By parallel series of terms, often of revealing etymology, but which differ from language to language, Iranian, Indic, Greek and Italic testify to a common Indo-European heritage: that of an hierarchical society, structured according to three fundamental functions, those of priests, warriors and tillers of the soil.

According to Indo-Iranian traditions society is organized into three classes of activity, priests, warriors and farmers. In Vedic India these classes were called “colors,” varṇa. In Iran, they have as their name pištra ‘craft’, the etymological sense of which is also “color.” We must understand the word in its literal sense: they are indeed colors. It was by the color of their clothes that in Iran the three classes were distinguished—white for priests, red for warriors and blue for farmers, according to a profound symbolism, which is taken from ancient classifications known in many cosmologies and which associates the exercise of a fundamental activity with a certain color that is itself connected with a cardinal point.

The same classes and the members of these classes are not called by the same terms in India and Iran. Here are the respective words:

India	Iran
(1) brahmán (brāhmaṇa-)	(1’) āθravan
(2) kṣatriya (rājanya)	(2’) raθaēštā
(3) vaiśya	(3’) vāstrō fšuyant
(4 śūdrá)	(4’ hūiti)

These words do not correspond; however, the organization is the same and also the mode of classification rests on the same distinctions. It is in their true meanings, as in their relations within the social system, that we must examine the terms.

Here, briefly, are the lexical meanings of the two series:

India

Iran

We see that both in India and Iran these terms, although distinct, are organized in the same way and refer to the same activities. This social structure was maintained longer in Iran than in India.

This terminology is basic to the problem which dominates the whole organization of Indo-European society. The two groups of terms are different in their lexical character, but they agree in their social reference. The tripartite division of society to which they testify is the oldest to which we can attain. Its survivals in historical times have not always been recognized, especially in Indian society. It was the merit of Emile Senart to show that the Indian castes should not be explained by internal rules but are in fact the continuations of much older divisions which India has inherited and which did not originate on Indian soil. The Indian castes are the much fossilized systematization of divisions which go back certainly to the Indo-Iranian period, if not to Indo-European society itself. The problem is to examine the words which define in India and Iran this division into castes, and then to see if, in other societies of the Indo-European family, we can recognize a similar system.

If we review the various terms, we find that for the most part they can be interpreted directly and have a signification which is still accessible to us. We can show this by taking them in succession.

The Iranian term for “priest,” Avestan āθravan, has its Vedic correspondent in atharvan which, to tell the truth, is not quite what one would expect, but the two words can be superimposed without great difficulty, the difference between –θr– in Iranian and –thar– in Indic not constituting any serious obstacle to the comparison. The derivatives are symmetrical in both Indic and Iranian: Av. aθauruna-, which denotes the function of a priest, and Vedic ātharvaṇá ‘relating to the atharvan’; the detail of the structures is evidence for the concordance of the original meanings. Only the etymological analysis of the word remains uncertain.

It has long been thought that āθravan– and atharvan– can be explained by the word for “fire,” which is ātar in Iranian. Although the connection is plausible from a formal point of view, we run into great difficulties with the meaning itself: it is by no means certain that āθravan is the fire-priest. In Mazdaean Iran he is responsible for religious ceremonies; in India, the atharvan is endowed with magical powers. This conception finds expression in the collection of magical hymns called precisely the Atharva–Veda. The function of this personage is divided thus: in Iran the exclusively religious side is shown, in India we see the magical aspect. But there is nothing we can see in this role which particularly relates to “fire.” There never existed in Iranian any etymological relationship between ātar and āθravan; and, the second difficulty, this word for fire, Av. ātar-, is quite unknown in India, where fire as a material concept and as a mythological figure is called agni-, a term corresponding to Latin ignis and to Old Slav. ognjǐ. We cannot therefore regard the connection between ātar– and the word for “priest” āθravan as anything like certain.

