As part of the pre-site survey in Dronning Maud Land for the European Project for Ice Goring in Antarctica (EPICA), the spatial variability of snow-layer thickness and snow chemistry was studied at two geographically different ice-core drill sites. The study aimed to quantify error bars on accumulation rates derived from firn and ice cores. One site is located on the polar plateau at Amundsenisen (76° S, 8° W) and the other in the coastal area at Maudheimvidda (73° S, 13° W). Medium-deep ice cores (100 m) and shallow firn cores (10-20 m) were drilled and snow pits (0.5-2 5 m) were dug at each site. At Amundsenisen a large (16 m x 6 m x 2.5 m deep) snow pit was dug. Snow structure in this large snow pit was mapped using optical surveying equipment, and photographically documented. Samples for analysis of nine ions and oxygen isotopes were collected along one depth profile. Density and in situ electrical conductivity measurements were made along three depth profiles! Snow-layer variability was studied in two different areas and at two different scales. At a regional scale, measured by snow-radar soundings, the variability was 8% on the polar plateau and 45% in the coastal area. The variability at a micro-scale in the large snow pit was 9%. The results indicate that ice cores from the polar plateau are more representative for a larger area than ice cores drilled in the coastal area There is no doubt that there are significant error bars on high-resolution accumulation data received from firn and ice cores, especially from the coastal area, but averaging over tens of years reduces the error in accumulation estimates.