The A word

Revisiting the etymology of one of the most controversial words in history

Nobody knows what the Proto-Indo-Europeans called themselves. Reconstructions and comparison produce many different words for “people” and “tribe”, but the search for a definitive autonym has frustrated scholars for over a century.

It’s of course possible they didn’t call themselves anything, and drew their identity from their individual tribes. This seems likely to me. The picture painted by genetic evidence is that the groups ancestral to modern Indo-European languages and peoples expanded into and subjugated areas which had, in turn, been occupied by their cousins at a thousand years remove. These weren’t people especially sentimental about their distant relatives.

But everyone knows the most popular historical name for the language family and the people who spoke it.

In the beginning

The term Aryan is the ancient autonym of the groups we now call “Indo-Iranians”. It entered the western academic lexicon in the late 18th century to describe this family at the time when the serious comparative study of Sanskrit, Avestan, Greek, and Latin were in their infancy.

It wasn’t used to describe the whole macrofamily until long after its two most common names were well established. Indo-Germanic and Indo-European were both independently coined in the early 19th century by reference to its easternmost and westernmost members and its easternmost and westernmost geographical extremities.

In keeping with regional chauvinism, “Indo-European”, proposed by an Englishman, is the preferred option in the English speaking world. “Indogermanisch” remains the default in German scholarship, despite having been originally proposed by a Frenchman.

“Aryan” wasn’t extended to encompass the entire Indo-European family until much later, after several other terms had been proposed and rejected. Japethic was suggested only a few years after I.E. and I.G. as a counterpart to Semitic, but sadly failed to catch on. Other, stranger options have included Indo-Teutonic (“Indisch-Teutisch”), Sanskritish, Indo-Celtic, Tocharo-Celtic, and, perhaps most ignominiously of all, Indo-Hittite and Indo-Anatolian.

The less said about these the better. (Especially Indo-Anatolian). But whence “Aryan”?

A matter of honour

The initial proposal to extend “Aryan” as a term beyond the Indo-Iranian languages was made in 1819 by Friedrich Schlegel. His proposal compares arya in Indic and Iranic languages with German Ehre, Old English ār, and Old Norse eir, all meaning “honour” or “grace”. This root appears in many archaic Germanic names, and on this basis, and the basis of its close similarity in meaning to arya “noble, nobleman”, Schlegel proposed that *arya- must reflect the original Indo-European autonym, or at least that of a greater, Aryan-Germanic clade.

Long after Schlegel’s death, other authors attempted to link *arya- to ever widening sections of the family. Éire was also suggested as a cognate, which would bring Celtic into the mix. The Greek term aristos also looks similar enough to warrant further investigation even from a determined sceptic.

But alas, most of these are false cognates. Ehre and its cousins are now definitively traced to Proto-Germanic *aizō and P.I.E. *h₂eys- “to respect, revere”. Éire actually derives from Proto-Celtic *Φīweryū and is a distant cousin of Sanskrit पीवरी (pīvarī), both meaning “fat”, “rich” or “fecund”.

Aryan revisionism

Various suggestions for the etymology of aryan have been suggested over the years. Most of them aren’t worth paying attention to. More interesting are the arguments that it shouldn’t be considered an Indo-European word at all. This argument runs as follows:

1. An initial *a- in Indo-Iranian can potentially be derived from one of 5 different Proto-Indo-European sources. P.I.E’s two concrete vowels e and o both merge as a in Proto-Indo-Iranian, as in many positions do the syllabic resonants *m̥ and *n̥. In addition to these, we also have *a as a reflex of P.I.E h₂+e, which became *Ha at an early stage in P.I.Ir.

   1. In principle, this means that a P.I.Ir. *áryas could be traced back to one of three potential P.I.E. sources: *Heryos, *Horyos, or *n̥ryos.

