Evolutionary linguistics

Why Are We Hairless Bipeds?

Babel's Dawn - Fri, 08/28/2020 - 22:19

Lake Manyara National Park

My recent posts have discussed Donald M. Morrison’s new book The Coevolution of Language, Teaching, and Civil Discourse among Humans. I am continuing with that theme now even though today’s topic is pretty speculative and not-directly related to language or cultural origins,  but every now and then this kind of investigation encourages other thoughts about human origins, especially the reason we walk on two legs.

Bipedalism is so distinctive for the whole Homo line and yet there is no clear reason to justify it. I mean, it is convenient to free hands for carrying tools and such, but that came well after bipedalism was established, and is not much of a reason to change a body plan, especially one like the quadruped that has been so successful. I mean even bats kept all four legs. They did not get rid of their front legs the way birds did.

Morrison makes these remarks without providing much follow-up: “location and habitat must have been coevolutionary. An increasingly efficient bipedal gait, along with other traits, including the ability to swim and dive, gave our ancestors access to an expanding foraging territory, which further shaped our brains and bodies. An increasingly bipedal ape could wade deeper …” (69) so it i not his fault that I'm going on a tangent. His talk about water reminds me if a theory I’ve encountered only once before, 50 years ago when I read a pop science book by Desmond Morris called The Naked Ape. Morris made an off-hand remark about a theory that the reason our few body hairs grow in the opposite direction from that of other apes is that our ancestors had gone through an aquatic period. I had never heard that before and never did again, until coming across the reference to water in Morrison.

Well, one has to be skeptical. I don’t think there are any fossils of our ancestors that indicate that we went through an aquatic period. But wait… I once had the opportunity of asking Mary Leakey what Olduvai Gorge had looked like when it was home to the Zinjanthropus fossil she had found. She replied that Olduvai must have been much like Lake Manyara.

That idea stayed with me because Lake Manyara National Park in Tanzania is one of my favorite places on earth. It is nestled directly beside the wall of the Great Rift Valley and boasts a woodland, open stretches of parkland, lake shore, and a large blue lake. The lake itself is a bit of a hell hole, an alkaline spread of water directly open to the furnace sun with huge waves of heat shouting stay away. Water birds, especially flamingoes, are often in super-abundance, but the bitter water and heat keep the lions away, even though the prey is quite visible from the shady shore.

On the other hand, if some protohumans were able to adapt to the conditions, they would have a source of food that would keep them fed without becoming lion food themselves. Presumably, they would not have to live on the water, but like hippos, hang out between lake and shore. It is a pretty good habitat, a kind of niche found in reasonable abundance in East Africa, and yet not exploited by that many predator species. Furthermore, it offers a reason for several of our anatomical peculiarities: our bipedalism, which has never been explained; our lack of body hair; and maybe even our protruding noses. I thought of that bit about the noses while pondering the fact that if we were once aquatic, why did we not replace our noses with blow holes. But hold on, neither did hippos, and come to think of it, we do not have apelike noses. Chimps and gorillas have flat noses while humans have showy, stick up noses. Most are not as fancy as the nostrils on a hippo, but they are not bad for a primate. Also, when we go underwater we automatically hold our breath. Even very young human babies do that. Do chimps? I’ll leave that for another to explore.

You may have noticed that this post has nothing to do with language, but that is my point. Here we have a whole suite of features—bipedalism, hairlessness, prominent nose, ease under water (?)—not shared by our ape relatives and yet having no role in language or culture. Surely that oddity suggests that between the time we split from our last common ancestor with the chimps and the day we became trusting, talking, teaching humans we got to doing something unusual and unrelated to whom we seem to be today. When we think of ancestral habitats we think of the Serengeti, a grassy savanna with woodland and forest patches. But East Africa also has a bunch of soda lakes, so we should not just ignore them. And maybe we never gave up entirely on a water niche. I always loved splashing in the ocean. Maybe the pleasure has deep roots. Maybe I will learn a little more before another 50 years roll by.

Survey on researcher’s estimations of translatability of different semantic domains

A replicated typo - Tue, 08/25/2020 - 11:06

Last week, Bill Thompson, Gary Lupyan and I published a paper using word embeddings to look at semantic similarity between languages (copy of paper here). We showed that some semantic domains are more closely aligned (i.e., are more translatable) than other domains.

