That only applies to the description of the typical person with that type indicator, which is different than the type indicator itself. I.e., it may be possible that the categories (E/I)(N/S)(T/F)(P/J) have some validity, without the descriptions being accurate.
I would start by creating really tight definitions. With the exception of the E/I factor, which has already been validated, I'm not sure if the other factors are well-defined enough to measure.
http://en.wikipedia.org/wiki/Forer_effect
and a very good reason not to believe Myers-Briggs type results, which have never been carefully validated.