Y2K BCE: Problems in the chronology of the early Middle Bronze Age and the role of radiocarbon dating
Lecture delivered by E.
Marcus to the Middle Bronze Age Study Group, March 30, 2000 .
(note to fellow presenters and respondents: this text is an English version in the spirit of what will be an illustrated 30 minute presentation in Hebrew at HUC. Although you can expect some variation in particular wording, the questions, approach, data used, and conclusions, will be largely the same.)
“we have . . . been preparing time-tables; let us now have some trains.”
Sir Mortimer Wheeler
(cited by Manning 1993:
Thus, in an oft-cited quotation, Sir Mortimer Wheeler, one of the founding fathers of modern archaeological method, begged us to cease the endless scholarly preoccupation with chronology and turn to issues of substance and process. On the one hand, his call was, and remains, justified, and has to a large extent been followed by numerous scholars. However, the study of cultural, political, and economic process in any period is of little utility, if it lacks a temporal yardstick against which it may be measured. In general, for the study of regional process in isolation, a relative sequence based on a stratigraphically-based typological sequence can provide a convenient and, to some extent, meaningful framework. However, when considering inter-regional or cross-cultural interactions or comparisons, relative chronological frameworks are of limited utility. In order to study such interactions and make meaningful comparisons, a means of synchronizing relative frameworks is necessary, and may take some form of bi-lateral relationship, or a relationship to a third independent framework, such as another region, culture, natural phenomena, or our own modern calender. For the Middle Bronze Age IIa in the southern Levant, mutual relationships, in the form of artifactual comparisons between the Levantine relative sequence and the Egyptian or Syro-Mesopotamian sequence have been the mainstay of modern research. Establishing a synchronic relationship with either of these cultures should have provided, provisionally, an indirect calendrical chronology for the Levant, but a few salient problems have remained. Among these are the fact that while the synchronic evidence is patchy at best, it is extant primarily in the latter part of each sequence, leaving the early phases without firm chronological anchors. Moreover, while both the Egyptian and Syro-Mesopotamian archaeological sequences are synchronized with our modern calender by astronomically-based historical chronologies, a number of still unreconcilable issues has resulted in several competing schemes, dubbed variously the “high”, “middle”, and “low” chronologies. As an independent Levantine historical chronology, based on king-lists and astronomical observations, remains, and may always be, an elusive wish, it seems the calendrical chronology of the southern Levant will always be at the mercy of every change in the Egyptian and Syro-Mesopotamian chronologies. One approach that has been neglected in the past and may contribute to resolving aspects both problems, that is the complete dependence on the calendrical chronologies of Egypt or Syro-Mesopotamia and the dearth of synchronic evidence in the early phases of the period, is the creation of a radiocarbon calendrical chronology. The goal of this presentation is to consider the state of southern Levantine chronology in the early Middle Bronze Age (MB IIa), and consider some currently available and potential future approaches to utilizing radiocarbon determinations to resolve the problem of when in the early part of Y2K this material culture began.
The relative chronological sequence of the southern Levant during the Middle Bronze IIa is based on the stratigraphical sequences from a limited number of settlements sites, most containing extensive intramural mortuary remains, with which extramural cemeteries and isolated shorter-lived sites may be synchronized. Among the sites, from which sufficient comparative material is available, are Dan, Kabri, Megiddo and others, but the four-phase stratigraphically-based ceramic sequence devised by Kochavi and Beck, z”l, at Aphek has been the most relied upon relative framework for the last 25 years. Despite Beck’s concerns already in 1985 that inter-regional differences limit the utility of the Aphek sequence, a characteristic that has been further underscored by the evidence from Kabri and Dan, some of her seminal observations still hold, particularly her identification of the so-called Levantine Painted Ware as a ceramic hallmark of the early MB IIa. In what manner, temporally and regionally, this ceramic type developed and at what pace the red-slipped and burnished ceramics of so-called “classic” MB IIa became dominant, are still matters for our further consideration and research. At present, based primarily on the intra-site typological development at Aphek, there is more or less an accepted scheme of relative synchronizations (chronological chart), with a tripartite division that is considered MB IIa proper, and a fourth phase used to mark the transition to the MB IIb period.
