Assessment of genetic variability within a natural population of a pathogen is dependent on obtaining a genetically diverse collection of isolates for analysis. Collections of isolates of Ustilago tritici are traditionally made by sampling one smutted head of wheat per field from a large number of fields over a large area. However, there is little evidence that extensively sampling one smutted head from each of many fields over a large area yields a more genetically diverse collection of isolates than intensively sampling many smutted heads from each of a few fields. The objective of this study was to compare genetic diversity derived from intensive sampling of U. tritici (80 isolates from 4 fields) with that derived from extensive sampling (81 isolates from 81 fields). Genetic diversity was measured using virulence data and amplified fragment length polymorphism (AFLP). The virulence of the isolates was assessed on five differential hosts, and the AFLP data were obtained using 10 selective primers that yielded 23 polymorphic bands. Genetic diversity was similar within the extensive- and the intensive-sampling collections, when measured using virulence data, but it was more variable within the intensive-sampling collection than within the extensive-sampling collection, when measured using AFLP. In general, the results of AFLP analysis indicated a higher degree of genetic variability than did the virulence data. The results also indicated that host genotypes can strongly influence the genetic variability within populations of an obligate pathogen like U. tritici when this variability is measured on virulence data, but that host genotype has a weaker influence on diversity measured by AFLP.