In the absence of a statistically based assessment of duster assignments of
observations in terms of their similarity along selected variables the val
idity of a set of dusters could be developed if clusters withstand both emp
irical scrutiny and "make sense" to expert informants.