Academia.eduAcademia.edu
Marek Jankowiak (Newton Fellow, Wolfson) Medieval Seminar, All Souls, 27 Feb 2012 Dirhams for slaves. Investigating the Slavic slave trade in the tenth century The idea that a massive trade in Slavic slaves underpinned the economic growth of Europe in the 9th and 10th centuries is not new. It is, however, most often only implicit; and at any rate, it is very rarely discussed. Even though this was an obvious idea to the 19th-century Orientalists1, the interest in slave trade gradually faded away, and virtually disappeared in the 70s and 80s. Recently, however, the concept of a large-scale slave trade that allowed the Western European economies to emerge from the Dark Ages has resurfaced in the Anglo-Saxon scholarship. Voiced emphatically by Michael McCormick ten years ago, it was readily accepted in a number of recent syntheses; but the reaction on the Continent was less enthusiastic. The debate which was thus reignited has already led to some progress, but – as can be seen from a map of Michael McCormick2 – there are still some gaps to fill. The key area for the understanding of the mechanisms of the medieval slave trade has been disregarded. As рдб улоа sиaтб indicates, no such study can be complete without an effort to understand what was happening in the Slavic lands. And this is not any more a vain pursuit: the archaeological material accumulated in the last 20 years allows a foray into this still virtually unexplored area. The objective of this paper is to get a sense of how the trade in Slavic slaves operated, of its scale, and of its impact. I will use the numismatic evidence and written sources as a starting point for a risky interpretation of the archaeological material. Dirhams are good quality silver coins weighing on average 2.9 g. What makes them an invaluable source for a historian, is the inscription of the place and date of their minting. As opposed to early medieval Western European coinages, where a significant effort is needed to establish general chronological outlines, Arabic dirhams provide absolute and usually exact datings. We can thus follow in detail their flow into the Northern Lands, where they are found in large quantities in hoards. It is difficult to estimate the total quantity of dirhams discovered since records started in the 18th century. The biggest existing database contains 200,000 coins3, but many hoards had been scattered before they could be recorded. Over 1,000 hoards are known, and they may have comprised in total as much as 1 million coins. This is a very significant number. The Islamic dirhams would probably constitute 1 D.A. Khvolson, И ѣ ія х хъ, хъ, хъ, ья хъ, я хъ хъ… , St. Petersburg 1869;; G. Jacob, Welche Handelsartikel bezogen die Araber des Mittelalters aus den nordisch- baирiscдбn Länабоn?, 2. Auflage, Berlin 1891; J. Markwart, Osрбсолмäiscдб сnа лsрasiaрiscдб Sробiвzсгб, Leipzig 1903. 2 Origins of the European economy. Communications and commerce, A.D. 300-900, Cambridge 2001, p. 762. 3 Esрimaрб basба лn R.K. Kлтaибт, A.C. Kaбиin, Ciоcсиation of Arab Silver in Medieval Afro-Eurasia: Preliminary Obsботaрiлns , History Compass 5 (2007), p. 560–580. 1 the largest body of statistically usable material for the early Middle Ages that would allow invaluable insights into the history of Eurasia – if only they were adequately published. Unfortunately, no exhaustive list of finds exists for Sweden, the numerous new finds from all over Northern Europe are only exceptionally published, and the immensely rich collections of the coin cabinets in Stockholm and St. Petersburg remain virtually unavailable. Few hoards have been fully studied, and cases such as the recent find of the huge hoard in the settlement of Spillings in Gotland which weighed almost 70 kg, where extensive archaeological excavations have been conducted on the find spot, are exceedingly rare. The dramatic increase in new finds brought about by metal detecting goes hand in hand with the shrinking number of specialists capable of documenting them. Overall, it is probably fair to qualify the research on the Islamic coins in Northern Europe as moribund. My observations will be mainly based on a detailed list of hoards that I am compiling, and on a database of 48,000 coins, mainly from the Swedish collections, created by Gert Rispling that he kindly made available to me. If we look at the geographical distribution of the hoards (Fig. 1), the most salient feature is the cluster of ca 350 hoards on the Baltic island of Gotland. Fig. 1 – Ninth- and tenth-century dirham hoards. A similar number is recorded in Russia and Ukraine; big concentrations of hoards can also be found in Poland and Estonia. This is not a map of the Viking world: few dirhams have been found in Norway, 2 and the hoards from Denmark and mainland Sweden cannot stand comparison with Gotland or Poland. I will pass