Unconstrained Verbal Drift

This week Babel’s Dawn  has been looking at evolution and speech origins, but so far we have looked only at biological evolution. That cannot be the whole story behind language as a whole or even speech because language’s cultural component is too strong. Yet the basic facts about organisms (see the first post on evolutionary processes here) have their verbal analogies as well.

1. Speech is filled with random deviations from the rules.
2. The rules of language are well adapted for generating sentences.

Long ago Noam Chomsky noted the contradictions between these two facts, and argued that our understanding of the rules of language (our linguistic “competence) must be inborn, since the language we experience (our linguistic “performance”) provide such an impoverished example of the rules. If we learned through experience alone, we would never learn the rules.

It’s a nice try, but it suffers from the same problem that intelligent design faces when it ignores the need to maintain the species over time: along with intelligent design, we need intelligent maintenance. This blog is very strong on the notion that biology (and therefore Darwinian evolution) is an important element in the story of speech origins, but biology alone cannot explain how the language we see remains meaningful over time. The basic problem is this: language keeps changing, but it does not become meaningless. Just as natural selection provides an explanation for how life can continue for billions of years without becoming extinct, we need an explanation for how language can continue for thousands of generations without losing its ability to express meaning.

Let’s look at how language drifts along all three of its dimensions: phonology (the sounds of speech), semantics (the meaning of words), and syntax (the rules of sentences):

phonological drift: unconstrained drift should eventually lead to speech that consists only of um uh uh um, but obviously that cannot happen. Charles de Gaulle once made a critical speech about the waning French Empire that began, “I speak to the French peoples,” and then he paused to add, “In the plural.” That was because the French distinction between peuple and peuples, while retained in writing, is no longer voiced in speech. Obviously, when the syntax of a critical distinction has to be named, drift has gone too far. But this much drift is unusual, and, as the meaning of words is generally unrelated to their sound, constraints on phonology can be lax. The chief source of phonological drift seems to be the instability of speech output. A single speaker may pronounce the vowel sound in “boot” as a long u one time and then as a short i, the next, usually without anybody remarking on the difference. Presumably, if it mattered, we could have evolved a more stable vocal tract. Why didn’t we?

semantic drift: in this case unconstrained drift should make more and more of a vocabulary meaningless, but that seems not to happen. The meaning of words change, but do not hollow out. New words arise, old words disappear, and existing words are routinely redefined. The need for this semantic evolution is obvious. It is the only way we can continue to speak appropriately about our changing fields of attention, but how does our speech manage to keep up with our attention? Why doesn't change merely overwhelm us?

syntactic drift: changes in syntax threaten to obscure the relationship between words. This is the point where ape signing breaks down. When they put more than a couple of words together, even sympathetic onlookers find it impossible to tell how the words are related, or whether they are related. The syntax of languages seem surprisingly unstable. A thousand years ago, English expressed many relationships by adding inflection marks to the ends of words. Today the inflections are mostly gone and relationship is indicated by word order. Something similar happened in the change from Latin (which is highly inflected) to French (which depends much more on word order). Yet syntax does have some long-lasting elements as well. Indo-European languages have a certain “feel” that makes them instantly distinguishable from, say, Semitic languages. Ancient Akkidian syntax is more like modern Arabic than modern Arabic syntax is like modern English. And beyond that there is still more stability. All languages have subjects and verbs. So if there is some stability, why isn’t all of syntax more stable?

Something must be maintaining language so that it remains fundamentally meaningful even as all its surface characteristics change. We understand that natural selection maintains fitness over time, but language changes much too rapidly for its processes to be at work. So what is maintaining language over time?

Comments

[TLTB]

"Obviously, when the syntax of a critical distinction has to be named, drift has gone too far." Drift or no drift, languages vary in which morphological categories they mark grammatically (vs marking it lexically which is what the French leader was doing). Some languages get along just fine without any grammatical plural marking at all. In certain Native American languages, you mark on the verb how certain you are about the action you are discussing. In English, we can't do that except to say "I heard.., I saw..., I think..."

