The signal success that had been achieved in the decipherment of the Persian column of the Achaemenian inscriptions greatly facilitated the accomplishment of the difficult task that still remained. In the other two columns it was comparatively easy, especially in the short inscriptions, to identify the combinations of signs that corresponded to the proper names in the Persian, and to check them by their recurrence in the positions where they were again to be expected. The satisfactory application of this process left no doubt that the writing in both columns was from left to right; and that they were translations of the same text. When the groups of characters composing the proper names were ascertained, the next process was to separate them into letters or syllables, and to identify each with the corresponding letter in Persian. The result of this inquiry was to show clearly that the writing was partly syllabic and to raise the suspicion that it was also partly ideographic. It was seen that the signs were too numerous to be limited to an alphabet, and that long words could be expressed with comparatively few signs. In some cases its ideographic character was illustrated by the occurrence of only one sign to represent an entire word—such as ‘King.’ It was also observed that in the second column a vertical wedge usually preceded proper names as a determinative.

From the time when Niebuhr pointed out that there were three essentially different styles of writing, and that each style was uniformly reproduced in the same relative position in all the inscriptions, the subject had given rise to much speculation. It was at first thought that the three columns repeated the same text in the same language written in different characters.[611] Grotefend, however, recognised that the languages also were different, but he thought they were dialects closely related to each other. The first, as we have seen, he considered to be Zend, which he called the Median language; the second he thought was Parsi, or the language of the Persians; and the third another dialect of Persian, possibly Pehlevi.[612] Subsequently he changed the order of the last two, and described them respectively as resembling Pehlevi and Parsi.[613] As regards the signs, Münter thought that in the second column they were syllabic and in the third ‘hieroglyphic’; Tychsen and Grotefend thought that both had signs for vowels and consonants, which were at times replaced by an ideogram. Grotefend further saw that the second included signs for the combination of a consonant and vowel; the third he considered had no vowel signs, but used signs for the triple combination of consonant, vowel and consonant.[614] He entirely rejected the idea that either system was purely ideographic. In 1824, he prepared a Table for the third edition of Heeren, showing some words that corresponded to each other in the three languages. The inscriptions he selected were the G of Niebuhr, the inscriptions at Murgab and on the Caylus vase. These he arranged word for word in parallel columns opposite to one another. He used a full stop to indicate the combination of wedges that went to form each letter or syllable; indeed at that period it required scarcely less skill to divide the words into letters than to distinguish the words themselves. No attempt was made to assign values to the characters, and for many years no farther progress was made. In 1837 he still thought the three columns represented dialects of Old Persian, though they might not exactly correspond to Zend, Pehlevi and Parsi. The two first he considered nearer to each other as regards language; but he remarked that the two latter presented a closer resemblance to one another as regards the writing. Still he said the resemblance was by no means so close as that between the third column and the Babylonian inscriptions. He saw indeed that the writing of the third column was a mere simplification of the Babylonian; and he hazarded the useful conjecture that the writing of the second might be only an arbitrary modification of the third. He would not even yet admit that either could be, strictly speaking, described as syllabic; and he entirely rejected the idea that the third was a Semitic language.[615]

But the study was now upon the point of entering on an entirely new phase. We have already seen the success with which Burnouf and Lassen applied the discovery made by Grotefend to the long list of names in the I inscription. The visit paid in 1843 by Westergaard to Naksh-i-Rustam resulted in the recovery of a farther list of provinces from the Tomb of Darius. On his return to Germany, he made over his copy of the first column to Lassen, who was best qualified to turn it to account, and he devoted himself to an attempt to decipher the language of the second column. The result appeared in the ‘Zeitschrift für die Kunde des Morgenlandes’ of 1844; and the same year, in English, in the ‘Mémoires des Antiquaires du Nord.’[616] In 1845, the Memoir was republished in German, along with Lassen’s Second Memoir, and this, as the latest, must be regarded as the most authoritative version.[617] The investigation is based upon an analysis of the various proper names contained in the inscriptions. Westergaard began with the well-known names of the Kings which Grotefend had turned to such good account, and afterwards reviewed those of the provinces recently deciphered by Lassen. From these he obtained a sufficient number of values to attempt a transliteration of the more common words occurring in the Persian version, and especially in the well-known form that opens so many of the inscriptions. He has given us a transliteration of this passage, which is the first ever made into Roman characters of a Susian inscription.[618] His work was necessarily based upon Lassen’s defective decipherment of the Persian signs; and it was therefore inevitable that it should reproduce the same erroneous values. He transliterated all the various inscriptions, beginning with those of Xerxes as the shortest and simplest, and proceeding to those of Darius, including the one at Naksh-i-Rustam, which he was the first to copy.[619] It is curious to compare the earliest attempts at transliteration with those subsequently made by Oppert and Weisbach. It will be seen that few of the words in his glossary could now be recognised. He had little opportunity of displaying his skill as a translator, for in this department he implicitly followed Lassen’s rendering of the corresponding Persian column, and we find the same errors in both, from ‘Arça’ downwards; but he concludes with the K inscription, which has no Persian counterpart, and here he achieved a respectable success. He was able to make out sufficient to show that Darius laid claim to the foundation of the Persepolitan platform.[620] In the German edition he omits his adventurous transliterations (as well as the declination of Ku, ‘king,’ which it will surprise some scholars to hear takes the forms of ‘Kuyoni’ in the accusative singular, and ‘Kuthin’ or ‘Kuthrar’ in the genitive plural), and he has consequently lessened the interest, if not the value, of that work.[621]

The result of his investigations was to show that the language was partly alphabetical and partly syllabic. Down to 1837, Grotefend had recognised about sixty different signs.[622] Westergaard rightly calculated, from the relation of the vowels to the consonants, that they must exceed a hundred, but his actual identification only reached eighty-two. He noticed two other probable groups, but he was unable to decide whether they formed one sign or two. One of them has now gained admission.[623] In consequence of the great care he exercised in the collation of his copies only four defective signs found their way into his list.[624] He noticed the recurrence of twelve other signs, which he ascribed to the error of the copyist, though he thought three of them might be genuine. It turned out that three were genuine, but only one of those that seemed to him likely to be so.[625] In addition to the determinative sign before proper names, which Grotefend had pointed out, he recognised another, the horizontal wedge, which is sometimes interchangeable with it. It will remain a curiosity in the history of decipherment that Westergaard should have gone out of his way to declare that the words for God, Ormuzd and Heaven were not preceded by a determinative—‘their importance being, no doubt, thought too great to need any such distinction.’[626] He, in fact, mistook the determinative sign itself for the vowel which happens to precede his rendering of these three names.