over the reasons for hoarding, which is probably best understood as a cultural phenomenon resulting from the ways of displaying status in Scandinavian societies. Many hoarding zones in Eastern Europe can indeed be connected with areas of Scandinavian settlement; in others, however, a large-scale Scandinavian presence cannot be proved, for instance in Estonia and in the Polish interior. The stream of the dirhams flowing to the Northern Lands fluctuated significantly over time. The aggregated data for all the northern hoards show two periods of hoarding: a moderately dynamic in the 9th century, anа a гбnсinб siитбо осsд in the 10th century, climaxing in the 950s with almost 120 hoards in a single decade. The two periods are separated by a lull in the 880s and 890s. On a regional level, variability is even greater. With the exception of Gotland, where dirhams were buried throughout the 9th and 10th centuries, no other region was able to attract the flow of dirhams for a longer period of time. For instance, all hoards from East Prussia are dated to the years 810-830, Danish hoards are as a rule later than 920, Estonian finds are concentrated between 940 and 980, whereas hoards clustered in Eastern Poland have termini post quem between 890 and 910. This is a highly complex and dynamic picture. It tells us something on the nature of the trade between the Islamic world and the Northern Lands. The system as a whole was resilient enough to survive for almost two centuries, but its actors were in constant flux. They seem to emerge abruptly in seemingly random places, and to disappear even more rapidly. Only around the mid-10th century do the trade routes start to stabilize, when hoards clearly concentrate in areas such as Greater Poland, the region of Kiev and Svealand, where written sources would soon record the existence of early states. A link between the hoarding of the Islamic silver and state formation in Northern Europe can thus be posited. If dirhams were flowing to unpredictable areas of Northern Europe, it is easier to identify where they were minted. In the 9th century Baghdad and other Iraqi mints produced most of the coins that ended up in the hoards, while in the 10th century over 80% of dirhams originated in the Samanid emirate centred in Transoxania. Almost no coins of the Spanish Umayyads are known from Northern Europe, despite the well-attested trade connections. The shift in the main producing centres of the dirhams flowing to the Northern Lands coincides with the two phases of hoarding we have already observed. In what follows, I will concentrate my attention on the second, 10th-century phase of the Oriental trade. One category of coins deserves a special mention: the so-called imitations. At first sight, they are just poor copies of the original Abbasid or Samanid dirhams. They are found in significant quantities in the Northern European hoards: probably more than 10% of hoarded dirhams are imitations. Gert Rispling from Stockholm studied their dies extensively, and was able to build remarkably long chains of dies, resulting from the combination of different obverse dies with different reverse dies. These die-links show that imitations were produced in a limited number of mints, certainly outside the Islamic world, as their blundered legends demonstrate. Their silver content, perhaps surprisingly, was not inferior to that of regular Islamic issues. If they were not a means of introducing debased metal into circulation, why were they produced? Risмиinг s sрсаiбs suggest that the 9th-century imitations were chiefly 3 produced by the Khazars, while the 10th-century imitative coinage can be attributed mainly to the Volga Bulgars. They seem to have been produced ad hoc – as frequent permutations of the dies demonstrate – probably in order to correct some market disequilibrium. If we imagine that goods were exchanged at yearly fairs – and the account of Ibn Fadlan does suggest a yearly rhythm of trade – we can conceive that the Bulgars and the Khazars were acting as middlemen between the Northern and the Islamic merchants. This is confirmed by the written sources to which I will now turn. There is no shortage of written sources on the slave trade in the 10th century. The Arab geographers are explicit that dirhams were paid mainly for slaves and furs. Ibn Rusta, who was writing at the beginning of the 10th century, describes how the Rus, which at this stage means Scandinavians settled in Eastern Europe, were making profits on the Slavic – in Arabic Saqāliba – slaves: Tдб Rсs (…) raid the Saqaliba, sailing in their ships until they come upon them, take them captive and sell them in Khazaria and in Bulgar. They have no cultivated fields and they live by pillaging the land of the Saqaliba. (…) They have no dwellings, villages or cultivated fields. They earn their living by trading in sable, grey squirrel and other furs. Tдбy sбии рдбm вло siитбо cлins удicд рдбy sбр in bбирs anа убaо олсnа рдбiо уaisрs. (…) They treat their slaves well and dress them suitably, because for them they are an article лв роaаб. 