Like many American kids, a staple of the winter holiday season was watching How The Grinch Stole Christmas (I mean the 1966 cartoon, of course). One of the best parts is the song You’re a Mean One, Mr. Grinch, with its many memorable descriptions of how awful the Grinch really is. Among these are the lines:
The three words that best describe you are as follows, and I quote
"Stink, stank, stunk!"
It's the kind of relationship between stink, stank, and stunk that I want to dig into today. The difference between the present tense stink, the past tense stank, and the past participle stunk are signalled only through a change in the vowels. These kinds of vowel alternations are known as ablaut, a German term going back to the formidable nineteenth-century philologist and fairy-tale collector Jacob Grimm. As Dr. Seuss shows us, ablaut is still very much a part of English grammar, especially in the so-called strong verbs. But ablaut has a very long history, and goes back all the way to Proto-Indo-European (and probably even earlier). It’s the last major thing we need to look at to round out our multi-post survey of the sounds of this proto-language.
If we work back through well-established sound changes, the ‘short i’ [ɪ] of stink goes back to the Indo-European vowel *e. We can reconstruct the Proto-Indo-European form as something like *stengʷ-. Why? First off, Grimm’s law doesn’t affect stop consonants after s, so st- generally just comes from *st-, as far back as we can reconstruct. The final consonant, k, looks like it should go back to *g by Grimm’s law, but we know by comparing with forms like Gothic stinqan that the k actually goes back to a labiovelar stop: Proto-Germanic *kʷ (spelled q in Gothic). By Grimm’s law, this comes from Indo-European *gʷ, giving us *stengʷ-. This is related to the Latin verb stinguō, which we’ve borrowed in words like extinguish (from ex + stingu-ere).
The a of stank goes back (in this case) to Proto-Indo-European *o, so now we have two forms of the same word: *stengʷ- and *stongʷ-. There’s still one thing missing from these reconstructions: the accent. Proto-Indo-European words tended to have one syllable that was more ‘prominent’ in some way than the others (higher pitch, louder volume, and other phonetic emphases – it’s often supposed that pitch was the main feature of the Indo-European accent, but the actual situation was possibly more complicated). The present tense stink comes from forms like *sténgʷ-eti ‘pushes’ (yes, ‘pushes’: there’s been a good deal of semantic change with this verb!), while the past tense comes from something like *(ste)-stóngʷ-e ‘has pushed’. In both forms, the vowel of the verbal root itself, whether *e or *o, is accented, which we traditionally indicate with the acute mark.
The third form, stunk, is a touch more complicated. Here the vowel + consonant sequence un actually goes back just to a single sound, a ‘syllabic nasal’ that is often written *n̥. Similar sounds are found in modern English (though they don’t continue the Indo-European syllabic nasals directly). The way I say the name Elton (as in Elton John), there is no full vowel, and the core of the syllable consists of just the n on its own. In Indo-European-style notation, I might write Elton as eltn̥. The ring underneath the n indicates that it’s being used syllabically (the IPA uses a vertical bar underneath for this instead, [n̩]). For stunk, we can reconstruct an earlier form along the lines of *stn̥gʷ-nó-. Notice that the *n̥ is unaccented: the accent instead falls on the suffix *nó (which is ultimately the source of the -en suffix in words like written). The position of the accent will be important in a minute.
So, taking stock, we have three distinct forms of the same word. The most basic is *sténgʷ-, with an e-vowel; very creatively, we call this the e-grade of the word, and this is what is usually cited in dictionary forms and etymologies. We also have a variant with *o, *stóngʷ-, which predictably receives the label o-grade. Finally, we have what’s called the zero or reduced grade: there’s no full vowel at all, and the consonant n has to take over instead: *stn̥gʷ-.
In Proto-Indo-European as reconstructed, exactly when you find which ablaut grade is partly a matter of grammar. For instance, one major class of basic verbs tends to have e-grades in the singular forms of the present tense, and zero-grades in the plural. We’ve even seen this pattern already, with is in post 3. This very basic verb had, in Proto-Indo-European, a singular form *h₁és-ti ‘is’ (e-grade) and a plural form *h₁s-énti ‘are’ (with zero-grade of the first syllable). Certain grammatical forms, such as the ‘perfect’ formation (*(ste)-stóngʷ-e ‘has pushed’), tend to have o-grades.
From a historical perspective, ablaut is interesting for a number of reasons. One is that it might shed some light on ‘pre-Proto-Indo-European’, a stage of the language a bit earlier even than the standard Proto-Indo-European that we reconstruct through the Comparative Method (which I discussed in post 4, fart). We do this through an approach called Internal Reconstruction, which basically means looking at hints of patterns in the proto-language, and positing an even earlier stage where those patterns applied much more systematically.
