Correlating one method with another would seem straightforward, but logical problems arise. Because a new measurement is often not designed to replicate precisely the other scales with which it is being compared(indeed, it may be intended to be superior), the expected correlation may not be perfect. But how high should it be, given that the two indexes are inexact measurements of similar, but not identical, concepts? Here lies a common weakness in reports construct validation: few studies declare what levels of correlation are to be taken as demonstrating adequate validity. The literature contains many examples of authors who seem pleased to interpret arbitrarily virtually any level of correlation as supporting the validity of their measure construct validation should begin with a reasoned statement of the types of variable with