Abstract

Two chest radiographs of each of 687 coal-face workers, obtained during two surveys two and a half years apart at eight pits in South Wales, have been assessed according to their technical quality as “acceptable”, “black”, “grey”, or “white”. The effect of variations in quality on the reading of the category of simple pneumoconiosis (independently carried out) is reported; the effect on change of category between the two surveys is considerable.

From the distributions of categories in each class of technical quality within a pit on each survey, it appears that the extreme categories 0 and 3 are read proportionately more often on acceptable films than on imperfect films. This phenomenon is described quantitatively by fitting a continuous distribution of abnormality to each of the sub-groups, and transforming the ϰ-ray continuum scale into a scale that approximately normalizes the distributions. Measured in this way, the black films are found to have more than twice as much effect on the reading as the grey and white films. A method for approximately correcting for this effect is given.

The findings are interpreted in terms of random observer error in assessing abnormality on a continuous scale, and a recently published model for observer error is criticized.