Old Persian Language and Script

Three distinct things travel under the name Old Persian. The first is a language: a southwestern Iranian tongue, spoken in Fārs, related to the other Iranian languages of the empire and ultimately to Sanskrit, Greek and Latin. The second is a writing system, a cuneiform syllabary of about forty signs used for that language and, on present evidence, for no other. The third is a register: the formal, heavily patterned idiom of royal proclamation in which nearly every surviving text is cast. Most long-running disputes about Old Persian turn on how language, script and royal register relate to one another.^[1][3]

The association between the language and its cuneiform script is not exclusive. Baghbidi, following Herzfeld and citing Frye, interprets an effaced inscription beside the entrance to Darius's tomb as Old Persian written in Aramaic letters. He reports the sequences ḥšʾyty wzrk and mʾhy, understood as Old Persian xšāyaθiya vazṛka, “great king”, and māhyā, “in the month”, and assigns the text to Artaxerxes II or III.^[11] Kent had catalogued the twenty-five lines below the Elamite panel as an Aramaic version. Baghbidi's interpretation separates the language of those lines from the alphabet used to write them.^[12]

Production of an inscription also involved different skills. In discussing the detection of modern forgeries, Schmitt distinguishes the scholarly knowledge available to a forger from the copying of a prepared exemplar by an ancient engraver. That comparison concerns the preparation and reproduction of a text; it does not establish the linguistic competence of every individual who worked on a royal monument.^[13]

Whether the inscriptions are composed in a literary idiom, and whether anyone spoke it, has been contested for a century, and the positions do not line up neatly.

Kent allowed that the inscriptions contained recognisable stylistic features, while judging that Old Persian “had no developed literary style” and that “probably many of its fashions of expression are taken over direct from Aramaic, lingua franca of the ancient Orient”. His discussion goes on to identify features of expression, omissions, chiasmus and rhyme. Schmitt's later assessment places this restricted treatment beside Meillet and Benveniste's rejection of literary form in the inscriptions.^[14][15]

Schmitt's study of 2016 exists to overturn that judgement. He calls a systematic treatment of the figures of speech in the Old Persian corpus an urgent desideratum, judges Kent's two paragraphs far too brief, and states the conclusion directly: the inscriptions stand apart from plain prose through the use of particular stylistic devices, which, along with certain words and word-forms, lend them kunstsprachliche Züge, the features of a crafted literary language.^[15] The book catalogues forty-three figures, from alliteration to zeugma, among them chiasmus, ring composition, merism, polyptoton, rhyme and the law of increasing members. Some are unremarkable in any prose; others are hard to reach by accident. Behaghel's law, which places the bulkiest member of a series last, Schmitt takes as a general mark of elevated speech, exemplified by hamātā hamapitā, which he glosses “von derselben Mutter (und) demselben Vater”. In English, the phrase denotes the same mother and the same father.^[15] The figure he calls statement plus negated counter-statement he connects to inherited Indo-European poetic diction. For DB IV 44f. he prints yaθā ima hašiyam, nai duruxtam adam akunavam and glosses it “daß ich dies wirklich, ungelogen getan habe”. For DNb 7f. he prints taya rāstam daūštā ami, miθa nai daūštā ami and glosses it “daß ich dem Recht(en) freund bin, dem Unrecht(en) (aber) nicht freund bin”.^[15]

A third position accepts the artificiality and pushes it further. Rossi holds that the inscriptional variety "is likely to correspond to a literary, artificial form" of the dialects spoken in Fārs, citing Stolper and Jacobs, and adds Schmitt's view that Old Persian as it is known was restricted to royal usage.^[1][2] Stolper had put it more sharply: the language of the inscriptions "was not what the kings and satraps spoke, or indeed what anyone spoke", but an artificial idiom drawing on several Iranian dialects and already archaic in its grammar when the texts were cut.^[16]

Tavernier, in the same Companion, calls Old Persian “the king's and the Persian elites' vernacular”, whose knowledge “was not widespread in the empire”.^[17] This describes the social distribution of a spoken language, whereas Rossi's formulation addresses the variety represented by the inscriptions. Stolper and Tavernier themselves distinguish the Persian dialect probably spoken by administrators from the literary dialect of royal texts.^[9] Van Bladel rejects the stronger claim, attributed to Herzfeld, that the language of the inscriptions was already dead in Darius's reign. Whether a text is stylistically shaped and whether its grammar reflects a living language are therefore separate questions.^[10]

The narrow range of surviving genres limits comparison between written registers. Rossi's survey records proposals for an interdialectal literary language, a lingua franca and diglossia, rather than a single agreed account of the relation between inscription and speech.^[1] The administrative tablet from Persepolis supplies a non-royal context for Old Persian writing, but its obscure vocabulary and uncertain transaction prevent it from settling the register question. It establishes an additional use of the script and language while leaving ordinary administrative prose and conversation largely undocumented.^[8][9]

Old Persian and Avestan are the only Old Iranian languages directly attested, and they are separate languages with very different corpora: one a body of royal proclamations cut in stone and metal, the other a religious tradition transmitted orally for centuries before it was written. Comparison between them, and with Indo-Aryan and the wider Indo-European family, supplies much of what is known about Old Persian morphology, and it supplied the readings that broke the script. A comparison still has to fit the signs actually cut and the sense of the particular sentence; a resemblance between isolated words settles nothing.^[18][19]

Greek explanations of Persian names form a separate strand of evidence. Herodotus writes: “Darius signifies the Doer, Xerxes the Warrior, Artoxerxes the Great Warrior; and such the Greeks would rightly call them in their language.”^[20] These are Herodotus's renderings for a Greek readership. Reconstructing the Iranian names requires comparison with their inscriptional and documentary forms, rather than treating the Greek glosses as readings of Old Persian signs.^[19]

Most Old Iranian vocabulary is known only at second hand. Tavernier's dossier of Iranica sets out the category: proper names and loanwords "reconstructed on the basis of their reflections in the target languages", recovered from Elamite, Akkadian, Aramaic, Egyptian and Greek transmission rather than from any Iranian text. He notes that beyond Old Persian and Avestan "various dialects are indirectly attested", their existence inferred from philological features that separate them from both, and that Old Persian cuneiform is the only contemporary Old Iranian script.^[19] The consequence for the lexicon is severe: an everyday Persian word may be recoverable only because an Elamite scribe at Persepolis wrote down a Persian name or technical term.

