1. Amphora (Double-Handled Clay Jars)

1. Amphora (Figs. 3, 4, 5). The amphorae - usually large double-handled clay jars with elongated bodies. and spike bases - were commonly used in the ancient Mediterranean as containers of liquid stock such as wine, olive oil, fish sauce and other products. Production centres of various types of amphorae proliferated in the littoral regions of the Mediterranean. During the B.C./A.D. changeover, wine and other products were regularly exported in amphorae to India along the Egypt-India maritime route. The Periplus (sec 39,49,56,60) informs us of the arrival of Roman wine at ports serving north and south India Reference to Yavana (Mediterranean) wine arriving by sea is to be found in the Tamil Sangam poetry (Cimino and Scialpi 1974: 22). The wide distribution of

24 amphora find-spots in the subcontinent corroborates the textual evidence for Mediterranean wine imports into India (Fig. 4). Distribution Mediterranean amphorae have been found in all littoral-regions except Southeast Asia. In the Red Sea littoral amphorae have been recovered in substantial quantity on the Egyptian coast. Excavations at the Roman port-site of Quseir alQadim (Fig. 1) have yielded a number of amphora plugs and substantial quantity of amphorae (Sidebotham 1986 a plate. 10, pp. 56). The site of another major trade-port, Berenice, has also yielded amphorae on the surface (Meredith 1957:58). Prospections carried out by Sidebotham et al. (1989:127-166) at the Roman/Byzantine site of Abu Shah'r on the northern Egyptian Red Sea coast has brought to light a major amphora assemblage. However, most of the Abu Shah'r amphorae are late from our point of view, being dated between 3 rd-7 th century A.D. A major percentage of the Early Historic ceramic collection made by the Centre National de la Recherche Scientifique, Centre D'Etudes Alexandrines on the land routes connecting Egyptian coast with the Nile comprises Egyptian and Mediterranean amphorae of the Dressel 2-4 variety. This amphora type made up approx. 50 % of the total amphora finds (Dr.J.Y.Empereur, C.N.R.S., Alexandria: Alexandria: personal communication). Egyptian and Mediterranean amphorae have come to light at Coptos, the main inland transhipment centre on the Nile connecting the Red Sea coast with the entrepot of Alexandria (Monica Smith, University of Michigan: personal communication). An integrated survey and classification of amphorae from the Red Sea coast of Egypt has yet to be published. Riley (in Sidebotham et al. 1989:149) points to the nascent status of ceramic studies in this region when he says: "The Red Sea coastal region (is) one of the last frontiers for pioneering Roman pottery research..." The intensity of fieldwork in other parts of the Red Sea has been less than in Egypt. In particular, the vast stretches of Saudi Arabian, Sudanese and Ethiopian coasts remain under-prospected However, large corded amphorae of Egyptian origin dated to 3 rd-4 th century A.D. were recovered from the ancient port-site of Adulis in Ethiopia in the early part of this century (Fig. 3; Casson 1984: 208) Mediterranean amphorae have come to light in both the South Arabian and Somali coasts of the Gulf of Aden (Fig. 3) Specifically, the sites yielding amphorae are Qana (Yemen), Khor Rori (Dhofar, Oman), Shabwa (Yemen), Ras Hafun (Somalia), Socotra Island (Yemen) and Masirah Island (Oman)

