Tuesday, October 25, 2005
Just Possibly Like So but Maybe Not So Much Stories
The Transition to Language,
ed. Alison Wray, Oxford, 2002
If DNA analysis has secured the there-that's-settled end of the evolutionary biology spectrum, language origins lie in the ultra-speculative. As a species marker and, frankly, for personal reasons, language holds irresistable interest; unfortunately, spoken language doesn't leave a fossil record, and neither does the soft tissue that emits it. In her introduction, Alison Wray, while making no bones about the obstacles faced by the ethical researcher, suggests we use them as an excuse for a game of Twister.
Advanced Twister. Forget about stationary targets; the few points of consensus among Wray's contributors are negative ones:
- Language didn't develop for the sake of syllogisms, though it couldn't have developed without some assurance of truthfulness.
- We can't even be sure language developed for purely practical purposes, since the earliest trace of non-utilitarian artifacts (ground red ochre pigment) matches the earliest appearance of homo neanderthalensis and sapiens fossils.
- Cultural change happens faster than biological change and is more susceptible to the influence of statistical outliers.
- Further muddling genetic causality, infants go through a lot of biological changes while immersed in human culture.
- Chomsky's proposal that Universal Grammar is programmed into our brains is just plain silly.
- Still, the combination of syntax and semantics kind of seems slightly unique to the human species.
The most solid lesson to take away from the book is a sense of possibility. Such as:
- "Comparative Vocal Production and the Evolution of Speech: Reinterpreting the Descent of the Larynx" by W. Tecumseh Fitch
I think it was David Hume who defined man as the only animal that shoots Coca-Cola out its nose if you tell it a joke while it's drinking. In all other mammals, the larynx is set high in the throat to block off nasal passages for simultaneous nose-breathing and mouth-swallowing.
The same holds for newborns, which is why they can suckle without pausing for breath. In about three months, our larynx starts moving down our throat and we begin our life of burps and choking. About ten years after that's finished, boys' voices break as their larynxes lower a bit more.
Aside from the comic potential, what we gain from all this is a lot of volume, a freer tongue, and a much wider range of vowel sounds.
Got that? Good, because it's wrong! In the year 2000 Fitch realized that dogs and cats sometimes manage to produce sounds above a whimper. Embarrassingly, living anatomy's more flexible than dead anatomy. When a barking or howling dog lifts its head, its larynx is pulled about as far down its throat as an adult human's, thus allowing that dynamic range the neighbors know so well.
However, humans are unique in having a permanently lower larynx.
Almost. As it turns out, at puberty the males of some species of deer permanently drop their larynxes and start producing intimidating roars as needed.
Why would evolution optimize us for speech before we became dependent on speech? To generalize from the example of deer and teenage boys, maybe the larynx lowered to make men sound bigger and more threatening?
That would explain why chicks dig lead singers. It fails to explain why chicks particularly dig tenors, or why chicks can talk. As has happened before in science, I fear someone's been taking this "mankind" thing a bit too literally. Mercifully, Fitch goes on to point out that in bird species where both sexes are territorial, both sexes develop loud calls, and so there may be a place for the female voice after all.
On a similar note....
- "Sexual Display as a Syntactical Vehicle: The Evolution of Syntax in Birdsong and Human Language through Sexual Selection" by Kazuo Okanoya
We understand how a vocabulary can be built up gradually. But how can syntax?
Having, like the other contributors, rejected genetic programming as an option, Okanoya thinks syntax began as a system of meaning-free sexual display before being repurposed: grammar as melody. The bulk of his article is devoted to the male Bengalese finch, each of whom hones an individualized song sequence over time, listening to its own progress rather than relying on pure instinct or pure mimicry — sorta like a cooing babbling infant, right? Right?
Drifting further from shore, Okanoya speculates that "singing a complex song may require (1) higher testosterone levels, (2) a greater cognitive load, and (3) more brain space." And a bit further: "Since the ability to dance and sing is an honest indicator of the performer's sexual proficiency, and singing is more effective than dancing for broadcasting...." And as we wave goodbye: "... the semantics of a display message would be ritualistic and not tied into the immediate temporal environment and, hence, more honest than the news-bearing communication that dominates language today."
