The origin of the vowel that is sometimes found before sonorants, especially before /r/, in Armenian and Greek has been a traditional subject of debate in Indo–European Linguistics. While some authors explain it as a mere prothetic vowel, the followers of the so–called laryngeal theory consider it rather as another evidence of this odd phoneme.
In his Dictionnaire de la langue gauloise X. collects the existing alternative reconstructions for ‘name’ in its laryngeal version: *h1néh3-men-, *h1n̥h3-men-, h1nóm-en- describing them later as «laryngeal juggling with metathesis» (), reconstructions that displays in its prelaryngeal modality: «en(om)n̥-, (o)nomn̥-, nōmn̥» for a root that would be attested with such a meaning in Albanian Tosk emën, Armenian anun, Avestan nāma, Breton ano, Cornish hanow, Old Slavic imę, Gothic namo, Greek ὄνομα, Old Indian nā́ma, Hitite lāman -la‑a‑ma-an), Latin nōmen, Iron Ossetian nom and Digor Ossetian non (), Old Prussian emnes, Turfan Tocharian ñom and Kucha Tocharian ñem or Umbrian nome, a root considered also present as a copy from the Indo-European part in the Finno-Ugric group of the Uralic group, with variants such as nimi in Finnish and nēv in Hungarian.
Regularly and based on testimonies such as those of the Armenian anun and the Greek ὄνομα, or eventually on testimonies provided by some Celtic languages (Breton ano, Cornish hanow…), laryngalists systematically reconstruct ―almost by medical prescription, one might say― a laryngeal. In reference to the Armenian, , for example, maintains: «The most notable coincidences with Greek are the reflexes of the Indo-European laryngeals in anteconsonantal initial position (“prosthetic vowel[s]”». Similarly for : «A peculiar phenomenon of Armenian is the significant development of prosthetic vowels: the presence of such vowels in Armenian can be partly explained by admitting the existence of a laryngeal in the reconstructed Indo–European word».
Thus, some, as we have seen, would reconstruct here a laryngeal H1 , which is considered vocally associated with a timbre /e/, precisely, by the way, the only timbre in which we do not see this case documented either in Armenian, which presents /a/, or in Greek, which presents /o/. Perhaps for this very reason any of the three basic laryngeals, as observes, have finally been proposed for the initial element, decanting this same author (ibidem) for the laryngeal H3 and offering the two alternative roots of h3neh3-mn and h3nh3-men.
Necessarily simplifying the presentation now, let us say that both the Armenian and the Greek groups would be characterized, in effect, by being able to present in some words inherited from Indo–European an additional previous vowel where the other groups usually present an initial sonant: /l- m- n- r‑/. A very relevant fact is that in Greek this vowel historically only appears regularly before /r/ and that likewise in Armenian the «consonant that is systematically avoided in initial position is /r-/, but sporadically the vowel attack is also used to avoid initial /l‑/, /m‑/ and /n‑/» ().
Thus, the Greek corresponds with ὀρέγω ‘[ex]tend’ to the Irish rigim ‘I extend’ and Latin regō ‘I direct - I straighten’, the Greek responds with ἐρεύγομαι ‘I vomit – I burp’ to the Latin rūctō ‘I burp’; in contrast to the Latin ‘red’ rubrum (accusative), Western Tocarian or B ratre and Vedic rudhirá‑, Greek presents ἐρυθρός etc. For this reason, in historical Greek an initial /r/ in a patrimonial word ―that is, of Indo-European etymology― regularly betrays the drop of another previous consonant. Thus, the verb of the famous sentence by the philosopher Heraclitus of Ephesus (540‑480 a.C.) πάντα ῥεῖ “everything flows” ―that is to say: everything is temporary― has its correspondence in the ancient Indian srávati ‘flows’, where the initial /s/ is maintained, which must therefore have fallen in Greek at a time after the intolerance of initial /r/. Also for Greek ῥῖγος ‘cold’ the correspondence with Latin frigus ‘cold’ invites postulating a primitive initial sequence /sr/ ( s. srīg‑; s. u.; lege infra). Similarly, the initial /r/ in the historical classical form ῥόδον ‘rose’ of Ionic-Attic Greek was preceded by a consonant in preclassic times, as evidenced, among other facts, by the form of the Aeolian dialect βρόδον. In short, in Greek the examples of «initial ῥ‑ […] come mostly from the evolution of groups like *sr- or *wr‑» ().