25 The most interesting results from our point of view have been obtained from Qana (ancient Kane of the Periplus). This site, being excavated by a joint RussianYemeni expedition since 1985 has revealed three occupational phases: Lower, Middle and Upper Significantly, 80% of the pottery recovered from the three periods at Qana is imported and the bulk of it consists of amphorae (Prof. A. Sedov: personal communication). Dressel 2-4 amphorae provide the primary date (1 st-2 nd century A.D.) for the Lower Period which falls within the chronological range of the present study. The Dr.2-4 type comprise the bulk of total amphora yield from the Lower Period (Prof A. Sedov: personal communication). To quote from the preliminary report on the excavation: "During the lower' period storage vessels were dominant in the pottery corpus. These included both imported amphorae (over 56 % of the total pottery recovered) and locally produced, hand-made vessels, so-called zirs (25 % of the finds). Most of the amphorae fragments are of the so-called 'Koan type' (Dressel 2-4)" (Sedov 1992: 116, emphasis mine). According to Sedov (1996:15-21) other types of amphorae excavated from Qana (Lower and Middle periods dated to 1 st-4 th century A.D.) comprise Rhodian amphorae of the 1 st century A.D. (Peacock and Williams Class 9), Dressel IB, Dressel 7-11, North African Tripolitinian type, Africana type, the North African-Gallic type (including the Dressel 30 variant) and amphorae of Palestinian origin. The excavator is unsure about the classification of some amphora fragments (Sedov 1996: Fig. 5:1-2). He compares these to a similar sherd recovered from Socotra Island (Fig. 5). In this study, we attempt to identify handles (grooved with ovoid cross-section) similar to one discovered by Sedov at Socotra Amphorae handles similar to the Socotra sherd are published in the excavation reports of the sites of Maheshwar, Baroda and Nevasa in India All these handles are comparable to the Dressel 12 amphora type (Fig. 5; discussion below). Fragments of Mediterranean amphorae dated to the second half of the 1 st millenium B.C. were excavated at Shabwa (Salles 1996: footnote 13). Excavations conducted at the port-site of Khor Rori (identified with ancient Moscha of the Periplus) have yielded amphorae fragments (Yule and Kervran 1993: 80). From the island of Socotra (ancient Dioskorida of the Periplus) "fragments of the handle of a massive Roman amphora" and "body sherds of large amphora type

26 vessels" were recovered during prospections by a joint Russian-Yemeni team between 1985-87 (Naumkin and Sedov 1993: 601-605). The illustrated handle fragment carries multiple grooves. It may belong to a Dressel 12 type amphora (see Peacock and Williams 1986:113 for parallels, also compare with similar grooved handles excavated from Maheshwar, Baroda and Nevasa in Fig. 5). Mediterranean amphora fragments have been excavated from the coastal site of Hafun (Opone of the Periplus) (Stiles 1992:27-36; 1994:59) A complete Dressel 2-4 amphora found at Hafun is believed to have entered a private collection (Prof J.F. Salles: personal communication). From the island of Masirah, Salles (in Sedov 1996:19-20) discovered two complete amphorae of Palestinian origin similar to those recovered from the Middle and Upper periods at Qana. In the Persian Gulf littoral Mediterranean amphorae have been recovered from the sites of Thaj (Potts 1990:44), Mleiha (Potts 1990: 265-271; Salles 1993: 493- 523) and Failaka (Gachet and Salles 1993:73-75). All amphora fragments from these sites are of Rhodian manufacture and dated to the 3 rd-2 nd century B.C. (see below). In India, Mediterranean amphorae have been reported from over 50 sites so far (Fig. 4, see also Appendix-I for bibliographical details). The distribution of amphora find-places in Fig. 4 shows a concentration in the western and southern parts of India. A review of the descriptive amphorae recovered in course of prospection and excavation from various sites in India is undertaken below. The review leaves out, in most cases, the body-sherds which are difficult to identify on the basis of published photographs and drawings of Mediterranean amphorae. In course of perusal of fieldwork reports the author found that a number of unidentified handles published are typical to some amphora-types This review presents a tentative classification of the unidentified amphora handles (Fig. 5). The published catalogues used in the identification are typological chart of amphorae made by Dressel (in Cimino 1994 in Plate XXXVIII), the paper on Roman wine amphorae by Tchernia (1989), the catalogue of Roman amphorae by Peacock and Williams (1986) and studies on the Arikamedu amphorae by Will (1992:151-156, in Begley et al. 1996:318-349) and Slane (1992:204-215). Besides, the review includes observations on amphorae studied by the author in collaboration with experts. Arikamedu (Pondicherry). The most detailed studies to date on amphorae from Arikamedu have been undertaken by Dr. Elizabeth Lyding Will She has published two reports on part of the corpus of amphorae recovered from Arikamedu

