Five different types of surgical mask of varying design and composition of natural and synthetic fibres were tested for their efficiency in vivo by means of a special test chamber. Contaminated particles escaping through or around the mask during speech by the wearer could be collected and sized. Analysis of the data showed that the gross efficiency of all the masks was high, but that some masks were distinctly better at small particle 'filtration' than others. There was a significant difference in efficiency between the best and worst masks. The best masks contained more fabric, were softer and were pleated, while the worst were stiffer, smaller and not pleated. Reusable cotton fabric masks were as effective as synthetic fabric masks when made to a good design.