In April of this year, evolutionary anthropologist turned psychologist, Quentin Atkinson, had a paper in Science titled, “Phonemic Diversity Supports a Serial Founder Effect Model of Language Expansion from Africa.” In it, he provided a mathematical evaluation of phoneme counts (phonemes are the sounds that make up speech — vowels and consonants) from a selection of the world’s languages, indicating that the number of phonemes decreases the further you get from Africa along previously proposed human migration routes. That is, languges with the highest phoneme counts were in Sub-Saharan Africa, and languages with the lowest phoneme counts were in South America and the South Pacific. Phoneticists, of course, immediately questioned how you can accurately compare phoneme counts from Sub-Saharan Africa (which often sport doubly articulated consonants) to languages that employ only single articulated consonants, but, leaving this aside, the math in the paper looks to be pretty solid. And if you want to complain that the 500+ languages compared in the paper are too few to be properly representative of the diversity of linguistic expression around the world, that’s fine, but I’m going to ignore you. Let’s get to the bottom line: the paper provides some neat evidence in favor of estimated routes of human migrations through the unlikely channel of human language.
Why is this so surprising? Well, for those who don’t know much about historical linguistics, the basic premise of that field is that languages change… a lot. Some parts of language are particularly fickle, like vocabulary items. For example, when did you last hear something described as “boffo” or as “the elephant’s instep”? Probably not that recently, excepting ironic usage by your local barista. Other parts change a bit more slowly, like the grammar of a language. Here, words like “whom,” which is notoriously on its way to extinction, find themselves only infrequently replaced. At even slower rates of change, there are modifications to the sound-system of a language, especially in terms of addition or loss of particular sounds. Consider the distinction that used to exist in English between “w-” and “wh-“; although this distinction (as between words “which” and “witch”) is still maintained in some areas of the world, it is more common to find that the two have merged to the “w” pronunciation. In those areas, “wh-” has been lost, resulting in a smaller inventory of available sounds, and it is this particular phenomenon that Atkinson adresses. (Other examples of historical change abound, and a good text on historical linguistics will provide many more examples and a much deeper level of discussion than I can here.)
So, given that languages change a lot, a lot of people were caught by surprise at the time depth that Atkinson suggests is available in the current data. In the paper, he proposes that a common linguistic ancestor-system must have occurred about 100,000 to 150,000 years ago. In terms of historical change, that’s a long time! When you consider the number of changes that occurred in the span between Old English to Modern English (a period of around 1000 years), a span 2 orders of magnitude larger would seem impossibly long to discuss in terms of linguistic change. After all, the languages involved should have been far too prone to having changed to a dregree that any evidence of human origins would be erased, but, no — in spite of those millenia, a glimpse at origins is indeed what we seem to find.
Amazing, really. So, what’s the concern? I’ll admit that, to me, the math seems solid enough. I personally can’t complain about the numbers. But, as a trained linguist, I can (and do!) take some issue with his proposed mechanisms for progressive phoneme loss along human migration routes. Atkinson’s suggestions is that language changes in the same way that species do: that there is repeated application of the founder effect for progressive offshoot populations. In evolutionary genetics, the way this works is that large populations contain a lot of genetic diversity. Since a lot of this variation occurs at the level of the individual, any given individual will express only a subset of the available variability in the entire population. Think of it this way: let’s say that 10% of the people in a population carry a gene for green eyes. If a handful of individuals from this population decide to go off and start a new colony, there’s a fair chance that none of them will carry that gene. Over progressive generations and progressive offshoot communities, the communities that are most distant from the original community (most distant in terms of being offshoots of offshoots of offshoots, etc., etc., etc.) will have the smallest subset of genes.