In contrast, the Greek treatment of the other initial sonants is much less regular and even occasionally oscillating. For ‘light - mild’ we have Old Slavic lьgъkъ with Polish lekki, Gothic leihts with German leicht, Latin leuis, Lithuanian leñgvas or lengvùs and Sanskrit laghú-, but Greek ἐλαχύς ‘small’. For ‘free’ there is Latin līber but Greek ἐλεύθερος. To the ‘fog’ of Armenian mēg, Old Slavic mьgla with Polish mgła, Old Indian mēgháḥ (‘cloud’) or Lithuanian miglà Greek counterposes ὀμίχλη. Armenian inn and Greek ἐννέα presents additionally for ‘nine’ an initial vowel where the common language no longer has /n/: Albanian nëntë, Breton and Cornish nau with Welsh naw, Gothic niun, Old Indian náva, Latin nouem, ñu in both Tocarian languages… However, initial sonants are also sometimes perfectly stable in Hellenic words of very good Indo-European ancestry such as λείπω ‘I leave’ (cfr. Latin [re]linquō ‘I leave’), μήτηρ ‘mother’ (cfr. Latin māter) or νέος ‘young’ (cfr. Latin nouus ‘new’).
Actually ―and the detail seems quite significant― this treatment also extends to the initial sibilant, so one would be tempted to say that the vowel prosthesis occurred in Greek and Armenian before continuous consonants and specifically before typical and basic ones: /l m n r s/. However, this last modality actually occurs only before the so-called liquid s, that is: before the preconsonantal /s/. Thus, for ‘star’ both Armenian (astł) and Greek (ἀστήρ) present again an initial vowel as opposed to the Cornish steyr, Gothic stairnō, Latin stella, Eastern Tocarian śreñ, or Vedic stár. Aparently the Hittite could in this point align with the Greek, since we have records of the type ispānt- or sipānt- ‘libate’ for the same root that we find in Latin spondēre ‘promise - to commit - to marry’ and Greek σπένδω ‘libate’ and σπονδή ‘libation’, but because of the ambiguity of the writing we cannot here be sure of the Hittite testimony (). Unlike which, as we shall now see, happened with the historical vowel before /r/, the vowel before liquid s did not occur in Mycenaean. So we have, for example, pe‑ma for σπέρμα ‘seed’, since «In initial the /s/ before a consonant is not marked» (). The few exceptions, given the syllabographic nature of the Mycenaean script, do not allow us to suppose that the corresponding vowel was actually pronounced (), but rather it points to a graphic record to note, when deemed appropriate, the real presence of the simple initial /s/.
As said, from the laryngeal theory these vowels are seen as the reflection of an old… yes, of course, laryngeal! Thus, with regard to the three supposed vowel reflexes of the laryngeals in Greek observes that these «are also found in initial position (where they were formerly treated as prothetic vowels) as in érebos ‘darkness’ from *h1regw ‑, anḗr ‘man’, and omíkhlē ‘mist, fog’ from h3migh‑». Note incidentally the risk of circular reasoning involved in the automatic assignment to ἔρεβος of a laryngeal H1 by beginning with /e/, to ἀνήρ of an H2 by beginning with /a/, to ὀμίχλη of an H3 by beginning with /o/ and so on… The fact is that, in effect, non-laryngalist Indo-European Linguistics explained ―and also continues to explain― the phenomenon, on the other hand, as a case of vowel prosthesis, thus, for example, , who presents the historical vowel as «probably the result of a particular change made in still pre-Greek times. To such secondarily created initial vowel we today call prosthetic vowel».
