Alternaria spp. were sampled from two rough lemon (RL) and two Minneola tangelo (MIN) groves in a limited geographic area in central Florida to test for host-specialized forms of the pathogen. Isolates of Alternaria spp. were scored for variation at 16 putative random amplified polymorphic DNA (RAPD) loci and for pathogenicity on both hosts. Subpopulations on each host were differentiated genetically and pathogenically, which was consistent with the hypothesis of host specialization. Highly significant genetic differentiation was detected among all four subpopulations (Nei's coefficient of gene differentiation [G(ST)] = 0.292, P = 0.000); most of the differentiation occurred between hosts (G(ST) = 0.278, P = 0.000). Phenograms of qualitative similarities among isolates within subpopulations revealed two or three distinct clusters of isolates within each subpopulation. The majority of isolates sampled from RL were pathogenic on RL and not on MIN, although a few RL isolates were able to induce disease on MIN, and 44% were nonpathogenic on either host. In contrast, isolates from MIN were pathogenic only on MIN, never on RL, and only 3% of the isolates were nonpathogenic. Overall, three genetically distinct clusters of isolates were detected on both hosts. One of the clusters (cluster A) sampled from RL was pathogenic on RL and not on MIN and consisted almost entirely of one RAPD genotype. This cluster also contained two isolates that were 93% similar to the majority genotype but were pathogenic on MIN and not RL. In isolates from MIN, two distinct clusters of isolates were found in one subpopulation (clusters B and C), and three distinct clusters were found in another subpopulation (clusters A, B, and C). Clusters A and B were found on both hosts, while cluster C was limited to MIN. Populations of Alternaria spp. sampled from RL and MIN showed a high degree of host specificity; however, the specificity obscured a high level of genetic variation within subpopulations.