Reconstructing Unwritten Languages
The Germanic languages seemed to undergo a unique consonant shift. But did the same thing happen in Armenia?
The Father Tongue
How did our linguistic ancestors speak? What would you hear if you took a time machine to the Jutland in 1000 B.C.? When searching for words from this era, you’ll find a little asterisk in front of each one of them. For example, here are the old pronouns from that epoch:
No, these folks didn’t litter their speech with one-letter swear words. Rather, they didn’t use letters at all. This language, Proto-Germanic, was never written down. While places like Germany, the Netherlands, and Denmark boast powerful economies today, they weren’t up to speed on the whole “represent sounds with symbols” thing back then.
If it wasn’t written down and Stephen Hawking disproved time travel, how do we know what this language sounded like? In short, we don’t. That’s what the asterisk is for. It signifies an un-attested word that derives from linguistic models rather than historical records. You can think of that asterisk as a “probably” or a “maybe.” How, you might ask, do linguistics come across these probablies and maybes? Let’s run through an example.
Imagine, in an alternate history, you’re a colonist from Sub-Saharan Africa, Meso-American, Australia, or your favorite non-European spot. You find yourself among three primitive tribes in Northern Europe, and (according to one tribe) the three groups are called the “English,” the “Dutch,” and the “Germans.” You find a member from each tribe, show them a tooth, and ask them what they call it. You’d hear these three words:
Tooth
Tand
Zahn
If my Dutch and German-speaking readers can forgive me, I’ll write these three words in a manner that best indicates their pronunciation to us English-only folks:
Tooth
Tahnt
Tsahn
You Meso-Afro-Australio travelers can’t find any written documents. No stone tablets or papyrus scrolls to help you understand their history. However, you hear enough similarities to determine that each language stems from a single tongue spoken thousands of years in the past. What would you think the word for “tooth” was in that old language?
Let’s start with that “n,” which you hear in the mainland languages but not among the English. It’s possible, of course, that the other languages added the “n” later on. Adding consonants isn’t unheard of. The Latin nummerus became the English number, and you can still hear the old b-less version in numerical and numerous. However, the added “n” seems unlikely in this case. It doesn’t link together sounds in the way that “b” in number does, and the quick trip to the North can teach you that the Swedish word sounds a lot like the Dutch one. It’s beginning to seem a bit implausible that three separate languages would somehow add an extra “n,” so you conclude that the English version lost this sound.
What about the start of the word? You heard a “t” in English, Dutch, and Swedish, but a “ts” in German. Again, we could either conclude that the Germans added an “s” or that everyone else lost theirs. After running through some more words, you’ll discover that the Germans display this pattern in a lot of different words. English sleep and Dutch slapen correspond to German schlafen. English make and Dutch maken correspond to German machen. You notice this pattern enough to discover the High German Consonant Shift, and you conclude that the Germans shifted the “t” to a “ts” while everyone else kept the original “t.”
Now, for the last consonant sound. German doesn’t have anything after the “n,” but you know, from your home continent, that people tend to drop the last sound of a word. You hear a “t” in the Dutch version, but you’re not sure how long that “t” has been there. After all, the Dutch (and the Germans, for that matter) never end a word with a voiced consonant. Voiced consonants involve a vibration of the voice box (compare “s” (unvoiced) to “z” (voiced.) Hence, Dutch words never end with a “d.” Even though you hear a “t” today, it could have been a “d” some short time ago. After all, you hear a voiced “d” in the plural: tanden. That still leaves us with a choice between two sounds: “th” and “d.” Which came first? Perhaps, back on your old continent, you’ve noticed that those “th” sounds seem unstable. Languages tend to do away with them in relatively short order. You even find accents among the English tribe that seem to be phasing them out as we speak. Thus, you conclude that a change from th→d makes more sense than a change from d→th.
Our linguistic intuition leaves us with a reconstruction that looks like t__nth. I never spend much time on vowels, so I’ll just assume that the old vowels sounded like the ones Dutch, German, and Swedish. My counterfactual colonial linguist therefore believes that the ancestors of the English, Dutch, and Germans referred to a calcified food chewer as a “tanth.”
