Spatial structure of both nuclear and mitochondrial RFLPs were studied in several populations of the chestnut blight fungus, Cryphonectria parasitica, using a variety of spatial autocorrelation tests designed to detect nonrandom patterns. Fungal individuals were sampled from cankers on infected chestnut trees, and the location of each tree was mapped. Single-locus nuclear RFLPs, nuclear fingerprints, and mitochondrial DNA haplotypes were determined for each individual. Individuals with the same DNA fingerprint genotypes occurred closer together than would be expected at random in four of the five plots, while mitochondrial DNA haplotypes were aggregated in all five plots. Genetic distances between individuals, expressed as one minus the proportion of shared restriction fragment size classes for fingerprints and mitochondrial haplotypes, were significantly correlated with Euclidean distances between individuals in four of the five populations, but these correlations were very weak (r < 0.18). The same DNA fingerprint and single-copy nuclear RFLP alleles occurred on the same trees or immediately neighbouring trees more often than would be expected at random. Most of the aggregation for all three genetic markers occurred among individuals within the same cluster of chestnut stems or on neighbouring trees. Lack of spatial autocorrelation in one population was probably due to sampling on a larger scale that was too coarse to detect any patterns. Significant aggregation of genotypes in C. parasitica is most likely caused by some degree of restricted dispersal within populations. The implications of restricted dispersal are discussed in relation to the breeding system and isolation by distance in populations of. C. parasitica.