19-31 .16 32 * 10 ⑉33 - 39 45 34 35 44 -47 #54 55- 51 52 AMPHORA FINDS IN INDIA 37 1. TAXILA 2 GILUND 3. BALATHAL? 4. MARMI 5. MATHURA 6. SHAMALAJI 7. DEVNIMORI 8. SODANGA 9. DANGWADA 10. UJJAIN 11. MANDVI 12. BET DWARKA 13. DWARKA 14. NAGARA 15. LOTHAL? 16. BARODA 17. VALLABHIPUR 18 PRABHAS PATAN 19. SATHOD 20. JALAT 21. AJABPURA 22 SHRIMALA 23. AMRELI 24. FATEHPUR 25. KOTHU 26. ARLA 27 BEYLI 28 SHEDAVADA 29 SANAND 30 MODHERA 31 JUNAGADH 32. KAMREJ 33. DHATVA 34. PAUNAR 35. ADAM 36. PAUNI 37. TAMLUK? 38. KARNAJA 39. BHOKARDAN 40. SOPPARA 41. NEVASA 42. ELEPHANTA 43. JUNNAR 44. TER 45. KOLHAPUR 46. NAGARJUNAKONDA 47. CHANDRAVALLI 48. DHARANIKOTA 49. BHATKAL ? 50. VASAVASAMUDRAM 51. ARIKAMEDU 52. KARAIKADU 53. ALAGANKULAM 54. KARUR 55. URAIYUR Fig.4.

AMPHORA TYPES RHODIN Nevasa UNIDENTIFIED AMPHORA HANDLES FROM FIELD REPORTS 81 HTT 1.1 Arikamedu DRESSEL 12 1 Socotra 1.1 Baroda 1.2 Maheshwar 1.3 Novasa DRESSEL 21-22 1 Nevasa 1.1 Nasik H Fig. 5. Preliminary identification of amphora handles from India and Socotra V L 1.4 Nevasa 28

29 so far (see Will 1992:151-156; Will in Begley et al. 1996:318-349). Will (in Begley et al. 1996:320) provides a catalogue of 82 amphora pieces from Arikamedu. According to her, this number comprises 42% of the total yield of amphorae from Arikamedu before 1989-90. Most of the amphorae studied come from French excavations at Arikamedu in the early forties. The various amphora types identified by Will are geniune Koan, pseudo-Koan/Dressel 2-4; Knidian; Late Rhodian, Dressel 6, Dressel 7/8; Dressel 20; Dressel 22; Tunisian (Will type 21 c) and Punic (Will type 49-53). Of the double-handled sherds from Arikamedu studied by Wills (1992: 151- 156), nearly half were found to be similar to the pseudo-Koan or Dr.2-4 type. In her recent study, Will (in Begley et al. 1996:318) specifies the number of Dr.2-4 pieces as 16. She also identifies 22 fragments of genuine Koan amphorae Slane (1992: 204- 205) studying the amphorae illustrated by Wheeler et al. (1946:41-45) opines: "There is no doubt that the long-necked examples from Arikamedu are Italian Dr. 2-4 amphoras of the 1 st century A.D." Tchernia (personal communication letter of 25 th June 1992) is of the view that double-handles of the Dr.2-4 type are common at Arikamedu. As Will could study only 5 sherds from Wheeler's excavation (Will in Begley et al. 1996:320) we cannot be sure about the relative propotions of genuine and imitation Koan amphorae recovered from Arikamedu. However, the observations above by experts suggest that both the Koan and pseudo-Koan amphorae formed strong components of the total amphora corpus of Arikamedu. Brahmapuri (Dist. Kolhapur, Maharashtra) From the excavated Brahmapuri pottery, two rim-and- (bifid) handle fragments of amphora were identified by Sankalia (1957: 207-208) According to Sankalia the handles 'resemble the type illustrated by Sir Mortimer Wheeler from Arikamedu (Fig. 10, type 70)...' As this particular type from Arikamedu has the typical features of a Dressel 2-4 type amphora, it follows that the Kolhapur handles most probably also belong to the Dr.2-4 variety Nevasa (Dist. Ahmadnagar, Maharashtra): From Nevasa, a total of 63 amphora sherds were excavated from Pd. IV/V levels of the site (Sankalia et al. 1960: 280-281). Presently only 14 of the original 63 sherds are available for study in the Deccan College, Pune. Out of the 14 sherds, 6 are descriptive and the rest are body sherds. The descriptive sherds comprise 1 rim and double-handle portion (NVS 439, Pl. V.d), 1 body and double-handle portion (NVS 4426, Pl. V.c), 1 doublehandle fragment (NVS 4426 A; Pl.V.a), 1 rib of a broken double-handle (NVS 3342; Pl.V.b), 1 single handle (NVS 2803) and 1 carinated shoulder portion (NVS 1709). Prof. A Tchernia (Centre Camille Julian, Marseilles, France) has examined the first four sherds detailed above. He compares these sherds to the Italian Koan

