^{page 517 note 1} On the provenance of the inscription, see below, p. 524.

^{page 517 note 2} Les inscriptions sémitiques de Loyanq conservées au Musée Gouvernemental de Pékin: contribution à l'étude de la question juive en Chine. Pékin: Imprimerie des Lazaristes, 1926Google Scholar. There is in fact only the one inscription involved, and the use of the plural in the title is due merely to an excess of caution on the part of the author. Although he thought it probable that the three stones formed parts of one inscription, he had observed that they were not all of the same breadth (p. 11), and for this reason he considered it more prudent to imply by his title that the case for a single original inscription was not proven.

^{page 517 note 3} ‘Nous sommes bien éloignés de prétendre que toutes les lettres peuvent facilement se référer à un type de Palmyre: des difficultés surgissent’ (p. 6). And with reference to the middle-sized fragment, ‘En passant à la traduction de ce fragment nous ferons remarquer qu'il offre incontestablement plus de difficultés que le premier, d'abord parce qu'il n'est pas complet, ensuite parce qu'il contient des signes difficilement identifiables avec des lettres des alphabets sémitiques’ (p. 16). These difficulties might even have sufficed to persuade the author that his investigations were in the wrong direction, had it not been that the circumstances offered an apparently reasonable excuse: the carving was ‘fort probablement l'oeuvre d'un lapicide chinois accomplissant en quelque sorte une besogne purement matérielle’.

^{page 517 note 4} Loewenthal, R. in his bibliography (‘The Jews in China’, Chinese Social and Political Science Review (Peking), XXIV, 2, 1940) listed three items: pp. 176–7, no. 158Google Scholar, Prévost's original pamphlet, with a reproduction of the photograph; p. 203, no. 223, ‘Les inscriptions hébraïques du Musée de l'Université du Gouvernement chinois à Pékin’, Bulletin Catholique de Pékin, XI, 134, 1924, pp. 407–10Google Scholar; and pp. 211–12, no. 250, Van den Brandt, Joseph, Catalogue des principaux ouvrages sortis des presses des Lazaristes à Pékin de 1864 à 1930, Pékin, Imprimerie des Lazaristes, 1933Google Scholar, which naturally includes the publication by Prévost, with a brief additional annotation. More recently, White, W. C., Chinese Jews, Toronto, 1942, I, 193Google Scholar, quoted Loewenthal's note on Prévost from the article mentioned above, but without the safeguard of the photograph.

^{page 518 note 1} Information from Dr. Loewenthal, in a personal communication.

^{page 518 note 2} I am indebted to Professor Henning, not only for bringing the inscription to my notice, but also for his kindness in placing at my disposal the information which he had collected concerning it. I should like also to express my sincere gratitude to Dr. Loewenthal, who, in addition to giving us the relevant bibliography, generously supplied a photographic copy of Prévost's work, which otherwise would have remained out of reach, and admirably clear photographs of rubbings from the stones, which are here reproduced. Without this additional material, much would have remained uncertain in the reading of the text; and all the available information concerning the provenance of the inscription is derived from a few remarks by Prévost, supplemented by some particulars told by Brother Van den Brandt to Dr. Loewenthal when the latter visited Peking in 1940.

^{page 518 note 3} Konow, S., Kharoehṭhī inscriptions (Corpus inscriptionum Indicarum, II, 1), p. 65Google Scholar.

^{page 520 note 1} Prévost thought that the missing stone was ‘employée très probablement dans des travaux de construction’, while Van den Brandt suggested that it might have been thrown away by one of the porters on the journey. This discrepancy, trivial in itself, is disturbing; since, if reliable information had been available in the notes of the expedition which brought the stones to Peking, there would have been no need to resort to conjecture. We must bear this in mind with reference to the more important question of the origin of the stones.

^{page 520 note 2} Konow, , op. cit., pp. 65, 79, 145, 160, 171Google Scholar. A sixth example, in the Ārā inscription (ibid., p. 165), was read by Konow as ‘25’, where the sign read as ‘20’ is badly defaced by weathering of the stone, and remains doubtful. The final ‘5’, however, is certain.

^{page 521 note 1} Or, more probably, two on the 1st and three on the 8th, if in the “Takht-i-Bāhī” inscription (Konow, op. cit., plate XII) we read ['a]thama in place of the extremely bizarre restoration [pra]tham[e], which would seem to imply a derivation from a form such as *praṣṭama.

^{page 521 note 2} Of the three instances which do not conform to this pattern, one is on the 16th (Fatehjang, plate IV), one on the 28th (Sui Vihār, plate XXVI: in the translation of the latter, p. 141, ‘eighteenth’ is a slip of the pen), and one probably on the 27th (Loriyān Tangai, plate XXI.1). I have left out of account the Skārah Ḍherī inscription (plate XXIII.8), where Konow's reading seems purely speculative.

