Quick viewing(Text Mode)

KHAZAR TURKIC GHULÂMS in CALIPHAL SERVICE1 Arab

KHAZAR TURKIC GHULÂMS in CALIPHAL SERVICE1 Arab

KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE1

BY

PETER B. GOLDEN

Arab — Khazar relations, in particular their military confrontations for control of the , have long been the subject of investigation. As a consequence of that sustained period of military encounters, the enjoyed a fierce reputation in the Islamic world. Thus, al-Balâd- hurî, in his comments on the massacre of some of the populace of MawÒil () in the aftermath of the ‘Abbâsid takeover in 133/750, says that “because of their evil, the people of Mosul were called the Khazars of the .”2 Much less studied is one of the side-products of those wars: the presence of Khazars in the itself. As with other peoples captured in warfare or taken or purchased on the periphery of the Caliphate, the Khazars in the Islamic heartlands were largely, but not exclusively, used as professional military men, slave-soldiers of the Caliphs, usually termed ghilmân (sing. ghulâm, lit. “boy, servant, slave”3). The relationship of these aliens to the larger Arab and Arabized

1 I would like to thank Michael Bates and Matthew Gordon for their thoughtful read- ings of this paper. Needless to say, I alone am responsible for any remaining errors or failure to take their advice. An earlier, non-updated Russian version of this article (“Khazarskie tiurkskie guliamy na sluzhbe Khalifata) is in press in Moscow, to be pub- lished in Khazary, Trudy Vtorogo Mezhdunarodnogo kollokviuma, ed. by I.A. Arzhant- seva, V.Ia. Petrukhin and A.M. Fedorchuk. 2 This passage is cited and translated in C.F. Robinson, Empire and Elites After the Muslim Conquest. The Transformation of Northern Mesopotamia (Cambridge, 2000), p. 138; the text is in al-Balâdhurî, Ansâb al-Ishrâf, ed. ’A. ad-Dûrî, Bibliotheca Islamica (Wiesbaden/Beirut, 1398/1978), III, p. 281. 3 H. D. Yıldız, Islâmiyet ve Türkler (Istanbul, 1976), p.81, suggests that this term did not really mean “slave,” but rather one who was at the disposal of another, commenting that it has different connotations from Arab. ’abd or raqîq both of which denote “slave.” These terms, he remarks, are never used with regard to the Turkic ghilmân. Yıldız further notes (pp. 84-86) that “Taking all these possibilities into account, the hitherto held view regarding the origin of the Turkic units in the ‘Abbâsid army, that is that they were slaves

Journal Asiatique 292.1-2 (2004): 279-309 280 P.B. GOLDEN society of the Near East was complicated. Al-JâÌi (d. 870), the brilliant essayist of the ‘Abbâsid era who was concerned with ethnic questions, using traditional genealogical formulas for the integration of aliens into Islamo-Arab society, pointed to the alleged common descent of the Turks and Arabs from Abraham/Ibrâhîm. The older Arab institution of clientage (walâ'), by that time the chief means of “naturalizing” non- Arab (Arabizing) ethnic elements into the Caliphal state, was also brought into play4. The ‘Abbâsids used a combination of patronage (iÒ†inâ') and military slavery to assure the loyalty of the servitors of their regime5. The beginning of the involvement of Turkic ghilmân in the affairs of the ‘Abbâsid Caliphate can be dated to the struggle of al-Ma'mûn (reg. 813-833) with his brother al-Amîn (reg. 809-813) for the rulership6. Operating out of Central Asia, al-Ma'mûn (who was in Marv when his who could be bought, should, in our opinion, be met with doubt.” Rather, he concludes, these were hired soldiers. This is not entirely accurate. Al-IÒ†akhrî, Kitâb Masâlik al- Mamâlik, ed. M.J. De Goeje (Leiden, 2nd ed., 1927), p. 318, refers directly to the raqîq Mâ warâ’n-Nahr (“the slaves of ”), one of the leading sources of the Turkic ghilmân, see below, note 11. Clearly, the ghilmân were viewed as slaves — although their social standing as military men was undoubtedly higher. A view very similar to that of Yıldız was put forward at about the same time by M.A. Shaban, see discussion below. 4 The essay, Manâqib Jund al-Khilâfa wa Fa∂â ‘il al-Atrâk, see Hilâfet Ordusunun Menkıbeleri ve Türkler’in Fazîletleri, . trans. R. ≤e≥en (Ankara, 1967), is not free of promoting certain ideological positions (it was addressed to the powerful Turkic courtier, al-FatÌ b. Khâqân) and has to be used with some caution, see discussion in J. Lassner, The Shaping of ‘Abbâsid Rule (Princeton, 1980), pp.116ff. The term walâ’ has a range of meanings. In addition to the older legal term denoting “clientage”(see Vl. V. Polosin, Slovar’ poetov plemeni ‘abs VI-VIII vv. Moskva, 1995, p. 523: “pokrovitel’stvo, otnosheniiia pokrovitelia i klienta; druzhba”) it also expresses “friendship, amity, benev- olence, good will, fidelity, fealty, allegiance, devotion, loyalty” in Modern Arabic, see J.M. Cowan (ed.), The Hans Wehr Dictionary of Modern Written Arabic, 4th ed. (Ithaca, 1994), p. 1289. In the medieval Islamic world it denoted the ongoing bonds of loyalty and allegiance between a freedman (mawlâ, pl. mawâlî) and his onetime master. Converts from the non-Arab population also entered into a walâ’ relationship with their Arab “patron,” see D. Pipes, Slave Soldiers and (New Haven, 1981), pp. 95, 107-109. 5 R.P. Mottahedeh, Loyalty and Leadership in an Early Islamic Society (Princeton, 1980), pp.82-83. 6 Turkic military slaves or Turks serving in a military capacity are sporadically noted in the pre-‘Abbâsid period, see Yıldız, Islâmiyet ve Türkler, p. 47; D. Pipes, “Turks in Early Muslim service” Journal of Turkish Studies 2 (1978), pp. 86-97; M.S. Gordon, The Breaking of a Thousand Swords. A History of the Turkish Military of Samarra (A.H. 200- 275/815-889 C.E.) (Albany, 2001), pp. 6, 162, n.21.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 281 father, Hârûn ar-Rashid died) was the first Caliph to gather to his person a substantial number of Turkic ghulâms7. His other brother and ally, al-Mu‘taÒim (reg. 833-842), early in Ma'mûn’s reign, began to build up a private military establishment8, often recruiting locally Turkic slaves who were already in serving in other households. These included the Khazars Îtâkh9 (purchased in 199/815) and Bugha10, the

7 Al-Mas‘ûdî, Murûj adh-Dhahab wa Ma‘âdin al-Jawhar, ed. Ch. Pellat (Beirut, 1962-1979), IV, p. 261. In the course of his struggle for the Caliphate with his brother al- Amîn, al-Ma’mûn even contemplated seeking the protection of “Khâqân, the king of the Turks,” see a†-™abarî, Ta’rîkh a†- ™abarî. Ta’rîkh ar-Rasûl wa’l-Mulûk, ed. M. Ibrâhîm (Cairo, 1967-69), VIII, pp. 403-404, Eng. trans. The History of al-™abarî, ed. E. Yar- Shater et al., 38 vols. (Albany, 1985-1992), vol. XXXI, The War between Brothers, trans. M. Fishbein (Albany, 1992), pp. 71-72. After initial full citation, all future citations to this series will give translator’s name, volume number, abbreviated title. 8 It has been suggested that al-Mu ’ta Òim’s choice of Turks for these units stemmed from his mother’s alleged Turkic or part-Turkic origins. This, however, has not been demonstrated, see discussion in H. Töllner, Die türkischen Garden am Kalifenhof vom Samarra (Beiträge zur Sprach- und Kultturgeschichte des Orients, 21 (Walldorf-Hessen, 1971), pp. 20-21. 9 This is a diminutive form deriving from Turkic it/ït “dog” = ïtaq “little dog,” see M. Erdal, “Ein unbermerkter Chasarische Eigenname” Türk Dilleri Ara≥tırmaları 1991 (Ankara, 1991), pp. 31-36. On these names, see P.B. Golden, “Khazar Turkic Ghulâms in Caliphal Service: Onomastic Notes” Archivum Eurasiae Medii Aevi 12 (2002-2003), pp. 15-27. 10 Turk. bugha “bull” (see Sir Gerard Clauson, An Etymological Dictionary of Pre- Thirteenth Century Turkish (Oxford, 1972, henceforth ED), p. 312. The evidence for Bugha’s Khazar ethnicity (he is also called at-Turkî in the Islamic sources) stems from an incident noted in the Georgian sources s.a. 851. Bugha had been campaigning in the Cau- casus, according to the K’art’lis Ts’xovreba, ed. S. Qaukhch’ishvili (T’bilisi, 1955), I, pp. 256-257) and the Caliph had become suspicious of him when “he learned that he was taking council with the Khazars, his tribesmen” (tomt’a mist’a). Georg. tomi can also mean “kinsman,” as well as “clan, family, (fellow) tribesman,” see I. Abuladze, Dzveli k’art’uli enis lek’sikoni, T’bilisi, 1973, p. 413. In Mod. Georg. (K. Tschenkéli, Geor- gisch-Deutsches Wörterbuch, Zurich, 1960-1974), Fasz. 15, p. 359) it denotes “(Volks-) stamm, Geschlecht, Sippe.” The Georgian evidence would appear to point to a more immediate blood relationship. AÒ-∑ulî, Kitâb al-Awrâq, ed. V.I. Beliaev and A. B. Khali- dov (St. Petersburg, 1998), pp. 148/457, noted also in Gordon, Breaking, p.19), in Bugha’s obituary notice, says that he was a famous warrior among the Turks who then came into the possession of al-Fa∂l b. Sahl, one of the men close to al-Ma 'mûn who was associated with Khurâsân. This would imply that he was taken in warfare in Central Asia (I am also indebted to Michael Bates for bringing this passage to my attention and for making this valuable source available to me). The language of the notice is vague. The term “Turks” included the Khazars. It is not clear when and how al-Fa∂l b. Sahl acquired him. The notice further reports that Bugha escaped from captivity but was recaptured by Ghassân b. ‘Abbâd, who was governor of Khurâsân 202-205/818-821 for al-Ma’mûn

Journal Asiatique 292.1-2 (2004): 279-309 282 P.B. GOLDEN famous commander Ashinâs (acquired while al-Ma’mûn was still caliph) and others. The Eastern Iranian Sâmânids who controlled the Transox- anian marches for the Caliphate became the leading supplier and trainer of Turkic ghulâms11, assuring a steady supply of recruits. By the time of his accession, al-Mu¨taÒim had created a private army of some 3-4000. Some of these soldiers were slaves, others, such as al-Afshîn, the scion of a ruling Central Asian (Ustrûshana/Ushrûsana) Iranian family, clearly were not12. The training of these slaves, noted by al-IÒ†akhrî and Ibn Îawqal, followed older Soghdian models which emphasized both martial arts and the formation of educated and well-bred courtiers13. The