As an aesthete, I'm charmed. As a skinny whiney guy with a big nose, I'm relieved to learn that Woody Allen really was the sexiest man in the world. And yet why does Mrs. Bush exhibit more coherent syntax than Mr. Bush? Does she really have more testosterone?
Perhaps we could broaden the notion of "sexual display" a bit. In a communal species, wouldn't popularity boost one's chance at survival and reproduction regardless of one's sex?
At any rate, Okanoya's flock of brain-lesioned songbirds should win him a Narbonic Mad Science Fellowship.
- "Serial Expertise and the Evolution of Language" by H. S. Terrace
A straightforward "everyone said that only human beings can do this but actually monkeys can do it too" piece. In this case, monkeys can learn how to enter a seven-digit PIN on a cash machine which changes all the positions of the buttons every time they use it, except it's a banana-pellet machine and photographs instead of digits. An ominous aside: "It is doubtful, however, that the performance described in this study reflects the upper limit of a monkey's serial capacity."
Towards the end, Terrace refers to recent research on language kind-of acquisition among bonobos. Unlike the common chimps on whom we've wasted so many National Geographic specials, bonobo chimps can learn some ASL and English tokens purely by observing how humans use them. Still, there's no evidence that their use reflects anything more than hope of reward. When it comes to utterly profitless verbiage, humanity still holds the edge!
- "ProtocadherinXY: A Candidate Gene for Cerebral Asymmetry and Language" by T. J. Crow
The candidate's a waffling policy wonk. Unelectable.
- "Dual Processing in Protolanguage: Performance without Competence" by Alison Wray
Language has words and grammar; communication has expressions. Wray focuses on units of expression which we never consciously break down into units of language, claiming that "a striking proportion" of formulas, idioms, cliches, and Monty Python recitations are manipulative or group defining signals rather than informative messages.
What we call "communication" among non-human species consists pretty exclusively of such signals, and so it is puzzling that human language doesn't deal with them more directly and efficiently. Wray's solution to the puzzle supposes a protolanguage that was all message, no words: "layoffameeyakarazy", "voulayvoocooshayavekmwasusswar", and so forth.
As any walk through a school cafeteria will remind us, the expressivity available to holistic formulaic language is pretty limited, which (says Wray) is why homonids stayed stuck in a technological rut for a million years. Meanwhile, analytic language developed slowly and erratically as a more or less dispensible, but very useful, supplement to holistic utterances.
Until it, um, was all we had and we became forced to cobble together holistic messages in our current peculiar way.
Thump. On the holistic side, there are tourists' phrasebooks, aphasics who can memorize (but not create) texts, and pundits who quote and name-drop in lieu of comprehension. But it seems problematic to claim that language derives from the holistic. On the contrary, Wray's evidence indicates that, although the need is there, language does a pretty poor job of meeting it.
- "Language and Revolutionary Consciousness" by Chris Knight
A prole in a poke. Despite the title and the opening citation from Marx & Engels, Knight's worried about how materialism might have blocked the development of language.
Human children become more linguistically skilled when treated pleasantly by their parents, but other great apes don't show much affection towards their offspring. Similarly, there wouldn't be much reason to learn language in a culture where everyone lied all the time, but a gorilla's most altruistic and cooperative signals tend to be the exclamations it can't repress. Homo nonrepublicanis is the only ape to evolve sincerity.
What caused this awful mishap? Well, Chris Knight has this theory that all of human culture all over the world began when women's genetic material realized that they'd have a better chance to win the Great Game if men couldn't tell when they were menstruating, since the men's genetic material would be inclined to seek out more reproductively active genetic collaborators at such times. Ding-dong, Red Ochre calling! — and the rest is history.
As for language? Hey, didn't you read the part about this explaining "all of human culture"? Isn't language part of human culture? Q.E.D.
[Inclusive as Alison Wray strove to be, some hurt feelings were bound to occur, and as far as preposterous anthropological mythmaking goes, Eric Gans may beat Knight. For one thing, Gans's story would be easier to get on the cover of a science fiction pulp. For another, it emphasizes the inhibitory aspect of non-mimetic representation.
