No abstract

No abstract

Brayfield and Crockett (1) have provided a valuable service in their recent review of the literature on employee attitudes and employee performance. Their well-substantiated conclusions as to the general absence of correlation between employee attitudes and such performance measures as productivity, absenteeism, accident rate, and job tenure are so out of keeping with commonly held assumptions that their paper will have a considerable impact upon the utilization of such attitude measures for both research and administrative purposes. In terms of the recent classification of validity concepts (4, 6), their findings point to an absence of both predictive and concurrent validity for morale surveys, for the typical purposes of their utilization. The present paper raises the question of the construct validity of morale surveys, and presents positive evidence of such validity. Before giving these data, we would like to state briefly our understanding of construct validity, in terms somewhat different from Cronbach and Meehl (4), although in fundamental agreement with them.A given scientific construct has multiple potential operational specifications. If, as sampled, these operational specifications concur, the construct and the sampled measurement techniques have validity. Constructs for which diverse operational specifications persistently fail to agree are in the long run modified or abandoned. In the physical sciences, agreement among methodologically independent operational specifications of the same construct may agree on the order of .99, if expressed as a correlation coefficient over a population of instances. In the psychology of individual and group differences,

No abstract

No abstract