30 amphorae excavated by Wheeler at Arikamedu (personal communication: letter of 25 th June, 1992). Besides the Koan types, other amphorae types could be recognised from the drawings of the Nevasa amphorae handles published by Sankalia et al. (1960:fig. 145). Two thick grooved handles (Sankalia et al. 1960: fig. 146, type 6,7) are tentatively identified in this study as belonging to Dressel 12 amphorae (Fig. 5, Peacock and Williams 1986:113-114). The point of similarity with Dr. 12 type can be noticed in the grooved, ovoid cross-section of the handles and in the sharp downward curve of the handles from below the rim. Similar grooved handles have been reported from Baroda Pd. II and Maheshwar Pd. VI (Fig. 5; discussion below). Furthermore, a doublehandle fragment (Sankalia et al. 1960: fig. 146, no.2) has the typical 'pointed' handles of the Rhodian amphorae of the 1 st century A.D. (Fig. 5; Peacock and Williams 1986:102-103). A similar handle was also excavated at Arikamedu (Wheeler et al. 1946:Fig. 10:81; Fig. 5). Another handle fragment from Nevasa illustrated in Fig. 5 is described by Sankalia et al. (1960:315) as 'clamp-like' is similar in cross-section (oval) and its near right-angle 'bent' with the Dressel 21-22 type amphorae (Peacock and Williams 1986:96-97). A handle-portion of amphora (NVS 4426, layer 4, Pd.V) not illustrated in the Nevasa report but which the author could photograph in the Deccan College reserve collections (Pl.V.c) seems to be part of either a Rhodian amphora or the type designated Class 11 by Peacock and Williams (1986:107-108) This inference is reached primarily on the basis of the sharp 'upward' curve of the bifid handle. However, it is not possible to say whether the bifid handle conceived a 'pointed' curve below the rim (as in Rhodian type above) or above the rim (as in Class 11 of Peacock and Williams). Nasik (Dist. Nasik, Maharashtra): An unidentified terracotta handle published in the Nasik excavation report (Sankalia and Deo 1955 fig. 49:11, p. 108) can be compared with the Dressel 21-22 amphora (Fig. 5). The Nasik specimen is also similar to an amphora handle identified from Nevasa as Dressel 21-22 type (Fig. 5). Like the Dr.21-22 type illustrated by Peacock and Williams (Fig. 5), both the Nasik and Nevasa handles have a near right-angle curvature. Furthermore, the Dressel 21-22 type is described by Peacock and Williams 1986: 96-97) as having an oval non-bifid handle cross-section (Peacock and Williams 1986: 96-97). Similarly, the Nasik handle is described by the excavators as having a "at the broader end a smooth rectangular surface. " (Sankalia and Deo 1955:108) (Fig. 5).

