases.

And a recent article suggests that not only language but at least one domain of thought may gain its expressive power from a discrete combinatorial system. The title, “A language of thought for the mental representation of geometric shapes,” sums it up. The authors develop a set of primitive geometric operations that can generate simple shapes, like straight lines, zigzags, circles, spirals, which are widely found in human cultural history going back to a 540,000-year-old etched shell from Java. The system can also be extended to generate more complex forms. The authors present evidence (judgments of complexity, ease of memory for different geometric shapes) that these operations are close to what humans use to conceptualize shapes.  

Later spectacular cave paintings will showcase a uniquely human talent for analog pictorial representation, but basic geometry may be another area of human uniqueness: other animals so far don’t seem to treat the same geometric shapes as special.

Hits, slides, and rings

423 – 400 thousand years ago

Part of the challenge of language is coming up with some way to distinguish thousands or tens of thousands of words from one another. It would be hard to come up with that many unique sounds. What human languages do instead is to come up with phonemes and rules for stringing phonemes together into syllables, and then create words by arbitrarily pairing up one syllable, or a few, with a meaning. Phonemes are the individual sounds of a language, roughly comparable to individual letters. There are about forty phonemes in most dialects of English. (English spelling does a pretty sloppy job of matching up phonemes and letters. Finnish comes close to one phoneme per letter.)

Often in evolution organisms don’t solve new problems from scratch, but instead harness preexisting adaptations. I argued earlier that the abstract “space” of possession (“The Crampden estate went to Reginald.”) may have developed by harnessing preexisting concepts of physical space. And our abilities to recognize speech sounds may harness our preexisting capacities for recognizing the sounds of solid objects interacting. At least that’s the argument of a recent book by Mark Changizi, Harnessed: How Language and Music Mimicked Nature and Transformed Ape to Man.

Changizi notes that even though we’re mostly not aware of it, we’re very good at using our hearing to keep track of what’s going on in our physical surroundings. For example, people easily recognize the difference between someone going upstairs and someone going downstairs, and we’re pretty good at recognizing individuals by their treads. The sounds that solid objects make can be broadly categorized as hits, slides, and rings. Hits: one object collides with another and sends out a sharp burst of sound. Slides: an object scrapes against another and sends out a more extended sound. Rings: an object reverberates after a collision. Changizi argues that these correspond to the major categories of phonemes.

• Hits = plosives, like p b t g k
• Slides = fricatives, like s sh th f v z
• Rings = sonorants, including sonorant consonants, like l r y w m n, and vowels

These are not the only sounds we can make with our mouths. We can do barks and pops and farts and so on. But our auditory systems are especially cued into solid object physics, so when we try to come up with easy-to-distinguish phonemes, that’s what we focus on. And a lot of rules about how phonemes hook up also follow from this principle – for example hits followed by rings are more common than the reverse. (Linguists find that CV syllables – one consonant followed by one vowel, like bee, two, or go – are found in pretty much every language, while other combinations – VC, ape, or CCV, spa, or CVCCC, sixths – have a more restricted distribution.)

So even if imitating nature is not the whole story of phonemes, it may at least be where they got started.

Later on when we talk about writing systems, we’ll see there’s a similar argument about how these are tuned to tickle our primate visual systems.

Learn This One Weird Trick (Part One)

… that humans use, and now you can too!

There are people who think that human beings are nothing special. Sure (the argument goes) people have uniquely large brains. But all sorts of creatures have unique features. Elephants are the only animals with trunks. Tamarins and marmosets are the only primates that give birth to twins. Platypuses are the only venomous mammals. Spotted hyenas are the only mammals whose females sport pseudo-penises (through which they give birth!). And so on. If we could ask members of these species they’d claim that they’re the special ones.

