Discriminant validity is not a statistic, it is a concept. The idea is that you want to make sure your construct is related to other constructs as you would expect (convergent validity; e.g., in the Big 5, Extraversion is positively correlated with Openness to Experience), but the constructs are not so correlated that they are redundant (divergent validity: the Extraversion--Openness correlation is small to moderate at best, so the items are not measuring the same construct).
Terrence D. Jorgensen
Postdoctoral Researcher, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam