Closing Remarks

We are grateful to Ann Killebrew, editor of JEMAHS, for highlighting our project, which underscores the impact the data has on current chronological and historical reconstructions in the southern Levant, and to all of the respondents for providing their perspectives. Below we address key themes and queries raised by them, including the reliability of the Gezer 14C dataset, the proper relationship between 14C dating and pottery, use of archaeomagnetism, some miscellaneous comments, and aspects of historical interpretation. We trust that our comments will be understood in the respectful manner that they are intended.We begin by responding to doubts raised about the reliability of the Gezer 14C dataset, since this is of primary importance. Finkelstein and Piasetzky question the Gezer data primarily in two aspects: (1) “quality” or reliability of contexts and their stratigraphic attributions; and (2) whether Gezer is consistent with other 14C datasets in the Shephelah and southern coastal plain. They conclude that a substantial number of contexts are questionable and proceed to exclude these from their revised Gezer model, although its output shows few substantial differences. Using their multi-site “Shephelah model” as a baseline for comparison, they note that the Gezer model is generally higher. Hence, they decide that the Gezer data is discrepant from its neighbors and cannot be trusted, at least not without additional information. While we welcome careful, critical evaluation of radiocarbon contexts, there are major problems with Finkelstein and Piasetzky’s methodology for assessing the Gezer dataset, including inconsistent judgments of context quality, “by eye” identification of single dates as outliers, and several aspects of the Shephelah multi-site model used to assess the compatibility of Gezer with the surrounding region.There is indeed a hierarchy of contexts for dating purposes, and large clusters of charred seeds with evidence of in situ burning are among the best, having an especially low but non-zero risk of residual or intrusive material. However, these are not the only contexts that can be dated with acceptably low risk, and in most 14C dating programs it is unrealistic to utilize solely “top” contexts. Our approach at Gezer was to restrict ourselves to a reasonable level of risk (i.e., ensuring samples are closely associated with surfaces or installations and avoiding singular items), prioritize the best contexts for multiple measurements, and run enough dates to ensure a consistent dataset and stable Bayesian model, all while staying within budget. We hesitate to freely use the term cluster, but we selected individual baskets containing 3–5 seeds or more and targeted contexts that were moderately or very seed-rich. This should help allay concerns that isolated seeds were dated. Most samples for Strata 12B and 11 were collected during collaborative fieldwork with a 14C specialist (LW), who also selected samples from previous seasons in close consultation with the excavation team (subsequent to a first round of dates with Beta Analytic). Our comments regarding some of the main contexts queried by Finkelstein and Piasetzky, are as follows:We reject the suggestion that the contexts we dated do not represent the respective strata, or that some kind of systematic error biases the Gezer dataset. We took considerable care to select samples and contexts with clear stratigraphic associations and provided reasonable detail to substantiate this. It is unrealistic to expect full publication of stratigraphic details in a journal article. While we will endeavor to provide more photos, plans, and sections for detailed independent scrutiny in the final publication, the notion that the Gezer dataset should be disregarded as unreliable in the meantime is unreasonable.As explained above, manual assessment of outliers is highly problematic except where they are obvious outliers that clearly deviate from other data of the same phase (no overlap at 2σ and well separated). The modeling process should be employed to assess groups of dates, homing in on the overall time signal and preventing poorly fitting data from skewing results—whether due to simple measurement statistics or unrecognized residual/intrusive items. Except where there are serious doubts about certain contexts, only obvious outliers should be excluded from modeling upfront. Best practice is to use OxCal’s outlier functionality to automatically identify and downweigh poorly fitting data (refer to the detailed explanation in Webster et al. 2023). We disagree with Finkelstein and Piasetzky’s removal of many Gezer dates before modeling on a methodological basis, not solely because we consider the contexts reliable.The Gezer dataset shows good overall agreement, and the modeling results are stable. This cannot be explained away by the current critiques of context, or by invoking rodents and stray seeds from brick debris. Risks for residual or intrusive material were managed carefully.Finkelstein and Piasetzky suggest that Gezer is the “odd-one-out” compared to other sites in the Shephelah. This is not the case. We already made comparisons of Gezer with other Shephelah and southern coastal sites based on single-site models: key boundary estimates are shown in Fig. 11 of Webster et al. (2023) (see also fig. 10 of the summary article above), and the OxCal code can be found in the supplementary files. We focused our comments and comparisons on specific issues, not strictly the absolute dating of a series of ceramic transitions. Following the publication of the pottery, one can take this further. However, as we argue below, pinning dates to pottery should not be the primary focus of radiocarbon research but instead the independent synchronization of sites and strata. The end of Gezer 12B fits perfectly with the end of Lachish VII, both considered to mark the end of LB II. As noted, our results for Philistine 2 pottery at Gezer are consistent with Ashkelon and with Tell es-Safi. Distinctly lower estimates from Tel Miqne VIB and especially Beth Shemesh 6 are based on more limited data. Gezer Strata 9 and 8 fit acceptably with nearby transitional Iron I/IIA strata: Khirbet Qeiyafa, Khirbet al-Rai VII, and Beth Shemesh 4. Comparison of 14C data indicates that the first two sites start before Gezer Stratum 8, which is compatible with the pottery evidence, since Gezer Stratum 8 is developed Iron IIA rather than transitional Iron I/IIA. The absolute dating of Gezer Stratum 8 is only a “discrepancy” if it is regarded as belonging with late Iron IIB horizons (see Finkelstein and Ben-Dor Evian this Forum discussion), but this is not warranted. Stratum 9 is ephemeral, and the first major Iron IIA phase is Stratum 8. Gezer Stratum 7 is well dated to the late tenth or early ninth century BCE, and an alignment with Tel Zayit, Tell es-Safi A3, and Hazael is highly improbable; this need not be viewed as problematic from a pottery point of view.Finkelstein and Piasetzky’s point of reference for considering Gezer a discrepancy is a Bayesian model that includes data from multiple sites, arranged in phases that are considered to have equivalent pottery typology (Finkelstein and Piasetzky 2015; updated 2023). In general, they are comfortable with this approach because: (1) they are confident in the ceramic equivalencies and consider it methodologically better to allow this to balance out any differences between datasets from individual sites; (2) it pools larger amounts of data (i.e., better statistics); and (3) they rule out potential for delays or differences in material culture change.In our view, single-site models should be preferred over multi-site models wherever there is sufficient data to allow a site to stand alone. This is because: (1) single-site (or single-area) models are solely or chiefly based on radiocarbon data and directly observable stratigraphic order (Bronk Ramsey 2013: 34); (2) making exact equivalencies based on pottery typology is questionable since strata at different sites probably overlapped considerably (synchronous destruction horizons are the exception rather than the rule!); (3) further, slow and subtle changes in pottery styles and/or limited assemblage sizes can create considerable uncertainty in equating strata, despite the most skilled analysis; (4) finally, we are much less confident than Finkelstein and Piasetzky in the assumption that differences or delays in material culture have negligible impact, whether between the coast and the Shephelah or north versus south. If one wishes to test this, the only way to do so is by comparing single-site models.Multi-site models are useful for gaining a coarse average assessment of a chronological issue where there is little data from individual sites. This is no longer the case for Iron Age chronology in the southern Levant, and by grouping strata together we risk distorting and blurring 14C-based indications for their relative dating and overlap. If outputs from multi-site models differ strongly from the individual single-site models, one should be especially wary. This is either due to incorrect or overly rigid assumptions when attempting to equate strata, or inadequacies in the underlying data or contexts. Both are distinct possibilities to consider.Our examination of the Shephelah model of Finkelstein and Piasetzky raises specific concerns. We limit our comments to omission of data and modeling issues that impact estimates for the start of early Iron I and middle Iron I (using Finkelstein and Piasetzky’s terminology). New dates from the Austrian expedition for Lachish VII, VI, and V are not utilized (see Webster 2021), but perhaps more importantly, datasets from Ashkelon (Asscher et al. 2021) and Tell es-Safi (Asscher et al. 2015) seem wholly omitted from consideration. Notwithstanding Finkelstein’s reservations regarding the Tell es-Safi dataset (Finkelstein 2016), completely omitting datasets from two major Philistine centers is highly problematic. Early Iron I and the dating of Philistine 1 pottery in their model