4 The mention of Khazaria and of Bulgar confirms the picture emerging from the study of the imitations, namely that exchanges were concentrated in a small number of marketplaces. From Bulgar, we have the famous eye-witness account of Ibn Fadlan, who visited it in 922, and who emphasizes the central importance of the slave trade: I saу рдб Rсs, удл дaа cлmб вло роaаб anа camмба by рдб оiтбо Iрiи. (…) Round their necks, [their women] wear torques of gold and silver, for every man, as soon as he accumulates 10,000 dirhams, has a torque made for his wife. When he has 20,000, he has two torques made [and so on]. (…) Wiрд рдбm, рдбоб aоб bбaсрiвси sиaтб гiоиs, вло saиб to the merchants. Each of the men has sex with his slave, while his companions look on. (…) As sллn as рдбiо bлaрs arrive at this port, each of them аisбmbaозs (…) anа prostrates himself before a гобaр iали, sayinг рл iр: Oд my илоа, I дaтб cлmб волm a вaо country and I have with me such and such a number of young slave girls, and such and such a number of sable skins (…). I улсиа иiзб yлс рл ал рдб вaтлсо лв sбnаinг mб a merchant who has large quantities of dinars and dirhams and who will buy everything that I уanр anа nлр aогсб уiрд mб лтбо my моicб . 5 Ibn Faаиan s insisрбncб лn sиaтб гiоиs is мбодaмs mлоб рдan тлyбсоism. Iр probably tells us something about the nature of the demand for Slavic slaves in the Islamic world. Saqaliba slaves are quite conspicuous in Arabic sources. In accordance with the then prevailing views on the merits of 4 P. Lunde, C. Stone, Ibn Faаиān and the land of darkness : Arab travellers in the far north , London 2012, p. 126. 5 P. Lunde, C. Stone, Ibn Fadlān anа рдб иanа лв аaозnбss : Aоab роaтбиибоs in рдб вaо nлорд , London 2012, p. 45- 48. 4 individual races, they were thought to be suitable for specific types of service. Thus, in Iraq Saqaliba are unattested as slave soldiers or administrators, unlike the Turks, or even as agricultural or mining workforce similar to the Zanj slaves. They are, however, amply attested in the urban literature of the bourgeoning cities of Iraq as domestic servants, concubines and eunuchs. To give an example, the works of al-Jahiz, a prolific writer from Basra who died in 868/9, are full of allusions to the Saqaliba. In his Book of Animals he describes, for instance, how to improve a Slavic slave: Iв рдбоб aоб рул Sиaтic bолрдбоs волm рдб samб mлрдбо anа вaрдбо, even if one of them is the twin brother of the other, when one of them is castrated, he becomes a better servant and smarter in all kinds of activity and manual work. He will be more skilled in them and more fitting for them. You will also find him more intelligent in conversation – these are all his qualities. His brother will remain in his innate ignorance, natural stupidity and Slavic simple-mindedness; he will also be unable to understand foreign languages. His hand will be clumsy and he will not become skilful, because his intellect will not be trained. He will not be abиб рл бxмобss дimsбив воббиy anа билнсбnриy, nло рл молnлсncб cибaоиy. (…) Tдб first result of the castration of a Saqlabi is the purification of his intelligence, sharpening of his acumen, strengthening of his nature and stimulation of his mind. 6 This passage probably explains the shortcomings of my paper, but more importantly it highlights the substantial presence of the Slavic eunuchs in Basra around the mid-9th century. It would seem that trade in the Slavic slaves consisted mainly of high-value individuals, such as eunuchs and young women. This peculiar profile of the slaves imported from the Slavic lands may explain the apparent profitability of this very long-distance trade. The Arab sources thus create the impression of ubiquity of the Slavic slaves. The question of the scale of this slave trade is crucial for the understanding of its mechanisms. Therefore, I would like to propose the risky exercise of recalculating dirhams into slaves. For all the uncertainty involved in this kind of calculation, I think that it is useful to estimate the order of magnitude of the slave traffic to which the number of dirhams found in the Northern Lands corresponds. The details of this calculation are illustrated below (Fig. 2); the result is in the order of tens of thousands, and not hundreds of thousands, of Slavic slaves sold to the eastern part of the Islamic world in the 10 th century. This may seem intellectual charlatanry, but I think it is unlikely that we err here more than by an order of magnitude. Assumptions are as prudent as possible, including the one concerning the percentage of the total original flow of silver that is preserved in the recorded hoards. This ratio can be estimated on the basis of the die studies of the imitations, if we assume that virtually all of them were exported to the Northern Lands. Overall, this is still a high number of slaves, but, on the other hand, it can be recalculated into several hundred, at best a thousand and several hundred, of slaves sold per year, a scale which seems perhaps less spectacular. 