But what would linguists actually predict? Before the paper was released, Bill and Gary ran a survey of linguists, asking them to predict our results. Gary tweeted the results, and I’ve collected them here (text and graphs by Gary).

Survey results

Prior to the paper being published, we conducted a survey asking people to indicate what they thought were the most and least translatable domains:

Our primary goal was to see whether people who subscribed to more universalists vs. relativist views on language ranked the domains differently. We measured universalist/relativist leanings by having people respond to questions like these:

And we also asked what kinds of things people research/work on. We had 97 complete responses. (Thank you!!) spanning various language research disciplines.

Here are the correlations between how respondents rated the importance of methods and the importance of topics (using 1st principal component to cluster). There’s not much that’s surprising here. (Colored squares are p<.05).

We can also see: People who think concepts depend on language more likely to think that some ideas cannot be translated. Those who think social context is important tend to deny linguistic universals. Those who think language is innate think concepts are independent of language.

We next created a few discrete groups. “Universalists” are people who think that there are ling universals, that our concepts are independent of natural language, and language is largely innate. “Relativists” are those who scored on the other side of the median on those questions.

Where do researchers who study different things fall on their beliefs about innateness, universality, etc.? Formal semanticists think social context doesn’t matter much. Researchers studying linguistic relativity are more likely to think there aren’t linguistic universals. Again, not surprising. But nice to see numbers.

Now on to translatability ratings of domain. Overall, the correlation is rather low and primarily driven by Quantity which is ranked as the most translatable both by our respondents and in our data.

The graph below shows the size and direction of people’s mis-estimates. Kinship terms are, in reality, quite translatable, but people think they are not. Terms relating to the house (e.g., bed, ladder, chair) are rated by people as highly translatable, but our data indicate this is not so.

The graph below shows the correlations between participant groups and observed data. Some groups came closer to the observed ranking than others. Researchers studying/doing translation had the numerically higher correlation to the observed data. Notably people respond similarly regardless of their theoretical stance.

If we exclude the quantity domain, people’s ratings remain similar to one another, but correlations with the observed data drop to 0 and in some cases turn negative.

Below is a graph showing the respondent’s ratings compared to the concreteness of the words in the domain. As some may suspect, the domains people think of as highly translatable, tend to be the more concrete ones: animals, household objects, the physical world. Universalists especially think that these domains are highly translatable.

But translatability as estimated by our semantic alignment measure turns out to be unrelated to concreteness, helping to explain why people’s estimates deviate so systematically.

Interestingly, not all respondents were equally driven by concreteness. E.g., people studying word learning tended to rely quite heavily on concreteness when estimating translatability.

Responses of people in typology and/or linguistic anthropology were much less driven by concreteness.

Compared to “universalists”, who rate seemingly objective domains (animals, food/drink, basic actions) as very translatable, people with more experience in translation rate these domains much lower.

“Relativists” also tend to recognize that the language used to describe food and drink can vary substantially.

What did we Learn?

With the exception of the domain of Quantity (judged by most to be very translatable), people’s ratings of which domains are translatable deviated substantially from what we obtained using the semantic alignment method described in our paper. But for the most part, people deviated from the observed data in quite similar ways with many being driven by concreteness: People think concrete words are easy to translate and more abstract words are harder. Our data suggests that this intuition may not be correct. Many seemingly concrete categories are represented quite differently in different languages and may thus be hard to translate. More relational/abstract domains (e.g., number, kinship, time) may be structured very similarly by different langs.

The analysis script and (anonymised) data is available on request from Gary Lupyan.

The Language Race

Babel's Dawn - Mon, 08/24/2020 - 21:44

My last post reviewed Donald M. Morrison’s book The Coevolution of Language, Teaching, and Civil Discourse among Humans. This entry is a brief meditation inspired by my reading.

