Recent revision of the 0 serotyping scheme for Serratia marcescens has allowed the definitive serological identification of a collection of 51 1 epidemiologically distinct strains in terms of both lipopolysaccharide (0) antigens and capsular (K) antigens. High levels of typability were achieved, 88% and 91y0 respectively, with only 2% failing to type with either method. In most cases, non-typability was due to a lack of antigen, i.e., the strains produced only rough LPS or were acapsular, suggesting that typability would be little improved by the discovery of additional serotypes. The distribution of the 58 O:K serotypes was very uneven, with 014:K14 accounting for 30% of the 423 clinical strains in the collection, but only 5% of the 88 non-clinical, environmental strains. Thus, the prevalence of 014:K14 strains in hospitals is not reflected in the environment. Similar conclusions were valid for 027:K14, 021:K3 and 021:K14 strains, as well as those with rough lipopolysaccharide. Conversely, the proportions of 06:K3, 06:K14, 08:K14 or 028:K28 strains were significantly lower among the clinical collection than among their environmental counterparts (12% in total rather than 65%). This suggests that 014:K14 may have a selective advantage in colonising or infecting hospitalised patients and, therefore, that the 014 and K14 polysaccharides themselves may contribute towards the apparent pathogenicity of these serotypes.