An Introduction to Laryngeal Theory
Laryngeal Theory is a theory about Proto-Indo-European that posits the existence of certain sounds in the proto-language that (at the time they were first posited, any rate) were only known by their effects on other sounds that could be seen in attested languages - instead of knowing about the sounds by way of their own evolution, they are known by way of the evolution of other sounds. Since the original proposition of the theory, the Hittite language has been discovered and deciphered, and Hittite is the only attested language in which PIE laryngeals have been preserved at all - and very partially, at that. What follows is an attempt to explain Laryngeal Theory to a reader who has little or no knowledge of Indo-European Linguistics. At some points I will make us of the International Phonetic Alphabet (IPA), and the wikipedia article I linked to could serve as a useful reference.
What is Proto-Indo-European?
The Proto-Indo-European language, commonly abbreviated PIE, is a hypothetical language that is the ancestor of the Indo-European languages - a massive language family including some of the most widely spoken languages today - English, Spanish, Hindi, Bengali, Russian, Persian, French, among many others - as well as some of the more influential languages of cultures past that are no longer spoken, but exist in written corpuses of various sizes - Latin, Ancient Greek, Old Church Slavonic, Sanskrit, Gothic, Avestan (the liturgical language of Zoroastrians, related to Old Persian), Hittite, Old Irish, as well as a few others.
The Comparative Method
Proto-Indo-European is an entirely hypothetical language, proposed to have existed (and now universally believed to have existed) because of the overwhelming similarities between the various languages into which it developed over time and in different places. Linguists have reconstructed much of the language using the comparative method, that is to say, comparing corresponding items to see what the correspondences are. As a quick and fairly straightforward example, let's look at the word for 'father' in English, Latin, Ancient Greek and Sanskrit (A quick note on that: Except for English, I have only chosen ancient languages as examples. This is because ancient languages generally serve as better examples for reconstruction because - and perhaps this is obvious - they are much closer to what we are trying to reconstruct. English is on the chart instead of Gothic - the oldest attested language of the Germanic branch - because in this instance it happens to be very conservative and illustrate the point perfectly.)
Before getting down to business, I would like to point out the vowels of the first syllable in each word (English 'a', Latin 'a', Greek 'a' Sanskrit 'i'). I'll get back to that in a bit. Now, let's ignore that and everything else except for the first letter of each word (English 'f', Latin 'p', Greek 'p' and Sanskrit 'p'). It doesn't take much thought to see that English is the odd one out in this bunch. It probably doesn't come as a surprise that the consensus on how to reconstruct this word is *pH2-tēr. "What the hell is that?" you are no doubt shouting. Ok, maybe that does come as a surprise. It does look pretty ridiculous. For a long time it was written *pə-tēr, which probably looks like a slightly friendlier beast. For right now, though, just ignore everything after the 'p' and know that the asterisk at the beginning of a word means that the word has no actual attestation anywhere. Because PIE is a language that has no written attestation, all PIE words are written with an asterisk at the beginning.
From the above, one might be tempted to deduce that the method was simply to see which sound is the most common result and assume that it was the original sound. That would be a mistake. After all, Germanic languages could simply be more conservative than all the rest. Perhaps the group that became Germanic separated off before a sound change that shifted *f > *p. Why linguists favor *p as the original sound in PIE has to do with more complicated issues in Phonetics than I really want to get into here - about how sound changes happen, and about what kinds of changes are more common, and also about the bigger picture of how PIE changed to become Germanic. To rely simply on taking the most frequent result and assuming it was the original might lead one to the right answer with "father". It doesn't, however, work in other examples. Take Eng. "come", Latin "ven-iō", Greek "βάινω (bain-ō)" and Sanskrit "gam-", which, if you can even manage to trust me that they are related, come from a root "*gʷṃ-" - that is to say, the original consonant (gʷ) is preserved in none of the attested languages as such.