“Median” is conventionally used for Iranian forms whose phonology differs from what is regarded as typical Old Persian. Kent describes a Median influence on the inscriptional language and groups the proposed examples by their sound changes and their occurrence in geographical names, personal names, royal titles and other vocabulary.^[21] Rossi questions how securely that category identifies a particular language: in his account it is a conventional label for indirectly preserved remnants, not a separately recovered Median corpus (see the Medes).^[22] Stolper offers an interpretation of the mixed vocabulary in terms of royal identity, suggesting that non-Persian Iranian forms helped the kings present themselves as Iranian as well as Persian. That is an explanation of the choice of forms, distinct from establishing their linguistic origin.^[16]

Elamite and Akkadian stand in a different relation again. Elamite is not an Iranian language and has no established relatives; Akkadian is Semitic. Three languages in cuneiform on one monument represent three writing traditions and three grammars, not three ways of spelling the same speech.^[1][2]

The corpus is dominated by royal texts on durable materials. Rossi's survey gives the distribution: most inscriptions come from Persis (Persepolis, Naqš-e Rostam, Pasargadae), from Elam (Susa) and from Media (Bisitun, Hamadan); from outside the central regions there are three inscriptions of Darius from the Suez canal, a rock inscription of Xerxes at Lake Van in Armenia, and fragments from Babylonia.^[2] The geographical pattern records acts of royal commemoration; by itself it cannot map the distribution of Persian speech.

The genres are few. A text may name the king and his ancestry, list the lands, record a building and the materials brought for it, invoke Auramazda, set out the qualities expected of a ruler, or, once, narrate a year of civil war at length. The Bisitun inscription is the great exception in scale. Schmitt's editions count five columns of Old Persian totalling 414 lines in seventy-six sections, against three surviving Elamite columns of 260 lines covering only the first sixty-nine sections, and a single Babylonian column of 112 very long lines; a further set of short captions runs from DBa to DBk, the last of them bilingual in two lines each.^[23][24] The thirty-six-line fifth column exists in Old Persian alone, which is a plain demonstration that a version could be produced in one language and not the others.^[24][2]

Darius the king proclaims: This (is) what I have done in the second and third year, after I became king: a country called Elam became rebellious; a man called Athamaita, an Elamite – him they made their chief. After that, I sent an army; a man called Gaubaruva, my servant – him I made its chief. Then Gaubaruva went to Elam with the army; he joined battle with the Elamites. Afterwards, Gaubaruva defeated the Elamites and decimated them; and he took their leader prisoner, brought him to me. After that I killed him. After that the country became mine.

Darius I, DB §71^[5]

Monolingual, bilingual, trilingual and four-language texts all occur. Baghbidi classes the inscriptions as sometimes monolingual, often in Old Persian, sometimes bilingual, mostly trilingual and "rarely quatrilingual (Old Persian, Elamite, Babylonian, Hieroglyphic Egyptian)".^[25] A surviving example is the stone jar published by Westenholz and Stolper: an Egyptian text in hieroglyphs on one shoulder, dated to Darius's thirty-sixth year, and on the opposite shoulder a different text in Old Persian, Elamite and Akkadian, each giving the king's name and title.^[26] The editors call the object familiar and its texts familiar, while the combination of four versions on a piece from Darius's reign is unique; they also warn that the two sides may have been inscribed at different times and in different places.^[26] Rossi adds the hieroglyphic texts on the statue of Darius from Susa.^[2]

Quantity is unevenly distributed even within one text. Kuhrt records that the Susa foundation charter survives in "thirteen pieces of the Old Persian text (one almost complete), twelve Elamite and twenty-seven Akkadian", on clay tablets, barrels, stone tablets and glazed bricks from numerous places at Susa.^[27] Repeated copies of a known text enlarge the manuscript base without enlarging the vocabulary.

What the corpus contains shapes what can be known. Royal first-person statement supplies forms for "I", "king", "country", "made" and "protected" many times over, while whole domains of ordinary life go unrepresented. Kent's grammar says as much of itself: some categories are "scantily represented", and for each class "all extant forms are listed", except where examples are numerous.^[18] van Bladel puts the same limit more starkly for the nouns: "Not one declension is completely known."^[28]

The phonetic inventory contains thirty-six signs. Kent divides them into three vowel signs, a, i and u; twenty-two consonant signs with inherent a; four with inherent i, di, ji, mi and vi; and seven with inherent u, du, gu, ku, mu, nu, ru and tu. The thirty-three consonantal signs thus do not provide a complete set of vowel-bearing alternatives for every consonant.^[3] Weissbach describes the distribution by consonant: only d and m have all three forms, two others have a- and i-bearing signs, and k, g, t, n and r have a- and u-bearing signs. Baghbidi likewise counts thirty-six phonetic signs.^[29][25]

Most consonants have only the sign conventionally described as carrying inherent a. Reading a sequence requires the orthographic rules as well as the sign values, since the written combination can represent more than one sequence of sounds. The description as a syllabary refers to these consonant–vowel values; the relatively small repertoire explains why alphabetic comparisons also recur. Neither terminology removes the need to reconstruct the phonological form from the spelling.^[3][30]

Variation in the shape of a sign must also be distinguished from a new sign value. Schmitt identifies particular forms of y and of the divider at Bisitun, using their difference from later inscriptions in his argument about the monument's position in the history of the script. Such observations depend on comparison of particular inscriptions rather than on an undifferentiated count of characters.^[31]

Word division is one respect in which the system is notably disciplined. Weissbach observed that word separation in Old Persian is carried through with great accuracy, and that the script is much the better for it against Elamite and Babylonian, neither of which has a word divider at all.^[29]

Besides the phonetic signs, Kent identifies five lexical word signs: XŠ for xšāyaθiya, “king”; DH for dahyāuš, “land”; BU for būmiš, “earth”; AM for Auramazdā; and BG for baga, “god”. He adds “what seems to be a ligature for AM-ha” at DSk 4. Thirty-six phonetic signs, five word signs and this ligature total forty-two, before the divider and numeral signs are included.^[3][4] The alternative count of eight logograms treats the two graphic forms of DH and the two of AM separately and includes AMha. Kent's chart displays the variants, while Baghbidi uses the eight-logogram count. The difference concerns how lexical identities, graphic alternatives and the ligature are counted.^[3][25]

A word sign could be completed phonetically. Kent records that an ideogram without added syllabic characters stands for the nominative singular, while other forms are indicated by writing the last character or characters of the full word after the ideogram.^[4] The device identifies the word and its ending at once, so two copies of the same text may reach the same Persian word by different graphic routes, and the presence of logograms never suspends the grammar.