He thought he had discovered signs for six vowels (a, â, i, e, u and o), and eighteen consonants.[627] Of the former only one (u) turned out to be correct. His a and i have both syllabic values (an and in). His long â is a defective sign (No. 10). His e should be i and his o an a. Only four of his consonants were ultimately found to be used as alphabetical signs, and for these his values are correct (t, kh, s and m). A few of the others represented correct consonantal values, though used syllabically. For example, his p is ap, his r is ra, his s is as, his n is na. He was more fortunate in his syllabic values, sixteen of which have been accepted as strictly correct, if we include two where w is substituted for the more usual m; and eight others may be passed as approximately correct (Nos. 5, 11, 12, 22, 23, 60, 69, 81). The values he attached to about fifteen other signs were, however, highly inaccurate and misleading; and he was unable to assign any value whatever to twenty-two out of his eighty-two signs.[628]

He could not identify the sign for l; no doubt because there was no corresponding sound in the Persian by which he was guided, and he thought that no syllable began with m. He considers all the syllables begin with a consonant, though he detected two possible exceptions, where the syllables appear to be as (Nos. 8 and 82). Neither of these are in reality exceptions, but there are numerous other instances where the vowel is the initial. None of his syllables exceed two letters, though later investigation has shown that there are several cases where a vowel separates two consonants. He noticed that the syllabic sign was frequently replaced by two signs to express the simple consonant and the vowel. He observed the peculiarity that the language knows of no difference between the sounds of m and w, both being written with the same letter; and he also showed that the sonants are generally absent, for while k, p, t are represented, he could not find g, b, or d. He saw that the consonants are sometimes doubled in the middle of a word; and he conjectured that this form was adopted to indicate the distinction. Thus, he thought the syllabic sign for pi might express either pi or bi; but if preceded by the alphabetical sign for p, the two together denoted the surd.

At a time when the affinities of the language were entirely unknown St. Martin had given it the name of Median, no doubt under the supposition that it was allied to Persian and spoken by the Median branch of the common Aryan race.[629] The investigations now concluded by Westergaard showed for the first time that its affinities were with the Scythic or Turanian family of languages, and the inconvenience of retaining the name of Median first became apparent.[630] There could be no doubt that the Median conquerors were Aryans, as may be shown from the names of their leaders, and their close relationship to the Persians in race and language was accepted by all the early historians as self-evident. It is clear, therefore, that the language of the second column is not theirs, yet Westergaard determined to adhere to the name of Median as its most appropriate designation. He was led to adopt this course chiefly because there appeared to be no other district except Media where it could have been spoken, and also because its graphic system seems to indicate contact with Assyria. There can be little doubt that his decision was fully justified at the time, and, to avoid unnecessary confusion, in referring to the earlier writers we have in this chapter generally followed their description of the language of the second column as Median. On the other hand, modern discovery has tended to increase the importance of Susa, by showing that at least a portion of its territory had become the hereditary dominions of Cyrus, before his accession to the Persian throne. At the same time the connection between the language of the second column and that of the ancient inscriptions found in the district has been more completely recognised. There appears good ground to believe that the Aryan race had been established throughout Media long before they rose into historical importance. The inconvenience of calling a Scythian language after the Aryan Medians has therefore manifestly increased, while the name of ‘Susian,’ though not free from objection, has become more appropriate, and is the one now generally adopted. How far the language was spoken beyond the limits of Susiana, among the subject tribes of Media, may still be open to conjecture. There is, as we shall see, some evidence in favour of its extension.

It will be recollected that early in 1846 Rawlinson had been able to translate the whole of the last paragraph of the second column of the Behistun inscription from the Median or Susian text, the Persian version of that passage having been found illegible.[631] Several years later Mr. Norris, with all his special knowledge, was obliged to confess that he was ‘unable to give a better translation than Colonel Rawlinson has prepared’ of that very paragraph.[632] In the same year Rawlinson announced the opinion his studies had led him to adopt upon this subject. It does not appear that he was as yet acquainted with Westergaard’s essay; but he had arrived at the same conclusion with reference to the Scythic affinities of the language. He likewise estimated the number of the signs at about a hundred, ‘the vowels, unless they commence a syllable, being for the most part inherent.’ He does not appear to have noticed the absence of sonants, but he saw that there must be considerable interchangeability in letters of the same class, and perhaps even between n and l. He added that the language evinced a repugnance to r. He held that ‘it resembles the Scythic in the employment of post-positions and pronominal possessive suffixes.’ In the declension of nouns it uses post-fixed particles that are frequently the same as in modern Turkish, and he notified the existence of a Tartarian gerund. The pronouns are, he says, Semitic; the adverbs Aryan; the vocabulary a strange agglomeration of Turkish and Semitic. Although its construction is more akin to Aryan than Scythic, yet upon the whole he decides that its affinity is with the Scythic, and suggests that that would be a more appropriate name for it than Median.[633] He thought some of its peculiarities might be explained by comparison with the Georgian, which, when time permitted, he proposed to undertake.[634] He suggested that it was the language of the aboriginal race whom the Aryan Medes had conquered, and whose settlements reached at least to Behistun, where an inscription had been found without the Persian translation, apparently indicating that it was locally comprehensible.[635]

When Rawlinson had finished his Persian Memoir he devoted himself to a more elaborate study of the Median, and he appears to have nearly completed an essay upon the subject. Meanwhile, however, he was drawn from this branch of inquiry to the more attractive and useful study of the Babylonian column: and his work on Median was never published.[636] He did not, however, lose his interest in the subject. He handed over his copies and other materials to Mr. Norris, who undertook the investigation of the subject and to whom he continued to give valuable assistance as occasion arose.

