Chapter 14 THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH* THOMAS L. MCCLELLAN The Oriental Institute The University of Chicago PART I: EVIDENCE FOR THE ABSOLUTE CHRONOLOGY OF ALALAKH INTRODUCTION Alalakh is one of those rare sites where historical, textual, archaeological, and art historical data converge to provide opportunities for formulating integrated hypotheses about absolute chronology that have consequences for northern Syria as well as Mesopotamia and Anatolia. In addition to the tablets found in levels VII and IV which provide genealogies of the local dynasts of Yamhad and Alalakh itself, the information contained on the tablets also provides historical evidence relating to monarchs of Hatti, Babylonia, Man, Mitanni, and Egypt. Archaeologically, Bichrome ware, Habur ware, Nuzi ware, Gray ware, Amuq-Cilician Painted ware, and the Cypriote, Mycenaean, and Minoan wares, almost all the distinctive pottery wares in the second millennium B.C. Levant, glyptics and other small objects, have been marshaled to support premises about the dating of strata at Alalakh. These discussions cover the entire second millennium B.C. and in terms of chronology they affect our understanding of its beginning (with Amuq-Cilician pottery at issue) and its end (the significance of Mycenaean IIIB pottery found at Alalakh). But it is the question of the dates of levels VII and IV that has generated the most intense investigations since they are related to the problem of high, middle, and low chronologies of Mesopotamia. New information is emerging from different quarters that threaten (or perhaps one should say promise) to make the Alalakh material obsolete, but at the moment Alalakh still plays an important role in our understanding of chronology. Alalakh has long been a key element in the widely used middle chronology which in a sense was canonized by its adoption in the prestigious Cambridge Ancient History} In recent years re-examination of the * Helene Kantor's analysis of Alalakh chronology (H. J. Kantor, "Syro-Palestinian Ivories," JNES 15 [1956]: 158-60) is still basic to any new consideration. In her honor I am pleased to submit a few more words about Alaiakh. 1. I. E. S. Edwards, C. J. Gadd, and N. G. L. Hammond, eds., CAlP vol. 1:1 (Cambridge, 1970), pp. xix-xx. .1.82 THOMAS L. MCCLELLAN archaeological data from Alalakh has seriously challenged that prevailing view in favor, in some quarters, of the low chronology. How valid is this reinterpretation and docs the archaeological data independently favor one chronological scheme over another? To put it another way, what is the evidence for the dates of levels VII through IV and how precisely can the archaeological material in them be dated, either as evidence used to date the levels or to be firmly dated by their presence in those presumably chronologically fixed levels? HISTORY OF MODERN SCHOLARSHIP (Fig. 26) Destruction of Level VII In reviewing the extensive literature for the dates of the end of level VII and the beginning of level IV there have been several turning points in the discussions on the fixing of these dates. Originally in 1940 Smith, followed by Woolley in the 1950s, fixed the date for the destruction of level VII to ca. 1750 B.C., Smith based his assessment on ceramic and glyptic evidence and on a synchronism of Yarim-Lim of level VII with Hammurapi of Babylon.2 Habur ware, Smith believed, could be dated 1800-1600 B.C., assuming its beginning was fixed at Chagar Bazar, and its end by its stratigraphic relationship with Nuzi ware at Tell Biila, a ware whose date was fixed at Nuzi where it occurs in a context dated to SauStatar. The glyptics of level VII showed Egyptian influences dating to the 12th Dynasty. Smith attributed the destruction of level VII to general unrest related to the Kassite invasion of Babylonia, and he suggested level VI was destroyed by Mursiii I.3 After the publication of the Alalakh tablets in 1953 it was quickly demonstrated by Landsberger that five generations of rulers were mentioned in the tablets, and that Hammurapi I of Yamhad, a contemporary of Hammurapi of Babylon, lived considerably earlier than the destruction of level VII.4 In 1956 Kantor and Albright lowered the date for the destruction of level VII by 100 to 150 years.5 In addition to adopting this new understanding of the historical synchronisms they both correlated the pottery of level VII with the Middle Bronze IIB period in Palestine and specifically with Megiddo strata XII-X, and discussed the seals. For many years these were the most detailed ceramic analyses of level VII apart from Woolley's final report; of the two, Kantor provided the fuller comparisons and also discussed the seals mentioned by Smith: To sum up, the glyptic from Alalakh VII can now be added to the other evidence which supports the view that Canaanite art, as a clearly recognizable and coherent school of craftsmanship, came into being in the final stage of the Middle Bronze period, the phase contemporary with the late First Dynasty of Babylon and the Second Intermediate Period of Egypt. ^ The basis for dating level VII by Albright and Kantor remained the ceramics and glyptics, though Kantor attributed its destruction to Mursiii I, following Landsberger. Albright saw Alalakh level VII as 2. S. Smith, Alalakh and Chronology (London, 1940); C. L. Woolley, A Forgotten Kingdom (Baltimore, 1953); idem Alalakh (London, 1955). 3. Smith, Alalakh and Chronology, pp. 3-4, 13-16, 35-36. 4. D. J. Wiseman, The Alalakh Tablets (London, 1953); B. Landsberger, "Assyrische Königsliste und 'Dunkles Zeitalter,"' JCS 8 (1954): 31-73, 106-133. 5. Kantor, "Syro-Palestinian Ivories"; W. F. Albright, "Stratigraphic Confirmation of the Low Mesopotamian Chronology," BASOR 144 (1956): 26-30. 6. Ibid., p. 160; for a recent study of level VII see A. Kempinski, Syrien und Palästina (Kanaan) in der letzten Phase der Mittelbrome IIB-Zeit (1650-1570 v. Chr.) AAT 4 (Wiesbaden, 1983). Fig. 26. Alalakh: Proposed Dates for Levels. 184 THOMAS L. MCCLELLAN "Stratigraphic Confirmation of the Low Mesopotamian Chronology" as the title of his 1956 article stated, but he was vague about the time between the end of level VII and the beginning of VI. Goetze, who dated the destruction of level VII to 1650 B.C., or within twenty years of Albright, thought "that the two dates proposed, reduced to their real meaning, actually coincide," and found that the Alalakh data fit well with the middle chronology.7 In his discussion of the dating of level VII, the destruction of which he attributed to MurSili I, Goetze appealed to an argument that links its date with the duration between it and level IV; this argument was later taken up by Rowton (see below). In 1957 the discussion about the destruction of level VII took a new turn with the discovery of a tablet at Bqgazkoy which states that Hattusili destroyed Alalakh. The correlation of a general ZukraSi in the level VII tablets with the same who was mentioned in a Hittite document independently dated to the reign of HattuSili I has led most scholars to conclude it was level VII that HattuSili I destroyed.8 However this synchronism did not provide a firm absolute date for the stratum's destruction or HattuSili I's campaign. The end of this first phase of scholarship was marked by two chronological studies of Rowton that linked the date for the destruction of Alalakh level VII with Mesopotamian middle chronology.9 In the first instance he simply adopted Albright's date of 1640 B.C. and Kantor's 1600 B.C., based on their pottery studies. He noted, however, that since the pottery cannot provide a more precise date than ca. 1640-1600 B.C. for the destruction of level VII, it is "compatible with the [Mesopotamian] low chronology, as well as with the higher chronologies"; thus it was not Alalakh VII that persuaded him to favor the middle chronology.10 A few years later he modified his views; Alalakh VII could not have been destroyed later than 1650 B.C. because it would unduly