Isolated as it seems to be, this term may nevertheless go back far into history. That it is confined to Indo-Iranian does not prove that it is of recent creation. In any case, to regard it as Indo-Iranian is perhaps to simplify the problem, because even within Indo-Iranian, as we have seen, the forms do not exactly coincide. Their relationship is perhaps not that of common forms which have been inherited in parallel ways by both members. A morphological detail suggests a different and more precise relationship. As against Vedic átharvan-, Avestan presents a root with inflectional variation, āθravan– in the strong cases (nominative and accusative), athaurun– (i.e. aθarun-) in the weak cases (genitive, etc.). If we posit for Iranian a primitive flexion āθarvan– (altered into āθravan– under the influence of ātar-), genitive aθarunō, etc., we get a regular structure, whereas the Vedic declension átharvan-, átharvanaḥ is not, and seems to have been recast. It would then be possible to regard the Vedic form átharvan– rather as a borrowing from Iranian āθarvan– than an authentic Indic correspondent. This would explain better the relative rarity of átharvan– in the Rig-Veda as compared with brahman-, and also its specialization in the world of charms and deprecatory rites, while the term in Iranian keeps its ancient value as a term for a social class.

To designate the functions and the class of priests in India, the hallowed term is brahmán. It raises a problem which is still more difficult. The exact signification and origin of the word has provoked long debates which are not yet at an end.

There are in fact two forms, differentiated by the place of the accent, by their gender, and meaning: bráhman (neuter), brahmán (masculine), the first designating a thing, the second a person. This shift of the accent from the root to the suffix is a regular procedure which, because the Indo-European tone preserved its discriminatory and phonological function, served to distinguish an action noun from an agent noun.

What is the meaning of the well-known term bráhman? It is almost impossible to define it precisely and in a constant fashion; in the Hymns it admits of translation in a disconcerting number of different ways. It is a mysterious fluid, a power of the soul, a magic and mystical power; but it is also a hymn, a religious practice, an incantation, etc. Consequently, how can we characterize with any exactitude the masculine brahmán that is “the person vested with bráhman,” who is also designated by the derived noun brāhmaṇa?

There is nothing in Indian tradition to guide us in a reconstruction either of the form or the notion it designated; what we lack is a concrete sense to which we could attach the diversity of usage. India itself does not supply this firm pointer: bráhman is tinged with a meaning of a mystical character; it is one of the notions on which Indian speculation exercised itself at an early period and this obliterated the point of departure. The analysis of the form has fared no better: the origin of bráhman is one of the most controversial questions in Indo-European etymology. For a century now the most varied suggestions have succeeded one another and have been the object of dispute. Since the fluid sense of bráhman admits of any interpretation, the textual exegesis of the Vedic uses itself reflects in turn various tentative etymologies. Let us recall the principal ones.

It has been proposed to connect bráhman with a group of ritual terms in Indo-Iranian of which the principal ones are Vedic barhiṣ– ‘sacrificial grass’, Avestan barəziš– ‘cushion’ and especially Avest. barəsman– ‘bundle of branches which the priest holds in his hand during the sacrifices’. There has in fact been a formal proposal to make the etymological equation Ved. bráhman = Av. barəsman-. However, without even insisting on the difference of the structure in the root syllable, a point which is not without importance, the gap in sense is so marked in Vedic itself between the notion of “sacrificial strewing” (barhiṣ-) and that of bráhman– that it would be vain to attempt to reconcile them. The technique of oblation to which barhiṣ– in Vedic and bráhman– in Avestan refer has never had any extensions in the abstract sense, religious or philosophical, which is the exclusive sense of bráhman. In fact, barəsman in Avestan is only a ritual term without religious implications, designating an instrument, the use of which is prescribed along with that of other cult accessories. The characteristic association of barəsman– with the verb star– ‘spread’, to which the Vedic phrase barhiṣaḥ star– ‘to spread out the sacrificial grass’ exactly corresponds, shows that these terms had from their origin only a material and strictly technical sense, to which they remained confined. They had nothing in common with bráhman.

Of quite a different kind is the ancient connection between Vedic bráhman and Lat. flāmen, which once was in considerable favor. In this concordance we were supposed to have evidence of ancient terms preserved both in Latin and Indic; an ancient neuter coined by means of the same suffix (-man, Lat. –men) is supposed to have become in both languages the word for a cult officiant. Added to this were supposedly the remarkable resemblances in the functions of brāhmaṇa and flāmen respectively. But this equation encounters numerous objections. The comparison of the essential element of the form, the root brah– in Indic and flā– in Latin, causes grave difficulties; we should have to posit for Latin *flags–men-, a form difficult to justify, with the additional disadvantage that it does not yield any precise sense either in Italic or in Indo-European. This is why this equation has been abandoned.