2. Complicating matters, our initial a is long in Sanskrit ārya and short in Old Persian (𐎠𐎼𐎡𐎹arya) and Avestan (𐬀𐬌𐬭𐬌𐬌𐬀 arija). Sanskrit does make use of apparent variants of this lemma with a short vowel, but the relationship between the various forms is uncertain. It is also not entirely certain what these different forms mean, with proposed meanings in translation ranging from “friend” to “stranger” to “guest” to “enemy.”

3. A comparable cognate in the Nuristani languages (Indo-Iranian’s third and most marginal branch) is hard to come by, as are comparable words in other branches.

These considerations have discouraged further searches for a pan-Indo-European origin for the term. The association with the Nazis obviously doesn’t help matters, and has lead to a general freeze in the investigation of the term and its likely cognates.

However, the interests of scientific enquiry sometimes require us to ignore matters of decorum.

I believe that we now have sufficient evidence from multiple branches to prove that this term is not confined solely to Indo-Iranian (including Nuristani), can be reconstructed as a Proto-Indo-European word. Importantly, I believe I can also demonstrate the original meaning of this word, and the significance of the concept in Proto-Indo-European society.

The long and short of a

As mentioned above, Sanskrit principally shows a form with a lengthened ā (a vriddhi-formation with no clear explanation), which isn’t easily reconciled with the short a in Avestan and Old Persian. Vedic Sanskrit does however show several words very similar in form with a short a in this position. These include árya, aryá, and arí among others. There has, until recently, been no clear consensus on how these words relate to each other, and in many instances they are translated across various different hymns with English words that are near opposites of each other, a fact which has caused considerable confusion.

As also mentioned above, it has generally been claimed that the Nuristani languages – a small independent branch of Indo-Iranian confined to the border between Afghanistan and Pakistan – lacked any clear cognate to the Iranic and Indic terms arya and ārya.

But before we discuss this any further we need to talk about feasting

Nelson (2018) has recently revisited this issue and argued persuasively that the words arā and aro in some Nuristani languages are directly cognate to Vedic arí. Arí in the vedas is an enigmatic term that is variously translated as “stranger” or “guest”.

He demonstrates that its unusual range of semantic associations are the result of its origin in a system of prestige feasting analogous to that seen in other “big man” societies, in which prestige and authority is centred on individual heroic achievement and generosity. An arí, which Nelson renders as “prestige feast giver” but which we might more simply render “lord” (an Anglo-Saxon term with almost identical etymological meaning), can be a source of both generosity and animosity, hospitality and depredation, depending on whether or not one is a guest at his feasts or on the receiving end of his warriors’ attentions.

In pre-Islamic Nuristani society, an arā or mü moč (“man of mü”) represented the more senior of two ranks of “big men”, one who has satisfied the ritual requirements around hosting at least one prestige feast or mü. These are lavish affairs involving the sacrifice and consumption of prescribed numbers of certain livestock animals, usually goats and cattle. The more junior rank in this system is that of the “hero” or “warrior” (šura, cognate to Sanskrit śura, or lei moč). Status as a šura is obtained through success in raiding and warfare.

For both the arā and the šura, specific levels of prestige are attached to the specific number of feasts given and number of enemies killed. Nelson notes that some names of personages in the Rigveda, such as Navagva and Daśagva, may represent fossilised terms from a similar prestige feasting system, respectively referring to “one who has given nine cows” and “one who has given ten cows”.

A šura is generally a younger man still in the process of establishing himself, while an arā is generally an older man, aged roughly between 30 and 50, who plays sponsor to younger hopefuls and mediates in their disputes. Older, more senior “inactive” arā, who are themselves retired from the business of amassing and distributing spoils but nevertheless retain their prestige, play a similar mediating and advisory role to their more junior “active” counterparts.

Perceptive readers will note that this system bears striking similarities to the society portrayed in the Iliad, where younger warriors take to the battlefield in search of glory and fame and older, more established “kings” dole out spoils and administer sacrifices, and are expected to offer hospitality to non-hostile travellers passing through their territory. (Note that one of the most plausible etymologies for Agamemnon’s title anax is “bringer of spoils”). We even have a counterpart to the “inactive” or “retired” arā in the figures of Nestor and Laertes.

