At first sight the Pahlavi books present the strangest spectacle of mixture of speech. Purely Semitic (Aramaic) words—and these not only nouns and verbs, but numerals, particles, demonstrative and even personal pronouns—stand side by side with Persian vocables. Often, however, the Semitic words are compounded in a way quite unsemitic, or have Persian terminations. As read by the modern Zoroastrians, there are also many words which are neither Semitic nor Persian; but it is soon seen that this traditional pronunciation is untrustworthy. The character is cursive and very ambiguous, so that, for example, there is but one sign for n, u, and r, and one for y, d, and g ; this has led to mistakes in the received pronunciation, which for many words can be shewn to have been at one time more correct than it is now. But apart from such blunders there remain phenomena which could never have appeared in a real language; and the hot strife which raged till recently as to whether Pahlavi is Semitic or Persian has been closed by the discovery that it is merely a way of writing Persian in which the Persian words are partly represented —to the eye, not to the ear—by their Semitic equivalents. This view, the development of which began with Westergaard (Zendavesta, p. 20, note}, is in full accordance with the true and ancient tradition. Thus Ibn Mokaffa', who translated many Pahlavi books into Arabic, tells us that the Persians had about one thousand words which they wrote otherwise than they were pronounced in Persian. For bread he says they wrote lhma, i.e., the Aramaic lahtnd, but they pronounced nan, which is the common Persian word for bread. Similarly bsra, the Aramaic besrd, flesh, was pronounced as the Persian gbsht. We still possess a glossary which actually gives the Pahlavi writing with its Persian pronunciation. This glossary, which besides Aramaic words contains also a variety of Persian words disguised in antique forms, or by errors due to the contracted style of writing, exists in various shapes, all of which, in spite of their corruptions, go back to the work which the statement of Ibn Mokaffa' had in view. Thus the Persians did the same thing on a much larger scale, as when in English we write £ (libra) and pronounce “pound” or write <5r j or & (et) and pronounce “and.” No system was followed in the choice of Semitic forms. Sometimes a noun was written in its status absolutus, sometimes the emphatic â was added, and this was sometimes written as S sometimes as ∏. One verb was written in the perfect, another in the imperfect. Even various dialects were laid under contribution. The Semitic signs by which Persian synonyms were distinguished are sometimes quite arbitrary. Thus in Persian khwésh and khwat both mean “self”; the former is written nfsL·h (nafshâ or nσfsheh), the latter BNFshκ with the preposition be prefixed. Personal pronouns are. expressed in the dative (i.e., with prepositional I prefixed), thus lk (lakh) for tu, “ thou,” lnh (land) for amd, “ we.” Sometimes the same Semitic sign stands for two distinct Persian words that happen to agree in sound; thus because hdnd is Aramaic for “this,” hna represents not only Persian é, “this,” but also the interjection é, i.e., “O” as prefixed to a vocative. Sometimes for clearness a Persian termination is added to a Semitic word; thus, to distinguish between the two words for father, pit and pitar, the former is written ab and the latter abitr. The Persian form is, however, not seldom used, even where there is a quite well-known Semitic ideogram.

These difficulties of reading mostly disappear when the ideographic nature of the writing is recognized. We do not always know what Semitic word supplied some ambiguous group of letters (e.g., pun for pa, “to,” or ht for agar, “if ”); but wealways can tell the Persian word— which is the one important thing—though not always the exact pronunciation, of it in that older stage of the language which the extant Pahlavi works belong to. In Pahlavi, for example, the word for “female” is written mdtak, an ancient form which afterwards passed through mddhak into mddha. But it was a mistake of later ages to fancy that because this was so the sign T also meant D, and so to write T for D in many cases, especially in foreign proper names. That a word is written in an older form than that which is pronounced is a phenomenon common to many languages whose literature covers a long period. So in English we still write though we do not pronounce the gutteral in through, and write laugh when we pronounce Zα∕.

Much graver difficulties arise from the cursive nature of the characters already alluded to. There are some groups which may theoretically be read in hundreds of ways; the same little sign may be t?, χ∖ ∏*, fc∏, m, χj, ∏J, and the ∏ too may be either h or kh.

In older times there was still some little distinction between letters that are now quite identical in form, but even the fragments of Pahlavi writing of the 7th century recently found in Egypt show on the whole the same type as our MSS. The practical inconveniences to those who knew the language were not so great as they may seem; the Arabs also long used an equally ambiguous character without availing themselves of the diacritical points which had been devised long before.