For a third, it deals with a central riddle of language evolution (as opposed to the evolution of language). Some linguistic changes seem reliably unidirectional. For example, highly inflected languages are harder to learn than subject-verb-object ordered languages; when cross-cultural contact (or cultural catastrophe) occurs, languages downgrade inflection in favor of word order; and there are no known examples of a order-based language evolving more reliance on inflection.
So where did those inflected languages come from? An even more inflected, difficult, and unwieldy language? That doesn't sound like a very practical invention.
Gans has a simple fix: Language wasn't meant to be practical. Luther and Tyndale shouldn't have gotten so exercised over Greek New Testaments and Latin Masses; incomprehension's the original sacred point.
Not that I believe any of this. I just think it's cool. Jock-a-mo fee-nah-nay.]
- "Did Language Evolve from Manual Gestures?" by Michael C. Corballis
- Spoken languages are accompanied and aided by gestures.
- Language learning depends on the distinctively human gesture of pointing — an act of reference, not quite mimetic, not quite expressive, depending on a mirrored sense of other.
- Broca's area both plays a central role in language use and contains gestural mirror neurons.
- Sign languages (if enouraged) evolve from iconic shapes and gestures to arbitrary symbols connected by "natural" analogues to the syntactic ordering and inflection of spoken languages.
- In other primates, gestures are social and under voluntary control, whereas vocal noises tend to be involuntary.
- The hominid shift to an upright stance immediately freed up the hands for greater mischief.
All of this suggests (or at least doesn't disprove) that "language" could've evolved gesturally long before it became vocal. Once audible intentional vocalizing was biologically possible, there'd be good reasons to switch: yelling would cover a wider distance; semantic tokens would be more stable; it would allow conversation during tool manufacture and use.
And as proven by infants and tourists, it's possible to add vocalization gradually to gesturally based communication, avoiding that awkward "everything at once or nothing at all" scenario.
Thumbs up, as they say.
- "The Finished Artefact Fallacy: Acheulean Hand-axes and Language Origins" by Iain Davidson
A glum warning against reading too much into much-too-selected evidence. In this case, the too much is complex planning that would require language's help, and the much-too-selected are so-called "hand-axes" which might, from raw statistical evidence, be accidental by-products rather than intentional products of an industry.
- "Foraging versus Social Intelligence in the Evolution of Protolanguage" by Derek Bickerton
Bickerton wants to get back to the real reason for human communication: better food and plenty of it. As a student of menu French and Italian, I'm in no position to argue.
Actually, he's pretty mild-mannered about it. Elsewhere, he's guessed that syntax is rooted in reciprocal altruism. But since there's no evidence that hominids dealt with any more social complexity than other primates, he doesn't believe social conditions alone could've triggered a change as drastic as predicated language.
The conditions which did radically distinguish our ancestors from their primate relatives were environmental. Instead of living large in the forest, hominids roamed savannahs full of predators and fellow scavengers, and did so successfully enough to expand out of Africa. Also, unlike the socially-focused great apes, humans are capable of observing and drawing conclusions from their surroundings. (To put it in contemporary terms: Driving = environmental interaction; road rage = social interaction.) Any ability to observe and then to reference would be of immediate use to a foraging and scavenging species. Predication might develop from a toddler-like combination of noun and gesture ("Mammoth thisaway"), and lies would be easily detected and relatively profitless.
- "Methodological Issues in Simulating the Emergence of Language" by Bradley Tonkes and Janet Wiles
It's unlikely that human language's primal goal was to accurately communicate an arbitrary two-digit number. At that level of abstraction, about all computer simulations can do is disprove allegations that something's impossible. So, ignoring the metaphors, this paper shows it's possible to improve communication of two-digit numbers across generations of weighted networks without benefit of Prometheus.
Bringing the metaphors back in, they report that smaller populations and an initially restricted but growing number of inputs are helpful in when establishing a stable "language", and point out that "because many sensory capabilities are not available at birth, the child learns its initial categorizations in what is effectively a simplified perceptual environment." (We'll come back to this in a bit.)
- "Crucial Factors in the Origins of Word-Meaning" by L. Steels, F. Kaplan, A. MacIntyre, and J. Van Looveren
More computer simulation; worse anthropomorphizing. Sponsorship by the Sony Corporation might have something to do with that, and with inviting the public to interfere through a web page and at various museums. The number of breakdowns introduced by this complexity is left vague, but the project seems to have earned a Lupin Madblood Award for Ludicrously Counterproductive Publicity Stunts.