(see Gordon, Breaking, pp. 19 and n. 49, p. 172) which would seem to place him there. However, he may have entered the Caliphate from Khazaria and his escape came while with al-Fa∂l b. Sahl in Khurâsân. Khazar slaves also came to the Caliphate from Central Asia (Khwârazm, see below). In any event, al-Mu¨taÒim, we are told, purchased him “and his sons,” from al-Îasan b. Sahl for 10,000 . Were these children born in captiv- ity or was the entire family enslaved at the same time? Our source provides no informa- tion on this. 11 Al-IÒ†akhrî, ed. De Goeje), p. 318 remarks that “in Samarqand is the place of assembly of the slaves (majma‘raqîq) of Transoxiana and the best of the slaves of Tran- soxiana are (those) schooled (in) Samarqand;” see also Ibn Îawqal, Kitâb ∑ûrat al-Ar∂ (Beirut, 1992), p. 407. D. Ayalon, “The Mamlûks of the Seljuks: Islamic Military Might at the Crossroads” Journal of the Royal Asiatic Society, 3rd ser. 6/3 (1996), p. 313, cites this passage and remarks that these slaves were superbly outfitted. See also H. Kennedy, The Early (London, 1981), p.167. 12 On the Iranian title afshîn, see F. Justi, Iranisches Namenbuch (Marburg, 1895, reprint: Hildesheim, 1963), pp. 252-253; R.N. Frye, The Golden Age of Persia. The Arabs in the East (London, 1975), p. 48. In purchasing as many ghulâms as he could acquire in Baghdad, al-Mu¨taÒim often pestered their owners into selling them to him, see al-Ya¨qûbî, Kitâb al-Buldân, ed. M.J. De Goeje (Leiden, 1892), pp. 253-256; Pipes, “Turks in Early Muslim Service,” Journal of Turkish Studies, 2 (1978), pp. 88-90; Y ıld ız, Islâmiyet ve Türkler, pp. 66-79 (who estimates that al-Mu‘taÒim eventually built up a force of 25,000 Turkic military slaves); H. Kennedy, The Prophet and the Age of the (London, 1986), pp. 158-159; Lassner, Shaping, pp. 113-114. This slave bodyguard was associated with al-Mu‘taÒim as early as 818. These included such subse- quently renowned figures as Ashinâs/Ashnâs (who was given the Persian name Ashinâs when he saved Mu‘taÒim’s life in combat against Kharîjî rebels, see a†-™abarî, Ta’rîkh, ed. Ibrâhîm, VIII, p. 558, Eng. trans. The History of al-™abarî, ed.Yar-Shater et al., vol. XXXII, Storm along the Northern Frontiers of the ‘Abbâsid Caliphate, trans. C.E. Bosworth (Albany, 1987, henceforth cited as ™abarî/Bosworth, XXXIII, Storm, pp. 67- 68), Îtâkh, WaÒîf and Sîmâ ad-Dimishqî. Not surprisingly, given the importance of the Sâmânids in training and supplying the Caliphate with Turkic ghilmân, Persian appears to have been spoken by a number of the prominent Turks (including Khazars) in ¨Abbâsid service, see discussion in Gordon, Breaking, pp. 18, 103, 137. 13 See E. De la Vaissière, Histoire des marchands sogdiens (Paris, 2002), p. 305.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 283 recruitment of these Central Asian Iranian “volunteers” was begun by al-Ma'mûn14. Clearly, there were two distinct sources of acquiring troops that came from the eastern borderlands of the Islamic world: slaves (largely from Turkic Eurasia) and professional soldiers from - ian Transoxiana (including both Soghdian and Turkic military men). Contemporary Byzantine rulers were engaged in similar activities. The , formed ca. 840, was one of the groupings of the Imperial bodyguard. It consisted of three subgroupings, one of which was largely composed of Khazar and Farghânian (Fargánoi) mercenaries15, the same sources of recruitment that we see for the Caliphal ghulâms. The ninth century Kletorologion speaks of the “foreigners” (êqnikoí of the Hetai- reia, such as Turks, Khazars and the rest, numbering 54.”16 Al-Maqdisî reports that the “majority of the ghilmân of the King (= the Byzantine ruler, pbg) are Turks and Khazars.”17 Ibn Rusta places the number of Khazar and Turkic ghilmân in Byzantine service at 10,000. He also notes the presence of Khazar bowmen at the “Mankabâ” Gate (very likely ÉÑ∏µæªdG al-Manklabâ = tò maglábion or magklábion) who were probably

14 Al-Balâdhurî, FutûÌ al-Buldân, ed. R.M. Ra∂wân (Cairo, 1959), pp.418-419; H. Kennedy, The Armies of the Caliphs (London-New York, 2001), pp. 124-125. 15 W. Treadgold, Byzantium and its Army 284-1081 (Stanford, 1995), pp. 110,115; M. Whittow, The Making of Byzantium 600-1025 (Berkeley, 1996), pp.169-170; A. Kazhdan (ed.), The Oxford Dictionary of Byzantium, (Oxford, 1991, henceforth ODB), 2, p. 925. See also J. Marquart, Osteuropäische und ostasiatische Streifzüge (Leipzig, 1903), p. 227 and D. Ludwig, Struktur und Gesellschaft des Chazaren-Reiches im Licht der schrift- lichen Quellen (Münster, 1982), p. 205; Constantine Porphyrogenitus, De cerimoniis aulae byzantinae, ed. I.I. Reiske (Bonn, 1820), II.15, p. 576, noting the “Farghânians and Khazars…all bearing swords and carrying shields;” II.44, pp. 660-661, where he men- tions the 45 Farghanians, 47 Khazars and 84 Turks in the royal Hetaireia. Farghanians and Khazars are also noted in II.49, p. 693 and II.52, p. 749 where a fuller listing of the foreign troops serving in the imperial guard is given: “Farghanians, Khazars, Hagarenes (Arabs, Muslims), Franks (Frággoi, i.e. Western Europeans)” and others who draw salaries from the imperial beneficence. 16 A Klêtorologion is a treatise giving listings of the various offices, titles and ranks of state of the Byzantine administrative hierarchy, see J.B. Bury, The Imperial Adminis- trative System in the Ninth Century with a Revised Text of the Kletorologion of Philoth- eos (London, 1911, reprint New York, n.d.), p. 171. This phrase is virtually identical to the comments by Constantine Porphyrogenitus, De cerim., ed. Reiske, II.52, p. 772. Khazars are also listed (p. 160 of the Klêtorologion) among the “foreigners” in royal ser- vice who receive salaries. 17 al-Maqdisî, Kitâb al-Bad’ wa’l-Ta’rîkh, ed. Cl. Huart (Paris, 1899-1919), IV, p. 68.

Journal Asiatique 292.1-2 (2004): 279-309 284 P.B. GOLDEN part of the Hetaireia18. It is hard to square this number with the rather modest figures noted by the Byzantine sources. The latter, however, may have only recorded the actual numbers serving in the Emperor’s guard. Others may have served in other units. Nonetheless, it is unlikely that their numbers were as high as Ibn Rusta reports. It is interesting to note here that al-Maqdisî and Ibn Rusta call these foreign professional military men in Byzantine service ghilmân, a term of some ambivalence (see below). In any event, Muslim contemporaries, it would appear, saw these groups in Byzantine and ‘Abbâsid societies as essentially the same. When these foreign troops do come into the purview of the Byzantine historians, they appear to be associated with the ruler’s personal forces. Thus, we are told by the anonymous “Continuator” of Theophanes19, that some Khazar troops in the personal guard of the Emperor, Leo VI (886-912) were cap- tured in a battle with the Bulgarian tsar, Symeon (893-927) in 89420. The Îamdânid poet Abû Firâs (d. 357/968, his mother was a Byzantine and he had spent some time as a captive in ) mentions a certain “Tuznîq [≥«fõJ] the Khazar, commander of the Khazars” against whom he broke two lances in combat in 347/95821. Presumably, this Tuznîq was a commander of the Khazar contingent in the Byzantine forces of that era (the age of Constantine VII Porphyrogenitus, 913-959). We have no direct information on how these Khazars entered Byzantine service. As we know from the mid-tenth century Khazar Hebrew correspondence there had been some considerable enmity between Constantinople and Khazaria several decades earlier22. Constantine’s , had a decidedly anti-Khazar orientation. ****

18 Ibn Rusta, Kitâb al-A‘lâq an-Nafîsa, ed. M.J. De Goeje (Leiden, 1892), pp. 120, 124; Marquart, Streifzüge, pp. 219, 226 19 Actually a composite work of several authors, the last of whom wrote not long after 961, see ODB, 3, pp. 2061-2062. 20 These captives were subjected to mutilation, their noses being cut off, and were then sent to Constantinople to humiliate the Byzantines, Theophanes Continuatus, Histo- riae, ed. I. Bekker (Bonn, 1838), p. 358. 21 S. Dahan (ed.), Le Diwan d’Abu Firas al-Hamdani (Damascus, 1363/1944), II, p. 279; see also A.A. Vasiliev, Byzance et les arabes (Bruxelles, 1935, 1950, 1968), II/2, pp. 349-350, 369 who reads this name as T.zî.b.q (≥«fõJ) 22 See N. Golb, O. Pritsak, Khazarian Hebrew Documents of the Tenth Century (Ithaca, 1982), pp. 112ff., 136.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 285

In the Islamic orbit, M.A. Shaban, Richard Frye, Christopher Beck- with and others have argued23 that this essentially comitatus-type struc- ture, had at its root an older, Central and Inner Asian guard-corps system known in Iranian (Soghdian) Central Asia as the châkars24 [cf. Pers. châkar “a servant, apprentice; maid,”25 Pahl. châkar, Oset. tsaghar “rab,” Soghd. c’gr, Scythian qiagarov ]26. Chinese sources note this institution (Zhe-jie: Early Middle Chinese: *tçiah *kiat27 = châkar) among the ruling and mercantile elite of Soghdia28. Narshakhî, in his History of Bukhârâ remarks that “two hundred youths,” from the aris- tocracy, “girded with gold belts and swords” reported daily to the ruler of that city “for service.”29 Shaban suggests that the “whole institution was now transplanted from the East into the heart of the empire. The châkars continued in the service of their own leaders, who in turn were created mawâlî of the ruler.”30 Thus, this personal guard of al-Ma'mûn and his brother al-Mu‘taÒim, was adapted to the older Arab system of

23 M.A. Shaban, Islamic History. A New Interpretation, 2, A.D. 750-10554 (A.H. 132- 448) (Cambridge, 1976), pp. 63-66; R. Frye, History of Ancient Iran (München, 1984), pp. 352-3 and his Heritage of Central Asia (Princeton, 1996), pp. 195-196; C. Beckwith, “Aspects of the Early History of the Central Asian Guard Corps in Islam,” Archivum Eurasiae Medii Aevi, IV (1984), pp. 29-43. 24 De la Vaissière, Histoire, p. 282, who notes the importance of “Soghdian military elites” modelled on the châkar (“gardes du corps nobles du souverain”) in al-Ma’mûn’s entourage after his stay in Merv and subsequent assumption of the caliphal throne.. 25 F. Steingass, A Comprehensive Persian-English Dictionary (1892, reprint: Beirut, 1970, henceforth noted as Steingass), p. 386; A.A. Kerimova et al., Giliansko-russkii slovar’ (Moskva, 1980), p. 59: châk¢r “sluga, rab.” 26 V.I. Abaev, Istoriko-etimologicheskii slovar’ osetinskogo iazyka (Moskva- Leningrad, 1958), I, 286, who derives it from Iran. kar-“to do.” 27 See E.G. Pulleyblank, Lexicon of Reconstructed Pronunciation of Early Middle Chi- nese, Late Middle Chinese, and Early Mandarin (Vancouver, BC, 1991), pp. 154, 401. 28 See S. Beal, Si-Yu-Ki. Buddhist Records of the Western World (London, 1884 ), I, p. 33 (Xuanzang) and E. Chavannes, Documents sur les Tou-kiue (Turcs) Occidentaux suivi de Notes Additionnelles (Paris, 1941), pp. 137, 147, 313; Beckwith, “Aspects,” p. 37. 29 Abu’l-Bakr MuÌammad Narshakhî, Ta‘rîkh-i Bukhârâ,ed. M.M. RaÂavî (Tehran, 1363/1943-44, reprint n.d.), pp. 16-17, Eng. trans.: R. N. Frye, trans. The History of Bukhara (Cambridge, Mass., 1954), p. 9. V.V. Bartol’d (W. Barthold), v epoxu mongol’skogo nashestviia in his Sochineniia (Moskva, 1963-1977), I, p.238, however, suggests that this was only an “honor guard” drawn from the aristocracy. This may have been true for a later period, but the ealier Chinese accounts underscore the military prowess of the Zhe-jie. 30 Shaban, Islamic History, 2, pp. 64-65.