Further comparisons can also be drawn to early Germanic society as described by Tacitus and other Roman authors, and to raiding and feasting habits present in the literary record from Old Irish and Old Welsh sources, and the archaeological record from Mycenaean Greece.

Beyond this, Jonathan Fenno notes that the named victims of the Iliad’s heroes are almost always killed in batches of nine. Fenno posits this as a mnemonic device employed by an oral poet to both provide structure and create running totals. I would suggest it might also be a memory of a similar sort of “prestige” system, in which a hero’s “ranking” was determined by his kill count in multiples of nine, paralleling the Nuristani traditions in which different ranks of šura status are reached by killing specific numbers of enemies. Another survival may come in the form of Achilles’s statement that he has taken twelve cities by sea, and eleven by foot (with Troy itself the obvious “missing” city in the second dozen).

Rebuilding arya

Nelson uses his elaboration of the Nuristani system of prestige feasting to posit a similar, but much altered, system in Rigvedic society, which is already undergoing evolution or collapse by the time of the poems themselves. He concludes that the terms arí, aryá, and ārya (with and without initial accent) and their close associates must form a derivational family.

He concludes that the adjectival aryá and its substantivised forms must derive from arí, with a general sense of “productive, generous, hospitable, noble” which extends to “fecund” in other derivatives.

I would like to offer a few tentative modifications to this schema:

First, I do not think we should see arya- as an extended version of arí, but both as different Caland-modifications of the same root. In P.I.E., the suffix *-is is well attested suffix which forms nouns from verb stems. The suffix *-yós forms adjectives (generally but not always from nominal roots), and the related suffix *-yōs similarly forms intensive adjectives.

We can probably discount a putative pre-form *n̥ryós as this would suggest an underlying root rey-. P.I.E. roots beginning with *r are extremely uncommon, and generally indicate a vanished laryngeal. At a stretch we might suggest a root *ley- obscured by the Indo-Iranian r/l merger, but no such root appears to exist.

This leaves us with the only possible P.I.E. candidates being *Her- and *Hor-, as all vowel-initial P.I.E. roots must be assumed to have once contained a vanished laryngeal. The identify of this laryngeal needn’t be confirmed for the moment, as there are two candidates that fit this general shape: *h₂er- and h₃er-. We’ll explore the relative merits of each below.

For now, however, we need only concern ourselves with hypothetical pre forms to our various Vedic reflexes: *Herís > arí “lord, giver of feasts” (nominal singular aríḥ < P.I.Ir *-iš and P.I.A. *iṣ), *Heryós > aryá “lordly, generous, hospitable, productive”, *Heryōs > aryā “very lordly, extensively generous, etc”, *Heryos > ārya (substantivised) “noble(man), hospitable or generous, productive person”.

To explain the vriddhi-formed ārya, with a lengthened grade initial vowel, we have two options, one of which is possibly too clever by half, and one of which is possibly too simple.

1. By analogy to aryā, and possibly conditioned on the basis of an accent on the final syllable, a marginal zero-grade form *Hryós or *Hryōs was restored by hypercorrection to *Hēryos (or possibly *Hōryos).

2. Metrical considerations.

Both of these options could potentially explain why substantive forms with both the lengthened and non-lengthened initial a exist in Vedic. I will leave this as something to be considered by those more competent and interested in the matter than I.

All this is very well and very interesting, but we still can’t be sure if this term, or variants of it, are native to P.I.E. To do that, we need to find some likely cognates in other languages. Fortunately, I can point to at least half a dozen of them.

Let’s look for cognates

First, we have to decide what our posited root Her- actually is. I think we can summarily discount h₁er-; this is a purely nominal root meaning “earth” and does not intuitively relate to any of the meanings or concepts we have discussed above. This leaves us with *h₂er- (“to fit or join together”) or *h₃er- (“to move, rise, spring, quarrel, or fight”).