The First Stop
Stepping outside the thought experiment, you’ll find that scholars came to the same conclusion, though they spell the “th” with the funky þ. Linguists refer to this process as the comparative method, and it’s the main source of our pre-historical philological knowledge. Luckily, in our universe, we’re not reliant on interviews with Germanic. We don’t need to speculate about a lost “d” in the German “Zahn,” since we can see it in older versions of the language. Furthermore, we don’t need to ask if the English vowel changed since we can see a different vowel in Old English. We can also exploit the fact that the Germanic tongues belong to a large language family with thousands of years of written history. No assembly is required to produce the Sanskrit danta, Latin dentem, or Ancient Greek edontes1.
We can run the same sort of comparative analysis on these words and try to reproduce the OG word for “tooth.” Linguistics have done so, and they concluded that the word started as h₃dónts in Proto-Indo-European, the ancestral language spoken in the Russian steppes around 6,000 years ago. If you’re wondering what h₃ sounded like, you’ll have to keep wondering until someone figures it out.
Scholars have reconstructed more than that just the word for “tooth.” Through these vocabulary comparisons, we have a good idea of which consonant sounds were available in the original Proto-Indo-European language. I’ll focus on stops - the consonant sounds that involve a blockage of airflow followed by a sudden release of air (English stops are “b,” “p,” “d,” “t,” “g,” and “k.”) The academic literature organizes Proto-Indo-European stops into three series. While Linguists can reconstruct six-thousand-year-old languages using advanced computational methodologies, Substack can’t support basic tables, so here’s a screenshot from Google Sheets:
Note that I’m skipping some details here. There were plain, palatal, and rounded versions of the “Back of Mouth” stops, but that doesn’t matter for the forthcoming analysis. The aspirated consonants (indicated with an ʰ), featured a strong, outward burst of air. To illustrate this, keep a hand in front of your mouth and say “pay” and “bay.” You’ll feel a gust for the first, but not the second. That “p” (or, more precisely, “pʰ”) is aspirated. In fact, we never you’ll never hear “bʰ” in modern English, even though we see it in the chart above. If you’re a regular reader of this blog, you might know what happened to “bʰ.” If you don’t, let’s return to our analysis of the word “tooth.”
The original word, the one you would have heard 6,000 years ago in Proto-Indo-European, was h₃dónts. Note that the h₃ has nothing to do with aspiration. In fact, let’s just ignore it for now. It’s scaring me. Compare the dónts in Proto-Indo-European to tanth from Proto-Germanic. At least two consonant shifts must have occurred to create the word tanth. The original “d” downgraded to a “t” and the original “t” smoothed out into a “th.” It’s not remarkable that this change occurred in the Proto-Germanic word for tooth. Consonants change all the time. What’s remarkable is that this occurred among words for everything else.
By the time we get to Proto-Germanic, the three series of stops changed to sounds shown in the chart below:2
Around 2-3 thousand years ago, among those early Germanic people, every stop consonant shifted. Starting with series 3, we see that all the aspirated ones lost their voice. That’s why we don’t say “bʰay.” In series 2, all the voiced consonants lost their voice. That’s why we see a voiced “d” in dónts and an unvoiced “t” in tanth. Series 1 experienced the most drastic change. They converted to fricatives, which involve a narrowing of the airflow, in contrast to the blockage seen among stops. The shift from “t” to “th” explains that the last sound in tanth. Together, these consonant changes are known as Grimm’s Law.
Don’t let the nomenclature confuse you. When I first heard the term Grimm’s Law, I thought it represented a universal law of language, an oral F=ma. That’s not how Grimm intended it. A better name for Grimm’s Law might be something like the Proto-Germanic Consonant Shift. In other words, think of Grimm’s Law as a historical event. It’s something that happened among the Germanic tribes in the first century B.C. and not something we generalize across humanity. It would be highly improbable for a different language from a different culture in a different historical epoch with different phonology to somehow undergo the same consonant shift.
…Improbable, but maybe not impossible.