compress the duration of the following strata. Using the low chronology, level VII would have been destroyed around 1575 B.C., but that is too low because "... we have seen that the end of level IV is to be dated not later than 1473 B.C. A total of only one hundred years for the three levels, VI, V, and IV, is very improbable."11 This is one of three arguments he presented against the low chronology, but figure 26 illustrates that Rowton's seemingly fixed date for the destruction of level TV is much higher than the consensus of opinion; we will review this dating of level IV below. On the other hand Rowton argued that the Mesopotamian ultra-high chronology of 1750 B.C. for the destruction of Babylon by MurSili I is ruled out because that would mean HattuSili I, who destroyed Alalakh VII, did it fifty years earlier, or 1800 B.C. But that would mean the ceramics and glyptics of VII would then be contemporary with Egypt's 12th Dynasty, which Kantor and Albright showed to be impossible.12 He 7. A. Goetze, "Alalah and Hittite Chronology," BASOR 146 (1957): p. 25; idem, "On the Chronology of the Second Millennium B.C,"/CS 11 (1957): pp. 63-73; also see W. F, Albright, "Further Observations on the Chronology of Alalakh," BASOR 146 (1957): pp. 26-34. 8. J. D. Muhly, "Near Eastern Chronology and the Date of the Late Cypriote I period," in N. Robertson, ed., The Archaeology of Cyprus: Recent Developments (Park Ridge, New Jersey, 1975), pp. 76-89; N. Na'aman, "Syria at the Transition from the Old Babylonian Period to the Middle Babylonian Period," UF 6 (1974): p. 273; Landsberger, "Assyrische Konigsliste," p. 52. 9. M. B. Rowton, "The Date of Hammurabi," JNES 17 (1958): pp. 97-111; Idem, "Chronology: Ancient Western Asia" in Edwards, Gadd and Hammond, eds., CAN3 vol. 1:1, pp. 193-239, [published in fascicle form in 1962). 10. Rowton, "The Date of Hammurabi," p. 100. 11. Rowton, "Chronology: Ancient Western Asia," p. 232. 12. Ibid., p. 61. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 185 marked the interval between HattuSili I and MurSili I at fifty years, but this is an estimate which he said was supported in part by the dynastic genealogy of the rulers of Yamhad and Alalakh.13 Over a decade passed before the stratification of Alalakh was again discussed in detail, but in the space of two years, 1975-76, studies about Alalakh appeared on glyptics, genealogy, the pottery and artifacts of levels VI-V, and, in part of a much larger work, pottery from the earliest levels XVII-VIII.14 Both Collon and Na'aman presented new evidence and interpretations for genealogies found in the tablets and sealings of level VII. In particular Na'aman proposed the existence of two new figures in the ruling line at Alalakh, using the principle of papponomy. This thesis, he believed, best fits the internal evidence for the tablets and sealings, but also is necessary to span the long time duration of level VII that is known to have existed through correlations with the 1st dynasty of Babylon. Collon, from her study of the sealings, concluded most of them date to the latter part of level VII, or about 1700-1650 B.C., but Na'aman strongly argued that they cover the entire span of level VII which he dated to 1720-1620 B.C.15 In a later work Collon lowered her estimate for the destruction of VII to the "late 17th century."16 Most significantly her decision not to use glyptics as a primary dating tool has effectively removed them from the chronological debate over Alalakh. They both accepted the middle chronology, but whereas Collon referred to Rowton to confirm that view, Na'aman recited arguments against the low chronology. One argument stated that the names found in tablets from levels VII and IV differ greatly, and "thus it is difficult to assume that such a sharp change of personal names at Alalakh took place over a span of only a few decades, and it seems more likely to us that a much longer period was required."17 His other two arguments cited archaeological data. First, the pottery of level VII is Middle Bronze Age II and not later than 1600 B.C., according to the expert opinions of Albright, Kantor, and Kempinski. Secondly, the low chronology would place the destruction of level VII only fifty years before the beginning of level IV and Idrimi, i.e. about 1500 B.C., but that is too short a time for levels VI and V according to the archaeological evidence.18 In accepting the middle chronology, Na'aman assumed MurSili I destroyed Alalakh VII, but he reduced Rowton's estimate of fifty years between events and argued for a date of 1620 B.C. for the destruction of level VII. In marked contrast is the position of Gates who dated the destruction of level VII much lower, to 1575 B.C. on the basis of archaeological data.19 She attacked the problem indirectly by concentrating on the dates for levels VI and V rather than level VII (see below). Levels VI and V span a period from 1575 to 1460 13. Ibid., pp. 43^15. 14. D. Collon, The Seal Impressions fromTellAickanalAlalakh AO AT 27 (Neukirchen-Vluyn, 1975); N. Na'aman, "A New Look at the Chronology of Alalakh Level VII," AnSt 26 (1976): pp. 129-43; M.-H. Gates, Alalakh-Tell Atchana, Levels VI and V: A Re-examination of a Mid-Second Millennium B.C Syrian City, Unpublished Ph. D. dissertation, Yale University, (Ann Arbor, 1976); B. Williams, Archaeological and Historical Problems of the Second Intermediate Period. Unpublished Ph. D. dissertation. University of Chicago (Chicago, 1975). 15. D. Collon, "A New Look at the Chronology of Alalakh Level VII: A Rejoinder," AnSt 27 (1977): pp. 127-31; N. Na'aman, 'The Chronology of Alalakh Level W Once Again," AnSt 29 (1979): pp. 103-113. 16. D. Collon, The Alalakh Cylinder Seals, BAR International Series 132 (Oxford, 1982), p. 3. 17. Na'aman, "A New Look," p. 142. 18. Na'aman, "Syria at the Transition," pp. 265-74; Idem, "A New Look," p. 141. 19. See now M.-H. Gates, "Alalakh Levels VI and V: A Chronological Reassessment," Syro-Mesopotamian Studies 4/2 (1981); idem, "Alalakh and Chronology Again," in P. Astrom, ed., High, Middle or Low? (Gothenburg, 1987), pp. 60-86. 186 THOMAS L, MCCLELLAN B.C., determined primarily on the basis of Cypriote and Palestinian synchronisms. These levels contain numerous Cypriote imports whose absolute dates have been established by Egyptian chronology determined in Palestinian contexts. Especially important to levels VI-V is Bichrome ware that dates from 1575/60 to 1475/60 B.C.20 Furthermore, examination of levels VII and VI shows that there was no gap between them, so level VII was destroyed about 1575 B.C. Gates points out that we can no longer discuss Mesopotamian dates in terms of high, middle, and low chronologies since re-examination of the Venus tables of Ammisaduqa of Babylon show them to be unreliable for those purposes.21 Another recent study of the archaeological data from Alalakh is that of Kempinski from whom Na'aman found archaeological support for the middle chronology.22 Although published in 1983, it is a modified version of his earlier doctoral dissertation and does not take into consideration the works of Collon and Gates or the new study of the Venus tablets. Kempinski is concerned with a rather narrow time span in Palestine: 1650-1570 B.C., the chronology of which he fixes with reference to five points: 1) Alalakh VII was destroyed by HattuSili I around 1620/15 (middle) or 1555/50 (low), 2) Bichrome and Black-Impressed wares are found in Alalakh VI-V, 3) synchronism of Alalakh with Megiddo and Tell el-Ajjul may be established: Alalakh VI, Megiddo X and Ajjul II are coeval, as are Alalakh V, Megiddo IX and Ajjul I, 4) the time span of Ajjul II may be estimated, and 5) the time span of Megiddo IX helps determine the date of stratum X there.23 In terms of absolute dating, the middle chronology must be used because the low chronology, which places the destruction of Alalakh VII to ca. 1550 B.C., unduly compresses levels VI-V into a fifty year period delimited by the Idrimi inscription of level IV.24 Furthermore it would place the destruction of VII one year after the fall of Dynasty 15 in Egypt, and the emergence of Bichrome ware, the floruit of Black-Impressed ware, and the "Hyksos" scarabs would fall in that late, short fifty year period 1550-1500 B.C., which he says is impossible. The discussion of the archaeological assemblages of Alalakh by Kempinski is detailed and needs careful consideration. Innovative aspects of it are his equation of Black-Impressed ware with Tell el-Yahudiyeh ware and his analysis of "Hyksos" design motifs on scarabs and pottery of levels VI-V.25 But it is the broad design of Kempinski's argument that is most surprising. In his effort to date the Palestinian Middle Bronze Age IIB period he turned to what he considered to be the well-dated strata at Alalakh, levels VI-V. In fact this is a reversal of the line of argument by Albright, Kantor, Rowton, and Gates who seek dating support for those Alalakh levels from synchronisms with the well-dated Palestinian sequence. In other words we have a classic tautology.26 20. Gates, Alalakh Levels, p. 22. 