The problem of surface finish and its influence on cavitation inception is reviewed in terms of both theoretical aspects and practical implications. A field study indicates that the roughness problem may be broken down into two categories, namely, isolated asperities and distributed roughness. All the available data for isolated asperities are collated into a single equation involving Reynolds number and relative roughness height. The effects of distributed roughness are linearly proportional to the local skin friction coefficient. New data are presented for slots and wire screen roughness. The use of the collated data is illustrated by several examples.