Abstract
In anthracology, taxon saturation curves enable optimal sample sizes (at least 250 charcoal fragments per stratigraphic level) to be determined from the domestic fuel dispersed from hearths; however, a broad spatial sampling of these scattered fragments, recorded using an excavation grid, is necessary for the frequency spectrum of taxa to be reliable; the synchronous levels of a site, therefore, provide the same frequency spectra of taxa, within confidence intervals; in other words, the observations are reproducible, which is a requirement of palaeoecological interpretation. Using sites in southern France, including Lattara and l'Abeurador (Hérault), I propose to further exploit the spatial distribution of taxon frequencies in archaeological levels, with the aim of obtaining information on the heterogeneity of ancient woodlands. Spatial analysis relies on levels of charcoal fragments which have been well sampled through sieving and avoiding charcoal concentrations. At first sight, the frequency distribution of taxa may look remarkably uniform, due to numerous stochastic processes including wood collection, hearth management, combustion, dispersion, burial and fragmentation; typically, the average taxon frequency is used for palaeoecological interpretation. However, the frequencies of most taxa have a unimodal distribution in the grid squares, close to the normal distribution, which is why I hypothesise that the frequency distribution of each taxon (per grid square) might represent its sampling distribution in its past environment. Taxon proportions fluctuate in each firewood assemblage due to its probability of collection, which is relative to its distribution heterogeneity (or accessibility) in the environment. The ratio s2/m (variance/mean), labelled “distribution index”, which estimates the regularity of the taxon distribution in the grid squares, can be used as a proxy of past environmental heterogeneity. Other taxon frequencies have a plurimodal distribution in the squares, meaning that they likely derive from several populations of data. In this case, for taxa identified to genus level, I discuss the possible existence of several species with different frequencies derived, or not, from independent habitats.