21. Ibid., p. 37; E. Reiner and D. Pingree, The Venus Tablet of Ammisaduqa, Bibliotheca Mesopotamia 2/1 (Malibu, 1975), p. 23. 22. Kempinski, Syrien und Palästina. 23. Ibid., pp. 79-80. 24. Ibid., pp. 220-21. 25. See Gates, "Alalakh and Chronology Again," pp. 79-80. 26. I have not considered the dating of levels XVII to Vffl, but the extremely low dates suggested in Williams, Archaeological and Historical Problems, for the latest of these levels affect the date of level VII. It was previously shown that the earliest stratum, level XVII, could not predate the end of Amuq J, i.e. circa 2000 B.C. due to the presence of Amuq-Cilician Painted ware through these earliest Alalakh levels, and the absence of earlier Amuq material; see M. J. Mellink, Review at Alalakh, by C. L. Woolley, AJA 61 (1957): pp. 395-400. It is generally THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 187 DATING OF LEVELS VI AND V Until the studies by Gates the dating of Levels VI and V was usually an afterthought to the issues of levels VII and IV. Once those crucial levels were pinpointed, then VI and V were sandwiched in between them. The early exception was Kantor's detailed list of ceramic similarities between Alalakh VI-V and' Megiddo IX.27 Gates argued that in fact VI-V provide better chronological evidence than does level VII: However, that date [the destruction of VH] cannot be given precisely by any archaeological features of Level VH, whose pottery is typically MB II B-C in Syro-Palestinian terms, without any eccentricities, or — more significantly — without any imported wares which might better define it. In contrast, Level VI ... is the earliest level at Alalakh with imported wares which can be used as reliable dating indices.28 Specifically it is the imported Cypriote, "Syrian," and Palestinian pottery that is most useful for dating, while she found Aegean and Mesopotamian imports less so 29 Among the Cypriote pottery at Alalakh it is the Bichrome ware that is the best evidence: "... the Bichrome Ware examples provide the most specific chronological reference: by themselves they date Alalakh VI-V."30 The other Cypriote wares of chronological value are Monochrome, White Slip I and II, Red-on-Black, and Base Ring.31 In addition Gates assumed that the small sample sizes of these strata and the lack of stratigraphic control of material recovered below the modern water level means it is not possible or at least not wise to search for/distinguish chronological or typological differences in the assemblages. Nevertheless, Williams classified the levels into different periods of the Middle Bronze Age and proposed dates for them: Alalakh Period _ 2000 B.C. xvn XVI MB I 1800 XV XIV MBUA 1725 xm XII XI X 1650 K vra MBIHA 1600 VII MRTTTR _ 1550/1540 The rationale for Williams' dates is not entirely clear to me; at one point (Williams, Archaeological and Historical Problems, p. 1162) he stated that the dating of Alalakh VTI is determined by correlations with Egyptian chronology, but he did not state what those correlations are. Elsewhere, he provided a fuller discussion of the pottery from strata XVII-VUI, B. Williams and R. Hassert, "Some Aspects of the Excavation at Tell Atchana, Part I: A Critical Review of the Pottery from Levels XVII-VTI," Serapis 4 (1978): pp. 41-56. 27. Kantor, "Syro-Palestinian Ivories," p. 159. 28. Gates, Alalakh-Tell Atchana, p. 3. 29. Ibid., pp. 17-27. 30. Ibid., p. 19. 31. Ibid., pp. 18,22. las THOMAS L. MCCLELLAN found glyptics, mold-made figurines, glass vessels, and glazed earthen ware to be useful for dating; yet it is clear they are secondary next to the Cypriote wares and Bichrome ware.32 How can the date of Bichrome ware be so precisely established? In the main, by its presence in Megiddo Stratum IX, whose date is chronologically fixed by its destruction at the hands of Tuthmose III in 1468/1467 B.C..33 DATE OF LEVEL IV Construction of the palace in stratum IV was attributed to Niqmepa by Woolley because it was originally believed that the earliest and greatest number of tablets found in the palace were from his reign/4 but that is at issue, and it depends on his relationship to Idrimi. In turn the dating of Idrimi presents its own set of problems. Line 43 of the inscription on the statue of Idrimi provides a synchronism with a Mitannian king. Originally Smith read the name as Bara *Sutarna, which has subsequently been read as Barrattarna. A further problem is whether Idrimi was the grandson of Niqmepa or his father. Again, following Landsberger, the consensus today is that Idrimi was the father of Niqmepa.35 The date of Idrimi has been linked to the campaign of Thutmose III to the Euphrates. Rowton pointed out that there should be reference to the Egyptian presence in the Idrimi inscription if he reigned during or after the raid. Therefore Idrimi must have reigned before Thutmose Ill's campaign, which Rowton dated to 1473 B.C.36 On the other hand, Oiler, whose assessment of the dating of Idrimi and of level IV is the most detailed study of the past decade, suggests that the campaign of Thutmose III could have occurred after the Idrimi inscription was composed.37 In any event the impact of the Egyptians at Alalakh and in northern Syria may not have been as great as has been suggested.38 While most scholars would date Idrimi to about 1500 (± 50) B.C., Oiler believed the evidence allowed for a more precise date in the lower part of that range — i.e. ca. 1475 B.C., on the basis of four points: Alalakh texts, Hittite evidence, Nuzi evidence, and archaeological evidence from level IV. In the Alalakh texts the genealogical data shows the line of descent to be Ilimilimma, Idrimi, Niqmepa, and Ilimilimma. Furthermore, there are from three to six tablets from the level IV archives which belong to Idrimi, contrary to the original assessments of Smith and Woolley.39 From these tablets Oiler 32. Ibid., pp. 27-28. 33. Ibid., p. 21, but see below. 34. Woolley, A Forgotten Kingdom, p. 101; idem, Alalakh, pp. 110-111, 392. 35. S. Smith, The Statue of Idri-Mi (London, 1949); Woolley, Alalakh, p. 392; Landsberger, "Assyrische Konigsliste," pp. 53-55. 36. Rowton, "Chronology: Ancient Western Asia," pp. 229-30. 37. G. H. Oiler, The Autobiography of Idrimi: A New Text Edition with Philological and Historical Commentary, Unpublished Ph. D. Dissertation, University of Pennsylvania (Ann Arbor, 1977), p. 169. He dated the campaign of Thutmose m to 1457 B.C., following the accession date of 1490 for Thutmose III instead of 1504 B.C. An ingenious suggestion has been made that the statue and inscription were actually crafted much later by nationalists tired of Hittite control. Thus the historical value of the inscription would be diminished, J. M. Sasson, "On Idrimi and Sarruwa, the Scribe," in M. A. Morrison and D. I. Owen, Studies on the Civilization and Culture of Nuzi and the Hurrians (Winona Lake, Indiana, 1981), pp. 309-24. 