Waterlogged sites usually contain large amounts of plant remains and it is impossible to analyse everything that is present. To get representative samples, different sampling programmes have been developed and tested in the last twenty years (Van Zeist & Palfenier-Vegter 1983, Maier 2001, Pétrequin 1997, Jacomet et al. 1989; Jacomet et al. 2004), which should provide informations about:
In addition to these, intra-house and intra-site pattern (Jacomet & Brombacher 2005) should be identified and analysed including:
Systematic sampling (grid sampling) often forms the basis of a sampling strategy. These regular samples might be taken over the whole excavation area, for example one in every square metre. In case of wells or latrines the fills may be sampled in a vertical sequence and horizontally within levels. The samples will mainly contain accidentally formed plant assemblages, resulting from the different human activities. In addition to these samples, judgement samples (Jacomet & Kreuz 1999, 97; van der Veen 1987: fig. 105) or “riddled” and “spot” samples (Kenward & Hall 1995) should be selected from obvious plant concentrations (stores, chaff concentrations, coprolites, moss polsters and others). For on-site stratigraphical investigations, vertical sample sequences are taken as block samples, profile columns or core samples to answer questions concerning settlement history and development of the site (as an example see Jacomet 1985 or Jacomet et al. 2004).
The Environmental Archaeology Unit and the York Archaeological Trust recommend that for a general biological analysis, every significant layer should be sampled (AEA 1995; Kenward & Hall 1995):
While sampling, the following should be considered (after Jacomet & Brombacher 2004, p. 69):
Because of the good preservation conditions, waterlogged sediments contain many more plant remains than dry soils: on average wetland sites produce more than 1000 items per litre sediment (Jacomet, Kreuz 1999, p. 103). The number of samples needed is dependent on the questions posed and the preservation of the material (Jacomet et al. 1989). Of course there is a big range in the concentration of plant remains, depending on the type of sediment and on preservation conditions. For example contents of latrines, rubbish heaps and waterlogged human coprolithes may produce extremely high concentrations of items, while waterlogged samples from housefloors or pathways usually produce only low values.
As a sample should be representative for a special archaeological layer, for a structure or maybe for the site itself, it should produce so many items, that their amount corresponds with the real ratio of taxa in this unit. A statistical approach for that was made by Van der Veen & Fjeller (1982).
On the other hand, some empirically established values may give an impression of the range of sample size used in different sites. For example, for a house floor a minimum of 10 samples may be needed while, 3 to 5 samples may be sufficient for the areas between houses (Hosch and Jacomet 2001 and 2004). The volume of samples should depend on the concentration of plant remains in the deposit but a minimum of 1-5 litres is required (Jacomet and Kreuz 1999) If, for good measure, bigger samples have been taken, they can be split into subsamples before processing or after sieving (AEA 1995; Kenward et al. 1980).
At Roman Stonea, a sample of about 5 litres was collected when a deposit was recognised as containing waterlogged plant remains. Of these samples, usually a 1 litre subsample was analysed (van der Veen 1996). Kenward & Hall (1995) had taken 5-10 kg samples for the general biological analyses at Coppergate, York, while at Bremetenacum, Roman Ribchester, 500 ml of the waterlogged samples were wet sieved to 0.5 mm (Huntley & Hillam 2000). In the neolithic site Hornstaad-Hörnle IA, Germany, nearly 4600 systematic and judgement samples have been analysed with volumes between 10 and 1000 ml (Maier 2001).
While even very small samples (<300 ml) may produce a representative amount of smaller plant remains, large items (fruit stones from sloes or plums, nuts etc.) could be or will probably be under-represented. For that, coarse-sieving programmes may be used in addition to the main sampling programmes, where large samples are sieved only with one big-mesh sieve (for example 3 mm, Maier 2001, 20). Hosch & Jacomet (2001, 66) found, that to get a statistically representative record of items >2 mm in the wetland site Arbon Bleiche 3, CH, 3 litres of sediment were needed.
For more information in German on sampling see University of Basel fieldnotes.