We shall not linger over other attempts which have come to naught, but we think that a new fact has come to light which must put an end to this discussion. We now have at our disposal a firm foothold for the determination of the original sense of bráhman. It is the Iranian correspondent which supplies it, since in an inscription in Old Persian the word brazman– figures, which corresponds exactly to Vedic bráhman. The sense of the Old Persian word has been established by W.-B. Henning, [1] who has shown that brazman develops to brahm in Middle Parthian and Middle Persian, and that brahm signifies “form, (decent) appearance” and is applied sometimes to clothing and sometimes to deportment and conduct.

In fact brazman in Old Persian refers to cult and may indicate the “appropriate form,” the “rite” which this cult demands. This would also be the sense of bráhman in Vedic; all the usages of the term have in common the notion of “ceremonial form” in the behavior of the priest who makes the offering and in the operations of the sacrifice. It is along these lines that we should define the proper sense of the term bráhman, which later was charged with mystical and speculative values.

Consequently, the Indic brahmán (or brāhmaṇa-) is he who ensures the performance of the rites in the prescribed forms. This is the definition which, at the conclusion of this analysis, harmonizes the functions of the cult official with the now assured sense of the fundamental Vedic term bráhman, Old Persian brazman. The conceptual basis is now established in Indo-Iranian, even if the root of the word does not recur elsewhere.

We are still too poorly informed of the Persian religion of the Achaemenids to assess the role of the brazman in cult. There is no proof that this abstract noun ever produced in ancient Iranian an agent noun, parallel to Vedic brahmán, to designate the person who knows and carries out the operations of cult. This is one reason for believing that brahmán is a purely Indic term which has its equivalent a different term in Iran: the āθravan of the Avesta.

The words for the other two classes are derivatives or compounds which are easy to interpret; they do not give rise to such complex problems as those which were raised by the term for the priest. But each is tied up with an important concept and because of this they deserve a brief comment.

The designation for the warrior class in India is Skt. kṣatriya, rājanya. The first word is a derivative form of kṣatra ‘power’, a notion which will be studied in greater detail in the Iranian world; [2] the second, rājan(i)ya– ‘of royal stock’ comes from the word for ‘king’ rāj(an)-. These two words are not applied to dignitaries but to the members of a class and designate them by the privilege attached to their condition. They do not refer to the profession of arms; both evoke the concept “power,” “royalty.” We discern in these two clear terms the manner in which the word for “warriors” was orientated in India: if there was a connection between “warriors” and “power,” this is because temporal power was not the necessary attribute of the rāj.

We shall see in fact that, when examining the concept of rēx as it is defined both in ancient Rome and India, that the “king” was not endowed with the real power. [3] What we learn from the words kṣatriya and rājanya is that power, defined by kṣatra and rāj(an)-, was associated with the profession of arms.

In Iranian society, the equivalent term to kṣatriya is, in its Avestan form, raθaēštā-. More frequently, raθaēštar– is encountered, a secondary analogical form of agent nouns in –tar (a type corresponding to Gr. -τωρ, -τήρ and Latin –tor); *-star– as an agent noun from stā– is impossible, as roots with an intransitive sense, like stā– ‘to keep upright’, do not supply agent nouns. The formation of the compound justifies the analysis raθaē–štā-, which signifies “he who stands upright in the chariot,” just like the corresponding Vedic ratheṣṭhā, the epithet of the great warrior god Indra. This descriptive term goes back to a heroic age with its idealization of the warrior and its celebration of the young fighter who, standing upright in his chariot, hurls himself into the fray. Such is the Indo-European conception of the noble warrior. It was not on foot or on horseback that the Indo-European warrior went into battle. The horse is still a draft animal attached to the war chariot. It needed a long history and a number of inventions before the horse could become a mount and so transform the conduct of war. But long after the revolution in technique and culture represented by the appearance of the mounted warrior, the vocabulary was still to testify to the priority of the chariot as compared with equitation. Thus the Latin expression equo vehi, that is, “go on horseback” continued to employ the verb vehere ‘to transport in a vehicle’. The ancient verb which was appropriate to the technique of the chariot was adapted to the new practice of horse-riding. In Homer eph’ híppōn baínō (ἐφ’ ἵππων βαίνω) signifies not “to mount a horse” but always “to get into the chariot.” The sole function of the horse was to pull the chariot. To mount a horse was no more conceivable to a warrior of the Indo-European age than to ride an ox would have been for the people of the classical period. In calling the “warrior” by the term “fighter in a chariot,” Iran was more faithful than India to the Indo-European ideology of the warrior class.