The study of Median was taken up in the meantime by Dr. Hincks from the point where it was left by Westergaard and Rawlinson. Hincks’s contribution is contained in three papers read before the Irish Academy between June 1846 and January 1847; and the last two appeared after he had seen the opinions of Rawlinson to which we have referred.[637] These papers were communicated without delay to the Asiatic Society in London, as we learn from a note by the Secretary, who says he had just received (June 1846) a communication on the Median cuneiform from ‘a learned clergyman in a remote part of Ireland.’[638]

Hincks differed in some material points from Westergaard. He would not admit the long vowels, and limited them to a, i and u, adding, however, er to their number. He restricted the simple consonants to m, p, t, k, s, and n; and from these he thought all the other signs were combined. He rightly admitted the compound syllabic form, consisting of two consonants joined by a vowel, which Westergaard had overlooked, and he even allows the occasional use of a double syllable, such as ‘ersa’ and ‘washa.’ He showed that the vowel is never omitted, though not always necessarily pronounced. Thus the group ‘an-na-ap’ reads simply ‘anap’; and he did not consider that the simple consonants preceding its syllabic form altered the sound as Westergaard had suggested.[639] Hincks adopted the eighty-two signs distinguished by Westergaard, and attempted the identification of sixty-seven, or seven more than Westergaard. His identification of the three vowels, a, i and u, was correct: and his limited number of consonants were also correct so far as they went, with the exception of p, which is the syllable ‘pir’ (No. 40). He, however, allotted more than one sign to some of his consonants: thus k is represented by three signs, t by two; and conversely he allowed that several characters might express the same value. He accepted fifteen of Westergaard’s correct syllabic values and added nine of his own, to which we may add fifteen others nearly correct, or thirty-nine in all; thus, including his three vowels and five consonants, he had forty-seven signs available for transliteration, as opposed to the twenty-nine of Westergaard. Singularly enough he made little improvement in the misleading syllabic signs of his predecessor. Some of them, indeed, he made considerably worse, and added others of his own, so that, notwithstanding a few corrections, their number rose to seventeen. He, however, pointed out that there was a determinative before the words for ‘god,’ ‘Ormuzd’ and ‘heaven,’ contrary to the opinion of Westergaard; and that the group the latter had mistaken for the vowel a was precisely that sign. He was less successful in his classification of the language, which he could not accept as Scythic. He considered its affinity was with the Aryan family, but he could not find that any of the Indo-European languages had similar inflections.[640]

A few years later De Saulcy wrote two articles in the ‘Journal Asiatique’ on the Median, without, however, making any important contribution to the subject.[641] His papers deserve notice chiefly on account of his eminence in other departments of study, and because he was the only Continental writer whose attention was directed at that time to this special branch of inquiry.[642] He thought he could recognise that it bore a close relationship to Persian, sufficient to justify the opinion of Strabo that the two languages were the same. Both Westergaard and Rawlinson had already observed that some of its grammatical forms and vocabulary could be best explained by reference to the modern Georgian and Turkish; and De Saulcy, notwithstanding his opinion of its affinity with Aryan, fully recognised that it had left traces in other quarters, including Kurd, Mongol, Armenian and the Gipsy tongue, but nowhere to a greater degree than in Turkish. He supports his opinion with a wealth of illustration drawn from these sources that must have fairly distracted his printer, and the indiscriminate use of the mechanism of philological dictionaries has in fact led the writer into many serious errors. He does not refer to Hincks, and notwithstanding all the resources at his command, he has fallen far short of the Irish writer whose country rectory was ill provided with these artificial appliances. De Saulcy has not, in fact, added a single correct value to those already known, and has failed to recognise several already established as correct. The utmost generosity cannot concede to him the possession of more than twenty-one correct values and nine nearly correct, all previously known; so that he had not more than thirty available for transliteration, as opposed to the twenty-nine of Westergaard and the forty-seven of Hincks. But he introduced a host of errors that are wholly his own. He assigned no less that thirty values that are absolutely wrong, although he only attempted sixty-two out of the eighty-two in Westergaard’s list. But it is not only in the details of decipherment that he went astray; his error covers the whole conception of the structure of the language. He has no less than fifteen different signs for vowels representing many fantastic gradations of sound. Different modifications of a and ou monopolise each three signs, besides the concession of one each for hou and ô. A simple u is not suffered to appear; but ha, he or e?, ya, aï?, and oui are classified among the vowels. Some difference still exists as to the treatment of the vowels. The latest authority limits them to a, e, i and u, and excludes o.[643] M. Oppert ranks a or ha among the consonants; and in addition to e, i and u he admits o and such sounds as yi, ya, ah among vowels.[644]

