4.3. Consonants of Proto-Indo-European, part 2: places of articulation; nasals

The consonant phoneme inventory of Proto-Indo-European, as usually reconstructed. IPA transcriptions are based upon their manifestations in the surving daughter languages, as exemplified in e.g. SoundCorrespondences.html and discussed in detail below.

Stops: Places of articulation

Modern Indo-European languages have inherited bilabial, alveolar and velar stops with various patterns of voicing and aspiration, as discussed in the previous section. Proto-Indo-European *b is particularly uncommon in reconstructions, but is attested in a few words such as *dʰeub- > English deep, in which *b is expected as the ancestor of English [p], according to Grimm's Law.

Alveolar place is referred to as “dental” in many works; as we can't know for sure whether they were dental or alveolar, it doesn't matter greatly. However, as the reflexes of PIE *t, *d etc in modern Indo-European languages are usually alveolar, not dental, it seems more likely that they were alveolar in PIE. Where dental reflexes exist, as in French /t/ = [t̪], or dental fricatives such as Old English þ [θ] or Spanish intervocalic d [ð], e.g. nada [naða], those could be due to developments that post-date Proto-Indo-European. Even so, the Germanic development *t > þ [θ] (part of Grimm's Law) might suggest that in the dialect of PIE from which Germanic developed, *t had a dental place of articulation, [t̪].

As well as the “plain velars” *k, *g and *gʰ, a series of pre-velar stops *ḱ, *ǵ, and *ǵʰ is needed, motivated by the fact that in a number of branches of Indo-European, the latter developed into pre-palatal, postalveolar, or alveolar affricates and fricatives (so-called satemization), whereas words containing “plain velars” did not develop in that way. For example:

There is further discussion of these changes in the section on satemization.

The formation of such affricates or fricatives from fronted or palatalized velars is quite common across Indo-European languages, and is witnessed historically by e.g. waves of palatalizations in Slavic, “velar softening” in English: a synchronic phonological alternation such as electri[k] ~ electri[ʃ]ian ~ electri[s]ity and analo[g], analo[ʤ]y arising from earlier sound changes between Latin and French. Since such precedents attest to true velars developing into postalveolar or alveolar affricates and fricatives, it is not necessary to suppose that the pre-velar stops were originally palatal. Rather, a natural chain of incremental changes (natural in the sense of phonetically incremental and cross-linguistically well-attested) would be from pre-velar to palatal to postalveolar to alveolar, e.g. *ǵʰ [g̟ʲʱ] > *[ɟ] ~ [ɟʝ] > [ʥ] ~ [ʤ] > [ʣ] and the corresponding fricatives.

In languages in which the pre-velar stops were not affricated, such as Germanic languages, the contrast between velars and pre-velars was lost or conflated, which implies that the articulatory difference between velars and pre-velars cannot have been too large. For such reasons, it seems to me most likely that the pre-velars were just slightly fronted velars, not palatal plosives (which are not common in surviving Indo-European languages); they may also have had palatalized secondary articulation, hence the transcriptions such as [g̟ʲ] in the table above, but I know of no sure evidence for that. However, as many Indo-European languages (e.g. Slavic, Irish, to some extent English) exhibit and retain systems of palatalized vs. plain (or palatalized vs. velarized) secondary articulation of consonants, by conventional philological reasoning it seems likely that a contrast between some palatalized vs. non-palatalized consonants was a feature of the proto-language. This hypothesis is also shaped by the supposed existence of a third group of velars, the “labiovelars” (more specifically, labialized velars).

Labialized velars

As with the pre-velars, a set of distinctively labialized velar consonants is motivated by differences from velars in the way that they developed in some of the daughter languages. Sometimes they develop into labial consonants proper, for example:

*kʷ [k̟ʷʰ] > [pʱ] in Ancient Greek πόλος [pʱólos] “pivot, axis”, undoubtedly from PIE *kʷel- “to turn”, which in Germanic developed into [xw] > [hw] (as in Old English hwēol “wheel”). Several other branches of Indo-European retain velar stops in reflexes of this stem. A historical development from [kw] > [p] is also attested in e.g. Proto-Celtic *kʷenkʷe > Welsh pump.

*gʷ [gw] > [b] in Ancient Greek βαίνω [bɑenɔ:], whereas English come, Lithuanian gimti, Sanskrit गमति [gɐmɐtʰi], Polish gać, and Albanian gatuaj attest to the original velar nature of the first consonant of this word.

Survivals of the labiovelar category are seen in *gʷʰ > Latin [gʷ] e.g. unguis and Armenian եղունգ ‎[jɛruŋɡʷʱ], both from the same root *h₃nogʷʰ- [ŏ̥nogʷʱ] “fingernail”. The Armenian pronunciation provides a very useful example of what may appear to some as an unusually “exotic” consonant, and shows us what *gʷʰ may have sounded like.