Synchronizing this sequence with the historical chronologies of Egypt and Mesopotamia has never been an easy task. Despite the general perception that a significant component of early MB IIa material culture owes at least some origin or inspiration to the northern coastal Levant and inland Syria, the absence of clearly datable northern material in the southern Levant coupled with the lack of any clearly southern Levantine material in any reliable sequence from any of the key sites in those regions has excluded any firm direct or indirect links to the Syro-Mesopotamian historical chronology. Only towards the end of the MB IIa sequence and into the MB IIb do we begin to find objects such as cylinder seals and impressions from various sites that can offer some means of synchronization. The cylinder seal impression from Lachish, published by Beck and the cylinder seal from T.984 at Kabri are apparently the earliest stratified examples. Sadly, the former, on a bowl assigned to the Aphek Palace II Phase, lacks clear contemporary parallels and a firm date. The latter, depending on which chronology is used, provides a terminus post quem in the 19th through the first half of the 18th century (middle Babylonian chronology) for possibly the third, but certainly for the transitionary phase of MB IIa-IIb.
Turning to synchronizations with Egypt, the situation is somewhat better, although not without some difficulties and limitations, both in the southern Levant and in Egypt (chart). Many Middle Kingdom finds in the southern Levant are either without secure stratigraphic assignment, such as those from Megiddo, or lack any clear chronological utility, such as the aspatharia shells from Tel Mevorakh, or this ceramic imitation of an alabastron from Aphek (illustrations). Scarab seals, which are widely used in later periods, are scarce and, as Daphna Ben-Tor and Weinstein have both showed, appear only in the latter phases of the MB IIa. This example from Aphek Palace II phase is the earliest found so far and is considered by Weinstein to date to no earlier than the 13th Dynasty. If I understand his argument correctly, the date of its design is based largely on a parallel in the Montet Jar from Byblos, the scarabs of which Daphna Ben-Tor now dates to no later than the mid-12th Dynasty. A scarab impression on a loomweight from a less-than-secure context at Tel Nami, which I published with M. Artzy, also provides a broad, middle to late, 12th Dynasty terminus post quem for the same phase, although Brandl argues that even this upper bound is too high. As a reflection of the Egyptianization process upon local Levantine culture, the scarcity of scarabs until the transitional MB IIa-IIb is a significant factor. However, that alone should not preclude the possibility that a significant portion of the MB IIa was coevel with the 12th Dynasty, even if it is not supported by numerous synchronizing finds. Evidence to such contemporaneity has been found at Tel el-Ifshar, where Y. Porat and Sam Paley excavated what seems to be the longest MB IIa sequence distinguished so far. Their provisional phasing, based largely on the main architectural developments and destructions levels in and associated around a large building in Area C, is divided into eight phases, A-H. Based on the limited reports on the ceramics, it seems that the early phases A-C are characterized by the Levantine Painted Ware, while the “classic” red-slipped and burnished wares characterize Phase E and afterwards. Until further data is available the following relative sequence with Aphek can be suggested (chart). Of salient importance is the Egyptian marl clay jar which was found just south of the large building. Based on parallels in Egypt, D. Arnold dated this vessel to somewhere before the reign of Senusret II through Senusret III (High: 1878-1853), i.e., somewhere in the first half of the 19th century. Other fragmentary Egyptian ceramics were found in dumps of Phases B and C, but for these Arnold was only able to provide a vague date range from the 11th through 13th Dynasties. Unfortunately, the precise stratigraphic horizon of the complete vessel is not clear and it may range from Phase C to Phase E. Depending on the phase to which it is assigned and whether it is interpreted as a terminus post or ante quem has great implications for the synchronization of the MB IIa sequence. As a terminus post quem either Phase C or Phase E should be dated to 1900 or later. As the vessel is in mint condition, apart from a small hole caused by a handpick, and it is not a luxury item typical of the later Egyptianizing phenomenon, I suggest that it should not be viewed as a long-lived heirloom artifact. Thus, as a terminus ante quem, Phases A-C or Phases A-E should be dated prior 1850. The implication for the beginning of the MB IIa sequence depends on how one calculates the lifespan of the preceding phases. At present, no clear firm criteria exist for making such a calculation. Some, such as Manfred Bietak in the case of Tell Daba, have argued for the length of a generation per architectural phase, i.e., approximately 20-25 years, but this is an assumption for which a testable method has yet to be devised.