But it is in the treatment of the consonants that De Saulcy has most departed from the earliest and the latest scholars. He fully admits the syllabic character of the language, but he has done the utmost under the circumstances to conceal this peculiarity.[645] In his list the only sign that appears in syllabic form is the semi-vowel ar attributed to a defective sign (No. 10), and the eye already accustomed to the appearance of syllabic combinations is struck by the singular bareness produced by purely alphabetical letters. He admits altogether twenty-one distinct consonantal sounds, of which fourteen are ‘quiescent’ or simple consonants, and each is represented by one sign only, except m, w, to which three are allotted. He allows six gutturals to a language that has at most but two, and fills the other classes with scarcely less profusion. He recognises that the signs for m and w are interchangeable, and that d and t, as well as b and p, have several signs common to both; but he gives both p and t the exclusive use of others; and dè and dh have each a sign reserved to themselves. His consonantal sounds are given different signs according as they are supposed to be followed by a, by e or i, and by o or ou, an idea no doubt suggested by the restricted use of the same principle in Old Persian. These letters may be said to be practically syllabic signs, as they are only used in conjunction with a vowel; but in reality the classification involves serious error. It is now admitted that, with few exceptions, the same sign never conveys the sound of more than one vowel, and in the isolated cases where it occurs it is generally at the end of a word. In the whole Syllabarium of Weisbach there is only one instance where the same sign is given an optional sound of pe, pat, and in that of Oppert there are only three where a can be exchanged for i.[646] In neither authority is there a single instance where e and i are interchangeable. It is therefore a fundamental error of the gravest character to represent the same sign as systematically employed to express either of two vowel sounds, even if the two selected were ever interchangeable. It is, moreover, doubtful whether there was any difference allowed between the sounds of o and u, except in the attempt to express foreign words. De Saulcy’s system also excluded the compound syllables when two consonants are divided by a vowel. While Hincks did good service in adding a third determinative to the two acknowledged by Westergaard, De Saulcy actually refused to admit the determinative character of the horizontal wedge, and thereby reduced the number to one. On the other hand, he thought he had detected a sign to indicate the plural, which is in fact nothing but the syllable sin.

Perhaps the most useful contribution he made was the remark that, while a few of the Median signs bore a certain resemblance to those of the same value in Persian, the larger number of them were actually ‘identical’ with the Assyrian.[647] It must be confessed that it was long before this identity was detected, and even such a practised copyist as Westergaard declared that the five different species of cuneiform writing then known ‘differed from one another in the shape of nearly every letter or group,’[648] and considerable practice is still required before their identity can be brought home to the eye. The similarity—to employ a less emphatic expression—proved, however, of great assistance in enabling future inquirers to fix the value of the signs. Some time had yet to elapse before they could take advantage of this discovery, for when De Saulcy wrote, the Assyrian characters were themselves still undeciphered.

De Saulcy’s work was criticised by Löwenstern in the ‘Revue Archéologique’ (1850), where he sought to trace the affinity of Median to the Aramaic branch of the Semitic family, while he admitted that some of the endings might be traced in Pehlevi and New Persian. He did not consider it to be the Median language, which, he maintained, is properly represented by Zend; and he suggested, as Mr. Rich had done before, that it might have been the dialect of Susa.[649] Holtzmann fully recognised that it contained Semitic elements, but still he considered that it should be regarded as the mother of Pehlevi; and he subsequently added that it was probably the language of the court at Susa.[650] Still the opinion of its Scythic origin, first suggested by Westergaard and Rawlinson, continued to gain ground, supported, no doubt, by the numerous similarities to this linguistic stock that were pointed out in the learned Memoirs of De Saulcy. In 1852, M. Oppert, writing on the Persian column, went so far as to propose to drop the designation Median altogether, and, following the suggestion of Rawlinson, to substitute ‘Scythic.’ At that time he considered it to be the language of the hordes who overran Western Asia, and who, after twenty-eight years of domination in Media, were finally expelled by Cyaxares. It must be admitted that they were fortunate to leave such a permanent memorial of their passage.[651]

A more important contribution than any of these was made by Luzzato, who showed with sufficient clearness that twenty-four of the Median signs corresponded to the Babylonian (1850).

Down to the time we have now reached the progress made was disappointing. We have seen that much speculation was indulged as to the affinity of the language and the people to whom it should properly be attributed; but very little knowledge of the language itself had been acquired. Dr. Hincks had read forty-eight signs with sufficient correctness, and De Saulcy thirty-one, out of the one hundred and eleven with which the language is written; but both scholars were encumbered by the multitude of unknown or erroneous values. In 1852, however, a considerable step was made in advance. On July 3 of that year, Mr. Norris read a paper to the Asiatic Society on the result of his study of the Median column of the Behistun inscription with which Colonel Rawlinson had entrusted him; but the essay did not appear in the Journal in its completed form till 1855. The much greater diversity of material at his command enabled him to dispose of many difficulties that had obstructed earlier scholars. In the first place, he was able to increase the number of signs from the eighty-two that had puzzled his predecessors to one hundred and three, and to all of these he endeavoured to assign values. He was successful with fifty-seven, and twenty-one others were sufficiently correct for purposes of transliteration, making a total of seventy-eight in addition to two determinatives. One other sign has not even yet been satisfactorily settled,[652] and he still had twenty-four incorrect values. Unlike many decipherers, however, he was able to distinguish the gradations of certainty attaching to the values he assigned. Those that appear in his list allied to any vowel other than a, i or u are to be regarded as questionable. He estimated their number with tolerable accuracy at about twenty. He fully recognised the true syllabic character of the language. ‘Each character,’ he says, ‘represents a syllable which may be either a single vowel or a consonant and vowel, or two consonants with a vowel between them.’[653] He has, however, given six signs in his list a purely consonantal value, m, r, s and t, the two latter being each represented by two signs.[654] Four of these are correct, but the second signs for s and t are both wrong. He limited the vowels to a, i and u, but he has also allowed e to appear in his list, because he found a sign in Median that corresponds exactly to the Babylonian for that letter, and it is now admitted as correct. He reckoned eleven initial consonantal sounds: they include the consonants already named with the exception of m, together with th, ch, p, k, v, l, n and y. With these united to one of the three vowels all the syllables in the language are formed. The double syllables are composed of the same consonantal sounds (excluding th and y), separated by a. He did not consider that any of the signs represented the union of two consonants separated by i or u, an opinion that has been since over-ruled. He recognised that there was no difference between the surd and sonant consonants at the beginning of a word, and he agreed with Westergaard that the double letter in the middle emphasised the hard sound. He saw also that the same sign was used equally to express the Persian m and w; and that ‘the aspirate, which is quite uncertain, must also be disregarded.’ He accepted the two determinative signs already admitted by Westergaard, and also the determinative before the words for ‘god’ and ‘heaven’ which had been pointed out by Hincks.[655] He also recognised the ideogram for ‘month.’ The gradual recognition of the similarity of a large number of the Median and Babylonian signs was at length beginning to bear fruit. Norris indicates a resemblance between nearly fifty in his list of one hundred and three signs.[656] Twenty of these are indeed ‘identical’ in form, and independent investigation proved that they conveyed the same sound in both languages. In the others the similarity was sufficiently striking to afford an important confirmation of the values arrived at by the study of the text. Indeed in a few cases—such, for example, as the e already mentioned, the similarity to the form of a Babylonian character was the only clue to the value of the Median sign.[657] If he had allowed himself to be more influenced by the Babylonian script he would have slightly increased the number of his correct values;[658] though, on the other hand, a too rigorous adherence to this rule might have landed him in some errors. He hazarded the important statement, subsequently so remarkably confirmed, that there are evident signs that the syllabarium had been originally devised to express a Scythic tongue,[659] for, he said, ‘the unchangeable roots, the agglutinative structure and the simple syllabisation of such tongues are perfectly suited to such a mode of writing, while the Semitic and Indo-Germanic cannot without the most awkward and unsystematic arrangements be represented by it.’ He considered that the language of the second column must have been that of the pastoral tribes of Persia; and he pointed out that the omission from the Median text of the names of the districts in which Otiara and Rhages are placed showed that those towns were well known where the language was spoken, and it also afforded some evidence of the area it covered. On the other hand, the discovery of Scythic inscriptions at Susa seemed to point no less emphatically to its southern range, and the possibility presented itself that that was the source from which it spread. De Saulcy farther showed that a close analogy might be traced to Turkish; but Norris was the first to point out that its nearest modern relationship is with the Volga-Finnish branch of the Scythic family, and that it bears a close resemblance to the language now spoken by the Finns.[660]