The Last Stop
You ever noticed that Greek words use “ph” to make the “f” sound? Examples include philosophy, phobia, and phonology. What’s up with that? In ancient times, those “ph” represented an aspirated “pʰ,” just like the one we use in “pay.” Over time, that sound shifted to the “f” that we hear today. Similarly, the ancients pronounced the theta as tʰ, before it degraded into the familiar “th” sound. In both cases, the changes match the ones seen among series 1 stops under Grimm’s Law. Pretty crazy, right?
But this section isn’t about Greek. It’s about another branch of the Indo-European-Language family: Armenian. And tennis.
The Problem
Since I hate when articles make me scroll up and down, I’ll paste the original set of stops in Proto-Indo-European below:
The Proto-Germanic ones:
And finally, here’s what seems to have happened in Proto-Armenian3:
Series 2 and 3 stops experienced the same shifts in both Proto-Armenian and Proto-Germanic. Series one looks a bit different, but Greek shows that those aspirated, unvoiced stops (pʰ/tʰ/kʰ) can turn into fricatives. In fact, linguists suggest that this occurred during Grimm’s Law. The Proto-Indo-European “p” changed to a “pʰ” before finally landing as an “f.” In other words, those aspirated sounds could represent a literal last-stop trip to Fricative Station. Hence, the Armenian Series 1 shift looks like an intermediate version of Grimm’s Law.
Yes, I know, false crabs. That was a great analogy; thank you very much. This, however, looks less like a false crab and more like a false identical twin. Proto-Germanic speakers never spoke with Proto-Armenian ones, so there’s no historical reason for the symmetry. Some suggest that Grimm’s Law occurred due to contact with Semitic traders, but most scholars consider this a crackpot theory. Maybe it is a crackpot theory, but is it less crackpot than the alternative? Two nearly identical changes across vastly different branches of the Indo-European family? You cannot be serious.
Linguistic Hawk-Eye
For most of tennis’s history, all calls were made by umpires who each watched one of the court’s various lines. One would hear a confident “OUT” call as if the line judges had any idea where a 120 MPH volley landed. You still see these umpires in most professional tournaments, but others have eliminated them in favor of an automatic technology called Hawk-Eye. Initially, the US Open allowed players to challenge a limited number of calls. Today, Hawk-Eye calls every shot, and you won’t find any human umpires on the hard court. Hawk-Eye also shows the crowd a neat visual for otherwise difficult-to-call shots. The technology shows a dramatic 3D model of the ball’s descent and landing point. Or, well, it seems to.
In fact, Hawk-Eye doesn’t know where the ball lands. It doesn’t even try to determine that. Instead, Hawk-Eye tracks the trajectory and spin of the ball and predicts where it will land. The computer determines whether that recorded voice will yell “OUT” before the ball touches the court. That may feel wrong, but the technology maintains a margin of error of around 2-4MM, far less than that of a shouty line judge.
What does this have to do with the Armenian Consonant Shift? Remember, no one ever listened to Proto-Indo-European, Proto-Germanic, or Proto-Armenian. No one witnessed Grimm’s Law or the Armenian Consonant Shift. They used the comparative method, the Hawk-Eye of historical linguistics, to determine each of these ancient languages. Tennis players sometimes argue with Hawk-Eye, but we know the system to be more accurate (not to mention less biased) than the human eye. It’s much harder to calculate a margin of error for a given linguist’s comparative analytics. Here, Linguistic Hawk-Eye has determined that two different shots landed in the same spot. Let’s explore two ways that it might missed the call.
Solution 1: The Standard Model of Proto-Indo-European is Wrong
Enter the Glottalic Theory. This suggests a slightly different series of stops in Proto-Indo-European.
Unfortunately, the written word betrays me here. I can’t explain ejective consonants with text, so you’ll need to listen to them in this Wikipedia article. They amount to speakers spitting the sound out, and one can hear them in dozens of languages across the world.