From within Egypt itself, synchronizations with the MB IIa Levantine sequence have been proposed based on depictions at Beni Hasan, and ceramic imports in at Dahshur, el-Lisht, and especially Tell Daba. Bietak has argued for dating the beginning of the MB IIa to not much earlier than 1900, based on a sequence of well-dated 12th Dynasty tomb depictions at Beni Hasan. During the reigns of Amenemhet I and Senusret II, Asiatics are depicted bearing IBA, or earlier, epsilon or crescentic axes, while in the sixth year of Senusret II’s reign (High: ca. 1872) the duckbill axe makes its appearance in the famous tomb scene of Khnumhotep. However compelling such a clear correlation between material cultural change and a historically-based sequence may be, we cannot assume that Egyptian tomb depictions behaved like catalogs from a weapon’s supplier. All that we can be certain of is that by ca. 1900 the artistic idiom of Egyptian tomb painters had assimilated the material cultural change. This change was gradual, and overlapped, as discussed by Alaa el-din Shaheen in a rarely cited 1990 study in GM. Nevertheless, from the Egyptian perspective, we can use the appearance of the duckbill axe as evidence of a terminus post quem for the beginning of the MB IIa period. This correlates well with the evidence from T. 756 from el-Lisht, where D. Arnold, and others, have noted that two Levantine Painted Ware juglets are associated with Egyptian pottery that could date to as early as Amenemhet II to as late as Senusret III (High: 1911-1853). She also notes the depiction of a Levantine juglet in the aforementioned tomb of Khnumhotep. Unfortunately, a single MB bowl sherd found in a gap between stones in the pyramid context of Amenemhet I, which could have been the basis for the synchronization of the beginning of the 12th Dynasty with MB IIa, could have been deposited by 13th Dynasty tomb robbers. Arnold, with the aid of P. Beck, has also noted other imported later “Palace II” MB IIa pottery, from no earlier than late 12th Dynasty contexts (Amenemhet III) at Dahshur, somewhat earlier than the date proposed by Weinstein, who notes the long continued use of this complex well into the 13th Dynasty.
The site, which should provide the key to understanding the synchronization of the Levant and Egypt, has been the most contentious and divisive. While the sequence at Tell Daba has the largest quantity and variety of MBA Levantine ceramics outside of the Levant, finding a consensus to the nature of relationship has been nearly impossible (comparative synchronization chart). Explanations for this discrepancy include regional variability and lack of full publication. For the present purposes I will limit myself to observations regarding the earliest phases at the site, Stratum H and earlier. Until recently, it was generally perceived that the earliest level at Tell Daba that contain MB IIa Levantine artifacts was Stratum H. Renewed excavations at Ezbet Rushdi, below the temple of Senusret III, have revealed levels preceding that of Stratum H, which also contain imported Levantine pottery. Thus, Weinstein and others, were correct in noting that Daba stratum H does not represent the earliest phases of the MB IIa, something that was certainly inferred by Bietak as recently as 1992. Moreover, the Levantine Painted Ware, which provided the key to the synchronization of Stratum H with early MB IIa, has now been found in Stratum e/2 or e/3 (general Stratum K?), i.e. the reign of Amenemhet II (1911-1876). T. Bagh has noted parallels to these examples both from the southern and northern Levant. Thus at Daba, we can now speak of the presence of Levantine Painted Ware from Amenemhet II through the end of 12th Dynasty, which is quite a longer lifespan than previously observed. How then do we view the examples from Stratum H? In my unpublished doctoral dissertation and in conversations with Prof. Bietak and the Austrian team, I have suggested that the Levantine Painted Ware sherds found in a Stratum H pit are not in a primary context, but are refuse from an earlier phase and are thus not contemporary with Stratum H. This explanation would resolve some of the compression of disparate Levantine types (e.g., Ovoid Tell el-Yahudiya), which are reported from that stratum. Needless to say, my suggestion was not warmly received, although I think it should be considered along with the possibility that some of the mortuary context in later strata contain “curated” MB IIa objects (examples). The publication of the settlement levels from Tell Daba in comparison with the tombs from Daba V may offer an opportunity to test this hypothesis. In the meantime, Stratum H should be viewed as coevel with late MB IIa Levantine contexts. Based on the seriation of Egyptian pottery at Dahshur, Bietak has synchronized this stratum with the late 12th Dynasty (Amenemhet III). I am unaware of any other Egyptian chronological evidence from Stratum H. Thus, Stratum H correlates well with the late 12th Dynasty being coevel with later MB IIa (Palace II?), but this cannot be refined until further evidence is available both from Daba and the southern Levant.