I believe that *h₂er- is our most obvious candidate. *h₃er-, because of the a/e/o merger in Indo-Iranian, appears to have been blended and confused with *h₂er- at a very early date, and the two cannot always be easily separated. However, cognates in other branches can be confidently traced to *h₂er-.

*h₂er- is a very well attested root across almost all branches. Its most widespread descendants are forms derived from *h₂er-mó-s ~ *h₂r̥-mó-s or h₂ér-mn̥ ~ *h₂r̥-mén-s which supply words for bodily members (most obviously “arm” in English, but also more generally “forequarter” “limb”, “shoulder” in Slavic), but also for some vehicles or other things which “fit together” (e.g. Latin arma for “armour”), and we can presume that the sense of “arm” or “forequarter” derives from a sense of “that which links (a person to something else)”.

Forms in *-tis, *-tos, and *-tus are also common; these produce a set of terms closely related in meaning:

*-tus: Latin artus “joint”, Armenian արդ ard “shape, order” (but possibly also conflated with a similar *-tis formation meaning “result, gain”), Greek ἀρτύς “arrangement, bond, friendship, order”, Sanskrit ऋतु ṛtú “fixed or appointed time or order, a season”, Avestan 𐬭𐬀𐬙𐬎 “a period of time, a judgement”.

*-tis: P. Germanic *ardiz “inborn quality, characteristic, type, manner, district”, Sanskrit ऋति ṛtí “prosperity”, Greek ἄρτι “just now, the present moment” (adverb), and Latin ars “skill, trade”, P. Toch. *ārtsä “each”.

*-tos: Latin artus “narrow, strait” figuratively “strict”, Avestan 𐬀𐬴𐬀 aša “truth”, Sanskrit ऋत ṛtá “right, proper, the law”.

These all reflect fixed qualities of one type or another, and suggest something which has already been fully “joined”. The nunances between these forms is often difficult to pinpoint, but along with interjections common among many branches descended from the bare form of this root, *h₂er or *h₂r̥ (“it is so”, “so”, “thus”, “indeed”) serve to demonstrate that this root carries a sense of completeness or finality, and we can therefore presume that its sense of “fit or slot together” often quoted in etymologies also extends to “to complete, to fully join up”.

I believe positive identification for *h₂er- can be made via the comparative method with reference to cognates in Celtic, Greek, and Germanic. I will describe these as “certain cognates”, because I do not see any argument that can reasonably be made against them.

Certain Cognates

In the Celtic, Germanic, and Phrygio-Hellenic families, we can reconstruct the following:

Proto-Celtic: *aryos “nobleman” (whence Old Irish aire “nobleman” and Gaulish arios “nobleman, freeman”)

According to ordinary phonological changes between P.I.E. and P. Celt., this word can only be a reflex of *h₂er-yo-s, with initial *h₂e- > *a and the rest of the word left unchanged.

Proto-Germanic: *arjaz “esteemed, prestigious, distinguished”

This adjective is only attested in its Proto-Norse superlative form ᚨᚱᛃᛟᛊᛏᛖᛉ arjoster < *arjōstaz. It may also be preserved in the early Germanic names Ariovistus (perhaps *arjo-wistiz, “of distinguished pedigree”) and Ariomanus (*arjo-mannaz, “man of distinction”).

Once again, this word can only be a reflex of *h₂er-yo-s by normal pre-Germanic sound shifts (*h₂e > *a, merger of short *o with *a, final *-s > *-z).