But of course we can’t ask them, and in any case, this isn’t a very convincing argument. Human beings have an absolutely outsize impact on the Earth, and the advent of human beings looks like one of the major evolutionary transitions, comparable in importance to the origin of the eukaryotic cell or multicellular life. But even if we buy this, it still leaves open the question of whether there’s a key adaptation – a One Weird Trick – that accounts for the exceptional course of human evolution. Here are some candidates that being are being batted around these days:

1) The cognitive niche. The basic idea is at least as old as Aristotle, that human brings are defined by their capacity for Reason. A modern version of this is advocated by evolutionary psychologist John Tooby and cognitive scientist Steven Pinker. Pinker in particular has elaborated the argument that humans are uniquely adapted to acquire and share knowledge, by virtue of a suite of cognitive, social, and linguistic adaptations. We’ve already touched on several aspects of this: Human beings seem to have taken the capacity for thinking about physical space and retooled it for thinking about the abstract cognitive space of possession – a social relationship. (Other abstract cognitive spaces include kinship, time, and change-of-state.) And humans seem to harness the machinery for processing the sounds of interacting solid objects in creating major categories of phonemes. For a more complete exposition, here’s an academic article by Pinker, and a talk on youtube.

2) Culture. Rob Boyd and Pete Richerson, who’ve done a lot of mathematical modeling of cultural evolution, are skeptical about the “cognitive niche” argument. Too much culture, they argue, is things that have been learned by trial-and-error, and are passed on from one generation to the next without people understanding why they work. Boyd and Richerson appeal, as anthropologists have for generations, to the importance of culture. We mentioned earlier their argument that the frequency of climate change in the Ice Age was nicely calibrated to favor social learning rather than individual learning or instinct. Joseph Henrich provides a recent defense of the importance of culture. Contra Pinker, he thinks humans often don’t have a good cause-and-effect understanding of the things they do, but depend heavily on imitation and the accumulated wisdom of the elders. And see this post, for the importance of High Fidelity cultural transmission in the evolution of animal and human intelligence.

Coming up: Part Two. Recursion and Shared Intentionality

Hits, slides, and rings

425 – 403 thousand years ago

Part of the challenge of language is coming up with some way to distinguish thousands or tens of thousands of words from one another. It would be hard to come up with that many unique sounds. What human languages do instead is to come up with phonemes and rules for stringing phonemes together into syllables, and then create words by arbitrarily pairing up one syllable, or a few, with a meaning. Phonemes are the individual sounds of a language, roughly comparable to individual letters. There are about forty phonemes in most dialects of English. (English spelling does a pretty sloppy job of matching up phonemes and letters. Finnish comes close to one phoneme per letter.)

Often in evolution organisms don’t solve new problems from scratch, but instead harness preexisting adaptations. I argued earlier that the abstract “space” of possession (“The Crampden estate went to Reginald.”) may have developed by harnessing preexisting concepts of physical space. And our abilities to recognize speech sounds may harness our preexisting capacities for recognizing the sounds of solid objects interacting. At least that’s the argument of a recent book by Mark Changizi, Harnessed: How Language and Music Mimicked Nature and Transformed Ape to Man.

Changizi notes that even though we’re mostly not aware of it, we’re very good at using our hearing to keep track of what’s going on in our physical surroundings. For example, people easily recognize the difference between someone going upstairs and someone going downstairs, and we’re pretty good at recognizing individuals by their treads. The sounds that solid objects make can be broadly categorized as hits, slides, and rings. Hits: one object collides with another and sends out a sharp burst of sound. Slides: an object scrapes against another and sends out a more extended sound. Rings: an object reverberates after a collision. Changizi argues that these correspond to the major categories of phonemes.

• Hits = plosives, like p b t g k
• Slides = fricatives, like s sh th f v z
• Rings = sonorants, including sonorant consonants, like l r y w m n, and vowels

These are not the only sounds we can make with our mouths. We can do barks and pops and farts and so on. But our auditory systems are especially cued into solid object physics, so when we try to come up with easy-to-distinguish phonemes, that’s what we focus on. And a lot of rules about how phonemes hook up also follow from this principle – for example hits followed by rings are more common than the reverse. (Linguists find that CV syllables – one consonant followed by one vowel, like bee, two, or go – are found in pretty much every language, while other combinations – VC, ape, or CCV, spa, or CVCCC, sixths – have a more restricted distribution.)

So even if imitating nature is not the whole story of phonemes, it may at least be where they got started.

Later on when we talk about writing systems, we’ll see there’s a similar argument about how these are tuned to tickle our primate visual systems.

