This paper examines the sample-free test calibration claim, derived from Rasch's structural model for test items, by comparing item difficulty and log ability estimates between different samples of ability levels. Three tests from the Ball Aptitude Battery--Inductive Reasoning, Paper Folding; and Vocabulary--were administered to three samples of different ability levels: (1) 353 high school freshmen; (2) 112 high school seniors; and (3) the same 112 high school seniors four years later as young adults. The Rasch model fit all three tests reasonably well in all three samples, and the claim of sample-free test calibration was supported in that the Rasch item difficulty and person ability parameters were relatively independent of the ability level of the calibrating samples. (Author/BS)