The question of human antiquity in Australia has recently become a debate largely about dating. On one side, Allen (1994) and Allen & Holdaway (1995) argue that humans have been in Australia for some 40,000 years, essentially because no 14C ages greater than 40,000 b.p. have been obtained from Australian archaeological sites. On the other side, Roberts et al. (1990a; 1994) have published thermo- and optically stimulated luminescence (TL and OSL) age measurements that indicate human presence 53-60,000 years ago. To support their view, Allen & Holdaway (1995) show that ages significantly greater than 40,000 b.p. have been reported from many geological sites in Australia but there are none from archaeological sites.
We believe that the reliability of 14C ages, when close to the dating limit, has not been sufficiently examined. There is no more practical way of testing any dating method than to compare it with another method, using the same or stratigraphically associated specimens. Concordant ages may indicate that both methods are correct; disparate ages demonstrate that at least one of the methods is incorrect. Unfortunately, the 14C and luminescence dating methods have not been well tested against each other at Australian archaeological or geological sites, for the critical age-range of about 35,000 to 50,000 b.p.
The fundamental assumption in 14C dating is that a specimen has remained closed to all exchanges of carbon, from the time of its formation to the time of its measurement. Although a variety of methods are used to isolate the original carbon in a sample, it is not easy to discover whether decontamination is perfect when 14C itself is the only tracer of the process. In this paper, we examine the reliability of Late Pleistocene 14C age estimates derived from various sites, in the light of both 14C data per se and comparisons of 14C dates with ages derived by other techniques, and we show that a large number of 14C dates in the range 30-45,000 b.p. do not represent the true age of their parent deposits. This is not a new conclusion, but by documenting the data we hope that procedures for obtaining reliable ages for late Pleistocene sites - archaeological and geological - will be clarified. Although most of our case-studies are drawn from Australia, our conclusions probably have a wider relevance.
A problem revisited
The question of the age of human antiquity in Australia is reminiscent of an earlier debate, concerned with the position of sea-level some 30-40,000 years ago. FIGURE 1 shows sea-level estimates in the age range of 20-38,000 b.p., from more than 60 studies reviewed by Thom (1973). Sample materials included shells, corals, wood and other organic matter. Many of the putative sea-level estimates are within 10 m of the present level but altogether they scatter over more than 50 m. The scatter does not reflect tectonic displacements of individual sites; most of the data come from places that are tectonically stable on Pleistocene time-scales.
Thom (1973) considered that most - if not all - the sea-level estimates represented in FIGURE 1 are unreliable, except perhaps for the three points around -45 to -50 m at about 30,000 yr b.p. Except for the same three points, none are remotely compatible with widely accepted data which show that sea-level varied between -50 to -130 m, 40,000 to 20,000 years ago, according to 230Th/234U-dated raised coral terraces in Huon Peninsula, Papua New Guinea, and oxygen-isotope data from deep-sea cores (Chappell & Shackleton 1986; Shackleton 1987; Chappell 1994). In some cases, Thom (1973) considered that dated samples may have been misassociated with sea-level events; in most cases the 14C dates themselves were in error. In short, FIGURE 1 represents a set of late Pleistocene 14C ages that have missed the target.
We now examine sources of dating error that have been revealed by selected cross-dating …