Thus, in summarizing the current state of affairs, the evidence for synchronization with Egypt does not support a date for the beginning of the MB IIa that is much earlier than 1910. Does this correlate with the stratigraphic evidence from Tell Ifshar, as discussed above? Can we place Strata A-C or A-E within the period of Amenemhet II through Senusret III, i.e., as few as three or as many as five principal phases within sixty years? I see no reason why not, but I also see no reason why yes. This is the crux of our Levantine Y2K problem. Can we push back the beginning of MB IIa any further in time without relying on Egyptian chronology. At present, using the available evidence and conventional methods, no. But, I think we can test this date through the use radiocarbon determinations.
Radiocarbon determinations are only beginning to enter the foray of chronological debates in this region. The key issue facing archaeologists, who seek to utilize radiocarbon determinations is whether the results are reliable and meaningful. First and foremost, the types of samples must be chosen from organic materials that most closely reflect the time of deposition in the archaeological context. A case in point is the study by Lev-Yadun, Lifschitz, and others, of wood panels and beams from the Al-Aqsa mosque, which demonstrates the use of timbers that contained rings going back over a thousand years. Thus, while even old wood may be useful in the dating of prehistoric contexts, for historic periods the radiocarbon measurement of a piece of charcoal may only provide a crude and usually meaningless terminus post quem. However, a beam of wood with sufficient rings (50+) could be linked to the increasingly refined dendrochronological sequence being produced Peter Kuniholm or other planned sequences. In lieu of such a direct link, even a small number of rings (20-30) could be subjected to radiocarbon measurement and wiggle-matched to the calibration curve, which I will consider shortly. Suffice it to say, such a match would provide a terminus post quem with a precision as high as 7-10 years! All we need is for one of us to successfully extract such a beam, say from one of the MB II palaces. In the meantime, ideally, one would like to measure an organism that has a short life span, such as a seed or fruit stone, which grows during a single year, and whose death may be presumed to be close if not identical to that of its deposit in the archaeological context. Until two decades ago, unless a large sample of single-year cultigens were found, such as from a granary, the range of contexts that could be subjected to a conventional radiocarbon study was quite limited. The advent of Accelerator Mass Spectrometry radiocarbon determinations, which allows the measurement of minute samples, even individual seeds, has revolutionized research. Now, individual seeds from a granary, from a jar, and even sandwiched between well-stratified and sealed living surfaces, can be measured to a high degree of precision, accuracy, and meaning. Now that reliable and meaningful measurements can be obtained, how do we interpret them? First, we must recognize that radiocarbon measurements are not calendrical dates or date ranges. Due to fluctuations in the production of atmospheric radiocarbon there is no direct one-to-one correlation between radiocarbon “years” and calendrical years. In order to synchronize the radiocarbon measurement, we need a key to the amount of radiocarbon in all preceding years. Such a key, called a calibration curve, represents the measurement of radiocarbon in a sequence of known years. Although there are many naturally occurring annual patterns of growth or accumulations, such as peat bogs or coral reefs, from which samples may be extracted and measured, the most commonly used and accepted key is the dendrochronological sequence. Counting, sampling and radiocarbon measuring of tree rings back in time, has produced internationally accepted curves, the most recent of which was published in 1998. Unfortunately, none of the curves has ever been very user-friendly for the archaeology of historical periods. It is wiggly, and is not at always at the most amenable