He was the first to treat the grammar systematically. He pointed out with sufficient accuracy the case-endings of nouns,[661] and explained that, in common with Finnish, they distinguished more cases than the Indo-European languages, such as the allative and the locative. He showed that there was no distinction of gender, and that the adjectives had the same case-endings as the nouns. He explained that the plural was formed by the addition of pa (which, however, ought to have been p, pe, or ip), and that the case-ending was suffixed to it. He distinguished the pronouns with tolerable correctness, and showed their relationship to the Tartar languages. He also proved that the analogies with these languages are more obvious in the verb than in the other parts of speech.

He has given an admirable transcript of the Median text of the Behistun inscription in eight plates, accompanied by a transliteration that shows a remarkable improvement on all previous efforts, though of course it has since undergone considerable alteration. His translation follows the rendering of the Persian column made by Rawlinson, and has received little alteration from later scholars. He brought his essay to a close by giving the Median transliteration and translation of the remaining inscriptions, including two that were lately found by Mr. Loftus at Susa, and which now appear for the first time. The unilingual at Persepolis was at length intelligibly rendered. Westergaard had made out its general purport; but even this imperfect result escaped the efforts of De Saulcy.[662] The last paragraph was, however, found one of unusual difficulty, and Norris suggested two alternative versions:

or

In 1879, Oppert reads the same passage:

The latest attempt was made by Weisbach, in 1890, and runs thus:

The work of Norris excited some controversy, and Holtzmann was especially concerned to refute the Finnic-Tartar hypothesis. M. Haug revived the theory of its closer relationship to Turkish, proposed by De Saulcy, and he suggested that the Persians themselves were originally Tartars.[664]

M. Oppert had now become the chief representative of cuneiform studies upon the Continent. His essays on the Persian column of the inscriptions (1851-2), evinced a complete mastery of the subject and considerable independence in the treatment of doubtful passages. The reputation he had earned led to his being attached to the French scientific expedition to Babylonia. On his return he undertook to write an account of the expedition, and his second volume, which made its appearance in 1859, contains an elaborate account of the work of decipherment.[665] Although the book is chiefly concerned with Assyrian, he has given a Syllabarium of the Median, with the object chiefly of comparing it with the Babylonian and Assyrian systems of writing.[666] The great importance resulting from such a comparison now becomes apparent. The values of the Assyrian signs were already ascertained in a large number of cases, and it was recognised that, with some exceptions, the similar sign in Median generally expressed the same value. The principle was also definitely admitted that each sign has only one value,[667] and that an independent sign may be looked for to express the combination of the vowels a, i and u before and after each of the principal consonants, k, p, t, m, r, l and s, and therefore we may expect a sign for each of such forms as ka, ki, ku, ak, ik, uk, and so on. The application of this system enabled Oppert to make a very decisive improvement in the Syllabarium. It will be recollected that Norris had twenty-four wrong values, and twenty-one only approximately correct. Oppert now corrects twelve of the former and eighteen of the latter. The result was that he could dispose of eighty-three values absolutely and six nearly correct. He was doubtful as to the sound of one, wrong as to seven, and he omitted seven. Thus, when the determinatives are added, all of the 106 signs in Menant’s list were deciphered except fifteen. He also added a fourth to the list of determinative signs. He showed that the one mistaken by Norris for s was in fact used simply to indicate that the following letter was an ideogram (No. 66 of Hincks). It had long ago been observed that a single sign was employed for ‘king,’ and Norris added another for ‘month.’ Oppert points out that the determinative before ‘god’ was also an ideogram for ‘god,’ and that ‘man,’ ‘water,’ ‘animal,’ or ‘horse’ and ‘road’ were likewise indicated by ideographic signs. He considered that the grammatical forms show analogy first to Magyar, then to Turkish, Mongol and Finnish. He gave it the name of Medo-Scythic, and he now considered that it was spoken by the tribes at Persepolis and Behistun—more particularly by those in the north of Media. The student of the second column had at his command some ninety-nine proper names, besides a large number of Persian official terms and titles transliterated into the Median script; and with their assistance the pronunciation of about a hundred and forty words was already known.

