Abstract
In a replication of the well-known Goldberg (1972) study, sets of inventory scales were con structed from a common item pool, using variants of what are here called the Inductive, Deductive, and External strategies, respectively. Two addi tional scales were also written. Peer ratings for 21 traits, employing two different scale formats, served as criteria. Subjects were 138 students of both sexes. In spite of a number of procedural dif ferences, most notably a concentration on "trait- relevant" criteria and the use of zero-order correla tions as indices of validity, Goldberg's main finding stood unchallenged: Very little variation in validity was attributable to construction strategies. It is pointed out, however, that the Deductive strategy accomplishes its feats with much less effort and considerably shorter scales. Further work with this methodology is urged, as is increased attention to the measurement of criterion variables.