thus relies solely on Tel Miqne, which tends toward lower estimates for the start of Philistine 1 and 2. Beth Shemesh 6–5 are included in middle Iron I, and it is noteworthy that Stratum 6 (the first appearance of Philistine 2 at the site) is represented by three measurements on a single olive pit (Bunimovitz and Lederman 2016: 677; also an unburnt context). A single-site model for Beth Shemesh puts the start of Level 6 in the late eleventh or even early tenth century BCE, but we highly doubt that this transition is adequately captured. Given that Tel Miqne and Beth Shemesh are dominant elements of Finkelstein and Piasetzky’s Shephelah model, the low estimates they reach for early and middle Iron I are not surprising.Some colleagues (Finkelstein and Piasetzky, Ben-Dor Evian) have expressed the sentiment that 14C dates should not be published without pottery plates, or that the results are of minimal value without them. While comparing these two lines of evidence is important, let us be clear: 14C dating (and archaeomagnetism) are capable of setting the absolute chronological framework of a site’s strata without reference to pottery. Utilizing only the long stratigraphic sequences commonly available in the Near East and aided by statistical tools, one can build quite precise single-site models. After more than two decades of 14C dating applied to Bronze and Iron Age sites of the southern Levant, we are increasingly able to directly compare these individual site chronologies and have less need to transfer absolute chronologies using pottery typology. In other words, the primary goal of 14C dating should now shift from trying to pin particular pottery types or assemblages (e.g., “Iron IIA”) to the calendar and instead focus on directly establishing the absolute dates of individual sites and strata, which can then be directly synchronized in absolute terms. We agree with the statement that “we need to abandon the traditional phasing and switch to absolute, real-time, high-precision dates” (Dever, this Forum discussion). In any case, material culture terms should lose their centrality in the language we use to discuss chronology. 14C dating and pottery typology are separate lines of evidence that should be compared when both analyses are complete. Delaying the publication of 14C dating until pottery analysis is complete does the discipline a disservice and makes no more sense than delaying the publication of pottery analysis until 14C dates are available.We provided our assessment of key implications of 14C data for pottery dating and acknowledge that this cannot be properly critiqued by others until our pottery analysis is published. In an attempt to partially address this issue, we offer here a representative sample of the pottery from Strata 8 and 7 (Figs. 3–4). Some “specialty items” in the assemblages should be mentioned. In Stratum 8 we identified 4 sherds of Cypriot White Painted pottery, which might be expected in tenth century contexts (Thomas, Keimer, and Garfinkel 2021: 407, table 12); 5 sherds of Black-on-Red pottery which is similarly found elsewhere in the tenth century (Thomas, Keimer, and Garfinkel 2021), somewhat earlier than the date proposed in Kleiman et al. (2019); 3 black burnished juglet sherds; and several Ashdod Ware sherds, all of which fit comfortably within the tenth century. All of these specialty items are found in similar numbers in Stratum 7: 4 black burnished juglet, 7 Cypriot White Painted, and 2 Black-on-Red sherds. Given the similarity of pottery types between the Iron IIA early and Iron IIA late (generally dated to the tenth century BCE and ninth century BCE, respectively), we are not sure that presenting the pottery will be a deciding factor in arguing the date of our Strata 8 and 7; our ongoing analysis of the pottery assemblages, however, may prove otherwise. On the other hand, contrasting the assemblages of Stratum 8 and 7 will highlight the development of the pottery sequence within the tenth century. Preliminary observations indicate that while the main types continue from Stratum 8 to Stratum 7, a contrast is seen in the ratio of surface treatments of bowls and kraters. In Stratum 8, red “wash” (weak red slip without burnish) predominates while red slip and hand-burnishing appears, whereas in Stratum 7, the opposite is observed: red-slip and hand burnishing predominates, and red “wash” is found in lesser numbers.Archaeomagnetism is mentioned three times (by Ben-Yosef, Levy, and Maeir). We are certainly in favor of advancing the application of archaeomagnetism alongside radiocarbon dating. In fact, we submitted samples for archaeomagnetism in conjunction with Y. Vaknin two years ago. First, we wanted to find independent dating for what we interpret as the Tiglath Pileser III destruction of the Stratum 6 Four Room House (Wolff 2021). Vaknin came to the field and took samples, but it turned out that the mudbrick was contaminated by materials used in restoring the building, so that effort was unsuccessful. As a backup, however, and as