Now, back to the word "father" for one last point. I said to take note of the vowel of the first syllable in each word. In English, Latin and Greek they are <a> (<α> in Greek if we want to be nitpicky about scripts), and in Sanskrit it is <i>. (I won't be nitpicky about Sanskrit because, on the one hand, Sanskrit has a standardized system for transcription into the Roman alphabet, and on the other hand, my computer doesn't support a single one of the many, many other scripts that are used or have been used to write Sanskrit. Getting back to business...) The lack of correspondence in the above vowels caused problems for some of the linguists who began doing work on reconstructing the PIE vowel system.
At first, most linguists (they probably still called themselves 'philologists' at this point - it was a different time!) simply didn't touch the PIE vowel system. Vowels seemed too shifty, too soft, too inscrutable. It shouldn't come as a shock to anyone that most of these guys were German. They liked their consonants. You could rely on consonants. Vowels must have seemed so, I dunno, so Mediterranean. Some people were convinced that vowels didn't operate or change in any kind of regular or systematic way. Finally, people got over it and decided to work on reconstructing vowels. When they did, they ended up coming to the conclusion that Latin <a> should corresponded to Greek <α> and Sanskrit <a>. And most of the time it does. Sometimes, though, you get a situation like 'father', where Sanskrit has <i>. There were still enough people around who thought that vowels were entirely unreliable, and so just shrugged it off, but someone had the bright idea to suggest that perhaps we're really dealing with two separate sounds in PIE: one (PIE *a) gives 'a' in Latin, Greek and Sanskrit, and another, that evolved into 'a' in Latin and Greek (and as you saw, sometimes English) and 'i' in Sanskrit. Lacking any real information about what this sound was, people decided to use the character 'ə' (schwa) to describe it - sometimes referred to as 'schwa indogermanicum', but rarely nowadays, and I think we are just as well for that. Don't, however, make the assumption that by writing *ə, that people were suggesting that the sound had the phonetic value [ə], like the 'uh' sound in English. They chose the symbol as a variable to stand for [sound-about-which-we-only-know-that-it-
And now, if you've managed to come along with me this far, please store all that information away for just a little bit, while I introduce yet another topic. "Ablaut" is a German word that can be very literally translated as "Down sound" or "Away sound". It was conceived as a term that would contrast with the term "Umlaut" (a term that is used specifically within the realm of Germanic languages) which means "Around sound". None of this, as far as I have been able to surmise, is remotely helpful in understanding what these terms mean.
Both Ablaut and Umlaut are examples of what in Linguistics is called 'apophony' or 'vowel gradation' - an alternation of vowels within the stem of a word, which communicates grammatical information. You may not think about this all the time, but you actually use it all the time in English. A word like 'swim' has a past tense form 'swam' and a past participle 'swum'. That is actually a preservation of the Indo-European Ablaut system (albeit with some phonetic changes), something which most Indo-European languages have done away with entirely. I can't really blame them, though; I don't think there is an English speaker alive who hasn't messed up the vowel of some verb at some point, or said something liked 'swimmed' instead of 'swam' in an almost admirable attempt at regularity. But English (and Germanic languages generally) persist in their idiosyncracies while other Indo-European branches have seen fit to trade in the system for things newer, cleaner and more efficient.
A quick description of how Ablaut functioned in PIE is as follows. A word root in PIE follows this formula without fail: one or two consonants, followed by a vowel, followed by one or two consonants. One or both of the consonants on either end can be a "sonorant", which means it is a consonant that can behave like a vowel if its surrounded by consonants (or a vowel that behaves like a consonant if it is next to a vowel - it's just semantics. Still with me?). Sometimes people write this as C(C)V(C)C, which if you can parse it, is actually very useful shorthand. The vowel in the root can either be 'e', 'o', or nothing, from which fact we say that a root can have an 'e-grade', an 'o-grade' or a 'zero grade', depending on which vowel or lack thereof is operational. Onto this root can be fixed suffixes and prefixes that determine the function of the word. Let's see a concrete example.
