LAUSR

Abstract This article presents a systematic methodological comparison of three archaeobotanical proxies (phytoliths, pollen and seeds) applied to an assemblage of dung pellets and corresponding archaeological refuse deposits from Early Islamic contexts at the site of Shivta. We set out with three main methodological questions: one, to evaluate the relative input of botanical remains from dung in refuse assemblages; two, to evaluate each archaeobotanical dataset and to test whether they are comparable, complementary or contradictory in their interpretations from dung; and three, infer herding practices at the site during the Early Islamic period. Our findings show that ovicaprine dung accumulated in Early Islamic Shivta during at least two periods: mid-7th–mid-8th centuries CE, and late-8th–mid-10th centuries CE. Methodologically, we see incomplete and incompatible reconstructions arise when each method is considered alone, with each proxy possessing its own advantages and limitations. Specifically, the amount of preserved seeds in dung pellets is low, which restricts statistical analysis and tends to emphasize small or hard-coated seeds and vegetation fruiting season; yet this method has the highest taxonomic power; pollen preserves only in uncharred pellets, emphasizes the flowering season and has an intermediate taxonomic value; phytoliths have the lowest taxonomic value yet complete the picture of livestock feeding habits by identifying leaf and stem remains, some from domestic cereals, which went unnoticed in both seed and pollen analyses. The combined archaeobotanical reconstruction from samples of the mid-7th–mid-8th centuries suggests that spring-time herding at Shivta was based on free-grazing of wild vegetation, supplemented by chaff and/or hay from domestic cereals. For the late-8th–mid-10th century samples, phytolith and pollen reconstruction indicates autumn-winter free-grazing with no evidence of foddering. Unlike the dung pellets, macrobotanical remains in the refuse deposits included domestic as well as wild taxa, the former mainly food plants that serve for human consumption. Plant remains in these refuse deposits originate primarily from domestic trash and are only partially composed of dung remains. The significance of this study is not only in its general methodological contribution to archaeobotany, but also to lasting discussions regarding the contribution of dung remains to archaeological deposits used for seed, pollen and phytolith analyses. We offer here a strong method for determining whether deposits derive from dung alone, are mixed, or absolutely do not contain dung. This has important ramifications for archaeological interpretation.