Greek: ἀρείων, ἄριστος “superior”, “best”

As with the Celtic and Germanic reflexes, these two words can only be derived from a preceding form in *h₂er-. In the case of ἄριστος this is a quite simple archaic superlative in -ιστ-, *h₂éristos. ἀρείων represents a development of a comparative in -*isō~*is-n-es, a variant of the intensive adjectival *-yōs suffix mentioned above that is also the ultimate source of English -er. *h₂er(e/h₁)isō yields *ar(e)ihō in Proto-Greek, which in turn yields ἀρείω (+ν paragogue). I include the additional e in brackets here as it’s not clear to me at time of writing if this is an “original” vowel (*-yōs is sometimes preceded by *-é-) a laryngeal *h₁ (which would have become an *e in this position in Greek) or if it is a secondary compensatory lengthening of *i triggered by the loss of *h.

Once again we only have comparative and superlative forms, but as I’ve stated before, I believe there is a good explanation for this: *h₂eryos, would have been realised in Greek as *ἀιρος as the result of Proto-Greek’s “second palatalisation”, and would have been confused in most instances with a reflex of the verb αἴρω, ἀείρω “to raise up, praise, win, or remove”.

There may be one instance of *ἀιρος surviving in its original sense preserved in the personal name Αἴροπος (perhaps “of noble countenance”), but this name is a hapax, and thus hard to analyse. This could also, plausibly, be glossed as a continuation of *h₂eris, which would have seen similar development in Greek.

*h₂er-say (other potential close relatives)

Tocharian: Direct reflexes of *h₂eryos and *h₂eris would appear in Proto-Tocharian as **arye and **arä respectively. **arye would continue unchanged in Toch. B. and become **ari in Toch. A., while **arä would become **ar in both A. and B. We find no evidence of these words in Tocharian text. However, the circumstances of their absence are interesting, as are the reflexes of *h₂er that we do find.

These include the verb ārt- (A) and ārtt- (B), “to praise, love, or rejoice in”. These suppose a P.Toch. *ārtw-, which must itself reflect a verbalisation of a noun *ārtwe, a thematicised variant of the *-tus formations discussed earlier. This noun, *ārtwe, appears to be preserved in a fossilised form in the compound verb ārtte tärk- “to overlook, be indifferent”, with the sense of “leave off that which is proper”.

We also find an adjective ārwar, ārwer “ready” (< *ārwer) with a fossilised form arwāre (< *arwōro), suggesting a past-participle form of a verb *ar- which is otherwise unattested.

It is quite likely that in early Tocharian, as in Indo-Iranian, the verb *ar- became confused with the verb *ār- <h₃er- “to leave, renounce, cease, come to an end”. That said, the traces we do have suggest that it might have originally carried the sense of “prepared, in fit state”, supporting the notion of the antiquity of this meaning. (Moreover, the fact that *ārtwe has re-evolved the sense of “praise” suggests that these connotations were still, to some extent, linked).

(Note that we also see apparently “spurious” vriddhi-forms with lengthened initial a in Tocharian just as we do in Sanskrit.)

Armenian: արի ari “manly, valiant, corageous” is unfortunately a false friend, being a derivative of *aynr and ultimately P.I.E. *h₂nḗr. This said, it may be a doubly false friend, as both *h₂eryos and *h₂eris would have resolved as **ari in Armenian as well and would thus likely have been lost at an early stage. *h₂er- also continues directly in Old Armenian as առնեմ arnem with a meaning “do, form, compose, make, craft, accomplish”.

Anatolian: The Hittite adverb āra “right, properly” appears to continue an o-grade formation in *h₂or-o, an environment in which an initial h₂ disappears without a trace (we’d otherwise expect to see *ḫara). This is supported by a Lycian cognate era “rite”, which shows a typical merger of P.Anat. *o with *e rather than with *a as in Hittite. O-grade forms of *h₂er- are attested otherwise in Italic (e.g. ōrdō < *ordō) and Indo-Iranian (e.g. अलम् áram).

The related Hittite adjectives arāwa- “free” and the Lycian noun arawa- (< erawa-) “freedom”, which are generally taken to be derivatives of the adverb, both point to an underlying verb signalling a state of “correctness” or “completeness”, also with reference to social rank and customs. Another (apparent) derivative, arāš “friend”, also hints at either a parallel or convergent evolution of semantic range to the Indo-Iranian word family of the same root.