The remarkable success of Oppert was due almost entirely to the successful comparison of the Median and Babylonian signs; and it had something of an accidental character, for it appeared in a work chiefly devoted to Assyrian and without special reference to this particular branch of the subject. Indeed the attention of scholars was now so thoroughly absorbed by the study of Assyrian and the many new discoveries it opened to their view, that the second column fell into comparative neglect. Mordtmann wrote papers upon it in 1862, and again in 1870, in the ‘German Oriental Gazette,’ in which he appears to have ignored the results already attained, and to have given different values to some of the signs. He called the language Susian, in consequence of the order in which the provinces of ‘Persia, Susiana and Babylon’ occur in the Behistun inscription, and also because Susa bears an entirely different name in the Median from that given to it in the Persian, while the other names are alike in both. In support of his opinion he was the first to show that the inscriptions on some bricks found at Susa, which were then beginning to attract attention, though written in a different dialect, were evidently similar in speech and writing to the second column.

It will be recollected that Rawlinson visited Susa in 1836, and observed a few bricks and a broken obelisk bearing the peculiar inscriptions to which we have just referred. He considered the style of writing to be ‘the farthest removed of any from the original Assyrian type,’ and he surmised that the language is ‘not even, I think, of the Semitic family.’[668] In 1852, Loftus collected a few other inscriptions in the same character and language, which were sent to Rawlinson. Mr. Norris, who announced this acquisition in his Median Memoir, stated that Rawlinson still thought that the characters were those of ‘the Assyrian alphabet,’ but in a different language, and that he had made out sufficient to show that they belonged to Susian kings who were anterior to Darius.[669]

When Mr. Layard visited the same neighbourhood, in 1841, he was fortunate enough to be able to copy two inscriptions at Malamir, one of thirty-six lines and the other of twenty-four.[670] Rawlinson, in his classification of the cuneiform inscriptions, called them the Elymaean, and from the differences they presented, he considered that they ‘are entitled to an independent place,’ apart from Babylonian or Assyrian. In 1850, he again points out their dissimilarity from either of the two last mentioned, but he adds that they are not so difficult to read as those he had found on the bricks at Susa.[671] The surprising discovery of De Saulcy that the Median and Babylonian characters are ‘identical,’ notwithstanding their apparent diversity, naturally stimulated the ingenuity of other writers to widen the sphere of the ‘identical’; and Mordtmann was among the first who laboured in this direction. The work was continued by Lenormant, who made his appearance within the circle of cuneiform scholars in 1871, by the publication of his first series of ‘Lettres Assyriennes,’ followed, in 1873, by the ‘Choix de Textes.’ The result of the minute comparison he instituted was to show that the Old Susian script closely resembled Old Babylonian, while the Elymaean of Malamir is simply an earlier form of the Median or New Susian character. The development towards ‘identity’ had now gone so far that Bertin describes the difference between Old Susian and Old Babylonian as very slight, while Elymaean and Median are simple ‘variants of the same.’[672]

Mordtmann and Lenormant were successful also in showing that both these two newly-discovered languages, or dialects, are closely related to Median, and belonged therefore to the Scythic family. The remarkable discovery that had recently been made that a Scythic language—the Akkadian—was the primitive speech of Chaldæa, gave a very unexpected extension to the range of the Turanian races; and it was now beginning to be recognised that the civilisation of Western Asia is to be referred to them, and not, as heretofore supposed, to a Semitic people. The effect of these discoveries was to stimulate once more the flagging interest in the writing of the second column, and efforts were now directed to determine the nature of its relationship to the newly-found dialects, and more particularly to ascertain the people to whom each might be attributed. The first to enter upon this new field of inquiry after Lenormant was Oppert, who submitted a tentative translation of an Old Susian inscription to the Congress of Orientalists in 1873. In the following year, Mr. Sayce attempted two short inscriptions published by Lenormant. If we may judge by comparison with a later version given by Oppert, no great measure of success was yet attained.[673] Indeed three years afterwards, Oppert himself admitted that the Susian could not yet be read. The inscriptions of Malamir caused less difficulty, and Sayce declared that it was ‘the same as the Median with a few unimportant variations.’ It must be confessed, however, that his subsequent analysis tended to show that these ‘variations’ had a considerable range both in grammar and vocabulary.[674]

Meanwhile the language of the second column continued to receive a great variety of names which has produced immense confusion. Lenormant, however, heroically adheres to ‘Median,’ because one of the Median tribes is specially distinguished as Aryan, and it was therefore reasonable to suppose that the mass of the people were of different, and presumably of Turanian, race. Professor Sayce admits that it must have been ‘the vernacular of the lower classes of Persia: in other words, of the Medes’; but he preferred to call it Elamite, ‘as less likely to lead to ambiguity and misconception.’ He suggests that Amardian would be still better, for Susiana is always called ‘Khapirti’ in the Median text; and this is evidently the same as the country of the Amardians of Strabo. He showed that its relation to the dialect of Malamir was closer than to that of Old Susian; and he thought there was no doubt that the latter was related to the Akkadian.

One of the most interesting parts of Professor Sayce’s essay was the publication of ‘a revised list of the powers of the Elamite’ (i.e. Susian) ‘character, which can now be determined by means of the Assyrian syllabary.’[675] Written in 1874, this document illustrates the process of development that occurred between the two works of Oppert, that of 1859 and the one of 1879, to be mentioned shortly. It will be recollected that we left him with seven wrong values, six nearly correct and seven omitted altogether. We find that when Sayce wrote, four of the first and one of the second had been corrected, and three signs omitted by Oppert were now provided with correct, or nearly correct, values. At the same time a plentiful crop of fresh errors was introduced. Six or seven values, correctly ascertained in 1859, were now rejected and erroneous ones substituted. At least three of the values proposed for the signs omitted by Oppert were very far indeed from the mark. The ideogram for ‘horse’ was rejected in favour of the syllable az, which may have suggested to Weisbach the substitution of ‘donkey’ in place of the nobler quadruped preferred by Oppert.

