Bacterial shedding, wound contamination and clinical-infection rates in clean wounds are influenced by operating-theatre dress. The aim of this study was to clarify the relative contribution of hats, masks and clothing to the control of wound contamination in both ultraclean (enclosed vertical laminar-flow) and conventional (plenum ventilated) airflow theatres. Personnel wore varying combinations of dress in both types of theatre. Colony forming units (cfus) were measured on settle plates at head and waist height, and in the air by a centrifugal air sampler. Bacterial counts in conventional theatres were consistently high and were not significantly influenced by theatre dress. There was a 22-fold increase in cfus on settle plates at waist height when neither hat nor mask were worn, a 15-fold increase when a hat but no mask was worn and a fourfold increase with a mask but no hat in vertical laminar airflow enclosures, although air sample counts remained low. When balloon-cotton clothing was worn, rather than cuffed polyester with microfilament barrier-fabric gowns, cfu counts rose by a factor of six. The bacterial inoculum in conventionally ventilated theatres, or in ultraclean theatres if hat or mask are omitted or balloon-cotton clothing worn, is theoretically sufficient to infect a prosthetic arthroplasty. Theatre-air sampling alone does not reflect local contamination when a surgeon stands over a wound in a vertical laminar-flow enclosure, and both hats and masks are an important part of dress in such environments.