The production of neuraminidase by a classical strain of Clostridium welchii (C. perfringens) type A was studied. Good yields were produced in 5% Proteose Peptone-water medium (PPW5); the enzyme was essentially extracellular but some further neuraminidase could be released by ultrasonic disintegration of the cells. This also released N-acyl neuraminic acid-aldolase (NAN-aldolase) and the degree to which this interferes with the assay for neuraminidase was evaluated. Forty-one British reference food-poisoning strains of C. welchii type A were examined for extracellular neuraminidase production in PPW5. Twelve of 17 strains that produce so-called heat-sensitive spores were neuraminidase positive whereas 20 of 24 strains that are non-haemolytic and produce very heat-resistant sporeswere neuraminidase negative. Variation was found in the ability to produce neuraminidase among strains of a single Hobbs' serotype; four Hobbs' type-13 strains produced neuraminidase but a fifth did not. Disruption of the cells of a Hobbs' type-2 strain that did not produce any extracellular neuraminidase released NAN-aldolase but there was no evidence of cell-associated neuraminidase. British food-poisoning strains of C. welchii type A thus include some that are clearly neuraminidase positive and some that still cannot be shown to produce neuraminidase. There is no correlation between lack of neuraminidase production and the ability to cause food poisoning, although the majority of non-haemolytic heat-resistant strains do not produce neuraminidase. It remains possible that neuraminidase may play a part in C. welchii gas gangrene; it is suggested that the ability to define neuraminidase-negative strains may now be of value in investigating this possibility.