6 Lewicki, Źоóаła aоabsзiб ал аziбjóу Słлуiańszczyzny , тли. 1, Wолcłaу 1956, м. 166-167. 5 Fig. 2 – Simulation of the number of slaves sold into slavery in the 10th century. This calculation does not include Saqaliba slaves exported to Spain. Spain is a special case. On the one hand, Saqaliba slaves are well attested in Spain, again mainly as domestic servants: at least several thousand of them are said to have lived in Madinat al-Zahra under the caliph Abа aиRahman III in the mid-10th century7. But the Saqaliba also played in Spain a role which in the eastern part of the Islamic world was reserved for the Turks. They constituted a significant part of the caliphal administration and guard, and used these positions as springboards to spectacular careers. A recent prosopography lists by name over 100 Saqaliba from Umayyad Spain 8, a result impressive both by the number of the known individuals and by the diversity of positions in the army and administration they were filling. Spain was thus another important zone of demand for Slavic slaves. We are here getting close to a crucial point. Even though thousands of Slavic slaves made their way to 10th-century Spain, only several dozen of 10th-century Spanish coins were found in the northern hoards. Moreover, very few 10th-century silver hoards are known from the region of Prague, which is 7 Ibn Idhari, Kiрāb aи-bayān aи-mсгдоib вī aздbāо aи-Andalus wa-al-Maghrib, ed. G.S. Cлиin, E. L‘тi-Pолтбnçaи, Leiden 1948-1951, II, 232. 8 M. Meouak, Ṣaнāиiba, бсnснсбs бр бscиaтбs à иa cлnнс’рб ас млстлiо. G‘лгоaмдiб бр дisрлiоб абs ‘иiрбs politiques "marginales" dans l'Espagne umayyade , Helsinki 2004 (Annales Academiae Scientiarum Fennicae 331), 156-207. 6 associated in our sources with the exportation of the Slavic slaves to Spain. If dirhams were indeed flowing to the Northern Lands as the payment for slaves, why, in a region were slave trade is attested by written sources, do we get no silver at all? We find a hint in the description of al-Andalus by yet another geographer, Ibn Hawqal, who travelled extensively between 943 and 973 and visited both Spain and Central Asia: Onб лв рдб famous items of their merchandise is handsome slave-girls and slave-boys captured in the land of the Franks and in Galicia, as well as Saqaliba eunuchs. All the Saqaliba eunuchs on the surface of the earth are imported from al-Andalus, because they are castrated near that country, and this is done by Jewish merchants. (...) The country [of the Saqaliba] is long and wide. (…) The sea-arm stretching from the ocean towards the country of Gog and Magog traverses their country (...) cutting it into two halves. Thus half of their country, along its whole length, is raided by the Khurasanis who take prisoners from it, while its northern half is raided by the Andalusians. (…) In these areas, many captives can still be obtained 9. A look at the world map appended to the work of Ibn Hawqal may help us to understand this account (Fig. 3). If we turn it upside down, to orientate it northwards, we see a body of water, which extends from Constantinople to the north, presumably through the Black Sea, somewhere to the Outer Ocean. This concept serves to explain how the two extremities of the Islamic World, Spain and Khorasan, could possibly have access to the same pool of Saqaliba slaves. Ibn Hawqal conjectures that this was bбcaсsб рдб иanа лв рдб Saнaиiba is илnг anа уiаб , удicд is nлр вaиsб. Bср in обaиiрy, when saying that the Spanish and the Khorasanis were getting their Slavic slaves at the opposite two ends of the Slavic lands, he implies the existence of two separate trade systems. We have already seen the eastern one, where the Rus were selling Slavic slaves for Samanid dirhams at the markets of Bulgar and of Itil. As for the other one, sources suggest that the Spanish market was supplied by Jewish merchants who were buying Slavic slaves at the market of Prague. 9 Ibn Hauqal, Configuration de la terre (Kitab surat al-Ard), tr. J. H. Kramers, G. Wiet, Beyrouth 1964, vol. 1, p. 109. 