ABSTRACT
The analysis by Sells and his colleagues of the 600 × 600 matrix of correlations among Cattell and Guilford personality items was unquestionably a remarkable achievement. But an important technical issue remains as to whether a collection of items is better factor analyzed by items or on the basis of homogeneous subtests or clusters of items. In fact, Cattell (1957), Guilford (1975), Comrey (1961), and Cronbach and Gleser (1953) have all recommended the use of what are variously labeled clusters, parcels, and homogeneous item dimensions. Others, like Eysenck and Eysenck (1969) and Howarth and Brown (1971) have emphasized an item-based factor analysis.