Too bad, because it's a great idea. Instead of modeling perception and language evolution separately, the project combines the two with gesture in an "I Spy" guessing game. Two weighted network simulations have access to visual data through a local video camera, have a way to "point" at particular objects (by panning and zooming), and can exchange messages and corrections to each other. The researchers monitor.
With the usual caveat about how far analogies should be carried, some of the results are enjoyably suggestive. A global view isn't needed to establish a shared vocabulary. Communication can be successful even with slightly varying interpretations and near synonyms. Again, it helps if the initial groups are fairly small, and if the complexity of the inputs increases over time.
- "Constraints on Communities with Indigenous Sign Languages: Clues to the Dynamics of Language Genesis" by Sonia Ragir
Between Creole formation, sign languages, and computer simulations, we now have a few examples of language evolution to look at. Could it be that grammar isn't genetically programmed? Could it be that social conditions play a part in the development of syntactical language!?
Well — yes. But Ragir's attack on genetic programming is kind of a MacGuffin anyway — a good excuse to cover some interesting ground.
Ragir compares nine sign languages, and, where possible, their histories and the circumstances of their users-and-originators. That (limited) evidence shows it's possible for a context-dependent quasi-pidgin to go for some generations. Grammaticalization and anti-semantic streamlining of illustrative gestures seem to happen gradually rather than catastrophically. They're introduced by children rather than adults, and only when peer contact is encouraged. The emergence of syntax is socially sensitive.
Returning to her MacGuffin, Ragir proposes:
that we consider 'language-readiness' as a function of an enlarged brain and a prolonged learning-sensitive period rather than a language-specific bioprogram. In other words, as soon as human memory and processing reached a still unknown minimum capacity, indigenous languages formed in every hominine community over a historic rather than an evolutionary timescale. As a result of species-wide delays in developmental timing, a language-ready brain was probably ubiquitous in Homo at least as early as half a million years ago. [...] As for what triggered the increase in brain size that supports language-readiness...
Here's where we come back to that thing I said we'd come back to. A human newborn is in pretty bad shape compared to the newborns of a lot of other species, and stays in pretty bad shape for a pretty long time. As Nature vs. Nurture combatants can't seem to get through their now-hardened skulls, this lets human infants and children undergo more physical — and specifically neurological — transformation while immersed in a social context.
Although that can be entertaining, maintenance is an issue. And in a savannah environment, dependent on wandering and surrounded by predators, maternity or paternity leaves would be hard to procure. It's nice that our plasticity encourages language, but what would've encouraged our plasticity?
Definitionally, hominids are featherless bipeds. But, as some readers will vividly recall, bipedalism raises a difficult structural engineering problem: If you're going to walk on two legs, there's a limit to how wide your hips can get; narrow hips limit what you can give birth to. Mother Nature's endearingly half-assed solution was to make what we give birth to more compressible.
And since the kids were going to be useless anyway, they might as well be smart.
- "The Slow Growth of Language in Children" by Robbins Burling
An attack on the all-or-nothing idea of syntax which so exercised Chapter 3. While we're growing up, syntax development isn't catastrophic, and Burling says it's even more gradual than it looks. Infants comprehend some syntactic clues long before they can reproduce them. And command of syntactical rules continues to grow long after children are reading and writing recognizable sentences. (Hell, sometimes I'm still faking it.) So why think it had to be all-or-nothing species wide?
- "The Roles of Expression and Representation in Language Evolution" by James R. Hurford
Another oppositional piece. Did language develop purely from primate calls? Or purely as a representation of our own mental activity as sum fule say? Or purely as an excuse for alliteration?
I threw in that last choice myself, but you see the problem. The options aren't exclusive, and introspection doesn't yield universally applicable results. For example, it may be true that "devices such as phonology and much of morphology" "make no contribution to reasoning" as experienced by Pinker and Bloom and Hurford, but they surely do to mine.
Still, not a bad resource when you're bored by the usual arguments against Sapir-Whorf: If our thinking was determined by language, we'd all be completely batshit.
- "Linguistic Adaptation without Linguistic Constraints: The Role of Sequential Learning in Language Evolution" by Marten H. Christiansen and Michelle R. Ellefson
So, have you heard that Universal Grammar might not be genetically programmed?