Journal Asiatique 292.1-2 (2004): 279-309 286 P.B. GOLDEN walâ' and Islamicized upon their return to Baghdâd. Here, châkar was rendered as ghulâm. In this regard, Shaban suggests that these “ser- vants” were really the equivalent of “European knights” and the use of servile terminology (e.g. khâdim, waÒîf, ‘abd) was simply a means of expressing loyalty31. In Beckwith’s view, al-Ma'mûn acquired “a guard corps of châkars” as “a result of his long-term residence in Central Asia itself.”32 The “deeply ingrained concept of the guard corps bondage system” was adapted by the Arabs “to their native walâ' system of bondage.” A great variety of terms came to be employed to designate these various alien groups in caliphal service, reflecting how these military bondsmen “could be fitted into the existing Islamic social system.”33 Beckwith concluded that “the Turks and Sogdians of Central Asia passed on the guard corps, as part of their common heritage, to the Arabs, who then made it ‘Islamic’; it was not the other way around.”34 Matthew Gordon has disputed these findings, noting that there is a clear distinction between the Shâkiriyya (the Arabic rendering of châkar), which is often linked with the Jund (“army”) and as such were free and professional soldiers35 and the Turks who were subsequently found as caliphal ghulâms in Sâmarrâ. Many of the Turks began their careers as slaves. Moreover, it is not at all clear that al-Ma’mûn brought back a ready-made Turkic châkar corps when he secured his hold on the caliphate. Rather, Gordon contends, the “initial phase of the guard’s his-

31 Shaban, Islamic History, 2, pp. 63-4. His assertation that the bulk of the ghulâms were not slaves and were not purchased has been sharply criticized by Beckwith, see above. F. Amabe, The Emergence of the Abbâsid Autocracy: The ‘Abbâsid Army in Khurâsân and Adharbayjân (Kyoto, 1995), pp. 140-169, is also inclined towards Sha- ban’s view. See a balanced discussion in Kennedy, The Armies of the Caliphs, pp. 121- 122, 143 n. 26. Kennedy, although finding Amabe’s arguments (see above) “in some ways persuasive” remains nonetheless cautious with regard to his use of the evidence. He concludes that the “evidence is not clear either way.” 32 Beckwith, “Aspects” AEMAe, IV (1984), p. 38. 33 Ibid, pp. 38-39. 34 Ibid, p. 40. 35 Gordon, Breaking, pp. 40-41. Gordon concludes that the Jund at this stage probably consisted largely of non-Arab elements. See also Kennedy, The Armies of the Caliphs, p. 31 who notes that the term also designated the districts in which the soldiers were sta- tioned.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 287 tory” is to be placed “within the confines of the Islamic Near East.” Although, he does not exclude Turkic and Central Asian influences in shaping the caliphal guard corps, Gordon is inclined to view them as “elements” that “may simply have struck a familiar chord” in their Near Eastern milieu. Al-Mu‘taÒim, with whom this institution flourished and matured, had never spent time in Central Asia. Inspiration, Gordon con- cludes, came from the Near Eastern traditions with which he was famil- iar.36 As noted earlier, there can be no doubt that some of the figures asso- ciated with the ghulâm forces who were later prominent in Baghdad and Sâmarrâ came into the Islamic world as slaves. Other prominent mili- tary-political figures (such as Afshîn) did not. Moreover, the ghulâms, after their purchase, appear to have been manumitted and brought into or at least presented as part of the clientage system (walâ'). As Kennedy remarks, they were not called mamlûks or ‘abîd, the usual terms denot- ing “slave.” Instead, we find the term ghulâm, a term which does not necessarily mean that they were “unfree.”37 Given the enormous influence of Iranian governmental and other usages and the not inconsiderable cultural baggage that the Turks, what- ever their status, brought with them, one must allow, perhaps, for greater Turko-Iranian influences than Gordon seems prepared to accept. The language used to describe this institution and those directly involved in it is instructive. In Pre-Islamic Iran, the ruler was surrounded by elite groupings of warriors, e.g. the asbârân (“cavalrymen, knights”), who, like the members of the Germanic tribal comitatus, “were bound to the king by ties of inviolable loyalty; they pledged themselves to remain faithful to their chief even unto death.”38 The Sâsânid Shâhs had a “select company” of royal bandagân (a term denoting “bondsmen”39)

36 Ibid, pp. 7-8, 156. 37 Kennedy, The Armies of the Caliphs, p. 122. 38 M. Zakeri, Sâsânid Soldiers in Early Muslim Society. The Origins of ¨Ayyârân and Futuwwa (Wiesbaden, 1995), pp. 57-59, 66, 69-70. 39 The use of the term bandagân (Mod. Pers. sing. banda “bondsman, servant, slave” > Old Pers. sing. bandaka “subject, servant” < Old Pers. band- “to bind,” see Steingass, p. 202; R. G. Kent, Old Persian. Grammar, Texts, Lexicon, rev. 2nd ed. (American Ori- ental Series, 33, New Haven, 1953), p. 199; V.S. Rastorgueva and D.I. Edel’man, Etimo- logicheskii slovar’ iranskikh iazykov (Moskva, 2000-2003), 2, pp. 68-75, cf. bandaka

Journal Asiatique 292.1-2 (2004): 279-309 288 P.B. GOLDEN drawn from the asbârân who “formed an imperial guard (jund-i shâhan- shâh), a corps of cavalry, probably consisting of ten thousand, called the ‘corps of the immortals’” a name that harkens back to Achaemenid times40. Sâsânid bandag (= Aramaic ‘abdâ “servant, slave”) denoted “all forms of subordination of a servant to his master including the ser- vice rendered at the court to the king as his companion.”41 In the Fârs- nâma of Ibn al-Balkhî, it is noted that the Sâsânid Shâhs “treated all generals, officers, and rank and file of the army as if they had bought them as bandagân.”42 These royal servitors, moreover, were expected to wear distinguishing articles of attire (earrings and/or belts of servitude) attesting to their status. Extraordinary acts of service could be rewarded with an âzâd-nâma (“certificate of manumission”)43. The bandagân, thus, were not really slaves. In both Achaemenid and Sâsânid times servile terminology could designate very high officials personally bound to the ruler. This was the terminology of personal loyalty. In the pre- Islamic Turkic world, as I have noted elsewhere, “political dependence was expressed in the vocabulary of slavery.”44 A number of Turkic ghulâms had names which contain the title chor/chur45 which in the con- temporary Soghdo-Turkic milieu of Central Asia denoted someone who was in the retinue of the ruler46: e.g. Bilge47 chor QƒLɵ∏H [blkâjwr], a Turkic commander in the 860s48, among others.

“podchinennyi, podvlastnyi, vassal.” According to Zakeri, Sâsânid Soldiers, pp. 70-71, this term, recorded in the Behistun inscription to designate the “most trustworthy consul- tants and assistants” of Darius I, “cannot mean that they were all, literally, his slaves. The word here means a nobleman “bound” to the king in a relationship which, though subordinate, was freely accepted and probably sealed with an oath.” 40 Zakeri, Sâsânid Soldiers, p. 77. 41 Ibid, p. 71. 42 Ibid, p. 71. The passage cited from Ibn al-Balkhî is from The Fârs-nâma of Ibnu’l- Balkhî, ed. G. Le Strange and R.A. Nicholson (Bibb Memorial Series, N.S. 1, Cambridge, 1921), p. 43. 43 Zakeri, Sâsânid Soldiers, pp. 71-74. The girdle was another symbol of this relationship. 44 P.B. Golden, “The Terminology of Slavery and Servitude in Medieval Turkic” in D. DeWeese (ed.), Studies on Central Asian History in Honor of Yuri Bregel (Blooming- ton, 2001), p. 29. 45 Clauson, ED, pp. 427-28, remarks on its relatively rare appearance in texts and places it as “something smaller than a xagan and more than a beg.” 46 See N. Sims-Williams and J. Hamilton, Documents turco-sogdiens du IXe-Xe siècle de Touen-houang in Corpus inscriptionum iranicum. Part II, Inscriptions of the Seleucid- Parthian Periods and of Eastern Iran and Central Asia, vol. III. Sogdian (London, 1990),

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 289

Another Sâsânid term for these servitors of some relevance to our theme is reteg (Mod. Pers. rîdak “young boy, servant boy”49) which corresponds to Arab. ghulâm.50 Zakeri concludes that Sâsânid cavalry units that took service with the Arabs “were chiefly responsible for the introduction of private retainers into (the) early Muslim military estab- lishment.”51 These forces were termed in the early Arab sources ghulâm, mamlûk (“owned, slave”), shâkirs (< Pers. châkar, see above), ‘abd (“slave, bondsman, servant”) and fatâ (“youth”)52. The Arabicized shâkir was used in much the same sense as the bandagân, refer- ring to the “privately armed soldiers” or “military detachments” of the king or great lords. In this sense, according to Zakeri, the term may be translated as “follower.”53 What is the relationship of the châkar institution to the caliphal troops called Shâkiriyya? The latter were a force of elite cavalry toops, closely associated with the physical safety of the caliph, deriving, it appears, from Eastern Iranian (Khurâsânian) elements that had entered ‘Abbâsid service54. In the sources (cf. a†-™abarî), the Shâkiriyya are noted as a p. 82: cwr “titre attribué vraisemblement à quelqu’un faisant partie de la suite du sou- verain: antrustion, leude, page; nom de personne.” 47 Turk. bilge “wise,” see Clauson, ED, pp. 340-41. 48 al- ™abarî, ed. Ibrâhîm, IX, pp. 327, 330,Eng. trans. The History of al-™abarî, ed.Yar-Shater et al., vol. XXXV, The Crisis of the ‘Abbâsid Caliphate, trans. G. Saliba (Albany, 1985), pp. 87, 91. 49 Zakeri, Sâsânid Soldiers, pp. 79-80. 50 Zakeri also notes the ayyâr (from which he would also derive the Arabic ‘ayyâr) > Old Pers. *ady-âva-bara (“bringing help, succor, giving support”) > MP adyâwar > adyâr > ayyâr > yâr “assistant, helper, friend.” The term ayyâr originally designated the “companions or followers of the Great King or a high official on his travel…It also des- ignated a community of free warriors who, through a ceremonial oath, voluntarily took upon themselves to remain faithful to a lord and constitute his subordinates and follow- ers,” Zakeri, Sâsânid Soldiers, p. 87. 51 Ibid, pp. 182-183. 52 The use of terms denoting “boy, youth” were meant to show the non-adult status, at least officially, of these servitors. In Middle Turkic, the term oghlan (< oghul “off- spring, child, son”) acquired this status only under the influence of Islamic usage, see Golden, “The Terminology of Slavery and Servitude in Medieval Turkic” in DeWeese (ed.), Studies on Central Asian History in Honor of Yuri Bregel, p. 52. 53 Zakeri, Sâsânid Soldiers, pp. 79, 84-87, 89. 54 Kennedy terms them “élite but dependent troops” whose name was an assertion of their Iranian heritage, see the discussion in Kennedy, The Armies of the Caliphs, pp. 199- 203, who reviews the notices on this term and its contexts, since Umayyad times.