38. Ibid., pp. 169-73. 39. Ibid., pp. 150-52; Collon, The Seal Impressions, pp. 160-70. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 189 concluded that Idrimi reigned in level V and possibly for a short period in level IV, which means that the palace of level TV might have been constructed by Idrimi, not Niqmepa.40 Using Hittite and Nuzi data, Oiler adduced two other arguments for the dating of Idrimi to ca. 1475 B.C. Through the treaty between Idrimi and Pilliya, known from an Alalakh tablet, and through a Hittite treaty with a Pilliya, king of Kizzuwatna, and the Hittite monarch Zidanta, Idrimi may be synchronized with the Hittite ruler and given a date of ca. 1475 B.C. That is only one of several possible dates for the Hittite ruler, but it is most acceptable, it is argued, because Kizzuwatna did not exist before ca. 1500 B.C.41 Idrimi was contemporary with Barrattarna and Niqmepa with SauStatar. At Nuzi Barrattania may be synchronized with Tehip-tilla, and SauStatar with Winnerke, Tehip-tilla's mother. Winnerke and Tet)ip-tilla belong to the first and second generation of a family that ended in its fifth generation with the destruction of Nuzi, which may be dated to ca. 1350 B.C. Assuming a generation spanning twenty-five to thirty years, then Winnerke, and consequently Barrattarna and Idrimi, date to about 1500-1475 B.C.42 Unfortunately there is some uncertainty over the date for the destruction of Nuzi. According to Oiler, archaeological evidence from level IV provides a chronological range for that stratum and in turn for Idrimi whose reign partially overlaps with it. The absolute dates for level IV may be bracketed by a terminus post quern and a terminus ante quern. A Late Helladic IIIA sherd and Base Ring II sherds found in level IV provide a terminus post quern of no earlier than 1400 B.C., following the assessment of Merrillees and Collon.43 The terminus ante quern is derived from Gates' dating of level V, which ends around 1475/1460 B.C. It may be observed at this point that in accepting the possibility that Idrimi was the builder of the level IV palace we are faced with yet another point of ambiguity in the dating of that level. A more serious aspect is that while Oiler's arguments are quite acceptable within the context of his discussion, his arguments are not acceptable to us because they border on tautologies from the archaeological point of view. A terminus ante quern for level IV based on presumed fixed dates for levels VI and V cannot be used by those who would be concerned about the floating levels VI and V being compressed against the presumably fixed date of level IV. Finally we cannot use the archaeological data to support the date of Idrimi because in most of the discussions Alalakh IV is dated by the presumably fixed chronology of Idrimi. SUMMARY AND DISCUSSION Throughout the discussions three main archaeological arguments have been used to date Alalakh level VII: 1) lowering the date of level VII too much will unduly compress levels VI and V against level IV, the date of which has been established, 2) level VII is contemporary with Megiddo strata XII-X on the basis of ceramic parallels (and therefore is fully Middle Bronze Age II in characteristics and date), 3) the imported Palestinian and Cypriote pottery in Alalakh VI and V and consequently those strata themselves can be dated from the mid sixteenth to the mid fifteenth centuries B.C. (1550-1450 B.C. — Albright, 1575/60-1475/60 B.C. — Gates). 40. Oiler, The Autobiography, pp. 152, 158-59. 41. Ibid., pp. 162-65. 42. Ibid., pp. 165-66. 43. Collon, The Seat Impressions, 169. 190 THOMAS L. MCCLELLAN 1. Compression of Levels VI and V against Level IV This argument is based on two assumptions: a) the date of level IV may be fixed, and b) the layers between VII and IV, i.e. levels Via, VIb, Va, and Vb, are too numerous and too substantial to have existed in a very short period of time. Neither argument is compelling. First, we do not have precise absolute dates for the reign of Idrimi or level IV (see above). Second, one of the most difficult and dangerous procedures is to estimate the absolute duration of an archaeological deposit by the number or thickness of its layers or by the grandeur or flimsiness of architectural elements. Thick debris layers can accumulate in a very short period, as can substantial rebuildings. By the same token, the reverse is true; thin layers and flimsy architectural remains may span long time periods. In short, archaeological strata in themselves are not reliable indicators of chronological duration. 2. Synchronisms of Alalakh with Megiddo (fig. 27) The comparison of the Alalakh and Megiddo assemblages is the foundation for the archaeological dating of Alalakh, but the foundation is shaky for three reasons: 1) the problem of distance, 2) the problem with the stratigraphy of Megiddo, and 3) uncertainty of the absolute date for Megiddo stratum IX. To be sure the general procedures follow accepted archaeological assumptions — the greater the similarities, the closer the assemblages are in date. But the distance between the two sites is so great that geo-cultural differences are substantial. Gates attempted to overcome this problem by concentrating on comparison of imports that come from Cyprus, a third source common to both sites. Despite the geo-cultural differences, there has been genera! agreement about the synchronisms of strata at Alalakh and Megiddo, at least until recently (fig. 27). But how valuable are these correlations? As long ago as 1969 Kenyon pointed out major problems with the Megiddo sequence, especially in the correlation of Areas AA and BB within Megiddo.44 Most recently this has been reinforced by Gonen who suggests stratum IX in Area AA should be equated with stratum X in area BB.45 In her effort to sort out the stratigraphy Kenyon discussed the Middle Bronze Age and Late Bronze Age sequence in terms of selected loci called phases H to S and AA to AH, and pottery groups A to H. In other words, the Megiddo strata no longer have the value and meaning we once thought they possessed. If we are to relate Alalakh to Megiddo, it should be correlated with specific pottery groups and/or phases established by Kenyon, Müller, or Gonen.46 More than ever this is necessary since the previous consensus of Kantor and Albright for the Megiddo-Alalakh correlation (fig. 27) has broken down. Now Gates and Kempinski correlate Alalakh level VI to Megiddo stratum X, while Dever compares level VII to Megiddo X. 44. K. M. Kenyon, "The Middle and Late Bronze Age Strata at Megiddo," Levant 1 (1969): pp. 25-60; also see U. Müller, "Kritische Bemerkungen zu den Straten XIII bis IX in Megiddo," ZDPV 86 (1970): pp. 50-86; H. E. Kassis, "The Beginning of the Late Bronze Age at Megiddo: A Re-examination of Stratum X," Berytus 22 (1973): pp. 5-22. 45. R. Gonen, "Megiddo in the Late Bronze Age-Another Reassessment," Levant 19 (1987): pp. 83-100. 46. As, for example, is done in the chart by W. G. Dever, "Relations between Syria-Palestine and Egypt in the 'Hyksos' Period," in J. N. Tubb, ed., Palestine in the Bronze and Iron Ages (London, 1985): fig. 1. However Gonen, "Megiddo," p, 84, notes serious problems with Kenyon's study; Gonen's analysis should be more reliable. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 191 MEGIDDO ALALAKH Kenyon Kantor Albright Epstein Gates Kempinski Dever R XIII ...........1....... XII lili XIII D ■ XI •X'XiXiivdX'X" ■X'XiXv'Xk* VIII iüll Al sllli E •i'C'X'i'X*:":*:' VI F X VI llllll G H IX VI v VI V VI V V V VI VIM Fig. 27. Megiddo and Alalakh: Proposed Synchronisms Between Levels. Closely related is the question of absolute dates for Mcgiddo and for the Palestinian sequence in general. It is often assumed that stratum IX at Megiddo was destroyed in the famous campaign of Tuthmose III (1468/67 B.C.).47 This assumption is crucial to Gates' argument48 but it is doubtful, according to Gonen: However, now that it has been convincingly pointed out that the Egyptian king never boasted of destroying Megiddo (Shea 1979, 4-5), the siege need not be tied to a destruction level. It could well have taken place during the lifetime of layer VIQ. 49 But not only the absolute dates for Megiddo are questioned; the chronology for the entire Palestinian Middle Bronze Age II sequence is currently in dispute. Dcver provides a chronological scheme that is in the 47. Gates, "Alalakh and Chronology," p. 65; but see now W. Helck, "Was kann die Ägyptologie wirklisch zum Problem der absoluten Chronologie in der Bronzezeit beitragen?" in P. Ästrom, ed.. High, Middle or Low? (Gothenburg, 1987), p. 26; K. A. Kitchen, "The Basics of Egyptian Chronology in Relation to the Bronze Age," in P. Ästrom, ed., High, Middle or Low? (Gothenburg, 1987), pp. 40-41. 48. Gates, Alalakh Levels, p. 21; idem, "Alalakh and Chronology," p. 65. Gales is now aware of this serious challenge to the date of stratum IX — personal communications. 49. Gonen, "Megiddo," p. 97. 192 THOMAS L. MCCLELLAN mainstream of Palestinian archaeology, for example see Kenyon's absolute dates and Cole's relative chronology.50 But this scheme has been challenged by Bitak who wishes to lower the absolute dates for the Palestinian sequence on the basis of his interpretation of the Tell Da'ba stratification in the eastern Delta, and other Egyptian data.51 One important dating criterion is the scarab, and here too there is strong disagreement as to absolute dates.52 While these new debates do not necessarily mean dates used in the past in comparing Alalakh with Middle Bronze Age Palestine are incorrect, they must be used with increased caution. 3. Cypriote Pottery It is assumed that the distribution of imported Cypriote ware at Alalakh will manifest the same chronological range as at other sites. Of these the best dated is thought to be Bichrome ware, and it is Megiddo, especially stratum IX, that provides its date. We have already seen that the absolute date for Megiddo IX has broken away from its fixed mooring. The occurrence of Bichrome ware at Megiddo has been subject to several studies that locate it primarily in stratum IX.53 At Alalakh Bichrome ware is found in levels VI and V, but at most there are no more than twelve sherds54 In contrast, larger quantities of other Cypriote wares were found. I suggest it may be better to evaluate the whole range of Cypriote wares and their general pattern of distribution, rather than relying heavily on Bichrome ware which, because of its scarcity, may provide a misleading sense of its date. The chronological distribution in northern Syria may differ somewhat from that of Palestine; such is the case for Nuzi ware at Alalakh when compared to northern Mesopotamia. 50. Dever, "Relations between Syria-Palestine," fig. 1; cf. idem, "The Beginning of the Middle Bronze Age in Syria-Palestine," in F. M. Cross, W. E. Lemke, and P. D. Miller, Jr., eds., Magnaiia Dei: The Mighty Acts of God (Garden City, New York, 1976); K. M. Kenyon, "Palestine in the Middle Bronze Age," in I. E. S. Edwards, et al., CAHZ vol. 11:1 (Cambridge, 1973), p. 94; D. P. Cole, Shechem I: The Middle Bronze 1IB Pottery (Winona Lake, Indiana, 1984), p. 84. 51. M. Bietak, "Problems of Middle Bronze Age Chronology: New Evidence from Egypt," AJA 88 (1984): pp. 471- 85. 52. D. O'Connor, "The Chronology of Scarabs of the Middle Kingdom and the Second Intermediate Period," JSSEA 15 (1987): pp. 1-41; W. Ward, "Scarab Typology and Archaeological Context," AJA 91 (1987): pp. 507-32. 53. C. M. Epstein, Palestinian Bichrome Ware (Leiden, 1966); M. Artzy, I. Pcrlman, and F. Asaro, "Imported and Local Bichrome Ware in Megiddo," Levant 10 (1978): pp. 99-111; B. Wood, "The Stratigraphic Relationship of Local and Imported Bichrome Ware at Megiddo," Levant 14 (1982): pp. 73-79. 54. Epstein, Palestinian Bichrome Ware, pp. 134-37; Gates, Alalakh Levels, p. 19. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 193 PART II: ANALYSIS OF ALALAKH POTTERY ASSEMBLAGES THE PROBLEM The comparison of the strata from Alalakh with other Near Eastern sites continues to be hindered by our incomplete understanding of the Alalakh material itself, a problem that arises from the way it was originally published. Its pottery is illustrated in a series of 168 types for seventeen levels, which makes it difficult to recognize differences in assemblages from one level to the next. To understand the characteristics of the assemblages one consults distribution tables, but the meaning of the tables' figures are obscured by the unequal size of the assemblages; for example, there are 617 pots from level IV and only two from level XI and four from level XV.55 Drawing inferences from the raw numbers without standardizing the assemblages is misleading. Compounding the issue is the fact that the Cypriote and Mycenaean pottery are not listed in the tables. The classification system is also problematic. Often three or four variants of a type are illustrated, for example type 23a-e, but the shapes of the variants may be significandy different, as in the case of types 4a and 4b, which may lead to a sense of unease about the reliability of Woolley's typology.56 Sometimes a single drawing represents scores of specimens; eighty-two specimens for type 3b were found in level IV alone, but there is only one illustration. Thus our understanding is vague of how representative the illustration is of the constituted type and of the range of variation within the type. Gates' study of levels V and VI, in which she illustrates many new drawings of pots and sherds, goes far in helping clarify the nature of those assemblages and their relevant types.57 In the following section I wish to present another approach which is to analyze the type distributions through frequency curves, and the assemblages of various levels by frequency curves and multivariate statistics.58 DATA PREPARATION Using Woolley's distribution tables quantitatively is something of a risk, as was noted above, for the published types may be so poorly defined as to render them unreliable, and the counts of specimens may present problems.59 Nevertheless it is worth the effort to experiment with them and, as we will see, meaningful patterns are produced. However, it must be kept in mind that there are likely to be unavoidable distortions in the results that can only be recognized and rectified by fresh fieldwork. Throughout it is assumed here that Woolley's distribution tables record only whole, restorable, and partially complete vessels. Except for level VI, Woolley divided levels I to X into a and b sublevels, but usually the bulk of the pottery from a level is from only one sublevel. To establish data sets of pottery for the levels, the choice was faced either to omit the smaller sublevel or group it with the larger sample. Unfortunately, the Cypriote and 55. Woolley, Alalakh, pp. 332-40. 56. Ibid., pis. 109-10. 57. Gates, Alalakh-Tell Atchana, idem, Alalakh Levels. 58. For another example see T. L. McClellan, "Chronology of the 'Philistine' Burials at Tell el-Far'ah (South)," JFA 6 (1979): 57-73. 59. Cf. Williams and Hassert, Some Aspects. 