The assignment of the divider's two shapes differs between editions. Kent writes that it “in the Behistan text has the form of an angle with the point to the left, and in other texts is a single slanting wedge running from upper left to lower right”.^[4] Schmitt's description reverses that distribution: in DB “the word divider is not an angle, as in all later texts, but an oblique wedge half as high as the line”; he also identifies a distinctive first vertical wedge in the y sign.^[31] Kent discusses where dividers occur, including the beginnings of columns and sections at Bisitun, and exceptions in the Susa texts. The administrative tablet provides a further specific convention: no divider separates its numerals from the following words.^[4][8] Weissbach's account likewise records variation in the treatment of numerals. A general statement that the script separates words therefore requires these local qualifications.^[29]

Kent describes numerical forms for 1, 2, 3, 10, 20 and 100. One is a long vertical wedge; two consists of two short verticals; three adds a long vertical on their right. Ten is an angle with its point to the left, twenty two such angles, and one hundred two short horizontal wedges meeting above a vertical. Smaller units stand to the right of larger ones. Cardinals are normally written with these signs, although aiva-, “one”, is spelled out in a formulaic phrase. Kent adds: “But the ordinals are written in full, with the regular characters.”^[4] Stolper and Tavernier compare the administrative tablet's possible year formula with Bisitun's duvitiyāmca çitiyām θardam, “in the second and third year”. Their interpretation of the tablet's final form remains conditional.^[8]

A modern encoding of the repertoire exists and is worth distinguishing from the ancient one. The Unicode Old Persian block encodes thirty-six phonetic signs at U+103A0 to U+103C3, eight logograms at U+103C8 to U+103CF, one word divider at U+103D0 and five numerals at U+103D1 to U+103D5, fifty characters in all.^[32] The arrangement follows the eight-logogram convention and is a decision about digital text rather than an independent witness to what Achaemenid scribes recognised as a sign.

The signs underdetermine the language. Stolper put the ambiguity in figures: with only a small, consistently applied repertoire, "the five characters that spell the verb 'is' could be read, in theory, in any of seventy-two ways", of which a Persian speaker would reject seventy-one as nonsense; reading the word correctly as asti requires the language as well as the script.^[16] The observation cuts against any account in which learning the signs amounts to learning to read.

Nasals before consonants are regularly not written, and the standard normalisations show it. Kent prints the name of Cambyses with the unwritten nasal raised as a superscript letter within the normalised transcription and with no length mark on the u; other editions normalise the same name by other conventions, so a macron on it belongs to whichever edition supplies one.^[30] Vowel length is also unevenly represented, and the value of a vowel sign can depend on the preceding consonant sign. A normalised form is therefore an argument about the language, reached from the signs by rule and comparison.

Kent is explicit that the three operations differ: transliteration records the signs, normalisation proposes a linguistic form, translation gives the sense; and "the normalized text is not necessarily a phonetic text, but only an approximation to such a text".^[30] Two editions may print different normalisations of a word while agreeing entirely about what the sentence means. For historical reading, the practical warning concerns names: Darius, Xerxes and Hystaspes reached English through Greek and Latin, and their Old Persian forms need not resemble them.

The script also carries diagnostic information about the language written in it. Stolper and Tavernier identify the language of the Persepolis tablet as Persian rather than another Iranian dialect on a single orthographic ground, the use of the character ç rather than θr, and report Schmitt's observation, made in a personal communication, that its consonant-plus-vowel sequences conform to Old Persian orthographic rules.^[8] Spelling conventions, applied consistently, identify a dialect even where the vocabulary is obscure.

Old Persian shares the wedge-shaped elements of the surrounding cuneiform traditions. Whether particular signs were derived from Mesopotamian prototypes has been a separate question. Weissbach's 1911 edition reviews a series of attempted derivations and argues against their proposed correspondences. His treatment is evidence for the history of that debate, not a reason to regard every later formulation as settled in advance.^[29]

He begins from the shapes themselves. In Old Persian the oblique wedge running from upper left to lower right occurs only as the word divider, and the wedge running from lower left to upper right not at all; no vowel or syllabic sign has fewer than two or more than five elements, or more than four horizontals, three verticals and two Winkelhaken; crossing wedges occur exactly once, in the sign vi.^[29] These are the formal constraints from which Weissbach develops his argument, rather than a demonstration by themselves of how the repertoire was invented.

The derivations went ahead regardless. Grotefend in 1802 and Oppert in 1859 compared the Old Persian king-ideogram with its Babylonian counterpart; Menant in 1869 went further; and Oppert in 1874 built a complete acrophonic system, on which the inventors selected thirty-six Babylonian ideograms, translated their meanings into Old Persian, and took the initial syllable of each Persian word as the new sign's phonetic value. Halévy rejected the scheme in detail in 1885, and Weissbach dismantled it column by column: of the seven columns of Oppert's comparative table only three are unobjectionable, about two-thirds of the ideographic values in the third are wrong, and of the thirty-six Old Persian words in the sixth only seven are actually attested. Scarcely a line of the table, he concluded, escapes an objection of that kind.^[29] His criticism extends to the rival schemes: Deecke in 1878, Sayce in 1884 and Halévy agree on only seven of the thirty-six signs; Peiser in 1900 compared fifteen signs, in nine of which the Old Persian reading differs from that of the supposed Babylonian ancestor; Hüsing's derivation through Parsua and Media rests on two unknown quantities; Hommel in 1904 traced the script instead to a cursive alphabet akin to Phoenician and Aramaic. Fossey in 1904 held the Babylonian origin to be indisputable while judging the individual derivations arbitrary, since almost any Babylonian sign could be turned into an Old Persian one by sufficient force.^[29]

Weissbach's verdict on the whole literature is blunt: the attempts of Oppert, Deecke, Sayce, Halévy, Peiser and Hommel have all alike failed, and "Regeln, die fast nur aus Ausnahmen bestehen, können als Regeln nicht anerkannt werden" — rules consisting almost entirely of exceptions cannot be recognised as rules. His own conclusion is that the script, as it now presents itself, can only be regarded as "ein selbständiges, frei erfundenes, künstliches Gebilde", an independent, freely invented, artificial construction which borrowed from the older cuneiform systems nothing but the elements of writing, the wedge and the Winkelhaken.^[29]

Later descriptions continue to distinguish the relationship at different levels. Kent says that the syllabary “quite obviously goes back to” the Akkadian cuneiform syllabary, while adding that “its simplicity as compared with its parent syllabary shows that it has been specially drawn up for its present purpose”. His footnote refers to Weissbach's critique of the proposed conversions.^[3] Stolper instead states that the signs “have no formal connection with the cuneiform characters used to write Akkadian and Elamite”.^[16] These formulations differ over graphic descent. They need not imply disagreement that Persian scribes worked in a region where wedge-writing was long established. An account of the script's origins must specify whether it proposes the borrowing of sign forms, of phonetic values, or of the general practice of writing in wedges.^[29]

The question is usually approached through the last paragraph of Bisitun, and the paragraph is damaged.

Darius the king proclaims: By the favour of Auramazda, this (is) the form of writing, which I have made, besides in Aryan. Both on clay tablets and on parchment it has been placed. Besides, I also made the signature (?); besides, I also made the lineage. And it was written down and read aloud before me. Afterwards, I sent off this form of writing everywhere into the countries. The people strove (to use it? abide by it?).