We can specify the pre-Greek character of the process: it dates back to at least the 2nd millennium BC, since in the Mycenaean Greek of that time one can well collate, for example, the classical Greek ἐλεύθερος ‘free’ with the Mycenaean ereuterose ( s. ἐλεύθερος. item : e‑ru‑ta‑ra ‘red’ cfr. ἐρυθρός). Even authors very prone to laryngalism have been in this case in favor of the explanation as a vowel prosthesis, as is the case of Bernabé: «the presence of a prosthesis cannot, without more ado, be considered as evidence of a lost laryngeal. Attributing the prosthesis to the presence of a laryngeal and then reconstructing the laryngeal because there is a prosthesis is a flagrant petitio principii» (; seconded here by ).
The question, therefore, to be elucidated lies in trying to determine if said historical vowel would truly come from a laryngeal segment, whatever its coloratura (or numbering), in which case Armenian and Greek would maintain a more archaic stage than other languages, or if it were a real prosthesis, in which case the vowel would not represent any old Indo-European phoneme but an innovation now shared, now carried out independently in Armenian and Greek. In the latter case, secondarily, it would be convenient to inquire under what conditions or contexts such historical and at least three-thousand-year-old vowel could have been generated.
A first objection ―of course, not definitive but indicative, like all those that will follow― to the laryngalist proposal would be that of the scant economic simplicity of such an assumption, since it accepts as more relevant the minority testimony of two linguistic groups against the testimonial majority of the ten of the remaining Indo-European groups.
Secondly, the laryngalist position obviously implies accepting that the minority testimony is the conservative one and that all the other languages would have innovated in unison, suffering an apheresis or loss of a short initial vowel. The process itself is not problematic, even Modern Greek lost a fair amount of the unstressed open-syllable-initial vowels that Ancient Greek had: ἐροτῶ ‘I ask’ ροτώ; ὑψηλός ‘tall’ ψηλός… In the Romance sphere there are also examples of analogous processes: Italian vescovo, Portuguese bispo or Valencian bisbe have lost the initial e- of episcopu- ‘bishop’ and also the Spanish migraña with respect to its Greek etymon ἡμικρανία or the Friulian legri ‘joyful’ ( Latin alacre-). The /i/ of Greek-Latin historia disappeared from the Italian storia ‘history’ like the /a/ of amor in the Spanish phrase por mor de. It is further believed that the international ghetto procedes ultimately from Latin Ægyptu- ‘Egypt’, so it would have lost the initial vowel, already monophthongized at the time (Italian ghetto, Provencal guet; s. u.). Also in Logudorese Sardinian we have apheresis, for example, in [krˈezja] from Latin ecclesia- ‘church’ ()… etc. But what, as we see, normally constitutes a sporadic or marginal process, in the Indo-European linguistic ensemble it would have nevertheless systematically affected all the languages in approximately 85/90% of the linguistic groups, without seeing what reason could have made it lose without more ado so many laryngeal entities and without leaving a single trace in said groups.
Thirdly, conservatisms are as a rule more typical of the languages or dialects of the periphery and it is difficult to see how this peripheral condition could be applied to Armenian or Greek, languages that at least historically do not occupy marginal geographic positionsa as do, on the other hand, the westernmost Celtic, the eastern Indian and Tocharian or the northern Baltic and Germanic, all of them groups where the initial vowel in question does not appear.
Fourthly, the phonotactic foundation of an initial laryngeal-plus-sonant sequence is not well seen either. Even accepting that H or a form of H once represented the most frequent laryngeal, the glottal fricative /h/, initial sequences of the type /hl hm hn hr/ or /hs/ followed by consonant are quite unusual in languages. In short, it is hard to see how something similar to the banal /h/ could be generalized in a phonetic context that is so inauspicious, so hardly natural, or at least so unusual.