The root *likʷ- means 'leave, abandon'. It is actually the ancestor of the word 'leave' in English. It looks a little mangled on its way, but trust me. At this point, the astute reader might think he or she has detected an inconsistency. "Stephen!" he or she exclaims, "you said that the vowel in the root could only be 'e' or 'o' or zero. How, then, does '*likʷ-' have an 'i' in the root?" Well, at the risk of blowing your mind, 'i' actually counts as one of the consonants - one of those 'sonorants' I mentioned. When 'e' or 'o' is present in the root, the 'i' is pronounced as a semi-consonant or 'glide' similar to English 'y' in 'boy'. In Greek, *likʷ- gives us the verb λείπω (leip-ō), "I leave" (trust me, that 'p' comes from *kʷ). That form is built of the e-grade of the root. The simple past tense is ἔλιπον (e-lip-on), "I left", built off the zero grade. The perfect tense is λέλοιπα (le-loip-a), "I have left", built off the o-grade.
Just to reinforce the pattern, let's look at the PIE root *drḱ-, which means 'see, watch'. Incidentally, it's where we get the word "dragon" - apparently for whomever decided to call it that, the point he or she wanted to make was that dragons watch you, but how that was more important than fire-breathing is utterly beyond me. Anyways, like *likʷ above, this word is well preserved in Greek (also in Sanskrit, not sure about other branches). The present tense is δέρκομαι (derk-omai) "I see". Once again, the present tense form is built off the e-grade root. The simple past is ἔδρακον (e-drak-on) "I saw". Here it helps to know that 'r' could also be a vowel in PIE (and continued to be so in Sanskrit). In Greek, however, vocalic 'r' became 'ra', so when we see ἔδρακον we can assume it came from PIE *e-dṛk-om (putting a little circle beneath something is a standard note that means 'this thing is a vowel, trust me'). And finally, without much surprise, we have the perfect form δέδορκα (de-dork-a), "I have seen", built off the o-grade.
So the point of going through all those examples from the roots '*likʷ-' and 'drḱ-' was this: not only do variant forms of a single root exist, but they exist in recognizable patterns. There are certain semantic and morphological spaces where one can reasonably expect to find an e-grade and not an o-grade, or a zero grade and not an e- or o-grade, etc. It isn't the only way that grammatical information is communicated (since it is reinforced by prefixes and suffixes that communicate much more specific information - which is why it was easy to scrap the system in the long run), but its variations are regular.
Laryngeal Theory arose from the inability of philologists to explain the ways in which certain roots seemed to deviate from the Ablaut pattern I've just explained. One of the more common roots that falls into this group is *dō-, which means "give", and ought to be fairly recognizeable to any English speaker in words its gotten from Latin sources, like 'donate' or 'dowry', even though the root that gives us 'give' has taken over in the Germanic branch. Now, let's look at the present tense form (where we've established that we expect to see the e-grade) and the past participle form (where, I'll tell you, we expect to find the zero grade); in Latin, i'll use a noun built of the root instead of the present tense form because it better illustrates the point. In all examples the hyphens are used to separate out the root from suffixes or prefixes.
|expected e-grade||dō-num ('gift')||δίδωμι (di-dō-mi)||da-dā-mi|
|expected zero-grade||da-tus||δοτός (do-tos)||di-taḥ|
What you see is a present tense forms that supports the root *dō- (in Sanskrit, practically everything becomes 'a' - both 'e' and 'o' do, which makes it not very helpful to look for ablaut patterns since it can't distinguish e- and o-grades). The past participle, however, is all over the board: da- in Latin, do- in Greek, and di- in Sanskrit. That's essentially useless for determining what the original zero-grade root is. Let's look at two more of these roots - *stā-, "stand, put", and *dhē-, "place, make" that similarly seem to be lacking a final consonant. Let's do more or less the same thing as before, and look at the same forms in Latin, Greek and Sanskrit. Ignore that the root begins sth- in Sanskrit. That is a Sanskrit phenomenon that doesn't need to be dealt with here.