Journal Asiatique 292.1-2 (2004): 279-309 290 P.B. GOLDEN distinct grouping, separate from the Maghâriba and the Turks. They appear to have constituted a special element in the regular army along- side the jund55. This, in a number of respects, looks like the châkars and their role in Iranian Central Asia. However, the term shâkirî […ôcÉ°T] in Arabic came to denote a “hireling, mercenary,”56 a connotation that does not seem to be present in the Iranian Central Asian institution. Terms may be borrowed, but their specifics and contours in the host society may undergo change. Undoubtedly, this ancient Iranian tradition, part of the comitatus insti- tution that was spread across Eurasia57, deeply influenced, if it did not provide the direct model for, the ghulâm institution within the Muslim world. The situation is somewhat confused by the fact that the ‘Abbâsids had several such groupings, the Shâkiriyya and the Turkic ghulâms, each with different ethnic bases and — when necessary — each used to check the other. Gordon sees the “inspiration” for the Turkic ghulâm system as stemming from “Near Eastern traditions” (see above). There is no doubt that they were given local coloration, but the Shâkiriyya, among others, derived from Iranian traditions, a source for many practices of the ‘Abbâsid realm. The Iranian traditions, in turn, had strong Central Asian as well as Mesopotamian roots. These issues are, by no means, resolved to the satisfaction of all scholars studying this institution, especially with regard to the Central Asian influences on the Muslim ghilmân. The Châkar institution, i.e. the comitatus, as we have noted, was not unique to the Iranians. There are Turkic models for this as well. Indeed, the comitatus can be found thoughout Eurasia. Sharaf -Zamân ™âhir Marwazî reports that the Toquz Oghuz Qaghan had 1000 shâkirî (i.e. châkars) and some 400 female attendants58. The personal body-guard

55 ™abarî, ed. Ibrâhîm, IX, pp. 130, 155, 165, Eng. trans. The History of al- ™abarî, ed.Yar-Shater et al., vol. XXXIV, Incipient Decline, trans. J.L. Kraemer (Albany, 1989, henceforth: Kraemer, XXXIV, Incipient Decline), pp.19,63 (and n. 225), 79. 56 See A. de Biberstein, Kazimirski, Dictionnaire arabe-français (Paris, 1860), I, p. 1258, shâkirî < Pers. Châkar “Employé au service, salarie, à gages, mercennaire” 57 See P.B. Golden, “Some Notes on the Comitatus in Medieval Eurasia with Special Reference to the Khazars” Russian History/Histoire Russe, 28, Nos. 1-4 (Spring-Sum- mer-Fall-Winter 2001), pp. 153-170. 58 Sharâf al-Zamân Tahir Marvazi on , the Turks and , ed. trans. V.F. Minorsky (London, 1942), Arabic text, p. 18/Eng. trans., p. 29. The shâkirî were fed and given drink three times a day by their lord.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 291 corps of the Türk Qaghans was called “Fu-li” (= Mid. Chin. bu¢h li¢h = Turk. Böri “wolf”) according to the Zhou-shu59. György Györffy60 with following him, have concluded that the Old Türk name for this retinue was buyruq61. Minorsky attempted to connect châkar with Ottoman çaÈar “a palisaded enclosure, a camp of in such a fortified enclosure.”62, A similar meaning is noted for Chaghatay châqâr/chârqâr in the sense of “a form of external fortification of a fort.”63 Räsänen connected this with Kkirg. (Qazaq) shakar (‘desperate, mischievous (man), dare-devil” etc64.) and Yakut chaqar (“bol’shoe semeistvo”). All of this seems doubtful. Ottoman (çaker “servant”), Qïrghïz (cheker “prikazchik, sluga”), Uzbek (châkar “khizmatkâr, chârakâr, malay”65), Qumuq and Qarachay-Balqar (chaghar “krepostnoi krest’ianin”)66 have all assimilated the Persian term. Medieval Qïpchaq

59 See Pulleyblank, Lexicon, pp. 101, 189; Liu Mau-tsai, Die chinesischen Nach- richten zur Geschichte der Ost-Türken (T’u-küe) (Wiesbaden, 1958), I, p. 9. 60 Gy. Györffy, “Die Rolle des buyruq in der alttürkischen Gesellschaft,” Acta Orien- talia Hungarica, 11 (1960), pp. 169-79. 61 O. Pritsak, “The Distinctive Features of the Pax Nomadica” in Popoli delle Steppe: Unni, Avari, Ungari (Settimane di Studio del Centro Italiano di Studi sull’Alto Medioevo, Spoleto, 1988), pp. 768-9. According to Pritsak, the Soghdian equivalent of buyruq was gwrg’p’ynt. Clauson derives the Turkic term from Turk. buyur- “to order, command” and renders it as “commanded (to do something)’; in the early period a title of office, appar- ently a generic term for all persons commanded by the xagan to perform specific duties, civil or military,” see Clauson, ED, p. 387. 62 Minorsky/Marvazi, p. 94; see Sir James W. Redhouse, A Turkish and English Lex- icon (Constantinople, 1890, reprint: Beirut, 1974), p. 705. Minorsky ignores Osm. Çaker (Redhouse, p. 707, “ a servant,” a borrowing from Persian. Minorsky suggested that in Turkic (and Mongol, the tribal name Tsakhar) the meaning was extended to encompass the “persons grouped round a court…” This connection is uncertain. 63 Sheykh Süleyman Efendi, Lughat-i Chaghatay ve Türkî ‘Osmânî (Istanbul 1298/ 1889-81), p. 147 and L. Budagov, Sravnitel’nyi slovar’ turetsko-tatarskikh narechii (St. Petersbrug, 1869, 1871), I, p. 459. 64 M. Räsänen, Versuch eines etymologischen Wörterbuchs der Türksprachen (Helsinki, 1969), p. 96, who follows Minorsky connecting it with Mong. Tsakhar and deriving all from Pers. châkar; B.N. Shnitnikov, Kazakh-English Dictionary (London-The Hague- Paris, 1966) p. 231; see also H. Oraltay et al. (trans.), Kazak Türkçesi Sözlügü (Istanbul, 1984), p. 305: ≥aqar “kendi i≥ini bilen, çakal, gözü açık.” Qïrghïz chaqar “bedovyi, vidavshii vidy, khitryi, bol’shoi” (K.K. Iudakhin, Kirgizsko-russkii slovar’ (Moskva, 1965), p. 836 would seem to be unrelated. 65 Z.M. Ma ‘rufov, Özbak tilining izâhli lughati (Moskva, 1981), II, 377. 66 Z.Z. Bammatov, Kumyksko-russkii slovar’ (Moskva, 1969), 351; E.R. Tenishev et al., Karachaevo-balkarsko-russkii slovar’ (Moskva, 1989), 720.

Journal Asiatique 292.1-2 (2004): 279-309 292 P.B. GOLDEN shâkird or shâkirt is unrelated etymologically, but similar sounding (<. Pers. Shâgird “scholar, student, apprentice, disciple…boy, servant, groom”). In the Mamlûk-Qïpchaq glossaries it is defined as al-ghulâm67. Châker et al. meaning “servant, serf” etc., clearly come from Persian. *** In the Khazar Qaghanate, the Qaghan was guarded by a standing army of Khwârazmian Muslim immigrants, the *Orsiyya [á«°SQ’G] from whose ranks, according to al-Mas¨ûdî (d. 956-7), in his day, the wazîr was cho- sen. At the time of his writing (930’s-940’s), this man was AÌmad b. Kûyah68. The Orsiyya were separate from the rest of the army. They were free to practice their religion and were not required to participate in Khazar wars on Muslims. The Orsiyya had migrated to Khazaria “in olden times, after the appearance of Islam,” driven thither by draught and plague69. Ibn Rusta reports that the Khazar “king” ( the Îshâ = the -Beg, ranking beneath the sacral Qaghan, goes out with an army of 10,000 mounted warriors, some of whom are salaried troops (probably a reference to the Orsiyya)70. Ibn Fa∂lân, who was in Bulgharia in 921-922, remarks about this Muslim leader, probably the predecessor of AÌmad b. Kûyah, that “over the Muslims is a man from the ghulâms (ghilmân) of the King. He is called õN [khz].”71 Although we cannot pinpoint the date in which the Orsiyya army/comitatus was formally created, given its “antiquity” in Khazaria it seems likely that it was already in existence by the time men like Îtâkh came into the Islamic

67 Cf. Ed-Dürretü’l-Mudiyye fi’l-Lugati’t-Turkiyye, ed. trans. R. Toparlı (Erzerum, 1991), f. 14a/59/126. See Pers. Steingass, p. 724. 68 Omeljan Pritsak (see Golb and Pritsak, Khazarian Hebrew Documents, pp. 51,53- 55) sees in this wazirial dynastic name the origin of the toponym Kiev: áHÉjƒc [Kûyâba], bviiq (Kiyob], Kioába. 69 Al-Mas ’ûdî, Murûj adh-Dhahab ed. Pellat I, p. 213. The name, later Orus/Urus, may stem from the Old Iranian ˆAorsoi *Aurusha, see P.B. Golden, “Cumanica III: Urusoba” in D. Sinor (ed.), Aspects of Altaic Civilization III (Bloomington, 1990), pp. 33-46. 70 Ibn Rusta, ed. M.J. de Goeje, pp. 139-140. 71 Perhaps, Turk. *qïz “costly, expensive… rare,” see Clauson, ED, p. 680. It is also possible, perhaps even more likely, that this is an Iranian, i.e. Khwârazmian name, cf. Khwârazm. gh’z “to run, hasten” W.B. Henning, A Fragment of a Khwarezmian Dictio- nary, ed. D.N. MacKenzie (London, 1971), p. 38, cf. the name Kharkhîz (see Justi, Iranisches Namenbuch, pp. 170, 500).

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 293 world and certainly could have served as a model for how they concep- tualized the nature of their service to the ‘Abbâsids. Interestingly enough, al-Mas‘ûdî refers to this grouping of servitors of the Khazar Qaghans as ghulâms. *** The obedience and respect of the Turks for their elders and their “sense of service towards their great ones” were held in high regard by the ‘Abbâsids who “invited them into their service where they became their soldiers,…commanders and their retinue and their elite.”72 This passage raises a number of questions: were the Turks commanded by their own nobles or by Iranian nobles from Transoxania (the dihqâns)? There is only sparse evidence to support the latter supposition73. Did these nobles enter Caliphal service freely (for which there is some evi- dence) or did they, too, undergo enslavement, training and subsequent manumission? Whatever their social status (slave or free) when they entered Islamic society, the ghilmân, functionally, were very different from household slaves or slaves working in state enterprises (e.g. the Zanj in Southern who would soon erupt in a rebellion that convulsed the Caliphate). The function of these “military slaves” in ‘Abbâsid society was to serve as a “foreign legion” against both internal and external foes and to strengthen the central government74. Drawn from the periphery of the Islamic world, they experienced what Orlando Patterson has termed “social death” in an alien environment in which they were cut off from their family, clan and tribe75. Their only loyalty was to be directed, it was hoped, toward their master, the Caliph. The Caliph al-Mu¨taÒim (833-42) discouraged their establishing marital or

72 al-IÒ†akhrî, ed. M.J. De Goeje, pp. 292-292. Ibn Îawqal, Kitâb ∑ûrat al-Ar∂ (Beirut, 1992), pp. 387-388, has a slightly different phrasing of this passage: “the Turks became their soldiers (rijâl) because of their (military) virtues over the other soldiers and their nobles (dahâqînuhum) the commanders (umarâ’) among them … The nobles (dahâqîn) of Transoxania are their officers, their retinue and the elite..” 73 See discussion in Gordon, Breaking, pp. 32-33. 74 Kennedy, The Prophet and the Age of the Caliphates, p. 159-161 who stresses that that the “military caste became separate from the rest of society…by origin, language and custom…” This was part of a caliphal program to achieve greater centralization. 75 See Orlando Patterson, Slavery and Social Death (Cambridge, Mass., 1982).