Nasals

There are abundant examples of *n and *m in syllable-initial and syllable-final position, attested across the whole Indo-European family (see  SoundCorrespondences.html). Variation between [m] and [n] in Proto-Indo-European is seen in *pel-mo > film vs. *pel-no > Bosnian pelena “diaper”.

Assimilation

*h₂enk- [ħaŋk] “angle” > Iron Ossetian æнгуыр [aŋgur], Sanskrit अङ्क [aŋka], Ancient Greek ἄγκος [aŋgos]. See Allen (1987: 35) for evidence from Latin authors that [ŋ] was the pronunciation of Ancient Greek γ in this context.

*h₂eng-ul- [aŋgʊl] “ankle” > Sanskrit अंगुली [aŋguli], Ancient Greek ἀγκύλος [aŋgýlos] > Modern Greek αγκύλος [aŋgílos]. The stem *h₂eng- [ħæŋg] > Persian انگشت [aŋgʊ́ʃt], Siraiki [aŋgútʰɑ:].

*pénkʷe [péŋkwe] “five” > Lithuanian penki [peŋkʲɪ], Armenian հինգ [hiŋg], Latin quinque [kwiŋkwe] > Italian cinque [ʧiŋkwe], Spanish cinco [θiŋko, siŋko]. See Allen (1978: 27‒28) for evidence from Latin authors that [ŋ] was written with Latin n in this context. Compare these with assimilation to a following bilabial plosive in Welsh pump [pɪmp], Oscan 𐌐𐌖𐌌𐌐𐌄 pumpe; assimilation to a following alveolar plosive in Ossetian фондз fondz, Pashto پنځه pindza, Ancient Greek πέντε pente > Modern Greek [pede]; and assimilation to a following postalveolar or palatal plosive in Urdu پانچ panch [paɲʧ], European Romani panzh [paɲʒ].

*ǵʰengʰ- “gang” > Sanskrit जङ्घा [ʤʌŋhɑ:], Romani chang [ʧɑŋ], Urdu ٹانگ [ta:ŋg̊].

*ḱónk-e- “hang” > Sanskrit शङ्क [ʃʌŋkʌ] “doubt”.

*ḱr-n-go- “horn” > Sanskrit शृङ्ग [ʃɹ̩ŋgʌ]; cf. the unassimilated alveolar nasal in *ḱr̥-n [k̟ʲr̩n] > English horn.

*spré-n-gʰ-e > spring.

Another example of bilabial assimilation to a following velar is seen in *h₂emǵʰ-u- [ħaŋgʱu] “narrow” > Eng(lish), hang(nail), Latin ango [aŋgo:], German Angst. Evidence that the nasal in the Proto-Indo-European root was bilabial comes from Hittite ḫamanki ~ ḫaminkanzi “to tie, betroth” < * h₂m-ón-ǵʰ-ei, * h₂m-n-ǵʰ-énti. Such a derivation implies a combination of anticipatory assimilations: first, the alveolar *n assimilates to the following *ǵ in the cluster *mnǵ > [mŋg], and then possibly [mŋg] > [ŋŋg], or possibly the [m] is deleted or “eclipsed” by the [ŋ] in a double articulation [m͡ŋ].

A third and quite secure example of bilabial nasal assimilation is seen in the PIE prefix kom- > Proto-Italic *kom > Old Latin com > Latin cum > variant forms con-, col- in e.g. conditiō “putting together” > condition, colligō “to draw together” > collect.

Syllabic nasals

In some phonotactic contexts, such as in zero ablaut stems (which have no vowel), the nasals [m], [n], and possibly [ŋ] could be syllabic. Perhaps a little confusingly, philological works write syllabic consonants using an under-circle, like this: *n̥, whereas the IPA notates syllabic consonants with a subscript short vertical line, like this: [n̩].

For example:

Syllabic [n̩]

*kn̥h₂-onk-o-s [kn̩aoŋk̟os] “honey” > Hindi कनक kanak “gold”

*mn̥-ti‑ [mn̩ti], a form of *men- “mind” > Proto-Germanic *mundi > English “mind”

*seh₁-mn̥ [se:mn̩] > Latin semen “seed”

*tn̥h₂-u- [tn̩:u] “thin” > Persian تنک tunuk, Balochi تنک tanak “narrow”

*wed-n̥s [wedn̩s] “wet” > Sanskrit उदन् udan, Latvian ūdens “water”

Possibly syllabic [ŋ̍]

*dn̥ǵweh₂ [dŋ̍gwɐħ] “tongue” > Old Latin dingua, Latin lingua.

Syllabic [m̩]

*dḱm̥t‑, *déḱm̩-t‑ “ten” > Lithuanian dešimt, Latin decem.

*septm̥ [septm̩] “seven” > Latin septem

In Indo-Iranian languages, syllabic nasals sometimes developed into the vowel [a], e.g. *n̥dʰér [n̩dʱér] “under” > Sanskrit अधर adhara; *dḱm̥t‑ > Persian صد‎ sad, which adds further indirect evidence that the underlying nasals in such words were syllabic.