|expected e-grade of *stā-||stā-re ('to stand')||ἱστᾱμι (hi-stā-mi)||ti-ṣṭhā-mi|
|expectged zero-grade of *stā-||sta-tus||στατός (sta-tos)||sthi-taḥ|
|expected e-grade of *dhē-||fē-ci ("I made/did")||τίθημι (ti-thē-mi)||da-dhā-mi|
|expected zero-grade of *dhē-||fa-ctus||θετός (the-tos)||dhi-taḥ|
The way the roots vary here is very similar to the way the roots varied with *dō-, except that with the root of the Greek past participle has an 'e' vowel with '*dhē-', making the roots fa- in Latin, the- in Greek, and dhi- in Sanskrit; and an 'a' vowel with 'stā', making the roots sta- in Latin, sta- in Greek, and sthi- in Sanskrit. Wait a second. We've seen that before. That's the same set of results from our old friend schwa indogermanicum. So the zero grade of *stā- can be reconstructed as *stə-. But this doesn't really resolve anything. Nowhere else is there evidence to suggest that 'ā' should alternate with 'ə'. It still stands in contrast to the ablaut pattern generally observed. So at this point, a very clever man by the name of Ferdinand de Saussure started the way of thinking about this problem that resulted in what we know as Laryngeal Theory.
Basically, Saussure's idea was this. If *stə- is a zero grade form, then we would expect the e-grade form to be *steə- (remembering that 'ə' is still nothing more than a variable, so we could assume it takes the properties of a resonant - becoming consonant in the presence of a vowel). We could then assume that that was the original form, but that at some point still before PIE split up, '*eə' became '*ā', that is to say it merges with an 'e' in such a way that it changes the quality of the vowel to an 'a' and lengthens it.
We could do the same with both *dhē- and *dō- too, really. Let's reconstruct the zero grade of '*dhē-' as *dhə1-, where 'ə1' is defined as a sound that becomes 'a' in Latin, 'e' in Greek and 'i' in Sanskrit. The e-grade would then be *dheə1-, and the effect of 'ə1' would be to merge with the 'e' vowel in such a way that the vowel quality does not change, but just lengthens. I'll give away a bit of the story and say that linguists have rechristened the original schwa indogermanicum (the one from 'father') 'ə2'. I guess they felt that the 'ə' of '*dhē-' deserved to be 'ə1' because it is the neutral schwa, that has no qualitative change on its surrounding vowels. Then just to finish the story, you have to suppose a third schwa, 'ə3', that would have the effect of changing *deə3- to *dō-.
Now that's a terrible lot of information, so let's review with a little chart. But from now on, i'm going to change the symbol we use for the variable. H is a better variable, since 'ə' already has a real phonetic value, and we've basically moved away from any kind of implication that [ə] is the actual phonetic value when we're supposing that there are three different sounds at play.
|Laryngeal||Function||Result of 'e' + laryngeal||Result of laryngeal in Greek|
So now, if nothing else, you might be able to see how we arrived at '*pH2-tēr' way back when. Laryngeal Theory explains a lot, but it took quite a while for it to gain general acceptance among Indo-Europeanists. The main thing that riled everyone up was that this was a pretty drastic methodological shift. At the beginning of this I made a big deal about the Comparative Method. That's pretty much the basis of historical linguistics. Making a statement about something unattested without having at least two attested forms to compare, so the general wisdom went, was little more than a stab in the dark. Laryngeal Theory required people to rely on internal reconstruction , which (while logical) doesn't have anything outside of the language to compare itself to and check against. That is, until the discovery of Hittite (and the rest of the Anatolian language branch) in the early 20th century.
The Anatolian branch of languages is an entirely extinct branch of PIE that gave rise to several languages spoken in and around what is now Turkey, attested as far back as 1600 B.C. The whole branch has so many archaic features, that most linguists assume that it split from PIE at an earlier stage than the other Indo-European languages groups. Among these archaic features is the partial preservation of Laryngeals as consonants. Specifically, H2 and H3 are preserved, written as one letter (transliterated as 'ḫ'), and H2 doesn't show up all the times we expect it should; prompting some linguists to posit the existence of a fourth laryngeal simply to explain the discrepancy (a bit extreme, if you ask me).
