The Origin of Language: Tracing the Evolution of the Mother Tongue

Who would have thought that we could reconstruct any part of the earliest language? Yet, it seems that we can, as Merritt Ruhlen shows in this book. Others have already described the contents, so I just want to respond to some of his critics.

1. Apparently, his earliest critics complained that he wasn't using a method approved by professional linguists. His response is that linguistics only made progress AFTER someone (Sir William Jones in 1786) used the same method that he was using, and if it is illegitimate for him to use it, then it was illegitimate for that person to use it, too, which would throw out all of linguistics and not just his work.

2. "Languages change too much over time for the earliest language to have left any traces." But we don't know this. Artistic styles in ancient Egypt didn't change at all over a few millennia, so why MUST language change? If language was thought of as sacred, then it may not have changed for thousands of years.

3. "Ruhlen chose the data that would give him the conclusions he wanted." Ruhlen used data to show that in the earliest language, "tik" means one and "pal" means two. Anyone who disagrees can do their own study showing that "pal" means one and "tik" means two. As far as I know, no one has done this.

4. "Going back that far leaves too much to chance." Let's say that the probability of our reconstruction of Proto-Indo-European is 90% accurate. Going back to its proto language would give 81% accuracy, and going back yet further would give 73%. 73% is of course as not as accurate as 90%, but on the other hand, it means that the reconstruction is more likely to be right than not, and that is more than I would have expected. Of course, if the original reconstruction were only 75% accurate, then by the third reconstruction we would be at 42%, which isn't very impressive. Now if mainstream linguists want to say their work is at best 75% accurate, then they've got a good objection, but I suspect they want to say their own work is better than that.

5. "Ruhlen whines." The logic that Ruhlen used in the first point I listed above I found to be quite clear and compelling. Nevertheless, another reviewer, Robert L. Trask, utterly failed to understand it. Trask was a professor of linguistics at the time he wrote his review, whereas the back cover of this book doesn't even give an affiliation for Ruhlen. The fact that a professor of linguistics couldn't understand the simple logic that someone who wasn't a professor used is, I think, an excellent reason to complain. I don't think of it as whining.

By the way, when I first heard of Ruhlen, he was teaching high school, but he is now apparently at Stanford, a move up that is well deserved.

I have given this book 5 stars because of the potential significance of the ideas it contains. The book is a bit quirky and I can see why those trained in linguistics might give it the thumbs down. The methods and approach that are described in the book are not rigorous. However, they are designed to present a few important concepts and in that regard, I think the book succeeds.

First of all, this book and its author have been seriously challenged by mainstream academics. But it is worth pointing out that there are two sides to any debate. In a nutshell, the author bases his ideas (in part) on those of his mentor Greenberg and he presents the idea that it is possible to identify protolanguages that existed and form the basis for the language families of today. In itself, that is nothing terribly new or controversial. It is well understood that languages like Hindi, German, Welsh, Latin, English, Greek, Hittite etc. all fall into a common language family known as Indo-European and that a proto-indo-european (PIE) language may have been spoken somewhere in eastern Europe or the middle east thousands of years ago.

Rather than review the book in depth, I would like to take the space provided here to argue against the critics of this book. The book becomes controversial when the suggestion is made that remnants of an original common world language can be identified from a few common words found in many of the world's existing languages.

The major criticisms of this idea are (as I understand it):

1) Ruhlen does not use methodology that is considered appropriate by most linguists. He uses a global comparison approach which many linguists believe favours the identification of chance similarities. Ruhlen himself argues elsewhere that the length of the words he has identified rules out the possibility of chance comparison. Because I do not know any way of easily evaluating this statistically, I cannot comment on this argument. However, as a scientist, I would never dare to argue that something was not true solely because the only method that was accepted by the academic community provided no evidence for its truth.

2) It is argued that the rate of language change is too great for there to be any possibility of comparison between major language families. Here I have to differ. Of course I am not a linguist. But it seems to me obvious that the rate of change of languages depends on many social and geographic factors. A larger population might be expected to have a more stable language structure than a smaller population for example. It might be argued for example that language would evolve faster among small mobile populations spread out over a large geographical space (e.g. the America 12,000 years ago). It is argued that over just a few hundred years, French and English have changed almost to the point of being different languages. This may be true, but obviously we can still easily discern similarities between Indo-European languages.

So, how old is PIE? It is argued on the linguistic evidence (the reasoning seems a little circular here) that PIE is at least 3500 years old and probably about 5000-6000 years old. The problem here is that the archaeological evidence just doesn't support this. We now know, for example, that the "Celts" moved into Ireland soon after the end of the last Ice Age (not during the Iron Age as once thought) and the genetic evidence argues against the mass movement of people (and therefore language) in recent millenia. For this reason alone, we might date PIE to at least 11,000 years ago.

Of course the arguements linguists use to date languages are sophisticated and I am cautious in concluding that they are wrong. However, their dating is only as good as their methodology. They would argue that movement of people does not necessarily coincide with movement of language. Hmmm... yes, a ruling elite can impose its language on a majority subservient population (Anglo-Saxons, Normal French etc.) but there is not archaeological evidence for this sort of thing during the neolithic period, say in Ireland).

Linguists use techniques such as glottochronology to date languages. I am not an expert on this technique, but most of the graphical presentations of the data I have seen do not suggest that statistically significant conclusions can easily be reached.