194 THOMAS L. MCCLELLAN Mycenaean pottery is listed only by level, not by sublevel.60 Consequently, the decision was made to group sublevcls, and some levels, together to constitute the data sets used. In our figures levels are labeled with roman numerals except for the level 10 assemblage which represents grouped Alalakh levels XI to XVII that separately have very small assemblages: XI-(2 pots), XII-(31), XIII-(ll), XIV-(27), ,XV-(4), XVI-(O), XVII-(8). Alalakh level X was omitted. Woolley's 168 pottery types were reduced to eighty types, which will be referred to as variables, by eliminating some infrequent types and by grouping many of the types on the basis of shape similarities (see the variable-pottery type conversion lists, fig. 28a, b).61 This was done for ease of handling on the computer, but it also may be more realistic to use broader types when there is such uncertainty about Woolley's typology. Variable v81 is a miscellaneous category that lumps all the unused types (a total of eighty-seven types and sub-types); despite this large number of omitted types, together they generally represent less than ten percent of the vessels Woolley listed in each assemblage, except for levels 2 (10.06%) and 10 (13.25%). In the multivariate analyses the miscellaneous vessels (v81) were omitted and percentages recalculated (not illustrated). The other eighty variables are based on shape except for Red Lustrous ware and the imported Cypriote and Mycenaean wares (i.e., variables v24-v30).62 Figure 29 lists the percentage presence of each variable in the assemblages. Except for the total number of vessels in each assemblage (end of fig. 29), the count of vessels in each assemblage is not listed since they are found in Woolley's tables. DATA ANALYSIS Multivariate statistics were used to search for patterns in the relationship of the levels (cases) and the pottery (variables) from Alalakh. The distribution of variables (pottery types) in the levels is used to establish measures of similarity and correlation between the levels, and by inverting the data matrix the similarity or correlation of variables (pottery types) is determined by their patterns of distribution in find spots (levels). Two measures of association were used: the Pearson correlation coefficient and Euclidean distance. In the program with the Pearson correlation coefficient the data is standardized so that variables with large numbers of vessels or large percentages of vessels and those with small numbers and percentages have equal weight or influence in determining the strength of the correlations. When the Euclidean distance is used on a matrix of percentages of types, it gives more weight, in determining measures of similarity, to those pottery types (variables) that constitute high percentages of individual assemblages. 60. Woolley, Alalakh, pp. 354-76. 61. Hereafter the term types associated with numbers refers to Woolley's typology; variables with numbers have the prefix v. The term variable is widely used to describe the elements or attributes that something (a city, a flower, a pot, a stratum) possesses that makes it similar to or different from another thing; the degree to which these variables are present or absent in something are used to measure the similarity between one thing and another. In computer programs, variables are usually organized in columns and the thing being measured, cities or flowers, are listed in the rows, and are usually called the cases. A matrix is the listing of rows and columns (cases and variables); to invert a matrix is to to make the rows the columns and the columns the rows, or to make the cases variables, and the variables the cases. 62. The Alalakh data sets were analyzed on a Macintosh Plus with a 20 megabyte hard disk and the software package Systat: L. Wilkinson, SYSTAT: The System for Statistics (Evanston, Illinois, 1988). THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 195 ALALAKH Variable Type 01 002i 01 002b Dl 0OÍJ 01 005 01 006i 01 006b 02 003b o: 004b 0! 02li 04 0251 04 025) 04 023» 04 023d 04 119 OS 043 OS 044i 05 044b 05 044c 06 0481 06 046b 06 046c 07 055i 07 055b 06 060i 08 060b 09 0611 06 062i 06 062b 09 063i 09 063b 09 063c 09 069i 09 069b 10 070 II 032 12 033 12 084i ALALAKH Variable 002i 01 002b 01 OOJi 01 003b 02 004b 02 005 01 0061 01 006b 01 0071 31 007b 31 006 47 009l 48 00% 49 009c 49 Oil 32 01! 30 015 33 021 a OS 021c 51 022 34 0231 04 023b 04 023c 04 0i5o 04 024 32 0261 35 029 33 032 36 040 34 0411 54 041b 34 041c 54 0421 37 042b 37 043 OS 044i 05 044b 05 ALALAKH Variable Type 12 064b 12 0840 12 085 13 091i 13 091b 13 093b 13 0930 13 0941 13 094b 14 103» 14 103b I4 1041 I4 104b 15 106b 16 1131 16 1130 16 113c 17 119« 18 1181 18 118b 19 137 20 153 21 158 22 163b 23 1671 24 Monochrome 23 Base Ring 1 26 Base Hing II 27 Red Luslrous 28 White Sip 1 29 While Sip II 30 Mycenaean 31 0071 31 007b 32 Oil 33 015 34 022 ALALAKH Typo Variable 044c 05 045 53 046 35 047 35 0481 06 04Bb 06 048c 06 049 56 050 56 051 56 032 56 0S3 36 0341 36 034b 56 0331 033b 07 07 056 57 057l 57 037b 058 37 37 060i 08 060b 08 0611 03 062i 08 062b 03 063 38 0641 38 064b 38 067l 39 067b 39 O681 09 068b 09 068e 09 0691 09 069b 070 09 10 072b 58 ALALAKH Variable Type 33 0261 36 032 37 0421 37 042b 38 063 38 0641 38 064b 39 067i 39 06Tb 40 0931 40 093b 41 097i 41 09Tb 42 099i 42 099b 42 099c 43 too 44 102i 44 102b 45 120 46 132i 47 008 48 009i 49 009b 49 009o 50 013 31 0210 52 024 S3 029 54 040 54 0411 34 041b 54 04lo 55 045 33 046 55 047 56 049 a ALALAKH Type Variable 07Í1 59 073b 59 073 60 082 11 083 12 0841 12 034b 12 084c 12 085 12 086 61 087 61 038 61 039 61 09Ii 13 091b 13 093b 13 093o 13 094i 13 094b 13 0951 40 095b 40 097i 41 097b 41 098 62 0991 42 099b 42 099c 42 100 43 I02i 44 102b 44 I03i 14 103b 14 1041 14 104b 14 106b 15 1071 63 107b 63 b ALALAKH Variable Type 56 050 56 051 56 032 56 033 56 0541 56 054b 37 036 37 0371 57 057b 37 038 38 072b 39 0731 39 073b 60 075 (1 086 61 087 61 088 61 089 62 093 63 1071 63 107b 64 I08i 64 108b 64 108c 63 110 66 111 67 117i 68 121 68 122 68 1231 63 123b 68 124 63 125 63 126 69 127 70 128 71 1291 ALALAKH Type Variable 108a 64 SOS» 64 1030 64 tio 63 111 66 1131 16 113b 16 113c 16 115. 17 1171 67 1131 18 118b 18 119 04 120 45 121 68 122 68 1231 68 123b 68 124 68 125 63 126 68 127 69 128 70 1291 71 1296 72 129b 72 I30i 72 1301, 72 130s 72 131a 73 1321 74 136 73 137 19 139 76 146 77 147 77 149 78 ALALAKH Variable Type 72 129b 72 1290 72 1301 72 130b 72 130c 73 1311 74 1321 75 136 76 139 77 146 77 147 78 149 79 159 79 160 80 165 ALALAKH Type Variable t32l 46 153 20 153 21 139 79 160 79 163b 22 163 30 I67l 23 BasB Ring 1 23 Base Ring II 26 Monochrome 24 Mycenaean 30 Red Luslrous 27 While Slip 1 28 While Slip 11 29 Fig. 28. Alalakh: Conversion Lists (a) Sorted by Variables and (b) Sorted by Type. 196 THOMAS L. MCCLELLAN THE ALALAKH LEVELS Two cluster analyses and a factor analysis were run on the data. For the cluster analyses two dendrograms (fig. 30a, b) represent two runs to produce hierarchial clusters using single linkage method. In the first dendrogram (fig. 30a) the distance measure was Euclidean. It shows that levels VI and VII are most closely clustered, then levels V, VI, and VII, then levels IV, V, VI, and VII, and another cluster of levels I and II. Level III does not cluster closely with levels I to VII and levels VIII, IX, and 10 are increasingly farther away, in terms of Euclidean distance, from other levels. In the second dendrogram (fig. 30b) the Pearson correlation coefficient was used. Levels VIH and IX are the most closely clustered, then levels V and VII, etc. Overall there are two large clusters: 1) levels I-III and 2) levels IV-10. But in that second cluster, level IV joins it at some distance from the main group. These differences between the two clustering schemes can be explained in part by their different measures of distance. In the dendrogram on figure 30a levels IX and 10 stand in isolation because they have a few variables (types) that represent very high percentages of their assemblages, and the Euclidean distance measure reflects this aspect. In the dendrogram on figure 30b the clustering reflects the standardized weight of types regardless of the percentage of an assemblage they represent. Factor analysis (fig. 30c) was run on the same data set using the Pearson correlation coefficient.63 Levels V, VI, and VII load highly on factor 1; VIII, IX and 10 on factor 2; and I, II, and III on factor 3. Level IV loads only moderately on factors 1 and 3. The results of the factor analysis correspond more closely to the dendrogram on figure 30b which not surprisingly uses the same Pearson correlation coefficient. The analyses using the Pearson correlation coefficient indicate three distinct groups: one — levels VIII, IX, and 10; two — levels V, VI and VII; and three — levels I, II and III. Conversely there is a break between levels VII and