Darius I, DB §70^[5]

Kuhrt's translation marks uncertainties in “signature (?)” and “strove (to use it? abide by it?)”. Kuhrt’s commentary nevertheless describes the interpretation of the passage as the first creation of a script for Old Persian as widely, though not universally, accepted. Uncertainty over the wording of particular clauses therefore coexists with a substantial scholarly case for a statement about a new script; it does not by itself disprove that interpretation.^[6]

The dispute is old. Rossi reports that argument over the interpretation of the passage, with its Elamite counterpart, "has been going on for more than 100 years, and there is no consensus to date".^[7] Schmitt, whose edition is standard, restored the key phrase as ima dipiciçam taya, "this (is) the form of writing, which", and read Darius as speaking first of the script's shape and then of its inner form, which made it capable of rendering Aryan. He also closed his treatment with a disclaimer: "As a last remark I would like to add that I myself regard neither the restoration nor the translation given here for this last paragraph as final."^[33][6] His grounds for rejecting rival restorations are physical as much as philological: some proposals "require too much space", and Gershevitch's reading was set aside "since the traces visible on the photograph do not support such a reading".^[33]

Baghbidi disputes the identification of “Aryan” with a newly invented script. He finds “no firm basis for the prevailing opinion that 'Aryan' refers to the Old Persian cuneiform script, and thus concluding that this script did not exist before the Bisotun inscription”. He points to the other uses of ariya- for people or language, including Darius's self-description as Aryan in DNa, and to the circulation of the content in Babylonian and Aramaic.^[25][34] Rossi advances a different objection: the creation of a technical writing system would, in his view, fall outside the usual subject matter of royal discourse.^[2] These arguments address the meaning and genre of §70, whereas Schmitt combines his reading of the passage with palaeographic evidence for Bisitun's priority.^[31]

The versions are not parallel, which is part of the problem. There is no Babylonian version of the paragraph at all, although Kuhrt records that there was space for it; the Elamite exists only as the separate minor text DBl, fitted into free space on the panel above the relief because no room remained at the end of the main text, and it carries a clause the Old Persian lacks: "this is my text, which I made, in Aryan, as previously was not".^[6][24]

Rossi summarises Weissbach as supporting invention under Darius, but Weissbach's own 1911 discussion leaves the age of the script formally unsettled. Rossi writes that Darius's invention of the script "became fashionable to claim after Weissbach's siding" in 1911, and that Weissbach changed his mind before 1930.^[2] Weissbach did side that way, and he also refused to close the question: "Ich kann die Frage nach dem Alter der ap. Keilschrift noch nicht für entschieden halten" — the age of the Old Persian cuneiform script, he wrote, he could not yet regard as settled — adding only that he found it probable that Darius claimed to have been the first to make inscriptions in Aryan script and language, and that the script was accordingly introduced by that king.^[35] The negative evidence he thought firm: of the kings before Darius "besitzen wir überhaupt keine Zeile", not a single line survives, and against Herzfeld's use of the Murghab trilingual he set three points, that the only two inscriptions securely attributable to Cyrus the Great are purely Babylonian, that the short Murghab text would be the sole pre-Darius trilingual known if it were Cyrus's, and that it stands on a palace rather than a tomb.^[35]

Schmitt's palaeographic argument concerns the two characters that have distinctive forms at Bisitun. He treats those differences as supporting the claim that Old Persian cuneiform was used there for the first time.^[31] Rossi's survey preserves alternatives: Diakonoff proposed that the system might originally have served Median or another Iranian language, and Huyse critically examined the arguments for invention under Darius. Rossi presents economy of sign creation as an explanation offered by scholars who reject Diakonoff's proposal: only signs necessary to avoid ambiguity would have been created. This is a proposed rationale for the incomplete consonant–vowel inventory, rather than a documented instruction issued to its inventor.^[7][2]

One classical passage has been cited in this debate, but its pseudepigraphic date sharply limits what it can establish about fifth-century Persian writing. A letter transmitted under the name of Themistocles asks for silver mixing-bowls and gold censers inscribed with the old Assyrian characters, and not those which Darius, the father of Xerxes, had lately written for the Persians. Weissbach accepted that the letters were universally held to be spurious and relied on Herzfeld's own admission that such texts can preserve good historical notices. They cannot be treated as direct testimony from the historical Themistocles. Herzfeld nevertheless argued that a Greek could not have told Persian cuneiform from Assyrian and that the new script Darius gave the Persians was Aramaic.^[36]

Short inscriptions at Pasargadae carry the name of Cyrus, and they have to be dated one by one. Kuhrt's corpus sets them out under the heading "Cyrus the Great inscribed into Darius' lineage": CMa, trilingual, on the columns of Palaces S and P and above the winged figure of Gate R, called Palace R in Kuhrt’s account, reading "I am Cyrus, the king, the Achaemenid"; CMb, a fragmentary trilingual naming Cyrus the great king, son of Cambyses; and CMc, on the garment folds of the Palace P figure. Her note treats all three as added by Darius and claiming Cyrus for the Achaemenid line, following Stronach, who proposed re-siglating them DMa, DMb and DMc.^[37] Schmitt's nomenclature requires a distinction between two objects. He retains CMa and CMc for texts naming Cyrus, even while accepting a Darian date, and records at least eighteen CMa exemplars on Stronach's count, five surviving. Kent's and Kuhrt's CMb is his DMa (ex-CMb), a group of fragments found near Palace P whose trilingual text is tentatively reconstructed. Schmitt uses CMb for a different object, the Zendan-North relief fragment preserving two Old Persian signs.^[23] In his published survey of Iranian writing systems, Skjærvø presents invention under Darius and later addition of the ancestral inscriptions as a proposed explanation.^[38]

A minority disputes this. Rossi reports Nylander's model, on which Cyrus wrote the original texts in Elamite and Babylonian and Darius added Old Persian versions in a recently invented script, and records that Vallat and Rossi have supported the possible authenticity of all the versions, largely on palaeographic grounds, while accepting that Old Persian writing would then have remained highly restricted before Darius.^[2]

CMc illustrates the need to distinguish surviving languages from an inscription's reconstructed original form. Kuhrt presents Elamite and Akkadian versions and notes that it was probably originally trilingual. Schmitt describes a trilingual text with possible remnants of Old Persian, referring to Schaudig's edition.^[37][23] The two descriptions do not establish opposite original language arrangements. Both require attention to the fragmentary survival of the inscription before it can support a chronology of the script.