My handaxe

1,043 – 986 thousand years ago

By today’s date, Acheulean tools are well developed in Africa, and found in India too. Sophisticated tools like the Acheulean hand axe probably tell us something not just about cognition in relation to tool making, but also about social cognition. You wouldn’t make a hand axe, use it, and abandon it. Nor would you go to all the trouble if the biggest, baddest guy in the group was immediately going to grab it from you. So there is probably some notion of artifacts-as-personal-possessions by the time Acheulean appears.

Possession is a social relationship, a relationship between two or more individuals with respect to the thing possessed. Robinson Crusoe didn’t “own” anything on his island before Friday came along.

Linguists have noted something interesting about the language of possession that maybe tells us something about the psychology of possession: Expressions for possession are often similar to expressions for spatial locations. Compare spatial expressions:

João went to Recife.
Chico stayed in Rio.
The gang kept Zezinho in Salvador.

and corresponding constructions for possessions:

The Crampden estate went to Reginald.
The Hampden estate stayed with Lionel.
Thag kept axe.

Of course the Crampden estate didn’t go anywhere in physical space, but it still traveled in the abstract social space of possession. In some cases just switching from inanimate to animate subject will switch the meaning from locative to possessive. The Russian preposition y means at/near when applied to a place (People are at Nevsky street) but possession when applied to a person (Hat is “at” Ivan = Ivan has hat.)

What may be going on here: people (and many other creatures) have some mental machinery for thinking about physical space. That machinery gets retooled/borrowed/exapted for thinking about more abstract relationships. So the cognitive psychology of space gets retooled for thinking about close and distant social relationships, or time ahead and behind. In other words, we may be seeing a common evolutionary phenomenon of organs evolved for one purpose being put to another purpose – reptile jaw bones evolve into mammalian inner ear bones, dinosaur forelimbs evolve into bird wings. You can find Steve Pinker making this argument in his book The Stuff of Thought. And Barbara Tversky’s recent Mind in Motion: How Action Shapes Thought seems to make the argument at greater length. For a while most of the evidence of repurposing spatial cognition for more abstract relationships came from linguistics, but there’s now some corroboration from neurology.

And I’ve made the argument for the particular case of kinship: regularities in kin terminology across cultures tell us something about pan-human ideas of “kinship space.” (My kin and mybody parts are arguably the most basic, intrinsic primitive sorts of possessions, since long before my handaxe.) This implies that the evolutionary psychology of kinship has not just an adaptive component (adaptations for calculating coefficients of relatedness and inbreeding), but also a phylogenetic component  (homologies with the cognitive psychology of space).

We’ll see other possible examples, involving e.g. the evolution of speech sounds, as we move along.

Hits, slides, and rings

438 – 415 thousand years ago

Part of the challenge of language is coming up with some way to distinguish thousands or tens of thousands of words from one another. It would be hard to come up with that many unique sounds. What human languages do instead is to come up with phonemes and rules for stringing phonemes together into syllables, and then create words by arbitrarily pairing up one syllable, or a few, with a meaning. Phonemes are the individual sounds of a language, roughly comparable to individual letters. There are about forty phonemes in most dialects of English. (English spelling does a pretty sloppy job of matching up phonemes and letters. Finnish comes close to one phoneme per letter.)

Often in evolution organisms don’t solve new problems from scratch, but instead harness preexisting adaptations. I argued earlier that the abstract “space” of possession (“The Crampden estate went to Reginald.”) may have developed by harnessing preexisting concepts of physical space. And our abilities to recognize speech sounds may harness our preexisting capacities for recognizing the sounds of solid objects interacting. At least that’s the argument of a recent book by Mark Changizi, Harnessed: How Language and Music Mimicked Nature and Transformed Ape to Man.

Changizi notes that even though we’re mostly not aware of it, we’re very good at using our hearing to keep track of what’s going on in our physical surroundings. For example, people easily recognize the difference between someone going upstairs and someone going downstairs, and we’re pretty good at recognizing individuals by their treads. The sounds that solid objects make can be broadly categorized as hits, slides, and rings. Hits: one object collides with another and sends out a sharp burst of sound. Slides: an object scrapes against another and sends out a more extended sound. Rings: an object reverberates after a collision. Changizi argues that these correspond to the major categories of phonemes.