VIII and another between levels III and V. Recognition of three groups in these analyses confirms a tripartite division made earlier by the author through visual inspection of selected frequency curves (see below and fig. 38) in which breaks between levels VII and VIII and levels III and IV were observed. What was not recognized in the visual inspection was the intermediate position of level IV which in the multivariate analyses does not group closely with either the levels before or after it. It is assumed that time is the main influence causing the levels to cluster or to load as they do. Thus we can speak of the following periods: Period 1: levels 10 (=XVII-XI), IX, and VIII; Period 2: levels VII, VI, and V; Transitional Period 3: level IV; and Period 4: levels III, II, and I. 63. Rotation was by varimax method; loadings on the factors were sorted. A loading of less than ±0.5 is the rule of thumb cutoff point for assigning a variable to a factor. For discussions of factor analysis see R. J. Rummel, Applied Factor Analysis (Evanston, Illinois, 1970); J. Kim and C. W. Mueller, Factor Analysis: Statistical Methods and Practical Issues (Beverly Hills, 1978); and M. Norusis, SPSS/PC+: Advanced Statistics (Chicago, 1986), pp. B40-B69. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 197 THE ALALAKH POTTERY TYPES The pottery from Alalakh may be analyzed using factor analysis, cluster analysis, and frequency curves. Factor and cluster analyses, utilizing the Pearson correlation coefficient, produce similar results. In factor analysis 9 factors were extracted (fig. 31).64 To determine whether the factor loadings were caused by chronological differences, and if so what those distinctions are, the percentages of pottery variables assigned to the factors were calculated and graphed as frequency curves (fig. 37a); the factors are arranged to read from left to right, earliest to latest. Also included on the graph are three of the six variables (see bottom of fig. 31) that did not load highly on any of the nine factors. Pottery drawings (figs. 33-36) illustrate what kinds of pottery load highly on the different factors. The pottery was also subjected to cluster analysis (fig. 32) using the Pearson correlation coefficient and the single linkage method of clustering. A number of clusters correspond closely to specific factors of the factor analysis, and at one point of dissimilarity most of the eighty variables (pottery types) fall into eight clusters, labeled A to H. Cluster A contains most of the variables that load on factors 2 and 4; cluster B contains mostly variables of factor 5; large cluster C contains elements from factors 1, 6, and 7; cluster G has two variables from factor 9 and one from factor 4; cluster H mainly contains variables from factor 8. Again, frequency curves (fig. 37b) illustrate the distribution of pottery belonging to the eight clusters, plus four individual pottery types still unclustered in the point selected on the dissimilarity scale (cluster E was omitted from fig. 37b). Finally, frequency curves (fig. 38) of selected individual pottery types are presented, as well as frequency curves of painted and decorated wares (fig. 37c). In the latter figure some of the calculations are based on incomplete and impressionistic data. In the Alalakh report there is no accurate statistic for the distribution of the Amuq-Cilician painted ware, though it usually occurs on two vessel types: bowls (types 23a-c and 119) and pitchers (type 70). The frequency curve is based on the assumption that all of those vessels were decorated in the Amuq-Cilician style, but in fact how many were undecorated is not known. In the case of the local painted ware and Habur ware there are no figures, nor is there for Nuzi ware in level III, although its presence in levels IV and II should be accurately reflected. The frequency curve for Gray ware should also be reasonably accurate. No attempt was made in figure 37c to subdivide Mycenaean and Cypriote wares, but see figure 38 for a breakdown of Cypriote wares. In both figures the mainland Red Lustrous ware was grouped with Cypriote wares, since I used Astrom's classification for it too.65 64. The rotation was by varimax and the factors were sorted. Choice for the number of factors to be extracted was by using the rule of thumb of eigenvalues greater than 1. 65. For classification of Cypriote pottery I followed P. Astrom, "Relative and Absolute Chronology, Foreign Relations, Historical Conclusions," in P. Ástrom, ed., The Swedish Cyprus Expedition vol. 4:1D: The Late Cypriote Bronze Age (Lund, 1972). 198 THOMAS L. MCCLELLAN TABLE OF LEUEL BV TVPE (coLurms) KU PERCENTS 1 2 3 4 5 6 l 1. 14 2.27 00 .00 3.41 1. 14 2 .00 2.S2 00 .00 3.77 .63 3 4.07 i 63 .01 .00 .00 6.50 4 0.32 10 43 . 16 00 .81 «7 5 34.67 73 .00 73 36 CO 6 14.29 00 .00 3.17 00 1.59 7 19.35 6.45 7.26 OG 00 CO 8 18.82 00 14. 12 21 18 00 .00 g 7.04 ,00 39.44 36.62 .00 .00 10 00 1.20 1.20 43.37 .00 .00 TOTAL 11.50 8.24 3.08 4.98 .89 t.01 7 8 9 10 11 12 1 .00 .00 12.50 .00 2.27 4.55 2 .00 2.52 8. n .00 1.89 13.21 3 1.63 «88 3.25 .00 61 8. 13 4 1.94 2 43 3.24 00 1.13 2.92 5 1.82 .73 2. 19 36 .36 .36 1 .00 3 17 .00 00 .00 4.7» 7 .X .81 .00 ai .00 .00 t .00 CIO .00 2.33 00 CO a .00 00 .00 .00 .00 .00 ig .00 00 00 7.23 00 .00 TOTAL 1. 13 1.78 3.20 .39 .83 3.38 13 14 13 19 n 18 r 4.5! 2.27 .00 00 .DO .00 2 4.40 1.2t .00 63 .00 5.03 3 4.88 1.23 .00 .00 .00 4.88 4 7.94 1.94 00 .00 00 2.11 3 4.01 4.^4 1 82 .36 00 .73 6 4.76 9 32 4 78 00 .00 1.59 7 .61 4 84 8 87 6.43 00 .01 i 1. 18 00 00 5 88 3.53 .00 9 ■ X 00 .00 .00 4.23 .00 10 1.20 .00 .00 3.61 0 4.1 .00 TOTAL 4.92 2.67 1 13 1.07 .77 1.94 19 20 21 22 23 2* 1 00 1. 14 1.14 .00 1, 14 .00 2 .00 63 12 58 00 3.77 .00 3 .00 1 63 .81 .00 1 63 .00 4 . 16 1 04 .65 5. 19 1. 13 .49 5 1.46 36 CO 36 .30 .36 s 3.17 1.59 .00 3.17 1 39 1.59 7 4.03 1.61 CO .61 00 .00 S 7.06 CO .00 00 CO .00 1.41 00 .00 00 DO .00 10 .00 .00 .00 00 .00 .00 TOTAL 1. 13 1.19 1,54 2. 13 1.01 .30 25 26 27 28 29 30 1 00 .00 .00 .00 .00 22.73 2 00 .00 i.n .00 00 4.40 3 .00 CO 00 .00 00 .00 4 1.62 2 59 2.76 1.94 1.94 .16 3 1.09 1 09 1.46 00 00 .00 6 3.17 00 .00 00 .00 00 : .00 00 1.61 .00 .00 00 8 .00 .00 .00 .oc 00 .00 9 .00 .00 .00 CO .00 .00 10 .00 .00 00 00 .00 .00 TOTAL .99 1.13 1.54 .71 .71 1.66 31 32 33 34 35 36 1 3.41 2.27 ,00 .00 .00 00 2 .63 1.26 3.14 1.26 .00 1.89 3 .00 1 63 ,00 81 .81 23.38 4 1.13 49 .0! .91 97 .00 5 .73 .36 4.01 .73 1 46 ,00 S 00 1.90 I.S9 CO .00 1.39 7 .00 .00 .00 CO 00 .00 8 .00 00 1.18 00 8.24 .00 9 .00 .00 .00 00 00 00 IC 1.20 .00 .00 .00 4.82 1.20 TOTAL .83 .63 1 36 .39 1.30 2.02 37 39 39 40 41 42 1 4,55 3.41 1. 14 3.41 .00 1.14 2 .63 1.26 .00 00 .00 .63 3 1.63 81 .00 3.23 .00 .00 4 .65 81 .32 .00 CO 1.30 3 .00 .00 3.63 00 36 .73 0 .00 1.59 .00 00 1 59 .00 7 .00 .00 1.61 .00 .00 00 8 .00 00 .00 .00 1 ie .00 9 00 00 .00 00 1 41 .00 10 .00 .00 00 .00 .00 .00 TOTAL .63 .71 .99 .41 .24 .71 43 44 43 46 47 48 1 oo .00 .00 1.14 CO .00 2 00 30 00 .63 00 .00 3 .00 1.63 OG CO .00 .00 4 .00 63 ou . 15 . 16 .00 I ^3 00 1.S2 ,73 .Ä .00 6 00 4.76 .00 .00 3 17 .00 7 .91 00 oc ei .00 00 B 7.06 .00 CO oc .00 1.18 g 4,23 00 .00 .00 CO 4 23 10 .00 00 .00 ,oo 00 .00 TOTAL 71 .53 .30 .36 .24 .24 Fig. 29. Alalakh: Matrix of Percentages of Eighty-One Variables in Ten Levels. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 199 49 50 51 52 53 94 1 .00 .00 oo .00 .00 2.27 2 .00 .00 00 .00 .30 .63 3 .00 .00 00 .00 00 .00 4 .00 00 .00 .00 .00 .81 5 3.65 00 .36 1.09 00 .00 S 1.91 1» .00 1.54 00 .00 7 .00 00 5.63 00 .00 .oo 6 .00 .00 .00 00 .00 .00 0 .00 .00 .00 00 .00 .00 10 .00 6.02 .00 1.20 3.C1 .00 TOTAL .65 .30 47 .30 . 10 .47 55 56 37 58 59 60 1 .00 .00 .00 .00 1. 14 .00 2 .00 1 26 .00 00 .00 .00 3 .00 .30 .00 00 61 .OC 4 .00 .31 65 65 .32 .00 3 .73 1.44 .00 .00 .00 .73 6 1.» 00 00 1.59 00 3. 17 J .DO CO .81 .31 00 .00 e 1. te oo 2.35 .00 00 .00 8 .00 00 .00 .00 00 .00 ID .00 .00 1.20 00 .00 .00 TOTAL .24 .65 .47 .36 .24 24 61 62 63 64 63 66 1 .00 1.14 .00 .00 .00 1.14 2 .00 1.09 .63 .00 .00 .63 3 .81 .