For the third class, the Indic term is vaiśya, which literally means “man of the viś,” which is approximately “man of the people.” This establishes a connection between this last class and membership of a social division, called viś.

It is quite different in Iran, where the complex, and not always well understood, designation is composed of two associated words designating one and the same person: vāstryō fšuyant.

The first is a derivative from vāstra ‘pasture’, cf. vāstar ‘herdsman’. These two terms (vāstra, vāstar) are very common in the Avesta and are endowed with great importance. We have had occasion elsewhere [4] to analyze the etymology and to study the sense which they assume in the pastoral way of life and the religious ideology of Iran; they are among the most significant words of Zoroastrian doctrine. The second, fšuyant, is a present participle from the root fšu– ‘to rear stock’. The class is thus named analytically by a combination of the two words, one of which refers to “pasturing” and the other to “stock-breeding.”

A double expression like this belongs to a category of compounds known under the name of dvandva. These are double words, the two components of which are in asyndeton, simply juxtaposed, both in the plural or, more frequently, in the dual. The two terms, closely associated, form a conceptual unit. This type is illustrated in Vedic by Mitra Varunā, which unites the two juxtaposed gods; dyāvā pr̥thivī (dyaus/pr̥thivī) ‘heaven-earth’; and also mātā–pitarā(u) ‘the two, mother and father’. The dvanda subsumes the unity of the concept in its two distinct species. It may also appear in looser forms and simply associate two qualifications. For instance in Latin the expression Patres conscripti only makes sense if we recognize it as two juxtaposed nouns, patres on the one hand, and conscripti on the other; that is to say, here we have two groups of persons, originally independent, who together constituted the Senate. It is an expression of the same type which we have here in Iranian: the vāstryō and the fšuyant are two different kinds of persons: one has to do with pastures, the other is in charge of livestock. Then, since each forms part of a single class, a single term serves to indicate them: vāstryō fšuyant. This Iranian class has an explicit functional denomination in contrast to the Indic term vaiśya, which simply indicates their belonging to a tribe.

For completeness’ sake we must mention a fourth class which appears in the most recent lists. In India, the fourth estate is called śūdrá, the etymological sense of which escapes us; it is applied to people of the lowest category, ethnically mixed, people without a well-defined profession or a precise function.

In Iran, too, after the three traditional classes, one text mentions the hūiti, a term which seems to signify “occupation, craft” and which is applied to artisans. We do not know the date when this new social distinction came about which lumped all the artisans together and made them into a distinct class.

To estimate the importance of this triple classification it should be noted that it did not only apply to groups of human beings. It was extended to the groups of concepts which were thus brought into relation with the several classes. This is not easy to recognize at first sight; it is indirectly revealed in expressions which appear to be of little significance, but which are understood in their full sense once they are brought into connection with what are essential social concepts. We read in an Achaemenid Persian inscription of Darius the expression for a prayer to avert three calamities from the country: dušiyārā ‘bad harvest’, hainā ‘the enemy army’, draugā ‘the lie’, that is to say, the perversion of moral and religious order. This is not a chance formulation. These three calamities correspond to a necessary order. The first, “bad harvest,” ruins the farmer; the second, the attack of the enemy, affects the warrior; the third, the “lie,” concerns the priest. We find here again, transposed into three kinds of misfortune, this same hierarchy of the three classes which we have found implicit in the words for their representatives. Society cannot be conceived, the universe cannot be defined, except by this triple order. Is this division, which embraces the whole people, limited to Indo-Iranian society? It might be thought to be very old, going back to the Indo-European period. In fact, it has left its traces everywhere. We recall in particular in Greek the legendary tradition about the original organization of Ionian society. A reflection of it survived in the myth concerning Ion, the eponym of the race. A legend (preserved by Strabo, 383) attributes to Ion the division of society into four classes:

Plato in the Critias also alludes to it when he enumerates:

hiereîs	dēmiourgoí	geōrgoí	mákhimoi
(ἱερεῖς)	(δημιουργοί)	(γεωργοί)	(μάχιμοι)
“priests”	“artisans”	“farmers”	“warriors”

On the other hand we know the names of the four great Ionian tribes, headed by the four sons of Ion. These four proper names may be related to the four social classes. Unfortunately they are cited in a different order in different authors, which makes the comparison difficult and prevents the direct equation of each name with one of the four functions.