• Hits = plosives, like p b t g k
• Slides = fricatives, like s sh th f v z
• Rings = sonorants, including sonorant consonants, like l r y w m n, and vowels

These are not the only sounds we can make with our mouths. We can do barks and pops and farts and so on. But our auditory systems are especially cued into solid object physics, so when we try to come up with easy-to-distinguish phonemes, that’s what we focus on. And a lot of rules about how phonemes hook up also follow from this principle – for example hits followed by rings are more common than the reverse.

So even if imitating nature is not the whole story of phonemes, it may at least be where they got started.

Later on when we talk about writing systems, we’ll see there’s a similar argument about how these are tuned to tickle our primate visual systems.

My handaxe

101- 959 thousand years ago

By today’s date, Acheulean tools are well developed in Africa, and found in India too. Sophisticated tools like the Acheulean hand axe probably tell us something not just about cognition in relation to tool making, but also about social cognition. You wouldn’t make a hand axe, use it, and abandon it. Nor would you go to all the trouble if the biggest, baddest guy in the group was immediately going to grab it from you. So there is probably some notion of artifacts-as-personal-possessions by the time Acheulean appears.

Possession is a social relationship, a relationship between two or more individuals with respect to the thing possessed. Robinson Crusoe didn’t “own” anything on his island before Friday came along.

Linguists have noted something interesting about the language of possession that maybe tells us something about the psychology of possession: Expressions for possession are often similar to expressions for spatial locations. Compare spatial expressions:

João went to Recife.
Chico stayed in Rio.
The gang kept Zezinho in Salvador.

and corresponding constructions for possessions:

The Crampden estate went to Reginald.
The Hampden estate stayed with Lionel.
Thag kept axe.

Of course the Crampden estate didn’t go anywhere in physical space, but it still traveled in the abstract social space of possession. In some cases just switching from inanimate to animate subject will switch the meaning from locative to possessive. The Russian preposition y means at/near when applied to a place (People are at Nevsky street) but possession when applied to a person (Hat is “at” Ivan = Ivan has hat.)

What may be going on here: people (and many other creatures) have some mental machinery for thinking about physical space. That machinery gets retooled/borrowed/exapted for thinking about more abstract relationships. So the cognitive psychology of space gets retooled for thinking about close and distant social relationships, or time ahead and behind. In other words, we may be seeing a common evolutionary phenomenon of organs evolved for one purpose being put to another purpose – reptile jaw bones evolve into mammalian inner ear bones, dinosaur forelimbs evolve into bird wings. You can find Steve Pinker making this argument in his book The Stuff of Thought. And Barbara Tversky’s recent Mind in Motion: How Action Shapes Thought seems to make the argument at greater length; I’m looking forward to reading it. For a while most of the evidence of repurposing spatial cognition for more abstract relationships came from linguistics, but there’s now some corroboration from neurology.

And I’ve made the argument for the particular case of kinship: regularities in kin terminology across cultures tell us something about pan-human ideas of “kinship space.” (My kin and mybody parts are arguably the most basic, intrinsic primitive sorts of possessions, since long before my handaxe.) This implies that the evolutionary psychology of kinship has not just an adaptive component (adaptations for calculating coefficients of relatedness and inbreeding), but also a phylogenetic component  (homologies with the cognitive psychology of space).

We’ll see other possible examples, involving e.g. the evolution of speech sounds, as we move along.

Accounting for brains

Why do humans have such big brains? Partly because we’re primates, and primates in general have big brains – not just big brains but an exceptional density of neurons, especially in big primates. But a recent article by González-Forero and Gardner offers some more specific ideas.

Before getting down to the article, a general reflection on statistics and methodology:

In many areas of science, you’ve got a lot of data and you want to sort out cause and effect. This happens in evolutionary biology, for example, when you want to determine what selective pressures have caused brains to evolve to different sizes. And it happens in medicine, when you want to find out what lifestyle choices generate what health problems. It also happens in social science and public policy, when you want to find out what programs generate what social outcomes. A common method in these cases is to use multivariate regression, looking for the strongest correlates of your dependent variable. This has its limits however. You often find that a lot of your variables are correlated with one another, and it’s hard to figure out what is cause and what’s effect.