slope. All this means is that there are many calendrical years, decades and even centuries, for which the same radiocarbon date will be obtained (examples). Thus, there are time periods for which we cannot hope to obtain a meaningful calendrical date, even with the utmost precision. Therefore, examining the radiocarbon calibration curve is a prerequisite to determining the feasibility of a meaningful study. Looking at the Middle Bronze Age, there are many periods that seem unusable, other than establishing broad calendrical ranges that might have some meaning as crude terminus post quems (examples). Fortunately, there is a very amenable unwiggly period in the 20th century BCE, which suggests a priori that radiocarbon can potentially distinguish between the death of organisms, i.e., harvests, in the 20th and 19th centuries. The goal should be to obtain high precision radiocarbon measurement of single-year growth organism from the early phases of the MB IIa Levantine sequence, which can be calibrated to meaningful calendrical ranges (simulated examples).
In order to determine if this is an achievable goal we must consider the calendrical meaning of a calibrated radiocarbon determination. This issue has been a point of contention, particularly in “historical” periods, among radiocarbon specialists, statisticians, and archaeologist for some time, and should likely be an issue that will be considered at the forthcoming radiocarbon conference to be held here in Israel in early June of the current year. The crux of the issue is whether to relate to calibrated radiocarbon determinations in terms of possibility, probability, or both. A raw radiocarbon measurement is expressed as a mean and standard deviation of one or two degrees, expressed by the Greek letter sigma. The standard deviation represents the degree of precision that can be achieved with amount of isotopic carbon in the sample and current laboratory methods. It may be depicted graphically as a normal, Gaussian distribution, and is, as such, not a measurement of radiocarbon years, but a range of possible true measurements. The actual true measurement is somewhere within this range, with the mean being the most probable and the two tail ends being the least probable, but still possible answers. One and two sigma represent, respectively, approximately 68% and 95% confidence that the true measurement lies within that range. In other word, the possible answer must be within this range and the probability increases the closer we get to the center. The smaller the overall range of a single sample, the more meaningful and useful such a measurement is for our uses. Multiple measurements from the same contexts and, it can be argued the same phase can be averaged in order to achieve higher precision and thus more meaningful ranges. This is particularly crucial given the nature of the radiocarbon curve (examples), which when we convert the raw measurement into calendrical ranges can create quite broad and wiggly date ranges. Note, for example, the difference between a simulated date in an amenable versus a wiggly part of the curve for the Middle Bronze Age. Of the many approaches taken to radiocarbon “dating”, one is called the “intercept method”, because it is defined by the widest points of interception on the calibration curve. Others have argued that this is not a realistic approach as it treats all possible answers within this range as equally likely. This deficiency has led to numerous probabilistic approaches, which emphasize the higher probability areas of the curve. There are as many approaches as there are scholars. It becomes even more complicated when we go beyond single measurements, or multiple measurements form single phases, and carry out measurements of samples from stratified sequences. How do we choose between the “possibilistic” and “probabilistic” approaches? I will illustrate one probabilistic approach using the Bayesian statistical method pioneered by Caitlin Buck and others, and popularized by Christopher Ramsey’s Oxcal program. In order to do that I will present this and other issues I will use the pilot project I initiated with samples from Tel Nami and Tell el-Ifshar.
