These studies were designed to investigate the production of neuraminidase by a wide range of species of strictly anaerobic bacteria and to explore the possibility that the enzyme may play a part in the pathogenic mechanisms of these organisms. A glycoprotein fraction was prepared from pooled human plasma for use as substrate in assays for neuraminidase. Sensitive and reliable procedures were developed for characterisation of neuraminidase production by Clostridium perfrinaens and Bacteroides fragilis, and these studies were extended to a survey of well characterised strains of other species in these genera. There was variation amongst strains of C. oerfrinaens type A; many heatresistant food-poisoning strains did not produce the enzyme whereas most haemolytic, heat-sensitive strains did. The results were consistent for all strains of the other clostridial and bacteroides species examined and it is suggested that tests for neuraminidase production should be a valuable addition to currently used biochemical tests in certain areas of taxonomic studies in both genera. The ability of various Clostridia to produce experimental myonecrosis was tested in guinea pigs. Strains classed as neuraminidase-positive in vitro also produced the enzyme in vivo; neuraminidase-negative strains did not. Pathogenic Clostridia produced extensive myonecrosis and death in 24-48 h; virulence was assessed by the challenge dose required to produce fatal infection or by the severity of local muscle infection in surviving animals. Although a number of pathogenic Clostridia produce neuraminidase, there was no clear-cut correlation between virulence and ability to produce the enzyme. Large amounts of neuraminidase were produced in the tissues by virulent strains of C. oerfrinoens and C. septicum but none by the equally pathogenic but neuraminidasenegative C. novyi: at least one neuraminidase-negative strain of C. perfringens was able to produce fatal infection. Various equine and rabbit antisera were assessed for their ability to protect animals challenged with a virulent strain of C. perfringens: the protective effect was found to correlate with the content of anti-oe-toxin. C. perfringens neuraminidase proved to be a poor antigen and it was difficult to produce sera with high levels of neutralising activity; however, there was no evidence that anti-neuraminidase contributed to the protective effect of the antisera tested. The occurrence and roles of neuraminidase and its substrates are reviewed, and various ways in which neuraminidase produced during infection might contribute to tissue damage are discussed. The pathogenesis of C. perfringens gas gangrene remains incompletely understood and a number of features of the disease are not adequately explained by theories that attribute the major role to oc-toxin. Neuraminidase is one of a collection of digestive enzymes that are thought to be of value to commensal or saprophytic C. perfringens strains and it is probable that it has a similar role during infection; however, there is no evidence that neuraminidase is an important virulence factor for the organism.