31 Nagara (Dist. Kaira, Gujarat): Recently, Dr. A. Marangou-Lerat (personal communication: letter of 4 th August 1994), in her review of amphora from the coastal site of has identified a small fragment of a double-handle as probably from Kos or from some other region of the eastern Mediterranean. This handle-fragment therefore is either geniune Koan or psuedo-Koan/Dressel 2-4 (Pl. VI.b). A base-portion of an amphora excavated at the site is tentatively identified here as belonging to a Dressel 1 type amphora or Fabric 2 of Class 6 in the catalogue of Peacock and Williams. (1986:93-95). The base-fragment from Nagara (Pl. VI.d.e) has been identified on the basis of Peacock and Williams' description of Fabric 2/Class 6 having a creamy-white fabric and a solid spike. The shape of the spike also matches the example illustrated by Peacock and Williams (1986:93). The two pink handles of amphora from Nagara illustrated in Pl. V.a.c. could not be classified on the basis of records available to me. Prabhas Patan (Dist. Junagadh, Gujarat) A neck and double handle portion of a Koan type amphora was excavated from the coastal site of Prabhas Patan in Gujarat. The amphora fragment was recently discovered in the Prabhas Patan pottery corpus deposited at the Deccan College, Pune (personal communication with Dr VS Shinde, Deccan College). Baroda (Dist. Baroda, Gujarat): An unidentified grooved handle published in the Baroda excavation report is unmistakably that of an amphora (Subbarao 1953 fig. 15:17). The angle of curve of the Baroda handle and its grooved, ovoid cross-section seems to be typical of the Dressel 12 type (Fig. 5). Maheshwar (Dist. W. Nimar, Madhya Pradesh): An unidentified grooved handle published in the Maheshwar excavation report seems to be part of a Dressel 12 amphora The excavators describe the handle as being of "a coarse red fabric, convex in outline with three deep, parallel and vertical grooves on the outer flat surface of the handle." (Sankalia, Subbarao and Deo 1958: Fig. 103, n.1, p.200). The curvature of the Maheshwar handle also appears to be like the Dressel 12 type (Fig. 5). Pauni (Dist Bhandara, Maharashtra): In the recent excavation at Pauni conducted by the Archaeological Survey of India in 1995-96, a double-handle fragment of an amphora has come to light (Amarendra Nath: personal communication) Mathura (Dist Mathura, Uttar Pradesh). A double-handle of an amphora carrying the stamp of the manufacturer on top of the twin ribs was recovered from Saka-Kushana levels (1 st-3 rd century A.D.) in course of excavations at Ambarish Tila conducted by the Archaeological Survey of India (Joshi and Sinha 1991: 255-259). This sherd has been examined by Will (in Joshi and Sinha 1991: 255-256, personal

32 communication) who sources the stamp to ateliers near Pompei and relates the form to the Dressel type 3. Gilund (Dist. Udaipur, Rajasthan): While exploring the archaeological mound at Gilund, Dr. V.S. Shinde of the Deccan College found a thick handle fragment of an amphora. The handle, illustrated in Pl. VIII.a has a 'curvature' and cross-section typical of amphora. The shape of the amphora handle approximates to the Dressel 20, 23, 25, 28 types. However, a more precise classification was not possible by me Taxila (north Pakistan): At Taxila, a double-handled portion of an amphora recovered from first centuries B.C./A.D. levels at the Indo-Parthian city of Sirkap by Sir John Marshall can be compared with the Koan type amphorae from Arikamedu (Wheeler et al. 1946: 45). Manikpatna (Dist. Puri, Orissa): A photograph of a body fragment of amphora excavated from the site of Manikpatna on the Orissa coast was provided to me by Dr. D.R. Pradhan. The sherd, illustrated in Pl. VII shows the typical 'tapering' shape of the container. The corrugations in the interior of the body also reveal the pot sherd to be part of an amphora. However precise classification could not be done. Karnaji (Dist. Midnapore, W.Bengal): A photograph of a completely intact amphora discovered at Karnaji village was recently published by Sengupta (1996:128) without specific type identification. A comparison with the typological forms catalogued by Dressel (in Cimino 1994: Pl. XXXVIII) suggests the Karnaji amphora to be a Dressel 28 type. This brief review of Mediterranean amphorae deposited along the Red SeaIndian Ocean littoral reveals distribution of a number of types. The following types of amphora have been identified in the Indian corpus: Genuine Koan, Dressel 1, Imitation Koan or Dressel 2-4, Rhodian amphorae of the 1 st century A.D., Knidian, Spanish Dressel types 7, 8, 20, Istrian Dressel type 6, Dressel 12, Dressel 21-22, Benghazi ER amphora 2 (or Class 11 of Peacock and Williams 1986:107), Dressel 28, Tunisian amphora (Wills type 21 c) and Punic amphora (Wills type 49-53). The substantial quantity of amphorae sherds, especially body sherds, which still have to be examined by experts promises to expand the database and make it precise. Of the amphorae identified, the Dressel 2-4 variety seems to proliferate along the Egypt-Arabia-India route. The high incidence of Dr.2-4 finds on the Egyptian Red Sea coast and the strong Dr.2-4 profiles revealed at Qana, Arikamedu and Nevasa leads us to consider that this type of amphora may have constituted the bulk of Mediterranean wine-amphora exports to the Orient. However, the quantitative profile