Kent assigned substantial weight to the association of the inscriptions with Cyrus's palace and inferred that the syllabary existed in Cyrus's time.^[39] The later debate separates the building's date from the possible addition of inscriptions. An older monument can carry later writing, while the case for a Cyran text requires evidence specific to the inscription as well as to its architectural setting. Rossi and Vallat's defence of possible authenticity and Schmitt's Darian attribution should therefore be read as competing assessments of the material, not as a settled building date automatically transferred to the script.^[2][23]

Authenticity questions affect both the apparent beginning and the later history of the corpus. Some inscriptions claim authors earlier than their proposed date of production; others are considered modern fabrications. These possibilities must be assessed for each object, rather than separated into two categories with dates assumed in advance.^[23][13]

The ancient problem is pseudepigraphy. Two gold tablets from Hamadan carry Old Persian texts in the names of Ariaramnes and Arsames, ancestors of Darius; if they were what they claim, a developed Old Persian written tradition would stand several generations before Bisitun. Schmitt describes them from the objects: AmHa, a fragment of a gold tablet with a monolingual Old Persian text breaking off in the middle of a word in line 10 and possibly trilingual originally; AsHa, a second gold tablet preserved in three pieces and carrying fourteen lines in Old Persian. He judges both inauthentic and produced only in late Achaemenid times, and reports a consensus that the inscriptions of Ariaramnes and Arsames do not come from those kings.^[23] His study of style reaches the same date by another route: the anacoluthon in AmHa is genuine, and the inscription arose in the late Achaemenid period on account of its faulty language.^[40] Rossi calls the two laminae hardly authentic, citing their linguistic corruption and their appearing and disappearing from the international scene.^[2]

The date is not agreed, and the disagreement is about how late. Kent held that the orthography points to approximately the time of Artaxerxes II, noted that the two kings are styled kings in Pārsa although the tablets were found in Media, suggested the pieces might be anti-Cyrus propaganda, and raised the possibility of a third tablet in the same series.^[39] Baghbidi reports Schmitt's dating to Artaxerxes III.^[25] Skjærvø's published writing-systems survey includes modern forgery among the explanations proposed for inscriptions attributed to Darius's ancestors, alongside later Achaemenid additions.^[38] Four kinds of argument are in play — orthography, the mismatch between find-spot and titulature, the objects' modern provenance history, and their physical form — and they do not all point to the same century.

The modern problem is a trade in fakes. Schmitt's survey dates the recognition of forged Old Persian cuneiform to the end of the nineteenth century and charts the growth of the corpus: "Kent (1953: p. 115b) listed eight and Schmitt (2007) pointed out 26 clear examples of forged inscriptions", and the number has continued to rise.^[13] The bulk of them are unexcavated pieces said to come from the region of Hamadan, and they appear in quantity from the 1930s onwards. The diagnostics are specific: unattested sign shapes, which are suspect in a system as simply constructed as this one; irregular, clumsy or deformed characters; wavy lines; and a small repertoire of signs repeated, often drawn only from the letters needed to spell the name of Darius. Individual exposures turn on the same kind of detail — a sheet of metal copying DPd from a modern edition and preserving an old erroneous reading corrected in the 1960s; clay tablets with DPa and XPe turned over like a modern book rather than upside down; four silver bowls whose word for "silver" is a back-translation of the Persian sīm rather than the Old Persian term; a coffin whose forger worked from a draft in modern Persian and transferred it into cuneiform with a model syllabary.^[13] Schmitt also reverses one of his own verdicts in print, a folded gold sheet from near Hamadan that he had judged late Achaemenid now being, in his view, a modern fake.^[13]

That trade has reached the scholarly corpus. van Bladel reports that Fattori has demonstrated that several of the inscriptions alleged to have been discovered at Hamadan, including inscriptions attributed to the reigns of Artaxerxes I, Darius II and Artaxerxes II, and long used by specialists, are frauds produced roughly between the 1930s and the 1950s; A2Hc is probably among them, and D2Ha and A2Hc are described as modern fakes modelled on known texts. The forgers, on this account, learned ancient grammatical solecisms from modern publications and reproduced them, creating what van Bladel calls inauthentic mistakes; he excludes the unauthenticated pieces from his own study.^[41] The group named there does not include the Ariaramnes and Arsames tablets, and the argument should not be extended to them. Any statement about "the inscriptions of Artaxerxes II" now has to say which inscriptions it means.

Old Persian inflects. Kent's statement of the noun is precise about what the language had lost: all the cases found in Sanskrit and Avestan occur "except the dative, which has been lost, its functions being assumed by the genitive form"; the ablative has no distinctive form, having merged with the instrumental and locative except for the single form Bābiraus; the accusative plural has become identical with the nominative except in the enclitic pronouns; both singular and plural are well represented, and a few duals survive.^[18] van Bladel describes the same system as close to Young Avestan and slightly reduced against Old Avestan and Vedic, with up to six case endings across three genders, verbs formed from often unpredictable stems, and the aorist and perfect thinly attested — the perfect stem occurring only once in the whole of Bisitun.^[28] These mergers are already present in the earliest substantial texts, so the language of Darius was some way from an inherited system in full working order.

The royal formulae put the case endings to work. "King of kings" is a noun governing a genitive plural, xšāyaθiyānām, the relation carried by the ending rather than by a preposition (see the King of Kings). Names of lands appear as places ruled, territories taken or points of departure according to construction. A genealogy is a chain of patronymics, and the chain is what made the first decipherment possible, since a man named as king in one text appears as a king's father in another.

Darius the king proclaims: These (are) the men who were there at that time, when I slew Gaumata the magus, who called himself Bardiya. At that time, these men strove as my followers: (one) called Vindafarna, son of Vahyasparuva, a Persian; (one) called Hutana, son of Thukhra, a Persian; (one) called Gaubaruva, son of Marduniya, a Persian; (one) called Vidarna, son of Bagabigna, a Persian; (one) called Bagabukhsha, son of Datavahya, a Persian; (one) called Ardumanish, son of Vahauka, a Persian.

Darius I, DB §68^[5]

The passage names the conspirators of 522 (see the Seven) in a fixed frame — personal name, patronymic, ethnic — repeated six times. In each identification, the father’s name has a genitive ending and is followed by puça, son.^[42] Reading such a sentence securely establishes what Darius claimed about the men beside him and about the magus he says he killed. Whether the claim is true is a separate question, answered from other evidence.

Fragmentary texts are restored partly from endings. A surviving ending can indicate that a broken word was a personal name, a place-name or a noun in a particular role, which narrows the possibilities without closing them. The commentary on the Persepolis tablet does exactly this, repeatedly distinguishing a recognisable grammatical shape from an uncertain identification of the word itself.^[8]

Kent's account of the verb opens by naming the categories and their unevenness: the Old Persian verb "shows the expected form-categories, though some of them are scantily represented" — present, aorist and perfect tense stems; active, middle and passive voices; indicative, subjunctive, optative, injunctive and imperative moods; present, imperfect, aorist and perfect tenses; three persons, three numbers, one infinitive and four participles.^[43] The inherited apparatus is there; the corpus exercises only part of it, and some categories are known from a handful of forms.