The form in which these names have been transmitted has been affected by the interpretation: it is clear, for instance, that Plutarch intends his list to designate the warriors, artisans, farmers and goatherds. All the same, this list of names may well roughly cover the four classes. We can try to establish some correlations, but we must discard Plutarch ‘s interpretation, which is too transparent to be anything but a late adaptation of terms which were no longer understood.

Hóplētes (hóplēs) is known from a number of inscriptions: e.g. from Miletus (fifth century BC) hoplḗthōn (ὁπλήθων), genitive plural with an orthographic variant; in Dacia, we encounter a phylḕ hopleítōn (φυλὴ ὁπλείτων). The name is doubtless to be connected with hόplon, plural hόpla, not in the sense of “arms,” which is secondary, but with the proper sense of “instruments, tools.” On this interpretation the word would designate artisans.

Argádēs (confirmed by epigraphic reference from Cyzicus and Ephesus as a name given to a khiliostús, a group of one thousand men) has a resemblance to the name of Argos, the meaning of which we know. Argos signifies τὸ πεδίον ‘ground’, ‘plain’ in the language of the Macedonians and the Thessalians, according to Strabo. Argádēs, if it refers to the ground or soil, would then designate the farmers. Such is the second identification which we can make with some probability.

Geléōn and Aigikoreús would then correspond to the noble functions, and we should expect them to head the list, as in fact they do in Herodotus. For Aigikoreús we are struck by the resemblance of this compound to aigís, the “aegis” of Athena. It is also relevant to recall that the four classes were respectively put into relation with Zeus, Athena, Poseidon and Hephaistos. We may link the last two classes to the latter two gods, Hóplēs as “artisans” to Hephaistos, Argádēs as “farmers” to Poseidon, who was patron of agriculture among his other functions. There remain the two classes attributed to Zeus and Athena. The Aigikoreús may be linked with the latter. As for Geléōn, we recall that he is under the patronage of Zeus according to an inscription (IG II², 1072), mentioning Zeùs Geléōn. This testimony associates the last term with the only divine name left at our disposal, that is, Zeus.

It is certain that we have here survivals which were no longer understood at the time when this tradition was recorded, and their interpretation remains hypothetical. However, the manner in which the different persons divided the social activities among themselves conforms with the explicit traditions of India and Iran. The fourth activity is that of the artisan, as it is in Iran. Finally, this distribution is regulated by divine order. We may therefore suppose that here, in a legendary form, the old social divisions have survived and this would in itself be a reason for considering it as Indo-European and not merely Indo-Iranian.

This analysis may also find confirmation in the Italic world, notably in the Iguvine Tablets, a ritual formulated in the Umbrian language for the use of the Atiedian priests of Iguvium (Gubbio) in Umbria.

The tablets describe the ceremony of the annual lustration performed by the priests; it consists of a circumambulation of the territory of the city. The procession is interrupted by stations at each gateway of the town, each one occasioning oblations and recitations of formulae. Now, in the prayers which are repeated in the form of litanies, certain expressions recur which are worth analyzing. They appeal for divine protection over creatures or things which are enumerated in six consecutive words, divided into three groups of two:

The first term, ner–f (accusative plural of ner) corresponds to Skt. nar, Gr. anḗr (ἀνήρ); these are the men of war, the chiefs; arsmo is the term designating the rites, the sacred; uiero = Lat. virōs ‘the men’; pequo = Lat. pecus ‘livestock’; castruo, which corresponds to Lat. castra, designates the cultivated land, the fields; fri–f = Lat. fructus. We have thus: the chiefs, the priests; the people, the herds; the fields, the products of the earth; three groups of two words or, one might say, three successive dvandva. One of these dvandva, ueiro pequo ‘men-animals’, recurs in Iranian in the form pasu vīra ‘animals-men’; this correspondence, which has been long noted, illustrates the antiquity of the rite and the formulation itself of the Iguvine Tablets.

Each of the three is concerned with a department of social life: first, the priests and chiefs; then, man and the animals; finally, the earth and its fruits.

This division corresponds, although in a somewhat different manner, to the ancient scheme, with an extension. It mentions not only the society of men, but also the products of the soil. This addition apart, the principle of classification remains the same: the priests, the warriors, the farmers (men and herds).

We limit our study to an enumeration of the proofs of this social organization, where these proofs consist of specific terms or of onomastic data. The other pointers which may be gathered from a study of the religious and mythologies lie beyond the limits of our subject. In any case, it is the domain in which George Dumézil has contributed works of fundamental importance which are too well known to need citation here. [5]