But in regards to your general point, there is no reason that biological conservation can't be behind linguistic change. If it is true that a single set of genetically-programmed principles is behind all the variation of the modern languages of the world, there is no reason it can't be behind all the variation of the states of a single language over time. Just as biology places the constraints on how much languages can vary, it can place constraints on how much languages can change.

----------------
THE BLOGGER RESPONDS

I hope readers noticed this clause buried in the middle of the second paragraph, for it states the faith on which this comment rests: “If it is true that a single set of genetically-programmed principles is behind all the variation of the modern languages of the world,” IF IT IS TRUE…

Programmers will notice the basic problem with this doctrine: you don’t program principles, you program procedures. If X, then do Y, Else do Z. I’d like to see the procedures for constraining the transition from Old English to Modern English while also constraining the transition from Sanskrit to Urdu.

[TLTB]

It cannot possibly be denied that a single set of genetic endowments makes the acquisition of any human language possible, and that therefore all linguistic variation is constrained by that endowment. The only thing that can be argued about is how much of linguistic variation (synchronic or diachronic) is constrained by that genetic endowment (no one would claim it is 100%) and how much is due to cultural influence or random, unconstrained change. Clearly both nature and nurture play a role.

But acquisition IS genetically- endowed. We cannot NOT learn language, as I think you have acknowledged. So why do we need a mechanism other than acquisition for guaranteeing that language doesn't become meaningless over time? Could one acquire a meaningless language? Could there be a meaningless language? The notion seems incoherent to me, so maybe I'm missing something.

--------------------
THE BLOGGER RESPONDS

“The only thing that can be argued about is how much of linguistic variation (synchronic or diachronic) is constrained by that genetic endowment (no one would claim it is 100%) and how much is due to cultural influence or random, unconstrained change.”

To put that in terms of this blog: how much of the origin of speech reflects biological evolution and how much reflects cultural evolution?

We will see where the blog takes us, but at this point my suspicions are that biology was fundamental in giving us the machinery for producing speech and almost irrelevant to the content of the speech, i.e., the meaning of what we have to say. I’m not entirely alone in this opinion. For example, the linguist Derek Bickerton writes, “The most crucial thing to grasp about the emergence of symbolic representation is that it must have been primarily a cultural rather than a biological  event.” Bickerton says that it could arise culturally because “the minimal necessary biological equipment was already in place.” I don’t always agree with Bickerton, but at this point we do seem to share common ground.

As a thought experiment: try to imagine what would happen if (a) vervets found themselves in a setting where their innate vocalizations were no longer appropriate to their circumstances, or if (b) a computer programmed to converse in one setting was required to converse in quite a different setting.

[TLTB]

Bickerton could be right, but that just removes the question to another domain: how did we get the machinery in the first place? What adaptionist pressures (if we are sticking with an adaptionist account) led to its development? If not for speech (and I am sympathetic here), then for what?

If that's your view, then you aren't far from Chomsky, Hauser & Fitch who clearly stated that they think the basic machinery of language (for them recursion) may have evolved for a purpose other than speech.

-------------
BLOGGER: Of course your questions ("how did we get..." etc) are the issues the blog tries to explore.

[David]

I'd actually like to suggest that the fundamental evolved characteristic of human cultural learning is replication fidelity, paralleling the mechanisms for genetic replication fidelity in cells. The basic mechanism for copying fidelity of cultural behaviours is scaffolded learning, which is the process by which infants acquire language from their caregivers. It seems likely that this capacity developed early in the evolution of Homo, as it could explain the appearance and extreme conservatism of the Olduwan and Acheulian tool traditions (1m years each with next to no variation).

Another relevant point is that language strata vary at different rates. Articulation can change very rapidly, in generations, as soundings can change without affecting meaning. Lexicogrammar (vocab and syntax) takes longer, centuries to millenia, as words and their patterns directly encode meanings. Discourse semantic patterns seem to be most persistent, perhaps for tens of millenia, as they most directly interface with the ecosocial contexts for which language is adapted. Indeed discourse semantic patterns only seem to change significantly in response to radical ecosocial changes, such as the European colonisation and industrialisation of the last few centuries, in which modern written modes have evolved, in science and adminstrative fields. In contrast, spoken languages of all cultures seem to mean in basically very similar ways, with myriad peripheral variations that tend to attract the attention of linguists.