METHODS

The cores were sampled approximately every 1.5 m (one sample per section). Selected intervals, where there was obvious lithologic cyclicity, were sampled more closely (every 0.2 to 0.4 m). Sample size was 10 cm3.

Sample preparation was carried out at the British Antarctic Survey. Biogenic silica (in the form of diatoms, radiolarians, and rare silicoflagellates) was measured by point counting a smear slide made from each sample. This method is quick but tends to overestimate silica compared with the true weight percentage (Pudsey, 1993). For textural analysis, each dried sample of ~5 g was wet sieved at 63 µm (4 ) to measure sand percentage and the sand fraction weighed and retained. Fine-fraction size distribution down to 0.5 µm (11 ) was measured on a Sedigraph 5100 particle size analyzer at Royal Holloway and Bedford New College, University of London. The chosen Sedigraph output was the weight percentage within each 0.25- interval from 4 to 11 (63-0.5 µm). The very high clay content of most samples (commonly >70% finer than 8 (4 µm) and >40% finer than the measurement limit of 11 ) necessitated extrapolation of the fine end of the cumulative frequency curves to obtain values of 16 to calculate the standard grain-size parameters of sorting (G) and skewness (SkG) (Folk, 1974). Modal size was measured to the nearest 0.1 from each size-frequency histogram.

The data for each site are plotted against age using the sedimentation rate curves based mainly on magnetic reversal data of Shipboard Scientific Party (1999a, 1999b, 1999c).

NEXT