33 of Dr. 2-4 occurrence needs to be elucidated for many more sites (especially in India) before we can be conclusive on this issue. Chronology A comprehensive chronological classification of amphorae based on typology is now available (see especially catalogue of Peacock and Williams 1986). Such a classification makes it possible to use descriptive amphorae pieces and in many cases even body-sherds as viable chronological markers'. In this section the implications of the chronology provided by the Dressel 2-4 type the most prolific amphora on the Egypt-India route - is discussed in the light of dating emerging from other associated amphora types and indigenous 'markers' such as Satavahana coins The life-span of the Dressel 2-4 amphora is taken from the end of the 1 st century B.C. to the mid-2 nd century A.D. (Peacock and Williams 1986: 105-106). According to Woolf (1992: 286), the 'Dressel 2-4 was the last amphora to be produced on a Mediterranean-wide scale. Even during the main period of its production, from the end of the last century B.C. to the mid-second century A.D., it co-existed with other types. ' The main areas of Dr.2-4 production were in Italy, Spain, France, England (?) and Alexandria in Egypt (Peacock and Williams 1986:105- 106; Woolf 1992: 285-286, Dr. J.Y. Empereur: personal communication). According to Prof. J.F. Salles (personal communication) the lower date of end of 1 st century B.C. holds true for all Dr.2-4 production-areas. However, the upper chronology varies for Dr. 2-4 sub-types upto mid-2 nd century A.D. A major implication of the Dr.2-4 chronology is that the lower date of this amphora type is in consonance with the date of establishment of a number of key port-sites and settlements involved in Indo-Mediterranean maritime exchange. For instance, on the Egyptian coast the important trade-port of Quseir has revealed Roman foundations dating to the beginning of 1 st century A.D. No stratified evidence of the earlier Ptolemaic period has yet come to light (Sidebotham 1986 a 53-54, Dr.J.Y. Empereur: personal communication). Similarly, excavations at Qana show commencement of activity in the B.C./A.D. changeover. In the words of the excavators: 'It is not impossible that buildings dating to the second half of the 1 st millenium B.C. do exist somewhere in the "lower" city or among the buildings of the 'Urr Mawiyat fortress However, as noted above, we have found no evidence of occupation at Qana which dates to such an early period' (Sedov 1992 110-137). In fact, the Lower Period at Qana has been dated primarily on the basis of the chronology of the Dressel 2-4 amphorae which, as noted above, constitute the majority of the imported ceramics from this period.