The fossil skulls of our Homo ancestors show an amazingly rapid growth of the Homo brain. Homo habilis (2.8 million years ago) had a brain only slightly larger than that of a chimpanzee. By 200 thousand years ago, the brain had tripled in size. 2.6 million years is long by most standards but is very fast for an organ to triple in size, and is especially impressive when you realize that sustaining neurons requires much greater support in calories than say muscle. The growth of the brain indicates a steady improvement in diet during those same 2.6 million years. How does that kind of steady growth persist for millions of years?

Morrison argues, reasonably enough, that the pressure to evolve so steadily over a prolonged time is usually an arms race and as predators gain speed so prey species improve their speed too and  the circle continues. Predators kill and eat the slower prey, so the faster prey is selected. The slow predators starve while the faster predators survive to have descendants. The enlarged brain permits the Homo genus to remember more, develop language, technology, and culture to a much higher level than our chimpanzees. Morrison's Chapter 4, “An Evolutionary Explosion,” does a fine job of considering the suite of changes made to the body of our ancestors to produce us, not just our brain size but our bipedal, gracile shape and reworked hands. (Some how he omits the loss of body hair that must have produced a serious crisis in a species that uses mutual grooming as a major means for social bonding.)

Morrison points out that there is a kind of trap in arms-race evolution. As you develop brains that require a changed diet, there can be no going back to the original diet. If the new foods become scarce, other high-power foods are required. If finding the food requires some improbable solution that calls for more than instinct, the species must be able to teach its children how the solution works. There is a cycle of dependency. Smart solutions demand that the solutions be passed on through the generations. They must be  taught. “Without language, it would not have been possible to engage in the kinds of cooperative, joint attentional activities that allowed hominin ancestors to afford their enlarging brains, thrive in the ancestral environment, spread across the face of the Earth, and fly to the moon.” (p. 79) Thus, language has to have already existed in some form while the brain was enlarging. It cannot be just the cherry on top of a long growing brain.

Missing in this discussion is a consideration of who we were racing against. It takes two to tango through an arms race. And there is a peculiar thing about evolutionary arms races; the species do not benefit so much from the process. The predators are faster, but the prey is faster too. That’s why prey are only a little bit faster than their predators. When naturalists encounter something like the pronghorn antelope of the American prairie which is far faster than any of its predators, they know to check the fossil history. The plains used to include cheetahs which could give the pronghorns a run for their money.  The cheetahs are gone and inevitably the pronghorns will eventually slow down.

But who were our ancestors competing against to get so smart? Homo habilis was already slightly larger brained than chimpanzees so they did not have to get much smarter to have a decisive intellectual advantage over their cousins. There is no point in getting much larger brains, and there may have been disadvantages. Smart as we are, we are physically much weaker than a chimpanzee. Chimps can tear us limb from limb and rip off our faces. You do not want to engage a chimp in a round of fisticuffs. Yet we got weaker and weaker and smarter and smarter.

So let’s take a hint from the speedy pronghorns and look to the fossils. There the evidence shows something remarkable. At first, varieties of the Homo genus began to multiply. Not that long ago, there were Neanderthals, Denisovans, and our own line thriving at the same time. The others died out leaving only us. Somehow we have outlasted them all. It suggests that populations of Homo rather than individuals were doing the competing. Neanderthals were apparently doing fine until our line joined them in Europe. Then, over many generations, the  Neanderthal population dwindled to nothing. A similar story can be told about the Denisovans of central Asia.

Morrison says at one point, “Groups with larger numbers of innovators outcompeted other groups,” (111) putting the competition at the group level. But he also says, “in a snowballing effect, more  complex technologies would have selected for individuals with the cognitive capacity to use them,” (112) putting the competition back on the individual level. I am very disappointed in this ambiguity because I thought I had found some company. Talk of an arms race that left us so much smarter than our nearest living primates, implies a race with competitors who are no longer around. The likeliest groups are other populations of big-brained proto-humans, so that’s likely to be who we were  racing against. Then, Morrison puts such an emphasis on the group benefits of language, especially teaching, that at first I assumed he was talking about group (or to put it technically: multi-level) competition. After all, that is how human history works today.