a substitute for 14C analysis from Stratum 6 (cf. Levy’s Halstatt Plateau comment), we decided to sample loom weights found in that building. In addition, we wanted to receive independent dating results for the tenth century, so Vaknin and Wolff chose sherds from Strata 8 and 7 to sample. Finally, we chose to sample a few mud stoppers from Stratum 10. Unfortunately, none of these samples have been prepared or run. Archaeomagnetism in Israel is essentially a one-man show (Vaknin) and a one-lab show (Hebrew University that is and has to and to results are to we need more in this whether from or the should be mentioned that took three archaeomagnetism samples from Bronze and Iron Age in our the results have not been many respondents the Gezer 14C dataset has implications for the Iron Age chronology and associated historical Gezer a of 14C evidence for the dating of Philistine pottery and for an early of Iron IIA in the south. However, we that the of research for 14C dating and Iron Age chronology is not about a or chronology. sites, most large datasets at Tel and Tel favor a low chronology et al. and about Tel This the an both in terms of our of the pottery typology and the historical Finally, chronological should who is or this only to what is representative of good to certainly do not to strata and phases to specific We to by of their there are many for a destruction even if its date fits the of a particular historical On the other hand, we be not to consider the possibilities raised in is indeed but without to suggest test them, and then have no in our with of the by absolute dating is the to test the of proposed between and from a chronological without the need to take regarding the of the can help to the possibilities of might be Evian that generally to not have been to and their On the other hand, we at Gezer Stratum 12B is not a destruction of the site (see 2023). The destruction of the represent limited the of the during an by as Ben-Dor Evian it might have been the at a when it was we only that the destruction of the dates to that it directly or to is only a reasonable not to our data and the appearance of Philistine 2 pottery at Gezer most in the BCE the destruction of Stratum 12B and a phase We to by Ben-Dor Evian that Philistine 2 pottery at Gezer as the date for the appearance of this pottery or that Philistine 2 early BCE This is not by the 14C that data is compatible with the middle since Philistine 2 at Gezer in the middle or the of the century. The Gezer data does not to a Philistine but it does seem to rule out a low are chronological observations and of the of historical III and should be and comments regarding the of the chronology on and to balance historical by the of However, we find comments the Gezer their and these are We suggest that by the of our rather than by the by the for an article in a The that was the only to this point that are Evian and are to highlight the current of radiocarbon data for As we note in our et al. and in more detail in Webster and and Wolff estimates for the of I are using the New 14C models. However, I is at the very there is no data for and dates for the are also quite While samples that directly date the of I and the should be a for one with the evidence at hand rather than This is what we in our article. because of associated with both the 14C dating of I and the historical dating of we that an one or with Gezer Stratum 8 is despite some between the estimates for I and Stratum 8 versus the BCE date for to was a not an attempt to prove a Evian that Gezer not mentioned in the of This is not a Gezer has been identified by some with (e.g., others this as has that 11 might be although this is as Gezer might be among the We do not destruction based on the I but consider it a (see 2023). The of and is indeed a and but if a indeed then Gezer be a to in the certainly not site on as some earlier it is perhaps to that all were with minimal and no to with the and to over have raises the of our discipline with destruction It is and to destruction and with single (e.g., the I or While this may well be partially we can increasingly with the help of is by the limited of by and the of to certain Gezer Stratum 7 an of where 14C evidence clearly a and Piasetzky and suggest that a with Hazael might be As above, Finkelstein’s of is is that some dates at the end of their However, the of Stratum 7 dates be considered no is of multiple measurements, from the 2 shows that can be expected for a of measurements BCE and BCE one should expect a of much of the tenth or ninth while for BCE the of dates will in the ninth century. The dates of Gezer Stratum 7 as a are more consistent with what one expect for BCE than for BCE, whereas the opposite is of Tell es-Safi. In Webster et al. fig. (see also fig. 10 of the summary we compared end for Gezer Stratum 7 Tell es-Safi A3, and Tel in 5 we the same but show the individual as well as the end It is highly that Gezer Stratum 7 Tell on the other hand, fit The Tel dataset and model suggest that this might