•30 .00 00 2.44 .00 4 1.62 .32 . 16 91 .49 .49 5 .73 .00 i 46 .00 .00 .00 6 4.76 .00 00 .00 .00 1.59 7 .00 .00 .00 .00 00 31 8 .00 .00 00 00 .00 .00 8 .00 .00 00 00 .00 .00 10 .00 .00 .00 .00 1.20 .00 TOTAL .99 .36 .36 .30 .41 41 67 68 B9 70 71 72 I .00 00 .00 3.41 1. 14 .00 2 .00 63 .00 0C .63 1.26 3 .00 .00 .81 00 .81 .00 4 .00 .01 00 00 .00 65 5 73 i .09 2 55 .00 .00 .00 £ 00 .00 00 .00 .00 1.59 ? 00 .81 .00 .00 00 8.06 H 00 .00 .00 .00 00 00 3 00 00 .00 .00 oo .00 '0 .00 .00 .00 .00 .00 .00 TOTAL .12 .59 47 IB . 18 1 01 73 74 75 76 77 78 1 .00 .00 .00 .00 1. 14 .00 2 00 CO 00 .00 2.32 1.26 3 .00 00 .00 .00 00 .81 4 .00 .00 .16 .00 4S . 16 5 .00 .00 Í.46 36 1.09 .00 6 1.59 CO .00 OO .00 .00 7 1.61 1.61 .00 2.42 .00 .00 a .DO .00 00 .00 1. 18 .00 0 .00 00 .00 .00 .00 .00 10 .00 00 .00 .00 .00 .00 TOTAL 18 12 .30 .24 .71 .24 79 SO 81 Torn. 1 1. 14 .00 6.92 100.00 2 .00 00 10.06 100.00 3 .81 1.M 8 13 100 00 4 .65 32 9.40 IX. 00 S .00 .00 5.57 100.00 8 .00 00 3.17 too.00 7 .00 00 9.68 100.00 8 .00 00 2.33 100.00 9 .00 00 1.41 100.00 10 .00 .00 13.25 100.00 TOTAL .36 .24 8.06 100.00 NUMBER OF VESSELS IN EACH ASSEMBLAGE 1 = 88 2 - 159 3 - 123 4-617 5 • 274 6 > 63 7 =r 124 6 - 85 9 - 7 1 10. 83 Fig. 29. Alalakh: Matrix of Percentages of Eighty-One Variables in Ten Levels, (cont.) 200 THOMAS L. MCCLELLAN DISTANCE METRIC IS EUCLIDEAN DISTANCE SINGLE LINKAGE METHOD (NEAREST NEIGHBOR) TREE DIAGRAM 2.000 DISTANCES 4.000 111- |-1 II-I IV-"| v-1_ -— VI-1 vii-' — VIM—-1 ix—---J 10- DISTANCE METRIC IS 1-PEARSON CORRELATION COEFFICIENT SINGLE LINKAGE METHOD (NEIGHBOR) TREE DIAGRAM 0.000 DISTANCES 1.000 I -II- III- IV-V- VII-VI -VIII-IX-10- ROTATED LOADINGS 1 2 3 L5 0.901 0.075 -0.002 L7 0.867 0.128 -0.088 L6 0.823 0.083 0.113 L9 0.077 0.911 -0.058 L10 -0.115 0.880 -0.020 L8 0.457 0.837 -0.098 L2 -0.048 -0.054 0.850 L1 -0.092 -0.016 0.701 L3 0.180 -0.031 0.528 L4 0.446 -0.104 0.382 C Fig. 30. Alalakh: (a) Dendrogram for Cluster Analysis of Levels Using Euclidean Distance, (b) Dendrogram for Cluster Analysis of Levels Using Pearson Correlation Coefficient, and (c) Factor Loadings of Levels on Three Factors (High Loadings are Bracketed). THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH (70) (30) (37) (54) (31) (38) (9) (59) (79) (11) (VI) (32) m (46) (5) 142) Í62) (47) (60) (24) (61) (44) (25) (14) (58) (55) (66) (52) (73) (22) (67) (45) (75) (09) (63) (40) (39) (1) (33) (56) (68) (7) (72) (51) (74) (76) (15) 06) (29) (28) (64) (2) (26) (27) (13) -Sin (36) (6) (65) (8) (18) (21) (23) (78) (12) (77) (34) (53) (50) (10) (17) (57) (35) (19) (43) (4,S) 0) (20) (4) (41) -0.965 -0.962 -0.957 -0.936 -0.900 -0.895 -0.864 -0.832 -0.791 -0.774 -0.752 -0,734 -0.733 -0.677 -0.675 -0.619 -O.505 0.093 0.100 0.100 0.098 0.087 0.102 -0.042 0.136 0.214 -0,468 0.193 0.154 0.073 0.068 0.068 0.069 0.071 0.079 0.100 -0.173 0.300 0.157 0.075 0.138 0.054 0.146 0.127 0.122 0.133 0.169 0.301 0.004 0.004 0.004 -0.053 0.032 0.109 -0.383 0.011 0.027 -0.139 0.081 0.102 0.079 -0.059 -0.217 0.032 -0.248 -0.274 0.064 0.150 0.150 0.228 0.354 0.292 0.274 0.350 0.365 0.359 D.402 -0.285 0.413 0.388 0.071 0.098 0.124 0.102 0.144 -0.307 0.170 0.121 0.112 0.179 0.186 -0.394 0.128 0.191 0.160 0.091 0.161 I-Ü.987 I -0.973 -0.973 -0.967 -0.934 -0.915 -0.855 -0.849 -0.716 -0.716 -0.684 -0.638 -0,544 0.004 0.004 0.002 0.035 0.059 -0.403 0.074 -0.177 -0.202 0.089 0.123 0.052 -0.056 0.118 0.117 0.118 -0.379 0.280 -0.014 -0.014 -0.014 0.070 -0.012 0.124 -0.366 0.098 0.054 -0.118 0.159 -0.407 -0.050 0.151 -0.210 0.178 -0.124 0.230 0.158 0.141 0.141 0.207 0.312 0.257 0.240 -0.148 0.305 0.300 0.342 -0.430 0.297 -0.473 0.075 0.086 0.116 0.120 -0.046 0.186 -0.023 0.124 0.140 0.023 0.122 0.082 0.101 -0.381 0.036 -0.292 0.103 0.019 -O.101 -0.069 0.016 0.157 -0.156 -0.265 0.188 -0.192 0.236 -0.349 0.154 0.103 0.122 0.016 0.079 -0.069 -0.053 0.181 0.092 0.092 0.092 0.114 -0.312 -0.290 -0.177 0.077 0.080 0.125 0,141 0.071 0.046 0.068 0.177 0.091 0.144 -0.248 -0348 0.144 0.144 0.147 0.291 0.246 0.056 0.074 0.164 0.264 0.296 0.148 0.320 0.155 0.042 0.048 0.086 0.083 0.141 0.103 0.105 0.105 0.123 0.123 0.091 0.134 0.085 -0.480 0.073 0.142 0.061 0.041 0.042 0.072 0.083 0.066 0.092 -0.380 -0.392 0,186 -0.347 0.135 -0.710 -O.03O 0.060 0.060 0.071 0.082 0.079 0.070 -0.334 -0.257 0.154 0.107 -0.444 0.151 0.097 0.097 0.097 -0.206 0.118 -0.348 0.184 0.091 0.084 0.106 0.150 -0.000 -0.002 0.039 0.108 0.075 0.110 0.184 0.138 0.135 0.135 0.101 0.343 -0.015 0.312 -0.233 0.267 0.346 0.193 -0.457 0.396 0.349 0.123 0.135 0.042 -0.199 -0.145 0.021 -0.028 -0.054 -0.289 -0.198 0.228 0.137 0.191 0.063 -0.001 -0.641 0.004 0.097 0.150 -0.155 -0.168 0.052 -0.329 0.057 -0.228 0.221 -0.019 0.192 0.127 -0.824 0.029 0.029 -0.080 0.070 -0.036 0.088 -0.007 0.057 0.005 -0.341 -0.487 -0.600 -O.009 0.029 0.027 0.032 0.103 0.180 -0.990 -0.990 -0.990 -0.966 -0.939 -0.740 -0.651 0,136 0.173 -0.188 0.182 0.209 -0.005 0.014 -0.486 -0.464 -0.O40 -0.408 -0.352 -0.609 0.270 0.255 0.189 0.246 0.056 0.063 0.060 -0.110 -0.284 0.063 -0.165 0.073 0.157 0.221 0.113 0.085 0.036 0.064 0.064 0.064 -0.242 0.140 0.081 0.130 0.140 0.238 0.175 0,168 -0.538 0.106 0.069 0.065 0.075 0.097 0.203 -0.005 -0.005 -0.005 -0.019 0.021 0.162 -0.326 0.063 -0.979 0.157 -0.970 0.052 -0.952 0.014 -0.896 -0.189 -0.750 -0.110 -0.590 0.068 0.140 0.043 -0.169 0.051 -0.394 0.052 0.326 0.039 0.328 0.357 -0.331 0.107 0.068 0.107 0.068 0.158 0.127 0.241 0.242 0.063 0.193 0.102 0.097 0.231 0.257 0.242 0.296 0.225 0.302 0.244 0.309 0.497 -0.456 0.280 0.298 0.319 0.307 0.145 -0.041 0.002 -0.112 0.051 -0.203 -0.439 0.109 0.109 -0.539 -0.372 -0.325 0.085 -0.195 -0,659 -0.211 -0.795 0.027 0.035 0.041 0.030 0.006 0.052 0.024 0.083 0.122 -0.166 0.156 0.110 0.052 0.086 0.086 0.089 0.073 -0.316 0.088 0.184 0.294 -0.541 -0.573 -0.277 0.050 -0.016 0.136 0.130 0.144 0.146 0.191 0.021 0.021 0.021 -0.034 0.055 -0.428 -0.346 -0.032 -0.105 -0.100 0.046 -0.432 -0.773 8 0.042 0.045 0.083 0.058 -0.273 0.091 0.081 0.104 0.107 0.089 0.099 0.125 0.101 0.086 0.049 0.067 0.042 0.035 0.034 0.044 0.061 0.068 0.050 0.119 0.070 0.194 0.095 -0.527 0.068 0.060 0.010 0.010 0.014 0.039 0.019 0.023 0.051 0.290 0.103 0.033 0.069 0.087 0.072 0.060 0.059 0.061 0.073 -0.175 0.036 0.036 0.036 0.013 0.039 0.071 0.054 0.101 0.050 0,119 -0,342 0.136 0.115 0.026 0.089 0.071 0.099 0.159 0.084 -0.936 -0.844 -0.583 0.032 0.050 0.066 0.061 0.066 0.096 0.108 0.065 0.066 0.118 0.104 0.131 0.065 0.085 0.102 0.084 0.103 0.040 0.043 0.047 0.055 0.055 0.051 0.092 0.034 -0.494 0.082 0.055 0,020 0.034 0.038 0.038 0.040 0.055 0.076 0.051 0.044 -0.284 -0.111 0.098 0.074 0.068 0.015 -0.005 -0.008 -0.002 0.017 -0.608 0.014 0.014 0.014 0.035 0.029 0.077 0.030 0.059 0.067 0.083 0.062 0.124 0.136 0.100 0.141 0.116 0.144 -0.265 0.130 0.013 0.013 -0.267 -0.047 -0.175 -0.255 0.369 0.464 0.457 0.463 0.174 -0.388 0.432 0.811 -0.796 -0.684 -0.467 0.349 0.282 0.109 0.077 -0.054 Fig. 31. Alalakh: Table of Factor Loadings of Eighty Variables (Pottery Types) on Nine Factors (High Loadings are Bracketed). THOMAS L. MCCLELLAN Fig, 32. Alalakh: Dendrogram for Cluster Analysis of Eighty Variables (Pottery Types). Clusters A to H are Bracketed. THE CHRONOLOGY AND CERAMIC ASSEMBLAGES OF ALALAKH 203 13{v50) 119(v4) 29(v53) 70(v10) Y 115a(vi7) FACTOR 8 ^) ^P> 23c(v4) 23b(v4) 23a(v4) 10Q(v43) 21a{v3) FACTOR 9 26a(v35) 56(vS7) 57b(v57) 57a(v57) S8(vS7) 137(v19) I d) * 4 129b(v72) 130a(v72) 129c(v72) 21c(vS1) 139(v76) 106b{v15) 130b(v72) 130c(v72) FACTOR 4 113a(vl6) 113b(v16) 113c(v16) 132a(v74) 8(v47) 24(v52) 163b(v22) 72b(v58) 45(v55) 47(v55) 46(v55) FACTOR 2 (3 ® (D 102a(v44) 103a(v14) I04b(v14) 103b(v14) 104a(v14) 102b(v44) 75(v60) Fig. 33. Alalakh: Drawings of Variables (Pottery Types) That Load Highly on Factors 8,9, 4, and 2, 204 THOMAS L. MCCLELLAN FACTOR 2 131a(v73) 111(v66) B9(v61) B8(v61) 2a(v1) 2b(vl) 15(v33) 9c(v49) 9b(v49) 5(v1) 6a(v1) 6b(v1) 3a(v1) 117a(v67) 127(v69) <3> 107b(v63) 107a{v63) 136(v7S) €5 E FACTOR 3 122{v68) 126(v68) 124(v68) 125(v68) 123b & ^ u> 98(v62) 95a(v40) 95b(v40) 160