The two writers who have brought the knowledge of the Median syllabary to its present standard are Oppert and Weisbach: the former in a special treatise written in 1879 (‘Le Peuple des Mèdes’) long remained the leading authority on the subject, and his conclusions have in the main derived confirmation from the more recent investigations of Weisbach on the language of the second column, which appeared in 1890. With very few exceptions, to be noticed later, both scholars are in substantial agreement as to the values of the signs; and it is this agreement that forms for the present the standard of right and wrong, by which the efforts of their predecessors have been judged. Both writers were guided to a large extent by the values of the corresponding Babylonian characters.[676] Oppert, as was said, thought he could trace a resemblance between ninety-six of the Median signs and their Babylonian equivalents. In each case, the sound as well as the form of the character was appropriated. Weisbach is much less struck by the general application of this law. He fully admits the strong similarity of many of the signs, but some have, he says, so far diverged from their original types as to be hardly recognisable. Others, he thinks, were borrowed from New Assyrian, and a few from older forms. Indeed he is, on the whole, indisposed to derive the syllabary direct from the New Babylonian. He thinks it is more properly to be traced through the writing of Malamir to the Old Susian, and the development from the Old Babylonian of the latter was a parallel and independent process similar to that which produced the New Babylonian. But as the Old Susian character has not yet been sufficiently investigated, he restricts his comparison to the New Babylonian, and he points out some of the principles that were followed in the evolution of the new script. For example, the vertical wedge that crosses the horizontal in the Babylonian is generally placed before them in Median, and the same rule applies to the horizontal crossing the vertical. There is an evident effort to simplify both the writing and the language. The signs preserve the same signification in both, but when it happens in Babylonian that the same sign has many different values (sometimes no less than nine) it has never more than two in Median. So also the number of homophones, or different signs with the same values, are strictly limited. Indeed, he considers that the Median was an early effort to approach an alphabetical system.[677]

We have seen that several of the errors made by Oppert in his list of 1859 were corrected from various sources before 1874.[678] He was still left with three wrong values, five only nearly correct and four omitted altogether. Of the first class he now gives one a value that accords with that of Weisbach—11ᵇ, ‘ko’ for ‘kam,’ the ‘gau’ of Norris and Weisbach.[679] Of the second class he corrects four,[680] and he supplies correct values to three of those previously omitted.[681]

He now presents us with a list of a hundred and twelve distinct signs, and no less than fifteen ideographs; but when their syllabic values are also known the majority are included in the hundred and twelve signs. Four, however, appear among the ideographs for the first time, and raise the total number of signs to a hundred and sixteen. Six of these are,[682] however, repeated twice over to express different syllabic values, thus reducing the number of distinct signs to a hundred and ten. He accepts a hundred and five of the hundred and six signs that were already known, rejecting only one (No. 21). He completes his number by the addition of five other signs,[683] of which Weisbach has accepted three. With these two exceptions, the whole of Oppert’s signs are to be found in Weisbach: that is, a hundred and eight out of the hundred and ten. Weisbach, however, includes in his list the No. 21 of Hincks, omitted by Oppert.

The two writers are also substantially agreed as to the values of the signs. Of the hundred and six given in Menant’s list, we find they differ only as to seven. Of these, two are omitted by Oppert[684] and one by Weisbach.[685] In four cases only have they arrived at absolutely contradictory values.[686]

The discovery announced in 1859 of a determinative sign to indicate that the one preceding it should be read ideographically led to the identification of a number of signs to which ideographic values may be attached, and their number is raised from the seven known in 1859 to sixteen. The ones now added by Oppert are ‘town,’ ‘mountain,’ ‘race,’ ‘arch’ (a window), ‘sea,’ ‘house,’ ‘head,’ ‘ship,’ ‘camel’; and Weisbach has since contributed a fifth determinative which, he says, is placed before articles of wood.[687]

The difficulty of the transliteration is greatly enhanced by the fact that the same sign represents both m and w, and the three gradations of sound—the surd, aspirate and sonant—are not distinguished by separate signs. It is therefore impossible to say in many cases whether to transcribe m or w, k or g, p or b, f or d; and Oppert has enumerated no less than six different modifications of sounds—č (tch), ḡ (dj), z̆, ts, dz, and z—that are all represented by a single sign.[688] Weisbach points out that g, is only clearly distinguished from k when it occurs before i. The Median z represents the Persian z, c, j, and the Babylonian s (and o?); the r and n are sometimes found to be interchangeable. When the sound cannot be checked by its occurrence in a proper name, Oppert generally adopts the hard sound; but he allows himself a certain latitude in the application of this rule, and he shows a decisive preference of w to m. Weisbach follows the uniform practice of transcribing the gutturals, dentals and labials by the tenues; he always uses z for the Persian c, j and z, and for the Babylonian z and s; and, contrary to Oppert, he prefers m to w.

Oppert recognises the five vowels a, e, i, o and u; and, following De Saulcy, he admits yi, ya and ah in his list, among vowels. Weisbach excludes the o, for he considers there was no difference of sound between it and u; the y he includes among the semi-vowels, with both r and l. Although Oppert found that all the five vowels follow the consonant to form the single syllable, he considered that only a, i and u precede it. Both writers agree to limit the consonantal sounds to about eleven. Oppert was the first to treat the grammar with elaborate care, and when he claimed that his work was ‘une création entièrement nouvelle,’ the statement was probably more strictly accurate than he imagined.

The nouns do not admit of any distinction of gender, and have only the singular and plural number. Oppert distinguishes no fewer than twelve different cases, indicated by suffixes, a profusion limited by Weisbach to eight, viz. nominative, genitive, accusative, dative, ablative, allative, locative and comitative.

Oppert, however, did good service in unravelling the mysteries of the verb, though his passion for systematising and ‘restoration’ has carried him too far; and his verb, declined through all its moods and tenses, presents a very different appearance from the skeleton which, according to Weisbach, is all that can be strictly collected from the texts themselves. For example, we are presented with the six persons of a complete past tense, where Weisbach can only find authority for three; and we get a complete imperfect, although only one termination is really known, viz. that of the first person singular. With no less confidence we find two verbs—a reciprocative (‘je me sus’), and an ‘intensive’ (‘savoir bien’)—of which Weisbach can see no trace; and the same remark may be applied to his desiderative (‘je veux savoir’) and his factitive (‘je fais savoir’).[689]

Both writers agree as to the personal pronouns; but in the possessive Weisbach can only find the third person singular and the first person plural, while Oppert supplies us with the series complete.[690]

Weisbach calls attention to the dialectical differences in the Naksh-i-Rustam inscriptions, and to the evidence of decay visible in the language of Artaxerxes at Susa.