During the past 300 years many human populations have shrunk profoundly or died out altogether in the competition with expansive, imperial peoples. The Comanches, for example, did not meet their fate because individually they were not as tough as the newcomers. Quite the opposite; in one-to-one duels with the white newcomers, the Comanches almost always won. Yet within a human lifetime they were almost completely eradicated because they as a group could not match the technological and organizational strengths of their group-conscious enemies.[1] Just as we can ask when humans began to use language, we can wonder when their evolutionary story became more dependent on multi-level selection than on the selfish gene. I lean toward the proposition that language and group-competition are about equally old, and I am sorry Morrison does not tell us where and why he stands on the issue.

Instead he wants it both ways, “As the use of language became essential to both individual and group survival, individuals born with genetic programs that made them, in one way or another, slightly better at using language (e.g., slightly better short-term memory, slightly better at taking another person’s perspective) became slightly more likely to survive into adulthood, find suitable mates, and raise offspring to adulthood.” (113) I have never found a convincing argument that selection would actually favor the better speaker. I’d like to think that brains beat brawn and the pen is mightier than the sword, but I doubt the victory when we are speaking on a one-vs-one level. To counter this doubt, Morrison puts forward a remark by the psychologist Geoffrey Miller, “Language puts minds on display where sexual choice could see them clearly for the first time in evolutionary history.” (169) It is a fair point, one to think about, but, especially in the age of Donald Trump, we cannot assume that the revelation of a twelfth-rate mind will cause a person to live  the life of a Darwinian loser.


[1] See Empire of the Summer Moon, S.J. Gwynne

Teaching As a Human Trait

Babel's Dawn - Wed, 08/19/2020 - 14:55

What did people have to talk about when language was new? They had been getting along fine without words, and suddenly they had a few, but what was there to say?

Donald M. Morrison has  written a book (The Coevolution of Language, Teaching, and Civil Discourse among Humans) that proposes language got up and running as a teaching system. Speculation about teaching is common, but usually limited to teaching how to make stone tools. Opinions are mixed as to whether language was necessary to teach how to make the early tools, especially Oldowan tools. Showing without talking might well have been enough to teach how to make these simple tools. And then these tools stayed unchanged for eons and could not have helped that much in advancing language. But Morrison takes the subject of teaching more seriously and imagines it had a role in much more than telling apprentices, smash here.

One thing I really like about this approach is that it provides a natural means for the evolution of the speech triangle, which, on this blog, is the defining relationship of language. The speech triangle is formed by a (1) speaker and a (2) listener paying joint attention to a (3) topic. Normal animal and machine communications are limited to a one-dimensional relationship in which a (1) controller manipulates a (2) controlee. There is no obvious way to get from the one-dimensional relation to the speech triangle, or at least I never spotted one until Morrison came along.

The teacher student relationship combines both the simple controller-controlee relation with the human speech triangle relation. The feat that makes this leap to the speech-triangle possible is joint-attention. To teach, the controller-controlee relation is a little more complicated than usual because it requires the two points to pay attention to the same thing. The speech triangle is a bit simpler because the speaker and listener are not changing places so much.

Morrison says: "Once a group of early human ancestors, for whatever reason, became accustomed to combining pointing with other signals—either vocal or gestural, or both—they would have been vastly better at engaging in joint attentional activities. And because joint attentional activities (crucially, cooperative foraging for difficult-to-acquire food sources) were becoming increasingly important to the survival of the group, individuals who were genetically even a  tiny bit better at using the new system would have been more successful…"  (p. 149)

So the story goes: our ancestors were growing more cooperative and combined pointing with other actions to get everyone thinking on the same page, resulting in groups capable of paying joint attention to things. Then language could be used to teach members of the group (especially younger members) and establishing the speech triangle as a unique communication relationship. At last language was off and running.

This covers plenty of good ground, and gets the unique, relational aspects of language into the scenario. Like all scenarios, it is speculative and cannot be proven, but it does establish that a process was possible. When I first was snagged by the mystery of language’s origins, the point that held my attention was that it seemed impossible to have a beginning. So whenever I am satisfied that some detail or other is possible, I feel I am making progress. We will never know the details of the narrative, but we can at least say it is not an impossible mystery that was somehow overcome.

That, by the way, is one of my objections to Morrison’s tale. He is perfectly willing to argue that something was unlikely but could have happened once in a sky-blue moon and the unlikely development became fact. He actually states, “The emergence of language in our ancestors was, in this account, a highly unlikely biocultural accident that might never have happened, but did.” (152) I much prefer things to be overdetermined.