The distribution follows the business of the texts. A narrative of rebellion needs past tenses in quantity: men revolted, armies fought, a pretender was seized, the king restored order. A building text needs verbs of making and bringing. A prayer needs the optative, and the blessing-and-curse pairs at the end of Bisitun run through wishes in both directions. Exhortation needs the imperative and the second person, directed at kings not yet born.

Darius the king proclaims: You who hereafter shall behold this inscription, which I have inscribed, or these sculptures, do not destroy (them); as long as you shall have strength, care for them! Darius the king proclaims: If you shall behold this inscription or these sculptures, (and) shall not destroy them and, as long as you have strength, care for them, may Auramazda be friendly to you, may you have offspring in great number, may you live long, and what you shall do, may Auramazda make it successful for you!

Darius I, DB §§65–66^[5]

A repeated reporting formula, "Darius the king proclaims", divides the whole narrative into sections and reasserts the speaker at every turn. Direct speech is also used for opponents, so that a rival's claim can be stated before it is denied; the pretenders of 522 are made to say who they say they are. The arrangement gives the text much of its force, independently of the meaning of any single verb. Schmitt finds the same shaping at the level of the composition: DB I 27f. and I 72 frame the whole pre-accession narrative by ring composition. Schmitt’s German rendering makes the voice shift explicit: “Dies (ist), was von mir getan (worden ist), nachdem ich König geworden war” at DB I 27f. is resumed as “Dies (ist), was ich getan habe, nachdem ich König geworden war” at I 72.^[15]

The grammatical first person represents royal authorship; the physical production of an inscription required trained staff. Bisitun says that the text was written down and read aloud before the king, which attests oversight; the drafting, the preparation of exemplars and the cutting required trained staff whose work the inscriptions do not describe.^[5][2]

Some Old Persian words are attested once, and a hapax is a translator's problem with no internal solution. Kuhrt notes of yaumaini- in Darius's tomb inscription that "the adjective only occurs here", and presents three renderings of that single word, from "mit siedender Vergeltungskraft" through Schmitt's "I am fervent in counter-attack" to her own "I am furious in the strength of my revenge".^[44] The disagreement concerns the sense of one word within an agreed passage.

Other words have a range that modern categories cut differently. Kuhrt observes that viθ- means "house" and "family" and could mean "court" or "palace", as Kent and Schmitt take it, while Lecoq compared the Latin idiom for home and war.^[44] The same applies to the terms usually rendered country, people, law and kingship: the translation is serviceable and the Persian word's attested range remains part of the evidence.

The sharpest case is the address near the end of DNb. Kuhrt's note sets out the problem: Old Persian marika- "has usually been translated as 'menial'", following Kent, "which has an odd ring here, on the basis of the Babylonian rendering qallu, 'slave'", and she refers to Schmitt's suggestion that the passage could be an exhortation to the king's successor, which fits what follows.^[44] The two translations stand side by side in the held editions. Kent renders the vocative "O menial, vigorously make thou known of what sort I am, and of what sort my skillfulnesses, and of what sort my superiority"; Kuhrt has "Oh young man, actively make known what kind you are, of what kind your skills are, what kind your conduct."^[45][44] The Babylonian rendering records how one ancient translator understood the word in that place and therefore remains part of the philological problem; it does not by itself settle the Persian word's range.

The passage has a transmission history as well. Kuhrt records that DNb contains ten more lines than the Xerxes version XPl, separated from the main text by a space and containing the moralistic exhortation, and that this material was incorporated into the penultimate section of the Aramaic version of Darius's Bisitun text from Elephantine; she adds that the divergences between the two royal texts are primarily scribal, on Schmitt's assessment.^[44] What travels is a block of exhortation, taken into a different composition in a different language at Elephantine, rather than two complete inscriptions copied side by side. Whether the addressee of the transplanted passage remained the same is not something the Aramaic can settle.

Trilingual publication invites the assumption that the versions are interchangeable, and they are not. They differ in length, in clause structure and sometimes in what they assert, as the disagreement between §70 and its separately positioned Elamite counterpart illustrates.^[6][24] A version may also be a translation of a translation.

How the versions were produced is modelled in several ways, and each model has a name attached. Schmitt held that a dictation by the king in Old Persian lies at the root of the oldest written version, the Elamite, and allowed that the text may have been written down in the royal chancellery "in both Aramaic and Elamite" on that dictation.^[24] Rossi reports Sims-Williams’s inference of an Aramaic stage from two Old Persian words transcribed phonetically in an Aramaic DNb fragment. Rossi also reads Sancisi-Weerdenburg’s 1999 position as apparently returning to the idea of an Aramaic original.^[2] Schmitt elsewhere considered the Old Persian of DB a retranslation from Elamite; Rossi assigns all the royal texts to the chancellery's scribal staff, responsible for producing them multilingually.^[24][2][44]

For the administration rather than the monuments, Tavernier reconstructs a chain of six steps from the formulas that close letter-orders in Aramaic, Egyptian demotic and Elamite, attested from Darius I to Artaxerxes III in the Aršāma and Pherendates correspondence, the Persepolis archives and the Akhvamazda letters: a satrap gives a command in Old Iranian; an official passes it in Old Iranian to an interpreter; the interpreter writes it in Aramaic; he translates it if necessary into Egyptian or Elamite; he hands that to an indigenous scribe; the scribe produces the surviving document. The interpreter-scribes, teppir in Elamite, spr in Aramaic, held a considerable position, as their high rations at Persepolis show. Tavernier's resulting model is three-tiered — Old Persian for king and satraps, Aramaic at the top of the administration, regional languages below — and he notes that the evidence covers only satrapal correspondence, so that outside that level the empire's handling of multilingualism cannot be reconstructed.^[17] Rossi's reply in the next chapter is that the reconstruction entails many crucial assumptions, that on Tavernier's own showing the process can be envisaged for only a minority of the tablets, and that most translation most likely ran directly from Old Persian into Elamite; his summary of the state of the question is that the details are much better understood than a generation ago while "the overall framework still eludes us".^[1]

Gershevitch's alloglottography thesis addresses a further possibility: Elamite texts functioning as written representations of Persian speech. Rossi reports that the thesis was never effectively demonstrated and conflicts with substantial Elamite evidence; he also identifies its supporters and critics.^[1][2] That assessment does not remove the wider question of Persian speech behind documents written in other languages. Van Bladel describes administrative Aramaic as a standardised chancellery language and cautions against treating its written use as proof of a universal spoken lingua franca. His objection to the modern constitutional sense of “official language” should be distinguished from the evidence of administrative functions. A surviving empire-wide declaration of language status is not supplied by these sources.^[46]