In my opinion, this is where we run into immediate trouble. Language doesn’t function in the same way that a collection of genes does. When a small group of people leave a community to forge ahead and create a new settlement, they leave speaking the same language that the community spoke. The phonemes they leave with will not be merely a subset of the available phonemes from the initial population (barring founders’ speech impediments). Change should thus occur after initial migrations start. In order to accomplish that, two mechanisms spring to mind:
1. Isolation and Drift
Offshoots over time will experience some amount of natural language change and drift. Since there is a large amount of variability in phonation anyway (think of all the different vowel sounds you use to say “Oh!” to express various things), over time, these variances can build in such a way that various speech sounds will be produced differently after a few generations. Where one might expect large populations to wash out some of the variability on average, interconnectivity does not affect change and drift into new dialects. For example, in the United States, you might expect there to be a diminishment of regional differences with the advent of mass-communication — television, radio, the Internet — but, surprisingly, the opposite is true: American dialects are becoming increasingly different. Small population size may thus not have the particular effect that we’re looking for, which sort of devalues the founder effect. The other small issue with this sort of change is that drift can indeed eliminate previously existing distinctions by merging them, but it can likewise lead to separation of speech sounds as they diverge into two close but different phonemes, a point brought up by Trudgill (2004). For example, the “w-” and “wh-” merger mentioned earlier was actually the end result of an earlier split, with Old English possessing only the “wh-” sound. The “w-” / “wh-” distinction seems to arise somewhat later, during the period when Middle English was spoken, as, around that time, the two sounds started seeing different use in different words. (further discussion).
As for the other mechanism of phoneme drop, the one that I feel would better fit the data is:
2. Loss through Progressive Pidginization/Creolization
This may take slightly more explanation. The basic idea here is one that comes from contact linguistics. That is: languages change most when they are in contact with other languages. As pointed out by Thomason and Kaufman (1988), so many features can be changed by linguistic contact that it becomes difficult to cleanly extract genetic relationships from languages that have been in contact, making basic Stammbaum theory a little less applicable. Taking this into account, languages with the greatest amount of contact are going to be the ones with the greatest amount of apparent change, which is, interestingly, the basic finding of a paper that Atkinson previously published. Now, at the extreme end of contact, where change is most radical, there is the process of pidginization, which occurs in the development of a basic, limited-use lingua franca between groups that do not otherwise share a common language. Since it almost always occurs when 3 or more languages come into contact, what usually happens is that one language ends up donating its lexicon (vocabulary) for usage (this language is referred to as the ‘lexifier’). One hallmark of pidgins is that the phonology of the pidgin is notably reduced from the phonology of the lexifier, tending to maintain only the phonemes that overlap in the contributing languages. This is the ultimate in linguistic accomodation (again, Thomason and Kaufman (’88)).
In the textbook definitions of creolization (n.b. – the field of creolistics has not yet fully settled on a rigid definition of creolization), this pidgin language may become established as a first language of the community, at which point the basic, limited-use lingua franca expands to be employed for everyday use. As described in Mühlhäusler’s classic 1986 text, Pidgin & Creole Linguistics, when this happens, there is often some expansion of the phonological inventory of the language (generally borrowing some of the original distinctions from the lexifier), but the final phoneme inventory is still not as vast as that of the lexifier to begin with. (For additional discussion of the complexities of creole phonology, see Norval Smith, “Creole Phonology” (2008)) What this suggests is that as populations collide, there will be progressive shrinking of phoneme inventories.
Ok. So here are two mechanisms. We still haven’t adressed two big questions! First, why does the original population have so many phonemes? And, secondly, is there anything we can derive about human population movements from these two mechanisms of loss? Let’s adress the latter first:
Human population movements
The above two mechanisms surprisingly predict two differing styles of exodus in the Edenic diaspora. Isolation and Drift, (1) above, seems to be fairly non-directional, so instead we’ll substitute Atkinson’s proposed mechanism of founder effects for this part of the discussion. If we assume founder effects, then we would expect to see mostly frontier expansion. Again, according to Atkinson, small groups would promote greater shifts (as there is not a significant background of variation to determine what is meaningful variation and what is noise), and this would suggest movement mostly limited to groups migrating along the edges. This also suggests that expansion occurs for the most part because there are new lands to take advantage of, a trend which would no doubt be intensified and augmented by the growing phenotypic wanderlust of the people at the expanding edge, in line with Shine, et al, (2011).