In the case of ablaut, the ‘hints’ are the partial correlation of ablaut with the position of the accent. If we look at *stn̥gʷ-nó- or *h₁s-énti, we can see that the unaccented syllables are in the zero-grade, while the accented ones have a full vowel, either *e or *o. Using Internal Reconstruction, we can speculate that zero-grades also used to have fuller vowels, but that these were reduced through lack of stress. So *stn̥gʷ-nó- might have come from an earlier form like **stengʷ-nó- (using the double asterisk to mark this as an Internally Reconstructed form). At some point, the unstressed *e was lost, and the *n was left as the only sonorous, vowel-like element in the first syllable. This kind of development would be perfectly normal. The name Elton once had a full vowel in its second syllable (actually this is etymologically the same as the word town), but in my speech, through lack of stress, it’s been worn down just to a syllabic n.
This kind of hypothesis, of zero-grades coming from highly reduced unstressed syllables, would also seem to fit nicely with the forms of ‘to be’ we just looked at: singular *h₁és-ti versus plural *h₁s-énti. Perhaps these come from something like **h₁éseti and **h₁esénti, with the only difference being the place of the stress. Then there was a period of vowel loss (technically known as syncope), where the unaccented e’s were lost. So **h₁éseti became *h₁ésti, and **h₁esénti became *h₁sénti.
This kind of Internal Reconstruction is broadly plausible, but I should emphasize again that if there was ever this kind of really neat correlation between accent and ablaut, it belonged to a stage of pre-Proto-Indo-European. For the language as we reconstruct it using the Comparative Method, things are different. For one thing, we find plenty of examples of unaccented full vowels. I’ve already mentioned that stink should go back to something like *sténgʷeti, where the second syllable is an unaccented *e, and there are plenty of other examples. Another kind of issue is that we also find zero-grade syllables that are accented. These include words like ‘wolf’, which goes back to Proto-Indo-European *wĺ̥kʷos. The ring under the l marks it as syllabic (this also has a close parallel in some varieties of English, including in my pronunciation of the word bottle). But in this particular word, *wĺ̥kʷos, this syllabic l is accented.
So if our Internal Reconstruction hypothesis is correct, there’s been a good deal of change between that early stage of pre-Proto-Indo-European, and Proto-Indo-European proper. And there are other questions too. One is where the difference between e- and o-grades come from. We’ve seen that one and the same root can sometimes be in the e-grade (like *stengʷ-) and sometimes in the o-grade (*stongʷ-). Why? This question has been much debated, but so far we don’t have a really compelling, generally accepted answer (though there are some good ideas out there). Another issue is why accents move around. Even if we Internally Reconstruct a singular **h₁éseti and a plural **h₁esénti, why is the accent on the first syllable in one form, and the second in the other? Again, there are theories about this, but not a general scholarly consensus.
All in all, the situation is summed up nicely by Paul Kiparsky, who wrote in an important paper that: ‘There are deep connections between accent and ablaut, but they are not transparent, even in the proto-language’ (p. 7).
In any case, Internal Reconstruction aside, by the time we get to ‘standard’ Proto-Indo-European, ablaut and accent shift had become important grammatical tools. They were used in inflection (in the different forms of verbs of nouns) and in derivation (making one word from another). We’ll see plenty more examples of accent and ablaut issues as we continue on with the later history of English (and the same would be true if we were looking at the history of any other Indo-European language, be it Breton or Kurdish or Albanian).
To wrap things up for today, I want to spend a moment looking at the Proto-Indo-European vowels, in light of ablaut. We have two ‘basic’ or ‘full’ vowels, *e and *o, which can interchange with one another through ablaut. Both of these could also be long, *ē and *ō, though these long forms were less common in Proto-Indo-European (and may not have existed at all at an earlier stage of ‘pre-Proto-Indo-European’). In addition, there was probably also a vowel *a (and just possibly a long *ā), though some researchers have denied that a ‘basic *a’ ever existed. Either way, the sequence *h₂e was ‘coloured’ to *h₂a, which meant there were plenty of *a vowels floating around in actual speech.