Journal Asiatique 292.1-2 (2004): 279-309 294 P.B. GOLDEN other ties with the local population. Wives were provided from the slave women taken from abroad76. In these circumstances, the “socially dead” would seek to establish new links. These, ultimately, would be based on self-interest, but other factors could be involved as well. In a foreign environment, speakers of a common or closely related tongue might be expected to group together if for no other reason than communication. ‘Abbâsid military society was already organized along “ethnic” or regional lines, e.g. the Farâghina (military forces from Farghâna), the Maghâriba military forces from the Maghrib, i.e. “West” = North Africa). Although there were ghulâms drawn from other regions, the predominant ethno-linguistic element was Turkic. ‘Abbâsid sources of recruitment, by the early ninth century, were varied: the Sâmânid- patrolled frontiers of Transoxiana, including Khwârazm which did a brisk trade with the Turkic world, in particular with Volga Bulgharia, a vassal state of the Khazar Qaghanate and with the latter as well77. While the former may have been expected to provide the bulk of Turkic slaves (the Khazars having easier access to the Slavic and Finnic worlds and hence slaves from those peoples), there were nonetheless a number of ghulâms that surface in prominent positions in Caliphal politics that have associations with the Khazar realm. We cannot tell whether they were ethnic Khazars, came from subject peoples of the Khazars or sim- ply entered the Islamic world through Khazaria, all of which could have gained them the nisba “al-Khazarî.” Not all individuals thus designated necessarily came from Khazaria. In any event, Khazar captives may have come to the Islamic heartlands from Khazaria or the East. In addition to distinguishing themselves from the non-Turks in Caliphal service, the Turks also formed cliques among themselves. Unlike later Mamlûk history, for which we have many more sources, we cannot point to institutions such as the khushdashiyya which help to explain internal Mamlûk alliances. Shaban believes that men like Îtâkh

76 Shaban, Islamic History, 2, p. 66; al-Ya ’qûbî, Kitâb al-Buldân, ed. de Goeje, pp. 258-9. 77 See T.S. Noonan, “Volga Bulghâria’s Tenth Century Trade with Sâmânid Central Asia” Archivum Eurasiae Medii Aevi 11 (2000-2001), pp. 140-218. Ibn Îawqal (Beirut ed.), p. 398 says that most of the slaves from Khwârazm, which was involved in consid- erable trade with the Turks, are “∑aqâliba, Khazars and (from lands) beyond that, together with the slaves of the Turks…”

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 295 and Bugha were “leaders” of the Khazar forces that had been brought into the caliphal armies78. Did such ethnic cliques exist among the ‘Abbâsid ghulâms? If so, when did they come into existence? To see this more clearly, a brief look at the complicated politics of the ninth century Caliphate is necessary. The ‘Abbâsids were, as we have seen, looking for a stable military force that was not tied ethnically to the pop- ulation over which they ruled. Reducing their dependence on Arab and Persian elements (in particular the ™âhirids) allowed for greater central- ization. Outstanding Turkic ghulâms were soon promoted to important administrative positions. Thus, Ashinâs became governor of Egypt (219- 230/834-844) and Îtâkh served as governor of Yemen (225-230/839-40- 844) and then succeeded Ashinâs. However, there were constant ten- sions between different factions, especially as the ghulâms increasingly began to take on the attributes of a praetorian guard and to interfere in state affairs to promote their own personal interests. Moreover, the pop- ulation of Baghdad soon turned against the overbearing and occasionally riotous Turkic ghulâms, attacking them on the street. As a consequence, in 220/835, al-Mu‘taÒim founded the city of Sâmarrâ, lying some 97 km to the north of Baghdad, as a secure base for his Turkic troops, away from the tension-filled capital79. Even here, the Caliph sought to segre- gate the Turks from the rest of the population. According to al-Ya’qûbî: “they separated the sections of the Turks from the people completely. They placed them in areas separate from the others and they were not to mix with the muwalladûn80 nor were they to neighbor with others except the Farghânians. He allotted to Ashinâs and his companions the place known as al-Karkh. And he joined to him a number of Turkic commanders and men (rijâl [soldiers]). He ordered mosques to be built and markets. Khâqân ‘Ur†ûj and his companions were allotted the lands lying beyond the Jawsaq

78 Shaban, Islamic History, 2, p. 66. 79 a†-™abarî, ed. Ibrâhîm, IX, pp. 17-18, ™abarî/Bosworth XXXIII, Storm, pp. 25-28; Lassner, Shaping, p. 114-115 and his “Sâmarrâ” Dictionary of the , ed. J. Strayer et al. (New York, 1982-1989), 10, pp. 641-644; Gordon, Breaking, p. 50. 80 Defined by R. Dozy (Supplément aux dictionnaires arabes (Leiden, 1881, reprint: Beirut, 1968), II, p. 849) “celui qui, sans être d’origine arab, est né parmi les Arabes et a reçu une éducation arabe.” It also had the additional meanings of “adopté, né de parents esclaves, Métis, mulâtre.” Gordon, Breaking, p. renders it “local population” which is, perhaps, the best sense of the term here, i. e. the Arabo-Aramaic population of this old Sâsânid territory.

Journal Asiatique 292.1-2 (2004): 279-309 296 P.B. GOLDEN

al-Khâqânî. He ordered his companions to be gathered together and barred them from mixing with the people. He alloted to WaÒîf and his companions the area beyond al-Îayr… And it turned out that the allotments of the Turks and the Iranian Farghâ- nians81, completely, were at a distance from the markets and the throngs in the broad streets…. There is no one of the populace, merchants and others, who may mix with them in their streets. Subsequently, he (Mu‘taÒim, pbg) bought them female slaves and married them off to them. He forbade them to marry or become related by marriage with any of the muwalladûn until they have offspring. And they married one another. He provided a steady sustenance for the female slaves of the Turks. He inscribed their names in the dîwân. None of them can divorce or separate from his wife…When he gave Ashinâs the Turk his allotment on the western edge of the settlement, he provided allotments to his companions who were with him. It was called al-Karkh. He ordered him to not give free rein to any stranger from among the merchants or any other of those neighboring them nor to permit social intimacy (mu‘âshara) with the muwalladûn… The fourth street is called (the street of) Barghâmush the Turk. In it are the allotments of the Turks and Farghânians. The roadways of the Turks are separate and the roadways of the Farghânians are separate. The Turks are in the roadways which face the qibla, the Farghânians, facing them, are in the roadways which are north of the qibla. Each roadway faces the other. No one mixes with them. The end of the dwellings of the Turks and their allotments (are at) the allotments of the Khazars which lie to the east…”82 Our source appears to distinguish between three distinct groupings: the Turks, the Farghânians and the Khazars, each with their own dis- tricts. The Khazars were Turks and were noted as such by Muslim, Chi- nese, Byzantine, and Georgian sources.83 Indeed, individuals, such as Îtâkh, were labelled at-Turkî “the Turk” in some accounts and al-Khaz- arî “the Khazar” in others. While Turkî could refer to a Turkic-speaker coming from Eastern Khurâsân, Transoxiana or the Khazar lands, Khaz- arî would have only referred to one coming from or perhaps via the Khazar Empire. Nonetheless, for linguistic84, geographical or perhaps

81 I.e. al-farâghina al-‘ajam. Arabic ‘ajam “denotes “barbarous” and “Persian.” It can also mean “non-Muslim” (Kennedy, The Armies of the Caliphs, p. 125). Al-Ya ’qûbî is probably engaging in word-play here. 82 al-Ya ’qûbî, Buldân, ed. de Goeje, pp. 258-259, 262. 83 See P.B. Golden, Khazar Studies (Budapest, 1980), I, pp. 57-58. 84 The Khazars or substantial elements among them may have spoken Oghuric (Bul- gharic) rather than standard Turkic, see L. Ligeti, A magyar nyelv török kapcsolatai a

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 297 other reasons, the Khazars were distinguished from the other Turks. The Farghânians, in all likelihood, were composed, at least in part, of some Turkic elements as well. Their elite, however, men like Afshîn, descended from the hereditary ruling class of Iranian Central Asia. The “Turks” were most probably from Central Asian groupings (Oghuz, Qarluqs and Toquz Oghuz) — although the term, as we have seen, sometimes, also included Khazars85. Thus, the distinctions among these forces were largely geographic and to a lesser extent, linguistic (i.e. they spoke different forms of Turkic, and some spoke various forms of Iran- ian, i. e. Soghdian, Khwârazmian — there were, undoubtedly, bilingual elements as well) and ethnic. These living patterns would only serve to foster closer ties among those springing from a common place of origin. Presumably, prominent officers of Khazar origin, domiciled among their fellow-countrymen, tended to have Khazars as their most trusted subor- dinates. However, too close an association with one’s former fellow- countrymen outside of the Islamic world was, understandably, viewed with suspicion (e.g. the case of Bugha the Elder who, despite his victory over the Khazars in 851, was withdrawn from the Caucasian theater of operations after he had been reported to have “taken council with the Khazars, his tribesmen/kinsmen”86). Matthew Gordon has also under- scored the distinction between those Turks acquired earlier and recruited in Baghdad and those that came directly from the steppe. The former, already acculturated, at least to some extent, could assist in the training of the latter.87 Again, common life experiences may have bound those men into groups or factions. **** Turkic ghilmân are first noted in 202/817 accompanying Abû IsÌâq (the future al-Mu’taÒim) in a campaign against the Khârijites. It was

honfoglalás elött és az Árpád-korban (Budapest, 1986), pp. 475-487. Golden, Khazars Studies, I, pp. 123ff., concludes that the sparse Khazar vocabulary known to us does not yet provide a convincing case for either. 85 Kennedy, The Armies of the Caliphs, p. 120. 86 Golden, Khazar Studies, I, p. 78; K’art’lis Ts’xovreba, ed. Qaukhch’ishvili, I, pp. 256-257 and note 10 above. 87 Gordon, Breaking, pp. 17, 23.

