From 1992 to 1999, 18 sets of aecial or uredinial isolates of Puccinia coronata were collected from two sites in Minnesota for analysis of virulence associations. In addition, one set of aecial isolates was collected in Ontario and one in New York. Also, aecial isolates and uredinial isolates collected from scattered locations in Minnesota in 1994 and again in 1995 were bulked for comparison with populations from discrete sites. Isolates were tested for virulence on 26 single-Pc gene oat lines. Virulences to 14 pairs of differential lines were found to be significantly (P < 0·05) associated in linkage disequilibrium in at least six of the 24 site × year populations. The significant virulence associations were found among both uredinial and aecial isolates. Linkage disequilibria normally dissipate with repeated generations of sexual reproduction. Finding the same virulence associations repeatedly over years and locations for sexual populations of P. coronata indicates that certain pairs of virulence genes (or avirulence genes) contribute to increased fitness when they occur together, even in the absence of the corresponding resistance genes in the host. Mean virulence complexity did not differ significantly between a site with no known Pc genes in the host population and a site with low frequencies of Pc genes, suggesting little or no selection pressure against unnecessary virulence per se. Means and standard deviations of virulence complexity were similar for aecial and uredinial isolates within sites, which suggests that selection did not strongly favour either heterozygotes or intermediate virulence complexity during uredinial generations.