7 Fig. 3 – Ibn Haунaи s maм лв рдб улоиа. We have a travelogue of one of these merchants, Ibоaдim b. Ya нсb from Tortosa, who travelled around 960 to Prague. He stresses the importance of its market, and lists slaves among its chief commodities. Other sources confirm the connection between the slave market of Prague, the Jewish merchants, and Spain. But why there are no dirhams in Prague? Several explanations may be proposed. First of all, there are also no dirhams in the Carolingian Empire and its successor states, which indicates some control over currency crossing the borders. With some bewilderment, Ibrahim b. Ya нсb noted that trade in Prague was conducted not with monetary means of exchange, but by means of kerchiefs, small pieces of cloth, which had an exchange rate versus silver. The availability of silver was clearly lower in Prague than in regions trading with the Samanids; and we should not dismiss a cultural reason for the absence of hoarding: Czech warriors apparently did not display their status through the possession of vast quantities of silver. Thus two distinct systems of trade in the Slavic slaves were in operation in the 10th century, one run by the Rus and other Scandinavians, and marked by hoards of dirhams; the other dirhamless, centred on Prague and supplied by the Czech dukes (Fig. 4). Even if other slave markets operated in the 9th and 10th centuries – for instance, Constantinople, apparently supplied at least partly by the Magyars – these two trade systems are likely to have been the most massive. They may also have left the highest number of material traces, as I will try to demonstrate now. 8 Fig. 4 – Two systems of trade in the Slavic slaves in the 10th century. Can we hope for any archaeological traces of the trade in Slavic slaves? Can we say more precisely where the slaves originated, how they were enslaved, and how they were transported to the slave markets of Bulgar, Itil and Prague? Slave trade is usually thought to be invisible archaeologically. Its direct traces are indeed difficult to identify. Twenty years ago, Joachim Henning composed a catalogue of iron fetters from medieval Central Europe10. Of the 32 finds he listed, only 6 can be dated to the 9th-11th centuries. They all come from the fringes of the Slavic world, from regions such as Khazaria that, as we have seen, played a pivotal role in the slave trade. They thus attest to the reality of the slave trade, but it is difficult to draw any other conclusions. To my knowledge, other forms of direct evidence, such as distinctive burial types, have not yet been identified, neither in the Slavic lands nor in the Islamic world. Methods such as strontium analysis or ancient DNA sampling cannot thus be applied. Genetic studies of the modern populations may also prove to be ultimately inconclusive, given that Central Asian populations were probably exposed to contact with Slavs also in later times, as a result of Mongol invasions or of the modern Russian colonization. Recent studies on the archaeology of slavery in Africa are informed by a different approach. A collection of studies published several months ago begins, on the first page, with the following 10 J. Hбnninг, Gбвanгбnбnвбssбиn im sиaуiscдбn Siбаисnгsоaсm сnа або бсолмäiscдб Sзиaтбnдanаби im 6. bis 12. Jaдодсnабор , Germania 70 (1992), p. 403–26. 9 statement: nбу обsбarch focuses at the landscape scale, rather than attempting to find physical evidence for slavery per se, assessing instead the settlement systems of slavery-based economies, and the depopulation and abandonment which followed from wars of enslavement 11. This approach involves thus, in the first place, searching for evidence of sudden depopulation or population movements. Taking the example of Poland, I will try to show that the idea of applying such a methodology to Slavic archaeology is not eccentric. Polish archaeology is in a post-revolutionary period. The traditional chronological framework based on the periodization of pottery collapsed after a substantial number of sites had been redated by dendrochronology. To give an example, Gniezno, the capital of the first Polish state, which was thought to have been built at the end of the 8th century, has now been shown to have been built 150 years later, in the year 940. If we take a wider perspective, almost no sites can currently be dated to the 8th century, and quite few to the 9th century. The situation changed abruptly in the 10th century, where fortresses were built on an impressive scale and silver was hoarded in several areas of the modern Poland. It is, however, unclear what caused this sudden acceleration. A particularly promising area for the landscape approach lies at the southern edge of the core territory of the Piast state in Greater Poland. Population studies, based on palynological data and archaeological surveys, show that around the year 1000 the population was concentrated around the central fortresses of the Piast state such as Gniezno. Outer parts of Greater Poland were less densely populated, in particular the western part of the region. Now, in the previous phase of settlement the picture was exactly the opposite: it was precisely the western part of Greater Poland which had the highest population density, whereas the future Piast domain was almost empty (Fig. 5)12. The dating of this earlier phase is controversial: this study was published when the old chronological framework was still in place, hence the dating to the 7th-8th centuries. The difficulty is that almost no settlements at all can currently be dated to this period, therefore a date in the 9th, or perhaps even in the early 10th century is much more likely. 