Journal Asiatique 292.1-2 (2004): 279-309 298 P.B. GOLDEN there that Ashinâs earned his distinctive Persian sobriquet88. The ghulâm commanders came to prominence in the , during the wars against the Khurramiyya sectarian movement led by Bâbak that convulsed Azarbayjan ca. 210/816-817-222/837, occasionally spilling over into Byzantine lands as well89. Bâbak was largely crushed by the Central Asian general al-Afshîn Khaidhar (Îaidar in other sources) b. Kâwûs of Ustrushana/Ushrûsana. Îtâkh, Afshîn and Ashinâs also figured promi- nently in the famous campaign of 233/838 in which Amorion, the ances- tral town of the Byzantine ruling house, was taken90. At one crucial point, Îtâkh led a force of Turks and Maghâriba that enlarged the breach in the Byzantine fortifications91. Increasingly, however, there developed rivalries between the Central Asian Iranians and Abnâ’ (older partisans of the ‘Abbâsids from Khurâsân who had spearheaded the ‘Abbâsid revolution), on the one hand, and the Turks on the other. The unsuccessful plot by the former to overthrow al-Mu’taÒim in 223/838, in support of al-‘Abbâs, son of al- Ma’mûn, led to a massive purge of commanders prominent in the Amor- ion campaign92. Bosworth suggests that this “mass slaughter” weakened

88 a†-™abarî, ed. Ibrâhîm, VIII, p. 558, Eng. trans. The History of al-™abarî, ed.Yar- Shater et al., vol. XXXII, The Reunification of the ‘Abbâsid Caliphate, trans. C.E. Bos- worth (Albany, 1987, henceforth: Bosworth, XXXII, Reunification), p. 68. Ashinâs is Persian, deriving from the words he used when placing himself in front of Abu IsÌâq’s attackers: ashinâs mâ-râ “recognize me! (Pers. shinâkhtan “to know, to be acquainted with, to acknowledge, to discern, ” see Steingass, p. 762). 89 See Kennedy, Early Abbasid, pp. 170-174 (on the early phase of the war); R. Mot- tahedeh, “The Abbâsid Caliphate in Iran” in The Cambridge History of Iran, vol. 4, The Period from the Arab Invasion to the Saljuqs, ed. R.N. Frye (Cambridge, 1975, hencforth CHIr), p. 75 and B.S. Amoretti, “Sects and Heresies” in CHIr, VI, pp. 505-509; Kennedy, The Prophet and the Age of the Caliphates, pp. 166-167. For the impact on Byzantium, see W. Treadgold, The Byzantine Revival 780-842 (Stanford, 1988), pp. 232- 233, 272-273, 282-286, 290-295. Îtâkh, for example, is mentioned somewhat derisively as al-Mu‘taÒim’s “cook” (a position he held at one time) in a letter that Bâbak, in 223/837, now desperate, sent to the Byzantine Emperor seeking his aid, see a†-™abarî, ed. Ibrâhîm, IX, p. 57/ ™abarî/Bosworth, XXXIII, Storm, p. 94. 90 See a†-™abarî, ed. Ibrâhîm, IX, pp. 57ff/ ™abarî/Bosworth, XXXIII, Storm, pp. 97ff; Treadgold, The Byzantine Revival, pp. 297ff. and the extensive discussion in Vasiliev, Byzance et les arabes, I, pp. 144-177. 91 a†-™abarî, ed. Ibrâhîm, IX, pp. 66-67/ ™abarî/Bosworth, XXXIII, Storm, pp. 113-114; Gordon, Breaking, p. 49. 92 a†-™abarî, ed. Ibrâhîm, IX, pp. 71-79/ ™abarî/Bosworth, XXXIII, Storm, pp. 121- 134

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 299 the “Khurâsânian and Transoxanian element among the leading com- manders” and resulted in a “corresponding rise in the influence of the Turkish slave element,” although a†-™abarî says, quite pointedly, that “the remainder of the commanders, comprising Turks, men of Farghâ- nah, and others…were killed in their entirety.”93 Clearly, there were divisions and factions among the Turks. This was also part of the background to the fall, in 840, of Afshîn, the victor over Bâbak in Azarbayjan and a rival of Îtâkh and Ashinâs. Hav- ing failed to act quickly against a rebel (who was his kinsman), the Caliph now grew suspicious of Afshîn who reciprocated these senti- ments. Afshîn considered flight to Ushrûsana via Khazaria and Armenia. Thwarted in this, he plotted against al-Mu‘taÒim. Îtâkh played a key role in protecting the caliph. Afshîn was toppled from his position of promi- nence, imprisoned and starved to death in 226/840-41. Shortly there- after, in 227/841, Îtâkh successfully led troops against a Kurdish rebel (Ja‘far b. Mihrj?sh)94. Îtâkh (the onetime cook) and Ashinâs now became the key figures during the Caliphate of the weak al-Wâthiq (227-232/842-847)95. It is interesting to note here that it was during the latter’s reign that two caliphal embassies were dispatched to Khazaria and thence to the north- ern lands. One was led by MuÌammad b. Mûsâ al-Khwârazmî, the noted astronomer, mathematician and geographer. His nisba indicates family origins in Khwârazm, although he was born near Baghdad96. The other, which ventured to the land of “” was headed by Sallâm the Interpreter, who knew Turkic97. Did these missions to Khazaria

93 Turks had been recruited to carry out the assassinations of al-Mu‘taÒim and some of the leading military figures (such as Afshîn and Ashinâs) associated with him. The assas- sins were drawn from the entourages of the intended victims, see a†-™abarî, ed. Ibrâhîm, p. 79/ ™abarî/Bosworth, XXXIII, Storm,, pp. 133-134 n.371. 94 a†-™abarî, ed. Ibrâhîm, IX, p. 118// ™abarî/Bosworth, XXXIII, Storm, p. 206. 95 a†-™abarî, ed. Ibrâhîm, IX, p. 102, 104-110/ ™abarî/Bosworth, XXXIII, Storm, pp. 175-176, 180-193; aÒ-∑ûlî, pp. 23/588 (who says he ran the affairs of state for al- Wâthiq) and a brief account in al-Ya‘qûbî, Ta’rîkh al-Ya‘qûbî (Beirut, 1390/1970), II, p. 477-478. For an overview of the events of the latter half of the ninth century, see Kennedy, The Prophet and the Age of the Caliphate, pp. 159ff. 96 T. Lewicki, Zród¥a arabskie do dziejów S¥owianszczyzny (Wroc¥aw-Kraków- Warszawa, 1956-1977), I, pp. 13-14. 97 See al-Muqaddasî, AÌsan at-Taqâsim fî Ma‘rîfat al-Aqâlîm, ed. M.J. De Goeje (Lei- den, 1877, 2nd ed. 1906), p. 362; Ibn Khurdâdhbih, Kitâb Masâlik wa’l-Mamâlik, ed.

Journal Asiatique 292.1-2 (2004): 279-309 300 P.B. GOLDEN reflect the influence of leading Khazar figures, such as Îtâkh, at the caliphal court? As with al-Mu’taÒim, Îtâkh remained in charge of the security police in Sâmarrâ under the new Caliph. A†-™abarî comments that “whoever al-Mu’taÒim or al-Wâthiq wished to have killed would be imprisoned and killed by Îtâkh.”98 Al-Ya’qûbî reports that al-Wâthiq also made him governor of Khurâsân, Sind and the districts around the Tigris. He had also served as governor of Yemen and later became the chamberlain (Ìâjib) of al-Wâthiq’s successor, al-Mutawakkil (232-247 /847-861). For the latter he was, according to a†-™abarî, “responsible for the regular army, the Maghâribah, the Turks, the mawlâs, the Postal and Intelli- gence service (barîd), the office of the chamberlain, and the Caliphal Palace.”99 In either 233/847-848 or 234/849, he fell from grace after a drunken quarrel started by al-Mutawakkil which so incensed Îtâkh that he contemplated murdering the Caliph. Although relations were patched up on the surface, al-Mutawakkil later had him confined and ultimately starved to death (or deprived of water, died of thirst) in 235/849100. This was a patrimonial state. Despite his many offices and substantial accom- plishments101, Îtâkh’s power derived not from an institutional base but

M.J. De Goeje (Leiden, 1889), pp. 162ff; Ibn Rusta, ed. De Goeje, p. 149. See also dis- cussion in D.M. Dunlop, History of the Jewish Khazars (Princeton, 1954), pp. 190-192. 98 a†-™abarî, ed. Ibrâhîm, IX, pp. 166-167, Eng. trans. The History of al-™abarî, ed.Yar-Shater et al., vol. XXXIV, Incipient Decline, trans. J.L. Kraemer (Albany, 1989, henceforth: Kraemer, XXXIV, Incipient Decline), p. 81 99 a†-™abarî, ed. Ibrâhîm, IX, p. 167/ ™abarî/Kraemer, XXXIV, Incipient Decline, pp. 81-82. 100 See the general outline of his career and death, see Gordon, Breaking, pp. 9, 49-50, 76-84, 109-110 and the remarks of Kraemer in The History of al-™abarî, Kraemer, XXXIV, Incipient Decline, p. 9 n. 17 and pp. 81-86 (= a†-™abarî, ed. Ibrâhîm, IX, pp. 166-170); al-Ya‘qûbî, Ta’rîkh, II, p. 479. The latter also comments (p. 481) that Îtâkh gained the ranks and estates of Ashinâs when the latter died in 230/844-45; a†-™abarî, ed. Ibrâhîm, IX, p. 103/ ™abarî/Bosworth, XXXIII, Storm, p. 179. P.M. Cobb, White Ban- ners. Contention in ‘Abbasid , 750-880 (Albany, 2001), p. 161 n. 84, suggests that Îtâkh may also have held the governorship of Syria. He is not noted, however, among the amîrs of Damascus, see ∑alâÌ ad-Dîn aÒ-∑afadî, Umarâ ‘ Dimashq fî ‘l-Islâm, ed. S. al- Munajjad (Damascus, 1374/1955). On Îtâkh’s demise, see a†-™abarî, ed. Ibrâhîm, IX, pp. 168ff, ™abarî/Kraemer, XXXIV, Incipient Decline, pp. 83ff.; Shaban, Islamic His- tory, 2, pp. 2,72,73; Ya‘qûbî, Ta ‘rîkh, p. 486. 101 Although Îtâkh had substantial military achievements, his main strengths appear to have been in the area of administrative offices, see Gordon, Breaking, p. 113.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 301 from his close personal relationships with the Caliphs and the military might of his followers among the ghilmân — an uncertain, fluid base. As Mutawakkil plotted his Itâkh’s death, he duplicitously apologized to the latter for the drunken quarrel, saying “you are my father and you have reared me.”102 Careers were made and broken on the basis of these personal ties. This was true not only with regard to the relations between the Caliphs and their nominal servants, but among those Turks who had acquired power as well. The Caliphs could bring about the fall of one or another of these servants, but increasingly this would have to be part of an intricate political process in which the Caliphs, themselves decreasing in power, would have to play off the competing factions among the ghilmân. These “networks,” although led by Turks, comprised non- Turks (Arabs and others), as well, particularly in bureaucratic posts. The complexities of these relationship (involving marital ties as well) have begun to be investigated103. In addition to Îtâkh and Ashinâs104, new, powerful Turkic comanders were appearing: the Khazar Bugha the Elder (Arab. al-Kabîr), Bugha the Lesser (aÒ-∑aghîr) and WaÒîf. Al-Mutawakkil (232-247/847-861), was chosen by a grouping of powerful figures that included Îtâkh and WaÒîf105. The new caliph, however, after initially enmeshing the Turks even more deeply into caliphal politics, wanted to be free of his “guardians” and in 235/849, as we have seen. Îtâkh’s doom was encom- passed (Ashinas having already died in 230/844-45) and WaÒîf became chamberlain106. These policies had led to conflicts within the Turkic ghilmân as well107. There was now an attempt to remove the Turks (or at least certain fac- tions of them, it should be remembered that one of the closest advisers of the Caliph was the Turk FatÌ b. Khâqân) from their positions of

102 a†-™abarî, ed. Ibrâhîm, IX, p. 167/Kramer, XXXIV, Incipient Decline, pp. 82. 103 On the “networks” of the various prominent commanders, see Gordon, Breaking, pp. 111ff. 104 For an outline of his career, see Kramer, XXXIV, Incipient Decline, p. 5 n. 6; Gor- , Breaking, pp. 17-18, 50, 76-79. 105 a†-™abarî, ed. Ibrâhîm, IX, p. 154, ™abarî/Kraemer, XXXIV, Incipient Decline, pp. 61-62; Gordon, Breaking, p. 80. 106 Ya ’qûbî, Ta’rîkh, p. 479 (his service to al-Wâthiq), p. 481 (the death of Ashinâs and Îtâkh assumes many of his posts). Gordon, Breaking, p. 83. 107 Gordon, Breaking, p. 82.