I have never forgiven Chomsky for appealing to the miraculous origin of concepts, and I’m not about to accept miracles in this case either. Evolution, even quick evolution like the rise of the Homo genus from habilis to sapiens takes a long time and many generations. Something unlikely may happen, but for it to stay around it needs to survive many bad breaks. Luck evens itself out. A bad bounce can determine a World Series, but not a whole season of baseball. There ought to be a statistical reason for some novelty to persist through the ages.

The emergence of language, I would say, was so likely an event that it probably happened many times, where one group on, say, the Serengeti started talking and another group near maybe the Mombasa coast quite independently started using words and another group up around somewhere else like the modern Kenya/South Sudan/Ethiopia border hit upon a third unrelated vocabulary.

The key achievement of Morrison’s book is to establish teaching as one of the defining characteristics of the human lineage and its critical presence at language’s start.

I like to imagine what I call two-dimensional teaching. One dimension is practical, the other humanizing. Practical teaching is not limited to humans, although as Morrison points out, we engage in a great deal more practical teaching than any other lineage, but the humanizing side is the astonishing part. It deals with things a person need not know in order to live a successful life of feeding and reproducing. Those are the Darwinian requirements that every species must master.

Humanizing knowledge is the cultural stuff that can be irrelevant to survival or even impede it. Before the invention of writing, this was the knowledge that was always one generation away from being lost. It includes things like manners, i.e., the proper way of doing something. Shaking hands is an example of this kind of learning. In my own case, I was taught by my father to look the person in the eye, use a firm grip, and express pleasure at greeting that person. As a result, I am always aware when the person I shake hands with looks the other way. We shall never know, but it is possible that as long ago as Homo habilis (2.3 to 1.65 million years ago) our lineage had some taught ritual for greeting one another and folks were judged according to how well they reflected similar training in the greeting. That is what I call a humanizing lesson. Once it is lost; it is gone forever, but while it lasts it is a peculiarly human bit of behavior.

This kind of little thing can define great splits. In the film Zorba the Greek, Zorba says of a French woman living in the town that she was forever an outsider because she crossed herself the wrong way. The Orthodox in the town finished the action by touching the right side of the chest and then the left. The Catholic woman from France touched the left side first and then the right. So every time she prayed she showed her foreignness. Surely differences of that nature go way way deep in the lineage, reflecting different teaching traditions that had gone on for generations.

Morrison puts the possible origin of language at 3 million years ago, which would bring us back to the australopithecine era. Of course, whenever they began, the original languages were far simpler than any modern language (which Morrison suggests may have been as recent as 200 thousand or even 100 thousand years ago). So the humanizing lessons of manners, mythology, and specialized knowledge are at least as old as Homo sapiens and verbal instruction of some sort is much older.

By focusing on teaching, Morrison hits on a trait that is especially important in thinking about humans. We are able to evolve at a much greater rate than Darwinian processes allow. A mutation spreads generation by generation. But ideas can spread among a talking and teaching species at the speed of sound. Culture outpaces biology so rapidly that, by now, very little of our behavior is innate. Even gasps of astonishment vary from culture to culture. What is inborn is our ability and desire to learn from our elders.

At some point teaching becomes less common, to be replaced by mutual thinking. Morrison calls this a shift to “civil discourse.” His book continues to be interesting and provocative, but it moves a bit beyond the focus of this blog. So I will stop here. There are some other things to note about the book, so I will have a few more posts about it.

NOTE Morrison replies: " believe we're actually agreed on the role of something like "luck" in language origins. The fact that humans alone, of all remaining primate species, managed to escape Bicketon's "prison of animal communication" was not, I think we both agree, inevitable. But neither was it, as I think we also both agree, a "miracle." Nature, as I've written somewhere, doesn't need miracles to perform its work, just known evolutionary processes and mechanisms. But, while the spark that triggered the evolution of language must have been a kind of accident...a confluence of unlikely events...the explosive coevolution of brains, language, technology, etc. that followed was, like any explosion, unstoppable. "

Syndicate content