Modern MSS., following Arabic models, introduce diacritical points from time to time, and often incorrectly. These give little help, however, in comparison with the so-called Pâzand or transcription of Pahlavi texts, as they are to be spoken, in the character in which the Avestd itself is written, and which is quite clear and has all vowels as well as consonants. The transcription is not philologically accurate; the language is often modernized, but not uniformly so. Pâzand MSS. present dialectical variations according to the taste or intelligence of authors and copyists, and all have many false readings. For us, however, they are of the greatest use. To get a conception of Pahlavi one cannot do better than read the Minoi-Khiradh in the Pahlavi with constant reference to the

Pâzand. Critical labour is still required to give an approximate reproduction of the author’s own pronunciation of what he wrote.

The coins of the later Sâsânian kings, of the princes of Tabaristân, and of some governors in the earlier Arab period exhibit an alphabet very similar to Pahlavi MSS. On the older coins the several letters are more clearly distinguished, and in good specimens of well-struck coins of the oldest Sâsânians almost every letter can be recognized with certainty. The same holds good for the inscriptions on gems and other small monuments of the early Sâsânian period; but the clearest of all are the rock inscriptions of the Sâsânians in the 3d and 4th centuries, though in the 4th century a tendency to cursive forms begins to appear. Only r and v are always quite alike. The character of the language and the system of writing is essentially the same on coins, gems, and rocks as in MSS.—pure Persian, in part strangely disguised in a Semitic garb. In details there are many differences between the Pahlavi of inscriptions and the books. Persian endings added to words written in Semitic form are much less common in the former, so that the person and number of a verb are often not to be made out. There are also orthographic variations; e.g., long â in Persian forms is always expressed in book-Pahlavi, but not always in inscriptions. The unfamiliar contents of some of these inscriptions, their limited number, their bad preservation, and the imperfect way in which some of the most important of them have been published leave many things still obscure in these monuments of Persian kings; but they have done much to clear up both great and small points in the history of Pahlavi.

Some of the oldest Sâsânian inscriptions are accompanied by a text belonging to the same system of writing, but with many variations in detail, and an alphabet which, though derived from the same source with the other Pahlavi alphabets (the old Aramaic), has quite different forms. This character is also found on some gems and seals. It has been called Chaldaeo-Pahlavi, &c. Olshausen tries to make it probable that this was the writing of Media and the other that of Persia. The Persian dialect in both sets of inscriptions is identical or nearly so.

The name Pahlavi means Parthian, Pahlav being the regular Persian transformation of the older Parthava. This fact points to the conclusion that the system of writing was developed in Parthian times, when the great nobles, the Pahlavâns, ruled, and Media was their main seat, “the Pahlav country.” Other linguistic, graphical, and historical indications point the same way; but it is still far from clear how the system was developed. We know indeed that even under the Achaemenians Aramaic writing and speech were employed far beyond the Aramaic lands even in official documents and on coins. The Eranians had no convenient character, and might borrow the Aramaic letters as naturally as they subsequently borrowed those of the Arabs. But this does not explain the strange practice of writing Semitic words in place of so many Persian words which were to be read as Persian. It cannot be the invention of an individual, for in that case the system would have been more consistently worked out, and the appearance of two or more kinds of Pahlavi side by side at the beginning of the Sâsânian period would be inexplicable. But we may remember that the Aramaic character first came to the Eranians from the region of the lower Euphrates and Tigris, where the complicated cuneiform character arose, and where it held its ground long after better ways of writing were known. In later antiquity probably very few Persians could read and write. All kinds of strange things are conceivable in an Eastern character confined to a narrow circle. Of the facts at least there is no doubt.

The Pahlavi literature embraces the translations of the holy books of the Zoroastrians, dating probably from the 6th century, and certain other religious books, especially the Minôi-Khiradh (see above) and the Bundehish. The Bundehish dates from the Arab period. Zoroastrian priests continued to write the old language as a dead tongue and to use the old character long after the victory of a ne>v empire, a new religion, a new form of the language (New Persian), and a new character. There was once a not quite inconsiderable profane literature of which a good deal is preserved in Arabic or New Persian versions or reproductions, ι>articularly in historical books about the time before Islam. Very ittle profane literature still exists in Pahlavi; the romance of Ardashir has been mentioned above (p. 135, note 1). The difficult study of Pahlavi is made more difficult by the corrupt state of our copies, due to ignorant and careless scribes. A Pahlavi grammar is of course an impossibility. The necessary preparation for the study is a sound knowledge of New Persian, with which one easily finds the clue to the inconsiderable grammatical variations of the older language. The lexical peculiarities of the texts are more considerable, and partly due to the peculiarities of priestly thought and speech. Of glossaries, that of West (Bombay and London, 1874) is to be recommended; the large Pahlavi, Gujarati, and English lexicon of Jamaspji Dastur Minocheherji (incomplete, 3 vols., Bombay and London, 1877-82) is very full, but has numerous false or uncertain forms, and must be used with much caution.