For the reasons already mentioned there is considerable diversity observable in the transliterations made by the two writers, but so far as we have observed they are in substantial agreement with regard to the meaning of the text, as may be seen by a comparison of their rendering of the unilingual inscription.[691] Occasionally, however, Weisbach finds himself unable to follow the more imaginative flights of his predecessor. He will have nothing to do with the ‘restoration’ of the concluding paragraph of the Suez inscription of Darius.[692] He is equally unable to accept the interesting completion of the detached inscription at Behistun, marked L by Norris. Norris reads: ‘I made another tablet in the Arian language, such as did not exist before, and I made a large ... and a large ... and ... and....’ This not very promising attempt is perfected by Oppert as follows: ‘I have made also elsewhere a book in Arian language, that formerly did not exist, and I have made the text of the Divine Law (Avesta) and a commentary of the Divine Law and the prayer and the translation.’ He observes truly that ‘the passage is of first-rate importance,’ and he adds encouragingly: ‘The explanation which I give is sure.’ It is therefore somewhat disconcerting to find that the latest writer cannot get much beyond the crudity of Norris. Weisbach reads: ‘Machte ich Inschriften in anderer Weise (?) (nämlich) auf Arisch, was vormals nicht war, und das grosse ... und das grosse ... und das ... und das ... machte ich.’ Alas for the Divine Law and the commentary and the prayers![693]

We have already said that in 1859 Oppert had abandoned the hypothesis that the language of the second column was introduced by the Scythic hordes who were expelled by Cyaxares. He recognised that it was connected with the ancient Akkadian, and could not, therefore, be a new importation, but an original native dialect. He considered that the Median name itself is Turanian and related to ‘Mada,’ the Akkadian for ‘land.’ He thought it is clear from Herodotus that the six tribes who composed the Median nation were not all of the same origin. The dominant caste were, he still held, of the same Indo-European race as the Persians, but a large part of the population were Turanians, and the language of the second column was that of the agricultural and nomadic tribes of Media, especially those of the north. He accordingly gave it the name of Medo-Scythic to distinguish it from the Aryan Median of the classical writers.

This very tenable hypothesis he, however, abandoned in its turn, and advanced one much more hazardous. He relies upon a fable of Herodotus to show that the Aryans occupied the country from a much earlier period than has been commonly supposed, and did not, therefore, make their first appearance with Dejoces. Both Aryans and Turanians, he now thinks, were long settled together in the same country, and it is impossible to say which of the two were the first comers, though he inclines to give precedence to the Aryans. The various tribes were upon an equality, sometimes the Turanian and sometimes the Aryan gaining the ascendency. He considers the former were known as the Medes, a word essentially Turanian in its origin, while the others retained their proper designation of Aryans. He holds that the Median dynasty of Dejoces was Turanian. He carefully analyses the names of their kings, and he has succeeded in affording a fresh illustration of the peculiar power of philology to prove any thesis whatever, when employed by a skilful manipulator. Not many years before, he laid it down as self-evident that these same names were pure Aryan.[694] Now it becomes no less apparent that they are pure Turanian. The dynasty of Dejoces marks the ascendency of the Turanian Medes, and the language of the second column is that which was spoken by them. He accordingly drops his previous qualification of Scythic, and gives it simply the name of Median. The rise of the Persian power enabled the Aryan Medes to recover the position they had temporarily lost, and hence all the names that occur from the time of Darius clearly belong to that race. Mada became a geographical name which embraced the whole population of the country now under an Aryan aristocracy, and Herodotus was therefore fully justified in speaking of the Medes and Persians as one in speech and descent. The theory of the Turanian origin of the Median dynasty has been almost universally abandoned.[695] The attack upon it was led by the Jesuit scholar Delattre,[696] and the whole controversy has been ably summarised by Weisbach. Recent writers have thought it so necessary to insist upon the Aryan race of the Median kings that they decline to give to the Turanian language of the second column the name of Median. It has accordingly come into fashion to indicate in an unmistakeable manner the source from which it has sprung. Delattre called it ‘Anzanisch,’ from the name of the territory ruled by the Malamir kings. Halévy adopted a suggestion made by Mr. Sayce, and calls it ‘Amardian.’[697] Hommel speaks of it as ‘Susian-Median’ and ‘Susian,’[698] a term which Weisbach has qualified by calling it New Susian. The more probable opinion seems to be that it was the language of Susa at the time of the Persian conquest, and possibly also of some of the subjugated tribes in Media. The name of Susian is therefore more appropriate than one that might confuse the people who spoke it with the Aryan conquerors of their country. Still it is very far from satisfactory. The great importance of the inscriptions recently discovered in the Old Susian language will tend more and more to reserve to them the designation of Susian; and considerable confusion will arise from its extension also to the language of the second column. The latter may possibly be a descendant of the true Susian, but both in the system of writing and in the language the connection is remote.

The relationship of the Median is now placed almost beyond the sphere of controversy. M. Gobineau, who wrote in 1859, maintained indeed that it was connected with Pehlevi, half Semitic and half Aryan: and M. Mohl still earnestly hoped that we might ‘get rid of the Scythic hypothesis and all the complications it involves.’[699] But this desire was not destined to be realised, and its affinity to the Altaic branch of the Turanian family is now admitted. Some doubt is felt as to whether it has left any successor, and which of the modern languages approaches the nearest. Oppert inclines to Turkish:[700] Weisbach is more guarded, and considers that it exhibits marked differences from all the living representatives of its Turanian relatives.[701]