^{page 522 note 1} Konow, , op. cit., p. 160Google Scholar.

^{page 522 note 2} ibid., p. 145.

^{page 522 note 3} ibid., p. 65.

^{page 522 note 4} Or possibly, ‘a gift to the Vihāra’.

^{page 522 note 5} Konow, , op. cit., p. 79Google Scholar. The photograph (plate XV.1) seems to have danaṃ- rather than Konow's reading daṇaṃ-.

^{page 522 note 6} It now seems preferable on palaeographical grounds to follow the editors of the Niya documents in reading tsa for the conjunct character which Konow and others have interpreted as tśa (see Gāndhārī Dharmapada, introduction, pp. 73–4); and transcriptions such as saṃvatśaraye are altered accordingly.

^{page 522 note 7} Konow, op. cit., plate XV.1.

^{page 522 note 8} Mathurā Lion Capital inscription, Konow, , op. cit., plate VII, A 14–15Google Scholar.

^{page 522 note 9} Lamp from Swat: Burrow, T., Journal of the Greater India Society, XI, 1944, 15Google Scholar.

^{page 522 note 10} Avadāna-śatalca, i.266, ii.109; Divyāvadāna, 274.13. In this last, the reading in Cowell and Neil's edition is cāturdiśāryabhikṣusaṃghāya, which—excellent though the sense may be— must assuredly be a corruption, and should be restored to cāturdiśāya bhikṣusaṃghāya.

^{page 523 note 1} For example, Konow, op. cit., plates XII and XIII.2.

^{page 523 note 2} See further my edition of the Gāndhārī Dharmapada, introduction, pp. 81–2.

^{page 523 note 3} Konow, op. cit., plate XXXV.

^{page 523 note 4} Gāndhārī Dharmapada, pp. 92, 177.

^{page 524 note 1} op. cit., pp. 161–2, plate XXXI.

^{page 524 note 2} ibid., plates I.6, IIC, XIII.4, XXXIII.1, XXVII.4, XXX.2, XXXII.2.

^{page 525 note 1} Even the date was reported wrongly, since the first printed report appeared in 1924: see above, p. 517, n. 4.

^{page 525 note 2} See p. 520, n. 1.

^{page 525 note 3} See Maspero, H., ‘Les origines de la communauté bouddhiste de Lo-yang’, JA, CCXXV, 1934, 87–107Google Scholar; Züroher, E., The Buddhist conquest of China, Leiden, 1959Google Scholar. (I am indebted to Professor D, C. Twitchett for these references.)

^{page 526 note 1} Zürcher, , op. cit., 23 ffGoogle Scholar.

^{page 526 note 2} For indications of contacts between the Indians of Lou-Ian (Krorayina) and the Sogdians, see Henning, W. B., ‘The date of the Sogdian Ancient Letters’, BSOAS, XII, 3–4, 1948, p. 603, n. 3Google Scholar.

^{page 526 note 3} Pelliot, P., ‘Les noms propres dans les traductions chinoises du Milindapañha’, JA, 09–10 1914, 379–419Google Scholar; Weller, F., ‘Über den Aufbau des Pāṭikasuttanta’, Asia Major, V, 1, 1928, 104 ff.Google Scholar; Waldschmidt, E., Bruehstücke buddhistischer Sūtras aws dem zentralasiatischen Sanskritkanon, I, 1932, 149–206 and 229 ffGoogle Scholar. (with reference to the version of the Mahāsamāja-sūtra in the Chinese Dīrghāgama). The earlier discussions of the question, prior to the new edition of the ‘MS Dutreuil de Rhins’ by Bailey, H. W. (‘The Khotan Dharmapada’, BSOAS, XI, 3, 1945, 488–512)CrossRefGoogle Scholar were to a large extent hindered in the identification of the specific dialect as Gāndhārī (the ‘North-western Prakrit’) by systematic errors in the first edition of this text (Senart, E., ‘Le manuscrit kharoṣṭhī du Dhammapada’, JA, 09–10 1898, 193–308)Google Scholar. See further Bailey, H. W., ‘Gāndhārī’, BSOAS, XI, 4, 1946, 764–97CrossRefGoogle Scholar, and my edition of the Gāndhārī Dharmapada, introduction, pp. 49–54.

^{page 526 note 4} We cannot, of course, categorically exclude the possibility that yet other early Chinese Buddhist texts may be shown to have been translated from versions in some other Prakrit dialect or dialects; but there is at present, so far as I am aware, no positive reason for suggesting that this might be so.

^{page 526 note 5} loc. cit., 93.

