AN EXPERIMENT IN FILM READING

Abstract

In a first trial of an experimental classification of simple pneumoconiosis, each observer was asked to record, for each reading, not only the (I.L.O.) category into which he would have placed the radiograph but also whether he had seriously considered placing it in an adjacent category. Three skilled readers took part in the experiment, in which 120 films (consisting of three, taken in 1949, 1952, and 1956, for 40 subjects who had been at one colliery throughout the period) were read independently by each reader on two occasions and, subsequently, in a joint reading session. Despite the element of arbitrariness in the definition of when an alternative category should be recorded, the proportion of films for which alternates were quoted was very similar (just about half) for all three readers.

Complete agreement in the finer classification which is produced with the experimental method of reading was naturally less common than when conventional categorization was considered. However, when account was taken of the extent of the disagreements, both intra-observer error and inter-observer error were found to be considerably reduced in the experimental classification.

In studying the progression of pneumoconiosis in the subjects' films from one survey to another, it was found that there was an apparent understatement of progression when the conventional classification was adopted. This arose because, where both films for one subject were placed in the same (I.L.O.) category, there was a considerable tendency for progression to be indicated by a change in the categorization according to the experimental classification.