Fort. 1208-101 supplies an administrative use of Old Persian outside royal display. It was excavated with the Persepolis Fortification tablets in 1933, reached Chicago in 1936 and was published by Matthew Stolper and Jan Tavernier in 2007. Its tongue-shaped form, string traces and seal impressions resemble those of Elamite administrative tablets. The seal scene is similar, but the individual seal was not identified in the Fortification repertory. The editors therefore leave open whether the tablet belonged to the archive's own operations or was intrusive.^[47][8][9] Their statement of its significance retains the distinction between reading and handling: “In the reign of Darius I, at least one Persian in Persia wrote Persian language in Persian script and expected someone else to know, if not how to read it, then at least where to file it.”^[48] Rossi's later summary places the emphasis on an expected reader.^[1]

What it says is largely unknown. The editors record that the script is neat and clear and the characters certainly Old Persian, and then that "most of the text is legible, but very little of the content is recognizable"; they found little parallel in the royal inscriptions or in Avestan, and little help in the Iranian administrative vocabulary preserved through Elamite transmission.^[8] The numerals are the securest part. Line 1 carries a numeral of at least 604, written with a possible higher element lost in the break, which, on the editors’ tentative identification of a dry measure, would represent six thousand litres or more; and the numeral 100 in that line, they note, occurs in only one other Old Persian text known at the time of publication, as part of the numeral 120 in Wc:1.^[8] Legibility and interpretation diverge sharply in this tablet, and the commentary keeps them apart throughout.

The publication explicitly reassesses earlier accounts of the script’s function. Stolper and Tavernier quote Stolper's own 2005 account of its use for royal inscriptions and labels, alongside Gershevitch's view of a very small readership and Schmitt's statement that Old Persian played no part in administration.^[48][16] The estimate of between eight and twenty readers or writers from Darius to Rawlinson is the editors' summary of Gershevitch's argument, not a verbatim quotation from his study. Their juxtaposition shows how a new administrative text challenged earlier generalisations. The single survivor changes the boundary of what is attested but gives no measure of frequency.^[48][9]

The editors then set out two readings of their own find and choose between them. On the "tip of the iceberg view", the tablet is the sole surviving trace of a larger body of Old Persian records, of the kind that administrative documents with short-term utility rarely leave; on the "general literacy view", Old Persian writing was more widely known than the surviving monuments suggest. Their verdict: "As matters stand, it seems more parsimonious to take this tablet as the product of extraordinary behavior than as evidence of widespread practice."^[9] They ground that in sampling. As of April 2007, Stolper reported that he had re-examined many of the boxes Hallock had drawn on and about 360 others, so that "at least 1,500 of the boxes have been examined in ways that would have been likely to identify other Old Persian tablets", out of some 2,350 in the collection.^[9] That figure describes the state of examination in 2007 and is not a statement about the archive as it now stands.

Literacy in Old Persian is argued from indirect evidence on all sides. The minimal position is Gershevitch's, on which the readership was a handful of individuals in two and a half millennia.^[48] van Bladel argues the other way from the character of the script and the placement of the texts: his modern students "read its simple cuneiform letter-signs easily within a week of practice before they know even the basics of the grammar", from which he infers that native speakers could have learnt the signs with equal ease had they the time and the interest, especially where the signs ornamented palace walls and columns; and Old Persian writing was not confined to buried foundation tablets and inaccessible cliffs, appearing on inscribed bowls, weights and fittings.^[49]

The same author is careful about what that does not show. Teaching the language to students without Greek, Sanskrit or Assyriology behind them, he writes, demonstrates that "comprehending the grammar of Old Persian, even in its limited corpus, is far from intuitive and is indeed quite difficult"; the corpus is small enough to fit with a translation into a single volume of two hundred pages, and the grammar is still hard.^[28] The signs are the easy half.

The inscriptions themselves presuppose readers, or at least address them. Bisitun instructs future kings to preserve the text and the sculptures, and elsewhere announces that the writing was sent out to the lands, which implies an expectation that it would be received and understood somewhere.^[5] Rossi endorses Tuplin's distinction between the public display of a text with an iconographic component and the more extravagant scenario of §70, in which a text alone travels to the provinces.^[2]

Speaking Persian, recognising its cuneiform signs and interpreting an administrative record required different skills. The editors of Fort. 1208-101 hypothesise multilingual personnel capable of moving between scripts, but do not recover the training or literacy of a whole population from the tablet. Its existence demonstrates a practical use, while the scale and institutional setting of that use remain uncertain.^[8][9]

The texts of the later Achaemenid kings depart from the usage of Darius and Xerxes, and the departures can be quantified. From the reign of Artaxerxes I the range of noun cases in use contracts almost entirely to three, nominative, genitive-dative and accusative, with very few residues: one locative, wiθiyā, "in the house", lexicalised under Artaxerxes I, and one ablative surviving in a formulaic prayer after a preposition, hačā wispā gastā, "from all that is foul", three times under Artaxerxes II.^[50] In the Persepolis inscription of Artaxerxes III known as A3Pa, "One quarter of the words taking nominal and pronominal endings exhibit different kinds of grammatical problems", and, where the copies remain legible, the same departures occur in all four, "showing that they were based on a common master-copy".^[50] Set against the formulas of the Susa charter and the Elvend inscription of Xerxes, the same phrases appear with altered endings: būmim becomes būmām, asmānam becomes asmānām, šiyātim becomes šāyatām, and genitives are formed as though agglutinated, dārayawahuš-ahyā.^[50]

Kent catalogued the phenomena in detail, distributing them across sections on metathesis, omissions, additions, alteration of characters, miswritings, syntactical misuse and the creation of new incorrect forms.^[51] His explanation is often overlooked, because the taxonomy is remembered as a list of errors. Kent wrote that most of these errors belong to inscriptions later than those of Xerxes, since by that time the development to Middle Persian was under way:

OP had ceased to be a vernacular, and the scribes who composed the inscriptions had no experience of the language as it had been. They were thrown back upon the use of words and forms found in the older records, the use of which they often failed to understand.

R. G. Kent, Old Persian, §50^[51]

He says it a second time in his chapter on syntax, where he calls Old Persian by that date "virtually a dead language employed only in writing ceremonial official records, its spoken form having suffered a wearing down of the endings".^[52] Van Bladel's genealogy of the scholarship shows that a simple opposition between an older philology of error and a modern philology of change is misleading: Kent already linked many later irregularities to developments in spoken Persian.