^{page 526 note 6} The term in question (ssŭ ) is rendered here by Maspero as ‘lieu de culte’, and elsewhere in the same article as ‘temple’, in accordance with a theory of the semantic history of the Chinese word; but the generally accepted sense of ‘monastery’ seems preferable: see Zürcher, , op. cit., 38–9Google Scholar. Here, however, a new complication seems to be raised by the fact that Zürcher, if I am not mistaken, appears to reserve the translation ‘monastery’ for , and ‘vihāra’ for ching-shê , the two terms being also given separately in the index, without a cross-reference between them. The ‘vihāra within the palace’ established by the emperor Hsiao-wu in A.D. 381 must indeed be a monastery, since ‘monks were invited to dwell in it’ (p. 151); but on p. 189 a reference back to this same passage calls the same palace-vihāra a ‘chapel’. I must of course leave it to Sinologists to decide what the differing implications of the two Chinese terms may be, if in fact a difference is intended. (Both arc given as renderings of vihāra in Hôbôgirin, s.v. bikara.) But in order to prevent misunderstanding, I must make it clear that in the present article ‘monastery’ is accepted as a translation of vihāra. The latter, indeed, is used in the earlier texts to denote the dwelling-place (in principle, temporary) of a single bhikṣu, and a number of such dwellings may be grouped into a sanghārāma, a term which continued to be familiar in Central Asia and China, and which is also often appropriately rendered by ‘monastery’ (cf. Lamotte, E., Histoire du bouddhisme indien, I, 64Google Scholar). But from a fairly early period, the etymological sense of ‘dwelling’ allowed the term vihāra to be used for the place of residence of a community. In two of the avadāna-passages cited above (p. 522, n. 10), a vihāra is presented to the Sangha of the four quarters; and for this, ‘monastery’ seems to be a convenient translation. (Professor Henning has given me the information that the same word ssû was used for ‘monastery’ in Manichaean texts also: Chavannes, and Pelliot, , ‘Un traité manichéen retrouvé en Chine’, JA, 01–02 1913, 106Google Scholar (reprinted in book form, 130) translated as ‘monastére’ on p. 108 (132), where a list of rooms is given, although a few lines later the word was rendered as ‘temple’, p. 113 (137). ‘Monastery’ is however to be preferred: see Henning, W. B., BSOAS, XI, 1, 1943, p. 218, n. 4.Google Scholar)

^{page 527 note 1} Demiéville, P., in L'Inde classique, II, pp. 411–13Google Scholar.

^{page 527 note 2} Zürcher, , op. cit., 255, 413Google Scholar, citing a work of the mid sixth century, Lo-yang ch'ieh-lan chi ‘Account of the monasteries (sanghārāma) at Lo-yang’ (Taishō Tripiṭaka, vol. 51, no. 2092), where it is stated that in A.D. 316 there were 42 (or 32) monasteries in Lo-yang. The burning in A.D. 190 was not the end of the city, and the evidence quoted by Maspero (see above, p. 526, n. 5) shows that the monasteries in question either survived, or were founded or re-established after this date. Lo-yang was still in existence to be burned again in A.D. 311, and between 530 and 535: Henning, W. B., BSOAS, XII, 3–4, 1948, p. 603, n. 3Google Scholar.

^{page 527 note 3} Gāndhārī Dharmapada, introduction, p. 50; Waldschmidt, E., Bruchstücke buddhistischer Sūtras, I, p. 229Google Scholar.

^{page 527 note 4} The Dīrghāgama was translated in A.D. 413: Demiéville, P., in L'Inde classique, II, p. 418Google Scholar.

^{page 527 note 5} See Konow, op. cit., plates XXVI (Sui Vihār and Zeda) and XXVII (Mānikiāla). The highly individual writing of document 661 from Endere (see following note) shows a slight tendency to flattening in ha, and an angular shape.

^{page 528 note 1} Kharoṣṭhī inscriptions [i.e. documents] discovered by Sir Aurel Stein at the Niya, Endere, and Lou-Ian sites, ed. Boyer, A. M., Rapson, E. J., Senart, E., and Noble, P. S., Oxford, 1920–1929Google Scholar.