So there’s been a lot of interest lately in a different approach, where you start out at the beginning with an explicit model of cause-and-effect pathways and use your data to estimate the strength of causal connections. An excellent popular introduction to this rapidly developing field comes from Judea Pearl, in The Book of Why: The New Science of Cause and Effect. Pearl makes the case here that statistics needs to move beyond pattern recognition, to testing causal models and counterfactual reasoning. Pearl sees counterfactual reasoning in particular as a human specialty, One Weird Trick that distinguishes humans from other creatures, and he is skeptical about current work in Artificial Intelligence, impressive as it is, that is mainly about pattern recognition.

Richard McElreath, commenting on González-Forero and Gardner, puts it this way:

“Automobile engineering can provide an analogy for studying this type of system. It would be difficult to understand racing-car design through regression analysis of how engine size varies depending on changes in other features, such as the mass and shape of the car. Instead, a model is needed that uses physical laws to predict optimal combinations of the variables under different criteria. Understanding brain evolution poses a similar challenge in that an organism’s features co-evolve under biological constraints.”

So turning to the article itself, what the authors do is to test an explicit model in which a developing organism has to allocate energy to growing a brain, growing a body, and reproducing. They ask what sorts of evolutionary challenges would lead to the particular combination of brain size, body size, and reproductive life history that we see in Homo sapiens. The challenges might be ecological (e.g. securing more food). They might be social (outwitting competitors). They might be solitary or cooperative (working with others to secure more food, or banding with others to defeat rival bands). Their conclusion: the best fit to their model comes when they assume that the evolution of big brains is 60% a result of individual ecological adaptation, 30% a result of cooperative ecological adaptation, and 10% a result of group-versus-group social adaptation. More specifically, what mostly drives the evolution of brain size in their model is that marginal returns to investing in ecological skills don’t decline as quickly for humans as for our close relatives. Spending extra years learning stuff continues to have a payoff for us, maybe because culture and language mean that there are a lot more useful tricks floating around to learn.

These results have to be considered pretty tentative at this point. Note however that they count strongly against the view that human brain evolution is mostly about being Machiavellian and outsmarting the other guys, although they do allow a modest role for inter-group competition. And they count against the view, advocated by Geoffrey Miller, that the human mind evolved as a sexual display, like the peacock’s tail. So it may be true that “sexual love … lays claim to half the powers and thoughts of the youngest portion of mankind” (Schopenhauer). But (at least according to González-Forero and Gardner), whatever the claims of love on our hearts, we owe our big brains to our work.

Learn This One Weird Trick (Part Two)

… that humans use, and now you can too! (Continued from the previous post.)

 3) Recursion. What if you have one mirror facing a second mirror, so the first mirror shows what’s in the second mirror, which shows what’s in the first mirror …? What if you take a chameleon, which tries to take on the color of its surroundings, and put it on a mirror? What if you point a video camera at the very screen that’s showing what the video camera is pointing at? What if (getting mathematical) you use a function in defining that same function? What if you use the cleaning attachment from your vacuum cleaner to suck dust off the vacuum cleaner itself? (Okay, the last one is a bit lame.) The basic idea in each of these cases is called recursion, which is a major concept in mathematics and computer science. Douglas Hofstadter’s Gödel, Escher, Bach is all about recursion. Some people think recursion – nesting ideas about ideas inside one another in a potentially infinite hierarchy, or (for syntax) phrases inside phrases — is central to human uniqueness. Noam Chomsky has lately been pushing a hard-core version of this argument. Here he is with Robert Berwick defending his view.

Related to the idea of recursion is the idea of “meta-representation”: not just having ideas about the world but having ideas about ideas, being able to put a box around a proposition, and then attaching a tag to it that says the equivalent of “This is true” or “This is false” or “This will be true later” or “Suppose this were true,” and then manipulating it accordingly. A nice little essay in “imagination,” elaborating this idea, is here from Simon Baron-Cohen, best known as an authority on autism.