Three explanatory mechanisms recur in these studies, supported by different kinds of evidence. The first is a stonemason copying a draft he could not read, which Schmitt names as the normal condition of the ancient engraver.^[13] The second is an error already present in the written exemplar: Schmitt explains the anomalous position of upari at DNb 49, against its placement three lines earlier, as a mistake by the scribe of the Vorlage and expressly not by the stonemason, since the anticipating supplement could only arise in a written draft.^[53] The third is change in the spoken language surfacing through a conservative written norm. On Mancini's reading, reported by van Bladel, the late spellings are regular historical masks for Proto-Middle Persian forms, which "would mean that the scribes of the late Achaemenian Persian inscriptions were not so much making mistakes as they were consciously employing a conventional and somewhat conservative orthography for their present language". The evaluations of those inscriptions as corrupt or barbaric would then be, in van Bladel’s phrase, doubly wrong.^[10]

One misspelling carries a datable consequence. Artaxerxes III's šāyatām for šiyātim can only be a back-formation in which the sequence -āya- is used to write a long ā, which means that the contraction had already happened in speech; with the loss of the final syllable and the regular changes xš- to š- and θ to h, *xšāyaθya has by then become šāh. van Bladel's conclusion is that "the word for 'king' had already become šāh by the time of Alexander's invasion beginning in 334 bce".^[54]

The field's framing of the stage has settled while its name has not. van Bladel states that the language of the late Achaemenian inscriptions "is correctly understood as Proto-Middle Persian" and that this "is also the consensus of current scholarship", citing Schmitt, Lazard, Huyse, Skjærvø, Mancini, Korn and Fattori; the competing labels are Schmitt's Spätaltpersisch, "late Old Persian", Skjærvø's "Pre-Middle Persian", on the ground that the language of the later inscriptions is neither Old nor Middle Persian, "Proto-Middle Persian", and van Bladel's own "late Achaemenian Persian".^[10] The older explanations that competed with language change are named and dated: Meillet excluded the late inscriptions from the description of Old Persian altogether; Sturtevant blamed Babylonian scribes who had lost the case endings of their own language, while conceding that highly inflected ancient languages are otherwise remarkably free of gross errors in case usage, Old Persian forming the single exception; Brandenstein and Mayrhofer described the late texts as faulty Old Persian no longer mastered.^[50][49] A fourth, narrower mechanism applies to individual texts: Stolper and Tavernier point out that of Kent's own examples of erroneous long writing, two are better explained on morphological grounds, following Schmitt, and that the fifth, in the daiva inscription of Xerxes, is a scribal error ascribed to the underlying draft.^[8]

Style changes with grammar. Schmitt finds in the genealogies of Darius II and of Artaxerxes II and III a figure with no model in the classical texts of Darius I and Xerxes, and reads it not as a new stylisation but as "stilistischer Unbeholfenheit als Folge mangelnder Beherrschung der Inschriftensprache und ihres Formulars", stylistic clumsiness following from an imperfect command of the language of the inscriptions and of its formulary.^[40]

Authentication affects which texts can support a linguistic reconstruction. Van Bladel explicitly excludes the Hamadan inscriptions that he judges modern fabrications on Fattori's argument; van Bladel’s analysis proceeds from the remaining evidence. The important distinction is between anomalies repeated across authenticated ancient exemplars and anomalies copied into modern objects from published editions. The latter cannot supply independent evidence for an ancient linguistic development.^[41][50]

Political succession and linguistic periodisation describe different developments. The categories Old Persian and Middle Persian organise a history inferred from unevenly surviving written evidence; a change of dynasty does not by itself supply a linguistic boundary. Van Bladel's use of a late Achaemenian stage makes that distinction explicit while connecting the later inscriptions to the subsequent development of Persian.^[10]

The script had to be deciphered without a surviving ancient explanation of its signs or an accompanying version in a script already understood. Kent nevertheless records the resources available to the early decipherers: copies of the monuments, knowledge of royal genealogies, comparison with Iranian languages and the recognition of repeated titulature. The task began with structural clues from names, genealogy and related Iranian languages, rather than with a ready bilingual key.^[55]

The first stages were about structure. The earliest complete publication of an inscription, DPe, came in 1711; Carsten Niebuhr's copies of 1778 established that the Persepolis texts used three systems of writing and ran from left to right, the direction shown by two copies of XPe with differing line divisions. O. G. Tychsen in 1798 recognised that the three systems represented three languages and that a recurring diagonal wedge in the simplest of them divided words, although he dated the texts to the Parthian period; Friedrich Münter identified the divider independently in 1802, guessed that a recurring group meant "king", and assigned the texts correctly to the Achaemenid period.^[55]

Grotefend, in the same year, compared DPa and XPe in Niebuhr's copies. Assuming Achaemenid royal names and a language close to Avestan, and helped by Silvestre de Sacy's recent reading of the Pahlavi royal titles, he exploited the genealogical pattern: one text names a king whose father is not given a royal title, another a king whose father is the king of the first text, which fits Hystaspes, Darius and Xerxes. The result was a foothold rather than a solution. Kent records that of the twenty-two different signs in DPa and XPe Grotefend "got only 10 correctly", admitting two values each for two of them, and that apart from the three names, "king" and "great" were the only words he identified correctly; he added the name of Cyrus in CMa in 1815.^[55]

The inherent vowels took another generation. Lassen supplied the inherent a in 1836 and saw in 1839 that some characters occurred only before i and others only before u; H. C. Rawlinson, Hincks and Oppert independently recognised in 1846 and 1847 that those consonant signs carried inherent i and u, and Oppert at the same time worked out the writing of diphthongs and the omission of n and m before consonants.^[55] Henry Creswicke Rawlinson's copying and publication of the Bisitun text belong to this sequence of investigations, alongside the linguistic contributions of the other decipherers.^[55][56]

Getting the text off the rock was a physical operation, and Rawlinson described it himself. The Babylonian panel stood on a block he had judged unapproachable, and the work was done by a local climber whom he paid.

At length, however, a wild Kurdish boy, who had come from a distance, volunteered to make the attempt, and I promised him a considerable reward if he succeeded… passing over a distance of twenty feet of almost smooth perpendicular rock in a manner which to a looker-on appeared quite miraculous… Here with a short ladder he formed a swinging seat, like a painter's cradle, and, fixed upon this seat, he took under my direction the paper cast of the Babylonian translation of the records of Darius.

H. C. Rawlinson, 1852^[57]

King and Thompson, sent by the British Museum in 1904, record that Rawlinson's squeezes presented to the Museum “crumbled and fell to pieces”, while the rock inscriptions themselves continued to deteriorate. They also describe A. V. W. Jackson's access to the ledge in 1903 and his collation of part of the Old Persian text.^[57] Schmitt records surviving notebooks in the British Library and squeezes of the minor texts DBb–DBj in the British Museum, alongside the destruction of other squeezes in the 1930s or 1940s.^[56] His edition uses photographs taken by the German Archaeological Institute in the winter of 1963–64. On the available photographs of column III, scaffolding obscures a number of signs, preventing their fresh collation from those photographs. Earlier editors' recorded readings consequently remain part of the evidence; the obstruction does not mean that the characters had never been examined.^[58]

Old Persian then became the instrument for the rest. Weissbach put the dependence at its strongest: it is a question whether the Elamite and Babylonian scripts would ever have been opened at all, had scholars not possessed in the already deciphered Old Persian originals the keys to those labyrinths.^[35] Each of the other two required its own decipherment, and the process was a chain of connected investigations rather than a single key (see the Bisitun inscription for the monument's own history).^[55][2]