¹ Shackle, C., ‘“South-Western” elements in the language of the Ādi Granth’, BSOAS, XL, 1, 1977, 37–50Google Scholar: and ‘Approaches to the Persian loans in the ādi Granth’, BSOAS, XLI, 1, 1978, especially pp. 81–3.Google Scholar These are referred to here as ‘SWE’ and ‘APL’ respectively.

² In whose preparation I owe a particular debt of gratitude to Professor J. C. Wright, for many valuable criticisms and suggestions.

³ Outside this tradition, the attribution is of course regarded as more than questionable: cf. the most recent study of Jayadeva, , Miller, Barbara Stoler (ed. and tr.), Love song of the Dark Lord, Jayadeva's Gītagovinda, New York, 1977, p. 39Google Scholar, n. 1. The other hymn attributed to Jaidev in the AG—Mārū Jaideu (p. 1106)Google Scholar—is in the usual AG language: the remark of Macauliffe, M. A., The Sikh religion, Oxford, 1909, VI, 16Google Scholar, that it is ‘perhaps one of the most difficult of human compositions’ must be taken to refer to its complex metaphorical use of the technical vocabulary of yoga.

⁴ This uncertainty of attribution is not in itself remarkable within the AG, where there is a certain blurring of boundaries between the compositions attributed to the first two Gurūs: compare, e.g., the final salok of Gurū Nānak, 's Japu (p. 8)Google Scholar, which appears without any indication of separate authorship, with the almost identical verses specifically attributed to Gurū Aṅgad as Mājh vār M1, 18Google Scholar, salok M2, 2 (p. 146).Google Scholar The more or less slight differences between otherwise identical compositions appearing in different parts of the AG pose important textual problems which have never been squarely faced: the best known, and perhaps most illuminating, instance is the appearance of the So daru in three different versions (pp. 6, 8–9, 347–8).

⁵ The second member of each pair being written in the South-Western variety (‘SWE’, 37).Google Scholar

⁶ ‘SWE’, 38Google Scholar; ‘APL’, 74.Google Scholar

⁷ But their use in all printed editions suggests a substantial, if confused, MS tradition. As before, the readings of the Śabāarath Srī Gurū Granth Sāhib jī, fourth ed., Amritsar, 1969Google Scholar, are followed here.

⁸ It is often observed that one feature of the Sahaskritī salok is their partial freedom from the normal AG requirement of rhyme. But many of the salok do in fact rhyme, and in most other cases there is at least an assonance of this kind in final -ah, sometimes -aṃ.

⁹ A nice instance of the confusion to which usual Gurmukhi practice of writing separate letters, not conjuncts, can give rise to is provided by alapa, generally understood as ‘small’ (11, 46)Google Scholar [> alpa-], but apparently to be taken as ‘untouched’ in one instance (40) [? < alipta-].

¹⁰ Besides such forms without true OIA originals as thakanta (49) [CDIAL 13737, *sthakk-].

¹¹ More puzzling is the apparent addition of -r in cātrika- ‘papiha’ (J 12), drusaṭāṃ ‘enemies’ (37)—confused with √druh- ?, and the interjection dhrigaṃ (2), for the AG form dhrigu (63).

¹² The cluster is resolved in the rhyme-form samajayā ‘openly’ (48).Google Scholar

¹³ But compare nihaphalah (65)Google Scholar, and nihaphalaṃ (SS 2).

¹⁴ Compare, for example, the distorted rhyme-forms in -phā in the hymn Prabhamacr;tī M4, 5 (pp. 1336–7).Google Scholar

¹⁵ Though visarga is naturally lost in the usual AG dukha- (24, 40, etc.): an initial h- is lost in rida- ‘heart’ (13, 19, etc.), and the same change may account for rikhiaṃ ‘the senses’ (13) [? < √hṛṣ-].

¹⁶ ‘APL’, 76–8Google Scholar: compare also pp. 74–6 for many of the consonantal changes just described.

