The Gezer Discrepancy: Comments on Recently Published Radiocarbon Dates

A set of radiocarbon results for samples extracted at Gezer in Israel have recently been published (Webster et al. 2023). They cover the Late Bronze IIB to Late Iron IIA strata unearthed by the Tandy excavation in the southern sector of the site, near the Iron Age gate and casemate wall. The authors consider the results against important events and cultural processes in the history of the region: Merenptah’s campaign in the late thirteenth century BCE, the appearance of Philistine pottery, King Solomon’s building activities referred to in 1 Kgs 9:15, Sheshonq I’s campaign to Canaan in the tenth century BCE, and Hazael of Aram Damascus’s assault on the Northern Kingdom in the late ninth century (see also Webster, Wolff, and Ortiz, this issue). They conclude that “We have been able to help place developments at Gezer in regional perspective and check the chronological viability of multiple proposed historical correlations” (Webster et al. 2023: 27). This article focuses on two words in this statement: “regional perspective.”1The strategic importance of Gezer, and possible historical correlations, call for a careful evaluation of the radiocarbon determinations from the site. One way to do this is to compare the Gezer data to radiocarbon results from nearby sites (map in Fig. 1; Table 1), that is, to the large number of determinations from seven sites in the Shephelah that have accumulated in recent years (for the data and our updated chronology model, see Finkelstein and Piasetzky 2023).Stratum 12A: Following Mazar (1990: 290–91, 295–96) and others, the Tandy team label this layer Iron IA. We and others label this phase Late Bronze III, meaning the end phase of Egyptian rule in Canaan, until ca. 1130 BCE (Ussishkin 1985; Martin, Finkelstein, and Piasetzky 2020).A gap between Strata 12A and 11? We label Philistine monochrome-only layers (“Philistine 1” in Dothan and Zukerman 2004) “early Iron I,” and bichrome layers (“Philistine 2” in Dothan and Zukerman 2004) “middle and late Iron I” (for Philistine pottery and its date, see also Killebrew 2005: 219–27, 232). The former is apparently missing in the Gezer sequence, at least in the Tandy excavation area. If so, there may have been a short occupational gap at that time.Stratum 10A, which was destroyed by fire, is described by Webster et al. as Iron I; perhaps it should be equated with the destruction of nearby Khirbet er-Rai (also Khirbet Qeiyafa and Beth-Shemesh 4). We label these layers “Transitional late Iron I / early Iron IIA” (explained in Finkelstein and Fantalkin 2012).Stratum 9 should seemingly be placed in the early Iron IIA (contemporaneous with Lachish Level V—the type layer of this period in the Shephelah) and Stratum 8 in the early days of the late Iron IIA (early days of Lachish Level IV?). According to the authors, Stratum 8 features the six-chambered gate and the casemate city wall (on this, more below). For Stratum 8, the authors describe a destruction with no accumulation of finds on the floors; perhaps this “destruction” should be interpreted as leveling the old structures in preparation for the construction of Stratum 7.The pottery of Stratum 7, destroyed by fire, is described by the excavators as Iron IIA (Webster et al. 2023: 12); it has not been presented yet. Logic calls for the comparison of this devastation with the late Iron IIA destructions at nearby Tell es-Safi/Gath, Tel Zayit, and Aphek (for the latter, Kleiman 2015; more below).The comments below are based on our interpretation of Webster et al. 2023, table 3.Stratum 12B: The two last items (GrM 13317, GrM 13321) seem to the naked eye to be earlier than the other samples. They are described by the authors as “large seed concentrations that were certainly burnt in situ and should be reliable; we suspect this may be simply a matter of measurement statistics, or the seeds include material from slightly earlier in the life of Stratum 12B” (Webster et al. 2023: 18). Calculating the two samples alone results in a date in the fourteenth century BCE (Late Bronze IIA). We wonder if they do not represent a layer below Stratum 12B.Stratum 12A: There are no samples from this layer for radiocarbon dating (Webster et al. 2023: 14). Hence a slot was entered into the model. If this layer is contemporary with Lachish VI, it may have continued until at least 1130 BCE—the generally agreed-upon date for the pullout of Egypt from the Levant (the Gezer team dates it until the mid-twelfth century [19]).Stratum 11: Note that the only sample marked as “several olive pits” (OZV275) provides a date that in fact fits the Late Bronze destruction; it comes from a lamp and bowl (foundation?) deposit. The nature of the other samples—single olive pits or a cluster—is not clear (see below on the risk of dating single items).Stratum 10B: Samples are good: concentration of olive pits.Stratum 10A: All samples are good: clusters in destruction debris.Stratum 9: There is only a single sample for this layer (Beta 436533). Table 3 in Webster et al.’s article describes it as “seeds,” but it is not clear whether these were found in a concentration or as single items in different places on the floor.The authors introduced an extra boundary between Strata 9 and 8 “due to the especially marked change in architecture and possible gap.” We do not deem this necessary, as the plan of Stratum 8 may represent gradual development.Stratum 8: Two results (Beta 436538, Beta 436540) fit the Late Bronze; they were treated as outliers (Webster et al. 2023: 18).Stratum 7: For its end, only the samples from the destruction debris on floor of Unit D Room 5 (Locus 15/81002) should be consulted (more below). Tabun 15/81034 could have gone out of use before the destruction and hence its samples may represent the early days of this layer, if not an earlier phase. Burial 16/91050 may also represent an earlier period.In our chronology model for the Iron Age Shephelah (Finkelstein and Piasetzky 2015; updated 2023) the results from the different sites are quite consistent with each other. Table 2 compares transition dates between periods (ceramic phases, that is, relative chronology) in our Shephelah model with the results of the Gezer model.In such proximity one cannot expect pottery repertoire of sites to change at a different pace. Hence the Gezer results should have provided consistent results with the nearby Shephelah sites. This is not the case. Each of the Gezer dates is not in clear-cut disagreement with the Shephelah dates. However, all Gezer dates are systematically and significantly earlier than the Shephelah model results; in fact, there is almost no overlap between the 68.3% ranges of the two models. Note that the early limit of three Gezer ranges marked with an asterisk is earlier by 72, 134, and 112 years compared to the late limit determinations for the nearby Shephelah sites.In our 2023 article we also presented results for transitions at Megiddo and Tel Rehov, sites with good control over stratigraphy and ceramic typology and a dense sequence of 14C results, located only ca. 85 km air distance to the north and northeast of Gezer, meaning that in this case as well differences in the pace of changes in the ceramic repertoire are unlikely. Megiddo provides the strongest 14C evidence for the Late Bronze and the Iron I in the southern Levant, while Rehov presents good data (from secure contexts) for the Iron IIA. All in all, the Megiddo and Rehov models are consistent with our Shephelah model (Finkelstein and Piasetzky 2023) and inconsistent with the Gezer model.The most obvious case of discrepancy is the destruction of Stratum 7 at Gezer compared to Late Iron IIA destruction layers in the Shephelah. The best way to check synchronization of destructions of different sites is to compare their uncalibrated results rather than their calibrated dates, as the former have smaller uncertainties and Gaussian-like distributions. Due to the nature of the calibration curve, close uncalibrated dates do not necessarily result in similar calibrated dates. However, two uncalibrated dates for destruction layers that do not agree mean that the events did not take place simultaneously.Accordingly, the results for Tell es-Safi/Gath A3 (2707±16) and Aphek 8 (2667±20; Sharon et al. 2007) fit well, seemingly pointing to the military campaign of Hazael of Aram Damascus (e.g., Maeir 2004). The Gezer Stratum 7 and Tel Zayit dates (2773±20 and 2771±15, respectively) are consistent with each other, but there is a standard deviation of ca. 3 from the Tell es-Safi and Aphek dates. Statistically, such a difference between two simultaneous events may occur, but the probability is very low. The more likely interpretation of the data would be that there were two different events, possibly different campaigns during the reign of Hazael (842–800 BCE).Yet, there are other ways to explain this inconsistency: (1) the ceramic assemblages from the Gezer and Tel Zayit destructions are earlier than the assemblages from the Tell es-Safi and Aphek destructions; (2) in the case of Gezer, Stratum 7 samples originated from Stratum 8 contexts. Note in this connection the aerial photographs of the Tandy excavation area (Webster et al. 2023: fig. 3; and Webster, Wolff, and Ortiz, this issue, figs. 5, 9) in which the six-chambered gate looks as if planted from Stratum 7 into the architecture of Stratum 8.2 No less important, attention should be given to Sample Beta-436536, described as “dismantling the plaster surface of Room 5” (Webster et al. 2023: 14), which provides the latest date for Stratum 7, namely, 2730±30 = 901–831 BCE. Though a single item, if this seed indeed comes from the “floor matrix” (Webster et al. 2023: table 3), it gives the earliest possible date for this unit. Remarkably, this date better fits the dates for the neighboring Shephelah sites.Trying to better understand the Gezer Discrepancy, we constructed our own Bayesian model for the site, using OxCal v4.4 (Bronk Ramsey 2009) and the IntCal20 calibration curve of Reimer et al. (2020). Our model adheres to our comments above; Table 3 indicates the differences between the Gezer-team model and our model.Our model yielded a low Oxcal “model agreement” of less than 50%. The system then indicated that removal of one more datum—GrM1990 (Stratum 11)—will improve the agreement to ~75%. We removed this datum too.The results of Webster et al.’s Model B and our own Gezer model are compared in Table 4. Despite the differences in the construction of the models (see Table 3), the results are close. This hints that the problem is in the archaeology of Gezer—the association of the strata with the 14C samples—rather than in the measurements or construction of the model.The following possibilities for the Gezer Discrepancy should be considered:3A verdict on the Gezer results can be reached only once the full datasets have been published. For now, in view of the consistent discrepancy described above, the 14C results from Gezer should not be taken at face value. No less important, in any attempt to establish links between the finds on the ground and textual data, the Gezer dataset must be considered as part of the full body of 14C results from the Shephelah.