…none of which belong to the Indo-European language family. There’s no relic of these sounds in the Indo-European languages today. Why, then, do linguists hypothesize them? For one, something feels a bit off about the series 2 and 3 stops in the standard model. Did Proto-Indo-European speakers really differentiate aspirated and non-aspirated consonants? Imagine if “bʰay” and “bay” had different meanings. Could you pick those apart in a real-time conversation? More importantly, the Glottalic Theory fixes our “false identical twin” problem. Under this framework, there wasn’t much of a shift at all. The charts below show how pedestrian Grimm’s Law looks under the Glottalic Theory.
Under this framework, Series 3 stops didn’t change at all. Series 2 stops lost their ejective quality, as they seem to have done in all Indo-European languages. The Series 1 shifts remain the same as in the standard model. Now, if we compare Proto-Germanic to Proto-Armenian, the similarities look less remarkable. In both branches:
Series 3 stops did not change
Series 2 stops lost their ejective quality. This is a major change, sure, but remember that, under the Glottalic Theory, every branch of the Indo-European family has lost them.
The Glottalic Theory and Standard Model agree regarding the Series 1 changes.
Thus, the Glottalic Theory produces a world where the symmetries between Proto-Armenian and Proto-Germanic look less remarkable.
Solution 2: The Standard Model of Proto-Armenian is Wrong
Solution 1 seems a bit extreme. Sure, it solves the “false identical twin” issue, but it does by smashing the consensus model of Proto-Indo-European and introducing a new type of consonant that none of its ancestors maintain. Wouldn’t it be easier to accept that we got Proto-Armenian wrong?
Earlier in the article, I presented the Armenian Consonant shifts, which were theorized to occur in the passage between Proto-Indo-European and Proto-Armenian. I should take a step back, though, and just present the standard theory behind Proto-Armenian. You can see that below, with the “change” column blacked out.
Few of today’s Armenian dialects share this set of stops. That’s not unique to the Armenian language, of course. Dutch lost the “g” sound, and most Germanic languages have discarded “th.” Still, it’s worth noting that the symmetry between Grimm’s Law and the Armenian Consonants shift only appears when we compare Proto-Indo-European with Proto-Armenian. We wouldn’t hear any symmetry if we listened to today’s Armenian speakers. When we look at the standard model Proto-Armenian, we’re trusting Hawk-Eye. Linguist Andrew Garrett, the source of all this article’s information about these Proto-Languages, thinks that Hawk-Eye missed the call.
The smoking gun comes our this article’s last piece of academic jargon: Adjarian’s Law. This “law” shows that Armenian vowels became “fronted” after Series 3 stops. What do I mean by “fronted?” Essentially, the vowels move your tongue closer to your teeth and lips. An English example would involve a shift from the “a” in “alms” to the one in “cat.” Vowels move all the time, of course, but Garret can explain why this fronting only occurs in certain Armenian syllables of certain Armenian dialects.
Say “it” and “hit.” Notice a difference in the vowel? Neither do I, but linguists with high-tech equipment and fancy mathematical models do. In the latter case, the vowel is a bit more fronted4. The breathy “h” sound pushes the vowels towards the front of our mouths.
Which brings us back to Armenian. Why would the vowel fronting occur after Series 3 stops? Simple: They used to have a breathy sound to them. Thus, Garrett presents an alternative model for Proto-Armenian.
Under this framework, the Series 3 stops never shifted in Proto-Armenian, breaking its symmetry with Proto-Germanic. We still see parallels with Grimm’s Law among Series 1 and 2 stops, but Garrett’s model moves outside the dangerous universe of false identical twins and back into the comfortable world of false crabs. Garrett, in other words, built his own, personal Linguistic Hawk-Eye, and you can decide if its calls make more sense.
These languages have case systems, so there’s more than one form of each noun. I picked the forms that illustrated my point most clearly.
The Series 3 “Back of Mouth” shift was actually to /x/ rather than /h/, with /x/ being the consonant sound in German auch. The shift from /x/ to /h/ occurred later.
All forthcoming analyses from this paper: https://linguistics.berkeley.edu/garrett/garrett-1998-bls.pdf.
In more technical terms, there’s a [+ART] change among the F2 factor.