¹⁷ The table of AG declensions given in ‘SWE’, 40, may be consulted.Google Scholar

¹⁸ Other expansions are more puzzling, like garabatah ‘pride ?’ (J 10), and—especially— kuṭambyate, a genitive, ‘of the family’? (1).Google Scholar

¹⁹ Perhaps compare the odd rikhiaṃ (p. 301, n. 15, above).Google Scholar

²⁰ The only parallel to this stylistically inappropriate use of a Persian loan is [Ar. amr >] amara-hīṇaṃ ‘without authority’ (56).Google Scholar

²¹ For another small distinctive feature of these salok, compare the change t > th, described on p. 301 above.

²² Although this could also be taken as a straightforward realization of hdaya-. nisacai ‘surely’ is an unremarkable AG adverb [> niścaya-], occurring alongside nisacau (SS 1): they appear respectively as nisacaü and nihacaü in SV 1.

²³ Perhaps lokakah (15) represents a peculiar archaization of AG loka kā, rather than a mere metrical extension of the type described in p. 304, n. 18.

²⁴ A similar metrical adaptability is provided by the employment of many variants of the usual AG 3 sg. present termination -ai: this may appear as -i, -aī, or -ei.

²⁵ It is perhaps worth underlining the extent to which these versions, when compared with SS, show a much greater approximation to AG norms. As suggested in p. 298, n. 4, though, it is impossible to say whether SV is an imperfectly Sahaskritized version of SS, or whether SS is a systematization of SV made by Gurū Arjan when including them in the AG before his own salok Sahaskritī.

²⁶ Thus in one line it is 2 sg. and 3 sg.: bikhayanta jīvaṃ, vasyaṃ karoti, niratyaṃ karoti jathā marakaṭah ‘you [anger] get the sinful soul in your power, and it dances about like a monkey’ (47). Of course, if jīvaṃ, is taken as a pl., then the second instance of karoti will be 3 pl.!

²⁷ The Sikh religion, III, p. 430, n. 1.Google Scholar

²⁸ Singh, Kānh, Mohān koś, second ed., Patiala, 1960, 103Google Scholar, followed—as often—in the commentary of Singh, Sāhib, Srī Gurū Granth Sāhib darpaṇ, Jalandhar, 1971, X, 12.Google Scholar

²⁹ Only in the -oending, relatively prominent in G (cf. supra, p. 305)Google Scholar does the hypothesis of a double system of classicization, to Prakrit as well as to Sanskrit, seem at all attractive.

³⁰ Rāmakalī M1, 1, 1 (p. 876)Google Scholar: the only Sahaskritī form in this hymn is the rather common concluding formula praṇavati nānaku ‘Nānak humbly submits that …’.

³¹ So Śabdārath, ad loc.

³² So Mohān koś, 103Google Scholar, followed by Singh, Sāhib, VI, 442.Google Scholar

³³ The suspicious ring of over-ingenuity attaches itself unmistakably to the comment in Trumpp, E., The ādi Granth, London, 1877, pp. cxxxiii–ivGoogle Scholar: ‘As they are neither Sanskrit Shlōks nor composed in the Shlōka metre the question is, what is meant by salok Sahaskritī. The easiest solution would be to understand by it the Sanskriti metre, which is much used in Prākrit. In Prākrit poetry the Sanskriti is a stanza consisting of four verses, which together contain the number of 96 moras, each verse averaging between 18 and 28 moras. Very likely by the word Sahaskritī a variation of this metre is intended, as the Slōks in the Granth do not contain the same number of moras [!]’. As Trumpp later admits, there is no uniform metre to be discerned in the salok Sahaskritī (although the ending is general), nor is the number of lines constant.

³⁴ Śabdārath, p. 1353Google Scholar (cf. p. 705): the example is unfortunately chosen, since karanti nowhere occurs in the AG, which has only karanta, and karoti (see p. 308, n. 26).Google Scholar The use of the term ‘Gāthā’ as an assumed language-name is reverted to below.