On the other hand, (2), Progressive Pidginization, would maintain that the forces promoting expansion are fairly constant throughout the population, reliant more on environmental factors than anything else. In this case, we can expect movement of groups from any of the already settled areas. Thus, when drought strikes in one place, people move until there are enough resources to support them, which will often take them on a leap-frogging path across already settled areas, bringing them into contact with many other groups along the way. This would heavily promote language shifts, and the more settled areas a group has to pass over to get to unsettled land, the more opportunity for loss and change by the time they settle down. This also suggests that in large-scale natural disasters and crises, many groups would come into simultaneous contact, and thus would increase the likelihood of pidginization and chances of phonemic drop essential to the model Atkinson provides.
One could also provide some sort of mixture of these two styles of expansion, but, frankly, that’s less interesting. Likewise, these basic movements are doubtless overly simplified, and there is at least one other factor that sort of confounds these hypotheses, but it will take some explaining —
Other Population Interactions / Initial Phoneme Inventory
A discerning reader might have noticed that this post is categorized under “fringe science.” There is, as with all things, a reason for this. One offshoot of (2), above, is that there is another factor that likely influenced human migrations to some extent. As this particular factor is definitionally without supporting evidence, it seemed essential to categorize this post as “fringe.” However, given mounting data from related disciplines, I feel that the claim I’m about to make is not unfathomable nor even unreasonable. This factor, this offshoot of (2), is that modern humans, when migrating out of Africa, were not just running into open land; they were running into other hominids that had previously left Africa, and were probably interacting with them vocally. What these interactions were like, of course no one can fully answer, but we can make some scientific guesses. First off, to establish our bases, can we really suggest that Homo sapiens was even remotely interested in interacting with other species of Homo? With increasing confidence, we can say: yes. Evidence continues to mount that humans interbred with other hominids. Most recently, this paper published last month finds evidence of Neanderthal genes in our genome, making it the last in a fairly long line of papers advancing the same view. (For more work on Neanderthal genetics, see The Neandertal Genome Project.) Likewise, this 2010 paper presents some evidence suggesting that modern humans in migration into Melanesia first came into contact with and interbred with the Denisova hominin, which are dated as being an earlier (and therefore more distant) offshoot of the Homo line. Additionally in the region, having started to leave Africa 1.5 million years ago, was Homo erectus, a relative even slightly more distant. But does interbreeding and contact suggest communication? Not necessarily. After all, nature provides us with many examples of a diverse range of species that manage to breed without so much as a single word exchanged between mates. (And even humans, on occasion.) However, the relatively advanced cultural knowledge that these other hominids possessed suggests a range of communication that Homo sapiens might have interacted with. Even erectus may have had the ability to create fires and tools, which suggests a level of transference of procedural cultural knowledge indicative of basic communication skills. Perhaps even enough to converse with modern humans.
If this is the case, then the first question that arises is: what were the communication systems of other hominids like? Here, we move even deeper into the realm of fringe science and speculation, so I won’t make too many predictions, but I will propose that other groups of hominids likely had fairly stable phoneme inventories that were much smaller than the inventory of our ancestor population in Sub-Saharan Africa. This begs the question: why the difference? This, even in speculation, is unclear. One possibility is that of genetic mutations related to our speech capacity. FOXP2 is a gene that has been implicated in the complexity of human communication and reasoning abilities. People with mutations in this gene tend to have trouble with symbolic reasoning and difficulty with structurally complex language. Interestingly, the modern copy of the gene is the result of a mutation about 200,000 or fewer years ago, meaning that it is limited to modern humans — more distant hominids did not share this mutation. Given Atkinson’s time depth of up to 160,000 years, this would have been, in evolutionary terms, not terribly long after the mutation became ubiquitous in the population. Would this matter? In this regard, we need to take a short detour through the formation of new vocabulary items in signed languages. When a deaf child is born to a hearing family that knows no sign, what often springs up is a system of “homesign” or “kitchen sign”, where gestures are reinforced between the child and the siblings and caregivers, such that an elementary linguistic system is forged. The individual signs tend to be highly iconic, where a sign for