Although the impact of their dissent's weakened by its placement, Christiansen and Ellefson do well with the set-up:
Whereas Danish and Hindi needed less than 5,000 years to evolve from a common hypothesized proto-Indo-European ancestor into very different languages, it took our remote ancestors approximately 100,000-200,000 years to evolve from the archaic form of Homo sapiens into the anatomically modern form, sometimes termed Homo sapiens sapiens. Consequently, it seems more plausible that the languages of the world have been closely tailored through linguistic adaptation to fit human learning, rather than the other way around. The fact that children are so successful at language learning is therefore best explained as a product of natural selection of linguistic structures, and not as the adaptation of biological structures, such as UG.
Their eclectic research is held together by one common ingredient: learning an "artificial language" with no semantics outside its visual symbols. This reduces "language" to the ability to pick up and remember an arbitrary rule behind sequences. Admittedly, that's not much of what language does, but it includes some of what we call grammar.
Strengthening the association, in a clinical study, agrammatic aphasics did no better than chance in absorbing the rules behind the sequences. And brain-imaging studies have found similar reactions to grammatical errors, game rule violations, and unexpected chords in music.
Next, Christiansen and Ellefson look at a couple of common grammatical tendencies: putting topic words at the beginning or end rather than the middle of a phrase, for example, or structuring long sequences of clauses in orderly clumps. In both cases, we've picked patterns that reduce the cognitive load. Artificial grammars which followed these rules were learned more easily than ones which didn't, both by human subjects and by computer simulations.
- "Uniformitarian Assumptions and Language Evolution Research" by Frederick J. Newmeyer
Newmeyer begins by agreeing with the general consensus that you can't tell much about a culture from its language. That doesn't mean there are no major differences between languages, though. Or that there weren't even more drastic differences between prehistoric languages and the languages we know. Or that we really know anything about the prehistoric cultures themselves. Or when language started. Or how often. Or the physical capabilities of the speakers.
In fact, we have no facts. We're fucked.
- "On the Evolution of Grammatical Forms" by Bernd Heine and Tania Kuteva
Heine and Kuteva soldier on, trying to boil down a fairly reliable set of rules for language change and deduce backwards from them. Each visible grammatical element in turn is shown (in the examples they choose) to be derivable from some earlier concrete noun or action verb. The basic principle should be familiar from ethno-etymologically crazed types like Ezra Pound: the full weighty penny of meaning slowly worn down by calloused palms into a featureless devalued token....
Fitting the general tone of the book, though, they close with a warning that their approach is based on vocabulary rather than syntax, and so, even assuming one-way movement away from "a language" consisting only of markers for physical entities and events, we still can't say much about how they might have been put together.
And so, in conclusion, say anything.
Doesn’t Chomsky’s proposition sound more or less valid if you describe it the other way around: UG isn’t programmed into the brain, but it’s the way the brain’s evolved to handle language. Sure, I’ve buried damn near the entire argument beneath “evolved,” but still...it doesn’t sound nearly so absurd when stripped of unseemly hardware/software metaphors.
Also, wasn’t Bickerton the one who claimed language evolved with thumbs? The idea that the same part of the brain which controls gestures expanded in neuronal density and/or complexity with the development of the thumb...or am I misremembering something I read elsewhere? (And have I misremembered this particular fact before, or is this a genuine deja vu?)
A lot depends on just what one might mean by “programmed into the brain.” Does the brain’s grammar consist of things like Noun-Phrase, and Verb-Phrase and are these things genetically encoded? Steven Pinker said as much in The Language Instinct. I don’t buy that at all. I suspect that syntax is mostly minimal and that most of language structure is derived from cognitve structure.
In the end, the only way these issues are going to be resolved is through deeper knowledge of the brain. Just when that’ll arrive is anyone’s guess. I’m not holding my breath.
It’s really unfortunate that Pinker has come to stand for the entire discipline. His psycholinguistic hypotheses (as logical extensions of his theoretical work) are quite out there.