³⁵ AG, pp. 938–46Google Scholar, where Sahaskritī forms include the nominal -aṃ, as in alāraṃ (8)Google Scholar; sunnaṃ, punnaṃ (51)Google Scholar; alakhaṃ (59)Google Scholar: many nouns in -o, as saṃsāro, adhāro (58)Google Scholar; nāmo (59)Google Scholar; ahaṅkāro, (60)Google Scholar: and a few 3rd sg. verb forms, as utarasi (18)Google Scholar, jāpasi (50, 66)Google Scholar, besides the common praṇavati (49).Google Scholar The picture given of the Gorakhnāthī yogīs in Briggs, G. W., Gorakhnāth and the Kānphaṭa yogīs, repr., Delhi, 1973 (especially, e.g., p. 251)Google Scholar encourages one to think that most of them would be much more at home in Sahaskritī than in Sanskrit, although this may be an unfair extrapolation of twentieth-century observations to fit early sixteenth-century circumstances!

³⁶ The word sahaskta- has a long ancestry in a comparable sense, being listed in Grassmann, H., Wōrterbuch zum Rig-Veda, fourth ed., Wiesbaden, 1964Google Scholar, S.V., as ‘kräftig gemacht, gekräftigt, von Indra’, although this pedigree hardly explains how it reached Gurū Nānak! An interesting etymological analogy is suggested by the AG word sahasā [< saṃśaya-], which appears more frequently than the expected sansā, but a direct derivation of sahaskritī from saāhib Singh, X, 12: ‘sahas is the Prakrit form of Skt. saṃs, just as sahasā is the Prakrit form of saṃśaya.… The word ‘sahaskritī’ is the Prakrit form of samstta, and all these salok are in the Prakrit language’.

³⁷ The Sikh religion, III, p. 434, n. 1.Google Scholar

³⁸ Mohān koś, 303Google Scholar, followed by Singh, Sāhib, X, 54.Google Scholar

³⁹ Śbdārath, p. 1360.Google Scholar

⁴⁰ Trumpp, , pp. cxxxvii–viiiGoogle Scholar, recalling the convoluted argument quoted in p. 310, n. 33, above. Metrical analysis is unable to establish any clear-cut distinction between the salok Sahaskritī and the Gāthā, or, indeed, a common pattern for either.

⁴¹ The relevant passages are Amacr;sa M5, 22Google Scholar, lr. (p. 376)Google Scholar: rāri karata jhūṭhī lagi gāthā = ‘false talk’; Mārū M5, 30, 2 (p. 1007)Google Scholar: suṇi nānaka jīvai gātha [sic] = ‘teaching’; and three examples in the Gāthā M5 themselves, where the word clearly appears as an alliterative line-heading, and seems to be used in the sense of ‘praise’. The three passages are: gāthā gumpha gopāla-kathaṇ (G 6); gāthā gūṛa apāraṃ (G 10); and gāthā gāvanti nānaka ‘Nānak sings His praises’ (G 18).

⁴² Kohli, S. S., A critical study of Adi Granth, New Delhi, 1961Google Scholar, is rather unconvincing in the suggestion that Jai. is Sanskritized Eastern Apabhraṃśa (pp. 31–4), while the other Sahaskritī compositions are examples of a Sanskritized Western Apabhraṃśa (pp. 34–40): the final definition of the latter as an ‘imitation Sanskrit’ is nearer the mark.

⁴³ cf. the phrase guramukhi nādaṃ, guramukhi vedam, in Japu, 5 (p. 2)Google Scholar; or the verbal 3 sg. ending -asi in tūṭari, chūṭasi, in Bilāval M1 thitī, 2 (p. 839)Google Scholar. There is a particular concentration of such forms in the hymn headed Gūjrī M1 a., 5 (pp. 505–6)Google Scholar, referred to below in p. 312, n. 46.

⁴⁴ If a local term is preferred for this, if only for reasons of symmetry, Arjan's, Gurū own ‘ḍakhaṇa’Google Scholar (‘SWE’, p. 37Google Scholar, and n. 8) is the obvious choice.