34 Besides the Dr. 2-4 type, the dates of other amphorae-types recovered from the Lower and Middle period at Qana have also been applied to attribute a chronology of 1 st-4 th century A.D. for the two periods. The chronology for the Upper Period at Qana has been drawn from amphorae-types dated between 5 th-7 th century A.D. (Sedov 1996:21-23) In contrast to the prolific Dressel 2-4 amphorae found in the Gulf of Aden, in the Persian Gulf littoral the amphorae so far reported are of Rhodian provenance and dated to the 2 nd century B.C. (Potts 1990:44, 265-271; Salles 1993:493-523, Gachet and Salles 1993:73-75). Though Roman glassware and coins of the early centuries A.D. have been recovered from the Persian Gulf area no amphora of the B.C.-A.D. transition has yet come to light in this region. In particular, no Dr.2-4 type amphora has been discovered on Persian Gulf sites (personal communication: Prof. J-F. Salles) Coming to the sites yielding the Dr.2-4 amphora in India, we do know that amphorae occur without break along 8 consecutive layers of Nevasa Pd. IV/V (Sankalia et al. 1960: 65; Fig. 41). Interesting from our point of view is the fact that the lowest and uppermost amphora-yielding layers have also provided "Early" and "Late" coins of the Imperial Satavahanas (Sankalia et al. 1960:65, 176-185; Fig.41). The Early Satavahana coins are found mostly in the lower layers (Sankalia et al: 1960: 171-176). These are largely the issues of King Satavahana whom Shastri (1987: 89-92; forthcoming) has identified as the Satavahana monarch Simuka and placed his regnal years in the last quarter of the 1 st century B.C. The upper amphorayielding layer D 3 provides a coin of of the Satavahana King Gautamiputra Yajna Sri Satakarni (Sankalia et al. 1960: 181-182) whose regnal years have been attributed to the later half of the 2 nd century A.D. Therefore, here we have a situation where the "indigenous" context provides a chronology to the amphora-sequence. The end of 1 st century B.C. - end of 2 nd century A.D. chronology indicated for the amphora-yielding layers at Nevasa on the basis of Satavahana coinage equates to the chronological span of the Dressel 2-4 form. We know that 4 of the available 14 sherds have been identified as Italian Koan type by Tchernia (above). However, the identified Dr. 2-4 sherds are distributed in the middle layers of Nevasa Pd.V and are of no help in dating the termini of the amphora sequence. Among the other available sherds we have 1 body-sherd from the lowest layer H 8 and 2 body-sherds from the upper layers C 2, C 3 The precise identification of these sherds is necessary to corroborate the chronology offered by the Satavahana coins.

35 For Arikamedu, the chronology of initial occupation proposed by Wheeler et al. (1946: 24) was for a beginning at the "end of the first century B.C. or beginning of 1 st century A.D." However, Will's identification of more than half the double-handles from Arikamedu as Koan of the 2 nd century B.C. serves to 'push down' Wheeler's chronology for initial Mediterranean maritime contact with the settlement However, before we accept a revised chronology based on Will's identification, questions relating to the stratigraphic 'positions' of the genuine and imitation Koan amphorae need to be answered. Will does not provide any information on the 'context' of the excavated amphorae. From her account we do not know at which stratigraphic point the genuine Koan sherds give way to the later imitation Koan or Dressel 2-4 amphorae. At least we would expect the genuine Koan double-handles of 2 nd-1 st century B.C. vintage to occur below Wheeler's 'Arretine layers' where the terra sigillata has been firmly dated to the 1 st century A.D. (Wheeler et al. 1946: 34-41; Comfort 1992: 134-150, Prof. D. Gabler, Hungarian Academy of Sciences: personal communication) However we find no amphora double-handles reported by Wheeler in the pre-Arretine phase. Only three sherds are recorded from these levels: 2 rim portions and 1 carinated shoulder portion (Wheeler et al. 1946: 43). None of these amphorae have been studied by Will. The discussion on the chronology of amphorae from Quseir, Qana (Lower Period), Nevasa and Arikamedu points to the evidence of Dressel 2-4 amphorae as crucial to the periodization of Indo-Arabian-Mediterranean contact While focusing on the Dr 2-4 evidence it is equally necessary to place in counterpoint the dates of other amphora-types found in association. In this regard we have already discussed the geniune Koan amphorae identified by Will at Arikamedu Other associated amphora-types recovered from Arikamedu are: 2 stamped handle-pieces of Knidian origin dated to 1 st century B.C. (Will 1992: 153, Slane 1992: 205-206), Rhodian sherds of the 1 st century B.C. 1 st century A.D. and Spanish Dr. 7 & 8 amphorae of the late 1 st century B.C. - early 1 st century A.D. (Will 1992:153), Spanish Dr.20 amphora of the 1 st century B.C - 1 st century A.D. (Will 1992: 154), Istrian Dr.6. amphora fragment dated to later 1 st century B.C. to 3 rd quarter of 1 st century A.D (Will 1992:154); Dr 22 amphorae pieces dated to 1 st century A.D. (Will in Begley et al. 1996:319), fragment of Punic amphora dated between 3 rd-6 th century A.D (Will in Begley et al. 1996 349), and fragment of Tunisian amphora dated to 5 th century A.D (Will in Begley et al. 1996 349)