In the fifth place, it turns out that, seen from the point of view of the proposal of a prothetic vowel, the process in its essence would be a phenomenon of linguistic recurrence or recidivism in Armenian, since in this language copies from Middle Iranian with initial r present prothetic [e] or [a] (generally [eɾ] and [ar]); thus, Middle Iranian rang Armenian erang ‘colour’ or Middle Iranian rāt Armenian ar̄at ‘abundant’. However, the most recent copies no longer have the prosthetic vowel and «in later loans initial r- is preserved» (). Mutatis mutandis, the same can be said about the vowel in question before liquid s, since «Iranian loanwords also present a phenomenon of sporadic vowel prosthesis also in the case of sp- and systematic in the case of the consonantal link typical of Iranian xš- which gives Armenian šx-: cfr. ašxarh ‘world’» (). Thus, in a real, historical situation, what is evident is that the Armenian does not preserve any vowel ―or laryngeal― but that simply performs a banal phenomenon of prosthetics. Therefore, here the laryngealist proposal would violate the principle ―also, of course applicable to the linguistic reconstruction― of «scientific unitarianism, according to which we must assume for prehistoric times the same behaviour in the processes of evolution, contact and diffusion of languages that we observe in the historically documented periods» ().
Sixth, the explanation of the vowel as a prosthesis before initial /r/ in Armenian is consistent with the general avoidance of the stop group plus /r/ before a vowel (ȻrѴ), sequence that in this language mechanically generates a metathesis (ѴrȻ) and both initial and interior: Vedic śubrá- ‘brilliant’ but Armenian surb ‘saint’ (), while also in this case the explanatory capacity of the laryngalist proposal remains again as an isolated phenomenon and ad hoc.
Seventh, the fact that in Greek the additional vowel is always short is more consistent with the hypothesis of a prosthetic vowel than with that of a laryngeal, which according to traditional laryngalist theory is regularly at the origin of most of the historical long vowels of the Indo-European languages. Again, the circular reasoning that precisely because we always have short vowels in Greek in this case, here by chance the laryngeal would never have come into contact with any vowel, is clearly refutable. Because of its non-distinction between long and short vowels, the Armenian testimony cannot be used here.
In eighth place, since in real phonetic descriptions the phonemes usually described as laryngeal are associated with laryngeal or glottal sounds and being /h/, the glottal fricative, the most frequent phoneme in this series and also very present in both ancient Greek as in the Armenian languages, if we accept the laryngalist perspective, then it is not understood why these languages would not have developed this solution here and presented, therefore, initial aspiration. Indeed, one might expect, for example, rather /hˈonoma/ (†ὅνομα) with aspiration than /ˈonoma/.
In ninth place, at least for Armenian, the prosthesis is phonologically very congruent with its general tendency to vowel supplementation or spontaneous generation of a vowel ―very generally [ə]― a phenomenon in this language, as is known, very extended in internal position, that is to say: as an epenthesis. Thus, grel ‘write’ is pronounced [gərˈel] in contemporary Armenian.
In tenth place, the evaluative comparison of proposals for very analogous phenomena that are more observable in other languages also advocates the explanatory hypothesis of the prosthesis. Thus, before a couple of toponyms like Spanish Rentería - Basque Errenteria, we know thanks to the etymology (derived from the Spanish renta ‘rent - income’) that it is not the first language that would lose an initial before-sonant vowel ―or laryngeal― but rather the second language that would seek a prosthetic vowel, since Basque is reluctant to all initial /r/ and has been historically, so that old Latin words of the type rege- ‘king’ have been regularly copied as errege.
Along the same lines and eleventh, the hypothesis of vowel prosthesis before liquid s- is congruent with the typologically banal phenomenon of prosthesis that, without going literally any further, we find in Western Romance: Latin stella French étoile, Spanish estrella, Portuguese estrela… The reluctance to initial /s/ before a consonant remains fully operative in contemporary Spanish, a language that still automatically generates a prosthetic /e/ in this position. (cfr. English standard estándar; English slogan eslogan and even the brand Movistar is pronounced /mobiestˈaɾ/! for feeling like a composite of English star etc.). Similarly in the Indo-European ensemble the sequence /s/ before /r/, a group that tends to be complex in terms of articulation, is treated by means of formulas that avoid it, such as /str/ or /fr/. Thus, in front of the ancient Indian srávati ‘flowing’ or srava- ‘flood’ or Lithuanian sravė́ti ‘distill - filter’ we have German Strom ‘torrent - stream’, Latvian straume ‘torrent’, Polish strumień ‘torrent’ or the Thracian name of river Στρύμων. For its part, the transition from /sr/ to /fr/ is found in Latin (for example, frigus ‘cold’, cfr. Greek ῥῖγος) and in Brithonic Celtic (for example Breton fron ‘nose’ as compared to the Irish srón). Again, the passage from zero to the prothetic vowel (0 Ѵ) or the emergence of another consonant (0 Ȼ or Ȼ1 Ȼ2) offers many more linguistc parallels than the unprecedented passage of a laryngeal segment to a vowel (H Ѵ).
