Thirty-five spore-forming bacilli were isolated from evergreen oak leaves at four stages (one-year-old and two-year-old leaves, upper layer and underlayer litter leaves) and over the four seasons within one year. These isolates, plus five reference strains, were characterized morphologically and physiologically by a total of 100 tests and genetically by DNA/DNA hybridization. Phenotypic similarities of all strains were determined by numerical taxonomy, with the unweighted average linkage (UPGMA) algorithm and simple matching (SsM) and Jaccard (S j) coefficients used as measures for similarity. Three groups (A to C) were defined at a similarity level of 71% (SsM) or 57% (S j). They contained leaf isolates phenotypically related to Bacillus cereus, B. pumilus and B. circulans species respectively. The majority of the leaf isolates were assigned to B. cereus (34%) and B. pumilus (63%). DNA/DNA hybridization also discriminated three groups (genomic groups 1, 2 and unclassified strains) which presented a good correlation with the numerical analysis. Yet, DNA/DNA hybridization grouping revealed a higher degree of discrimination by defining four subgroups (la, lb, 2a and 2b). Genomic subgroup la contained leaf isolates belonging to the B. cereus species; isolates from genomic subgroup lb belonged to B. mycoides species and isolates from genomic subgroup 2a belonged to B. pumilus species. Subgroup 2b consisted of a new genomic subspecies of B. pumilus which exhibited a degree of homology ranging from 53 to 64% with the B. pumilus type strain and a coefficient of divergence (ATm) ranging from 5.5 to 7°C. The different genomic groups presented different substrate metabolism capacities and a different spatial distribution on evergreen oak leaves. B. cereus strains (group 1) were predominantly located on litter leaves whereas B. pumilus strains (group 2) were found on the phyllosphere. In contrast with group 1, group 2 was able to metabolize some sugars and pectin, while group 1 isolates were able to hydrolyze starch and glycogen. Thus, our hypothesis is that group 1 succeeds group 2 when the leaves are littered.