Italic: Italic has various extended forms of this root with *-t- as briefly listed above. In addition we also have ōrdō, which most likely finds its origin in a combination *h₂or-dʰéh₁t, “(that which has been) placed correctly”, once again showing the o-grade with an adverbial sense with a definite meaning “correct(ly), proper(ly)”.

*h₂or-dʰéh₁t

All this considered, we can make the following conclusions:

1. Indo-Iranian ārya- and its near homonyms are explicable as reflexes of the P.I.E. root *h₂er-.

2. This root is intimately associated across almost all branches with physical completeness and by extension to social prestige.

3. This link to social prestige is attested directly in no less than 5 branches, with plausible inferences of similar ancient meanings for another 3.

By extension, I tentatively propose that the form *h₂eryos (and perhaps *h₂eris) were terms common to P.I.E. vocabulary to describe individuals of “rank” in a social system analogous to that seen in pre-Islamic Nuristani societies.

There has clearly been a great deal of reluctance to link any of the terms discussed above to arya for political reasons, and many alternative etymologies have been proposed, none of them particularly satisfying. However, I think in light of the Nuristani evidence summarised above, it is impossible not to see a connection here.

If we make the (merited) assumption that early Indo-European society was one based on a similar system of prestige ranking, governed by success in war and the giving of feasts, we can begin to see how meanings like “esteemed”, “praiseworthy”, and “noble” (along with the extended range of meaning around “hospitable”, “refined”, “generous”) can arise for an adjective based on the root *h₂er- “to fit, join together” and perhaps “to be or to become complete, perfected, actualised”. A *h₂eris, or a person who is *h₂eryos, is one who has satisfied the requirements necessary to hold a prestige rank, who has “joined together” all the necessary qualifications or “completed” a specific course of actions.

Terms pointing to similar states can be seen in almost daughter languages – not necessarily direct reflexes of *h₂eryos, but with similar underlying meanings. Examples include Latin nobilis (< noscō + bilis, “famed, excellent”, literally “worthy of being known”), Greek ἀγαθός (most likely “whose deeds are great”, compare also Latin magnificentus from a similar etymology), possibly Germanic *haliþaz (“hero”, perhaps related to Slavic *xvala “praise, thanks”). Similar terms also abound for the “junior” prestige rank, most of them distantly related to proficiency in arms or the act of warfare or raiding.

*h₂eryos may have been only one of a number of words used to describe these “ranks”, which may not have been entirely formalised at the time of the protolanguage.

However, it does seem to have been one of the most significant, as seen by its spread across four of the most significant families for the purposes of dating and reconstruction. (In fact, the only branch in which we find almost no evidence of *h₂er- used in this fashion appear to be Albanian and Balto-Slavic. If anyone is aware of any, please let me know.)

As of yet, we can make no assumptions as to the existence of a “two-tiered” system as seen in the pre-Islamic Nuristani society, although there is significant circumstantial evidence. A similar society, however, much changed and perhaps in the process of breaking down, is observable in both Vedic and Homeric texts. Roman traditional narratives and anthropological descriptions of Germanic and Celtic society also allow us to make similar inferences, as do Greek descriptions of the Illyrians, and, marginally, in Ancient Near Eastern evidence of Pre-Zoroastrian Iranic societies like the kingdom of Mitanni.

It is clearly a very ancient term which was likely superseded early as systems of prestige governance evolved, were recontextualised, or collapsed. In some, it has disappeared entirely, in others, it is elevated to a substantive rank, in others, it becomes a particularly archaic word for “fame” or “success”. In Indo-Iranian, we see it elevated to an endonym, perhaps because this people persisted as a highly mobile prestige-based society centred on warriors and their patrons far longer than in any others in the Indo-European family. It is also interesting that traces of this word-family and the institutions associated with it seem to persist most strongly in branches which maintained long-lasting traditions of “piratical” warfare.