Prior to this project, there were only five available radiocarbon measurements from Tell el-Hayyat and Gesher, the latter on wood. For the pilot project, suites of samples were taken from secure contexts within a stratified sequence (chart). The measurements were carried out at the Oxford Radiocarbon Accelerator Unit and the results are available in the handout I have distributed. Turning to the suite of determinations from Tel Nami, note that all three calibrated ranges overlap considerably. This result is due to the nature of the calibration curve and the broad error range from the raw measurement - something I hope to improve upon in the future. Meanwhile, if we were to consider only the mean, or the mean with either the one or two sigma ranges, we might think that the stratigraphy is reversed. As the individual harvests, which what the radiocarbon measurement reflects, are in relative chronological sequence, we must assume that the true answer for the earliest phase must be in the beginning of its range, the second in the middle of its range, and the last at the end of its range. Where in the beginning, middle, and end, is what Bayesian statistics attempts to resolve by finding the best statistical match for the given sequence. The result is a sequence of truncated distribution curves or date ranges that best fit the given evidence. The main critique of this method is that the “best fit” treats each measurement as equally distributed in the sequence. How do can we know that each of the harvests are equidistant in time from phase to phase? In the case of a measurement of rings from a dendrochronological sequence, the difference in time between the samples is a constant and allows us to find the best fit to the calibration curve. As to seeds in the archaeological sequence, the archaeological difficulty is not readily resolved, but I am optimistic that a statistically valid and archaeologically meaningful method may yet be developed.
Using the intercept method, a possible interpretation of the radiocarbon evidence from Tel Nami, which I propose, is to argue that the lower bound (ca. 1730) for the date range of the legume, provides the lowest possible terminus ante quem for the Palace II pottery in the destruction level in which it was found. Such a date does not conflict with Weinstein’s dating of that phase as late as the early 13th Dynasty, or the evidence from Dahshur, if using Kitchen’s high chronology. If using the low Egyptian chronology, the harvest would have to occur within the first 29 years of the 13th Dynasty. The other date ranges would simply be considered earlier, although by how much is not clear.
Returning to Tel Ifshar, we have five determinations from Phase B though E. I will not repeat the Bayesian analysis of the sequence, but note the overlapping nature of the broad ranges, and the questions they raise regarding the calendrical date of the sequence of harvests. However, note again the lower bounds of the measurements from Phase F/G and G, i.e., both considered by the excavators to be of the late MB IIa Palace II phase, are no lower than 1740. In addition, the upper bounds do not preclude a calendrical date for the early MB IIa in the 20th century, although they cannot, at this level of precision, confirm an early date. If we include the chronological evidence of the Egyptian jar, that is a terminus post quem of 1853 (High) and 1817 (low), the radiocarbon ranges are not in conflict with any of the Egyptian chronologies, although clearly the curves would be truncated considerably.
Thus, the pilot project of radiocarbon determinations has shown that we can obtain measurements that for the later MB IIa are extremely consistent with the high Egyptian chronology, but do not exclude the low Egyptian chronology. For the beginning of the MB IIa, the available dates do not have the precision necessary to distinguish between the 20th and 19th centuries. However, samples from the incipient Phase A of Tell el-Ifshar were not measured. To improve precision, and to increase the number of measurements form early MB IIa contexts, such as Ifshar Phase A-C, Tell el-Hayyat Phase 5, etc., a new project has been initiated. Funding obtained with the aid of the SCIEM project run by Manfred Bietak, I am collecting samples for measurement. Unfortunately, although I thank all our colleagues for being very forthcoming in offering samples, apart from sites already on the list you have, I have found that hardly any regular botanical sampling has been carried out during the excavations of the last three decades. Given the limited available cultigens, I am focusing on obtaining higher resolution for the contexts already measured in the pilot project and have expanded the sample to include in situ bones, such as articulated animal bones in tombs, such as those from Kabri and Gesher. Presumably, these were slaughtered shortly before they were interred. Unfortunately, successful measuring of bones depends on the preservation of collagen, which isn’t always the case in this region. Finally, I can only urge you to make botanical sampling a priority in the future. Sampling is best planned and executed in the field, when the contexts and stratigraphy are in front of you rather than as afterthought. I think the potential contribution justifies the effort, which must be collective if, using Wheeler’s analogy, we are to succeed both in running trains and making accurate time-tables.