Twelfth, the laryngeal theory involves here an internal, regular and general change, but the dissimilar tolerance to the various sonants involved suggests rather a contact phenomenon, since, fundamentally due to diachronic, dialectal, diastratic and diatopic differences, in the phenomena of linguistic contact the results tend to be less regular and sometimes even sporadic and contradictory. Thus, for example, in Latin both ampulla ‘ampoule - little bottle’ and amphora ‘amphora’ come from the same Greek root with the aspirated labial (/ph/), a phonetic type that was introduced into Latin only in a second phase of greater contact with Greek. Likewise, for example, in an ancient phase of Latin, lampada, ‑æ ‘lamp - luminary’ was obtained from the Greek accusative λαμπάδαν, in a way that, as attested by Spanish lámpara, remained in popular speech in front of the most recent and cultist word lampas, ‑adis from the nominative λαμπάς with genitive λαμπάδος (cfr. also French top Spanish tope but in more recent times English top Spanish top etc.). Thus, while, as we saw, intolerance was greatest for initial /r/, it could be said to be least for initial /m/: Armenian mard ‘man [mortal]’, Greek μορτός ‘man - mortal’ in a gloss of Hesychius (c1057 Schmidt) like German Mord ‘murder’, Avestan mərəta-, Old Church Slavonic mrьtvъ ‘dead’, Ancient Indian mr̥tá- ‘dead’ and martá– ‘mortal’, Old Irish marb ‘dead’, Latin mortis (genitive) ‘dead’ and mortuus ‘dead’ or Lithuanian mirtìs ‘death’.
In thirteenth place, the typological material also supports the prosthetic explanation, since the restriction of sonants in the initial and especially of the liquid ones and especially of the vibrant one is a phenomenon documented in very diverse and distant linguistic groups, which suggests that it is a resource with a certain universal scope and, therefore, essentially based on common human articulatory or phono-acoustic characteristics. To continue with the emblematic and most frequent case of restriction of initial /r/, let us remember that, in addition to Basque, also in Iberian its two vibrants «never appear in initial position» (). Also the vibrant *r was not allowed in initial position in Proto-Uralic either, «but numerous borrowings from Indo-European and non-Indo-European languages, even from remote times, totally changed the situation […] Still today, however, in a Finnic-Ugric language of the Permian group, Udmurt or Votyak, the presence of [r] is not allowed in word–initial position» (). In addition, the prohibition of initial /r/ is still operative in the Samoyedic group of the Uralic linguistic ensemble and even more: in Proto-Samoyedic «there seems to have been a tendency to extend this restriction to the other liquid ., as well, for most of the Uralic items with *l- show *y- in proto-Samoyedic» (), although the alternative explanation is possible here that we are facing a simple process *[l j]. Likewise, words begining with /r/ and /n/ are also, for example, very doubtful in the Nilo-Saharian languages ().