knife might be based on a cutting or sawing motion, the sign for shovel might be a shoveling motion, and so on and so forth. Once the sign becomes established, another process takes over — taking established symbols and making them more efficient. In this manner, lengthy or complex signs are reduced to some of their more basic components to allow greater rapidity and complexity of communication through combinations of symbols. Though no exact correlate has yet been described in spoken languages (given the lack of situations in which this would happen (raising children without communicating with them is unethical)), we can estimate that early hominids with new FOXP2-enhanced symbolic representation abilities might suddenly find the sounds around them to be more meaningful. These sounds might then be taken into the newly evolving language. A word for bird might start out incorporating whistles and chirps, the word for rain might start out mimicking the patter of raindrops, and so forth. It would take only a short time for these vocal symbols to become canonized, at which point, pushes to increase efficiency may take over. Some sounds that are unwieldy to make in fluent conversation (like whistles) may drop out, whereas others, like ingressives (clicks) which can be made in fluent speech might be kept. If you think this is a bit of a reach, consider the linguistic subfield of sound symbolism — the basic premise being that words tend to evoke some aspects of the things they represent. A sharp object would therefore more likely be labeled “tschikk”, which sounds abrasive, than “mub”, which sounds blob-like. (For recent support to these claims, see this recent paper comparing basic vocabulary in nearly half the world’s languages.)
However, this may not be the whole story… after all, Neanderthals bear the same FOXP2 mutation that we do. If this is at the heart of our increased appeal to iconicity, we might have to make an exception for our closest relatives. However, this is not to say that they would make the same choices of symbolic incorporation that we did. If this is the case, then later interactions would again lead to the same sort of pidginization process where only the intersection of phonemes is used in the medium of inter-group communication, leading to a reduction in overall phoneme count. Or, it’s also possible that other groups merely kept the basic set of speech sounds that are endemic to our species.
Right, then… what do I mean by sounds endemic to our species? Well, babies babble. Even deaf babies. What’s interesting is that the phonemes that babies produce tend to be limited to a fairly small set of regularly occurring phonemes, which also happen to be the most common speech sounds in the world’s languages. Assuming that babbling evolved not initially as a linguistic form, but as a phenomenon co-opting a number of previously existing vocal signs for use in attracting caregiver attention, the suggestion then pops forth that other hominids may have been using these inventories for their communication, as well. If this is the case, then interactions of modern man with these other hominids with those small consonant inventories would have tended towards loss of some of the larger inventory of the sapiens line through contact.
Of course, there is another possibility that Atkinson does not bring up, but which also seems reasonable to suggest. The founder effect relies on reduction in diversity. However, this, again, is in regard to genetic material which operates somewhat differently from phoneme production. Taking the mouth and throat as tools to divert and manipulate the airstream requires us to consider how many ways these tools can be configured. The musclature of the mouth, after all, limits the number of ways the tongue can be positioned. On the cognitive side, close differences in phonation tend to be clumped together — something called categorical perception. Taking these factors together suggests that there may be a limit to the number of phonemes that humans will find useful for communication. However, that limit is fairly large, and well above the number of basic phonemes an infant babbles or that are found in the world’s most recently established languages. What this suggests is that we might not be dealing with progressive loss at all, but rather progressive gain, a concept that John McWhorter has been pushing in his Creole Prototype Hypothesis. The simplest languages are the newest, and complexity is something that grows as a language becomes more established and matures. The Sub-Saharan African langauges are so phonemically complex because they have had time to mature to that level of complexity.
What is interesting about this concept is that is actually creates the same pattern of results that Atkinson finds! The cline of variability will still find its way back to the origins of modern humans in Sub-Saharan Africa. And what’s also nice about this is that it would work under the premise of slow changes, rather than positing some sort of massive phoneme creation event from which most of us have been moving away. Rather, it would fall well in line with other well-studied phenomena, like human personality, making language, again, look like an autonomous organism, growing and becoming more complex and varied as it ages.
Languages thus can be born, they can die, and it would also seem they can grown and mature throughout their entire lives.