Building on the work of Mael (1991), Gandy, Outerbridge, Sharf & Dye (1989)
and Asher (1972), 10 attributes or dimensions of biodata items were define
d operationally. The 160 items of a biodata form undergoing validation were
then rated reliably on the 10 item attributes. Biodata responses and super
visory ratings were obtained for a sample of 528 clerical workers, and an e
mpirical scoring key was developed using an option-keying method. Degree of
item validity was associated with items that were Indirect (opinions of ot
hers about the respondent) and-to a lesser degree-Verifiable and Job Releva
nt. Five of the seven indirect items concerned putative or projected assess
ments of personality attributes by the respondents' supervisors. Items repr
esenting Non-controllable events at first also seemed related significantly
to valid prediction; however, that was owing to the strong association bet
ween Controllability and Directness. The other six dimensions were not rela
ted to item validity, including the degree to which an item was Historical
in nature, which has in the past virtually defined what is meant by a bioda
ta item. The factor structure of dimensions for valid items was different f
rom that of non-valid items. Implications for the construction of biodata f
orms and the limitations of the study are discussed.