Deoxynivalenol (DON) and ochratoxin A (OTA) in agricultural commodities present hazards to human and animal health. Bulk lots are routinely sampled for their presence, but it is widely acknowledged that designing sampling plans is particularly problematical because of the heterogeneous distribution of the mycotoxins. Previous studies have not explicitly looked at the interactions between the spatial distribution of the mycotoxin and the strategy used to take samples from bulk. Sampling plans are therefore designed on the assumption of random distributions. The objective of this study was to analyse the spatial distribution of DON and OTA in bulk commodities with geostatistics. This study was the first application of geostatistical analysis to data on mycotoxins contamination of bulk commodities. Data sets for DON and OTA in bulk storage were collected from the literature and personal communications, of which only one contained data suitable for geostatistical analysis. This data set represented a 26-tonne truck of wheat with a total of 100 sampled points. The mean concentrations of DON and OTA were 1342 and 0.59 microg kg(-1), respectively. The results showed that DON presented spatial structure, whilst OTA was randomly distributed in space. This difference between DON and OTA probably reflected the fact that DON is produced in the field, whereas OTA is produced in storage. The presence of spatial structure for DON implies that sampling plans need to consider the location of sample points in addition to the number of points sampled in order to obtain reliable estimates of quantities such as the mean contamination.