36 Except for the genuine Koan sherds from Arikamedu placed by Will in the 2 nd century BC, the other amphora-types from the site approximate to dates in the changeover from 1 st century B.C. to 1 st century A.D. The dates of various amphorae-types (apart from Dr.2-4) identified from Nevasa are Rhodian (late 1 st century B.C. - 2 nd century A.D., Peacock and Williams 1986:102-103), Dressel 21-22 (1 st century A.D., see Peacock and Williams 1986:96- 97) and Class 11 of Peacock and Williams (1 st-2 nd century A.D. Peacock and Williams 1986:107-108). Contextually, at Baroda and Maheshwar the Dr. 12 handles are found in levels dated to the early centuries A.D. (Fig. 5). The Dressel 21-22 fragment from Nasik has a 1 st century A.D. date (Fig. 5; Peacock and Williams 1986:96-97). The baseportion from Nagara tentatively identified as Dressel 1/ Fabric 2, Class 6 can be dated from Late Republican to 79 A.D. with the majority of finds in the Augustan Period (Peacock and Williams 1986 93-95). This review of chronology of amphorae deposited in Indian Ocean lands reveals that most of the amphorae forms recovered are those produced in the B.C-A.D. transition years,i.e, late 1 st century B.C to 2 nd century A.D Provenance Recent prospections along the Egyptian Red Sea coast have yielded substantial quantity of Alexandrian and Italian Dressel 2-4 amphorae (Dr.J.Y. Empereur personal communication). In the Gulf of Aden littoral, the amphora evidence from the port-site of Qana has been best studied in terms of provenance Though we are not informed about the quantity of amphorae at Qana which can be traced to various production centres in the Mediterranean, nevertheless the preliminary report on the excavations provides details about three Dressel 2-4 amphorae carrying a Palmyrene, Syriac and Greek inscription each. All three vessel fragments have been attributed an eastern Mediterranean origin (Sedov 1992: 116-121) Furthermore, Prof. A.V Sedov, head of the Russian team at Qana, has indicated that many of the amphorae from the Lower Period at the site are Alexandrian (Egyptian) Dr. 2-4 type The multiple-grooved amphora handle from Socotra, tentatively identified as belonging to the Dressel 12 group on the basis of the classification of Peacock and Williams (1986 113), can be attributed to a Spanish origin with reference to their catalogue In India, the recorded Dr 2-4 amphorae from Mathura, Nevasa and Arikamedu have been attributed to production centres in the western Mediterranean

37 Specifically, the Dressel 3 double-handle from Mathura has been traced by Will (in Joshi and Sinha 1991: 255-259) to the workshops of Campania in central Italy Tchernia (personal communication) believes that the four Koan type double-handles from Nevasa (discussed above) 'come more probably from the western part of the Mediterranean than from Greece. Italy and Spain are possible origins' (extract from letter of 25 th June, 1992). Of the double-handles from Arikamedu, Will (1992: 151) says that about half the collection studied comprises 'Italian imitations of Koan amphoras (Will type 12, Dressel forms 4 and 3...)." There is a likelihood of an amphora sherd from Nagara being a Dr. 2-4 variety from the eastern Mediterranean. A Marangou-Lerat (personal communication) has recently identified a double-handle from Nagara as being either from Kos or some other region of the eastern Mediterranean' (extract from letter of 4 th August 1994) Of the non-Dressel 2-4 varieties found in India we have from Arikamedu the amphorae produced in Knidia, Rhodes, Istria and Spain and north Africa. From Nevasa, we have identified a handle piece of Rhodian origin The Dr. 12 type from Nevasa, Baroda and Maheshwar can be attributed a Spanish origin (Peacock and Williams 1986:113) The Dr 21-22 type from Nevasa and Nasik may have come from Italy (Peacock and Williams 1986:96). The probable Class 11 amphora from Nevasa may come from the Eastern Mediterranean or even North Africa or Italy (Peacock and Williams 1986:107). Will (in Begley et al. 1996:349) has identified amphora fragments of north African origin (Punic and Tunisian) in the Arikamedu corpus