11 P.J. Lane, K.C. MacDonald, Slavery in Africa. Archaeology and Memory , Proceedings of the British Academy 168, Oxford 2011, p. 1. 12 S. Kсоnaрлуsзi, Sрсаia лsaаniczб naа sросзрсоą zasiбаибnia i глsмлаaозą , Ziemie polskie w X wieku i ich znaczбniб у зszрałрлуaniс się nлубj maмy Eсолмy , ба. H. Samsлnлуicz, Kоaзóу 2000, м. 329-350, esp. p. 339340 n. 50, and maps 17 and 18. 10 9th/10th c.? ca 1000? Gniezno Poznań Gniezno Poznań rnatowski, “tudia osadnicze nad strukturą zasiedlenia i gospodarką , Ziemie polskie w X wieku i ich znacz kształtowaniu się “tudia Fig. 5 –Population distribution in Greater Poland. kształtowaniu się strukturą gospodarką , Whatever the precise dating, the implication of this study is that there were significant population movements in Greater Poland before the year 1000, when entire tribes were apparently disappearing. The new archaeological material supports such a reading. A group of hillforts built on the banks of the river Obra, which was the axis of the area which was depopulated before the year 1000, has been dendrochronologically dated to the turn of the 9th and 10th centuries. According to the current state of research, this is the earliest compact group of hillforts in this part of Poland. They predate, in particular, the fortresses of the core Piast territory, where only two very small forts are currently known to have existed around the year 900. The earliest of the big Piast hillforts was built around 920, while Gniezno and several other big fortresses can be dated to 940. Another group of hillforts was built in the 950s along a line situated only 20 kilometres south of the Obra river; but here datings are less secure. The dates of hoards interestingly correlate with these groups of hillforts. The earliest hoard from Greater Poland has the terminus post quem of 941 and comes from the area of the oldest hillforts; this and another roughly contemporary small hoard are the only hoards from this territory; no silver hoard is known from after around 950. In the Piast zone, hoards are slightly later and can be dated to the late 940s; the inflow of silver then accelerates in the 950s and continues for decades. Dendrodates and termini post quem of hoards correlate very closely also in other regions of Poland, which suggests some relation between the inflow of silver and the construction of hillforts. It is tempting to connect the construction of fortresses on either side of the string of hillforts along the Obra river with the destruction of the tribe which had dwelt in them. Traces of destruction are indeed abundant: most of the Obra hillforts suffered a violent end documented by burnt layers where human bones are often scattered. These layers are dated by the archaeologists to mid-10th century or slightly later. Written sources suggest a terminus ante quem of 965, when the Piasts were in direct contact with the dukes of Prague, which presupposes that the intervening tribes had been by that time 11 eliminated or subjected. In my view, these types of evidence concur to suggest the violent elimination of the Obra tribe somewhere in the 940s. 9th/10th c.? S. Kurnatowski, “tudia osadnicze nad strukturą zasiedlenia i gospodarką , Ziemie polskie w X wieku i ich znaczenie w kształtowaniu się nowej mapy Europy, ed. H. Samsonowicz, Kraków 2000, map 17 Fig. 6 – Hillforts (circles), destroyed hillforts (crossed circles) and hoards (squares) in Greater Poland. This type of approach, based on the combined evidence of dendrochronology, hoards, surveys and excavations, suggests that also other regions in the western Slavic lands – such as Mazovia in Poland or Galicia in Western Ukraine – suffered depopulation in the second half of the 10th century. Further research is needed to confirm this intuition, but it can perhaps be conjectured that Slavic slaves on the markets of Prague, Bulgar and Itil originated from this kind of unfortunate tribes, reduced to slavery by neighbouring bands of warriors, such as the early Piasts or Premyslids. My last point is that a long-distance slave trade involves sophisticated logistics. How these Slavic slaves were transported from their homelands to the slave markets? This brings me to the hillforts. They are the main feature of the early medieval landscape in the western Slavic lands. 