Another technique that is used to date languages is to examine words that are common to all members of the family. "Horse" for example in PIE. On the basis of this common word it has been concluded that the horse must have been domesticated at the time of the invention of PIE. This, of course, is not necessarily the case. It is quite possible that the horse was known, but not domesticated, at this time. I would recommend a book on the Origins of Vedic Culture for a thorough analysis of this subject The Quest for the Origins of Vedic Culture: The Indo-Aryan Migration Debate

3) It has been argued that Proto-World (the original language) must be hundreds of thousands of years old and that this date is well beyond the period that similar words would be preserved. Here I must differ most strongly. Both the archaeology and the genetics suggest that language could have been invented approximately 50,000 years ago. This period coincides with the start of the great leap forward in artistic endeavor and the migration of people out of Africa. It also coincides with evidence for a major linguistic event (leap forward) as defined by a genetic sweep that occurred in the FoxP2 gene which is linked to language acquisition. If we believe an older date for PIE of say about 10,000 BCE, then 50,000 BCE does not seem so distant.

So in summary, the hypothesis that a Proto-World language can be rediscovered through a global comparison approach is feasible, although unproven. Therefore, this book and other work by Ruhlen needs to be taken seriously.

The book presents some of the concepts of language families in an easy-to-digest fashion. But this is also its weakness. If you are not trained in linguistics, this book will seem easy to read and enjoyable. If you are trained in linguistics, then its methodology may annoy you. Either way, it is ground-breaking stuff.

I can also recommend The Unfolding of Language: An Evolutionary Tour of Mankind's Greatest Invention by Guy Deutscher as an intelligent read and a great exposition of the mechanisms of language change.

I hoped a more scientific book, it starts with too many assumptions, but I have like and I think is worth reading it, at least reading it, since it's not a book to consult frquently.

This is Ruhlen's point. Based on modern similarities, all languages are related, some more distantly than others. At the end there is a "tree" showing the purported relationships between our language families. Ruhlen offers his case; he does not prove it. And with the exception of a handful of words, he does not attempt to recognize *Proto-World. He argues that reconstruction is unnecessary.

This book is for the lay reader. It offers a small group of words from a group of languages, and asks the reader to compare them. In each case the careful reader will find the intended relationships. Thus "The Origin of Language" guides the cooperative reader to accept the existence of language families, and then, and here's the controversial part, to accept the existence of links between families.

At first I found this infuriating. I was not a cooperative reader. After all, I have always known that there are several major language families and several isolates, each separate from the others. Ruhlen hand-picked words that would make his point. But as I read on, I accepted first that he was making a case that I disagreed with and that was likely wrong. And by the end, my previous thinking had been shaken.

"New Synthesis"

"The Origin of Language: Tracing the Evolution of the Mother Tongue" falls into a category of scholarship that seems to go by the name "new synthesis." Mutually supporting bits of linguistic, archaeological, genetic, and social evidence are woven together to tell the story of humans leaving Africa and spreading, first along the shores of the Indian Ocean and on to New Guinea and Australia, later to the interior and western parts of Eurasia (and even later to the Americas). Authors including Renfrew and Cavelli-Sforza have written, in their own fields, books which fit into this new synthesis. A nice introduction would be Steve Olsen's "Mapping Human History."

Worth a Look

You may not agree with Merritt Ruhlen's thesis. But if you curious about the origins of language, you should take a look at this short volume. Reject it if you will, but at least you will know what the "lumpers," the single origin people are claiming.

This book, written for lay readers, ventures two arguments. One, which seems plausible enough, is that existing linguistic families correspond to genetic markers in the peoples who originally spoke them, and are related in a similar tree that relates human populations. The other, far more controversial, is that just as *Homo sapiens sapiens* is the offspring of a single stock and a common group of ancestors, so also are all human languages related; and in fact some root words of Proto-Human can be reconstructed by the comparative method.

Amateur philologists like myself will naturally jump at such a tantalizing suggestion; and perhaps one test of how good an amateur you are is to see how many problems you can find with his development of the themes.

His basic method, like Joseph Greenberg's, is bulk comparison of vocabulary. Few attempts to reconstruct underlying forms are made here, with some conspicuous exceptions. His method is to present words in lists, and invite readers to perceive similarities themselves; an interesting rhetorical ploy that makes the reader the accomplice in setting up the main thesis. In these lists, reconstructed forms from protolanguages appear unmarked next to vocabulary items taken from wide ranges of existing lnaguages; there's no reassurance that any of the words in one of his rows are from the same language, or are actually attested forms. The same is true of the long lists of words cited in text: protolanguages appear besides dozens of obscure languages. Nouns, verbs, and grammatical particles or inflections appear alongside one another in the same list.

The rest of the book is taken towards presenting an argument that explains why reconstructions or analyses of the history of his roots will not be forthcoming. Ruhlen argues that no such analysis was necessary to identify Indo-European, which was discovered on the strength of word lists alone. One flaw in this argument is that scientific hypotheses are not thumbs-up or thumbs-down propositions; rather, they gain or lose confidence depending on the depth, power, and detail of their explanatory power. It isn't that these word lists aren't enough to get to "maybe," it's that they aren't enough to go further than that.

In any case, phonological and historical rules *do* appear in the text. Ruhlen notes correctly that often a /k/ sound is voiced to /g/, or that /t/ sounds can be palatized to /c/ or /s/ sounds. These phonological rules let Ruhlen cast a wider net, allowing more and more words from different languages to be mustered as evidence. They *never* appear as historical grounds to reject a purported cognate, on the ground that the inherited form in this particular branch must have changed in an intermediate stage prior to the observed form in one language.

The bottom line is that Ruhlen's hypothesis stalls at the "maybe" stage. Someone is going to have to do the harder work of actual historical linguistics here if this hypothesis is going to be able to move past that stage.