Although Serratia marcescens is a well-known nosocomial pathogen, investigation of its hospital ecology has been limited by the lack of available typing techniques. During an investigation of the occurrence of this organism in a neonatal intensive care unit, we evaluated a number of such techniques. Using a selective medium, we conducted prospective surveillance of neonatal rectal colonization and environmental contamination with S. marcescens. In 8 months of surveillance, 5.1% (20 of 394) of the infants admitted to the unit became colonized. Most sink surfaces and drains were also culture positive. Differences between isolates could not be detected in biotypes from a commercial identification system (MicroScan) or by antibiograms, total protein fingerprints, or plasmid profiles. Serogrouping and genomic DNA restriction endonuclease analysis revealed the presence of six strains that colonized infants and a similar number of environmental strains. These two methods were concordant, with the exception that genomic DNA analysis demonstrated lack of relatedness between some strains within the same serogroup. DNA restriction endonuclease analysis was practical and reliable. The differences this method detected between environmental and neonatal strains provided strong evidence that the environment was not an important reservoir for S. marcescens in our neonatal intensive care unit.