Excavation Sample Size: A Cautionary Tale

Abstract

Frequently, only five percent or less of a midden site is excavated for environmental-analysis purposes before it is turned over to the bulldozers for destruction. Such exceptionally small sample sizes have become accepted in cultural-resource-management work as adequate for gaining a good understanding of the chronology and cultural activities at a site. This assumption was tested by the author with a 63 percent excavation sampling fraction from a southern California midden. The data indicate that a far-from-complete understanding of a site may result from small sampling fractions and that more carefully designed sampling strategies and statistical manipulation of the data may not overcome this problem.

Focused, intense and long-term: evidence for granular ark (Anadara granosa) exploitation from late Holocene shell mounds of Blue Mud Bay, northern Australia

“…These excavation squares were further divided into smaller 0.5 m 2 quadrants (25% of the excavated area) to enable sampling of the sieve residue after excavation. This level of sampling was required in the examination of these large mound sites due to the timeconsuming nature of laboratory sorting and analysis of the large volume of material recovered from each site (Ambrose, 1967;Casteel, 1970;O'Neil, 1993).…”

“What do I do with all these shells?” Basic guidance for the recovery, processing and retention of archaeological marine shells

“…How much to sample is difficult to resolve (Bellhouse and Finlayson, 1979); sampling an increasing amount of a midden until no new types of shell or artefacts are produced can mean excavating most of the midden (O'Neil, 1993). Meadow (1980) gave a minimum 10% sample by volume for any site, not just middens.…”

Reassessing Marine Fishery Intensification in Southeast Queensland

“…Various screening experiments have demonstrated that mesh sizes create significant biases in faunal recovery (e.g. O'Neil 1993;Shaffer 1992;Shaffer and Sanchez 1994). Experiments which have included fish bones indicate that mesh sizes between 0.5mm and 2mm are required for maximum representative recovery (Casteel 1970(Casteel , 1972(Casteel , 1976aColley 1990;Gordon 1993).…”

“…The high degree of spatial heterogeneity evident at the two middens which have been subjected to larger-scale excavation, Sandstone Point ( Figure 6) and Toulkerrie (Figure 10), should serve as a caution against accepting results from small samples as representative of entire sites (see O'Neil 1993).…”