For example, in Latin we have the word ante 'before, facing' (from an earlier *anti) and in Greek we have ἀντί (anti) 'against, facing'
(both of these are well known in many words borrowed into English: antebellum, antidote, etc.) and in Sanskrit (just to round everything out nicely) anti 'in the presence of, near'. All of these appear to be related to a noun in Hittite ḫants 'face', of which 'ḫanti' would be a normal derivative meaning 'in/at the face (of), facing'. Seizing once again on an opportunity to reëvaluate a root in terms of ablaut, linguists now reconstruct *anti as *H2ent-i.
To drive the point home, in Latin we have the word ovis 'sheep', in Greek ὄϊς (oïs) 'sheep', Sanskrit avi- 'sheep' (all of these cognate to the English word 'ewe'). In Luwian (another language of the Anatolian branch) there exists the word ḫawi-, also meaning sheep. Now, if I let you know that the change *o > a is normal in both Sanskrit and Anatolian, it would make sense to reconstruct *H3ewi- as the PIE root from which all these have sprung.
Among the many things I've asked the reader to swallow without much explanation is the term 'laryngeal' istelf. 'Laryngeal' refers to a sound articulated at the larynx, an organ in the throat. The term 'laryngeal' came to be associated with these sounds after Hittite showed that these sounds were preserved as something that was probably a laryngeal sound. Furthermore, it seemed like laryngeal sounds were good candidates for the type of sounds that would have the effects on vowels that were proposed (although both pharyngeal and glottal sounds - other points of articulation in the throat - were and are discussed as well). They are breathy, and easily lost (like 'h' in English), although the particular ways in which they were differentiated is entirely a matter of debate.
I'll bring up one suggestion as to the nature of H3 for no other reason than that I find it compelling. It is highly likely, in my opinion, that H3 would have had lip-rounding - that is to say, that when the sound was being articulated in the throat, it would have been accompanied by a rounding of the lips. 'H3' is the laryngeal that has the effect of turning 'e' into 'o', which means essentially that it turns it from a vowel that doesn't have lip-rounding to a vowel that does. Given that, it would be very surprising if 'H3' didn't have some lip-rounding. To see a parallel example, let's look at the sound [hʷ] in English, written 'wh' and in most varieites of English nowadays pronounced as [w] or [h], depending on its phonetic environment. Specifically let's look at the word 'who' in English, pronounced [hu:] in all varities of English with which I'm familiar. In Old English, the word was spelled hwa, and the consensus is that it was pronounced [hʷa:]. The first sound shift that happened in this word is the one we can see preserved in the spelling 'who'. The vowel shifted from 'a' to 'o' (another shift that required the addition of lip-rounding). The consensus is that this word would have been pronounced in Middle and Early Modern English has [hʷo:] (rhyming with its Latin relative 'quo'). After this, the lip-rounding became redundant, because it was articulated both in the consonant [hʷ] and the vowel [o:], and so the consonant was simplified to [h]. The vowel changed as well (to [u:]), but preserved the lip-rounding (we can assume that the shift from [hʷ] to [h] happened before the shift from [hʷ] to [w] in words like 'what' or 'white', which would explain why we pronounce [hu:] and not [wu:]) . Given all this, it wouldn't be out of bounds to suggest that H3 was [hʷ] , except that it could just as easily be [χʷ] or [ħʷ] or almost any other sound with lip-rounding that is articulated in the throat or in the back of the mouth.
So that's basically it. Laryngeal Theory is fascinating to me because it draws on an ability to see sounds through their relationships to other sounds in their phonetic environments, and also because it requires a shift away from (or, depending on your point-of-view, an addition to) traditional methodology in Historical Linguistics. It is also fascinating because ultimately it is an unaswerable question. What were these sounds? We can't really know. But we can learn more about what we do know because we know one more way (three more, i guess) in which sounds can effect each other.