It was hypothesized that field dependence and attention measures would differentiate accident-loaded and accident-free employees of a petroleum services company. Analysis of variance revealed main effects and a three-way interaction for the Rod-and-Frame Test. Main effects occurred for the Attention-Diagnostic Method and Embedded- Figures Test. No differences occurred for the Closure Test. Regression analysis produced an R (76) = .41, p < .01, with the Attention-Diagnostic Method contributing more to prediction. Equality-symmetry violations occurred in the data. Cautious interpretation was advised because of the assumption violations. The accident-loaded subjects produced consistently greater performance variances, which suggested general performance characteristics in several respects. Future research should be longitudinal-predictive, oriented from Kerr 's complementary safety theories.