In zero-grade positions, we find a wider set of syllabic sounds, all of which interchange with consonants. We’ve already seen *n (consonant) and *n̥ (syllabic), as well as *l and *l̥. Beside these are *m and *m̥, *r and *r̥. Furthermore, back in post 5, on yoke, I mentioned that there’s an interchange in Indo-European between the consonants *j and *w and their corresponding vowel forms *i and *u, which works basically the same way. So we have roots like *wéd- ‘water, wet’ (note Grimm’s law!), which has a derivative *ud-rós (whence otter, an ‘aquatic creature’). Finally, we have the schwa, a vowel inserted especially next to laryngeal consonants in zero-grade positions to break up awkward consonant clusters. The form I’ve been giving as *h₁s-énti ‘are’ probably was actually something like *h₁əsénti. (Some people instead write *h̥₁sénti, as if the laryngeal itself were syllabic, but schwa-insertion seems more likely.)
An otter eating from my hand.
These dozen or so syllabic sounds would go on to change quite a great deal in most of the various Indo-European languages, including Germanic and then English. We’ve already seen that none of the vowels in stink, stank, stunk is exactly like its Indo-European source. But the system of vowels and consonants that we’ve gone over in the past few posts are the starting point for many of the linguistic developments we’ll be looking at in the rest of this series.
The basics of the vowel system can be found in any of the textbooks I’ve been mentioning in the past few posts. You’ll find some differences in perspective between them. Beekes & de Vaan (2011, 141), for instance, emphasize the minimality of the Indo-European vowel system, by focus on underlyingly phonemic vowels as the only ‘real’ vowels, and furthermore deciding that alternating sounds (like *u~*w) are underlyingly consonants rather than vowels. In this way, they claim that ‘Proto-Indo-European had only two vowels’. This is ‘true from a certain point of view’ (to quote Obi-Wan), but also leads, in my view, to a certain distortion. The ‘underlying’ structure of vowels does matter, but so does the actual realization, and the latter may actually be the more important in many respects.
I didn’t talk too much about the issue of whether the vowel *a actually exists. Its existence is denied especially by the ‘Leiden School’. Beekes & de Vaan, in the same textbook just mentioned (141-143) make the basic case against PIE *a, and a fuller argument is developed in a 1989 article by Alexander Lubtosky. Don Ringe gives a handy short list of words possibly showing PIE *a in his book From Proto-Indo-European to Proto-Germanic (pages 12-13 in the second edition from 2017, or pages 10-11 in the first edition, from 2006). Roland Pooth has a more extensive discussion (in German) on *a as characteristic of certain semantic classes of words in a 2015 paper (in German), which includes a useful long appendix of roots and words with *a, sorted by semantic type (pages 112-117).
The workings of ablaut, and its possible relationship to the accent, are very complicated issues, and discussions of them can get extremely technical. One popular framework involves talking about different types of accent-ablaut patterns (given rather verbiose names like proterokinetic and acrostatic; these kinds of terms are now standard parlance in this area of research). Much of this apparatus was really developed in the 1960s and 70s by people like Johanna Narten and Jochem Schindler (among others), building on work from earlier in the 20th century. Fortson gives a pretty accessible overview of all this in his textbook (especially in chapter 6).
There are various Internal Reconstruction approaches to how ablaut might have originated in ‘pre-PIE’. There’s a relatively readable discussion in Winfred Lehman’s Theoretical Bases of Indo-European Linguistics (chapter 6 focuses on pre-PIE phonology). For a more recent approach, with some very different perspectives on things like the origin of the o-grade, I’d again recommend the same paper by Martin Kümmel that I mentioned in the last post (the paper has two sections, one on vowels and the other on consonants).
A good deal of very recent research has focused less on the early origins of ablaut, and more on what the synchronic rules of Proto-Indo-European proper were. (‘Synchronic’ refers to rules operating in the language at a given point in time, as opposed to ‘diachronic’, which are processes viewed through time in a historical perspective.) I’ve already quoted from one influential article taking this approach, by Paul Kiparsky. There has also been some criticism of Kiparsky's analysis, including this response by Roland Pooth (who favours a more strongly morphological analysis of ablaut).
There are also quite a few relevant papers in the essay collection Indo-European Accent and Ablaut, edited in 2013 by Götz Keydana, Paul Widmer, and Thomas Olander, representing several different approaches and perspectives. There’s a relatively accessible (though these kinds of discussions are inevitably rather technical) overview of this kind of approach to ablaut in the chapter ‘The Morphology of Proto-Indo-European’ by Jesse Lundquist and Anthony Yates, especially section 3 (pages 30-41 of the online PDF, or 2121-2137 in the printed handbook).
(The handbook that the Lunquist and Yates chapter appears in is a very useful thing, if you have access to it. The third volume also has a chapter on ‘The Phonology of Proto-Indo-European’ by Andrew Miles Byrd, which expands on quite a few of the issues discussed over the last several posts. Unfortunately the three-volume work is extremely expensive.)