Scott, insofar as there’s evidence to work with, the weight of evidence might suggest that the brain didn’t so much evolve to handle language as evolve into something that was capable of desiring and developing language, followed (perhaps) by an intensive weeding out period. I admit it’s not a lot of weight, though; and there are those larynxes....
bbenzon, since it’s just a matter of disproving a hypothesis, wouldn’t some researchers say that this specific issue could be resolved (in favor of your own suspicions) by approaches other than driving full reductionist comprehension all the way up our conscious fundamentals? Thus the computer simulations on the one hand and the studies of actual language evolution on the other. Showing that something isn’t necessarily genetically determined is simpler than showing how it is determined.
Of course, all these folks pick on the strongest possible statement of “programmed into the brain” to make the job of disconfirmation as easy as possible. It’ll be a lot trickier—maybe impossible—to separate the possibly particularly human aspects of categorization (abstracted the way they picture it and program it into their simulations) from other possibilities.
What interests me the most, though, is something else, and I have no idea how tricky it would be to figure out because so few people are trying. Certainly not these people, who I greatly respect, or the people who are trying to dumb down aesthetics to match the dumbest of pop science, who I don’t. But it seems clear enough that what most researchers assume language is doing is not what language is “optimized” for. Yes, when cultures meet (or are devastated) for economic purposes, some sort of optimization goes on; we’d expect that. Still, there’s too much mess to be accounted for by the accounting systems being used.
There, in this persistent and reflorescent residue of the asystematic, this excess of deficiencies—here, “continental philosophy”, linguistics, social psychology, biology, politics, and the cognitive sciences all might have a place. But in a territorial war, grounds for collaboration tend to be left deserted.
Oh, the opposable thumb: oddly, it didn’t show up in this collection, although it’s often added to the mix when people talk about gestures, Broca’s area, and brain asymmetry. (But then Corballis’s article was very compact.) I don’t remember Bickerton rotating around that one in particular, although he’s sometimes a bit heavy-right-handed.
Ray, before discussing FoxPro 2.0, you should read all of the FHC vs. PJ in Cognition. You probably have, though. But the authors of this volume could not have. The sentiment you attribute them attributing to Chomsky misses the point fairly spectacularly in any case.
I do enjoy Cognition, but haven’t had enough reading time lately to catch the latest episode of the serial—thank you for the pointer.
But, as I hope I hinted, I think Chomsky is less a truly hissable villain in these things than he is a convenient excuse to spin a yarn. Of course, either way, he’s being hissed, and so I can understand why he might take offense.
My only stake in this argument is that something music-like evolved before language. It’s an old enough notion—I’m told it goes back to Rousseau, certainly Darwin. I argued the point in my 2001 book, Beethoven’s Anvil: Music in Mind and Culture. Steven Mithen has made much the same argument in his 2005 The Singing Neanderthals: The Origins of Music, Language, Mind and Body. I’ve got an essay-review of the Mithen in-press (whatever that means for a web publication) in which I bump the argument up a notch by arguing that any evolutionary trajectory leading to language must necessarily pass through quasi-music first.
Whether or not I’m right on that last argument, it is the kind of argument we need. & it depends, not on guesses about the adaptive value of language, but on the basic requires of getting two nervous systems to “talk” with one another.
It is interdependence between language and cognition that I find the most fertile fallow ground (though hardscrabble for AI). Empathetic fallacy? Literature (aesthetic not trade however scholarly) may be seen as epiphenomenal in this context (and music choraled into the mix).
But I also like that the first of Amazon’s associated Statistically Improbable Phrases (SIPs) is <a href="http://www.amazon.com/gp/phrase/ref=sip_bod_0/104-7792820-8766311?_encoding=UTF8&src=0199250669&checkSum=VCiYmRTnv8+eCsnELowjfwMi4uTzndvZUa304OOo4ls=&phrase=communicative error">communicative error</a>. But then I would.
Any opinion on Terrence Deacon‘s work?
On a dissenting note, Oliver Sachs’ “Recalled to Life” in the Hallowe’en NYer suggests mimesis is more fundamental.
Terrence Deacon linkfix. (Sorry.)
I read big hunks of Symbolic Species back when I was researching my music book. I found it quite useful. I’m sympathetic to the notion that we don’t need an innate universal grammar—though an aweful lot depends on just what one means by that—and that language and cognition have co-evolved. Beyond that . . . I don’t much recall.