⁴⁵ It needs to be said that the corpus used in the preparation of ‘SWE’ was a rather restricted one, and that the South-Western variety is in fact rather more fully represented in AG, particularly in the compositions of Gurū Nānak, than was suggested there. To the compositions of the two Gurus analysed in ‘SWE’—Sirī MS ch., 3, salok (pp. 80–1)Google Scholar; Jaitsarī vār MS, salok (pp. 705–10)Google Scholar; Mārū M1 a., 9–10 (pp. 1014–15)Google Scholar; Mārū vār M5, salok (pp. 1094–1105)Google Scholar—the following may be added: Sirī M1, 24 (p. 23)Google Scholar; Sirī M1 a., 5 (pp. 55–6)Google Scholar; Gaüṛī vār M5, salok (pp. 318–23), 19, 1 (p. 322)Google Scholar actually being headed salok ḍakhaṇā; Āsā M5, 105, 107, 110 (pp. 397–8)Google Scholar; Āsā M1 a., 14 (p. 418)Google Scholar; Gūjrī vār M5, salok (pp. 517–24)Google Scholar; Vaḍhans M1, 1–3 (pp. 557–8)Google Scholar; Vaḍhans M5 ch., 1–2Google Scholar, salok (pp. 576–8)Google Scholar; Jaitsarī M5 ch., 1 (pp. 703–4)Google Scholar; Sūhī M5 a., 4 (p. 761)Google Scholar; Sūhī M1 kucajjī, 1 (p. 762)Google Scholar; Sūhī M5 guṇavantī, 3 (p. 763)Google Scholar; Sūhī M5 ch., 1 (p. 777)Google Scholar; and Rāmakalī vār M5, salok (pp. 957–66)Google Scholar. The detailed findings of ‘SWE’ are, however, also applicable to this quite large body of additional material. The whole corpus is examined from a stylistic viewpoint in my later article, ‘The South-Western style in the Guru Granth Sahib’, Journal of Sikh Studies, V, 1, 1978.Google Scholar

The greater consistency of Gurū Arjan emerges in his handling of this and the Sahaskritī varieties, while in Torkī the exemplar is Nānak', Gurūs Tilaṅg M1, 1 (p. 721)Google Scholar, quoted and translated in ‘APL’, 82.Google Scholar

⁴⁶ The traditional assumption is that the bhagat bāṇīwas introduced into the Sikh scriptural tradition in the mysterious volumes known as the Goindval, pothīsGoogle Scholar, compiled under the auspices of the third Gurū Amar Dās (d. 1574): cf. McLeod, W. H., The evolution of the Sikh community, Oxford, 1976, 60–1Google Scholar. These pothīs were then used by Gurū Arjan in his compilation of the AG in 1603–4.

⁴⁷ The sources from which the bhagat bāṇī entered the Goindval pothīs must remain speculative. It is of course true that Gurū Nānak must have been familiar with the poetry of the Bhagats, and creative imitation would seem a more likely hypothesis than independent creation of the different stylistic varieties: but, when Gurū Nānak writes better pseudo-Persian than ‘Kabīr’, whose was the imitation?

The attempt is made, in Singh, Sāhib, IV, 181–2Google Scholar, to establish that Jai. furnished a direct model for hymn, Gurū Nānak's, Gūjrī M1 a., 5 (pp. 505–6)Google Scholar: but the case is not really made. This hymn does in fact have only a few Sahaskritī features, including—beside the quite frequent -aṃ ending— just ārādhitaṃ (1), [vyāpta- >] biāpitaṃ (5), and the unusual 3rd sg. bharamāti ‘wanders’ (4).

⁴⁸ ‘SWE’, 36–7.Google Scholar

⁴⁹ The Sikh religion, I, p. viGoogle Scholar: ‘Hymns are found in Persian, medieval Prakrit, Hindi, Marathi, old Panjabi, Multani, and several local dialects. In several hymns the Sanskrit and Arabic vocabularies are freely drawn upon’.

⁵⁰ It has not been thought necessary to illustrate this concern here, but an obvious example is Gurū Nānak's castigation of Brahmanical dress and ritual in SS 1/SV 1 (translated in Macauliffe, , I, 237)Google Scholar: an idea of the themes treated by Arjan, Gurū in his salok SahaskritīGoogle Scholar; and gāthā can be gained from the selected translations given in Macauliffe, , III, 430–8.Google Scholar

⁵¹ ‘APL’, 88 ff.Google Scholar