Journal Asiatique 292.1-2 (2004): 279-309 302 P.B. GOLDEN power. This led to a counter-conspiracy to depose the Caliph. The crisis came to a head when the latter confiscated the property of WaÒîf, the very wealthy and last major Turkic leader108. The Turkic elite retaliated, probably with the connivance of al-Mutawakkil’s son al-MuntaÒir (861- 862) by murdering al-Mutawakkil and FatÌ b. Khâqân (to whom WaÒîf’s property had been given) on 4 Shawwâl 247/-10, 861. According to a†-™abarî, the assassins were Baghïr [ôZÉH], Mûsâ, the son of Bugha the Elder (al-Kabîr), Hârûn b. Suwârtegin and Bugha al- Sharâbî. Al-Ya’qûbî has a slightly different cast: Bugha the Younger, Ötemish [¢ûeÉJhG], Baghïr, Baghlu, Yrd [Oôj?] Wâjin, Sghlqh [¬Ø∏©°S for ¬≤∏¨°S?] and Kündash/Kündesh [¢TGóæc].109 A number of these men, it appears, were of Khazar origin: Mûsâ b. Bugha al-Kabîr, Ötemish b. Kh†rkîn (ø«cô§N *Qut erkin)110 and most probably Hârûn b. Suwârtegin [ø«µJôGƒ°S], who was later closely associated with the Khazar IsÌâq b. Kundâjiq111. Was this a Khazar clique? The murder of al-Mutawakkil ushered in an age of internecine strife lasting until 256/870 in which weak caliphs, the now factionalizing Turks and other foreign military groupings competed for power. Ini- tially, the two Bughas and Ötemish were running the affairs of the Caliph al-Musta’în (248-252/862-866). Bugha al-Kabîr died, however, in Jumâdâ II 248/August 2-30, 862 (aged 90, according to some accounts). His son, Mûsâ, succeeded to many of his father’s duties. Gor- don views this as the date of the “emergence of the second generation of Samarran Turks.”112 This was a different generation, literate, acculturated

108 See aÒ-∑ûlî, pp. 119-120/485. 109 a†-™abarî, ed. Ibrâhîm, IX, pp. 222ff., ™abarî/Kraemer, XXXIV, pp. 171ff.,; al- Ya‘qûbî, Ta ‘rîkh, II, p. 492; Kennedy, The Prophet and the Age of the Caliphate, p. 170; Shaban, Islamic History, 2, p. 97. 110 According to al-Mas ’ûdî, Tanbîh, ed. de Goeje, p. 363, he was “the son of Bugha (the Elder’s) sister.” He was of the second generation of Turks, a man born in Iraq and one who was considered a man of culture, see Gordon, Breaking, p. 94. On this name, see Golden, “Khazar Turkic Ghulâms in Caliphal Service: Onomastic Notes” AEMAe 12 (2002-2003), pp. 23-24: qut “heavenly good fortune” and erkin/irkin a title. 111 Hârûn would appear to be the brother of WaÒîf b. ∑uwârtegin al-Khazarî noted by al-Mas ’ûdî, Tanbîh, pp. 275, 376. A†- ™abarî, ed. Ibrâhîm, X, pp. 124-125 knows him as WaÒîf b. ∑uwârtakîn at-Turkî. ∑uwâr or ∑awâr refers to the Sawar < Sabir tribal group- ing associated with the Khazars and the Volga Bulghars, see Golden, Khazar Studies, I, pp. 256-257; Ludwig, Struktur, pp. 359-360.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 303

(at least at the leadership level), but still well aware that its power derived from access to or control over the Caliph and the troop factions they could muster. In the kaleidoscope of caliphal machinations and ghulâm internal pol- itics that quickly manifested itself, Mûsâ b. Bugha, Ötemish and AÌmad b. ™ûlûn constituted a powerful faction, based, perhaps, on kinship ties (Mûsâ and Ötemish) and political interest. Ötemish’s star was only briefly ascendant, having become the wazîr of al-Musta’în (248-252/ 862-866), he had grown overly greedy and was killed by the Turks on 14 Rabî¨ II 249/June 6, 863, an outcome engineered by WaÒîf and Bugha aÒ-∑aghir113, a faction/coalition that did not have a strong Khazar com- ponent114. Then, for a time, WaÒîf and Bugha aÒ-∑aghir, dominated the scene. Equally meteoric was the career of Baghïr, one of the assassins of al-Mutwakkil whose popularity with the Turkic troops and growing closeness to the Caliph brought the suspicious interest of WaÒîf and Bugha aÒ-∑aghir. He was killed in 251/865, an act that precipitated a period of internecine strife for much of that year. The Caliph al-Musta’în did not long survive these events and was deposed the next year. WaÒîf was, in turn, murdered by rivals within the Turkic camp in 253/867 (the soldiers were unhappy over delays in their pay) and his ally Bugha aÒ-∑aghir, although briefly able to hold great power, was executed in 254/868 by the Caliph al-Mu’tazz (252-255/866-869) who was allied with the Turkic commander Bâyakbâk115. Their faction now came under the leadership of WaÒîf’s son, ∑âliÌ and Bâyakbâk who appointed AÌmad b. ™ûlûn (founder of the ™ûlûnid dynasty) to the governorship of Egypt. ∑âliÌ b. WaÒîf’s failures in the fiscal-tax area led to his replacement by Mûsâ b. Bugha the Elder. The Caliph al-Muhtadî (255-256/869-870, brought to power by ∑âliÌ b.

112 Gordon, Breaking, p. 94. 113 a†-™abarî, ed. Ibrâhîm, IX, p. 263, Eng. trans. The History of al-™abarî, ed. Yar- Shater, XXXV, trans. J.L. Kraemer, The Crisis of the ’Abbâsid Caliphate (Albany, 1985, henceforth cited as ™abarî/Kraemer, XXXV, Crisis) pp. 11-12; Gordon, Breaking, pp. 94- 95. 114 According to aÒ-∑ûlî, pp. 153-154/452-451, Bugha al-Kabîr’s sons Mûsâ and MuÌammad played a key role in Ötemish’s death — clearly a break had occurred in this “family” alliance. 115 The events are discussed in Gordon, Breaking, pp. 95-98.

Journal Asiatique 292.1-2 (2004): 279-309 304 P.B. GOLDEN

WaÒîf (who himself perished not long after), MuÌammad (another son of Bugha the Elder) and Bâyakbâk116, wanted to reassert Caliphal authority. In the course of his brief and tumultuous reign, al-Muhtadî tried, unsuc- cessfully, to turn Bâyakbâk against Mûsâ b. Bugha. The killing of Bâyakbâk and MuÌammad b. Bugha at caliphal behest led to another round of bloodshed that ended with the al-Muhtadî’s gruesome demise. However acculturated MuÌammad b. Bugha may have been, the steppe traditions were still observed at his funeral. “The Turks,” a†-™abarî reports, “broke a thousand swords over MuÌammad b. Bugha’s grave in accordance with their custom when a chief dies.”117 Al-Muhtadî’s successor was al-Mu’tamid (256-279/870-893), a son of al-Mutawakkil. His administration proved to be more adept at dealing with the Turks. The role of the latter, as in the other depositions, was cru- cial. Al-Mu’tamid was aided and to some degree overshadowed, espe- cially after 269/882, by his brother al-Muwaffaq (d.278/891) who had a good relationship with Mûsâ b. Bugha, now the major power figure of his generation. After the latter’s death in 264/877, his role was taken over by Kayghalagh [≠∏¨«c, in aÒ-∑ûlî: ≠∏≤«c Kayqalagh] a Khazar118, and IsÌâq b. Kundâjîq (d.278/891). The latter was a well-established member of the Khazar grouping within the Turks surrounding the ‘Abbâsid caliphs. Ibn Kundâjîq’s line was already one long situated in caliphal service, dating, perhaps, to the early eighth century (see below). He would also serve as a counterpoise to the increasingly powerful governor of Egypt, AÌmad b. ™ûlûn (d. 270/884). Under al-Mu’tamid and al-Muwaffaq and then the lat- ter’s son, al-Mu’ta∂id (279-289/892-902), a period of relative equilibrium between the Turkic troops and the government had been established. The seat of government was also returned to Baghdad. The increasingly inde- pendent ™ûlûnids were conquered in 292/905119 Mention is still made, at this time, of Khazar forces in caliphal service. There were Khazars

116 Gordon, Breaking, pp. 100-103. 117 a†-™abarî, ed. Ibrâhîm, IX, pp. 456-469, Eng. trans. The History of al-™abarî, ed. Yar-Shater, XXXVI, trans. D.Waines, The Revolt of the Zanj (Albany, 1992, henceforth cited as ™abarî/Waines, XXXVI, Revolt), pp. 91-107; gives a number of different accounts. 118 According to aÒ-∑ûlî, pp. 240/365, he was a brother of Mûsâ and MuÌammad b. Bugha al-Kabîr. 119 Kennedy, The Prophet and the Age of the Caliphate, pp. 174-186; Gordon, Break- ing, 144-145.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 305 present among the caliphal “chamber pages.” A†-™abarî, in a notice recorded s.a. 293/906, reports their presence along with other “pages” in a battle against the Qarma†ian rebels near Qâdisiyya. They “remained steadfast and fought to the death, inflicting great losses upon the Qarma†ians.”120 A certain Sîmâ al-Khazarî is noted among the ‘Abbâsid military men of this same period (early tenth century)121, but these are no longer commanding figures in the realm. This was a brief respite before the disastrous caliphate of the hapless al-Muqtadir (295-6, 296-317, 317-320/908,908-929, 929-932) who was driven from office several times during a time of domestic revolt and declining revenues. The old Turkic elite and their descendants no longer figured as prominently in affairs. Newer groups of Turks, entering the caliphal heartland in the 930’s via the Dailamites of Iran, would now play a central role122. The era of Khazar ghilmân as a discernible force in caliphal service was over. This coincided with a series of Rus’ raids down the Volga through Khazaria into the Muslim communities on the Caspian (beginning in 910, after an earlier raid sometime between 864- 884)123. Baghdad and Atïl were now fading imperial seats.