2,312 hillforts have been catalogued in Poland, aолсnа 640 aоб абscоibба in рдб обcбnр Encycилмaбаia лв дiиивлорs in рдб Czбcд Rбмсbиic , several hundred were counted in Germany and in Ukraine. Some were built in the Bronze Age, some as late as in the 13th century, but those erected in the 9th and 10th centuries are a sizeable group if not a majority. They are often thought to be a response to a general feeling of insecurity, but it has not been explained what caused this sudden surge in insecurity. Hillforts varied 12 in size and hence no doubt in purpose (Fig. 7). Small hillforts, rarely exceeding a hectare or two, were mainly built in Germany and Northern and Western Poland. In some areas there are clusters of bigger fortifications, enclosing up to 5 ha; the two most prominent such clusters coincide with concentrations of hoards. A different kind of hillfort is common in Southern Poland, the Czech Republic and Slovakia, where hillforts were fewer and bigger, often reaching 20 ha and more. Western Ukraine is an exception: in addition to small and medium hillforts, several gigantic fortresses, extending over hundreds of hectares, are known. They can apparently be dated to the 10th century. mainly small hillforts (<1 ha) big hillforts (up to 5 ha) small and very small hillforts big hillforts (up to 5 ha) very big hillforts (up to 25 ha) gigantic hillforts (200-400 ha) Naszacowice Fig. 7 – Hillforts in Poland. In order to understand what was the purpose of the big and very big hillforts, I would like to take the example of Naszacowice in southern Poland. With its 15 ha it is the second biggest hillfort in Poland. It is situated on the high bank of the Dunajec river, an important communication corridor. As you can see the site commands the valley and offers excellent views. It has a relatively complicated plan, with a central section surrounded by two bigger sections and a third smaller one, which do not communicate with each other (Fig. 8). This site has been thoroughly excavated over a period of 20 years, but yielded very few finds13. Apparently the only house or two was built against the inner side 13 J. Poleski, Naszacлуicб. Ein воüдmiррбиaирбоиicдбо Bсогуaии am Dсnajбc, 2 vols., Kоaзów 2004. 13 of the inner rampart. Other finds are limited to a skeleton of a young individual and a couple of ornaments of late Avar, Moravian and post-Moravian origin. Fig. 8 – Naszacowice. What does this mean? Naszacowice was not a fortified settlement, because it is empty. It was also not a refuge, because, on the one hand, it was situated on the main communication artery in this region – precisely the kind of environment which would be abandoned when enemy was approaching – and, on the other hand, no early medieval settlements are known from its vicinity. Still, substantial effort was invested in its construction. I propose to interpret it as a slave camp, garrisoned by warriors based in its central section who were also certainly keeping watch over the outer ramparts. It would have been periodically filled with captives on their way to a slave market, most probably that of Prague. I am tempted to explain in the same way the other big empty hillforts from southern Poland, the Czech Republic and western Ukraine. It is time to bring together the dirhams, written sources and hillforts. The zone of the big empty hillforts coincides not only with the area almost devoid of dirham hoards, but also with the zone of influence of the dukes of Bohemia. We can perhaps consider this territory to be the catchment zone of the slave market of Prague. In that case, the big empty hillforts can be understood as nodal points of its system of supply, which relied on land transport over significant distances. Big caravans – such as the one 14 Ibn Fadlan joined on his way to Bulgar, which was composed of 3,000 horses and 5,000 men – were probably, we would say, the most cost-effective means of conveying slaves by land over hundreds of kilometres. So why there are no big hillforts in the dirham area? Because of the navigable rivers, which enabled frequent, smaller shipments of the human cargo. Both the logic and the logistics of the two trade systems must have been very different. I recapitulate. Written sources show that slaves were the most important commodity traded in the 10th-century Northern Europe; the sheer quantity of dirham hoards proves that its scale was far from negligible. I suggest that this slave trade left a much deeper imprint in the archaeological material than is usually thought. I realize that much work remains to be done, in the first place recording and publishing the numismatic material, and reviewing a larger body of archaeological evidence. But I think that this interpretation has the advantage of providing a coherent explanation for the extraordinary flow of Islamic dirhams to Northern Europe. It was the underlying slave trade with the Islamic world, and not with other parts of the Medieval world, that led to an unprecedented accumulation of capital, which was then reinvested in creating the Northern European states that we know today. 15