Secrets of the Words You Know
Blog 11: The Consonants of Proto Indo European
First, an answer to the two questions in the previous blog:
Question 1: English words borrowed from Greek with the element ‘pod’ include Tripod; Gasteropod (slugs and snails with a ‘belly-foot’); Cephalopod (Squids and Octopus are termed ‘head-foot’, and the ‘pus’ of the eight-footed ‘Octopus’ is from the same root, like the Platypus we saw earlier); Podiatrist (‘foot-healer’); Chiropodist (‘hand and foot specialist’)…..I’m sure there are others!
These words are all modern coinages except for the genuinely ancient word for a three-legged cauldron, stool or other item of furniture, the tripod. (Three legs are of course ideal for stability on uneven floors.) This word has a special claim to fame as a key element in the decipherment of the ‘Linear B’ script used in palaces in Crete and on the Greek mainland over 3,000 years ago. Signs assessed as reading ‘ti-ri-po-de’ in this script, which has one sign per syllable, were found next to a drawing of such a cauldron, thus confirming that the language of Linear B was Greek.
Question 2: It would be un-settling if you found it im-possible to find this a-typical ‘a’! Yes, ‘a’ is the Greek reflex of the Indo-European negative particle *n̥ that shows up both as the ‘un-’ that English has inherited from Germanic and in our borrowing from Latin and French of ‘in-’ and ‘im-’.
Now we turn to the consonants usually assumed to have been in use in PIE. Four features of these consonants need some explanation.
Proto Indo European, like English, uses the presence or absence of voicing to distinguish several consonants (‘t’ voiceless; ‘d’ voiced etc).
Evidence from Sanskrit shows however that Proto Indo European, unlike English, must also have made use of aspiration as an additional distinguishing feature.
The generally-accepted view is that for each of the main groups of stop consonants, there were three realisations in PIE: voiceless (eg *t), voiced (eg *d) and voiced aspirate (eg *dh- as in our borrowing from Sanskrit ‘dharma’, a word hard to translate, but broadly about correctly observing moral or religious practice). For pronunciation of ‘dh’, think a ‘d’ followed by a strong breath. The voiced aspirates have survived only in Indic languages.
These sounds, which we mentioned in Technical Note 1, are by their nature complex, as they combine two articulations, one of the lips and one of the soft palate.
Of course we cannot know exactly how sounds were produced 5,000 or so years ago, but we have reasons to believe that the PIE labio-velars would have been rather different from, say, the initial ‘kw’ sound of ‘quell’, which is a close parallel to the consonant clusters ‘tw-’ or ‘dw-’ in ‘twelve’ and ‘dwell’.
The labio-velars were perhaps closer to the West African sound often transliterated as ‘gb’, for example in the word ‘Igbo’, referring to the people and language of that name rooted in South Eastern Nigeria. To make this sound, your lips and the back of your mouth close and release simultaneously, whereas with ‘quell’ the velar ‘stop’ is released just before the breath reaches the lips and your lips articulate only a ‘w’ sound, and not a ‘b’.
Labio-velars have evolved very differently in different daughter-languages of PIE. For example, in Greek, though they were still written in Linear B, they had completely vanished by merging into other pre-existing consonants before our earliest alphabetical records of Greek. In English-inherited words, on the other hand, some of them are still visible in writing to this day in our inherited words beginning with ‘qu’ and most (but not all) of those beginning with ‘wh’. Some dialects, but not standard English, still indeed pronounce ‘wh’ with the aspiration shown in writing.
Labio-velars are often involved when you see some of the less obvious equivalences between words of common Indo-European heritage.
- How Many Types of ‘Guttural’ Consonants?
It was clear to researchers very early on that a sweeping ‘fronting’ from guttural stops to fricatives (‘s’ or in some cases ‘sh’, and their voiced and voiced-aspirate equivalents) had taken place even before back vowels in an important set of Indo-European language groups, most notably in Indo-Iranian and Balto-Slavic. This is known as the ‘centum/satəm’ division, based on the words for 100 in Latin (‘centum’ had a hard ‘c’ in Classical Latin) and Avestan (a form of Old Persian). The ‘ə’ is a symbol for the ‘neutral’ sound of the first vowel of English ‘alone’, and we will come back to it shortly.
For some other words with gutturals, however, the ‘satəm languages’ did keep *k and its related voiced and voiced aspirate velars as hard sounds before a PIE back vowel. Before a PIE front vowel (‘i’ or ‘e’) they typically palatalized such sounds to affricates such as ‘tsh’, rather than to fricatives.
Some of these conserved velars were clearly related to words in the ‘centum languages’ that showed evidence of ‘labio-velar’ origins. For example, Latin conserves a descendant of the PIE labio-velar *kʷ sound in its ‘question words’, like ‘cui’ in the legal phrase ‘Cui bono? ’ (literally ‘to whom for something good?’, in other words, ‘In whose interest is it?’). In modern ‘satəm languages’ like Hindi or Russian, ‘question words’ typically begin with a ‘k’.
