Testing feldspar luminescence dating of young archaeological heated materials using potshards from Pella (Tell Tabqat Fahl) in the Jordan valley

Recent developments in the use of more stable feldspar signals in the luminescence dating of sediments offer the possibility of obtaining accurate feldspar luminescence ages for ceramic artefacts; this is especially interesting in locations which do not provide suitable quartz extracts. Here we examine the application of the stable infrared stimulated luminescence signal measured at elevated temperature (in this case 290°C; pIRIR290) after stimulation at about room temperature to Levantine pottery samples. A total of 52 potsherds were collected from three superimposed iron-age units at Pella (Jordan); based on C dating, typology and seriation these units were deposited between 700 and 900 BCE. Sand-sized quartz extracts were unsuitable, and there was insufficient sand-sized feldspar, and so polymineral fine grains were chosen for dating. Various tests for reliability were undertaken (dose recovery, dependence of De on first stimulation temperature etc.). The pIRIR signals are weak, and 14 potsherds were rejected on this basis. Of the remainder, 3 were confidently identified as outliers. Based on those sherds for which IR signals were sufficiently intense, we use the ratio of the IR50 to pIRIR290 signals to argue that these outliers do not arise from incomplete resetting during manufacture. The ages from each layer are considerably over dispersed (typically by ~25%) but average ages for each unit are consistent with each other and with the expected age range. The average OSL age for the site is 2840 ± 220 years (n = 35), with the overall uncertainty dominated by systematic uncertainties; this average is consistent with the range of C ages from 970–1270 BCE reported from across the destruction horizon. We conclude that the pIRIR290 signal is delivering accurate ages, but that the variability in age from shard to shard is much greater than would be expected from known sources of uncertainty. This demonstrates the need for site ages to be based on multiple samples; individual shard ages are unlikely to be sufficiently accurate.