^{page 528 note 2} Such a pronouncement is admittedly based on a subjective judgment, and I have deliberately expressed the matter in terms to indicate that I am aware of the fact, as I am also aware that such a subjective judgment is an open invitation to criticism. Nevertheless, it must not be thought that an impressionistic judgment on palaeographical questions is of necessity inadequate as a matter of principle, and that it is merely the resort of one who shirks the labour needed for a palaeographic analysis of detail. It is important to see that neither of these two things is a substitute for the other, and that both are necessary. The manner of integration of details into the style of the whole is difficult to talk about, since in most fields where a visual judgment is important no adequate terminology for such a discussion has been evolved. This is particularly obvious in some aspects of the history of the visual arts, where much has been written which will appear to a scientifically trained reader to be deplorable nonsense. Some of it may well be; but in other cases the presence of the visual object itself may convince the reader that the written account is not nonsense at all, even though it may be inadequately expressed. There are obviously intrinsic difficulties in framing an adequate terminology to deal with ‘style’; and perhaps there is little incentive to make the effort in fields where it is always easier to say to the student, ‘Go and look at it’. The two points to be made in the present context are, first, that my subjective judgment is the judgment of one individual, which is now open to examination and, if necessary, to correctioII, by others qualified; and, secondly, that if this particular judgment should turn out to be wrong, as it may, this would not show that an assessment based on style is in principle a mistake, but merely that in the present instance the underlying palaeographical experience had been too restricted in range in relation to the material under discussion.

^{page 528 note 3} On the two characters transcribed sa and sa, and their distribution in the texts, see Gāndhārī Dharmapada, introduction, pp. 67–70.

^{page 529 note 1} For a comparable example of mini, see Gāndhārī Dharmapada, plate III, line 39.

^{page 529 note 2} The transfer from pen-writing to an inscription of such an ingressive junction is most unusual. On the other hand, at the foot of a vertical, a pen-drag in the direction of the following character is very common. Even in the Aśokan Kharoṣṭhī this feature of pen-writing is often carried over to the inscription; and although it appears only rarely on stone thereafter (see for example Konow, op. cit., plates XXXIV.1 and XXXVI.2,3: in the latter, artificially elaborated into an ornament which disguises its origin), it is relatively common on smaller inscribed objects, such as the Kaniṣka casket (ibid., plate XXV). In the present inscriptioII, an example with an acute angle can be seen in va in saṃvatsara'e, a rounded form in va in bhavatu; and several others. The appearance of such ‘semi-cursive’ junction-features is a natural development in rapid pen-writing in many scripts besides Kharoṣṭhī, and in itself hardly calls for comment. In the Dharmapada manuscript, successive characters are sometimes completely joined: for example, line 3 deśida; 9 -kamu; 277 medhavi; 287 vijadi; 342 mañati; and a single character such as vi, usually written with two strokes, may sometimes be produced without lifting the peII, with the result that a faint curve joins the bottom of the vertical to the top of the vowel stroke: line 289 vi. If this were slightly more frequent, we should probably agree to call such writing ‘cursive’; but in the circumstances a classification (necessarily arbitrary) into ‘cursive’ and ‘semi-cursive’ is not particularly useful. It should be said, however, that the term ‘cursive’ has already been much misused with reference to Kharoṣṭhī, and has shown a tendency to intrude even into palaeographic discussion in unsuitable applications, although we may presume that it was first applied to Kharoṣṭhī merely as a casual gesture by some amateur epigraphist in the nineteenth century who meant nothing very definite by the word. A particular character written with a rapid pen may indeed be more cursive than another in a more deliberate calligraphy. But the adjective has often been evoked by contrasts resulting more from the difference between pen and stilus than from contrasting manners of using the same instrument. And, with reference to the inscriptions, it seems an abuse of terminology, to put it no more strongly, to use the term at all when there is much less difference between the most ‘cursive’ of the epigraphic forms and the most rigidly ‘monumental’ than there is between the Latin square capitals and rustic capitals. Can we plead that the term might be kept in reserve, in case one day a manuscript should be found written in a style which might more appropriately be designated ‘cursive’?

^{page 529 note 3} Stein, A., Ancient Khotan, 369–70Google Scholar; Rapson, E. J., in Kharoṣṭhī inscriptions, III, p. 325Google Scholar.

^{page 530 note 1} Among those of which photographs have been published elsewhere, one which may be some 15 years earlier (no. 656) can be seen in Stein, Serindia, plate XXVI.

^{page 530 note 2} The length of 70 years suggested for the period in question depends on assumptions which, though not proved, seem sufficiently probable to allow us to adopt this figure as a working hypothesis. On the same assumptions, and with the addition of four or five years at either end to allow for the probability that a few of the many undated documents may lie outside the limits of the extreme regnal years, the period covered by the total collection in the edition would be approximately A.D. 180–275 on the earlier reckoning, and approximately A.D. 230–325 on the later.

^{page 530 note 3} Rapson, , op. cit., p. 296Google Scholar, remarked in passing that this manuscript ‘cannot be much later than the Niya documents’, which is surprising. No reason was given, however, and we are left to suspect that ‘later’ may have been a slip of the pen. The arguments for a later date on linguistic grounds (Burrow, T., BSOS, VIII, 2–3, 1936, 427 ff.Google Scholar) are inconclusive: see Gāndhārī Dharmapada, introduction, p. 55, n.