But there are also words where there no evidence of a labio-velar in the ‘centum languages’ but where ‘satəm languages’ still have a hard sound. A case in point, this time with *g rather than *k, is Sanskrit ‘yoga’, related to English ‘yoke’, both deriving from a root with a basic meaning of ‘join’.
It is therefore generally accepted that PIE had three types of broadly ‘guttural’ sounds:
- one articulated rather forward in the mouth (conventionally notated as *k̂, *ĝ and *ĝh), as in the word for 100. We will call this a ‘palatal’.
- one also a simple sound, but articulated further back (notated *k, *g and *gh), as in yoke/yoga. We will call this a ‘velar’.
- one a ‘labio-velar’, notated *kʷ, *gʷ and *gʷh, as in the ‘question words’.
The case for this conclusion has been strengthened by the discovery of three corresponding sets of guttural sounds in Luwian, an extinct Anatolian language.
In the earliest known stages of the satəm languages, the first is already fully palatalized (usually to a sibilant such as ‘s’, ‘sh’ or their voiced equivalents), and the second and third are merged and are palatalized only before PIE *i and *e, and then only to affricates like ‘tsh’.
For the centum languages, like Germanic, Celtic, Latin or Greek, the first two types merge as velars and evolve identically, while the labio-velars have their own separate history.
- Laryngeals’ and their implications
Minor anomalies in many fields of enquiry often turn out to be very significant for understanding the nature and origin of what we perceive.
In the case of PIE, one such anomaly was seen to exist in vowel-gradation. We have looked briefly at the triad of e-grade and o-grade short vowels and zero-grade, but there are other examples involving an alternation between a long *ē, *ā or *ō and – where one would expect a zero grade – a set of different outcomes between daughter languages that seemed best explained by the ‘neutral’ vowel ‘ə’ (mentioned earlier when we used the word ‘satəm’).
In 1878 a Swiss graduate student, Ferdinand de Saussure, proposed that the long vowels reflected an original short vowel followed by a sonorant consonant whose later disappearance would lengthen the preceding vowel. In zero grade, as we saw earlier with ‘n̥’ and other sonorants, this consonant would itself become a vowel sound (in this case ‘ə’) in zero grade. This makes possible a neat match with the basic ablaut system, in which the sonorant consonant – notated as ‘A’ by de Saussure – would correspond exactly to any other sonorant, let’s say ‘n’:
e-grade zero grade
Basic ‘e’ –
Followed by ‘n’ ‘en’ ‘n̥’
Followed by ‘A’ ‘eA’, yielding a long ‘ā’ ‘Ḁ’, yielding ‘ə’
For simplicity, I have left the ‘o-grade’ out of this little table. De Saussure in fact proposed two sonorants (written ‘A’ and ‘O’), which would ‘colour’ an ‘e’ to either ‘a’ or ‘o’, so that ‘eA’ would appear as a long ‘ā’ and ‘eO’ appear as ‘ō’.
De Saussure did not ascribe a phonetic value to ‘A’ and ‘O’ but it was suggested very early on that such a sound was likely to be some fricative in the back of the throat. This came to be called a ‘laryngeal’ (with reference to the larynx). Hence de Saussure’s conjecture came to be known as ‘laryngeal theory’.
Laryngeal theory did not find wide support until the decipherment of Hittite in the early 20th century. It was then found that Hittite (and another Anatolian language, Luwian), conserved in writing a sound which appeared to correspond to some cases where a laryngeal might have been expected. Since then, the theory has gained general acceptance, and though most scholars would agree that the sounds are not well-named as ‘laryngeals’ the name has stuck.
Debate continues over exactly how many laryngeals should be posited for PIE. Most scholars have opted for at least three laryngeals, two, as proposed by de Saussure, with a ‘colour’, one of ‘a’ (nowadays often notated *h₂) and the other of ‘o’ (*h₃); and one without a colour, which would lengthen a preceding ‘e’ but not change it to another vowel sound (*h₁). When ‘vocalised’ between two consonants, all these laryngeals would give the ‘neutral’ vowel ‘ə’, except in Greek, where it seems that the original colour is preserved. In this set of blogs, for simplicity, I will spare you the subscripts and notate any laryngeals as just ‘H’.
Some would go on to argue that all PIE long vowels should be analysed as a short vowel plus a laryngeal; and/or that ‘a’ did not exist in PIE, and could always be derived from *e following an a-coloured laryngeal.
This would, among other things, radically reduce the number of roots apparently starting with an *a, such as the root behind the English word ‘arm’ and its various cousins that we met in the last blog. Hittite shows, indeed, that many PIE roots apparently beginning with *a did start with a laryngeal. However, this is not the case for all.
