Bob, you are absolutely correct about sample size and statistical significance. You would really need seven of each to attain 95% accuracy (degrees of freedom for a t-value of .95), which would really narrow your certainty down about which is better because the result will be biased toward the true measurands, in this case "straight" and "square". But do not underestimate the value of a grab sample, as Mr. Pete has done. He used random chance for each sample- they were what he had, random from the others made before and since. The relative difference turned out to be very small, which would naturally grow smaller as the sample population grows. Errors would occur on both sides of the true value, so positive and negative to square, meaning the central tendency for a large population will be close to true square. Any variance will be deviation from square, so the average should also be true square. At the end of the day, I'm just surprised at how comparable the squares really were. I have two Chinese ones that move around when I tighten the thumb screw. I have a Lufkin that is rigid and repeatable, but not very true. I was beginning to think that all combo squares were crap. Maybe I need a Starrett...