Sources of Recruitment of Khazar Ghulâms

The early Khazar-Arab military contacts began in the transition period in which the Khazar Qaghanate emerged out of the Western Türk state (ca. 643-653). In engagements in Azarbayjan in 99/717-718, some fifty Khazar captives were taken by Îâtim b. al-Nu’mân al-Bâhîlî and brought to the Caliph ‘Umar II (717-720) in ÎunâÒira. Among the prisoners may have been the ancestor of IsÌâq Ibn Kundâjîq124. The

120 a†-™abarî, ed. Ibrâhîm, X, p. 126, Eng. trans. The History of al-™abarî, vol. XXXVIII, The Return of the Caliphate to Baghdad, trans. F. Rosenthal (Albany, 1985), pp. 164-165. These forces were perhaps associated with the commander Ra’iq al-Khazarî (see a†- ™abarî, ed. Ibrâhîm, X, p. 125/trans. Rosenthal., pp. 163-164. 121 Ibn Miskawaih, Tajarub al-Umam. The Eclipse of the ‘Abbâsid Caliphate, ed. H.F. Amedroz, trans. D.S. Margoliouth (Oxford, 1920-1921), I, pp. 19, 46-47; Ludwig, Struktur, p. 358. 122 See discussion in Kennedy, The Prophet, pp. 187-199. 123 Briefly discussed in Golden, Khazar Studies, I, pp. 80-81. 124 See Yıldız, Islâmiyet ve Türkler, pp. 25-26. Presumably this ancestor was one of

Journal Asiatique 292.1-2 (2004): 279-309 306 P.B. GOLDEN

Arab-Khazar wars, punctuated by brief truces, continued until 737125. In that year the Umayyad general and future caliph, Marwân scored an impressive victory over the Khazar Qaghan, forcing him and other mem- bers of the royal house and some elements of the Khazar population to convert to Islam126. Some “40,000 captive Infidels” were settled by Marwân in the Lezgin lands (ar∂- al-Lakz) around the Samur river127. The ethnicity of these captives and their potential as a subsequent source of Khazar ghulâms is unclear. If they had converted to Islam, they could not, in theory, be enslaved. In 749, that same Marwân, now Caliph (745- 650), is said to have dispatched his “∑aqlâb” (Slav128) chamberlain (Ìâjib) with a squad of twenty of his mawâlî consisting of “” (∑aqâliba), Khazars and Rûm (Byzantines) who carried out an assassi- nation for him129. Perhaps these Khazars and Slavs were acquired during his 737 campaign. On the whole, the documentation on Khazar slaves is very spotty130. Ludwig has reviewed the meager data, noting the presence of Khazar the fifty Khazar prisoners brought to ’Umar as reported in a†-™abarî, ed. Ibrâhîm, VI, pp. 553-554, Eng. trans. The History of al- ™abarî, ed. Yar-Shater, /XXIV, The Empire in Transition, trans. D.S. Powers (Albany, 1989, henceforth cited as ™abarî/Powers, XXIV), pp. 74-75. The account is repeated in Ibn al-Athîr, Al-Kâmil fî’t-Ta’rîkh, ed. C.J. Torn- berg (Leiden, 1851-76, reprint: Beirut, 1965-67), V, p. 43. Dunlop, History pp. 60-61, notes the event but does not make this connection in Ibn Kundâjiq’s ancestry. As noted by Dunlop, the poet al-BuÌturî says that IsÌâq b. Kundâj’s fame in Iraq was added on to the glory he had already achieved in al-Bay∂a (an Arab designation for “the White/Yellow Fort,” an important Khazar fort) and Balanjar (one of the old Khazar cen- ters in the North Caucasus). This underscored his Khazar origins. Sarkel, however, was built in 838 (Golden, Khazar Studies, I, pp. 239-243), well after ‘Umar II’s era. Other Khazar forts may have been known by this designation. 125 On the wars, see Dunlop, History, pp. 41-87. 126 a†-™abarî is curiously reticent about this. The fullest account is found in Ibn A‘tham al-Kûfî, Kitâb al-FutûÌ, ed. S. Zukkâr (Beirut, 1312.1992), III, pp. 252-255. 127 Ibn A‘tham al-Kûfî, ed. Zukkâr, III, p. 255; al-Balâdhurî, FutûÌ, ed. Ra∂wân, p. 210. 128 Pl. ∑aqâliba a term usually denoting “Slav” (< Greek Sklábov) but also occa- sionally for the various populations of Northern , see P.B. Golden, “∑aÈâliba” Encyclopaedia of Islam, 2nd ed. (Leiden, 1960-), VIII, fasc. 143-144, pp. 872- 878; D.E. Mishin, Sakaliba. Slaviane v islamskom mire (Moskva, 2002); M. Meouak, ∑aqâliba, eunuques et esclaves à la conquête du pouvoir (Helsinki, 2004). 129 Al-Balâdhurî, Ansâb al-Ashrâf, ed. Dûrî, III, p. 121 noted by Ludwig, Struktur, p. 204. 130 Gordon, Breaking, pp. 21-22.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 307 troops in the Caliphal army besieging Amida in 767 and the participa- tion of Khazars in a slave revolt in Îarrân that included Indians and oth- ers131. He considers Darband and Khwârazm as the principal markets for Khazar slaves. Moreover, although considerable numbers of slaves entered the Islamic world from Khazaria, as we know from the Islamic geographers132, our sources do not tell us how many of them were actual Khazars, how many were from subject peoples and — the most likely source- how many were acquired through Khazar military activity against their neighbors. Indeed, the Îudûd al-‘Âlam states quite categor- ically that most of the slaves coming from Khazaria to the Muslim world stemmed from the Turkic Pecheneg groupings that were subject to the Khazars or within the Khazar orbit133. Al-IÒ†akhrî notes that only pagans are slaves among the Khazars since the , Christians and Muslims of that land all prohibited the enslave- ment of one of their correligionists134. In light of this and the growing social and economic differentiation between the members of the monotheistic faiths and the pagans, Ludwig suggests that the bulk of the slaves coming into the Islamic world must have come from this pagan, lower class135. This is an interesting supposition one that requires still further substantiation from archaeological sources. In any event, it would probably be most relevant for the period after the latter half of the ninth century, the period from which our sources derive much of their

131 Ludwig, Struktur, pp. 204-205, citing the Syriac source Dionysios of Tell MaÌrê (Chronique de Tell-MaÌrê, Quatrième partie, ed. J.-B. Chabot, Paris, p. 85, trans. p. 72). Ludwig also notes a Muslim commander who had the nisba “al-Khazarî” who partici- pated in a 737 Arab campaign against the Türgesh in Khurâsân. The reference is perhaps to AÒim al-‘Uqaylî al-Jazarî iQõ÷G not iQõÿG) who is reputed to have been able to speak Khazar, see a†- ™abarî, ed. Ibrâhîm, VII, p. 114, Eng. trans. The History of al-™abarî, vol. XXV, The End of Expansion (Albany, 1989), p. 132 and Bal‘amî’s Persian rendering of a†-™abarî in Tabari, Les Omayyades, trans. H. Zotenberg (a partial reprint of Chronique de Abou Djafar Mohammed ben Djarir ben Iezid Tabari, trans. H. Zotenberg, Paris, 1867-1874, Paris, 1983), p. 241. Dunlop, History, p. 73 n.70, however, considered the form “al-Khazarî” to be correct. 132 Thus, the Îudûd al-’Âlam. The Regions of the World, ed. trans. V.F. Minorsky (London, 1937, reprint with addenda: 1970), p. 161 on Khazaria: “This is a very pleas- ant and prosperous country with great riches. From it come cows, sheep and innumerable slaves.” 133 Îudûd/Minorsky, p. 160; Ludwig, Struktur, p. 206. 134 al-IÒ†akhrî, ed. De Goeje, p. 123. 135 Ludwig, Struktur, pp. 206-207.

Journal Asiatique 292.1-2 (2004): 279-309 308 P.B. GOLDEN information and the period by which the various monotheistic faiths were firmly established. Another source of recruitment of Turkic soldiers were the children of the ghulâms already in service. A†-™abarî reports, s.a. 251/865 that the Caliph al-Musta‘în (862-866) chastised a group of Turkic commanders, remarking: “didn’t you bring your children for me to put in the military rolls, about two thousand of them, and didn’t I agree to let them join you?”136 Presumably, these are the offspring of the ghulâms already residing in Sâmarrâ for whom the Caliphate had provided wives. The twelfth century genealogist, ’Abd al-Karîm as-Sam’ânî, lists sev- eral Muslim worthies with the nisba “al-Khazarî” and the indication that they were of Khazar stock. None of these figures bears a Turkic name137. Clearly whatever their origins, they were long and far removed from Khazaria. The Khazar onomastic material has been dealt with elsewhere138. Not surprisingly, it shows, in addition to the expected Turkic and Arabic ele- ments, names of Iranian origin as well. Some of these may, perhaps, point to Khwârazmian groupings that formed the comitatus of the Khazar rulers, or names acquired in Sâmânid schools.

SUMMARY The Khazars were the leading opponents of Umayyad expansion into the North Caucasus and beyond. One of the little-studied areas of Khazar-Arab interac- tions is the role played by ghulâms of Khazar origin in the ‘Abbâsid Caliphate. The nature of the ghulâm system remains the subject of controversy. Was it truly a slave system or was it simply a means of expressing political-personal loyalty? Were its origins native to the Islamic Near East or was it imported from Central Asia (cf. the Soghdian châkar system)? The author is inclined to see important Iranian and Turkic notions about the ruler’s comitatus as key factors in shaping this institution. The ghulâms in a new political-social environment,

136 a†-™abarî, ed. Ibrâhîm, IX, pp. 283-284, Eng. trans. The History of al- ™abarî, vol. XXXV, The Crisis of the ’Abbâsid Caliphate, trans. G. Saliba (Albany, 1985), p. 35. 137 ’Abd al-Karîm b. MuÌammad b. Man Òûr at-Tamîmî as-Sam ’ânî, Al-Ansâb, ed. ’Abdur RaÌmân al-Yamânî ( Îaidarâbâd, 1966), V, pp. 121-122. 138 Golden, “Khazar Turkic Ghulâms: Onomastic Notes” AEMAe 12 (2002-2003), pp. 15-27.

Journal Asiatique 292.1-2 (2004): 279-309 KHAZAR TURKIC GHULÂMS IN CALIPHAL SERVICE 309 segregated from the rest of the population, may have formed ethnic-based cliques. There is some evidence for the existence of a Khazar faction. Some Khazar ghulâms (e.g. Îtâkh, Bugha al-Kabîr, IsÌâq b. Kundâjîq) attained posi- tions of great prominence. Khazars were recruited from prisoners of war, the slave trade via Khazaria, Volga Bulgharia, Khwârazm and the larger Sâmânid world. The number of Khazar ghulâms in Caliphal service declined by the early tenth century (era of Rus’ Volga-Caspian raids).

RÉSUMÉ Les Khazars furent les principaux opposants à l’expansion omeyyade au nord du Caucase et au-delà. L’un des domaines les moins connus des relations entre Arabes et Khazars concerne le rôle joué par les ghulâms d’origine khazare dans le califat abbaside. La nature du système des ghulâms demeure un sujet de dis- cussion: ce système était-il basé sur l’esclavage ou n’était-ce qu’un moyen d’ex- primer une loyauté personnelle et politique? Est-il né au Proche-Orient ou provient-il d’Asie centrale? (cf. le système sogdien des Châkars)? L’auteur considère que la notion de comitatus si importante dans les mondes turcs et iraniens a joué un rôle clé dans la constitution du système des ghulâms. Les ghulâms, dans un nouvel environnement social et politique, séparés du reste de la population, ont pu former des groupes de solidarité à base ethnique. On a quelques indices de l’existence d’une faction khazare. Certains ghulâms khazars (Îtâkh, Bugha al-Kabîr, IsÌâq b. Kundâjîq) sont parvenus à de hautes positions. Ils étaient recrutés parmi les prisonniers de guerre et les esclaves commerciali- sés via la Khazarie, la Bulgarie de la Volga, le Khorezm et l’empire samanide. Leur nombre déclina passé le début du dixième siècle, période des raids Rus’ par la Volga et la Caspienne. Mots clefs: Khazars, Ghulâms, Châkars, Comitatus, ‘Abbâsid Caliphate

Journal Asiatique 292.1-2 (2004): 279-309