In the fourteenth and last place, the typological material also supports the prosthetic explanation in the sense that the addition of an initial before-sonant vowel is a phenomenon not only documented in a good number of languages but also precisely in languages geographically not far from Armenian and historic Greek. So the qualitative clue is added here to the quantitative. Thus, the initial vibrant was not tolerated in the ancient language of the Hittites either. Likewise and exactly as in Spanish, in Kurdish the strong vibrant in initial [r] is allowed but not the simple [ɾ] (). Likewise, the two great and related non-Indo-European languages of ancient Anatolia, about which we have some information, have similar restrictions. in Hurrian the «liquids /l/ and /r/ do not appear in word-initial position» () and in Urartian there are no «words with initial [r‑]» (Wilhelm 2008b: 108). Similar phonotactic restrictions are found in the Turkic ensemble: «Initial r- is assumed not to have existed in Proto-Turkic, and it still does not occur in native words» (). The same happens with the other liquid: «Initial l- is assumed not to have existed in Proto-Turkic, and it still only occurs in copied words or after loss of an initial vowel» (). Also in South Siberian Turkic «The liquids /l/ and /r/ and the nasals /m/ and /ŋ/ do not occur initially in native words» (). Already in Old Turkic «Nasals appear word–initially only in interrogative elements» (). In short, in the Turkic linguistic ensemble «Word-initial n, m, ŋ, l, r are avoided, the only seemingly native exception being the interrogative ne ‘what’ […] Loanwords and liquids are often provided with prothetic vowels» (). Similarly : «In the Turkic and Mongolic languages, there was an absolute prohibition of liquids in word beginnings, although now, in most cases, they are allowed in loanwords (often ancient) from Arabic or Indo–European languages».
Certainly a language in which neither initial /l/ nor /n/ nor initial /r/ nor anteconsonantal initial /s/ were admitted offers a very little Indo-European phonological pattern. Consequently, the series of arguments exposed invites us to reject the hypothesis of the laryngeal theory: the Indo-European root would actually begin with a sonant, possibly a form like *náman- ‘name’ or similar, and it would be the speakers of Armenian and Greek who would have innovated by adding directly and almost automatically in these cases simply a prosthetic vowel. A corollary to this would therefore be the reasonable suspicion that when these events occurred, these languages would have been in contact with non-Indo-European languages that were reluctant to use sonants in word initial position.
Thus, the simplest explanation consists in supposing that we are dealing with an ordinary phenomenon of contact between Greek and Armenian with a language[s] with more severe restrictions than the Indo-European group for some initial sonants or for some initial consonant groups. Conversely, the oddest and less plausible explanation would be to see here an archaism concerning those two languages, Greek and Armenian, and much more an archaism that implied the presence of a laryngeal. Anatolia or its area of influence seems, of course, the most propitious scenario for this phenomenon of linguistic contact that occurred at least shortly before the Mycenaean period and both because of the intermediate position of Anatolia between historical Armenian and Greek.
[2] «fenomeno peculiare dell’armeno è costituito dal notevole sviluppo di vocale protetiche: la presenza di queste vocali in armeno si può spiegare in parte ammettendo l’esistenza nella forma ie. ricostruita di una laringale».
[3] «consonante che sistematicamente è evitata in posizione iniziale è /r‑/, ma sporadicamente l’attaco vocalico viene impiegato anche per evitare /l‑/, /m‑/, /n‑/ iniziali».
[6] «prawdopodobnie wynik szczególnej zmiany, która się dokonała jeszcze w epoce pragrecki. Taką wtórnie wytworzoną samogłoskę początkową nazywamy dziś samogłoską protetyczną».
[7] «la presencia de prótesis no puede, sin más, considerarse como testimonio de una laringal perdida. Atribuir la prótesis a la presencia de una laringal y reconstruir luego la laringal porque hay prótesis es una flagrante petitio principii».
[8] «prestiti dall’iranico inoltre presentano un fenomeno di protesis vocalica sporadica anche nel caso di sp- e sistematica nel caso del nesso consonantico tipico dell’iranico xš– che dà arm. šx–: cfr. ašxarh ‘mondo’».
[9] «unitarismo científico, según el cual debemos suponer para épocas prehistóricas el mismo comportamiento en los procesos de evolución, contacto y difusión de lenguas que observamos en los periodos documentados históricamente».
Downloads (Last 12 months)
Catedrático de Filología Latina
Facultad de Filología
Universidad de Valencia