4) Shared intentionality. Suppose you and I are friends with a couple, Fred and Wendy Smith. I tell you “I saw Wendy Smith kissing a man in the park yesterday.” Logically speaking, there’s nothing to say the man wasn’t Fred. But you’ll probably assume that I meant she was kissing someone other than Fred. Why? Well if the man had been Fred I could just as easily have said “I saw Wendy Smith kissing Fred in the park yesterday.” Since I didn’t say that, you assume I mean to convey the man wasn’t Fred. Note this only works if both of us try to pack as much relevant information into our sentences as possible and know the other person is doing the same. (If you think this sounds like recursion, you’re right.) Back in the 1950s, Paul Grice, a philosopher, worked out a lot of how we pack non-literal meanings into sentences. But the same principles are at work even when people are communicating non-linguistically. This leads to another theory of human uniqueness: human beings are uniquely good at developing shared intentions with one another: each party knows the other party is trying to communicate something, so they converge on the correct answer. People may have been doing this even before language evolved. Following up on this can quickly get you into game theory, where a central concept is “common knowledge”: not just “I know X” and “You know X,” but “I know X,” and “I know X is common knowledge to us,” and similarly for you. The cc option on your email generates common knowledge: if you see someone’s address there then they see yours, and you know they know you know, etc. Here’s a philosophical treatment.

But you can skip the philosophy if you want and move on to a telling little piece of anatomy that’s relevant here. In most mammals, including chimpanzees, the sclera (white of the eyes) is not visible. It’s hard to tell where a chimpanzee is looking, easy for a human. Human eyes make it easy to cooperate in sharing attention, a first step in developing shared intentions. If you know your card games, chimpanzees are playing poker, humans are playing bridge.

Our discussion of human uniqueness on Logarithmic History has been frustratingly short on specific dates. But human sclera are probably a fairly simple trait genetically, and we may soon enough discover the genes involved and even tell how long ago they mutated.

Learn This One Weird Trick … (Part One)

… that humans use, and now you can too!

There are people who think that human beings are nothing special. Sure (the argument goes) people have uniquely large brains. But all sorts of creatures have unique features. Elephants are the only animals with trunks. Tamarins and marmosets are the only primates that give birth to twins. Platypuses are the only venomous mammals. Spotted hyenas are the only mammals whose females sport pseudo-penises (through which they give birth!). And so on. If we could ask members of these species they’d claim that they’re the special ones.

But of course we can’t ask them, and in any case, this isn’t a very convincing argument. Human beings have an absolutely outsize impact on the Earth, and the advent of human beings looks like one of the major evolutionary transitions, comparable in importance to the origin of the eukaryotic cell or multicellular life. But even if we buy this, it still leaves open the question of whether there’s a key adaptation – a One Weird Trick – that accounts for the exceptional course of human evolution. Here are some candidates that being are being batted around these days:

1) The cognitive niche. The basic idea is at least as old as Aristotle, that human brings are defined by their capacity for Reason. A modern version of this is advocated by evolutionary psychologist John Tooby and cognitive scientist Steven Pinker. Pinker in particular has elaborated the argument that humans are uniquely adapted to acquire and share knowledge, by virtue of a suite of cognitive, social, and linguistic adaptations. We’ve already touched on several aspects of this: Human beings seem to have taken the capacity for thinking about physical space and retooled it for thinking about the abstract cognitive space of possession – a social relationship. (Other abstract cognitive spaces include kinship, time, and change-of-state.) And humans seem to harness the machinery for processing the sounds of interacting solid objects in creating major categories of phonemes. For a more complete exposition, here’s an academic article by Pinker, and a talk on youtube.

2) Culture. Rob Boyd and Pete Richerson, who’ve done a lot of mathematical modeling of cultural evolution, are skeptical about the “cognitive niche” argument. Too much culture, they argue, is things that have been learned by trial-and-error, and are passed on from one generation to the next without people understanding why they work. Boyd and Richerson appeal, as anthropologists have for generations, to the importance of culture. We mentioned earlier their argument that the frequency of climate change in the Ice Age was nicely calibrated to favor social learning rather than individual learning or instinct. Joseph Henrich provides a recent defense of the importance of culture. Contra Pinker, he thinks humans often don’t have a good cause-and-effect understanding of the things they do, but depend heavily on imitation and the accumulated wisdom of the elders. And see this post, for the importance of High Fidelity cultural transmission in the evolution of animal and human intelligence.

Coming up: Part Two. Recursion and Shared Intentionality