Because laryngeals have usually vanished, many roots that were in fact of the standard Consonant-Vowel-Consonant shape (Blog 10) appear instead:
- as a vowel followed by a consonant (when the first consonant would have been a laryngeal), as with ‘arm’ or, to give another example, ‘acre’ and the Latin equivalent which underlies the borrowed term ‘agriculture’ [literally ‘field cultivation’]. from a root *Hag- with a suffixed ‘r’)
- as a consonant followed by a long vowel (where the second consonant would have been a laryngeal, whose loss would then lengthen the preceding vowel). A good example is the root *dō- (‘to give’ as in our borrowing ‘donation’ from Latin, which would originally have been *deH-, where in this case the laryngeal would have been ‘o-coloured’.
The disappearance of laryngeals is a further reason why looking at the first two surviving consonants to check relationships can only be a broad-brush approach.
For the purpose of establishing relationships among Indo-European daughter languages, we will very seldom find ourselves having recourse to laryngeals. However, there is one case which we will see quite often.
Some familiar Greek-based words start with a vowel where other Indo-European languages have a consonant, as we saw with ‘asterisk’ versus ‘star’ in Technical Note 1. ‘Astrology’ and ‘astronomy’ are two further borrowings from the Greek word for ‘star’, which is even more clearly to be seen in the star-like flower, the ‘aster’.
As another example, let us take the Greek equivalent of English ‘name’, which is familiar from the word ‘onomato-poeia’ [literally ‘name making’]. The ‘n’ and ‘m’ of the Greek word are as we would expect from the English word and from its Latin equivalent which underlies the word ‘nominal’. The initial ‘o’, like the initial ‘a’ of ‘aster’ etc, is unexpected.
Such examples are generally agreed to be the legacy of an initial laryngeal in the PIE root which has disappeared in nearly all other extant language groups. In the case of ‘star’, for example, the Hittite word ‘hasterza’, confirms this. We will come across several other examples of this Greek initial vowel as we progress.
(For completeness, a similar situation is found in Armenian, but is not pursued here, due to the lack of Armenian borrowings into English.)
The Hittite word for ‘star’ also leads us to another interesting link. We mentioned in Blog 10 that the typical ‘root’ of a set of words is usually of the form consonant-vowel-consonant, and that suffixes are then very often added to make words. The Hittite word for ‘star’ suggests that the root was of the form *Hes-, to which the common suffix *-ter (meaning a person or thing that delivers the action implied by the root) has been added.
If we then search for words in English that might have been inherited from such a root, we can expect the first letter to be a vowel (as we saw with ‘acre’ above) and the next consonant to be a sibilant: the consensus is that ‘ash’, the residue of burning, is indeed derived from *Hes-. [The ’ash tree’ has a quite different origin.]
We will discover in a later blog that in certain circumstances a Proto Indo European *s can become an ‘r’ in Latin. This explains words borrowed from Latin which also link to the concept of burning, which appears to be the meaning of the root *Hes-, such as ‘ardent’ or ‘arson’, or the result of burning, such as ‘arid’. From Greek, we have also the ‘azalea’, which prefers well-drained (and so dry or arid) acidic soil.
So, to the speech community whose language has given rise to English and its many cousin-languages, a ‘star’ was literally ‘a burning thing’, while for us the four letters of ‘star’ now appear just an arbitrary designation. Only by looking across related languages have we been able to see the first letter as the relic of the same root about ‘burning’ that also gave us ‘ash’, and the last three letters as a suffix about a ‘doer’.
The Proto Indo European Sound System
In the light of what has been said above, we will work with the following assumed system of PIE consonants:
Sonorants (other than pure vowels)
Nasals m, n
Liquids l, r
Semi-vowels w/u, y/i
Fricatives
Sibilant s
Laryngeals H
Stop Consonants
Articulation Voiceless Voiced Voiced Aspirate
Labial p [b][1] bh
Dental t d dh
Palatal k̂ ĝ ĝh
Velar k g gh
Labiovelar kʷ gʷ gʷh
Compared to many other languages, this reconstructed system is rich in stop consonants and quite poor in fricatives (even if we allow for 3 laryngeals). There are, for example, five different ways of articulating stop consonants versus just three (labial, dental and guttural) in English, but no fricative sounds like English ‘f’, ‘v’, ‘th’, ‘sh’ or its voiced equivalent, the ‘zh’-sound of ‘seizure’.
In subsequent blogs, we will explore the sound-codes that lie behind the relationships between English words which have either been inherited directly from these sounds or borrowed from other languages that have similarly inherited them ultimately from that speech community over 5,000 years ago.
In Technical Note 2, we explained that ‘sounds have been shown to evolve in a much more regular fashion than do meanings’. You will see many examples of the variety of meanings that are shown by the word-families English has acquired from these two sources, both ultimately from Indo European roots.
As in the comparison of the initial consonants of English and German in Blog 9, we will start with the sonorants. First up, in next week’s Blog, will be Proto Indo European *m.
But first, there is one cross-cutting question to consider: how do I know if an English word is inherited rather than borrowed? That’s the question addressed in Technical Note 4, which offers you five practical suggestions.
[1] The square bracket will be explained later.